Characterization of N-linked oligosaccharides attached to recombinant human antithrombin expressed...

YeastYeast 2002; 19: 1191–1202.Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/yea.914

Research Article

Characterization of N-linked oligosaccharidesattached to recombinant human antithrombinexpressed in the yeast Pichia pastorisMasaaki Hirose*, Shoju Kameyama and Hideyuki OhiProtein Research Laboratory, Pharmaceutical Research Division, Mitsubishi Pharma Corporation, Osaka, Japan

*Correspondence to:Masaaki Hirose, ProteinResearch Laboratory, MitsubishiPharma Corporation, 2-25-1,Shodai-Ohtani, Hirakata, Osaka573-1153, Japan.E-mail: [email protected]

Received: 19 April 2002Accepted: 22 July 2002

AbstractWe studied the structures of four N-linked oligosaccharide chains of the recombi-nant human antithrombin (rAT) expressed in the yeast Pichia pastoris. rAT was fullyglycosylated at Asn 96 and Asn 155, whereas the glycosylation on Asn 135 and Asn192 was partial. The glycosylation level on Asn 135 was only 12% and this reductionis assumed to be one of the reasons for a higher heparin-binding affinity of rATthan plasma-derived human antithrombin (pAT). In order to determine the sizes andelectrostatic charges of the N-linked oligosaccharides, rAT was treated with PNGaseF, and the reduced ends were labelled by pyridylamination followed by analysisusing anion exchange and amide adsorption columns. The N-linked oligosaccharideswere 78% neutral and 22% phosphomannosylated. The neutral oligosaccharides werethought to be Man9 – 12GlcNAc2 as their major components. The phosphomannosy-lated oligosaccharides were then subjected to mild acid hydrolysis and/or digestionwith alkaline phosphatase, and their charge shifts were analysed by the affinity to ananion exchange column. Among phosphomannosylated oligosaccharides, monophos-phate diester type was predominant, whereas negatively charged diphosphate diesterand monophosphate monoester types were minor components. The mannose residuesat the non-reducing end(s) of Man9 – 12GlcNAc2 were phosphomannosylated or phos-phorylated and these are the major components. Because rAT is less negativelycharged than pAT, which has disialyl biantennary N-glycans, it might be less repul-sive to pentasaccharide-bearing anticoagulantly active heparan sulphate proteoglycanmolecules exposed on the surface of the damaged vascular vessels. Copyright 2002John Wiley & Sons, Ltd.

Keywords: antithrombin; Pichia pastoris; N-glycosylation; oligosaccharide; phos-phomannosylation

Introduction

The methylotrophic yeast Pichia pastoris has beenwidely used as a host for the high expression offoreign genes (Cereghino and Cregg, 2000). As thenumbers of the reports of foreign gene expressionusing P. pastoris increases, interests in the struc-tures of oligosaccharides attached to the polypep-tide chains produced by this yeast have beenraised. In the secretion pathway of eukaryotic cells,oligosaccharide chains attach to the Asn residuesin the consensus sequence, Asn–X–Ser/Thr (X �=

Pro) (Kornfeld and Kornfeld, 1985). The presenceof oligosaccharide chains, and their length andstructures, may contribute to the functional activi-ties of the expressed proteins.

Although yeast cells possess a N-glycosylationmechanism similar to that of mammalian cells,the structures of oligosaccharides are different:yeast cells do not produce complex or hybridtypes of oligosaccharides, which are often foundin mammalian glycoproteins. In Saccharomycescerevisiae, glycoproteins with high-mannose typeoligosaccharides are synthesized as follows. First,

Copyright 2002 John Wiley & Sons, Ltd.

1192 M. Hirose, S. Kameyama and H. Ohi

Glc3Man9GlcNAc2 is transferred to polypeptidechains in the endoplasmic reticulum, as observedin mammalian cells, and several sugar residuesare trimmed to become a core structure,Man8GlcNAc2. Finally, mannose residues areadditionally attached to the core chain in theGolgi apparatus (Herscovics and Orlean, 1993;Kukuruzinska et al., 1987). In some cases,30–150 mannose residues are attached to anouter chain (Nagasu et al., 1992). In addition,phosphomannosylation often occurs at the non-reducing end of the core and outer chains(Ballou, 1990). Recombinant proteins with suchglycosylation are unsuitable for therapeutic usein humans because hypermannosylation andphosphomannosylation on oligosaccharide chainsmay cause the development of neo-antibodies(Ballou, 1982, 1990).

It is known that P. pastoris-derived recombi-nant glycoproteins are also synthesized with high-mannose type oligosaccharides, but the sizes ofoligosaccharides are shorter than those derivedfrom S. cerevisiae (Bretthauer and Castellino,1999; Cereghino and Cregg, 2000; Grinna andTschopp, 1989; Montesino et al., 1998). In mostcases, P. pastoris produces high-mannose typeoligosaccharides with 8–18 mannose residues asthe major components, e.g. Man8–14GlcNAc2 in S.cerevisiae invertase (Grinna and Tschopp, 1989),and Man9–12GlcNAc2 in the kringle 2 domainof tissue-type plasminogen activator (Miele et al.,1997b). In a few cases, hypermannosylation isobserved in HIV-1 envelope glycoprotein gp120(Scorer et al., 1993), neuraminidase of A/Victoria/3/75influenza virus (Martinet et al., 1997) and α-N-acetylgalactosaminidase (Zhu et al., 1998). Asthe minor components, phosphomannosylation isobserved in a few P. pastoris-derived recombinantproteins, e.g. one-third of the N-linked oligosac-charides are negatively charged in S. cerevisiaeinvertase (Grinna and Tschopp, 1989), and 20% ofoligosaccharides on the kringle 2 domain of humantissue-type plasminogen activator are phosphoman-nosylated (Miele et al., 1997a). On the other hand,Mucor pusillus aspartic proteinase is phospho-mannosylated, but other five glycoproteins anal-ysed in the same report did not contain phospho-mannosylated oligosaccharides (Montesino et al.,1998). From these reports, it is considered that thesize of N-linked oligosaccharides derived from P.pastoris is relatively small, and the possibilities

of hypermannosylation and phosphomannosylationare low. However, the mode of glycosylation isdependent on the nature of the polypeptide chainsof the expressed proteins. Therefore, the structuresof oligosaccharide chains should be analysed toevaluate the biological activities of each expressedglycoprotein.

Plasma-derived human antithrombin (pAT) is a58 kDa glycoprotein circulating in blood at a con-centration of 125 mg/l (Murano et al., 1980), play-ing an important role in haemostasis in associa-tion with pentasaccharide-bearing anticoagulantlyactive heparan sulphate proteoglycan (HSPG)molecules (de Agostini et al., 1990; Rosenberget al., 2000). pAT belongs to the serine proteinaseinhibitor (serpin) family (Carrell et al., 1991), andinhibits several activated coagulation factors, suchas thrombin and factor Xa. Its inhibitory activityis strongly enhanced (∼1000-fold) in the presenceof heparin (Rosenberg and Damus, 1973). It con-tains 15% carbohydrate with a biantennary com-plex type (Franzen et al., 1980; Mizuochi et al.,1980) of oligosaccharide chains attached to fourAsn residues at positions 96, 135, 155 and 192(Bock et al., 1982). Two isoforms of pAT arepresent (Peterson and Blackburn, 1985): the α-isoform is fully glycosylated and the β-isoformlacks the glycosylation at Asn 135. The β-isoformhas a higher heparin-binding affinity than the α-isoform.

pAT has been used for the treatment of dissem-inated intravascular coagulation and hereditary ATdeficiency (Schipper et al., 1978; Samson et al.,1984). In the past, recombinant human antithrom-bin (rAT) has been expressed in several host cells:S. cerevisiae (Broker et al., 1987), Schizosaccha-romyces pombe (Broker et al., 1987), insect cells(Gillespie et al., 1991), COS-1 cells (Stephenset al., 1987; Wasley et al., 1987), Chinese hamsterovary cells (Wasley et al., 1987), baby hamster kid-ney cells (Fan et al., 1993), and transgenic goat intomilk (Edmunds et al., 1998). In order to providerAT as a pharmaceutical reagent, use of a high levelof yeast expression systems is desirable. Accord-ingly, rAT has been expressed in S. cerevisiae andSz. pombe but unfortunately the yeast-derived prod-ucts had only 10% of the pAT heparin co-factoractivity (Broker et al., 1987).

We have produced rAT in P. pastoris with asignificantly higher expression level than thosereported with the above yeasts (Mochizuki et al.,

Copyright 2002 John Wiley & Sons, Ltd. Yeast 2002; 19: 1191–1202.

N-linked oligosaccharides on rAT 1193

2001). The inhibitory activity of P. pastoris-derived rAT was half that of pAT, while theheparin-binding affinity was 10-fold higher. It hasan apparent molecular weight of 56 kDa, indicat-ing that the size of the carbohydrate chains ofrAT is smaller than of pAT. The major compo-nents of the carbohydrate chains are N-linked high-mannose type oligosaccharides (Mochizuki et al.,2001). Since the glycosylation was shown to affectthe heparin-binding affinity of rAT (Bjork et al.,1992; Ersdal-Badju et al., 1995), we thought thatthe higher binding affinity of rAT derived from P.pastoris is also caused by the unique glycosyla-tion in this yeast. We have extended our study onthe structure of P. pastoris-derived rAT; here wereport the glycosylation level at the four Asn sites,the sizes of the major high-mannose type N-linkedoligosaccharides and the presence of minor phos-phate ester oligosaccharides.

Materials and methods

Materials

P. pastoris-derived rAT was prepared as describedpreviously (Mochizuki et al., 2001). Briefly, it wasexpressed under the control of the mAOX2 pro-moter, and purified by chromatography on heparin-affinity column followed by hydroxyapatite andcation exchange columns. Finally, contaminantswith molecular weights higher than 60 kDa wereremoved by gel permeation chromatography ona Sephacryl S-200 HR column (Amersham Bio-sciences). The purity of the purified protein wasover 99%.

Endoproteinase Asp–N was from Roche Diag-nostics. Dithiothreitol (DTT), monoiodoacetic acidand phenylisothiocyanate (PITC) were from WakoPure Chemicals Industries Ltd. E. coli C75 alka-line phosphatase, PNGase F and pyridylaminated(PA)-oligosaccharides were from Takara Bio, Inc.(Shiga, Japan).

Methods

Chemical modification

S -Carboxymethylation of rAT was conducted asfollows: rAT was reduced with 20-fold molarexcess of DTT (based on the contents of disulphide

bonds) in 0.5 M Tris–HCl, pH 8.0, 6 M guani-dine–HCl, 10 mM EDTA at 50 ◦C for 4 h. Afterthe reaction mixture had cooled to room tempera-ture, it was mixed with monoiodoacetic acid (twomolar equivalent to DTT) resolved in 0.1 N NaOH,and incubated for 20 min at room temperature indark. It was dialysed in 3% acetic acid, and thenin 50 mM Tris–HCl, pH 7.8, with three time bufferchanges. Mild acid hydrolysis was conducted in0.1 N HCl at 100 ◦C for 30 min (Thieme and Bal-lou, 1971). Reduced ends of the oligosaccharidesproduced by PNGase F digestion were labelledwith 2-aminopyridine using a pyridylamination kit(Takara Bio, Inc.) according to the manufacturer’sprotocol (Hase et al., 1984). To remove excessreagents, PA-oligosaccharides were adsorbed to acellulose cartridge (Takara Bio, Inc.) equilibratedwith 67% n-butanol/17% ethanol and eluted with50% ethanol, as instructed by the manufacturer’sprotocol. The solvents were dried out.

Enzymatic digestion

S -Carboxymethylated rAT was digested with endo-proteinase Asp–N (substrate/enzyme molar ratio =100) in 0.1 M Tris–HCl, pH 7.8, at 37 ◦C for 16 h.The phosphate groups attached to the non-reducingends of PA-oligosaccharides were removed bytreatment with 1 unit alkaline phosphatase in50 mM Tris–HCl, pH 9.0, 1 mM MgCl2 at 65 ◦Cfor 3 h (Karson and Ballou, 1978). To release N-linked oligosaccharides, 0.25 mg rAT was digestedwith 1 mU PNGase F in 0.2 M Tris–HCl, pH 8.6,0.2% SDS, 1% Nonidet P-40 at 37 ◦C for 16 h.After confirming the complete conversion of rATto a 49 kDa core protein by SDS–PAGE analysis,the released oligosaccharides were purified usingthe cellulose cartridge as described above.

Other chemical analyses

For amino sugar analysis, an aliquot of peptidesamples was transferred to a glass tube, driedunder vacuum and hydrolysed at 110 ◦C for 4 hin the gas phase of 4 N HCl (Cefalu et al., 1991).The hydrolysates were derivatized with PITC,and phenylthiocarbamoyl-glucosamine was anal-ysed using PICO-TAG (Waters Co.), accordingthe manufacturer’s instructions. N-terminal amino



acid sequence analysis was performed on a gas-phase protein sequencer, Model PPSQ-10, Shi-madzu (Kyoto, Japan) equipped with an onlinephenylthiohydantoin (PTH) analyser.

Separation of peptides by HPLC

The separation of peptides was conducted witha Cosmosil 5C18-AR300 column (Nacalai TesqueInc., Kyoto, Japan, 4.6 × 150 mm) using a lineargradient composed of solvents A and B (0–80%solvent B over 50 min at a flow rate of 1.0 ml/minat room temperature; while solvent A, 0.1% TFA;solvent B, 0.1% TFA in acetonitrile). Peptideswere detected by monitoring the absorbance at215 nm and they were collected manually. Neu-tral and acidic PA-oligosaccharides were anal-ysed using an amide adsorption column, TSKgelAmide-80 (Tosoh, Tokyo, Japan, 4.6 × 250 mm)(Tomiya et al., 1988), using a linear gradient of0–100% solvent D over 50 min with a flow rateof 1.0 ml/min at 40 ◦C (solvent C, 3% aceticacid/65% acetonitrile adjusted to pH 7.3 with tri-ethylamine; solvent D, 3% acetic acid/50% ace-tonitrile adjusted to pH 7.3 with triethylamine).PA-oligosaccharides were separated on an anionexchange column (TSKgel DEAE-5PW column,7.5 × 75 mm, Tosoh) (Nakagawa et al., 1995).Elution was conducted at a flow rate of 1.0 ml/minat 30 ◦C using solvents E and F (solvent E, 10%acetonitrile adjusted to pH 9.5 with triethylamine;solvent F, 3% acetic acid/10% acetonitrile adjustedto pH 7.3 with triethylamine). After injection ofPA-oligosaccharides, solvent F was kept 0% for10 min, and then PA-oligosaccharides were elutedwith a linear gradient of 0–20% solvent F for40 min. Each fraction was collected manually.In both HPLC systems, PA-oligosaccharides weredetected by monitoring fluorescence (excitation at320 nm and emission at 400 nm).

Results

Glycosylation of rAT

pAT is a glycoprotein with four carbohydratechains linked to Asn residues, 96, 135, 155 and192 (Bock et al., 1982). We first studied whetherthese sites are glycosylated in rAT expressed in P.pastoris. S -Carboxymethylated rAT was digested

with endoproteinase Asp–N. The resulting peptideswere separated on RP–HPLC (Figure 1) and eachwas subjected to the detection of amino sugar(glucosamine) and to sequence analysis. Amonga number of fractions, D1, D2, D3, D6, D7 andD8 were glucosamine-positive, indicating that thesefractions contained N-glycosylated peptides.

Some of the peaks, especially small peaks, con-tained several peptides, but major peptides wereunequivocally identified by sequence analysis. Thesequences of D1–D8 included one of the four Asnresidues located at the N-glycosylation sites. D6and D8 contained peptides starting from Asp 74and ending Asn 96, although PTH–Asn 96 wasnot detected due to its glycosylation. No otherfractions contained peptides that included Asn 96.These results show that Asn 96 was fully glyco-sylated. Similarly, sequence analysis of D1 andD2 found that these fractions contained peptidesfrom Asp 149 to Gln 159 including Asn 155. Bothfractions were glucosamine-positive and PTH–Asn155 was not detected. Since no other fractions con-tained peptides that included Asn 155, this site wasalso fully glycosylated. The difference between thepairs of these two glycopeptides is attributed tothe different chain length and/or heterogeneity oftheir carbohydrate attached to these Asn residues,as described below.

Two fractions D5 and D7 contain peptides, Asp117–Gly 148, which includes Asn 135. PTH–Asn135 was not detected in D7, but it was detectedin D5. Fraction D7 was glucosamine-positive butD5 was negative. These results indicated that Asn135 was partially glycosylated. To determine thelevel of glycosylation on Asn 135, the amountsof these two peptides in the HPLC fractions werecalculated from the yields of PTH–Ile at the thirdcycle in the sequence analysis. PTH–Ile was usedfor the quantitation because PTH-derivatives ofhydrophobic amino acids are stable in sequenceanalysis and quantitatively recovered. Fractions D7and D5 contained 10.5 pmol and 74.0 pmol ofpeptides, respectively, indicating that only 12% ofAsn 135 is glycosylated. Two fractions D3 and D4contain peptides, Glu 180–Thr 199, including Asn192. Fraction D3 was glucosamine-positive but D4was negative. PTH–Asn 192 was detected in D4but it was not detected in D3 by the sequenceanalysis. The yields of PTH–Ala at the fifth cyclewere 117.4 pmol in D3 and 13.3 pmol in D4. These



Figure 1. Separation of the peptides produced by endoproteinase Asp-N digest of rAT. S-Carboxymethylated rAT wasdigested with endoproteinase Asp-N and the resulting peptides were separated on a Cosmosil 5C18-AR300 column(Nacalai tesque, 4.6 × 150 mm) using a linear gradient elution, as indicated in Methods. Fractions that contain possibleN-glycosylation sites were designated as D1–D8

Table 1. Summary of N-glycosylation of rAT

N-Glycosylation sitea 96 135 155 192

Peptide fragmentb D6 D8 D5 D7 D1 D2 D3 D4Positiona Asp 74–Asn 96 Asp 117–Gly 148 Asp 149–Gln 159 Glu 180–Thr 199Glucosaminec + + − + + + + −N-glycosylation (%) 100 12 100 90

a N-glycosylation sites are the residues numbered from the N-terminus of AT.b Peptide fragment numbers are taken from Figure 1.c + and − designate glucosamine-test positive and negative, respectively.

results indicate that the level of glycosylation atAsn 192 is 90%.

From the above data, we concluded that rAT wascompletely glycosylated on Asn 96 and Asn 155,while glycosylation at Asn 135 and Asn 192 arepartial with 90% at Asn 192 and only 12% at Asn135 (Table 1).

In the course of the sequencing analysis ofAsp–N peptides, partial O-glycosylation sites thatwere determined in our previous work (Mochizuki

et al., 2001) were also confirmed to be Ser 3, Thr9 and Thr 386.

The size of the N-linked oligosaccharides ofrAT

rAT was digested with PNGase F, and thereducing ends of the released oligosaccharideswere labelled by pyridylamination, as describedin Methods. The resulting PA-oligosaccharides



Figure 2. Size distribution analysis of PA-oligosaccharides by an amide adsorption column. rAT was digested withPNGase F and reduced ends of oligosaccharides were labelled by pyridylamination. PA-oligosaccharides were appliedto TSKgel Amide-80 column (Tosoh, 4.6 × 250 mm) and eluted by a linear gradient elution, as indicated in Methods.PA-oligosaccharides were detected by monitoring fluorescence. Elution time was converted to DU standardized by theretention time of authentic PA-glucose oligomers

were subjected to HPLC equipped with an amideadsorption column. By detecting fluorescence, fourmajor discrete peaks, M9, M10, M11 and M12,were found (Figure 2), indicating that these fourcarbohydrate chains are different in size. Theretention time is converted to dextrose units(DU), which was standardized by commercialPA-glucose oligomers (Takara Bio, Inc.). Usingthe same column as we used in this study,the DUs of five PA-high-mannose type oligosac-charides, Man5GlcNAc2-PA, Man6GlcNAc2-PA,Man7GlcNAc2-PA, Man8GlcNAc2-PA and Man9

GlcNAc2-PA, were reported to be 6.2, 7.1, 8.0,9.0 and 9.7, respectively (Tomiya et al., 1988). Inthis definition, the addition of one mannose moietyto carbohydrate chains increases close to 0.9 DU.Peak M9 eluted at 9.7 DU, which is exactly thesame position as reported for Man9GlcNAc2-PA.

P. pastoris attaches high-mannose type oligosac-charides to polypeptide backbones but it does notproduce complex-type sugar chains (Cereghino andCregg, 2000). Thus, we concluded that the structureof M9 is Man9GlcNAc2-PA. Three other peaks,M10 at 10.7 DU, M11 at 11.5 DU and M12at 12.4 DU, eluted following M9 by an intervalof about 0.9 DU. Therefore, their structures arethought to be Man10GlcNAc2-PA, Man11GlcNAc2-PA and Man12GlcNAc2-PA. These are the sizes ofthe four major structures of the carbohydrate chainsattached to rAT. Several minor peaks appearingbetween and after the major peaks contained mod-ified forms of the four major oligosaccharides, asdiscussed below.

Neutral and acidic oligosaccharides of rATPA-sialyl oligosaccharides were separated on ananion exchange (DEAE) column by the degrees



Figure 3. Charge analysis of PA-oligosaccharides by an anion exchange column. PA-oligosaccharides were analysedusing a TSKgel DEAE-5PW column (Tosoh, 7.5 × 75 mm). Elution position of the authentic PA-sialyl oligosaccharidesPA sugar 021, 023, 024 and 025 (Takara Bio, Inc.) were indicated by arrows with numbers. PA sugar 021; monosialyltriantennary sugar, 023; disialyl triantennary sugar, 024; trisialyl triantennary sugar and 025; tetrasialyl triantennary sugar.(A) PA-oligosaccharides were separated into the neutral fraction N and the acidic fractions A1 and A2. (B) Fraction A1 wastreated with mild acid hydrolysis, and the converted oligosaccharides were analysed on the same column. A1 shifted to theposition shown by a broken arrow, and it was designated as fraction A1′. (C) A1′ was digested with alkaline phosphataseand analysed on the same column. The shifted fraction was designated fraction A1′′, as indicated

of negative charges (Nakagawa et al., 1995). P.pastoris is known to produce phosphomannosyloligosaccharide but sialyl-oligosaccharides (Grinnaand Tschopp, 1989; Miele et al., 1997a). PA-oligosaccharides were prepared from rAT as aboveand applied to the DEAE column, as detailed inMethods. The major portion did not bind to thecolumn and appeared in the non-adsorbed frac-tion, fraction N (Figure 3A), showing that these areneutral oligosaccharides. Fraction N was appliedto the amide adsorption column and four peaks,M9, M10, M11 and M12, appeared (Figure 4A).The DUs of these peaks are the same as the fourmajor peaks shown in Figure 2. This result is con-sistent with the proposed neutral structures of the

four major oligosaccharides as mentioned above.The minor portion contained acidic oligosaccha-rides and eluted by the pH gradient system in twofractions, A1 and A2 (Figure 3A). The proportionof 78% neutral and 22% acidic oligosaccharideswas calculated from the dimensions of the peak(Figure 4).

The elution position of A1 is close to that of PA-monosialyl triantennary oligosaccharide, and A2eluted at a similar position to PA-disialyl trianten-nary oligosaccharide, indicating A1 has one and A2has two net negative charges. The acidic fractionA1 was also applied to the amide adsorption col-umn and it gave four peaks (Figure 4B) with appar-ently the same DUs as the minor peaks observed in



Figure 4. Size distribution analysis of the neutral and acidicPA-oligosaccharides using an amide adsorption column.PA-oligosaccharides fractionated with the DEAE columnshown in Figure 3A were analysed using a TSKgel Amide-80column (Tosoh, 4.6 × 250 mm) using the same conditionas Figure 2. Analysis of the neutral PA-oligosaccharides,fraction N (A), and the acidic PA-oligosaccharides, fractionA1 (B) were shown. The four major peaks were designatedM9, M10, M11 and M12 (A)

Figure 2, indicating that most of the minor peaks(Figure 2) obtained from total PA-oligosaccharidesare acidic.

Monophosphate diester type oligosaccharidesof rAT

The presence of phosphomannosylation on high-mannose type oligosaccharides in the glycoproteinsof yeast cell wall mannan was reported (Ballou,1990; Thieme and Ballou, 1971; Trimble et al.,1991). In these studies, mild acid hydrolysis wasused to convert monophosphate diester type into

monophosphate monoester type (Thieme and Bal-lou, 1971). We took the same approach to findphosphomannosylation in the acidic fractions. Theacidic fraction A1 was subjected to mild acidhydrolysis, and the resulting product was appliedto the DEAE column. The product, A1′, waseluted in the similar position to the acidic A2fraction (Figure 3B). The acidic fraction A1 wasalso treated with α-mannosidase (jack bean), andit shifted to the same position as A1′ (data notshown), indicating that A1 contained high-mannosetype oligosaccharide diesterified with mannose atthe non-reducing end. Following the hydrolysiswith mild acid, fraction A1′ was digested withalkaline phosphatase and re-applied to the column.The products, A1′′, did not bind to the column andappeared in the non-adsorbed fraction (Figure 3C).This result shows that phosphate group attachedto the carbohydrate chains before the treatments.When A1 was treated with alkaline phosphatasewithout prior acid treatment, it eluted in the orig-inal position (data not shown). This indicates thatthe phosphate group is not linked to the end ofthe chain, since phosphatase is known to cleaveonly the terminal phosphate. Fractions A1, A1′and A1′′ were analysed on the amide adsorptioncolumn and several peaks were eluted from eachfraction (Figure 5). The four peaks, M9′, M10′,M11′ and M12′ from A1′′ were eluted later thanthose from A1′ by about 0.3 DU, while a groupof the peaks from A1′ eluted earlier than thosefrom A1 by about 0.6 DU. The elution positionsof M9′, M10′, M11′ and M12′ are the same asthose of the neutral fraction N (see Figure 4A).Thus, these four are the same as M9, M10, M11and M12. From these results, we conclude that thestructures of PA-oligosaccharides in the acidic frac-tion A1 are monophosphate diester types, Man–P-Man9–12GlcNAc2, with one negative charge. Bymild acid hydrolysis, they were converted to A1′,P–Man9–12GlcNAc2, with two negative charges,shifting its elution position on the DEAE col-umn toward the end of the pH gradient, whichis shown by the DU shift by 0.6 on the amideadsorption column. P–Man9–12GlcNAc2 in A1′was then converted to neutral oligosaccharides,Man9–12GlcNAc2, by the phosphatase treatmentwith the loss of phosphate moiety, which is shownby the DU shift by 0.3 on the amide adsorp-tion column.



Figure 5. Size distribution analysis of the acidic oligosac-charides in the fraction A1. Fraction A1 (Figure 3A)was analysed using a TSKgel Amide-80 column (Tosoh,4.6 × 250 mm) with the same elution condition as Figure 2(A1). Fraction A1 was treated with mild acid hydrolysis andthe product, fraction A1′, was analysed using the same col-umn (A1′). A1′ was then digested with alkaline phosphataseand the product, fraction A1′′, was analysed using the samecolumn. Four major peaks in A1′′, designated M9′, M10′,M11′ and M12′, emerged (A1′′)

Monophosphate monoester and diphosphatediester oligosaccharides of rAT

The acidic fraction A2 (Figure 3A) is actuallycomposed of two types of oligosaccharides, A2aand A2b, as shown in Figure 6A. By the treat-ment of A2 with mild acid, fraction A2a shiftedto A2a′, which is near the position of tetrasialyloligosaccharide on the DEAE column (PA sugar025, Takara Bio, Inc.), but fraction A2b did notshift (Figure 6B). When the fraction A2 (mixtureof A2a and A2b) was digested with alkaline phos-phatase without prior acid hydrolysis, A2a did notshift, but A2b moved to the non-adsorbed fraction(Figure 6C). When the fraction A2 was first treatedwith mild acid and digested with phosphatase,both were converted to neutral oligosaccharides

and moved to the A2′′ fraction (Figure 6D). Fourpeaks were separated from A2′′ on the amideadsorption column at the same DUs as M9–M12or M9′ –M12′ (data not shown), indicating thatA2′′ is a mixture of four neutral oligosaccharides,Man9–12GlcNAc2. These results suggest that A2acontains diphosphate diester type oligosaccharides,(Manx –P-)2Man9–12GlcNAc2. The fraction A2bwas shifted to the neutral position with phosphatasedigestion without prior acid treatment. Thus, it con-tains monophosphate monoester oligosaccharides,P–Man9–12GlcNAc2. The retention time and theshape of A2b was similar to A1′ in Figure 3B,also showing that A2b contains monophosphatemonoester oligosaccharides.

Discussion

Recently, we expressed rAT in P. pastoris and stud-ied its biochemical properties: rAT has a weakerinhibitory activity (50%) than pAT due to the inter-ference of the inhibitor/enzyme binding by partialO-glycosylation near its active site, and rAT hasa 10-fold higher binding affinity to heparin thanpAT (Mochizuki et al., 2001). The present studysuggests that the higher binding affinity of rAT isattributed to the lack of glycosylation at Asn 135.This site is partially glycosylated in rAT with only12%. The glycosylation at this Asn is also partialin pAT (85–90%), and the molecule lacking in gly-cosylation at this site is called β-isoform (Petersonand Blackburn, 1985). rAT has been expressed inmammalian and insect cells, and the β-isoform ofrAT has a higher binding affinity than pAT regard-less of their host cells (Fan et al., 1993; Ersdal-Badju et al., 1995). The higher heparin-bindingaffinity of rAT is probably attributed to the absenceof glycosylation at Asn 135 of rAT. This sug-gests that the addition of the glycosylation on Asn135 causes the conformational change of AT todecrease the heparin-binding affinity. Another rea-son for the higher heparin-binding affinity of rATis that the oligosaccharides on rAT have fewernegative charges than pAT. Since HSPG exposedon the surface of the damaged vascular vesselsare negatively charged (Rosenberg et al., 2000;de Agostini et al., 1990), pAT are thought to bemore repulsive than rAT. The major oligosaccha-rides on pAT are disialyl biantennary type (Franzenet al., 1980; Mizuochi et al., 1980), while the major



Figure 6. Charge analysis of the oligosaccharides in the acidic fraction A2 by an anion exchange column. (A)PA-oligosaccharides, A2, were separated using a TSKgel DEAE-5PW column (Tosoh, 7.5 × 75 mm) with the sameelution condition as shown in Figure 3. Fraction A2 was separated into fractions A2a and A2b. (B) A2 was treated withmild acid hydrolysis and the products were analysed on the same column. A2a shifted to the position shown by a brokenarrow, and it was designated fraction A2a′. (C) A2 was digested with alkaline phosphatase and the products were analysedon the same column. A2b shifted to the position shown by a dotted arrow. (D) A2 was treated with mild acid hydrolysisfollowed by the digestion with alkaline phosphatase. The products were analysed on the same column. Shifted fractionswere designated fraction A2′′, as shown by arrows with broken and dotted lines

oligosaccharides of rAT are neutral high-mannosetype. The negative charges of rAT are provided byphosphomannosyl oligosaccharides present as theminor components (22%). If the phosphomannosyloligosaccharides are avoided, the heparin-bindingaffinity of rAT could be even higher.

It is said that three features of the glycosyla-tion occurring in S. cerevisiae hamper the practicaluse of rAT produced in this yeast: (a) hyperman-nosylation exceeding over 30 mannose residues inthe outer chain; (b) terminal α1-3 mannosylation;and (c) phosphomannosylation. The previous studyshowed that a small amount of hypermannosylatedform is present in rAT produced in P. pastoris.This problem can be solved by removing the largerhypermannosylated molecules by a gel permeation

chromatography, as shown in the previous paper(Mochizuki et al., 2001). α1,3-Mannosylation doesnot occur in P. pastoris because of the deficiencyof α1-3 mannosyltransferase (Trimble et al., 1991).We found approximately 20% of rAT is phospho-rylated form, which is not present in pAT. Thus,it may cause the development of neo-antibodieswhen it is intravenously administered. In order toavoid this possibility, there are several ways to pre-pare fewer phosphomannosylated glycoproteins inP. pastoris. The most effective way is to knock outthe gene(s) responsible for transphosphomannosy-lation. In S. cerevisiae the phosphomannosylation-related gene, MNN4, was identified, but the dis-ruption of the gene did not completely eliminatethe formation of phosphorylated oligosaccharides



in this yeast (Odani et al., 1996). We are nowtrying to isolate and characterize a gene respon-sible for phosphomannosylation in P. pastoris tosee whether the disruption of the gene can helpeliminate the phosphomannosylated oligosaccha-rides produced in this yeast. Since partial O-glycosylation at Thr 386 reduces the inhibitoryactivity of rAT produced in P. pastoris (Mochizukiet al., 2001), an alanine-charged mutation at thissite could avoid the problem. In combination withpreparation of fewer phosphomannosylated rAT inP. pastoris, Ala 386-rAT may be a more efficientanticoagulant reagent with high affinity to HSPG.

Acknowledgements

The authors thank our colleagues Mr Yoshio Tsuda and DrNobuaki Hamato for helpful discussions. We thank Profes-sor Kazuo Fujikawa (University of Washington, Seattle) forcritical reading of the manuscript.

References

Ballou CE. 1990. Isolation, characterization, and properties ofSaccharomyces cerevisiae mnn mutants with nonconditionalprotein glycosylation defects. Methods Enzymol 185: 440–470.

Ballou CE. 1982. Yeast cell wall and cell surface. In The MolecularBiology of the Yeast Saccharomyces cerevisiae: Metabolism andGene Expression, Strathern JN, Jones EW, Broach JR (eds).Cold Spring Harbor Laboratory Press: New York; 335–360.

Bjork I, Ylinenjarvi K, Olson ST, Hermentin P, Conradt HS,Zettlmeissl G. 1992. Decreased affinity of recombinantantithrombin for heparin due to increased glycosylation.Biochem J 286: 793–800.

Bock SC, Wion KL, Vehar GA, Lawn RM. 1982. Cloning andexpression of the cDNA for human antithrombin III. NucleicAcids Res 10: 8113–8125.

Bretthauer RK, Castellino FJ. 1999. Review: glycosylation ofPichia pastoris-derived proteins. Biotechnol Appl Biochem 30:193–200.

Broker M, Ragg H, Karges HE. 1987. Expression of humanantithrombin III in Saccharomyces cerevisiae and Schizosaccha-romyces pombe. Biochem Biophys Acta 908: 203–213.

Carrell RW, Evans DL, Stein PE. 1991. Mobile reactive center ofserpins and the control of thrombosis. Nature 353: 576–578.

Cefalu WT, Bell-Farrow A, Wang ZQ, Ralapati S. 1991. Determi-nation of furosine in biomedical samples employing an improvedhydrolysis and high-performance liquid chromatographic tech-nique. Carbohydr Res 215(1): 117–125.

Cereghino JL, Cregg JM. 2000. Heterologous protein expressionin the methylotrophic yeast Pichia pastoris . FEMS MicrobiolRev 24: 45–66.

de Agostini AI, Watkins SC, Slayter HS, Youssoufian H, Rosen-berg RD. 1990. Localization of anticoagulantly active heparansulphate proteoglycans in vascular endothelium: antithrombinbinding on cultured endothelial cells and perfused rat aorta. JCell Biol 111: 1293–1304.

Edmunds T, Van Patten SM, Pollock J, et al. 1998. Transgenicallyproduced human antithrombin: structural and functionalcomparison to human plasma-derived antithrombin. Blood91(12): 4561–4571.

Ersdal-Badju E, Lu A, Peng X, et al. 1995. Elimination ofglycosylation heterogeneity affecting heparin affinity ofrecombinant human antithrombin III by expression of a β-like variant in baculovirus-infected insect cells. Biochem J 310:323–330.

Fan B, Crews BC, Turko IV, Choay J, Zettlmeissl G, Gettins P.1993. Heterogeneity of recombinant human antithrombin IIIexpressed in baby hamster kidney cells. J Biol Chem 268(23):17 588–17 596.

Franzen LE, Svensson S, Larm O. 1980. Structural studies on thecarbohydrate portion of human antithrombin III. J Biol Chem255(11): 5090–5093.

Gillespie LS, Hillesland KK, Knauer DJ. 1991. Expression ofbiologically active human antithrombin III by recombinantbaculovirus in Spodoptera frugiperda cells. J Biol Chem 266(6):3995–4001.

Grinna LS, Tschopp JF. 1989. Size distribution and generalstructural features of N-linked oligosaccharides from themethylotrophic yeast, Pichia pastoris. Yeast 5: 107–115.

Hase S, Ibuki T, Ikenaka T. 1984. Reexamination of the pyridy-lamination used for fluorescence labeling of oligosaccharidesand its application to glycoproteins. J Biochem (Tokyo) 95(1):197–203.

Herscovics A, Orlean P. 1993. Glycoprotein biosynthesis in yeast.FASEB J 7: 540–550.

Karson EM, Ballou CE. 1978. Biosynthesis of yeast mannan. JBiol Chem 253(18): 6484–6492.

Kornfeld R, Kornfeld S. 1985. Assembly of asparagine-linkedoligosaccharides. Ann Rev Biochem 54: 631–664.

Kukuruzinska MA, Bergh MLE, Jackson BJ. 1987. Proteinglycosylation in yeast. Ann Rev Biochem 56: 915–944.

Martinet W, Saelens X, Deroo T, et al. 1997. Protection of miceagainst a lethal influenza challenge by immunization with yeast-derived recombinant influenza neuraminidase. Eur J Biochem247: 332–338.

Miele RG, Castellino FJ, Bretthauer RK. 1997a. Characterizationof the acidic oligosaccharides assembled on the Pichia pastoris-expressed recombinant kringle 2 domain of human tissue-typeplasminogen activator. Biotechnol Appl Biochem 26: 79–83.

Miele RG, Nilsen SL, Brito T, Bretthauer RK, Castellino FJ.1997b. Glycosylation properties of the Pichia pastoris-expressedrecombinant kringle 2 domain of tissue-type plasminogenactivator. Biotechnol Appl Biochem 25: 151–157.

Mizuochi T, Fujii J, Kurachi K, Kobata A. 1980. Structuralstudies of the carbohydrate moiety of human antithrombin III.Arch Biochem Biophys 203(1): 458–465.

Mochizuki S, Hamato N, Hirose M, et al. 2001. Expression andcharacterization of recombinant human antithrombin III inPichia pastoris. Protein Expr Purif 23(1): 55–65.

Montesino R, Garcıa R, Quintero O, Cremata JA. 1998. Variationin N-linked oligosaccharide structures on heterologous proteinssecreted by the methylotrophic yeast Pichia pastoris . ProteinExpr Purif 14: 197–207.

Murano G, Williams L, Miller-Andersson M, Aronson DL,King C. 1980. Some properties of antithrombin-III and itsconcentration in human plasma. Thromb Res 18(1–2): 259–262.



Nagasu T, Shimma Y, Nakanishi Y, et al. 1992. Isolation ofnew temperature-sensitive mutants of Saccharomyces cerevisiaedeficient in mannose outer chain elongation. Yeast 8(7):535–547.

Nakagawa H, Kawamura Y, Kato K, Shimada I, Arata Y, Taka-hashi N. 1995. Identification of neutral and sialyl N-linkedoligosaccharide structures from human serum glycoproteinsusing three kinds of high-performance liquid chromatography.Anal Biochem 226: 130–138.

Odani T, Shimma Y, Tanaka A, Jigami Y. 1996. Cloning andanalysis of the MNN4 gene required for phosphorylationof N-linked oligosaccharides in Saccharomyces cerevisiae.Glycobiology 6(8): 805–810.

Peterson CB, Blackburn MN. 1985. Isolation and characterizationof an antithrombin III variant with reduced carbohydrate contentand enhanced heparin binding. J Biol Chem 260(1): 610–615.

Rosenberg RD, Damus PS. 1973. The purification and mechanismof action of human antithrombin-heparin co-factor. J Biol Chem248(18): 6490–6505.

Rosenberg RD, Edelberg JM, Zhang L. 2000. The heparin/anti-thrombin system: a natural anticoagulant mechanism. InAntithrombin III and Heparin Co-factor II in Hemostasisand Thrombosis, 4th edn, Colman RW, Hirsh J, Marder VJ,Clowes AW, George JN (eds). Lippincott Williams & Wilkins:Philadelphia, PA; 712.

Samson D, Stirling Y, Woolf L, Howarth D, Seghatchian MJ, deChazal R. 1984. Management of planned pregnancy in a patientwith congenital antithrombin III deficiency. Br J Haematol56(2): 243–249.

Schipper HG, Jenkins CS, Kahle LH, ten Cate JW. 1978.Antithrombin-III transfusion in disseminated intravascularcoagulation. Lancet 1(8069): 854–856.

Scorer CA, Buckholz RG, Clare JJ, Romanos MA. 1993. Theintracellular production and secretion of HIV-1 envelopeprotein in the methylotrophic yeast Pichia pastoris . Gene 136:111–119.

Stephens AW, Siddiqui A, Hirs CHW. 1987. Expression offunctionally active human antithrombin III. Proc Natl Acad SciUSA 84: 3886–3890.

Thieme TR, Ballou CE. 1971. Nature of the phosphodiesterlinkage of the phosphomannan from the yeast Kloeckera brevis.Biochemistry 10(22): 4121–4129.

Tomiya N, Awaya J, Kurono M, Endo S, Arata Y, Takahashi N.1988. Analyses of N-linked oligosaccharides using a two-dimensional mapping technique. Anal Biochem 171: 73–90.

Trimble RB, Atkinson PH, Tschopp JF, Townsend RR, Maley F.1991. Structure of oligosaccharides on Saccharomyces SUC2invertase secreted by the methylotrophic yeast Pichia pastoris .J Biol Chem 266(34): 22 807–22 817.

Wasley LC, Atha DH, Bauer KA, Kaufman RJ. 1987. Expressionand characterization of human antithrombin III synthesized inmammalian cells. J Biol Chem 262(30): 14 766–14 772.

Zhu A, Wang Z-K, Beavis R. 1998. Structural studies of α-N-acetylgalactosaminidase: effect of glycosylation on the level ofexpression, secretion efficiency, and enzyme activity. Arch BiolBiophys 352(1): 1–8.


Characterization of N-linked oligosaccharides attached to recombinant human antithrombin expressed...

Documents

Transcript of Characterization of N-linked oligosaccharides attached to recombinant human antithrombin expressed...