O-GlcNAcylation of the Tumor Suppressor FOXO3 Triggers Aberrant Cancer Cell Growth

Pancreatic ductal adenocarcinoma (PDAC) is a type of malignant tumor associated with extremely poor outcomes (1, 2). Tumorigenesis of PDAC has been attributed to many factors, such as loss-of-function mutations or posttranslational modifications (PTM) of tumor suppressor genes (e.g., FOXOs, p53, and RB). FOXO is a subclass O member of the forkhead family of proteins (3), which are known to influence diverse and important biological events such as the cell cycle and apoptosis by regulating target gene expression (e.g., p21, p53; refs. 4, 5). Several types of FOXO (FOXO1, 3, 4, and 6) are present in mammals, and of these, FOXO3 and FOXO4 have been associated with various cancers. In particular, FOXO3 is a known tumor suppressor that represses cell-cycle continuation via p21 and thus effectively represses abnormal cell division (6). Several PTMs that control FOXOs activity have been identified, including glycosylation (N- or O- type; refs. 7–9), phosphorylation (10, 11), N-terminal acetylation (12), and methylation (13). However, the potential correlation of FOXO3 O-GlcNAcylation with PDAC or other cancers has not been established.

O-GlcNAc is an O-linked β-N-acetylglucosamine moiety that reversibly attaches to the side-chain hydroxyls of serine or threonine residues in both cytoplasmic and nuclear proteins (7, 14–16). Protein O-GlcNAcylation is regulated by two enzymes, O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA). OGT uses UDP-GlcNAc as a substrate for O-GlcNAcylation, whereas OGA removes O-GlcNAc from side chains. In cancer cells, the high rate of anaerobic glycolysis (i.e., the Warburg effect) provides a main source for O-GlcNAcylation via the hexosamine biosynthetic pathway (HBP; ref. 17), and a causative link has been identified between cancer mortality and abnormal glucose metabolism (hyperglycemia; ref. 18). This type of modification usually competes with phosphorylation at the same serine or threonine sites on proteins targeted for regulation (16). However, the number of targetable sites located within FOXO3 and the potential for site-specific O-GlcNAcylation within FOXO3 to affect PDAC tumorigenesis remain unknown.

We aimed to explore whether site-specific O-GlcNAcylation would regulate FOXO3 target gene expression in the context of pancreatic cancer cell proliferation. Because O-GlcNAcylation serves as a cancer hallmark (19), with increased levels in various cancers relative to adjacent nontumor tissues (20), we hypothesized that site-specific O-GlcNAcylation within FOXO3 might reduce the tumor-suppressive activity of FOXO3 via its target genes (e.g., p21) and thus trigger the tumorigenic activities of pancreatic cancer cells. Here, we show that the site-specific O-GlcNAcylation within FOXO3 Ser284 termed S284 O-GlcNAc seems to work as a critical block of FOXO3 tumor suppressor activity, which appears to be via a part of p53 pathway composed of MDM2-p53-p21 axis (21) in triggering an accelerated cancer cell growth.

Tumor and adjacent nontumor tissues were surgically collected from human patients and immediately frozen at −78°C until use. O-GlcNAc modification within FOXO3 was evaluated in human pancreatic ductal carcinoma (n = 24), hepatocellular carcinoma (n = 10), gastric tubular carcinoma (n = 10), cholangiocarcinoma (n = 5), colon carcinoma (n = 10), and lung carcinoma tissues (n = 10). All tissue samples were obtained from the archives of the Department of Pathology, Yonsei University College of Medicine (Seoul, Korea). This study was approved by the Institutional Review Board (IRB, 4-2015-0474) of Yonsei University of College of Medicine, and PDAC diagnoses were made by pathologists at the Severance Hospital of Yonsei University. All subjects received a written informed consent that the studies were conducted in accordance with recognized ethical guidelines.

Pancreatic cancer cell lines (PANC-1, BxPC-3, and HPAC) were purchased from ATCC, human pancreatic duct epithelial cell line (HPDE) was kindly provided by Dr. Ming-Sound Tsao (University of Toronto). All cell lines were authenticated by short tandem repeat (STR) profiling by Korean Cell Line Bank and shown to be negative in mycoplasma test using Mycoprobe Mycoplasma Detection Kit (R&D System). The following ATCC-specified cell culture media were used: DMEM for PANC-1, RPMI1640 for BxPC-3, and DMEM-F12 (1:1) for HPAC with 10% FBS. HPDE cells were grown in serum-free media supplemented with bovine pituitary extract and epidermal growth factor. All cells were cultured in a 37°C incubator with a 5% CO₂ atmosphere. PANC-1 cells were transiently transfected using Lipofectamine 3000 (Invitrogen) and selected with G418 (Sigma-Aldrich) according to the manufacturer's instructions. Human FOXO3 DNA was purchased from Origene (Rockville), and human OGT DNA was kindly provided by Dr. Jin Won Cho (Yonsei University).

Antibodies specific for FOXO3 (2497, 12829), p21 (2947), CDK2 (2546), CDK-4 (12790), and cyclin D1 (2978) were purchased from Cell Signaling Technology. Antibodies specific for O-GlcNAc (ab2739), OGA (ab124807), MDM2 (ab10344), CDK1 (ab18), and cyclin G2 (ab54901) were purchased from Abcam. Antibodies specific for GAPDH (sc-25778), p53 (sc-126), cyclin A (sc-751), cyclin E (sc-481), anti-mouse (sc-2005), and anti-rabbit (sc-2030) were purchased from Santa Cruz Biotechnology. An OGT-specific antibody (O6264) was purchased from Sigma-Aldrich. Thiamet G [1009816-48-1, 3aR,5R,6S,7R,7aR)-2-(Ethylamino)-3a,6,7,7a-tetrahydro-5-(hydroxymethyl)-5H-pyrano[3,2-d]thiazole-6,7-diol, OGA inhibitor] was kindly provided by Dr. Jin Won Cho (Yonsei University).

Cell lysates were incubated with specific antibodies and lysis buffer. Subsequently, agarose beads (Sigma-Aldrich) were added and incubated. The beads were washed three times with PBS-T, and eluted using PBS pH 2.0, neutralized with 1 N NaOH. The resulting samples were separated by SDS-PAGE (10%) and transferred onto nitrocellulose membranes using an iBLOT dry blotting system (Invitrogen). The membranes were then blocked with 5% skim milk in TBS-T, incubated for anti-FOXO3 or O-GlcNAc antibody, and subsequently incubated with secondary antibody (Santa Cruz). Immunoreactive proteins were detected using ECL Plus detection reagents (GE Healthcare) and detected using a Typhoon 9400 scanner (GE Healthcare). The band intensity was densitometrically evaluated using ImageQuant TL software (GE Healthcare).

Liquid chromatography separation was performed on an Ultimate 3000 RS nano LC system (Thermo Scientific). A C18 Easy nano column was used for peptide separation. Mass spectra for peptide identification or quantification were acquired using an Orbitrap Q Exactive mass spectrometer (Thermo Scientific). Peptide fragmentation was performed via high-energy collision dissociation with a set energy of 27 NCE. The ion selection intensity threshold was set at 1.0 × 10⁵ with charge exclusions of z = 1 and z > 7. MS/MS spectra were acquired at a resolution of 17,500, with a target value of 2 × 10⁵ ions or maximum integration time of 120 milliseconds and a set isolation window of 2.0 m/z.

Specific primers for serine (S) and threonine (T) to alanine (A) and aspartic acid (D) mutations of FOXO3 were designed and used to insert point mutations in a plasmid vector. A PCR-amplified DNA fragment of pCMV6-FOXO3 was generated using Phusion DNA polymerase (NEB). After PCR, the nonmutated sequences were cleaved using DpnI (NEB). The mutated vectors were transformed into Escherichia coli (DH5α) that were cultured and prepared using a Plasmid Maxi Kit (Qiagen).

Total RNA was extracted using TRIzol (Invitrogen) and purified using an RNeasy Kit (Qiagen) according to the manufacturers' instructions. Complementary DNA (cDNA) was obtained via reverse transcription and subsequently combined with primers, a 10 mmol/L dNTP mixture, and SYBR Green PCR Mix (Applied Biosystems). iQ SYBR Green Supermix (Bio-Rad Laboratories) was used for qRT-PCR analyses. The reaction products were run on a CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories). Serial cDNA dilutions were used to generate GAPDH intensity reference standards.

Nano LC–MS/MS analysis was performed using an Easy n-LC (Thermo Fisher) and an LTQ Orbitrap XL mass spectrometer (Thermo Fisher) equipped with a nano-electrospray source. Samples were separated on a C18 nanobore column. Mass spectra were acquired using data-dependent acquisition with a full mass scan (350–1200 m/z), followed by 10 MS/MS scans. For the MS1 full scans, the Orbitrap resolution was 15,000, and the AGC was 2 × 10⁵. For MS/MS in the LTQ, the AGC was 1 × 10⁴.

Three independent experiments were performed per assay, and the data are presented as means ± SDs. Data were analyzed using Student t test, and P values of <0.05 were considered statistically significant.

Additional information on detailed analysis procedures, reagents, methods for cell culture, DNA transfection, immunohistochemistry, protein fractionation, MS identification, Site-directed point mutagenesis, RNA isolation, qRT-PCR, total cell counting, WST-1 assay, crystal violet staining, FACS analysis, nuclear fractionation, label-free MS quantifications, GO analysis, KEGG analysis, and chromatin immunoprecipitation (ChIP) are provided in Supplementary Materials and Methods.

To determine quantitative differences in global levels of O-GlcNAc between tumor and adjacent nontumor regions of PDACs, we measured the relative levels of O-GlcNAc in paired tumor/adjacent nontumor tissue samples pooled from 12 PDAC patients. The global levels of O-GlcNAc in PDAC tissues were much higher in tumor regions relative to those in adjacent nontumor regions (9.2 ± 0.52-fold, n = 3; Fig. 1A). However, the immunohistochemical staining assay showed that there is essentially no difference in FOXO3 protein level in the nuclei of both tumor and normal adjacent tissues in 16 paired PDAC tumor tissue sets (Fig. 1B). This result suggests that the level of FOXO3 expression remains steady, regardless of malignancy, as opposed to that of global O-GlcNAc. We then assessed via IP the levels of O-GlcNAc that are predicted to have originated from FOXO3 in tumor tissues with a FOXO3-specific monoclonal antibody (mAb), followed by Western blotting (IP/WB analysis) with an anti-O-GlcNAc mAb. A clear band on the blot indicated that the increase in O-GlcNAc was probably attributable to FOXO3 protein binding (Fig. 1C). The FOXO3 O-GlcNAcylation levels were much higher in the tumor regions of pancreatic tissues than in adjacent nontumor regions (7.37 ± 1.68-fold, n = 3; Fig. 1C). With quadrupole time-of-flight tandem mass spectrometry (Q-TOF MS/MS; Agilent 6530), we confirmed that the excised protein from the gel used for Western blotting was indeed FOXO3 (MW: 71,517 Da; score: 1455; pI: 4.99; queries matched: 227; sequence coverage: 77%). Reproducibility of our observation was confirmed from the repeated experiment using the 12 paired tissue samples from individual PDAC patients in which all tumor regions exhibited higher O-GlcNAcylation within FOXO3 (4.61 ± 0.38-fold, n = 12; Fig. 1D). Interestingly, when we performed a subsequent IP/WB screen for the O-GlcNAcylated form of FOXO3 in malignant tumors from six different tissues (pancreas, liver, stomach, bile duct, colon, and lung, n = 69), we found notably elevated O-GlcNAc levels that probably originated from FOXO3 in two other tumor tissues, gastric adenocarcinoma (GC) and hepatocellular carcinoma (HCC). Their relative levels of O-GlcNAc are as follow: PDAC, 4.61 ± 0.38-fold, n = 12; GC, 2.65 ± 0.31-fold, n = 10; and HCC, 1.76 ± 0.46-fold, n = 10 (Fig. 1E). We focused our remaining investigations on the possible consequential role of enhanced FOXO3 O-GlcNAcylation in the context of PDAC tumorigenesis, in that the most significant changes in FOXO3 O-GlcNAcylation level were identified in PDAC tumors (vs. GC or HCC tumors).

After confirming the presence of significantly enhanced FOXO3 O-GlcNAcylation in PDAC (4.61 ± 0.38-fold, n = 12), we wanted to map specific O-GlcNAcylation sites and further quantify the O-GlcNAc levels of FOXO3 so as to elucidate the potential roles of these sites relative to cancer cell proliferation. To this end, we generated FOXO3-overexpressed cell lines for the mapping of specific sites and the quantification of O-GlcNAc levels within FOXO3. Judging that this exploration would be more experimentally feasible in FOXO3-overexpressed PDAC cell lines, we selected a few pancreas-origin cell lines (e.g., HPDE, PANC-1, BxPC-3, and HPAC) for our stepwise O-GlcNAc site-mapping procedure. First, we used standard methods (22) to generate a FOXO3-overexpressing PANC-1 cell line, FOX-OV, wherein a pCMV-FOXO3-GFP expression vector was stably transfected into PANC-1 cells (Supplementary Fig. S1); subsequently, we performed a protein separation series using OFFGEL and a proteomic analysis using a Thermo Q Exactive LC–MS (Fig. 2A). Western blotting of the cell lysates obtained from FOX-OV cells revealed a clear band corresponding to FOXO3 (MW = 71.2 kDa; top part in Fig. 2B), whereas FOXO3 was not detectable in other cell lines (e.g., PANC-1, mock, GFP cells). IP/WB analysis was also performed to confirm the O-GlcNAcylation of FOXO3 in FOX-OV cells (bottom part in Fig. 2B). Second, to obtain purified FOXO3 proteins for O-GlcNAc site mapping by MS, the FOX-OV lysates were fractionated on an Agilent 3100 OFFGEL fractionator to remove nonspecific proteins, and the purified proteins were then subjected to IP/WB. As anticipated, the enriched proteins were confirmed to be FOXO3, with a theoretical pI of 5.24. The Western blot data support our conclusion that the recombinant FOXO3-GFP protein overexpressed in FOX-OV cells matched the expected size and was highly enriched (Fig. 2C). Third, to verify the chemical nature of O-GlcNAc enrichment on FOXO3 in FOX-OV cells, we subjected the purified FOXO3 proteins to liquid chromatography-mass spectrometry (LC–MS)/MS analysis on a Thermo Q Exactive LC–MS. We conducted 13 independent experiments after treating the proteins with a combination of endoproteases (Trypsin, AspN, and LysC). Finally, we were able to map seven O-GlcNAcylation sites, which are located at the transactivation domain of the overexpressed FOXO3 (S284, S286, S411, T475, S476, S482, and S551), to which the transcription coregulator usually attaches (Fig. 2D). Repeated rigorous MS analyses yielded an average recombinant overexpressed FOXO3 score of 10,969.46 ± 4691 (n = 13), with 92.72% coverage (Supplementary Table S1). In addition, the LC–MS/MS peaks matched those of the predicted O-GlcNAcylated peptides (Supplementary Fig. S2). Thus, FOX-OV cells enabled the precise mapping of O-GlcNAcylated sites and provided a basis for the further molecular characterization of the potential role of site-specific O-GlcNAc residue(s) in PDAC cell proliferation.

To analyze FOXO3 target gene expression in the site-specific O-GlcNAc-deficient FOX-OV cells, we constructed mutant cell lines where each of the seven potential O-GlcNAc sites was mutated to alanine (S284A, S286A, S411A, T475A, S476A, S482A, S551A). All mutated recombinant proteins were well expressed in the PANC-1 cell line with evidence of O-GlcNAc attached (Supplementary Fig. S3). We then evaluated how O-GlcNAcylation at each site affected the expression of various FOXO3 target genes. Although the expression of most FOXO3 target genes remained unchanged, the levels of p21, a cyclin-dependent kinase inhibitor 1 (CDKN1A), were significantly increased (57.6 ± 2.91%, P = 0.0015) in FOX-S284A mutant cells relative to those in FOX-OV cells in which all seven amino acids are predicted to be O-GlcNAcylated (Fig. 3A). To eliminate any possibility that S284 phosphorylation might also affect p21 expression, we also generated FOX-S284D (Ser to Asp) cell lines, which harbored a phospho-mimetic FOX-OV, to distinguish the effect of a loss of O-GlcNAcylation from S284 phosphorylation. Our qRT-PCR and Western blotting results revealed that p21 expression increased only in FOX-S284A cells lacking O-GlcNAc, and not in those expressing FOX-S284D (Fig. 3B and C), suggesting that O-GlcNAc (S284 O-GlcNAc), not phosphorylation, might have suppressed p21 expression as a result of the loss of FOXO3 tumor suppressor activity (23).

To corroborate a functional link between a PTM change at FOXO3-S284 and the cell-cycle–regulatory activity of p21 in PDAC cells, we assessed the cell proliferation rates and cell-cycle–regulatory functions of cells exhibiting S284 O-GlcNAc (e.g., FOX-OV) versus those of mutant cells lacking S284 O-GlcNAc (e.g., FOX-S284A and FOX-S284D). After transfecting FOXO3-GFP and mutant recombinant DNA (FOX-S284A and FOX-S284D) into PANC-1 cells, we counted the total cell numbers over time. At 5 days posttransfection, when changes in cell numbers were optimally observed, we counted much higher numbers of FOX-OV cells relative to either FOX-S284A or FOX-S284D cells (Fig. 4A), suggesting a clear stimulating effect of O-GlcNAc at FOXO3 S284 on PDAC cell proliferation. Similarly, a quantitative WST-1 assay of cell proliferation revealed reduced proliferation rates in both FOX-S284A and FOX-S284D cells relative to those in FOX-OV cells (Fig. 4B). In addition, when viable cells were stained with crystal violet (24), the FOX-S284A and FOX-S284D cell populations fixed at 5 days posttransfection remained essentially the same, whereas the numbers of FOX-OV cells increased considerably under the same conditions (Fig. 4C). Altogether, these results suggest that the attachment of O-GlcNAc at FOXO3 S284 could shift the cell phenotype to highly proliferative by interrupting FOXO3 tumor suppressor activity via p21 repression. Next, we asked how this restored p21 protein might inhibit the activities of cyclin-CDK family complexes during cell-cycle progression from the G₁ to the S phase in this cell line. We analyzed the cell cycles of FOX-OV, FOX-S284A, and FOX-S284D mutant cells at 5 days posttransfection using FACS. In contrast to FOX-S284A mutant cells, which underwent G₁ arrest, the phospho-mimetic cell line FOX-S284D underwent G₂ arrest (Fig. 4D), indicating that S284 O-GlcNAc has a clear disruptive effect on cell cycle regulation, probably by suppressing FOXO3-directed p21 activity.

To explore which proteins or oncogenic signaling pathways might be strongly affected by changes in S284 O-GlcNAc levels in PDAC cell lines, we profiled protein levels in both FOX-OV and FOX-S284A cells, and the relative protein levels between the cell lines using label-free LTQ Orbitrap XL LC–MS/MS. Of the 1,051 detected proteins (FDR <1% at peptide and protein levels, peptide count ≥ 2, peptide sequence number ≥ 9), 339 were found to be differentially expressed (>3-fold) and were subjected to further filtration through the SIEVE program, followed by a heat map–based peptide analysis (Fig. 5A). The highly confident scores (0.940–0.971) obtained for these selected proteins were supported by a volcano plot peak shape (Fig. 5B). Using these results, we constructed a cluster of metabolic proteins, including oncogenic proteins (e.g., MDMs, MAPKs, and c-MYC), tumor suppressors (e.g., BRCA1, NF-κB, HSPs), and p53 pathway–related proteins (e.g., p21) that might have been altered by S284 O-GlcNAc (Fig. 5C). Next, we selected 249 proteins that exhibited approximately ±1.5-fold changes with two or more peptides (increase: 113 proteins, decrease: 136 proteins). After quantifying these selected proteins, we further analyzed their potential roles by a cross-examination of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (Supplementary Table S2) and Gene Ontology (GO) annotation (Fig. 5D) using the PANTHER DB (Protein ANalysis through Evolutionary Relationships). The KEGG database combines various types of information about systemic cellular and organismal functions, including metabolic and cellular processes and human diseases (25, 26). We classified the proteins into various system categories according to molecular function, biological processes, and pathways using the PANTHER DB (27–29). From this cancer signaling analysis, we found that S284 O-GlcNAc possibly causes transcriptional dysregulation of FOXO3-interacted proteins (p53) involved in cell-cycle control (Fig. 5D).

To determine which cell signaling pathway(s) regulate the above-described differences in cell arrest (G₁ vs. G₂ with respect to the O-GlcNAcylation sites, we subjected mock, FOX-OV, FOX-S284A, and FOX-S284D cells to a qRT-PCR and Western blot analysis of proteins with potential involvement in cell-cycle regulation. First, when the transcriptional analysis of three major genes (MDM2, p53, and p21) involved in p53 pathway (Fig. 5C), the expression of both p53 and p21 genes was highly activated in S284A (lacking O-GlcNAc residues), not S284D mutant cells (phosphomimetic), compared to those in FOX-OV cells (containing O-GlcNAc residues), whereas the expression of MDM2 was highly suppressed (Fig. 6A). Second, when the expression of these three proteins were accessed by Western blotting, their expression pattern remained essentially the same as those observed in the transcription assays (Fig. 6B, see bottom panel for graphic illustration). Third, whereas the expression of FOXO3 protein remained unchanged in any of cells examined (i.e., FOX-OV, FOX-S284A, FOX-S284D), the major regulatory proteins involved in the oncogenic cell cycle (e.g., CDK2, CDK4, and cyclin D1) in the S284A, not S284D mutant cells, were downregulated (Fig. 6B; Supplementary Fig. S4A). Taken together, results from our comprehensive analysis using label-free MS quantification, qRT-PCR, and Western blot analysis strongly suggest that S284 O-GlcNAc indeed causes activation of MDM2 expression at both mRNA and protein levels, which consequently suppresses the p53 expression, results in inhibition of p21 expression (Fig. 6C, top; refs. 30, 31). In other words, a lack of S284 O-GlcNAc may allow the cell to inhibit a cell proliferative state in which p53 regulates the function of p21, resulting in cell-cycle arrest (Fig. 6C, bottom).

In this study, we provide evidence to support S284 O-GlcNAc as a critical block of FOXO3 in the context of PDAC tumorigenesis. Accordingly, our results clearly demonstrate the molecular interplay between the site-specific PTM of the tumor suppressor and tumorigenesis, which suppresses the cell-cycle regulator p21 via MDM2–p53–p21 axis in cancer cells (Fig. 7). To date, no reports have described a link between quantitative changes in site-specific O-GlcNAcylation of FOXO3 and pancreatic cancer. Therefore, to the best of our knowledge, our report is the first to describe on the direct consequence of O-GlcNAcylation at S284 of FOXO3 in tumorigenesis.

Given the above features, we have identified a few important points worth noting. First, our work has identified S284 O-GlcNAc as a critical block of FOXO3 that triggers proliferation of pancreatic cancer cells (Fig. 2 and 4) and thus paves the way for similar studies involving other types of cancer (e.g., HCC and GC) that we have found to harbor elevated global O-GlcNAc levels (Fig. 1E). Second, our results provide clear evidence that S284 O-GlcNAc activates the MDM2 oncoprotein, which then suppresses the MDM2-p53-p21 axis (32). Our findings clearly set the possible role of modified FOXO3 that appears to activate MDM2 and consequently degrades p53, which provides additional view on the relationship between FOXO3 and MDM2 as reported earlier (30). Third, our quantitative proteomic profiling analysis of cancer-related proteins in FOX-S284A mutant cells indicates the presence of a cohesive MDM2-p53-p21 axis within the p53 pathway that appears to work in concert in response to the S284 O-GlcNAc, resulting in the downregulation of many oncogenic proteins (Fig. 5C; refs. 33, 34).

Despite clear increases in the O-GlcNAc levels of FOXO3 in FOX-OV cells, we observed no specific changes in the OGT or OGA levels that could be attributed to a loss of the FOXO3 tumor suppressor function (Supplementary Fig. S4B and S4C). Therefore, to investigate this phenomenon, we measured the relative expression levels of key genes involved in the HBP, including phosphofructokinase 1 (PFK1; ref. 35), which is known to direct the reaction flux to HBP by generating fructose-6-phosphate (F6P), and glutamine:fructose-6-P amidotransferase (GFAT; ref. 16), which converts F6P to Glc-6-P (key precursor substrate of OGT) and serves as the rate-limiting step in the HBP. Interestingly, the expression of both genes was notably increased in FOX-OV cells relative to FOX-S284A cells (Supplementary Fig. S4B and S4C), which suggests a link between HBP activation and FOXO3 O-GlcNAcylation in the absence of altered OGT or OGA activity. Therefore, HBP pathway enzymes might stimulate FOXO3 O-GlcNAcylation in pancreatic cancer cells (Fig. 7).

Within cells, FOXOs are usually phosphorylated by kinases such as Akt in response to environmental stress (e.g., carcinogenesis, heat shock, oxidative stress; ref. 36). This modification induces changes in FOXO localization and its resulting transcriptional activities. To test whether site-specific S284 O-GlcNAcylation or phosphorylation would cause FOXO3 to shuttle between the nucleus and cytoplasm, we isolated both nuclear and cytosolic fractions from PDAC cells (37) with different FOXO3 S284 statuses and examined changes in the localization of FOXO3 proteins. Accordingly, we found that changes in the FOXO3 S284 status did not appear to affect the nucleocytoplasmic localization of this protein (Supplementary Fig. S5A). The lack of altered nucleocytoplasmic localization suggests the likelihood of different efficiencies in binding to promoter regions of the target gene (p21; ref. 38). However, this suggestion does not seem to be accurate, because we observed no notable difference between FOX-OV and FOX-S284A in terms of binding efficiencies to the p21 gene promoter in a ChIP experiment (Supplementary Fig. S5B). Our preliminary results regarding cellular behavior or phenotyping suggest that the O-GlcNAcylation of FOXO3 S284 is unlikely to affect either cell migration or invasion. Moreover, FOX-S284A mutant cells, which lack O-GlcNAcylation at S284, exhibited a lower proliferation rate and cell stability and greater dose-dependent sensitivity to mitomycin C (i.e., reduced tumor cell growth) than FOX-OV cells. However, a human apoptosis array of FOX-S284A cells found that the expression of most proteins involved in the apoptosis signaling pathway did not differ noticeably from FOX-OV cells. Accordingly, the effect of site-specific O-GlcNAcylation of FOXO3 more likely controls cell proliferation without affecting cellular apoptosis.

In conclusion, it would be interesting to test whether a similar type of S284 O-GlcNAc found in pancreatic cancer cell can also present in both GC and HCC, which have been shown to exhibit abnormally high levels of FOXO3 O-GlcNAcylation (Fig. 1E). However, the mechanism that underlies the strong O-GlcNAcylation of FOXO3 remains puzzling because no change occurred in the expression of either OGT or OGA in pancreatic cancer cells that harbor the highly O-GlcNAcylated FOXO3-S284 (Supplementary Fig. S4B and S4C). From the perspective of potential clinical applications, our study may be the first to demonstrate the potential of S284 O-GlcNAc as a target for cancer drugs such as gemcitabine. Therefore, this link between S284 O-GlcNAc and cancer cell proliferation could also be applied to translational studies and may lead to the discovery of potential therapeutic and tumor-suppressive targets (39).

No potential conflicts of interest were disclosed.

Conception and design: Y.-K. Paik

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Kim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Shin, H.-J. Cha, J.-Y. Cho, Y.-K. Paik

Writing, review, and/or revision of the manuscript: H. Shin, H. Kim, Y.-K. Paik

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Shin, H.-J. Cha, K. Na, M.J. Lee, J.-Y. Cho, C.-Y. Kim, E.K. Kim, C.M. Kang, Y.-K. Paik

Study supervision: Y.-K. Paik

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the article.

This work was supported by a grant from the Korean Ministry of Health and Welfare (HI13C2098 and HI160274 to Y.-K. Paik). We thank Prof. Jin Won Cho (Yonsei University) for his kind suggestions and for providing the human OGT vector and Thiamet G (OGA inhibitor). We also thank Prof. Jaewhan Song (Yonsei University) for his many helpful suggestions and kind assistance with FACS-based cell-cycle measurements.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.