GAPDH Enhances the Aggressiveness and the Vascularization of Non-Hodgkin’S B Lymphomas Via NF-Κb-Dependent Induction of HIF-1Α

Home , Enolase, PFKFB3, PKM2

ORIGINAL ARTICLE GAPDH enhances the aggressiveness and the vascularization of non-Hodgkin’s B lymphomas via NF-κB-dependent induction of HIF-1α

J Chiche1,2, S Pommier1,2,3, M Beneteau1,2, L Mondragón1,2, O Meynet1,2, B Zunino1,2, A Mouchotte1,2, E Verhoeyen1,2, M Guyot2,4, G Pagès2,4, N Mounier5, V Imbert2,6, P Colosetti2,7, D Goncalvès2,7, S Marchetti2,7, J Brière8, M Carles1,2,3, C Thieblemont8 and J-E Ricci1,2,3

Deregulated expression of glycolytic enzymes contributes not only to the increased energy demands of transformed cells but also has non-glycolytic roles in tumors. However, the contribution of these non-glycolytic functions in tumor progression remains poorly deﬁned. Here, we show that elevated expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), but not of other glycolytic enzymes tested, increased aggressiveness and vascularization of non-Hodgkin’s lymphoma. Elevated GAPDH expression was found to promote nuclear factor-κB (NF-κB) activation via binding to tumor necrosis factor receptor-associated factor-2 (TRAF2), enhancing the transcription and the activity of hypoxia-inducing factor-1α (HIF-1α). Consistent with this, inactive mutants of GAPDH failed to bind TRAF2, enhance HIF-1 activity or promote lymphomagenesis. Furthermore, elevated expression of gapdh mRNA in biopsies from diffuse large B-cell non-Hodgkin’s lymphoma patients correlated with high levels of hif-1α, vegf-a, nfkbia mRNA and CD31 staining. Collectively, these data indicate that deregulated GAPDH expression promotes NF-κB-dependent induction of HIF-1α and has a key role in lymphoma vascularization and aggressiveness.

Leukemia (2015) 29, 1163–1176; doi:10.1038/leu.2014.324

INTRODUCTION glyceraldehyde-3-phosphate dehydrogenate (GAPDH) have linked Although cancer is a heterogeneous group of diseases with glycolytic enzymes to the control of cell growth and modifications 6–10 varying characteristics specific to each type of tumor, there are of signaling pathways. However, the contribution of these criteria common to all of these tumors, such as metabolic changes. functions to tumorigenesis in vivo remains poorly defined. Tumors in growth phase have a metabolic demand that exceeds The glycolytic pathway involves the conversion of glucose to the reduced oxygen tension (hypoxia)/nutrients they receive until lactate and the generation of ATP via a series of 10 enzymes. One new vessels are formed (angiogenesis). Low pressure of oxygen of them, GAPDH, which catalyzes the reaction of glyceraldehyde- + will stabilize the hypoxia-inducing factor-1α (HIF-1α) that will lead 3-phosphate (G3P)+NAD +Pi into 1,3-diphosphoglycerate+NADH+ + to the expression of genes involved in cell survival and adaptation, H , is a key enzyme of this metabolic pathway. Although GAPDH including vascular endothelial growth factor (VEGF) that will favor has long been considered to be an enzyme of seemingly little angiogenesis during cancer progression.1 Therefore, VEFG signal- interest, recent studies have demonstrated that in addition to its ing has emerged as a clinical attractive strategy in oncology. well-characterized glycolytic role in energy production, GAPDH is a However, insufficient efficacy and resistance limit its success.2 multifunctional protein with numerous significant non-glycolytic – Fundamental metabolic changes, controlled in part by HIF-1, are functions.7,10 12 Although the role of GAPDH in tumorigenesis is implemented by tumor cells to meet their high-energy still unclear, some studies have linked high levels of gapdh mRNA requirements.3 This increased dependence for glycolysis (Warburg to a poor prognosis, specifically in breast cancer and lung effect) that helps to meet the energy needs and to facilitate the cancer.13,14 However, how GAPDH expression could be related to uptake and incorporation of nutrients into the biomass of tumor aggressiveness is still elusive. developing tumors is a common feature of malignant tumors The present study was conducted to investigate whether and is hypothesized to be a key step in tumorigenesis.4,5 specific glycolytic enzymes could play a role in tumor aggressive- Investigation into the mechanisms driving the tumor metabolic ness. This study was performed using the Eμ-Myc transgenic shift has created interest in the identification of new functions of mouse model of non-Hodgkin’s lymphoma15 and results were glycolytic enzymes involved in cancer progression, whatever they confirmed using non-Hodgkin’s lymphoma patient biopsies. We are or not related to their enzymatic properties. Recently, the revealed that only GAPDH, but not other tested glycolytic non-metabolic activities of 6-phosphofructo-2-kinase/fructose-2,6- enzymes, contributes to an increase in lymphoma growth and biphosphatase 3, hexokinase II, pyruvate kinase M2 (PKM2) and vascularization. Mechanistic investigations revealed that GAPDH

1Inserm, U1065, Centre Méditerranéen de Médecine Moléculaire (C3M), équipe ‘contrôle métabolique des morts cellulaires’, équipe 3, Nice, France; 2Université de Nice-Sophia- Antipolis, Faculté de Médecine, Nice, France; 3Centre Hospitalier Universitaire de Nice, Département d’Anesthésie Réanimation, Nice, France; 4Institute for Research on Cancer and Aging, CNRS UMR 7284/U INSERM 1081, Centre A. Lacassagne, Nice, France; 5Centre Hospitalier Universitaire de Nice, Département d’Onco-Hématologie, Nice, France.; 6Inserm, U1065, Centre Méditerranéen de Médecine Moléculaire (C3M), équipe ‘inflammation, cancer et cellules souches cancéreuses’, Nice, France; 7Inserm, U1065, Centre Méditerranéen de Médecine Moléculaire (C3M), équipe ‘mort cellulaire, différenciation, inflammation et cancer’, Nice, France and 8AP-HP-Hôpital Saint-Louis, Service d’hémato- Oncologie, Université Paris Diderot, Sorbonne Paris Cité, F-75010 Paris, France. Correspondence: Dr J-E Ricci, Inserm U1065, Centre Méditerranéen de Médecine Moléculaire (C3M), équipe ‘contrôle métabolique des morts cellulaires’, équipe 3, 151 route de St Antoine de Ginestière, Batiment Archimed, BP 23194, Nice 06204, France. E-mail: [email protected] Received 3 July 2014; revised 4 November 2014; accepted 10 November 2014; accepted article preview online 14 November 2014; ; advance online publication, 16 December 2014 GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1164 regulates HIF-1α by mediating nuclear factor-κB (NF-κB) activation, (Promega, Madison, WI, USA). The luciferase assay was performed as a process driven by the interaction of GAPDH with tumor necrosis previously described.10 factor receptor-associated factor-2 (TRAF2). NF-κB activity. HeLa cells stably expressing GAPDH-V5 or GAPDH double mutant (DM-V5) control (pMIG) were transiently co-transfected with a MATERIALS AND METHODS vector-encoding cyan fluorescent protein (0.5 μg) and a luciferase reporter μ κ Cell culture and hypoxic exposure gene (2 g) controlled by a minimal tk promoter and six reiterated B sites (κBx6 tk luc). Twenty-four hours after transfection, cells were exposed to HeLa cells were obtained from American Type Culture Collection and either normoxia or hypoxia for 24 h before analyzing by fluorescence- cultured as recommended. Mouse primary Eμ-Myc lymphomas were activated cell sorting for the transfection efficiency and harvesting and 16 isolated as described previously. Incubation in hypoxia was carried out at analyzing as previously described.19 For NF-κB activity in mouse primary B 1% O2 (as opposed to normoxic, 21% O2 incubation), 37 °C in 95% lymphomas, Eμ-Myc cells were stably transduced with the κB-luciferase- 20 19 humidity and 5% CO2/94% N2 in a sealed anaerobic workstation (Whitley IRES-GFP lentiviral construct and analyzed as previously described. hypoxystation H35, Shipley, UK). Patients and tissue sample preparation Reagents and antibodies A total of 13 patients who underwent biopsies at the diagnosis stage for Mouse anti-V5 was from Invitrogen (Carlsbad, CA, USA), rabbit anti-GAPDH diffuse large B-cell lymphoma (DLBCL) between May 2007 and May 2011 at was from (Abcam, Cambridge, UK), mouse anti-Erk2 was from Santa Cruz the Saint-Louis Hospital (Department of Onco-Hematology, Hospital Saint- Biotechnology (Santa Cruz, CA, USA). Mouse anti-GAPDH used for Louis, Paris, France) were selected. The patients received the necessary immunoprecipitation was from Santa Cruz Biotechnology. Rabbit anti- information concerning the study and consent was obtained. The main GAPDH used for immunohistochemistry was a ‘Prestige antibody’ from clinical and histopathological data of the patients are summarized in Sigma (St Louis, MO, USA). Rabbit anti-HIF-1α was a gift of Dr Pouysségur. Supplementary Table 1. The morphologic classification of the tumors was 21 Anti-CD19-PE was from BD Bioscience (Franklin Lakes, NJ, USA), Anti-CD31 carried out according to the World Health Organization criteria. Ann Arbor stage was assigned to the tumor extensions, and all patients were was from BD Bioscience. Other antibodies were from Cell Signaling 22 23 Technology (Beverly, MA, USA). GAPDH-specific inhibitor, Koningic acid scored by IPIaa. RNA was extracted as previously described. (KA), was from Euromedex (Souffelweyersheim, France). 2-Deoxyglucose was from Toronto Research Chemicals (Toronto, ON, Canada). Statistical analysis Pimonidazole- and HP-Red549-labeled anti-pimonidazole were purchased Continuous variables and binomial variables, expressed as mean (s.d.), from Hypoxyprobe Inc. (Burlington, MA, USA). were analyzed with Student's t-test or one-way analysis of variance and compared, respectively, between groups with Fisher’s exact test or χ2 test. Plasmids Continuous variables (proliferation index), expressed as mean (s.d.) after 7 log transformation, were analyzed with linear regression and compared GAPDH plasmids were described previously. Using a cDNA library and between groups with Fisher’s exact test. For time-to-event variables, the following classical methods, ENO1 (NP_001419) and PKM2 (NP_872270) survival functions were estimated with Kaplan–Meier method and were cloned by PCR and transferred into a pMIG-GFP viral vector. compared with log-rank tests. All statistical analyses were done with R The sequence of short hairpin RNA (shRNA)-gapdh targeting human project software (version 2.15.1). AP-value of less than 0.05 was considered (shgapdh (h)) and mouse gapdh (shgapdh (m)) are provided upon request. to indicate statistical significance (*Po0.05, **Po0.01 and ***Po0.001). Complementary sense and antisense oligonucleotides were annealed into Continuous biologic variables were dichotomized by applying the standard BglII/HindIII–cut pSUPER retro.Neo+GFP vector (oligoengine). shRNA- split-sample approach to determine the best cutoff using an unbiased targeting luciferase was used as a control shRNA (shctl),17 excepted for method. The resulting thresholds were checked by including cubic luciferase assay (empty pSUPER retro.Neo+GFP vector was used). smoothing splines in the risk function of the Cox model. A spline curve was used to determine the best cutoff point to discriminate the DBCL ‘low’ ‘ ’ RNA extraction and real-time quantitative PCR and high expressers of gapdh. Total RNA was extracted from cells using the RNA extraction kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. After RESULTS reverse transcription-PCR, the relative mRNA expression level of gapdh, hif-1a, ca9, vegf-a, ldh-a (mouse and human) were obtained by real-time Increase in GAPDH expression accelerates c-Myc-driven quantification PCR, using the TaqMan PCR Master Mix (Eurogentec, lymphomagenesis Seraing, Belgium) and TaqMan assay primer set (Applied Biosystems, By monitoring the overall survival of Eμ-Myc mice that sponta- 15 Foster City, CA, USA) on the 7500 Fast and the Step One (Applied neously develop clonal B-cell lymphoma, we observed that most Biosystems) according to the manufacturer’s instructions. For in vitro of them could be regrouped in two categories: those mice that experiment, all samples were normalized to rplp0. All mRNA samples from developed highly aggressive lymphomas (‘high,’ median survival human tumor tissues were normalized by ppia (cyclophilin-a). o11 weeks), and those mice that developed less aggressive lymphomas (‘low,’ median survival 420 weeks; Figure 1a). Lactate measurement Strikingly, we found that the highly aggressive lymphomas fl Intact lymph nodes (LNs) were dissociated in a lysis buffer A (see consistently displayed higher glycolytic ux than did the less Supplementary Legends). Lactate concentration was determined by an aggressive ones, as they produce significantly more lactate, the enzyme-based assay using 900 μM β-NAD (Sigma), 175 μg/ml L-lactate end product of glycolysis (Figure 1b). Surprisingly, when we dehydrogenase (Sigma) and 100 μg/ml glutamate-pyruvate transaminase monitored the protein expression of several glycolytic enzymes, (Roche, Basel, Switzerland) diluted in a sodium carbonate (620 mM)-L- including 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 24 glutamate (79 mM) buffer adjusted to pH 10 and was then normalized by 3, only one enzyme, GAPDH, was significantly overexpressed the total protein. Lithium lactate was used as a standard. in ‘highly’ aggressive lymphomas compared with the ‘low’ aggressiveness group (Figure 1c and Supplementary Figure 1A, Luciferase assays B), as confirmed by a higher GAPDH activity (Supplementary HIF-1 activity. HeLa cells stably expressing the p3HRE-Dptk-LUC vector, Figure 1C). Of note, the level of gapdh mRNA observed in LNs of which contain three copies of the hypoxia-responsive element (HRE) from ‘low’ aggressive lymphomas was equivalent to the one observed the erythropoietin gene,18 were transiently transfected with control vector in the LNs of wild-type healthy littermates (data not shown). (pMIG or empty pSUPER vector) or vectors (2 μg) to either overexpress or Some tumors in Eμ-Myc have either a disruption of the p19ARF- to silence GAPDH. The following day, transfected cells were exposed to MDM2-p53 pathway or an overexpression of the anti-apoptotic normoxia or hypoxia for 24 h before cell lysis in a reporter lysis buffer proteins BCL-2. It is worth mentioning that we did not find any

Survival (%) aggressiveness: Lactate in LN 20

low (10) (mM/mg of protein) high (15) 0 0 low high 0 5 10 15 20 25 30 (10) (10) Weeks after birth aggressiveness

low high : aggressiveness

: Mice # 3.0 ** c-Myc 2.5

p53 2.0

PFKFB3 1.5

GAPDH 1.0

ENO1 0.5 Relative quantification of GAPDH level (GAPDH/Erk2) PKM2 0

Erk2 low high aggressiveness

100 pMIG (10) GAPDH-V5 (10) 80 PKM2-V5 (10) ENO1-V5 (10) Bcl-xL (10) 60 *** ns 40 * Survival (%)

0 15 20 25 30 35 40 Days post-iv Figure 1. GAPDH expression accelerates Eμ-Myc lymphomagenesis. (a) Kaplan–Meier curves of transgenic Eμ-Myc mice showing a group of mice displaying accelerated lymphomagenesis, named ‘high’ aggressive lymphomas (n = 15), compared with the ‘low’ aggressive lymphomas (n = 10). (b) Measurement of lactate within axillary lymph node tumors of transgenic Eμ-Myc mice presented in a (n = 10 per group). (c) Whole- cell lysates prepared from axillary lymph node tumors of transgenic Eμ-Myc mice presented in a were analyzed by immunoblots with the indicated antibodies. Each lane represents an independent mouse. Erk2 is used as a loading control. Right panel: Quantification of GAPDH expression levels relative to Erk2 expression between the two groups of transgenic mice. (d) Kaplan–Meier curves of syngenic C57BL/6 mice intravenously injected with primary Eμ-Myc cells transduced with retroviral green fluorescent protein (GFP)-encoding constructs to stably express either a V5-tagged-glyceraldehyde-3-phosphate dehydrogenase (GAPDH-V5) or -enolase-1 (ENO1-V5) or -pyruvate kinase M2 (PKM2- V5) or Bcl-xL, compared with control vector (pMIG;). N = 10 per group (*Po0.05, **Po0.01 and ***Po0.001, ns ¼ not significant).

correlation between GAPDH and BCL-2 or p53 expression such as hypoxia and glucose deprivation.25 We did not observe a (Figure 1c and Supplementary Figure A1). In the same line, MYC correlation between GAPDH and MYC expression (Figure 1c). is a well-known master regulator of the re-routing of major To assess the speciﬁc role of GAPDH in lymphoma aggressive- catabolism and anabolism nutrients in response to conditions ness, primary Eμ-Myc lymphoma cells were transduced to

80 3

60 2 g of protein) l/ weight (mg) 40

g/ 1 (axillary lymph nodes)) (axillary lymph 20 ( Hemoglobin within LN ( 0 0 pMIG pMIG ENO1-V5 PKM2-V5 GAPDH-V5 GAPDH-V5

de * * CD31/DAPI (X10) pMIG GAPDH-V5 2 mRNA

1.5 vegf-a

0.5

Fold induction of 0 pMIG ENO1-V5 PKM2-V5 GAPDH-V5

f p=0,004 1.5 2 1

0.5 vessels/mm

Number of CD31 positive 0 pMIG GAPDH-V5 Figure 2. GAPDH expression promotes a vascularized phenotype that is associated with an induction of vegf-a mRNA in vivo.(a) Mouse primary Eμ-Myc cells stably expressing a control pMIG, GAPDH-V5, ENO1-V5, PKM2-V5 or Bcl-xL constructs were intravenously (i.v) injected into wild-type syngenic C57BL/6 mice. Fifteen days post i.v. injection, the axillary LN tumors were harvested and weighted (n = 8 per group). (b) Illustration of the phenotype observed in axillary LN 15 days following injection of Eμ-Myc cells overexpressing GAPDH-V5 compared with control Eμ-Myc cells (pMIG). (c) Hemoglobin quantification within axillary LN presented in b (n = 4 per group). (d) Relative mRNA levels of vegf- a in axillary LN tumors harvested as in a were determined by quantification PCR and normalized by the level of rplp0 (n = 4 per group). (e) CD31 immunofluorescence of LN tumors harvested as in a. Fifteen days after i.v. injection of Eμ-Myc cells overexpressing GAPDH (GAPDH-V5) or not (pMIG). (f) Quantification of CD31-positive vessels/mm2 of LN tumors presented in e (n = 6 per group) (*Po0.05, **Po0.01).

Leukemia (2015) 1163 – 1176 © 2015 Macmillan Publishers Limited GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1167 overexpress either GAPDH, the anti-apoptotic Bcl-xL (which is than the controls (Supplementary Figure 1F). Upon killing the considered to be an inducer of very aggressive lymphomas), mice, we verified that the level of GAPDH expression was lower in another non-limiting glycolytic enzyme enolase-1 (ENO1) or the the LN obtained from Eμ-Myc-shgapdh cells compared with that last limiting glycolytic enzyme PKM2 (Supplementary Figure 1D). obtained from control Eμ-Myc-shctl cells (Supplementary Figure GAPDH overexpression significantly decreased the lifespan of the 1G). Altogether, our results suggest that high expression of mice compared with the control (pMIG) group (Figure 1d). GAPDH, but no other glycolytic enzyme tested, is inducing a very Moreover, lymphoma's aggressiveness provided by GAPDH over- aggressive lymphoma. expression in Eμ-Myc cells is as important as that obtained upon Bcl-xL overexpression, further underlying the role of GAPDH in GAPDH expression drives a vascularized phenotype in vivo lymphomas aggressiveness (Figure 1d). In contrast, overexpression of ENO1-V5 or PKM2-V5 did not affect the lifespan of the mice We repeated the in vivo experiments and killed the mice compared with control cells (Figure 1d) while each enzyme was bearing control (pMIG)-, GAPDH-V5-, ENO1-V5- or PKM2-V5- μ overexpressed to the same extend (Supplementary Figure 1D). We expressing E -Myc lymphomas, 15 days post transfer. In agreement verified that the proportion of infected B cells in vivo was with the overall survival (Figure 1d), the size of Eμ-Myc-GAPDH-V5 equivalent between the groups (Supplementary Figure 1E). Of was increased compared with control LNs (pMIG) or ENO1-V5- or note, similar results were obtained using independent Eμ-Myc PKM2-V5-overexpressing Eμ-Myc LN (Figure 2a). Upon killing the clones. mice, we verified that there was an increase in specific To further confirm the implication of GAPDH on lymphoma GAPDH activity only in GAPDH-V5-expressing LN (Supplementary progression, we reduced its expression in primary Eμ-Myc Figure 2A). In contrast, mice injected with shgapdh-transduced lymphoma cells using specific shRNA. We observed that mice Eμ-Myc cells exhibited smaller axillary LNs (Supplementary Figures injected with shgapdh-transduced Eμ-Myc cells survived longer 2B-D). Moreover, we not only observed that the mice injected

* * HeLa - Hx E -Myc - Hx 4 mRNA 3 pMIG GAPDH-V5 PKM2-V5 ENO1-V5 pMIG GAPDH-V5 hif-1 2 HIF-1 HIF-1

GAPDH-V5 1 V5 GAPDH

Fold induction of 0 Erk2 Hsp90 pMIG ENO1-V5 PKM2-V5 GAPDH-V5

N Hx HeLa- Hx E -Myc- Hx 100 ** shgapdh (h) ) shgapdh (h) 3 80 shctl #546 #675 shctl #7.1 #46

HIF-1 60 HIF-1

40 GAPDH HIF-1 activity GAPDH 20 Actin Erk2 (RLU/ g of protein x10 (RLU/ g of protein

0 1 0.57 0.54 : HIF-1 /Actin: 1 0.64 0.62 pMIG GAPDH-V5 1 0.67 0.53 : GAPDH/Actin: 1 0.63 0.70 Figure 3. GAPDH increases HIF-1α expression and HIF-1 activity in hypoxia (Hx). (a) Relative mRNA levels of hif-1α in axillary lymph node (LN) tumors harvested 15 days post i.v. (as in Figure 2a), were determined by quantiﬁcation PCR and normalized by the level of rplp0 (n = 4 per group). (b) Extracts from HeLa cells transfected with a control vector (pMIG) or a GAPDH-V5-, PKM2-V5-, ENO1-V5-encoding constructs and exposed to Hx for 24 h were immunoblotted for the indicated proteins. (c)Asinb using mouse primary Eμ-Myc lymphoma cells stably transduced with control (pMIG) or a GAPDH-V5-encoding vector. (d) HeLa-HRE-luciferase cells were transfected with pcDNA3- or GAPDH-V5- encoding pcDNA3 construct before exposure to normoxia (N) or Hx for 24 h. Luciferase activity was normalized by the respective protein amount. (e) Total cell extracts from HeLa cells stably transduced with a shgapdh-GFP-encoding vector (shgapdh (h) #546 and #675) and then exposed to Hx for 24 h were analyzed for HIF-1α expression. (f)Asine using primary Eμ-Myc lymphoma cells and an independent sequence shRNA-targeting mouse gapdh (shgapdh(m) #7.1 and #46). Actin or Erk2 were used as loading controls (*Po0.05, **Po0.01).

© 2015 Macmillan Publishers Limited Leukemia (2015) 1163 – 1176 GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1168 with GAPDH-V5-expressing Eμ-Myc lymphoma cells had bigger LN, content, a read-out of functional vascular structures, as it was but in addition, the LN were more vascularized than the controls increased in Eμ-Myc-GAPDH-V5 LN compared with controls (Figure 2b). This was conﬁrmed by the measurement of hemoglobin (Figure 2c).

Eµ-Myc HeLa N Hx N Hx ++--: pMIG ++--: pMIG --++: GAPDH-V5 --++: GAPDH-V5 P-Akt (Ser 473) P-Akt (Ser 473)

Akt Akt

P-I B (Ser 32-36) P-I B (Ser 32-36)

I B I B GAPDH-V5 GAPDH-V5 GAPDH GAPDH

Hsp90 Hsp90

N Hx pMIG GAPDH-V5 200 *** --++: Akti 160 P-Akt (Ser 473) 120 **

Akt g of protein) µ /

3 80 B activity P-I B (Ser 32-36) 40 (RLU x 10 Hsp90 0 pMIG pMIG GAPDH-V5 GAPDH-V5 *** 120 ** pMIG GAPDH-V5 +-+-: pCDNA3 100 -+-+: Myc-I B S32-36A *** 80 HIF-1

g of protein) g of protein) 60 µ /

3 GAPDH-V5

B activity 40 GAPDH

20 Myc-I B S32-36A

(RLU x 10 I B 0 +- +-: pCDNA3 Hsp90 -+- +: Myc-I B S32-36A pMIG GAPDH-V5 Figure 4. GAPDH expression regulates hif-1α mRNA levels through the activation of the NF-κB pathway. (a) Total cell extracts from mouse primary Eμ-Myc cells (left panel) or HeLa cells (right panel) stably expressing control (pMIG) or GAPDH-V5-encoding pMIG vectors and exposed to either normoxia (N) or hypoxia (Hx) for 24 h were immunoblotted for the indicated proteins. Hsp90 is used as a loading control. (b) Total cell extracts from stable HeLa-GAPDH-V5 or control HeLa-pMIG cells treated (+) or not ( − ) with Akt inhibitor (Akti, 1 μM) in Hx for 24 h were immunoblotted for the indicated proteins. Hsp90 is used as a loading control. (c) Stable HeLa-GAPDH-V5 and control HeLa-pMIG cells were transiently transfected with a κB promoter-luciferase construct prior incubation in N or in Hx for 24 h. Luciferase activity was determined and normalized to the protein quantity. (d)Asinc, transfected with a κB promoter-luciferase construct and either a control pcDNA3 or a dominant negative of IκBα-encoding pcDNA3 (Myc-IκBαS32-36A) prior incubation in N or in Hx for 24 h. Luciferase activity was then quantiﬁed. (e)Asind, looking at the expression of indicated proteins. Hsp90 is used as a loading control (**Po0.01 and ***Po0.001).

IP-TRAF2 IgG control

++--: pMIG --++: GAPDH-V5 IP GAPDH IgG control V5 TRAF2 IP IP GAPDH TRAF2

TRAF2 V5

GAPDH TRAF2 OUTPUT OUTPUT Hsp90 Hsp90

IP-TRAF2 IgG control *** 160 + -- + --: pMIG -++ -++: GAPDH-V5 140 --+ --+: KA 120 V5 100 * 80 IP / g of protein) / g of protein) TRAF2 3

B activity 60 40 V5

(RLU x 10 20

TRAF2 0 +- +-: DMSO OUTPUT : KA Hsp90 -+ - + pMIG GAPDH-V5

25 ** ** 200 *** *** ns 20 160

15 120 g of protein)

/ 5 10 80 Lactate kB activity 5 40 ( mol/mg protein) (RLU x 10 0 0 N - KA 2-DG N - KA 2-DG Hx Hx Figure 5. GAPDH binds TRAF2 in hypoxia (Hx), a mechanism that is prevented by GAPDH-speciﬁc inhibition with koningic acid (KA). (a) Total lysates from mouse primary Eμ-Myc cells exposed to 24 h of Hx were subjected to immunoprecipitation of endogenous GAPDH or a control IgG. Presence of co-immunoprecipitated TRAF2 was determined by immunoblotting. (b) Total lysates from primary Eμ-Myc cells overexpressing GAPDH-V5 exposed to 24 h of Hx were subjected to TRAF2 immunoprecipitation or a control IgG. Presence of co-immunoprecipitated GAPDH-V5 was assessed by immunoblotting. (c) Total lysates from stable HeLa cells overexpressing GAPDH-V5 treated or not with 0.2 μg/ml of KA before exposure to 24 h of Hx were subjected to co-immunocprecipitaion as presented in b. (d) Stable HeLa cells expressing control (pMIG) or GAPDH-V5 were transiently co-transfected with a κB promoter-luciferase construct and a CFP-encoding vector. 24 h after transfection, cells were treated (+) or not (dimethyl sulfoxide (DMSO)) with 0.2 μg/ml of KA just before exposure to 24 h in Hx. Luciferase activity was then determined and normalized to the protein quantity for each condition. (e, f)Eμ-Myc cells stably expressing κB promoter-luciferase lentiviral construct were treated or not with KA (0.2 μg/ml) or 2-deoxyglucose (2-DG; 0.25 mM) just before they were exposed to normoxia (N) or Hx for 24 h. Luciferase activity (e) and lactate produced in the supernatant (f) were determined (*Po0.05, **Po0.01 and ***Po0.001, ns ¼ not signiﬁcant).

© 2015 Macmillan Publishers Limited Leukemia (2015) 1163 – 1176 GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1170 As VEGF is a key regulator of angiogenesis, we measured its regulation of the master transcription factor controlling its expression. We observed that among the glycolytic enzyme tested, expression, the hypoxia-inducible factor-1, HIF-1. Very surprisingly, only GAPDH expression led to an increase in vegf-a expression in LN we observed that GAPDH-overexpressing LN display higher levels of lymphoma-bearing mice (Figure 2d). This observation was further of hif-1α mRNA compared with that obtained in control (pMIG), supported by the increase in the vascularized structures stained for ENO1- or PKM2-overexpressing LN (Figure 3a). This increase in the endothelial marker, CD31, within the LNs of Eμ-Myc-GAPDH-V5 mRNA was associated with an increase in HIF-1α protein levels compared with controls (Figures 2e and f). (Figures 3b and c) and in HIF-1 transcriptional activity (Figure 3d Taken together, our results suggest that GAPDH overexpression, and Supplementary Figure 3A) after 24 h of hypoxia at 1% O2.In but not that of other glycolytic enzymes tested, participates in addition, the increase in HIF-1α expression observed upon GAPDH increased lymphoma aggressiveness and vascularization, which is expression could not be obtained upon the expression of PKM2 or associated with an increase in vegf-a mRNA levels. ENO1 (Figure 3b), indicating a specificity of the effect (as observed Elevated GAPDH expression increases VEGF secretion by in vivo, Figure 1d). Importantly, we verified in vivo using the lymphoma cells via its binding to TRAF2 resulting in an increase in pimonidazole probe than hypoxia could be observed in enlarged NF-κB and HIF-1 activities LNs of Eμ-Myc mice (Supplementary Figure 3B). As GAPDH is leading to an increase in vegf-a and vascularization To further support our conclusion, we verified in HeLa and in (Figure 2), we investigated the role of GAPDH expression on the primary Eμ-Myc cells that a decrease in GAPDH levels (using

* IP-GAPDH Vehicle KA (5 mg/kg) 14 IgG control

12 Mice #: 422 423 428 429 430

10 TRAF2

8 IP activity in LN GAPDH 6 1 0.87 0.5 0.46 0.48 : TRAF2/GAPDH 4

2 TRAF2 ( absorbance/mg of protein) GAPDH 0 0.5 5 GAPDH

KA INPUT

Vehicle (mg/kg) Hsp90

1.5 mRNA in LN

1 hif-1

0.5 Fold induction of

000 05 0 5 0

0.5 5

KA Vehicle (mg/kg) Figure 6. GAPDH binds TRAF2 in vivo.(a) GAPDH activity measured in lymph nodes (LNs) of mice bearing Eμ-Myc lymphomas, 24 h after a single i.p. injection of phosphate-buffered saline (PBS; vehicle) or koningic acid (KA, indicated doses). (b) Total cell extracts were isolated from the spleen of mice presented in a to perform endogenous GAPDH-TRAF2 co-immunoprecipitation. Quantiﬁcation of densitometric ratio of TRAF2/GAPDH is provided. (c) Real-time quantiﬁcation PCR analysis of hif-1α mRNA levels in LNs obtained as in a. Transcript levels were determined relative to rplp0 mRNA levels and to vehicle condition (n = 3 per group) (*Po0.05).

IP-TRAF2 IgG control * 40 ** +-- +- - : pMIG -+- -+- : GAPDH-V5 35 - - + - - + : DM-V5 30 25 V5 pMIG GAPDH-V5 DM-V5

g of protein) 20 IP /µ

3 HIF-1 15 TRAF2 B activity 10 V5 5 GAPDH

V5 (RLU x 10 0 Hsp90 TRAF2 pMIG INPUT DM-V5 Hsp90 GAPDH-V5

ns 100 ***

) 100 3 80 80 60 60 ** 40 40 g of protein x10 Survival (%) µ HIF-1 activity 20 20 pMIG (10)

(RLU/ 0 GAPDH-V5 (10) DM-V5 (10) 0 pMIG DM-V5 1618 2022 24 26 28 30

GAPDH-V5 Days

** 160 ** * 500 140

120 400 100 300 80 60

/mg of protein) 200 weight (mg) -1 40 100 ( (axillary lymph nodes))

20 (pg.ml Total VEGF in axillary LN Total 0 0 pMIG pMIG DM-V5 DM-V5 GAPDH-V5 GAPDH-V5 Figure 7. Mutation in GAPDH catalytic site reduces its binding to TRAF2 and does not enhance lymphoma aggressiveness. (a) Total cell extracts from Eμ-Myc cells expressing a green ﬂuorescent protein (GFP)-encoding vector (pMIG), GAPDH-V5 or GAPDH C152S/H179F-V5 (DM-V5) exposed to hypoxia for 24 h were subjected to TRAF2 immunoprecipitation or an IgG control. The presence of co-immunoprecipitated GAPDH-V5 was assessed by immunoblotting. (b) Stable HeLa cells expressing GAPDH-V5 or DM-V5 were transiently transfected with a κB promoter-luciferase construct prior incubation in normoxia (N) or hypoxia (Hx) for 24 h. Luciferase activity was determined as previously. (c)Eμ-Myc cells presented in a were cultured in Hx for 24 h and the level of the indicated protein was analyzed by immunoblotting. Hsp90 was used as a loading control. (d) Stable HeLa-HRE-luciferase cells were transiently transduced with a pMIG constructs to overexpress GAPDH-V5 or DM-V5 before exposure to Hx for 24 h. Luciferase activity was determined as in b.(e) Kaplan–Meier curves of syngenic C57BL/6 mice intravenously injected with primary Eμ-Myc cells infected with pMIG constructs to overexpress GAPDH-V5, DM-V5 compared with control Eμ-Myc cells transduced with empty pMIG vector. N = 10 per group. (f) Weight of axillary lymph nodes of the mice presented in e, 16 days post injection (n = 7 per group). (g) Secreted VEGF in axillary lymph nodes isolated in f. Total VEGF measured was normalized by the quantity of protein in each sample (n = 5 per group) (*Po0.05, **Po0.01 and ***Po0.001, ns ¼ not signiﬁcant).

© 2015 Macmillan Publishers Limited Leukemia (2015) 1163 – 1176 GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1172 several independent shRNAs targeting either the murine or the and failed to activate NF-κB (Figure 7b) and to elevate HIF-1α human form of GAPDH) led to a decrease in HIF-1α protein expression and HIF-1 activity (Figures 7c and d) in hypoxia, further expression (Figures 3e and f and Supplementary Figure 3C) under suggesting that the catalytic site of the enzyme has a role in the hypoxic conditions. observed effects. As NF-κB is a key regulator of hif-1α transcription, controlled in Accordingly, DM-overexpressing Eμ-Myc cells did not show part by Akt, we investigated NF-κB pathway in GAPDH- increased lymphoma progression in vivo, as opposed to cells overexpressing Eμ-Myc and HeLa cells. Although we have shown overexpressing wild-type GAPDH (Figures 7e and f and that GAPDH could stabilize the active form of Akt,10 we observed Supplementary Figures 6A–C). Finally, GAPDH expression but not – that GAPDH increased Ser32 36 phosphorylation of IκBα in DM expression, led to an increase in total VEGF secretion normoxia and in hypoxia (Figure 4a), in a Akt-independent (Figure 7g) as a consequence of NF-κB and HIF-1 activation. manner (Figure 4b), leading to an increase in NF-κB activity κ (Figure 4c). In the opposite, silencing of GAPDH reduced NF- B DLBCL patient biopsies presenting elevated GAPDH expression activity (Supplementary Figure 4), reinforcing the link between κ α κ have enhanced activation of the NF- B/HIF-1 axis leading to an GAPDH and NF- B pathway. increase in vegf-a expression and vascularization To further investigate the contribution of NF-κB in the α To further strengthen our observations, we analyzed the level of regulation of HIF-1 by GAPDH, we used a dominant negative κ α form of IκBα,IκBαS32-36A. Its expression reduced NF-κB activity expression of the NF- B/HIF-1 axis in DLBCL patient biopsies (Figure 4d) and HIF-1α protein expression (Figure 4e), indicating obtained at diagnosis (Supplementary Table 1). The pre-clinical κ mouse model we used recapitulates certain aspects of human that the GAPDH-dependent activation of NF- B is required for the 27 GAPDH-mediated regulation of HIF-1α. Burkitt lymphoma. However, as c-MYC upregulation is a feature κ of a proportion of follicular lymphomas undergoing aggressive As we demonstrated that GAPDH expression enhances NF- B 28 activity in hypoxia, and knowing that TRAF2 is a key regulator of transformation to DLBCL and as Burkitt lymphomas are far less fi NF-κB activation, recently shown as able to interact with GAPDH frequent, we con rmed our pre-clinical observations using DLBCL albeit in a different context,26 we investigated GAPDH-TRAF2 patient samples. ‘ ’ binding in Eμ-Myc and HeLa cells under hypoxic culture condition. Out of the 13 DLBCL biopsies obtained, 6 were low and 7 were ‘ ’ We investigated its interaction upon hypoxia as this binding has high expressers of gapdh mRNA (Figure 8a) upon unbiased 26 analysis of level of expression (see Materials and methods section been shown to be enhanced upon stress and as hypoxia can be 29 observed in enlarge LNs of sick Eμ-Myc mice (Supplementary for more details). Using publicly available data set and upon Figure 3B). We found that both endogenous GAPDH or transfected similar and unbiased analysis of the level of gapdh expression in fi GAPDH-V5 could be co-immunoprecipitated with TRAF2 in primary 414 samples of DLBCL patients, we con rmed that 50% of the Eμ-Myc cells (Figures 5a and b) and HeLa cells (Figure 5c) exposed to patients present higher levels of gapdh than the other half of the hypoxia. Interestingly, the use of non-toxic doses of a specific patients (not shown). μ GAPDH inhibitor, KA, reduced GAPDH activity by 50–60% As observed in E -Myc mice, biopsies expressing the highest high κ (Supplementary Figure 5A) and prevented GAPDH binding to TRAF2 level of gapdh mRNA (gapdh ) also had high levels of the NF- B- κ targeted gene nfkbia, indicating an increase activity of NF-κBin (Figure 5c) and NF- B activity (Figure 5d) under hypoxic condition. 19 As a consequence, GAPDH inhibition with KA failed to increase those samples (Figure 8b). Interestingly, mRNA levels of hif-1α, HIF-1α expression (Supplementary Figure 5B) and HIF-1 activity which controls vegf-a expression, and vegf-a mRNA itself was high low (Supplementary Figure 5C) to the same extend to that observed higher in gapdh compared with gapdh patients (Figure 8b). with shRNA-targeting GAPDH. It is worth mentioning that another member of the HIF family (hif- To determine if GAPDH role in glycolysis could be separated 2α) was not significantly modulated (Figure 8b) within the same from GAPDH effects on NF-κB activation, we analyzed lactate biopsies, underling the specificity of the observed effect. production and NF-κB activity in Eμ-Myc cells in the presence of KA Finally, immunohistochemical analysis of those DLBCL samples or in the presence of the glycolytic inhibitor 2-deoxyglucose (Figures confirmed that compared with biopsies expressing low level of 5e and f). Although KA efficiently reduced GAPDH (Supplementary gapdh mRNA, biopsies displaying high levels of gapdh Figure 5D) and NF-κB activities (Figure 5e, in agreement with its express more GAPDH protein and present higher number of ability to prevent GAPDH binding to TRAF2, Figure 5c), it did not functional vessels as shown by CD31-stained vascular structure lead to a reduction of lactate production (Figure 5f, Supplementary (immunohistochemical analysis of three patients of each group are Figure 5D). On the opposite, 2-deoxyglucose reduced lactate presented in Figure 8c and quantified in Supplementary Figure 7A) production (Figure 5f and Supplementary Figure 5E) but did not and by the presence of tumor cells and red blood cells within the modulate NF-κB activity (Figure 5e and Supplementary Figure 5E). vessels (see insets in Figure 8c and Supplementary Figure 7B). Altogether indicating that GAPDH controls NF-κB pathway indepen- Taken together, our results suggest that GAPDH, as opposed to dently of its glycolytic functions. the other tested glycolytic enzymes, can increase lymphoma We further tested GAPDH–TRAF2 interaction in vivo by treating growth and vascularization in vivo. This effect is mediated by the Eμ-Myc-lymphoma-bearing mice with KA. Very interestingly, we ability of GAPDH to bind TRAF2, which in turn contributes to confirmed that endogenous GAPDH could bind endogenous NF-κB activation and subsequent HIF-1α induction. TRAF2 in vivo and that an efficient dose of KA (5 mg/kg, leading to a 50% reduction of GAPDH activity in LN), but not an inactive dose DISCUSSION of KA (0.5 mg/kg; Figure 6a), reduced the interaction observed between those proteins (Figure 6b). Moreover, we confirmed Differences in metabolism were among the first identified in vivo that GAPDH inhibition with KA (5 mg/kg) reduced hif-1α variations between normal and cancer cells.4,30 Although most, mRNA (Figure 6c) levels, underscoring the link between GAPDH, if not all, glycolytic enzymes are found to be overexpressed in the TRAF2, NF-κB and HIF-1α expression. vast majorities of cancers,31 their precise roles in oncogenesis are To understand the contribution of GAPDH activity in the regulation far from being understood. It has long been considered that those of NF-κB/HIF-1 pathway and in lymphoma aggressiveness, we used a enzymes are only overexpressed to meet the energy demand of GAPDH double mutant (DM, GAPDH C152SandH179F) that is unable cancer cells and to facilitate the uptake and incorporation of to sustain the glycolytic function of the enzyme7 and that is known nutrients into the biomass.5 However, non-glycolytic roles of these to disrupt GAPDH/TRAF2 interaction.26 Compared with wild-type enzymes are beginning to emerge in several settings, although GAPDH, we demonstrated that DM bound less to TRAF2 (Figure 7a) rarely in the context of cancers.6–10,32

Leukemia (2015) 1163 – 1176 © 2015 Macmillan Publishers Limited GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1173 We established that GAPDH but not other glycolytic enzymes In addition, we demonstrated that GAPDH, but not the other was upregulated in highly aggressive Eμ-Myc LN tumors. tested glycolytic enzymes, is a key regulator of vascularization and The reason of this specificity of expression remains to be made. of c-myc-dependent lymphomagenesis. Indeed, transcription factors such as HIF-1 or c-Myc will Several lines of evidence suggested that GAPDH is involved in control the expression of most glycolytic enzymes.33 It is therefore the modulation of NF-κB signaling,26,35 but this is the first unlikely that their activation will lead to a specific increase in demonstration of the GAPDH-dependent activation of NF-κBon GAPDH but no other glycolytic enzymes. In this line, we did not HIF-1α expression, vascularization and lymphoma aggressiveness. observe any correlation between c-Myc and GAPDH expression in As GAPDH is a HIF-1-targeted gene, our results highlight the our models (Figure 1c). One possibility is that a transcription existing positive feedback loop between HIF-1 and GAPDH. factor, which remains to be identified, has some specificity We also demonstrated that GAPDH expression increases mRNA for GAPDH expression compared with other glycolytic levels of nfkbia, vegf-a, hif-1α but not hif-2α. We showed that this enzymes. Another attractive hypothesis that deserves further process could be explained in part by the modulation of the NF-κB investigation comes from the observation that glutamine activity that is indeed known to induce hif-1α but not hif-2α. downstream metabolites can activate GAPDH expression.34 However, GAPDH is known to interact with nucleic acids11 and to As glutamine metabolism is exacerbated in some cancers, regulate mRNA stability and consequently controls the expression it is possible that the accumulation of those metabolites of proteins, such as endothelin-1,36 colony-stimulating factor-1 could participate in the observed increase in GAPDH (ref. 37) and interferon-γ.32 Therefore, we cannot exclude that on expression. top of the ability of GAPDH to activate the NF-κB/HIF-1 pathway

a gapdhlow gapdhhigh 4

3.5

mRNA 3

2.5 gapdh 2

1.5

0.5 Fold induction of 0 3 9 1 7 4 6 14105 8 131112 Patient biopsy # ns * b ** ** 4 2.5 4 7

mRNA 2 mRNA in mRNA In 3 mRNA In 3

a - 5

nfkbia 1.5 vegf hif-2 hif-1 4 2 2 1 3 DLBCL biopsies DLBCL biopsies 1 DLBCL biopsies 2 In DLBCL biopsies 1 0.5 1 Fold induction of Fold induction of Fold induction of 0 Fold induction of 0 0 0

Figure 8. High levels of gapdh mRNA in DLBCL biopsies correlates with an upregulation of hif-1α mRNA and with a vascularized phenotype. (a) Real-time quantification PCR (qPCR) analysis of gapdh levels in biopsies of DLBCL patients revealed two groups of patients displaying a ‘low’ (n = 6) or a ‘high’ (n = 7) level of gapdh (see Materials and methods section). Transcripts levels were determined relative to cyclophilin-A (ppia) mRNA levels. (b) Real-time qPCR analysis of nfkbia, hif-1α, hif-2α and vegf-a in DLBCL from patients displaying ‘gapdhlow’ (n = 6) versus ‘gapdhhigh’ (n = 7) levels of gapdh mRNA. Transcript levels were determined relative to cyclophilin-A (ppia) mRNA levels. (c) Serial sections of paraffin-embedded DLBCL biopsies from different patients (patients #1, #4, #9 are ‘low’ gapdh, while #5, #8 and #11 are ‘high’ gapdh mRNA) were stained for GAPDH and CD31. The white scale bar represents 100 μm (x26). Zoom (right panels): 50 μm (x60) (*Po0.05, **Po0.01 and ns ¼ not significant).

#1 gapdh

#4 low

#5 gapdh high #8

#11

GAPDH CD31 Figure 8. (Continued)

upon hypoxia, the RNA-binding ability of GAPDH could also a sick mice, Supplementary Figure 3B) GAPDH-mediated increase contribute, in part, to stabilize hif-1α and/or vegf-a mRNA. of NF-κB/HIF-1 activities will lead to a global increase in glycolytic Upon hypoxia, GAPDH expression further stimulates vegf-a metabolism resulting in an increase in lactate production expression and total VEGF secretion in vivo, suggesting that (Figure 1b). Altogether, our results indicate that GAPDH may play through its ability to enhance NF-κB/HIF-1 activities, GAPDH a broader role in cancer progression than has previously been controls two central signaling pathways involved in pro- appreciated. It is appealing to speculate that some tumors may angiogenic cytokines production as well as in metabolism adopt glycolytic metabolism not only for tumor proliferation but adaptation of tumor cells. Very importantly we could establish also to facilitate tumor survival through cytokine production. that GAPDH-mediated modulation of NF-κB/HIF-1 activities are We could established that GAPDH inhibition using KA independent of GAPDH glycolytic functions (Figures 5e and f, prevented TRAF2-GAPDH interaction very likely, although the Supplementary Figures 5D and E). Therefore, as hypoxia is ability of this compound to covalently bind to the essential observed in physio-pathological settings (that is enlarged LN of cysteine residue of GAPDH38,39 (residue mutated in the DM

Leukemia (2015) 1163 – 1176 © 2015 Macmillan Publishers Limited GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1175 mutant that also prevent GAPDH-TRAF2 binding). Our work also 11 Colell A, Green DR, Ricci JE. Novel roles for GAPDH in cell death and carcino- brings a new light on the use of GAPDH-specific inhibitors in genesis. Cell Death Differ 2009; 16: 1573–1581. clinical setting. In most attempts on the use of those inhibitors as 12 Lavallard VJ, Pradelli LA, Paul A, Beneteau M, Jacquel A, Auberger P et al. a new cancer treatment, the goal was to limit glucose metabolism Modulation of caspase-independent cell death leads to resensitization of imatinib 69 – in cancer cells, which will very likely lead to unacceptable toxicity mesylate-resistant cells. Cancer Res 2009; : 3013 3020. 13 Revillion F, Pawlowski V, Hornez L, Peyrat JP. Glyceraldehyde-3-phosphate in vivo. Here we provide evidences that the use of lower doses of dehydrogenase gene expression in human breast cancer. Eur J Cancer 2000; 36: such inhibitors, leading to a partial decrease in GAPDH activity, will 1038–1042. not significantly impact on glucose metabolism for cells in the 14 Wang D, Moothart DR, Lowy DR, Qian X. The expression of glyceraldehyde-3- oxygenated areas of the tumors (therefore reducing their toxicity) phosphate dehydrogenase associated cell cycle (GACC) genes correlates with but will mainly act on the central site of tumor adaptation, that is cancer stage and poor survival in patients with solid tumors. PLoS One 2013; 8: the hypoxic regions, by reducing the activity of the NF-κB/HIF-1 e61262. axis, thus limiting metabolic changes and production of pro- 15 Harris AW, Pinkert CA, Crawford M, Langdon WY, Brinster RL, Adams JM. The E mu-myc transgenic mouse. A model for high-incidence angiogenic cytokines. Therefore, the use of lower doses of GAPDH spontaneous lymphoma and leukemia of early B cells. J Exp Med 1988; 167: inhibitors (that remains to be determined) for limited amount of 353–371. time in combination with conventional chemotherapy could 16 Lindemann RK, Newbold A, Whitecross KF, Cluse LA, Frew AJ, Ellis L et al. Analysis improve their efficiency by limiting the crosstalk between tumor of the apoptotic and therapeutic activities of histone deacetylase inhibitors by cells and its microenvironment. Moreover, the GAPDH-dependent using a mouse model of B cell lymphoma. Proc Natl Acad Sci USA 2007; 104: control of vegf-a expression is of potential relevance to the 8071–8076. treatment of several cancers. Accordingly, high GAPDH expression 17 Beneteau M, Zunino B, Jacquin MA, Meynet O, Chiche J, Pradelli LA et al. Combination of glycolysis inhibition with chemotherapy results in an antitumor is associated with an unfavorable prognosis in breast and lung 109 – 13,14 immune response. Proc Natl Acad Sci USA 2012; : 20071 20076. cancer. Therefore, GAPDH inhibitors combined with conven- 18 Dayan F, Roux D, Brahimi-Horn MC, Pouyssegur J, Mazure NM. The oxygen sensor tional chemotherapies could represent an interesting option factor-inhibiting hypoxia-inducible factor-1 controls expression of distinct genes not only to target the glycolytic metabolism of the tumor but through the bifunctional transcriptional character of hypoxia-inducible factor- also to reduce the GAPDH-dependent pro-angiogenic cytokine 1alpha. Cancer Res 2006; 66: 3688–3698. production. 19 Bottero V, Imbert V, Frelin C, Formento JL, Peyron JF. Monitoring NF-kappa B transactivation potential via real-time PCR quantification of I kappa B-alpha gene expression. Mol Diagn 2003; 7:187–194. CONFLICT OF INTEREST 20 Garaulet G, Alfranca A, Torrente M, Escolano A, Lopez-Fontal R, Hortelano S et al. IL10 released by a new inflammation-regulated lentiviral system The authors declare no conflict of interest. efficiently attenuates zymosan-induced arthritis. Mol Ther 2013; 21: 119–130. 21 Campo E, Swerdlow SH, Harris NL, Pileri S, Stein H, Jaffe ES. The 2008 WHO ACKNOWLEDGEMENTS classification of lymphoid neoplasms and beyond: evolving concepts and practical applications. Blood 2011; 117:5019–5032. We gratefully acknowledge the Centre Méditerranéen de Médecine Moléculaire 22 project TIN-Hslpf. A predictive model for aggressive non-Hodgkin's lymphoma. animal and microscopy facilities. We thank Drs Tanti JF, Cormont M and Giorgetti- The International Non-Hodgkin's Lymphoma Prognostic Factors Project. N Engl J Peraldi S for the hypoxic chamber, Larbret F for cell sorting, Dr Pouysségur J for Med 1993; 329: 987–994. HIF-1α antibody, Escoubas C, Paquet A and Barbry P for their help. Acknowledgement 23 Thieblemont C, Briere J, Mounier N, Voelker HU, Cuccuini W, Hirchaud E et al. to the Tumorothèque of Hopital Saint-Louis for the preparation of the tumoral The germinal center/activated B-cell subclassification has a prognostic impact for samples of the patients. This work was supported by the Fondation ARC (Association response to salvage therapy in relapsed/refractory diffuse large B-cell lymphoma: pour la Recherche sur le Cancer), the Agence Nationale de la Recherche (ANR-09- a bio-CORAL study. J Clin Oncol 2011; 29: 4079–4087. JCJC-0003, LABEX SIGNALIFE ANR-11-LABX-0028-01), by la Fondation de France, the 24 De Bock K, Georgiadou M, Schoors S, Kuchnio A, Wong BW, Cantelmo Centre Scientifique de Monaco and the Cancéropole PACA. JC was supported by the AR et al. Role of PFKFB3-driven glycolysis in vessel sprouting. Cell 2013; 154: Fondation ARC and by la Fondation de France. LM was supported by Fondation pour 651–663. la Recherche Medicale (FRM) and la Ville de Nice. 25 Le A, Lane AN, Hamaker M, Bose S, Gouw A, Barbi J et al. Glucose-independent glutamine metabolism via TCA cycling for proliferation and survival in B cells. Cell Metab 2012; 15: 110–121. 26 Gao X, Wang X, Pham TH, Feuerbacher LA, Lubos ML, Huang REFERENCES M et al. NleB, a bacterial effector with glycosyltransferase activity, targets 1 Semenza GL. Targeting HIF-1 for cancer therapy. Nat Rev Cancer 2003; 3:721–732. GAPDH function to inhibit NF-kappaB activation. Cell Host Microbe 2013; 13: 2 Potente M, Gerhardt H, Carmeliet P. Basic and therapeutic aspects of angiogen- 87–99. esis. Cell 2011; 146:873–887. 27 Adams JM, Harris AW, Pinkert CA, Corcoran LM, Alexander WS, Cory S et al. 3 Kroemer G, Pouyssegur J. Tumor cell metabolism: cancer's Achilles' heel. Cancer The c-myc oncogene driven by immunoglobulin enhancers induces lymphoid Cell 2008; 13: 472–482. malignancy in transgenic mice. Nature 1985; 318: 533–538. 4 Galluzzi L, Kepp O, Vander Heiden MG, Kroemer G. Metabolic targets for cancer 28 Lossos IS, Alizadeh AA, Diehn M, Warnke R, Thorstenson Y, Oefner PJ et al. therapy. Nat Rev Drug Discov 2013; 12: 829–846. Transformation of follicular lymphoma to diffuse large-cell lymphoma: alternative 5 Lunt SY, Vander Heiden MG. Aerobic glycolysis: meeting the metabolic require- patterns with increased or decreased expression of c-myc and its ments of cell proliferation. Annu Rev Cell Dev Biol 2011; 27: 441–464. regulated genes. Proc Natl Acad Sci USA 2002; 99: 8886–8891. 6 Majewski N, Nogueira V, Bhaskar P, Coy PE, Skeen JE, Gottlob K et al. Hexokinase- 29 Lenz G, Wright G, Dave SS, Xiao W, Powell J, Zhao H et al. Stromal gene signatures mitochondria interaction mediated by Akt is required to inhibit apoptosis in the in large-B-cell lymphomas. N Engl J Med 2008; 359:2313–2323. presence or absence of Bax and Bak. Mol Cell 2004; 16:819–830. 30 Jones RG, Thompson CB. Tumor suppressors and cell metabolism: a recipe for 7 Colell A, Ricci JE, Tait S, Milasta S, Maurer U, Bouchier-Hayes L et al. GAPDH and cancer growth. Genes Dev 2009; 23:537–548. autophagy preserve survival after apoptotic cytochrome c release in the absence 31 Altenberg B, Greulich KO. Genes of glycolysis are ubiquitously overexpressed in of caspase activation. Cell 2007; 129: 983–997. 24 cancer classes. Genomics 2004; 84: 1014–1020. 8 Yang W, Xia Y, Ji H, Zheng Y, Liang J, Huang W et al. Nuclear PKM2 regulates beta- 32 Chang CH, Curtis JD, Maggi LB Jr, Faubert B, Villarino AV, O'Sullivan D et al. catenin transactivation upon EGFR activation. Nature 2011; 480: 118–122. Posttranscriptional control of T cell effector function by aerobic glycolysis. Cell 9 Luo W, Hu H, Chang R, Zhong J, Knabel M, O'Meally R et al. Pyruvate kinase M2 is a 2013; 153: 1239–1251. PHD3-stimulated coactivator for hypoxia-inducible factor 1. Cell 2011; 145: 33 Kim JW, Zeller KI, Wang Y, Jegga AG, Aronow BJ, O'Donnell KA et al. Evaluation of 732–744. myc E-box phylogenetic footprints in glycolytic genes by chromatin 10 Jacquin MA, Chiche J, Zunino B, Beneteau M, Meynet O, Pradelli LA et al. GAPDH immunoprecipitation assays. Mol Cell Biol 2004; 24: 5923–5936. binds to active Akt, leading to Bcl-xL increase and escape from caspase- 34 Claeyssens S, Gangneux C, Brasse-Lagnel C, Ruminy P, Aki T, Lavoinne A et al. independent cell death. Cell Death Differ 2013; 20: 1043–1054. Amino acid control of the human glyceraldehyde 3-phosphate dehydrogenase

© 2015 Macmillan Publishers Limited Leukemia (2015) 1163 – 1176 GAPDH accelerates lymphomagenesis via NF-κB/HIF1 J Chiche et al 1176 gene transcription in hepatocyte. Am J Phys Gastrointest Liver Physiol 2003; 285: 37 Zhou Y, Yi X, Stoffer JB, Bonafe N, Gilmore-Hebert M, McAlpine J et al. The G840–G849. multifunctional protein glyceraldehyde-3-phosphate dehydrogenase is both 35 Mookherjee N, Lippert DN, Hamill P, Falsafi R, Nijnik A, Kindrachuk J et al. regulated and controls colony-stimulating factor-1 messenger RNA stability in Intracellular receptor for human host defense peptide LL-37 in monocytes. ovarian cancer. Mol Cancer Res 2008; 6:1375–1384. J Immunol 2009; 183: 2688–2696. 38 Endo A, Hasumi K, Sakai K, Kanbe T. Specific inhibition of glyceraldehyde-3-phosphate 36 Rodriguez-Pascual F, Redondo-Horcajo M, Magan-Marchal N, Lagares D, dehydrogenase by koningic acid (heptelidic acid). J Antibiot 1985; 38:920–925. Martinez-Ruiz A, Kleinert H et al. Glyceraldehyde-3-phosphate dehydrogenase 39 Sakai K, Hasumi K, Endo A. Identification of koningic acid (heptelidic acid)-mod- regulates endothelin-1 expression by a novel, redox-sensitive mechanism ified site in rabbit muscle glyceraldehyde-3-phosphate dehydrogenase. Biochim involving mRNA stability. Mol Cell Biol 2008; 28: 7139–7155. Biophys Acta 1991; 1077:192–196.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)