Endogenously Produced Endothelial Lipase Enhances Binding and Cellular Processing of Plasma Lipoproteins via Heparan Sulfate Proteoglycan-mediated Pathway*

Endothelial lipase (EL) is a new member of the triglyceride lipase gene family, which includes lipoprotein lipase (LpL) and hepatic lipase (HL). Enzymatic activity of EL has been studied before. Here we characterized the ability of EL to bridge lipoproteins to the cell surface. Expression of EL in wild-type Chinese hamster ovary (CHO)-K1 but not in heparan sulfate proteoglycan (HSPG)-deficient CHO-677 cells resulted in 3–4.4-fold increases of 125I-low density lipoprotein (LDL) and 125I-high density lipoprotein 3 binding (HDL3). Inhibition of proteoglycan sulfation by sodium chlorate or incubation of cells with labeled lipoproteins in the presence of heparin (100 μg/ml) abolished bridging effects of EL. An enzymatically inactive EL, EL-S149A, was equally effective in facilitating lipoprotein bridging as native EL. Processing of LDL and HDL differed notably after initial binding via EL to the cell surface. More than 90% of the surface-bound 125I-LDL was destined for internalization and degradation, whereas about 70% of the surface-bound 125I-HDL3 was released back into the medium. These differences were significantly attenuated after HDL clustering was promoted using antibody against apolipoprotein A-I. At equal protein concentration of added lipoproteins the ratio of HDL3 to VLDL bridging via EL was 0.092 compared with 0.174 via HL and 0.002 via LpL. In summary, EL mediates binding and uptake of plasma lipoproteins via a process that is independent of its enzymatic activity, requires cellular heparan sulfate proteoglycans, and is regulated by ligand clustering.

Lipoprotein lipase (LpL) 1 and hepatic lipase (HL), two members of the triglyceride lipase gene family, have well estab-lished roles in regulating lipid and lipoprotein metabolism and are implicated in atherosclerosis (1)(2)(3). LpL is synthesized primarily in adipose and skeletal muscle and is transported to the endothelial surface, where it is bound to heparan sulfate proteoglycans (HSPGs). LpL is predominantly a triacylglycerol hydrolase, and its enzymatic action is mainly related to hydrolysis of triglycerides (TG) in TG-rich apolipoprotein B-containing lipoproteins, chylomicrons and very low density lipoproteins (VLDL) (reviewed in Ref. 4). HL is synthesized primarily in hepatocytes (5,6) and is bound mostly to hepatic and endothelial HSPGs in the hepatic sinusoids (7). Like LpL, HL has substantial TG lipase activity, but unlike LpL, HL also has significant phospholipase activity (8). This increased phospholipase activity may play an important role in the ability of HL, as opposed to LpL, to directly modulate HDL metabolism (2).
In addition to their lipolytic activities, LpL and HL have been shown to mediate "bridging" between lipoproteins and HSPGs on the cell surface, which results in increased cellular uptake and degradation of lipoproteins (9 -16). In several studies LpL dramatically enhanced binding, internalization, and degradation of VLDL and LDL by cultured cells (10 -14, 17, 18). In contrast, LpL had relatively small, if any, effects on HDL binding and holoparticle uptake, but significantly increased selective uptake of cholesterol esters from HDL particles (19 -22). Like LpL, HL has been shown to enhance binding and/or uptake of chylomicrons, chylomicron remnants, VLDL, and LDL by different cell types in vitro (15,16,23,24) via a HSPG-dependent process (15,16,23). In addition, in several studies HL efficiently increased holoparticle cellular uptake of HDL as well as selective uptake of cholesterol esters (16,25,26). These bridging effects of LpL and HL require HSPGs and heparin-binding domains of the lipases but do not depend on their catalytical activities. There is also evidence for nonenzymatic effects of LpL and HL on lipoprotein metabolism in vivo (27)(28)(29)(30)(31)(32).
Endothelial lipase (EL), a 480-amino acid protein (M r ϳ68,000), is a new member of the triglyceride lipase gene family (33)(34)(35)(36). Many typical features of this gene family are conserved in EL: the catalytic triad residues, the lid that controls access of substrate to the hydrolytic pocket, and the cysteine residues that form intramolecular disulfide bonds. EL is unique in the triglyceride lipase family in that it is synthesized by endothelial cells, however, a number of other cell types also express EL (33,34). EL has detectable triglyceride lipase ac-tivity, but this activity is significantly less relative to its phospholipase activity compared with HL and especially LpL (33,37). In vitro, conditioned medium containing EL had the ability to substantially hydrolyze HDL phospholipids but had little activity toward LDL phospholipids, suggesting relative selectivity for HDL (37). In vivo, even low levels of EL overexpression in the livers of wild-type and human apoA-I transgenic mice dramatically reduced HDL cholesterol and apoA-I levels and increased HDL catabolism (33). As a result of EL overexpression, plasma levels of apoB-containing lipoproteins were also reduced, although to a lesser extent (33). Recent studies in EL knockout mice (38) and in mice injected with specific antibody against EL (39) provided additional evidence for a physiological importance of EL for lipid and lipoprotein metabolism in vivo.
The clusters of positively charged residues in LpL and HL have been implicated in binding of these molecules to heparin and HSPGs. Because the putative heparin-binding and lipoprotein-binding sites present in LpL and HL are highly conserved in EL, we hypothesized that EL may also serve as a bridging molecule between lipoproteins and cell surface and matrix HSPGs. Therefore, the primary focus of this study was to examine the ability of EL to mediate binding and holoparticle uptake of plasma lipoproteins and to compare EL-dependent metabolism of apoB-versus apoA-I-containing lipoproteins in vitro. We also compared EL with HL and LpL in their abilities to facilitate bridging of different major classes of plasma lipoproteins with cells. In this study, we demonstrated that EL can function as an efficient bridging molecule between plasma lipoproteins and cells in a process that requires intact cell surface HSPGs, but does not depend on EL enzymatic activity. Moreover, we found that compared with LpL and HL, EL has distinct preferences in bridging individual classes of lipoproteins.
Cultured Cells-COS-7 and two types of Chinese hamster ovary cell lines, wild-type CHO-K1 and the CHO mutant line pgsD-677 (CHO-677), which is specifically deficient in both N-acetylglucosaminyltransferase and glucuronosyltransferase activities and hence lacks heparan sulfate (42,43), were obtained from the American Type Culture Collection (Manassas, VA). COS-7 cells were grown and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Both CHO cell lines were maintained in Ham's F-12 medium supplemented with 10% fetal bovine serum. These three cell lines were chosen because they are readily infected with recombinant adenoviral constructs and were successfully used in the past to characterize effects of LpL and HL on lipoprotein metabolism (16,37,44,45).
Expression of Enzymes-Wild-type EL or enzymatically inactive EL, in which the active site serine was substituted with alanine (EL-S149A), LpL, and HL were expressed using recombinant adenoviruses. Recombinant adenoviruses encoding human EL (AdEL), human EL-S149A (adEL-S149A), human HL (AdHL), or GFP (AdGFP), used as a negative control in all experiments, were constructed as previously described (37). A recombinant adenovirus encoding human LpL (AdLpL) was a generous gift from Dr. Nicolas Duverger (Aventis Pharmaceuticals).
Adenoviral infections of cells were performed as described (37). Briefly, prior to infection cells were grown to 80 -90% confluence in 12-well plates (22 mm/well). Cells were then incubated with recombinant adenoviruses encoding EL, EL-S149A, LpL, HL, or GFP (control virus) in 0.3 ml of serum-free medium at a multiplicity of infection of 3000 particles/cell. Two hours later, 0.7 ml of fresh growth medium containing serum was added to each well and incubations were continued for 2 days. At 48 h post-infection, cells were washed twice with phosphate-buffered saline and used for ligand binding experiments with 125 I-labeled lipoproteins.
Expression of lipases by infected cells was confirmed by Western blotting and by assay of triglyceride lipase activity of conditioned medium according to the protocols described previously (37). Antibody against human EL was generated to a peptide in the N-terminal region of EL, as previously described (33). Antibody to rat HL that cross-reacts with human HL was also previously described (46). A polyclonal antibody to human LpL was a generous gift from Dr. Mark H. Doolittle.
Cellular Metabolism of 125 I-VLDL, 125 I-LDL, and 125 I-HDL 3 -For binding studies, cells grown in 12-well plates were incubated with 0.5 ml of the serum-free medium supplemented with 0.2% bovine serum albumin (Sigma number A-8806) and 125 I-labeled lipoproteins (5 g of protein/ml) unless otherwise stated. To assess the effects of endogenous expression of lipases on surface binding of lipoproteins, cells were incubated with labeled lipoproteins for 1 h at 4°C. To measure cellular uptake and degradation of lipoproteins, cells were incubated for 3 h at 37°C in the presence of labeled lipoproteins, then surface-bound (heparin-releasable), intracellular (heparin-resistant), and degraded (assessed by trichloroacetic acid-soluble, CHCl 3 -insoluble radioactivity in medium) ligand was measured as described (40,41).
To examine the time course of cellular processing of 125 I-LDL versus 125 I-HDL 3 , labeled lipoproteins were incubated with cells in serum-free medium for 1 h at 4°C to allow surface binding without further catabolism. Cells were washed at 4°C to remove unbound material. Fresh medium at 37°C with no ligands was then added, and incubations were continued at 37°C for the indicated times. Assays for surface-bound, intracellular, and degraded ligands were then performed as above. Additionally, trichloroacetic acid-precipitable radioactivity in the medium was quantified as an indication of lipoproteins released into the medium via retroendocytosis or desorption from the cell surface during the incubation at 37°C.
To test potential effects of ligand clustering on cellular metabolism of 125 I-HDL 3 , cells were incubated with 125 I-HDL 3 at 4°C for 1 h, to allow cell-surface binding. Cells were rinsed at 4°C to remove unbound ligand and incubated for an additional 30 min at 4°C with or without antibody against human apoA-I (10 g/ml final concentration). Cells were then washed briefly and incubated with pre-warmed media at 37°C for the indicated times followed by measurements of surfacebound, intracellular, degraded and medium released radioactivity.
All results for 125 I-lipoprotein metabolism were normalized to cellular protein. Lipase-dependent catabolism was calculated by subtracting the values obtained in control cells infected with adGFP from those obtained in cells expressing lipases (total catabolism). HSPG-mediated catabolism was calculated by subtracting the values obtained in the presence of heparin (100 g/ml) from those obtained in its absence.
Assay of EL Protein-In parallel with ligand binding experiments, a separate set of cells was used to measure EL or EL-S149A protein available for bridging of labeled lipoproteins. For this purpose, cells were not exposed to radiolabeled lipoproteins, but instead were incubated in the presence of heparin (100 g/ml) to release surface-bound lipases into the medium. EL protein concentration was evaluated by densitometric analysis of Western blots of the conditioned medium using a Bio-Rad imager (model GS-700) and Quantity One program (Bio-Rad). Protein concentration was defined in arbitrary units as determined by densitometric analysis. Several dilutions of conditioned medium were used to make sure the intensity of the bands was proportional to the amount of protein applied on the gel.
Role of HSPGs in Lipoprotein Metabolism-We used three approaches to test the contribution of HSPGs in EL-mediated processing of lipoproteins. First, we blocked sulfation of cellular proteoglycans by preincubating cells for 18 h at 37°C in sodium chlorate (50 mM), an inhibitor of sulfate adenyltransferase (47), thus preventing sulfation of glycosaminoglycan side chains of HSPGs. Control cells were exposed to 50 mM sodium chloride. Second, we compared lipoprotein metabolism in wild-type versus HSPG-deficient CHO cells. Third, we incubated cells with labeled lipoproteins in the presence or absence of heparin at 100 g/ml, a concentration that specifically blocks interactions with HSPGs (40), but not lipoprotein binding to the members of the LDL receptor family or other lipoprotein receptors (48).
Statistical Analyses-All results are displayed as mean Ϯ S.E., n ϭ 3. Error bars that appear absent indicate S.E. values smaller than the drawn symbols. Role of Cell Surface HSPGs in EL-mediated Increase of 125 I-LDL and 125 I-HDL 3 Metabolism-We next tested if cell surface HSPGs are responsible for the observed effects of EL on binding of lipoproteins by incubating cells with labeled lipoproteins in the presence of 100 g of heparin/ml. At this concentration heparin selectively blocks ligand binding to heparan sulfate side chains of HSPGs (40) but not lipoprotein binding to the members of the LDL receptor family or other lipoprotein receptors (48). The EL-dependent component of 125 I-LDL or 125 I-HDL 3 binding in control CHO-K1 cells infected with adEL was completely abolished by heparin (Fig. 1, panels A and B). Cellsurface binding of both 125 I-LDL and 125 I-HDL 3 via EL was also inhibited by more than 90% when cells were pretreated for 18 h in the medium containing 50 mM sodium chlorate, which blocks sulfation of glycosaminoglycan side chains of HSPGs (data not shown). Additional support for the role of HSPGs in EL-mediated binding of lipoproteins was obtained when we tested effects of adEL infection on 125 I-LDL and 125 I-HDL 3 binding in HSPG-deficient CHO-677 cells. The level of EL secretion by HSPG-deficient cells was similar or slightly higher than the level of EL secretion in control cells, but there was no EL-dependent increase in 125 I-LDL and 125 I-HDL 3 binding in HSPG-deficient cells (Fig. 1, panels A and B). As expected, heparin (100 g/ml) also had no effect on lipoprotein binding by HSPG-deficient cells (Fig. 1, panels A and B). Taken together, these results indicate that HSPGs are crucial for EL-dependent increases in binding of 125 I-LDL and 125 I-HDL 3 to the cell surface.

Effects of EL Expression on Binding of 125 I-LDL and 125 I-
To verify if enzymatic activity is required for the bridging effects of EL, we compared binding of 125 I-LDL and 125 I-HDL 3 by cells infected with adEL versus those infected with recombinant adenovirus encoding enzymatically inactive EL (adEL-S149A). In four independent experiments, adEL-S149A-infected cells demonstrated significant increases in binding of 125 I-LDL and 125 I-HDL 3 compared with cells infected with control virus (Fig. 2). When normalized to protein expression, increases in binding of 125 I-LDL and 125 I-HDL 3 caused by expression of EL-S149A were 99.2 Ϯ 2.0 and 102.8 Ϯ 3.90%, respectively, of those resulting from the expression of wild-type EL. This indicates that enzymatic activity of EL is not required for bridging of either 125 I-LDL or 125 I-HDL 3 .
Intracellular Processing of 125 I-LDL and 125 I-HDL 3 by Cells Expressing EL-To compare effects of endogenous EL on cellular metabolism of 125 I-LDL and 125 I-HDL 3 cells were incubated with labeled lipoproteins for 3 h at 37°C, a temperature that allows internalization and degradation of lipoproteins after their binding to cell receptors. Under these conditions, expression of EL by COS cells resulted in the increased binding, intracellular accumulation, and degradation of both 125 I-LDL and 125 I-HDL 3 compared with control cells infected with adGFP (Fig. 3, A and B, respectively). EL-dependent components of binding, cellular accumulation, and degradation of lipoproteins were inhibited by more than 95% when cells were incubated with labeled lipoproteins in the presence of heparin (100 g/ml). Interestingly, the relative portion of 125 I-LDL that was internalized via the EL-mediated process in 3 h was much higher compared with that of 125 I-HDL 3 , suggesting that the nature of the ligand may affect metabolism of the ligand after initial binding to the cell surface. Additionally, in the experiments performed at 37°C the values for the EL-mediated binding of 125 I-LDL, expressed in nanograms of metabolized lipoprotein per mg of cell protein, were much higher compared with those for 125 I-HDL 3 suggesting significantly more efficient intracellular processing of LDL versus HDL via an EL-dependent pathway.
We next compared kinetics of cellular processing of 125 I-LDL versus 125 I-HDL 3 . About 75% of 125 I-LDL initially bound to the cell surface through the EL-mediated pathway was internalized, and ϳ50% of internalized material was degraded by the end of the 2-h incubation (Fig. 4A). Less than 10% of 125 I-LDLrelated radioactivity was recovered from the medium as trichloroacetic acid-precipitable material. In contrast, the majority of 125 I-HDL 3 was released back into the medium in trichloroacetic acid-precipitable particles and only 20 -25% of 125 I-HDL 3 was eventually internalized (Fig. 4B). By the end of the 2-h incubation ϳ80% of internalized 125 I-HDL 3 was degraded. Interestingly, compared with 125 I-LDL, degradation of 125 I-HDL 3 started earlier and proceeded at a higher rate when expressed as a percentage of the internalized material, but it could be related to a considerably lower rate of internalization of 125 I-HDL 3 .
We sought to explain the observed differences in the cellular processing of LDL versus HDL by comparing affinities of binding of these lipoproteins to EL. For this purpose, COS cells expressing EL were incubated for 1 h with increasing concentrations of 125 I-LDL or 125 I-HDL 3 . Incubations were performed at 4°C to minimize the enzymatic activity of EL. At all concentrations tested (0.37-90 g of lipoprotein protein/ml), binding of 125 I-LDL or 125 I-HDL 3 to EL-expressing cells was signifi- cantly higher than binding to control cells. For analysis, ELdependent binding of lipoproteins was estimated by subtracting values obtained in control adGFP-infected cells from those obtained in EL-expressing cells. The molar concentrations of LDL and HDL were calculated assuming an average molecular mass of apolipoproteins of 550 and 85.5 kDa, respectively (see Refs. 49 -51). Scatchard analysis of EL-dependent binding is shown in Fig. 5. The curve for EL-dependent binding of 125 I-LDL was nonlinear (Fig. 5A), indicating the presence of more than one binding site. Nonlinear regression analysis of the concentration curve for 125 I-LDL revealed two types of binding sites, one with somewhat lower affinity (apparent K d of 161.9 nM; B max of 1.80 pmol/mg of cell protein) and another with much higher affinity but lower capacity (apparent K d of 1. ticipation of HSPGs. Endocytosis via HSPGs depends on their clustering by large, multimeric ligands (40,52,53). Thus, relatively larger LDL particles that also have higher affinity toward EL could be able to trigger clustering of HSPGs more efficiently than HDL. To establish whether clustering of HDL might affect lipoprotein internalization and degradation, we tested effects of goat polyclonal antibody against apoA-I (anti-apoA-I) on cellular processing of HDL 3 . First, we assessed the effect of anti-apoA-I on surface distribution of DiI-HDL 3 . COS cells expressing EL were incubated with DiI-HDL 3 for 1 h at 4°C followed by incubation at 4°C in the absence or presence of anti-apoA-I. Cells incubated in the absence of anti-apoA-I had a very low level of fluorescence dispersed over the cell surface with slightly higher signal at the edges of the cells (Fig.  6A). Treatment of cells with anti-apoA-I resulted in a dramatic increase of the intensity of fluorescence with most of the signal localized in clusters on the periphery of the cells (Fig. 6B). In a parallel experiment, DiI-HDL 3 was substituted with 125 I-HDL 3 to test if cross-linking of the surface-bound 125 I-HDL 3 with antibodies at 4°C would affect lipoprotein binding. COS cells were incubated with 125 I-HDL 3 for 1 h at 4°C, unbound lipoproteins were washed away and cells were incubated for an additional 30 min at 4°C without or with anti-apoA-I. Anti-apoA-I had no effect on the amount of radioactive lipoproteins that remained bound to the cell surface. 125 I-HDL 3 binding was 104.5 Ϯ 2.9 and 108.0 Ϯ 2.2 ng/mg of cell protein for cells treated without or with anti-apoA-I, respectively. Taken together, these data indicate that the striking difference in the intensity of fluorescence between control and anti-apoA-Itreated cells in Fig. 6 is because of antibody-induced clustering of DiI-HDL 3 , which makes the fluorescent signal easier to detect.
We next checked if clustering by anti-apoA-I would affect cellular metabolism of 125 I-HDL 3 at 37°C. As displayed in Fig.  7A, by the end of the 40-min incubation at 37°C, less than 15% of 125 I-HDL 3 initially bound to the cell surface via EL was internalized or degraded by the cells kept in the absence of anti-apoA-I (Control). The rest of the radioactivity was released into the medium or stayed on the cell surface in the heparinreleasable pool in agreement with the results in Fig. 4B. Treatment of cells with nonspecific IgG had no effect on 125 I-HDL 3 distribution (IgG). In contrast, incubation of cells for 30 min at 4°C in the presence of anti-apoA-I before incubation at 37°C resulted in internalization of ϳ40% of the surface-bound ligand accompanied by the corresponding decrease of HDL fraction released into the medium (anti-apoA-I). Stimulation of 125 I-HDL 3 uptake by anti-apoA-I was completely abolished in the presence of heparin (100 g/ml) implying a crucial role for HSPGs in the process. Interestingly, after clustering with anti-apoA-I, kinetics of 125 I-HDL 3 internalization and degradation are very similar to that of 125 I-LDL (compare Figs. 4A and 7B).
Comparison of Lipoprotein Bridging Preferences of EL Versus LpL and HL-EL, HL, and LpL discretely hydrolyze different lipoprotein classes (37). We next explored whether lipoprotein bridging preferences of the three members of the lipase family might also be different. Parallel sets of COS-7 cells were infected with equal amounts of control vector or adenoviral constructs encoding EL, HL, and LpL. Expression of the active enzymes was verified by Western blotting and by measuring triglyceride lipase activity. TG lipase activities in the medium from adGFP-, adEL-, adHL-, and adLpL-infected cells were 26.5 versus 161.1 versus 337.5 versus 215.9 nmol/ ml/h, respectively. Infected cells were incubated with 125 I-VLDL, 125 I-LDL, or 125 I-HDL 3 at equal protein concentrations (10 g/ml). Lipase-dependent cellular metabolism of lipoproteins was measured by subtracting values obtained in control, adGFP-infected, cells from those obtained in EL-, HL-, or LpLexpressing cells. As shown in Fig. 8A, expression of EL resulted in very similar increases in metabolism of 125 I-VLDL and 125 I-LDL, whereas its effect on 125 I-HDL 3 was relatively lower. Similar to EL, HL was able to enhance the metabolism of all three lipoprotein classes (Fig. 8B). Under our experimental conditions, LpL caused pronounced increases of 125 I-VLDL and At the end, cells were washed to remove unbound ligands and total radioactivity associated with the cells was measured. Lipoprotein binding to control cells infected with adGFP was used to estimate nonspecific binding. EL-dependent binding of lipoproteins was calculated by subtracting values obtained in adGFP-infected cells from those obtained in EL-expressing cells. The data are plotted as bound/free versus bound, where bound is picomole of lipoproteins bound per mg of cell protein and free is nanomole of added lipoproteins per ml of medium. Dotted and dashed lines represent curves for low and high affinity binding sites, which were calculated using nonlinear regression analysis.
FIG. 6. Effect of anti-apoA-I on distribution of HDL 3 on the cell surface. COS-7 cells grown in the glass-bottom 35-mm Petri dishes (MatTek corporation, Ashland, MA) and infected with adEL were incubated with DiI-labeled HDL 3 (10 g/ml final concentration) in serumfree medium for 1 h at 4°C. Then, cells were washed to remove unbound material and incubated for an additional 30 min at 4°C without (A) or with anti-apoA-I (B). At the end of the incubation, cells were washed briefly, fixed with 3% paraformaldehyde in phosphate-buffered saline overnight at room temperature, and blocked in phosphate-buffered saline/bovine serum albumin for 1 h at room temperature. Samples were analyzed using a Zeiss Axiovert 100TV microscope (Germany) with 40X Plan-Apochromat lenses, a precisely controlled XYZ stage (Applied Precision), and a scientific grade cooled CCD camera (Micro-Max, Princeton Instruments, Trenton, NJ). To maximize resolution along the optical axis, illumination from a mercury lamp was directed through a fiber optic scrambler to provide high intensity, homogenous illumination to the back aperture plane of the objective lens. 125 I-LDL metabolism but not of 125 I-HDL 3 (Fig. 8C). Of note, comparison of EL effects versus those of HL or LpL based on absolute values should be viewed very cautiously as the differences might reflect different amounts of lipases available for bridging. Instead, the ratios of metabolism of HDL 3 versus metabolism of VLDL calculated for each lipase (Fig. 8D) can be directly compared because these ratios are independent of the protein expression. Among the three lipases, HL had the highest HDL 3 to VLDL ratio of 0.174, whereas the same values for EL and LpL were 0.092 and 0.002, respectively (Fig. 8D). These results indicate that EL, HL, and LpL have very distinctive preferences in bridging different classes of plasma lipoproteins.

DISCUSSION
The results presented in this study demonstrate that EL, a recently discovered member of the triglyceride lipase family, is able to mediate efficient binding and cellular metabolism of both apoB-and apoA-I-containing lipoproteins. Stimulation of lipoprotein metabolism by EL is independent of EL enzymatic activity because a catalytically inactive EL mutant, EL-S149A, had bridging capacity similar to that of wild-type EL (Fig. 2). Thus, our results with EL and EL-S149A extend the existing body of evidence for structural nonenzymatic effects of members of the triglyceride lipase family on metabolism of different classes of plasma lipoproteins in vitro (10,12,25,26).
Absence of EL-mediated bridging to HSPG-deficient CHO cells and to CHO-K1 cells treated in sodium chlorate, as well as the ability of low concentrations of heparin to inhibit the action of EL, indicated that EL-mediated binding of lipoproteins depends on the cell surface HSPGs and particularly on the heparan sulfate (HS) side chains but not on the core protein of HSPGs. Interestingly, the major characteristics of HS side chains are determined mainly by the cell type or physiological state of the cell rather than by the core protein of HSPGs (54,55). Thus, it is likely that all three classes of HSPGs (syndecans, glypicans, and perlecan) could participate in EL-mediated binding of plasma lipoproteins to the cell surface. At the same time, the nature of the core protein could play a crucial role in the mechanism of HSPG-mediated uptake of lipoproteins because it is the structure of the core protein that determines interactions of HSPGs with different components of endocytic machinery of cells (40, 45, 56 -60). Indeed, metabolism of LpL-enriched 125 I-labeled LDL was kinetically and biochemically distinct depending on whether syndecan, glypican, or perlecan was the predominantly expressed HSPG (40,45,61).
For this study we used CHO-K1 and COS cells, which express primarily syndecans. Of note, endothelial cells express significant amounts of syndecan as well (62). In our experiments, the uptake and degradation of lipoproteins via an ELmediated pathway proceeded at a relatively slow rate. By the end of a 3-h incubation with labeled lipoproteins, about 65% of 125 I-LDL was recovered from the intracellular compartments and 16% of lipoproteins was degraded, whereas the remainder was on the cell surface (Fig. 3). These values and the results of kinetic studies presented in Fig. 4 are consistent with the rate of ligand internalization and degradation via syndecan-mediated pathway described previously (40,52,63), suggesting that HSPGs play a major role not only in surface binding, but also in the uptake and degradation of lipoproteins via an EL-mediated process.
The fate of lipoproteins bound to the cell surface HSPGs via EL may also depend on characteristics of the lipoproteins themselves. For example, apoB-containing VLDL and LDL were taken very efficiently into the cells after initial cell-surface binding via the EL-mediated process. In contrast, more than 70% of HDL 3 bound to the cell surface after incubation at 4°C was subsequently released into the medium during incubation at 37°C (Fig. 4). A number of factors may contribute to the observed differences. We found that compared with HDL 3 , LDL is able to bind much more avidly to EL. Interestingly, analysis of 125 I-LDL interaction with cell surface EL revealed the presence of two types of binding sites, which differ significantly by their apparent K d and B max . We speculate that the low affinity, high capacity type reflects interactions between EL and lipid components of LDL, whereas interaction between EL and apolipoprotein B is likely to be associated with the second class, characterized by high affinity and low capacity. Furthermore, our finding, that interaction of HDL with EL could be described by only one class of binding sites with low affinity and high capacity is in line with the inability of the structurally and functionally related HL and LpL to bind directly to protein components of HDL (49,64). Several studies demonstrated that HL and LpL had significantly higher affinities toward LDL than HDL (49, 64 -66), although there was disagreement on the relative importance of the protein-protein interaction between these lipases and apoB.
Different affinities of LDL versus HDL toward EL could be one of the factors responsible for the differences in uptake and degradation of these lipoproteins. The lower efficiency of HDL uptake may also be related to a relatively smaller HDL surface area and, therefore, fewer sites available for binding of EL on a FIG. 7. Effect of anti-apoA-I on EL-mediated catabolism of 125 I-HDL 3 by COS cells. A, fractional distribution of 125 I-HDL 3 after preincubation in the absence or presence of anti-apoA-I. COS-7 cells infected with adGFP or adEL viruses were incubated with 125 I-labeled lipoproteins (10 g/ml final concentration) in serum-free medium for 1 h at 4°C to allow surface binding without further catabolism. Cells were washed at 4°C to remove unbound material and incubated for an additional 30 min at 4°C in fresh medium in the absence of IgGs, in the presence goat polyclonal antibodies against human apoA-I (10 g/ml) or matching concentrations of irrelevant goat IgGs. Then, cells were briefly washed and incubations were continued for 40 min at 37°C in fresh medium with no ligands. Measurements of EL-dependent surface binding, intracellular accumulation, degradation and medium release of trichloroacetic acid-insoluble ligands were then performed as described above. The results are displayed as percentage of total amount of lipoproteins initially bound via the EL-mediated process to the cells during incubation at 4°C. B, kinetics of EL-dependent catabolism of 125 I-HDL 3 after cross-linking with anti-apoA-I. Cells were treated as described in the legend to Fig. 4, except a 30-min incubation at 4°C with anti-apoA-I antibody (10 g/ml) was included before incubation at 37°C in the medium with no ligands. For each time point the differences between catabolism of 125 I-HDL 3 by COS cells infected with adEL and adGFP were calculated. single HDL particle. This, in turn, may prevent efficient clustering of the cell surface HSPGs, which is required, for example, for the syndecan-mediated internalization (40,52). The fact that cross-linking with anti-apoA-I changed fractional distribution of 125 I-HDL 3 and dramatically increased its uptake and degradation via the EL-dependent pathway (Fig. 7), strongly supports the importance of the clustering event for efficient endocytosis.
The observed differences in processing of apoB-versus apoA-I-containing lipoproteins via the EL-dependent pathway may potentially have different impacts on the intracellular lipid metabolism. For example, it is unlikely that EL alone will mediate any significant selective uptake from apoB-containing lipoproteins, VLDL or LDL, because more than 90% of VLDL and LDL particles that were initially bound to the cell surface were then internalized and degraded (Figs. 3A and 4). On the other hand, it is plausible to assume that EL alone, even in the absence of scavenger receptor BI, will be able to mediate selective uptake of cholesterol esters through lipid exchange at the cell surface with subsequent release of HDL particles into the medium in the absence of clustering antibody. Indeed, in the study published while this manuscript was under review, Strauss et al. (67) showed that EL-mediated bridging of HDL particles to the cell surface resulted in increased selective uptake of cholesterol esters.
Side-by-side comparison of EL versus HL and LpL, presented in Fig. 8, implies that the three members of the TG lipase family have different abilities to promote interaction between individual classes of plasma lipoproteins and cells. In our experiments, EL and HL were able to bridge efficiently all three lipoprotein classes. At the same time, LpL had prominent effects only with apoB-containing lipoproteins, VLDL and LDL, but not with HDL 3 , in agreement with the majority, but not all previously published studies (19 -22). Some differences in the effects of LpL may be because of the use of endogenous versus exogenous LpL and/or reflect variations in the LpL concentrations used. The HDL/VLDL bridging index, which is independent of lipase expression, was the highest for HL and the lowest for LpL (Fig. 8D). Of note, the relative capacity of EL to bridge different classes of plasma lipoproteins apparently does not directly correlate with its enzymatic activity toward lipid components of the same classes of lipoproteins. McCoy et al. (37) have shown recently that EL was able to hydrolyze lipids from HDL 3 more efficiently than from LDL. In contrast, here we found that EL bridges more efficiently LDL than HDL 3 (Figs. 3,  4, and 8), when lipoproteins were normalized based on their protein concentrations or based on the percentage of lipoprotein particles taken by the cells from the medium. These data further support the notion that the effects of EL on hydrolysis of lipoprotein lipids and on holoparticle uptake of lipoprotein particles are not necessarily related. These observations together with the recent in vivo studies (33,38,39) and data from Strauss et al. (67), suggest that EL, through its enzymatic and structural action, may have a profound effect on HDL-mediated reverse cholesterol transport, whereas the primarily bridging effect of EL on apoB-containing lipoproteins may be more relevant for the local metabolism of these particles in the vessel wall. Additional experiments using EL knockout animals could provide more direct evidence for the relative physiological importance of enzymatic versus structural activity of EL.
Overall, our data indicate that EL is able to bridge plasma lipoproteins with the cell surface in a process that requires intact HSPGs but does not rely on enzymatic action of the lipase. Moreover, we demonstrated that compared with HL and LpL, EL has distinctively different bridging preferences toward plasma lipoproteins offering further evidence of a unique role of EL in lipoprotein metabolism.