Regulation of Cytokine Receptors by Golgi N-Glycan Processing and Endocytosis

See allHide authors and affiliations

Science  01 Oct 2004:
Vol. 306, Issue 5693, pp. 120-124
DOI: 10.1126/science.1102109


The Golgi enzyme β1,6 N-acetylglucosaminyltransferase V (Mgat5) is up-regulated in carcinomas and promotes the substitution of N-glycan with poly N-acetyllactosamine, the preferred ligand for galectin-3 (Gal-3). Here, we report that expression of Mgat5 sensitized mouse cells to multiple cytokines. Gal-3 cross-linked Mgat5-modified N-glycans on epidermal growth factor and transforming growth factor–β receptors at the cell surface and delayed their removal by constitutive endocytosis. Mgat5 expression in mammary carcinoma was rate limiting for cytokine signaling and consequently for epithelial-mesenchymal transition, cell motility, and tumor metastasis. Mgat5 also promoted cytokine-mediated leukocyte signaling, phagocytosis, and extravasation in vivo. Thus, conditional regulation of N-glycan processing drives synchronous modification of cytokine receptors, which balances their surface retention against loss via endocytosis.

Co-translational modification of proteins in the endoplasmic reticulum by N-glycosylation facilitates their folding and is essential in single-cell eukaryotes. Metazoans have additional Golgi enzymes that trim and remodel the N-glycans, producing complex-type N-glycans on glycoproteins destined for the cell surface. Mammalian development requires complex-type N-glycans containing N-acetyllactosamine antennae, because their complete absence in Mgat1-deficient embryos is lethal (1, 2). Deficiencies in N-acetylglucosaminyltransferase II and V (Mgat2 and Mgat5) acting downstream of Mgat1 reduce the content of N-acetyllactosamine, and mutations in these loci result in viable mice with a number of tissue defects (3, 4). N-glycan processing generates ligands for various mammalian lectins, but the consequences of these interactions are poorly understood. The galectin family of N-acetyllactosamine-binding lectins has been implicated in cell growth, endothelial cell morphogenesis, angiogenesis (5), cell adhesion (6), and cancer metastasis (7). Galectins are expressed widely in metazoan tissues, located in the cytosol and nucleus, and secreted by a nonclassical pathway to the cell surface (8).

Galectin-3 (Gal-3) binds to poly-N-acetyllactosamine (i.e., a polymer of Galβ1,4GlcNAcβ1,3) with higher affinity than to the more ubiquitous N-acetyllactosamine antennae (9), and Mgat5 controls production of these larger polymers by producing the preferred intermediate for their addition (fig. S1A) (10). The nonlectin N-terminal domain of Gal-3 mediates pentamer formation in the presence of multivalent ligands, thereby cross-linking glycoproteins in proportion to ligand concentrations (11). The resulting superstructure of galectins and glycoproteins at the cell surface generates a molecular lattice. Mgat5-modified N-glycans on T cell receptors bind Gal-3, which opposes antigen-dependent clustering and suppresses autoimmune disease (12). Here, we examine the possibility that Mgat5-modified N-glycans on cytokine receptors oppose constitutive endocytosis by retaining surface receptors where membrane remodeling is active, notably in tumor cells and monocytes (model in fig. S1B).

The Mgat5-deficient background suppresses the oncogenic potency of a polyomavirus middle T oncogene (PyMT) transgene in mice (4). The PyMT protein is a cytosolic scaffold that promotes Src, phosphatidylinositol (PI) 3-kinase, and Shc/Ras activation (13), but activation of these intracellular pathways remains partially dependent on host cell responses to extracellular stimuli and to Mgat5-modified N-glycans (4). To explore this interaction, we isolated mammary epithelial tumor cells lines from PyMT transgenic mice on Mgat5–/– and Mgat5+/+backgrounds and compared their responsiveness to growth factors by measuring phosphorylation and nuclear translocation of extracellular signal–regulated kinase (Erk) (Fig. 1A and fig. S2). Mgat5–/– tumor cells were less responsive than Mgat5+/+ cells to epidermal growth factor (EGF), insulin-like growth factor (IGF), platelet-derived growth factor (PDGF), basic fibroblast growth factor (bFGF), and fetal calf serum, but infection of the cells with a retroviral vector encoding Mgat5 restored responsiveness (Fig. 1B and fig. S3). These cytokine receptors are all highly N-glycosylated with 8 to 16 N-glycosylation sites (N-X-S/T). Characterization of EGF receptors (EGFR) in carcinoma cells has revealed 10 to 12 occupied sites, and a subset of the N-glycans are Mgat5-modified and extended with poly-N-acetyllactosamine (14). The transforming growth factor–β (TGF-β) receptors TβRI and TβRII have only one and three N-X-S/T consensus sites, respectively. Mgat5–/– cells displayed a two- to threefold decrease in sensitivity to TGF-β compared with the ∼100-fold decrease in sensitivity to EGF, PDGF, IGF-1, and FGF, supporting the notion that both Golgi processing (i.e., Mgat5 and poly-N-acetyllactosamine) and the number of N-glycans per receptor are important (Fig. 1, B and C, and fig. S3).

Fig. 1.

Mgat5-modified N-glycans promote cytokine signaling. (A) Nuclear translocation of Erk-P 10 min after stimulation with cytokines (100 ng/ml) or 5% fetal calf serum, measured by scan array immunofluorescence imaging (Array-Scan, Cellomics, Incorporated, Pittsburgh, PA). (B) Erk-P nuclear translocation 10 min after stimulation with EGF. (C) Smad2 and 3 nuclear translocation 45 min after stimulation with TGF-β1.

Next we probed the cell surface with a chemical cross-linker, which revealed that EGFR was associated with Gal-3 on the surface of Mgat5+/+ cells whereas this interaction was greatly reduced on Mgat5–/– cells (Fig. 2A). Pretreatment of Mgat5+/+ tumor cells with lactose, a competitive inhibitor of Gal-3 binding, blocked EGFR–Gal-3 cross-linking, confirming that the carbohydrate-reactive domain of Gal-3 is required. Lactose pretreatment also depleted surface Gal-3 on Mgat5+/+ cells, producing a phenocopy of the Mgat5–/– cells. Furthermore, lactose and, with greater effect, N-acetyllactosamine dampened EGF-dependent activation of Erk in Mgat5+/+ cells (Fig. 2B). Thus, surface residency of both Gal-3 and glycoprotein receptors displays a dependency on Mgat5 modification of the receptor N-glycans.

Fig. 2.

Cell-surface EGFR binds to Gal-3 in an Mgat5-dependent manner. (A) Mgat5+/+ and Mgat5–/– cells were pretreated as indicated, then subjected to DTSSP (dithio-bis-suffosuccinimydyl propionate) cross-linking and biotinylation of surface proteins. Biotinylated proteins were captured on streptoavidin-agarose beads. Sucrose and untreated cells were identical, and data shown are representative of three independent experiments. (B) EGF stimulated Erk-P nuclear translocation after a 24-hour pretreatment with lactose, N-acetyl-lactosamine, or sucrose in Mgat5+/+ cells. (C) Scatchard plot of EGFR 125I-EGF binding to Mgat5+/+ and Mgat5–/– cells revealed 14,000 and 2500 binding sites per cell, with affinities of 2.7 and 2.5 nM, respectively. (D) Co-localization of EGFR with the endosome protein EEA-1 by immunofluorescence. (E) Erk-P nuclear translocation in cells pretreated with buffer (left) or nystatin plus K+ depletion (right) before addition of EGF (100 ng/ml). Mgat5–/– cells were infected with retrovirus vectors for expression of either Mgat5 (rescued) or a mutant form of Mgat5 (L188R). (Inset) A blot of cell lysates probed with L-PHA lectin to reveal Mgat5-modified N-glycans.

Receptor density at the cell surface is influenced by rates of de novo production, endocytosis, recycling, and degradation (15). Scatchard analysis revealed sixfold fewer EGFRs at the cell surface in mutant cells (Fig. 2C), although total EGFR amounts were not different between Mgat5+/+ and Mgat5–/– cell lysates (Fig. 2A). The affinity of EGF binding (2 to 3 nM) was the same in Mgat5+/+ and Mgat5–/– cells, consistent with previous observations that ligand affinity is not markedly altered by variations in N-glycan processing (16). Mgat5–/– cells displayed fourfold greater co-localization of EGFR with EEA-1, an early endosomal marker (Fig. 2D and fig. S6A). Pretreatment of Mgat5+/+ cells with lactose but not sucrose promoted receptor accumulation in the endosomes, mimicking the distribution observed in untreated Mgat5–/– cells and providing further evidence that receptors are anchored at the cell surface by the lattice. In the absence of the lattice, receptor amounts increase in the early endosomes, possibly because of reduced trafficking downstream of receptor tyrosine kinase signaling (17). Lastly, the Mgat5–/– signaling deficiency could be rescued by K+ depletion and nystatin treatment, agents that inhibit endocytosis (Fig. 2E and figs. S2C and S4). This chemical rescue of signaling was comparable to a genetic rescue using Mgat5, whereas a mutated Mgat5 [Lys188 → Arg188 (L188R)] that could not enter the Golgi failed to rescue signaling (Fig. 2E).

Ligand binding and autophosphorylation of EGFR stimulates clathrin-dependent EGFR endocytosis, where signaling is briefly amplified and then followed by transit of receptors to recycling or proteolytic compartments (18). In contrast, TβRII internalization is ligand-independent and occurs via both clathrin- and caveolae-dependent pathways (19). In spite of the differences between TαR and EGFR receptor dynamics, their regulation by the lattice was qualitatively similar. TβRII receptors associate with Gal-3 in an Mgat5-dependent manner, and the interaction could be blocked by lactose (fig. S6A). TGF-β signaling in mutant cells was restored by infection with the Mgat5 retroviral vector and by blocking endocytosis with the use of K+ depletion and nystatin (Fig. 3A and fig. S5B). Total TβRII amounts were similar in mutant and wild-type cells, but more were localized to endosomes and less were at the surface in mutant cells (figs. S5C and S6B). Binding of 125I-labeled TGF-β1 to surface TβRII was reduced 2.3-fold, whereas binding to TβRI was unchanged, an effect consistent with the number of N-X-S/T sites per receptor (i.e., three and one, respectively) (Fig. 3, B and C). We conclude that cytokine receptors are cleared from the surface of Mgat5–/– cells more rapidly but are delayed in the early endosomes compared to Mgat5++ cells.

Fig. 3.

EMT and the malignant phenotype are dependent on Mgat5. (A) Smad2 and 3 nuclear translocation in cells pretreated with buffer (left) or nystatin plus K+ depletion (right) before addition of TGF-β (5 ng/ml). Mgat5+/+, Mgat5–/–, and mutant cells were infected with Mgat5 retroviral vectors. (B) Autoradiograph showing 125I-TGF-β1 binding to cell-surface TβR. The identities of TβRI and TβRII were confirmed by immuno-precipitation. (C) 125I-TGF-β1 bound to TGF-β receptors was quantified by densitometry. (D) E-cadherin (green) in tight junctions of Mgat5+/+ and Mgat5–/– tumor cell monolayers revealed by immuno-fluorescence. Nuclear DNA is revealed by Hoechst staining. Cells plated at low density in second row are stained for actin microfilaments with rhodamine-phalloidin (red) and antibodies against vinculin (green). In the third row, contact inhibition is assessed by scratch-wounding confluent cell monolayers. The arrow indicates direction of cell movement, showing closure of the wound at 24 hours by Mgat5+/+ but not by Mgat5–/– cells. (E) Cell motility on fibronectin-coated wells measured by scan array microbead clearance over 18 hours and expressed as mean ± SE for 100 cell tracks. (F) Spontaneous lung metastases (mean ± SE, n = 6) in wild-type mice (*P < 0.01).

Epithelial-mesenchymal transition (EMT) in epithelial cancers is characterized by the loss of adhesion junctions, increased membrane remodeling, and metastasis (20). The Mgat5+/+ tumor cells displayed EMT with loss-of-adhesion junctions, a fibroblastic morphology with lamelipodia containing active signaling, and reduced contact inhibition in a scratch-wound assay (Fig. 3D). In contrast, Mgat5–/– tumor cells retained an epithelial morphology characterized by E-cadherin localization in cell adhesion junctions, cortical actin stress fibers, small focal adhesions, and strong contact inhibition of growth. EMT could be induced in Mgat5–/– by infecting the cells with a retroviral vector for expression of Mgat5. In vitro, wild-type and Mgat5-rescued mutant cells displayed greater cell motility (Fig. 3E), and in vivo these cells produced significantly greater numbers of lung tumor metastases, than Mgat5–/– cells (Fig. 3F). We conclude that Mgat5 is necessary for EMT and supplies positive feedback through cytokine receptors to Ras, PI3 kinase, and Smad2 and 3 signaling (20). The Ras-Raf-Ets pathway positively regulates Mgat5 transcription (21). TGF-β signaling also stimulates Mgat5 expression in mesenchymal cells (22), and we confirmed that TGF-β increased the surface expression of Mgat5-modified N-glycans (fig. S5E). TGF-β induces its own gene expression as well as that of other cytokines, and, cumulatively, these sources of positive feedback appear to maintain the EMT-invasive phenotype. Suppressors of constitutive endocytosis other than the lattice described here may promote tumor progression. In this regard, the metastasis suppressor protein Nm23-H1 is a nucleotide diphosphate kinase that supplies guanine triphosphate (GTP) to the small guanine triphosphatase (GTPase) dynamin required for membrane invagination (23).

These observations suggest that a transformation-associated increase in membrane remodeling and endocytosis may require the lattice to protect receptors at the surface. We found that nontransformed Chinese hamster ovary (CHO) cells and the lectin-resistant cell lines Lec1 and Lec8 all displayed similar responses to TGF-β, suggesting the lattice was not required (fig. S7A). The Lec1 and Lec8 cell lines are deficient in Mgat1 and Golgi uridine 5′diphosphate–Gal transporter activity, respectively, and consequently depleted for poly-N-acetyllactosamine and presumably the lattice. However, when these cell lines were transformed with polyomavirus large T (designated CHOP), signaling was significantly greater in the wild-type cells compared to the transformed Lec mutants (fig. S7B). To explore this with other cell lines, we treated transformed and non-transformed cells with swainsonine, a Golgi α-mannosidase II inhibitor that blocks processing upstream of Mgat5. Swainsonine reduced TGF-β responsiveness in B16F10 murine melanoma cells and SW620 human colorectal cancer cells but not in nontransformed murine epithelial NMuMG cells (fig. S7, C to F).

The motile and highly endocytic phenotype of activated macrophages is similar in this regard to tumor cells and may require the lattice to retain surface cytokine receptors. We examined this possibility by using lipopoly-saccharide (LPS)-elicited peritoneal macrophages and determined that endogenous phospho-Smad2 and -Erk2 and 3, as well as acute responses to TGF-β and serum, were suppressed in Mgat5–/– cells (Fig. 4, A and B). Furthermore, binding of 125I-labeled TGF-β1 to surface receptors was reduced in Mgat5–/– macrophages, characteristic of the Mgat5-deficient phenotype observed in tumor cells (fig. S5F). Early in the injury response, cytokines stimulate leukocyte migration, and subsequently TGF-β attains amounts that suppress the inflammatory process (24). Consistent with this chronology, we observed that skin inflammation induced by either arachidonic acid or phorbol ester was delayed in both its onset and its resolution phases in Mgat5–/– mice (Fig. 4C). Similarly, the initial rate of leukocyte extravasation into the peritoneal cavity in response to an injection of thioglycollate was impaired in Mgat5–/– mice (Fig. 4D), demonstrating that Mgat5 regulates leukocytes motility. Phagocytosis of latex beads by Mgat5–/– macrophages was reduced. The PI3 kinase inhibitor wortmannin reduced phagocytosis in wild-type macrophages but did not suppress further in Mgat5–/– cells (Fig. 4E). Thus, the lattice promotes responsiveness to extracellular cytokines in this example of a nontransformed but endocytic cell type. Mgat5 modification of N-glycans on integrins and other adhesion receptors may also influence membrane remodeling and extracellular matrix assembly (25).

Fig. 4.

Macrophage signaling, migration, and phagocytosis are dependent on Mgat5. (A) Smad2 and 3 and (B) Erk-P nuclear translocation in LPS-elicited peritoneal macrophages from Mgat5+/+ and Mgat5–/– mice stimulated TGF-β1 (2 ng/ml) and 5% FCS, respectively. (C) Ear swelling induced by topical application of arachidonic acid in Mgat5–/– (◯)) and Mgat5+/+ (+) mice (*P < 0.001). (D) Leukocytes recruitment into the peritoneal cavity 3 hours after an injection of thioglycolate (*P G 0.001). (E) Phagocytosis of five or more fluorescent latex beads by thioglycolate-elicited macrophages, either untreated or treated with wortmannin (100 nM) for 1 hour ex vivo during bead phagocytosis (*P < 0.01). Values are mean ± SE for five mice per genotype.

Lattice-dependent regulation of receptors occurs primarily at the cell surface, is dependent on Golgi enzyme activities and the number of N-glycans per receptor, and opposes receptor loss to endocytosis (fig. S1B). Receptor tyrosine kinases activate Rab5, a small GTPase required for endocytosis (17), thereby promoting receptor loss, whereas our results show that positive feedback from signaling to the Golgi pathway strengthens the lattice, thereby maintaining cytokine responsiveness. Lastly, we speculate that genetic and environmental factors that regulate the integrity of the lattice should influence the decline in surface cytokine receptors known to occur with aging (26, 27) and thereby progenitor cell and tissue renewal.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Table S1

References and Notes

View Abstract

Navigate This Article