Subjects were categorized into four groups based on having high or low HDL-C and the presence or absence of IHD (Table 1). Serum from each subject was subjected to dextran sulfate/MgCl₂ precipitation to prepare a serum fraction containing HDL but depleted of LDL and other apoB-containing lipoproteins. Results of liquid chromatography/mass spectrometry (LC-MS-MS) composition analysis of sphingolipids in HDL-containing fractions are summarized in Table 2. The major sphingolipids associated with the HDL-containing fractions were S1P, C24 ceramide and C24:1 ceramide. As compared to total serum (Table 3), the HDL-containing fraction contained ~76% of the S1P, 21% of the C24 ceramide and 20% of the C24:1 ceramide found in total serum.

Statistical analysis of the LC-MS-MS data showed that S1P and DH-S1P in HDL-containing serum fractions were significantly lower (17% and 37% reduction, respectively; p < 0.0001) in individuals with high HDL-C having IHD as compared to individuals with high HDL-C having no evidence of IHD (Figure 1A and 1B). Furthermore, S1P and DH-S1P in HDL-containing fractions were significantly lower (22% and 33% reduction, respectively; p < 0.0001) in individuals with low HDL-C and having IHD as compared to individuals with low HDL-C having no evidence of IHD (Figure 1A and 1B).

The ability to correctly classify patients with and without IHD by using levels of S1P and DH-S1P was assessed by receiver operating characteristic (ROC) analysis (Figure 2). S1P and DH-S1P levels in the HDL-containing fraction were both found to discriminate subjects with IHD from those without IHD better than would be expected by chance alone, as characterized by the area under the curves (0.756 for S1P and 0.773 for DH-S1P) (Figure 2).

Among the other sphingolipids analyzed in the HDL-containing fractions from the CCHS subjects, C24:1-ceramide levels were also significantly lower (p < 0.003) in individuals with IHD as compared to their control groups for both the high and low HDL groups (Figure 1C).

S1P, DH-S1P and C24:1-ceramide levels in the HDL-containing fractions were also assessed relative to the concentration of apolipoprotein A-I (apoA-I) in the samples. The ratios of these sphingolipids to apoA-I were all significantly lower (p ≤ 0.001) in individuals with high HDL-C having IHD as compared to individuals with high HDL-C having no evidence of IHD (Figure 1D-F). Furthermore, these ratios were also significantly lower (p ≤ 0.006) in samples from individuals with low HDL-C and having IHD as compared to individuals with low HDL-C having no evidence of IHD (Figure 1D-F).

To examine the quantitative relationship of the concentration of these lysosphingolipids associated with the HDL-containing fraction and IHD, subjects were grouped as shown in Table 4 by their quartile of [S1P] in the HDL-containing fraction of serum (upper limits for Q1 = 1.21, Q2 = 1.45, Q3 = 1.73, Q4 = 2.38 μM). A Cochran-Armitage test for linear trend was conducted to determine whether an increase in [S1P] corresponded to a decrease in the proportion of subjects with IHD. The result was statistically significant (Z = 6.0887, p-value < 0.0001), indicating that the proportion of subjects with IHD decreases as [S1P] (as defined by the corresponding quartile) increases. The analysis was repeated using quartiles of [DH-S1P] (upper limits for Q1 = 0.075, Q2 = 0.113, Q3 = 0.149, Q4 = 0.244 μM) (Table 5). The result of the Cochran-Armitage test of these data was also highly statistically significant (Z = 6.4654, p-value < 0.0001), indicating that the proportion of subjects with IHD decreases as [DH-S1P] (as defined by the corresponding quartile) increases.

Sphingolipids in total serum from the CCHS subjects were also evaluated. As shown in Table 3, levels of S1P in total serum were not significantly different between IHD and non IHD individuals for either the high HDL-C or low HDL-C groups. Among the other sphingolipids analyzed in total serum, C18-Cer, C22-Cer and C22:1-Cer were significantly higher in the total serum from individuals with IHD as compared to individuals without IHD irrespective of their HDL levels (Table 3). Levels of several other sphingolipids were statistically different between IHD and non IHD individuals in the high HDL-C and low HDL-C groups, but the direction of the difference was opposite between the two groups.

While the effects of S1P on endothelial cell functions have been implicated as a basis for the anti-atherogenic effects of HDL, rather little is known as to the effects of DH-S1P on vascular cell behaviors. Generally, DH-S1P has been found to be an agonist of S1P receptors [^11-13]. We used electrical cell substrate impedance sensing (ECIS) to assess the ability of DH-S1P to influence endothelial barrier activity, a major physiological function of the endothelium. DH-S1P (on an albumin carrier) induced a dose-dependent increase in transendothelial electrical resistance (TEER) (Figure 3A). The magnitude of the response was similar to that achieved using S1P (Figure 3B). By contrast, C24:1-ceramide, the third sphingolipid found to be significantly lower in HDL-containing fractions of serum from individuals with IHD, did not effect TEER (Additional file 1, Figure S1).

We next evaluated the effects of S1P1 receptor antagonists on the process of DH-S1P-induced barrier enhancement. The S1P1 antagonist, W146, and the S1P1/S1P3 antagonist, VPC23019, both inhibited the TEER response to DH-S1P, similar to the inhibitory effects these agents exert on S1P-mediated barrier enhancement (Figure 3C and 3D). ECIS analyses were also used to evaluate the effect of DH-S1P on endothelial cell migration. The results showed that DH-S1P also promoted S1P1-dependent endothelial cell migration (Figure 4).

Findings from the LC-MS-MS analysis of samples from subjects with and without IHD suggested that S1P and DH-S1P levels on HDL particles might relate to the biological activity of HDL in the context of its anti-atherogenic effects. To address this we evaluated the effect of HDL enriched with varying amounts of S1P on endothelial cell barrier function. Transendothelial electrical resistance monitoring showed that HDL enriched with S1P elicited a dose dependent increase in maximal impedance and sustained a higher level of impedance than native HDL (Figure 5). These results are consistent with the possibility that subjects with higher levels of HDL-associated S1P might have lower vascular permeability, which would be expected to decrease their susceptibility to atherosclerosis.

The study involved the analysis of blood serum samples existing in the Copenhagen City Heart Study (CCHS) collection [²¹,³⁰] which includes a group of patients with documented coronary atherosclerosis and IHD (the Copenhagen Ischemic Heart Disease Study) [³¹,³²]. The CCHS is a prospective cardiopulmonary study of 20- to 93-year-old Danes of both sexes sampled from the general population in 1976-1978 and reexamined in 1981-1983, 1991-1994 and 2001-2003 [³³]. Informed consent was obtained from all participants. The Danish Ethics Committees of Copenhagen and Frederiksberg approved the study (study No. 100.2039/91 and KA93125). Based on analysis of the total population (individuals from CCHS without ischemic heart disease (IHD), n = 5,911 [women = 3,384; men = 2,527]) the average normal level of HDL-C for women is 62.1 ± 18.9 mg/dL and 50.7 ± 16.4 mg/dL for men. Testing was performed on four groups of CCHS samples, including 55 samples from individuals with high HDL-C (80.3 ± 14.3 mg/dL) and no evidence of IHD, 53 samples from gender and age matched individuals with high HDL-C (78.8 ± 14.2 mg/dL) and verified IHD, 54 samples from individuals with low HDL-C (33.7 ± 5.8 mg/dL) and no evidence of IHD, and 42 samples from gender and age matched individuals with low HDL-C (31.8 ± 5.3 mg/dL) and verified IHD. All individuals had LDL-C <160 mg/dL, triglycerides <150 mg/dL, and none were treated with LDL-C-lowering medications. Subjects had an average age of 62 years, approximately 30% were women and 8.5% had diabetes mellitus. The characteristics of subjects from which samples were derived are summarized in Table 1. Additional apolipoprotein and lipid information pertaining to the same groups of subjects as analyzed herein is published in a recent study [³⁴]. This study also showed that for all of the criteria described in Table 1, there was no statistically significant difference between individuals with and without IHD for both the high and low HDL-C groups.

Blinded liquid chromatography-tandem mass spectrometry (LC-MS-MS) sphingolipid analysis was performed on aliquots of total serum samples and on aliquots of LDL- and VLDL-depleted, HDL-containing preparations from CCHS serum samples. To prepare LDL- and VLDL-depleted, HDL-containing preparations, CCHS serum samples were subjected to magnetic bead-dextran-sulfate/MgCl₂ precipitation (Reference Diagnostics, Inc., Bedford, MA) to remove apolipoprotein B (apoB)-containing particles (i.e., LDL and VLDL) [³⁵]. Total cholesterol levels in the supernatants were measured by an enzymatic method using a commercially available kit (Wako Pure Chemical Co., Osaka, Japan). ApoA-I levels were quantified enzymatically in HDL-containing preparations on a Cobas Fara analyzer (Roche Diagnostics Systems, Inc) using reagents from Sigma Aldrich (St. Louis, MO).

Aliquots of total serum or the LDL- and VLDL-depleted, HDL-containing fractions were subjected to LC-MS-MS analysis on a Thermo Finnigan TSQ 7000 triple quadrupole mass spectrometer, operating in a Multiple Reaction Monitoring (MRM) positive ionization mode, using a modified version of the protocol described by Bielawski et al. [³⁶]. Briefly, CCHS samples (50 μl diluted 1:2 with Dulbecco's phosphate-buffered saline [DPBS]) were fortified with the internal standards (ISs: C17 base D-erythro-sphingosine (17CSph), C17 S1P (17CS1P), N-palmitoyl-D-erythro-C13 sphingosine (13C16-Cer) and heptadecanoyl-D-erythro-sphingosine (C17-Cer)), and extracted with ethyl acetate/iso-propanol/water (60/30/10 v/v) solvent system. After evaporation and reconstitution in 100 μl of methanol, samples were injected on the HP1100/TSQ 7000 LC/MS system and gradient eluted from the BDS Hypersil C8, 150 × 3.2 mm, 3 μm particle size column, with 1.0 mM methanolic ammonium formate/2 mM aqueous ammonium formate mobile phase system. Peaks corresponding to the target analytes and internal standards were collected and processed using the Xcalibur software system. Quantitative analysis was based on calibration curves generated by spiking an artificial matrix with known amounts of the target analyte synthetic standards and an equal amount of the internal standards (ISs). The target analyte/IS peak areas ratios were plotted against analyte concentration. The target analyte/IS peak area ratios from the samples were similarly normalized to their respective ISs and compared to the calibration curves, using a linear regression model.

A one-way ANOVA model was fit to each response in order to look for differences in group means. The Studentized residuals were calculated and assessed for the model assumptions of normality and constant variance. Following a significant result, Tukey's adjustment for multiple comparisons was used to control the Type I error rate associated with the corresponding pairwise comparisons. Hypothesis tests were conducted at level of significance 0.05.

HDL (1.063-1.21 g/ml density fraction) was purified from human plasma by sodium bromide density-gradient centrifugation as described in Wasan et al. [³⁷] and dialyzed against DPBS, 0.03 mM EDTA. Following dialysis, total cholesterol, LDL-C and HDL-C were determined using a lipid profile chip (Cholestech, Hayward, CA). No LDL-C was detected in the purified HDL fraction. Protein content of HDL was measured by Bio-Rad DC assay (Bio-Rad, Hercules, CA). Sphingolipid content of HDL was measured by LC-MS-MS by the MUSC Lipidomics Core Facility according to previously reported methods [⁵]. S1P- and DHS1P-fortified HDL were prepared by preincubation of native HDL overnight at 4°C with varying amounts of S1P or DH-S1P (0-1.38 μg S1P/mg HDL protein) followed by dialysis against DPBS, 0.03 mM EDTA to remove free sphingolipds. Sphingolipid content of fortified HDL was measured by LC-MS-MS.

TEER was measured by electrical cell substrate impedance sensing (ECIS) [³⁸] as described previously [⁵]. Briefly, human umbilical vein endothelial cells (HUVEC) (Cascade Biologics, Inc., Portland, OR) maintained in EGM-2 medium were seeded into wells of ECIS 8W10E+ electrode arrays at a density of 1 × 10⁵ cells per well. Arrays were pre-coated with human plasma fibronectin (Invitrogen, Carlsbad, CA) at 100 μg/ml in 0.15 M NaCl, 0.01 M Tris, pH 8.0. Cells were cultured in EGM-2 medium and impedance measured every five min at 15 kHz frequency. When the electrical resistance reached a maximal plateau (~3 days) the medium was replaced with serum-free endothelial basal medium (EBM; Lonza) containing 1X penicillin-streptomycin-glutamine (Invitrogen). Electrical resistance was monitored until a minimal plateau was reached (~24 h). Effectors (i.e., S1P fortified-HDL, dihydro-S1P (DH-S1P)-fortified HDL, S1P-albumin (fatty acid free bovine serum albumin from Sigma), DH-S1P-albumin or albumin in DPBS) were introduced into the culture medium by removing a volume corresponding to that of the effector to be added. S1P (i.e., D-erythro-sphingosine-1-phosphate) and DH-S1P (i.e., dihydro-D-erythro-sphingosine-1-phosphate) were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). The maximum volume of each effector added did not exceed 1/25 of the 400 μl volume of conditioned culture medium in each well. For experiments evaluating the effects of the S1P1 antagonist W146 (Avanti Polar Lipids) and the S1P1/S1P3 antagonist VPC23019 (Avanti Polar Lipids) on TEER, antagonist stocks (1 mM in 5% acidified DMSO, 4 mg/ml BSA) were diluted 1:100 into the conditioned EBM at the same time that S1P or HDL was added.

All values are original measurements from the above-mentioned studies. Selection and matching for the present study were based on these values. All individuals had LDL-C <160 mg/dL, triglycerides <150 mg/dL, and none were treated with LDL-C-lowering medications. Group 1: Females n = 16 Males n = 37; Group 2: Females n = 13 Males n = 29; Group 3: Females n = 16 Males n = 39; Group 4: Females n = 15 Males n = 39. Group 1 had high HDL-C (>90th percentile) and verified IHD; this group was compared with Group 3 without IHD, but matched by age, sex, and similar HDL-C levels (>90th percentile HDL-C levels for Group 1 and 3 individuals were females:≥ 73.5 mg/dL; males:≥ 61.9 mg/dL). Group 2 had low HDL-C (<10th percentile) and verified IHD; this group was compared with Group 4 without IHD, but matched by age, sex, and similar HDL-C levels (<10th percentile HDL-C levels for Group 2 and 4 individuals were females:≤ 38.7 mg/dL; males:≤ 34.1 mg/dL). All individuals without IHD were selected from The Copenhagen City Heart Study's 4th examination. Patients with IHD were selected from individuals referred to the Copenhagen University Hospital, Rigshospitalet, Denmark for coronary angiography.

¹Abbrevations used: Sph, sphingosine; DH, dihydro; Cer, ceramide; P, phosphate; BDL, below detection limit, NA, not applicable since at least one group in the pair was below detection limit.

²Sample sizes are High HDL-C, IHD, n = 53; High HDL-C, no IHD, n = 55; Low HDL-C, IHD, n = 42; Low HDL-C, no IHD, n = 54.

³p-values are derived from a Tukey's comparison following a one-way ANOVA test conducted at level of significance 0.05.

¹Abbrevations used: Sph, sphingosine; DH, dihydro; Cer, ceramide; P, phosphate; BDL, below detection limit; O-NS, Overall not significant by ANOVA; NA, not applicable since at least one group in the pair was below detection limit.

²Sample sizes are High HDL-C, IHD, n = 53; High HDL-C, no IHD, n = 55; Low HDL-C, IHD, n = 42; Low HDL-C, no IHD, n = 54.

³p-values are derived from a Tukey's comparison following a one-way ANOVA test conducted at level of significance 0.05.