Antagonism of Glycolysis and Reductive Carboxylation of Glutamine Potentiates Activity of Oncolytic Adenoviruses in Cancer Cells

Hanahan and Weinberg's elegant exposition of the “Hallmarks of cancer” in 2000 (1) provided a timely insight into essential aspects of the tumor phenotype that commonly arise from the plethora of acquired genetic mutations. Originally, they described just six cancer hallmarks; however, the pace of scientific progress was so great that in 2011 they published an updated version including two enabling characteristics, namely, genetic instability and tumor-promoting inflammation as well as the two emerging hallmarks, immune evasion and, crucially, the reprogramming of energy metabolism (2). The most well-known aspect of metabolic reprogramming of cancer cells is the phenomenon known as the Warburg Effect, where cells uncouple glycolysis and the tricarboxylic acid (TCA) cycle and undertake disproportionately more glycolysis even in the presence of oxygen, a process known as “aerobic glycolysis” (3, 4). Uncoupling glycolysis and the TCA cycle yields less ATP per glucose molecule, but allows glycolytic metabolites to be used in macromolecular synthesis (e.g., via the pentose phosphate pathway to produce nucleotide precursors), a feature that is important in both proliferating cells and cancer cells (5, 6).

Glucose and glutamine are the two primary sources of carbon for energy homeostasis and macromolecular biosynthesis in all mammalian cells. Normally, glucose is considered to be the main contributor of cellular ATP production via glycolysis and the TCA cycle; however, when glucose-derived carbon is directed away from the TCA cycle (e.g., in aerobic glycolysis or the Warburg effect), glutamine has often been shown to replenish the TCA cycle via α-ketoglutarate (αKG), a process termed anaplerosis (7–9). Tumors commonly harbor mutations in the TCA cycle or electron transport chain that disable normal oxidative mitochondrial function. In such cells, glutamine-dependent reductive carboxylation rather than oxidative metabolism is the major pathway of citrate formation, which provides acetyl coenzyme A for lipid synthesis as well as other TCA cycle intermediates and related biosynthetic precursors. This reductive, glutamine-dependent pathway is the dominant mode of metabolism in rapidly growing malignant cells both in vitro and in vivo and also in cells with normal mitochondria subjected to hypoxia. Consistent with these concepts, recent studies have shown that MYC activation downstream of oncogenic KRAS pushes tumor metabolism toward a “Warburg” phenotype of increased lactic acid production and a shunting of glucose metabolites toward the nonoxidative pentose phosphate pathway for ribose generation (10–12). This “metabolic transformation” has been described as a vulnerability of cancer, and recently, many treatments have been developed to exploit it (7–9, 13, 14).

Viruses are professional intracellular parasites that rely on manipulation and hijacking of host cell metabolism to provide both energy and the molecular building blocks needed for the production of a large number of viral progeny. In many ways, their demands for macromolecular synthesis appear similar to those of a proliferating or cancer cell. Metabolomic analysis of virally infected cells was first conducted with cells infected with human cytomegalovirus (HCMV; ref. 15). During the lytic portion of its life cycle, HCMV was found to induce a range of metabolic changes including an upregulation of the TCA cycle, nucleotide biosynthesis, and glycolysis. Such alterations are now known to be critical for replication of many viruses, including Kaposi's sarcoma-associated herpesvirus, herpes simplex virus 1, hepatitis C virus, human immunodeficiency virus, and dengue virus (15–20). To date, studies that have investigated the impact of nutrient deprivation on virus production, including HCMV, HSV-1, and poliomyelitis virus, have found that both glucose and glutamine are required for production of new infectious viral particles (21–23).

In two seminal articles, Thai and colleagues reported that E4ORF-1–induced MYC-dependent upregulation of glucose and glutamine consumption and glutaminase (GLS) activity are necessary for optimal adenovirus replication in primary lung epithelial cells, and that GLS activity is required for optimal replication of HSV-1 and influenza A as well (12). Whether both HSV-1 and influenza A also rely on MYC activation to create the dependence on GLS however remains to be determined. Interestingly, Thai and colleagues also report an increase in glutamine uptake and reductive carboxylation in normal cells infected with wild-type adenoviruses (11, 12).

Oncolytic viruses represent an emerging new class of cancer therapeutics, with the first agent (Imlygic) receiving its product license in 2015 (24). Given the current high profile of this field (25), surprisingly few studies have analyzed the possible involvement of tumor-associated metabolic transformation in contributing to the cancer selectivity of oncolytic viruses. The availability of energy and biosynthetic intermediates within cancer cells could provide a very supportive environment for oncolytic viruses, relieving the requirement to achieve reprogramming of cellular metabolism themselves.

In this study, we set out to determine how metabolic transformation of tumor cells may be suitable to support activity of wild-type and oncolytic adenoviruses, and how virus infection may further modify tumor cell metabolic processes. For clinical relevance, we have explored how virus-infected tumor cells behave when glucose and glutamine are available at only very low levels or absent, and we have studied two distinct cancer cell lines in depth—one (A549 human lung carcinoma cells) chosen for its support of a rapid adenovirus infectious cycle shows a classical “Warburg” phenotype with aerobic glycolysis not coupled to oxidative phosphorylation (OXPHOS), and the other (SKOV3 ovarian carcinoma) selected because the virus normally takes several days to complete its life cycle and the cell line couples glycolysis closely to OXPHOS (26).

A549 and SKOV3 cells (ATCC) were maintained in DMEM (Sigma) with 25 mmol/L glucose supplemented with 10% FBS (Gibco) unless otherwise stated. Cells were maintained in humidified atmosphere in 5% CO₂ at 37°C. All cell lines were authenticated by short tandem repeat profiling (CRUK Cambridge Institute). For virus infections cells were infected in DMEM supplemented with 2% FBS. Ninety minutes after infection, medium was removed and replaced with DMEM supplemented with 2% FBS. All cell lines were routinely tested for Mycoplasma using MycoAlert Mycoplasma Detection Kit and were free from contamination during the course of these studies.

EnAd-CMV-GFP, EnAd-SA-GFP, and EnAd-SA-FLuc were kind gifts from PsiOxus Therapeutics. EnAd-CMV-GFP is a replication-competent derivative of EnAd, a chimera of Ad3 and Ad11p with a 2444 bp deletion in E3ORF and a 25 bp deletion in E4ORF4 containing a CMV immediate-early promoter-driven GFP cassette inserted after the fiber gene. EnAd-SA-GFP and EnAd-SA-FLuc are derivatives of EnAd containing GFP or luciferase under the control of the viral major late promoter/splice acceptor. Concentrated and purified virus stocks were prepared through a cesium chloride gradient. Crude or purified virus stocks were used to infect HEK 293 cells grown to confluency in a Corning HYPERFlask. Purified stocks were characterized using the Quant-iT PicoGreen dsDNA Assay Kit (Thermo Fisher Scientific) and TCID₅₀ using A549 cells. Infectious particles were measured by TCID50 assay on A549 cells using 1:10 serial virus dilutions. Plates were observed day 5 after infection. The highest dilutions showing cytopathic effect (CPE) were counted, and infectious particle unit [plaque-forming units (PFU) per milliliter] calculated using the Karber formula.

A549 cells or SKOV3 cells were seeded in 24-well plates at 100,000 cells per well in 24-well plates and left to adhere at 37°C and 5% CO₂ overnight in 1 mL high-glucose DMEM + 10% FBS. Cells were then infected with 100 virus particles per cell (ppc) Ad5, Ad11p, or EnAd in DMEM + 2% FBS, and after 90 minutes, the medium was removed and replaced with 1 mL DMEM high glucose + 2% FBS and 300 μmol/L 2-NBDG and then incubated at 37°C and 5% CO₂. Every 24 hours, cells were detached using nonenzymatic dissociation buffer in triplicate, fixed in 4% formalin washed in PBS, and then resuspended in MACS buffer. Fluorescence in FL-1 represents glucose uptake. Fluorescence was analyzed by flow cytometry using a BD FacsCalibur (BD Biosciences) and data processed with FlowJo v10.0.7r2 software (Treestar Inc.).

Lactate secretion into medium was measured using the Lactate-Glo assay kit (Promega) according to the manufacturer's protocol. Briefly, cells were plated in a 96-well plate at 10,000 cells per well and left to adhere at 37°C and 5% CO₂ overnight. Cells were then infected in triplicate with 100 ppc Ad5, Ad11p, or EnAd in DMEM + 2% FBS, and after 90 minutes, the medium was removed and replaced with 200 μL DMEM high glucose + 2% FBS. Every 24 hours post-infection (hpi), 10 μL of medium was removed and diluted in 1,000 μL of PBS and then stored at −20°C until all the samples were collected. Note that 50 μL of each sample was transferred to a sterile white 96-well assay plate. Fifty microliters of lactate detection reagent was added to each well, and the plate was placed on a shaker for 60 seconds to mix and then incubated for 60 minutes at room temperature. Luminescence was recorded using the victor plate reader (1.0s per well). The data are presented as fold increase in luminescence of uninfected cells.

Cells were seeded at 100,000 cells per well in a 24-well plate and incubated overnight to adhere in a humidified atmosphere in 5% CO₂ at 37°C. Cells were then infected in triplicate with 100 ppc of EnAd-CMV-GFP (containing a CMV immediate-early promoter-driven GFP cassette inserted after the fiber gene representing early viral protein production) or EnAd-SA-GFP (containing GFP under the control of the viral major late promoter/splice acceptor representing late-stage protein synthesis) in DMEM + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 1 mL DMEM + 2% FBS was added to the wells. At set time points, cells were washed with PBS detached using nonenzymatic cell dissociation buffer, fixed in 4% formalin for 20 minutes, washed and resuspended in MACS buffer, and analyzed for fluorescence in FL-1. The percentage of fluorescent cells was gated against uninfected cells, and the mean fluorescence represents the geometric mean.

Cells were seeded at 100,000 cells per well in a 24-well plate and incubated overnight to adhere in humidified atmosphere in 5% CO₂ at 37°C. Cells were then infected in triplicate with 100 ppc of EnAd, Ad11p, or Ad5 in DMEM + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 1 mL DMEM + 2% FBS was added to the wells. At set time points, wells were scraped, and the lysate was stored at −20°C until analysis. Samples were extracted with the GenElute mammalian genomic DNA Miniprep Kit (Sigma) according to the manufacturer's protocol. Primers and FAM-TAMRA probes (Sigma) recognized the fiber gene. Analysis of Ad11p and EnAd was conducted using forward primer TACATGCACATCGCCGGA, reverse CGGGCGAACTGCACCA, and probe CCGGACTCAGGTACTCCGAAGCATCCT. Analysis of DLD1 cells was conducted using sequences against human beta-2-microglobulin using forward primer ATCCGACATTGAAGTTGA, reverse primer CCCACTTAACTATCTTGGG, and probe TCACACGGCAGGCATACTCA. Primers and probe were added to QPCRBIO probe mix Hi-Rox (PCR Biosystems) master mix. The qPCR was run on an ABI PRISM 7000 (Applied Biosystems).

Cells were seeded at 100,000 cells per well in a 24-well plate and incubated overnight to adhere in humidified atmosphere in 5% CO₂ at 37°C. Cells were then infected in triplicate with 100 ppc of EnAd, Ad11p, or Ad5 in DMEM + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 1 mL DMEM + 2% FBS was added to the wells. At set time points, wells were scraped, and the lysate was stored at −20°C until analysis. A549 cells were seeded in 96-well plates at 20,000 cells per well in 200 μL DMEM High Glucose + 10% FBS and incubated in humidified atmosphere in 5% CO₂ at 37°C. Note that 1:10 serial virus dilutions were created in DMEM high glucose + 2% FBS and used to infect the A549 cells. Cells were incubated in humidified atmosphere in 5% CO₂ at 37°C for 5 days and then scored for the presence of viral plaques. The highest dilutions showing CPE were counted, and infectious particle unit (PFUs per milliliter) was calculated using the Karber formula.

A549 cells were seeded at 100,000 cells per well in a 24-well plate and incubated overnight to adhere in humidified atmosphere in 5% CO₂ at 37°C. Cells were then infected in triplicate with 100 ppc of EnAd, Ad11p, or Ad5 in DMEM + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 1 mL DMEM + 2% FBS was added to the wells in the presence or absence of 20 mmol/L glycolysis inhibitor 2-deoxy-D-glucose (Sigma) and incubated in humidified atmosphere in 5% CO₂ at 37°C for 96 hours. Images were taken using a Zeiss Axiovert 25 inverting light/fluorescence microscope, and images were recorded by a Nikon DS-U2 camera and processed with NIS-element AR 3.00 software.

Cells were seeded at 10,000 cells per well in E-Plate 16 plates (ACEA Biosciences) and incubated overnight to adhere in humidified atmosphere in 5% CO₂ at 37°C. Impedance measurements were taken every 10 minutes for the duration of the experiment and displayed as Cell Index, which is used to represent cell viability in real time. Cells were then infected with 100 ppc of EnAd in the presence or absence of glycolysis inhibitor 2-deoxy-D-glucose (Sigma) in DMEM + 2% FBS.

Lactate secretion into medium was measure using the glutamine-glutamate-glo-assay assay Kit (Promega) according to manufacturer's protocol. Briefly, A549 were seeded at 10,000 cells per well in a 96-well plate and left overnight to adhere and then infected with 100 ppc of enadenotucirev, Ad5, and Ad11p or left uninfected and cultured in DMEM (high glucose) + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 200 μL DMEM + 2% FBS was added to the wells. Every 24 hours, medium was removed from wells followed by a wash with PBS. Fifty microliters of inactivation solution was added to the samples followed by 50 μL Tris solution, and samples were stored at −20°C. Twenty-five microliters of aliquots of samples were pipetted into a white 96-well plate. Twenty-five microliters of GLS buffer was added to the wells followed by a 40-minute room temperature incubation, and 50 μL of glutamine detection reagent was added to all wells, and luminescence was measured using the victor plate reader (1.0s per well).

A549 cells were seeded at 100,000 cells per well in a 24-well plate and incubated overnight to adhere in humidified atmosphere in 5% CO₂ at 37°C. Cells were then infected in triplicate with 100 ppc of EnAd-CMV-GFP or EnAd-SA-GFP in DMEM + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 1 mL DMEM + 2% FBS was added to the wells in the presence or absence of 4 mmol/L glutamine and incubated in humidified atmosphere in 5% CO₂ at 37°C for 96 hours. Images were taken using a Zeiss Axiovert 25 inverting light/fluorescence microscope using a 400 ms exposure, and images were recorded by a Nikon DS-U2 camera and processed with NIS-element AR 3.00 software.

Lactate dehydrogenase (LDH) levels in medium were analyzed using the CytoTox 96 Non-Radioactive Cytotoxicity assay (Promega) according to the manufacturer's protocol.

Primary human malignant peritoneal ascites were received from the Churchill Hospital, Oxford University Hospitals (Oxford, UK) following written-informed consent from patients with indications of advanced carcinoma, predominantly ovarian cancer. This work was approved by the research ethics committee of the Oxford Centre for Histopathology Research (Reference 09/H0606/5+5), and studies were conducted in accordance with the CIOMS guidelines. Upon receipt, cellular and fluid fractions were separated, and fluid used immediately or aliquots stored at −80°C for future analysis. The cellular fraction was treated with red blood cell lysis buffer (Roche) following the manufacturer's instructions. Cell number and viability were determined by trypan blue stain.

Cells were seeded at 100,000 cells per well in a 24-well plate and incubated overnight to adhere in humidified atmosphere in 5% CO₂ at 37°C. Cells were then infected in triplicate with 500 ppc of EnAd-SA-GFP (containing GFP under the control of the viral major late promoter/splice acceptor representing late-stage protein synthesis) in RPMI + 2% FBS. Ninety minutes after infection, medium was removed from the cells, and they were washed with PBS, and then 1 mL RPMI + 2% FBS was added to the wells. Seven days after infection, cells were imaged using the celigo image cytometer (Nexcelom Biosciences), and the number of GFP-positive cells was analyzed by the celigo software.

All animal experiments were performed in accordance with the terms of UK Home Office guidelines and the UKCCCR Guidelines for the Welfare of Animals and were approved by the University of Oxford Animal Welfare and Ethical Review Body. All animals were held in individually ventilated cages in specific pathogen-free barrier units and allowed to acclimatize for 1 week prior to any procedures being carried out.

For in vivo coadministration studies, 6-week-old female athymic nude mice Crl:NU(NCr)-Foxn1^nu (Charles River) were injected s.c. with 5 × 10⁶ DLD1 cells. Once xenografts developed to approximately 100 mm³, mice were randomized and dosed intratumorally with 10 μL containing 5 × 10⁹ particles of enadenotucirev or vehicle control in two separate locations. Ninety-six hpi, mice received 500 mg kg⁻¹ 2-deoxyglucose (2DG) every other day for 2 weeks, whereupon they were sacrificed and the xenografts were removed. Tumors were homogenized in 1x reporter lysis buffer (Promega) using a motorized homogenizer (Ultra Turrax IKA T18 Basic; Fisher Scientific) to obtain a 150 mg/mL homogenate. Luciferase assays were conducted according to the manufacturer's protocols (Promega) and normalized to protein content using the bicinchoninic acid assay (QuantiPro, Sigma).

In all cases of more than two experimental conditions being compared, statistical analysis was performed using a one-way ANOVA test with Tukey post hoc analysis or two-way ANOVA test using Bonferroni post hoc analysis. All data are presented as mean ± SD. The significant levels used were P = 0.01–0.05 (*), 0.001–0.01 (**), and 0.0001–0.001 (***). All in vitro experiments were performed in triplicate, unless stated otherwise.

We set out to characterize how virus infection affects cellular energy metabolism. Glucose uptake was measured in two different cancer cell lines infected with three adenoviruses (Ad5, Ad11p, and EnAd) at 100 ppc, which equates to a multiplicity of infection of 1–3 infectious virus particles per cell (Supplementary Table S1). A549 cells showed an increase in uptake of the fluorescent glucose analogue 2-NBDG following infection with each of the viruses, although the stimulation of uptake by EnAd was faster and more pronounced (Fig. 1A and B). Cells incubated with a lower viral dose (10 ppc) showed a similar pattern (Supplementary Fig. S1).

Adenovirus-infected cells are well known to show increased acidification of their culture medium (Fig. 1C; ref. 12). To assess whether this reflected aerobic glycolysis, we measured secretion of lactic acid by A549 cells infected with Ad5, Ad11p, and EnAd (Fig. 1D). All viruses caused increased lactic acid secretion, with Ad5 and EnAd showing significant effects (up to 1.4-fold rises) by 72 hpi and Ad11p showing an increase after 96 hpi. Cells infected with EnAd showed a fall in lactic acid production, perhaps reflecting earlier cell death. This was reflected by earlier release of LDH from cells infected with EnAd (Supplementary Fig. S1).

Each of the viruses also promoted increased glucose uptake in SKOV3 cells, with 40% to 50% of cells showing more uptake of 2-NBDG than uninfected control cells at 48 hpi (Fig. 1E). Interestingly, the mean levels of 2-NBDG uptake were approximately 10-fold higher following EnAd infection than for the wild-type Ad5 and Ad11p (Fig. 1F). Finally, whereas wild-type Ad5 and Ad11p showed no obvious effects of virus dose on increased glucose uptake, EnAd showed a clear dose effect with higher doses giving increased glucose uptake at much earlier times (Fig. 1G).

In order to assess whether glycolysis was essential to the successful EnAd life cycle, the competitive inhibitor of glucose-6-phosphate isomerase, 2DG, was used to inhibit glycolysis. The presence of 2DG inhibited glycolytic flux in EnAd-infected A549 cells, as shown by the dose-dependent loss of acidification of the medium (Supplementary Fig. S2). The treatment also increased the basal glucose uptake of A549 cells as previously reported (Supplementary Fig. S2).

Virus activity was probed using two EnAd-based GFP reporter viruses, with GFP controlled by a CMV promoter (EnAd-CMV-GFP) or a splice acceptor under the virus Major Late Promoter (EnAd-SA-GFP). Low concentrations of 2DG (up to 10 mmol/L) had little effect on the frequency of GFP-positive cells or the level of GFP expression using EnAd-CMV-GFP (Fig. 2A and B). However, the highest concentration of 2DG (20 mmol/L) decreased both the percentage of cells positive for GFP and the signal intensity.

We also assessed the effects of 2DG on virus genome replication and production of infectious viruses. Although the number of genomes produced was decreased by the presence of 2DG (Fig. 2C), EnAd-SA-GFP showed similar effects of 2DG on the frequency of infection, but the intensity of GFP expression was enhanced, in a dose-dependent fashion, reaching at least twice the level of control cells at 5 mmol/L 2DG and above (Fig. 2D and E). This was unexpected and suggested that antagonism of glycolysis might have a positive effect on the activity of the MLP or could be boosting the translation of late virus transcripts. The number of infectious viruses produced was essentially unaffected (Fig. 2F). When analyzing the efficiency of virus genome packaging by comparing the amount of viral DNA to the number of infectious particles, increasing concentrations of 2DG gave fewer unpackaged virus genomes, suggesting greater efficiency of genome packaging (Fig. 2G). This fits with a more balanced production of virus genomes and capsid proteins in the presence of 2DG.

Light microscopy images of EnAd-infected A549 cells at the point of death (4 days after infection) show that the presence of 2DG increased the formation of single-membrane blisters that are typically observed when cells are killed by EnAd-mediated oncosis, indicating that viral death mechanisms are not changed by restricting glycolysis (Fig. 2H; ref. 27).

The effects of glycolysis inhibition on virus cytotoxicity were measured in real time in A549 cells using the xCELLigence system (Fig. 2I) giving a real-time analysis of cell viability.

2DG treatment alone was not cytotoxic up to 25 mmol/L and in fact increased the amount of viable cells in a dose-dependent manner (Fig. 2I). Cells infected with EnAd (100 ppc) showed the onset of cytotoxicity from 60 hours after seeding (36 hpi), and cytotoxicity occurred around 10 hours earlier in the presence of 2DG (Fig. 2I).

We also assessed the effects of removing glucose from the media of virally infected cells. Removal of glucose from the medium at the point of infection decreased viral protein expression in A549 cells; however, cells allowed to acclimatize to 0 mmol/L glucose for 48 hours prior to infection showed elevated virus transgene expression and increased infectious particles (Supplementary Fig. S3). This further substantiates the observation that inhibiting glycolysis improves virus activity. For further studies into the mechanism, we made use of 2DG rather than glucose removal, because, in static culture, the presence of 2DG is better tolerated.

SKOV3 cells grow slower than A549 and couple glycolysis to OXPHOS, shedding less lactate (Supplementary Fig. S2; ref. 26). EnAd replicates relatively slowly in SKOV3 cells, and the presence of 2DG showed little effect on transgene expression from EnAd-CMV-GFP (Fig. 3A–C). Treatment with 20 mmol/L 2DG brought forward the peak expression of EnAd-SA-GFP to 96 to 150 hpi compared with 150 to 200 hpi without 2DG (Fig. 3D–F). This differential effect on these two reporter viruses suggests that inhibition of glycolysis may accelerate expression of late virus proteins (expressed under control of the MLP), perhaps reflecting shortening of the virus replication cycle.

The effects of 2DG on SKOV3 cells were assessed in real time (Fig. 3G and H). In mock-infected cells, growth was inhibited by increasing doses of 2DG, although no toxicity was observed (Fig. 3H). In cells infected with EnAd without 2DG, cytotoxicity began at roughly 170 hours after seeding (5 days after infection). The presence of 2DG accelerated the onset of cytotoxicity, with 5 mmol/L 2DG causing death at 3 days after infection. These data confirm that inhibition of glycolysis can accelerate production of late virus proteins, shorten the virus life cycle, and hasten the onset of virolysis.

The effects of 2DG on virus genome replication in SKOV3 cells were measured using EnAd (Fig. 3I) and Ad11p (Supplementary Fig. S2). Five days after infection, the total number of EnAd genomes (measured by qPCR) increased from 4 × 10¹¹ to around 7 × 10¹¹ genomes/mL with the greatest effect observed after treatment with 5 mmol/L 2DG. Similarly, Ad11p genome production rose from 1 × 10⁹ to 3.5 × 10⁹ genomes/mL following addition of 5 mmol/L 2DG. The number of infectious progeny viruses produced in SKOV3 cells was measured by TCID₅₀ assay. Treatment with 2DG caused approximately 100-fold increased production of EnAd progeny (Fig. 3J) with similar effects on Ad11p (Supplementary Fig. S2).

SKOV3 cells use glucose to fuel the TCA cycle (Supplementary Fig. S2), and this appears not to create a good environment for virus activity. In contrast, glycolysis inhibition by 2DG, leading to anaplerotic fuelling of the TCA cycle, provides a superior environment. Pusapati and colleagues (28) produced glycolysis-independent cell lines by serial passage in 2DG, with the resultant cell lines fuelled predominantly by anaplerotic use of glutamine. We created a similar SKOV3 cell line through multiple passages in 20 mmol/L 2DG. As expected, this glycolysis-independent SKOV3 cell line is dramatically more supportive of the viral life cycle with increased viral transgene production and over 50-fold increased production of virus genomes compared with parental SKOV3 cells (Supplementary Fig. S3).

These data show that, in A549 cells, where glycolysis and the TCA cycle are not closely coupled, viral lysis and genome production occur rapidly, and inhibition of glycolysis causes a modest improvement in virus activity. This may be because these cells already exhibit a Warburg phenotype. In SKOV3 cells, where glycolysis is closely coupled to OXPHOS, the normally low levels of virus activity can be dramatically augmented by inhibition of glycolysis driving cells toward anaplerotic use of metabolites.

Clinical samples of malignant peritoneal ascites containing a range of tumor cells were incubated with 500 ppc EnAd-SA-GFP under a range of metabolic conditions. Ascites samples treated with increasing doses of 2DG showed a significant increase in the number of GFP-positive cells (Fig. 4A). Similarly, RT-qPCR analysis shows a doubling in viral genomes produced in ascites cells treated with 20 mmol/L 2DG (Fig. 4B). These results suggest that inhibition of glycolysis boosts virus activity in primary cancer cells as well as in cell lines

We assessed the effects of 2DG treatment on the activity of EnAd-SA-FLuc after a single intratumoral injection of 5 × 10⁹ vp/tumor into pre-established DLD1 human colon carcinoma xenografts in athymic nude mice. Intraperitoneal injections of 2DG (500 mg/kg) were given every 48 hours for 2 weeks after infection, after which time, mice were sacrificed, the tumors removed for measurement of late viral protein expression via luciferase activity and virus genomes by qPCR. The combination of 2DG and EnAd in vivo led to a 10-fold increase in late viral protein production (Fig. 4C) and over 10-fold rise in virus genomes produced per cell (Fig. 4D). These results show that, as with the in vitro models, restriction of glycolysis produces a favorable metabolic phenotype, which improves the adenovirus life cycle.

Because we observed that EnAd produces more infectious virus particles when glycolysis is restricted, we hypothesized that other nutrients were being used either for energy metabolism or to provide biosynthetic precursors. Glutamine provides an alternative energy source and supports many metabolic functions needed for cell survival, growth, and proliferation. It is particularly important in tumors where glucose is often present at low levels with some tumor cells regarded as “glutamine-addicted.”

We showed an increase in glutamine uptake by A549 cells following infection with Ad5, Ad11p, or EnAd, although the stimulation of uptake by EnAd was faster and greater (Fig. 5A). Infection with Ad5 or Ad11p only showed a significant increase in uptake at around the point of death (96 hpi), whereas EnAd caused a significant increase in glutamine uptake by 48 hpi.

To assess whether glutamine was essential to successful EnAd life cycle, A549 cells were infected with EnAd-CMV-GFP or EnAd-SA-GFP. Cells infected with either virus in the presence of glutamine (4 mmol/L) showed high expression of GFP expression by 72 hpi. In the absence of glutamine, very little GFP expression was observed with either virus (Fig. 5B). We confirmed this using flow cytometry, showing that around 90% of cells infected with EnAd-CMV-GFP or EnAd-SA-GFP are GFP positive by 96 hpi in the presence of glutamine, whereas less than 10% of cells were positive if infected in media without glutamine (Fig. 5C–F). In GFP-positive cells, the levels of GFP expression per cell were 10- to 50-fold lower in the absence of glutamine (Fig. 5D and F).

We also analyzed the effect of limiting glutamine on viral genome replication (Fig. 5G) and the production of infectious virus particles (Fig. 5H). Removal of glutamine resulted in over 1,000-fold decrease in viral genomes and a 100,000-fold reduction in infectious viral particles. A similar effect was also observed using the parental virus Ad11p (Supplementary Fig. S4). This result was confirmed using the GLS inhibitor CB839 (Supplementary Fig. S4).

Glutamine appears to be essential to support adenovirus replication in A549 cells. Many cancer cell types utilize exogenous glutamine, via its conversion to αKG, as a TCA cycle intermediate. This anaplerotic role of glutamine might allow the cell to use more glucose-derived glycolytic intermediates for biosynthesis instead of feeding into the TCA cycle, and may provide additional TCA cycle intermediates for biosynthesis. To assess the importance of this pathway during adenovirus infection, we attempted to supplement the missing glutamine by adding alternative anaplerotic TCA cycle intermediates to EnAd-infected cells.

Following infection of A549 cells in glutamine-free media with EnAd-GFP viruses, only dimethyl-αKG (dmαKG) and not pyruvate or oxaloacetic acid (OAA) caused an increase in the frequency and the average intensity of GFP expression. This was more obvious for EnAd-SA-GFP than for EnAd-CMV-GFP, suggesting that αKG may play a particular role at later stages of the virus life cycle (Fig. 6A–D).

Glutamine-free A549 cells showed very low levels of virus genome synthesis (Fig. 6E), and this was slightly increased following addition of pyruvate or OAA. The addition of dmαKG showed a complete recovery in viral genomes produced. Similarly, the addition of exogenous dmαKG to glutamine-depleted virus-infected A549 cells allowed them to produce almost as many infectious virions as complete medium. Addition of other exogenous TCA cycle intermediates did not have this effect (Fig. 6F).

One aspect of glutamine metabolism is the possibility for production of biosynthetic intermediates such as citrate by reductive carboxylation, without any depletion of cellular NAD⁺. Production of citrate from glucose consumes two equivalents of NAD⁺, which can be deleterious for the redox balance of the cell if biosynthesis is intense. When we measured levels of cellular NAD⁺ following virus infection, we found that Ad11p causes only a relatively modest fall, whereas EnAd led to substantial depletion of NAD⁺ (Fig. 6G). It may be that the significance of reductive carboxylation in virus infection is to facilitate biosynthesis without causing fatal depletion of cellular NAD⁺ levels, although this remains to be elucidated.

Similar results were seen in EnAd-infected SKOV3 cells in the absence of glutamine, where only addition of dmαKG caused a restoration of GFP expression (Fig. 7). The addition of dmαKG was particularly beneficial for synthesis of late virus proteins, because the effects on GFP expression were more dramatic using EnAd-SA-GFP than EnAd-CMV-GFP, restoring the frequency of infection to levels observed in complete medium, and giving a dramatic increase in the level of GFP expression per cell that even surpassed the level of expression in complete medium (Fig. 7A–D).

In parallel experiments, we assessed whether the addition of anaplerotic TCA cycle intermediates to complete medium would affect the activity of EnAd GFP reporter viruses (Supplementary Fig. S5). Addition of pyruvate, OAA, and dmαKG showed no effects on the frequency of infection of A549 cells, nor on the level of CMV promoter-driven expression of GFP, but dmαKG significantly increased the median expression of SA-GFP (Supplementary Fig. S5). Again this suggests that boosting reductive carboxylation, even in the presence of high glucose, can enhance late-stage viral protein production.

In this study, we explored the effects of adenovirus infection of cancer cells on their metabolic pathways and how this affects the lytic cycle. We elected to work in two well-characterized cancer cell lines, one (A549) known to support a rapid adenovirus life cycle, and the other (SKOV3) chosen because the virus life cycle is longer. Previous studies have shown that A549 cells take up high levels of glucose that is mainly secreted as lactate (26). In contrast, SKOV3 cells route virtually all of their glucose into the TCA cycle, secreting little lactate (26). A549 cells might be seen as more metabolically transformed toward a “Warburg” phenotype, and this may contribute to their greater permissiveness for adenovirus infection (Supplementary Fig. S2C and S2D).

Previous studies of cancer metabolism in vitro have often been conducted in complete media, with nutrients present at concentrations in excess of the levels likely to be achieved in vivo. For greater clinical relevance, we designed our approach to include studying the consequences of limiting supply of glucose and/or glutamine.

The observation that limiting glucose supply enhanced virus activity was counterintuitive, but was clearly apparent in both cell lines, in human tumor biopsies ex vivo and in a xenograft model in vivo. This contrasts with previous findings with some other viruses where inhibiting glycolysis blocks the life cycles of both HCMV and HSV-1 and can induce apoptosis in cells infected with KSHV during the latent stage of infection (29–32). Our initial experiments observed this by simply decreasing glucose (Supplementary Fig. S3); however, low levels of glucose were hard to maintain in static culture due to rapid consumption by cells. We therefore decided to use 2DG as a glycolysis antagonist in the presence of high glucose levels and obtained similar results.

Using two GFP reporter systems for EnAd showed that the beneficial effects of low glucose were particularly important using virus expressing GFP under transcriptional control of the virus MLP (EnAd-SA-GFP). Early adenovirus genes are expressed immediately upon infection and translated by cap-dependent mechanisms, and late genes are expressed only following virus replication, in a cap-independent manner. It therefore seems that restricted glycolysis may be effective in allowing the virus to express its late, cap-independent, structural proteins, an observation that is supported by the literature, showing an increase in cap-independent translation in glucose deprivation (33–35). This empowering of virus life cycle was particularly obvious in SKOV3 cells, so it therefore seems that restricting glycolysis is more effective in cells such as SKOV3 as creating an artificial Warburg phenotype mirrors the favorable metabolic phenotype of A549 cells.

Given that all of the adenoviruses showed improved performance when glucose was restricted, we were surprised to observe that glucose uptake into cells was significantly increased following infection. EnAd caused greater increases than either of the wild-type viruses, presumably reflecting its high rate of replication. Exactly why both cells exhibit this behavior remains unclear and will be the subject of further investigation. Previous studies in Ad5 have shown that E4ORF1 activates cMYC and promotes enhanced glycolysis and glutamine uptake along with an increase in reductive carboxylation in cultured normal epithelial cells (11, 12). It seems likely that this role of E4ORF1 may have evolved for activity where extracellular nutrient levels are lower than used in tissue culture and where increased glucose entry is indeed beneficial for virus activity. Because adenovirus infection is known to deregulate the cell cycle, perhaps rerouting biosynthetic intermediates from cell growth to virus production, it was feasible that 2DG treatment or glucose restriction might boost virus activity through a similar mechanism. However, xCELLigence data showed that 2DG did not affect growth of A549 cells (Fig. 2I), and although it caused a transient inhibition of SKOV3 cell growth (Fig. 3G), this recovered within 48 to 72 hours. Hence, we consider cell-cycle disturbance is unlikely to represent the fundamental factor underpinning the improved virus activity. In addition, the glycolysis-independent SKOV3 cell line we produced is able to divide and grow robustly in the presence of 2DG (Supplementary Fig. S3) and shows dramatically improved virus production compared with parental SKOV3 cells. Hence, we feel that effects on the cell cycle alone cannot explain our observations, rather there is a qualitative change in metabolism due to glucose restriction that provides a superior environment for virus production.

Whereas restricted glycolysis was beneficial for the virus life cycle, restriction of glutamine had powerful negative consequences, with virus genome replication virtually ablated. In addition to being essential to the virus life cycle, uptake of glutamine was strongly increased following adenovirus infection, with EnAd causing greater increases than the other adenoviruses.

We focused on whether glutamine was essential as a source of building blocks, particularly those containing nitrogen, or whether its main role is for anaplerosis. Following cellular uptake into the cytoplasm, glutamine is hydrolyzed by GLS producing glutamate and free ammonium. Glutamate can play a variety of roles in metabolism, but can be deaminated to produce αKG, which can enter the TCA cycle as an anaplerotic metabolite, and may be involved in energy generation by OXPHOS or may undergo reductive carboxylation to produce isocitrate, citrate, and acetyl-CoA for biosynthesis outside of the mitochondria (7, 14, 36–38).

Cells were cultured in glutamine-free medium supplemented by membrane-permeable metabolites that could enter the TCA cycle, namely αKG, OAA, and pyruvate. These agents have all been previously shown to penetrate cells and provide anaplerotic substrates to rescue vaccinia virus from glutamine deficiency (39). The ability of dmαKG to fully rescue production of infectious EnAd particles in glutamine-free cells suggests that EnAd normally relies predominantly on the ability of glutamine to enter the TCA cycle anaplerotically, not as a source of nitrogen. It was also interesting that only dmαKG, and not pyruvate or OAA, showed significant rescue of virus activity. This is different from vaccinia (39). When there is a high ratio of αKG to citrate, glutamine is metabolized reductively, with αKG producing isocitrate, citrate, and acetyl-CoA (40). In contrast, pyruvate and OAA, although able to enter the oxidative TCA cycle, will produce citrate before they produce αKG, therefore prompting oxidative metabolism. Citrate and acetyl-CoA are important cataplerotic substrates for cellular biosynthesis; hence, the essential role of glutamine may be to produce sufficient levels of these metabolites for synthesis of amino acids and lipids (Fig. 7E).

The observation that αKG can enhance virus activity in A549 cells beyond using glutamine alone suggests that cellular uptake of glutamine may be a limiting factor in these cells, and the use of membrane-permeable alternatives could overcome this. This was substantiated by the observation that improving EnAd protein production also occurred when dmαKG was added to complete medium (Supplementary Fig. S5).

In assessing these findings, we wondered why producing metabolites by reductive carboxylation appears to be essential for enhanced virus activity. One consequence of producing acetyl-CoA or citrate from glucose is NAD⁺ consumption during glycolysis. Producing citrate and acetyl-CoA from glutamine is net neutral for NAD⁺. It is established that low cellular levels of NAD⁺ (and high levels of NADH) inhibit glycolytic flux. For this reason, rapidly proliferating cells uncouple glycolysis and OXPHOS, and use glutamine as an anaplerotic source of biosynthetic intermediates (41–44). In EnAd-infected A549 cells, we observed a rapid fall in NAD⁺ levels (Fig. 6G) to approximately 50% of control levels with no obvious fall following infection with Ad11p.

Throughout these studies, EnAd behaved in a similar way to its wild-type parent Ad11p, although the metabolic changes it caused were more pronounced, probably reflecting its rapid replication and shorter life cycle exposing the host cells to greater metabolic demands. We propose the hypothesis that virus-infected cells can produce a basal level of biosynthetic TCA intermediates from glucose, but falling NAD⁺ levels means that any excessive demand has to be met by reductive carboxylation of glutamine. EnAd may therefore experience particular benefit when the host cell has already undergone metabolic transformation, and this may contribute to its cancer selectivity. This hypothesis also raises the link between Warburg phenotypes and permissiveness to viral infection. This was confirmed through the creation of glycolysis-independent SKOV3 cells through long-term exposure to high levels of 2DG, giving a cell line with dramatically increased susceptibility to EnAd infection and replication (Supplementary Fig. S3).

This requirement for reductive carboxylation provides a good deal of phenotypic overlap between rapidly-proliferating cancer cells and cells infected with adenoviruses. Metabolic transformation and indeed reductive carboxylation are hallmarks of many cancer cells (2–4, 6, 44), and is likely to contribute to the cancer tropism exhibited by wild-type and oncolytic adenoviruses. The widespread nature of this cancer hallmark suggests that oncolytic viruses could be engineered for deficiency in their ability to hijack cellular metabolism. Finally, this phenotypic overlap between tumor cells and virus-infected cells may also be useful in helping identify which patients will be most appropriate for treatment with oncolytic viruses, because glutamine utilization can be imaged using PET based on ¹⁸F-(2S,4R). It follows that glutamine PET may provide an important noninvasive biomarker to guide patient selection in oncolytic virus clinical trials.

L.W. Seymour reports receiving commercial research grant from, has an ownership interest (including stock, patents, etc.) in, and is a consultant/advisory member for, Psioxus Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Dyer, B. Schoeps, J. Freedman, L.W. Seymour

Development of methodology: A. Dyer, B. Schoeps, P. Jakeman, E.J. Jacobus

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Dyer, B. Schoeps, S. Frost, E.M. Scott

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Dyer, B. Schoeps, S. Frost, L.W. Seymour

Writing, review, and/or revision of the manuscript: A. Dyer, B. Schoeps, S. Frost, L.W. Seymour

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Dyer, E.J. Jacobus

Study supervision: A. Dyer, L.W. Seymour

The authors gratefully acknowledge Alison Carr and Dr. Philip Miller for facilitating the use of ascites cells and the kind patients for their donation of ascites fluids. The authors are also grateful for financial support from Cancer Research UK (C552/A17720 to A. Dyer, B. Schoeps, E.M. Scott, and L.W. Seymour) and the Medical Research Council Doctoral Training Partnership (MR/N013468/1 to S. Frost; MR/K501256/1 to J. Freedman).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.