Forty-eight patients with diverse cardiovascular risks were recruited for this study. Diverse health conditions may influence the composition and functions of HDL, and differences between HDL are needed for appropriate analysis of our study. Therefore, we included patients with various cardiovascular risk factors, unless they were in uncontrolled status (Supplementary Data 1). The study protocol was approved by the Institutional Review Board of Severance Hospital, Seoul, Korea (number 4-2013-0281). The investigation was conducted according to the principles stated in the Declaration of Helsinki. All participants provided written, informed consent. All patients enrolled in the study were statin-naïve or had discontinued prior lipid lowering agents for at least 3 months. Only patients without history of cardiovascular disease were eligible for wash-out of prior agents. Patients enrolled in this study and had coronary artery disease were all diagnosed for the first time and statin-naïve. Exclusion criteria are described in the Supplementary Data 1. At the time of enrollment, the patients were interviewed to obtain medical histories, followed by laboratory assessment.

The patients were instructed to fast and avoid drinking alcoholic beverages or smoking for ≥12 hours before the collection of blood samples, which were analyzed within 4 hours. Lipid levels were analyzed using an auto-analyzer at a local laboratory certified by the Korean Society of Laboratory Medicine. HDL was isolated by ultracentrifugation as described in the Supplementary Data 1.

Measurement of cholesterol efflux capacity, NO production, and VCAM-1 expression is described in the Supplementary Data 1.

For proteomic analysis, HDL samples were pooled into equal amounts according to the highest and lowest quartiles of cholesterol efflux capacity, NO production, and VCAM-1 levels. Before mass spectral analysis, 80 μg of each pooled HDL sample was separated based on molecular weight in a NuPAGE gel (4–12% Bis-Tris gel; Invitrogen, Carlsbad, CA, USA). Methods for sample preparation and in-gel digestion are described in the Supplementary Data 1.

Ten microliters of each sample were injected into a reversed-phase Magic C₁₈AQ column (15 cm × 75 µm) on an Eksigent MDLC system (Eksigent Technologies, Dublin, CA, USA). The nano HPLC system was coupled to an LTQ XL-Orbitrap mass spectrometer (Thermo Scientific, Waltham, MA, USA). Each liquid chromatography-mass spectrometry (MS)/MS file was analyzed using the SEQUEST algorithm in Proteome Discoverer 1.4 (Thermo Scientific). MS and MS/MS data were searched against the SwissProt human database (August 2016) with 20121. Detailed methods used for mass spectral analysis are available in the Supplementary Data 1.

The relative abundances of the proteins identified in pooled HDL samples showing the highest and lowest quartile values of cholesterol efflux capacity, NO production, and VCAM-levels were calculated based on the normalized spectral abundance factor (NSAF). Spectral count data (number of MS/MS spectra matched to the protein) were first extracted using Proteome Discoverer 1.4 software and exported into MS-EXCEL. The NSAF for each protein was calculated as follows:

in which the total number of MS/MS spectra matching peptides from protein k (SpC) was divided by protein length (L) and then divided by the sum of SpC/L for all N proteins.

For statistical analysis of the dataset, natural logarithm of each NSAF value was calculated, followed by Student's t-test of ln(NSAF) values obtained from the three replicates of the highest and lowest quartile values of each parameter. For the proteins showing relative abundance ratio >2 depending on three functional properties of HDL, Benjamini and Hochberg method-based false discovery ratio (FDR)-adjusted p values were calculated. A p value <0.05 was considered statistically significant. Additional analytical methods are described in the Supplementary Data 1.

The demographic data of the study subjects are presented in Table 1. Of the 48 patients, 34 (71%) were male, with patient ages ranging from 30 to 80 years. The mean high-density lipoprotein-cholesterol (HDL-C) level was 45.7±13.2 mg/dL, whereas the mean low-density lipoprotein-cholesterol (LDL-C) level was 127±32 mg/dL. Patients with the highest quartile value of cholesterol efflux were younger and showed higher total- and LDL-C levels than those with the lowest quartile value. The characteristics and lipid profiles of the patients with the highest and lowest quartile values for NO production and VCAM-1 expression were not significantly different. Test results of each functional parameter according to quartile groups are presented in Figure 1.

For proteomic analysis, 80 μg of each HDL sample pooled according to the highest and lowest quartile values of the functional parameters was separated based on molecular weight. In-gel tryptic digestion of the stained protein bands was performed to identify the proteins. Through mass spectral analysis of the extracted peptides from each pooled HDL sample, 213, 184, and 199 proteins were identified, as shown in the Venn diagrams in Supplementary Figure 1.

Label-free quantification based on the spectral count was performed to identify HDL-associated proteins that were differentially abundant in samples with high functional parameter values. Only proteins with p<0.05 are shown in Figures 2, 3, 4. Table 1 lists the proteins that were highly ranked according to their differential abundance (relative abundance ratios ≥2 or ≤0.5, p value <0.05 from Student's t-test) in HDL with the highest quartile for each parameter. The spectral count and relative abundance ratios of all differentially abundant proteins (p value <0.05 from Student's t-test) are listed in Supplementary Tables 1, 2, 3. The functional categories of the proteins were classified and are presented in Supplementary Tables 4, 5, 6.

Twenty-seven proteins were more abundant and 33 were less abundant in HDL with the highest quartile compared to in HDL with the lowest quartile of cholesterol efflux capacity (Figure 2 and Supplementary Table 1). Fourteen proteins showed relative abundance ratios ≥2. However, when the p values were adjusted by multiple testing, only 6 proteins, thyroxine-binding globulin, alpha-1B-glycoprotein, plasma serine protease inhibitor, vitronectin, angiotensinogen, and serum amyloid A-4 (SAA4), met the criteria. In contrast, 14 proteins showed relative abundance ratios <0.5. After FDR adjustment by Benjamini and Hochberg method, only 3 proteins, complement C4-A, alpha-2-macroglobulin, and immunoglobulin mu chain C region, met the criteria (Table 1). When proteins that were more abundant in HDL with highest cholesterol efflux capacity were classified by functional categories, those belonging to “lipid metabolism” were most common. Among proteins less abundant in the same sample, those belonging to “immunity” were most common (Supplementary Table 4).

Nineteen proteins were significantly more abundant and seven were less abundant in HDL with the highest versus lowest quartile of NO production (Figure 3 and Supplementary Table 2). Four proteins showed relative abundance ratios ≥2. In contrast, no proteins showed relative abundance ratios of <0.5. When the p-values were adjusted by multiple testing, no proteins showed abundance ratios ≥2 or <0.5 (Table 2). The functional categories of proteins showing higher or lower abundance in HDL with high NO production varied considerably (Supplementary Table 5).

Fifteen proteins were significantly more abundant and nine were less abundant in HDL with the highest quartile than in HDL with the lowest quartile of VCAM-1 expression (Figure 4 and Supplementary Table 3). Three proteins showed relative abundance ratios ≥2. Six proteins showed relative abundance ratios <0.5. After adjusting the p values, no proteins showed abundance ratios of ≥2 or <0.5 (Table 1). When the proteins that were more abundant in HDL showing high VCAM-1 expression were classified by functional categories, those belong to “immunity” were most common (Supplementary Table 6).