Glycosylation regulates Notch signalling

Intracellular post-translational modifications such as phosphorylation and ubiquitylation have been well studied for their roles in regulating diverse signalling pathways, but we are only just beginning to understand how differential glycosylation is used to regulate intercellular signalling. Recent studies make clear that extracellular post-translational modifications, in the form of glycosylation, are essential for the Notch signalling pathway, and that differences in the extent of glycosylation are a significant mechanism by which this pathway is regulated.     

The role of palmitoylation in signalling, cellular trafficking and plasma membrane localization of protease-activated receptor-2

Protease-activated receptor-2 (PAR2) is a G protein coupled receptor (GPCR) activated by proteolytic cleavage of its amino terminal domain by trypsin-like serine proteases. This irreversible activation mechanism leads to rapid receptor desensitization by internalisation and degradation. We have explored the role of palmitoylation, the post-translational addition of palmitate, in PAR2 signalling, trafficking, cell surface expression and desensitization. Experiments using the palmitoylation inhibitor 2-bromopalmitate indicated that palmitate addition is important in trafficking of PAR2 endogenously expressed by prostate cancer cell lines. This was supported by palmitate labelling using two approaches, which showed that PAR2 stably expressed by CHO-K1 cells is palmitoylated and that palmitoylation occurs on cysteine 361. Palmitoylation is required for optimal PAR2 signalling as Ca²⁺ flux assays indicated that in response to trypsin agonism, palmitoylation deficient PAR2 is ∼9 fold less potent than wildtype receptor with a reduction of about 33% in the maximum signal induced via the mutant receptor. Confocal microscopy, flow cytometry and cell surface biotinylation analyses demonstrated that palmitoylation is required for efficient cell surface expression of PAR2. We also show that receptor palmitoylation occurs within the Golgi apparatus and is required for efficient agonist-induced rab11a-mediated trafficking of PAR2 to the cell surface. Palmitoylation is also required for receptor desensitization, as agonist-induced β-arrestin recruitment and receptor endocytosis and degradation were markedly reduced in CHO-PAR2-C361A cells compared with CHO-PAR2 cells. These data provide new insights on the life cycle of PAR2 and demonstrate that palmitoylation is critical for efficient signalling, trafficking, cell surface localization and degradation of this receptor.     

Wnt-11 and Fz7 reduce cell adhesion in convergent extension by sequestration of PAPC and C-cadherin

Wnt-11/planar cell polarity signaling polarizes mesodermal cells undergoing convergent extension during Xenopus laevis gastrulation. These shape changes associated with lateral intercalation behavior require a dynamic modulation of cell adhesion. In this paper, we report that Wnt-11/frizzled-7 (Fz7) controls cell adhesion by forming separate adhesion-modulating complexes (AMCs) with the paraxial protocadherin (PAPC; denoted as AMCP) and C-cadherin (denoted as AMCC) via distinct Fz7 interaction domains. When PAPC was part of a Wnt-11-Fz7 complex, its Dynamin1- and clathrin-dependent internalization was blocked. This membrane stabilization of AMCP (Fz7/PAPC) by Wnt-11 prevented C-cadherin clustering, resulting in reduced cell adhesion and modified cell sorting activity. Importantly, Wnt-11 did not influence C-cadherin internalization; instead, it promoted the formation of AMCC (Fz7/Cadherin), which competed with cis-dimerization of C-cadherin. Because PAPC and C-cadherin did not directly interact and did not form a joint complex with Fz7, we suggest that Wnt-11 triggers the formation of two distinct complexes, AMCC and AMCP, that act in parallel to reduce cell adhesion by hampering lateral clustering of C-cadherin.     

Glycosylation of human CRLR at Asn123 is required for ligand binding and signaling

Calcitonin receptor-like receptor (CRLR) constitutes either a CGRP receptor when complexed with receptor activity-modifying protein 1 (RAMP1) or an adrenomedullin receptor when complexed with RAMP2 or RAMP3. RAMP proteins modify the glycosylation status of CRLR and determine their receptor specificity; when treated with tunicamycin, a glycosylation inhibitor, CHO-K1 cells constitutively expressing both RAMP2 and CRLR lost the capacity to bind adrenomedullin. Similarly, in HEK293 EBNA cells constitutively expressing RAMP1/CRLR receptor complex CGRP binding was remarkably inhibited. Whichever RAMP protein was co-expressing with CRLR, the ligand binding was sensitive to tunicamycin. There are three putative Asn-linked glycosylation sites in the extracellular, amino terminal domain of CRLR at positions 66, 118 and 123. Analysis of CRLR mutants in which Gln was substituted for selected Asn residues showed that glycosylation of Asn123 is required for both the binding of adrenomedullin and the transduction of its signal. Substituting Asn66 or Asn118 had no effect. FACS analysis of cells expressing FLAG-tagged CRLRs showed that disrupting Asn-linked glycosylation severely affected the transport of the CRLR protein to the cell surface on N66/118/123Q mutant, and slightly reduced the level of the cell surface expression of N123Q mutant compared with wild-type CRLR. But other single mutants (N66Q, N118Q) had no effect for other single mutants. Our data shows that glycosylation of Asn66 and Asn118 is not essential for ligand binding, signal transduction and cell surface expression, and Asn123 is important for ligand binding and signal transduction rather than cell surface expression. It thus appears that glycosylation of Asn123 is required for CRLR to assume the appropriate conformation on the cell surface through its interaction with RAMPs.     

Palmitoylation of Hedgehog proteins by Hedgehog acyltransferase: roles in signalling and disease

Hedgehog acyltransferase (Hhat), a member of the membrane-bound O-acyltransferase (MBOAT) family, catalyses the covalent attachment of palmitate to the N-terminus of Hedgehog proteins. Palmitoylation is a post-translational modification essential for Hedgehog signalling. This review explores the mechanisms involved in Hhat acyltransferase enzymatic activity, similarities and differences between Hhat and other MBOAT enzymes, and the role of palmitoylation in Hedgehog signalling. In vitro and cell-based assays for Hhat activity have been developed, and residues within Hhat and Hedgehog essential for palmitoylation have been identified. In cells, Hhat promotes the transfer of palmitoyl-CoA from the cytoplasmic to the luminal side of the endoplasmic reticulum membrane, where Shh palmitoylation occurs. Palmitoylation is required for efficient delivery of secreted Hedgehog to its receptor Patched1, as well as for the deactivation of Patched1, which initiates the downstream Hedgehog signalling pathway. While Hhat loss is lethal during embryogenesis, mutations in Hhat have been linked to disease states or abnormalities in mice and humans. In adults, aberrant re-expression of Hedgehog ligands promotes tumorigenesis in an Hhat-dependent manner in a variety of different cancers, including pancreatic, breast and lung. Targeting hedgehog palmitoylation by inhibition of Hhat is thus a promising, potential intervention in human disease.     

Asparagine-linked glycosylation of human chymotrypsin C is required for folding and secretion but not for enzyme activity

Human chymotrypsin C (CTRC) plays a protective role in the pancreas by mitigating premature trypsinogen activation through degradation. Mutations that abolish activity or secretion of CTRC increase the risk for chronic pancreatitis. The aim of the present study was to determine whether human CTRC undergoes asparagine-linked (N-linked) glycosylation and to examine the role of this modification in CTRC folding and function. We abolished potential sites of N-linked glycosylation (Asn-Xaa-Ser/Thr) in human CTRC by mutating the Asn residues to Ser individually or in combination, expressed the CTRC mutants in HEK 293T cells and determined their glycosylation state using PNGase F and endo H digestion. We found that human CTRC contains a single N-linked glycan on Asn52. Elimination of N-glycosylation by mutation of Asn52 (N52S) reduced CTRC secretion about 10-fold from HEK 293T cells but had no effect on CTRC activity or inhibitor binding. Overexpression of the N52S CTRC mutant elicited endoplasmic reticulum stress in AR42J acinar cells, indicating that N-glycosylation is required for folding of human CTRC. Despite its important role, Asn52 is poorly conserved in other mammalian CTRC orthologs, including the rat which is monoglycosylated on Asn90. Introduction of the Asn90 site in a non-glycosylated human CTRC mutant restored full glycosylation but only partially rescued the secretion defect. We conclude that N-linked glycosylation of human CTRC is required for efficient folding and secretion; however, the N-linked glycan is unimportant for enzyme activity or inhibitor binding. The position of the N-linked glycan is critical for optimal folding, and it may vary among the otherwise highly homologous mammalian CTRC sequences.     

Glycosylation sites in the atrial natriuretic peptide receptor: oligosaccharide structures are not required for hormone binding.

Atrial natriuretic peptide (ANP) is a hormone involved in cardiovascular homeostasis through its natriuretic and vasodilator actions. The ANP receptor that mediates these actions is a glycosylated transmembrane protein coupled to guanylate cyclase. The role of glycosylation in receptor signaling remains unresolved. In this study, we determined, by a combination of HPLC/MS and Edman sequencing, the glycosylation sites in the extracellular domain of ANP receptor (NPR-ECD) from rat expressed in COS-1 cells. HPLC/MS analysis of a tryptic digest of NPR-ECD identified five glycosylated peptide fragments, which were then sequenced by Edman degradation to determine the glycosylation sites. The data revealed Asn-linked glycosylation at five of six potential sites. The type of oligosaccharide structure attached at each site was deduced from the observed masses of the glycosylated peptides as follows: Asn13 (high-mannose), Asn180 (complex), Asn306 (complex), Asn347 (complex), and Asn395 (high-mannose and hybrid types). Glycosylation at Asn180 and Asn347 was partial. The role of glycosyl moieties in ANP binding was examined by enzymatic deglycosylation of NPR-ECD followed by binding assay. NPR-ECD deglycosylated with endoglycosidase F2 and endoglycosidase H retained ANP-binding activity and showed an affinity for ANP similar to that of untreated NPR-ECD. Endoglycosidase treatment of the full-length ANP receptor expressed in COS-1 cells also had no detectable effect on ANP binding. These results suggest that, although glycosylation may be required for folding and transport of the newly synthesized ANP receptor to the cell surface, the oligosaccharide moieties themselves are not involved in hormone binding.     

Mutational analysis of mouse Wnt-1 identifies two temperature-sensitive alleles and attributes of Wnt-1 protein essential for transformation of a mammary cell line.

The proto-oncogene Wnt-1 encodes a cysteine-rich, secretory glycoprotein implicated in virus-induced mouse mammary cancer and intercellular signaling during vertebrate neural development. To attempt to correlate structural motifs of Wnt-1 protein with its function, 12 mutations were introduced singly and in several combinations into the coding sequence of Wnt-1 cDNA by site-directed mutagenesis. Mutant alleles in a retroviral vector were tested for their ability to transform the mouse mammary epithelial cell line C57MG in two ways: by direct infection of C57MG cells and by infection of NIH3T3 cells that serve as donors of Wnt-1 protein to adjacent C57MG cells in a secretion-dependent (paracrine) assay. In addition, the synthesis and secretion of mutant proteins were monitored in multiple cell types by immunological assays. Deletion of the signal peptide demonstrated that transformation in both direct and paracrine assays depends upon entry of Wnt-1 protein into the endoplasmic reticulum. Changes in potential proteolytic processing sites (two basic dipeptides and a probable signal peptidase cleavage site) did not adversely impair biological activity or protein processing and uncovered a second site for cleavage by signal peptidase. Replacement of each of the four asparagine-linked glycosylation sites did not affect transforming activity at normal temperatures, but one glycosylation site mutant was found to be temperature-sensitive for transformation. An allele encoding a protein that lacks all four glycosylation sites was also transformation competent. In two of four cases, substitution of serine for a cysteine residue impaired transforming activity at the usual temperature, and transformation was temperature sensitive in a third case, implying that at least some of the highly conserved cysteine residues are important for Wnt-1 function.     

Post-translational processing events in the secretion pathway of human protein C, a complex vitamin K-dependent antithrombotic factor.

Human protein C (HPC) undergoes several post-translational modifications, including gamma-carboxylation, N-linked glycosylation, and internal proteolytic processing. We have utilized a recombinant human kidney cell line (293) secreting correctly modified HPC (rHPC) to study the processing reactions for the modification of this complex protein. gamma-Carboxylation was shown to proceed via a vitamin K-dependent pathway and was required for both efficient secretion and anticoagulant activity. rHPC was rapidly secreted following the addition of vitamin K to depleted cells, and secretion was not inhibited by cyclohexamide indicating that non gamma-carboxylated rHPC accumulates as an intracellular releasable pool. However, in cells grown in the presence of vitamin K, the majority of intracellular rHPC was gamma-carboxylated, suggesting that this post-translational modification is not rate limiting for secretion under conditions optimal for vitamin K-dependent carboxylation. Nonglycosylated rHPC was found to be secreted inefficiently, and processing of the N-linked core in the endoplasmic reticulum, but not in the Golgi, was required for secretion. Further, the intracellular rHPC present in vitamin K-supplemented cells was core glycosylated, but not processed past the high mannose step. gamma-Carboxylation occurred after core glycosylation, indicating that this modification is not cotranslational. Further, glycosylation and gamma-carboxylation were not coupled and did not need to proceed sequentially. Proteolytic processing of the internal KR dipeptide was found to occur late in the secretion pathway, and the cleavage was calcium-dependent. The secretion rate of rHPC was also calcium-dependent but was independent of the calcium effect on internal KR dipeptide removal, indicating that cleavage is not required for efficient secretion. Our results define the sequence of processing events, the subcellular localization of the processing reactions, and the rate-limiting steps in the secretion pathway for this complex protein.     

Secretion of heparanase protein is regulated by glycosylation in human tumor cell lines

The endo-beta-d-glucuronidase, heparanase, is capable of specifically degrading heparan sulfate, and this activity is associated with the metastatic potential of tumor cells. The predicted amino acid sequence of heparanase includes six putative N-glycosylation sites; however, the precise biochemical role of glycosylated heparanase remains unknown. In this study, we examined the link between glycosylation and the function of heparanase in human tumor cell lines. Heparanase protein was glycosylated at six Asn residues in human tumor cell lines. Treatment with a glycosylation inhibitor demonstrated that glycosylation was not required for the activity of heparanase. However, glycosylation affected the kinetics of endoplasmic reticulum-to-Golgi transport and of secretion of the enzyme.     

N-glycosylation of the human melanocortin 1 receptor: occupancy of glycosylation sequons and functional role

The melanocortin 1 receptor (MC1R), a major determinant of skin pigmentation and phototype, mediates the actions of α-melanocyte-stimulating hormone on melanocytes and is critical for melanocyte proliferation and differentiation. MC1R has two putative N-glycosylation targets, Asn15 and Asn29. It has been shown that MC1R is a glycoprotein with an unusual sensitivity to endoglycosidase H digestion. However, the occupancy and functional importance of each specific glycosylation sequon remains unknown. We demonstrate that MC1R is N-glycosylated at Asn15 and Asn29, with structurally and functionally different glycan chains. N-glycosylation is not necessary for high affinity agonist binding or functional coupling but has a strong effect on the availability of MC1R molecules on the plasma membrane, most likely by a combination of improved forward trafficking and decreased internalization. Finally, we found that MC1R variants exhibit different degrees of glycosylation which do not show a simple correlation with their functional status or intracellular trafficking.     

N-linked glycosylation of G. mellonella juvenile hormone binding protein - comparison of recombinant mutants expressed in P. pastoris cells with native protein

Juvenile hormone (JH) regulates insect growth and development. JH present in the hemolymph is bound to juvenile hormone binding protein (hJHBP) which protects JH from degradation. In G. mellonella, this protein is glycosylated only at one (Asn(94)) of the two potential N-linked glycosylation sites (Asn(4) and Asn(94)). To investigate the function of glycosylation, each of the two potential glycosylation sites in the rJHBP molecule was examined by site-directed mutagenesis. MS analysis revealed that rJHBP overexpressed in the P. pastoris system may appear in a non-glycosylated as well as in a glycosylated form at both sites. We found that mutation at position Asn(94) reduces the level of protein secretion whereas mutation at the Asn(4) site has no effect on protein secretion. Purified rJHBP and its mutated forms (N4W and N94A) have the same JH binding activities similar to that of hJHBP. However, both mutants devoid of the carbohydrate chain are more susceptible to thermal inactivation. It is concluded that glycosylation of JHBP molecule is important for its thermal stability and secretion although it is not required for JH binding activity.     

N-linked glycosylation sites affect secretion of cryptococcal phospholipase B1, irrespective of glycosylphosphatidylinositol anchoring

Secreted phospholipase B enzymes (PLB1) with high levels of N-linked glycosylation are proven fungal virulence determinants. We demonstrated that removal of N-linked glycans from secreted cryptococcal PLB1 leads to loss of enzyme activity. To determine if individual N-glycan attachment sites affect secretion of active enzyme, we altered three along the entire length of the protein, by site-directed mutagenesis, namely Asn56, Asn430 and Asn550 to Ala, in wild-type PLB1 (full length) and a glycosylphosphatidylinositol (GPI) anchorless version (PLB1(GPI-)) that is hypersecreted due to lack of membrane association. Alteration of Asn56 and Asn550 in both PLB1 and PLB1(GPI-) abolished enzyme secretion while alteration of Asn430 reduced secretion by 60%, following expression in Saccharomyces cerevisiae. Reduced secretion coincided with reduced enzyme in membranes and cell walls confirming a reduction in the rate of PLB1 transport to the cell surface. Deglycosylation of cryptococcal PLB1 increased its susceptibility to proteolysis suggesting that the absence of full glycosylation status leads to degradation of unstable PLB1, resulting in reduced traffic through the secretory pathway. We conclude that individual N-linked glycans are required for optimal transport of PLB1 to the cell surface and optimal secretion of both PLB1 and PLB1(GPI-).     

Analysis of the role of oligosaccharides in the apoptotic activity of glycodelin A

Glycodelin A, also known as placental protein-14, is a multifunctional glycosylated protein secreted by the uterine endometrium during the early phases of pregnancy. It is a known suppressor of T cell proliferation, inducer of T cell apoptosis, and inhibitor of sperm zona binding. Unlike in contraceptive activity, where the glycans on the molecule have been shown to play a crucial role, mutagenesis of the asparagines at sites of N-linked glycosylation (Asn(28) and Asn(63)) to glutamine shows that the apoptogenic activity of glycodelin A is executed by the protein backbone. Glycosylation at Asn(28) appears to play a role in the extracellular secretion of the molecule, as mutation of Asn(28) resulted in a significant decrease in the amount of secreted protein, and loss of both glycosylation sites reduced the secretion drastically. Our results also suggest that the loss of glycosylation does not affect the dimerization status of the molecule.     

Purified Wnt-5a increases differentiation of midbrain dopaminergic cells and dishevelled phosphorylation

The Wnt family of lipoproteins regulates several aspects of the development of the nervous system. Recently, we reported that Wnt-3a enhances the proliferation of midbrain dopaminergic precursors and that Wnt-5a promotes their differentiation into dopaminergic neurones. Here we report the purification of hemagglutinin-tagged Wnt-5a using a three-step purification method similar to that previously described for Wnt-3a. Haemagglutinin-tagged Wnt-5a was biologically active and induced the differentiation of immature primary midbrain precursors into tyrosine hydroxylase-positive dopaminergic neurones. Using a substantia nigra-derived dopaminergic cell line (SN4741), we found that Wnt-5a, unlike Wnt-3a, did not promote beta-catenin phosphorylation or stabilization. However, both Wnt-5a and Wnt-3a activated dishevelled, as assessed by a phosphorylation-dependent mobility shift. Moreover, the activity of Wnt-5a on dishevelled was blocked by pre-treatment with acyl protein thioesterase-1, indicating that palmitoylation of Wnt-5a is necessary for its function. Thus, our results suggest that Wnt-3a and Wnt-5a, respectively, activate canonical and non-canonical Wnt signalling pathways in ventral midbrain dopaminergic cells. Furthermore, we identify dishevelled as a key player in transducing both Wnt canonical and non-canonical signals in dopaminergic cells.     

Glycosylation of human apolipoprotein E. The carbohydrate attachment site is threonine 194

The glycosylation of human apolipoprotein (apo) E was examined with purified plasma apoE and apoE produced by transfected cell lines. The carbohydrate attachment site of plasma apoE was localized to a single tryptic peptide (residues 192-206). Sequence analysis and amino sugar analysis of this peptide derived from asialo-, monosialo-, or disialo-apoE indicated that the carbohydrate moiety is attached only to Thr194 in monosialo- and disialo-apoE and that asialo-apoE is not glycosylated. Mammalian cells that normally do not express apoE were transfected with human apoE plasmid expression vectors to test the utilization of potential carbohydrate attachment sites and the role of apoE glycosylation in secretion. Site-specific mutants of apoE, designed to eliminate or alter glycosylation sites, were expressed in HeLa cells by acute transfection. Apolipoprotein E(Thr194----Ala) was secreted exclusively as the asialo isoform, confirming that Thr194 is the site of carbohydrate attachment in these cells and indicating that glycosylation of apoE is not essential for secretion. Apolipoprotein E(Thr194----Asn,Gly196----Ser), which introduces a potential site for N-glycosylation at position 194, was secreted with a higher apparent molecular weight than native, O-glycosylated apoE. Studies with tunicamycin indicated that this apoE was N-glycosylated at Asn194. Stably transfected cell lines expressing human apoE were prepared from wild-type Chinese hamster ovary (CHO) cells and from CHO ldlD cells, which are defective in glycosylation. The transfected wild-type cells secreted multiply sialylated apoE. The transfected ldlD cells also secreted high levels of apoE even in the absence of glycosylation, which confirms that glycosylation is not essential for secretion of apoE.     

The role of N-linked glycosylation in determining the surface expression,G protein interaction and effector coupling of the alpha (alpha) isoform of the human thromboxane A(2) receptor

In humans, thromboxane (TX) A(2) signals through two TXA(2) receptor (TP) isoforms, termed TPalpha and TPbeta, that diverge exclusively within the carboxyl terminal cytoplasmic domains. The amino terminal extracellular region of the TPs contains two highly conserved Asn (N)-linked glycosylation sites at Asn(4) and Asn(16). While it has been established that impairment of N-glycosylation of TPalpha significantly affects ligand binding/intracellular signalling, previous studies did not ascertain whether N-linked glycosylation was critical for ligand binding per se or whether it was required for the intracellular trafficking and the functional expression of TPalpha on the plasma membrane (PM). In the current study, we investigated the role of N-linked glycosylation in determining the functional expression of TPalpha, by assessment of its ligand binding, G protein coupling and intracellular signalling properties, correlating it with the level of antigenic TPalpha protein expressed on the PM and/or retained intracellularly. From our data, we conclude that N-glycosylation of either Asn(4) or Asn(16) is required and sufficient for expression of functionally active TPalpha on the PM while the fully non-glycosylated TPalpha(N4,N16-Q4,Q16) is almost completely retained within the endoplasmic reticulum (ER) and remains functionally inactive, failing to associate with its coupling G protein Galpha(q) and, in turn, failing to mediate phospholipase (PL) Cbeta activation.     

N-linked glycosylation regulates human proteinase-activated receptor-1 cell surface expression and disarming via neutrophil proteinases and thermolysin

Proteinase-activated receptor 1 (PAR(1)) induces activation of platelet and vascular cells after proteolytic cleavage of its extracellular N terminus by thrombin. In pathological situations, other proteinases may be generated in the circulation and might modify the responses of PAR(1) by cleaving extracellular domains. In this study, epitope-tagged wild-type human PAR(1) (hPAR(1)) and a panel of N-linked glycosylation-deficient mutant receptors were permanently expressed in epithelial cells (Kirsten murine sarcoma virus-transformed rat kidney cells and CHO cells). We have analyzed the role of N-linked glycosylation in regulating proteinase activation/disarming and cell global expression of hPAR(1). We reported for the first time that glycosylation in the N terminus of hPAR(1) downstream of the tethered ligand (especially Asn(75)) governs receptor disarming to trypsin, thermolysin, and the neutrophil proteinases elastase and proteinase 3 but not cathepsin G. In addition, hPAR(1) is heavily N-linked glycosylated and sialylated in epithelial cell lines, and glycosylation occurs at all five consensus sites, namely, Asn(35), Asn(62), Asn(75), Asn(250), and Asn(259). Removing these N-linked glycosylation sequons affected hPAR(1) cell surface expression to varying degrees, and N-linked glycosylation at extracellular loop 2 (especially Asn(250)) of hPAR(1) is essential for optimal receptor cell surface expression and receptor stability.     

Structural requirements for dimerization,glycosylation, secretion,and biological function of VPF/VEGF

Vascular permeability factor (VPF) also known as vascular endothelial growth factor (VEGF), is a dimeric protein that affects endothelial cell (EC) and vascular functions including enhancement of microvascular permeability and stimulation of EC growth. To investigate the structural features of VPF/VEGF necessary for efficient dimerization, secretion, and biological activities, we employed site-directed mutagenesis with a Cos-1 cell expression system. Several cysteine residues essential for VPF dimerization were identified by mutation analysis of the Cys-25, Cys-56, and Cys-67 residues. Mutant VPF isoforms lacking either of these cysteines were secreted as monomers and were completely inactive in both vascular permeability and endothelial cell mitotic assays. VPF Cys-145 mutant protein was efficiently secreted as a glycosyaated, dimeric polypeptide, but had a reduction in biological activities. The site of N-linked glycosylation was directly identified as Asn-74, which, when mutated produced an inefficiently secreted dimeric protein without post-translational glycosylation, yet maintained full vascular permeability activity. Finally, we found that one VPF mutant isoform Cys-101 was not secreted and this mutant functioned as a dominant-negative suppressor of wild-type VPF secretion as demonstrated by co-expression assays in Cos-1 cells.