Structural Elucidation of the N - and O -Glycans of Human Apolipoprotein(a) ROLE OF O -GLYCANS IN CONFERRING PROTEASE RESISTANCE

,

Lipoprotein(a) (Lp(a)) 1 consists of a low density lipoprotein (LDL) particle covalently linked through a single disulfide bond to apolipoprotein(a) (apo(a)) (1). Apo(a) contains multikringle domains that are homologous to plasminogen kringles IV and V (2). This homology is thought to underlie the atherogenicity of Lp(a) because apo(a) and its naturally occurring proteolytic fragments compete with plasminogen for fibrin(ogen) binding (3,4). The predicted apo(a)-mediated decrease in fibrinolysis may contribute to the accumulation of fibrin at sites of atherosclerotic lesion development. Apo(a) also exhibits additional proatherogenic properties including binding to fibronectin (5) and decorin (6), direct inhibition of tissue-type plasminogen activator 1 action (7), and modulation of fibrinolysis after cellsurface binding (8,9).
The concentration of plasma and urinary apo(a) proteolytic fragments is increased concomitantly with elevated plasma Lp(a) levels (10,11). In some cases, the apo(a) fragments have a higher atherogenic potential than the intact Lp(a). For example, when the apo(a) moiety of Lp(a) is cleaved by protease to generate "mini-Lp(a)" (12), the resulting particle binds fibrinogen (12) and fibronectin (13) more avidly than undigested Lp(a). Because apo(a) fragments have potentially atherogenic properties and because they are also known to accumulate in atherosclerotic lesions (14), it is important to understand the factors that regulate apo(a) proteolysis. Previous studies have focused on the role of serine proteases and metalloproteinase as relevant enzymes (5,12,15,16); however, the factors related to the proteolytic susceptibility of the various apo(a) domains have not been addressed.
The presence of O-glycans is known in some cases to decrease the sensitivity of certain polypeptides to proteolytic cleavage (17). Based on monosaccharide analysis, apo(a) is predicted to contain a high degree of O-glycosylation (18), and the high content of Ser and Thr residues in the interkringle linker domains suggested that these were the most likely sites for glycan attachment (2,19). Furthermore, studies of apo(a) glycopeptides revealed that the O-glycans are clustered in the kringle IV linker domains (20). This raised the possibility that apo(a) O-glycans could restrict proteolytic cleavage of the interkringle linkers. In the present work we have used exoglycosidase digestion and mass spectrometry to elucidate the structures of apo(a) glycans, and we investigated the potential importance of the major glycan species in modulating the sensitivity of apo(a) to thermolysin digestion.

EXPERIMENTAL PROCEDURES
Materials-Acetonitrile (chromosolv) and hexane were from Riedel-de Haen, Haen, Germany. Methanol, chloroform, and KBr were from BDH, Poole, UK. Acetone was from Fisher. Phosphate-buffered saline was prepared using Oxoid (Basingstoke, UK) tablets. All reagents for hydrazinolysis were from Oxford GlycoSciences, Abingdon, * This work was supported by a Wellcome Trust fellowship and Grant 058833 (to B. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Study Subjects-The plasma lipid and Lp(a) concentrations of the two male and one female subjects under study are given in Table I. Blood was drawn on EDTA (0.1%) after an overnight fast. Total cholesterol and triglycerides were determined by nephelometry. Plasma Lp(a) concentration was determined by enzyme-linked immunosorbent assay, and apo(a) isotyping was performed as described previously (21).
Isolation and Purification of Lp(a) and Apo(a)-Lp(a) and apo(a) were isolated by sequential ultracentrifugation and lysine-Sepharose affinity chromatography as described previously (12). After purification, Lp(a) (1 mg/ml) was incubated with dithiothreitol (10 mM) for 3 h at 37°C. After this treatment, apo(a) was separated from the remaining LDL-like particles on heparin-Sepharose according to Ref. 11 using 1 mg of Lp(a) per 2 ml of resin. All subsequent analyses were performed on individual Lp(a) or apo(a) samples.
Release of N-Linked Glycans by Automated Hydrazinolysis-Purified apo(a) was subjected to automated hydrazinolysis (22) using a Glyco-Prep 1000 instrument (Oxford GlycoSciences Ltd.). For the N-glycan release, hydrazinolysis was performed at 100°C for 5 h in order to achieve maximal recovery. Since these conditions resulted in a partial degradation of O-glycans via a ␤-elimination reaction occurring at the glycosidic linkage between the core GalNAc and sialylated Gal residues, a recognized characteristic of O-glycan instability also known as "peeling" (23,24), milder conditions of 60°C for 6 h were manually employed (see below). The recovered glycan solutions were evaporated to dryness using a vacuum centrifuge, and their reducing termini were fluorescently labeled by reductive amination with 2-aminobenzamide (2-AB) (25) using a LudgerTag kit (Ludger, Oxford, UK).
Release of O-Linked Glycans by Manual Hydrazinolysis-Purified apo(a) was lyophilized and then cryogenically dried before hydrazinolysis (26). Samples were incubated with hydrazine for 6 h at 60°C to release the O-linked glycans (27). Excess hydrazine was removed by evaporation, and the glycans were re-N-acetylated with acetic anhydride in 0.2 M sodium acetate, pH 8.0. Sodium salts were removed with a column containing 5-fold binding excess of Dowex AG50 ϫ 12(H ϩ ) 200 -400 mesh (Bio-Rad) followed by elution with 5 volumes of water. Peptides were removed by descending paper chromatography on Whatman 1MM chromatography paper in butanol:ethanol:water (8:2:1 v/v/v) for 48 h. Glycans were recovered from the paper (Ϫ1 to ϩ2 cm from origin) by washing with water. A rotary evaporator was used to concentrate samples before 2-AB labeling and analysis by HPLC and mass spectrometry.
Simultaneous Exoglycosidase Sequencing of the Released Glycan Pool-The 2-AB labeled glycan pools were evaporated to dryness, and standardized enzyme solutions were added to individual aliquots of each sample (28). The indicated mixtures were incubated for 16 -24 h at 37°C in 100 mM citrate:phosphate buffer (pH 5) containing 0.2 mM zinc acetate and 0.15 M NaCl. ABS and BTG were used at a concentration of 1-2 units/ml. HPLC Systems-Normal phase (NP) and reversed phase (RP) HPLC separations were performed on Waters (Watford, UK) 2690 Alliance separations modules equipped with Waters temperature control modules and Waters 474 fluorescence detectors. External degassers were also used (Douglas Scientific, Southampton, Hampshire, UK). Systems were controlled via Waters Millenium 32 software.
LC-ESI MS-A Waters CapLC system was interfaced with a QTOF hybrid quadrapole time-of-flight mass spectrometer with electrospray ionization in positive mode, fitted with a Z-spray ion source (Micromass UK, Ltd., Wythenshawe, Manchester, UK). A 1 ϫ 150 mm microbore NP-HPLC column was packed with stationary phase material from a GlycoSepN column (Glyko). The same solvents were used as for standard NP-HPLC with a gradient from 80 to 50% acetonitrile over 120 min at a flow rate of 10 l/min. Positive ion mass spectra of the sialylated O-glycans were recorded under the following conditions: source temperature 90°C; desolvation temperature 150°C; desolvation gas flow 200 liters/h; nebulizer gas 40 liters/h capillary voltage 3000 V; cone voltage 30 V; mass range 50 -3500.
MALDI-TOF MS-Positive ion MALDI-TOF mass spectra were recorded with a Micromass TofSpec 2E reflectron-TOF mass spectrometer (Micromass UK, Ltd., Wythenshawe, Manchester, UK) fitted with delayed extraction and a nitrogen laser (337 nm). The acceleration voltage was 20 kV; the pulse voltage was 3200 V, and the delay for the delayed extraction ion source was 500 ns. Samples were prepared by mixing 0.5 l of the aqueous glycan solution and 0.5 l of a saturated solution of 2,5-dihydroxybenzoic acid on the stainless steel MALDI target plate and allowing the mixture to dry at room temperature. The sample: matrix mixture was then recrystallized from ethanol (30).
Deglycosylation and Proteolytic Fragmentation of Apo(a)-Limited proteolysis of apo(a) was carried out by incubation with thermolysin (from Bacillus thermoproteolyticus, 40 units/mg, Sigma) at a mass ratio of enzyme to substrate of 1:200 in 0.125 M Tris-HCl, 0.15 M NaCl, 10 mM CaCl 2 (pH 7.8) for 30 min at 37°C (15). The reaction was stopped by addition of the polyacrylamide gel electrophoresis loading buffer. Deglycosylation of Lp(a) was performed as follows: Lp(a) (10 g) was incubated with CPS (0.05 units) and/or O-glycosidase (5 milliunits, Roche Molecular Biochemicals) in 100 l of phosphate buffer (pH 7.3) overnight at 37°C.
Electrophoresis and Immunoblotting-Sodium dodecyl sulfate-polyacrylamide gel electrophoresis was performed on 4 -10% acrylamide gradient slab minigels (Mini Protean II, Bio-Rad), using a discontinuous buffer system (31). Before electrophoresis, samples (50 -200 ng of apo(a) protein) were combined with glycerol, bromphenol blue, and EDTA to final concentrations of 2%, 0.01%, and 0.05 mM, respectively. Reduced samples were prepared by boiling at 100°C for 4 min in the presence of 20 mM dithiothreitol and 2% sodium dodecyl sulfate. Proteins were then electroblotted onto nitrocellulose and revealed by immunoblotting (15). The uniform transfer of all material was confirmed by Coomassie staining gels after electroblotting and by reversible staining of the nitrocellulose sheets with Ponceau S solution. Fragments of apo(a) were revealed using a peroxidase-conjugated polyclonal antiapo(a) antibody (32). The polyclonal antibody reacts strongly with all apo(a) fragments generated either in vivo or in vitro (15,31). Visualisation was achieved by enhanced chemiluminescence detection (ECL, Amersham Pharmacia Biotech).   3  M  59 167  48  95  120  50  29 a TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; TG, triglyceride; Lp(a), lipoprotein (a); apo(a) ISO, apolipoprotein(a) isotype, given as number of kringle IV repeats.
b Subject 2 displayed two apo(a) isotypes.

Assignment of Apo(a) N-and O-Glycan Structures by
Exoglycosidase Sequencing-Human Lp(a) was isolated from fasted plasma and apo(a) purified as described previously (12). Apo(a) was then subjected first to automated hydrazinolysis to remove all N-and O-linked glycans, and the released sugars were fluorescently labeled and analyzed by NP-HPLC. Two predominant N-linked glycans were detected that accounted for more than 85% of the material eluting in positions of common Nlinked glycans (Fig. 1). The G.U. values of the two major peaks and the products resulting from their digestion with the listed exoglycosidases indicated that they were biantennary complex oligosaccharides present in either a mono-(A2G2S1) or disialylated (A2G2S2) state (Fig. 1A). Treatment of the N-glycans with ABS resulted in the loss of both peaks and formation of a single peak with a G.U. value identical to the neutral biantennary structure, A2G2 (Fig. 1B). In the presence of both ABS and BTG, the terminal galactose residues were also removed leading to the formation of the common core structure GlcNAc 2 Man 3 GlcNAc 2 , which is also abbreviated as A2G0 (Fig.  1C). Because the conditions employed for hydrazinolysis of N-glycans can lead to the partial destruction of O-glycans, apo(a) was also subjected to a milder manual hydrazinolysis method to maximize the recovery of O-glycans. This procedure yielded several compounds that were labeled with 2-AB and analyzed by HPLC. An NP-HPLC profile of the recovered peaks is shown in Fig. 2. The peaks labeled with asterisks (as well as a compound that coeluted with peak 1) were also detected in hydrazinolysis "blanks" and were subsequently found to be derived from carbohydrates associated with the filter paper used in the glycan purification method. The apo(a) O-glycans detected were all of the core type 1 structure (Gal␤1-3Gal-NAc␣1-R) and were found to be present in different sialylated states. The predominant O-glycan detected was a monosialylated trisaccharide (peak 3) that was closely related to the disialylated (peak 4) and nonsialylated (peak 1) structures that were also present ( Fig. 2A). Peak 2 accounted for ϳ4% of the O-glycan pool and was identified as NeuNAc␣2-3Gal which is a degradation product caused via a ␤-elimination reaction occurring at the glycosidic linkage between the core GalNAc and sialylated Gal residues, a recognized characteristic of O-glycan instability also known as "peeling" (23,24).
Treatment of the O-glycan pool with ABS resulted in the loss of the sialylated structures and a concomitant increase in the level of the nonsialylated core type 1 structure, Gal␤1-3GalNAc (Fig. 2B). In the presence of ABS and BTG the Gal␤1-3GalNAc peak was partially removed, and the remaining nondigestible material appeared to be due to a contaminant introduced during the hydrazinolysis procedure. In order to achieve an accurate determination of the amount of Gal␤1-3GalNAc normally present on apo(a), the glycan mixture was also run on RP-HPLC to separate the contaminant from Gal␤1-3GalNAc. The RP-HPLC profiles are shown in Fig. 3. The compounds labeled with asterisks again represent peaks that were present in the hydrazinolysis "blank," and the remaining peaks are numbered using the same nomenclature as shown in Fig. 2A. The Gal␤1-3GalNAc (Fig. 3A, peak 1) was clearly separated from the other compounds and was found to account for 11% of the O-glycan pool. Note that the order of elution of peaks 2 and 3 changed under the RP-HPLC conditions as compared with the NP-HPLC conditions (Figs. 3A and 2A, respectively). The identity of peak 1 was confirmed by ABS digestion of the sialylated structures (peaks 3 and 4) that were quantitatively recovered as Gal␤1-3GalNAc on the RP-HPLC system (Fig. 3B).
Confirmation of Glycan Structural Assignments by Mass Spectrometry-We next confirmed the structural assignments made for all of the apo(a) glycans by mass spectrometry. The sialylated O-glycans were analyzed by (NP) LC-ESI MS (Fig.  4). The two peaks eluted in their predicted order with the monosialylated structure eluting at 60 min and the disialylated structure eluting at 82 min (Fig. 4A). The corresponding mass spectra for these peaks revealed molecular ions at m/z values of 795.3 (Fig. 4B) and 1086.4 (Fig. 4C) Human apo(a) was subjected to hydrazinolysis, and the released glycans were labeled with 2-AB and analyzed by NP-HPLC (A). The glycan pool was also treated with ABS (B), and ABS ϩ BTG (C), in order to sequence the oligosaccharide chains (see "Experimental Procedures" for details). The digestion of the two major complex biantennary structures (peaks 1 and 2) to form the predicted truncated products are indicated by the arrows. A2G2S2, disialylated complex biantennary glycan containing 2 galactose residues; A2G2S1, monosialylated complex biantennary glycan containing 2 galactose residues; A2G2, nonsialylated complex biantennary glycan containing 2 galactose residues; A2G0, nonsialylated complex biantennary glycan containing no galactose. G.U. values were derived from a dextran ladder (29). The glycan structures are represented by the following symbols: E, mannose; f, N-acetylglucosamine; छ, galactose; ࡗ, N-acetylgalactosamine; ૺ, N-acetylneuraminic acid. determined by analyzing the glycan pool by MALDI-TOF MS, and the masses of the most abundant ions detected are given in Table II along with their corresponding glycan structures.
The structures of apo(a) glycans have not been determined previously; however, predictions based on apo(a) monosaccharide composition have been made (18,34). A comparison of the published monosaccharide composition of apo(a) compared with the predicted monosaccharide composition based on our analysis of intact N-and O-glycans is given in Table III. The values for the three studies listed in Table III are   Apo(a) samples were subjected to a manual hydrazinolysis procedure at 60°C for 6 h and the recovered glycans 2-AB labeled and analyzed by RP-HPLC. The peaks were assigned according to the numbering scheme for the NP-HPLC analysis (Fig. 2). Identity of the structures was confirmed by digestion with ABS and by comparison of the product and precursor arabinose unit (A.U.) values with those of known standards (29). See main text for further details. protect potential protease-sensitive regions from enzyme action. In order to test this hypothesis, we pretreated Lp(a) with CPS or CPS plus O-glycosidase (which removes core Gal-GalNAc from Ser and Thr residues) and assessed apo(a) susceptibility to limited thermolysin digestion. In the presence of thermolysin, apo(a) was degraded to form two peptides (Fig. 6A, lane 2) that have been characterized previously and found to be the result of a cleavage between Ala 3532 and Phe 3533 of the linker 4 domain that links kringle IV4 and kringle IV5 (12). When Lp(a) was first treated with CPS to remove terminal sialic acids, two dominant peptides were again formed after thermolysin treatment, and their molecular weights were decreased compared with the peptides generated without CPS pretreatment (Fig. 6A, compare  lanes 2 and 3). This shift in molecular weight is consistent with the loss of approximately 16 -20 residues of NeuNAc from each peptide. Interestingly, the larger N-terminal fragment was also more degraded as evidenced by the smearing pattern below the major band (Fig. 6A, lane 3). This suggests that desialylation alone may alter the conformation of the interkringle linkers to such an extent that thermolysin action is enhanced or that the charge associated with the sialic acids normally repels access of the apo(a) polypeptide to thermolysin. Evidence for lower molecular weight peptides was also seen below the C-terminal fragment, again consistent with a conformational change that favors thermolysin action (Fig. 6A, lane 3). When Lp(a) was pretreated with CPS and O-glycosidase and then incubated with thermolysin, the vast majority of the N-terminal peptide was degraded with less than 5% of the material remaining in its original position (Fig. 6A, lane 4). The C-terminal peptide was also further degraded, although a band remained with a molecular mass ϳ15 kDa less than the original C-terminal peptide (Fig. 6A, lane 4).
This may be partially due to the loss of the remaining Gal-GalNAc residues or to the formation of deglycosylated fragments of the original N-terminal peptide.
The pretreatment of apo(a) with the O-glycosidase (which specifically cleaves Gal-GalNAc) alone did not alter apo(a) susceptibility to thermolysin digestion nor did it result in a change in molecular weight of the two thermolysin proteolytic products (Fig. 6A, compare lanes 2 and 5). This final result indicates that only a very small proportion of the apo(a) glycans was present as Gal-GalNAc and that their removal does not make a significant impact on the susceptibility of apo(a) to thermolysin digestion. The fact that the molecular weight of the initial N-terminal and C-terminal peptides remains constant after O-glycosidase treatment also supports our oligosaccharide sequencing and mass spectrometry data ( Fig. 3 and Table II) showing that the predominant O-glycan present is NeuNAc-Gal-GalNAc (which is not a substrate for O-glycosidase). Fig. 6B shows an additional Western blot from a separate experiment. In this case the gel was intentionally overloaded in order to detect any minor apo(a) fragments generated under the different digestion conditions. The data are in close agreement with those shown in Fig. 6A and confirm that desialylation of apo(a) results in a partial increase in its sensitivity to thermolysin digestion, whereas complete removal of apo(a) O-glycans dramatically increases its subsequent fragmentation.
Prediction of O-Glycosylation Potential of Interkringle Linker Domains-Previous work has revealed a thermolysin-sensitive site for apo(a) cleavage in the kringle IV4 linker domain, i.e. between kringles 4 and 5 (12). Our present data suggest that O-glycans present in the interkringle linker domains may stabilize apo(a) by preventing thermolysin proteolytic action. By using a computer algorithm that compares polypeptide sequences with those of glycoproteins known to contain O-glycans (33), we compared the O-glycosylation potential for the interkringle linker domains IV2, IV4, and IV7. In agreement with our hypothesis that O-glycosylation can modulate apo(a) protease resistance, the kringle IV4 linker was predicted to have no sites occupied by O-glycans (Fig. 7). In contrast, the kringle IV2 and IV7 linkers (as well as the other kringle IV linkers, data not shown) were predicted to be highly glycosylated (Fig. 7). DISCUSSION The present studies utilized exoglycosidase sequencing techniques and mass spectrometry to determine the complete structure (including glycosidic linkage analysis) of human apo(a) oligosaccharides. The predicted monosaccharide composition was in general agreement with data published previously (18,34). However, the disialylated and non-sialylated O-glycan structures that were predicted previously to occur (18) were found to account for only 10 -20% of the apo(a) O-glycan pool (Fig. 3). The most abundant oligosaccharide was, in fact, a monosialylated core type 1 O-glycan (Fig. 5), a structure that can also be deduced from the previous monosaccharide analysis (18,34).
The slightly lower ratio of NeuNAc:Gal observed in our study is due to the finding that almost half of the biantennary N-glycans (which accounted for ϳ17% of total apo(a) glycans) were present in a monosialylated state. In earlier studies (18), NeuNAc and Gal were present at equimolar ratios, suggesting an absence of monosialylated N-glycans. Possible reasons for this discrepancy may be related to the high variation in sialic acid concentrations reported (as determined by gas chromatography) or due to the nonspecificity of the thiobarbituric acid method used in the NeuNAc quantitation (18). Use of the thiobarbituric acid or "Warren" method can lead to an overestimation of lipoprotein sialic acid levels (35,36). It is unlikely that the apo(a) N-glycans were desialylated during processing in the present work as the O-glycan pool (which also contained NeuNAc in the ␣2-3 linkage to Gal) did not contain a predominant nonsialylated structure. We have also observed that human apoB100 contains a high proportion of its N-glycans as a monosialylated complex biantennary structure, and this was noted when the glycans were removed from apoB100 by either hydrazinolysis or treatment with peptide N-glycosidase F, 2 indicating that under our hydrazinolysis conditions the NeuNAc␣2-3Gal glycosidic linkage is stable. When apo(a) was treated with O-glycosidase (which cleaves Gal-GalNAc but not NeuNAc-Gal-GalNAc from Thr/Ser) followed by thermolysin, there was no significant change in the molecular weight of the resulting peptides compared with the peptides resulting from treatment of apo(a) with thermolysin alone. This supports our exoglycosidase sequencing data indicating that the Gal-GalNAc content of the original apo(a) samples was low.
Having established the apo(a) oligosaccharide structures, we went on to investigate the potential influence of the O-glycans on apo(a) sensitivity to protease digestion. It is well known that human plasma and urine contains apo(a) fragments (31,37). The factors that control the formation of these fragments are, however, poorly understood. It appears that specific domains of apo(a) are more susceptible to proteolytic cleavage than others (12). For example, the major cut site for both thermolysin and neutrophil elastase (both serine proteases) is in the kringle IV4 linker (12,13). This domain is predicted to be devoid of glycan chains (Fig. 7B), and we hy-  pothesized that the removal of O-glycans from other linkers may result in increased susceptibility to protease digestion. This was shown to be the case when O-glycans were trimmed by sialidase treatment, and particularly when the entire oligosaccharide structures were removed by treatment with sialidase and O-glycosidase. The role of O-glycans in protecting apo(a) from protease digestion may also explain why the apo(a) peptide generated by cleavage at the elastase cut site in the kringle IV7 linker requires an extended incubation time or increase in enzyme concentration (13). Other examples of O-glycans conferring protease resistance have also been reported (see Ref. 17). Interestingly, the Drosophila melanogaster "mucin-D" glycoprotein contains a large amount (40% by mass) of Gal␤1-3GalNAc that renders it highly resistant to protease action (38), consistent with our data showing that the same disaccharide alone (generated after treatment of Lp(a) with sialidase) provided resistance to thermolysin digestion.
It is likely that apo(a) O-(and N-) glycans play additional functional roles, for example in intracellular processing (39,40), maintaining the tertiary structure of apo(a), and preventing aggregation (17). Because apo(a) O-glycans are hydrophilic, they may also play a role in ensuring that the bulk of apo(a) is extended out into the aqueous phase, as has been observed in structural studies of recombinant Lp(a) (41).
Megalin/gp 330 has recently been identified as an endocytic receptor for Lp(a) (42). This receptor also appears to be involved in the cellular uptake of other glycosylated apolipoproteins including apoE, apoJ, apoH, and apoB100 (43)(44)(45)(46). Since megalin is highly expressed in the brush border of renal proximal tubules and in the coated pits of glomerular epithelial cells (47,48), it provides a plausible control mechanism for the generation of specific urinary apolipoproteins (49). Interestingly, plasminogen is also a ligand for megalin (50), yet its urinary excretion is extremely low compared with apo(a) (49). Since the kringle type IV and type V structures of apo(a) are homologous to plasminogen (2), it is tempting to speculate that the higher excretion of apo(a) may be related to the almost 14-fold higher content of carbohydrate associated with the apo(a) kringle linkers. This possibility has not been addressed as far as we are aware.
The presence of terminal Gal residues on apo(a) may also confer an ability to bind to the macrophage asialoglycoprotein receptor. An analogous pathway has been proposed for apoB100 (LDL) endocytosis by macrophages (51,52). Arguing against such a pathway, one study has shown that the removal of sialic acid residues from recombinant apo(a) did not alter its binding to a partially characterized macrophage foam cell surface receptor (53). Investigation of the full range of functions for apo(a) oligosaccharides will clearly require further study.
The generation of apo(a) fragments in vivo has been suggested to be potentially atherogenic due to the possibility of C-terminal peptides interfering with fibrinolysis (15,19). Of potential relevance, apo(a) fragments accumulate in atherosclerotic lesions where they may promote thrombogenesis (14). In the microenvironment of the artery wall, it is possible that macrophage-derived glycosidases and proteases act in concert to degrade apo(a). This might have the unfortunate consequence of generating macromolecular complexes between "mini-Lp(a)" and extracellular matrix components that could then be taken up by macrophages to form foam cells and thereby promote the development of the atherosclerotic lesion. Since macrophages can release proteases (54,55) and glycosidases (56), our discovery that apo(a) glycans normally limit the extent of proteolytic fragmentation might explain why, in the macrophage-rich atherosclerotic lesion (57), apo(a) fragments are found to accumulate (14). It would be interesting to assess the degree of glycosylation present on such lesion-derived apo(a) fragments.
In conclusion, the present study has directly elucidated the structure of human apo(a) oligosaccharide chains and shown that the O-glycans play an important role in maintaining the stability of apo(a) by conferring resistance to degradation by the serine protease, thermolysin.