Spontaneous remodeling of HDL particles at acidic pH enhances their capacity to induce cholesterol efflux from human macrophage foam cells.

HDL particles may enter atherosclerotic lesions having an acidic intimal fluid. Therefore, we investigated whether acidic pH would affect their structural and functional properties. For this purpose, HDL2 and HDL3 subfractions were incubated for various periods of time at different pH values ranging from 5.5 to 7.5, after which their protein and lipid compositions, size, structure, and cholesterol efflux capacity were analyzed. Incubation of either subfraction at acidic pH induced unfolding of apolipoproteins, which was followed by release of lipid-poor apoA-I and ensuing fusion of the HDL particles. The acidic pH-modified HDL particles exhibited an enhanced ability to promote cholesterol efflux from cholesterol-laden primary human macrophages. Importantly, treatment of the acidic pH-modified HDL with the mast cell-derived protease chymase completely depleted the newly generated lipid-poor apoA-I, and prevented the acidic pH-dependent increase in cholesterol efflux. The above-found pH-dependent structural and functional changes were stronger in HDL3 than in HDL2. Spontaneous acidic pH-induced remodeling of mature spherical HDL particles increases HDL-induced cholesterol efflux from macrophage foam cells, and therefore may have atheroprotective effects.

sizes of the HDL-containing supernatants were analyzed using a high-resolution SEC Superose HR6 column connected to the Amersham-Pharmacia (GE Healthcare) AKTA chromatography system. Typically, a 50 µl aliquot was injected into the column and eluted with PBS buffer at a fl ow rate of 0.5 ml/min, and fractions (0.5 ml) were collected for protein and lipid analyses. Kinetics of the elution profi le changes in SEC were analyzed using the Unicorn 5.2 software. The particle size was assessed by a calibration curve ( R 2 = 0.98) using a gel fi ltration calibration kit (GE Healthcare).

Analysis of HDL composition
Proteins were resolved in 12.5% SDS-PAGE under reducing conditions, transferred to nitrocellulose, and immunoblotted either with anti-human apoA-I polyclonal antibody ( 21 ) or with anti-human apoA-II monoclonal antibody ( 22 ). The amounts of total cholesterol were measured using the Amplex Red cholesterol assay kit (Invitrogen) according to the manufacturer's instructions. Phospholipids were measured by a fl uorometric assay ( 23 ) using Amplex Red reagent to enhance its sensitivity. Essentially, in this method, choline-containing phospholipids (PCs) are hydrolyzed by phospholipase D (Sigma P0065) into free choline, which is further oxidized with choline oxidase to yield H 2 O 2 . The formed H 2 O 2 then reacts stoichiometrically with the Amplex Red reagent (10-acetyl-3,7-dihydroxyphenoxazine) in the presence of HRP, and ultimately forms the fl uorescent compound resorufi n. Finally, the fl uorescence was measured with a VICTOR3 multilabel plate reader (Perkin Elmer; Finland) using an excitation wavelength of 544 nm and an emission wavelength of 595 nm.

Analysis of HDL particle size and quantitation of pre ␤ -HDL
HDL samples (8 µg as total protein) were loaded onto selfprepared 4-30% polyacrylamide gradient gels (8.0 cm × 8.0 cm) and run at 125 V under nondenaturing conditions overnight at 4°C to reach equilibrium and then stained with Coomassie blue. HDL particle size was determined based on the use of high-molecular-weight electrophoresis calibration standards as molecular size markers. Particle size was also assessed by negative staining electron microscopy ( 24 ). For this purpose, HDL samples (3 l from HDL stock, 1 mg/ml) were dried on carboncoated grids, after which 3 l of 1% potassium phosphotungstate, pH 7.4, was added and also dried on the grids. The samples were viewed and photographed in a JEOL 1200EX electron microscope at the Institute for Biotechnology, Department of Electron Microscopy, Helsinki, Finland. For the determination of the size distribution of the lipoprotein particles, the diameters of 100 randomly selected particles were measured from the electron micrographs.
Agarose gel electrophoresis (0.6% gel) was carried out using the Paragon electrophoresis system according to the instructions of the manufacturer (Beckman Coulter, Inc.). Proteins were transferred from the agarose gel to the polyvinylidene difl uoride (PVDF) membrane by pressure blotting. ApoA-I was identifi ed using a monoclonal anti-human apoA-I antibody (Abcam; UK), followed by an HRP-conjugated anti-rabbit IgG as the secondary antibody (Dako; Denmark). To further examine the effect of pH on the formation of pre ␤ -HDL, HDL 2 or HDL 3 (0.1-1 mg/ml) was incubated in either 20 mM MES (pH 5.5-6.5), 20 mM PIPES (pH 6.5 or 7.0), or 20 mM Tris (pH 7.5) buffer containing 150 mM NaCl, 2 mM MgCl 2, and 2 mM CaCl 2 at 37°C for different periods of time (0-48 h). The pre ␤ -HDL and ␣ -HDL contents were quantifi ed by two-dimensional crossed immunoelectrophoresis ( 25 ), and the amount of pre ␤ -HDL was expressed as a percentage of the sum of the pre ␤ -and ␣ -mobile areas. lesions may be exposed to acidic pH. On the basis of the fact that HDL particles are sensitive to various physiological perturbations, we hypothesized that the structural and functional properties of the particles are also sensitive to changes in pH of the medium in which they are suspended. Our data show that at acidic pH, HDL particles undergo spontaneous remodeling, with formation of lipid-poor apoA-I displaying pre ␤ mobility and fusion of the ␣ -migrating HDL particles, and that the generated mixture of pre ␤ -HDL and fused ␣ -HDL possesses an enhanced ability to promote cholesterol effl ux from cultured human macrophage foam cells. The enhanced effl ux-inducing capacity was shown to be primarily due to the de novo generation of the pre ␤ -HDL particles.

Isolation of human lipoproteins
LDL (d = 1.019-1.050 g/ml), HDL 2 (d = 1.063-1.125 g/ml), and HDL 3 (d = 1.125-1.210 g/ml) were prepared from freshly isolated plasma of healthy volunteers obtained from the Finnish Red Cross by sequential fl otation ultracentrifugation using KBr for density adjustment ( 14 ). Ultracentrifugation was carried out in a Beckman Optima TM TLX system table top ultracentrifuge using a Beckman fi xed-angle rotor (TLA-100.3) at 541,000 g . This method subfractionates HDL from plasma in only 6 h, and it yielded HDL preparations that were almost totally devoid (<0.7% of total HDL) of the minor pre ␤ -migrating component composed of lipid-poor apoA-I found to be present in the HDL 3 subfraction isolated by the traditional lengthy ultracentrifugation protocol (typically 5% to 8% of total HDL) ( 15 ). The HDL 2 and HDL 3 subfractions were washed by refl otation for 2 h at densities of 1.125 and 1.21 g/ml, respectively. Lipoprotein purity was assessed by size-exclusion chromatography (SEC), agarose gel electrophoresis, and nondenaturing gradient polyacrylamide gel electrophoresis (NDGGE). No detectable activities of LCAT, PLTP, or CETP were present in the HDL 2 or HDL 3 preparations, when determined using assays described previously in detail ( 16,17 ). Lipoprotein stock solutions were dialyzed against PBS, pH 7.4, containing 1 mM EDTA, fi ltered, purged with nitrogen, and then stored at 4°C, and used within 2 weeks, during which no changes in protein or lipid composition or particle size were observed. The amounts of lipoproteins are expressed in terms of their total protein concentrations, which were determined by bicinchoninic acid protein assay kit (Pierce; Rockford, IL) using BSA as the standard. LDL was acetylated by repeated additions of acetic anhydride ( 18 ) ( 19 ). The specifi c activities of the [ 3 H] CE acetyl-LDL preparations ranged from 50 to 100 dpm/ng protein. Lipid-free apoA-I was purifi ed according to a previously published procedure ( 20 ). The results are presented for lipoprotein preparations derived from a single donor, and similar data were also obtained for lipoprotein preparations derived from three other donors.

SEC of acidic pH-treated HDL particles
HDL 2 or HDL 3 (0.1-2 mg protein/ml) was incubated in either 20 mM MES (pH 5.5, 5.75, 6.0, 6.25, or 6.5), 20 mM PIPES (pH 6.5 or 7.0), or 20 mM Tris (pH 7.0 or 7.5) buffer containing 150 mM NaCl, 2 mM MgCl 2 , and 2 mM CaCl 2 for the indicated times. All incubations were carried out at 37°C. After incubation, the samples were centrifuged at 10,000 g for 10 min, and the particle percentage radioactivity in the medium relative to the sum of total radioactivity present in the medium and the cells. Cholesterol effl ux to the incubation medium in the absence of any added cholesterol acceptors was considered as basal effl ux, and was subtracted from the effl ux values obtained in the presence of acceptors.

Statistical analysis
Results are reported as mean ± SD. Statistical signifi cance ( P < 0.05) was determined by two-tailed Student's t -test.

RESULTS
Effect of pH on HDL particle size HDL 3 and HDL 2 were incubated at different pH values (5.5-7.5) for 24 h, after which aliquots of the incubation media were analyzed by SEC. Figure 1A shows that by using this methodology, non-incubated HDL 3 (nHDL 3 ) eluted as a single peak, and that the same applied to HDL 3 preparations incubated for 24 h at pH 7.5, 7.0, or 6.5. In contrast, incubation of HDL 3 at pH 6.0 and below led to substantial changes in the elution profi le of HDL 3 , the major peak (I) having a shorter elution time and a minor peak (II) appearing only slightly earlier than did the isolated and purifi ed lipid-free apoA-I, which served as a standard. Thus, the SEC analysis revealed that when compared with the original Peak I (nHDL 3 ), the particles in the major peak had an increased size, whereas the particles in the newly appeared Peak II were much smaller. Analysis of pH-treated HDL 3 using the highly sensitive two-dimensional crossed-immunoelectrophoresis revealed that incubation at pH 6.5 and below resulted in the formation of pre ␤ -migrating particles and that the lower the pH, the higher was the amount of pre ␤ -HDL formed ( Fig. 1B ). To estimate the sizes of the HDL 3 particles during size conversion, the fractions of the major peak were pooled and analyzed by 4-30% NDGGE. This analysis confi rmed the presence of enlarged HDL particles after incubation at pH 5.5 ( Fig. 1C ). Compared with HDL 3 , similar but less-remarkable effects were observed when HDL 2 was incubated at pH 5.5 (see supplementary Fig. I ). In contrast, LDL, which lacks apoA-I but contains a nonexchangeable and nonreleasable apolipoprotein, apoB-100, was fully resistant to the above-shown acidic pH-induced effects on HDL (see supplementary Fig. IIA ).
We next used negative staining electron microscopy to visualize the particle morphology both before and after incubation of HDL 3 at acidic pH. Before the incubation, HDL 3 appeared as spherical particles with a mean diameter of 8.3 ± 1.2 nm (range 6-11 nm) ( Fig. 1D ). Incubation of HDL 3 at pH 5.5 caused a signifi cant change in the particle size distribution (mean diameter 10.8 ± 2.3 nm; range 7-16 nm), and revealed a remarkable increase in the population of larger particles. Indeed, we found that 60% of the acidic pH-treated particles were within the size range of 9-13 nm. We also observed that large particles with diameters of 12-16 nm, not present in the nHDL 3 , appeared. Taken together, these results show that upon exposure to pH 5.5, of the ultracentrifugally isolated HDL 3 and HDL 2 , Circular dichroism spectroscopy Samples of HDL 3 or LDL (50 g/ml) were analyzed by circular dichroism (CD) before and after their incubation at pH 7.5 or 5.5. The treated samples were placed in 0.1 cm quartz cuvettes, and CD spectra were recorded on a JASCO J-715 spectropolarimeter (Japan Spectroscopic Co.; Tokyo, Japan) in the region of 190-250 nm with a step size of 0.5 nm, scan speed of 50 nm per min, band width of 1 nm, and 1 s response. The cell-holder compartment was thermostatically maintained at 37 ± 0.1°C. For each sample, fi ve spectra were averaged, and blank measurements were subtracted. The kinetics of apolipoprotein unfolding were monitored by recording CD signals at 222 nm. Molar ellipticity ([⍜]) was calculated from the equation: [⍜] = (MRW)* ⍜/10lc, where ⍜ is a measured ellipticity in degrees, l is the cuvette path length (0.1 cm), c is the protein concentration (g/ml), and the mean residue weight (MRW) is obtained from the molecular weight and the number of amino acids. The ␣ -helix contents were calculated from the equation using [⍜] at 222 nm: percent ␣ -helix = [(-[⍜] 222+ 3,000)/(36,000 + 3,000)] * 100 ( 26,27 ).

Proteolysis of acidic pH-modifi ed HDL by chymase
HDL 2 or HDL 3 (1 mg/ml) was incubated in either 20 mM MES (pH 5.5) or 20 mM Tris (pH 7.5) buffer containing 150 mM NaCl, 2 mM MgCl 2 , and 2 mM CaCl 2 at 37°C. After 24 h incubation, samples were dialyzed extensively against 5 mM Tris (pH 7.4) containing 150 mM NaCl, 1 mM EDTA. Aliquots of the dialyzed HDL fractions (1 mg/ml) were then incubated in the absence or presence of 200 BTEE units/ml of recombinant human chymase (kindly provided by Teijin Ltd. Hino, Tokyo, Japan). The enzymatic activity of chymase was measured and 1 U of chymase activity was defi ned, as described by Woodbury et al., ( 28 ). After incubation, the samples were kept on ice, and 50 µl aliquots were analyzed by SEC. Aliquots (200 µl) were withdrawn, and soybean trypsin inhibitor (SBTI; fi nal concentration: 100 µg/ml) was added to the samples to fully inhibit chymase activity ( 15 ). Aliquots of the various chymase-treated preparations and their corresponding nontreated HDL preparations were then added to macrophage foam cell cultures, and their abilities to promote cholesterol effl ux were examined.

Primary monocyte-macrophage cultures
Human monocytes were isolated from buffy coats (Finnish Red Cross Blood Transfusion Center; Helsinki, Finland) by centrifugation in Ficoll-Paque gradient as described ( 29 ). Washed cells were suspended in DMEM supplemented with 100 U/ml penicillin and 100 µg/ml streptomycin, counted, and seeded on 24 well-plates (1.5 million cells per well). After 1 h, nonadherent cells were removed and the medium was replaced with macrophage-serum-free medium (Gibco) supplemented with 1% penicillin-streptomycin and 10 ng/ml of granulocyte macrophage colony-stimulating factor (Biosite; San Diego, CA). The medium was then changed every 2 to 3 days throughout the culture period.

Cholesterol effl ux from macrophage foam cells
The monocyte-derived macrophages were incubated in DMEM (pH 7.4) containing 25 g/ml of [ (7.5 ± 0.2 nm), a fi nding consistent with previous reports on the size of pre ␤ 1-HDL ( 30,31 ). To quantitatively determine the generation of pre ␤ -HDL, the HDL particles were analyzed by two-dimensional crossed-immunoelectrophoresis. Incubation of HDL 3 at pH 5.5 resulted in the formation of pre ␤ -HDL in a time-dependent manner ( Fig. 2C , left). After 6 h incubation at pH 5.5, the fraction of the newly generated pre ␤ -HDL fraction amounted to 11% of the total HDL, and it increased to 27% after 48 h incubation. In contrast, at neutral pH 7.5, even after 48 h incubation, the majority of HDL 3 migrated in the ␣position and only a negligible amount (always <1%) of pre ␤ -HDL was detected ( Fig. 2C ; right). Similar, but lessprofound, data were also found for HDL 2 (see supplementary Fig. IIIA, B ), again revealing that HDL 3 is more susceptible to acidic pH remodeling than is HDL 2 .

Distribution of proteins and lipids in acidic pH-modifi ed HDL
We next examined whether incubation of HDL at pH 5.5 would exert differential effects on HDL protein and lipid the HDL 3 particles in particular undergo remodeling, which consists of two stages: release of smaller apoA-Icontaining particles with pre ␤ -mobility and formation of enlarged ␣ -migrating HDL particles.

Acidic pH-induced HDL remodeling and pre ␤ -HDL generation as a function of time
Next, we investigated the rate of the formation of enlarged HDL particles and the detachment of apoA-I. For this purpose, HDL 3 was incubated at pH 5.5 and aliquots of supernatants at different time points up to 48 h were analyzed by SEC. The formation of enlarged HDL particles (Peak I) and the release of apoA-I with pre ␤ -mobility (Peak II), were both time-dependent ( Fig. 2A, B ). To obtain more-precise information about the time-dependent effect of acidic pH on HDL size, the average sizes of particles in Peaks I and II were calculated using gel fi ltration calibration of proteins with known Stokes diameters. Native HDL 3 particles were 8.8 ± 0.1 nm in diameter, and incubation at pH 5.5 progressively increased their size ( Table 1 ). The size of the particles in Peak II remained unchanged Fig. 1. Effect of pH on HDL 3 size and pre ␤ -HDL formation. A: SEC Profi le of HDL 3 at different pHs. HDL 3 (1 mg/ml) was incubated in either 20 mM MES (pH 5.5-6.5), 20 mM PIPES (pH 6.5 or 7.0), or 20 mM Tris (pH 7.0 or 7.5) for 24 h, after which HDL samples (1 mg/ml) or purifi ed apoA-I (0.5 mg/ml) were centrifuged at 10,000 g for 10 min. Aliquots of the supernatants (50 µl for HDL 3 and 80 µl for apoA-I) were injected into a Superose HR6 column and eluted with PBS buffer (pH 7.4). Main fractions are labeled as Peaks I and II. nHDL 3 = nonincubated HDL 3 . Each profi le is representative of at least two independent experiments. B: Detection of pre ␤ -HDL by two-dimensional crossed immunoelectrophoresis. HDL 3 (1 mg/ml) was incubated at different pH values (5.5-7.5) for 24 h, after which ␣ -HDL and pre ␤ -HDL contents were measured using two-dimensional crossed immunoelectrophoresis. The amounts of pre ␤ -HDL are expressed as a percentage of the sum of the pre ␤ -and ␣ -mobile areas. The left panel shows two-dimensional crossed immunoelectrophoresis, and the right panel shows pre ␤ -HDL formation plotted as a function of pH. C: NDGGE analysis. Fractions in Peak I of HDL 3 at 0 h (nHDL 3 ) and 24 h at pH 5.5 were collected, and aliquots (8 µg) were analyzed by 4-30% NDGGE. D: Size distribution and electron microscopy of HDL 3 at acidic pH. HDL 3 (1 mg/ml) was incubated at pH 5.5 for 24 h, after which samples were negatively stained and photographed in the electron microscope as described in Materials and Methods. Data are representative of two independent experiments. nHDL 3 , the enlarged HDL 3 particles (Peak I) contained less protein and slightly more lipids, whereas the pre ␤ -HDL particles (Peak II) were rich in protein and had lower amounts of lipids ( Table 2 ). Immunoblot analysis (data not shown) revealed the presence of both apoA-I and apoA-II in the enlarged HDL particles (Peak I), whereas only apoA-I was detected in the pre ␤ -HDL particles (Peak II). Assuming that the average molecular mass of PC is 750 Da and that the arithmetical mean of the summated molecular mass of cholesteryl ester and of cholesterol is 550 Da, we could calculate that each apoA-I in the small HDL particles had bound, on average, 12 molecules of PC and 8 molecules of cholesterol. Thus, under acidic conditions, apoA-I was released as lipid-poor apoA-I-PC-cholesterol complexes.

HDL remodeling at acidic pH depends on particle concentration and is irreversible
Next, we examined the effect of HDL concentration on the size conversion of the particles observed at pH 5.5. As the concentration of HDL 3 increased from 0.1 to 2 mg protein/ml, Peak I showed a gradual shift toward shorter elution times, with progressively increased release of distribution. Before incubation, nHDL 3 elution profi les for protein, total cholesterol, and phospholipids fully coincided, and incubation at acidic pH induced similar shifts in all the elution profi les of the main peak (data not shown). However, the small HDL particles (Peak II) appeared to contain only minor amounts of phospholipids and cholesterol. Detailed analysis indicated that of the total protein, 88 ± 4% and 12 ± 0.6% were in Peaks I and II, respectively, and that the corresponding values for total phospholipids were 94 ± 3% and 6.4 ± 0.3%, and for total cholesterol, 95 ± 3% and 5.4 ± 0.3%. Compared with Fig. 2. Time-dependent effect of acidic pH on HDL 3 . A: SEC profi le of HDL 3 at different incubation times. HDL 3 (1 mg/ml) was incubated in 20 mM MES (pH 5.5) for the indicated times. Aliquots of the supernatants (50 µl) were analyzed by SEC after centrifugation, as described in Fig. 1A . Each profi le is representative of three to six independent experiments. B: Kinetics of acidic pH-induced remodeling according to Peak I disappearance and Peak II appearance in panel A. Peak areas were calculated by integration of the 280 nm absorbance using the Unicorn 5.2 software. The data (mean ± SD) are expressed as a percentage of the sum of the 280 nm absorbance of Peak I and Peak II. C: Time-dependent effect of acidic pH on pre ␤ -HDL level. HDL 3 (1 mg/ ml) was incubated for various periods of time at pH 7.5 or pH 5.5, after which ␣ -HDL and pre ␤ -HDL contents were measured using two-dimensional crossed immunoelectrophoresis. The amounts of pre ␤ -HDL are expressed as a percentage of the sum of the pre ␤and ␣ -mobile areas. samples were analyzed for the presence of pre ␤ -HDL and size changes, we found that pre ␤ -HDL particles appeared in the HDL 3 fraction after incubation at pH 5.5 for 30 min ( Fig. 3B , middle ; Fig. 3C ), i.e., before any signifi cant changes in HDL 3 size occurred, which, again, were observed only after 60 min of incubation ( Fig. 3B , bottom). Thus, the data indicated that after acidifi cation, unfolding of HDL apolipoproteins is the fi rst step, followed by apoA-I release and subsequent enlargement of the HDL particles.

HDL remodeling at acidic pH enhances their ability to facilitate cholesterol effl ux from cultured macrophage foam cells
As indicated in the above experiments, incubation of HDL at pH 5.5 resulted in changes of their structural properties. This prompted us to study the ability of acidic pH-modifi ed HDL to induce cholesterol effl ux from cultured macrophage foam cells. HDL 3 and HDL 2 were preincubated at pH 5.5 or pH 7.5 for different periods of time (0-48 h) and were then used as acceptors for cholesterol effl ux by incubating them for 16 h at pH 7.4 with the cholesterol-loaded cells. Interestingly, preincubation of HDL 3 or HDL 2 ( Fig. 4A, B ) at pH 5.5 increased their capacity to induce cholesterol effl ux: the longer the preincubation time of HDL particles at acidic pH, the higher was the rate of cholesterol effl ux induced by the acidic pHtreated HDL. In sharp contrast, preincubation of HDL 3 or HDL 2 at pH 7.5 did not change their ability to induce cholesterol effl ux, when compared with nHDL particles (0 h preincubation). Thus, our data demonstrated that acidic pH induces conversion of HDL to particle subpopulations with higher capacity to induce macrophage cholesterol effl ux.

Chymase treatment abolishes enhanced cholesterol effl ux induced by acidic pH-modifi ed HDL
Previously, we have demonstrated that particles with pre ␤ -mobility are responsible for the ABCA1-dependent cholesterol effl ux from macrophage foam cells induced by human plasma and HDL 3 , and, moreover, that selective proteolytic depletion of the pre ␤ -particles by mast cell chymase results in loss of the ABCA1-dependent cholesterol effl ux induced by plasma or by HDL 3 ( 15,36,37 ). These observations allowed us to test the relative contribution of the pre ␤ -particles to the observed acidic pH-dependent enhancement of cholesterol effl ux by HDL 3 and HDL 2 . Acidic pH-modifi ed HDL 3 and HDL 2 were neutralized by lipid-poor apoA-I in peak II (see supplementary Fig. IVA ), suggesting greater interactions between enlarged HDL particles at higher concentrations. Although Peak II was not detectable by SEC at concentrations below 0.25 mg/ml, two-dimensional crossed-immunoelectrophoresis analysis revealed substantial amounts of pre ␤ -HDL formed (see supplementary Fig. IVB ). Thus, our data indicated that acidic pH-induced HDL remodeling may occur even at the low concentration of HDL typically found in interstitial fl uids ( 32 ).
To examine whether the process of acidic pH-induced HDL remodeling is reversible, HDL subfractions were incubated for 24 h at pH 5.5 to trigger particle remodeling, and then the incubation mixtures were neutralized by extensive dialysis against neutral buffer of pH 7.5. After an additional 24 h incubation at pH 7.5, the acidic pH-induced changes in SEC profi les of HDL 3 and of HDL 2 were still present, demonstrating the irreversible nature of HDL interconversion at acidic pH (data not shown).

Unfolding of HDL proteins is the initiating step, followed by apoA-I release with ensuing enlargement of the apoA-Idefi cient HDL particles
To gain insight into the mechanism by which acidic pH induces remodeling of HDL, we investigated the effect of pH on the conformational characteristics of HDL apolipoproteins. First, far-UV CD spectra were recorded for nHDL and for HDL after incubation at pH 7.5 or pH 5.5 for 24 h. No signifi cant changes in CD spectra within this time period were observed at pH 7.5, when compared with spectra obtained for nHDL 3 . In contrast, CD spectra of HDL 3 at pH 5.5 showed reduced mean residue ellipticity signal intensity ( Fig. 3A ), indicating unfolding of HDL apolipoproteins. To better discern the process of acidic pH-induced HDL remodeling, we monitored the time courses of HDL unfolding, pre ␤ -HDL release, and particle size at early time points. We found that at pH 5.5, the unfolding of apolipoproteins on HDL 3 occurred immediately after acidifi cation of the medium, and that unfolding was already complete when the fi rst measurement was performed, i.e., not later than 1 min after the acidifi cation ( Fig. 3B ,  top). This fi nding well accords with previous reports on the chemical and thermal denaturation of HDL (33)(34)(35). In contrast, no changes in CD spectra were found when LDL was incubated at pH 5.5 (see supplementary Fig. IIB ), indicating that the acidic pH-induced unfolding does not apply to apoB-100. When aliquots of the same HDL 3 ( Fig. 5B ), and also abolished their ability to enhance cholesterol effl ux ( Fig. 5C, D ). Taken together, at acidic pH, HDL particles undergo remodeling into larger particles with formation of lipid-poor apoA-I particles with pre ␤mobility, the latter of which mainly covered the enhanced ability of the acid pH-modifi ed HDL to induce cholesterol effl ux.

DISCUSSION
In the present study, we have characterized the effect of acidic pH on structural and functional properties of ultracentrifugally isolated human HDL 3 and HDL 2 particles. The data indicate that at acidic pH, HDL particles belonging to either subclass undergo remodeling into larger particles with formation of lipid-poor apoA-I with pre ␤mobility. Importantly, when compared with native HDL, the ability of acidic pH-modifi ed HDL to promote cholesterol effl ux from cultured macrophage foam cells was enhanced. To the best of our knowledge, the present study is the fi rst report revealing the effect of acidic pH on both the structural and functional properties of human HDL.
The critical importance of pH on HDL stability became evident as we observed that in contrast to the effect of extensive dialysis at pH 7.4, and incubated in the absence or presence of chymase. SEC data showed that chymase treatment completely depleted the lipid-poor apoA-I particles (Peak II) in acidic pH-treated HDL 3 ( Fig. 5A ) and HDL 2 Fig. 3. Kinetics of acidic pH-induced HDL 3 remodeling. A: Far-UV CD analysis of HDL 3 . HDL 3 (1 mg/ml) was incubated in 20 mM MES (pH 5.5) or 20 mM Tris (pH 7.5) for 24 h. Aliquots of treated HDL 3 and nHDL 3 (non-incubated HDL 3 ) were diluted to 50 µg/ml for CD measurement as described in Materials and Methods. B: Sequence of HDL 3 remodeling at acidic pH. HDL 3 (1 mg/ml) was incubated in 20 mM MES (pH 5.5) for the indicated times. Aliquots of samples were used for CD measurement at 222 nm (top), for pre ␤ -quantitation by two-dimensional crossed immuneoelectrophoresis (middle), and for size by SEC (bottom). C: Detection of pre ␤ -HDL by agarose gel electrophoresis. Aliquots of samples (5 µg) in panel B were also analyzed by agarose gel electrophoresis as described in Materials and Methods. Data are representative of at least two independent experiments. 3 H]cholesterol-loaded macrophage foam cells. After incubation for 16 h with the different cholesterol acceptors, the media were collected and centrifuged to remove cellular debris, and the radioactivity of each supernatant was determined by liquid scintillation counting. Cells were solubilized by 0.2 M NaOH, and radioactivity was determined in the cell lysates. Effl ux of cholesterol was calculated as dpm medium /(dpm cells + dpm medium )×100. Results are representative as mean ± SD of triplicate wells from a donor. * P < 0.05 versus HDL at 0 h; ** P < 0.01 versus HDL at 0 h. Similar data were observed with macrophages derived from three different donors.
particles, contained both apoA-I and apoA-II as their major proteins and they migrated in ␣ -position, while the smaller newly formed pre ␤ -migrating particles contained only apoA-I, revealing selective release of this apolipoprotein from particle surfaces. This fi nding is consistent with the known lower lipophilicity of apoA-I, when compared with that of apoA-II ( 40,41 ).
The acidic pH-mediated HDL remodeling was a timeand concentration-dependent process ( Fig. 2 and supplementary Figs. IIIA, IVA ). The gradual increases in the quantities of both the lipid-poor apoA-I particles and the fused particles ( Fig. 2A , supplementary Figs. IIIA, IVA ) strongly suggest that these two processes are mechanistically connected. Previous reports have shown that exposure of HDL to chemical or thermal modifi cation induces HDL apolipoprotein unfolding and subsequent particle acidic pH, at neutral pH, the ultracentrifugally isolated HDL subfractions were stable and totally resistant to changes in particle size. The size conversion taking place at acidic pH was confi rmed by four different methods, i.e., SEC, NDGGE, agarose gel electrophoresis, and electron microscopy ( Figs. 1 and 2 ). Considering the changes in the size of the acidic pH-modifi ed HDL particles, we could calculate that the formation of larger particles (11-13 nm) had resulted from the fusion of two to three native HDL particles (7-9 nm), consistent with fusion of HDL particles induced by chemical and thermal denaturation ( 38,39 ). Further analysis indicated that about 50% of the particles analyzed individually by negative staining electron microscopy had an increased diameter; i.e., they were fused particles containing the core lipid of two or more native HDL particles. The enlarged particles, like the native HDL Dialyzed HDL 3 or HDL 2, which were treated at pH 5.5, were divided into two identical aliquots and then incubated with and without chymase (200 BTEE U/ml). As a control, dialyzed HDL 3 or HDL 2 , which were incubated at pH 7.5, was also incubated without chymase under the same conditions. After 6 h incubation, aliquots were taken and placed in ice, and aliquots (50 µl) were injected into the Superose HR6 gel fi ltration column for SEC analysis. Separately, aliquots (200 µl) were withdrawn, and SBTI (fi nal concentration: 100 µg/ml) was added to all samples to fully inhibit chymase. Aliquots (fi nal concentration: 25 µg/ml) were added to [ 3 H]cholesterolloaded macrophage foam cells cultured in medium containing SBTI (100 µg/ml). Cholesterol effl ux was measured after 16 h incubation, as described in Fig. 4 . Results are representative as mean ± SD of triplicate wells from a donor. Similar data were observed with macrophages derived from three different donors.
of apoA-I ( 48 ), whereas enrichment of HDL with phospholipids using the detergent dilution method ( 49 ) prevents serum opacity factor-induced apoA-I release ( 8 ). Finally, several human plasma factors (LCAT, PLTP, and CETP) consume or transfer lipids and induce remodeling of HDL (3)(4)(5), which can lead to destabilization of the HDL structure and, in some cases, also to the release of lipid-free/lipid-poor apoA-I and to particle fusion ( 7 ). We were unable to detect any measurable activities of LCAT, PLTP, or CETP in the HDL preparations either before or after incubation at acidic pH, thus excluding any significant involvement of these physiological factors in the observed remodeling of HDLs at acidic pH.
Regarding the functionality of the remodeled HDL, we found that acidic pH produces a progressive increase in the ability of HDL to induce cholesterol effl ux from macrophage foam cells ( Fig. 4A, B ). Similarly, the serum opacity factor-induced remodeling of HDL has been shown to exhibit an enhanced ability to promote cholesterol effl ux ( 50 ). Using mast cell chymase, which selectively and completely depleted pre ␤ -HDL in acidic pH-modifi ed HDL ( Fig. 5A, B ), we found that the gained capacity of the acidic pH-modifi ed HDL to induce cholesterol effl ux was fully abrogated ( Fig. 5C, D ). Thus, the data demonstrate that the pre ␤ -HDL particles generated upon incubation of HDL at acidic pH mainly accounted for the increased cholesterol effl ux capacity.
As suggested by the present fi ndings, in an acidic environment, HDL remodeling may lead to enhanced cholesterol effl ux. Atherosclerotic lesions have been shown to exhibit heterogeneity of extracellular pH, and pH values below 6.0 have been detected in some lipid-rich areas by fl uorescent microscopic imaging ( 13 ). In advanced atherosclerotic lesions, acidic environments develop in the deeper hypoxic areas as a result of local production and secretion of protons and other extracellular fl uid-acidifying agents ( 51 ). Microelectrode measurements of pH in microenvironments of activated cultured macrophages have demonstrated pH values as low as 3.6, and a rapid decrease in pH values, ranging from 0.5 to 2.0 pH units per h after activation of the cells has been observed ( 52 ). Importantly, when activated by oxidized LDL, macrophages can strongly acidify their local environment ( 53 ). Similarly, upon contact with aggregated LDL, macrophages can generate acidic extracellular surface-connected domains in which pH values below 6.5, and even down to pH 5.0, are present ( 54 ).
Based on the above information, we can envision that in the hypoxic areas of atherosclerotic lesions, acidic pHtriggered release of apoA-I from HDL at pH values of less than 6.5 may occur in the immediate vicinity of activated macrophages ( 55 ). Actually, generation of lipid-poor apoA-I in the immediate vicinity of macrophages appears to be optimal because, for the cholesterol effl ux-inducing effect of apoA-I, interaction between apoA-I and macrophage surface lipid transporters (notably ABCA1) is required. However, we need to be cautious when extrapolating the present in vitro fi ndings to in vivo conditions. Atherosclerotic arterial intima possesses a variety of powerful proteolytic fusion with concomitant apoA-I dissociation from the particles ( 35,38,42,43 ). Also, chaotropic detergent perturbation has been shown to transfer apoA-I from HDL particles to the aqueous phase with concomitant formation of larger HDL-like particles ( 38 ). However, no further experimental attempts have been made to defi ne whether apoA-I dissociation precedes fusion under chemically induced and thermally induced HDL remodeling. Importantly, our kinetic studies have provided the fi rst experimental evidence demonstrating that during acidic pH-induced HDL remodeling, the initiating step is the unfolding of HDL apolipoproteins, followed by apoA-I release and subsequent fusion of the apoA-I-defi cient HDL particles ( Fig. 3 ). Given the dominance of apoA-I in mature HDL particles, the changes in the CD spectra at acidic pH are likely to be attributable mainly to conformational changes of apoA-I rather than of other apolipoproteins present in HDL particles ( 44 ). The indispensability of the presence of apoA-I for HDL particle stability well agrees with the concept that HDL fusion can be triggered by an imbalance between the hydrophobic core and the particle surface, particularly when the latter becomes protein-defi cient ( 39 ). Actually, heat-induced and chemically induced HDL fusion ( 38,42,43 ) have been ascribed to dissociation of apoA-I. ApoA-I release has also been proposed to be responsible for PLTP-mediated HDL fusion ( 5,45 ). In contrast, LDL particles, which solely contain a nonreleasable apoB-100 on the particle surface, were totally resistant to acidic pH-dependent particle modifi cation (see supplementary Fig. II ). However, like the HDL particles, fusion of LDL particles has been shown to require reduction in the apolipoprotein content of the particle surface; i.e., in the case of LDL, release of protease-generated apoB-100 peptide fragments from the surface of proteolyzed LDL particles is the fusion-triggering mechanism ( 46 ). Taken together, release of apoA-I with ensuing generation of larger fused HDL particles appears to be a common feature of HDL in response to its physicochemical perturbations and also during its physiological remodeling.
As indicated by SEC and two-dimensional crossedimmunoelectrophoresis ( Fig. 2 and supplementary Fig. III ), the remodeling of HDL 3 is faster than that of HDL 2 at acidic pH, an observation similar to that seen in the remodeling of HDL 2 and HDL 3 by guanidine denaturation ( 38 ) or by the serum opacity factor ( 8 ). The greater susceptibility of HDL 3 to acidic pH remodeling suggests that these particles are less stable than HDL 2 particles. The faster remodeling of HDL 3 particles may simply be due to their smaller sizes, which make them less able to accommodate perturbations ( 38 ). Another susceptibility-increasing factor could be the phospholipid:protein ratio of HDL particles, which is related to particle size and is also known to affect particle stability ( 47,48 ). Indeed, compositional analysis of the HDL subfractions used in this study indicated that the phospholipid:protein ratio in the HDL 3 preparations was only about 67% of that in HDL 2 preparations, i.e., 0.6 and 0.9, respectively. Also, PLA 2 treatment of HDL, which results in decreased levels of particle phospholipids, does enhance the CETP-mediated dissociation enzymes potentially capable of degrading pre ␤ -HDL particles in the extracellular milieu of the tissue, at either neutral or acidic pH ( 10 ). Notably, activated macrophages secrete many powerful proteases with an acidic pH optimum, and it is known that lipid-poor apoA-I particles with pre ␤ -mobility are extremely sensitive to proteolytic degradation ( 37,56 ). Thus, the benefi t of generating lipid-poor apoA-I particles in an acidic extracellular microenvironment generated by activated macrophages may easily get lost. Furthermore, the enhanced cholesterol effl ux induced by acidic pH-modifi ed HDL might be partially counteracted by reduction in the ability of human macrophages to release cholesterol via the ABCA1 pathway in acidic culture conditions ( 27 ). Accordingly, we do not know the ultimate balance, either increase or decrease of ABCA1-dependent cholesterol effl ux in acidic regions present in advanced atherosclerotic lesions. Treating advanced acidic lesions with infusions containing mimics of apoA-I resistant to proteolysis ( 57 ) would be one strategy to support the cholesterol effl ux-inducing potential of spontaneously generated apoA-I in such acidic environments.