Cholesterol Efflux Efficiency of Reconstituted HDL Is Affected by Nanoparticle Lipid Composition

Cardiovascular disease (CVD), the leading cause of mortality worldwide is primarily caused by atherosclerosis, which is promoted by the accumulation of low-density lipoproteins into the intima of large arteries. Multiple nanoparticles mimicking natural HDL (rHDL) have been designed to remove cholesterol excess in CVD therapy. The goal of this investigation was to assess the cholesterol efflux efficiency of rHDLs with different lipid compositions, mimicking different maturation stages of high-density lipoproteins (HDLs) occurring in vivo. Methods: the cholesterol efflux activity of soybean PC (Soy-PC), 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), DPPC:Chol:1-palmitoyl-2-hydroxy-sn-glycero-3-phosphocholine (LysoPC) and DPPC:18:2 cholesteryl ester (CE):LysoPC rHDLs was determined in several cell models to investigate the contribution of lipid composition to the effectiveness of cholesterol removal. Results: DPPC rHDLs are the most efficient particles, inducing cholesterol efflux in all cellular models and in all conditions the effect was potentiated when the ABCA1 transporter was upregulated. Conclusions: DPPC rHDLs, which resemble nascent HDL, are the most effective particles in inducing cholesterol efflux due to the higher physical binding affinity of cholesterol to the saturated long-chain-length phospholipids and the favored cholesterol transfer from a highly positively curved bilayer, to an accepting planar bilayer such as DPPC rHDLs. The physicochemical characteristics of rHDLs should be taken into consideration to design more efficient nanoparticles to promote cholesterol efflux.


Introduction
Cardiovascular disease (CVD), the leading cause of mortality in industrially developed countries [1], is primarily caused by atherosclerosis, characterized by an abnormal lipid and inflammatory cell accumulation in the intima, the subendothelial layer of large arteries [2]. Atherosclerosis is associated with atheroma plaque formation and reduction in the vascular diameter, thus increasing the incidence of cardiovascular events [3]. Atherogenesis is initiated and maintained curve with the greatest atheroma regression occurring at a low concentration, while higher concentrations are inefficient in removing cholesterol due to the strong down regulation of the ABCA1 transporter [37].
Finally, the CSL-112 reconstituted HDLs arose as an improvement upon their predecessor, CSL-111. CSL-111 initially showed a potential therapeutic effect [38], but was disfavored due to its hepatotoxicity. On the contrary, CSL-112 was well tolerated and not associated with any significant alterations in liver or kidney function [31]. Moreover, CSL-112 has been found to enhance cholesterol efflux very efficiently. However, its beneficial potential in reducing major adverse cardiovascular events in a group of high-risk patients will be assessed in the on-going large phase III AEGIS-II study (NCT03473223).
Protein purity was first tested by SDS-PAGE and then dialyzed three times over 24 h, against 20 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM benzamidine hydrochloride, pH 8.0 buffer. Finally, protein aggregates were removed and 10% glycerol was added for protein storage. Concentration of apoA-I was determined from its extinction coefficient at 280 nm, 32,430 M −1 ·cm −1 .

Human ApoA-I HDL Reconstitution and Purification
Purified apoA-I was incubated with different lipids in a molar ratio of apoA-I to lipids of 1:125 (mol/mol). The lipid compositions used were Soy-PC, DPPC, DPPC:Chol:lysoPC (85:10:5% mol) and DPPC:CE:lysoPC (75:20:5% mol). First, lipid mixtures in chloroform:methanol (2:1 v/v) were dried with a stream of nitrogen followed by vacuum drying for 1.5-2 h. Lipids were then resuspended in TEN buffer (10 mM Tris-HCl, 1 mM EDTA, 150 mM NaCl, pH 8.0) at 42 • C and sodium cholate was added in a molar ratio of 1:1.2 (total lipid:cholate). Next, apoA-I was added and the resulting solution was incubated above lipid transition temperature (T m ) for 15 min and mixed vigorously every 5 min. The samples were then incubated overnight in agitation at the same temperature and subsequently dialyzed at 42 • C during 48 h against TEN buffer to remove the cholate.
rHDL purification was performed on a Superdex 200 10/300 GL (GE Healthcare) with TEN buffer as eluent. The column was first calibrated with molecular size standards from Amersham Biosciences. Samples were eluted at 4 • C at a 0.2 mL/min flow rate and elution profiles were expressed as retention volume.

Circular Dichroism
The secondary structure of apoA-I was analyzed in a thermostatized Jasco 810 spectropolarimeter in a 0.1 cm path length quartz cuvette. Spectra (200 to 260 nm) were obtained at 25 • C with a bandwidth of 1 nm, response time 1 s and 50 nm/scan speed. Each spectrum represents the average of 15 accumulations and was corrected subtracting the buffer spectra. The α-helicity of the protein for each rHDL composition was calculated using the mean residue ellipticity at 222 nm using the following equation: % α-helix = ((θ) 222 + 3000)/(36000 + 3000) × 100 [46].

Dynamic Light Scattering (DLS)
The size of the rHDL was determined by dynamic light scattering in a Nano-S Zetasizer (Malvern Instruments, Malvern, UK) as previously described [47]. Measurements were performed in triplicate (15 runs) at 25 • C. Viscosity and refractive index of TEN buffer were applied to the measurements.

Negative Stain Electron Microscopy (NS-EM)
Size and morphology of the rHDL were determined by adsorbing 10 µL of each rHDL preparation on a glow-discharged thin carbon-coated 300-mesh copper grid (Cu-300CN; Pacific Grid-Tech, San Francisco, CA, USA) as previously described [48]. The excess of solution was removed and the grid was washed three times in deionized water. Then, one drop (~30 µL) of 1% (w/v) uranyl acetate (UA) (pH 4.6) solution was applied and maintained for 1-3 min in the dark before excess stain was removed. Finally, the excess of solution was removed and the sample was air dried at room temperature.
Feret diameter was used to determine particle size by automatically selecting individual particles and then checked to remove overlapping or damaged particles. 1600 particle images from micrographs of each rHDL sample were used for the statistical analysis of particle size distribution.

rHDLs Transition Temperature: Steady State Fluorescence Measurements
Transition temperature of rHDL-lipid moiety was assessed by fluorescence anisotropy, which allows one to follow changes in the order of lipid bilayers taking advantage of the differences in the fluorescence polarization caused by orientation changes of a fluorophore in space. Briefly, the basis of the technique relies on the dependence of absorption and the emission of light on the orientation of the transition dipole moments. Hence, excitation with vertical polarized light can provide information on the rotational motion of a fluorophore because the emitted light will retain some of that polarization based on how fast it is rotating in solution. The extent of depolarization of the emission of a fluorophore in a lipid membrane reflects the degree to which a population of photoselected excited fluorophores loses its initial selective orientation and becomes randomized. Several studies revealed that anisotropy is mainly determined by the degree to which the fluorophore rotations are restricted by the molecular packing of the lipids [49,50].
The order of lipid bilayers in the different rHDLs was analyzed by measuring fluorescence anisotropy of diphenylhexatriene (DPH) [51]. DPH is a fluorescent probe that locates at the lipid-water interface in lipid bilayers and its fluorescence anisotropy is highly dependent on the lipid phase state, which decreases abruptly during the thermotropic gel-fluid phase transition.
DPH in methanol was added to nanoparticles in a final molar ratio of 1:75 (probe to lipid) and the mixture was incubated 1 h in agitation at 25 • C to incorporate DPH into the lipid bilayer of the nanoparticles.
Fluorescence anisotropy was determined as: where I VV and I VH are the intensities of vertically and horizontally polarized fluorescent light, respectively, when the excitation light is vertically polarized. G represents the compensating factor for the anisotropy sensitivity of the instrument, which is expressed as follows: where I HV and I HH refer to vertically and horizontally polarized light intensities, respectively, when excitation light is horizontally polarized.

Isolation of Human Plasma HDL and LDL
Human HDLs and LDLs were isolated from the human plasma of healthy individuals by ultracentrifugation as previously described [52]. Briefly, blood was collected in EDTA tubes and samples were centrifuged for 10 min at 3000× g and at 4 • C. Then, plasma density was adjusted to 1.4 g/mL using KBr and layered with cold PBS buffer, pH 7.4 to obtain two phases. Finally, samples were centrifuged at 27,000 rpm for 22 h at 4 • C and the bands corresponding to HDLs and LDLs were recovered carefully and stored at 4 • C until use.

LDL Acetylation
Human LDL acetylation was achieved by mixing 1 mL of concentrated LDLs in PBS with 1 mL of saturated sodium acetate solution under constant stirring at 4 • C, as previously described [53]. Little aliquots of acetic anhydride were added during an hour to get a 40 molar excess of acetic anhydride to lysine content of LDLs. The LDL solution was stirred for an additional 30 min and then dialysed with 12 L of PBS containing 0.3 mM EDTA, pH 7.4 for 24 h. Acetylation was confirmed by agarose gel electrophoresis.
J774A.1 macrophages were cultured in Dulbecco's modified Eagles Medium (DMEM, low glucose) supplemented with 10% FBS (v/v), 100 µg/mL streptomycin, 100 U/mL penicillin and MycoZap TM Prophylactic. This cell line was differentiated into foam cells after incubation with acetylated LDLs (LDLac). Briefly, J774A.1 cells were plated (10 5 cells/well) in a 24 well-plate and 24 h later, 125 µg/mL LDLac was added to each well and the cultures were maintained for an additional 24 h.
Human vascular smooth muscle cells (VSMCs) were isolated from carotid arterial atherosclerotic tissue samples. Carotid atheroma plaque samples were obtained by carotid endarterectomy. Samples were placed on ice and processed immediately. An enzymatic tissue digestion method was used to isolate and culture VSMCs from atherosclerotic tissue samples in two consecutive digestions.
First, tissue was digested for 3 h at 5% CO 2 and 37 • C with 300 U/mL of (Collagenase type I) followed by a second overnight digestion with 220 U/mL of ColI at 5% CO 2 and 37 • C. Digested tissue was filtered by a 100 µm nylon Falcon™ Cell Strainer (CLS431752-50EA, Sigma-Aldrich, St Louis, MO, USA) to remove undigested tissue and then, cells were plated in selective medium (2 ng/mL FGFb, 20 ng/mL IGF-1 and 0.5 ng/mL EGF, 5 ng/mL Heparin, 5% NCS, 0.2 µg/mL BSA, 2 mM l-glutamin, 100 µg/mL streptomycin and 100 U/mL penicillin in Gibco's Medium 231), which promotes selective VSMC growth. All the experiments were carried out with cells in passage zero at 70% confluence to reach a situation as close as possible to the real one.
When appropriate, cells were treated with T090, an LXR agonist, which induces ABCA1 expression. This study was approved by the local ethical committee (Ethical Committee of Clinical Research, Basurto University Hospital. Project identification code: PI2018015; approval date: 3 November 2019; name of the committee: Basque Country Research Ethics Committee (CEIm-E)). All carotid atheroma plaques were collected from patients who had signed written informed consent. This research was performed in agreement with the principles outlined in the Declaration of Helsinki.

Cholesterol Efflux Assay
Functionality of rHDLs was analyzed through their capacity to induce cholesterol efflux from cells loaded with TopFluor-Cholesterol as described before [54]. Labelling medium was prepared by complexing a mixture of cholesterol and TopFluor-cholesterol (3:1, M:M) with β-cyclodextrin (16 mM). Once mixed, cholesterol was dried and mixed with an 80-molar excess of β-cyclodextrin dissolved in Minimum Essential Medium Eagle (MEM)-Hepes 25 mM media; pH 7.4. Finally, the mixture was sonicated at 40 • C in a water bath for 30 min to re-suspend cholesterol and further incubated for 3 h at 37 • C in agitation.
After resting time, rHDLs were added to the cells in MEM-Hepes 25 mM (pH 7.4) containing 2 µg/mL ACAT inhibitor. The doses of rHDLs were defined by the quantity of human apoA-I present in the infusion. After 6 h incubation, media and cells were collected to assess cholesterol efflux. As an internal control of the experiment, FBS 20% and BSA 10 µg/mL were added to TopFluor-Cholesterol loaded cells to obtain maximum and non-specific efflux, respectively, in the absence of rHDL. Finally, media were collected and centrifuged for 15 min at 3800 rpm to remove cell debris. Cell monolayers were washed gently with MEM-Hepes 25 mM and solubilized with lysis buffer (50 mM Tris-HCl, pH 7.5, 0.1% SDS, 0.1% deoxycholic acid, 0.1 mM EDTA, 0.1 mM EGTA, 1% NP-40, 5.3 mM NaF and 1.5 mM NaP) during 30 min in a plate shaker at room temperature. Fluorescence of both culture media and cell lysates was measured using a Synergy™ HTX Multi-Mode microplate reader (λex: 485 ± 20 nm, λem: 528 ± 20 nm) and fluorescence intensities (FI) were used to calculate cholesterol efflux in each condition as follows: cholesterol e f f lux % = culture media FI culture media FI + cell lysate FI × 100 Culture media FI was obtained by subtracting the fluorescence intensity of the media with no acceptors. Specific efflux of each acceptor was calculated by subtracting the efflux to BSA.

Statistical Analysis
All measurements were performed at least 3 times, unless otherwise stated, and results are presented as mean ± SD. A Shapiro-Wilk test was performed to confirm that the data were normally distributed. The null hypothesis was verified, indicating that the data were normally distributed. As the intention was to compare HDL with each rHDL composition or DPPC rHDL with other rHDLs, that is, comparison of 2 variables, Student t-tests were employed for analysis. A 2-tailed Student's t test with a significance level of 0.05 was used to test differences in cholesterol efflux efficiency. All statistical analyzes were performed with the SPSS 25 (SPSS, Inc., Chicago, IL, USA).

Development and Biophysical Characterization of rHDL
HDL were reconstituted with different phospholipid mixtures (Soy-PC, DPPC, DPPC:Chol:lysoPC (85:10:5 mol%) and DPPC:CE:lysoPC (75:20:5 mol%) as indicated in the Materials and Methods section. The reconstitution ratio of apoA-I:lipid was optimized to 1:125. The rHDLs, aggregates, and free apoA-I were detected and separated by size exclusion chromatography on a Superdex 200 column as shown in Figure 1A. When applying the rHDLs samples, the aggregates were present in the void volume of the size exclusion column at 7-9 mL, and a rHDL homogenous peak was centered at 11-13 mL, preceding free apoA-I at 15 mL ( Figure 1A).
DLS was used to characterize rHDL size (hydrodynamic diameter) and homogeneity. The size distribution of nanodiscs indicated that rHDLs have an average diameter of~10 nm ( Figure 1B). We next evaluated, by circular dicroism (CD) measurements, α-helical structure in the purified rHDL and apoA-I ( Figure 1C). The higher α-helical content of rHDL shown by rHDL compared to free apoA-I (≈2.2-2.5 times, Table 1) indicates a correct protein conformation and well-structured protein within the nanodisc ( Figure 1C).
Negative stain electron microscopy (NS-EM) was also used to qualitatively examine homogeneity of the rHDLs and to measure particle diameter ( Figure 1D,E). The peak population of the selected 1600 particles was in the diameter range of 8-10 nm, confirming the values obtained by DLS ( Figure 1D). The determined diameters for the rHDLs were: DPPC 9.0 ± 1.6, DPPC:Chol:lysoPC 9.2 ± 2.4, DPPC:CE:lysoPC 10.8 ± 2.2 and Soy-PC 8.7 ± 2.3. Mostly, all nanodiscs appeared as single particles oriented randomly on the staining grid ( Figure 1E). The characteristic stacked nanoparticles were also observed by NS-EM but they appeared in a non-significant number. rHDL morphology was approximately circular, consistent with a discoidal shape ( Figure 1E).
Negative stain electron microscopy (NS-EM) was also used to qualitatively examine homogeneity of the rHDLs and to measure particle diameter ( Figure 1D,E). The peak population of the selected 1600 particles was in the diameter range of 8-10 nm, confirming the values obtained by DLS ( Figure 1D). The determined diameters for the rHDLs were: DPPC 9.0 ± 1.6, DPPC:Chol:lysoPC 9.2 ± 2.4, DPPC:CE:lysoPC 10.8 ± 2.2 and Soy-PC 8.7 ± 2.3. Mostly, all nanodiscs appeared as single particles oriented randomly on the staining grid ( Figure 1E). The characteristic stacked nanoparticles were also observed by NS-EM but they appeared in a non-significant number. rHDL morphology was approximately circular, consistent with a discoidal shape ( Figure 1E).  Transition temperature of the rHDLs lipid moiety was assessed by steady state fluorescence anisotropy using DPH, which localizes to the hydrocarbon core of the lipid bilayer [56]. The temperature-dependent fluorescence anisotropy changes of DPH allows determining phase transition temperature of the different lipid mixtures in rHDLs [57]. As shown in Figure 2, the phase transition temperature of DPPC rHDL obtained from our measurement is 42.9 ± 0.3 • C, which is similar to the literature value-range of phase transition temperature of DPPC nanodiscs [58]. The addition of Chol/lysoPC or CE:lysoPC to the nanodiscs increases the phase transition temperature by 1.7 and 4.1 • C compared to DPPC alone, respectively ( Figure 2). As shown in the Figure 2 inset, DPPC:Chol:lysoPC and DPPC:CE:lysoPC T m are 44.6 ± 0.6 and 47.0 ± 0.5 • C, respectively. As expected due to its lipid composition, HDL T m was 32.0 ± 0.3 • C in the range of the previously described transition temperature of lipoproteins (27-34 • C) [59]. Fluorescence anisotropy changes of DPH Soy-PC nanodisc were not assessed because they are already at liquid-crystalline state below 0 • C.  Transition temperature of the rHDLs lipid moiety was assessed by steady state fluorescence anisotropy using DPH, which localizes to the hydrocarbon core of the lipid bilayer [56]. The temperature-dependent fluorescence anisotropy changes of DPH allows determining phase transition temperature of the different lipid mixtures in rHDLs [57]. As shown in Figure 2, the phase transition temperature of DPPC rHDL obtained from our measurement is 42.9 ± 0.3 °C, which is similar to the literature value-range of phase transition temperature of DPPC nanodiscs [58]. The addition of Chol/lysoPC or CE:lysoPC to the nanodiscs increases the phase transition temperature by 1.7 and 4.1 °C compared to DPPC alone, respectively ( Figure 2). As shown in the Figure 2 inset, DPPC:Chol:lysoPC and DPPC:CE:lysoPC Tm are 44.6 ± 0.6 and 47.0 ± 0.5 °C, respectively. As expected due to its lipid composition, HDL Tm was 32.0 ± 0.3 °C in the range of the previously described transition temperature of lipoproteins (27-34 °C) [59]. Fluorescence anisotropy changes of DPH Soy-PC nanodisc were not assessed because they are already at liquid-crystalline state below 0 °C.

Cholesterol Efflux Promoted in Human and Murine Macrophages
The effect of rHDL lipid composition on promoting cholesterol efflux was assessed both in All measurements were carried out in TEN buffer pH 8. Concentration of HDL and rHDL was kept constant at apoA-I 2 µM for all measurements. Data points shown are means ± S.D. of at least three independent measurements. DPH: 1,6-diphenyl-1,3,5-hexatriene.

Cholesterol Efflux Promoted in Human and Murine Macrophages
The effect of rHDL lipid composition on promoting cholesterol efflux was assessed both in human THP-1 and murine J774A.1 macrophages, and in human VSMC-derived foam cells (Figures 3-5, respectively). Cells were loaded with TopFluor-cholesterol and cholesterol efflux was determined following incubation with rHDL of different lipid compositions. human HDL; in fact, DPPC rHDL particles were a 51% more efficient than HDLs and, DPPC:Chol:lysoPC rHDLs showed a 34% increased efficiency compared to HDLs. In contrast, DPPC:CE:lysoPC and Soy-PC rHDLs showed a similar cholesterol efflux compared to HDLs. Similar results, but to a lesser extent, were obtained in J774A.1 macrophages. As shown in Figure  3B, DPPC and DPPC:Chol:lysoPC rHDLs induced a significantly higher cholesterol efflux from the cells than those incubated with human HDL; in this case, DPPC rHDL particles were 35% more efficient than HDLs and, DPPC:Chol:lysoPC rHDLs showed increased efficiency by 24% compared to HDLs. In contrast, DPPC:CE:lysoPC and Soy-PC rHDLs showed similar cholesterol efflux compared to HDLs.
Cholesterol efflux induced by rHDLs in J774A.1 macrophage-derived foam cells showed similar results to those determined in THP-1 and J774A.1 cells ( Figure 4A). Incubation with DPPC rHDL induced a significantly higher cholesterol efflux when compared with HDL. Although the cholesterol efflux induced by HDL was lower when compared to non-foam J774A.1 cells, the efficiency of DPPC nanodiscs resulted in a higher efficiency when compared to that determined with DPPC rHDL in non-foam cells (57% vs. 35%, respectively). DPPC:Chol:lysoPC, DPPC:CE:lysoPC also showed a higher cholesterol efflux compared to HDL (≈20%, ≈25%, respectively) while Soy-PC rHDLs showed a similar cholesterol efflux than that determined for HDL. Next, the effect of ABCA1 overexpression on cholesterol efflux induced by rHDLs was examined. Abca1 mRNA expression was stimulated by incubating J774A.1 foam cells with TO90 and cholesterol efflux was determined in similar conditions as before. As shown in Figure 4A As shown in Figure 3A, incubation of THP-1 derived macrophages with DPPC and DPPC:Chol: lysoPC rHDLs showed a significantly higher cholesterol efflux than those incubated with human HDL; in fact, DPPC rHDL particles were a 51% more efficient than HDLs and, DPPC:Chol:lysoPC rHDLs showed a 34% increased efficiency compared to HDLs. In contrast, DPPC:CE:lysoPC and Soy-PC rHDLs showed a similar cholesterol efflux compared to HDLs. Similar results, but to a lesser extent, were obtained in J774A.1 macrophages. As shown in Figure 3B, DPPC and DPPC:Chol:lysoPC rHDLs induced a significantly higher cholesterol efflux from the cells than those incubated with human HDL; in this case, DPPC rHDL particles were 35% more efficient than HDLs and, DPPC:Chol:lysoPC rHDLs showed increased efficiency by 24% compared to HDLs. In contrast, DPPC:CE:lysoPC and Soy-PC rHDLs showed similar cholesterol efflux compared to HDLs. cholesterol efflux was similar to that induced by HDL. As shown in Figure 4B,C, TO90 almost induced twice the upregulation of ABCA1 transporter, indicating that cholesterol efflux induced by DPPC, DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs is efficiently enhanced by upregulating the transporter (Figure 4).  Figure 4B correspond to a representative western blot of n = 3. Levels of significance were determined by a two-tailed Student's t-test. * p < 0.01 compared to HDL and # p < 0.01 compared to DPPC.

Cholesterol Efflux Promoted in Human VSMC-Foam Cells
VSMCs extracted from carotid arterial atherosclerotic tissue samples showing foam cell phenotype obtained from carotid endarterectomy were used to determine the ability of rHDL to induce cholesterol efflux [60]. As shown in Figure 5, upon incubation with rHDLs, only DPPC nanodiscs induced a slight but significant increase in cholesterol efflux (22%) when compared with HDL. DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs showed similar cholesterol efflux to HDL. Upregulation of ABCA1 in VSMC by TO90 increases cholesterol efflux to HDL significantly when compared to nonstimulated cells ( Figure 5). In addition, cholesterol efflux to DPPC rHDL was also significantly increased compared to HDL upon TO90 treatment ( Figure 5). The effect of DPPC:Chol:lysoPC and DPPC:CE:lysoPCrHDLs in TO90 stimulated cells was similar to HDL ( Figure 5).  Figure 4B correspond to a representative western blot of n = 3. Levels of significance were determined by a two-tailed Student's t-test. * p < 0.01 compared to HDL and # p < 0.01 compared to DPPC.

Discussion
Reverse cholesterol transport from peripheral tissues to the liver is a major atheroprotective event, with cholesterol efflux as a rate-limiting step [61,62]. Two principal transporters contribute to this process: ABCA1 and ABCG1 [63]. ABCA1 results in the formation of discoidal HDL, while ABGC1 mediates cholesterol efflux through a diffusion mechanism that increases the pool of active cholesterol available for efflux [64].
Although the exact mechanisms of cholesterol efflux mediated by the ABCA1 transporter are not known, a central feature of cholesterol transfer is apoA-I interaction with ABCA1, which stabilizes the transporter and induces bending of the plasma membrane bilayer. This process creates a high curvature site that allows apoA-I to solubilize lipids by binding to exovesiculated plasma membrane domains [18,65]. Although the structural and physical features of apoA-I variants and mimetic peptides that are important in the formation of HDL-like particles have been previously investigated Figure 5. Effect of HDL, DPPC, DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs on cholesterol efflux in vascular smooth muscle cells (VSMC) derived foam cells stimulated or not with TO90. rHDLs were added to the stimulated and non-stimulated cells in MEM-Hepes 25 mM (pH 7.4) containing 2 µg/mL ACAT inhibitor and incubated during 6 h to promote cholesterol efflux. Cholesterol efflux was calculated as described in Methods. Data represent the means ± S.D. of at least three independent measurements. Levels of significance were determined by a two-tailed Student's t-test. * p < 0.01 compared to HDL and # p < 0.01 compared to DPPC.
Cholesterol efflux induced by rHDLs in J774A.1 macrophage-derived foam cells showed similar results to those determined in THP-1 and J774A.1 cells ( Figure 4A). Incubation with DPPC rHDL induced a significantly higher cholesterol efflux when compared with HDL. Although the cholesterol efflux induced by HDL was lower when compared to non-foam J774A.1 cells, the efficiency of DPPC nanodiscs resulted in a higher efficiency when compared to that determined with DPPC rHDL in non-foam cells (57% vs. 35%, respectively). DPPC:Chol:lysoPC, DPPC:CE:lysoPC also showed a higher cholesterol efflux compared to HDL (≈20%, ≈25%, respectively) while Soy-PC rHDLs showed a similar cholesterol efflux than that determined for HDL. Next, the effect of ABCA1 overexpression on cholesterol efflux induced by rHDLs was examined. Abca1 mRNA expression was stimulated by incubating J774A.1 foam cells with TO90 and cholesterol efflux was determined in similar conditions as before. As shown in Figure 4A, upregulation of the ABCA1 transporter significantly enhanced the cholesterol efflux induced by rHDL, with the cholesterol efflux induced by DPPC being 140% more effective than that induced by HDL. The effect of DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs on cholesterol efflux was 100% higher than HDL. On the other hand, Soy-PC rHDL induced cholesterol efflux was similar to that induced by HDL. As shown in Figure 4B,C, TO90 almost induced twice the upregulation of ABCA1 transporter, indicating that cholesterol efflux induced by DPPC, DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs is efficiently enhanced by upregulating the transporter (Figure 4).

Cholesterol Efflux Promoted in Human VSMC-Foam Cells
VSMCs extracted from carotid arterial atherosclerotic tissue samples showing foam cell phenotype obtained from carotid endarterectomy were used to determine the ability of rHDL to induce cholesterol efflux [60]. As shown in Figure 5, upon incubation with rHDLs, only DPPC nanodiscs induced a slight but significant increase in cholesterol efflux (22%) when compared with HDL. DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs showed similar cholesterol efflux to HDL. Upregulation of ABCA1 in VSMC by TO90 increases cholesterol efflux to HDL significantly when compared to non-stimulated cells ( Figure 5). In addition, cholesterol efflux to DPPC rHDL was also significantly increased compared to HDL upon TO90 treatment ( Figure 5). The effect of DPPC:Chol:lysoPC and DPPC:CE:lysoPCrHDLs in TO90 stimulated cells was similar to HDL ( Figure 5).

Discussion
Reverse cholesterol transport from peripheral tissues to the liver is a major atheroprotective event, with cholesterol efflux as a rate-limiting step [61,62]. Two principal transporters contribute to this process: ABCA1 and ABCG1 [63]. ABCA1 results in the formation of discoidal HDL, while ABGC1 mediates cholesterol efflux through a diffusion mechanism that increases the pool of active cholesterol available for efflux [64].
Although the exact mechanisms of cholesterol efflux mediated by the ABCA1 transporter are not known, a central feature of cholesterol transfer is apoA-I interaction with ABCA1, which stabilizes the transporter and induces bending of the plasma membrane bilayer. This process creates a high curvature site that allows apoA-I to solubilize lipids by binding to exovesiculated plasma membrane domains [18,65]. Although the structural and physical features of apoA-I variants and mimetic peptides that are important in the formation of HDL-like particles have been previously investigated [66][67][68][69], the effect of the lipid content of rHDL has been less well characterized. Therefore, in this study, we sought to explore the effect of the lipid composition of rHDL on cholesterol efflux in macrophages, macrophage-derived foam cells unstimulated or stimulated with TO90 and foam-VSMC unstimulated or stimulated with TO90 to upregulate ABCA1 expression. Efficiency of cholesterol efflux mediated by DPPC, DPPC:Chol:LysoPC or DPPC:CE:LysoPC and Soy-PC rHDLs with similar sizes has been assessed. Here, we have used three lipid compositions resembling different maturation stages of natural HDLs in vivo and Soy-PC, which is the major lipid composition constituent used in rHDL that are being tested in clinical trials [38]. The rationale of this study relies on the information provided by previous studies using rHDLs with different lipid mixtures, which has already indicated that lipid composition plays a significant role in cholesterol efflux from macrophages [70]. According to the results obtained in this work, DPPC rHDLs, mimicking nascent HDL are the most effective particles in inducing cholesterol efflux in all the cellular models used. When compared to HDL-induced cholesterol efflux, DPPC rHDLs were 20-40% more efficient depending on the cell culture tested (Figures 3-5). This effect can be attributed to the homogeneous composition of the DPPC rHDL as it has been previously shown that the phospholipid composition of HDL plays an important role in ABCA1-mediated cholesterol efflux and that enrichment of HDL with PC favors cholesterol efflux, in particular [71,72]. In addition, upregulation of ABCA1 with TO90 favored the cholesterol efflux induced by the nanoparticles, especially in macrophage derived foam cells (Figure 4).
Very interestingly, DPPC rHDLs also resulted in more efficiently favoring cholesterol efflux than Soy-PC, the lipid composition used in CSL-111 rHDLs [41,42] and in the smaller CSL-112 nanoparticles [23] currently being tested in humans. It has been described that rHDLs composed of saturated lipids exhibit greater cholesterol efflux from macrophages in vitro and cholesterol mobilization in vivo [73,74]. The more efficient cholesterol-efflux activity observed here with DPPC rHDL compared to Soy-PC rHDLs, can also be ascribed to the properties given by the different lipid composition of the nanodiscs. The main difference between Soy-PC and DPPC rHDL is that in the former, the existing 80% of PCs consist of a mixture of unsaturated fatty acids (C18:1, C18:2 and C18:3) and lyso-PC at 2.8% while composition in the latter is 100% DPPC. According to previously described data [75,76], the increased cholesterol efflux induced by DPPC rHDLs may be attributed to the higher physical binding affinity to cholesterol of saturated phospholipids compared with Soy-PC, in which the majority of phospholipids are unsaturated. Similarly, and in agreement with this, the different lipid composition of apoA-I Milano rHDL (ETC-216) and ETC-642 rHDL could explain the differences among the cholesterol efflux induced by the particles because they are constituted by POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) or a mixture of DPPC and sphingomyelin, respectively [43,77]. In addition, differences in the protein moiety may also contribute. It has already been shown that saturated long-chain-length phospholipids such as DPPC have higher physical binding affinity to cholesterol than POPC [75]. Additionally, the rigidifying effect of sphingomyelin present in ETC-642 rHDL can modify the physical properties of rHDLs resulting in a lower surface tension that could reduce the cholesterol exchange efficiency between membranes [78].
One of the goals of this work was to study the cholesterol efflux efficiency among rHDL which simulate different maturation stages. To do so, we have compared the effects of DPPC rHDL, with DPPC:Chol:lysoPC and DPPC:CE:lysoPC that can be considered particles in a more mature stage. DPPC:Chol:lysoPC rHDL resembling nascent HDLs having incorporated free cholesterol, and DPPC:CE:lysoPC resembling those in which cholesterol is esterified by the effect of Lecithin-cholesterol acyltransferase (LCAT). As shown in Figures 3-5, DPPC rHDL were the most efficient particles in inducing cholesterol efflux, an effect that may be mediated by the physicochemical characteristics of the nanoparticles. The role of membrane lipid composition in cholesterol exchange between membranes is not well understood, however phospholipids and fatty acyl chains have been shown to influence the rate of cholesterol movement between membranes [79][80][81]. In addition, curvature of the lipid bilayer that is imposed by the overall geometry of lipids shows a physiological significance in cholesterol transfer [82]. Lipid geometry is defined by the ratio between the size of the polar head group and acyl chain saturation. PC is a cylindrical lipid that forms flat monolayers [83]. Conversely, the large head group to acyl chain ratio in lyso-PC confers an inverted conical shape to the lipids, thereby favoring a positive curvature of the membrane by bending the monolayer away from the head groups [84][85][86]. Addition of cholesterol increases the packing of conical lipids (such as lyso-PC), and thus disfavors spontaneous curvature induced by the lysophosphospholipid [87] while the esterification of the 3 hydroxyl group of cholesteryl esters is structurally consistent with substantially increased positive curvature [88]. Recent studies have shown that membrane curvature is an active driving force in many processes involving membrane remodelling and cholesterol exchange [89]. It has been shown that cholesterol transfer is about 10 times faster from donor bilayers with high positive curvatures and when the acceptor bilayer is planar [82,90]. Although the curvature of biological membranes is very low, the bending of the plasma membrane bilayer by ABCA1 creates a high curvature that can facilitate cholesterol transfer [18,65]. The cholesterol transfer will be favored thermodynamically from the high curvature promoted by ABCA1 in the cell membrane to DPPC rHDLs instead of DPPC:Chol:lysoPC and DPPC:CE:lysoPC, because the former are planar bilayers and the latter two have positive curvatures [91].
The transfer of cholesterol between membranes is strongly dependent on temperature and is affected by the lipid composition, suggesting that membrane fluidity strongly influences the transfer rate [79]. In this work we have also assessed of thermotropic phase transition of HDL and rHDLs to compare their fluidity. Attending to the T m of DPPC, DPPC:Chol:lysoPC and DPPC:CE:lysoPC rHDLs, the nanoparticles are more rigid as Chol, and CE are incorporated, DPPC < DPPC:Chol:lysoPC < DPPC:CE:lysoPC, showing a T m increment of 1.7 and 4.1 • C compared to DPPC alone, respectively. This effect could indicate that particles with high transition temperatures could be less efficient in accommodating cholesterol from the plasma membrane due the intrinsic physical characteristics of the rHDL bilayer and that the higher T m of rHDL, the lower the cholesterol efflux rate induced by the rHDL. However, attending to this hypothesis, Soy-PC would be the most efficient particles inducing cholesterol efflux followed by HDL, which show Tm below 0 • C and 32.0 ± 0.3 • C, respectively. The lower capacity of promoting cholesterol efflux shown by HDL and Soy-PC indicates that rather than Tm, lipid composition favoring higher binding affinity of cholesterol (saturated acyl chains) and planar bilayer geometries such as shown by DPPC rHDLs are more favorable to promote cholesterol efflux.

Conclusions
In this work, the effect of the lipid composition of rHDL on cholesterol efflux in several cell models has been characterized to determine optimal parameters to achieve maximal cholesterol efflux rates. Three different lipid mixtures were used, mimicking different maturation stages of natural HDLs in vivo and Soy-PC, which is the lipid composition constituent used in the rHDL in clinical trials. Our results indicate that DPPC rHDLs, which resemble nascent HDL, are the most effective particles inducing cholesterol efflux in all the cellular models used. Among the mechanisms underlying their effects are: (1) the higher physical binding affinity of cholesterol to saturated long-chain-length phospholipids in pure DPPC rHDLs and, (2) geometry of lipids within a lipid bilayer influences the rate of cholesterol movement between membranes, and is favored when the donor bilayer has a high positive curvature and the acceptor bilayer is planar, as occurs with DPPC rHDLs.
In sum, the results presented here indicate that rHDLs with a lipid composition similar to nascent HDLs, are more efficient in promoting cholesterol efflux, and their physical characteristics should be taken into consideration to design more efficient rHDL to be used as a cholesterol efflux promoting nanodisc. In addition to promoting cholesterol efflux from cells and taking advantage of the biophysical features of the nanoparticles used in this study, functionalized rHDL could be used to remove the extracellular accumulation of cholesterol in lesions, thus constituting a potential therapeutical tool to avoid plaque progression.  Acknowledgments: We sincerely thank Haziq Siddiqi (Harvard Medical School) for his critical reading and editing of this manuscript. Technical and human support provided by SGIker (Analytical and High-Resolution Microscopy in Biomedicine Service of the UPV/EHU) and Rocío Alonso for excellent technical assistance are gratefully acknowledged.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.