Structural properties of high density lipoprotein subclasses homogeneous in protein composition and size.

We isolated native high density lipoprotein (HDL) subclasses homogeneous in size and in their protein content with the objective of investigating the differences and similarities in their apolipoprotein AI (apoA-I) structures. Defined particles were isolated from ultracentrifugally prepared HDL by immunoaffinity and gel-filtration chromatography. The isolated 88-A LpAI, 106-A LpAI, 96-A LpAI/AII particles (LpAI, particles contain only apoA-I; LpAI/AII, particles contain apoA-I and apoA-II), together with a 93-A reconstituted HDL were analyzed for purity, composition, and content of apolipoprotein molecules per particle, and were examined by far and near circular dichroism and intrinsic fluorescence spectroscopic methods, as well as by reaction kinetics with lecithin:cholesterol acyltransferase. The spectroscopic analyses indicated that the secondary structures and three-dimensional arrangements of apoA-I in all these particles are remarkably similar: their tryptophan residues are located in similar nonpolar environments and become exposed to increasing concentrations of guanidine hydrochloride in comparable denaturation steps; the 60-65% alpha-helical structures in apoA-I are denatured in similar patterns with 0-5 M denaturant concentrations. However, increasing surface lipid contents and the presence of apoA-II stabilize apoA-I on the HDL particles. The reaction kinetics with lecithin:cholesterol acyltransferase are similar and slow for the isolated HDL particles, reflecting product inhibition, and/or an apoA-I conformation that is unfavorable for the activation of the lecithin:cholesterol acyltransferase reaction.

sist of a heterogeneous population of particles containing different types and amounts of apolipoproteins and lipids. They are mostly spherical particles ranging in diameter from 70-120 A. Two main subclasses of HDL can be isolated by immunoaffinity chromatography; one of the subclasses contains both apolipoprotein A-I (apoA-I) and apolipoprotein A-I1 (apoA-11) (LpAI/AII), whereas the other subclass contains apoA-I, but no A-I1 (LpAI) (1)(2)(3)(4)(5)(6). Both of these subclasses of HDL include small amounts of other proteins and are heterogeneous in size. Several laboratories have investigated the metabolic behavior of these two heterogeneous HDL subclasses ('7-11). I n vitro studies of the binding of LpAI and LpAI/AII to various cells, and their ability to promote cholesterol efflux from cells enriched in cholesterol, have yielded conflicting results. For example, Barbaras et al. (8) found that LpAI and LpAI/AII particles bound equally well to preadipocytes loaded with cholesterol, but that only the LpAI particles promoted cholesterol efflux from the cells. In contrast, Johnson et al. (lo), using different cells and experimental conditions, were unable to find a significant difference in the behavior of the two HDL subclasses as cholesterol acceptors. Metabolic studies, in uiuo by Rader et al. (11) have shown that the catabolic rate of apoA-I on both types of particles is markedly different. Thus, because of the possible functional differences between the LpAI and LpAI/AII particles it is important to delineate their structural differences, particularly differences in the structure of apoA-I; however, to obtain unambiguous structural information, homogeneous particles are required.
During our investigations of reconstituted HDL (rHDL), we have found that apoA-I can exist in a few well defined conformational states in distinct rHDL particles (12)(13)(14). In discoidal rHDL containing apoA-I and palmitoyloleoylphosphatidylcholine (POPC), each apoA-I is arranged into 6 to 8 antiparallel cy-helical segments joined by fl turns or sheets (12, 15). The nonpolar side of the helices covers the edge of the lipid disc. The discrete diameters of these discoidal rHDL particles are defined by the number of apoA-I molecules per particle (two or more) (12, 13) and by the number of helices in each apoA-I that are in contact with the lipid. The structural differences in the apoA-I molecules were measured by fluorescence and circular dichroism spectroscopic methods, and the distinct structures of apoA-I were correlated with up to 15-fold differences in the reactivity of the rHDL with 1ecithin:cholesterol acyltransferase (LCAT) (12)(13)(14). We have also prepared a spherical rHDL of defined composition and size by reacting the discoidal rHDL with LCAT in the presence of low density lipoprotein (14). The 93-A rHDL product of this reaction, which contains large amounts of cholesterol ester and appears round by electron microscopy, has three apoA-I molecules in a structure comparable to that in the discoidal rHDL precursors.
Since we have firmly established, by using the rHDL models, that apoA-I can exist in well defined conformational states which can determine the functional properties of the rHDL, we set out in this work to isolate native LpAI and LpAI/AII particles of uniform size, by immunoaffinity chromatography and gel filtration, with the objective of characterizing their structures and establishing whether apoA-I exists in different conformations in the different native particles.

MATERIALS AND METHODS
Apolipoprotein A-I, LDL, and LCAT were prepared by routine methods (16-18) from human plasma donated by the Champaign County Blood Bank-Regional Health Resource Center. The purity of apoA-I and LCAT was checked by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (PAGE), and the gels were stained with either Coomassie Blue or silver stain using the Phast System (Pharmacia LKB Biotechnology). Sodium cholate, crystalline cholesterol, POPC, and bovine serum albumin (BSA) were purchased from Sigma; [4-"C]cholesterol was purchased from Du Pont-New England Nuclear; guanidine hydrochloride (GdnHC1) was the electrophoresisgrade F product from Fisher; bis(sulfosuccinimidy1)suberate (BS,) was purchased from Pierce.
Preparation of Anti-AI and Anti-AII Immunosorbents-Monoclonal antibodies against apoA-I and apoA-I1 were prepared by intraperitoneal immunization of BALB/c mice with intact HDL3 and characterized as previously described (19). A mixture (100 mg) of three different monoclonal antibodies (A05, A17, and A30) which recognize all forms of plasma apoA-I was covalently coupled to CNBractivated Sepharose 4B (Pharmacia LKB) as instructed by the manufacturer except that a ratio of 3.5 mg of antibodies/g of gel was used.
In an identical fashion a mixture (100 mg) of three different monoclonal antibodies (G03, G05, and Gll), which recognize all forms of plasma apoA-11, was coupled to CNBr-Sepharose (3.5 mg/g gel). The immunosorbents were packed into borosilicate glass columns of 5-cm internal diameter.
Preparation of Native HDL and rHDL-LpAI and LpAI/AII were isolated using a modification of the Cheung and Albers method (3). Human HDL (d = 1.063-1.21 g/ml) obtained by ultracentrifugal flotation was used in a two-step immunoaffinity chromatography procedure at 4 "C. Typically, 2.5 mg of HDL protein were injected on the anti-apoA-I1 column connected in series with an anti-apoA-I column at a slow flow rate (0.2 ml/min) during a 15-h period. Then the individual columns were extensively washed (2 ml/min) and each retained fraction, LpAI and LpAI/AII, was eluted with 20 ml of 3 M NaSCN. The eluted fractions were immediately desalted on Sephadex G-25 columns at a flow rate of 2 ml/min in order to minimize the time of contact with NaSCN. In all cases, more than 75% of the protein applied to the columns was retained by the immunoaffinity columns, and essentially all of the retained HDL protein was eluted with the NaSCN and recovered after the desalting step. The LpAI and LpAI/AII were concentrated using Centriprep 10 concentrators (Amicon) to 1 mg/ml and finally applied to a Superose 6 gel filtration column for further fractionation by size using a Pharmacia FPLC system. The losses of HDL protein on the Superose 6 column were insignificant, but each of the fractions with uniform sizes used in the subsequent experiments represented only IO-20% of the HDL protein applied to this column. The samples were stored in a standard 10 mM Tris-HC1, pH 8.0, buffer containing 0.15 M NaCI, 0.01% EDTA, and 1 mM NaN3. This buffer was used for all the experiments performed in this study unless otherwise specified.
In a separate experiment, we tested the effects of NaSCN on HDL particle size distribution and reactivity with LCAT. High density lipoprotein (2.5 mg of protein) was incubated with 3 M NaSCN for 3 h at 4 "C. After removal of the NaSCN by dialysis, the particle size distribution for exposed and unexposed HDL was shown to be essentially identical by gradient gel electrophoresis (method described below). Also the reaction kinetics with LCAT (described below) were found to be the same for HDL which had been exposed or not exposed to NaSCN. Therefore, we concluded that the effects of this salt on the structural and functional properties of HDL are insignificant as judged by these sensitive tests.
The 93-A spherical reconstituted HDL (rHDL) was synthesized using the sodium cholate dialysis method (20, 21). An rHDL mixture containing 5 mg of apoA-I was prepared from a molar ratio of 80:81:80, POPC/cholesterol/apoA-I/sodium cholate. This mixture of rHDL was incubated with LDL (1:2 weight ratio of rHDL to LDL protein) in the presence of 12 mg/ml BSA, 4 mM 0-mercaptoethanol, and 20 pg of LCAT at 37 "C for 48 h. Ultracentrifugal flotation was performed first at a density of 1.070 g/ml for two 24-h periods in order to remove the LDL, and then at a density of 1.21 g/ml for 24 h to float the rHDL. Residual LDL and small amounts of a 78-A particle were removed by gel filtration on a Superose 6 column as was previously described (14).
Electrophoretic Analyses of the rHDL Particle and the Native HDL Subfractions-The Stokes diameters of the rHDL and native HDL subfractions isolated after immunoaffinity and gel filtration chromatography were determined by nondenaturing gradient polyacrylamide gel electrophoresis (GGE) on Pharmacia precast PAA 4/30 gels as described (22). The following proteins were used as standards for the Stokes diameters of the HDL particles: BSA, 71 8,; lactate dehydrogenase, 82 8,; horse ferritin, 122 A; and thyroglobulin, 170 A.
The apolipoprotein composition of the native HDL subfractions was determined by SDS-PAGE and silver staining on 20% gel slabs.
Lipoprotein Analysis-The protein concentration was determined using the method of Lowry et al. (23) and the absorbance at 280 nm with an extinction coefficient of 1.13 ml/mg.cm for the rHDL and LpAI particles and an estimated extinction coefficient of 0.96 ml/ mg.cm for the LpAI/AII particles. The last extinction coefficient was estimated by assuming that the LpAI/AII particle contains an equimolar quantity of apoA-I and apoA-I1 and using molar extinction coefficients of 31,700 M" cm" and 12,000 M" cm" (24) and molecular weights of 28,000 and 17,500 for apoA-I and apoA-11, respectively. The content of apoA-I and apoA-I1 in LpAI/AII was also determined by an enzyme-linked immunosorbent assay (25). The Chen et al. method (26) was used to determine the phospholipid content, extending the standard curve down to 0.16 r g of inorganic phosphate. Total cholesterol was determined using the enzymatic assay of Heider and Boyett (27).
The number of apoA-I and apoA-I1 molecules per particle for the native HDL subfractions was determined by cross-linking with 10 mM BSs for 3.5 h in a modified version of the Staros technique (28).
The cross-linking reaction mixture was quenched using 250 mM ethanolamine and subsequently run on SDS-PAGE (IO-30%). Free apoA-I and free apoA-11, as well as a 1:l weight ratio of these two apolipoproteins, were also subjected to cross-linking with BSa and SDS-PAGE in order to serve as standards. In addition, mixtures of apoA-I to apoA-I1 in different molar ratios were run on the SDS-PAGE gel along with the 96-8, LpAI/AII particle. The intensity of the stained bands was quantitatively assessed using a Pharmacia-LKB Ultro Scan XL laser densitometer. The observed intensities indicated that the apoA-I and apoA-I1 were present in a 1:l molar ratio on the 96-8, LpAI/AII. The 93-8, spherical rHDL has been previously shown to contain 3 apoA-I molecules per particle (14).
Fluorescence Spectroscopy-Uncorrected tryptophan (Trp) fluorescence spectra were measured with a Perkin-Elmer MPF-66 fluorescence spectrophotometer at 25 "C. An excitation wavelength of 280 nm and 4-nm slit widths were used. The samples were adjusted to an absorbance value of approximately 0.05 at 280 nm with standard salt buffer which was shown to contribute little to the spectral region of interest, 330-355 nm. The same samples were also used to measure the intrinsic fluorescence polarization values at 25 "C with an SLM model 400 fluorescence polarization instrument using a 280-nm exciting light, 4-nm slit widths, and Corning glass 0-54 emission filters, Denaturation with solid GdnHCl was monitored by following the change in the wavelength of maximum fluorescence of the T spectra. Solid GdnHCl was added to 0.7 ml of free apoA-I, the 9 3 3 rHDL, or the native HDL subfractions directly into the cuvette in a sequential manner. The concentrations of GdnHCl ranged from 0.72 to 6.37 M. The time required for each addition of GdnHCl, mixing, and recording of spectra was approximately 3 min. This denaturation experiment was performed on two separate preparations of the 93-A rHDL and the native HDL subclasses, giving very similar results both times.
Circular Dichroism Measurements-Circular dichroic spectra were recorded with a Model CD6 Jobin Yvon, ISA (Longiumeau, France) spectropolarimeter at the Laboratory for Fluorescence Dynamics, University of Illinois, Urbana. In the far-ultraviolet region (200-250 nm) measurements were made in a 1-mm path length quartz cell. The rHDL and HDL samples were adjusted to an absorbance value at 280 nm of 0.1 for the denaturation experiments and the determination of the percentage of a-helicity. The mean residue ellipticity in units of deg.cm2 dmol" was calculated from the following equation: where Oh is the observed ellipticity in degrees at wavelength X, 1 is the optical path in cm, c is the concentration in g/ ml, and MRW is the mean residue weight. We used a value of 115 g/ residue for the rHDL and LpAI subfractions and a value of 114.6 g/ residue for the single purified subfraction of LpAI/AII (assuming that apoA-I and apoA-I1 are equimolar in this particle). The percentage of a-helicity was calculated from the empirical expression of Chen et al. (29). Denaturation with GdnHCl was also monitored using the change in the observed ellipticity at 222 nm. The conditions used were the same as those in the fluorescence denaturation experiment described above except that the absorbance at 280 nm of the samples was adjusted to 0.1. This experiment was repeated on two to three separate preparations of particles with very similar results.
For the CD measurements in the near-ultraviolet region (250-320 nm) a 10-mm path length quartz cell was used, and the optical density of all samples was adjusted to approximately 0.2 at 280 nm. The nearultraviolet CD spectra are the average of eight scans. Base-line runs were made prior to each sample run and the base-line was subtracted to obtain the final spectrum. The spectra were reproducible for at least two separate preparations of particles.
LCAT Reaction Kinetics-The rHDL and native HDL particles were incubated with aliquots of ['4C]cholesterol dispersed in 2% BSA containing approximately 4 X 10' cpm for 3 h at 37 "C, in order to label the substrate particles (30). The reaction mixtures for the kinetic analysis contained substrate concentrations ranging from 8 X lo-' to 2 X 10"j M apoA-I or apoA-I plus apoA-11, 2 mg of defatted BSA, 4 mM P-mercaptoethanol, and 0.4 or 0.5 pg of pure LCAT. The mixtures were incubated at 37 "C for 2 h. Lineweaver-Burk plots were constructed from data for four particle concentrations and the corresponding initial velocity results. The inverse slope of the Lineweaver-Burk plot gives the Vmax,spp)/Km(app), an indicator of the overall reactivity of the particles. The Vmax(spp)/Km(app) was adjusted for any differences in enzyme concentration.

RESULTS
This work represents one of the first attempts to isolate and characterize distinct size classes within the two main fractions of HDL, LpAI, and LpAI/AII. Fig. 1 shows a photograph of a 4-30% polyacrylamide gradient gel containing the LpAI and LpAI/AII fractions isolated by gel filtration. Panel A shows that LpAI has bee? successfully separated into two discrete sizes of 106 and 88 A as determined by gradient gel electrophoresis. These particles are free of contaminating apoA-I which appears as a double band at the bottom of the gel. Panel B illustrates the fractionation of the LpAI/AII. There appear to be three pain size classes in this fraction inclyding 96-, 87-, and 80-A particles. Lane F represents the 96-A size class which was purified and used in subsequent studies. The smaller particles shown in lane E were not effectively separated. Similar size classes of HDL have been observed in other laboratories starting with either plasma or ultracentrifugally isolated HDL in the immunoaffinity chromatography separation procedure (3,6,(31)(32)(33). Cheung and Albers (3) reported two mean Stokes diameters for LpAI of 10.8 and 8.5 nm and three mean Stokes diameters for LpAI/ AI1 including 9.6, 8.9, and 8.0 nm which agree very well with our results.
The photograph in Fig. 2  were largely eliminated by the gel filtration step, were not investigated, but should include the C apolipoproteins, and apoE. Table I lists the sizes of the native HDL subfractions and the Teconstituted rHDL sphere as determined by GGE. The 93-A spherical rHDL has been previously synthesized and characterized in our laboratory (14). Table I also contains the composition of the native and synthetic particles, as well as the number of molecules of apoA-I and apoA-I1 determined by cross-linking with BS3.
The cross-linking results indicate that the LpAI 88-8, particle contains 3 apoA-I molecules while the lot-8, particle contains 4 apoA-I. The cross-linking of the 96-A LpAI/AII gives a cross-linked protein with an equivalent migration to 2 apoA-I plus 2 apoA-I1 molecules or 3 apoA-I plus 1 apoA-I1 molecules on SDS-PAGE. In order to determine which of these ratios was correct, the intensity of protein staining for the 96-A LpAI/AII particle uersw that of standard molar  The diameters were obtained by GGE (k2 A).
'Phospholipid (PL), total cholesterol (TC), and protein (Prot) were determined in duplicate for a representative preparation. e Determined by cross-linking the apolipoproteins with bis(sulfosuccinimidy1) suberate (28) and analysis by SDS-PAGE.
The previously determined composition for a similar 93-A rHDL particle was 4412411 (14).
e An estimated extinction coefficient of 0.96 ml/mg.cm and the combined molecular weights of apoA-I and apoA-I1 were used in calculating the composition ratios. We assume two protein units (apoA-I plus apoA-11) per particle. ratios of apoA-I and apoA-I1 was determined on an SDSpolyacrylamide gel. A molar ratio of 1:1 (apoA-1:apoA-11) was found. Furthermore, enzyme-linked immunosorbent assays of this LpAI/AII fraction gave similar apoA-I and apoA-I1 concentrations: apoA-I = 3.21 PM and apoA-I1 = 2.94 p?. Thus, based on these results, it was concluded that the 96-A LpAI/ AI1 particle contains 2 apoA-I molecules and 2 apoA-I1 molecules per particle.
From the determined stoichiometries and the volumes of the components, the diameter of the particles as spheres was calculated. Ratios of cholesterol ester to free cholesterol were taken from James et al. (34) who subjected HDL, and HDL3 to a similar immunoaffinity chromatography fractionation and determined these lipid ratios for LpAI and LpAI/AII found in both the HDLz and HDLs subclasses.  (14) probably accounts for the 3-A difference in diameter. The diameters calculated for the native HDL subfractions are all smaller than the sizes determined by gradient gel electrophoresis. These differences in size could be due, in part, to the triglycerides present in the native HDL.
Apolipoprotein Conformation Based on Spectroscopic Results- Table I1 coptains the results from the spectroscopic studies on the 93-A rHDL, the native HDL subfractions, and free apoA-I. The wavelength of maximum intrinsic fluorescence indicates the relative polarity of the Trp residues in the apoA-I molecules. Since apoA-I1 does not contain any Trp residues, only the apoA-I Trp fluorescence is detected in the LpAI/AII particles. Because the wavelength values range from 331 to 334 nm, the Trp residues must reside in a fairly nonpolar environment.Jt seems that the native HDL subfractions, mainly the 106-A LpAI and 96-A LpAI/AII particles, may have their Trp residues in slightly more nonpolar environments especially compared to free apoA-I. The intrinsic fluorescence polarization values reflect the segmental motions of the Trp residues and their fluorescence lifetimes. As previously observtd, the free apoA-I has the highest polarization value. The 93-A rHDL and the native HDL subfractions seem to be similar in the dynamic behavior of their tryptophan residues.
The a-heljcal content is 55% for the free apoA-I and 65% for the 93-A rHDL. These values are within the expected error of the results published previously by our laboratory (12, 14). The LpAI subfrac:ions are quite similar in their a-helical content while the 96-A LpAI/AII particle has much lower ahelicity. This lower a-helicity is due to the presence of apoA-11, since it is well established that the percentage of a-helix is lower in apoA-I1 than apoA-I (37,38). Stoffel and Preissner (38) have shown that apoA-I1 in aqueous solution has 27% ahelical structure, and upon recombination with phospholipids, increases in helix content to 40%.
Denaturation Studies with Guanidine Hydrochloride- Fig.  3 show! the results of GdnHCl denaturation of free apoA-I, the 93-A rHDL, and the native HDL subfractions monitored by the change in the wavelength of maximum fluorescence. Free apoA-I denatures quite readily in 2 M GdnHCl as expected (12, 14,39,40). The results in Table 11 indicate that 50% denaturatiop has occurred at a 1.06 M GdnHCl concentration. The 8t-A LpAI subfra$ion is more readily denatured than the 106-A LpAI and 96-A LpAI/AII subfractions. It is possible that the apoA-I1 on the LpAI/+II subfraction stabilizes the structure of apoA-I. The 93-A rHDL is the most stable particle requiring 4.8 M GdnHCl in order to attain 50% denaturation (Table 11). From the plateaus in the denaturation curves one may speculate that the different regions of apoA-I containing the 4 Trp residues are denaturing independently from each other.
The results of the denaturation with the same concentrations of GdnHC1, monitored by circular dichroism, are given in Fig. 4. It appears that the loss of secondary structure followed by the change in the ellipticity does not occur at the same rate as the change in the wavelength of maximum fluorescence. Fifty percent of the secondary structure is lost at GdnHCl concentrations le$s than 2 M for the native HDL subfractions. Again, the 93-A rHDL appears more resistant to denaturation by GdnHCl, since it requires greater concentrations of GdnHCl to reach the 50% denaturation point.
Near-ultraviolet Circular Dichroism-The near-ultraviolet (250-320 nm) CD spectra (Fig. 5) were normalized to an optical density of 0.1 (at 280 nm)* for the three native HDL subclasses, the total HDL, the 93-A rHDL, and the free apoA-I. Two peaks at 284 and 291 nm are evident for the different native HDL a?d rHDL subfractions. The maximum at 284 nm for the 88-A LpAI is weaker. One minimum at 296 nm is observed for all of these particles, and a broad negative band with several extrema between 260 and 280 nm is observed for the particles that contain only apoA-I. A strong minimum at 272 nm characterizes the speFtrum of the 96-A LpAI/AII particle. The spectra of the 96-A LpAI/AII and the total HDL (1.063 g/ml< d < 1.21 g/ml) are very similar except that the strong minimum is blue-shifted at 269 nm for the total HDL. The spectrum of the free apoA-I is quite different. The peaks observed for the HDL particles at 291 and 284 nm are replaced by a minimum at 293 nm and a shoulder at 287 nm, and the minimum at 296 nm is no longer present. Also a positive plateau of low ellipticity replaces the negative band centered between 262 and 272 nm. As the extracted lipids of HDL have no optical activity between 250 and 320 nm, the observed differences in the spectra can be attributed to differences in protein conformation (37). ' Fluorescence polarization values were measured at 25 "C, and 280 nm was the excitation wavelength. The errors are k0.003.

TABLE I1 Fluorescence spectroscopy, circular dichroism, and GdnHCl denaturation results on free apoA-I, 93-A rHDL, and native HDL subclasses
'The % a-helix content was estimated from the empirical formula of Chen et al. (29) and molar ellipticity values at 222 nm. The  were determined from the inverse slopes of Lineweaver-Burk plots. The values shown are similar for the different particles and are very low compared to discoidal rHDL substrates of LCAT (12, 14). The 106-4 LpAI particles appear to be the least reactive and the 96-A LpAI/AII the most reactive particles in these preparations.

DISCUSSION
Prior to structural analysis of the three isolated subclasses of HDL, we needed to establish their homogeneity. They were shown (Fig. 1) to be quite homogeneous in size by nondenaturing GGE. Further proof of the homogeneity of the purified particles was the detection of a single band upon cross-linking of the proteins followed by analysis on SDS-PAGE. By SDS-PAGE we have also established that the LpAI fraction was completely depleted in apoA-I1 (Fig. 2, lune E ) and that apoA-I represented greater than 95% of the proteins contained in  the two size subclasses of LpAI (88 and 106 A). By the same technique apoA-I and apoA-I1 have beep shown to represent greater than 90% of the proteins in 96-A LpAI/AII particles.
The sizes given in Table I are similar to those reported by other laboratories (3,6,(31)(32)(33). However, comparisons between our compositional data and those of other laboratories are not easily accomplished since we have isolated pure size classes within the LpAI and LpAI/AII fractions where most other research groups have not. Only Nichols et ul. (3) have reported the compo$tion of a purified 10.6-nm LpAI size subclass. Our 106-A LpAI subfraction contains a similar amount of phospholipid, but less total cholesterol and more protein. These differences in composition may be due to differences in the homogeneity of the isolated particles and to the inherent ability of a fixed protein framework to contain more or less lipid, as shown for discoidal rHDL particles (14). The number of apoA-I molecules per LpAI 106-A particle agrees with the results of Nichols et al. (41) showing that 4 apoA-I molecules per particle are present on adult HDLzt, LpAI. We found 2 apoA-I and 2 apoA-!I molecules per particle (i.e. a 1:1 molar ratio) within the 96-A LpAI/AII subfraction. Other investigators have determined this molar ratio to range from 0.8:l to 3:l for ultracentrifugal fractions of HDL (31).  The Vmal(a~p)/Km(app) was obtained from Lineweaver-Burk analysis of the initial velocity uersus apolipoprotein concentration data. The results are the average of two experiments, each performed with four apolipoprotein concentrations ranging from 8 X to 2 X lo-' M.
The correlation coefficients from linear regression analysis were between 0.987 and 0.999. * The apolipoprotein concentration was calculated from the combined extinction coefficient and molecular weights for apoA-I and apoA-11, for a particle containing equimolar amounts of the two apolipoproteins.
The well-defined, highly reproducible sizes of the isolated LpAI subclasses and the 96-A LpAI/AII particles suggest the existence of fixed protein frameworks which determine the diameters of the different HDL subclasses. Close packing of the proteins on the surface of the particles is also suggested by the cross-linking results and the protein and lipid ratios on their surfaces. The efficient chemical cross-lipking (with BS3) of the 3 and 4 apoA-I molecules on the 88-A LpAI and 106-A LpAI pcrticles, respectively, and 2 apoA-I plus 2 apoA-I1 on the 96-A LpAJ/AII particles, indicates proximity between the protein monomers. Assuming a surface area around 4000 A* for apoA-I (36) the percent surface area occupied by protein is about 50 and 45% for the 88 and 106-A LpAI particles, respectively. This, together with the lipid/protein ratios, implies that protein-protein inter-and intramolecular contacts must be extensive. Assuming the extreme case where each a-helical segment of apoA-I is surrouaded by phospholipids (8 a-helical segments, each with 90-A of periphery in $ontact with pho!pholipids which occupy a surface area of 68 A* (36) and 8.2 A of linear distance) one can calculate that each apoA-I would be surrounded by 88 molecules of phospholipid. Since the LpAI particles contain only 20-23 molecules of phospholipid per apoA-I, it is clear that most of the a-helical segments of apoA-I must be in contact with each other in the protein monomers, and that some regions of the protein monomers must be adjacent to other monomers to account for the small amounts of surface lipids.

Wavelength (nm)
The present 93-;1 rHDL particle preparation contains a similar amount of phospholipid compared to our previous ones, but a substantially lower amount of cholesterol (14). Compared with the native HDL subclasses, the rHDL contains considerably more phospholipid aad less total cholesterol. In the past, we considered the 93-A rHDL to be a good model for mature HDL in the plasma compartment; powever, due to this difference in lipid composition, the 93-A rHDL probably represents an intermediate in the formation of a mature spherical rHDL. The synthesis of the 93-A rHDL involves adding exogenous heat-inactivated LDL as a source of free cholesterol, LCAT, and BSA to an incubation mixture at 37 "C. It is possible that the absence of lipid transfer proteins and other sources of cholesterol prevent the full maturation of the rHDL.
The fluorescence spectroscopy results indicate no significant differences in the environment and the mobility of the tryptophan residues of :he rHDL and native HDL subfractions including the 96-A LpAI/AII particle. Using Trp fluorescence quenching experiments with iodine, Talussot and Ponsin (42) also did not observe any differences between rHDL containing apoA-I and rHDL containing apoA-I and apoA-11. The circular dichroism spectra in the far-ultraviolet region showed ellipticity minima at 222 and 208 nm and maxima in the 190-195-nm region indicati9g that a-helix is the major secondary structure of the 93-A rHDL and the native HDL subfractions. The content of a-helical secondary structure did not differ qarkedly (61-65%) for the HDL particles except for the 96-A LpAI/AII particle which had an a-helix content of 52%. The lower percentage of a-helix in the 96-A LpAI/AII particle is the result of the equimolar amounts of apoA-I1 and apoA-I in the particle. As the ahelicity of apoA-I1 complexed with phospholipid has been established at about 40% (38), we can estimate from the total a-helicity of the 96-A LpAI/AII that the a-helix content of apoA-I at about 60%. Thus, the presence of apoA-I1 seems not to alter the secondary structure of apoA-I in the native 96-8, LpAI/AII.
In order to examine the structural stability of the apolipoproteins in the native HDL subfractions, the denaturation by GdnHCl was assessed by the change in the wavelength of maximum fluorescence and the ellipticity at 222 nm. The wavelength of maximum fluorescence monitors the NHz-terminal region of the apoA-I (up to residue 108) which contains the 4 Trp residues. Since apoA-I1 does not contain any Trp residues, only the structucal changes of the apoA-I molecules are monitored in the 96-A LpAI/AII particle in the fluorescence experiment. The ellipticity change at 222 nm follows the loss in secondary structure upon denaturation. The secondary structure represented mainly by a-helical segments is thought to start at residue 44 and to extend to the carboxyl terminus of the apoA-I. The apoA-I structure as previously described (12)(13)(14) and as shown in Figs. 3 and 4 is clearly stabilized by lipids while free apoA-I completely denatures by 2 M GdnHCl. Both Fags. 3 and 4 show a general pattern in the stability of the 93-A rHDL and the native HDL subclasses; the !8-A LpAI particles are tke least stable followed by the 106-A LpAI particles. The 96-A LpAI/AII particles are more resistant to qdnHC1 than the LpAI subfractions but less so than the 93-A rHDL. Possibly the 93-A rHDL are the most stable particles because of their higher content of surface lipids, and the absence of polyunsaturated lipids which may decrease the hydrophobic interactions within the particles. The apoA-11, even though it does not seem to alter the conformation of the apoA-I on the 96-A LpAI/AII particle, may stabilize the apoA-I conformation and the particles in general. Nichols et al. (43) have reported that ultracentrifugation of HDL in GdnHCl causes the dissociation of apoA-I alone between 2 and 3 M GdnHCl and the dissociation of apoA-I together with apoA-I1 between 5 and 6 M GdnHC1. Perhaps these separate pools of apoA-I represented LpAI and LpAI/AII and their relative stability due to apoA-11. In fact, Cheung and Wolf (32) investigating the stability of LpAI and LpAI/AII to ultracentrifugation found that the LpAI fraction lost more protein than the LpAI/AII fraction and changed the relative proportions of the different size subclasses. The LpAI/AII remained essentbally unchanged. These reports and our results with the 96-A LpAI/AII particles support the hypothesis that the LpAI/AII particles are more stable than the LpAI subclasses.
The curves in Figs. 3 and 4 suggest a multiphasic denaturation. Clearly, the secondary structure is lost before the apoA-I is completely denatured according to the fluorescence wavelength experiments. The results show a similar pattern of denaturation for apoA-I in all of the particles suggesting once again a similar protein structure. The first step in the denaturation curves shown in Fig. 3 may represent the exposure of 1 Trp residue in a rather unstable region of the protein.
We propose that this region involves Trp-108 which is located in one of the a-helical segments of apoA-I. Recent work in our laboratory on the denaturation behavior of the Lys-107 deletion mutant of apoA-I in discoidal rHDL complexes indicated that this section of apoA-I is structurally flexible, and is readily denatured (44,45). The next denaturation step in Fig. 3 from about 2.5 to 4.5 M GdnHC1, is still accompanied by changes in secondary structure and represents a wavelength change that could account for the exposure of the 2 Trp residues to solvent; therefore, we speculate that the Trp-50 and Trp-72 residues may be involved, since they are located in putative a-helical segments. Finally, we propose that the Trp-8 residue is exposed only at GdnHCl concentrations above 5 M. At this denaturant concentration all secondary structure is lost, yet one of the Trp residues is still protected, suggesting that the possible candidate is Trp-8 in a region of apoA-I which has no predicted a-helical structure. Furthermore, human apoA-IV, which has a linear sequence very homologous to apoA-I (46) and forms comparable rHDL particles to apoA-I (47), has a single Trp at position 12 which behaves in the presence of GdnHCl very much like the most stable Trp in apoA-I. While the secondary structure of apoA-IV in rHDL complexes is lost with 5.5 M GdnHCl, the single Trp residue in the NH2 terminus of the molecule is not yet exposed to solvent (47). These results strongly suggest a very stable NH2-terminal domain ine apoA-IV rHDL complexes. Similarly it is likely that the 93-A rHDL and the native HDL subclasses contain apoA-I with a stable NH2-terminal domain.
The near-ultraviolet circular dichroism spectra of the HDL particles and free apoA-I are very similar to the spectra measured by Lux et al. (37). The spectra of the different HDL particles display maxima at 284 and 291 nm and a minimum a t 296 nm (see Fig. 5). On the basis of their location and characteristic spacing, these maxima have been assigned to 1 or more of the Trp residues of apoA-I and correspond to two different vibrational states of the ' Lb electronic transition of Trp (37,(48)(49)(50). From studies with Trp model compounds, the minimum at 296 nm in HDL has been attributed to the 'La transition of Trp (50). The absence of 284-and 291-nm maxima (37,51) in the free and in the lipid-bound apoA-11, which contains no Trp residues, confirms their assignment to the Trp residues. In free apoA-I, these two peaks are reversed in sign and red-shifted to 287 and 293 nm, showing that the Trp residues responsible for the spectra have a different average conformation and are in a more polar environment.
Tryptophans, tyrosines (37,(48)(49)(50), and disulfides (52) may all contribute to the circular dichroism signal in the 260-280nm region. However, since Lux et al. (37) have shown that an increase in p H from 9.5 to 12.6 is accompanied by major changes of the ellipticity between 260 and 280 nm, we suggest that there is an important contribution of the tyrosine residues to the ellipticity in this region. Lux et al. (37) showed that the relipidation of HDL apolipoproteins with phosphatidylcholine alone was able to restore the 284-and 291-nm maxima which were inverted in the free apolipoprotein. The addition of cholesteryl esters intensified these bands and was required to reproduce the broad negative band between 260 and 280 nm present in native HDL. It is of interest to note that the intensity of the maxima at 284 nm for the HDL particles (Fig. 5) seem to be proportional to the total phospholipid content ( Table I). The presence, in the total HDL and in the 96-A LpAI/AII, of apoA-I1 containing 4 tyrosines per monomer contributes to the strong negative ellipticity observed at 269-and 272 nm, respectively. Thg addition of one part of the %-A0 LpAI, one part of the 106-A LpAI, and two parts of the 96-A LpAI/AII spectra between 250 and 315 nm approximately simulates the HDL spectrum, including the strong minima from 262 to 272 nm. The pear-ultraviolet circular dichroism spectra of the 88-and 106-A LpAI subfractions are very similar and indicate that the tertiary structure of the apoA-I in both these subclasses of LpAI is essentially identical. The section of the spectrum in Fig. 5 from 280 to 320 nm, due to the electronic tran$tions of the apoA-I Trp resigues, is also similar for the 96-A LpAI/AII as well as the 93-A rHDL. The shape, sign, characteristic wavelengths, and spacing of the spectrum for the 93-A rHDL particles is identical to the native LpAJ spectra; only the intensity of the 284nm band for the 93-A rHDL is somewhat higher probably because the rHDL contains double theo amount of phospholipid per apoA-I compared to the 88-A LpAI particle. Our results suggest very similar conformations of apoA-I in all of these native subfractions and the 93-A rHDL, including the same number of a-helical segments.
Finally, the LCAT reactivity studies indicate that the 93-A rHDL and the native HDL subfrastions are all poor substrates for LCAT compared to the 96-A rHDL discs which are at least 30-fold more reactive (14). Overall this low reactivity may be due to product inhibition by the cholesterol esters in these particles and/or to an unfavorable apoA-I structure for the activation of the LCAT reactioa. The 106-A LpAI is somewhat less reactive than the 88-A LpAI particle, which agrees with, but does not explain, the differential reactivity of HDL, uersus HDL, with LCAT (53). At this point we have 90 explanation for the marginally higher reactivity of the 96-A LpAI/AII particles.
From the above results the major differences between the 93-A rHDL and each of the native HDL s u b c l p e s is their phospholipid composition. Nevertheless, the 93-A rHDL may still be a good model for native HDL since the apoA-I structure appears to be quite similar for all of the particles studied. Thq major difference found is the increased stability of the 96-A LpAI/AII subfractions as compared to both LpAI subfractions. This property may underlie different functions of LpAI and LpAI/AII in the circulation and may be responsible for their different catabolism.