Hydrophobic Interactions of the apo-Gin-I Polypeptide Component of Human High Density Serum Lipoprotein*

Apo-Gln-I, the major polypeptide component of human serum high density lipoprotein, has four noninteracting hydrophobic sites which associate with alkanes, anionic detergents, and cationic detergents. Hexane and octane bind to these sites with association constants of 6.8 times 10-2 and 1.8 times 10-4 liters/mol, respectively, and compete with the anionic detergent, sodium dodecyl sulfate (C12OSO3-minus), at low detergent ligand binding ratios (i.e. smaller than or equal to 1.0 mol of C12OSO3-minus per mol of protein). At higher detergent binding ratios (larger than 2 mol of C12OSO3-minus per mol of protein) the polypeptide cooperatively binds alkanes and a conformational change is induced.

serum high density lipoprotein, has four noninteracting hydrophobic sites which associate with alkanes, anionic detergents, and cationic detergents. Hexane and octane bind to these sites with association constants of 6.8 x lo2 and 1.8 x lo4 liters/mol, respectively, and compete with the anionic detergent, sodium dodecyl sulfate (C120S08->, at low detergent ligand binding ratios (i.e. Il.0 mol of C120SOsper mol of protein). At higher detergent binding ratios ( >2 mol of C120S03 per mol of protein) the polypeptide cooperatively binds alkanes and a conformational change is induced.
Serum lipoproteins appear to be involved primarily in lipid transport (1) and are thought to be the source of the lipid deposits (i.e. plaques) in atherosclerosis (2, 3). Despite the physiological importance of these protein-lipid complexes, their molecular organization is not well defined. Reassembly studies have shown that the lipid-free polypeptide components of human serum high density lipoprotein (apoHDLl) interacts with phospholipids and cholesterol to give products that have a protein-lipid ratio similar to that of native HDL (4). (Native HDL is 50% protein, 27% phospholipid, 4.5% triglyceride, 14% cholesterol esters, and 4.5y0 cholesterol (4).) However, the molecular interactions between these lipids and the polypeptide components of apoHDL are not well understood.
Efforts in this laboratory have concentrated on characterizing the binding sites on the polypeptides from HDL in terms of their interaction with various amphiphilic ligands. Reynolds and Simon have demonstrated that the binding of the anionic amphiphile, C120SO;-, to apo-Gln-I and apo-Gln-II involves three to four discrete binding sites on the surface of each polypeptide in aqueous solvent (5). The association constants were found to be 2 X lo4 liters/m01 for both apo-Gln-I and apo-Gln-II.
In a fur- (Nutritional Biochemicals Corp.) at pH 5.0 was measured and found to be identical with that previously reported (12).
Methods-The binding of volatile hydrophobic ligands to proteins can be conveniently measured by use of a gas solubility cell as described by Wishnia and Pinder (13). An all glass apparatus was used having four interconnected chambers that permit exchange of the gas but not the liquid phase. The radioactive solute,  Aliquots of the aqueous phase were withdrawn with a 100-J calibrated Hamilton syringe-and injected into a scintillation vial and counted. The snecific activities of n-ll-aHloctane and n-[l-%1hexane were 7.47 X 1Om6 and 1.77 X lo-; pmol/cpm, respectively.

RESULTS
Renaturation of apo-Gln-I from 0.1% Sodium Dodecyl Suljate-The pooled fractions of apo-Gln-I from the 0.1% Cr20SOzcolumn were filtered through a Dowex anion exchange resin in order to rapidly remove the denaturing detergent. When the detergent-free apo-Gln-I was reapplied to the same column, without CLzOSO~-in the eluting buffer, a Stokes radius of 26 A was obtained compared to the 44 A observed in the presence of detergent. The change of 18 A in Stokes radius from the denatured state is in agreement with previous results and indicates that apo-Gln-1 has renatured to form a highly compact structure (5). The near ultraviolet circular dichroism of apo-Gln-I in aqueous solution, obtained from the Dowex anion exchange resin, is shown in Fig. 5 (s&d curzie 1) and is the same as that previously observed by stepwise renaturation from 6 M guanidine hydrochloride (6).
Sedimentation equilibrium studies of apo-Gln-I renatured from both 6 M guanidine hydrochloride (by stepwise dialysis) and from 0.1% Cr20S03 (by Dowex anion exchange resin) at pH 8.3 and p = 0.038, show an identical, but weak monomerdimer association. On a molar basis there is approximately 85% monomer (28,500 MW) and 15% dimer (57,000 MW) at a total protein concentration of 0.7 mg/ml. These sedimentation equilibrium results will be discussed in detail in a forthcoming publication. The above results reaffirm the previous observation that apo-Gln-I regains an ordered structure independent of the original denaturating solvent (5, 6). In the case of Cr20SOz-the renatured state appears to be the same whether the detergent is removed slowly by stepwise dialysis or rapidly by use of an anion exchange resin.
Binding of n-Octane and n-Hexane to apo-Gln-I-The binding data for n-[1-aHloctane and n-[1-r'C]hexane at 0" are presented in Figs. 1 and 2, respectively. Reversible binding was demonstrated as shown in Fig. 2 where points obtained by either increasing or decreasing the concentration of alkane fall on the same binding isotherm. In addition, the binding of both alkanes is independent of the apo-Gln-I concentration.
The binding isotherm for n-hexane was obtained over a g-fold range of protein concentrations, i.e. from 0.79 mg/ml to 7.02 mg/ml. This indicates that the alkane binding sites are the same for both monomer and dimer apo-Gln-I and are unaffected by monomerdimer association.
Binding isotherms for Cr20S03-and CllNMejt have previously been fit to four independent and identical binding sites with an association constant of 2 x lo4 liters/mol (5, 6). An analysis of the binding data in Figs. 1 and 2 by iterative curve fitting (15) has been made by assuming both one and four independent sites. The curves assuming four sites (solid) rather than one site (dashed) give a better fit to the experimental data. The calculated association constants, assuming four sites, are 6.8 x lo? liters/m01 and 1.8 x lo* liters/m01 for n-hexane and n-octane, respectively.
Although the assumption of four independent and identical sites adequately fits most of the data in Figs. 1 and 2 it is likely that a more complex combination of nonequivalent and slightly cooperative sites might fit the data equally well. In this regard it should be kept in mind that the free concentration of alkane is limited by its saturation vapor pressure at 1 atm in the solubility cell. The experimentally observable portion of the protein-alkane binding isotherm is thus limited by the competing process of alkane phase separation. (A complete discussion of hydrocarbon binding to water-soluble proteins is given by Tanford (16).) Alkane Binding to apo-Gln-1 in Presence of Detergent-The binding of n-hexane and n-octane to apo-Gln-I was measured at varying initial concentrations of C&OS03 as shown in Figs. 3 and 4. When 0.85 mol of Cr20S03-per mol of apo-Gln-I was initially bound, a noticeable reduction in hexane binding was observed compared to binding in the absence of detergent (see Fig. 3). This result is compatible with the suggestion that the alkane and detergent binding sites of apo-C&-I are identical. At higher initial C&OS03 concentrations, however, a coopera- tive interaction occurs which results in higher levels of alkane binding than are observed for apo-Gin-I without detergent present. Fig. 4 shows that for n-octane this effect was observable at initial binding levels of 2, 3, and 6 mol of C120S03-per mol of apo-Gln-I.
The near ultraviolet circular dichroism of apo-Gin-1 with an initial binding of 6.0 mol of Ci20S03-per mol of apo-Gln-I is shown in Fig. 5 and is nearly identical with that previously observed at 140 mol of Ci20S03-per mol of apo-Gin-I (6). It is important to point out that similar spectral changes have been observed in the presence of saturation binding levels of &a-NMe3+ and deoxycholate (6). The near ultraviolet CD of holo-HDL has been measured by Lux et al. (17) and bears a striking resemblance to the spectra of apo-Gln-I in the presence of 6 to 140 mol of C120SOsP per mol of polypeptide. In particular, the location and sign of the extrema at 290 nm and 283 nm are similar in location and sign and have been assigned by Lux ef al. (17) to 1 or more tryptophanyl residues of apo-Gin-I. The addition of octane (at its solubility maximum, 8.78 X 1OP M) to apo-Gln-I in the presence of detergent had little effect 911 the CD spectrum despite the fact (as shown in Fig. 4) that cooperative octane binding has occurred.
* The aqueous solubilities were 186 X 10-G and 5.2 X 10e6 M for n-hexane and n-octane, respectively. 1. The alkanes, n-hexane, and n-octane, as well as C&OS03and C14MNe3+ interact with approximately four hydrophobic binding sites on apo-Gln-I with association constants and free energies of transfer (pop -pow = -RT In K)z as shown in Table I. Reynolds et al. (18) have recently discussed in detail the proportionality between the free energy of transfer from water to hydrocarbon (poxc -pow)2 and the hydrophobic surface area of the solute. Fig. 6 (solid line) shows poHc -low for the transfer of n-alkanes as a function of the total surface area. The dashed line in Fig. 6 gives pop -pow for alkanes transferred from water to apo-Gln-I and it is apparent that this latter free energy is less favorable than that of transfer to a hydrocarbon medium. This result is in marked contrast to alkane binding to the hydrophobic sites of fl-lactoglobulin, myoglobin, and hemoglobin. Wishnia (19,20) found pop -pow for a series of n-alkanes binding to these proteins to be more favorable than the corresponding free energy of transfer to a hydrocarbon solvent. This result has been attributed to a nearly complete removal of the alkane from contact with water when bound to interior hydrophobic sites together with a small contribution from "strain" or locally suboptimal interactions in the unoccupied sites (20).
The free energy of transfer of alkanes to Ci20S03-micelles, however, is less favorable than poHC -pow, as is shown in Fig.  7, and this phenomenon has been attributed to a slightly smaller entropy of transfer (19) which reflects some molecular organization near the surface of the micelle (16). The hydrophobic binding sites on apo-Gln-I are clearly not like the interior "pockets" of fl-lactoglobulin, myoglobin, and hemoglobin, i.e. pop -pow > 1.1'~~ -pow for ape-Gln-I, whereas pop -pow < poHc -pow for P-lactoglobulin, myoglobin, and hemoglobin.
Most probably the apo-Gln-I sites are surfacelocalized and when occupied may still have some contact with water. Alternatively, the sites could have a slightly smaller entropy of transfer due to some degree of molecular organization, as has been found for &OS03-micelles. By studying the binding of ligands with varying alkyl chain lengths it was determined that the hydrophobic binding regions of serum albumin and &lactoglobulin have a limited size or binding capacity (13,15). We have already demonstrated that the anionic or cationic polar group on two detergent ligands of similar hydrophobicity (&,OSO, and ClaNMea+) has no measurable effect on the free energy of association to apo-Gln-I. The data in Table I for alkane binding (and in Fig. 6) may, therefore, be used to calculate the free energy of binding of C&OS03 to apo-Gln-I. Each A* of hydrophobic surface area contributes approximately 16 cal/mol to pop -pow at 0.0". The surface area of the dodecyl chain is -540 A2 (18), so the predicted pop -pow for C&OS03 binding to apo-Gln-I is --8.6 kcal/mol at 0.0" and -9.7 kcal/mol at 25.0".3 The experimentally determined value of pop -pow for C&OSOB-binding at 25.0" is -8.2 kcal/mol (see Table I), significantly less favorable than predicted. We must, therefore, conclude that the binding sites on this polypeptide, while primarily hydrophobic, exhibit an apparent restriction in size such that they can optimally accommodate alkyl chains shorter than Cs or 10.
2. As we have pointed out previously, these four sites cannot directly account for the large number of lipid molecules bound in uiuo. Furthermore, the conformation of apo-Gln-I in this state is different from that in viva as evidenced by circular dichroism data (Fig. 5, Refs. 6 and 17). This polypeptide, however, undergoes a conformational change in the presence of hydrophobic ligands at unbound ligand concentrations higher than those responsible for occupying the four sites discussed above. The final state of apo-Gln-I in the presence of C120SOs-, C14NMe3+, and deoxycholate at saturation levels is similar to that found in uivo (6) as evidenced by circular dichroism. (It should be noted that in Fig. 1 the onset of a highly cooperative binding phenomenon is observed at the highest levels of octane binding, but is limited by the solubility of octane itself.) We now observe that two hydrophobic ligands at low binding ratios can operate synergistically to induce the cooperative transition at lower concentrations of unbound ligand. Thus, as is shown in Figs. 3 and 4, the presence of low concentrations of C&OS03-leads to the cooperative binding of either octane or hexane to apo-Gln-I.
The implications of this phenomenon to the binding of lipid in uivo are significant. While the free energy of binding of long diacyl phospholipids to apo-Gin-I would not be sufficiently favorable to lead to significant interaction (5, 6), a synergistic effect could operate in tivo in which the essential conformational change results from the binding of a few moles of lysolecithin in the presence of other biological lipids. HoloHDL contains some 4 mol of lysolecithin molecules per HDL particles (1). Preliminary studies of lysolecithin binding to apo-Gln-I in this laboratory by Dr. M. E. Haberland indicates a cooperative interaction accompanied by a conformational change.