Differential Glycosylation Regulates Processing of Lipoprotein Receptors by γ-Secretase*

The low density lipoprotein (LDL) receptor-related protein 1 (LRP1) belongs to a growing number of cell surface proteins that undergo regulated proteolytic processing that culminates in the release of their intracellular domain (ICD) by the intramembranous protease γ-secretase. Here we show that LRP1 is differentially glycosylated in a tissue-specific manner and that carbohydrate addition reduces proteolytic cleavage of the extracellular domain and, concomitantly, ICD release. The apolipoprotein E (apoE) receptor-2 (apoER2), another member of the LDL receptor family with functions in cellular signal transmission, also undergoes sequential proteolytic processing, resulting in intracellular domain release into the cytoplasm. The penultimate processing step also involves cleavage of the apoER2 extracellular domain. The rate at which this cleavage step occurs is determined by the glycosylation state of the receptor, which in turn is regulated by the alternative splicing of an exon encoding several O-linked sugar attachment sites. These findings suggest a role for differential and tissue-specific glycosylation as a physiological switch that modulates the diverse biological functions of these receptors in a cell-type specific manner.

The low density lipoprotein (LDL) 1 receptor gene family comprises a class of multifunction cell surface receptors. They mediate the endocytosis of a diverse spectrum of extracellular ligands that include lipoproteins, lipases, proteases, and protease-inhibitor complexes, lipophilic vitamins, and their carrier proteins, growth factors, and other macromolecules with various biological functions (reviewed in Ref. 1). Recently, unexpected roles in cellular signal transduction have also been recognized (2) and our knowledge of the mechanisms by which LDL receptor family members regulate cellular behavior and growth continues to expand (3,4).
All the members of the LDL receptor gene family contain multiple N-linked glycosylation sites and some, such as the LDL receptor, the very low density lipoprotein receptor (VLDLR) and the apoER2, also possess a serine-and threoninerich region that directly precedes their single membrane-spanning domain. This so-called O-linked sugar domain serves as an attachment domain for numerous O-linked carbohydrates (reviewed in Ref. 5). The extent to which the different receptors undergo glycosylation is variable in different tissues, and distinct splice variants that contain or do not contain the O-linked sugar domain exist for apoER2 (6 -8) and VLDLR (9 -12).
Little is known about the functional importance of the carbohydrate chains on the different receptors. In general, one role of glycosylation is thought to be the stabilization of the receptor proteins. Several mutant cell lines with decreased levels of functional LDL receptors because of defects in glycosylation resulting in increased receptor turnover have been identified (13). Similarly, reduced stability of VLDLR splice variants lacking the O-linked sugar domain has been described (11). It is not clear, however, why apoER2 and VLDLR are differentially spliced in different tissues and which biological functions of the receptors might be regulated by the presence or absence of the O-linked sugar domain.
In the case of several other transmembrane proteins, modification of sugar side chains has been shown to regulate specific aspects of their metabolism. Cell surface receptors of the Notch family, for instance, are also single-pass transmembrane glycoproteins that are processed by a series of proteolytical cleavages. Ultimately, this results in the release of the Notch intracellular domain (ICD) into the cytoplasm, followed by translocation into the nucleus where it participates in the regulation of gene transcription (reviewed in Ref. 14). Extension of O-linked fucose monosaccharides by an N-acetylglucosamine transferase is crucial in determining which stimuli will preferentially induce the proteolytical processing of Notch molecules (15)(16)(17). Another example for specific proteolytic processing that is regulated by the glycosylation state of the protein is that of the amyloid precursor protein, which is also modified by its carbohydrate side chains (18).
Here, we report that the LDL receptor-related protein 1 (LRP1), which is processed in a Notch-like fashion by a series of proteolytical cleavage events that result in the release of its ICD, is differentially glycosylated in a tissue-specific manner. We hypothesized that this differential glycosylation of LRP1 might serve to regulate the rate at which LRP1 is being processed. Using reporter gene assays and immunoblotting for LRP1 cleavage products in the glycosylation deficient ldlD cell line we now show that the glycosylation state of this protein indeed determines the rate of processing. Specifically, the extracellular cleavage step that results in the shedding of the LRP1 ectodomain is enhanced when the protein is hypoglycosylated, which in turn results in increased substrate availability for the final ␥-secretase-mediated release of the LRP1 ICD. Furthermore, we demonstrate that the apoER2 can also be proteolytically processed in cultured primary neurons as well as in transfected cell lines and that the presence or absence of its alternatively spliced O-linked sugar domain modulates the rate at which the ICD of apoER2 is released. EXPERIMENTAL PROCEDURES D-(ϩ)-Galactose and N-acetyl-D-galactosamine were purchased from Sigma. The ␥-secretase inhibitor N-(19)-S-phenylglycine-t-butyl ester (DAPT) was synthesized as described in Ref. 19. L-685,458 was from Bachem (Torrance, CA).
Cell Culture and Transfection of CHO-K1, ldlD, and 293 Cells-The Chinese hamster ovary cell line CHO-K1 (CCL 61, ATCC), its mutant daughter cell line ldlD (25), and the human embryonic kidney cell line HEK 293 (CRL-1573, ATCC) were maintained in monolayer culture at 37°C in an 8% CO 2 atmosphere.
On day 0 cells were set up in Dulbecco's modified Eagle's medium/ F-12 (1:1) (CHO-K1 and ldlD) or Dulbecco's modified Eagle's medium low glucose (293 cells) (both Cellgro, Herndon, VA) containing 100 units/ml penicillin, 100 g/ml streptomycin sulfate (Cellgro, Herndon, VA), and 3 (CHO-K1 and ldlD) or 10% (v/v) fetal calf serum (Atlanta Biologicals, Norcross, GA). On day 2 cells were transfected by Fu-GENE TM (Roche Diagnostics) (CHO-K1 and LDL-D) or calcium phosphate precipitation (MBS Kit TM , Stratagene) (HEK 293) with the indicated amounts of plasmid DNA. Cells were harvested 48 h after transfection. For the treatment with ␥-secretase inhibitors a 20 mM stock solution of DAPT and a 10 mM stock solution of L-685,458 were prepared in Me 2 SO and used at final concentrations ranging from 100 nM to 10 M as indicated.
Preparation of Rat Embryonic Cortical Neurons and Neuronal Lysates-Animals were maintained in accordance with National Institutes of Health and institutional animal care guidelines. Cortical neurons from E16 wild type or apoER2 knockout mice (2) and E18 rat embryos (Sprague-Dawley) were prepared, cultured, and analyzed by immunoblotting as described in Ref. 26.
Preparation of Cell Membranes for Western Blot Analysis-Cells were grown to confluence in 100-mm culture dishes. For the preparation of membranes, culture medium was aspirated and cells were washed twice with ice-cold phosphate-buffered saline. Cells were scraped into 1 ml of phosphate-buffered saline per dish and centrifuged at 850 ϫ g at 4°C. Cells were resuspended in hypotonic Buffer A (10 mM HEPES-KOH, pH 7.8, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1 mM NaVO 3 ), left on ice for 20 min, and then passed 10 times through a 22-gauge needle. After centrifugation at 850 ϫ g for 5 min at 4°C, the supernatant was recentrifuged at 100,000 ϫ g for 30 min. Finally, the membrane pellet was resuspended in SDS buffer (10 mM Tris-HCl, pH 6.8, 100 mM NaCl, 1% SDS, 1 mM EDTA).
Preparation of Liver and Brain Membranes and Treatment with Glycosidases-Liver and brain membranes were prepared from adult Sv129Ev mice. Organs were resected and homogenized in membrane Buffer I (20 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM CaCl 2 , and EDTA-free protease inhibitor mixture (Roche Diagnostics)) (2 ml buffer per gram of tissue) immediately. Lysates were centrifuged at 1,000 and 1,500 ϫ g for 5 min each at 4°C. The supernatant was then subjected to centrifugation at 100,000 ϫ g for 30 min at 4°C. The pellet of this step was resuspended in Buffer II (50 mM Tris-HCl, pH 8, 80 mM NaCl, 2 mM CaCl 2 , 1% Triton X-100, 0.1% SDS, protease inhibitors as above) and centrifuged again at 100,000 ϫ g for 30 min. at 4°C. The protein concentration of the final supernatant was determined using Bio-Rad DC Protein Assay (Bio-Rad). 10 g of membrane proteins were used for the treatment with glycosidases. Neuraminidase (P0720S, New England Biolabs, Beverly, MA), PNGase F (New England Biolabs, Beverly, MA), and O-glycosidase (Roche Diagnostics) were employed according to the manufacturer's instructions. The treated membrane proteins were further analyzed by SDS-polyacrylamide gel electrophoresis and immunoblotting as described below.
SDS-Gel Electrophoresis and Immunoblot Analysis-SDS-PAGE and immunoblot analysis of membrane lysates were performed according to standard procedures on 4 -15% SDS-polyacrylamide gels. After electrophoresis and protein transfer to polyvinylidene difluoride membranes, immunoblot analysis was carried out with rabbit polyclonal antibodies against a carboxyl-terminal epitope of LRP1 (24), LDLR (27), or apoER2, or against the alternatively spliced cytoplasmic insert of the apoER2 (28). Bound antibodies were visualized by enhanced chemiluminescence (ECL) using SuperSignal CL-HRP substrate (Pierce) according to the manufacturer's instructions.
Reporter Gene Assays-CHO-K1, ldlD, and 293 cells were harvested for reporter gene assays 48 h after transfection. The culture medium was removed and the cells were washed once with phosphate-buffered saline. After the addition of "Reporter Lysis Buffer" (Promega, Madison, WI), the cells were processed according to the manufacturer's instructions. Luciferase and ␤-galactosidase gene expression were analyzed as described previously (20) using the "Luciferase Assay System" (Promega, Madison, WI) and the "Chemiluminescent ␤-Galactosidase Detection Kit" (BD Biosciences), respectively. The normalization of luciferase activity to ␤-galactosidase activity was done as described. All transfections for reporter gene assays were done in duplicate and repeated in at least two independent experiments.

The Glycosylation Pattern of LRP1 and Other Lipoprotein
Receptors Is Tissue-dependent-All members of the LDL receptor gene family are type I membrane glycoproteins. LRP1 is cleaved in a late secretory compartment by the processing proteinase furin (29). The mature, processed LRP1 consists of a membrane-bound 85-kDa subunit and a non-covalently associated extracellular subunit of ϳ515 kDa. The apparent molecular weight of LRP1 as judged by SDS-polyacrylamide gel electrophoresis varies between different tissues. As shown in Fig. 1A, the smaller LRP1 subunit from mouse brain migrates faster than the one present in mouse liver.
To determine whether differential glycosylation of LRP1 in brain and liver was responsible for the observed difference in molecular weight, we treated membrane preparations from either tissue with neuraminidase, O-glycosidase, neuraminidase ϩ O-glycosidase, or PNGase F (Fig. 1B). Removal of Nacetylneuraminic acid residues by neuraminidase reduced the difference in apparent molecular weight, and treatment with PNGase F, which removes all N-linked carbohydrate chains, resulted in identical electrophoretical migration of LRP1 from both tissues. Treatment with O-glycosidase, with or without pretreatment with neuraminidase, however, did not change the apparent molecular weight of LRP1 in either preparation. Thus, LRP1 is differentially N-glycosylated in liver and brain. This is mainly because of a higher degree of sialylation in the liver. In contrast to the LDL receptor, which is heavily Oglycosylated (boxed inset, Fig. 1B), there is no apparent Olinked glycosylation of LRP1 in either tissue. Several other members of the LDLR gene family, i.e. the LDL receptor, VLDLR, and apoER2, contain a serine-and threonine-rich domain, which directly precedes the transmembrane segment. This domain serves as an attachment domain for multiple O-linked carbohydrate chains. This so-called O-linked sugar domain is encoded by a separate exon in each of the respective genes and physiologically occurring splice variants that contain or lack the O-linked sugar domain have been found for VLDLR and apoER2. In agreement with these earlier findings we detected the described splice variants of both receptors by RT-PCR, whereas for the LDLR only transcripts that contained the exon encoding the O-linked sugar domain were detected. Intriguingly, there were also tissue differences in the relative abundance of apoER2 mRNA coding for the receptor form that lacks the O-linked sugar domain, which was more abundant in brain compared with placenta (Fig. 1C).
The Proteolytic Processing of LRP1 by ␥-Secretase Is Increased in Glycosylation Deficient Cells-The functional implications of this tissue-specific glycosylation of LDL receptor family members are largely unclear. It is known, however, that glycosylation affects the stability of some of the receptors. For other transmembrane proteins, which undergo regulated proteolytical processing, i.e. for receptors of the Notch family and for the amyloid precursor protein, regulation of proteolytic cleavage by their glycosylation state has been described. Furthermore, we recently reported that LRP, like amyloid precursor protein and Notch, can also undergo serial proteolytic processing, eventually resulting in the release of its ICD.
To investigate whether this processing might be modulated by the glycosylation state of LRP1 we used a CHO-K1 derived mutant cell line, ldlD, which lacks the enzyme galactose/Nacetylgalactosamine 4-epimerase. These cells are unable to synthesize galactose and its derivatives when maintained in galactose-free cell culture medium. Thus, the cells are unable to synthesize N-acetylgalactosamine-based O-linked sugar chains, as well as galactose-or N-acetylgalactosamine-containing complex N-linked oligosaccharides (25). ldlD cells were originally identified, because of their relative deficiency in LDL receptors. This protein is highly unstable when it is not Oglycosylated and steady state levels of the LDL receptor are greatly reduced in ldlD cells ( Fig. 2A). In contrast, LRP1 levels are only slightly reduced, but its electrophoretic mobility is increased because of hypoglycosylation in ldlD cells (Fig. 2B).
To examine the rate of proteolytic processing of hypoglycosylated LRP1, we transfected ldlD and CHO-K1 control cells with a plasmid encoding an LRP-Gal4/VP16 fusion protein. The Gal4/VP16 domain is fused to the LRP1 carboxyl terminus and drives transcription of a Gal4-luciferase reporter gene if released from the membrane by ␥-secretase-dependent cleavage of LRP1. Co-transfection of CHO-K1 cells with the LRP-Gal4/ VP16 and the Gal4-luciferase reporter plasmids resulted in a robust stimulation of luciferase expression, indicating that LRP1 is proteolytically processed and that its ICD is released in these cells. Under the same conditions severalfold higher reporter gene activity, indicating increased LRP1 processing, is found in ldlD cells. This suggests that LRP1 processing is increased if the protein is hypoglycosylated. With an analogous LDLR-Gal4/VP16 construct, no significant reporter gene expression was noted either in CHO-K1 or ldlD cells. This makes it unlikely that nonspecific protein turnover in ldlD cells could be the reason for the accelerated release of LRP-Gal4/VP16 ICD (Fig. 2C).

Hypoglycosylation of LRP1 Facilitates Extracellular Cleavage Resulting in Increased Production of ␥-Secretase Substrate
Precursor-To confirm that the mechanism of increased LRP1 ICD release in the ldlD line requires ␥-secretase activity, we treated LRP-expressing ldlD and CHO-K1 control cells with the ␥-secretase inhibitor L-685,458. This substance inhibits the final proteolytic cleavage step of LRP, which mediates the release of the ICD. Treatment with L-685,458 caused accumu-lation of a 25-kDa LRP1 fragment that is generated by ectodomain shedding and forms the normal substrate for ␥-secretase. Fig. 2D shows a dose-dependent increase in the accumulation of the 25-kDa fragment in membrane preparations from ldlD and CHO-K1 cells treated with various concentrations of the inhibitor. In the ldlD cells the 25-kDa fragment accumulated to a greater extent than in the CHO-K1 cells, indicating that the extracellular cleavage step that leads to ectodomain shedding is indeed enhanced in glycoslylation-deficient ldlD cells (Fig.  2D). Thus, the rate of ␥-secretase-dependent LRP1 ICD release is greater in the hypoglycosylated state because of increased substrate availability.
Proteolytic Release of the ApoER2 ICD-We next sought to determine the role of tissue-specific differential glycosylation for the metabolism of other members of the LDL receptor gene family. We focused on apoER2, because it is expressed in two splice forms in the brain, either containing or not containing the O-linked sugar domain. A first indication for a regulatory role of differential glycosylation for apoER2 processing came from the analysis of neuronal protein extracts from wild type and apoER2 knockout mice. An antibody directed against the carboxyl terminus of apoER2 detected not only the full-length receptor but also a considerably smaller protein of ϳ25 kDa in lysates of cultured neurons from wild type animals, but not in lysates from apoER2 knockout mice (Fig. 3A), indicating that the smaller band is indeed derived from apoER2. Because the extracellular domain of some splice variants of apoER2 can be cleaved within the secretory pathway (30), we hypothesized that apoER2, like LRP1, might also be proteolytically processed sequentially, possibly resulting in release of its ICD. To ad- dress whether ␥-secretase might be involved, we treated cultured wild type neurons with the ␥-secretase inhibitor DAPT to test whether the 25-kDa fragment would accumulate, as would be expected if it were a ␥-secretase substrate. As shown in Fig.  3B, treatment with the inhibitor did indeed lead to further accumulation of this fragment. Under these conditions a second smaller fragment of ϳ18 kDa was also detected with the carboxyl-terminal antibody. To determine whether this smaller fragment corresponded to a known form of the cytoplasmic domain that lacks a 59-amino acid insert, which is encoded by a single alternatively spliced exon (7), we probed the immunoblot with a specific antibody directed against this insert. As expected, only the larger 25-kDa fragment, but not the smaller 18-kDa fragment was detected by this antibody, confirming that the smaller fragment corresponded to the tail lacking the insert (Fig. 3C).
To obtain further evidence for ␥-secretase-mediated release of the apoER2 ICD from the plasma membrane, we used reporter gene assays similar to those described for LRP1 (20). Two different constructs were used to express an apoER2-Gal4/ VP16 fusion protein. They were based on apoER2 splice variants that either contain or do not contain a furin processing site following the ligand binding domain (21). Transfection of HEK 293 cells with either of these two constructs and a luciferase reporter plasmid revealed reporter gene activation that was caused by apoER2 ICD release (Fig. 3D). Reporter gene activation was comparable for both splice variants suggesting that processing of the extracellular domain by furin is not required for apoER2 ICD release.

Role of the O-Linked Sugar Domain for ApoER2
Processing-We next used the ldlD cells to examine the role of Olinked glycosylation on apoER2 ICD release. In membrane preparations from ldlD cells apoER2 migrates abnormally fast because of the absence of O-linked glycosylation. In addition, the steady state levels of the protein are slightly reduced (Fig.  4A). When CHO-K1 and ldlD cells were transfected with the apoER2-Gal4/VP16 constructs containing or lacking the furin processing site and with the luciferase reporter plasmid, reporter gene activation in ldlD cells was increased severalfold over the CHO-K1 cells. Again, reporter gene activation was comparable whether the furin site was present or not and comparably increased in the ldlD cells, indicating that both proteins are preferentially processed to a similar extent when they are hypoglycosylated (Fig. 4, B and C).
ldlD cells cannot synthesize O-linked sugar chains because they are unable to form N-acetylgalactosamine, the monosaccharide that is first linked to the protein chain. However, this defect also affects N-linked branched glycosylation, which could play a role in apoER2 processing. To better differentiate between the effects of both types of carbohydrate modifications on apoER2 processing, we constructed an apoER2-Gal4/VP16 fusion protein that lacked the O-linked sugar domain. Reporter gene activation in 293 cells transfected with this construct was several times higher than in cells transfected with the O-linked sugar domain containing receptor (Fig. 5). This result shows that the O-linked sugar domain does indeed play an important role in the regulation of apoER2 processing.
Hypoglycosylation Leads to Increased Extracellular Cleavage of ApoER2-When apoER2-expressing CHO-K1 and ldlD cells were treated with a ␥-secretase inhibitor, the ϳ25-kDa fragment accumulated to a much higher extent in ldlD than in the CHO-K1 cells (Fig. 6). The same was observed for LRP1 (Fig.  2D). This indicates that for apoER2, as for LRP1, the glycosylation state of the receptor regulates the extracellular cleavage step and thereby the generation of the COOH-terminal fragment, which serves as the substrate for ␥-secretase.

DISCUSSION
A growing number of cell surface proteins are now recognized to undergo sequential proteolytic processing that culminates in the release of their ICD (reviewed in Ref. 31). This property is also shared by some members of the LDL receptor gene family, i.e. LRP1 (20), and as shown in this study, the apoER2. How this processing of lipoprotein receptors, which involves cleavage by ␥-secretase, is regulated is largely unknown. In the present study we have shown that differential glycosylation of LRP1 and apoER2, combined with differential splicing in the case of the latter, can broadly regulate ICD release, and thus potential signaling pathways that depend upon the release of the cytoplasmic domain from the plasma membrane.
Treatment of cultured primary embryonic neurons with ␥-secretase inhibitors resulted in the accumulation of a carboxyl-terminal fragment of apoER2. The size of this fragment suggests that the penultimate proteolytic processing step that precedes cleavage by ␥-secretase occurs within the extracellular domain of apoER2 in close proximity to the plasma membrane. Processing of apoER2 thus occurs in a manner that is analogous to that of amyloid precursor protein, Notch, LRP1, and a rapidly growing number of other cell surface proteins that all undergo ␥-secretase-mediated processing after the bulk of their extracellular domains has been shed (31).
The functional consequence of this complex proteolytic machinery has so far been studied primarily on the example of the Notch family of developmentally regulatory proteins. After ligand-induced release from the plasma membrane, the Notch ICDs translocate to the nucleus where they regulate target gene transcription (reviewed in Ref. 14). Whether proteolytic processing of apoER2 has similar functional implications that may or may not be related to its role in Reelin signaling and the regulation of cortical lamination (1) remains to be elucidated. In preliminary experiments, however, we did not detect regulation of apoER2 processing by Reelin (data not shown).
Our results show that proteolytic processing of both LRP1 and apoER2 can be regulated by the glycosylation state of the receptors. All members of the LDLR family are post-translationally modified by N-linked glycosylation. In addition, the LDL receptor, VLDLR, and apoER2 also possess O-linked glycosylation domains that immediately precede the transmembrane segment on the extracellular side of the plasma membrane (32)(33)(34). It is encoded by a single exon, which is alternatively spliced in the VLDLR and apoER2 mRNA (6 -12). Glycosylation is known to promote the stability of the LDL receptor in particular (13), but neither the mechanism by which this is achieved nor the significance of the tissue-specific distribution of receptor splice variants containing or lacking the O-linked sugar domain have been investigated in any detail.
We have shown that LRP1 is also differentially glycosylated in different tissues, and that the rate of proteolytic processing of both, LRP1 and apoER2, is dependent upon their glycosylation state. If the proteins are hypoglycosylated, because of variances in carbohydrate additions such as in LRP, or through differential splicing of the O-linked sugar domain as in the case of apoER2, the receptors are more susceptible to proteolytic cleavage of their ectodomain, which in turn results in accelerated release of their ICD. Thus, a functional consequence of differential receptor glycosylation is regulation of proteolysis and ICD release.
The mechanisms and in particular the regulation of the priming cleavage of the receptors' extracellular domains and how this is modulated by the glycosylation state require further investigation. Members of the family of disintegrins and metalloproteinases (ADAMs) that are associated with the cell surface have been shown to be involved in some cases (35)(36)(37). It is likely that glycosylation serves as a barrier to limit access of these shedding enzymes to their substrates. The O-linked sugar domain illustrates this point particularly well. Its location close to the plasma membrane and the attachment of numerous carbohydrate side chains, resulting in an apparent FIG. 6. Increased generation of the apoER2 COOH-terminal fragment in ldlD cells. CHO-K1 and ldlD cells were transfected with 7.5 g of pApoER2 (pCLneo-LR5B-⌬-4 -6) per 100-mm dish. 12 h later cells were treated with 0, 100 nM, 1 M, or 2.5 M ␥-secretase inhibitor L-685,458 for 24 h. Cell membranes were then prepared and analyzed by SDS-gel electrophoresis and immunoblotting with the carboxyl-terminal apoER2 antibody. molecular weight increase of ϳ40,000 (7), suggest a role as a "ruff" that protects the "neck" of the receptors, as they protrude from the plasma membrane, from proteolytic attack.
Another mechanism of regulation has been described for Notch receptors. In this case, extracellular domain cleavage follows the binding of membrane-bound ligands on neighboring cells to the receptor. This is thought to induce a corresponding conformational change at the cleavage site that makes it accessible to the protease (discussed in Ref. 38). Differential glycosylation of the Notch receptors serves to control which ligands can bind most effectively. In particular, the extension of O-linked fucose residues by the N-acetylglucosamine transferase Fringe is crucial to achieve this regulation and further addition of galactose to the carbohydrate chain is also necessary (15)(16)(17)39). Although LRP1 does contain consensus motifs for O-linked fucosylation (C 2 X(3-5)S/TC 3 , where the second and third cysteine residues in epidermal growth factor repeats are separated by 3 to 5 amino acids and a serine or threonine preceding the third cysteine), we did not detect any classical O-linked (i.e. N-acetylgalactosamine-based) carbohydrate side chains in LRP1, and neither did we observe any evidence so far for modulation of extracellular proteolysis by ligand binding (20). It is possible that the primary role of glycosylation of the 85-kDa subunit of LRP1 is to constitutively control differential accessibility of the proteinase in different tissues.
Although the difference in molecular weight between LRP1 from liver and brain can be explained by differences in N-linked glycosylation, additional differential O-fucose-based modifications, which would possibly also be disturbed in the galactosedeficient ldlD cells, cannot be excluded. Further studies are needed to define the exact nature of the carbohydrate chains attached to LRP1 and to examine whether there are extracellular binding partners that recognize the glycosylation pattern and modulate receptor processing, analogous to the regulation of Notch receptors. Finally, tissue-specific modification of carbohydrate attachment sites by genetic recombination techniques may become necessary to refine our understanding of the physiological role of differential glycosylation of LDL receptor family members in vivo.