The bioenergetic signature of isogenic colon cancer cells predicts the cell death response to treatment with 3-bromopyruvate, iodoacetate or 5-fluorouracil

Colorectal cancer (CRC) is a common neoplasia which poses a heavy burden on public health systems worldwide [1]. Despite the establishment of CRC screening protocols, tailored therapeutic approaches are required to minimize the significant social impact of this disease [1]. At present, KRAS mutation status is the only validated predictive marker for targeted CRC therapy [2]. Thus, the development and clinical implementation of new predictive molecular markers are needed to aid in the selection of patients likely to respond to therapy and rationalized CRC treatments [2].

Cancer cells and tumors have a predominant glycolytic metabolism, even under aerobic conditions [3, 4]. Although the altered energetic metabolism of cancer cells has been proposed as a potential target for cancer treatment [3, 5‐7], it could also represent a therapeutic obstacle, because of its contribution to chemo- and radio-resistance [8]. In some tumors, this glycolytic phenotype is accompanied by a loss of bioenergetic activity in mitochondria [9, 10], which can be estimated by determining its bioenergetic signature[10, 11]. The bioenergetic signature is a protein ratio (β-F1-ATPase/GAPDH ratio) that assesses the expression of the catalytic subunit of mitochondrial H⁺-ATP synthase (β-F1-ATPase), a bottle-neck component required for the synthesis of biological energy, relative to the expression of glycolytic glyceraldehyde-3-phosphate dehydrogenase (GAPDH) [10]. Consistently, the bioenergetic signature has been observed to be significantly down-regulated in different human tumors compared to paired normal tissues [10, 12‐19]. Recent findings indicate that the bioenergetic signature also represents a functional index of metabolic activity because it correlates, both in vivo and in vitro, with the rate of glucose utilization by cancer cells and tumors [9, 11]. Moreover, according to large cohort studies of colon [10, 19], lung [9, 14] and breast [16, 20] cancer patients, low tumor bioenergetic signatures are associated with poor patient prognosis, strongly suggesting that impaired mitochondrial bioenergetics is at the heart of cancer progression.

Remarkably, down-regulation of β-F1-ATPase has been widely associated with the resistance of cancer cells to standard anticancer therapies [21‐23]. In the specific case of colon cancer cells, chemotherapeutic response to 5-fluorouracil (5-FU) [11, 21], as well as several metabolic inhibitors [23, 24], was assessed in cells with different genetic backgrounds: a condition that is likely to affect the cellular response to chemotherapeutic agents. The recent development of isogenic HCT116 colon cancer cell lines, representing different bioenergetic signatures[11], has provided an opportunity to unambiguously assess the influence of energetic metabolism on colon cancer therapy. In this study, we investigated cell death responses in metabolically different isogenic HCT116 cells and the regression of tumor xenographs, in response to the glycolytic inhibitors 3-bromopyruvate (3BrP) and iodoacetate (IA), and the classic chemotherapeutic agent, 5-FU. The small alkylating 3BrP and IA target the enzymes of glycolysis hexokinase [25] and GAPDH [26], respectively, although recent findings suggest that 3BrP also targets GAPDH [27].

Because of the regulated expression of β-F1-ATPase, development of HCT116 colon cancer cell lines, displaying low (G-cells), medium (M-cells) or high (SM-cells) bioenergetic signatures (see Figure 1, and additional file 1) was accomplished by modification of cell culture conditions [11]. As recently detailed [11], the bioenergetic signature of each cell line was found to inversely correlate with the rate of aerobic glycolysis, where G-cells > M-cells > SM-cells (Figure 1A-C). Evaluation of cell death responses were assessed using fluorescence microscopy after double labeling with Hoechst 33342 and propidium iodide (PI) (Figure 1A-C). Our results show that death responses (% PI positive cells) to both metabolic inhibitors (3BrP and IA) decreased as the bioenergetic signature of the cells increased. Thus, the lower the bioenergetic signature of a cell the greater the death response to the glycolytic inhibitor treatment (G > M > SM) (Figure 1D). In fact, significant inverse correlations were uncovered between the bioenergetic signature of a cell and the extent of cell death following 3BrP (R = -0.633; n = 36, P < 0.01) and IA (R = -0.616; n = 36, P < 0.01) treatment, supporting the relevance of these glycolytic inhibitors in cancer treatment [7, 23]. Specifically, in M-cells, 3BrP treatment was more effective than IA treatment at triggering cell death (Figure 1). In contrast, cell death in response to 5-FU treatment was found to directly correlate with bioenergetic signature (R = 0.519; n = 27, P < 0.01) (Figure 1D): as the activity of aerobic glycolysis is diminished cell death in response to 5-FU treatment is augmented (SM-cells > M-cells > G-cells), suggesting the participation of mitochondrial oxidative phosphorylation in the mechanism of 5-FU mediated cell death.

Flow cytometric analysis of plasma membrane exposure of phosphatidylserine (detected using an annexin V-FITC assay) was used as an index of apoptotic (annexin-positive) versus necrotic (PI positive) cell death [28] (Figure 2). This approach enables simultaneous estimation of the cell death pathway preferentially induced by each type of treatment. Upon treatment with the metabolic inhibitors 3BrP and IA, G-, M- and SM-cells all display a very large increase in the percentage of PI-positive cells (coupled with the absence of relevant changes in the percentage of annexin-positive cells) and thus appear to die by necrosis (Figure 2) [29, 30]. In agreement with these results, no activation of caspase 3 (an apoptotic indicator) was observed following any of these above treatments in any of the cell lines tested (data not shown). Therefore, inhibition of the activity of glycolytic enzymes appears to trigger necrotic cell death. Furthermore, this effect was observed to be more pronounced in cells that rely more heavily on glycolysis as a pathway for energy provision. In contrast, 5-FU treatment of G-, M- and especially SM-cells resulted in a significant percentage of annexin-positive stained cells compared to controls (Figure 2), suggesting induction of apoptosis in response to 5-FU treatment. Importantly, this induction of apoptosis following 5-FU treatment appears to be more pronounced in cells that rely less heavily on glycolysis. In agreement with this finding, caspase 3 activity was found to be significantly increased in 5-FU treated SM-cells (1.0 ± 0.2 vs. 2.2 ± 0.1 a.u./15000 cells for control and 5-FU treated cells, respectively, P < 0.05). Therefore, efficient activation of apoptosis following 5-FU treatment may be associated with cellular reliance on mitochondrial energetic metabolism for cellular energy provision, in agreement with previous reports [11, 21, 31].

To further confirm the cell death pathway activated in response to each of the treatments studied cellular ATP concentrations were determined (Figure 3). We observed that treatment of cells with 3BrP or IA was associated with a very large depletion of cellular ATP concentrations in all cell lineages (Figure 3), consistent with activation of necrosis by metabolic catastrophe in response to treatment with these metabolic inhibitors. In contrast, treatment of cells with 5-FU only marginally affected cellular ATP concentrations in G- and M-cells (Figure 3) and slightly, but significantly, promoted a 50% reduction in cellular ATP concentrations in SM-cells (Figure 3), indicating the absence of a compromised metabolic state following 5-FU treatment.

Previous findings have suggested that only highly glycolytic G-cells are able to develop tumors in nude mice [11]. Therefore, in order to test the in vivo tumor regression activity of the metabolic inhibitors analyzed in vitro, animals were implanted with G-cells. Animals that developed ~ 1 cm³ tumors were treated with daily doses of 8 μM 3BrP, 100 μM IA or 0.5 mM 5-FU over six consecutive days (Figure 4A). A control NaCl-treated group was also included for comparison (Figure 4A). Interestingly, from both a macroscopic (Figure 4B) and behavioral point of view, all treatments tested (except controls) seemed to affect the mice in a similar manner. Specifically, control animals developed a rapid 2.5-fold increase in tumor volume during the treatment period (Figure 4A). In contrast, animals treated with either 5-FU or IA revealed a significant ~ 30% decrease in tumor volume after 6 days of treatment (Figure 4A), while maximum tumor regression (> 50%) was observed in mice treated with 3BrP (Figure 4A), consistent with the higher cell-death trend associated with 3BrP treatment in vitro (Figure 1). However, the large differences in cell death triggered by 3BrP and IA compared to 5-FU in G-type cells (Figure 1A) were largely absent following in vivo treatments despite the fact that the tumors had a G-phenotype [11]. These results suggest that additional mechanisms may play a role in promoting tumor regression in vivo and that in vitro data should be extrapolated with caution.

In an effort to translate the bioenergetic signature to clinical practice, we have recently developed monoclonal antibodies against various markers of energetic metabolism [32]. We found that cancer abolishes cell-type specific differences in the bioenergetic signature[32], supporting its use as a generic target to combat different neoplasias [32]. Indeed, β-F1-ATPase expression has been shown to be a therapeutic response marker in different cancer cell lines, both for single and combined chemotherapy [19, 21‐24, 33]. In the present study, we document the correlation between the bioenergetic signature of a cell, which represents an index of the relative relevance of cellular energy provision pathways [9, 11], and the potential to execute cell death in response to the metabolic inhibitors, 3BrP and IA, and the anti-metabolite 5-FU. The correlations observed in this study cannot be ascribed to differences in the genetic background of the cells because: (i) all of the cells were derived from the same parental HCT116 cells and (ii) the energetic metabolism of HCT116 cells is a reversible phenotypic trait amenable to regulation [11, 34]. Furthermore, although some cancer cells can oxidize glutamine for energy production purposes [3‐5], glutamine contributes very little to the energetic metabolism of the highly glycolytic HCT116 cells used in this study. In fact, oxygen consumption rates, aerobic glycolysis rates and the bioenergetic signature of HCT116 cells are not affected by the presence of glutamine in the culture medium (see additional file 2).

Mechanistically, we propose that the cell death and tumor regression observed following administration of glycolytic inhibitors (3BrP and IA) may be due to induction of necrosis, whereas the cell death activity observed upon 5-FU treatment may occur through apoptosis (Figures 2 and 3). This later finding is consistent with the relevant roles played by oxidative phosphorylation [35] and mitochondrial H⁺-ATP synthase activity [33, 36] in the efficient execution of cell death. Indeed, the bioenergetic activity of mitochondria in colon cancer cells [11, 21] and tumors [19], has been associated with the ability to execute a ROS-mediated cell death response upon 5-FU treatment [11].

On the other hand, small alkylating agents have been shown to be able to kill cancer cells resistant to apoptosis by a process known as "programmed necrosis" through depletion of NAD⁺ via PARP1 activation [30]. However, the induction of necrosis in response to the glycolytic inhibitors 3BrP and IA is exerted independently of PARP1 processing (data not shown), and most likely results from a metabolic catastrophe due to cellular ATP depletion (Figure 3) [23, 24]. Overall, our studies suggest that the enzymes of glycolysis could represent therapeutic targets for the treatment of colon cancer that may be as effective as conventional treatments (5-FU) at promoting tumor regression, in agreement with findings by others [25, 37].

The use of glycolytic inhibitors as chemotherapeutic agents has pros and contras. One problem is the deleterious effects that these agents could trigger in cell types strictly dependent on aerobic glycolysis for energy (e.g. neurons, lymphocytes, erythrocytes, retina, renal medulla, etc). However, glycolytic enzymes do have highly specific active site residues that, in principle, could provide more specific drug targets than those of proteins involved in signal transduction pathways. Thus, the use of such inhibitors may be beneficial in combination therapy as enhancers of the action of current chemotherapeutic drugs [7, 23]. Targeting energetic metabolism might represent an alternative cancer treatment route in the near future, because tumor cells that are resistant to chemotherapy could effectively die by necrosis in response to different metabolic inhibitors. Whatever the case, the bioenergetic signature offers a reliable gauge to predict the cell death response (apoptotic or necrotic cell death) to cancer therapy.

In summary, we have demonstrated that the bioenergetic signature of colon cancer cells inversely correlates with the potential to execute necrosis in response to treatments with glycolytic inhibitors. In contrast, the bioenergetic signature directly correlates with the apoptotic response to 5-FU treatment. Overall, our results support the use of the bioenergetic signature as a gauge for predicting cell death in response to different therapeutic strategies in colon cancer.