Báo cáo Y học: Tyrosine sulfation and N-glycosylation of human heparin cofactor II from plasma and recombinant Chinese hamster ovary cells and their effects on heparin binding pot

Tyrosine sulfation and N-glycosylation of human heparin cofactor II from plasma and recombinant Chinese hamster ovary cells and their effects on heparin binding Christoph Bo¨hme1, Manfre

Tyrosine sulfation and N-glycosylation of human heparin cofactor II from plasma and recombinant Chinese hamster ovary cells

and their effects on heparin binding

Christoph Bo¨hme1, Manfred Nimtz2, Eckart Grabenhorst3, Harald S Conradt3, Annemarie Strathmann1 and Hermann Ragg1

1

Faculty of Technology, University of Bielefeld, Germany;2Molecular Structure Research and3Protein Glycosylation, Gesellschaft fu¨r Biotechnologische Forschung mbH, Braunschweig, Germany

The structure of post-translational modifications of human

heparin cofactor II isolated from human serum and from

recombinant Chinese hamster ovary cells and their effects on

heparin binding have been characterized Oligosaccharide

chains were found attached to all three potential

N-gly-cosylation sites in both protein preparations The

carbo-hydrate structures of heparin cofactor II circulating in blood

are complex-type diantennary and triantennary chains in a

ratio of 6 : 1 with the galactose being > 90% sialylated with

a2fi 6 linked N-acetylneuraminic acid About 50% of the

triantennary structures contain one sLex motif Proximal

a1fi 6 fucosylation of oligosacharides from Chinese

ham-ster ovary cell-derived HCII was detected in > 90%

of the diantennary and triantennary glycans, the latter

being slightly less sialylated with exclusively a2fi 3-linked N-acetylneuraminic acid units Applying the ESI-MS/ MS-MS technique, we demonstrate that the tryptic peptides comprising tyrosine residues in positions 60 and 73 were almost completely sulfated irrespective of the protein’s ori-gin Treatment of transfected Chinese hamster ovary cells with chlorate or tunicamycin resulted in the production of heparin cofactor II molecules that eluted with higher ionic strength from heparin–Sepharose, indicating that tyrosine sulfation and N-linked glycans may affect the inhibitor’s interaction with glycosaminoglycans

Keywords: heparin cofactor II; glycosylation; tyrosine sulfation; heparin binding; serpins

Heparin cofactor II (HCII) is a single-chain glycoprotein

with a carbohydrate content of about 10% circulating in

blood [1] The protein is a member of the serpin (serine

protease inhibitor) superfamily, most members of which

inhibit serine proteases by forming SDS stable complexes

with their target enzymes HCII functions as an inhibitor of

cathepsin G, chymotrypsin and thrombin After addition of

heparin or dermatan sulfate, the rate of HCII/thrombin

interaction is enhanced several orders of magnitude and the

inhibition rate approaches the value observed for thrombin

inhibition by antithrombin (AT) in the presence of heparin,

indicating that HCII may be an interesting anticoagulant

and antithrombotic agent Increased plasma levels of HCII

have been found under a variety of pathophysiological

conditions, suggesting that the protein may be involved in

the acute phase response [2,3]

HCII has been biochemically characterized in some detail As deduced from the cDNA sequence, the mature human protein consists of 480 amino acids and contains three potential N-glycosylation sites at positions Asn30, Asn169 and Asn368 [4,5] In the mouse, two HCII variants that appear to differ in number and/or structure of their glycans circulate in blood [6] In addition, two tyrosine sulfation signals with the characteristic accumulation of acidic amino acids are present close to the N-terminus of all HCII sequences known [7,8] Sulfation of these tyrosine residues has been demonstrated for HCII secreted from a human hepatoma cell line [9]

Composition and stoichiometry of the monosaccharide units of the inhibitor molecule have been analyzed [1,10] with mannose, galactose, N-acetylglucosamine, and sialic acid being the main elements of HCII glycans, the structure

of the carbohydrate chains, however, is unknown Glycosy-lation may have profound effects on a variety of biological features of proteins [11] This is also evident for human AT, which exists in two isoforms (AT-a and AT-b) These variants differ substantially in their affinity for heparin (elution from heparin–Sepharose at about 0.8 and 1.2M NaCl, respectively) due to differential N-glycosylation [12,13] As a consequence, AT-b is predominantly associated with endothelial and subendothelial cells [14,15], providing the vessel wall with a strongly antithrombotic surface HCII isolated from human plasma displays a low affinity for heparin as indicated by the low NaCl concentration required for its dissociation from heparin–Sepharose (about 0.25MNaCl) Various experiments, however, have shown that HCII has a considerably higher intrinsic propensity for

Correspondence to H Ragg, Faculty of Technology, University of

Bielefeld, Universita¨tsstr 25, D-33501 Bielefeld, Germany.

Fax: + 49 521106 6328, Tel.: + 49 521106 6321,

E-mail: hr@zellkult.techfak.uni-bielefeld.de

Abbreviations: AT, antithrombin; CHO, Chinese hamster ovary;

dhfr, dihydrofolate reductase; DMEM, Dulbecco’s modified Eagle’s

medium; GAG, glycosaminoglycan; HCII, heparin cofactor II;

HPAEC-PAD, high pH anion-exchange chromatography with pulsed

amperometric detection; MEM, minimal essential medium; MTX,

methotrexate; NeuAc, N-acetylneuraminic acid; PEG, poly(ethylene

glycol); PNGase F, polypeptide:N-glycosidase F; sLe x , sialyl Lewis X.

(Received 26 September 2001, revised 6 December 2001, accepted 11

December 2001)

heparin binding, approaching that of AT-a HCII mutants

devoid of an acidic region (positions 53–75), which

resem-bles the C-terminal tail of hirudin, require > 0.7MNaCl for

elution from heparin–Sepharose [16–18], suggesting that

this polyanionic domain may affect the heparin-binding

properties of HCII Here, we report on the structure of

post-translational modifications of plasma-derived and

recom-binant human HCII from CHO cells and their effects on

heparin binding

M A T E R I A L S A N D M E T H O D S

Outdated citrated human plasma was a gift of Centeon AG

Bovine insulin, MTX, PEG, polybrene, sodium chlorate,

sulfate-deficient DMEM/Ham’s F12 1 : 1 mixture, and

tunicamycin were obtained from Sigma Culture media and

fetal bovine serum were purchased from Life Technologies

Heparin-HiTrap (5 mL), Mono Q HR5/5 and Superdex

200 HR 10/30 were from Pharmacia Standard grade BSA

was supplied from Serva A Sartocon Micro 30-kDa

MWCO cross flow module was from Sartorius Human

transferrin was purchased from Bayer AG

Peptide-N4-(N-acetyl-4-D-glucosaminyl)asparagine amidase F (from

Flavobacterium meningosepticum) from recombinant

Escherichia coli was bought from Roche Molecular

Bio-chemicals Vibrio cholerae sialidase was from Calbiochem

Stable expression and amplification of HCII in CHO cells

Plasmid pWTBI2 was constructed by ligating a 1.6-kb

HindIII/EcoRI human HCII cDNA fragment into a

HindIII/EcoRI-cleaved expression vector of the pSV2

plasmid series [19] The HCII cDNA spanned the protein

coding sequence (including the signal peptide) down to the

EcoRI site at position 1559 in the 3¢ untranslated region

(numbering as described previously [4]) The expression of

the thrombin inhibitor cDNA in this construct is regulated

by the SV40 early promoter and the polyadenylation signal

for SV40 early mRNAs

Plasmid pWTBI2 and plasmid pSV2dhfr, which encodes

a mouse dihydrofolate reductase cDNA under the control

of the SV40 early promoter, were mixed at a ratio of 5 : 1

(w/w) and transfected into DHFR-deficient CHO cells

(routinely cultured in Ham’s F12 medium) by the polybrene

method, following established procedures [20] After

2–3 weeks of selection in nucleoside-free a-MEM

contain-ing 10% dialyzed serum, 2 mM L-glutamine and 0.3 mM

L-proline, individual colonies were picked and expanded

Cell lines expressing increased amounts of HCII were

isolated by augmenting the MTX concentration in

incre-ments (0.05, 0.25, 2.5, and 100 lM, respectively)

The amount of recombinant HCII secreted from

MTX-selected cell clones was determined by a sandwich ELISA

employing a monoclonal mouse anti-HCII capturing Ig and

a horseradish peroxidase-coupled monoclonal mouse

anti-HCII detection antibody Assays were performed in 96-well

microtiter plates (Nunc) with HCII from human plasma as

standard The enzyme-catalyzed oxidation of 3,3¢,5,5¢

tetramethylbenzidine was monitored at 405 nm with a

microtiter plate autoreader (BioTek Instruments) Cells

were counted using the trypan blue exclusion assay and the

specific HCII productivity (lg per 106cells per day) was

calculated

Treatment of HCII expressing CHO cells with inhibitors

of N-glycosylation and tyrosine sulfation Recombinant CHO cells producing human HCII were grown close to confluency in DMEM/Ham’s F12 medium After washing twice with serum-free medium, cells were incubated in serum-free medium in the presence of tunica-mycin at the concentrations indicated in the figure legends After 4 h, fresh medium containing the same concentration

of the glycosylation inhibitor was added Three days later, the conditioned medium was harvested and inspected by Western blotting for the presence of HCII as described previously [21] Treatment of cells with sodium chlorate (20 mM) followed the same procedure, except that cells were cultivated in sulfate-deficient DMEM/Ham’s F12 1 : 1 mixture during the presence of the sulfation inhibitor

Purification of HCII from plasma and from recombinant CHO cells

HCII from outdated citrated human plasma of a single blood donor was isolated essentially as described previously, including precipitation with barium chloride and poly(ethyl-ene glycol) [22], chromatography on heparin–Sepharose and

on Mono Q [6] Highly purified HCII suitable for glycoanalysis was obtained through inclusion of a final gel filtration step on Superdex 200 HR 10/30

For mass production of recombinant HCII, cells from the CHO clone Bi2/100/6/13/3/19 were trypsinized and cultured

in a Superspinner [23], consisting of a Duran flask (capacity

1 L) equipped with a magnetic membrane stirrer in order to improve the oxygen supply by bubble-free aeration The device was placed on a stirring plate in a CO2incubator together with a small membrane pump, which supplied the membrane stirrer with the incubator gas The starting volume (550 mL) was inoculated with 2.5· 105cellsÆmL)1 from five confluent T175-flasks, which had been propagated

in DMEM/Ham’s F12 medium with 2% fetal bovine serum To remove bovine HCII, the serum content was successively reduced further in a combined repeated/fed batch process To this end, the cell suspension culture was diluted to a cell density of 2.5· 105cellsÆmL)1, filled up to the maximum working volume (1 L) with serum-free DMEM/Ham’s F12 medium supplemented with 1 gÆL)1 BSA, 3 mM glutamine, 10 lgÆmL)1 bovine insulin,

10 lgÆmL)1human transferrin and antibiotics, and grown

to the early stationary phase After two further dilution steps, the cells were finally cultivated for 5 days in serum-free medium supplemented with 0.5 gÆL)1 BSA, amino acids, and glucose

For isolation of recombinant HCII, the cells were removed by centrifugation, and after addition of phenyl-methanesulfonyl fluoride (10 lM), the supernatant was concentrated about threefold in a cross-flow ultrafiltration module and dialyzed against 10 mM Tris/HCl, 1 mM EDTA, 10 lM phenylmethanesulfonyl fluoride, 5 mM NaCl, pH 7.4 The following isolation procedures included fractionation on heparin–Sepharose and chromatography

on Mono Q essentially as described above, except that the Tris/EDTA buffer system was used for heparin–Sepharose chromatography For glyco-analysis, HCII was rechromato-graphed on heparin–Sepharose with a linear 0–1M NaCl gradient (40 mL) and on Superdex-200 HR 10/30 Purity

was assessed by reducing SDS/PAGE on 10% gels run in a

Tris/glycine buffer system, according to the manufacturer’s

protocol (Novex), and subsequent visualization of proteins

by Coomassie blue staining The homogeneity of HCII from

plasma and from recombinant CHO cells was checked by

N-terminal sequence analysis using an Applied Biosystems

ProciseTMinstrument

Reduction, carboxamidomethylation and tryptic

digestion of HCII from plasma and from recombinant

CHO cells

One to two nanomoles of the purified protein were reduced,

carboxamidomethylated and digested with trypsin;

RP-HPLC separation of resulting peptides was performed

as described previously [24]

Separation of N-glycans by Mono Q anion-exchange

chromatography

Oligosaccharides were liberated quantitatively by treatment

of the glycoprotein preparations (2 mgÆmL)1) with 15 U of

recombinant PNGase F in 50 mMsodium phosphate buffer

(pH 7.2) for 8 h at 37°C

In order to separate native N-glycans according to

charge, desalted oligosaccharides were dried using a

Speed-Vac and redissolved in 0.5 mL of MilliQ water The

N-glycans were applied to a Mono Q HR 5/5 column at

room temperature and eluted at 1 mLÆmin)1with a mixture

of water (solvent A) and 0.5M NaCl (solvent B)

Chro-matographic conditions were: a 5-min isocratic run with

100% A, followed by a linear gradient to 10% B for 15 min,

an isocratic run using 10% B for 10 min and a final linear

gradient to 100% B over 1 min Oligosaccharides were

detected by their ultraviolet absorption at 206 nm

Frac-tions of 0.5 mL were collected

Enzymatic and chemical removal of sialic acid

Vibrio cholera neuraminidase (2.5 lL; 1 UÆmL)1) was

added to the samples containing 0.5 nmol of total

oligo-saccharides in 25 lL of sodium acetate, pH 5.0, 5 mM

CaCl2and 0.02% NaN3(w/v) The reaction mixture was

incubated for 2 h at 37°C For the chemical removal of

NeuAc, N-glycans were incubated in 100 lL of 0.2%

trifluoroacetic acid for 1 h at 82°C

High-pH anion-exchange chromatography with pulsed

amperometric detection

Purified native and desialylated oligosaccharides were

analyzed by high-pH anion-exchange (HPAE)

chromato-graphy using a Dionex BioLC system (Dionex, Sunnyvale,

CA, USA) equipped with a CarboPac PA1 column

(0.4· 25 cm) in combination with a pulsed amperometric

detector (PAD) [25,26] Detector potentials (E) and

pulse durations (T) were: E1: + 50 mV, T1: 480 ms;

E2: + 500 mV, T2: 120 ms; E3: ) 500 mV, T3: 60 ms,

and the output range was 500–1500 nA The

oligosaccha-rides were then injected into the CarboPac PA1 column that

was equilibrated with 100% solvent C Elution (flow rate of

1 mLÆmin)1) was performed by applying a linear gradient

(0–20%) of solvent D over a period of 40 min followed by a

linear increase from 20 to 100% solvent B over 5 min Solvent C was 0.1MNaOH in bidistilled H2O, solvent D consisted of 0.6MNaOAc in solvent C

Reduction and permethylation of oligosaccharides The enzymatically liberated N-glycans were reduced and permethylated as described previously [27]

MALDI-TOF MS The reduced and carboxamidomethylated tryptic peptides

of the HCII preparations were subjected to positive ion matrix-assisted laser desorption/ionization (MALDI) mass spectrometry, using a Bruker REFLEX time-of-flight (TOF) instrument equipped with delayed-extraction and reflectron systems and a N2laser (337 nm) operating with 3-ns pulse width and 107)108WÆcm)2 irradiance at the surface of 0.2 mm2spots In addition to the positive mode standard procedure for the detection of sulfated peptides using the reflectron for enhanced resolution, the peptide mixture was also analyzed in the positive and negative linear ion mode One-microliter samples containing equal volumes of peptide solution (10 pmolÆlL)1) and the ultraviolet-absorbing matrix [19 mg a-cyano-4-hydroxy-cinnamic acid in 400 lL acetonitrile and 600 lL 0.1% (v/v) trifluoroacetic acid in H2O] were spotted onto the stainless steel target and dried at room temperature Determination

of the molecular masses of reduced and permethylated oligosaccharides was carried out similarly in the positive ion mode using the reflectron One-microliter samples containing equal volumes of reduced and permethylated oligosaccharide solution ( 10 pmolÆlL)1) and a-cyano-4-hydroxycinnamic acid matrix were mixed and spotted on the target Sodium chloride was added to the matrix to a final concentration of 5 lM in order to guarantee the exclusive generation of sodium adducts of the carbo-hydrate molecular ions

Tandem electrospray ionization (ESI) mass spectrometry The peptide samples were dissolved in a 1 : 1 mixture of MeOH/H2O (the reduced and permethylated oligosaccha-ride samples in 9 : 1 MeOH/H2O) to a concentration of

 3 pmolÆlL)1, and gold-coated nanospray glass capillaries (Protana, Odense, Denmark) were filled with 3 lL of this solution The tip of the capillary was placed orthogonally in front of the entrance hole of a QTOF II mass spectrometer (Micromass, Manchester, England) equipped with a nano-spray ion source, and a voltage of 800 V was applied For collision-induced dissociation experiments, parent ions were selectively transmitted from the quadrupol mass analyzer into the collision cell Argon was used as the collision gas and the kinetic energy was set from )15 to )60 eV for optimal fragmentation The resulting daughter ions then were separated by an orthogonal TOF mass analyzer

R E S U L T S

Expression of HCII from CHO cells Human HCII cDNA was expressed under the control of the SV40 early promoter in DHFR-deficient CHO DUKX-B1

cells, which had been cotransfected with plasmid pSV2dhfr.

More than 20 individual clones growing in selection

medium were examined, but initially no cell lines producing

sufficient HCII detectable by the ELISA technique were

identified Therefore, individual clones were isolated at

random, exposed to 0.05 lMMTX, expanded, and checked

for the presence of human HCII mRNA by RT-PCR

Positive cell lines were picked and after several rounds of

selection with increasing MTX concentrations, cell lines

expressing up to 17 lg HCII per 106 cells per day were

isolated

Production, purification and characterization

of recombinant and plasma HCII

The high producer clone Bi2/100/6/13/3/19, selected in

medium containing 100 lM MTX was cultivated in

suspension in a 1-L superspinner by successively reducing

the serum concentration During the production phase,

the serum-free cultivated cells accumulated HCII to a

concentration of about 11 lgÆmL)1within a 5-day period

The recombinant inhibitor and HCII from plasma

were purified to apparent homogeneity as assessed by

SDS/PAGE (not shown) and subjected to sequence

analysis The N-terminus of the inhibitor consisted of a

single sequence (GSKGPLDQLEKGGE), irrespective of

whether HCII had been isolated from plasma or from

recombinant CHO cells Thus, these results are in agreement

with the accurate and efficient cleavage of the 19-amino-acid

signal peptide in CHO cells [5,28]

Characterization of the tryptic peptides

of serum-derived and recombinant HCII

Natural and recombinant HCII were characterized further

by peptide mapping After reduction and

carboxamidome-thylation, the proteins were digested with trypsin and the

resulting peptide mixtures were subjected to RP-HPLC

mapping The elution profiles of both protein preparations

were almost identical (data not shown) This result was

further corroborated by MALDI-TOF-MS analysis of the

tryptic peptide mixtures (compare Fig 1 and Table 1)

Most signals could be assigned, resulting in a sequence

coverage of almost 70% The pattern of peptides obtained

from both proteins was very similar, indicating an identical

polypeptide backbone of the natural and the recombinant

HCII Position 218, for which an amino-acid polymorphism

has been reported (Lys or Arg [4,5]), is populated with a

lysine residue in the plasma derived protein We note that

higher amounts of peptides containing oxidized methionine

were detected in the recombinant polypeptide preparation

(see legend to Table 1)

Identification of sulfated tyrosine peptides

and glycosylation site occupancy in trypsin digests

of serum-derived and recombinant HCII

Human HCII from HepG2 cells has been found to be

sulfated at two tyrosine residues [9] In our experiments,

however, solely the unmodified forms of the corresponding

tryptic peptides T43–65 and T66–101 were detected when

the cleavage products from both protein preparations were

analyzed by the standard MALDI-TOF-MS techniques

Even after desialylation by mild acid hydrolysis, no MALDI signals corresponding to N-glycopeptides were identified in the pertinent spectra (Fig 1) The tryptic peptides of both protein preparations were therefore additionally subjected

to ESI-MS analysis Using this ionization technique, the tyrosine containing peptides T43–65 and T66–101 were found predominantly in their post-translationally modified forms (Fig 2) These findings can be explained by the pronounced instability of tyrosine sulfate under the condi-tions routinely applied for positive ion MALDI-TOF-MS

of peptides

A complete series of sequence-specific fragment ions was only found with the smaller sulfopeptide upon ESI-MS/MS, whereas the larger one yielded an intense signal generated by elimination of SO3 and a relatively weak fragment due to peptide cleavage at the proline residue The daughter ion spectra of the sulfated and

ENTVTNDWIPEGEEDDDYL-DLEK(43–65) obtained after increasing the collision energy were found to be almost identical (data not shown), confirming the amino-acid sequence of the corresponding peptide from the recombinant and the serum-derived HCII

As no modified fragment ions were detected after collision-induced dissociation from the sulfated peptide, the position

of sulfation could not be deduced from this spectrum Such behaviour can be explained by the spontaneous elimination

of SO3 from any possible peptide fragment generated Therefore, the determination of the sequence position of sulfated tyrosine residues appears to be impossible by classical mass spectrometric peptide sequencing techniques However, in the case of the peptides under consideration here, the position of sulfation is unambiguous, as both peptides contain only a single tyrosine residue, and sulfation at Ser/Thr residues is improbable, in view of the characteristic consensus sequence for tyrosine sulfation [29,30] present in both peptides

In addition to the identification of tyrosine sulfate residues detected with the ESI technique, we were also able to identify the dominant glycoforms of all three potential glycopeptides [NLSMPLLPADFHK(30–42), DFVNASSK(166–173), SMTNR(T) (365–370)] after desi-alylation by mild acid hydrolysis (Table 1) As the corre-sponding unmodified peptides were neither detectable by MALDI nor by ESI-MS, we deduce from these results that all three potential HCII N-glycosylation sites are completely glycosylated This conclusion is also supported by the SDS/ PAGE pattern of HCII from CHO cells treated with limited concentrations of tunicamycin (Fig 3)

In contrast to the serum-derived protein, which predom-inantly contained a diantennary carbohydrate structure attached to each glycosylation site, the monofucosylated (plus 146 Da) derivative was observed for the glycopeptides

of the recombinant protein The expected linkage position

of this fucose unit to the proximal GlcNAc residue could be confirmed by MS/MS of the respective molecular ions by the identification of a fragment ion generated by the cleavage of the chitobiose bond of the N-glycan derived from serum HCII Generally, intense carbohydrate specific fragments were detected (fragment ions generated by elimination of monosaccharide residues from the molecular ion, as well as intense pure carbohydrate fragments), but only very weak peptide sequence specific fragment ions (data not shown)

Detailed characterization of the N-glycans from serum

and CHO cell-derived HCII

For a more detailed investigation of the N-glycan structures,

we performed additional mass spectrometric analyses of the

oligosaccharides liberated by PNGase F After reduction

and permethylation of the total glycan mixtures, the

MALDI-TOF spectra depicted in Fig 4 revealed the

presence of a mainly diantennary disialylated structure

and approximately 15% of triantennary trisialylated chains

in both proteins (compare legend to Fig 4), confirming the

results obtained from the ESI-MS analysis of the tryptic

peptides described above The CHO cell-derived N-glycans

were almost completely fucosylated at the proximal

GlcNAc residue, whereas only about 10% of the serum

protein showed fucosylation of the diantennary and

triantennary N-glycans Approximately 60% of the trian-tennary structures isolated from the serum HCII contained

a fucose residue In order to determine the linkage position

of the fucose to the triantennary structure, ESI-MS/MS analysis was performed on the triply charged molecular ion [M + 3Na]3+after isolation of the trisialylated oligosac-charide fraction by anion exchange chromatography on a Mono Q column [31] From the resulting daughter ion spectrum (Fig 5), we conclude that the fucose unit was not linked to the proximal GlcNAc residue, which is character-istic for the CHO cell-derived material, but was rather attached to a peripheral GlcNAc residue as is indicated

by the weak fragment at m/z 1021.6 [NeuNAc-Hex-(dHex)HexNAc + Na] and its more intense secondary fragment at m/z 646.3 [HO-Hex-(dHex)HexNAc + Na] (see fragmentation scheme and legend of Fig 5) This

Fig 1 MALDI-TOF-MS tryptic peptide fingerprints of natural (A) and recombinant (B) HCII Amino-acid sequence, calculated, and experi-mentally detected masses of the peptide fragments are summarized in Table 1 Both protein preparations yielded a very similar peptide pattern, suggesting an identical polypeptide backbone The oxidation rate of the methionine residues, however, was markedly higher for the recombinant protein, as can be clearly seen, e.g for the peptides T311-343 [m/z 3672.1) or T420-449 [m/z 3206.2], each containing a single methionine residue We failed to detect any of the N-glycopeptides using the MALDI technique, even after desialylation The dominant glycoforms of all three glyco-peptides, however, were readily observed employing the ESI technique Peptides T43-65 and T66-101, which had been reported to be sulfated [9], were observed only in the desulfated form due to elimination of SO 3 By ESI-MS (see Fig 2) it could be unequivocally demonstrated that serum HCII as well as the recombinant species are predominantly sulfated *, incompletely cleaved peptide.

Table 1 Amino-acid sequence and tryptic peptides of human HCII The peptides indicated below were detected by MALDI-TOF- and/or ESI-MS (see Figs 1 and 2, respectively) in tryptic digests of human HCII isolated from serum or recombinant CHO cells, respectively, indicating their identical polypeptide backbones In the serum-derived protein  10–20% of the methionine residues were oxidized, whereas in the recombinant polypeptide about 50% of these residues occurred in the oxidized form Peptides smaller than 650 Da were not detected (ND) With the inclusion of the glycopeptides detected only by ESI-MS, a sequence coverage of more than 80% was achieved N-glycosylation consensus sequences are underlined, sulfated tyrosine residues are typed in bold w, m, s ¼ weak, middle or strong signal, respectively GP, glycopeptide glyc, glycan.

a

Glycopeptide; using ESI-MS, the most abundant glycoforms could be detected (diantennary complex type with two NeuAc residues (serum protein) or diantennary complex type with proximal fucose and two NeuAc units (recombinant protein) b Peptide containing a sulfated tyrosine Due to the easy elimination of SO 3 , MALDI-MS allowed the detection of only the desulfated molecular ion With ESI-MS, however, the predominant presence of a sulfopeptide was unequivocally detected (see text and Fig 2). cPeptide bearing a carboxamidomethylated cysteine residue d The presence of two peptides with identical molecular masses but different amino-acid sequences could be unequivocally shown by ESI-MS/MS e Peptide with one missed cleavage site.

interpretation was further corroborated by the detection of

the intense fragment at m/z 316.2 [GlcNAc-ol + Na]

instead of the fragment characteristic for proximally

fucosylated N-glycans at m/z 490 [dHex-HexNAc-ol +

Na] Even the linkage position of the fucose residue at the

N-acetyllactosamine antennae could be assigned from the

daughter ion spectra, due to the well-known preferential

elimination of the 3-linked substituent of the GlcNAc

residue of the N-glycan antenna [32] A weak, but

repro-ducible signal at m/z 440.3 was detected, which is generated

by the elimination of fucose from the fragment ion at m/z

646.3 Therefore, fucose must be linked to O-3 of a GlcNAc

residue, indicating the presence of a sLexunit rather than a

sLea motif in the N-linked oligosaccharides from

serum-derived HCII

HPAEC-PAD mapping of the desialylated

charides enabled the quantitation of the basic

oligosac-charide chains present in both glycoprotein preparations

and demonstrated again the very similar sialylation

degree of both proteins Table 2 summarizes the

glycosy-lation characteristics of recombinant and serum HCII

based on mass spectrometry results and HPAEC-PAD mapping using oligosaccharides of known structure as standard

Heparin binding properties of HCII treated with inhibitors of post-translational modifications

In order to investigate the influence of post-translational modifications on heparin binding, recombinant CHO cells were treated with inhibitors of tyrosine sulfation and N-glycosylation under conditions that allowed partial inhibition of these modifications After dialysis, the condi-tioned medium was fractionated on heparin–Sepharose with

a linear NaCl gradient (Fig 6) HCII produced in the presence of 20 mMsodium chlorate dissociated in a bimodal manner from the affinity matrix The first peak is observed

at 280 mMNaCl, a concentration characteristic for HCII synthesized in the absence of the sulfation inhibitor

A second peak is present at  430 mM NaCl, and a considerable amount of HCII eluted at still higher ionic strength, a property not associated with HCII from cells

Fig 2 Mass region of the triply charged molecular ions of the two tyrosine sulfated tryptic peptides recorded by ESI-MS of (A) serum-derived HCII and (B) the recombinant protein from CHO cells The detected molecular ions of these peptides (43-ENTVTNDWIPEGEEDDDYLDLEK-65 and 66-IFSEDDDYIDIVDSLSVSPTDSDVSAGNILQLFHGK-101) are compatible with the presence of one tyrosine O4-sulfate ester in each peptide (calculated m/z of monoisotopic masses for the triply charged monosulfated molecular ions [M + 3H]3+: 940.4 and 1331.3, corresponding to a molecular mass of 2818.1 and 3990.8 Da, respectively) Arrows indicate the expected positions for the molecular ions of the triply charged unsulfated peptide species Approximately 10% of the smaller peptide were present in its unmodified form in both protein preparations In contrast, the larger peptide from the natural protein was completely sulfated, while about 20–30% of this fragment from the recombinant protein did not contain this modification.

grown in normal medium A shift towards elution at higher

salt concentrations, albeit less pronounced, was also

observed for the totally unglycosylated  64 kDa HCII

form, secreted from CHO cells grown in the presence of

1 lgÆmL)1tunicamycin Similar results were obtained with

HCII from HepG2 cells cultivated in the presence of

tunicamycin (not shown)

D I S C U S S I O N

In this investigation we have analyzed the structure of

post-translational modifications of human HCII from

circulat-ing blood and genetically modified CHO cells and their

effects on heparin binding All three potential

N-glycosy-lation sites were found to be populated with complex type

carbohydrate chains Interestingly, sLex motifs were

de-tected in triantennary oligosaccharides of plasma HCII,

whereas only trace amounts of this structural motif were

present in the diantennary glycan fraction It remains to be

determined which structural features discriminate

N-gly-cans for addition of this modification, and which types of

a1,3/4-fucosyltransferases [33] are able to decode the

involved signals The presence of sLexstructures in HCII

is an intriguing finding in the light of reports which indicate

that several proteins associated with the acute phase

response in humans contain altered glycostructures

[34,35] and that HCII levels in plasma are increased under

inflammatory conditions [2,3] Therefore, it could be

possible that the N-glycan structures of HCII are also

changed under certain (patho)-physiological situations We

have detected triantennary oligosaccharides containing

the Lex motif also in a1-microglobulin, hemopexin, and

in a-antitrypsin, whereas human serum AT contains this

structural motif almost exclusively in 5% of the dianten-nary oligosaccharides (H S Conradt & M Nimtz, unpublished observations) It remains to be established whether the amounts of sLex containing oligosaccharides synthesized by the liver are accomplished by a concomitant increase in the levels of the a2fi 3 sialyltransferase and a1fi 3 fucosyltransferase VI

sLexunits linked to cell-bound molecules on the surface

of leukocytes have been found to interact with selectins exposed on endothelial cells [36] It may be envisaged that sLex bearing HCII molecules could interfere with these processes resulting in decreased plasma levels of the inhibitor On the other hand, interaction of HCII molecules with receptors recognizing sLex-carrying struc-tures on vessel-lining cells could lower the risk of thrombotic events Blood from a single donor was used

in this work for the analysis of post-translational mod-ifications As glycan structures between individuals may differ, further investigations may determine whether qualitative or quantitative changes in the carbohydrate pattern of HCII correlate with specific (patho)-physio-logical situations

There were several features specific for human HCII expressed in CHO cells; compared to plasma HCII, only a1fi 6 fucosylation of the recombinant HCII was observed and sLexstructures were not detected (the trace amounts of difucosylated oligosaccharides contain an additional fucose residue a1fi 2 linked to galactose, thus constituting the Lewis H motif (for review see [37]) The NeuAc units are linked in a CHO cell-characteristic manner by a2fi 3 bonds, consistent with the lack of a2fi 6 sialyltransferase activity in CHO cells [38,39]

Tyrosine sulfate has been implicated in several biological roles like leukocyte adhesion and haemostasis [40] HCII contains two adjacent sequences (positions 53–62 and 69–75, respectively) with similarity to the consensus signals characteristic for tyrosine sulfation The presence of tyrosine O4-sulfate esters in this domain, which resembles the acidic C-terminal tail of hirudin, has previously been reported for HCII from HepG2 cells [9] We were not able

to detect this modification when routine positive ion MALDI-TOF conditions were used This issue has recently been addressed [41]; tyrosine sulfated peptides readily eliminate SO3 and therefore solely the desulfated peptide form was detected Even when applying negative ion MALDI in the linear mode, as has been recommended [41], we could detect only very small amounts of the sulfated molecular ion signal compared to the nonsulfated peptide signal Therefore, the detection and quantitation of polypeptide modification by sulfate provides a major challenge to mass spectrometric analysis In contrast to the situation with peptide phosphorylation, which can be detected in the positive as well as in the negative ion mode,

it is difficult or even impossible to detect tyrosine sulfation

by MALDI-TOF-MS With the electrospray technique applying very soft nozzle/skimmer conditions, degradation

of the peptide sulfates was minimized or almost avoided During MS/MS experiments on both sulfated peptides, we observed a very facile elimination of SO3even under very mild conditions where peptide bonds remained completely intact

The ESI-MS results presented here clearly show that serum HCII as well as its recombinant counterpart

Fig 3 Electrophoretic resolution of recombinant HCII from CHO cells

incubated with various concentrations of tunicamycin After

concentra-tion ( fivefold), the medium was fracconcentra-tionated by SDS/PAGE and

examined by Western blotting for the presence of HCII Lanes 1–5,

HCII from cells treated with the indicated concentrations of

tunica-mycin; lane 6, recombinant HCII from an independent experiment.

For comparison, PNGase F-treated HCII from CHO cells (lane 7) and

purified HCII from recombinant CHO cells (lane 8) were included.

expressed by CHO cells are almost quantitatively sulfated at

the two tyrosine residues at positions 60 and 73, respectively

The nearly complete tyrosine sulfation of the recombinant

HCII molecules was unexpected, as several reports showed

incomplete sulfate ester modification of recombinant

pro-teins expressed in CHO cell lines [42,43] Such individual

differences indicate that the efficiency of tyrosine sulfation

may depend on additional signals [44] and/or accessability

to the modifying enzyme

HCII from CHO cells incubated with inhibitors of

post-translational modifications eluted at higher ionic

strength from heparin–Sepharose than inhibitor molecules

isolated from untreated cells In the case of tunicamycin,

this may be the consequence of reduced sterical hindrance

of heparin binding due to the missing N-glycans, similar

to the situation observed with AT [45,46] This may

especially apply to the carbohydrate chain linked to

Asn169, a position in proximity to a site involved in

heparin binding by HCII [47,48] Inhibition of tyrosine

sulfation had an even more profound effect on the

interaction between HCII and heparin The observation

that higher NaCl concentrations are required for the dissociation of the unsulfated inhibitor from the heparin affinity matrix indicates that the N-terminal acidic domain may affect the GAG binding domain, although it can not

be excluded that this effect is due to the reduction in the protein’s overall negative charge

In summary, we present evidence for a very strong similarity of HCII from human serum and its recombinant counterpart from CHO cells with respect to tyrosine sulfation and N-glycosylation The remarkable identity of oligosaccharide antennarity and the extent of sialylation observed in both preparations represents an example of the importance of polypeptide structure governing protein N-glycosylation, as it is known that CHO cells have a high capacity to synthesize N-glycans of high tetraantennarity with considerable N-acetyllactosamine repeats, which however, were not detected in our recombinant glyco-protein We demonstrate that tyrosine sulfation and N-gly-cans individually affect heparin binding of the inhibitor These findings may be exploited to generate HCII variants with increased heparin affinity

Fig 4 MALDI-TOF mass spectra of the reduced and permethylated total N-glycans enzymatically liberated from human HCII isolated (A) from human serum or (B) produced by genetically engineered CHO cells The following complex type carbohydrate structures were assigned to the detected molecular ions [M + Na]+: 2622, diantennary monosialylated with one fucose residue; 2796, diantennary monosialylated with two fucose residues;

2809, diantennary disialylated; 2983, diantennary disialylated with one fucose residue; 3432, triantennary disialylated with one fucose residue; 3619, triantennary trisialylated; 3793, triantennary trisialylated with one fucose residue; 4242, tetraantennary trisialylated with one fucose residue (monoisotopic masses) The major difference between the oligosaccharides from the natural protein compared to its recombinant counterpart is the almost complete proximal fucosylation and a slightly lower degree of sialylation of the recombinant material The triantennnary trisialylated N-glycan with one fucose residue (marked by an arrow) is the only major molecular ion observed in both glycoproteins ESI-MS/MS (compare Fig 5), however, showed that the fucose residue in the triantennary structure from the serum protein is not linked to the proximal GlcNAc, but peripherally to an N-acetyllactosamine antenna, thus constituting a Lexmotif F ¼ fragment; * ¼ artefacts due to the insertion of CH 2 O or CO 2

HCII may provide a valuable tool to study the fidelity of

post-translational modifications in recombinant cell lines

engineered with new transferases, e.g fucosyltransferases or

sulfotransferases for the production of polypeptides which

are modified in novel ways and which may be used for the study of the functional significance of post-translational protein modifications

Table 2 N-glycan structures of serum-derived and recombinant HCII After enzymatic liberation of the N-glycans from both protein preparations, the carbohydrates were subjected to HPAEC-PAD after enzymatic desialylation The major oligosaccharide from serum HCII was identified as a diantennary type II N-acetyllactosamine oligosaccharide (Gal 2 GlcNAc 2 Man 3 GlcNAc 2 ), whereas the major glycan of the CHO cell-derived HCII contained the same structure with an additional proximally a1 fi 6 linked fucose ND, not detected.

Structure

% Oligosaccharides in

a2 fi 6 or a2 fi 3 sialylation of terminal

Fig 5 ESI daughter ion spectrum of the triply charged parent ion [M + 3Na]3+of a reduced and permethylated triantennary trisialylated N-glycan with one fucose residue isolated from serum HCII The detected fragment ions, in particular the signal at m/z 646.3 [HO-Hex-(dHex-Hex-NAc) + Na], as well as the accompanying signal at m/z 440.3, which is generated by the elimination of the fucose residue from the former ion, suggest a linkage to O-3 of the GlcNAc residue and therefore a Le x structure The signal at m/z 1021.6 [NeuAc-Hex-(dHex-HexNAc) + Na] again clearly indicates the presence of a peripheral fucose as shown in the fragmentation scheme This is confirmed by the detection of a signal at m/z 316.2, which is typical for an unfucosylated proximal GlcNAc-ol residue The presence of very small amounts of an isomeric carbohydrate structure including a proximal fucose is indicated by the weak signal at m/z 490.3 characteristic for this structural feature It should be noted that CHO cells predominantly produce proximally fucosylated structures and very small amounts of structures with an additional fucose linked a1 fi 2 to the galactose residue of an acetyllactosamine antenna (LeH-motif).