Glycobiology

Glycobiology Advance Access originally published online on February 9, 2007
Glycobiology 2007 17(6):600-619; doi:10.1093/glycob/cwm015

This Article Abstract FREE Full Text (PDF) All Versions of this Article: 17/6/600    most recent cwm015v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (1) Request Permissions Disclaimer Google Scholar Articles by Jongen, S. P Articles by Kamerling, J. P Search for Related Content PubMed PubMed Citation Articles by Jongen, S. P Articles by Kamerling, J. P Social Bookmarking          What's this?

© The Author 2007. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org

N-glycans of recombinant human acid -glucosidase expressed in the milk of transgenic rabbits

Susanne P Jongen⁴, Gerrit J Gerwig⁴, Bas R Leeflang⁴, Kate Koles^2,4, Maurice LM Mannesse⁵, Patrick HC van Berkel^3,5, Frank R Pieper⁵, Marian A Kroos⁶, Arnold JJ Reuser⁶, Qun Zhou⁷, Xiaoying Jin⁷, Kate Zhang⁷, Tim Edmunds⁷ and Johannis P Kamerling^1,4

⁴ Bijvoet Center for Biomolecular Research, Department of Bio-Organic Chemistry, Utrecht University, Padualaan 8, NL-3584 CH Utrecht, The Netherlands
⁵ Pharming Technologies BV, P.O. Box 451, NL-2300 AL Leiden, The Netherlands
⁶ Department of Clinical Genetics, Erasmus MC, P.O. Box 1738, NL-3000 DR Rotterdam, The Netherlands
⁷ Genzyme Corporation, Framingham, MA 01701

---

¹ To whom correspondence should be addressed; e-mail: j.p.kamerling{at}chem.uu.nl

Received on October 19, 2006; revised on February 2, 2007; accepted on February 3, 2007

	Abstract

Top Abstract Introduction Results Discussion Materials and methods Conflict of interest statement References

 
Pompe disease is a lysosomal glycogen storage disorder characterized by acid -glucosidase (GAA) deficiency. More than 110 different pathogenic mutations in the gene encoding GAA have been observed. Patients with this disease are being treated by intravenous injection of recombinant forms of the enzyme. Focusing on recombinant approaches to produce the enzyme means that specific attention has to be paid to the generated glycosylation patterns. Here, human GAA was expressed in the mammary gland of transgenic rabbits. The N-linked glycans of recombinant human GAA (rhAGLU), isolated from the rabbit milk, were released by peptide-N⁴-(N-acetyl-ß-glucosaminyl)asparagine amidase F. The N-glycan pool was fractionated and purified into individual components by a combination of anion-exchange, normal-phase, and Sambucus nigra agglutinin-affinity chromatography. The structures of the components were analyzed by 500 MHz one-dimensional and 600 MHz cryo two-dimensional (total correlation spectroscopy [TOCSY] nuclear Overhauser enhancement spectroscopy) ¹H nuclear magnetic resonance spectroscopy, combined with two-dimensional ³¹P-filtered ¹H-¹H TOCSY spectroscopy, matrix-assisted laser desorption ionization time-of-flight mass spectrometry, and high-performance liquid chromatography (HPLC)-profiling of 2-aminobenzamide-labeled glycans combined with exoglycosidase digestions. The recombinant rabbit glycoprotein contained a broad array of different N-glycans, comprising oligomannose-, hybrid-, and complex-type structures. Part of the oligomannose-type glycans showed the presence of phospho-diester-bridged N-acetylglucosamine. For the complex-type glycans (partially) (2-6)-sialylated (nearly only N-acetylneuraminic acid) diantennary structures were found; part of the structures were (1-6)-core-fucosylated or (1-3)-fucosylated in the upper antenna (Lewis x). Using HPLC-mass spectrometry of glycopeptides, information was generated with respect to the site-specific location of the various glycans.

Key words: N-glycosylation / recombinant human acid -glucosidase / transgenic rabbit milk

	Introduction

Top Abstract Introduction Results Discussion Materials and methods Conflict of interest statement References

 
Pompe disease, also known as acid maltase deficiency, is a glycogen storage disease, GSD type II, and belongs to the family of inherited lysosomal storage disorders (Hirschhorn 2001). The disease is characterized by a major deficiency of acid -glucosidase (1,4--D-glucan glucohydrolase; GAA; EC 3.2.1.3 [EC] /20) and is caused by mutations in the GAA gene. Glycogen accumulates predominantly in heart, skeletal, and smooth muscle where it causes cellular damage. Loss of mobility and weakening of the respiratory function are the main clinical findings. Cardiac failure occurs in the most severe cases. Disease severity is related to the amount of residual GAA activity (Reuser et al. 1995). Infants born with a complete lack of enzyme activity have a maximum lifespan of 2 years (Van den Hout et al. 2003). The onset of symptoms in adults can be delayed till the sixth decade of life with only 5–25% of GAA activity.

Human GAA is synthesized as a 110-kDa single-chain precursor protein containing seven N-glycosylation sites (Asn-140, -233, -390, -470, -652, -882, and -925) (Hermans et al. 1993). Its transport from the endoplasmic reticulum via the Golgi network to the lysosomes is mediated by the mannose 6-phosphate receptor (MPR) (Hasilik and Neufeld 1980). In this process the lysosomal enzyme oligomannose-type N-glycans with mannose 6-phosphate groups are essential (Kaplan et al. 1977; Kornfeld 1986). In the late-endosomal/lysosomal compartments the precursor is proteolytically converted into a multipeptide complex (Moreland et al. 2005). A small amount of the precursor is transported to the plasma membrane, and secreted (Wisselaar et al. 1993).

Recently, enzyme replacement therapy by intravenous injection of GAA has demonstrated promising results in the treatment of Pompe disease (van den Hout et al. 2000; Winkel et al. 2004; Kishnani et al. 2006). As one of the possibilities for large-scale production of therapeutic enzyme, recombinant human acid -glucosidase (rhAGLU) has been produced in the milk of transgenic rabbits (up to 8 mg/ml) (Bijvoet et al. 1999). The transgene construct consisted of the human GAA gene cloned behind the bovine S1-casein gene promoter (Bijvoet et al. 1998). The purified rhAGLU has an apparent molecular mass of 110 kDa, comparable to the natural GAA precursor secreted in human urine (Oude Elferink et al. 1984). It has been tested in phase I and II trials in babies and older patients with positive results (van den Hout et al. 2000; van den Hout et al. 2001; Winkel et al. 2004).

At present the glycosylation machinery of the rabbit mammary gland is rather unexplored. Recent data have become available from detailed studies on human C1 inhibitor (Koles, van Berkel, Pieper, et al. 2004; Koles, van Berkel, Mannesse, et al. 2004) but, so far, glycosylation patterns of lysosomal enzymes have not been reported. In this study, a detailed analysis of the N-glycans of rhAGLU expressed in the milk of transgenic rabbits is presented, and the glycosylation pathway is discussed.

	Results

Top Abstract Introduction Results Discussion Materials and methods Conflict of interest statement References

 
General
Monosaccharide analysis (Kamerling and Vliegenthart 1989) of rhAGLU revealed the presence of Fuc, Man, Gal, GlcNAc, and sialic acid (Table I), and a carbohydrate content of 12% (by mass). GalNAc was not detected, indicating the absence of GalNAc(ß1-4)GlcNAc-containing complex-/hybrid-type N-glycans, as well as the absence of conventional (mucin-type) O-glycans.

View this table: [in this window] [in a new window]   Table I. Monosaccharide composition analysis of rhAGLU and derived N-glycan fractions

 
Release and fractionation of N-glycans of rhAGLU
The N-glycans of rhAGLU were completely released by peptide-N⁴-(N-acetyl-ß-glucosaminyl)asparagine amidase F (PNGase F) digestion, as determined by SDS-PAGE, and by monosaccharide analysis. On sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), the native 110-kDa rhAGLU band migrated to a sharp 97-kDa band, after de-N-glycosylation (data not shown), in agreement with the de-N-glycosylation of seven Asn-sites. Monosaccharide analysis of the isolated N-glycan pool revealed a carbohydrate composition similar to that of the native glycoprotein (Table I). Sialic acid analysis of the glycan pool (and the native glycoprotein) showed, apart from Neu5Ac (97.9%), only trace amounts of Neu5Gc (0.2%) and Neu5,9Ac₂ (1.3%) (Figure 1).

View larger version (13K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 1. Sialic acid analysis using DMB derivatives of the rhAGLU N-glycan pool after enzymatic release with PNGase F.

 
1D ¹H Nuclear magnetic resonance (NMR) analysis of the glycan pool, i.e, inspection of the structural-reporter-group regions (Vliegenthart et al. 1983; Koles, van Berkel, Pieper, et al. 2004), demonstrated the presence of a mixture of oligomannose- and complex-type N-glycan structures (Figure 2A), as well as the occurrence of GlcNAc in a GlcNAc(1-P)Man unit (GlcNAc H-1, 5.481, J_1,2 = 3.5 Hz, J_1,P = 7.1 Hz) (Nimtz et al. 1995). Neu5Ac residues were exclusively (2-6)-linked (Neu5Ac H-3e, 2.668; Neu5Ac H-3a, 1.718), whereas Fuc residues occurred in (1-6) linkage to the Asn-bound GlcNAc residue [Fuc CH₃, 1.209 (GlcNAc-1) and 1.221 (ßGlcNAc-1)], and in (1-3) linkage as part of an antennary Lewis x element (Fuc CH₃, 1.178).

View larger version (19K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 2. One-dimensional 500 MHz 1H NMR spectra of (A) the enzymatically released N-glycan pool of rhAGLU; (B) the mono-charged Resource Q fraction N1; (C) the di-charged Resource Q fraction N2. The peak marked by an asterisk is stemming from the 2HO1H signal.

 
Fractionation of the N-glycan pool (80%) on Resource Q fast protein liquid chromatography (FPLC anion-exchange chromatography) yielded three carbohydrate-positive fractions, having elution positions corresponding to neutral (17%), mono-charged (43%), and di-charged (40%) N-glycans, denoted N0, N1, and N2, respectively (Figure 3).

View larger version (17K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 3. Elution profile at 214 nm of the enzymatically-released N-glycan pool of rhAGLU on a Resource Q column. The column was first eluted with 8 ml water, followed by a linear concentration gradient of 0–25% (v/v) 0.5 M NaCl over 3 min, and by a steeper linear concentration gradient of 25–100% (v/v) 0.5 M NaCl over 5 min, at a flow rate of 4 ml/min. The fractions marked by an asterisk did not contain carbohydrate.

 
Structural analysis of neutral N-glycans of N0
Monosaccharide analysis of neutral FPLC fraction N0 showed the presence of Man and GlcNAc only (Table I), supporting the presence of neutral oligomannose-type N-glycans, thereby eliminating the presence of neutral hybrid- and/or complex-type N-glycans. Its ¹H NMR spectrum revealed only structural reporters typical of oligomannose-type structures ranging from Man₅GlcNAc₂ to Man₈GlcNAc₂ (data not shown). Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis showed the presence of Man₅GlcNAc₂ to Man₉GlcNAc₂ ([M + Na]⁺, m/z 1257 /major peak, 1419, 1581, 1743, and 1906 /very minor peak) (data not shown). High-performance liquid chromatography (HPLC) profiling of 2-aminobenzamide (2-AB) labeled fraction N0 gave rise to five peaks, belonging to Man₅GlcNAc₂ (47%), Man₆GlcNAc₂ (27%), Man₇GlcNAc₂ (16%), Man₈GlcNAc₂ (8%), and Man₉GlcNAc₂ (2%) (data not shown).

Neutral FPLC fraction N0 was further preparatively separated by HPLC on LiChrosorb-NH₂, yielding five peaks, denoted N0.5–N0.9 (Figure 4A), belonging to Man₅GlcNAc₂ (44%), Man₆GlcNAc₂ (25%), Man₇GlcNAc₂ (18%), Man₈GlcNAc₂ (11%), and Man₉GlcNAc₂ (2%), respectively, as verified by MALDI-TOF MS.

View larger version (16K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 4. Elution profile at 206 nm of (A) the neutral Resource Q fraction N0 on a LiChrosorb-NH2 column. The column was eluted with a linear gradient of 20–45% (v/v) water in acetonitrile for 50 min at a flow rate of 2 ml/min; (B) the mono-charged Resource Q fraction N1 on a LiChrosorb-NH2 column. A 50-min gradient of 10 mM potassium phosphate buffer, pH 6.5, in acetonitrile (20–45%, v/v) was used at a flow rate of 2 ml/min; (C) the di-charged Resource Q fraction N2 on a LiChrosorb-NH2 column. A 50-min gradient of 10 mM potassium phosphate buffer, pH 6.5, in acetonitrile (20–45%, v/v) was used at a flow rate of 2 ml/min.

 
HPLC subfractions N0.5–N0.9 were further investigated by 1D ¹H NMR spectroscopy, and the structural-reporter-group data, typical for a series of oligomannose-type structures (Michalski et al. 1990; Stroop et al. 2000; Gutiérrez Gallego et al. 2004; Koles, van Berkel, Pieper, et al. 2004), are listed in Table II. Full structures, including the numbering of the Man and GlcNAc residues, are presented in Table III. Note that in case of oligomannose-type (and hybrid-type) N-glycans, the N-acetyl methyl signals of GlcNAc-2 are generally observed at approximately 2.064. The ¹H NMR data of fraction N0.5 correspond with a conventional Man₅GlcNAc₂ structure (cf. compounds QN2.1 in Stroop et al. 2000, and N0.2 in Koles, van Berkel, Pieper, et al. 2004). According to the ¹H NMR data of fraction N0.6, two isomers of Man₆GlcNAc₂, in a ratio of 21:4, occur. The major isomer N0.6.1 is an extension of N0.5 with Man-C, (1-2)-linked to Man-4 (Man6) (cf. compound QN2.2 in Stroop et al. 2000). The minor isomer N0.6.2 is an extension of N0.5 with Man-D2, (1-2)-linked to Man-A (Man6') (cf. compounds N0.3.3_2AB in Koles, van Berkel, Pieper, et al. 2004, and Man₆GlcNAc II in Michalski et al. 1990). For fraction N0.7, all three isomers of Man₇GlcNAc₂ as extensions of N0.6.1 (Man6) were indicated to be present: N0.7.1, with Man-D3 (1-2)-linked to Man-B (Man7); N0.7.2, with Man-D2 (1-2)-linked to Man-A (Man7'); and N0.7.3, with Man-D1 (1-2)-linked to Man-C (Man7'') (cf. compounds QN2.3A and QN2.3B in Stroop et al. 2000, and Man7, Man7', Man7'' in Gutiérrez Gallego et al. 2004); ratio 11:5:2. Similarly, for fraction N0.8 all three conventional isomers of Man₈GlcNAc₂ are present: N0.8.1, with both Man-D1 and Man-D3 (Man8); N0.8.2, with both Man-D2 and Man-D3 (Man8'); and N0.8.3, with both Man-D1 and Man-D2 (Man8'') (cf. compounds N0.4.5_2AB in Koles, van Berkel, Pieper, et al. 2004, QN2.4 in Stroop et al. 2000, and Man₈GlcNAc in Michalski et al. 1990); ratio 1:10:1. The amount of fraction N0.9 was too low for NMR analysis. HPLC profiling studies on the 2-AB labeled fractions N0.5–N0.9 confirmed the oligomannose-type structures Man₅GlcNAc₂-Man₉GlcNAc₂, as found in the NMR and MALDI-TOF MS analyses (Guile et al. 1996; Rudd and Dwek 1997).

View this table: [in this window] [in a new window]   Table II. 1H Chemical shifts of the structural-reporter-group protons of the constituent monosaccharides of neutral oligomannose-type N-glycans, liberated from rhAGLU

 

View this table: [in this window] [in a new window]   Table III. Relative amounts and structures of the identified neutral oligomannose-type N-glycans of rhAGLU

 
Structural analysis of mono-charged N-glycans of N1
Monosaccharide analysis of mono-charged FPLC fraction N1 showed comparable data with those obtained for the N-glycan pool (Table I). Its ¹H NMR spectrum (Figure 2B) is similar to that of the N-glycan pool (Figure 2A), including the Neu5Ac(2-6), Fuc(1-6), Fuc(1-3), and GlcNAc (1-P)Man structural-reporter groups. It should be noted that in complex-type N-glycans an (1-6)-fucosylated N,N'-diacetylchitobiose core element is recognized from the GlcNAc-2 H-1 and NAc signals at 4.66–4.67 and 2.09–2.10, respectively, together with the set of Fuc H-1, CH₃, and H-5 signals at 4.89–4.91, 1.21–1.22, and 4.08–4.13, respectively. The nonfucosylated N,N'-diacetylchitobiose unit is recognized from the GlcNAc-2 H-1 and NAc signals at 4.60–4.61 and approximately 2.081, respectively (Hård et al. 1992). Additionally, the presence of (2-6)-linked Neu5Ac is reflected by methylene signals at 2.67 (H-3e) and 1.70–1.72 (H-3a), and the presence of a Lewis x determinant by the set of Fuc H-1, CH₃, and H-5 signals at 5.11–5.13, 1.17–1.18, and 4.83–4.84, respectively (Kamerling and Vliegenthart 1992).

Mono-charged FPLC fraction N1 was further preparatively separated by HPLC on LiChrosorb-NH₂, yielding six fractions denoted N1.1–N1.6 (Figure 4B). Screening of the various fractions by ¹H NMR spectroscopy showed for N1.2, N1.3, and N1.4 complex mixtures, partially due to the presence of both Neu5Ac and phosphate residues. In order to separate (2-6)-sialylated glycans from nonsialylated glycans, these HPLC fractions were subfractionated by lectin affinity chromatography on agarose-bound elderberry bark lectin (Sambucus nigra agglutinin, SNA). After loading, nonsialylated glycans were eluted with phosphate buffered saline (PBS) buffer as unbound (flow-through) fractions, coded U. Then, (2-6)-sialylated glycans (bound fractions), coded B, were eluted with ethylenediamine (EDA). Subsequent elution with 50 mM lactose did not yield additional glycan material. The various U and B fractions were investigated by ¹H NMR spectroscopy, and the structural-reporter-group data are listed in Tables IV and V, respectively. Fractions N1.1 (Table V) and N1.6 contained no structural-reporter groups indicative for phosphorylation, so no further subfractionations were applied. Fraction N1.5 did not contain enough material for further analysis. Full structures, including the numbering of the various residues, are depicted in Table VI.

View this table: [in this window] [in a new window]   Table IV. 1H Chemical shifts of the structural-reporter-group protons of the constituent monosaccharides of the phosphorylated oligomannose-type N-glycans, liberated from rhAGLU

 

View this table: [in this window] [in a new window]   Table V. 1H Chemical shifts of structural-reporter-group protons of the constituent monosaccharides of the sialylated hybrid- and complex-type N-glycans, liberated from rhAGLU

 

View this table: [in this window] [in a new window]   Table VI. Relative amounts and structures of the identified mono- and di-charged N-glycans of rhAGLU

 
MALDI-TOF MS analysis of N1.1 revealed pseudomolecular ions at m/z 1789.94 and 1643.85, correlating with the [M-H + 2K] ⁺ adducts of Neu5Ac₁Hex₄Fuc₁GlcNAc₃ (M = 1789.55) and Neu5Ac₁Hex₄GlcNAc₃ (M = 1643.48), respectively. ¹H NMR analysis indicated the presence of an (1-3) mono-antennary, mono-(2-6)-sialylated complex-type N-glycan (Tables V and VI). The core GlcNAc-1 residue is only partially (1-6)-fucosylated (cf. structure N1.4 in Koles, van Berkel, Pieper, et al. 2004). HPLC profiling and exo-glycosidase studies on the 2-AB labeled material confirmed the occurrence of both structures (data not shown).

MALDI-TOF MS analysis of N1.2U, N1.3U, and N1.4U revealed pseudomolecular ions at m/z 1724.65, 1886.92, and 2049.17, respectively, correlating with the [M-H + 2Na] ⁺ adducts of Hex₆PGlcNAc₃ (M = 1724.50), Hex₇PGlcNAc₃ (M = 1886.56), and Hex₈PGlcNAc₃ (M = 2048.61), respectively.

¹H NMR analysis of fraction N1.2U indicated the presence of Man₆GlcNAc₂ structure N0.6.1 (Man6), extended with a GlcNAc(1-P) element. The presence of the GlcNAc(1-P) element is reflected by the occurrence of the GlcNAc H-1 and NAc signals at 5.482 and 2.074, respectively (Figure 5A). Going from N0.6.1 to N1.2U, an upfield shift is observed for Man-C H-1 ( –0.030 ppm) (Figure 6), and a downfield shift for Man-C H-2 ( +0.012 ppm). Furthermore, a downfield shift is observed for Man-4 H-1 ( +0.020 ppm), and an upfield shift for Man-4 H-2 ( –0.017 ppm). The structural-reporter-group data of N1.2U fit those of the earlier reported GlcNAc-P-Man₆GlcNAc₂ structure, having a GlcNAc(1-P-6)Man-C element, derived from BHK-21-expressed EPO (Nimtz et al. 1995). Here, the H-6a/H-6b signals of the phosphorylated Man-C residue have shifted out of the bulk region into the Man H-2 region, thereby complicating the spectrum at 4.00–4.02 ppm. It should be noted that also Man-B H-2 is influenced by –0.017 ppm probably demonstrating, by comparing N0.6.1 and N1.2U, the change in microenvironment. For a confirmation of the GlcNAc(1-P-6)Man linkage using ³¹P-filtered 2D ¹H-¹H total correlation spectroscopy (TOCSY), see the analysis of N2.2U, described in Structural analysis of di-charged N-glycans of N2. Previously, it has been shown that going from the trisaccharide Man-D1(1-2)Man-C (1-2)Man-4 to the GlcNAc-phosphorylated trisaccharide Man-D1(1-2)[GlcNAc(1-P-6)]Man-C(1-2)Man-4, Man-C shows an upfield shift for H-1 ( –0.065 ppm) and a downfield shift for H-2 ( +0.017 ppm) (de Waard et al. 1989).

View larger version (19K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 5. Partial one-dimensional 500 MHz 1H NMR spectra (H-1 region, 5.6- 4.4; H-2 region, 4.3- 3.9) of (A) fraction N1.2U; (B) fraction N1.4U; (C) fraction N2.2U; and (D) fraction N2.3U.

 

View larger version (22K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 6. Partial one-dimensional 500 MHz 1H NMR spectra (H-1 region, 5.6- 4.4) of fractions N0.6 [N0.6.1 (major) and N0.6.2 (minor)] (top) and N1.2U (bottom), indicating the shift at Man-C H-1 (N0.6.1 N1.2U) resulting from GlcNAc-phosphorylation, and the GlcNAc H-1 structural-reporter-group signal (GN-P).

 
¹H NMR analysis of fraction N1.3U revealed the occurrence of three GlcNAc-P-Man₇GlcNAc₂ structures derived from N0.7.1 (Man7) and N0.7.2 (Man7'), by extension with a GlcNAc(1-P) element (GlcNAc H-1, 5.484; GlcNAc NAc, 2.076). Going from N0.7.2 to N1.3U.1, an upfield shift is detected for Man-C H-1 ( –0.032 ppm). The observed downfield shift for Man-B H-1 ( +0.02 ppm) is not due to a GlcNAc-phosphorylation at Man-B, as substitution of this residue gives rise to upfield shifts (see e.g., N1.3U.3). N1.3U.2 is an extension of N0.7.1 with a GlcNAc(1-P)Man-C unit, indicated by the upfield shift for Man-C H-1 ( –0.032 ppm) (compare with N1.2U). Going from N0.7.1 to N1.3U.3, an upfield shift is detected for Man-B H-1 ( –0.037 ppm). Taking into account the NMR data for the GlcNAc(1-P-6) element (Nimtz et al. 1995) as well as the Man-B H-1 upfield shift, a GlcNAc(1-P-6)Man-B unit is suggested, yielding a Man-D3(1-2) [GlcNAc(1-P-6)]Man-B(1-6)Man-4' element in N1.3U.3 (for further details about Man-B GlcNAc-phosphorylation, see N2.2U).

The ¹H NMR spectrum of N1.4U showed the presence of Man₈GlcNAc₂ structure N0.8.2 (Man8'), extended with a GlcNAc(1-P) element (GlcNAc H-1, 5.483; GlcNAc NAc, 2.075) (Figure 5B). Going from N0.8.2 to N1.4U, upfield shifts are seen for Man-B H-1 ( –0.044 ppm) and Man-D3 H-1 ( –0.006 ppm), in agreement with the presence of a GlcNAc(1-P-6)Man-B unit (see N1.3U.3 and N2.2U).

HPLC profiling studies of the 2-AB labeled fractions N1.2U, N1.3U, and N1.4U before and after treatment with mild acid (removal of GlcNAc)/alkaline phosphatase (removal of phosphate) confirmed the presence of GlcNAc-phosphorylated Man₆GlcNAc₂, Man₇GlcNAc₂ and Man₈GlcNAc₂ structures, respectively.

MALDI-TOF MS analysis of fraction N1.2B showed pseudomolecular ions at m/z 1773.46 and 1919.54, corresponding with the [M-H + 2Na] ⁺ adducts of Neu5Ac₁Hex₅GlcNAc₃ (M = 1773.58) and Neu5Ac₁Hex₅Fuc₁GlcNAc₃ (M = 1919.64), respectively. The ¹H NMR spectrum of N1.2B reflected an (2-6)-sialylated hybrid-type N-glycan with a Man-A(1-3)Man-4'(1-6) unit (Man-A H-1, 5.106; Man-4' H-1, 4.896; cf. N1.5B in Koles, van Berkel, Pieper, et al. 2004); the core GlcNAc-1 residue is only (1-6)-fucosylated for 60%. These results are in agreement with HPLC profiling/exo-glycosidase studies of the 2-AB labeled fraction N1.2B.

MALDI-TOF MS analysis of fraction N1.3B revealed pseudomolecular ions at m/z 2123.72 and 2437.83, corresponding with the [M-H + 2Na]⁺ adduct of Neu5Ac₁Hex₅Fuc₁ GlcNAc₄ (M = 2122.72) and the [M-2H + 3Na]⁺ adduct of Neu5Ac₂Hex₅Fuc₁GlcNAc₄ (M = 2435.80), respectively. The ¹H NMR spectrum of N1.3B revealed the presence of a mixture of a monosialylated (<5%) and a disialylated (>95%) diantennary chain with core (1-6)-fucosylation. For the monosialylated form, only the (1-3) arm is sialylated (Man-4 H-1, 5.133; cf. N1.3 in Koles, van Berkel, Pieper, et al. 2004; Q1.2 in van Rooijen et al. 1998). The disialylated form is identical to the compound in fraction N2.2B, indicating an incomplete subfractionation. Due to its low amount, the structural-reporter-group signals of the monosialylated compound are not included in Table V. The various results are in agreement with HPLC profiling/exo-glycosidase studies of the 2-AB labeled fraction N1.3B.

MALDI-TOF MS analysis of fraction N1.4B showed a major pseudomolecular ion at m/z 2270.09, corresponding with the [M-H + 2Na]⁺ adduct of Neu5Ac₁Hex₅ Fuc₂GlcNAc₄ (M = 2268.78). ¹H NMR analysis indicated the major presence (>90%) of an (1-6)-fucosylated diantennary N-glycan, (2-6)-sialylated in the (1-3) arm and with a Lewis x determinant in the (1-6) arm (Fuc H-1, 5.132; Fuc CH₃, 1.178; Man-4' H-1, 4.914) (cf. structure N1.1 in Koles, van Berkel, Pieper, et al. 2004). HPLC profiling of the 2-AB labeled N1.4B on GlycoSep-N confirmed the presence of the difucosylated, monosialylated, diantennary N-glycan as the major structure.

MALDI-TOF MS analysis of N1.5 revealed pseudomolecular ions at m/z 2302.28, 2464.38, and 2243.18, correlating with the [M-H + 2K]⁺ adducts of Neu5Ac₁Hex₅Fuc₂GlcNAc₄ (M = 2300.72), Neu5Ac₁Hex₆Fuc₂GlcNAc₄ (M = 2462.78), and Hex₉PGlcNAc₃ (M = 2242.61), respectively. Taking into account the ¹H NMR data of fraction N1.5, it is proposed that Neu5Ac₁Hex₅Fuc₂GlcNAc₄ represents an (1-6)-fucosylated, (2-6)-monosialylated complex-type N-glycan with a Lewis x unit, Neu5Ac₁Hex₆Fuc₂GlcNAc₄ an (1-6)-fucosylated, (2-6)-monosialylated hybrid-type N-glycan with a Lewis x unit at the nonsialylated antenna, and Hex₉PGlcNAc₃ a GlcNAc-phosphorylated oligomannose-type N-glycan. The complex-type structure seems to stem from a chromatographic overlap with fraction N1.4.

MALDI-TOF MS analysis of N1.6 revealed a pseudomolecular ion at m/z 2813.52, correlating with the [M-H + 2K]⁺ adduct of Neu5Ac₁Hex₆Fuc₃GlcNAc₅ (M = 2811.91). Taking into account the ¹H NMR data, N1.6 is suggested to be a triantennary (1-6)-fucosylated, (2-6)-monosialylated N-glycan with Lewis x epitopes at the nonsialylated antennae.

Structural analysis of di-charged N-glycans of N2
Monosaccharide analysis of di-charged FPLC fraction N2 revealed only a gradual difference with that of fraction N1 (Table I). Its ¹H NMR spectrum (Figure 2C) demonstrated similar features as observed for fraction N1 (Figure 2B) and the N-glycan pool (Figure 2A), including the Neu5Ac(2-6), Fuc(1-6), and GlcNAc(1-P)Man structural-reporter groups, but only a trace amount of Fuc(1-3) could be observed.

Preparative HPLC separation of di-charged FPLC fraction N2 on LiChrosorb-NH₂ yielded four fractions, denoted N2.1–N2.4 (Figure 4C). As evidenced by ¹H NMR analysis, all fractions contained both Neu5Ac(2-6) and GlcNAc(1-P)Man elements, whereas fraction N2.2 also contained (1-6)-linked Fuc, and fraction N2.4 both Fuc(1-6) and Fuc(1-3) elements. Fraction N2.4 did not contain enough material (<2%) for further subfractionation. Subfractionations of fractions N2.1, N2.2, and N2.3 were achieved via lectin affinity chromatography on agarose-bound SNA, as explained for fraction N1 (see Structural analysis of mono-charged N-glycans of N1), yielding U and B fractions. The various U and B fractions were investigated by ¹H NMR spectroscopy, and the structural-reporter-group data are listed in Tables IV and V, respectively. Full structures, including the numbering of the various residues, are depicted in Table VI.

MALDI-TOF MS analysis of N2.1U, N2.2U, and N2.3U showed pseudomolecular ions at m/z 2030.15, 2192.32, and 2355.36, respectively, correlating with the [M-2H + 3Na]⁺ adducts of Hex₆P₂GlcNAc₄ (M = 2029.53), Hex₇P₂GlcNAc₄ (M = 2191.59) and Hex₈P₂GlcNAc₄ (M = 2353.64), respectively. Each of the three 2-AB labeled compounds showed only one signal by HPLC profiling.

The ¹H NMR spectrum of fraction N2.1U showed a close resemblance to that of N1.2U, with two major differences: a doubling of the intensity of the GlcNAc(1-P) H-1 and NAc signals at 5.479 and 2.077, respectively, and upfield shifts for Man-A H-1 ( –0.023 ppm) and Man-4' H-2 ( –0.013 ppm). Taken together the various NMR data, it can be concluded that N2.1U is an extension of N1.2U with an extra GlcNAc(1-P) unit at Man-A. Comparing all U fractions, it seems that GlcNAc-phosphorylation at O-6 of a Man residue leads to a clear upfield shift of its H-1 signal ( –0.03 till –0.06 ppm).

Based on its ¹H NMR analysis, fraction N2.2U can be considered as an extension of N1.3U.3 with one extra GlcNAc (1-P-6) residue at Man-C, meaning that starting from N0.7.1 (Man7) both Man-B and Man-C have such a substituent (GlcNAc H-1, 5.482; NAc, 2.076) (Figure 5C). Because of the noted shifts in value of several of the Man H-1 and H-2 signals, when going from neutral to GlcNAc-phosphorylated oligomannose-type structures, detailed 2D TOCSY (150 ms) and nuclear Overhauser enhancement spectroscopy (NOESY) (400 ms) experiments were carried out. NOESY cross-peaks between Man-C H-1 and Man-4 H-2; Man-A H-1 and Man-4' H-3; Man-B H-1 and Man-4' H-6; and Man-D3 H-1 and Man-B H-2; identified Man-4, -4'; -A, -B, -C, and -D3. These detailed assignments, together with the literature NMR data by Nimtz et al. (1995) for N1.2U, have been used in the identification of the various Man residues in the other GlcNAc-phosphorylated compounds. Application of 2D ³¹P selected ¹H-¹H NMR spectroscopy (Figure 7) confirmed the so far suspected GlcNAc(1-P-6)Man linkages (GlcNAc H-1, 5.482; Man H-6a, 4.125; Man H-6b, 3.959). Going from N0.7.1 to N2.2U, clear upfield shifts are detected for Man-B H-1 ( –0.050 ppm), Man-C H-1 ( –0.029 ppm), and Man-A H-1 ( –0.018 ppm). The relatively small upfield shift of Man-A H-1 (compare with N2.1U) is probably due to the GlcNAc-phosphorylation of both Man-B and Man-C.

View larger version (12K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 7. Partial 31P-filtered two-dimensional 1H-1H TOCSY spectrum of fraction N2.2U at 500 MHz recorded with a MLEV-17 mixing time of 10 ms. Peaks on the diagonal originate from the filter selection on the 31P–1H coupling. Cross peaks (nonsymmetrical) occur due to TOCSY transfer of the 31P-filtered signal on the diagonal (TOCSY transfers are visible on horizontal lines in the spectrum: dashed lines).

 
The ¹H NMR data of fraction N2.3U reflected an extension of N1.4U with one extra GlcNAc(1-P-6) residue at Man-C. Not only Man-B (H-1, –0.061 ppm) of N0.8.2 (Man8') is substituted by a GlcNAc(1-P) element, but also Man-C (H-1, –0.033 ppm) (Figure 5D). The two GlcNAc(1-P) H-1 and NAc signals coincide at 5.482, and 2.076, respectively. It is interesting to see that, comparing the ¹H NMR data of the different U fractions, coupling of a GlcNAc(1-P) element at Man-C generally leads to a downfield shift of Man-4 H-1 of approximately +0.020 ppm. Finally, the precise position of the GlcNAc-phosphorylated Man-B H-1 signal seems to be dependent on the presence or absence of a GlcNAc(1-P) element at Man-C: going from N1.4U to N2.3U Man-B H-1 shows an upfield shift of –0.017 ppm, whereas going from N1.3U.3 to N2.2U this upfield shift is –0.013 ppm.

MALDI-TOF MS analysis of fractions N2.1B and N2.2B showed pseudomolecular ions at m/z 2290.18 and 2436.79, correlating with the [M-2H + 3Na]⁺ adducts of Neu5Ac₂Hex₅GlcNAc₄ (M = 2289.74) and Neu5Ac₂Hex₅Fuc₁ GlcNAc₄ (M = 2435.98), respectively. HPLC profiling on 2-AB labeled N2.1B and N2.2B confirmed the presence of only one structure in each fraction. The ¹H NMR data of fraction N2.1B are in accordance with a conventional nonfucosylated, (2-6)-disialylated diantennary complex-type carbohydrate chain (cf. N2.3 in Damm et al. 1989, and Q2.4 in van Rooijen et al. 1998). Those of fraction N2.2B are in agreement with a conventional (1-6)-fucosylated, (2-6)-disialylated diantennary N-glycan (cf. Q2.2 in van Rooijen et al. 1998).

MALDI-TOF MS analysis of fraction N2.3B indicated a major pseudomolecular ion at m/z 2403.44, corresponding with the [M-2H + 3Na]⁺ adduct of Neu5Ac₁Hex₇PGlcNAc₄ (M = 2402.71). The ¹H NMR spectrum revealed N2.3B to be a hybrid-type N-glycan with an (2-6)-linked Neu5Ac in the Man(1-3) arm (Man-4 H-1, 5.123), and four Man residues, of which one bears a GlcNAc(1-P-6) substituent in the Man(1-6) arm (Man-4' H-1, 4.876). The Man-B H-1 signal resonates at 5.113, thereby indicating that Man-B bears the GlcNAc(1-P-6) elongation (compare with N1.3U.3).

MALDI-TOF MS analysis of N2.4 revealed pseudomolecular ions at m/z 2994.87 and 2567.89, correlating with the [M-2H + 3K]⁺ adducts of Neu5Ac₂Hex₆Fuc₂HexNAc₅ (M = 2994.91), and Hex₉P₂GlcNAc₄ (M = 2563.61), respectively. The ¹H NMR spectrum of fraction N2.4 revealed indications for an (1-6)-fucosylated, (2-6)-sialylated complex-type N-glycan with a Lewis x epitope, as well as a GlcNAc-phosphorylated oligomannose-type structure.

Site specificity studies
Liquid chromatography tandem mass spectroscopy (LC/MS) peptide map analysis of the site specific glycosylation (Table VII and Figure 8) demonstrated a glycoform heterogeneity comparable to that observed by ¹H NMR and MALDI-TOF MS experiments. Site specific glycosylation patterns were determined for all seven N-linked sites. Phosphorylated oligomannose- or hybrid-type structures were found at three of the seven sites, namely, Asn-140, Asn-233, and Asn-470. The majority of the phosphorylated oligomannose-type N-glycans was capped by a terminal GlcNAc group. A small amount of noncapped glycans were also observed. Remarkably, Asn-882 and Asn-925 were only occupied by complex-type N-glycans with varying degrees of sialylation. The accurate mass measurement capability of the Q-STAR qq-TOF MS system made it possible to differentiate the 1 Da mass difference between a diantennary/difucosylated/monosialylated glycan and a diantennary/disialylated glycan. The observed HPLC retention shift between the two glycoforms is also in agreement of the charge difference between them. The detection of the diantennary/difucosylated/monosialylated glycan confirms the existence of Fuc(1-3) in addition to core-Fuc(1-6), proposed by the ¹H NMR data.

View this table: [in this window] [in a new window]   Table VII. Mass spectrometry site specificity data performed on rhAGLUa

 

View larger version (22K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 8. Mass spectra of the glycopeptides of the seven N-glycosylation sites stemming from the site-specificity studies. (A) Asn-140; (B) Asn-233; (C) Asn-390; (D) Asn-470; (E) Asn-652; (F) Asn-882; and (G) Asn-925. For the coding system, see Table VII.

 

	Discussion

Top Abstract Introduction Results Discussion Materials and methods Conflict of interest statement References

 
In the literature so far only detailed glycan studies are available for native placental human GAA (Mutsaers et al. 1987), constituting the two active forms with molecular masses of 76 and 70 kDa. The 76-kDa form contains five and the 70-kDa form contains four N-glycosylation sites. Using ¹H NMR spectroscopy, it was found that the oligomannose-type N-glycans present were of intermediate size, Man₅GlcNAc₂ and Man₇GlcNAc₂ (missing Man-D1 and Man-D2), and of small size, Man₃GlcNAc(Fuc_0-1)GlcNAc and Man₂GlcNAc(Fuc_0-1)GlcNAc (missing Man-4). Furthermore, a tiny amount of sialylated diantennary N-glycans was detected. No indications for phosphorylation were found.

Here, we focus on the glycosylation pattern of rhAGLU, expressed in the mammary gland of transgenic rabbits and excreted in the milk. The methodologies applied comprised monosaccharide analysis, mass spectrometry, ¹H and ³¹P NMR spectroscopy, and HPLC-profiling/exoglycosidases. The structural results for rhAGLU show a highly complex N-glycosylation pattern, consisting of neutral oligomannose-type, GlcNAc-phosphorylated oligomannose-type, (2-6)-monosialylated [(1-6)-fucosylated] hybrid- and diantennary complex-type, (2-6)-disialylated [(1-6)-fucosylated] diantennary complex-type, and monosialylated GlcNAc-phosphorylated hybrid-type N-glycans. Four of the complex-type N-glycans were also shown to contain the Lewis x epitope. In addition trace amounts of glycans with an extra antenna on the Man(1-3) arm seem to be present. Sialic acid analysis demonstrated only trace amounts of N-glycolylneuraminic acid. The site-specificity studies indicated heterogeneous patterns for each of the seven N-glycosylation sites. It is interesting to note that only Asn-140, Asn-233, and Asn-470 contained phosphorylated oligomannose- and hybrid-type structures, whereas Asn-882 and Asn-925 were only occupied by complex-type N-glycans. No indications for O-glycosylation were found.

In rhAGLU, the endo-mannosidase pathway is competitively active (Moore and Spiro 1990), just like in the earlier discussed case of recombinant human C1 inhibitor, expressed in the milk of transgenic rabbits (Koles, van Berkel, Pieper, et al. 2004; Koles, van Berkel, Mannesse, et al. 2004). Inspection of the structures in the neutral fraction N0 (Table III) shows the presence of the three possible Man₈GlcNAc₂ isomers, namely, Man8 (major normal pathway; absence of Man-D2), Man8' (endo-mannosidase pathway; absence of Man-D1), and Man8'' (minor normal pathway; absence of Man-D3) (Weng and Spiro 1993, 1996; Verbert 1995; Ermonval et al. 2001). Additionally, the three isomers of Man₇GlcNAc₂ occur, with Man7 being the most abundant one, as well as the two isomers of Man₆GlcNAc₂, one with Man-C (Man6) and one with Man-D2 (Man6'). The smallest glycan is Man₅GlcNAc₂ (Man5).

Inspection of the various GlcNAc-phosphorylated Man_6-9GlcNAc₂ and hybrid-type N-glycans indicates that the major structures are derived from Man_6-8GlcNAc₂, missing Man-D1. The GlcNAc(1-P-6) units occur at Man-A, Man-B when substituted with Man-D3, and/or Man-C. No indications for phosphorylation were observed for Man-D1, Man-D2, and Man-D3. Previously, Varki and Kornfeld (1980) reported that GlcNAc(1-P-6) units (one or two) can occur on Man-A when missing Man-D2, Man-B, Man-C, Man-D1, and/or Man-D3 of Man_6-9GlcNAc₂ structures, as specified for ß-glucuronidase from mouse lymphoma cells. Even the hybrid-type structure N2.3B has been detected in P388D₁ mouse macrophage cells (Varki and Kornfeld 1983).

A unique event in the N-glycosylation of lysosomal enzymes is the mannose 6-phosphorylation. In this process, first, UDP-N-acetylglucosamine:glycoprotein N-acetylglucosamine-1-phosphotransferase (EC 2.7.8.17 [EC] ) transfers GlcNAc(1-P) from UDP-GlcNAc to one or more mannose residues on lysosomal enzymes to give rise to a phospho-diester intermediate (Reitman and Kornfeld 1981a, 1981b; Waheed et al. 1982). Then, the N-acetylglucosamine-1-phosphodiester -N-acetylglucosaminidase (EC 3.1.4.45 [EC] ), the uncapping enzyme, removes the GlcNAc residue to generate the active phosphomonoester (Varki and Kornfeld 1981; Waheed et al. 1981; Varki and Marth 1999). The free phosphate(s) can now be targeted by the MPR and transported towards the lysosomes. "Capped" phosphates cannot be targeted and no transport towards the lysosomes can occur via this way.

Detailed structural analysis of the glycosylation patterns of recombinant therapeutic glycoproteins is a prerequisite for their reliable use in patients. Undesirable immunogenic effects should be eliminated or at least reduced to a minimum. From phase I and phase II clinical trials, it is known that the transgenic human GAA (rhAGLU), described in this paper, is well accepted by patients with Pompe disease (van den Hout et al. 2000; van den Hout et al. 2001; van den Hout et al. 2004; Winkel et al. 2004; Klinge, Straub, Neudorf, Schaper, et al. 2005; Klinge, Straub, Neudorf, Voit, 2005). Evaluating the N-glycosylation pattern of the transgenic product, conventional oligomannose-, hybrid-, and complex-type N-glycans, regularly found in human glycoproteins are present. Neu5Gc, which may lead to immune reactions in humans, is only present in trace amounts. The major difference with the N-glycosylation of native lysosomal enzymes is the occurrence of capped phosphate groups (GlcNAc(1-P-6)Man) (27% of the total glycan pool) on part of the oligomannose- and hybrid-type structures. Indications for the presence of very minor amounts of noncapped phosphate groups in different rhAGLU batches were obtained from the included MS site-specificity studies as well as from in vitro experiments (A.J.J. Reuser et al. unpublished results). As discussed already, Man6P constituents are essential to be targeted by the MPR in order to reach the lysosomes (Gabel and Kornfeld 1982; Goldberg and Kornfeld 1983; Kornfeld and Kornfeld 1985; Kornfeld 1986). So, theoretically, taking into account the N-glycosylation pattern of rhAGLU, the recombinant product should be biologically inactive. However, the clinical trials have shown that it is biologically active after intravenous injection in patients (vide infra). Previously, it has been reported that, although N-acetylglucosamine-1-phospho-diester -N-acetylglucosaminidase resides in the trans-Golgi network (TGN), it cycles between the TGN and the plasma membrane, and is partially available at the cell surface (Kornfeld and Mellman 1989; Rohrer and Kornfeld 2001). Moreover, it has been demonstrated that human serum contains free N-acetylglucosamine-1-phospho-diester -N-acetylglucosaminidase, suggested to be derived from the Golgi enzyme (Lee and Pierce 1995). A possible explanation is that rhAGLU is uncapped in the bloodstream, then targeted by the MPR and transported towards the lysosome, starting the glycogen clearance. Following this assumption, it means that rhAGLU can be considered as a predrug, becoming fully biologically active after in vivo processing by endogenous enzymatic pathways. In this context, the finding of two GlcNAc-phosphorylated Man residues per glycan in rhAGLU is important, because diphosphorylated glycans are better recognized by the MPR than monophosphorylated ones (Goldberg and Kornfeld 1981; Natowicz et al. 1982).

	Materials and methods

Top Abstract Introduction Results Discussion Materials and methods Conflict of interest statement References

 
Recombinant human acid -glucosidase
rhAGLU, isolated and purified from pooled milk of a transgenic rabbit line expressing the glycoprotein on a 8 g/l scale (Bijvoet et al. 1999), was obtained from Pharming Technologies BV, Leiden, The Netherlands.

Release and isolation of N-linked glycans
Purified rhAGLU (70 mg) was dissolved in 5 ml 50 mM sodium phosphate buffer pH 7.3, containing 10 mM EDTA, 10 mM ß-mercaptoethanol, and 0.5% SDS (w/v), then denaturated for 5 min at 100°C. After having added Nonidet P40 (Sigma, Zwijndrecht, The Netherlands) to a final concentration of 1.5% (v/v), the N-glycans were released with recombinant PNGase F (from Flavobacterium meningosepticum expressed in Escherichia coli, EC 3.5.1.52 [EC] ; Roche Molecular Biochemicals, Indianapolis, IN). The digestion was carried out with 350 U PNGase F (added in two steps: 250 U at t₀ and 100 U after 12 h) for 24 h at 37°C. After filtration through 30-kDa cut-off centrifugal filters (Nalgene, Neerijse, Belgium), the de-N-glycosylated glycoprotein was recovered from the filters, and checked by SDS-PAGE (Laemmli 1970) and monosaccharide analysis (Kamerling and Vliegenthart 1989). The filtrate, containing the released N-glycans, was treated with Calbiosorb Adsorbent (Calbiochem, San Diego, CA) according to the manufacturer's protocol to remove detergents. Finally, the N-glycan pool was desalted and purified on a Bio-Gel P-2 column (43 x 1.5 cm, Bio-Rad Laboratories, Veenendaal, The Netherlands), eluted with 10 mM NH₄HCO₃, followed by lyophilization.

Monosaccharide analysis
Glycan (glycoprotein) samples were subjected to methanolysis (1.0 M methanolic HCl, 24 h, 85^oC) followed by re-N-acetylation and trimethylsilylation (Kamerling and Vliegenthart 1989). The mixture of trimethylsilylated (methyl ester) methyl glycosides was analyzed by gas-liquid chromatography (GLC) on an EC-1 column (30 m x 0.32 mm, Alltech, Breda, The Netherlands), using a Chrompack CP 9002 instrument (Chrompack, Middelburg, The Netherlands), and applying a temperature program of 140–240°C at 4°C/min. Confirmation of the data was established via combined GLC-EIMS on a GC 8060/MD 800 system (Fisons Instruments/Interscience, Breda, The Netherlands), equipped with an AT-1 column (30 m x 0.25 mm, Alltech), using the same temperature program as for GLC.

Sialic acid determination
Glycan (glycoprotein) samples were subjected to sialic acid analysis (Hara et al. 1989). Briefly, aliquots (10 µg) were hydrolyzed in 200 µl 2 M acetic acid for 3 h at 80^oC. After cooling to room temperature, 200 µl of a 1,2-diamino-4,5-methylene-dioxybenzene (DMB) solution was added, and the samples were heated in the dark at 50°C for 2.5 h. The DMB solution was prepared by dissolving 3 mg DMB in 1898.1 µl of a solution prepared by mixing 6 mg sodium hydrosulfite, 100.8 µl ß-mercaptoethanol, 1340.4 µl 2 M acetic acid, and 472.8 µl water. After cooling on ice, the generated DMB derivatives of sialic acids were immediately analyzed on a reversed-phase Cosmosil 5C18-AR-II column (4.6 x 250 mm, Waters, Eschborn, Germany), using a Spectroflow 400 HPLC system (ABI Analytical Kratos Division, Separations, H.I. Ambacht, The Netherlands), equipped with a Spectroflow 980 fluorescence detector (_exc.max = 373 nm, _em.max = 448 nm). The elution was carried out isocratically, using acetonitrile:methanol:water (9:7:84, v/v) as solvent.

Anion-exchange chromatography on Resource Q
The N-glycan pool of rhAGLU was fractionated into neutral and charged species on a Resource Q column (6 ml, Pharmacia, Uppsala, Sweden), using a Pharmacia FPLC system. The column was first eluted with water, followed by a linear concentration gradient of 0–0.5 M NaCl, at a flow rate of 4 ml/min; for gradient details, see Figure 3. The elution was monitored by UV absorbance at 214 nm. Individual fractions were lyophilized, then desalted on a Bio-Gel P-2 column (44 x 1 cm) using 10 mM NH₄HCO₃ as eluent, and lyophilized again.

HPLC fractionation on LiChrosorb-NH₂
The FPLC fractions were further fractionated on a LiChrosorb-NH₂ 10 µm column (250 mm x 4.6 mm, Alltech), connected with a LiChrospher Amino 5 µm guard column (7.5 x 4.6 mm), using a Waters 600 HPLC system. A 50-min gradient of water or 10 mM potassium phosphate buffer, pH 6.5, in acetonitrile (each 20–45%, v/v), at a flow rate of 2 ml/min, was used; for gradient details see Figure 4A–C. The fractionations were monitored by UV absorbance at 206 nm, and the individual fractions were lyophilized without desalting.

Labeling of N-glycans with 2-AB and HPLC profiling
Purified and lyophilized N-glycans were treated with 0.35 M 2-AB/1 M sodium cyanoborohydride in dimethyl sulfoxide:acetic acid (7:3, v/v) for 2 h at 65°C. The 2-AB labeled glycans were purified via paper chromatography on acid-pretreated QMA (Whatman) filter paper strips using acetonitrile (three times) as a mobile phase. Glycans (remaining at the base line) were eluted from the dried paper strips with water, and concentrated (Bigge et al. 1995; Kinoshita and Sugahara 1999).

Profiling was carried out on a GlycoSep-N column (50 x 4.6 mm, Oxford GlycoScience, Oxford, UK) at 30°C, using a Waters 2690XE Alliance system, equipped with a Waters 474 fluorescence detector (_exc.max = 373 nm, _em.max = 420 nm). A 100-min gradient of 50 mM ammonium formate, pH 4.4, in acetonitrile (25.2–55%, v/v) was used, at a flow rate of 0.8 ml/min, followed by a 3-min gradient to 100% ammonium formate, which was kept for 5 min at 1 ml/min before regeneration started.

Exoglycosidase digestion of suitable amounts of dried 2-AB labeled fractions, as part of the HPLC profiling, were carried out for 18 h at 37°C in 20-µl solutions, containing 10 µl 0.1 M sodium citrate buffer, pH 5.3, and a fixed amount of enzyme solution completed with water to 10 µl. Before use, the 2-AB labeled samples were purified through 5-kDa cut-off centrifugal filters (Millipore, Amsterdam, The Netherlands). A 2-AB labeled dextran hydrolysate served as external calibration standard on GlycoSep-N. Incubations of oligomannose-type glycans were carried out with 3 µl jack bean -mannosidase (19 U/ml, suspension in 3.0 M (NH₄)₂SO₄ and 0.1 M zinc acetate, pH 7.5; Sigma). Sequential and combined exoglycosidase digestions of hybrid- and complex-type glycans were carried out with 2 µl Streptococcus pneumoniae (2-3)-sialidase (250 mU/ml, in 25 mM NaCl, 20 mM Tris–HCl, pH 7.5; Calbiochem); 2 µl Arthrobacter sialidase (1 U/100 µl, in 10 mM Na-phosphate, 0.1% Micr-O-protect, 0.25 mg/ml bovine serum albumin, pH 7; Roche); 3 µl bovine testis ß-galactosidase (1–3 U/mg, suspension in 3.2 M (NH₄)₂SO₄, pH approximately 5.0; Sigma); 2 µl jack bean ß-N-acetylglucosaminidase (50 U/mg, suspension in 2.5 M (NH₄)₂SO₄, pH 7.0; Sigma); and/or 3 µl jack bean -mannosidase.

Mild acid hydrolysis/alkaline phosphatase digestion of phosphorylated glycans
Mild acid treatment of relevant samples was carried out in 100 mM HCl (0.5 ml) for 30 min at 100°C (Thieme and Ballou 1971). For dephosphorylation, samples, dissolved in 0.1 M NH₄HCO₃, were incubated overnight at 37°C with alkaline phosphatase (Sigma). Portions of 1 U/ml sample were added at t = 0 and at t = 4 h (Karson and Ballou 1978). Treatments were followed by HPLC profiling on a GlycoSep-N column, as described in Labeling of N-glycans with 2-AB and HPLC profiling.

Lectin affinity chromatography
Agarose-bound elderberry bark lectin (SNA) (2 ml; 3 mg lectin/ml gel) was obtained from Vector Laboratories (Burlingame, CA). After loading a small aliquot of the nondesalted mono- or di-charged HPLC LiChrosorb-NH₂ fractions in water onto the column (flow rate, 13 ml/h), the flow was stopped for about 15 min to achieve maximal binding; then, the column was washed with 20 ml PBS to elute the nonbinding fraction. The bound fraction [(