Mechanism of the Association of HDL3 with Endothelial Cells, Smooth Muscle Cells, and Fibroblasts

Human plasma high-density lipoprotein-3 (HDL3) has been shown to bind to a variety of cells and tissues. In order to investigate the nature of HDL3-cell association, we studied the interaction of 125I-HDL3 with porcine aortic endothelial cells, rabbit aortic smooth muscle cells, and normal human skin fibroblasts. At 37 degrees C, 125I-HDL3 association with endothelial cells was nonsaturable. Furthermore, 60% protein digestion of HDL3 by trypsin (T-HDL3) actually increased its ability, on a protein weight basis, to associate with endothelial cells and to displace 125I-HDL3 from all three cell types. Synthetic phospholipid-cholesterol discs containing either apo-A-I or apo-A-II were equally effective in displacing 125I-HDL3 from endothelial cells, and phospholipid-cholesterol vesicles containing no protein also displaced 125I-HDL3 from endothelial cells. Neither lipid-free apo-HDL3 nor apo-T-HDL3 was able to competitively displace 125I-HDL3. The above competitive displacement data, when expressed on a protein weight basis, showed differences in the ability of the competitors to displace 125I-HDL3 from cells in the following order of effectiveness: discs greater than T-HDL3 greater than native HDL3. When these data were expressed on a surface lipid weight basis, all three competitors, as well as the lipid vesicles, were approximately normalized to a single competitive displacement curve. Studies on the nature of the cellular mediators of HDL3-cell association revealed that the cell surface sites were resistant to proteolytic treatment. Furthermore, both 125I-HDL3 and 125I-T-HDL3 association with fibroblasts preincubated with varying concentrations of cholesterol increased in parallel with the free cholesterol content of the cells; although cycloheximide blocked this increase in HDL3-cell association, cycloheximide also prevented the increase in cholesterol content of cholesterol-treated cells. We conclude that the association of HDL3 with the cell types studied is not mediated by specific ligand and receptor proteins but rather involves the interaction of cellular surface lipids, possibly cholesterol, with the surface lipids of HDL3.


Mechanism of the Association of HDL3 with Endothelial Cells, Smooth
Muscle Cells, and Fibroblasts EVIDENCE AGAINST THE ROLE OF SPECIFIC LIGAND AND RECEPTOR PROTEINS* (Received for publication, April 17, 1984) Human plasma high-density lipoprotein-3 (HDL3) has been shown to bind to a variety of cells and tissues. In order to investigate the nature of HDL3-cell association, we studied the interaction of "1-HDL3 with porcine aortic endothelial cells, rabbit aortic smooth muscle cells, and normal human skin fibroblasts. At 37 "C, '2BI-HDL3 association with endothelial cells was nonsaturable. Furthermore, 60% protein digestion of HDL3 by trypsin (T-HDL3) actually increased its ability, on a protein weight basis, to associate with endothelial cells and to displace lZ61-HDL3 from all three cell types. Synthetic phospholipid-cholesterol discs containing either apo-A-I or apo-A-I1 were equally effective in displacing '251-HDL, from endothelial cells, and phospholipid-cholesterol vesicles containing no protein also displaced 1261-HDL3 from endothelial cells. Neither lipid-free apo-HDL3 nor apo-T-HDL, was able to competitively displace 12'I-HDL3. The above competitive displacement data, when expressed on a protein weight basis, showed differences in the ability of the competitors to displace '251-HDL3 from cells in the following order of effectiveness: discs > T-HDLs native HDL,. When these data were expressed on a surface lipid weight basis, all three competitors, as well as the lipid vesicles, were approximately normalized to a single competitive displacement curve. Studies on the nature of the cellular mediators of HDL3-cell association revealed that the cell surface sites were resistant to proteolytic treatment. Furthermore, both '261-HDL3 and l2"Z-T-HDL3 association with fibroblasts preincubated with varying concentrations of cholesterol increased in parallel with the free cholesterol content of the cells; although cycloheximide blocked this increase in HDL,-cell association, cycloheximide also prevented the increase in cholesterol content of cholesterol-treated cells. We conclude that the association of HDL, with the cell types studied is not mediated by specific ligand and receptor proteins but rather involves the interaction of cellular surface lipids, possibly cholesterol, with the surface lipids of HDL3.
There has been considerably recent interest in the S t N C t U I e * This work was supported by National Institutes of Health Grants HL 22682, 21006, 30898, and T-07343. The costs of publication of this article were defrayed in part by the payment of page charges.
This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Established Investigator of the American Heart Association. ___ and metabolism of HDL,' due, at least in part, to the epidemiologic evidence that plasma HDL levels are inversely correlated with coronary artery disease (1). The potential functional roles of HDL include removal of peripheral tissue cholesterol ("reverse cholesterol transport") (Z), stimulation of prostaglandin synthesis (31, and support of endothelial cell growth (4). In addition, there is evidence that HDL may supply cholesterol to rat steroidogenic tissue for hormone synthesis.
Except for the apo-E-containing HDL, and HDL, subfractions, which specifically bind to apo-E receptors (5), it is not known if HDL specifically binds to the tissues in question to subserve its roles. Several investigators have attempted to examine this relationship between binding and function, and there has recently appeared many published reports on HDL binding phenomena. The tissues and cells studied include human fibroblasts (6)(7)(8), endothelial cells (9), steroidogenic tissue and cells (10-lZ), hepatic tissue and cells (13, 14), smooth muscle cells (81, intestinal cells (151, and kidney membranes (16).
The observation of HDL binding to these tissues has led many workers to conclude that there exist specific HDL receptors (10)(11)(12)(13)(14)(15)(16)(17)(18). However, neither a cellular receptor protein nor a specific HDL ligand has been definitively identified in any system. Therefore, we chose to characterize the interaction of HDL, with three different tissue culture cells, porcine aortic endothelial cells, rabbit smooth muscle cells, and human skin fibroblasts, with an emphasis on identifying those determinants on HDLa which might be important in this interaction. Our data strongly suggest that no specific protein ligand is involved in this interaction and also place in doubt the existence of a specific receptor protein in the cell lines studied. Rather, our results suggest that the surface lipids of both HDL, and the cell plasma membrane play an important role in the interaction.  . Both studies showed that a large percentage of the cell association was "specific" (defined as displaceability by excess cold HDLR). Fig. 1 left shows that the cell association had an initial rapid phase and then increased more slowly. The concentration curve ( Fig. 1  right) showed a steeper slope at lower HDLR concentrations (<50 pg/ml) than at higher concentrations. Neither the time nor concentration dependence data showed true saturability. Camejo (35) has demonstrated that treatment of HDL with trypsin results in up to 70% hydrolysis of the HDL proteins and that the residual lipid-bound fragments contain no intact apoprotein. To determine whether the capacity of HDLR to associate with cells could be destroyed by such extensive apoprotein hydrolysis, we compared the ability of T-HDLR with that of native HDLa3 to associate with tissue culture cells. T-HDLR was prepared as described under "Experimental Procedures." Under the conditions employed (30-min digestion a t 37 "C), 60% of the Lowry-reacting material was removed, and all of the lipid components, unaltered in composition from native HDL3, remained with the 40% core peptides. Sepharose CL-GB chromatography of T-HDLa (Fig. 2 ) demonstrated no major size alteration of the hydrolyzed particle. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Fig. 2, inset) of the delipidated peptides from T-HDLR showed almost complete absence of intact apo-A-I or apo-A-I1 and the accumulation of low-molecular weight fragments.

Characterization
The ability of T-HDL3 to associate with endothelial cells is demonstrated in Fig. 3. Both total and specific '*'I-T-HDLacell association was greater than total and specific '2sII-HDLacell association, respectively, when plotted as a function of protein concentration. T o determine whether T-HDLa and HDL,, shared common cell-association sites, T-HDLa was compared with HDL3 in its ability to displace '"I-HDLa from tissue culture cells. Fig. 4 shows that in three different cell lines, porcine endothelial cells (Fig. 4, top panel), rabbit smooth muscle cells (Fig. 4, middle panel), and human fibroblasts (Fig. 4, bottom panel), T-HDL3 displaced '*'I-HDLR somewhat more effectively than native HDLa on a protein weight basis. Similar results were obtained whether relatively HDL, treated exactly as T-HDLs except for the addition of trypsin behaved identically to completely untreated, native HDLs both in terms of physical structure and cell associating capacity. high '*'II-HDLa concentrations (50 pg/ml) or low concentrations (2-5 pg/ml) were used; thus, if there are separate highand low-affinity sites for HDLR, as suggested in earlier studies (8), both are displaced equally well by T-HDL3. To investigate the possibility that T-HDLa interacts with cells in a nonspecific manner to broadly inhibit ligand-cellular interactions, the ability of T-HDL3 to competitively inhibit the association of "'I-LDL with cultured human fibroblasts was determined. The experiments showed no effect of T-HDLR on the cell association of '"I-LDL.
Another interpretation of the above competitive displacement data is that T-HDL3 exchanges apoproteins with '"I-HDL.?, thus decreasing the specific activity of the latter. This would result in decreased cell association of '251-HDL3 radioactivity even if the same number of '251-HDL3 particles were in fact being bound. To test this possibility, T-HDL3 and 1251-HDL, were preincubated for varying periods at 31 "C in the absence of cells and then exposed to cells for a relatively short time period (15 min). If there was significant apoprotein exchange, one would expect that with longer periods of preincubation, the specific activity and the apparent cell association of '251-HDL3 would decrease. However, as shown in Fig.  5, this was clearly not the case; the per cent inhibition of lz5I-HDL3 cell association remained constant and was independent of the time of preincubation. Although very rapid, apoprotein exchange (completed in less than 15 min) could give similar results, previous measurements of the kinetics of apoprotein exchange show that this is unlikely (36). (In the further competitive displacement experiments to be described below, employing HDL3 analogues other than T-HDL,, similar preincubation experiments were performed and gave the same result as that in Fig. 5.) Therefore, the data in Figs. 3 and 4 show that extensive trypsin treatment of HDL, actually enhances, on a protein weight basis, its ability to interact with HDL,-cell association sites on tissue culture cells. Procedures") was preincubated in the absence of cells at 37 "C for the indicated time periods and then exposed to endothelial cells for 15 min at 37 "C. lZ5I-HDL3 cell association was determined as described under "Experimental Procedures." The data, expressed as per cent inhibition of '251-HDL3-cell association, were calculated for each preincubation time point as follows: (control valueexperimental value)/(control value) X 100. The control value, which did not vary by more than lo%, averaged 76.5 ng of cell-associated '251-HDL3/mg of cell protein. Competitive Displacement of I2$I-HDL3 from Cells by Synthetic Discs and Vesicles-The above data suggest a lack of ligand specificity in the interaction of HDL3 with tissue culture cells. Further experiments were performed to determine if cell association of HDL3 was mediated by specific apoproteins. Synthetic phospholipid-cholesterol discs containing either apo-A-I or apo-A-I1 were prepared as described under "Experimental Procedures" and their ability to displace lZ51-HDL3 from cells was tested (Fig. 6). In two different tissue culture lines, apo-A-I and apo-A-I1 containing discs caused similar displacement of '251-HDL3. Thus there is a lack of specificity in the ability of HDL3 analogues to displace 1251-HDL3 from cells. Also note that the discs were actually more effective than native HDL3 in displacing lZ5I-HDL3. In further experiments we found that 20-fold excess (by protein weight) of lipid-free apo-HDL3 and apo-T-HDL, caused only 14.3% and 12.4% displacement of lz5I-HDL3 from endothelial cells (data not illustrated) in contrast to 75% displacement caused by 10-fold excess of apo-A-I and apo-A-I1 containing discs (Fig. 6). We therefore considered the possibility that it was not the apoproteins but rather the surface lipids of HDL, that were mediating the cellular interaction.
To examine the significance of lipid-cellular interaction, small unilamellar phospholipid-cholesterol vesicles containing no protein were tested for their ability to displace lZ5I-HDL, from porcine endothelial cells. The vesicles were almost as effective as native HDL, in their ability to displace lz5I-HDL, from the cells (Fig. 7, top panel). By contrast, the vesicles were completely unable to displace lz5I-LDL from human fibroblasts (not shown). To determine if the ability of T-HDL, and discs to displace lZ5I-HDL3 could be explained on the basis of lipid-lipid interactions, we replotted the data of Figs. 4 and 6 as fold excess of surface lipid weight of unlabeled HDL, or analogue rather than fold excess of protein weight (Fig. 7). Expressed in this way, all of the different particles showed approximately equal ability to displace na- tive HDL. The considerable differences in displacement ability when expressed on a per protein basis were clearly reduced. The normalization of the data to a single displacement curve was suggested particularly strongly by the smooth muscle cell data (Fig. 7, bottom panel). Similar normalization was obtained when the direct cell association data of '251-HDL3 and lZ5I-T-HDL3 (Fig. 3) was expressed on a per surface lipid weight basis (not illustrated). These results indicate that the lipids of HDL3 or its analogues probably play a major role in determining cell association of HDL,. Cellular Determinants of HDL,-Cell Association-Having concluded that HDL3 surface lipids and not a specific protein ligand mediate the interaction of HDL, with cells, we next sought to determine whether or not the cellular mediator of this interaction was a specific protein receptor. To initially address this issue, we determined if the cell surface binding sites were susceptible to proteolytic digestion. Porcine endothelial cells were treated with either trypsin or Pronase (at doses just below those found to detach the cells), and their ability to associate with lZ5I-HDL3 in comparison with untreated cells was examined (Fig. 8). The data show that treatment by neither enzyme affected 1251-HDL3-cell association (note that the three curves in Fig. 8 are partially superimposed). In the '251-LDL-fibroblast system, where a specific protein receptor-ligand interaction is known to occur, the lipoprotein-cell association is abolished by proteolytic treatment of the cells (37). Although the data in Fig. 8 do not rule out the existence of a specific HDL, receptor protein on porcine endothelial cells, since such a protein may simply be "hidden" from or resistant to the enzymes, they are consistent with the possibility that there is no HDLs receptor protein on these cells.
To further address the issue of the cellular determinants of HDL3-cell association, we examined cells with "up-regulated HDL3 binding sites. Oram  cellular association as compared with cells that had not been preincubated. We chose to examine the cholesterol-treated human fibroblasts because 25-hydroxycholestero1(50-100 pg/ ml) was toxic to our endothelial cells. When the fibroblasts were preincubated for 24 h with nonlipoprotein cholesterol, al. (18). We examined the specificity of the interaction of the up-regulated cells with HDL3 by testing the ability of these cells to interact with T-HDL3. As shown in Fig. 9, T-HDL3 was a better competitor than HDL, on a protein weight basis in its ability to displace 1251-HDL3 from cholesterol-treated fibroblasts, similar to the results obtained with non-cholesterol-treated cells (see Fig.  4). In addition, progressive enrichment of fibroblasts with free choiesterol by incubation with increasing amounts of nonlipoprotein cholesterol led not only to an increase in lz5I-HDL3-cell association, as reported previously (18), but also to an increase in lz51-T-HDL3 eeIi association (Fig. IO). Thus, up-regulated binding sites of cholesterol-treated fibroblasts show the same lack of specificity as the basal-state binding sites of nontreated cells and thus are probably not specific receptor proteins. In addition, the direct correlation between the free cholesterol content of fibroblasts, 90% of which is in the plasma membrane (39), and their ability to associate with T -H D L 3 and lZ5I-T-HDL3 may indicate that cell surface cholesterol is an important determinant of HDL, cell interaction; direct interaction of HDL, with cell surface cholesterol may explain why cholesterol-treated cells bind more HDL3.

fold, as reported by Oram et
Oram et al. (18) have shown that the protein synthesis inhibitor cycloheximide prevented the increased cell association of '251-HDL3 with cholesterol-treated fibroblasts. Since we have suggested that the mechanism of increased "'I-HDL3  cell association with cholesterol-treated cells is related to cholesterol enrichment of the plasma membrane and not to up-regulation of a receptor protein, we sought to determine if the effect of Cycloheximide might be due to prevention of the increase in free chotesterol content of cholesterol-treatedcells.
The data in Table I demonstrate this point; in parallel with cycloheximide's preventing the increased 1z51-HDL3-cell association with cholesterol-treated cells, cycloheximide also prevented the increase in the cells' free cholesterol content, which, as mentioned above, largely represents plasma membrane cholesterol (39). Thus, the prevention by cycloheximide of the increased 1261-HDL3-cell association with cholesteroltreated fibroblasts does not necessarily indicate that synthesis of a protein receptor is necessary for up-regulation but rather suggests, together with the previous data, that the cell surface cholesterol content is a major determinant of HDLs cell interaction.

DISCUSSION
We have concluded by using both direct cell association and competitive displacement experiments with HDL3, trypsinized HDL,, synthetic discs, and lipid vesicIes that the interaction of HDL3 with certain tissue culture cells is mediated not by a specific protein ligand but rather by HDL, surface lipids which probably interact with cellular surface lipids.
The experiments in this study were performed at 37 "C and therefore potentially measure both surface binding and internalization of HDLI. However, other investigators, using similar cell lines to those studied here, report very low levels of HDL3 internalization (6,8,9). For instance, both Tauber et al. (9) working with endothelial cells and Biesbroeck et al. (8) working with fibroblasts, found that less than 25% of cellassociated IZ5I-HDL3 was internalized in 4-6 h at 37 "C. Even if a significant fraction of the cell-associated '261-HDL3 were internalized, it is likely that the ability of a given analogue to competitively inhibit such internalization would be related to its ability to competitively inhibit the prior step of surface binding. These considerations indicate that valid conclusions  Fibroblasts (35-mm plates) were preincubated for 24 h in DMEM containing 0.2% bovine serum albumin alone or plus 50 pg/ml cholesterol and/or 0.5 mM cycloheximide. Some plates were then washed, lipid-extracted, and assayed for free cholesterol and phospholipid content as described under "Experimental Procedures." Parallel plates were washed, incubated with 5 pg/ml '*'I-HDL3 for 1 h at 37 "C, and assayed for cell association and protein content as described under "Experimental Procedures." Values are averages of duplicates, which varied by less than 10%. about surface binding can be drawn from our cell association studies.

24-h preincubation lmI-HDLs-
Our data show that the interaction of HDL, with certain cultured cells probably does not involve the recognition of a specific ligand site. This conclusion was partially drawn from the T-HDL, data, which shows that T-HDL, was more effective, on a protein weight basis, than native HDL in associating with cells (Fig. 3) and in displacing lZ51-HDL3 from cells (Figs. 4 and 10). If a specific HDL protein ligand did exist, it would very probably be destroyed or modified by the extensive trypsin treatment. Also, the equal efficiency of A-I and A-I1 discs in displacing HDL3 indicates nonspecificity of the HDL ligand. It might be argued that a specific ligand site resides in the T-HDL3 core peptides, perhaps a region of peptides with similar properties of A-I and A-11. However, since these core peptides probably represent the lipid-associated regions of the HDL3 apoproteins (35), it seems unlikely that the binding site of a specific protein ligand would be buried in the hydrophobic domain of the HDL particle where it would be relatively inaccessible to a cellular receptor. Our data showing that lipid vesicles without any protein can displace lZ5I-HDL3 from cells, whereas lipid-free apo-HDL, or apo-T-HDL3 cannot displace lZ5I-HDL3 from cells, further support the conclusion that a specific ligand protein does not mediate HDL cellular interaction in our cell culture systems.
Rather than the HDL, apoproteins playing a major role in HDL3 cellular interaction, the fact that our binding and competitive displacement data were approximately normalized when expressed on a per surface lipid basis provides positive evidence that the HDL, surface lipids subserve the major part of this role. The exact nature of the cell surface elements with which the HDL, surface lipids interact has not been definitively determined. The nonsaturability (Fig. 1) and the protease-resistance of the cellular sites (Fig. 8) suggest that a specific protein receptor is not involved. Our data ( Fig.  9 and Table I) and that of Oram et al. (18), which show a direct correlation between the free cholesterol content of the cell (and thus of the plasma membrane (39)) and the ability of the cell to associate with HDL3, may indicate that membrane cholesterol is an important determinant of HDL3 cell association. HDL3 may interact directly with cholesterol-rich patches in the cell membrane, or the plasma membrane cholesterol content may influence the arrangement of other cell surface lipids in a way to optimize their interaction with HDL, surface lipids.
The data from several previous studies also can be interpreted to show that specific HDLs receptor and ligand proteins do not exist. First, other studies, like ouys, show nonsaturability of HDL cellular association at 37 "C (9). Second, Miller et al. ( 6 ) demonstrated that HDL uptake in human fibroblasts could be accounted for solely by nonspecific adsorptive endocytosis and fluid pinocytosis without invoking specific receptor-mediated endocytosis. Third, although we were the first to show the protease-resistant nature of HDL3 cellular interactions with endothelial cells, others have shown protease-resistant HDL association in a variety of other tissues (6,7,11,13,19,20). Last, both Tauber  studying the interaction of apo-A-I and apo-A-I1 discs with ascites cells and erythrocytes, described cellular bound particles which were neither internalized nor removed by trypsin treatment. These data can be interpreted as suggesting nonprotein-mediated cell surface binding.
Other investigations of HDL binding phenomena have led to the conclusion that specific HDL receptor and ligand proteins do exist. Tauber et al. (17) working with endothelial cells, and Oram et al. (18) working with fibroblasts, demonstrated that cycloheximide blocked the increase in lZ5I-HDL3cell association that occurred with sterol treatment of the cells. Although the authors concluded from these data that up-regulation required the synthesis of a protein receptor, we believe these data indicate, in concert with our other data, that up-regulation requires cholesterol enrichment of the cell (specifically the cell membrane) since cycloheximide prevents the increase in the free cholesterol content of cholesteroltreated cells (Table I). Second, several investigators have demonstrated saturability of HDL, binding at 4 "C (8,9,13); however, the cell association became non-saturable at 37 "C (9). The only systems shown to have saturable '251-HDL3-cell association at 37 "C are steroidogenic tissues (11); the interaction of HDL, with these tissues may involve different mediators from those in the cells we studied (see below).
There have been several abstract reports in which the authors have concluded that the HDL-tissue interaction involves specific HDL ligands. Brinton et al. (23) reported that albumin-containing liposomes competed less efficiently than native HDL3 in displacing lZ5I-HDL3 from fibroblasts. HOWever, they also found that apo-A-I and apo-A-I1 liposomes could both effectively displace lZ5I-HDL3 from cells. Their report that tetranitromethane treatment of HDL3 or apo-A-I liposomes abolished their binding, whereas treatment with other protein-modifying agents did not, may be difficult to interpret since we have subsequently found that tetranitromethane causes apoprotein aggregation: In summary, we conclude that the nonsaturability, the protease resistance, and the nonspecificity of HDL3 cell as-I. Tabas and A. R. Tall, unpublished observations. sociation in our cell culture systems indicate that the mediators of this cell association do not include specific receptor and ligand proteins. Rather, based on data showing both excellent dsiplacement of lZ5I-HDL3 by lipid vesicles (and poor displacement by lipid-free apoproteins), equalization of competitive displacement abilities of HDL3 analogues on a per surface lipid basis, and a positive correlation between cell cholesterol content and HDL3-cell association, we conclude that the surface lipids of HDL3, probably interacting with cell surface lipid, possibly cholesterol, are the major mediators of HDL3-cellular association in our cell culture systems.
Our results do not exclude a relatively nonspecific role of the apoproteins in HDL3-cellular interactions in addition to the role of the surface lipids. For example, the HDL apoproteins, known to have prominent hydrophobic regions and lipid-binding properties (for review, see Ref. 40), may directly interact with cell surface lipids. Similarly, it is possible that cell surface proteins may play a relatively minor, nonspecific role in interacting with the surface of HDL3. These possibilities may explain previous data showing that trypsin treatment of cells to which HDL3 has been bound release a portion of the HDL, (9, 18). That trypsin releasability of a bound lipoprotein does not necessarily indicate specific protein receptor-ligand interaction was demonstrated by Brown et al. (41) who showed that trypsin could release 1251-LDL which had been bound to plastic Petri dishes that contained no cells.
The physiological significance of our conclusions about the mediators of HDL-cellular association remains to be determined. It is possible that lipid-lipid interactions mediate such proposed HDL-related processes as cellular cholesterol removal and HDL cholesterol ester delivery. In particular, Biesbroeck et al. (8) have suggested that HDL3 binding to cells is important in cellular cholesterol removal. Furthermore, Phillips et al. (42) have provided evidence that the diffusion of cholesterol through the cell's large unstirred water layer to the acceptor may be a rate-limiting step in cellular cholesterol removal. Thus, the binding of cholesterol acceptors (such as HDL) to cells could increase the rate of cellular cholesterol removal by increasing the penetration of the acceptor into the cell's water layer. These workers also showed that the ability of different recombinant HDL particles to remove cholesterol from tissue culture cells could be normalized on the basis of external surface area (Fig. 6 in Ref. 43). A possible interpretation of this observation, in view of our data (Fig. 7), is that surface lipids of the particles mediate their interaction with the cells, and that this interaction, in turn, facilitates cellular cholesterol removal. If HDL3 binding is, in fact, important in cellular cholesterol removal, and if HDL3 interacts with cell surface cholesterol, as we have suggested (see above), then an important mechanism in reverse chlesterol transport from cholesterol-loaded cells may involve increased HDL3 binding to the cholesterol-enriched surface of these cells.
If lipid-lipid interactions were physiologically important in the case of HDL cholesterol or cholesterol ester delivery, one wonders how much specificity such a process would require and what, in the absence of a specific protein receptor-ligand system, would confer such specificity. Bamberger et al. (44) have shown that HDL cholesterol delivery to hepatoma cells is mediated by hepatic lipase, probably secondary to hydrolysis of HDL surface lipids. Therefore, while the surface lipids of HDL could potentially bind to the surface lipids of all cells, perhaps only those cells with a specific mediator molecule, such as a lipase, would subsequently interact with the bound HDL in a physiologically relevant manner. Such a molecule could therefore confer a certain degree of specificity to these types of HDL-tissue interactions.
Although we have examined in this study three different cell types from three different species, it is possible that the association of HDL with other cells and tissues is, in fact, mediated by a specific receptor-ligand protein system. In particular, several steroidogenic tissues show saturable HDL association at 37 "C (ll), hormone-mediated increases in HDL binding (10,11,19), relative ligand specificity (14), and selective HDL uptake (45). However, even in these systems no specific receptor or ligand proteins have been definitively identified, and in several cases the binding has been shown to be protease-resistant (11,19). Thus, either a specific protein receptor-ligand system in steroidogenic tissues exists and has not yet been identified, or some mechanism other than a specific receptor and ligand confers selectivity to this particular HDL-tissue association. Interestingly, adrenal tissue has been shown to possess a lipase and thus may selectively interact with HDL in a similar manner to that proposed above.
The properties of HDL-cell association are markedly different from those of LDL-cell association, which is mediated by a specific ligand-receptor interaction: LDL-cell association shows saturability, specificity, and cellular protease-sensitivity (37). In addition, several investigators have tested the ability of trypsinized LDL to interact with fibroblasts (46, 47). In these studies, trypsin removed only 20% of the LDL protein instead of the 60% removed from HDL in our study, and gel electrophoresis of the trypsinized LDL revealed fragments larger than 70,000 daltons (46). Even with this comparatively limited proteolysis, the trypsinized LDL still did not displace lZ5I-LDL from fibroblasts as well as native LDL on a protein weight basis (46) in contrast to our results where T-HDL3 was a better competitive displacer than native HDL,.
Much of our knowledge of lipoprotein metabolism has come from the studies of the LDL-fibroblast system; the identified cellular protein receptor has been the subject of numerous and fruitful investigations (37, 48) into the function and regulation of LDL cellular interactions. Our study has demonstrated that HDL-cellular association, in which surface lipids play a more important role than surface proteins, is fundamentally different from LDL-cellular association. Further identification of the mediators of HDL-cellular interaction may lead to a deeper understanding of the function and regulation of HDL metabolism.