Overexpression of DPAGT1 Leads to Aberrant N-Glycosylation of E-Cadherin and Cellular Discohesion in Oral Cancer

Epithelial tissues maintain their mature architecture through the function of E-cadherin, a calcium-dependent N-glycoprotein cell-cell adhesion receptor (1–3). E-cadherin–mediated cell-cell contacts comprise multiprotein complexes known as adherens junctions (4) that play pivotal roles in cell polarity, cell-surface mechanics, cell migration, cell survival, cell differentiation (1, 5–8), and prevention of tumor spread (9).

E-cadherin is a single-span transmembrane N-glycoprotein containing five extracellular domains, or ectodomains, a transmembrane region, and a cytosolic tail (1). During homotypic cell-cell contact formation, the ectodomains dimerize in a Ca²⁺-dependent manner and interact with ectodomains on adjacent cells. The cytosolic tails of E-cadherins bind catenins that mediate the interaction with the actin cytoskeleton (10, 11). The interaction between E-cadherin/β-catenin complexes and α-catenin produces initial adherens junctions (4, 12, 13) that can further associate with actin-binding and actin-crosslinking proteins, adaptor proteins, and signaling molecules to produce diverse scaffolds (4, 14). Functional diversity of E-cadherin relies on its ability to form distinct adherens junctions, the composition of which is subject to dynamic rearrangements depending on the cell context (15).

N-glycosylation of E-cadherin has been implicated in the regulation of cell-cell adhesion (16–19). We and others have shown that complex oligosaccharides that modify E-cadherin in proliferating cells inhibit adhesion, whereas high-mannose/hybrid structures promote the formation of mature adherens junctions in dense cultures (18, 19). A likely upstream regulator of the N-glycosylation of E-cadherin is the DPAGT1 gene encoding the dolichol-P-dependent N-acetylglucosamine-1-phosphate transferase (20, 21) that initiates the synthesis of the lipid-linked oligosaccharide precursor for protein N-glycosylation in the endoplasmic reticulum. By regulating the abundance of lipid-linked oligosaccharide, DPAGT1 controls the extent of protein N-glycosylation (20, 21). High DPAGT1 expression is associated with cell proliferation and nascent adherens junctions, whereas down-regulation of DPAGT1 characterizes growth-arrested cells with mature adherens junctions (18, 19, 22).

In many cancers, progression to metastasis is accompanied by down-regulation of E-cadherin expression. Oral cancer, one of the most pernicious malignancies, has also been characterized by diminished E-cadherin levels (23–28). Nonetheless, a subset of oral cancers prominently expresses E-cadherin and maintains the ability to invade the surrounding tissue (27, 29). Because N-glycosylation of E-cadherin affects its ability to form mature adherens junctions, it is possible that aberrant N-glycosylation is the underlying cause for defective adhesion in some oral cancers. Indeed, N-glycosylation of cell-surface glycoproteins is frequently altered in cancer (30–32), with tumor progression linked to increases in the abundance of complex N-glycans (33–35).

In the present study, we examined the relationship between DPAGT1 expression and E-cadherin–mediated adhesion in oral cancer. We show that dense cultures of human salivary epidermoid carcinoma A253 cells displayed aberrantly high DPAGT1 expression and produced E-cadherin modified with complex N-glycans in nascent adherens junctions. Partial inhibition of DPAGT1 with small interfering RNA (siRNA) reduced the modification of E-cadherin with complex N-glycans and promoted the formation of mature adherens junctions as well as the assembly of functional tight junctions. Similar to A253 cells, nine archival and three fresh oral squamous cell carcinoma (OSCC) specimens exhibited elevated levels of DPAGT1, formed immature adherens junctions, and lacked detectable tight junctions. We conclude that N-glycosylation is one of the regulators of E-cadherin tumor suppressive activity by affecting the stability of adherens junctions and the assembly of tight junctions.

Reagents. Monoclonal antibodies to human E-cadherin cytoplasmic region, α-catenin, β-catenin, γ-catenin, PP2A, IQGAP1, ZO-1, Ki-67, and IgG isotype controls were obtained from BD. Polyclonal antibodies to claudin-1 and occludin and monoclonal antibody to cytokeratin AE1/AE3 were purchased from Zymed. Monoclonal antibodies to vinculin and pan-actin Ab-5 were from Upstate Biotechnology and Neomarkers, respectively. Rabbit polyclonal antibody to hamster DPAGT1 was prepared commercially (Covance Research Products). Secondary antibodies included goat anti-mouse and anti-rabbit IgG derivatized with either FITC or Texas red (Molecular Probes).

Cell culture and tissues. Cultures of primary human oral keratinocytes (HOK) and human salivary epidermoid carcinoma A253 cells were grown in keratinocyte medium and McCoy's 5a medium, respectively, supplemented with 10% fetal bovine serum, penicillin, and streptomycin. Total cell lysates were prepared by extraction with Triton/β-octylglucoside buffer (19).

Tissue specimens, representing scalpel-generated incisional biopsies from primary tumors, were obtained at the Boston University Medical Center and approved for studies by the institutional review board. Formalin-fixed, paraffin-embedded specimens of primary OSCC, including cytologically normal stratified squamous adjacent epithelia, were obtained from nine individuals with well-differentiated to moderately differentiated OSCC of the lateral tongue border, maxillary gingiva, and maxillary mucobuccal fold. Fresh tissues were obtained from three patients undergoing resection and radical neck dissection of moderately differentiated OSCC of the lateral tongue border. Regions of OSCC and adjacent epithelia, defined by an on-site pathology analysis, were snap-frozen at −80°C. One fraction of each was used for biochemical analyses, whereas another was embedded in OCT compound (Fisher Scientific) for section preparation as described below. Total lysates from adjacent epithelia and OSCC were prepared by extraction with Triton/β-octylglucoside buffer (19). Protein concentrations were determined using BCA assay (Pierce).

RNA interference. SMARTpool of siRNAs targeting DPAGT1 (S) was obtained from Dharmacon. Nonsilencing (NS) negative control siRNA was from Qiagen. A253 cells were transfected at 50% confluence with 200 nmol/L of either NS or S using Lipofectamine 2000 (Invitrogen) and cultured for 72 h. For determination of doubling times, cells were collected at 12 h intervals for a total of 72 h, and cell number was obtained using a hemocytometer.

Quantitative real-time PCR. Total RNAs isolated from A253 cells transfected with either NS or S were used for cDNA synthesis to assess DPAGT1 expression by real-time PCR. The gene expression profiles were generated by normalizing the Ct (threshold cycle numbers) of DPAGT1 with a housekeeping gene S29 and comparing the gene expression of cells treated with NS and S. Statistical analysis was done using real-time PCR from two independent RNA preparations, with each experiment repeated twice (n = 4). The P value between NS and S treatments was calculated using an unpaired t test.

Western blot. Cell and tissue lysates were fractionated on 7.5% SDS-PAGE, transferred onto polyvinylidene difluoride membranes, blocked with 5% nonfat dry milk, and incubated with primary antibodies to selected proteins. Protein-specific detection was carried out with horseradish peroxidase-labeled secondary antibodies and Enhanced Chemiluminescence Plus (Amersham Biosciences). P values were determined by ANOVA.

Peptide N-glycosidase F and endoglycosidase H digestions. Cell and tissue lysates were digested with 100 units of either peptide N-glycosidase F (PNGaseF) or endoglycosidase H (EndoH), both obtained from New England Biolabs, for 1 h at 37°C and analyzed by Western blot. Control samples were incubated without the enzymes.

Immunoprecipitation. Equal amounts of protein (200 μg) were precleared with antibody isotype controls and protein G beads (Sigma; Supplementary Fig. S1A). The resulting supernatants were incubated with 2.5 μg antibody against either E-cadherin or ZO-1 and 30 μL protein G beads for 2 h at 4°C. The beads were washed with lysis buffer and samples were analyzed by Western blot as described (19).

Transepithelial resistance. Transepithelial resistance (TER) was measured in Transwells (polycarbonate membrane, 12 mm diameter, 0.4 μm pore size; Corning Costar) using an epithelial voltohmmeter (World Precision Instruments). Values were calculated after subtracting background readings from blank Transwells with the media that were cultured in parallel. Statistical analysis was by ANOVA.

Section preparation. For immunohistochemical analyses, sections (3 μm) of archival tissues were placed on OptiPlus Positive-Charged Barrier Slides (BioGenex), deparaffinized, treated with Retrievit-6 Target Retrieval Solution (BioGenex), and processed for immunostaining.

OCT-embedded fresh tumor tissues were used for preparation of frozen sections (5 μm). One frozen section was set aside for H&E staining, whereas the remaining sections were processed for immunofluorescence analyses as described below.

Microscopy, immunofluorescence, and imaging. Morphologies of A253 cells transfected with NS and S were examined using a Nikon Eclipse TE300 microscope. For indirect immunofluorescence analyses, transfected cells were grown to confluence, fixed in 3.7% paraformaldehyde, permeabilized with 0.1% Triton X-100, blocked with 10% goat serum, and incubated with primary antibodies to E-cadherin. Cells were incubated with FITC-tagged secondary antibodies, counterstained for F-actin with rhodamine-phalloidin where indicated, mounted in Vectashield, and analyzed with a Zeiss LSM510 META confocal microscope.

For indirect immunofluorescence analyses, tissue sections were blocked with 10% goat serum and incubated with antibodies against selected proteins followed by secondary antibodies conjugated with either FITC or Texas red. Negative controls lacked primary antibodies. The slides were mounted in Vectashield and optical sections (0.74 μm) were analyzed by confocal microscopy. To compare fluorescence intensities between samples, settings were fixed to the most highly stained sample with all other images acquired at those settings.

A253 cells overexpress DPAGT1 and produce E-cadherin modified with complex N-glycans in nascent adherens junctions. Tumor cells frequently exhibit decreased intercellular adhesion and aberrantly high N-glycosylation. Because DPAGT1 is a key regulator of cellular N-glycosylation, we first examined the relationship between DPAGT1 expression and E-cadherin adhesion in dense cultures of primary HOK, HOK cells, and human salivary epidermoid carcinoma, A253 cells. The inverse relationship between DPAGT1 and E-cadherin–mediated adherens junctions (Fig. 1,A, diagram) was supported by immunofluorescence staining of E-cadherin and Western blot assessment of DPAGT1 abundance. Whereas in HOK cells E-cadherin displayed organization at cell-cell borders, in A253 cells it had mostly cytoplasmic distribution (Fig. 1,A, IF, arrows). This correlated with 2-fold higher DPAGT1 expression and greater molecular size of E-cadherin in A253 cells compared with HOK cells (Fig. 1A,, WB). To determine if increased molecular size of E-cadherin in A253 cells was caused by increased N-glycosylation, we examined its sensitivity to glycanases, EndoH and PNGaseF. EndoH removes high mannose and hybrid N-glycans, whereas PNGaseF removes N-glycans at asparagine residues with the exception of complex N-glycans modified by fucose at the chitobiose core. Mobility shifts before and after EndoH and PNGaseF treatments showed that E-cadherin from HOK cells was EndoH-sensitive and thus contained primarily high mannose/hybrid N-glycans (Fig. 1,B, lanes 1-3). E-cadherin from A253 cultures was mostly PNGaseF-sensitive, showing that it was modified by complex N-glycans (Fig. 1 B, lanes 4-6). Similar to A253 cells, OSCC of the tongue, CAL27 cells, also overexpressed DPAGT1 and hyperglycosylated E-cadherin (data not shown).

E-cadherin immunoprecipitates from A253 cells bound less adherens junction stabilizing components, γ-catenin and vinculin, and contained four times less PP2A, shown to be essential for adherens junction adhesive function (ref. 36; Fig. 1C). In contrast, the abundance of IQGAP1, known to destabilize adherens junctions (37), was increased in A253 cells (Fig. 1C).

A253 cells have compromised tight junction function. In epithelial cells, formation of mature adherens junctions has been shown to precede the assembly of tight junctions (38, 39). Overexpression of DPAGT1 in A253 cells correlated with a 2.4-fold decrease in the abundance of ZO-1 (Fig. 1A,, WB), a scaffold protein that recruits occludin and claudins for tight junction assembly. Whereas in HOK cells ZO-1 was in a complex with substantial amounts of occludin and claudin-1, this interaction was significantly diminished in A253 cells (Fig. 1D). Because claudins determine the barrier function of tight junctions, the low abundance of ZO-1/claudin-1 complexes indicated that A253 cells did not assemble stable tight junctions. This conclusion was supported by a greater interaction between E-cadherin and ZO-1 in A253 cells, known to mark nascent contacts (40). Furthermore, ZO-1 immunoprecipitates from A253 cells had more PP2A, shown to inhibit tight junction assembly (Fig. 1D). This decreased association of ZO-1 with occludin and claudin-1 corresponded to reduced integrity of tight junctions, as determined by TER, a measure of paracellular permeability. After growth to high density for 72 h, A253 cells formed tight junctions that displayed a steady-state TER of 160 Ohm-cm², a value significantly lower than a typical TER of >260 Ohm-cm² in HOK cell monolayers (Supplementary Fig. S1B; ref. 41). Thus, changes in the molecular organization of adherens junctions and tight junctions corresponded to reduced A253 cell TER.

Partial inhibition of DPAGT1 in A253 cells reduces E-cadherin N-glycosylation, stabilizes adherens junctions, and enhances tight junction function. Previously, we showed that hypoglycosylated E-cadherin variants formed more mature adherens junctions than extensively N-glycosylated E-cadherin (19). Therefore, we tested the possibility of stabilizing adherens junctions in A253 cells by reducing cellular N-glycosylation using siRNA to DPAGT1. Treatment of A253 cells with 200 nmol/L siRNA (S) resulted in down-regulation of DPAGT1 transcript levels by 45% and protein levels by 40% (Fig. 2A). E-cadherin from S cells exhibited faster mobility on SDS-PAGE compared with NS cells (Fig. 2B,, E-cad), suggesting that it was less N-glycosylated. E-cadherin junctional complexes from S cells contained more catenins and vinculin but less IQGAP1 compared with NS cells (Fig. 2B). Immunofluorescence localization of E-cadherin showed more colocalization with F-actin in S cells compared with NS cells (Fig. 2 C, S, merge, arrow). We note that similar levels of biotinylated E-cadherin were detected in NS and S cells, indicating that diminished N-glycosylation did not affect the targeting of E-cadherin to the cell surface (data not shown).

Importantly, S-cell monolayers formed tight junctions that maintained TER of 210 Ohm-cm², or 1.3-fold higher than NS cells (Supplementary Fig. S2A, TER), showing that partial inhibition of DPAGT1 caused an increase in A253 cell TER. This treatment did not affect cell viability and did not induce an unfolded protein response (data not shown). Partial inhibition of DPAGT1 did, however, substantially reduce the number of disorganized cells that lacked contact inhibition (Supplementary Fig. S2B). This was associated with diminished proliferation as indicated by an increase in S-cell doubling time (Supplementary Fig. S2C). Thus, by reducing DPAGT1 expression, it is possible to diminish the N-glycosylation status of E-cadherin and to drive the formation of mature adherens junctions and tight junctions in A253 cells.

Overexpression of DPAGT1 and defective intercellular adhesion are signatures of OSCC. Because extensive N-glycosylation of E-cadherin in A253 cells prevented the formation of mature adherens junctions and tight junctions, we examined whether a similar scenario was found in oral cancer tissues. Using nine archival specimens, we compared cytologically normal adjacent epithelia with OSCC for markers of proliferation, differentiation, and N-glycosylation using Ki-67, cytokeratin, and DPAGT1, respectively. H&E staining of adjacent epithelia revealed characteristic maturation from the basal epithelial cell layer to the spinous cell layer with subtle surface parakeratosis (Fig. 3,A, H&E, filled and unfilled arrows). Well-differentiated and moderately differentiated OSCC, on the other hand, exhibited invasive nests and cords of atypical squamous epithelium (Fig. 3,A, H&E, unfilled arrow) in addition to dyskeratosis, cellular pleiomorphism, and atypical mitotic figures (data not shown). In contrast to adjacent epithelia, where Ki-67 was detected only in the basal cell layer, in OSCC it was distributed throughout the neoplasm (Fig. 3,A, Ki-67, arrow). Furthermore, cytokeratin was insignificant within the basal layer of adjacent epithelia but abundant in the spinous and keratin layers, whereas in OSCC it displayed diminished staining within the spinous epithelium, indicating disruption of cellular maturation (Fig. 3,A, cytokeratin, arrows). Prominent DPAGT1 staining was detected in the basal layer of adjacent epithelia but was diminished in the cytodifferentiated cells of the spinous cell layer (Fig. 4,A, AE, DPAGT1, filled and unfilled arrows). In contrast, OSCC exhibited prominent DPAGT1 expression that ceased to be restricted to the basal layer (Fig. 3,A, OSCC, DPAGT1, arrow). In some tumor islands, intense DPAGT1 staining in the peripheral cell layer was associated with extensive cellular discohesion, suggesting that aberrantly high DPAGT1 expression was driving oral tumor spread (Supplementary Fig. S3). Thus, DPAGT1 expression was directly related to Ki-67 but inversely to cytokeratin, showing that it was a useful marker for dysregulated proliferation and differentiation during oral tumorigenesis (Fig. 3 A, DPAGT1/cytokeratin, merge, arrows).

Although E-cadherin was present at cell-cell interfaces in OSCC, its pattern was frequently more diffuse compared with adjacent epithelia, suggesting that it formed nascent adherens junctions (Fig. 3,B, E-cadherin, arrows, insets). Double staining for E-cadherin and α-catenin revealed that altered immunostaining patterns of E-cadherin correlated with a virtual loss of α-catenin from cell-cell contact sites (Fig. 3,B, α-catenin, arrows, insets), whereas β-catenin remained unaltered (Fig. 3,B, β-catenin). Thus, examination of nine archival cases of OSCC showed that increased cellular proliferation and loss of differentiation were associated with overexpression of DPAGT1 and compromised E-cadherin adhesion (Table 1). Furthermore, immunofluorescence staining for ZO-1 showed that it was mislocalized in OSCC, indicating lack of tight junctions (Fig. 3 C, ZO-1, arrow).

OSCC produces extensively N-glycosylated E-cadherin, nascent adherens junctions, and unstable tight junctions. To validate the relationship between overexpression of DPAGT1 and loss of intercellular adhesion in OSCC, we carried out biochemical analyses of E-cadherin N-glycosylation status and the composition of adherens junctions and tight junctions in freshly resected tissues from three patients with moderately differentiated OSCC of the lateral tongue border. DPAGT1 abundance was 4-fold greater in OSCC compared with adjacent epithelia (Fig. 4A,, WB), with all invasive tumor islands exhibiting high DPAGT1 immunofluorescence staining (Fig. 4,A, IF, OSCC, DPAGT1, arrow). In contrast, DPAGT1 expression in adjacent epithelia was differentiation-dependent (Fig. 4 A, IF, AE, DPAGT1, filled and unfilled arrows).

E-cadherin from OSCC migrated with a higher molecular size (Fig. 4B) due to increased modification with complex N-glycans based on EndoH and PNGaseF treatments (4,C, lanes 1, 3, 4 and 6, filled and unfilled arrows). Moreover, E-cadherin immunoprecipitates from OSCC exhibited diminished association with α-catenin, γ-catenin, and vinculin but a 2-fold greater association with IQGAP1 compared with adjacent epithelia (Fig. 4C). Also, there was less PP2A in complex with E-cadherin in OSCC than in adjacent epithelia (Fig. 4C). These biochemical studies linked the overexpression of DPAGT1 to increased N-glycosylation of E-cadherin and compromised intercellular adhesion in OSCC. Similar to archival specimens, fresh OSCC tissues displayed more diffuse immunofluorescence localization of E-cadherin that correlated with a pronounced cytoplasmic localization of α-catenin, indicating diminished junctional organization (Fig. 4,C, IF, arrows, insets). This was coincident with increased organization of IQGAP1 at cell-cell borders, suggesting that, in OSCC, adherens junctions were destabilized by the replacement of α-catenin with IQGAP1 (Fig. 4,C, IF, arrows). In addition, more PP2A was found in complex with ZO-1 in OSCC than in adjacent epithelia (Fig. 4D). Hence, excessive N-glycosylation of E-cadherin in OSCC was associated with diminished binding of PP2A to adherens junctions and an increased association of PP2A with ZO-1. Accordingly, expression of ZO-1 was reduced by 60% (Fig. 4A,, WB) and greatly diminished at the sites of tight junctions (Fig. 4 D, IF, arrow).

Aberrant N-glycosylation has been associated with virtually all types of cancers (34, 42). Likewise, loss of intercellular adhesion has been a hallmark of tumor progression to metastasis. The present study reveals a mechanism for how inappropriate up-regulation of a metabolic pathway of protein N-glycosylation is deleterious to oral epithelial homeostasis and a likely cause of disrupted intercellular adhesion in OSCC.

Our results indicate that a subset of oral cancer cases overexpresses DPAGT1 and produces extensively N-glycosylated E-cadherin. OSCC of well to moderate histologic grade exhibited abundant DPAGT1 expression throughout the neoplasm. Biochemical analyses revealed that overexpression of DPAGT1 was associated with extensive modification of E-cadherin with complex N-glycans. Characterization of E-cadherin scaffolds confirmed that tumor tissues were impaired in the formation of mature adherens junctions, displaying diminished association with γ-catenin, α-catenin, and vinculin. The cadherin-dependent tumor-suppressive functions of γ-catenin and vinculin have been described before (43, 44). Likewise, the involvement of α-catenin in tumor suppression has been shown to be a prognostic indicator of metastasis and survival (45). Thus, our results support the important roles of γ-catenin, α-catenin, and vinculin in the maintenance of intercellular adhesion in oral epithelia and suggest that it is through their recruitment to adherens junctions that N-glycans regulate, in part, the tumor-suppressive function of E-cadherin. Interestingly, whereas the association of α-catenin with E-cadherin junctions was diminished in OSCC, the abundance of IQGAP1 was dramatically increased. Because IQGAP1 associates with E-cadherin complexes by competing with α-catenin for the binding to β-catenin, its increased recruitment to adherens junctions in OSCC may be responsible, in part, for the diminished α-catenin binding to E-cadherin complexes. Another protein whose association with adherens junctions was reduced in oral cancer cells and tissues was PP2A. PP2A has been shown to enhance adherens junction function and to inhibit tight junction assembly (46, 47). Therefore, in oral tumorigenesis, E-cadherin N-glycans appear to destabilize adherens junctions through their inhibitory effect on the recruitment of PP2A. Diminished association of PP2A with adherens junctions is likely to promote the interaction between PP2A to ZO-1 and other tight junction components, leading to tight junction inhibition. Taken together, these studies indicate that N-glycosylation can be a determinant of the tumor suppressor function of E-cadherin in vivo.

Although the precise mechanism of how N-glycans inhibit the ability of E-cadherin to form mature adherens junctions remains unclear, our studies provide the first molecular explanation for the observed effect of N-glycosylation on intercellular adhesion. It is possible that, through steric hindrance, N-glycans interfere with the dimerization of the extracellular domains of E-cadherin and homotypic bond formation. In this scenario, removal of N-glycans would facilitate homotypic interactions and lead to conformational changes transduced in cis to the cytoplasmic region to promote association with various components of adherens junctions.

We note that although partial knockdown of DPAGT1 in A253 cells was likely to affect numerous cell-surface receptors, this study focused on E-cadherin because its hypoglycosylated variant V13 has been shown to enhance cell-cell adhesion in a manner similar to siRNA knockdown (19).

Although the underlying causes for overexpression of DPAGT1 in A253 cells and OSCC are unknown, our studies show that it is possible to down-regulate DPAGT1 with siRNA to effectively reduce N-glycosylation capacity of cells and increase intercellular adhesion. A model summarizing these conclusions is shown in Fig. 5. It is possible that partial inhibition of DPAGT1 will prove to be a useful functional approach for enhancing intercellular adhesion not only in OSCC but also in other epithelial tumors.

No potential conflicts of interest were disclosed.

Grant support: NIH/NIDCR grants 5 RO1 DE010183 and RO1 DE015304 (M.A.Kukuruzinska) and NIH/EY grant R24 EY014798 (A.S. Menko).

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.