O-GlcNAcylation destabilizes the active tetramericPKM2 to promote the Warburg effectYang Wanga,1, Jia Liua,1, Xin Jina,1, Dapeng Zhangb,c, Dongxue Lia, Fengqi Haoa, Yunpeng Fenga, Shan Gua,Fanlin Menga, Miaomiao Tiana, Yi Zhenga, Ling Xina, Xinbo Zhanga, Xue Hana, L. Aravindd, and Min Weia,2

aKey Laboratory of Molecular Epigenetics of the Ministry of Education, Northeast Normal University, Changchun, Jilin, 130024, People’s Republic of China;bDepartment of Biology, Saint Louis University, St. Louis, MO 63103; cProgram of Bioinformatics and Computational Biology, Saint Louis University, St. Louis,MO 63103; and dNational Center for Biotechnology Information, National Library of Medicine, National Institutes of Health, Bethesda, MD 20894

Edited by Gregg L. Semenza, Johns Hopkins University School of Medicine, Baltimore, MD, and approved November 13, 2017 (received for review March14, 2017)

The Warburg effect, characterized by increased glucose uptake andlactate production, is a well-known universal across cancer cells andother proliferating cells. PKM2, a splice isoform of the pyruvate kinase(PK) specifically expressed in these cells, serves as a major regulator ofthis metabolic reprogramming with an adjustable activity subjectedto numerous allosteric effectors and posttranslational modifications.Here, we have identified a posttranslational modification on PKM2,O-GlcNAcylation, which specifically targets Thr405 and Ser406, residuesof the region encoded by the alternatively spliced exon 10 in cancercells. We show that PKM2 O-GlcNAcylation is up-regulated in varioustypes of human tumor cells and patient tumor tissues. Themodificationdestabilized the active tetrameric PKM2, reduced PK activity, and led tonuclear translocation of PKM2. We also observed that the modificationwas associated with an increased glucose consumption and lactateproduction and enhanced level of lipid and DNA synthesis, indicatingthat O-GlcNAcylation promotes the Warburg effect. In vivo experi-ments showed that blocking PKM2 O-GlcNAcylation attenuated tumorgrowth. Thus, we demonstrate that O-GlcNAcylation is a regulatorymechanism for PKM2 in cancer cells and serves as a bridge betweenPKM2 and metabolic reprogramming typical of the Warburg effect.

PKM2 | O-GlcNAcylation | Warburg effect | cancer metabolism

Compared with normal cells, cancer cells have distinct meta-bolic features characterized by changes in nutrient consump-

tion and metabolite concentrations (1). One prominent feature isaerobic glycolysis, which leads to increased glucose consumptionand lactate production in cancer cells, a phenomenon referred toas the Warburg effect (2–5). It is believed that the Warburg effectnot only allows cancer cells to meet high biosynthetic demand inboth ATP and biomass production, but also provides a resistanceto acid-induced cell toxicity, thus promoting unconstrained pro-liferation and invasion (6).Recent studies have shown pyruvate kinase M2 (PKM2) to be

a major regulator for metabolic reprogramming in cancer whichcontributes to the Warburg effect (7–9). Pyruvate kinases (PKs)are the final rate-limiting enzymes in glycolysis. In humans, twoPK genes generate four isoforms via alternative splicing, PKL,PKR, PKM1, and PKM2, which display distinct tissue-specificexpression patterns (10, 11). The PKL and PKR isoforms areexpressed in liver and red blood cells, respectively. PKM1, con-taining exon 9 of the PKM gene, is expressed in most adult tis-sues, whereas PKM2, containing exon 10, is expressed duringembryonic development and in other highly proliferative cells,including cancer cells (2, 4, 11). The expression switch fromPKM1 to PKM2 is frequently observed in many cancer cells andis critical for cancer cell proliferation (7).The mechanism for this switch from PKM1 to PKM2, which

induces metabolic reprogramming and the Warburg effect in cancercells, remains largely unclear. It appears that, although the differ-ence between PKM1 and PKM2 is only 22 amino acids, PKM2displays a very unique regulatory property: whereas PKM1 forms astable, constitutively active tetramer, PKM2’s tetrameric status andactivity are closely controlled by numerous allosteric effectors and

posttranslational modifications. On the one hand, PKM2 is allo-sterically activated by fructose 1,6-bisphosphate (FBP) (12), anupstream intermediate in glycolysis, and serine, which is synthesizedfrom a glycolytic intermediate 3-phosphoglycerate (13). On the otherhand, PKM2 activity is regulated by several posttranslational modi-fications, such as phosphorylation, acetylation, and SUMOylation(14–19). These modifications, such as acetylation on Lys433 orphosphorylation on Tyr105, impair PKM2 activity by blocking theassociation of FBP (16, 17). Furthermore, several modificationsare frequently associated with the translocation of PKM2 fromcytoplasm to nucleus, where a nonmetabolic function of PKM2 iscritical in the establishment of the Warburg effect (16, 20). Thus, itappears that the permissive property of PKM2 function regulatedby allosteric effectors and posttranslational modifications and itsnuclear function allow PKM2 to serve as a key regulatory node inmetabolic reprogramming in cancer cells (21).O-GlcNAcylation of enzymes has been recently implicated as

an important regulatory mechanism for metabolic pathways (22).Two enzymes, phosphofructokinase 1 (PFK1) and glucose-6-phosphate dehydrogenase (G6PD), have been shown to undergoO-GlcNAcylation to regulate cell metabolism and proliferation(23, 24). In the present study, we investigated whether the keymetabolic regulator in cancer cells, PKM2, is also modified byO-GlcNAcylation and examined its potential role in the Warburgeffect. We discovered that, across diverse tumor cell types andpatient tissues, PKM2 was O-GlcNAcylated at two residues, Thr405

Significance

Cancer cells are characterized by a high rate of glycolysis evenunder normal oxygen availability to meet the demand of biomassproduction during rapid proliferation. An isoform of pyruvate ki-nase (PK), PKM2, preferentially expressed in cancers, was recentlyshown to be critical for this metabolic reprogramming with ad-justable activity and dynamic cellular relocalization. However,specific molecular mechanisms mediating PKM2’s role in cancer-specific metabolism remain largely elusive. We demonstrate thatO-GlcNAcylation of PKM2 on threonine/serine encoded by analternatively spliced exon disrupts the intersubunit interactionsin the active PKM2 tetramer. This causes the tetramer disas-sembly, reduced PK activity, and its nuclear translocation to fa-cilitate cell proliferation. Thus, our findings furnish a key piece inthe puzzle of aerobic glycolysis in cancer.

Author contributions: Y.W., J.L., X.J., Y.F., and M.W. designed research; Y.W., J.L., X.J.,D.L., F.H., S.G., F.M., M.T., Y.Z., L.X., X.Z., and X.H. performed research; Y.W., J.L., X.J., D.Z.,and D.L. contributed new reagents/analytic tools; Y.W., J.L., X.J., D.Z., D.L., F.H., Y.F., andM.W. analyzed data; and Y.W., D.Z., Y.F., L.A., and M.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1Y.W., J.L., and X.J. contributed equally to this work.2To whom correspondence should be addressed. Email: weim750@nenu.edu.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1704145115/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1704145115 PNAS Early Edition | 1 of 6

CELL

BIOLO

GY

and Ser406, of the region encoded by spliced exon 10. This modifi-cation detetramerized the active PKM2 and led to a dramaticallyreduced activity and increased nuclear translocation. We also ob-served that PKM2 O-GlcNAcylation promoted the Warburg effectwith an increased glucose consumption and lactate production.Thus, we demonstrated that O-GlcNAcylation is a regulatory modefor PKM2 in cancer cells and serves as a bridge translating genomicswitch from PKM1 to PKM2 toward metabolic reprogramming.

ResultsPKM2 Is Highly O-GlcNAcylated in Tumor Cells and Patient TumorTissues. We first set out to investigate PKM2 O-GlcNAcylationin different proliferating cell types, including activated T,Jurkat, MCF-10A, MCF-7, MDA-MB-231, HEK-293T, HeLa,A375, T98-G, U251, and U937. PKM2 expression and globalO-GlcNAcylation were pronounced in proliferating cells, butnot in quiescent naive T cells (Fig. 1A). Furthermore, throughtwo different methods (Click-iT O-GlcNAc Enzymatic Labelingand traditional immunoprecipitation), PKM2 O-GlcNAcylationwas detected in all tested cells, and it was strongly up-regulatedin malignant cell lines derived from leukemia and solid tumorscompared with untransformed cells, including naive T, acti-vated T, MCF-10A, and HEK-293T (Fig. 1B and Fig. S1A).Notably, no difference was found between activated T cells andtheir quiescent counterparts (Fig. 1B and Fig. S1A). These dataindicate that the up-regulation of PKM2 O-GlcNAcylationmight have a specific role in cancer cell proliferation.We further examined the level of O-GlcNAcylation of PKM2 in

both breast tumor tissues and tumor-adjacent normal tissues fromthe same patient. By studying a wide range of patient samples, wefound that PKM2 O-GlcNAcylation was significantly elevated inbreast tumor tissues compared with adjacent normal tissues(Fig. 1 C and D and Fig. S1B). The enhanced signal of PKM2O-GlcNAcylation was also consistent with the up-regulated ex-pression of PKM2 in tumor tissues compared with adjacent tissues(Fig. S1C). These results strongly indicate that, not only the highexpression of PKM2 in tumor cells, but also O-GlcNAcylation ofPKM2, is a characteristic feature for tumor cells.We examined the stoichiometry of PKM2 O-GlcNAcylation

using two chemoenzymatic labeling methods (a Click-iT bio-tinylation and an Alkyne-5K-PEG resolvable mass tag) (25, 26);both revealed that the basal O-GlcNAcylation level of PKM2 inMCF-7 cells was ∼7–10% (Fig. S2 A and B).

PKM2 Is O-GlcNAcylated at Thr405 and Ser406. We next sought toidentify the O-GlcNAcylation site(s) on PKM2. In Escherichia colicotransfected with GST-tagged PKM2 and O-GlcNAc transferase(OGT), Thr405 and Ser406 were detected asO-GlcNAcylation sites bymass spectrometry (Fig. 2A). Importantly, these two residues werelocated at the region encoded by exon 10, which is specifically in-troduced in the PKM2 transcript via alternative splicing in cancercells (Fig. 2A). When Thr405 or Ser406 was mutated into alanine(PKM2T405A or PKM2S406A), the O-GlcNAcylation level of PKM2in MCF-7 cells was decreased (Fig. 2B). The reduction of PKM2O-GlcNAcylation was more pronounced when Thr405 and Ser406 wereboth mutated (PKM2T405A/S406A) (Fig. 2B). We next investigatedthe issue of O-GlcNAcylation of PKM2 not being completelyabolished in these mutants by further exploring other potentialO-GlcNAcylation residues. By studying the available PKM2 crystalstructure (27), we selected all Ser and Thr residues located at thesurface of the protein (total of 12 additional residues), which aretheoretically accessible to OGT. However, mutating them to alaninedid not substantially decrease the level of PKM2O-GlcNAcylation asThr405 or Ser406 did (Fig. S2C). Therefore, both mass spectrometryand extensive mutation studies suggested that Thr405 and Ser406are the primary O-GlcNAcylation sites on PKM2.We next overexpressed wild-type PKM2 (PKM2WT) or

PKM2T405A/S406A in a series of tumor cells, includingMDA-MB-231,A375, U251, U2OS, T98-G, and HeLa. Compared with PKM2WT,PKM2T405A/S406A displayed greatly reducedO-GlcNAcylation (Fig. 2C).Next, we compared PKM2WT with PKM2T405A/S406A modification

upon OGT overexpression or O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarbamate (PUGNAc) treat-ment, which up-regulates O-GlcNAcylation by inhibitingO-GlcNAcase (24, 28) (Fig. 2D). In both cases, a clear el-evation of O-GlcNAcylation was detected in PKM2WT (Fig.2 E and F) or endogenous PKM2 (Fig. S3A), while poorlyO-GlcNAcylated PKM2T405A/S406A only exhibited a limited changeof O-GlcNAcylation under the same condition (Fig. 2 E andF). Stoichiometric analysis showed that, with PUGNAc, theO-GlcNAcylation level of PKM2 in MCF-7 cells was elevated to∼13% (Fig. S2B). We also examined the O-GlcNAc modification ofPKM2 in variable conditions. In response to a high supply of glucoseor glutamine, O-GlcNAcylation was up-regulated in PKM2WT, whileno evident changes were observed in PKM2T405A/S406A (Fig. S4).Together, these results suggest that Thr405 and Ser406 are thedominantO-GlcNAcylation sites on PKM2 across diverse tumor celltypes and that they are sensitive to fluctuations in nutritional supply.

O-GlcNAcylation Destabilizes PKM2 Tetramers and Reduces PKActivity. We then studied the impact of O-GlcNAcylation atThr405 and Ser406 on PKM2 structure and function. Based on thecrystal structure of PKM2 (27), we observed that Thr405 and Ser406

Fig. 1. PKM2 is highly O-GlcNAcylated in tumor cells and patient tumor tis-sues. (A) Detection of O-GlcNAc modification and PKM2 in human cell lines.The whole-cell lysates from indicated cells were analyzed by Western blotting(WB). All results shown in the study are representative of at least three in-dependent experiments. (B) Detection of PKM2O-GlcNAcylation in human celllines by the Click-iT O-GlcNAc enzymatic labeling system. O-GlcNAcylatedproteins were biotinylated, precipitated, and analyzed for PKM2 by WB. Rel-ative PKM2 O-GlcNAcylation level (Elusion) was normalized to the totalPKM2 protein level (Input) in each cell type. Error bars, SEM; n = 3 biologicalreplicas. One-way analysis of variance (ANOVA) and Bonferroni comparisonposttest: **P < 0.01; ***P < 0.001; ns, nonsignificant. (C) Analysis of PKM2O-GlcNAcylation in patient breast tumor tissues (T) and the matching adjacentnormal tissues (N) by the Click-iT system. (D) The quantification of PKM2O-GlcNAcylation level for C. Relative PKM2O-GlcNAcylation level (Elusion) wasnormalized to total PKM2 (Input). Median is shown; n = 29. Student’s t test:***P < 0.001. A.U., arbitrary units; GalT, β-1,4-galactosyltransferase.

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1704145115 Wang et al.

were located N-terminal to an α-helix which formed a prominentpart of the “C-C” interface during PKM2 tetramerization (29)(Fig. 3A). Thus, O-GlcNAcylation of Thr405 and Ser406 is likely todisrupt this tetramerization interface and thereby favors the di-meric or monomeric states of PKM2. Indeed, by modeling thePKM2 structure with O-GlcNAcylation at Thr405 and Ser406, wefound that one of the modified O-GlcNAc units, namely, that onThr405, can act as a direct barrier blocking a major stabilizing hy-drogen bond of the C-C dimerization interface (Fig. 3A), which isformed between a lysine (Lys422) from one monomer and a tyrosine(Tyr444) from the counterpart monomer (also known as the peg-in-hole mechanism) (29). The second unit Ser406, on the other hand,is likely to destabilize the oligomer further by interfering with otherhydrophobic interactions (Movie S1). Thus, this strongly predicts thatO-GlcNAc modification of PKM2 may induce disassociation of thetetramer by blocking the key interaction along the C-C interface.We tested this prediction by investigating the effects of

O-GlcNAcylation induced by PUGNAc or OGT overexpression onthe tetrameric state of PKM2. Whereas the treatment dramaticallyreduced the tetramer level of both PKM2WT (Fig. 3B and Fig. S5A) andendogenous PKM2 (Fig. S3B), the PKM2T405A/S406A tetramerswere unaffected in the presence of PUGNAc (Fig. 3B) or OGT(Fig. S5A). Moreover, the level of PKM2T405A/S406A tetramers wasclose to that of PKM2WT tetramers in the absence of PUGNAc (Fig.3B). This confirmed that O-GlcNAcylation of Thr405 and Ser406

was critical for the disassociation of the PKM2 tetramer inducedby PUGNAc or OGT overexpression. To further test the hy-pothesis that disruption of the Lys422–Tyr444 interaction wascentral to the destabilizing of PKM2 tetramers (29), we con-structed the single mutant PKM2K422A. As predicted, disrup-tion of the Lys422–Tyr444 interaction by mutating Lys422 resulted indestabilization of the PKM2 tetramer, like what we observed with theO-GlcNAcylation of Thr405 and Ser406 (Fig. 3C). Thus, these resultsfurther support our hypothesis that O-GlcNAcylation at Thr405

and Ser406 destabilizes PKM2 tetramers and leads to a shift of theoligomeric equilibrium toward dimers and monomers.It is known that tetramer conformation is required for the active

PK activity of PKM2 (14, 16). We found that both PUGNAc andOGT overexpression resulted in a lowered PKM2 activity in PKM2WT

as opposed to O-GlcNAcylation–deficient PKM2T405A/S406A

mutants (Fig. 3D and Figs. S3C and S5B). It should be noted thatthe mutant did not show an altered PKM2 activity with respect tothe wild type, indicating that the mutations alone do not causea loss of enzymatic activity (Fig. 3D and Figs. S5B and S6). To-gether, these data indicate that O-GlcNAc modification at Thr405

and Ser406 destabilizes PKM2 tetramers and reduces PK activity.

O-GlcNAcylation Is an Upstream Signal to Induce Nuclear Localizationof PKM2. In addition to the metabolic function in cytoplasm, recentstudies have shown that PKM2 can migrate into the nucleus to play a

Fig. 2. PKM2 is O-GlcNAcylated at Thr405 and Ser406.(A) Detection of the O-GlcNAcylation site(s) on PKM2.(A, Upper) A schematic depicting alternative splicing forPKM genes. The PKM1 and PKM2 isoforms are alter-natively spliced forms of the PKM gene. They differ bythe presence of either exon 9 (orange) in PKM1 or exon10 (green) in PKM2. Thr405 and Ser406 of PKM2 arederived from exon 10. (A, Lower) GST-tagged humanPKM2 and OGT were cotransfected in E. coli. TheO-GlcNAc-modified peptides were enriched for LC-MS/MS analysis. The data were processed by using theMASCOT engine, which identified the O-GlcNAc pep-tide (RLAPITSDPTEATAVGAVEASFK). The y and b frag-mentations were used to map theO-GlcNAcylation sitesto the Ser and Thr indicated in red. (B and C) Confir-mation of PKM2 O-GlcNAcylation at Thr405 and Ser406.Flag-tagged PKM2WT, PKM2T405A, PKM2S406A, orPKM2T405A/S406A was transfected into the indicated celllines and immunoprecipitated forWB. (D)O-GlcNAcylationis up-regulated by OGT and PUGNAc. MCF-7 cells weretransfected with HA-OGT or treated with PUGNAc.PKM2 and global O-GlcNAcylation were detected byWB. (E and F) OGT and PUGNAc up-regulate PKM2 O-GlcNAcylation through Thr405 and Ser406. MCF-7 cellsstably depleted for endogenous PKM2 by targetingshRNA and rescued by Flag-tagged PKM2WT orPKM2T405A/S406A overexpression were transfected withHA-tagged OGT or treated with PUGNAc. Flag-PKM2–associated proteins were immunoprecipitated andanalyzed by WB. Relative PKM2 O-GlcNAcylation levelwas normalized to the total PKM2 protein level foreach group. A.U., arbitrary units. Error bars, SEM; n =3 independent assays. Student’s t test: **P < 0.01; ns,nonsignificant.

Wang et al. PNAS Early Edition | 3 of 6

CELL

BIOLO

GY

nonmetabolic role in cancer cells (15, 16, 20, 30, 31). The migrationis facilitated by the binding of importin α5, a nuclear translocationfactor, to the nuclear localization signal (NLS) sequence of PKM2(9, 15). We examined the impact of O-GlcNAcylation on the dis-tribution of PKM2 in cells. Compared with cytoplasmic PKM2,nuclear PKM2 remained in a small amount (Fig. S7); however, itincreased dramatically upon boosting of O-GlcNAcylation ofPKM2 by PUGNAc (Fig. 4A and Figs. S3D and S7) or OGToverexpression (Fig. S5C). By contrast, the level of PKM2T405A/S406A

was lower in the nucleus regardless of PUGNAc treatment (Fig. 4Aand Fig. S7) or OGT overexpression (Fig. S5C). Consistentwith this, immunofluorescence demonstrated that both PKM2WT

(Fig. 4B) and endogenous PKM2 (Fig. S3E), but not PKM2T405A/S406A

(Fig. 4B), accumulated in the nucleus in response to PUGNActreatment. These suggest that the O-GlcNAcylation at Thr405 andSer406 is important for the nuclear translocation of PKM2.We next examined whether O-GlcNAcylation–induced PKM2

nuclear translocation involves the interaction of PKM2-NLS withimportin α5. We found that PKM2WT bound more importin α5 thanPKM2T405A/S406A did, and the up-regulation of O-GlcNAcylation byPUGNAc treatment further improved importin α5 association withPKM2WT (Fig. 4C). Interestingly, structural analysis revealed thatat the tetramer stage, key residues Arg399 and Arg400 of the NLS

sequence were buried at the C-C tetramerization interface (Fig. S8 Aand B) and could not be accessed by importin α5. However, asO-GlcNAcylation induces detetramerization of PKM2, the NLSregion is probably exposed to recruit importin α5 which facilitatesthe nuclear translocation of PKM2.ERK2-dependent phosphorylation of PKM2 at Ser37 has been

demonstrated to be required for its nuclear translocation (15).Hence, we wondered if there might be a cross-talk betweenO-GlcNAcylation and phosphorylation on PKM2. As previouslyreported, the PKM2 mutant whose Ser37 was replaced by alanine(PKM2S37A) failed to efficiently transport into the nucleus (Fig.4 A and B). Interestingly, the mutation at Ser37 also abolishedthe PKM2 nuclear translocation induced by O-GlcNAcylationupon PUGNAc treatment (Fig. 4 A and B). This implied that thenuclear translocation of PKM2 induced by O-GlcNAcylationrelies on the phosphorylation of Ser37. Furthermore, we foundthat when O-GlcNAcylation was increased in PKM2WT byPUGNAc treatment, PKM2-bound ERK2 responsible for Ser37

phosphorylation was increased, with concomitant increase inPKM2WT phosphorylation at Ser37 (Fig. 4D and Fig. S3F). Inparallel, the impairment of O-GlcNAcylation in PKM2T405A/S406A

caused decreased binding of ERK2 to PKM2 and a clear reductionof Ser37 phosphorylation regardless of PUGNAc (Fig. 4D). Incontrast, O-GlcNAcylation of PKM2 was largely unaffected whenPKM2 Ser37 was mutated to alanine or aspartate (Fig. 4E). Fur-thermore, we observed increased levels of PKM2 O-GlcNAcylation

Fig. 3. O-GlcNAcylation destabilizes PKM2 tetramers. (A) Modeling ofPKM2 structure with O-GlcNAcylation. (A, Upper) Dynamic equilibrium amongmonomers, dimers, and tetramers of PKM2. (A, Lower) Structural modelingshows that O-GlcNAc (blue) at Thr405 and Ser406 (green) disrupts the key sta-bilizing hydrogen bond between Lys422 (red) and Tyr444 (orange). (B and C)Change of PKM2 oligomer states upon O-GlcNAcylation. Transfection of Flag-tagged PKM2WT, PKM2T405A/S406A, or PKM2K422A along with or without con-comitant PUGNAc treatment was conducted in PKM2-depleted MCF-7 cells.The whole-cell lysates were analyzed by WB. GA, glutaraldehyde. (D)O-GlcNAcylation regulates PKM2 activity. Enzymatic activity of PKM2WT

and PKM2T405A/S406A was assayed. Error bars, SEM; n = 3 independent as-says. Student’s t test: *P < 0.05; ns, nonsignificant.

Fig. 4. O-GlcNAcylation promotes PKM2 nuclear translocation. Flag-taggedPKM2WT, PKM2T405A/S406A, PKM2S37A, or PKM2S37D was transfected inPKM2 depleted MCF-7 cells along with or without concomitant PUGNActreatment. (A) Fractionation analysis of PKM2. Nuclear, cytoplasmic fractions,and whole lysates were analyzed by WB. Lamin B1 and GAPDH are controls innuclear and cytosolic fractions. (B, Left) Cellular localization of PKM2 wild typeand mutants by immunofluorescence. (B, Left, Insets) Magnified nucleus. (Scalebar, 10 μm.) (B, Right) Percentage of cells positive for PKM2 in the nucleus.Error bars, SEM; n = 30 (10 fields from each of the three independent exper-iments). Student’s t test: ***P < 0.001; ns, nonsignificant. (C) The association ofPKM2 with importin α5. Proteins immunoprecipitated by Flag-PKM2 wereanalyzed by WB. (D and E) Interplay between O-GlcNAcylation and phos-phorylation on PKM2. Flag-PKM2–associated proteins were immunoprecipi-tated and analyzed by WB with the indicated antibodies.

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1704145115 Wang et al.

upon EGF treatment, which is known to boost Ser37 phosphory-lation through activating ERK signaling (Fig. S8C). However,in PKM2T405A/S406A, EGF-induced Ser37 phosphorylation wascompletely abolished (Fig. S8C). Collectively, these data showthat O-GlcNAcylation at Thr405 and Ser406 is an event upstreamof Ser37 phosphorylation.

PKM2 O-GlcNAcylation Promotes the Warburg Effect and Cell Proliferation.Having established that O-GlcNAcylation impairs the PKM2 ac-tivity in the cytoplasm and triggers its nuclear translocation, wenext wanted to know whether these O-GlcNAcylation–inducedchanges would promote theWarburg effect. Firstly, reduced PKM2activity might cause the accumulation of upstream glycolytic me-tabolites that can be used for lipid and nucleic acid synthesis (2, 4,32). Indeed, we found that the rate of lipid and DNA synthesiswas greatly up-regulated in cells expressing PKM2WT com-pared with that in cells carrying PKM2T405A/S406A (Fig. S9A).Thus, O-GlcNAcylation–induced reduction of PKM2 activitymay contribute to the Warburg effect by redirecting glycolyticflux toward anabolic pathways.Secondly, PKM2 migration into nucleus has been shown to

induce a HIF-1α– and c-Myc–mediated expression of a series ofglycolysis-related regulators, including glucose transporter 1 (Glut1)and lactate dehydrogenase A (LDHA), which are directly re-sponsible for the regulation of glucose consumption and lactateproduction (15). Indeed, we found that PKM2 O-GlcNAcylationsignaling was associated with the expression of Glut1 and LDHA:(i) upon PUGNAc treatment, PKM2WT enhanced both mRNA andprotein levels of Glut1 and LDHA, whereas PKM2T405A/S406A failedto do so (Fig. 5 A and B); and (ii) mutations of residues Arg399 andArg400 of the NLS sequence abolished PUGNAc-induced expres-sion of Glut1 and LDHA (Fig. S9 B and C). As expected, glucoseconsumption and lactate production were significantly in-creased by PKM2WT, but not PKM2T405A/S406A, in the presenceof PUGNAc (Fig. S9D). This induction could be mediated by c-Myc, as both mRNA and protein levels of c-Myc were increasedwith PKM2WT, but not PKM2T405A/S406A (Fig. S9 E and F), andan interaction of PKM2WT with the c-Myc promoter region wasalso observed (Fig. S9G). These observations demonstratedthat O-GlcNAcylation of PKM2 at Thr405 and Ser406 promotedthe Warburg effect by up-regulating the expression of Glut1and LDHA through a nuclear role via c-Myc. Together, weconclude that O-GlcNAcylation of Thr405 and Ser406 regulatesboth metabolic and nonmetabolic function of PKM2 to promotethe Warburg effect.We further investigated the impact of PKM2 O-GlcNAcyla-

tion on cell proliferation by constructing stable MCF-7 cancercell lines depleted for PKM2 and rescued by either PKM2WT

or PKM2T405A/S406A. Of note, the expression levels of exoge-nous PKM2 in these rescue cell lines were comparable (Fig. 5C).Compared with cells expressing PKM2WT, PKM2T405A/S406A rescuecells displayed a lower growth rate (Fig. 5D). We next injected nudemice with these rescue cell lines and assayed for tumor formation.Compared with mice injected with PKM2WT, mice injected withPKM2T405A/S406A exhibited decreased tumor volume and mass(Fig. 5 E–G). In line with the impaired tumor growth in mice injectedwith PKM2T405A/S406A, Ki-67, the classic marker of cell pro-liferation, was significantly decreased in those tumor tissues(Fig. 5H). These data support that O-GlcNAcylation of PKM2 atThr405 and Ser406 confers a growth advantage for tumor cells in vivo.

DiscussionHere, we have uncovered a previously unknown mechanism forregulation of PKM2 function by O-GlcNAcylation, which simulta-neously impinges on both metabolic and nuclear (nonmetabolic)functions of PKM2 to promote the Warburg effect. On the onehand, O-GlcNAcylation–dependent detetramerization of PKM2causes reduced PK activity, which directly rewires metabolic fluxestoward anabolic pathways for rapid cell proliferation. On the otherhand, O-GlcNAcylation at Thr405 and Ser406 directly destabilizesPKM2 tetramers to facilitate the exposure of NLS and downstream

Ser37 phosphorylation for PKM2 nuclear localization. This in turnstimulates c-Myc–dependent expression of two key glycolysis com-ponents, namely, Glut1 and LDHA, to enhance glucose consump-tion and lactate production.Notably, the O-GlcNAcylation residues of PKM2, Thr405 and

Ser406, are encoded by the alternatively spliced exon 10 and

Fig. 5. PKM2 O-GlcNAcylation is crucial for the Warburg effect and tumorgrowth. (A) O-GlcNAcylation of PKM2WT enhances the mRNA level of Glut1(Left) and LDHA (Right). Error bars, SEM; n = 4 biological replicas. Student’st test: ***P < 0.001; ns, nonsignificant. (B) Protein levels of Glut1 and LDHAwere induced by PKM2 O-GlcNAcylation. (C) Establishment of PKM2 rescuecell lines. MCF-7 cells depleted for endogenous PKM2 were stably reex-pressed with exogenous PKM2WT or PKM2T405A/S406A. (D) Proliferation curvesof PKM2WT and PKM2T405A/S406A rescue cell lines. PKM2WT and PKM2T405A/S406A

rescue cells were seeded at the same number in each well. Cell numbers werecounted every 24 h. Error bars, SEM. Student’s t test: *P < 0.05; ***P < 0.001; ns,nonsignificant. (E–G) Tumor formation in nude mice. PKM2WT and PKM2T405A/S406A

rescue cells were, respectively, injected into athymic nude mice. The xeno-graft tumors were sampled and photographed after 24 d. Images of micebearing tumor are as in E. Tumor volumes were measured at the indicatedtime points as in F. According to length (l), width (w), and height (h), vol-umes were calculated based on the equation v = lwhπ/6. The quantificationof the average mass of xenograft tumors was as in G. Mean is shown, n = 9.Student’s t test: *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant. (H)Immunohistology analysis for cell proliferation. Nine pairs of mice injectedwith PKM2WT or PKM2T405A/S406A rescue cells were costained with hema-toxylin and eosin (H&E) and Ki-67. (H, Left) Representative images depictingtumor tissues. (Scale bars, 50 μm.) (H, Right) Quantification of Ki-67 staining.Error bars, SEM; n = 15 (five fields from each of the three analyzed mice).Student’s t test: **P < 0.01.

Wang et al. PNAS Early Edition | 5 of 6

CELL

BIOLO

GY

located at the so-called C-C dimer interface of the PKM2 tet-ramer (29). Thus, O-GlcNAcylation appears to be a uniqueregulatory mode for PKM2 function which serves as a bridgetranslating genomic gene expression switch from PKM1 toPKM2 toward metabolic reprogramming. By providing anO-GlcNAcylation platform that can adjust the PK structure, theselective expression of PKM2 in rapidly proliferating cells mightthus confer certain advantages in the dynamic control of theWarburg effect.Two metabolic enzymes, PFK1 and G6PD, are known to be

regulated by O-GlcNAc modification (23, 24). O-GlcNAcylation ofPFK1 and G6PD redirects metabolic flux through the pentosephosphate pathway (PPP) and rebalances the metabolic need forribose 5-phosphate and NADPH (23, 24). Depending on thedemands for NADPH, ribose 5-phosphate, and ATP, cells areable to select the proper PPP metabolic mode to fine-tune glucose-6-phosphate flux (33). Given that PK catalyzes the last irreversiblereaction in glycolysis, PKM2 glycosylation probably facilitates globalmetabolic readjustment by redirecting metabolites to branching an-abolic pathways to meet the extensive needs of proliferating cells. Inaddition, compared with phosphorylation, acetylation, andSUMOylation, O-GlcNAcylation has been recognized as a uniquePTM sensing fluctuation in nutrient environment (28, 34–37). Bycontrolling PKM2 structure and function, O-GlcNAcylation cou-ples metabolic state to a dynamically changing environment. Thispreviously uncharacterized mechanism represents an effi-cient means for cells to coordinate metabolic regulation withnutritional dynamics.Thus, our findings furnish a key element for unraveling the

Warburg effect in cancer biology. Alteration of the PKM2 oligo-meric state by O-GlcNAcylation allows simultaneous control of the

metabolic and nuclear (nonmetabolic) roles of this key enzyme inproliferating cells. Importantly, we demonstrate the ubiquity of thisregulatory mode across a wide range of cancer cell types and pro-vide direct evidence for an in vivo role in tumor proliferation. Theseobservations make the O-GlcNAcylation apparatus a target of greatinterest, both for understanding tumor metabolism and for potentialintervention to limit tumor growth.

Materials and MethodsReagents, cell lines, patient tissues, cell fractionations and immunoprecipita-tion, PKM2 O-GlcNAcylation assay, PKM2 structure modeling, PKM2 activityassay, lipid and DNA synthesis assay, and immunohistochemistry analysis aredescribed in detail in SI Materials and Methods. Human blood samples wereobtained from the Jilin Blood Center according to the Standard for HealthExamination of Blood Donors (GB18467-2011). Breast tumor tissues andmatching tumor-adjacent normal tissues from patients were obtained fromthe Tissue Bank of China–Japan Union Hospital of Jilin University during sur-gery and stored at −80 °C. The present study was approved by the EthicalCommittee of Jilin Blood Center and China–Japan Union Hospital of JilinUniversity and conducted with the informed consent of all donors and pa-tients. All animal work procedures were approved by the Animal Care Com-mittee of the Northeast Normal University (Changchun, China).

ACKNOWLEDGMENTS. We thank Drs. Weiwei Yang and Gerald W. Hart forproviding useful reagents; Drs. Xianling Cong & Miao Hao (China-JapanUnion Hospital of Jilin University) and Dr. Lin Chen (Jilin Blood Center) forhuman tissues and blood samples; and Drs. Qunying Lei, Baiqu Huang, BaoLiu, Xianlu Zeng, Jun Lu, Shucai Wang, and Yuzhu Dong for critical com-ments. This work was supported by National Natural Science Foundation ofChina Grants 31170769 and 31270916; Program for Introducing Talents toUniversities Grant B07017; and the Fundamental Research Funds for theCentral Universities. L.A. is supported by the intramural research programof the National Library of Medicine, National Institutes of Health.

1. Sullivan LB, Gui DY, Heiden MGV (2016) Altered metabolite levels in cancer: Impli-cations for tumour biology and cancer therapy. Nat Rev Cancer 16:680–693.

2. Cairns RA, Harris IS, Mak TW (2011) Regulation of cancer cell metabolism. Nat RevCancer 11:85–95.

3. Koppenol WH, Bounds PL, Dang CV (2011) Otto Warburg’s contributions to currentconcepts of cancer metabolism. Nat Rev Cancer 11:325–337.

4. Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the Warburgeffect: The metabolic requirements of cell proliferation. Science 324:1029–1033.

5. Pavlova NN, Thompson CB (2016) The emerging hallmarks of cancer metabolism. CellMetab 23:27–47.

6. Lunt SY, Vander Heiden MG (2011) Aerobic glycolysis: Meeting the metabolic re-quirements of cell proliferation. Annu Rev Cell Dev Biol 27:441–464.

7. Christofk HR, et al. (2008) The M2 splice isoform of pyruvate kinase is important forcancer metabolism and tumour growth. Nature 452:230–233.

8. Dayton TL, Jacks T, Vander Heiden MG (2016) PKM2, cancer metabolism, and the roadahead. EMBO Rep 17:1721–1730.

9. Luo W, et al. (2011) Pyruvate kinase M2 is a PHD3-stimulated coactivator for hypoxia-inducible factor 1. Cell 145:732–744.

10. Noguchi T, Yamada K, Inoue H, Matsuda T, Tanaka T (1987) The L- and R-type iso-zymes of rat pyruvate kinase are produced from a single gene by use of differentpromoters. J Biol Chem 262:14366–14371.

11. Noguchi T, Inoue H, Tanaka T (1986) The M1- and M2-type isozymes of rat pyruvatekinase are produced from the same gene by alternative RNA splicing. J Biol Chem 261:13807–13812.

12. Jurica MS, et al. (1998) The allosteric regulation of pyruvate kinase by fructose-1,6-bisphosphate. Structure 6:195–210.

13. Chaneton B, et al. (2012) Serine is a natural ligand and allosteric activator of pyruvatekinase M2. Nature 491:458–462.

14. Lv L, et al. (2011) Acetylation targets the M2 isoform of pyruvate kinase for degra-dation through chaperone-mediated autophagy and promotes tumor growth. MolCell 42:719–730.

15. Yang W, et al. (2012) ERK1/2-dependent phosphorylation and nuclear translocationof PKM2 promotes the Warburg effect. Nat Cell Biol 14:1295–1304.

16. Lv L, et al. (2013) Mitogenic and oncogenic stimulation of K433 acetylation promotesPKM2 protein kinase activity and nuclear localization. Mol Cell 52:340–352.

17. Hitosugi T, et al. (2009) Tyrosine phosphorylation inhibits PKM2 to promote theWarburg effect and tumor growth. Sci Signal 2:ra73.

18. Spoden GA, et al. (2009) The SUMO-E3 ligase PIAS3 targets pyruvate kinase M2. J CellBiochem 107:293–302.

19. Anastasiou D, et al. (2011) Inhibition of pyruvate kinase M2 by reactive oxygen speciescontributes to cellular antioxidant responses. Science 334:1278–1283.

20. Yang W, et al. (2011) Nuclear PKM2 regulates β-catenin transactivation upon EGFRactivation. Nature 480:118–122.

21. Gui DY, Lewis CA, Vander Heiden MG (2013) Allosteric regulation of PKM2 allowscellular adaptation to different physiological states. Sci Signal 6:pe7.

22. Clark PM, et al. (2008) Direct in-gel fluorescence detection and cellular imaging ofO-GlcNAc-modified proteins. J Am Chem Soc 130:11576–11577.

23. Rao X, et al. (2015) O-GlcNAcylation of G6PD promotes the pentose phosphatepathway and tumor growth. Nat Commun 6:8468.

24. Yi W, et al. (2012) Phosphofructokinase 1 glycosylation regulates cell growth andmetabolism. Science 337:975–980.

25. Khidekel N, et al. (2003) A chemoenzymatic approach toward the rapid and sensitivedetection of O-GlcNAc posttranslational modifications. J Am Chem Soc 125:16162–16163.

26. Rexach JE, et al. (2010) Quantification of O-glycosylation stoichiometry and dynamicsusing resolvable mass tags. Nat Chem Biol 6:645–651.

27. Dombrauckas JD, Santarsiero BD, Mesecar AD (2005) Structural basis for tumor py-ruvate kinase M2 allosteric regulation and catalysis. Biochemistry 44:9417–9429.

28. Ferrer CM, et al. (2014) O-GlcNAcylation regulates cancer metabolism and survivalstress signaling via regulation of the HIF-1 pathway. Mol Cell 54:820–831.

29. Morgan HP, et al. (2013) M2 pyruvate kinase provides a mechanism for nutrientsensing and regulation of cell proliferation. Proc Natl Acad Sci USA 110:5881–5886.

30. Hamabe A, et al. (2014) Role of pyruvate kinase M2 in transcriptional regulationleading to epithelial-mesenchymal transition. Proc Natl Acad Sci USA 111:15526–15531.

31. Wang HJ, et al. (2014) JMJD5 regulates PKM2 nuclear translocation and reprogramsHIF-1α-mediated glucose metabolism. Proc Natl Acad Sci USA 111:279–284.

32. Christofk HR, Vander Heiden MG, Wu N, Asara JM, Cantley LC (2008) Pyruvate kinaseM2 is a phosphotyrosine-binding protein. Nature 452:181–186.

33. Stincone A, et al. (2015) The return of metabolism: Biochemistry and physiology ofthe pentose phosphate pathway. Biol Rev Camb Philos Soc 90:927–963.

34. Hart GW, Slawson C, Ramirez-Correa G, Lagerlof O (2011) Cross talk betweenO-GlcNAcylation and phosphorylation: Roles in signaling, transcription, and chronicdisease. Annu Rev Biochem 80:825–858.

35. Hart GW, Housley MP, Slawson C (2007) Cycling of O-linked beta-N-acetylglucosamineon nucleocytoplasmic proteins. Nature 446:1017–1022.

36. Wells L, Vosseller K, Hart GW (2001) Glycosylation of nucleocytoplasmic proteins:Signal transduction and O-GlcNAc. Science 291:2376–2378.

37. Caldwell SA, et al. (2010) Nutrient sensor O-GlcNAc transferase regulates breastcancer tumorigenesis through targeting of the oncogenic transcription factor FoxM1.Oncogene 29:2831–2842.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1704145115 Wang et al.

Supporting InformationWang et al. 10.1073/pnas.1704145115SI Materials and MethodsDNA Constructs and Cell Lines. PCR-amplified human PKM2 cDNAwas cloned either into the p3×Flag-CMV-10 vector betweenEcoRI and BamHI or into the pCDH/hygro vector betweenBamHI and NotI. Mutations, including PKM2T405A, PKM2S406A,PKM2T405A/S406A, PKM2S37A, PKM2S37D, PKM2K422A andPKM2R399A/R400A, PKM2T41A, PKM2T80A, PKM2T93A, PKM2S97A,PKM2S100A, PKM2S127A, PKM2T129A, PKM2S249A, PKM2S519A,PKM2S333A, PKM2T341A, and PKM2T346A, were made by using theQuikChange Site-Directed Mutagenesis Kit (TransGen Biotech).pCDH/hygro vector, pGIPZ/puro-shControl, and -shPKM2 werekindly provided by Weiwei Yang, Institute of Biochemestry andCell, Shanghai Institute for Biological Sciences, Chinese Academyof Sciences, Shanghai, China (1, 2). pcDNA3.1/HA-OGT wasgenerously provided by Gerald W. Hart, School of Medicine, JohnsHopkins University, Baltimore (3, 4). Transient transfection ofDNA constructs was performed by using Lipofectamine 2000 re-agents (Invitrogen) according to the vendor’s instructions.HumanMCF-10A immortalizedmammary epithelial cells were

cultured in DMEM/F12 medium (Sigma), supplemented with 5%horse serum (Gibco), 100 mg/mL epidermal growth factor (EGF;R&D), 1 mg/mL hydrocortisone (Sigma), 1 mg/mL cholera toxin(Sigma), and 10 mg/mL insulin (Sigma). Human MDA-MB-231 breast cancer cells were cultured in L15 medium (Sigma),supplemented with 10% FBS (Gibco) and 1% penicillin/strep-tomycin (Thermo Fisher Scientific). Human MCF-7 breast can-cer cells, A375 melanoma cells, HeLa cervical carcinoma cells,T98-G and U251 glioblastoma cells, U2OS osteosarcoma cells,and human embryonic kidney 293T (HEK-293T) cells werecultured in DMEM (Sigma) and supplemented with 10% FBSand 1% penicillin/streptomycin. Human U937 histiocytic lym-phoma cells and Jurkat leukemia T cells were cultured in RPMImedium 1640 (Sigma), supplemented with 10% FBS and 1%penicillin/streptomycin. All cell lines mentioned above werepurchased from the Cell Bank of Type Culture Collection of theChinese Academy of Sciences.The PKM2-depleted cell line was generated by lentiviral

transduction in MCF-7 cells. Briefly, lentivirus particles weregenerated in HEK-293T cells by cotransfection of the plasmidpsPAX and Pmd2.G along with either pGIPZ-shPKM2 or-shControl. Virus supernatant supplemented with polybrene(Sigma) was used to infect MCF-7 cells. Infected cells were thensubjected to DMEM containing puromycin (2 μg/mL; Sigma) for9 d selection before the expression test. To generate thePKM2WT and PKM2T405A/S406A rescue cell lines, pCDH-PKM2WT or -PKM2T405A/S406A along with psPAX and Pmd2.Gwere cotransfected into HEK-293T to prepare the lentiviruses.MCF-7 cells depleted for PKM2 were infected with lentivirusescontaining PKM2WT or PKM2T405A/S406A in the presence ofpolybrene and selected with puromycin (2 μg/mL; Sigma) andhygromycin (400 μg/mL; Sigma) for 2 wk.

Human T Cells and Breast Tissues. Naive T cells were purified fromwhole blood of healthy human subjects based on immuno-magnetic separation by AutoMACS System (Miltenyi Biotech).The blood was obtained from the Jilin Blood Center (Changchun,China). Isolated cells were plated at a density of 1 × 106 cells per mLin RPMI medium 1640, supplemented with 10% FBS and 1%penicillin/streptomycin. Activation and expansion of T cells wereperformed by using 30 U/mL recombinant human interleukin-2(Invitrogen) and Dynabeads Human T Activator CD3/CD28(Invitrogen), according to the manufacturer’s recommendation.

Following a 72-h incubation in a humidified CO2 incubatorat 37 °C, activated cells were confirmed by flow cytometry(FACSCantoII; BD). Breast tumor tissues and matching tumor-adjacent normal tissues from patients were obtained from theChina–Japan Union Hospital of Jilin University during surgeryand stored at −80 °C.

Click-iT O-GlcNAc Enzymatic Labeling and Stoichiometric Analysis ofO-GlcNAcylated PKM2. Chemoenzymatic labeling of proteins intotal cell lysates were carried out as described (5–11). Briefly,cell lysates (80–200 μg) were labeled by utilizing the Click-iTO-GlcNAc Enzymatic Labeling System (Invitrogen). The per-missive mutant β-1,4-galactosyltransferase was responsible forthe transfer of azido-modified galactose (GalNAz) from UDP-GalNAz to O-GlcNAc residues on O-GlcNAcylated proteins.Modified proteins were detected by utilizing the Click-iT BiotinProtein Analysis Detection Kit protocol (Invitrogen). Bio-tinylated proteins were resolubilized in binding buffer (0.1 Mphosphate, 0.15 M NaCl, 0.1% SDS, and 1% Nonidet P-40,pH 7.2). An appropriate amount of streptavidin resin (Pierce)was added to incubate with the mixture overnight at 4 °C. Thestreptavidin-bound complex was washed with binding buffer.Following the removal of supernatants, pellets were eluted byboiling with loading buffer (2% SDS, 10% glycerol, 2.5%2-mercaptoethanol, and 62.5 mM Tris·HCl, pH 6.8). For quan-tifying the O-GlcNAc stoichiometry of individual proteins, theintensity of the total PKM2 protein band (Input) and theO-GlcNAcylated PKM2 protein band (Elution) were measured.

Stoichiometric Analysis Using an Alkyne-5K-PEG Resolvable Mass Tag.O-GlcNAcylated proteins were labeled with GalNAz on theO-GlcNAc residues by the Click-iT O-GlcNAc Enzymatic La-beling System described above. Modified proteins were pre-cipitated with methanol, and the resulting proteins wereresuspended in 100 μL of PBS containing 0.4% SDS and 6 mMAlkyne-5K-PEG (polyethylene glycol; Sigma), premixed withBTTAA-CuSO4 complex (50 μM CuSO4; BTTAA:CuSO4 with2:1 molar ratio), and subsequently mixed with 2.5 mM freshlyprepared sodium ascorbate for 4 h at 37 °C. As O-GlcNAcmodified proteins in the lysate were labeled with a 5KD PEGmass tag, a shift of O-GlcNAcylated proteins in molecular masswas thus detectable in WB. Meanwhile, unlabeled lysates werediluted with loading buffer and loaded as controls for thequantitative evaluation of desired proteins.

Up- and Down-Regulation of O-GlcNAcylation. PUGNAc (150 μM,24 h; TRC) was used to up-regulate O-GlcNAc modification inthe desired condition, while OSMI-1 (50 μM, 24 h; Sigma) (12,13) and 6-diazo-5-oxo-L-norleucine [(DON) 100 μM, 24 h;Sigma] (14, 15) were used to down-regulate O-GlcNAcylation.

GST Pull-Down Assay. Following the cotransfection of GST-PKM2 and OGT, E. coli cells were induced by isopropyl-beta-D-thiogalactopyranoside (1 mM) at 16 °C for 4 h. Subsequently, thebacteria cells were resuspended in lysis buffer containing 50 mMTris·HCl, pH 8.0, 120 mM NaCl, 1 mM DTT, and 10% glycerol,plus 1 mM PMSF. After sonication and centrifugation, super-natant of the cell lysates was bound to GSTSep GlutathioneAgarose Resin (GE Healthcare) rolling at 4 °C overnight. Theagarose resin was washed extensively with lysis buffer beforeeluting for 1 h with lysis buffer (pH 7.5) containing 20 mMglutathione. Eluted proteins were then dialyzed extensively

Wang et al. www.pnas.org/cgi/content/short/1704145115 1 of 11

against dialysis buffer (50 mM NaCl, 10% glycerol, and 1 mMDTT, and 20 mM Tris·Cl, pH 8.0).

Mass Spectrometry Analysis of PKM2 O-GlcNAcylation Sites. Recombi-nantO-GlcNAcylated PKM2 proteins were analyzed by LC-MS/MS(Applied Protein Technology). Briefly, O-GlcNAcylated PKM2proteins were enriched by WGA lectin (Glycoprotein Isolation KitWGA; Thermo Scientific). The eluted O-GlcNAcylated protein(∼300 μg) was solubilized in 30 mL of SDT buffer (4% SDS,100 mM DTT, and 150 mM Tris·HCl, pH 8.0) at 90 °C for 5 min.DTT and iodoacetamide were added to reduce and block the cys-teine residues. Then trypsin was added and incubated at 37 °C for16–18 h. The digested product was performed on a Q Exactive massspectrometer that was coupled to Easy nLC (Thermo Fisher Sci-entific). The ESI mass spectrometry data were analyzed by usingMascot 2.2 MASCOT engine (Version 2.2; Matrix Science) againstthe human PKM database. For protein identification, the followingoptions were used: peptide mass tolerance = 20 ppm; MS/MStolerance = 0.1 Da; enzyme = trypsin; missed cleavage = 2;fixed modification: carbamidomethyl (Cys); variable modifica-tion: oxidation (Met); O-GlcNAc (Ser/Thr). The result wasfiltered by score >20.

Immunocytochemistry and Microscopy. Cells were fixed with 4%formaldehyde for 5 min at room temperature and then per-meabilized with 0.1% Triton X-100 for 3 min. PBS buffer with10% FBS was used to block nonspecific protein–protein inter-actions. Incubation with the primary antibody at 4 °C overnightwas followed by secondary antibody (Invitrogen) labeling. Cellnuclei were counterstained with 49,6-diamidino-2-phenylindole(Sigma). Images were taken by using a fluorescent confocalmicroscopy (Olympus FV1000). All images were acquired underroom temperature with the same settings and adjusted forbrightness and contrast identically by using Adobe PhotoshopCS5 software.

PKM2 Structure Modeling. The human PKM2 structure (ProteinData Bank ID code 1T5A) was used as a template to model theO-GlcNAcylation of Thr405 and Ser406 residues. The model wasbuilt using the GlyCam server (glycam.org/) and refined throughenergy minimization using AMBER 12 (ambermd.org/). Allos-Mod simulation was used to study conformations and movementof modified units (https://modbase.compbio.ucsf.edu/allosmod/).Pymol (https://pymol.org/2/) was used to conduct structuralanalysis and generate the structural figures.

PKM2 Oligomerization Assay. The whole-cell lysates (4 mg/mL)were cross-linked with 0.025% glutaraldehyde for 3 min at 37 °Cand terminated with Tris·HCl (pH 8.0, 50 mM). Subsequently,samples were analyzed by Western blotting with indicatedantibodies.

Cell Fractionations and Immunoprecipitation. Cells were lysed incold WB-IP lysis buffer (Beyotime) for 30 min and centrifuged(4 °C, 5 min at 14,000 × g) to remove cell debris. The supernatantwas collected and used as the whole-cell lysates. Nuclear andcytoplasmic fractions were extracted with the Nuclear and Cy-toplasmic Protein Extraction Kit (Beyotime) according to thevendor’s instructions. Flag-PKM2 protein was immunoprecipi-tated from whole-cell lysates by using anti-Flag antibody cou-pling to Protein A/G PLUS Agarose beads (Santa Cruz) for 3 hat 4 °C. Nonspecific mouse IgG antibody (Boster) was used as anegative control. Immunoprecipitated proteins were analyzed byWestern blotting.Antibodies used in this study include PKM2 (4053S; Cell

Signaling Technology; CST), GAPDH (5174; CST), Lamin B1(12586; CST), Ki-67 (9449; CST), HA (3274; CST), ERK1/2(4695S; CST), pERK1/2 (9101S; CST), LDHA (32182; Signal-

ingway antibody; SAB), PKM2 pS37 (11456; SAB), Glut1 (40967;SAB), Importin α5 (25191; SAB), Flag (F1804; Sigma), β-Actin(A4700; Sigma), β-Tubulin (T0198; Sigma), and O-GlcNAcyla-tion (CTD110.6; Sigma, O7764).

RNA Extraction and Quantitative Real-Time PCR. Total RNA wasextracted by using amammalian total RNAMiniprepKit (Sigma).cDNA was prepared with oligonucleotide (dT), random primers,and the Reverse Transcription Kit (Takara). Quantitative real-time PCR (qPCR) analysis was performed by using the SYBRReal-Time PCR Premixture (Takara) under the following con-ditions: 5 min at 95 °C followed by 40 cycles at 95 °C for 30 s,55 °C for 40 s, and 72 °C for 1 min by using a Roche LightCycler480 sequence detection system. Data were normalized to the expres-sion of β-actin ineach experiment. The following primer pairs wereused for qPCR: Glut1, 5′-CGGGCCAAGAGTGTGCTAAA-3′(forward) and 5′-TGACGATACCGGAGCCAATG-3′ (reverse);LDHA, 5′-TTGGAGGGCAGCACCTTACTTAGA-3′ (forward)and 5′-GCCTTAAGTGGAACAGCTATGCTGAC-3′ (reverse);c-Myc, 5′-ACACCCTTCTCCCTTCG-3′ (forward) and 5′-CCGCT-CCACATACAGTCC-3′ (reverse); and β-actin 5′-ATGGATGAC-GATATCGCTGCGC-3′ (forward) and 5′-GCAGCACAGGGTG-CTCCTCA-3′ (reverse).

ChIP Assay. ChIP was performed by using the SimpleChIP PlusEnzymatic Chromatin IP Kit (CST). Chromatin from 2 × 107 cellswas incubated with antibody against flag or IgG overnight. Thehuman c-Myc promoter specific primers used in qPCR were5′-CAGCCCGAGACTGTTGC-3′ (forward) and 5′-CAGAGCGTGG-GATGTTAG-3′ (reverse).

Glucose Consumption and Lactate Production Assay. After desiredtreatment, medium from cultured cells was collected for the assayof glucose and lactate. Glucose levels were determined by using aglucose (GO) assay kit (Sigma). Glucose consumption was thedifference in glucose concentration compared with DMEM.Lactate level was determined by using fluorescence-based assaykits (BioVision) according to the manufacturer’s protocols.

PKM2 Activity Assay. Whole-cell lysates or PKM2 proteinsimmunoprecipitated from the whole-cell lysates were incubatedwith reaction buffer (30 μM pyruvate, 6.6 μM NADH, and 0.2 MTris·HCl, pH 7.3) in the presence of FBP (500 μM) for 30 min atroom temperature. PK activity was then measured with acolorimetric-based PK activity assay kit (BioVision) according tothe manufacturer’s protocol.

Lipid and DNA Synthesis Assay. For lipid staining, 2 × 105 cellswere resuspended in 200 μL of BODIPY 493/503 (Invitrogen)at 0.5 mg/mL in PBS. Cells were stained for 15 min at roomtemperature and then washed twice with PBS before flowcytometry analysis. DNA synthesis was measured by 5-ethynyl-2′-deoxyuridine (EdU) incorporation as described (16). Briefly,following fixation and permeabilization, cells were pulsed withEdU for 20 min. Subsequently, EdU staining was performed byusing the Click-iT EdU Alexa Fluor High-throughput ImagingAssay Kit (Invitrogen) according to the manufacturer’s in-structions. The secondary antibody was conjugated with AlexaFluor 647 (Invitrogen). Cell populations were analyzed by flowcytometry (FACSCantoII; BD) equipped with CellQuest soft-ware. Data were analyzed and processed by FlowJo (Version7.6.5).

Cell Proliferation and Xenograft Analysis. For cell proliferationassay, 1 × 105 cells were seeded and cultured at 37 °C in an in-cubator. Cells were counted every 24 h over a 5-d period. Forxenograft assay, 5 × 105 PKM2WT or PKM2T405A/S406A rescuecells were resuspended with 50 μL of Matrigel (Corning) and

Wang et al. www.pnas.org/cgi/content/short/1704145115 2 of 11

100 μL of PBS. Subsequently, the mixtures were injected into theright flank (PKM2WT rescue cells) and left flank (PKM2T405A/S406A

rescue cells) of nude mice (nu/nu; 6- to 8-wk old males). The volumesof tumors were measured every 3 d. At 24 d after the injection, tu-mors were dissected and analyzed. The mice were obtained from theBeijing HFK Bioscience Co., Ltd. [Certification NO. SCXK (Jing)2009-0004]. All animal work procedures were approved by theAnimal Care Committee of the Northeast Normal University(Changchun, China).

Immunohistochemistry Analysis. Mouse tumor tissues were fixedand prepared for immunohistochemistry (IHC) as described (1).

Briefly, paraffin-cut sections of xenograft tumors were prepared,then stained with hematoxylin and eosin and Ki-67. To quantifythe IHC result of positive staining, five random areas of the tissuein each sample were microscopically examined, analyzed, andimaged (Olympus BX51).

Statistical Analysis. P values were calculated from Student’spaired t test when comparing within groups. Student’s unpairedt test was performed when comparing between groups. One-wayANOVA and Bonferroni comparison posttest were performed inthe indicated figures when more than two groups were compared.

1. Yang W, et al. (2012) ERK1/2-dependent phosphorylation and nuclear translocationof PKM2 promotes the Warburg effect. Nat Cell Biol 14:1295–1304.

2. Yang W, et al. (2011) Nuclear PKM2 regulates β-catenin transactivation upon EGFRactivation. Nature 480:118–122.

3. Iyer SP, Hart GW (2003) Roles of the tetratricopeptide repeat domain in O-GlcNActransferase targeting and protein substrate specificity. J Biol Chem 278:24608–24616.

4. Zeidan Q, Wang Z, De Maio A, Hart GW (2010) O-GlcNAc cycling enzymes associate with thetranslational machinery and modify core ribosomal proteins. Mol Biol Cell 21:1922–1936.

5. Yi W, et al. (2012) Phosphofructokinase 1 glycosylation regulates cell growth andmetabolism. Science 337:975–980.

6. Rexach JE, et al. (2010) Quantification of O-glycosylation stoichiometry and dynamicsusing resolvable mass tags. Nat Chem Biol 6:645–651.

7. Rao X, et al. (2015) O-GlcNAcylation of G6PD promotes the pentose phosphatepathway and tumor growth. Nat Commun 6:8468.

8. Clark PM, et al. (2008) Direct in-gel fluorescence detection and cellular imaging of O-GlcNAc-modified proteins. J Am Chem Soc 130:11576–11577.

9. Khidekel N, et al. (2003) A chemoenzymatic approach toward the rapid and sensitive de-tection of O-GlcNAc posttranslational modifications. J Am Chem Soc 125:16162–16163.

10. Qin W, et al. (2017) Quantitative time-resolved chemoproteomics reveals that stableO-GlcNAc regulates box C/D snoRNP biogenesis. Proc Natl Acad Sci USA 114:E6749–E6758.

11. Clark PM, Rexach JE, Hsieh-Wilson LC (2013) Visualization of O-GlcNAc glycosylationstoichiometry and dynamics using resolvable poly(ethylene glycol) mass tags. CurrProtoc Chem Biol 5:281–302.

12. Ortiz-Meoz RF, et al. (2015) A small molecule that inhibits OGT activity in cells. ACSChem Biol 10:1392–1397.

13. Park SK, et al. (2017) A conserved splicing silencer dynamically regulates O-GlcNAc transferase intron retention and O-GlcNAc homeostasis. Cell Rep 20:1088–1099.

14. Fujiki R, et al. (2011) GlcNAcylation of histone H2B facilitates its monoubiquitination.Nature 480:557–560.

15. Slawson C, et al. (2005) Perturbations in O-linked beta-N-acetylglucosamine proteinmodification cause severe defects in mitotic progression and cytokinesis. J Biol Chem280:32944–32956.

16. Groth A, et al. (2005) Human Asf1 regulates the flow of S phase histones duringreplicational stress. Mol Cell 17:301–311.

Wang et al. www.pnas.org/cgi/content/short/1704145115 3 of 11

Fig. S1. Detection of PKM2 O-GlcNAcylation by traditional immunoprecipitation. (A and B) Whole lysates from indicated cells and clinical human breast tumortissues were immunoprecipitated with PKM2 antibody, and then subjected to Western blotting (WB) analysis with antibodies against PKM2 and O-GlcNA-cylation (CTD110.6) respectively. Relative PKM2 O-GlcNAcylation level was normalized to the total PKM2 protein level for each cell type. A.U., arbitrary units.Error bars, SEM; n = 3 biological replicas. One-way analysis of variance (ANOVA) and Bonferroni comparison posttest: *P < 0.05; **P < 0.01; ns, nonsignificant.(C) PKM2 expression levels in 29 pairs of clinical human breast tumor tissues (T) and the adjacent normal tissues (N) examined by WB with indicated antibodies.

Wang et al. www.pnas.org/cgi/content/short/1704145115 4 of 11

Fig. S2. Evaluation of PKM2 O-GlcNAcylation stoichiometry and sites. (A and B) Stoichiometric analysis of O-GlcNAcylated PKM2 in MCF-7 cells. (A) O-GlcNAcylated proteins in whole lysates were biotinylated and precipitated by using the Click-iT O-GlcNAc enzymatic labeling system as described in SI Materialsand Methods. All samples were subjected to WB analysis for PKM2. To quantitatively evaluate O-GlcNAcylated PKM2, a series of lysate dilutions from inputindicated as 100%, 50%, 25%, 12.5%, 6.25%, 3.125%, and 1.5625% (A, Upper) or 14%, 12%, 10%, 8%, and 6% (A, Lower) were prepared and loaded.According to the loading controls, stoichiometry of O-GlcNAcylated PKM2 is ∼8–10%. (B) O-GlcNAcylated proteins were chemoenzymatically labeled withAlkyne-5K-PEG resolvable mass tag. Both labeled and unlabeled lysates were subjected to WB analysis for PKM2, and the labeled (O-GlcNAcylated) PKM2 willshow a 5KD shifted band. Based on two band densities, we calculated the O-GlcNAcylation level of PKM2, and it was 7.3%. We also introduced a similar set ofdilution controls in this experiment, including 100%, 50%, 25%, 12.5%, 6.25%, 3.125%, and 1.5625% of lysate dilutions from input; the comparison gave asimilar range of basal level to be 6.25–12.5%. We also examined the elevated level of PKM2 O-GlcNAcylation upon PUGNAc treatment, which was 13.4%.(C) Validation of O-GlcNAcylation sites on PKM2. The effect of mutation of a series of surface-located Ser and Thr residues on the level of PKM2 O-GlcNAcylation.Wild-type and single point-mutated (to Ala) PKM2 tagged with Flag were transfected into MCF-7 cells and immunoprecipitated for WB with indicated antibodies.

Wang et al. www.pnas.org/cgi/content/short/1704145115 5 of 11

Fig. S3. Impact of O-GlcNAcylation on endogenous PKM2. (A–C) Effect of O-GlcNAcylation on endogenous PKM2 tetramerization and enzymatic activity.Following PUGNAc treatment, endogenous PKM2 proteins in MCF-7 cells were immunoprecipitated and subjected to WB analysis as in A, oligomer assay as in B,and enzymatic activity analysis as in C. GA, glutaraldehyde. Error bars, SEM; n = 3 independent assays. Student’s t test: *P < 0.05. (D) Fractionation assay ofPKM2. Nuclear and cytoplasmic fractions as well as the whole lysates from indicated cells were subjected to WB analysis for PKM2. Lamin B1 and GAPDH wereincluded to indicate nuclear and cytosolic fraction respectively. (E) Immunostaining of PKM2. (E, Upper) In the presence or absence of PUGNAc, MCF-7 cellswere stained for PKM2. The white boxes in nuclei highlight the magnified Insets. (Scale bar, 20 μm.) (E, Lower) Percentage of MCF-7 cells positive for PKM2 innucleus was quantified. Error bars, SEM; n = 30 (10 fields from each of the three independent experiments). Student’s t test: ***P < 0.001. (F) Effect of O-GlcNAcylation on Ser37 phosphorylation of PKM2. Following PUGNAc treatment, endogenous PKM2 proteins in MCF-7 cells were immunoprecipitated for WBanalysis with indicated antibodies.

Fig. S4. O-GlcNAcylation of PKM2 at Thr405 and Ser406 in response to nutritional changes. MCF-7 cells transfected with Flag-tagged PKM2WT or PKM2T405A/S406A

were cultured in medium supplemented with glucose (Left) or glutamine (Right) for 24 h. Subsequently, immunoprecipitated proteins by Flag-PKM2were analyzedby WB.

Wang et al. www.pnas.org/cgi/content/short/1704145115 6 of 11

Fig. S5. OGT overexpression influences PKM2 tetramerization, enzymatic activity, and nuclear translocation. (A) Oligomer analysis of PKM2. Upon trans-fection of vector or HA-tagged OGT for 24 h, the whole-cell lysates from MCF-7 cells stably depleted for endogenous PKM2 coexpressing Flag-tagged PKM2WT

or PKM2T405A/S406A were collected for WB in the absence or presence of GA. (B) Enzymatic activity analysis of PKM2. Cell lysates were prepared as described inA, and Flag-PKM2 proteins were immunoprecipitated for activity assay. Error bars, SEM; n = 3 independent assays. Student’s t test: *P < 0.05, ns, nonsignificant.(C) Fractionation assay. Cell lysates were prepared as described in A and subjected to fractionation analysis for PKM2. Lamin B1 and β-Tubulin were included toindicate nuclear and cytosolic fraction respectively.

Wang et al. www.pnas.org/cgi/content/short/1704145115 7 of 11

Fig. S6. Oligomer and activity analysis of PKM2WT and PKM2T405A/S406A mutant in the absence of O-GlcNAcylation. (A and B) Inhibition of proteinO-GlcNAcylation in the presence of OSMI-1 or DON, which inhibits the generation of OGT substrate or OGT activity, respectively. Transfection of Flag-taggedPKM2WT or PKM2T405A/S406A along with or without concomitant OSMI-1 or DON treatment was conducted in MCF-7 cells stably depleted for PKM2. The whole-cell lysates were collected for WB as in A and immunoprecipitation (IP) as in B. (C and D) Both mutants and wild type specifically differ only under conditionselevating O-GlcNAcylation. Notably, the T405A/S406A mutation had no effect on PKM2 oligomerization in C or PKM2 activity in D under a reducedO-GlcNAcylation condition. This suggests that the observed difference in tetramer status and PK activity between PKM2WT and PKM2T405A/S406A induced byPUGNAc or OGT overexpression is due to the loss of the O-GlcNAcylation of these two residues (T405/S406), rather than their mutations. Error bars, SEM; n =3 independent assays. Student’s t test: ns, nonsignificant.

Fig. S7. Cellular distribution of PKM2 proteins. Flag-tagged PKM2WT or PKM2T405A/S406A was transfected in PKM2-depleted MCF-7 cells, along with or withoutconcomitant PUGNAc. Nuclear and cytoplasmic fractions as well as the whole lysates from indicated cells were subjected to WB analysis. Lamin B1 and β-Tubulinwere included to indicate nuclear and cytosolic fraction, respectively.

Wang et al. www.pnas.org/cgi/content/short/1704145115 8 of 11

Fig. S8. Detetramerization exposes Arg399 and Arg400 residues of the PKM2 NLS sequence. (A, Left) Structure of human PKM2 tetramer (PDB ID code 1T5A)with Arg399 and Arg400 structurally buried at the C-C tetramerization interface. (A, Right) Magnified view of the C-C tetramerization interface highlightingArg399 (purple) and Arg400 (yellow). (B, Left) Structure of PKM2 dimer (PDB ID code 1T5A) with Arg399 and Arg400 exposed. (B, Right) Magnified view high-lighting Arg399 (purple) and Arg400 (yellow) in the PKM2 dimer. (C) Following the transfection of Flag-tagged PKM2WT or PKM2T405A/S406A, PKM2-depletedMCF-7 cells were treated with or without EGF (100 ng/μL) for 30 min followed by IP and WB analysis.

Wang et al. www.pnas.org/cgi/content/short/1704145115 9 of 11

ns

PUGNAc

0.8

0.6

0.4

0.2

0

Lact

ate

Pro

duct

ion

(nm

ol/m

illio

n ce

lls/m

in)

WTT405A/S406A

Flag- + - +

***

PUGNAc

T405A/S406AWT

- + - +

c-Myc

Flag

F

A

PUGNAc

2.5

2.0

1.0

0.5

0

Glut1

mR

NA

Fo

ld E

nric

hmen

t

ns

**

WTR399A/R400A

- + - +

1.5

E

D

-PUGNAc

1.5

1.0

0.5

0Fold

Cha

nge

(MFI

) of

Lipi

d S

ynth

esis

ns

*

WTT405A/S406A

Flag+ - +

ns*

Fold

Cha

nge

(MFI

) of

DN

A S

ynth

esis

1.5

0.5

0

WTT405A/S406A

PUGNAc

1.0

Flag- + - +

ns

PUGNAc

2.5

1.5

1.0

0.5

0

ChI

P: F

lag

qPC

R: c

-Myc

***

WTT405A/S406A

- + - +

2.0

PUGNAc - + - +T405A/S406AWT

Glu

cose

Con

sum

ptio

n(n

mol

/mill

ion

cells

/h) 60

45

30

15

0

*

ns

Flag

2.5

2.0

1.0

0.5

0

LDHA

mR

NA

Fo

ld E

nric

hmen

t

ns

**

WTR399A/R400A

PUGNAc - + - +

1.5

T405A/S406A

- + - +

***

2.5

2.0

1.0

0.5

0

c-Myc

mR

NA

Fo

ld E

nric

hmen

t

ns

WT

PUGNAc

1.5

C

B

G

Fig. S9. PKM2 O-GlcNAcylation enhances Warburg effect-characteristic metabolic reprogramming and gene expression. (A) Assay of lipid and DNA synthesis.Transfection of Flag-tagged PKM2WT, PKM2T405A/S406A, or PKM2R399A/R400A accompanied with or without PUGNAc treatment was conducted in PKM2-depletedMCF-7 cells for 24 h. Lipid was detected by BODIPY (A, Left) and replicating DNA was pulse-labeled by EdU (A, Right). Following immunofluorescence, cellswere analyzed by flow cytometry. In each group, relative mean fluorescence intensity (MFI) of PUGNAc-treated cells was normalized to that from untreatedcells. Error bars, SEM. Student’s t test: *P < 0.05; ns, nonsignificant. (B and C) mRNA level of Glut1 or LDHA evaluated by qPCR in MCF-7 cells as treated in A.(D) Glucose consumption (Left) and lactate production (Right) in MCF-7 cells as treated in A. (E and F) Regulation of c-Myc expression by PKM2 O-GlcNAcylation. Cellswere treated as described in A, the protein level of c-Myc expression were analyzed by WB as in E and the mRNA level of c-Myc was evaluated by qPCR as in F.(G) PKM2 association with c-Myc promoter region. In cells treated as described inA, ChIP assay was performed by using antibody against Flag, followed by qPCR analysisof c-Myc promoter region. Error bars, SEM; n = 3 biological replicas. Student’s t test: *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant.

Wang et al. www.pnas.org/cgi/content/short/1704145115 10 of 11

Movie S1. The movie, derived from molecular dynamics simulations, depicts the movement of modified O-GlcNAc (blue) units along the C-C interface, whichpushes away the other monomer.

Movie S1

Wang et al. www.pnas.org/cgi/content/short/1704145115 11 of 11