Combination of glycolysis inhibition with chemotherapy results in an antitumor immune response

Abstract

Most DNA-damaging agents are weak inducers of an anticancer immune response. Increased glycolysis is one of the best-described hallmarks of tumor cells; therefore, we investigated the impact of glycolysis inhibition, using 2-deoxyglucose (2DG), in combination with cytotoxic agents on the induction of immunogenic cell death. We demonstrated that 2DG synergized with etoposide-induced cytotoxicity and significantly increased the life span of immunocompetent mice but not immunodeficient mice. We then established that only cotreated cells induced an efficient tumor-specific T-cell activation ex vivo and that tumor antigen-specific T cells could only be isolated from cotreated animals. In addition, only when mice were immunized with cotreated dead tumor cells could they be protected (vaccinated) from a subsequent challenge using the same tumor in viable form. Finally, we demonstrated that this effect was at least partially mediated through ERp57/calreticulin exposure on the plasma membrane. These data identify that the targeting of glycolysis can convert conventional tolerogenic cancer cell death stimuli into immunogenic ones, thus creating new strategies for immunogenic chemotherapy.

Although cancer represents a heterogeneous group of diseases with a wide variety of alterations, one common feature is the ability of cancer cells to use glycolysis instead of oxidative phosphorylation for their energy production (a process called the Warburg effect) (1, 2). This observation has led researchers to develop new therapeutic strategies using glycolysis inhibitors such as the nonmetabolizable glucose analog 2-deoxyglucose (2DG) (3). 2DG acts as a competitive inhibitor of glucose transport by sharing the same transporters; it is phosphorylated by hexokinase to form 2DG–6-phosphate, which is not metabolized further in the glycolysis pathway, thereby reducing it. In addition, 2DG prevents N-glycosylation of proteins leading to an unfolded-protein response (4, 5). This compound has been widely investigated in experimental and clinical settings and has been shown to be well tolerated by patients (6). Evidence from in vivo experiments indicates that some anticancer agents favor the induction of immunogenic cancer cell death leading to tumor-specific immune responses, a mechanism described as contributing to the eradication of chemotherapy-resistant cancer cells and cancer stem cells (7, 8). However, among the chemotherapy agents used in the clinic, only a few (mainly anthracyclines and gamma radiation) are highly efficient at triggering immunogenic cell death (9). This immunogenic cell death is triggered when dying cancer cells express “danger-associated molecular patterns” (DAMPs), which include calreticulin (CRT), on the plasma membrane (for review, see ref. 10). Once exposed on the plasma membrane, DAMPs elicit the recognition and removal of dead cells by antigen-presenting cells including dendritic cells (DCs), which upon maturation can signal the immune system to react against tumor antigens.

In the present work, we sought to determine whether a combination of glycolytic inhibitors with DNA-damaging agents could enhance the benefits of the cytotoxic agent. Only a handful of molecules that target metabolic pathways have been tested as a cancer therapy. Numerous reports have indicated that inhibition of glucose metabolism could potentiate or synergize with cytotoxic drugs to induce cell death. However, we demonstrate here that this combination of compounds could also lead to the induction of an efficient anticancer immune response, thus raising the concept that tumor cell metabolism could be linked to antitumor immune response.

Results

We first sought to determine whether glycolysis inhibition could synergize with chemotherapy to induce cell death. Therefore, we used the immunocompetent, preclinical model of Eμ-Myc mice (11–13). Deregulated c-myc expression under the control of the Ig heavy chain gene enhancer (Eµ) leads to clonal pre-B-cell or immature B-cell lymphoma. The genetics and histopathology of Eµ-Myc lymphomas resemble human non-Hodgkin lymphoma. Primary lymphoma cells were isolated from Eμ-Myc mice and treated with increasing amounts of the DNA-damaging agent etoposide (ETO) in the presence or absence of 2DG. ETO was used to treat Eμ-Myc mice as it is a typical treatment for aggressive non-Hodgkin lymphomas (as Burkitt or some diffuse large B-cell lymphomas). As shown in Fig. 1A, although 2DG induces only mild levels of cell death, in consistence with our previous work (14), it enhanced ETO-induced death in a caspase-dependent manner (Fig. 1B). We then established in Fig. 1C that cotreatment with ETO and 2DG resulted in a more than additive block in proliferation (the combination index was <1) (15), which was not correlated with an increase in ETO-induced DNA damage (Fig. S1).

Fig. 1.

Glycolysis inhibition synergizes with etoposide treatment to induce cell death in vitro. (A) Primary Eμ-Myc lymphoma cells were incubated for 20 h with either the indicated amount of etoposide (empty bars) or etoposide plus 2DG (100 μg/mL; black bars). The percentage of PI-positive cells was measured by flow cytometry. The predicted additive effect is represented in gray. (B) Eμ-Myc cells were treated as indicated for 6 h with or without 100 μg/mL 2DG, and specific DEVDase activity was measured. (C) XTT-derived isobolograms representing the concentration of each product leading to the indicated effect after a 20-h incubation (the number indicated on the top right part indicates the percentage of viable cells: 40 indicates 40% of viable cells in the XTT test). Results represent the mean ± SD of three independent experiments. PI, Propidium iodide; 2DG, 2-deoxyglucose; ETO, etoposide. **P < 0.01, ***P < 0.001.

We extended this observation to in vivo experiments. Syngeneic immunocompetent wild-type C57BL/6 mice were injected (i.v.) with lymphoma cells. Upon lymph node enlargement, mice were injected (i.p.) three times a week for 3 wk with PBS, 2DG, ETO, or ETO plus 2DG, and then killed. As expected, ETO treatment reduced lymph node enlargement; however, the ETO-plus-2DG combination treatment was significantly more efficient than the other treatments (Fig. 2A). We therefore repeated the experiments to assess the effects on mouse life span in each group (Fig. 2B). All mice in the control group (PBS) were killed for ethical reasons owing to the tumor size at ∼30 d. Whereas 2DG treatment alone did not extend mouse survival (consistent with the inability of 2DG to reduce lymph node enlargement; Fig. 2A), ETO-treated animals survived up to 84 d. In addition, the life span of the cotreated animals was extended up to 147 d (a 43% increase).

Fig. 2.

Glycolysis inhibition enhances the therapeutic benefit of chemotherapy in vivo. (A) C57BL/6 mice bearing Eμ-Myc lymphomas were injected thrice a week for 3 wk as soon as the lymph nodes became palpable with 2DG (500 mg/kg), ETO (2.5 mg/kg), or the combination. Harvested lymph nodes were measured after the third week of treatment. The left panel represents the inguinal lymph node weights. Images illustrating the protective effect of the cotreatment are presented on the right panel. (Scale bar, 1 cm.) (B) Kaplan–Meier survival curves of mice treated for 3 wk as in A. Vehicle-treated mice (blue line), 2DG (brown line), etoposide alone (orange line), and the combination (green line) are shown (n = 10 mice/group). (C) Mice were treated as in A and the time of relapse was measured. 2DG, 2-deoxyglucose; ETO, etoposide. *P < 0.05, **P < 0.01. Experiments were done twice leading to similar results.

During the experimental course, we observed lymph node shrinkage to nonpalpability after ETO or ETO-plus-2DG therapy. However, although in both cases the lymph nodes became palpable again (defined as the “time of relapse”), it took an additional 13 d (a 45% increase) for cotreated animals to relapse compared with the ETO group (Fig. 2C). We therefore hypothesized that the protection provided by the cotreatment could be linked to the ability of the immune system to control tumor growth. To test it, we i.v. injected lymphoma cells in immunocompromised (NMRI nude) mice and upon lymphoma apparition treated the mice as previously (Fig. 3A). The ETO-treated group was still significantly protected, as was expected, but the ETO-plus-2DG group was not further protected, strongly suggesting that the immune system was involved in the beneficial effect of the cotreatment. Accordingly, the lymph nodes of lymphoma-bearing immunocompetent mice contained significantly more tumor-infiltrating T cells, especially cytotoxic T cells (CD8⁺), when the mice were cotreated than when they were treated with a single agent (Fig. 3 B and C).

Fig. 3.

Cotreatment leads to tumor-specific T-cell activation. (A) Nude mice were injected (i.v.) with 0.5 million Eμ-Myc cells and treated as in Fig. 2 (n = 5 for 2DG, n = 6 for ETO or ETO-plus-2DG groups). (B) C57BL/6 lymphoma-bearing mice were treated as in Fig. 2 for 1 wk. The number of T cells isolated from axillary lymph nodes was analyzed by flow cytometry (n = 9 mice/group). (C) Same as in B, measuring the number of CD8⁺ cells (n = 9 mice/group). (D) Primary Eμ-Myc lymphoma cells treated for 24 h with 2DG (100 μg/mL), etoposide (30 ng/mL), or both were incubated with C57BL/6 monocyte-derived DCs and then cocultured with naïve syngeneic T cells. Ten days later, the quantification of IFN-γ–producing T cells upon tumor antigen presentation was determined by ELISPOT. (E and F) C57BL/6 lymphoma-bearing mice were treated as in Fig. 2 for 1 wk, and then CD8⁺ cells were isolated from the spleens of three independent mice and incubated with Eμ-Myc primary cells. The ability of CD8⁺ cells to kill Eμ-Myc cells was determined using LDH measurement. The ratio of CD8⁺:Eμ-Myc cells is indicated. Results represent the mean ± SD of three independent experiments and in vivo experiments were done twice. 2DG, 2-deoxyglucose; ETO, etoposide; NT, not treated. *P < 0.05, **P < 0.01.

To test whether cotreated tumor cells could lead to the activation of tumor-specific cytotoxic T cells, we induced the ex vivo differentiation of monocytes from C57BL/6 mouse bone marrow into DCs and loaded them with syngeneic Eμ-Myc primary tumor cells that had been treated with vehicle, 2DG, ETO, or ETO plus 2DG. Then DCs were maturated using lipopolysaccharide (LPS) and incubated for 10 d with syngeneic naive T cells. The ability of those T cells to be activated in the presence of tumor cell extracts was monitored by measuring IFN-γ production, a critical indicator of cytotoxic T-cell function that is important for antitumor immune responses. We carefully calibrated the stimulation so that the same extent of cell death was induced in ETO and ETO-plus-2DG conditions. We established that DCs loaded with ETO-plus-2DG–treated Eμ-Myc tumor cells led to significantly improved activation of tumor-specific T cells compared with other treatments (Fig. 3D). Finally, to test whether tumor antigen-specific T cells could be directly isolated from cotreated animals, we treated the Eμ-Myc–bearing mice for 1 wk with ETO or ETO plus 2DG, isolated CD8⁺ T cells from the spleen, and incubated them with tumor target cells. As presented in Fig. 3 E and F, CD8⁺ cells isolated from cotreated animals were significantly more efficient in killing Eμ-Myc tumor cells than the one isolated from ETO-treated mice.

Because the cotreatment led to an activation of tumor-specific T cells (Fig. 3), we hypothesized that we should be able to protect immunocompetent mice from tumor growth by performing vaccinations. Therefore, mice were either not vaccinated (control) or were immunized on one flank with dead tumor cells and challenged on the other flank a week later with the same tumor cells in viable form. Despite several attempts, we were unable to vaccinate naïve C57BL/6 mice with syngeneic, dead Eμ-Myc cells as tumors subsequently developed quickly from the vaccination site. It has been proposed that in the Eμ-Myc model, a single cell is sufficient to establish the lymphoma in a syngeneic mouse (16), preventing us from using this mouse model for the vaccination experiment. We therefore decided to use the CT26 colon carcinoma cell line. Those cells are syngeneic with BALB/c mice and are a recurrent model for vaccination experiments (9, 17). We carefully validated that CT26 cells were recapitulating all of the characteristics demonstrated by the cotreatment with ETO and 2DG that we have described so far (i.e., enhancement of cell death by ETO plus 2DG compared with ETO alone and tumor antigen-specific T-cell activation by DCs loaded with cotreated CT26 cells; Fig. S2 A and B). We vaccinated BALB/c mice with dead CT26 cells treated with ETO, ETO plus 2DG, or mitoxantrone (MTX) [used as a positive control for immunogenic cell death (9)]. One week later, the mice were challenged on the other flank with live CT26 cells. It was notable that none of the BALB/c mice developed significant tumors on the vaccinated flank. All of the mice in the nonimmunized group developed tumors at the challenge site and died within 26 d; however, 33% of the mice in the ETO-immunized group were tumor-free and alive at 90 d postchallenge. This rate significantly increased up to 66% in the ETO-plus-2DG group. The protective effect induced by the cotreatment was similar to the protective effect observed in the positive-control group (MTX) (Fig. 4). Taken together, our findings show that (i) the combination of 2DG with ETO leads to an increase in cytotoxicity of tumors cells compared with ETO alone and (ii) to the conversion of a tolerogenic to an immunogenic form of death.

Fig. 4.

Glycolysis inhibition enhances the vaccination potential of chemotherapy. Kaplan–Meier curve of a vaccination experiment. CT26 cells were treated for 48 h with etoposide (1 μg/mL), etoposide plus 2DG (100 μg/mL), mitoxantrone (1 μM), or PBS and injected s.c. on syngeneic BALB/c mouse flanks. One week later, live cells were injected into the opposite flank, and tumor appearance was monitored over time (n = 7 for nonvaccinated group and n = 12 for other groups). 2DG, 2-deoxyglucose; ETO, etoposide; MTX, mitoxantrone; N.S., nonsignificant. ***P < 0.001. Experiments were done twice, leading to similar results.

To identify the underling mechanism, we speculated that ERp57/CRT exposure on the plasma membrane could be involved in the observed protection induced by the cotreatment with ETO and 2DG. Indeed, ERp57/CRT translocation has been described as a key event in the immunogenic cancer cell death (9). In addition, as described previously (18), the exposure of ERp57/CRT on the plasma membrane requires caspase activation and an induction of endoplasmic reticulum (ER) stress. We found that ETO-plus-2DG treatment results in an increase in caspase activation compared with ETO treatment (Fig. 1B). In addition, 2DG is known to induce ER stress (19). We verified in CT26 and in Eμ-Myc primary cells cell line that 2DG could induce eIF2α phosphorylation, a typical ER stress marker (Fig. 5A and Fig. S3A). Therefore, we assessed the presence of ecto-CRT on the surface of tumor cells. We established that, whereas ETO led to a partial ERp57/CRT exposure on the plasma membrane as shown previously (9), ETO plus 2DG led to a significant increase in its exposure (to the same extent as the positive control MTX; Fig. 5B). This observation was confirmed in several cell types (primary Eμ-Myc cells, as well as in CT26, B16, and HeLa cells) using several DNA-damaging agents (ETO and mitomycin c) combined with several hexokinase inhibitors (2DG and lonidamine; Fig. 5B and Figs. S3 and S4). We generalized further our observations by comparing the ability of two independent cell lines (B16 and CT26) to expose CRT and to induce tumor specific activation of syngeneic T-cell in response to mitomycin c with or without 2DG (Fig. S5). We confirmed a close correlation between an increase in CRT exposure and an increase in IFN-γ production in response to cotreatment (Fig. S5).

Fig. 5.

Cotreatment leads to ER stress and an increase in calreticulin exposure. (A) CT26 cells were treated with 2DG (100 μg/mL) for 16 h. The phosphorylated or the total form of eIF2α was detected by immunoblotting. (Right) Quantification of densitometric analysis of three independent experiments. (B) CRT exposure was determined by flow cytometry analysis after incubation of CT26 with the indicated agents for 24 h (100 μg/mL 2DG, 1 μg/mL etoposide, 1 μM mitomycin c, and 1 μM mitoxantrone). The results represent the mean ± SD of three independent experiments. 2DG, 2-deoxyglucose; CRT, calreticulin. *P < 0.05. NT, Not treated.

Finally, to evaluate the role of ERp57/CRT in the protective effect induced by the cotreatment we knock down ERp57 using siRNA or shRNA (Fig. S6A and Fig. 6A) and validated that it reduced CRT exposure on the plasma membrane as previously shown (Fig. S6B and Fig. 6A) (20). Then we treated CT26 cells expressing shScr or shERp57 in presence of ETO or mitomycin c with or without 2DG and incubated them with DCs derived from syngeneic BALB/c monocytes (similar to what was done in Fig. 3D). We established ex vivo that reducing CRT exposure on the plasma membrane impaired the ability of DCs loaded with the cotreated dead tumor cells to activate tumor-specific T cells (Fig. 6B), especially upon mitomycin c-plus-2DG treatment. Finally, to demonstrate that CRT exposure induced by the cotreatment was required for the induction of tumor antigen-specific T-cell activation in vivo, we repeated the vaccination assay as in Fig. 4. For that matter, CT26 cells expressing shScr or ShERp57 were killed with mitomycin c and 2DG. We chose mitomycin c treatment as it is the clinical relevant treatment for colon carcinomas and because ERp57 knockdown blunted the ability of cotreated cells to induce tumor-specific T-cell activation (Fig. 6B).

Fig. 6.

Calreticulin exposure is involved in the induction of tumor antigen-specific T-cell response. (A) CT26 stably expressing sh scramble or shERp57 (knockdown as shown by immunoblotting in Inset; ERK was used a loading control) were treated as indicated and as in Fig. 5B for 24 h, and CRT exposure at the plasma membrane was determined by flow cytometry analysis. (B) IFN-γ–producing cells were measured as in Fig. 3D using CT26 cells expressing a control shRNA (shScr) or a shRNA targeting ERp57 (shERp57) (100 μg/mL 2DG, 1 μg/mL etoposide, 1 μM mitomycin c, 1 μM mitoxantrone). (C) Vaccination experiments are as in Fig. 4 using control (shSrc) or shERp57 CT26 cells treated with mitomycin c (1 μM) plus 2DG (100 μg/mL) for 48 h (n = 12 mice/group). Experiments were done twice leading to similar results. 2DG, 2-deoxyglucose; ETO, etoposide; Mc, mitomycin c; MTX, mitoxantrone; Scr, Scramble. *P < 0.05, **P < 0.01, ***P < 0.001. The results in A and B represent the mean ± SD of three independent experiments.

As presented in Fig. 6C, mice immunized with dead ShERp57 cells were significantly less protected from subsequent tumor challenge than mice immunized with dead shScr cells (Fig. 6C), demonstrating that the ability of the cotreatment to induce tumor antigen-specific T-cell response was, at least partially, mediated through CRT exposure to the plasma membrane of the tumor cells. It is worth noting that the vaccinated mice were alive and tumor free until the completion of the experiment at 120 d. We verified that shERp57 cells did not exhibit any changes in cell proliferation compared with parental or control (shScr) cell lines, as shown in Fig. S6C.

Discussion

One of the hallmarks of cancer cells is resistance to cell death mechanisms (21). Therefore, researchers and pharmaceutical companies have focused on finding ways to restore cell sensitivity to toxic signals. However, most chemotherapeutic treatments, although effective, are not curative, and the resistance of tumor cells to cytotoxic therapies is one of the most important reasons for treatment failure. So, a major challenge is to safely restore the antitumor immune response by favoring chemotherapies (or combination of drugs) that lead to the exposure/release of DAMPs, inducing immunogenic cell death (22). However, most chemotherapy agents induce “classical” (i.e., tolerogenic) apoptosis, and the resultant host immune response is relatively weak.

Here, we provide evidence that combining glycolysis inhibition (using 2DG) with a DNA-damaging agent could synergize in vitro to induce tumor cell death and increase the life span of immunocompetent but not immunodeficient mice that had been implanted with lymphoma cells isolated from Eμ-Myc mice (Figs. 1–3). We report that this combination of agents was more efficient than either agent administrated alone at inducing specific T-cell activation in response to tumor antigens (Fig. 3D) and was more efficient at vaccinating immunocompetent mice (Fig. 4), as illustrated in Fig. S7.

To understand how this drug combination could affect the activation of tumor-specific T cells, we speculated that, among the known DAMPs, ERp57/CRT exposure on the plasma membrane could be involved. Indeed, the role of CRT in immunological functions has been known for nearly 15 y (for a review, see ref. 22). Recently, the ability of CRT to be translocated to the plasma membrane of dying cells was associated with its ability to induce an anticancer immune response (9). The molecular mechanism governing the ecto-CRT/ERp57 exposure in dying cells was described as being dependent on caspase activation and on an induction of ER stress (20). We demonstrated an increase in ecto-CRT exposure on the plasma membrane when a glycolysis inhibitor (2DG or lonidamine) was combined with a DNA-damaging agent (ETO or mitomycin c). The ability of the cotreatment to induce CRT exposure and the activation of tumor-specific T cells was confirmed in several cell lines (CT26, HeLa, and B16) and in several primary Eμ-Myc clones using several DNA-damaging agents (ETO or mitomycin c), therefore suggesting that the antitumor response could be a general mechanism. Finally, the role of ecto-ERp57/CRT exposure was validated using CT26 cell line expressing a control or an ERp57-specific shRNA. Although there were no differences in the proliferation rate or the glycolytic capacities between these two groups of cells (Fig. S6C), only the CT26 cells presenting less ERp57 led to significantly reduced levels of CRT exposure upon treatment (Fig. 6A) and a decrease in the ability to activate tumor antigen-specific T cells (Fig. 6B) and to raise specific response to vaccination (Fig. 6C). These data indicate that the increase in ecto-ERp57/CRT exposure is a key event in the induction of an anticancer immune response in this context; however, this does not exclude the importance of other potential mechanisms that remain to be found. It is notable that, as presented in Fig. 4, the vaccination of mice with ETO-treated cells led to a partial protection against tumor growth, which is consistent with the partial CRT exposure induced by this treatment (Fig. 5B), as suggested previously (9). In addition, ∼33% of the mice vaccinated with dead shERp57-treated CT26 cells were still vaccinated against tumor growth (Fig. 6C), most likely because the knockdown of protein expression was partial (Fig. 6A).

The potential clinical relevance of our work relies on the fact that the glucose analog 2DG is the most widely investigated pharmacological agent for targeting glucose metabolism (for reviews, see refs. 3 and 6). Several reports have indicated that 2DG administrated orally was safe and well tolerated by patients (23), leading to the launch of several phase II and III clinical trials in cancer therapy. Our study suggests that such cotreatment might be efficient not only by increasing the cytotoxicity of the chemotherapy but also by allowing the immune system to react against the tumor. It is interesting to note that the dose of 2DG used in our study is equivalent to the one shown to be safe in human clinical trials (23, 24).

One concern about our work showing the ability of the cotreatment to induce tumor-specific T cells might be in the theoretical ability of 2DG to prevent T-cell activation and proliferation. Indeed, it is well described that upon activation T cells rapidly switch metabolic programs from fatty acid β-oxidation and pyruvate oxidation via the tricarboxylic acid cycle to aerobic glycolysis, pentose-phosphate pathway, and glutaminolysis (25). However, several in vivo studies described that the effect of 2DG on T-cell activation and proliferation was minor in the mouse strains we used in our study (C57BL/6 and BALB/c) (26, 27), even when those animals received several injections of 2,000 mg/kg of 2DG (four times more than the classical dose used in most studies, including ours) (26). In this line, we established that functional and active tumor-specific cytotoxic T cells could be isolated from the spleen of cotreated animals (Fig. 3 E and F), suggesting that the use of 2DG in vivo does not impair T-cell functions. Also, a recent study suggested that, under normoxic conditions, the primary activity of 2DG might not be to inhibit glycolysis but rather to interfere with N-linked glycosylation, leading to ER stress (4, 28).

It is of course important to keep in mind that escape from the immune-mediated surveillance of dying cells is among the characteristics of tumor cells. It is therefore likely that, even if the cotreatment gives a benefit to the patient, an aggressive form of tumor might relapse at some point. In addition, it is always possible that the induction of an excessive immune response to dying cells would be leading to autoimmune pathology as observed in systemic lupus erythematosus. So perhaps one challenge for future investigations would be to identify the therapeutic window and the conditions in which combining glycolysis inhibition with chemotherapy will reduce the extent of the cancer to allow the newly educated immune system to destroy any remaining tumor cells.

Materials and Methods

Cell Culture and Death.

CT26, B16, and HeLa cells were obtained from ATCC and cultured as recommended.

To induce cell death, cells were treated for 24 or 48 h with the indicated amount of ETO (TRC), mitomycin c (Sigma), MTX (Sigma), 2DG (TRC), or Lonidamine (Tocris). Cells were harvested, labeled with DAPI (Molecular Probes; 0.5 µg/mL) or with propidium iodide (PI) (Sigma; 0.5 µg/mL), and analyzed immediately by flow cytometry using a MACSQuant Analyzer (Miltenyi Biotec).

For DEVDase activity measurement, see ref. 29.

Cell proliferation was measured using XTT (Roche), and the resulting data were used for quantitative diagnostic plots known as “isobolograms” as described in ref. 15.

For lactate dehydrogenase (LDH) release experiments, CD8⁺ T cells were harvested from treated mouse spleens and then sorted using autoMACS (Miltenyi Biotec). The resulting purified cells were coincubated with tumoral Eμ-Myc cells in Iscove’s modified Dulbecco’s medium (IMDM) containing 10% inactivated FCS, penicillin/streptomycin (P/S), and 50 µM β-mercaptoethanol, with 10 U/mL IL-2 for 4 h at 37 °C. LDH release was then measured following the manufacturer’s instructions (Roche).

Detection of Ecto-CRT and γH2AX (Ser¹³⁹).

Cells were fixed at 4 °C with 0.25% paraformaldehyde in PBS, washed, and stained with rabbit anti-mouse CRT antibody (1:200; Abcam) for 30 min at 4 °C in 2% FCS PBS. For γH2AX, cells were fixed, permeabilized using 50 µg/mL digitonine, and stained using Phospho-Histone γH2AX (Ser¹³⁹) (1:400; Cell Signaling). In both cases, cells were incubated with anti-rabbit Alexa Fluor 488-conjugated secondary antibody (1:500; Molecular Probes), and the pellet was resuspended in PBS buffer with DAPI (0.5 µg/mL).

The samples were then analyzed by MACSQuant Analyzer (Miltenyi Biotec) and checked for CRT staining on DAPI-negative cells. Results are presented minus the value of the untreated condition.

Western Blot Analysis.

After treatment, cells were collected, washed, and lysed as described previously (29). Anti-Erk2, and -Hsp90 were purchased from Santa Cruz. Anti-ERp57 was from Abcam. Anti-phospho and total eIF2α were purchased from Cell Signaling Technology.

When indicated, Western blot quantification was made using either ImageJ or MultiGauge software.

RNA Interference and Cell Transduction.

Specific anti-ERp57 or scramble oligonucleotides were cloned in pSUPER-GFP-neo vector (Oligoengine) according to the manufacturer’s protocol (sequences are available upon request).

Transduction was made as previously described in ref. 30. Forty-eight hours later, infected cell populations were sorted by flow cytometer BD Aria for GFP⁺ cells, and expression of the targeted proteins was assessed by Western blot.

Transgenic Mice and Transplantation of Lymphomas.

All mice were maintained in specific pathogen-free conditions, and experimental procedures were approved by the Institutional Animal Care and Use Committee and by the regional ethics committee (approval references NCA/2007/11-04 from Comité Régional d'Éthique en matière d'Expérimentation Animale and NCE/2011-35 from Comité Institutionnel d'Éthique Pour l'Animal de Laboratoire - AZUR).

C57BL/6 Eμ-Myc transgenic mice were purchased from The Jackson Laboratory and were genotyped by PCR according to instructions from the supplier. Lymphoma-bearing animals were killed by cervical dislocation as soon as they presented signs of illness. A single-cell suspension was prepared from lymph nodes by teasing them on a 70-µm nylon filter. Cells were either resuspended in complete medium (DMEM supplemented with 10% FCS, 10 mM Hepes, P/S, 0.1 mM l-asparagine, and 50 µM 2-mercaptoethanol) for further ex vivo analysis or directly reimplanted in wild-type mice.

Lymphoma transfers were realized into syngeneic, nontransgenic, 6- to 10-wk-old C57BL/6 females by tail vein injection (0.5 × 10⁶ viable cells per mouse, in 200 µL of sterile PBS). As the inguinal lymph nodes became “well-palpable” (reaching around 5 mm in the longest diameter), the animals were i.p. injected, three times a week, unless stated otherwise, for the indicated period, with PBS, ETO (2.5 mg/kg), 2DG (500 mg/kg), or both compounds, and subsequently monitored for treatment response. The “time of relapse” reflects the time between remission and repalpability of a recurrent lymph node enlargement.

In some experiments, athymic NMRI nude mice were used instead of C57BL/6: injected i.v. with Eμ-Myc lymphoma cells (0.5 × 10⁶ viable cells per mouse) and treated as described above.

Antitumor Vaccination.

Eight-week-old BALB/c female mice were injected s.c. on the left flank with 3 × 10⁶ treated CT26 cells (killed either with MTX, ETO, or with 2DG and ETO). One week after, these mice were challenged by injecting 0.5 × 10⁶ live cells on the contralateral flank. The animals were then checked twice a week for tumor development using calipers.

DC Preparation and IFN-γ Measurement.

Bone marrow cells were flushed from femurs and tibias of young BALB/c or C57BL/6 wild-type mice, and monocytes were sorted using autoMACS (Miltenyi Biotec) and incubated with murine GM-CSF (Miltenyi Biotec; 100 ng/mL) and IL-4 (AbD Serotec; 100 ng/mL) for 1 wk in IMDM containing 10% inactivated FCS, P/S, and 50 µM β-mercaptoethanol. Immature DCs were then fed with tumor cells at a DC/tumor cell ratio of 1:1. Pulsed DCs were then stimulated with 100 ng/mL LPS (Sigma) for 24 h. T cells were purified sorted using autoMACS (Miltenyi Biotec) from spleen of naïve syngeneic mice and added to unpulsed or tumor-loaded DCs at a ratio of 1:20, in the presence of IL-2 (AbD Serotec; 0.1 ng/mL). Cocultures were tested for the presence of IFN-γ–producing T cells 10 d after the stimulation with DCs using an enzyme-linked immunospot (ELISPOT) assay (BD Biosciences) on T cells. Results are presented minus the value of the untreated condition.

Statistical Methods.

Data are expressed as mean ± SD. Differences in calculated means between groups were assessed by two-sided Student t tests. For experiments involving more than two groups, one-way ANOVA was performed and, in cases in which significant differences were detected, followed by pairwise t tests as indicated. Tumor-free survival time was calculated as the time from s.c. tumor cells injection until the first occurrence of a tumor volume of 1,500 mm³, death from any cause, or until 90-d (or 120-d) follow-up without any clinical event. Overall survival was the time from i.v. lymphoma cells injection until death. Kaplan–Meier survival analyses were performed using StatView 5.0 (Abacus Concepts), and comparisons of survival curves were made using two-sided log-rank tests. A value of P < 0.05 was considered significant.