O-GlcNAcylation of enolase 1 serves as a dual regulator of aerobic glycolysis and immune evasion in colorectal cancer

Significance

Colorectal cancer cells display key features such as metabolizing glucose at high rates, and escaping immune surveillance. However, how these two features are linked at molecular levels is largely unknown. Here, we show that an important glycolytic enzyme (enolase 1, ENO1) is dynamically modified by N-acetylglucosamine (O-GlcNAc) at multiple residues in colorectal cancer cells. The modification on threonine 19 of ENO1 increases glucose metabolism, while the modification on serine 249 promotes immune escape. The elimination of modifications on both sites synergistically inhibits colorectal cancer growth. Thus, our study reveals that O-GlcNAc modifications of ENO1 could serve as a dual regulator for glucose metabolism and immune evasion to promote colorectal cancer growth.

Abstract

Aerobic glycolysis and immune evasion are two key hallmarks of cancer. However, how these two features are mechanistically linked to promote tumor growth is not well understood. Here, we show that the glycolytic enzyme enolase-1 (ENO1) is dynamically modified with an O-linked β-N-acetylglucosamine (O-GlcNAcylation), and simultaneously regulates aerobic glycolysis and immune evasion via differential glycosylation. Glycosylation of threonine 19 (T19) on ENO1 promotes its glycolytic activity via the formation of active dimers. On the other hand, glycosylation of serine 249 (S249) on ENO1 inhibits its interaction with PD-L1, decreases association of PD-L1 with the E3 ligase STUB1, resulting in stabilization of PD-L1. Consequently, blockade of T19 glycosylation on ENO1 inhibits glycolysis, and decreases cell proliferation and tumor growth. Blockade of S249 glycosylation on ENO1 reduces PD-L1 expression and enhances T cell–mediated immunity against tumor cells. Notably, elimination of glycosylation at both sites synergizes with PD-L1 monoclonal antibody therapy to promote antitumor immune response. Clinically, ENO1 glycosylation levels are up-regulated and show a positive correlation with PD-L1 levels in human colorectal cancers. Thus, our findings provide a mechanistic understanding of how O-GlcNAcylation bridges aerobic glycolysis and immune evasion to promote tumor growth, suggesting effective therapeutic opportunities.

O-GlcNAcylation is a dynamic monosaccharide modification on serine or threonine hydroxyl moieties of intracellular proteins (1). The cycle of O-GlcNAcylation on protein substrates is mediated by a set of enzymes. O-GlcNAc transferase (OGT) is responsible for the addition of O-GlcNAc while O-GlcNAcase (OGA) catalyzes the removal of the GlcNAc moiety (2, 3). It is reported that thousands of proteins in cells are O-GlcNAcylated, which plays important roles in transcription, translation, cell cycle, and metabolism (4–6). The disorder of O-GlcNAcylation in cells is closely related to human diseases such as diabetes, cardiovascular diseases, and cancers (7–9). Mounting studies indicated that global O-GlcNAcylation is elevated in various types of cancer, and its overexpression links to the poor survival of patients (10–12). However, the molecular mechanisms of O-GlcNAcylation involved in regulating tumor growth and immune evasion remain elusive.

The ability for immune evasion is a common feature of cancer cells (13). The programmed death ligand 1 (PD-L1), a key immune checkpoint protein on cancer cell surfaces, plays a crucial role in cancer immune evasion (14, 15). The binding of PD-L1 with its receptor PD-1 on cytotoxic T lymphocytes induces T cell exhaustion and weakens antitumor activity (16, 17). In recent years, immune checkpoint inhibitors (ICIs) have underscored the profound therapeutic efficacy by disrupting the interaction between PD-L1 and PD-1 for the treatment of advanced melanoma, non–small cell lung cancer, renal cell carcinoma, and head and neck cancer (18, 19). However, the majority of patients do not achieve durable remission with the therapy, and some solid tumors are completely refractory to the treatment (20). Accumulated evidence demonstrated that the level of PD-L1 expression in cancer cells is a critical factor for assessing the efficacy of anti-PD-L1/PD-1 therapy (16, 21). Thus, a comprehensive understanding of PD-L1 expression regulation is essential for improving therapy efficacy and overcoming therapeutic resistance in cancer patients.

ENO1, which catalyzes the conversion of 2-phosphoglycerate (2-PG) to phosphoenolpyruvate (PEP) in the glycolysis pathway, plays a vital role in Warburg effect-dependent cancers (22). Apart from functioning as a glycolytic enzyme, ENO1 also has several moonlighting roles in multiple cellular processes (23). For example, ENO1 contributes to the gene transcription and protein translation by binding to DNA and mRNA in cancer cells (24). ENO1 promotes mRNA degradation of IRP1 to repress ferroptosis in hepatocellular carcinoma (HCC) cells (25). Additionally, ENO1 facilitated T cell-mediated antitumor immunity by promoting the proteasomal degradation of PD-L1 in lung cancer cells (26). A recent study showed that ENO1 promotes choline phospholipid metabolism and glioblastoma cell proliferation through the inhibition of CHKα degradation (27). As ENO1 expression is up-regulated in a variety of human cancers, ENO1 is considered a promising drug target for cancer treatment (28). Although ENO1 is tightly regulated by a myriad of posttranslational modifications such as phosphorylation (29), methylation (30), acetylation (31), and 2-hydroxyisobutyrylation (32), how O-GlcNAcylation modulates the metabolic and moonlighting functions of ENO1 to coordinate tumor growth and immune evasion is unknown.

In this report, we demonstrate a mechanism by which O-GlcNAcylation regulates colorectal cancer (CRC) growth and immune evasion. O-GlcNAcylation of ENO1 on T19 increases its enzyme activity by facilitating the formation of active dimers. On the other hand, glycosylation of S249 inhibits ENO1 interaction with PD-L1, leading to the decreased ubiquitin E3 ligase STUB1-dependent PD-L1 ubiquitination and degradation. Blockade of O-GlcNAcylation on ENO1 repressed glycolysis and promoted T cell antitumor activity, thereby inhibiting CRC growth and immune evasion in mouse models.

Results

ENO1 Promotes CRC Growth Through Its Enzymatic Activity.

CRC cells are highly dependent on aerobic glycolysis to generate metabolic intermediates to fuel macromolecule biosynthesis to sustain cell proliferation (33, 34). To explore the clinical relevance of ENO1 in CRC development, we analyzed the expression levels of ENO1 in 30 pairs of human CRC tissue samples and the corresponding peritumoral tissue samples by immunohistochemistry (IHC). The results showed that ENO1 expressions were significantly up-regulated in tumor tissues as compared to the peritumoral tissues, which was consistent with analyses derived from The Cancer Genome Atlas (TCGA) gene database of colon adenocarcinoma (COAD) (Fig. 1A and SI Appendix, Fig. S1A).

Fig. 1.

ENO1 is highly expressed in CRC and promotes cell proliferation. (A) Immunohistochemical analysis of ENO1 expression in primary CRC tissues and matched peritumoral tissues (n = 30 pairs). Representative staining images are shown. (Scale bar, 200 μm.) Comparative analysis of ENO1 expression between CRC tissues and their adjacent normal tissues was performed. (B) Immunoblotting of ENO1 expression in the indicated cells. (C) Cell proliferation of NCM460, HCT116, and RKO cells stably expressing scramble shRNA, ENO1-targeting shRNA, or reconstituted with WT or S40A ENO1 (n = 5). (D) ECAR in HCT116 and RKO cells stably expressing scramble shRNA, ENO1-targeting shRNA, or reconstituted with WT or S40A ENO1 (n = 3). (E) Image of tumors derived from the indicated cells in nude mice. The mice were killed and examined for tumor growth 20 d after injection. (F) Tumor volumes measured at indicated time points. (G) Tumor mass measured at the experimental endpoint (n = 5 mice). (H) Immunohistochemical analyses of tumor samples using an anti-Ki67 antibody. Ki67-positive cells were quantified using Image J (n = 5). Error bars in A, C, D, F, G, and H denote the mean ± SD. Statistical analyses in A and D were performed by unpaired two-tailed Student’s t test. Statistical analyses in C, F, G, and H were performed by one-way ANOVA with Tukey's multiple comparisons test.

A range of human CRC cell lines (LoVo, HCT-116, HT-29, RKO, and SW620) consistently showed higher ENO1 expressions compared with normal human colon epithelial cells (NCM460) (Fig. 1B). Depletion of ENO1 using small hairpin RNA (shRNA) repressed cell proliferation in both HCT-116 and RKO cells to a greater degree than in NCM460 cells. The inhibitory effect was completely reversed by reexpressing shRNA-resistant wildtype (WT) ENO1, but not the catalytically dead mutant (S40A) (35) (Fig. 1C and SI Appendix, Fig. S1B), suggesting that ENO1 enzymatic activity plays a critical role in cell proliferation. To confirm the role of ENO1 in cellular glycolysis, we analyzed the extracellular acidification rate (ECAR) in cells. ENO1 depletion significantly reduced ECAR compared with the control. Reconstitution of WT, but not S40A ENO1, restored ECAR (Fig. 1D). To further validate the importance of ENO1 enzyme activity in tumor development, we subcutaneously injected HCT-116 cells expressing WT or S40A ENO1 under the endogenous ENO1 depletion background into immune-compromised nude mice. Depletion of ENO1 expectedly inhibited tumor growth and Ki67 expression. These inhibitory effects were abrogated by reconstituted expression of WT, but not S40A ENO1 (Fig. 1 E–H and SI Appendix, Fig. S1C). Together, these results suggest that the enzymatic activity of ENO1 is required for CRC cell proliferation and tumor growth.

O-GlcNAcylation of ENO1 on T19 Enhances Its Enzymatic Activity.

Emerging evidence has shown that O-GlcNAcylation impacts activities of several key metabolic enzymes to regulate cellular metabolism (36–38). To determine whether ENO1 is modified by O-GlcNAcylation, we performed a well-established chemoenzymatic labeling experiment (39). The result showed that O-GlcNAcylation signals could be markedly elevated in the presence of OGA inhibitor thiamet-G (TMG) or overexpression of OGT (Fig. 2A). We further analyzed the O-GlcNAcylation stoichiometry of ENO1 in HCT116 cells by using the mass tagging approach (40, 41). In this assay, the PEG molecule of 5 kDa was conjugated to azido-tagged glycoproteins, which led to a molecular shift of the immunoblotting signal. By quantifying the ratio of intensity between the shifted and unshifted bands, we estimated the modification ratio of endogenous ENO1 to be about 27% under normal culture conditions (SI Appendix, Fig. S2A). The level of ENO1 O-GlcNAcylation increased in the presence of different concentrations of glucose, glutamine, or serum in the medium in a dose-dependent manner (SI Appendix, Fig. S2 B–D).

Fig. 2.

O-GlcNAcylation of ENO1 on T19 enhances its enzyme activity. (A) Analysis of O-GlcNAcylation levels of ENO1 in the presence of TMG (100 μM) or OGT overexpression in HCT116 cells. (B) Analysis and quantification of O-GlcNAcylation levels on ENO1 using various site-directed mutants expressed in HCT116 cells (n = 3 independent assays). (C) Comparison of enzyme activity of WT, T19A, and S249A ENO1 in the presence of TMG (100 μM) or OGT overexpression in endogenous ENO1-depleted HCT116 cells. (n = 5 independent assays). (D) Steady-state kinetics of purified WT, T19A, and S249A ENO1 upon OGT overexpression in E. coli (n = 3 independent assays). (E) Immunoblotting analyses of interaction between Flag-tagged ENO1 and HA-tagged ENO1 expressed in endogenous ENO1-depleted HCT116 cells in the presence or absence of TMG treatment. (F) Analysis of the oligomerization state of WT or T19A ENO1 in the presence or absence of TMG treatment using the crosslinking reagent 3-sulfo-N-hydroxysuccinimideester in endogenous ENO1-depleted HCT116 cells. Error bars in B, C, and D denote the mean ± SD. Statistical analyses in B were performed by one-way ANOVA with Dunnett's multiple comparisons test. Statistical analyses in c were performed by two-way ANOVA with Tukey's multiple comparisons test.

To map the O-GlcNAcylation site(s) on ENO1, we ectopically expressed Flag-tagged ENO1 in HCT116 cells. After immunoprecipitated with Flag-M2 magnetic beads and on-bead proteolytic digestion with trypsin, peptides were subjected to liquid chromatography coupled mass spectrometry (LC-MS) analysis. We identified four putative O-GlcNAcylation sites (T19, T26, S27, and S249) on ENO1 (SI Appendix, Fig. S2 E and F). To further validate the modification site(s), we generated singlet mutants by mutating these residues to alanine individually. The chemoenzymatic labeling assay showed that both T19A and S249A mutants significantly reduced the O-GlcNAcylation signal as compared to the control. Double site mutations (T19A/S249A) reduced the glycosylation signal by about 80%, suggesting that T19 and S249 are major glycosylation sites on ENO1 (Fig. 2B).

Sequence comparisons across different species showed that both T19 and S249 are evolutionarily conserved, suggesting that these residues may have important functions in ENO1 (SI Appendix, Fig. S2G). To detect the effects of T19 and S249 O-GlcNAcylation on ENO1 enzymatic activity, we generated stable HCT116 cells in which the endogenous ENO1 was depleted, and simultaneously reconstituted with Flag-tagged shRNA-resistant WT, T19A, or S249A ENO1(SI Appendix, Fig. S2H). Flag-ENO1 proteins were isolated from cell lysates for the subsequent enzyme activity assay. The T19A mutant, but not the S249A mutant, displayed a significant reduction of activity compared to WT ENO1 (Fig. 2C). Treatment with TMG or overexpressing OGT in cells similarly increased the activity of both WT and S249A ENO1 but had no apparent effect on the T19A mutant (Fig. 2C). To rule out the possibility that the altered activity resulted from a change of protein structure conferred by the mutation, we expressed and purified His-tagged WT, T19A, and S249A ENO1 from Escherichia coli host, which presumably lacks the endogenous O-GlcNAcylation machinery. Notably, WT, T19A, and S249A ENO1 all displayed similar enzymatic activities (SI Appendix, Fig. S2I). We further coexpressed OGT and critical enzymes in the UDP-GlcNAc synthesis pathway to generate O-GlcNAcylated ENO1 in E. coli (42) (SI Appendix, Fig. S2J). Interestingly, OGT overexpression increased the activity of WT and S249A ENO1, but not T19A ENO1 (SI Appendix, Fig. S2I). These results support the notion that T19 O-GlcNAcylation promotes ENO1 activity.

It is reported that O-GlcNAcylation engages in intricate crosstalks with protein phosphorylation (43). We then asked whether ENO1 O-GlcNAcylation had an interplay with phosphorylation. OGT coexpression or TMG treatment caused a slight reduction in phospho-threonine signals but not phospho-serine signals as detected with panantibodies, indicating that O-GlcNAcylation may affect ENO1 phosphorylation on threonine (SI Appendix, Fig. S3 A–D). It was reported that T19 at ENO1 was phosphorylated (44). Consistently, T19A but not S249A mutant reduced the phospho-threonine signal as compared to WT control (SI Appendix, Fig. S3 E–G). To further test the effect of T19 phosphorylation on ENO1 activity, we generated phosphorylation-mimicking mutant (T19D) by mutating threonine to aspartate on T19. Both T19D and T19A mutants showed similar enzymatic activities, indicating that phosphorylation on T19 has no impact on ENO1 activity (SI Appendix, Fig. S3H). Thus, these results reveal that the effect of O-GlcNAcylation on enzymatic activity was independent of phosphorylation on T19.

To gain a better understanding of the regulation of ENO1 activity by T19 glycosylation, we performed the enzyme kinetics study to determine the rate of enzyme-catalyzed reaction by varying substrate concentrations. Both WT and S249A ENO1 had a much higher maximum velocity (V_max) for 2-P-glycerate production than T19A ENO1 in the presence of OGT overexpression (Fig. 2D). The Michaelis–Menten constant (K_m) values of WT (93.52 μM) and S249A (117.80 μM) ENO1 for 2-P-glycerate were lower than that of T19A ENO1 (196.0 μM) upon OGT overexpression. Consistently, the catalytic efficiencies of WT (K_cat/K_m = 1.03 × 10⁶ s⁻¹ M⁻¹) and S249A ENO1 (K_cat/K_m = 1.12 × 10⁶ s⁻¹ M⁻¹) for 2-P-glycerate were about five-fold higher than that of T19A ENO1 (K_cat/K_m = 2.3 × 10⁵ s⁻¹ M⁻¹) (SI Appendix, Table S1). A previous study showed that the dimeric form of ENO1 has a higher enzymatic activity than the monomeric form (30). Thus, we subsequently examined whether T19 glycosylation impacts the oligomerization state of ENO1. We analyzed the interaction between HA-tagged and Flag-tagged ENO1 in HCT116 cells by immunoprecipitation assays, which showed that T19A ENO1 disrupted the binding between differently taggedENO1. Treatment with TMG enhanced the interaction between differently tagged WT, but not T19A ENO1 (Fig. 2E). We also employed 3-sulfo-N-hydroxysuccinimideester (BS3), a water-soluble cross-linker that reacts with primary amines (–NH₂) on proteins (45), to stabilize the oligomerization state of ENO1. O-GlcNAcylation increased the amount of dimeric WT ENO1 protein, but had no apparent effect on the T19A mutant (Fig. 2F). Collectively, these data demonstrate that T19 O-GlcNAcylation promotes ENO1 activity by facilitating the formation of ENO1 active dimers.

T19 O-GlcNAcylation of ENO1 Promotes Glycolysis and Colorectal Tumor Growth.

To investigate the effect of ENO1 O-GlcNAcylation on cellular glycolysis, we analyzed glucose uptake and lactate and ATP production in WT or T19A reconstituted HCT116 cells. OGT overexpression significantly increased glucose uptake, lactate and ATP production in WT ENO1 reconstituted cells, but only slightly augmented the effects in T19A ENO1 reconstituted cells (Fig. 3 A–C). Consistently, upon OGT overexpression, the ECAR was significantly elevated in WT ENO1 reconstituted cells compared to T19A reconstituted cells (Fig. 3D). In addition, cellular metabolite analysis by LC-MS showed that T19A ENO1 reconstituted cells accumulated higher levels of glucose-6-phosphate, fructose-6-phosphate, fructose-1,6-diphosphate, dihydroxyacetone phosphate, and 3-phosphoglyceric acid, metabolites at the early phase of the glycolysis pathway, but lower levels of phosphoenolpyruvate and pyruvate, compared to WT ENO1 reconstituted cells upon OGT overexpression, suggesting that OGT-mediated enhancement of ENO1 activity promotes aerobic glycolysis in CRC cells (Fig. 3E).

Fig. 3.

T19 O-GlcNAcylation of ENO1 promotes glycolysis and colorectal tumor growth. (A–C) Analysis of glucose uptake (A), lactate production (B), and ATP levels (C) in ENO1 WT or T19A rescue HCT116 cells in the presence or absence of OGT overexpression. (D) ECAR in ENO1 WT or T19A rescue HCT116 cells in the presence of OGT overexpression. (E) Relative abundance of metabolites derived from glycolysis in ENO1 WT or T19A rescue HCT116 cells in the presence of OGT overexpression. (F) Analysis of ENO1 glycosylation in indicated human colon cancer cell lines and a normal colon cell line NCM460. (G and H) Cell proliferation (G) and colony formation (H) of HCT116 cells stably expressing scramble shRNA, ENO1-targeting shRNA, or reconstituted with WT or T19A ENO1 in the presence or absence of OGT overexpression (n = 5). (I) Image of tumors derived from the indicated cells in nude mice. The mice were euthanized and examined for tumor growth 25 d after injection. (J) Tumor volumes measured at indicated time points. (K) Tumor mass measured at the experimental endpoint (n = 5 mice). (L) Immunohistochemical analyses of tumor samples using an anti-Ki67 antibody. (M) Ki67-positive cells were quantified using Image J (n = 5). Error bars in A, B, C, D, E, G, H, J, K, and M denote the mean ± SD. Statistical analyses in A, B, C, D, and E were performed by unpaired two-tailed Student’s t test. Statistical analyses in G, H, J, K, and M were performed by one-way ANOVA with Tukey's multiple comparisons test.

To gain a better insight of ENO1 glycosylation on CRC growth, we detected the level of ENO1 O-GlcNAcylation with CRC cell lines and NCM460 cells. ENO1 O-GlcNAcylation was significantly up-regulated in CRC cells compared to that in NCM460 cells (Fig. 3F). WT ENO1 reconstituted HCT116 cells displayed increased cell proliferation and colony formation as compared to T19A ENO1 reconstituted cells (Fig. 3 G–H). Reconstituted expression of WT ENO1 also fully rescued these effects resulting from depletion of endogenous ENO1. OGT overexpression significantly increased cell proliferation and colony formation in WT ENO1 reconstituted cells, but not in T19A ENO1 reconstituted cells (Fig. 3 G–H). We further subcutaneously injected these cells into nude mice and analyzed their ability to form tumors. Reconstituted expression of WT ENO1 reversed tumor growth retardation and reduced Ki67 expression caused by the endogenous ENO1 depletion. T19A ENO1 reconstituted cells showed a much weaker rescued effect. OGT overexpression further accelerated tumor growth with a corresponding increase in Ki67 expression in cells reconstituted with WT ENO1 but not T19A ENO1 (Fig. 3 I–M). Thus, these data demonstrate that T19 O-GlcNAcylation of ENO1 promotes colorectal tumor growth likely via upregulating aerobic glycolysis.

S249 O-GlcNAcylation of ENO1 Promotes PD-L1 Stability by Inhibiting the Binding of STUB1 and PD-L1.

It was reported that in addition to its canonical metabolic activity, ENO1 can enhance the association of PD-L1 with E3 ligase STUB1 by binding to PD-L1, resulting in increased PD-L1 degradation in cells (26). We analyzed the interaction between ENO1 and PD-L1 based on AlphaFold2 predictions and identified four key residues of ENO1 (N52, K54, K197, and E250) located at the interacting interface with PD-L1 (Fig. 4A and SI Appendix, Fig. S4 A and B). We then generated alanine substitutions for each of these residues and performed immunoprecipitation assays to evaluate their individual roles. Among these substitutions, the mutation of E250 to alanine (E250A) notably decreased the interaction between ENO1 and PD-L1 (SI Appendix, Fig. S4C). Further analysis found that E250 of ENO1 enabled the formation of critical salt bridges with R125 on PD-L1 to enhance the interaction stability (Fig. 4A). These results underscore the essential role of the E250 of ENO1 in mediating the interaction with PD-L1. We also performed pull-down experiments using GST-tagged full-length ENO1 and its various truncated mutants (N-terminal region, DNA-binding domain, and C-terminal region) with His-tagged PD-L1 obtained from the E. coli expression host. Notably, only the full-length ENO1 and the C-terminal region of ENO1 (containing the S249 glycosylation site) showed binding signals with PD-L1 (SI Appendix, Fig. S4D). Based on these data, we propose that a bulky GlcNAc modification on S249 adjacent to the key E250 residue would likely disrupt the interaction with PD-L1.

Fig. 4.

S249 O-GlcNAcylation of ENO1 represses PD-L1 degradation. (A) AlphaFold2-predicted models of ENO1 dimers (depicted in green and cyan) in complex with two PD-L1 molecules, each consisting of N and C domains (shown in red and magenta, respectively). The key interface residues (N52, K54, K197, and E250, illustrated as blue sticks) potentially critical for ENO1-PD-L1 interactions. The residue E250 of ENO1 is predicted to form salt bridges with R125 of PD-L1. (B) Immunoblotting of ENO1 and PD-L1 interaction in WT, T19, or S249 ENO1 reconstituted RKO cells in the presence or absence of OGT overexpression. (C) Quantification of the relative PD-L1 protein expression in WT, T19, or S249 ENO1 reconstituted RKO cells in the presence or absence of OGT overexpression. (D and E) Immunoblotting (D) and quantification (E) of PD-L1 levels in WT or S249 ENO1 reconstituted RKO cells by CHX treatment in the presence of inhibitors for lysosome (chloroquine), autophagy (3-MA), or proteasome (MG132). (F) Immunoblotting of STUB1 and PD-L1 interaction and PD-L1 ubiquitination in RKO or PD-L1 overexpressing HCT116 cells, with or without TMG treatment. (G) Immunoblotting of PD-L1 expression in WT or S249 ENO1 reconstituted RKO cells or PD-L1 overexpressing HCT116 cells infected with scramble shRNA or shRNA targeting STUB1 in the presence or absence of OGT overexpression. Statistical analyses were performed by one-way ANOVA with Tukey's multiple comparisons test.

To further analyze the effect of S249 glycosylation on ENO1 and PD-L1 interaction, we ectopically expressed Flag-tagged WT or S249A ENO1 and HA-tagged PD-L1 in 293 T cells in the presence or absence of OGT overexpression. Upon immunoprecipitation, a markedly reduced amount of PD-L1 was pulled down in cells expressing WT ENO1 upon OGT overexpression compared with cells without OGT overexpression. In cells expressing S249A ENO1, the amount of pulled-down PD-L1 had no apparent change in the presence or absence of OGT overexpression (SI Appendix, Fig. S5A). Immunoprecipitation experiments performed in the reverse order also showed consistent results (SI Appendix, Fig. S5B). In addition, immunofluorescence staining revealed that OGT overexpression reduced the colocalization of PD-L1 with WT ENO1, but not S249A ENO1 (SI Appendix, Fig. S5 C and D). We also explored the effect of ENO1 O-GlcNAcylation on the interaction with endogenous PD-L1 in WT, T19A, or S249A ENO1 reconstituted RKO cells. Coimmunoprecipitation showed that increasing O-GlcNAcylation by OGT overexpression reduced the interaction between ENO1 and PD-L1 in cells expressing WT or T19A ENO1 that was accompanied by increased PD-L1 protein levels, whereas no effect was observed in cells expressing S249A ENO1 (Fig. 4 B and C). Consistently, TMG treatment also increased PD-L1 protein levels in WT ENO1 reconstituted cells but not in S249A reconstituted cells (SI Appendix, Fig. S5 E–H). The effect of ENO1 S249 O-GlcNAcylation on PD-L1 expression was further confirmed by immunofluorescence staining and flow cytometry assays (SI Appendix, Fig. S5 I–L). OGT overexpression in reconstituted RKO cells did not alter the PD-L1 mRNA levels, suggesting a transcription-independent PD-L1 regulation mechanism (SI Appendix, Fig. S5M).

We next analyzed the half-life of PD-L1 and observed that the turnover rate of PD-L1 was much faster in S249A ENO1 reconstituted cells as compared to that in WT ENO1 reconstituted cells. Treatment with TMG promoted the stability of PD-L1 in WT ENO1 reconstituted cells, but not in S249A ENO1 reconstituted cells. In addition, proteasome inhibitor MG132, but not lysosomal inhibitor chloroquine (CQ) or the autophagy inhibitor 3-methyladenine (3-MA), rescued PD-L1 degradation in S249A ENO1 reconstituted cells, suggesting that ENO1 S249 O-GlcNAcylation represses PD-L1 expression through the proteasomal degradation pathway (Fig. 4 D and E).

Next, we investigated the mechanism by which S249 O-GlcNAcylation of ENO1 promotes PD-L1 stability. It was reported that ENO1 regulated the proteasomal degradation of PD-L1 by recruiting E3 ligase STUB1 (26). To verify this finding, we first depleted STUB1 and found that PD-L1 expression was increased, accompanied by a reduction of PD-L1 ubiquitination in RKO and HCT116 cells, confirming STUB1 as the E3 ligase of PD-L1 in CRC cells (SI Appendix, Fig. S6A). Moreover, the depletion of ENO1 in RKO cells and HCT116 cells increased PD-L1 levels accompanied with reduced PD-L1 ubiquitination. The effect was reversed by the reconstituted expression of WT ENO1 (SI Appendix, Fig. S6 B and C). In contrast, overexpression of ENO1 in RKO cells and HCT116 cells repressed PD-L1 expression. The inhibitory effect was abrogated by depleting STUB1 in cells (SI Appendix, Fig. S6D). Moreover, depletion of ENO1 inhibited the association of PD-L1 with STUB1 in RKO cells and HCT116 cells, while overexpression of ENO1 enhanced the interaction of PD-L1 and STUB1 in a dose-dependent manner (SI Appendix, Fig. S7 A–C). Pull-down assays showed that incubation of His-tagged ENO1 purified from E. coli enhanced the association of GST-tagged PD-L1 with Flag-tagged STUB1 isolated from E. coli and 293 T cells, respectively (SI Appendix, Fig. S7D). Thus, these results confirm that ENO1 facilitates the binding of PD-L1 to STUB1.

We speculate that ENO1 glycosylation represses the proteasomal degradation of PD-L1 by reducing the binding of PD-L1 to STUB1 (SI Appendix, Fig. S7E). To test this, we detected PD-L1 ubiquitination and its interaction with STUB1 in WT and S249A ENO1 reconstituted RKO cells, or HCT116 cells. TMG treatment reduced PD-L1 ubiquitination and its association with STUB1 in WT ENO1 reconstituted cells, but not in S249A ENO1 reconstituted cells (Fig. 4F). Furthermore, overexpression of OGT enhanced the expression of PD-L1 in WT ENO1 reconstituted cells, but not in S249A ENO1 reconstituted cells. This increase was abrogated by the depletion of STUB1 (Fig. 4G). Taken together, these data demonstrate that ENO1 glycosylation blocked its interaction with PD-L1, reduced STUB1 interaction with PD-L1, thus inhibiting STUB1-mediated ubiquitination and proteasomal degradation of PD-L1.

S249 O-GlcNAcylation of ENO1 Inhibits Antitumor Activity and Promotes Colorectal Tumor Growth.

As ENO1 O-GlcNAcylation stabilizes PD-L1 expression, we further ask whether ENO1 O-GlcNAcylation impacts T cell cytotoxicity for tumor cells. We restored PD-L1 expression in S249A ENO1 reconstituted RKO cells to the comparable level as that in WT ENO1 reconstituted cells (Fig. 5A). These cell lines showed comparable cell proliferation rates (Fig. 5B). When cocultured with activated human lymphocytes (peripheral blood mononuclear cells), S249A ENO1 reconstituted cells displayed an increased cell cytotoxicity compared to WT ENO1 cells. Restoration of PD-L1 expression in S249A ENO1 reconstituted cells repressed T cell-mediated cytotoxicity to the level observed for WT ENO1 cells (Fig. 5C). Further, compared to WT ENO1 reconstituted cells, CD8⁺ T cells cocultured with S249A ENO1 reconstituted cells displayed higher expressions of IFN-γ and TNF-α, which are signature cytokines secreted by activated T cells to kill tumor cells. Restoring PD-L1 expression also repressed the activity of CD8⁺ T cells cocultured with S249A ENO1 cells (Fig. 5 D and E). Similar results were observed in MC38 murine colon cancer cells cocultured with CD8⁺ T cells (SI Appendix, Fig. S8 A–F).

Fig. 5.

S249 O-GlcNAcylation of ENO1 inhibits antitumor activity and promotes colorectal tumor growth. (A) Immunoblotting of PD-L1 expression in stable RKO cells expressing WT, S249A ENO1, or S249A ENO1 with ectopic expression of PD-L1. (B) Cell proliferation analysis of stable RKO cells expressing WT or S249A ENO1 with ectopic expression of PD-L1. (C) Analysis of cytotoxicity in stable RKO cells expressing WT or S249A ENO1 with ectopic expression of PD-L1. (D and E) qPCR analysis of the mRNA expressions of IFN-γ (D) or TNF-α (E) in PBMCs after coculture with RKO cells. (F) Images of tumors derived from stable MC38 cells expressing WT or S249A ENO1 with ectopic expression of PD-L1 in immune-competent C57BL/6 mice. (G) Tumor volumes measured at indicated time points. (H) Tumor mass measured at the experimental endpoint (n = 5 mice). (I) Survival of mice injected with stable MC38 cells expressing WT or S249A ENO1 with ectopic expression of PD-L1. n = 10 mice per group. Log-rank test. (J and K) Immunohistochemical analyses of tumor samples using an anti-Ki67 (J) or anti-CD8 antibody (K). Ki67 or CD8 positive cells were quantified using Image J (n = 5). (L and M) MC38 cells expressing WT or S249A ENO1 with ectopic expression of PD-L1 were injected into immunocompetent C57BL/6 mice and further treated with IgG2b or CD8α mAb at the indicated time points. Images of the dissected tumors at the experimental end point (L). Growth curves of tumors measured at the indicated time points (M). (N) Immunohistochemical analysis of OGT expression in primary CRC tissues and matched peritumoral tissues (n = 30 pairs). Representative staining images are shown. (Scale bar, 200 μm.) Comparative analysis of ENO1 expression between CRC tissues and their adjacent normal tissues was performed. (O) Quantification of ENO1 O-GlcNAcylation levels from primary CRC tissues and matched peritumoral tissues (n = 30 pairs). (P) Immunohistochemistry staining of human CRC samples with the indicated antibodies. (Scale bar, 200 μm.) Semiquantitative evaluation of the staining of the tissue sections was performed according to intensity and area. A Pearson’s correlation test was used (two-tailed) (n = 30). Some of the dots on the graphs are overlapped and represent more than one specimen. Error bars in B, C, D, E, G, H, J, K, M, N, and O denote the mean ± SD. Statistical analyses in B, C, D, E, G, H, J, K, and M were performed by one-way ANOVA with Tukey's multiple comparisons test. Statistical analyses in N and O were performed by unpaired two-tailed Student’s t test.

To further investigate whether ENO1 O-GlcNAcylation represses T cell antitumor activity in vivo, we injected MC38 cell lines (WT ENO1 cells, S249A ENO1 cells, and PD-L1 restored S249A ENO1 cells) subcutaneously into immune-compromised nude mice or immunocompetent C57BL/6 mice. The results showed that all three cell lines developed tumors at a comparable rate and size in nude mice (SI Appendix, Fig. S9 A–C). Consistently, S249A ENO1 reconstituted HCT116 cells displayed similar tumor growth rate as compared to WT ENO1 reconstituted cells, and fully rescued the reduced tumor growth mediated by depletion of the endogenous ENO1, supporting that S249A mutant has no impact on tumor growth under immune-compromised conditions (SI Appendix, Fig. S9 D–F). In C57BL/6 mice, however, S249A ENO1 cells exhibited reduced tumor growth, tumor volume, and weight, and a corresponding decrease in Ki67 expression, and improved survival of mice (Fig. 5 F–J and SI Appendix, Fig. S9G). Immunohistochemistry (IHC) analyses of these tumor tissues showed that CD8⁺ T cell infiltration levels were up-regulated in mice inoculated with S249A ENO1 cells compared with those inoculated with WT ENO1 cells (Fig. 5K and SI Appendix, Fig. S9H). Restored expression of mouse PD-L1 in S249A ENO1 cells fully reversed these effects (Fig. 5 F–K and SI Appendix, Fig. S9 G and H). These data indicate that S249 O-GlcNAcylation of ENO1 in CRC cells impairs the antitumor activity of tumor-infiltrating CD8⁺ T cells by stabilizing PD-L1 expression.

To verify whether the antitumor effects of ENO1 glycosylation were dependent on CD8⁺ T cells, a CD8 monoclonal neutralizing antibody (CD8α mAb) was employed in the blockade of CD8⁺ T cell activity in C57BL/6 mice. Immuno-depletion of CD8⁺ T cells eliminated the tumor growth retardation of S249A ENO1 cells (Fig. 5 L and M and SI Appendix, Fig. S9I). Consistently, the number of infiltrating of CD8⁺ T cells in tumor tissues was negatively correlated with the tumor burden SI Appendix, Fig. S9J). Thus, these results confirm that ENO1 O-GlcNAcylation promotes tumor growth in a manner dependent on CD8⁺ T cell activity.

To access the translational relevance of inhibiting ENO1 glycosylation, we injected MC38 cell lines (WT ENO1 cells and S249A ENO1 cells) subcutaneously into immunocompetent C57BL/6 mice in the presence or absence of OGT inhibitor OSMI4 treatment (SI Appendix, Fig. S10A). The result showed that S249A ENO1 cells exhibited reduced tumor growth, tumor volume, and weight as compared to WT ENO1 cells. OSMI4 treatment also significantly inhibited tumor growth of WT ENO1 cells, but had no apparent effect on tumor growth of S249A ENO1 cells (SI Appendix, Fig. S10 B–D).

To further determine the clinical relevance of ENO1 O-GlcNAcylation, we performed western blotting and IHC analyses of 30 pairs of CRC tissues and adjacent normal tissue samples. We observed that the levels of OGT and ENO1 glycosylation were both significantly higher in tumor tissues compared to the matching adjacent normal tissues (Fig. 5 N and O and SI Appendix, Fig. S11). IHC analyses revealed that OGT levels were positively correlated with PD-L1 levels and the infiltration of CD8⁺ T cells (Fig. 5P). These data are in line with the cellular data and suggest that ENO1 O-GlcNAcylation is important in regulating CRC development and immune evasion.

Blockade of ENO1 O-GlcNAcylation Synergizes with PD-L1 mAb Therapy.

Since S249 O-GlcNAcylation of ENO1 represses T cell antitumor activity in CRC, we speculate that blockade of ENO1 O-GlcNAcylation on S249 might improve PD-L1 mAb therapy. To test this, we injected WT or S249A ENO1 reconstituted MC38 cells subcutaneously into C57BL/6 mice. When tumors were palpable, the PD-L1 mAb treatment was performed (BS12A). Combination of PD-L1 mAb treatment and blockade of ENO1 O-GlcNAcylation on S249 repressed tumor growth to a greater degree as compared to the solo treatment, suggesting a synergistic effect (SI Appendix, Fig. S12 B–D). To achieve a better therapeutic efficacy, we eliminated O-GlcNAcylation on both S249 and T19 by establishing stable MC38 cells in which endogenous ENO1 was depleted followed by concurrent reconstituted expression of shRNA-resistant Flag-tagged WT or double site mutations (T19A/ S249A) of ENO1(2A) (SI Appendix, Fig. S12E). These stable cell lines were injected subcutaneously into C57BL/6 mice (Fig. 6A). Blockade of ENO1 O-GlcNAcylation (2A ENO1) alone reduced the tumor size and inhibited tumor growth, which was similar to that with the PD-L1 mAb treatment of tumors derived from WT ENO1 cells. Strikingly, PD-L1 mAb treatment of tumors derived from 2A ENO1 cells inhibited tumor growth to a much greater degree and significantly prolonged the survival of mice (Fig. 6 B–E). In line with this observation, IHC analyses of tumor tissues showed that PD-L1 mAb treatment significantly enhanced the infiltration of CD8⁺ T cells and impeded Ki67 expression in tumors derived from 2A ENO1 cells compared to that in tumors derived from WT ENO1 cells (Fig. 6 F and G and SI Appendix, Fig. S12F). Additionally, the levels of IFN-γ and GZMB, the key effector molecules of cytotoxic T cells to induce cell apoptosis, were significantly up-regulated in tumors derived from 2A ENO1 cells, indicating enhanced T cell activity (Fig. 6 H–J). Taken together, our results suggest that blockade of ENO1 O-GlcNAcylation synergizes with PD-L1 mAb therapy to bolster antitumor T cell immunity.

Fig. 6.

Blockade of ENO1 O-GlcNAcylation synergizes with PD-L1 mAb therapy. (A) Schematic depiction of the combination therapy by blockade of ENO1 O-GlcNAcylation and using anti-PD-L1 antibody (200 μg/mouse). (B) Images of tumors harvested after 18 d were shown (n = 5). (C and D) Analysis of tumor weight (C) and tumor growth curve (D). (E) Survival of mice bearing syngeneic ENO1 O-GlcNAcylation depleted MC38 tumor following treatment with or without anti-PD-L1 antibody (n = 10 mice). Significance was determined by the log-rank test. (F and G) Immunohistochemical analyses of tumor samples using an anti-Ki67 (F) or anti-CD8 (G) antibody. Ki67 or CD8 positive cells were quantified using Image J (n = 5). (H–J) Flow cytometric analysis and quantification of intracellular cytokine staining of IFN-γ and granzyme B in CD8+ T cell populations separated group. Quantification of intracellular cytokine staining of IFN-γ (I) and Granzyme B (J) in CD8+ T cell populations in the lymph nodes. Error bars in C, D, F, G, I, and J denote the mean ± SD. Statistical analyses in C, D, F, and G were performed by one-way ANOVA with Tukey's multiple comparisons test. Statistical analyses in I and J were performed by two-way ANOVA with Tukey's multiple comparisons test.

Discussion

It is well known that cancer cells exhibit aberrant glucose metabolism characterized by increased glucose uptake and a high rate of glycolysis to support rapid proliferation in the hostile tumor microenvironment (46, 47). However, how the escalated glucose metabolism modulates aerobic glycolysis and immune evasion is still poorly understood. Here, we show that O-GlcNAcylation, derived from glucose metabolism, directly modifies the glycolytic enzyme ENO1, serving as a dual regulator of aerobic glycolysis and immune evasion in CRC cells. Mechanistically, O-GlcNAcylation of ENO1 on T19 enhances its activity through increased formation of dimers, thereby promoting aerobic glycolysis and tumor growth. On the other hand, S249 O-GlcNAcylation of ENO1 represses its interaction with PD-L1 and reduces STUB1-mediated ubiquitination and proteasomal degradation of PD-L1, leading to tumor immune evasion (Fig. 7). Thus, glycosylation of ENO1 mechanistically links aerobic glycolysis and immune evasion, two important hallmarks of cancer cells, to synergistically promote colorectal tumor growth.

Fig. 7.

A graphical model of ENO1 O-GlcNAcylation in regulating aerobic glycolysis and tumor immune evasion.

O-GlcNAcylation currently emerges as a key regulator of cellular physiology and plays a critical role in regulating various biological processes, including stress response, gene transcription, metabolism, and immune regulation (48). Multiple types of cancer display aberrantly high levels of O-GlcNAcylation (49). The role of O-GlcNAcylation in cancer development and progression has been actively investigated. Accumulated evidence has now indicated that aberrant O-GlcNAcylation reprograms multiple metabolic pathways, including glucose, lipid, amino acid, and nucleotide metabolism to support the energetic and biosynthetic demands of rapidly proliferating cancer cells (37, 50–52). Our recent study has demonstrated that O-GlcNAcylation directly regulates membrane protein endosomal trafficking machinery to impede the degradation of EGFR and PD-L1 by the lysosomal pathway to promote tumor growth and tumor immune evasion (53, 54). Our findings herein reveal that O-GlcNAcylation simultaneously regulates aerobic glycolysis and tumor immune escape by targeting ENO1 at different glycosylation sites. Blockade of ENO1 O-GlcNAcylation synergizes with PD-L1 mAb therapy to inhibit tumor growth and prolong mouse survival. Thus, targeting O-GlcNAcylation-mediated metabolic and immune adaptations could serve as a viable strategy for antitumor treatments.

Primarily recognized for its pivotal role in glucose metabolism as a glycolytic enzyme, ENO1 has recently garnered attention for its diverse nonmetabolic functions that facilitate tumor progression. In most studies, high expression of ENO1 promotes cancer cell proliferation and tumor growth, indicating ENO1’s function as a protumor factor. However, ENO1 has recently been shown to interact with PD-L1, which leads to ubiquitination and proteasomal degradation of PD-L1, thereby activating T cell immune activity to repress tumor growth (26). In this case, ENO1 appears to function as an antitumor factor. It is not clear how the seemingly opposing functions are reconciled in cells. Our study provides a possible mechanism that reconciles the two opposing functions. O-GlcNAcylation regulates both the metabolic and nonmetabolic functions of ENO1 by promoting ENO1 enzymatic activity on the one hand and stabilizing PD-L1 expression on the other hand. The impact of O-GlcNAcylation on ENO1 augments the aerobic glycolysis and represses tumor immunity, synergistically promoting CRC growth. It is also intriguing to investigate whether O-GlcNAcylation plays a similar role in regulating the development/progression of other types of cancers in a manner dependent on high expression of ENO1.

It is common that multiple glycosylation sites exist in the same protein (55). While it is often observed that specific glycosylation sites on proteins are associated with distinct functional outcomes, our study highlights the unique scenario where distinct sites of O-GlcNAcylation on ENO1 give rise to separate biological functions. Our work relies on mutational studies in which mutations at T19 and S249 of ENO1 alter its O-GlcNAcylation status, affecting aerobic glycolysis and immune evasion, respectively. Further investigation is required to confirm whether cancer cells indeed up-regulate glycosylation levels specifically at these sites to regulate glycolysis and immune response using O-GlcNAcylation site-specific antibody toward T19 or S249. It is also worth noting that the O-GlcNAcylation level of specific proteins is dynamic in response to different stress and nutrient levels (6). It is possible that glycosylation of ENO1 at these two sites is be subjected to regulation in different nutrient and stress states, and that glycosylation of the individual site is regulated independently. As an extension of this study, it would be captivating to explore how OGT differentially targets and glycosylates ENO1 at two different sites in response to the environmental niche.

Methods

Cell Culture and Tumor Tissues.

Cell lines 293 T, HT-29, HCT116, RKO, SW620, LoVo, NCM460, and MC38 were all sourced from American Type Cell Culture (ATCC). All of above cells were at 37 °C under 5% CO2 in Dulbecco’s Modified Eagle’s Medium (DMEM) (Gibco), supplemented with 10% fetal bovine serum (Gibco). Prior to use, all cells underwent testing to confirm the absence of Mycoplasma contamination and were authenticated through short tandem repeat fingerprinting. Colorectal cancer tissue and corresponding tumor-adjacent normal tissues were procured from patients at the first Affiliated Hospital of Zhejiang University (Hangzhou, China). Prior to the commencement of the study, informed consent was obtained from all participating patients. The procedures involving human subjects were approved by the Ethic Committee of the School of Medicine, Zhejiang University.

The Cancer Genome Atlas (TCGA) Analyses.

The data of ENO1 mRNA expression level in the colorectal cancer (COAD) and matching peritumoral tissues were obtained from the TCGA database and analyzed by using GEPIA2 (http://gepia2.cancer-pku.cn/#analysis). Statistical analyses were performed by ANOVA.

All other experimental methods are provided in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.