Glycosylation Profile of a Recombinant Urokinase-type Plasminogen Activator Receptor Expressed in Chinese Hamster Ovary Cells*

Association of urokinase-type plasminogen activator (uPA) to cells via binding to its specific cellular receptor (uPAR) augments the potential of these cells to support plasminogen activation, a process that has been implicated in the degradation of extracellular matrix proteins during cell migration and tissue remodeling. The uPA receptor is a glycolipid-anchored membrane protein belonging to the Ly-6/uPAR superfamily and is the only multidomain member identified so far. We have now purified the three individual domains of a recombinant soluble uPAR variant, expressed in Chinese hamster ovary cells, after limited proteolysis using chymotrypsin and pepsin. The glycosylation patterns of these domains have been determined by matrix assisted laser desorption ionization and electrospray ionization mass spectrometry. Of the five potential attachment sites for asparagine-linked carbohydrate in uPAR only four are utilized, as the tryptic peptide derived from domain III containing Asn233 was quantitatively recovered without carbohydrate. The remaining four attachment sites were shown to exhibit site-specific microheterogeneity of the asparagine-linked carbohydrate. The glycosylation on Asn52 (domain I) and Asn172 (domain II) is dominated by the smaller biantennary complex-type oligosaccharides, while Asn162 (domain II) and Asn200 (domain III) predominantly carry tri- and tetraantennary complex-type oligosaccharides. The carbohydrate moiety on Asn52 in uPAR domain I could be selectively removed byN-glycanase treatment under nondenaturing conditions. This susceptibility was abrogated when uPAR participitated in a bimolecular complex with pro-uPA or smaller receptor binding derivatives thereof, demonstrating the proximity of the ligand-binding site to this particular carbohydrate moiety. uPAR preparations devoid of carbohydrate on domain I exhibited altered binding kinetics toward uPA (a 4–6-fold increase in K d ) as assessed by real time biomolecular interaction analysis.

Association of urokinase-type plasminogen activator (uPA) to cells via binding to its specific cellular receptor (uPAR) augments the potential of these cells to support plasminogen activation, a process that has been implicated in the degradation of extracellular matrix proteins during cell migration and tissue remodeling. The uPA receptor is a glycolipid-anchored membrane protein belonging to the Ly-6/uPAR superfamily and is the only multidomain member identified so far. We have now purified the three individual domains of a recombinant soluble uPAR variant, expressed in Chinese hamster ovary cells, after limited proteolysis using chymotrypsin and pepsin. The glycosylation patterns of these domains have been determined by matrix assisted laser desorption ionization and electrospray ionization mass spectrometry. Of the five potential attachment sites for asparagine-linked carbohydrate in uPAR only four are utilized, as the tryptic peptide derived from domain III containing Asn 233 was quantitatively recovered without carbohydrate. The remaining four attachment sites were shown to exhibit site-specific microheterogeneity of the asparagine-linked carbohydrate. The glycosylation on Asn 52 (domain I) and Asn 172 (domain II) is dominated by the smaller biantennary complex-type oligosaccharides, while Asn 162 (domain II) and Asn 200 (domain III) predominantly carry tri-and tetraantennary complex-type oligosaccharides. The carbohydrate moiety on Asn 52 in uPAR domain I could be selectively removed by N-glycanase treatment under nondenaturing conditions. This susceptibility was abrogated when uPAR participitated in a bimolecular complex with pro-uPA or smaller receptor binding derivatives thereof, demonstrating the proximity of the ligand-binding site to this particular carbohydrate moiety. uPAR preparations devoid of carbohydrate on domain I exhibited altered binding kinetics toward uPA (a 4 -6-fold increase in K d ) as assessed by real time biomolecular interaction analysis.
The proteolytic potential of the plasminogen activation system is critically involved in vascular fibrinolysis (1) and tissue remodeling under pathological conditions such as cancer inva-sion and wound healing (2)(3)(4). Both plasminogen and its principal activators, tissue-type plasminogen activator and urokinasetype plasminogen activator (uPA) 1 are mosaic glycoproteins composed of a COOH-terminal chymotrypsin-like serine protease domain and a modular, non-catalytic NH 2 -terminal region, which exerts the specific cofactor activities of these proteins, exemplified by the cellular binding of uPA and fibrin-specific binding of tissuetype plasminogen activator. The modular composition of the aminoterminal fragment (ATF) of uPA includes an epidermal growth factor-like module and a kringle module, the former of which is responsible for the specific cell binding of uPA (5). The glycan structure of these proteins have been analyzed in great detail (6)(7)(8)(9) and in certain cases particular glycoforms of these proteins were even found to exhibit different biological properties (10)(11)(12)(13).
Binding of uPA to cells through the high-affinity association (K d of 0.1-1 nM) between the epidermal growth factor-like module of uPA and its membrane receptor facilitates cell associated plasmin generation catalyzed by uPA (14), a mechanism implicated in cancer invasion (3). The human uPA receptor (uPAR) is encoded as a single polypeptide chain of 313 amino acids including 5 potential attachment sites for N-linked carbohydrate (15). During post-translational processing an approximately 30-residue signal peptide is excised from the COOH terminus of uPAR with the concomitant addition of a glycolipid membrane anchor (16). Like plasminogen and its activators, uPAR also has a multidomain protein architecture being constructed from three homologous Ly6/uPAR-like (49) domains (17,18,49). These domains plausibly adopt an overall folding topology similar to those of CD59 (19,20) and the nonglycosylated snake venom ␣-neurotoxins (21).
A pronounced heterogeneity in the glycosylation pattern of uPAR expressed in various cell lines of neoplastic origin has been observed (22). It has, furthermore, been reported that cytokine treatment (transforming growth factor ␤1, epidermal growth factor, basic fibroblast growth factor, and phorbol 12myristate 13-acetate) of various cell lines (U937, A549, and HeLa) leads to a 5-20-fold increase in the number of surface exposed receptor molecules as well as a roughly proportional decrease in their ligand binding affinity (23)(24)(25)(26). Although the molecular mechanism underlying this phenomenon remains to be elucidated, differences in the glycosylation pattern might be involved, since phorbol 12-myristate 13-acetate induced differentiation of U937 cells is accompanied by an increase in the size heterogeneity of uPAR, presumably reflecting an altered processing of its glycan moieties (22,27,28). In accordance with this speculation is the finding that abolishment of the N-linked glycosylation of Asn 52 in uPAR domain I by site-directed mutagenesis (Asn 52 3 Gln) leads to an approximately 5-fold reduction in its binding affinity, when expressed in murine LB-6 cells (29). The present paper reports the first structural determination of the glycosylation pattern of uPAR and demonstrates in a purified system the direct influence of the carbohydrate moiety of uPAR domain I on real time receptor-ligand binding kinetics using surface plasmon resonance analysis.
Purified Proteins-A soluble recombinant variant of uPAR (residues 1-277) was expressed in Chinese hamster ovary (CHO dhfr Ϫ ) cells, which secrete this protein to the culture medium due to the truncation of the COOH-terminal signal sequence responsible for the glycolipid membrane attachment of uPAR (16). Cells were grown in the presence FIG. 1. Reversed-phase HPLC purification of uPAR domains II and III generated by pepsin treatment. Purified uPAR domain II ϩ III (residues 88 -277) obtained from chymotrypsin treatment of intact uPAR was dissolved in 0.2 M acetic acid and treated with low concentrations of pepsin. The enzymatic fragmentation was terminated by the reversed phase HPLC chromatography shown. The inset shows a Coomassiestained polyacrylamide gel (10%) after an SDS-polyacrylamide gel electrophoresis analysis of the collected fractions after reduction and alkylation.  a These peptides were labeled as described in the legend to Fig. 5. b These data represent two disulfide linked peptides (T4 ϩ T(7 ϩ 8)) examined in a previous study on the disulfide assignment of uPAR domain I (18). c This fraction contains approximately 30% of T(23 ϩ 24) thus accounting for relatively high levels determined for Cys, Tyr, and Lys (see Fig. 8). d Cysteine was determined as the mixed disulfide formed between cysteine and 3,3Ј-dithiodipropionic acid during acid hydrolysis (39). e Cysteine was determined as S-carboxymethyl-cysteine after derivatizing cystein with iodoacetamide. f ND, not determined. g Values represent data corrected for decomposition of N-acetylglucosamine during acid hydrolysis, assuming an average loss of 40%. These values are not as accurately determined as those measured concomitantly for the amino acids due to a variable decomposition of N-acetylglucosamine. of 10 nM methotrexate and secrete approximately 0.5 mg of uPAR per liter of harvest fluid (32). The uPAR protein was purified by immunoaffinity chromatography using a monoclonal anti-uPAR antibody (18). Active two-chain uPA (EC 3.4.21.31) was from Serono (Aubonne, Switzerland) and recombinant pro-uPA expressed in Escherichia coli was a kind gift from Dr. D. Saunders (Grü nenthal, Germany). The ATF of uPA was kindly provided by Drs. A. Mazar and J. Henkin (Abbott Laboratories, IL), whereas the growth factor-like domain (GFD) of uPA was produced by Glu-C digestion of two-chain uPA as described previously (33 (18). Cleavage in the linker region between domain II ϩ III was accomplished by pepsin treatment, a procedure recently devised to generate domain I ϩ II from intact uPAR (34). In the present experiment, lyophilized domain II ϩ III was dissolved in 0.2 M acetic acid and treated for 2 h at 37°C with pepsin at an enzyme to substrate ratio of 1:2000 (w/w). This digestion was terminated by injecting the sample directly onto a ProRPC TM HR5/10 column (Pharmacia) from which a sequential elution of the generated domains II and III was obtained by a 35-min linear gradient from 0 to 40% (v/v) 2-propanol in 0.1% (v/v) trifluoroacetic acid at a flow rate of 300 l min Ϫ1 (Fig. 1). Peptide Mapping of Domains II and III after Trypsin Degradation-Reversed-phase HPLC purified domain II and III (approximately 20 nmol) were taken to dryness by vacuum evaporation, redissolved in 150 l of 6 M guanidinium HCl, 0.5 M Tris-HCl, pH 8.0, and 2 mM EDTA. After flushing the samples with argon, reduction proceeded for 2 h at 50°C in the presence of 20 mM dithiothreitol. Generated thiol groups were subsequently alkylated by incubation with 50 mM iodoacetamide in the dark for 30 min after which the alkylated domains were purified by gel filtration on a Superdex 75 column using 0.1 M NH 4 HCO 3 as solvent.
These preparations were directly subjected to trypsin cleavage by addition of porcine trypsin at an enzyme to substrate ratio of 1:25 (w/w) followed by incubation overnight at 25°C. The resulting degradation patterns were visualized by reversed-phase HPLC chromatography using a Brownlee Aquapore OD-300 C18 column with a linear gradient (50 min) from 0 to 70% (v/v) acetonitrile in 0.1 and 0.085% (v/v) trifluoroacetic acid, respectively. The flow rate was 250 l min Ϫ1 and detection was at 214 and 280 nm. Identity of the eluted polypeptides was revealed by a combination of matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) and amino acid composition analyses.
Enzymatic Deglycosylation of Purified uPAR and Isolated Domains/ Glycopeptides Thereof-For structural analyses of the glycan moieties, sequential deglycosylation of reversed-phase purified domains of uPAR or isolated tryptic glycopeptides thereof was performed in 10 l of 50 mM NH 4 HCO 3 . Samples were initially incubated for 18 h with 5 milliunits of neuraminidase, followed by the withdrawal of 1 l for MALDI-MS analysis. After subsequent addition of 125 milliunits of N-glycanase, incubation was continued for another 18 h and 1 l was withdrawn for a second MALDI-MS analysis.
For functional binding studies of uPAR with differently processed glycan moieties, intact uPAR (approximately 1 nmol) was incubated for  The following experiments were conducted to explore whether the accessibility of the glycan moiety on Asn 52 in uPAR domain I toward enzymatic hydrolysis by N-glycanase was affected when the receptor was engaged in the formation of a bimolecular complex with various uPA derivatives. Receptor complexes were formed by preincubating uPAR (1 M) for 5 min at room temperature with buffer alone or a 4-fold molar excess of either bovine serum albumin (as control) or various receptor binding uPA derivatives (pro-uPA, ATF, or GFD). Synthetic peptides (AE78 and AE79) were, however, preincubated with uPAR at a 20-fold molar excess. These mixtures were treated with 20 milliunits of neuraminidase and 100 milliunits of N-glycanase for 20 min at 37°C, whereupon further deglycosylation was prevented by boiling for 5 min. The untreated sample also received the above deglycosylation mixture and was boiled immediately. Domain I was subsequently generated by incubation with 2 nM chymotrypsin for 10 min at room temperature and 1 l of these mixtures were finally subjected to MALDI-MS analysis. These experiments were performed in 50 mM Bistris-HCl, 100 mM NaCl, 5 mM EDTA (pH 7.1), including 2 mM APMSF, the presence of which allows the subsequent chymotrypsin cleavage.
Real Time Biomolecular Interaction Analysis (BIA)-The effects of various glycosidase treatments of uPAR on its ligand binding properties were determined by surface plasmon resonance using a BIAcore 2000 TM equipment (Pharmacia Biosensor, Uppsala, Sweden). The carboxymethylated dextran matrix (CM5 sensor chip) was preactivated with N-hydroxysuccinimide/N-ethyl-NЈ- (3-(diethylamino)propyl)carbodiimide, according to the manufacturers recommendations. Coupling of the ligands was achieved by subsequent injection of uPA (20 g/ml) or lectin AAA (50 g/ml) in 10 mM sodium acetate (pH 5.0) at a flow rate of 5 l min Ϫ1 for 6 min.
Just prior to the kinetic analyses by real time BIA, the various uPAR preparations from the deglycosylations experiments were re-purified by gel filtration on a Superdex 75 HR 3.2/30 column (Pharmacia, Uppsala, Sweden) using 50 mM Bistris-HCl, 45 mM NaCl, 5 mM EDTA (pH 7.1) as eluent to remove any aggregated or degraded material formed during the enzymatic deglycosylation (less than 10%). Sensorgrams (resonance units versus time) were recorded by the BIAcore 2000 TM at a flow rate of 10 l min Ϫ1 at 5°C, using 6 different concentrations of these uPAR preparations in the range of 10 -100 nM in running buffer (50 mM Bistris-HCl, 45 mM NaCl, 5 mM EDTA (pH 7.1) including 0.005% surfactant P-20). To reassure that no protein aggregation occurred while recording the sensorgrams, the individual samples were retrieved after analysis using the recovery function of BIAcore 2000 TM and subjected to analytical rechromatography on Superdex 75. After enzymatic deglycosylation samples were kept at 5°C and analyzed within 24 h. The sensor chip was regenerated at the end of each run by injection of 0.1 M acetic acid containing 0.5 M NaCl.
Specific detection of glycopeptides containing ␣(1-6)-linked fucose was performed by carbohydrate-specific surface plasmon resonance detection using immobilized lectin AAA (35). In such experiments regeneration of the sensor chip was obtained by injection of 100 mM fucose.
Data obtained from parallel mock-coupled flow cells (non-protein) served as blank sensorgrams for subtraction of bulk refractive index background. The obtained sensorgrams were analyzed by nonlinear least squares curve fitting using BIAevaluation 2.0 software (Pharmacia Biasensor, Uppsala, Sweden) assuming single-site association and dissociation models.
Electrospray Ionization Mass Spectrometry (ESI-MS)-ESI-MS was performed on a Sciex API III triple quadrupole instrument (Sciex, Thornhill, Canada) using an articulated ion-spray probe. The sample (2 mg/ml) was mixed with 1 volume of 1% (v/v) acetic acid in methanol and subjected to ESI-MS at a flow-rate of 3 l/min. A sprayer voltage of 5,000 volts and an orifice potential of 80 volts were used.
MALDI-MS-MALDI spectra were recorded on a linear time-of-flight instrument (Voyager, PerSeptive Biosystems, MA) equipped with a 1.2-m flight tube and a 337-nm nitrogen laser. To minimize prompt as well as metastable fragmentation of sialic acids during sample desorption (36), in particularly when ␣-cyano-4-hydroxycinnamic acid was used as matrix, laser power levels were kept as low as possible, typically 1% above the threshold value for desorption of matrix ions. Although metastable fragmentation of sialic acids undoubtly occurs to a significant extent during this desorption protocol, this would not give rise to artifacts in mass assignments, since the time-of-flight instrument was operated in the linear mode only, where fragment and parent ions possess identical initial velocities (36,37). The MS probes were precoated with a thin-layer of matrix by deposition of 1 l of freshly prepared matrix solution (20 mg/ml in 98% (w/w) acetone) followed by fast solvent evaporation. Sandwich sample preparation (38) occurred by consecutive deposition of 1 l of the following solutions to the precoated target: 2% (v/v) trifluoroacetic acid, sample to be analyzed and finally 15 mg/ml matrix dissolved in 50% (v/v) acetonitrile. After 1-2 min of crystallization, excess solvent was removed by aspiration and the probe surface was gently rinsed using 0.1% (v/v) trifluoroacetic acid. The majority of samples were analyzed using ␣-cyano-4-hydroxycinnamic acid as matrix, but sinapinic acid was used for intact uPAR to favor the formation of molecular ions (MH ϩ ) as well as reducing the MS induced fragmentation of sialic acid. One glycopeptide (T21) could, however, only be desorbed after deposition in 2,5-dihydroxybenzoic acid (2 l of 1:1 mixture of sample and matrix solution (10 mg/ml) was deposited on a clean MS probe surface).
Miscellaneous Analyses-Amino acid analysis was performed on a Waters amino acid analyzer with post-column o-phthaldialdehyde de- Theoretical trypsin cleavage of the human uPAR cDNA derived sequence yields 28 fragments, numbered T1 to T28; if a peak contains more than one peptide it will be assigned by two numbers, i.e. T1 ϩ T2; if a peptide on the other hand arises as a consequence of incomplete cleavage as an example between T1 and T2 it will be denoted T(1 ϩ 2).
tection. Acid hydrolysis was achieved by incubation in vacuo at 110°C for 20 h in 6 M HCl containing 0.05% (w/v) of both phenol and 3,3Јdithiodipropionic acid (39).

Generation and Purification of the Individual Domains of uPAR-
The NH 2 -terminal domain I of uPAR was purified by gel filtration after mild chymotrypsin treatment of a recombinant uPAR (residues 1-277) secreted by CHO cells, as reported previously (18). The remaining domain II ϩ III (residues 88 -277) was subsequently treated with pepsin and subjected to reversed-phase HPLC purification, which yielded two distinct components representing isolated domains II and III (Fig. 1). Amino acid composition analysis and mass spectrometry (ESI and MALDI) confirmed the purity and identity of these preparations of the individual domains of uPAR (Table I and Figs. 2 and 4). The pepsin treatment caused the liberation of domains II and III due to the specific cleavage in the interdomain linker region of the Glu 183 -Leu 184 peptide bond, in accordance with the major substrate specificity of this enzyme (40).
Although the theoretical molecular mass of the generated polypeptide chain corresponding to domain II (10,838.9 Da) is only slightly larger than that of domain III (10,185.3 Da) its migration in SDS-polyacrylamide gel electrophoresis is considerably slower and more heterogenous than that of domain III (Fig. 1, inset). As both domains have two consensus motifs for N-linked carbohydrate (Asn-Xaa-Thr/Ser) this might indicate that only one site is processed in domain III and/or that domain II in general carries more bulky carbohydrate moieties. The amount of N-acetylglucosamine determined by amino acid composition analysis also implies a more extensive glycosylation on domain II (Table I).
Glycosylation of the Individual Domains as Assessed by ESI and MALDI-MS-Purified uPAR domain I was amenable to analysis by ESI-MS, the spectra demonstrating that N-linked glycosylation of Asn 52 is the only post-translational modification detected ( Fig. 2 and Table I). By far the majority of this glycosylation is accounted for by a biantennary complex-type oligosaccharide carrying a deoxyhexose (most likely fucose) attached to the chitobiose moiety. The presence of an ␣(1-6)linked fucose moiety in this complex-type carbohydrate is verified by lectin-based surface plasmon resonance detection (35) using immobilized A. aurantia lectin. As shown in Fig. 3 only the glycosylated form of uPAR domain I binds specifically to the lectin and this interaction is completely inhibited by 0.5 mM fucose but not by 1 mM mannose, 1 mM galactose, or 1 mM N-acetylglucosamine.
The ESI-MS spectrum also reveals an apparent microheterogeneity in the sialylation pattern of domain I (Fig. 2, lower  panel). Some of this microheterogeneity may, however, be caused by ESI-MS induced fragmentation, since increasing the orifice voltage of the MS probe leads to a further fragmentation of the carbohydrate moiety. 2 Analysis of uPAR domain I by MALDI-MS after deposition in either ␣-cyano-4-hydroxycinnamic acid or sinapinic acid revealed a similar microheterogeneity in the sialylation pattern of uPAR domain I (data not shown). Although laser-induced prompt fragmentation may in principal generate some of this microheterogeneity, the predominant fragmentation mechanism by metastable ion forma-2 P. F. Nielsen, unpublished results.

TABLE II
Characterization of tryptic glycopeptides by mass spectrometry Masses were determined by ESI-MS (domain I) or MALDI-MS (glycopeptides) and are shown along with the theoretical masses in parentheses (when cysteine is present in the glycopeptides the mass of S-carboxamidomethyl cysteine is used). One atomic mass unit has been subtracted from the masses recorded of each molecular ion (MHϩ) by MALDI-MS due to the association of one proton during ionization. The relative abundance of the various glycoforms were estimated by comparison of peak intensities (peak heights) obtained within a single MALDI-MS spectrum for each glycopeptid after neuraminidase treatment.  (Fig. 4).

Characterization of Glycopeptides Derived from Domains II
FIG. 6. MALDI-MS spectra of the tryptic glycopeptide T20 from domain II. MALDI-MS spectra were recorded for HPLC purified T20 (residues 146 -169) before and after treatment with neuraminidase and N-glycanase. The average molecular mass of the deglycosylated T20 is 2697.96 Da (corrected for S-carboxyamidomethylation of cysteine and the Asn to Asp conversion by N-glycanase). An interpretation of the masses obtained for the neuraminidase-treated glycopeptide is shown in the middle spectrum. Peaks of low intensity with masses corresponding to an identical carbohydrate moiety carrying an additional core fucosylation (mass increase of 146.1 Da) have been highlighted. f, N-acetylhexosamine; q, hexose (most likely mannose); ⌬, hexose (most likely galactose).  N-glycanase). Since the NH 2 -terminal S-carboxyamidomethyl cysteine undergoes an intramolecular cyclization generating the corresponding thiazine derivative, the actual average molecular mass is 766.83 Da (compare with upper panel). An interpretation of the masses obtained for the neuraminidase-treated glycopeptide is shown in the middle spectrum. f: N-acetylhexoseamine; q, hexose (most likely mannose); ⌬, hexose (most likely galactose); ૺ, deoxyhexose (most likely fucose). and III-To obtain a detailed structural analysis of the Nlinked carbohydrates of uPAR domains II and III, including the positive identification of the putative unoccupied glycosylation site in domain III, the purified domains were reduced and S-carboxamidomethylated before trypsin cleavage and subsequent purification by reversed-phase HPLC (Fig. 5). The overall sequence coverage was greater than 95% of uPAR 1-277 , when the data on the identity of these HPLC purified peptides (Fig. 5, A and B) were combined by those obtained for domain I (Fig. 2). Only a single dipeptide (T13) and a single tripeptide (T18) thus escaped identification by MALDI-MS or amino acid composition analysis. Tryptic peptides encompassing both consensus motifs for N-linked glycosylation present in uPAR domain II (Asn 162 -Asp-Thr and Asn 172 -Thr-Thr) were indeed re-FIG. 8. MALDI-MS spectra of the tryptic glycopeptide T24 from domain III. MALDI-MS spectra were recorded for HPLC purified T24 (residues 199 -216) before and after treatment with neuraminidase and N-glycanase. The average molecular mass of the deglycosylated T21 is 2071.18 Da (corrected for Scarboxyamidomethylation of cysteine and the Asn to Asp conversion by N-glycanase). As indicated in Fig. 5B, this fraction also contains the incomplete tryptic peptide T(23 ϩ 24), the corrected molecular mass of which is 2898.14 Da. The NH 2 -terminal glutamine of T(23 ϩ 24) is partly converted to pyroglutamic acid (2881.11 Da). The characteristic double peaks (Ϯ17 Da) corresponding to the incomplete tryptic peptide T(23 ϩ 24) are indicated in the individual spectra by an asterisk (*). An interpretation of the masses obtained for the neuraminidasetreated glycopeptide T24 is shown in the middle spectrum. f, N-acetylhexoseamine; q, hexose (most likely mannose); ⌬: hexose (most likely galactose); ૺ, deoxyhexose (most likely fucose). covered as the corresponding glycopeptides T20 and T21 (Fig.  5A and Tables I and II), as predicted from the previous analyses of the purified receptor domain II (Fig. 4). Inspection of the MALDI-MS spectra recorded after neuraminidase or N-glycanase treatments of glycopeptide T20 (Asn 162 ) demonstrated that this glycan moiety is dominated by sialylated tri-and tetraantennary complex-type carbohydrate (Fig. 6). Similar experiments revealed that the second glycosylation site in uPAR domain II located at Asn 172 (T21) primarily carried the smaller sialylated biantennary complex-type carbohydrate including an additional deoxyhexose most likely as a core fucosylation (Fig. 7). Carbohydrate-specific surface plasmon resonance analysis, similar to that conducted for domain I in Fig. 3, confirmed the presence of fucose in purified intact domain II (data not shown).
Only a single glycopeptide (T24) was detected in the tryptic peptide map of uPAR domain III (Fig. 5B and Table I) and accordingly its sequence encompass one of the two glycosylation motifs present in this domain: Asn 200 -Ser-Thr. Subsequent glycoprofiling by MALDI-MS demonstrated that the glycan attached to Asn 200 primarily consisted of sialylated tri-and tetraantennary complex-type carbohydrate, approximately one-third of which contain an additional deoxyhexose presumably representing a traditional core fucosylation (Fig. 8). Accordingly, isolated domain III bound immobilized lectin AAA from A. aurantia as detected by surface plasmon resonance (data not shown). The average molecular mass of this glycan (3,075 Da) is comparable to the mass difference (2,875 Da) observed upon N-glycanase treatment of intact uPAR domain III (Fig. 4). The second cognate glycosylation motif present in uPAR domain III (Asn 233 -Gln-Ser) must therefore be unoccupied and the corresponding tryptic peptide (T26) was accordingly recovered quantitatively without carbohydrate as assessed by both amino acid composition analysis (Table I) and MALDI-MS (Table II).
In conclusion, site-specific microheterogeneity was evident from the analyses of all isolated tryptic glycopeptides. The glycan moieties on Asn 162 and Asn 200 were dominated by sialylated tri-and tetraantennary complex-type, whereas the glycan moieties on Asn 52 and Asn 172 primarily consisted of the smaller sialylated biantennary complex-type carbohydrate. A fifth potential glycosylation site located at Asn 233 was not processed at all.
Functional Impact of uPAR Glycosylation on Its Ligand Binding Affinity-As noted previously (18) the only carbohydrate accessible to cleavage by N-glycanase on native intact uPAR is that on domain I (Asn 52 ). To test whether this property is retained within bimolecular receptor-ligand complexes, we subjected preformed complexes between uPAR and pro-uPA, ATF, GFD, or a 17-mer synthetic peptide antagonist AE78 (AEPMPHSLNFSQYLWYT) to a combined enzymatic deglycosylation using neuraminidase and N-glycanase. When analyzed by MALDI-MS it was evident that the accessibility of the proximal N-acetylglucosamine linkage to Asn 52 to enzymatic hydrolysis is severely hampered by the presence of all ligands tested ( Fig. 9 and Table III). In contrast, the terminal sialic acids on this carbohydrate retained their sensitivity to neuraminidase (data not shown).
Defined preparations of intact uPAR containing truncated carbohydrate moieties were prepared by treatment with neuraminidase and N-glycanase under nondenaturing conditions. The manipulated proteins were purified by gel filtration and FIG. 9. Ligand induced protection on enzymatic deglycosylation of uPAR domain I as revealed by MALDI-MS spectra. Intact uPAR was preincubated with the ligands indicated before deglycosylation was attempted by incubation with a mixture of neuraminidase and N-glycanase. Domain I (residues 1-87) was finally released by a brief chymotrypsin treatment before the entire mixture was subjected to MALDI-MS. * indicates deglycosylated uPAR domain I, whereas ** represents the various glycoforms of uPAR domain I. The relative peak intensities corresponding to the glycosylated domain I (**) differ from that recorded previously by ESI-MS (Fig. 2), presumably because the activity of neuraminidase is not immediately and/or completely destroyed by the subsequent boiling. AE78 is a peptide antagonist of the uPA-uPAR interaction (AEPMPHSLNFSQYLWYT) and AE79 is an inactive scrambled version thereof (AEWSNLMQPYYPSTHFL). BSA, bovine serum albumin.

TABLE III
Ligand protection towards enzymatic deglycosylation of Asn 52 on intact uPAR Purified uPAR was incubated with 20 milliunits of neuraminidase and 100 milliunits of N-glycanase for 20 min, whereupon domain I was liberated by limited chymotrypsin cleavage and analyzed subsequently by MALDI-MS in the mixture. A relatively short incubation time was chosen to minimize dissociation of preformed complexes. This condition yields a 75% deglycosylation efficiency of uPAR domain I in the absence of added ligands, which defines the maximal (100%) deglycosylation efficiency achievable under these experimental conditions, cf. the equation presented in Footnote b. Representative MALDI-MS spectra for some of these experiments are shown in Fig. 9.

Sample
Relative peak heights a  (Table  IV). The binding kinetics between immobilized uPA and these uPAR preparations were measured directly by real time BIA using surface plasmon resonance (Table IV). The association and dissociation rate constants measured for the interaction between uPA and uPAR were comparable to those determined previously (31,34,42) and correspond to a dissociation constant of approximately 0.6 nM. Enzymatic removal of the sialic acids on uPAR by neuraminidase treatment did not change the binding kinetics significantly, whereas the specific removal of the entire carbohydrate moiety on Asn 52 in uPAR domain I by N-glycanase treatment led to an approximately 5-fold increase in the dissociation constant (K d of 3.2 nM versus 0.6 nM). This increase was caused by a combined effect on the association rate (a 3-fold decrease) and the dissociation rate (a 2-fold increase). DISCUSSION Several reports have proposed a possible functional role of the carbohydrate moieties of uPAR on its ligand binding properties (26,28,29). In the present study we have therefore examined the glycosylation pattern of a human soluble uPAR expressed in CHO cells. This recombinant protein escapes attachment to the plasma membrane due to a COOH-terminal truncation (residues 278 -313), which eliminates the signal sequence responsible for the normal glycolipid anchoring of the receptor (16). A similar soluble uPAR protein (presumably representing residues 1-283) is also secreted in vivo from peripheral blood cells affected by the disease paroxysmal nocturnal hemoglobinuria (43). The uPA binding properties of the soluble uPA receptor expressed in CHO cells are not impaired by the absence of the glycolipid moiety (44). In the present report we demonstrate that although the cDNA derived sequence of uPAR contains 5 consensus motifs for N-linked glycosylation (15), only 4 of these are actually utilized in the soluble uPAR secreted by CHO cells. The glycosylation profile determined for this recombinant uPAR is summarized in Fig. 10. The presence of O-linked carbohydrate in uPAR is precluded by comparison of masses derived from the cDNA sequence and those determined for the individual uPAR domains after treatment with N-glycanase (Figs. 2 and 4). In addition, no galactosamine was detected by amino acid composition analysis of intact uPAR after acid hydrolysis (16,22). The degree of microheterogeneity observed at the individual N-linked glycosylation sites in uPAR is comparable to that observed for other proteins expressed by CHO cells including tissue plasminogen activator (9, 45) and plasminogen (8), the most striking difference being the absence or very low abundance of high-mannose and hybrid-type oligosaccharides in the present uPAR preparation.
The latter observation has a bearing on the recently reported interaction between a recombinant uPAR and the cation independent mannose 6-phosphate receptor, since this complex formation was assumed to occur through a specific interaction with the carbohydrate moieties of uPAR (46). The recombinant uPAR analyzed in the present study also binds a purified mannose 6-phosphate receptor in ligand-blotting experiments via a mechanism that exhibited complete inhibition by preincubation with 1 mM free mannose 6-phosphate (data not shown). It is therefore most likely that the observed interaction with the mannose 6-phosphate receptor is governed by a very low abundance glycosylation variant in the present uPAR preparation carrying phosphorylated high-mannose or hybrid-type oligosaccharides. The relative abundance of such immature carbohydrates may, however, vary among uPAR preparations produced in different laboratories, since carbohydrate processing of recombinant glycoproteins is sensitive to several environmental conditions during large scale production, including the accumulation of ammonium ions (47).
Since the NH 2 -terminal domain I of uPAR (residues 1-87) plays a predominant role for uPA binding (17, 21, 31, 33, 44), we were particularly interested in testing a possible functional impact of the glycosylation at Asn 52 on ligand binding. As a FIG. 10. Deduced glycosylation profile of recombinant human uPAR expressed in CHO cells. The glycosylation profile as determined by mass spectrometry for the 5 potential carbohydrate attachment sites on human uPAR is shown for the desialylated recombinant protein (residues 1-277) expressed in CHO cells (see Table II). f, N-acetylhexoseamine; q, hexose (most likely mannose); ⌬, hexose (most likely galactose); and ૺ, deoxyhexose (most likely fucose).

TABLE IV
Kinetic and affinity constants for the interaction between immobilized uPA and intact uPAR subjected to various glycosidase treatments rule, carbohydrates can be considered as bulky hydrophilic structures that may shield a significant portion of the solvent accessible molecular surface area from participation in proteinprotein interactions. In this context it is therefore notable that the sole N-linked glycosylation site of uPAR domain I predominantly carries the smaller biantennary oligosaccharides (Fig.  10) and is located in close proximity of the ligand-binding site for uPA. The immediate vicinity of the N-linked oligosaccharide attached to Asn 52 and the ligand-binding site for uPA was demonstrated by the reduced accessibility of this particular glycan to enzymatic hydrolysis by N-glycanase under nondenaturing conditions, when the receptor was engaged in ligand binding to pro-uPA, ATF, or GFD (Table III and Fig. 9). Even the small 17-mer peptide antagonist (AE78) was capable of enforcing a similar constrain on the specific deglycosylation of Asn 52 upon interaction with its cognate ligand-binding site on uPAR (Table III and Fig. 9). Consistent with this topological relationship is the recent finding by site-specific photoaffinity labeling, that replacement of the single phenylalanine in AE78 with either p-benzoylphenylalanine or 4Ј-(trifluoromethyl-diazirinyl)-phenylalanine led to a photoprobe that inserted specifically into uPAR domain I upon UV-light exposure (31). The primary target residue for this photoaffinity labeling of uPAR has very recently been identified as Arg 53 in domain I. 3 Being juxtaposed to the glycosylation of Asn 52 this interaction site provides a reasonable molecular basis for the observed protection against deglycosylation of domain I in preformed uPAR⅐AE78 complexes. In addition, Tyr 57 has also been shown to participate directly in the formation of the receptor-ligand binding interface as demonstrated by protein-protein footprinting analyses (33).
A moderate negative modulation of the binding kinetics of the uPA-uPAR interaction was observed after enzymatic removal of the biantennary oligosaccharides attached to Asn 52 (Table IV). Although the molecular mechanism responsible for this relationship remains to be elucidated, these studies suggest, that an altered processing of the carbohydrate moiety attached to Asn 52 of uPAR domain I may in theory influence the binding kinetics of the uPA-uPAR interaction. Consistent with this speculation is the finding that prevention of the glycosylation of domain I by site-directed mutagenesis (Asn 52 3 Gln) in a glycolipid-anchored human uPAR expressed in murine LB6 cells led to a 4 -5-fold increase in the apparent K d for the cellular binding of uPA, compared with cells transfected with a wild type uPAR (29). Differentiation of monocytes and monocyte-like cell lines is accompanied by a decrease in their affinity toward uPA as well as an increase in the molecular size heterogeneity of their surface-exposed uPAR molecules (26,28,48). A shift in the glycan processing particularly at Asn 52 to the more bulky tri-and tetraantennary oligosaccharides could thus play a role for the in vivo modulation of the cellular binding of uPA observed upon monocyte differentiation. In future studies it should therefore be interesting to explore the binding kinetics between uPA and various recombinant uPAR molecules secreted by different expression systems, including baculovirus, Picia pastoris, Saccharomyces cerevisiae, and Aspergillus niger to test whether a correlation exists between the binding kinetics and the size of the glycan moiety attached to Asn 52 .