Perlecan Heparan Sulfate Proteoglycan

Cell surface heparan sulfate proteoglycans (HSPGs) participate in the catabolism of many physiologically important ligands. We previously reported that syndecan HSPGs directly mediate endocytosis, independent of coated pits. We now studied perlecan, a major cell surface HSPG genetically distinct from syndecans. Cells expressing perlecan but no other proteoglycans bound, internalized, and degraded atherogenic lipoproteins enriched in lipoprotein lipase. Binding was blocked by heparitinase, and degradation by chloroquine. Antibodies against β1 integrins reduced initial ligand binding, consistent with their roles as cell surface attachment sites for perlecan. By several criteria, catabolism via perlecan was distinct from either coated pits or the syndecan pathway. The kinetics of internalization (t 1 2 = 6 h) and degradation (t 1 2 ∼ 18 h) were remarkably slow, unlike the other pathways. Blockade of the low density lipoprotein receptor-related protein did not slow perlecan-dependent internalization. Internalization via perlecan was inhibited by genistein but unaffected by cytochalasin D, a pattern distinct from coated pits or syndecan-mediated endocytosis. Finally, we examined cooperation between perlecan and low density lipoprotein receptors and found limited synergy. Our results demonstrate that perlecan mediates internalization and lysosomal delivery that is kinetically and biochemically distinct from other known uptake pathways and is consistent with a very slow component of HSPG-dependent ligand processing found in vitroand in vivo.

cells, platelet secretory products, anticoagulants, extracellular matrix components, several molecules implicated in Alzheimer's disease, lipolytic enzymes, and certain lipoproteins (reviewed in Refs. [2][3][4][5][6][7]. Removal of cell surface heparan sulfate abolishes normal cellular handling of essentially every one of these ligands. Three models for the processing of ligands bound to cell surface HSPGs have been proposed. The simplest conceptually is endocytosis mediated directly by the HSPGs, a process for which substantial evidence now exists (reviewed in Refs. 4 and 8). The second model involves an initial binding of ligands to HSPGs followed by recruitment of auxiliary cell surface molecules, such as LDL receptor family members, that then mediate ligand internalization (9 -12). The precise mechanism for this cooperation between HSPGs and LDL receptor family members may involve ligand transfer or ternary complex formation, although no direct evidence to date can distinguish between these possibilities. The third model involves binding to HSPGs followed by conformational changes that then allow the ligand to interact with a high affinity, proteinaceous receptor, such as the HSPG-dependent binding of fibroblast growth factor-2 to its tyrosine kinase-linked receptor (13,14). For each of these three models, but particularly the first, the nature of the cell surface HSPG is a key determinant of subsequent cellular catabolism (4) (see also Refs. 15 and 16).
We previously reported that lipoprotein lipase (LpL), a heparin-binding protein, enhances cellular catabolism of atherogenic lipoproteins in vitro by bridging between the lipoproteins and cell surface HSPGs (17,18). These findings were promptly extended to other lipoproteins, different cell types, and additional bridging molecules (19 -23), and there is now evidence for HSPG-dependent catabolism of lipoproteins in vivo (4, 24 -28) (for a review of the development of this field, see Ref. 29). Interestingly, HSPG-mediated catabolism of LpL-enriched lipoproteins in vitro exhibits at least two kinetically distinct components, one leading to lysosomal degradation of ligand within 4 h and the other apparently much slower (18). Very slow catabolism of HSPG-bound ligands has also been found in the liver in vivo (28), particularly in the absence of LDL receptors (24). In exploring the molecular basis of these observations, we recently discovered that the syndecan family of transmembrane HSPGs directly mediates efficient endocytosis of multimeric ligands, with a t 1/2 for internalization of ϳ1 h, through a pathway that is triggered by clustering, acts independently of coated pits, and relies upon cholesterol-rich membrane rafts and the cytoskeleton (4,29,30). Importantly, syndecan-mediated ligand catabolism leads to nearly complete degradation in lysosomes within 4 h (29).
In the current study, we focused on ligand catabolism via the perlecan HSPG, a completely extracellular molecule that is genetically distinct from the syndecan family (31, 32). Perlecan is secreted but then adheres to the cell surface, in part by binding to integrins (33)(34)(35), or it incorporates into the basement membrane or other extracellular matrix (36). Perlecan is an attractive candidate for mediating ligand catabolism, because this molecule is abundant in the hepatic space of Disse and within the arterial wall (32,35,37,38), two important locations of HSPG-mediated ligand catabolism in vivo. Here, we demonstrate that cell surface perlecan HSPG mediates a distinct pathway for ligand internalization and lysosomal delivery that is kinetically and biochemically distinct from either coated pit internalization or syndecan-mediated catabolism but is consistent with the second, slower pathway for direct HSPGmediated processing of ligands.

EXPERIMENTAL PROCEDURES
Preparation of Reagents-Unless otherwise indicated, chemicals of analytical grade were purchased from Sigma. Bovine LpL (EC 3.1.1.34) was purified from milk by heparin-agarose chromatography (39,40) with minor modifications. 125 I-LpL, prepared by using lactoperoxidase and glucose oxidase enzymes (41), was kindly provided by Drs. S. K. Fried and S. Papaspyrou-Rao. LDL was isolated from fresh human plasma by ultracentrifugation (1.019 Ͻ d Ͻ 1.063 g/ml) and then radioiodinated by the iodine monochloride method (42). To allow examination of LDL receptor-independent pathways, most preparations of 125 I-labeled LDL were reductively methylated ( 125 I-mLDL) to modify approximately 35% of the lysine residues, thereby abolishing LDL receptor binding (43,44). The 39-kDa receptor-associated protein (RAP), which is the human homologue of mouse heparin-binding protein 44 (45) and a universal inhibitor for ligand binding to the LDL receptorrelated protein (LRP) (46), was expressed as a glutathione S-transferase fusion protein in bacteria, using a construct kindly provided by Dr. D. Strickland (47). Inhibitory antibodies against ␤ 1 integrins were from Transduction Laboratories (Lexington, KY). Genistein was from Calbiochem-Novabiochem.
Cultured Cells-Because of its extremely large size (ϳ467 kDa), no expression vectors to date exist for the perlecan core protein. Instead, we relied upon a variant colon carcinoma cell line, WiDr (ATCC no. CCL 218, also known as HT-29) (48), that synthesizes perlecan but no other proteoglycans and incorporates [ 35 S]sulfate nearly completely (Ͼ95%) into the heparan sulfate side chains covalently linked to the perlecan core protein (49 -51). Other proteoglycans, if present, are below the limits of detection. This key property of WiDr cells has been verified by several independent methods, including the identification of the perlecan core protein based on its unusual size and reactivity to specific antibodies and identification of the side chains by digestion with nitrous acid or heparitinase (49 -51). The Chinese hamster ovary (CHO) cell line, transfected with an expression vector for the human syndecan-1 core protein (CHO-Synd1), was described previously (29). The CHO mutant line pgsD-677 (CHO-677), which is specifically deficient in both N-acetylglucosaminyltransferase and glucuronosyltransferase activities and hence lacks heparan sulfate (52,53), was generously supplied by Dr. J. D. Esko. WiDr cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, and both types of CHO cells were maintained in Ham's F-12 medium supplemented with 10% fetal bovine serum.
Cellular Uptake and Degradation of 125 I-Labeled Ligands-Cells were grown in 15-mm wells to approximately 90% confluence in serumsupplemented media. For all experimental incubations, cells were changed to 0.5 ml of the corresponding serum-free medium supplemented with 0.2% bovine serum albumin (Sigma catalog no. A-8806), 125 I-labeled lipoproteins, or 125 I-LpL (5 g of protein/ml) and either unlabeled LpL (5 g of protein/ml) or a matching volume of lipase buffer (54). Cells were incubated at 37°C for 5 or 21 h in these media, and then surface-bound (heparin-releasable), intracellular (heparin-resistant), and degraded (assessed by trichloroacetic acid-soluble, CHCl 3 -insoluble radioactivity in media) ligand was quantitated (29,42,54). To assess the efficiency of endocytosis, we calculated ligand internalization as the sum of intracellular accumulation plus degradation, as described previously (29). Internalization calculated in this fashion takes into account ligand still within the cells, as well as ligand that had been taken up by cells but then degraded into amino acids, which the cells release to the culture media.
In initial experiments, we followed the common procedure of placing labeled ligands into the cell culture medium and then leaving a continuous supply of unbound ligands throughout the course of the incubations at 37°C (42). Subsequently, to examine in detail the kinetics of ligand internalization and degradation, labeled ligands were incubated with cells in serum-free medium for 1 h at 4°C, to allow surface binding without further catabolism, and the cells were washed at 4°C to remove unbound material. Fresh media at 37°C with no ligands were then added, and incubations were continued at 37°C for the indicated times (29). Assays for surface-bound, intracellular, and degraded ligand were then performed, and ligand internalization was calculated as above. Although deiodination of labeled protein has been shown to be at most a minor cellular process during incubations at 37°C lasting up to 5 h (42), our control experiments indicated accumulation of CHCl 3 -soluble radioactivity in culture media during prolonged incubations (overnight or longer), due to cell-mediated deiodination of extracellular 125 I-amino acids. Thus, our measurements of degradation during prolonged incubations to characterize the kinetics of the perlecan pathway included all trichloroacetic acid-soluble 125 I radioactivity. In some experiments, trichloroacetic acid-precipitable radioactivity in the media was also quantified, as an indication of retroendocytosis or desorption of intact ligand from the cell surface during the incubation at 37°C.
To verify the role of heparan sulfate side chains of perlecan in binding and catabolism of LpL⅐ 125 I-mLDL, cells were pretreated for 1 h at 37°C with heparitinase (5.5 units/ml; Sigma catalogue no. H 8891), and heparitinase was kept in the incubation medium at 37°C with labeled ligand until the end of the experiment, to prevent reassembly of HS side chains (see Ref. 18). Because heparan-binding components of serum may interfere with heparitinase digestion, these cells were placed into serum-free medium at least 1 h before the heparitinase pretreatment and kept serum-free until the end of the experiment (55). To verify the role of lysosomes in ligand degradation, we incubated cells at 37°C with ligand in the presence of chloroquine (150 M), an inhibitor of lysosomal proteases (42). To manipulate LDL receptors (42), some experiments with WiDr cells included an 18-h preincubation without serum supplementation, in the absence or presence of a mixture of 2 g of 25-hydroxycholesterol with 40 g of cholesterol per ml.
To examine intracellular processes involved in ligand catabolism via the perlecan HSPG, LpL⅐ 125 I-mLDL was bound to the cell surface at 4°C, and residual unbound ligand was washed away, as described above. Specific inhibitors were then added simultaneously with fresh medium at 37°C, without additional ligand, and the extent of internalization of LpL⅐ 125 I-mLDL from the cell surface was assessed after 2 h at 37°C. The inhibitors included freshly prepared genistein (0 -400 M), a tyrosine kinase inhibitor (56) that blocks syndecan-mediated endocytosis (29), and cytochalasin D (0 -2 M), which disrupts the cytoskeleton (57) and inhibits the syndecan internalization pathway (29). In control experiments to examine the effects of these inhibitors on internalization mediated by syndecan HSPGs or by LDL receptors, which proceed at different rates, we followed our previous approach of assessing the extent of ligand internalization after 45 min and 10 min, respectively (29). This design produces similar degrees of ligand internalization for all three pathways in the absence of inhibitors (ϳ30 -50% of the initially surface-bound ligand becomes internalized during the incubations at 37°C). Because coated pit-mediated internalization is so rapid (t 1/2 ϳ 5-10 min) (58,59) and may finish before inhibitors have time to act, our experiments examining the LDL receptor pathway also included a 30-min preincubation at 37°C with each inhibitor before chilling the cells to 4°C for binding 125 I-labeled native LDL to the cell surface (29). The effect of genistein has been reported to fade (29,60), so cells were exposed to this agent for a maximum of 2 h.
We also examined the effects of excess, unlabeled RAP on perlecanmediated internalization. To maximize its effect, RAP (50 g/ml) was present in these experiments during three periods: a 30-min preincubation at 37°C, the incubation at 4°C to allow surface binding of LpL-enriched 125 I-mLDL, and the final incubation at 37°C to allow cellular catabolism of surface-bound LpL⅐ 125 I-mLDL. Studies of the effects of antibodies against ␤ 1 integrins (10 g/ml) followed the same protocol used for RAP. All results for 125 I-lipoprotein catabolism were normalized to cellular protein (61), which averaged 181 g/well (WiDr) and 67 g/well (all types of CHO cells).
Supplementation of HS-negative CHO Cells with Perlecan HSPG-To perform these experiments, we took advantage of the fact that most perlecan from WiDr cells is secreted, not surface-bound, so WiDr-conditioned medium is a rich source of this molecule. Moreover, even cell surface-bound perlecan HSPG is still entirely extracellular, so other cell types can be enriched in this molecule simply by incubation in WiDr-conditioned medium. As our target cell for enrichment with perlecan HSPG, the CHO-677 mutant was chosen for several reasons. First, LDL receptors (62), the LRP (63), and direct syndecan-mediated endocytosis (29) have all been examined in CHO cells, thereby allowing comparison with perlecan-mediated catabolism in a single cell type.
Second, in our preliminary control experiments, Northern blotting of CHO mRNA revealed no detectable levels of the perlecan core protein message (data not shown). Third, the CHO-677 mutant makes no HS side chains (52), which eliminates background from other HSPGs.
WiDr-conditioned medium was prepared by incubating six 100-mm dishes of confluent WiDr monolayers for 24 h at 37°C in serum-free DMEM. This conditioned medium was centrifuged for 30 min at 1500 rpm to remove cell debris and then concentrated 10-fold using Macrosep centrifugal concentrators (50-kDa cut-off, Pall Filtron, Northborough, MA). To supplement HS-deficient CHO-677 cells with perlecan HSPG, we incubated these cells with the concentrated conditioned medium from WiDr cells for 1 h at 37°C to allow perlecan to bind to the cell surface, followed by three washes in phosphate-buffered saline/bovine serum albumin. Cellular catabolism of LpL⅐ 125 I-mLDL complexes was then measured in fresh serum-free medium, as described above, and results are reported as the portion attributable to the addition of perlecan.
Calculations of Catabolic Components-In the majority of our experiments, we examined 125 I-mLDL binding, internalization, and degradation in the presence and absence of unlabeled LpL, and LpL-dependent catabolism of this lipoprotein was calculated by subtracting the values obtained in the absence of LpL (LpL-independent catabolism) from those obtained in the presence of LpL (total catabolism), as described previously (18,29). In our later experiments comparing 125 I-labeled native LDL and 125 I-mLDL in the presence and absence of LpL, we computed the contributions of four possible catabolic components: 1) the LDL receptor-independent, LpL-independent component, which is usually referred to as nonspecific uptake or assay background and was measured by the catabolism of 125 I-mLDL in the absence of LpL; 2) the LDL receptor-dependent, LpL-independent component, which is the classical LDL receptor pathway and was computed by the difference between the catabolism of 125 I-LDL versus 125 I-mLDL; 3) the LDL receptor-independent, LpL-dependent component, which involves cell surface HSPGs (29) and was computed by the difference between the catabolism of 125 I-mLDL in the presence versus the absence of LpL, as just noted; and 4) a synergistic component, which requires cooperation between LDL receptors and LpL and was computed as the increase in 125  values. For comparisons between a single experimental group and a control, the unpaired, two-tailed t test was used.

Involvement of Perlecan in Ligand Catabolism-We began by
determining if WiDr cells, which express perlecan but no other proteoglycans, exhibit enhanced catabolism of 125 I-mLDL upon the addition of LpL, a molecule that bridges between lipoproteins and cell surface HSPGs. After a 5-h incubation, cell surface binding, internalization, and degradation of 125 I-mLDL by these cells were substantially increased by the presence of LpL (Fig. 1). Treatment of the cells with heparitinase inhibited LpL-dependent binding, internalization, and degradation of 125 I-mLDL by 93.5% Ϯ 0.5%, 81.6% Ϯ 3.5%, and 88.9% Ϯ 3.2%, respectively, confirming a key role for the heparan sulfate side chains of perlecan. The addition of chloroquine (150 M) inhibited ligand degradation by 87.5 Ϯ 3.4%, as assayed by the reduction in cellular release of 125 I-labeled amino acids into the medium, which was accompanied by a corresponding increase in the accumulation of intracellular radioactive material, indicating the involvement of lysosomes in this pathway.
Interestingly, after the 5-h incubation of LpL⅐ 125 I-mLDL with WiDr cells at 37°C, the ratio of surface-bound to internalized to degraded ligand was 58:42:6; i.e. a large portion of the ligand associated with cells still remained on the cell surface, and only small, although easily detected, amounts of internalized material had been degraded into amino acids (Fig.  1, right). Even after a 21-h incubation of LpL⅐ 125 I-mLDL complexes with WiDr cells, the distribution of this ligand was 31:69:21 (surface:internalized:degraded). These results are in striking contrast with the speed of endocytosis exhibited by known classes of lipoprotein receptors, such as LDL receptor family members (t 1/2 ϳ 5-10 min), which enter cells via coated pits (59), and the syndecan HSPG family (t 1/2 ϳ 1 h), which enters independently of coated pits (4,29,30). Moreover, endocytosis via all of these other receptors is quickly followed by lysosomal delivery and ligand degradation (29,59).
Kinetics of Perlecan-mediated Internalization-To accurately assess the kinetics of perlecan-mediated catabolism and compare it with other uptake pathways, we followed the cellular catabolism of 125 I-labeled lipoproteins that had been bound to the cell surface at 4°C, followed by warming to 37°C to allow the cells to process this material. Three types of cells were studied: WiDr cells; CHO 677 cells that we had supplemented with perlecan HSPG; and CHO-Synd1 cells, which overexpress the syndecan-1 HSPG. LpL-enriched 125 I-mLDL, which binds HSPGs but not LDL receptors, and 125 I-labeled native LDL (5 g/ml), which enters cells primarily via LDL receptors at this ligand concentration (42), were compared.
LpL-dependent internalization of 125 I-mLDL by WiDr cells was monoexponential and proceeded with t 1/2 of 6 h (Fig. 2, A  and B; see legend), while degradation of 50% of the ligand required roughly 18 h (Fig. 2C). There was no detectable retroendocytosis of internalized ligand or desorption from the cell surface after the initial 30 min at 37°C (data not shown). Catabolism of 125 I-labeled native LDL (5 g/ml) exhibited extremely rapid kinetics in WiDr cells (see legend to 2), suggesting distinct pathways. Syndecan-mediated catabolism of LpL-enriched 125 I-mLDL by CHO-Synd1 cells also exhibited kinetics consistent with prior literature: t 1/2 for internalization ϭ 1 h, and about 50% of the ligand was degraded after 2 h (Fig. 2, open diamonds, and Ref. 29). These parameters are completely distinct from the kinetics of perlecan-mediated catabolism in WiDr cells.
To verify that this catabolic pathway is determined by perlecan itself and is not peculiar to certain cell lines or types of ligands, we examined the perlecan pathway in CHO cells. Pretreatment of CHO-677 cells with concentrated WiDr-conditioned medium resulted in roughly a 2-fold increase in surface binding of LpL⅐ 125 I-mLDL complexes. As expected, this increase was nearly completely abolished (90.3 Ϯ 3.6% inhibition) when the concentrated conditioned medium had been digested with heparitinase beforehand. Most importantly, the kinetics of ligand internalization and degradation in the perlecan-treated CHO-677 cells were similar to the pattern seen in the WiDr line itself (compare filled triangles with filled squares in Fig. 2) and were clearly far slower than the pathways mediated by LDL receptors in WiDr cells (see the legend to Fig. 2) or by syndecan-1 in CHO-Synd cells (Fig. 2, open diamonds). Because LpL itself is an important ligand for cell surface HSPGs, yet this molecule is far smaller than LpL⅐ 125 I-mLDL complexes, we examined perlecan-mediated catabolism of 125 Ilabeled LpL in WiDr cells and found essentially the same kinetics as presented above for LpL⅐ 125 I-mLDL complexes (for simplicity, these data are not displayed). These results indicate that the slow perlecan-mediated pathway for ligand catabolism is distinct from other internalization pathways and appears to be independent of cell type or ligand size.
Because the HS side chains of internalized perlecan have been reported to undergo a partial degradation to oligosaccharides before entering the lysosomes, where complete degradation to monosaccharides then occurs (50), we sought to determine if the protein component of LpL⅐ 125 I-mLDL complexes that enter cells via perlecan also degrades in stages. LpLenriched 125 I-mLDL was bound to the surface of WiDr cells at 4°C, unbound ligand was washed away, and then the cells were incubated at 37°C for up to 22 h. At the end of this incubation, residual surface-bound material was removed with a heparin wash at 4°C, and the cells were solubilized in SDS and subjected to polyacrylamide gel electrophoresis, followed by autoradiography. As shown in Fig. 3, a significant amount of 125 I-labeled methylated apoB was still intact inside the cells even after 22 h, and there were no detectable traces of partially degraded products. Thus, degradation of the protein component of 125 I-mLDL internalized via perlecan does not appear to proceed in stages involving long-lived peptide fragments.
Cellular Mechanisms of Perlecan-mediated Internalization-We next sought to characterize the cellular mechanisms involved in the perlecan pathway for ligand internalization. Because endocytosis of ligands bound to cell surface HSPGs in a number of experimental systems has been reported to depend on LRP (9 -12), we examined the effect of RAP, the universal competitor for LRP binding, on perlecan-mediated internalization. Saturating concentrations of RAP (50 g/ml) slightly reduced the surface binding of LpL-enriched 125 I-mLDL to WiDr cells (by 20.7 Ϯ 3.3%), while having no effect on the rate of internalization (54.7 Ϯ 2.8 and 55.2 Ϯ 3.4% of surface-bound ligand was internalized by 5 h in control and RAP-treated cells, respectively, p Ͼ 0.9). These data do not support a role for LRP in the internalization of ligands bound to perlecan HSPGs.
Because adherence of cells to immobilized perlecan has been reported to be mediated in part by ␤ 1 integrins (33-35), we examined the role of these molecules in perlecan-mediated ligand catabolism. Pretreatment of WiDr cells with anti-␤ 1

FIG. 2. Time course of perlecan-mediated ligand catabolism by different cell lines, with comparisons with catabolism mediated by syndecan or by LDL receptors. WiDr cells (filled squares), perlecan-enriched CHO-677 cells (filled triangles), and CHO-Synd1 cells (open diamonds)
were incubated with 125 I-mLDL at 4°C in the presence or absence of LpL, unbound ligand was washed away, and then at t ϭ 0 the cells were warmed to 37°C to allow ligand catabolism. The difference between catabolism in the presence versus the absence of LpL is displayed for each of these cell lines. y axis units are LpL-dependent cell surface binding (A), internalization (B), or degradation (C) expressed as a percentage of total LpL-dependent cell association, where total LpL-dependent cell association was calculated as the sum of LpL-dependent surface binding plus intracellular accumulation plus degradation. A, which shows the amount of ligand remaining on the cell surface as a function of time, is displayed as a semilog plot. Linear regression of semilog transformed data from A was used to calculate the following parameters: perlecan pathway in WiDr cells, t 1/2 ϭ 6 h, r ϭ Ϫ0.999; perlecan pathway in enriched CHO-677 cells, t 1/2 ϭ 5 h, r ϭ Ϫ0.995; syndecan pathway in CHO-Synd1 cells, t 1/2 ϭ 1 h, r ϭ Ϫ0.989. B and C show linear plots. For an additional comparison, catabolism of surface-bound 125 I-labeled native LDL by WiDr cells exhibited a t 1/2 for internalization of ϳ9 min, and degradation was essentially completed by 2 h. In the three panels, the numerical values corresponding to 100% on the y axes were 568 Ϯ 46 ng/mg for CHO-Synd, 177 Ϯ 3 ng/mg for WiDr, and 34 Ϯ 1 ng/mg for perlecan-enriched CHO-677 cells. For simplicity, error bars are not shown, but errors were Ͻ12% of each mean value. antibodies reduced cell surface binding of LpL⅐ 125 I-mLDL complexes at 4°C by 34 Ϯ 2% compared with control cells preincubated with nonimmune IgG (p Ͻ 0.005). There was, however, no reduction in the percentage of surface-bound lipoprotein that subsequently became internalized during a 2-h incubation at 37°C. These results are consistent with the model that perlecan adheres to the cell surface in part through integrins.
We also examined the effects of metabolic inhibitors that have known actions on other internalization pathways. Genistein, a tyrosine kinase inhibitor, caused a dose-dependent inhibition of ligand internalization via perlecan (Fig. 4A, filled squares) but had at most a minor effect on coated pit-mediated internalization of 125 I-labeled native LDL (Fig. 4A, ϫ, and Ref. 29), indicating independence of these two pathways. Cytochalasin D, which disrupts the actin cytoskeleton, consistently caused small increases in perlecan-mediated internalization of about 5-20% by 5 h (Fig. 4B) but strongly inhibited the syndecan pathway and, to a lesser extent, coated pit internalization of 125 I-LDL, as previously reported (29,64). Thus, perlecanmediated ligand internalization depends on tyrosine kinases but does not require an intact cytoskeleton, which is a distinct pattern from other known internalization pathways.
Limited Cooperation between Perlecan and the LDL Receptor-The prolonged cell surface residence time of ligands bound to perlecan seems ideal to encourage interactions with other cell surface molecules. Therefore, we used our experimental system to revisit the controversial proposal that LDL receptors mediate the internalization of lipoproteins that have initially bound to cell surface HSPGs (see Refs. 18 -21, 29, 65, and 66). Here, we compared the catabolism of LpL-enriched 125 I-mLDL, which binds perlecan HSPGs but not LDL receptors, versus LpL-enriched 125 I-labeled native LDL, which is a ligand for both.
To maximize LDL receptor expression, WiDr cells were preincubated for 18 h in medium without cholesterol. As expected, this pretreatment resulted in much more cell surface binding of 125 I-labeled native LDL (33.5 Ϯ 0.8 ng/mg) compared with 125 I-mLDL (8.5 Ϯ 2.4 ng/mg) when these labeled lipoproteins were added to cells at 4°C in the absence of LpL. The addition of LpL substantially increased cell surface binding of both of these lipoproteins at 4°C, to 195 Ϯ 3.5 ng/mg for LpL⅐ 125 I-LDL and 165.7 Ϯ 6.1 ng/mg for LpL⅐ 125 I-mLDL. These values indicate that the increase in cell surface binding of 125 I-LDL at 4°C upon the addition of LpL (195.4 Ϯ 3.5 minus 33.5 Ϯ 0.8, equals 161.9 Ϯ 3.6 ng/mg) was statistically indistinguishable from the LpL-dependent increase in 125 I-mLDL binding (165.7 Ϯ 6.1 minus 8.5 Ϯ 2.4, equals 157.2 Ϯ 6.6 ng/mg, p Ͼ 0.5). In other words, LpL had the same quantitative effect on cell surface binding at 4°C, regardless of the presence of LDL receptorbinding motifs on the lipoprotein or prior treatment with methylation reagents. The only measurable difference between 125 I-LDL and 125 I-mLDL in this circumstance was their binding to cell surface LDL receptors, a finding that strongly supports the validity of side-by-side comparisons of these two lipoprotein preparations.
Using the arithmetic described under "Experimental Procedures," the initial binding at 4°C exhibited three detectable components: a nonspecific background (component 1 ϭ 8.5 Ϯ 2.4 ng/mg), classical LDL receptor binding (component 2 ϭ 33.5 Ϯ 0.8 minus 8.5 Ϯ 2.4, equals 25.0 Ϯ 2.5 ng/mg), and a component that requires LpL but does not involve LDL receptors (component 3 ϭ 165.7 Ϯ 6.1 minus 8.5 Ϯ 2.4, equals 157.2 Ϯ 6.6 ng/mg). There was, however, no statistically significant component that depends on cooperation between LDL receptors and LpL (component 4 ϭ 161.9 Ϯ 3.6 minus 157.2 Ϯ 6.6, equals 4.7 Ϯ 7.5 ng/mg, which is not significantly different from zero). In other words, there was no detectable synergy between LDL receptors and perlecan HSPGs in the initial binding of LpL-enriched 125 I-LDL to the cell surface at 4°C.
Next, unbound ligand was removed by washing, and cells were warmed to 37°C to allow ligand internalization and degradation at sequential time points. Fig. 5A displays the increases in internalization of 125 I-labeled native LDL (inverted filled triangles) and 125 I-mLDL (filled squares) that occurred upon the addition of LpL. Using the nomenclature under "Experimental Procedures," the inverted filled triangles in Fig. 5A represent Ln Ϫ n, and the filled squares represent Lm Ϫ m.
In contrast to cell surface binding at 4°C, in which the addition of LpL had essentially identical effects on 125 I-LDL and 125 I-mLDL, the effect of LpL on internalization and degradation was substantially different between the two labeled particles. LpL-dependent internalization of LpL⅐ 125 I-LDL proceeded with an initially higher rate than LpL⅐ 125 I-mLDL, although the two internalization curves became parallel from about 2-3 h until 15 h (Fig. 5A). Increases in degradation of the two lipoproteins upon the addition of LpL showed a similar pattern: initially greater LpL-dependent degradation of 125 I-LDL, followed by the curves becoming generally parallel between about 5 and 15 h (Fig. 5B).
Using the arithmetic described under "Experimental Procedures," the internalization and degradation of LpL⅐ 125 I-LDL exhibited all four possible components: a small, nonspecific background (component 1), which was Ͻ 10 ng/mg throughout the time course; classical LDL receptor-mediated catabolism (component 2), which peaked at ϳ30 ng/mg during the time course and exhibited the rapid internalization that is charac- teristic of coated pit pathways (t 1/2 Ͻ 10 min); a component that requires LpL but does not involve LDL receptors (component 3), which is displayed by the filled squares in Fig. 5, A and B, and exhibited slow internalization and degradation, as in Fig.  2; and a synergistic component that depends on cooperation between LDL receptors and LpL (component 4), which is displayed in Fig. 5C. The internalization curve in Fig. 5C was calculated as the gap between the two curves in Fig. 5A, i.e. (Ln Ϫ n) Ϫ (Lm Ϫ m). Likewise, the degradation curve in Fig. 5C is the arithmetic difference between the two curves in Fig. 5B.
Several features of this synergistic component (component 4) are apparent. First, although easily detected, it is only a minority of the increase in 125 I-LDL catabolism upon the addition of LpL. The two curves in Fig. 5C reach maxima of approximately 45 ng/mg, whereas the entire LpL-dependent internalization of 125 I-LDL reached approximately 160 ng/mg by the end of the time course (Fig. 5A). In other words, just under 30% of perlecan-bound LpL⅐ 125 I-LDL appeared to participate in component 4 in our system. Thus, most of the LpL⅐ 125 I-LDL complexes bound to perlecan behaved exactly like LpL⅐ 125 I-mLDL bound to perlecan. Second, the kinetics of the synergistic component were faster than perlecan-mediated internalization but slower than the classical LDL receptor pathway. Based on the data in Fig. 5C, it took roughly 45 min for half of the internalization via component IV to occur. Importantly, the synergistic pathway did not exhibit a long delay between internalization and degradation (compare the two curves in Fig.  5C), which is unlike pure perlecan-mediated catabolism and more like the classical LDL receptor pathway. Third, the synergistic component appeared to decrease by the later time points in Fig. 5C (t Ն 15 h), but this is an artifact of the calculations: the combined pathway was completed early in the experiment, whereas slow perlecan-mediated catabolism of LpL⅐ 125 I-mLDL continued throughout the entire time course, eventually allowing LpL⅐ 125 I-mLDL to partly catch up to LpL⅐ 125 I-LDL (Fig. 5, A and B).
To determine biochemical mechanisms of ligand internalization via component 4, we examined its sensitivity to specific inhibitors. For these experiments, WiDr cells with surfacebound ligands were incubated at 37°C for 2 h, a time point when the synergistic component of ligand internalization is easily detected (Fig. 5C). To verify the role of LDL receptors, WiDr cells were preincubated for 18 h in medium enriched in 25-hydroxycholesterol and cholesterol, which abolished most of the synergistic pathway (Fig. 6). Thus, component 4 appears to be a true hybrid pathway: it requires LpL, but also LDL receptor-binding motifs and cell surface LDL receptors. To determine if the intracellular processes of internalization are also hybrid, we examined the effects of genistein and cytochalasin D. Fig. 6 shows that genistein, which inhibits internalization via perlecan but not via the LDL receptor (Fig. 4A), had no inhibitory effect on component 4. Cytochalasin D, which impedes the LDL receptor but not perlecan (Fig. 4B), strongly inhibited internalization via component 4. These results indicate that the synergistic component uses molecular mechanisms distinct from internalization via perlecan alone (component 3) and more similar to LDL receptor-dependent pathways. DISCUSSION Based on the structure of its core protein and the mechanism for membrane attachment, perlecan is distinguished from the two other major classes of cell surface HSPGs, the transmembrane syndecan family and the phosphatidylinositol-anchored glypican family. We have hypothesized (4, 29) that this diver- sity among cell surface HSPGs may provide the molecular basis for our previous finding of at least two kinetically distinct HSPG-mediated pathways for catabolism of LpL-enriched atherogenic lipoproteins (18). A ligand may bind to the HS side chains of several different classes of HSPGs on a cell surface, but subsequent cellular processing appears to be dictated by the core proteins (4,15,29), which are quite diverse (2,4,7). Our current results indicate that the perlecan HSPG efficiently binds ligands at the cell surface and then mediates their uptake into cells through a pathway that involves cell surface ␤ 1 integrins and tyrosine kinases, but is independent of LRP or the actin cytoskeleton. The perlecan pathway for ligand internalization is kinetically and biochemically distinct from other known uptake mechanisms, such as coated pit internalization and endocytosis via syndecan HSPGs. Most unusually, perlecan allows its ligands to remain on the cell surface for comparatively long periods, and even after internalization, the ligands can remain intact for many hours within the cell before degradation in lysosomes.
The kinetics for internalization then lysosomal degradation of ligands via perlecan alone (component 3) is very similar to the turn-over of the heparan sulfate side chains of perlecan in the absence of added ligands (50). Perlecan itself undergoes slow internalization (t 1/2 of about 6 h) and then partial degradation of its HS side chains into oligosaccharides in a prelysosomal compartment and finally complete degradation of the HS chains to monosaccharides in lysosomes with t 1/2 of about 18 h (50). From the similarity to the kinetics of perlecan-mediated ligand catabolism (Fig. 2), we can draw three tentative conclusions. First, the t 1/2 for perlecan-mediated internalization is probably constitutive, with little if any stimulation by ligand binding. This pattern is in direct contrast to endocytosis via syndecan HSPGs, which is greatly accelerated by ligand clustering (29). Second, the extremely slow lysosomal degradation of ligand that had been internalized via perlecan (Figs. 1 and 2) could be the result of some structural interference with lysosomal hydrolases (cf. Ref. 67) but is more likely the consequence of an odd endocytic route involving a long delay in a nonlysosomal intracellular compartment (see Ref. 50). The nature of this putative compartment remains to be determined, although our inability to detect partial degradation products from internalized 125 I-methylated apoB (Fig. 3) suggests the absence of prelysosomal proteases. A prelysosomal compartment may also play a role in the degradation of the side chains of other proteoglycans, although with substantially different kinetics (68). Third, the similarity between perlecan HS turnover (50) and perlecan-mediated ligand catabolism (Fig. 2) is consistent with our other results indicating that transfer of ligand from perlecan to auxiliary cell surface molecules, such as LDL receptor family members, is not involved in the internalization of LpL⅐ 125 I-methylated LDL or 125 I-LpL.
We examined in detail the role of LDL receptor family members in the catabolism of ligands bound to cell surface perlecan. Because LpL has been reported to interact with LRP in vitro (9, 10), we measured cellular processing of LpL-enriched 125 I-mLDL in the presence of RAP, a universal inhibitor of binding to LRP. We found a small inhibitory effect of RAP, the human homologue of mouse heparin-binding protein 44 (45), on the surface binding of LpL⅐ 125 I-mLDL, which might be attributable to competition between LpL and RAP for sites on perlecan HS side chains (4,69). Nevertheless, in our system, RAP had no effect on the internalization rate of surface-bound LpL⅐ 125 I-mLDL. These results exhibit a pattern that is not consistent with models involving internalization of HSPG-bound ligands via LRP, which predict minimal or no effects of RAP on the initial binding of ligands to cell surface HSPGs but nearly complete inhibition of subsequent internalization (9).
Using several methods, we found a limited cooperation between perlecan HSPG and the LDL receptor in the catabolism, but not the initial cell surface binding, of LpL-enriched, 125 Ilabeled native LDL. Our studies using sterol suppression indicated that enhanced LpL-dependent catabolism of LpL⅐ 125 I-LDL compared with LpL⅐ 125 I-mLDL did, in fact, involve LDL receptors and was not primarily an effect of some sterol-insensitive structure, such as domain II of perlecan, which is homologous to the ligand-binding domain of the LDL receptor (31, 32). The synergy between perlecan and LDL receptors in our system is kinetically straightforward; roughly two-thirds of perlecan-bound LpL⅐ 125 I-LDL was unaffected by cell surface LDL receptors, while just under one-third showed accelerated internalization and degradation. The simplest explanation is propinquity; a perlecan HSPG that happens to be close to an LDL receptor, perhaps due to shared cytoplasmic NPXY sequences between ␤ 1 integrins and LDL receptors (70,71), is able to recruit the LDL receptor to assist in ligand catabolism. The majority of perlecan HSPG in our experimental system, however, does not properly encounter an LDL receptor and must rely on the pathway mediated directly by perlecan alone. An alternative explanation would be some sort of heterogeneity among the LpL-enriched 125 I-LDL particles, so that only a subpopulation was able to interact with LDL receptors after adhering to a molecule of perlecan. We do not favor this alternative, because we used a narrow density range for isolating LDL, a procedure that is known to generate preparations of uniform particles, and there is no evidence within such LDL preparations for a large fraction with impaired LDL receptor binding (42,44).
Ligand internalization by the synergistic pathway exhibited a t 1/2 of ϳ45 min, which is intermediate between the t 1/2 values for the classic LDL receptor pathway (ϳ9 min in these cells) and direct perlecan-mediated internalization (6 h). Nevertheless, our studies with metabolic inhibitors gave no indication that the biochemical machinery characteristic of direct perlecan-mediated internalization played any substantial role in the uptake of ligand through the synergistic pathway (compare Figs. 4 and 6). Presumably, then, it was the LDL receptor that drove internalization via the synergistic pathway. The fact that the t 1/2 for internalization was 45 min, not 9 min, indicates delays. For example, perlecan-bound ligands may take time to encounter a nearby LDL receptor, or an LDL receptor called upon to internalize perlecan-bound ligands might be slowed by the extra bulk, if ternary complexes are brought into the cells, or by the time it takes to tear a ligand off a nearby HS chain, if ligand transfer occurs. Interestingly, the synergistic pathway appeared to be more sensitive than classic LDL receptor-mediated internalization to cytochalasin D (compare the inhibitory effects in Figs. 4 and 6), which we speculate may reflect cytochalasin-induced alterations in cell surface distributions of perlecan and LDL receptors to reduce propinquity or, more likely, the formation of large complexes of ligands and receptors that become internalized by a process that resembles LDL receptor-dependent phagocytosis (64,72).
Several lines of evidence point toward physiologic and pathophysiologic roles for perlecan-mediated ligand catabolism. First, perlecan is present in sites where HSPGs are known to be involved in ligand catabolism, such as the liver and the arterial wall (32,35,37,38). Second, components of the direct perlecan-mediated pathway are known to be regulated. Synthesis of the perlecan core protein is regulated by inflammatory cytokines (73)(74)(75)(76). Cytokines also affect cell surface expression of integrins (77), which could change the amount of perlecan that adheres to the cell surface, as opposed to being released for incorporation into the basement membrane or other extracellular matrix. Side chain assembly is sensitive to growth conditions, which affect heparan sulfate chain length and the degree and pattern of heparan sulfate sulfation and epimerization (76,78), any of which may change ligand affinity. Third, cooperation with auxiliary cell surface receptors, such as LDL receptor family members, is likely to be affected by a variety of influences. In our experimental system, perlecan exhibited no cooperation with LRP and only a limited cooperation with LDL receptors, but other systems could show different results, based on different geometry and levels of expression of these molecules. For example, the liver can express very high levels of the LDL receptor (79), which might favor greater cooperation with perlecan. On the other hand, much of hepatic perlecan may not be directly adherent to any cellular surface (32,35,37), an arrangement that could reduce direct perlecan-mediated internalization as well as the propinquity of perlecan-bound ligands to LDL receptors. Fourth, HSPG-dependent hepatic catabolism of remnant lipoproteins has been described in vivo (24,25,28) and exhibits rapid binding and then extremely slow internalization kinetics in the absence of LDL receptors (24), consistent with the perlecan pathway or perhaps some combination of perlecan and syndecan (29).
Finally, the existence of several highly conserved molecules, such as LDL receptor family members, syndecan HSPGs, and perlecan HSPGs, that each mediate distinct internalization pathways suggests distinct but essential biologic functions. The roles of the perlecan pathway are presumably related to its ligand specificity as well as its unusual kinetics and intracellular itinerary. Lipoproteins (80,81) and other HS-bound ligands (13,14) can be altered during prolonged cell surface contact, particularly when hydrolytic enzymes are adherent nearby. Also, lengthy residence on the cell surface and the nature of the subsequent endocytic path can affect regulatory events within the cell, even when the different routes process similar amounts of ligand (82)(83)(84)(85).
Overall, our results demonstrate that the perlecan HSPG mediates a remarkably slow pathway for internalization and lysosomal delivery of its ligands that is kinetically and biochemically distinct from other uptake mechanisms and is consistent with the second, slower pathway for HSPG-dependent ligand catabolism. The known tissue distribution of perlecan, with high levels in the hepatic space of Disse and in the arterial wall, makes it an attractive candidate for interaction with plasma lipoproteins. Moreover, this distinct pathway may represent an important component in vivo of the catabolism of many other ligands that are known to bind HSPGs.