Characterization and response to exercise training of HDL-specific phospholipid efflux

Abstract

The HDL-specific phospholipid efflux (HDL-SPE) assay is a novel cell-free measure of HDL function that is inversely associated with coronary artery disease. However, the effect of exercise training on HDL-SPE is unknown. The purpose of this study was to examine the effect of exercise training on HDL-SPE in a large, diverse cohort free of overt disease. Clinical and functional measures of HDL were taken before and after 20 weeks of endurance exercise training in 508 participants from the HERITAGE Family Study. Associations of HDL-SPE with HDL-related traits were examined using Pearson's correlations at baseline and following exercise training (significance: P < 7.4 × 10⁻⁴). The effect of exercise training on HDL-SPE was examined using paired t-tests (significance: P < 0.05). Mean (SD) HDL-SPE was 1.40 (0.19) and higher in females compared with males and in White participants compared with Black participants. Baseline HDL-SPE was strongly associated with HDL-C (r = 0.45) and apoA-I (r = 0.43, both P < 6.9 × 10⁻²⁴) but not with measures of cholesterol efflux. Mean HDL-SPE increased (0.023, P = 0.002) following exercise training, but these increases only occurred in those with the lowest baseline HDL-SPE levels. Change in HDL-SPE was associated with changes in HDL-C (r = 0.27), medium HDL concentration (r = 0.24), and apoA-I and HDL size (r = 0.17, all P < 1.3 ×10⁻⁴). HDL-SPE increased following regular exercise, and changes in HDL-SPE were related to changes in HDL size and subclass concentrations. Our findings demonstrate that individuals at higher risk for coronary artery disease may experience the largest benefits from exercise training as related to this novel biomarker of HDL function.

HDL-C is a well-established risk factor of atherosclerotic CVD (ASCVD) (1) and is included in CVD risk calculations utilized for risk-stratification and risk-modification strategies (2, 3, 4, 5). However, the failure of pharmaceutical trials to improve cardiovascular outcomes despite raising HDL-C (6, 7) and of genetic studies to establish causal associations between HDL-C and CVD through Mendelian randomization (8) has led to the hypothesis that HDL function may serve as a better metric for CVD risk (9). One of the most important antiatherogenic functions of HDL is its ability to promote reverse cholesterol transport, which includes the efflux of cholesterol from peripheral tissues (10). Multiple studies have shown that cholesterol efflux capacity (CEC) is inversely associated with CVD independent of HDL-C and other traditional risk factors (11, 12, 13). However, the cell-based CEC assay is difficult to standardize and requires specific equipment not often available outside research settings. Therefore, our group and others have developed cell-free assays with the goal of reducing assay complexity to allow for potential routine clinical use (14, 15, 16, 17, 18). Specifically, we recently developed and validated a cell-free HDL-specific phospholipid efflux (HDL-SPE) assay and showed that HDL-SPE was inversely associated with prevalent and incident coronary artery disease (CAD) across three clinical studies (14). Proteomic analysis showed that HDL-SPE was mostly mediated by exchangeable HDL apolipoproteins.

While our initial results are promising, more must be known about this novel measure before standardized clinical use can be recommended. For example, though HDL-SPE is associated with CAD, it has not been examined in non-CAD populations. Furthermore, while the benefits of regular exercise on HDL and other lipoprotein profiles are well documented (19), less is known about the effect of exercise training on HDL function. Previous findings from our group showed that high doses of vigorous exercise increased cell-based CEC (20), but the effect of regular exercise on HDL-SPE is unknown. In the present report, we perform an in-depth characterization of HDL-SPE in a large study of Black and White adults free of CVD, including comparisons to cell-based and predicted measures of CEC and examination of changes in HDL-SPE in response to exercise training.

Materials and Methods

HERITAGE Family Study

The HERITAGE Family Study consisted of over 700 self-identified Black and White adults (56% female), recruited as family units, who completed 20 weeks of progressive supervised endurance training at one of four clinical centers (Indiana, Minnesota, Québec, and Texas; Clinical trial registration #NCT00005137). This was a single-arm intervention (i.e., nonrandomized) with no control group. The Consensus on Exercise Reporting Template (21) guidelines for this study are reported in Supplemental Table S1. Full details on study design have been previously described (22, 23). Briefly, participants were physically inactive at baseline (for at least 3 months prior to the start of the study), normotensive or mildly hypertensive (<160/100 mm Hg) without medications for hypertension, diabetes, or dyslipidemia, and with a BMI below 40 kg/m². The study protocol had been approved by the Institutional Review Boards at each of the participating centers of the HERITAGE Family Study consortium. Written informed consent was obtained from each participant. This study was conducted in accordance with the Declaration of Helsinki. The current work included 508 participants with available plasma samples that completed the intervention and had data on HDL- and lipid-related traits before and after exercise training.

Exercise training program

The training program consisted of three weekly sessions on a stationary cycle ergometer (Universal Aerobicycles, Cedar Rapids, IA) at the heart rate associated with 55% of baseline VO_2max for 30 min for the first 2 weeks. The duration and intensity were then gradually increased every 2 weeks until the heart rate associated with 75% of baseline VO_2max for 50 min was achieved. This level was maintained for the final 6 weeks of training. Power output was controlled directly relative to heart rate by using the Universal Gym Mednet (Cedar Rapids, IA) computerized system. The protocol was standardized across all clinical centers and supervised to ensure that the equipment was working properly and that the participants were compliant with the protocol. Adherence to the protocol was ≥95%.

HDL-SPE assay

The measurement of HDL-SPE was performed as previously described (14). Briefly, plasma samples were combined with saline and a lipid-coated calcium silicate hydrate solution in a 96-well plate. Following an hour incubation, the samples were centrifuged. Sample supernatants were then removed and combined with saline and 1% TX-100 in a black 96-well plate. Sample fluorescence was measured using a SpectraMax M3 plate reader. Results were normalized to the measured HDL-SPE of a pooled reference sample evaluated on every plate. All samples were run in duplicate, and the average value is reported. The HDL-SPE assay is highly reproducible, with an average coefficient of variation of ≈2.6% across all plates. Baseline HDL-SPE quartiles were created within each age quartile and race-sex group combination (16 groups in total) and then combined into four overall HDL-SPE quartile groups (24).

³H CEC assay

Measurement of the efflux of radiolabeled (³H) cholesterol from J774 macrophages to apoB-depleted plasma was performed as previously described (20, 25). Briefly, CEC was measured using J774 mouse macrophage cells in the presence and absence of cAMP, thus providing values for global efflux, as well as non-ABCA1-dependent efflux. Results were normalized to the measured efflux by a pooled reference apoB-depleted sample evaluated on every plate. All samples (baseline and postexercise) were run in duplicate, and the average value is reported.

HDL-associated apolipoprotein-predicted measures

HDL-associated apolipoprotein-predicted CEC (pCEC) and CAD risk (pCAD) were calculated as previously described (24, 26). Briefly, fasting serum samples were combined with recombinant ¹⁵N-His₆ApoA-I and then incubated, diluted, and purified. Enriched apoA-I-associated lipoproteins were heat denatured, digested with endoproteinase LysC (Santa Cruz Biotechnology, Dallas, TX), and internal standards were added. Targeted peptides were detected using an Agilent quadrupole mass spectrometer and quantified using MassHunter Quantitative Analysis (Agilent) using the internal standards as reference. A final panel of five apolipoproteins (apo A-I, C-I, C-II, C-III, and C-IV) was used to calculate pCEC and pCAD index (26, 27).

Determination of plasma lipids, lipoproteins, and postheparin lipolytic activities

Plasma samples were taken in the morning following a 12-h fast twice at baseline and 24 h and 72 h after the last exercise session. For eumenorrheic women, all samples were obtained in the early follicular phase to ensure samples were collected at the same phase of the menstrual cycle. Whole blood samples were ultracentrifuged to isolate VLDL. The HDL fraction was obtained after precipitation of LDL in the infranatant by the heparin manganese chloride method (28). Total cholesterol and triglyceride (TG) levels were determined in plasma and lipoprotein fractions by enzymatic methods using the Technicon RA-1000 analyzer. Lipoprotein-associated phospholipase-A2 mass was measured through a latex particle-enhanced turbidimetric immunoassay performed on a Roche P-modular analyzer (29). Concentrations of apoA-I and apoB in plasma and lipoprotein fractions were measured by the rocket-immunoelectrophoretic method. For each time point (baseline and post-training), the two values were averaged and used for analyses. Lipoprotein traits were adjusted for exercise-induced changes in hemodilution.

Postheparin LPL and hepatic lipase activities were measured on one occasion before and after completion of the exercise program, separate from the blood draw for lipid measures, after a 12-h overnight fast, 10 min after an intravenous injection of heparin (60 IU/kg body mass). The postheparin lipolytic activities were measured using a modification of the method of Nilsson-Ehle and Ekman, as previously described (30). Extensive quality-control procedures were implemented to ensure high-quality and reproducible lipid and lipase assays (31, 32).

The lipoprotein subclass profile was quantified via NMR spectroscopy at LabCorp, Inc (Morrisville, NC) using the LP4 deconvolution algorithm (33).

Measurement of cardiometabolic risk factors

Numerous clinical variables were measured at baseline and post-training, including objective body composition measurement via underwater weighing and computerized tomography scans, intravenous glucose tolerance tests, hemodynamics, and inflammatory markers. The methods and substantial quality control procedures implemented to ensure high quality and reproducible measurements have been previously summarized (23).

Given that many HERITAGE participants were below the age of 40, we used the Framingham Heart Study 30-year ASCVD prediction tool (5), implemented in SAS using a previously published macro (https://github.com/zmn0322/30-Year-CVD-Risk-Prediction), to estimate ASCVD risk.

Statistical analysis

Change with exercise training (delta) was calculated by subtracting the baseline value from the post-training value. The correlation of 41 lipid- and HDL-associated traits with HDL-SPE was tested at each timepoint (baseline, post-training) using partial Pearson's correlations adjusted for age, sex, and race, whereas post-training correlations included additional adjustments for baseline trait and HDL-SPE levels. Given the roles of HDL-C, apoA-I, and total TGs in HDL metabolism and their use as clinical measures, additional adjustments for HDL-C, apoA-I, and TG were also performed at each timepoint. Although the focus of this analysis was on lipid/lipoprotein-related traits, we also examined the correlation of HDL-SPE with an additional 27 cardiometabolic traits to further characterize HDL-SPE. The full list of 68 cardiometabolic traits can be found in Supplemental Tables S2 (baseline) and S3 (post exercise training change). To account for multiple testing, a Bonferroni correction of P < 7.4 × 10⁻⁴ (68 traits tested) was used to define significance. Paired t-tests were used to examine exercise-induced changes in HDL-SPE in the total cohort and stratified by sex, race, age quartile, and baseline HDL-SPE quartile. Differences between sex and race subgroups were tested via Student's t-tests, whereas a general linear model was used to examine HDL-SPE differences between age quartiles. A P value <0.05 was used to define significant within- and between-group differences in HDL-SPE, with P values ≤0.10 considered as trending toward significance.

Given the novelty of the HDL-SPE assay, exploratory multivariable regression models were performed to identify potential predictors. At baseline, any traits (excluding other measures of HDL function, i.e., predicted and cell-based CEC) that were nominally correlated (P < 0.05) with HDL-SPE were included in a forward regression model that also included age, sex, and race. To examine the predictors of change in HDL-SPE following exercise training, the change in nominally correlated traits was included, as well as age, sex, race, and baseline HDL-SPE.

Results

The 508 participants included in this study (56% female, 39% self-identified Black) had mean HDL-C, TG, and total cholesterol levels within normal ranges (Table 1). At baseline, mean HDL-SPE was 1.40 ± 0.19 in the total sample and was significantly higher in females compared with males and in White participants compared with Black participants (Table 2). On average, participants in the oldest age quartile (age, 47.3–65.9 years) had significantly higher HDL-SPE compared with participants in the other age quartiles. We did not find an interaction between sex and race (P = 0.38) or between age quartiles and sex (P = 0.22) and race (P = 0.38) on baseline HDL-SPE, thus, here forward, we report results for sex, race, and age quartiles separately.

Table 1.

Baseline clinical and demographic characteristics of HERITAGE Family Study participants (N = 508)

Trait	Mean ± SD
Age, years	34.9 ± 13.3
Race, % Black	39
Sex, % female	56
HDL-C, mg/dL	41.6 ± 10.8
Total cholesterol, mg/dL	170.5 ± 36.4
TGs, mg/dL	107.9 ± 61.1
BMI, kg/m2	26.5 ± 5.4
Fasting glucose, mmol/L	5.1 ± 0.6
Diastolic blood pressure, mm Hg	68.2 ± 8.9
Systolic blood pressure, mm Hg	118.4 ± 11.9
30-year ASCVD risk, %	15.8 ± 15.2

Table 2.

HDL-SPE values at baseline and following exercise training in the total sample and stratified by sex, race, and age-quartile subgroups

Group	HDL-SPE
	Baseline	Post-training	Delta	Delta-minimum	Delta-maximum
Total (N = 508)	1.40 ± 0.19	1.43 ± 0.19	0.023 ± 0.16a	−0.455	0.700
Females (n = 285)	1.42 ± 0.21b	1.44 ± 0.20	0.016 ± 0.16	−0.455	0.700
Males (n = 223)	1.38 ± 0.17	1.41 ± 0.17	0.031 ± 0.16a	−0.445	0.697
Self-identified Black (n = 196)	1.33 ± 0.19c	1.35 ± 0.17c	0.019 ± 0.15	−0.419	0.700
Self-identified White (n = 312)	1.45 ± 0.18	1.48 ± 0.18	0.025 ± 0.17a	−0.455	0.697
Age quartile 1 (n = 118)	1.37 ± 0.18a	1.40 ± 0.20a	0.036 ± 0.19a	−0.356	0.697
Age quartile 2 (n = 123)	1.38 ± 0.21a	1.41 ± 0.18a	0.028 ± 0.17	−0.445	0.698
Age quartile 3 (n = 135)	1.38 ± 0.19a	1.40 ± 0.18a	0.021 ± 0.14	−0.455	0.402
Age quartile 4 (n = 132)	1.48 ± 0.18b	1.49 ± 0.19b	0.007 ± 0.15	−0.423	0.608

HDL-SPE values are unitless because of normalization to a pooled plasma sample.

Delta calculated as post-training - baseline values.

Age quartiles calculated in the full HERITAGE cohort: age quartile 1: 15.9–22.5 years; age quartile 2: 22.6–30.7 years.

Age quartile 3: 31.2–47.2 years; age quartile 4: 47.3–65.9 years.

Different superscript letters denote significant differences (P < 0.05) across age quartile groups within a timepoint.

^aSignificant (P < 0.05) within-group change with exercise.

^bSignificant difference compared with male participants (P < 0.05).

^cSignificant difference compared with White participants (P < 0.05).

Baseline associations of HDL-SPE with cardiometabolic traits

The associations of HDL-SPE with HDL-related traits can be seen in Table 3. At baseline, HDL-SPE was most strongly correlated with HDL-C (r = 0.45) and apoA-I (r = 0.43, both P ≤ 6.9 × 10⁻²⁴). All remaining significant HDL-associated traits (e.g., total and large HDL, HDL-TG) were positively correlated with HDL-SPE (range r = 0.17–0.41, all P < 1.9 × 10⁻⁴). Baseline HDL-SPE was weakly (r = 0.13, P = 0.005) correlated with cell-based non-ABCA1 ³H-CEC and weak-to-moderately (r = 0.30, P = 1.3 × 10⁻¹¹) correlated with pCEC. However, HDL-SPE was not significantly correlated with cell-based global ³H-CEC (r = 0.08, P = 0.07). After additional adjustments for HDL-C, apoA-I, and TG, medium-sized and total HDL particle concentration remained associated with HDL-SPE, though at a weaker strength than in the original model (Supplemental Table S2).

Table 3.

Correlations of baseline HDL-SPE with lipid and lipoprotein traits in the total sample and by sex

	Total (N = 508)		Males (n = 223)		Females (n = 285)
	ra	P	rb	P	rb	P
ApoA-I, mg/dl	0.43	6.90E-24	0.41	3.59E-10	0.44	1.58E-14
Global 3H-CEC	0.08	0.07	0.20	0.003	−0.01	0.92
Non-ABCA1 3H-CEC	0.13	0.005	0.15	0.03	0.10	0.10
pCEC, %efflux/4 h	0.30	1.30E-11	0.31	2.48E-06	0.29	8.51E-07
HDL-C, mg/dl	0.45	3.05E-26	0.39	2.48E-09	0.48	8.22E-18
HDLP size, nm	0.31	9.72E-13	0.26	1.02E-04	0.34	3.13E-09
Total HDLP, μmol/l	0.40	3.58E-21	0.44	8.16E-12	0.38	5.14E-11
Large HDLP, μmol/l	0.31	5.08E-13	0.26	7.65E-05	0.34	5.00E-09
Medium HDLP, μmol/l	0.41	6.79E-22	0.37	1.77E-08	0.43	2.71E-14
Small HDLP, nmol/l	−0.04	0.40	0.07	0.34	−0.10	0.09
HDL-TG, mg/dl	0.17	1.81E-04	0.20	0.003	0.14	0.02
H1P, μmol/l	−0.03	0.54	0.03	0.63	−0.07	0.21
H2P, μmol/l	−0.03	0.52	0.06	0.42	−0.08	0.20
H3P, μmol/l	0.33	5.12E-14	0.31	3.22E-06	0.34	6.65E-09
H4P, μmol/	0.34	1.19E-14	0.29	1.62E-05	0.35	1.17E-09
H5P, μmol/	0.02	0.68	0.14	0.04	−0.05	0.44
H6P, μmol/l	0.29	2.41E-11	0.11	0.11	0.36	2.89E-10
H7P, μmol/l	0.26	2.36E-09	0.24	2.71E-04	0.30	3.66E-07
PH-HL activity, nm/ml/min	−0.04	0.38	−0.07	0.34	−0.04	0.53
PH-LPL activity, nm/ml/min	0.18	1.00E-04	0.21	0.002	0.14	0.02

apoA-I, apolipoprotein A-I; H1P-H7P, HDL particle subspecies 1–7: H7P (12 nm), H6P (10.8 nm), H5P (10.3 nm), H4P (9.5 nm), H3P (8.7 nm), H2P (7.8 nm), and H1P (7.4 nm); HDLP, HDL particle; PH-HL, postheparin hepatic lipase; PH-LPL, postheparin LPL.

Bolded text represents Bonferroni-corrected significant correlations (P < 7.4E-4).

^aPartial correlations adjusted for age, sex, and race.

^bPartial correlations adjusted for age and race.

Baseline HDL-SPE was not correlated with standard lipid panel measures or most of the included cardiometabolic traits such as blood pressure or any measure of body composition (Supplemental Table S2). In terms of CVD risk scores, baseline HDL-SPE was significantly, inversely correlated with pCAD index (r = −0.19, P = 2.1 × 10⁻⁵) and nominally inversely correlated with 30-year ASCVD risk (r = −0.14, P = 0.002).

In the total sample, 16 traits were nominally associated with HDL-SPE and entered into the regression model. Eight traits were retained in the final model that explained 42.7% of the variance in baseline HDL-SPE (Supplemental Table S4A). Among these, baseline apoA-I was the strongest predictor of HDL-SPE (partial R² = 24.6%), followed by self-identified race (partial R² = 6.0%), HDL-C (partial R² = 4.3%), and total HDL particle concentration (partial R² = 2.7%).

Given the significant differences in HDL-SPE between male and female participants, sex-stratified correlations with cardiometabolic traits were examined. The traits showing the strongest correlations with HDL-SPE in male participants were total HDL particle concentration, apoA-I, and HDL-C (r = 0.39–0.44, all P < 3.6 × 10⁻¹⁰, Table 3). Similarly, the strongest traits correlated with HDL-SPE in female participants were HDL-C, apoA-I, and medium-sized HDL particle concentration (r = 0.43–0.48, all P < 5.2 × 10⁻¹¹). HDL-SPE was nominally correlated with cell-based global (r = 0.20, P = 0.003) and non-ABCA1 ³H-CEC (r = 0.15, P = 0.03) in male but not female participants (Table 3). Additional adjustments for HDL-C, apoA-I, and TG led to total HDL particle concentration remaining the only trait significantly correlated with HDL-SPE in male participants (Supplemental Table S2). Conversely, visceral fat, medium HDL particle concentration, hepatic lipase activity, C-reactive protein, and H4P particle concentration were significantly associated with HDL-SPE in female participants with these additional adjustments.

Sex-stratified regression models included a total of 12 nominally associated traits for both male and female participants, with all entered traits identical between groups except for the inclusion of H5P and H6P concentrations for male and female participants, respectively. In male participants, only three traits were retained in the final model, explaining 35.5% of the variance in baseline HDL-SPE. Total HDL particle concentration (partial R² = 22.5) was the strongest predictor, followed by participant self-identified race (partial R² = 6.1%) and HDL-C (partial R² = 6.3%, Supplemental Table S4B). In female participants, seven traits were retained in the final model that explained 45.8% of the variance. The strongest predictor was apoA-I (partial R² = 27.4%), followed by self-identified race (partial R² = 6.3%), HDL-C (partial R² = 4.7%), H4P concentration (partial R² = 2.7%), age (partial R² = 1.9%), medium-sized HDL particle concentration (partial R² = 1.9%), and LPL activity (partial R² = 1.0%, Supplemental Table S4C).

Given the known and observed (data not shown) differences in HDL measures between males and females, to further explore potential sex differences in HDL-SPE, we performed additional models adjusting HDL-SPE for various HDL metrics along with age and race. These included HDL-C, apoA-I, and all HDL subclass particle concentrations, which were used as covariates in aggregate and individually. After adjustment for age and race, female participants still had higher baseline HDL-SPE than males with the additional individual adjustment for apoA-I (P = 0.005), total HDL (P < 0.0001), or small HDL (P = 0.0004) concentrations. However, HDL-SPE was different neither between males and females in the aggregate model (P = 0.53) nor with the individual inclusion of HDL-C (P = 0.46), medium HDL (P = 0.13), or large HDL (P = 0.15) concentrations as covariates.

Change in HDL-SPE with exercise training

On average, HDL-SPE significantly increased (0.023 ± 0.16, P = 0.002; ∼2.44%) following exercise training in the total cohort, though there was large interindividual variation in HDL-SPE response to exercise training (Table 2). Although HDL-SPE significantly increased following training in males only (0.031 ± 0.16, P = 0.004; 3.0%), there was no difference (P = 0.31) in the magnitude of change compared with female participants (0.016 ± 0.16, P = 0.10; 2.0%). No differences in exercise response were observed across self-identified race groups (P = 0.65), as mean HDL-SPE significantly increased with training in White participants (0.025 ± 0.17, P = 0.009; 2.5%), whereas Black participants trended toward an increase with training (0.019 ± 0.15, P = 0.08; 2.4%). On average, only participants in the youngest age quartile increased HDL-SPE following exercise training (0.036 ± 0.19, P = 0.04; 3.5%); however, no differences in the magnitude of change across age quartiles were observed (Table 2).

HDL-SPE values for each time point by age-, sex-, and race-specific quartiles of baseline HDL-SPE can be found in Supplemental Table S5. On average, those with the lowest levels of baseline HDL-SPE (Q1 and Q2) had the largest mean increases in HDL-SPE following exercise training (0.117 and 0.035, respectively, Fig. 1).

Fig. 1.

Change in HDL-SPE following exercise training by baseline age-, sex-, and race-specific HDL-SPE quartile. Box plots depict interquartile range of HDL-SPE change with median value shown as a line within the box. ∗Significant (P < 0.05) within group change with exercise training.

Associations of exercise-induced changes in HDL-SPE and cardiometabolic traits

Exercise-induced change in HDL-SPE was significantly associated with the changes in multiple HDL-related traits (Table 4), including HDL-C (r = 0.27, P = 1.6 × 10⁻⁹), total HDL particle concentration (r = 0.22, P = 9.9 × 10⁻⁷), and HDL particle size and apoA-I (both r = 0.17, P < 1.9 × 10⁻⁴). Change in HDL-SPE was significantly correlated with concomitant changes in global CEC (r = 0.16, P = 6.0 × 10⁻⁴) and pCEC (r = 0.23, P = 2.6 × 10⁻⁷) and nominally associated with change in non-ABCA1 CEC (r = 0.15, P = 7.5 × 10⁻⁴). Changes in 30-year ASCVD risk and pCAD index were inversely associated with the change in HDL-SPE (r = −0.15 and −0.16, respectively, both P < 2.5 × 10⁻⁴, Supplemental Table S3). Change in HDL-SPE was nominally inversely associated with the change in systolic blood pressure and positively associated with changes in LPL activity and total cholesterol. Additional adjustments for the changes in HDL-C, apoA-I, and TG reduced the number of significant correlations, with the changes in medium-sized HDL, H3P, total HDL particle concentrations, and pCEC, the only remaining significant correlations (Supplemental Table S3).

Table 4.

Correlations of the change in HDL-SPE with changes in lipid- and HDL-associated traits in the total sample and in sex-stratified subgroups

	Total sample (N = 508)		Males (n = 223)		Females (n = 285)
	ra	P	rb	P	rb	P
ApoA-I, mg/dl	0.17	1.88E-04	0.31	3.27E-06	0.07	0.25
Global 3H-CEC	0.16	6.05E-04	0.17	0.01	0.15	0.01
Non-ABCA1 3H-CEC	0.15	7.45E-04	0.08	0.28	0.21	3.67E-04
pCEC, %efflux/4 h	0.23	2.56E-07	0.33	7.42E-07	0.16	0.01
HDL-C, mg/dl	0.27	1.57E-09	0.24	3.10E-04	0.26	6.91E-06
HDLP size, nm	0.17	1.21E-04	0.13	0.06	0.19	0.001
Total HDLP, μmol/l	0.22	9.95E-07	0.30	9.37E-06	0.18	0.003
Large HDLP, μmol/l	0.19	3.02E-05	0.18	0.01	0.19	0.002
Medium HDLP, μmol/l	0.24	1.03E-07	0.30	7.69E-06	0.18	0.002
Small HDLP, nmol/l	−0.05	0.31	−0.02	0.80	−0.05	0.38
HDL-TG, mg/dl	0.09	0.053	0.10	0.15	0.09	0.15
H1P, μmol/l	−0.15	0.001	−0.16	0.02	−0.15	0.01
H2P, μmol/l	0.03	0.44	0.10	0.14	0.01	0.88
H3P, μmol/l	0.20	4.92E-06	0.28	4.13E-05	0.15	0.01
H4P, μmol/l	0.15	0.001	0.21	0.002	0.11	0.08
H5P, μmol/l	−0.02	0.74	0.04	0.53	−0.04	0.46
H6P, μmol/l	0.18	8.24E-05	0.12	0.08	0.21	5.39E-04
H7P, μmol/l	0.11	0.02	0.08	0.27	0.17	0.004
PH-HL activity, nm/ml/min	0.08	0.07	0.08	0.22	0.08	0.20
PH-LPL activity, nm/ml/min	0.12	0.01	0.10	0.15	0.13	0.03

apoA-I, apolipoprotein A-I; H1P-H7P, HDL particle subspecies 1–7: H7P (12 nm), H6P (10.8 nm), H5P (10.3 nm), H4P (9.5 nm), H3P (8.7 nm), H2P (7.8 nm), and H1P (7.4 nm); HDLP, HDL particle; PH-HL, postheparin hepatic lipase; PH-LPL, postheparin LPL.

Bolded text represents Bonferroni-corrected significant correlations (P < 7.4E-4).

^aPartial correlations adjusted for age, sex, race, baseline HDL-SPE, and baseline trait.

^bPartial correlations adjusted for age, race, baseline HDL-SPE, and baseline trait.

Changes in 13 HDL- and lipoprotein-associated traits were nominally associated with changes in HDL-SPE and entered into the regression model. A total of seven traits explained 31.7% of the variance in exercise-induced change in HDL-SPE, with baseline HDL-SPE levels being the highest predictor (partial R² = 19.9%, Supplemental Table S6A). The remaining significant predictors included change in HDL-C (partial R² = 4.6%), self-identified race (partial R² = 2.7%), and change in medium-sized HDL particle concentration (partial R² = 1.5%).

Sex-stratified analyses showed that changes in HDL-SPE were correlated with concomitant changes in different lipid and lipoprotein traits between male and female participants (Table 4). For example, change in HDL-SPE was most strongly correlated with changes in pCEC, apoA-I, and medium-sized HDL particle concentration in male participants (r: 0.30–0.33, all P < 3.3 × 10⁻⁶). In female participants, only the changes in HDL-C, non-ABCA1 CEC, and H6P particle concentration were correlated with changes in HDL-SPE. Sex-stratified correlations with additional adjustments for the changes in HDL-C, apoA-I, and TG showed that only the change in pCEC remained correlated with changes in HDL-SPE in males, whereas no traits remained correlated in females (Supplemental Table S3).

Sex-stratified regression models for change in HDL-SPE included a total of 10 nominally associated traits for both male and female participants, with most included traits common to both groups (for the full list of included traits, see Supplemental Tables S6, B and C). Five predictors remained in both models. The top three predictors were the same in both male- (total R² = 32.4%) and female-specific (total R² = 32.9%) regression models, specifically baseline HDL-SPE (partial R² = 20.5% and 20.1%, respectively), followed by the change in HDL-C (partial R² = 4.8% and 4.9%, respectively), and self-identified race (both partial R² = 2.9%, Supplemental Tables S6, B and C). Male-specific predictors included H1P and total HDL particle concentrations, whereas female-specific predictors were medium-sized HDL and H6P particle concentrations.

Discussion

In the present study, we demonstrated that 20 weeks of regular exercise training increased HDL-SPE in a diverse sample of adults free of CVD. Exercise-induced increases in HDL-SPE occurred, on average, only in participants with the lowest baseline levels. Female participants had higher HDL-SPE at baseline, but no sex differences in the magnitude of HDL-SPE response to exercise training were observed. We also extensively characterized the cardiometabolic traits associated with baseline and exercise-induced change in HDL-SPE, which included expected traits like HDL-C, apoA-I, and total HDL particle concentration at both timepoints. HDL-SPE was generally distinct from cell-based measurements of cholesterol efflux at baseline. Moreover, the change in HDL-SPE was weakly correlated with changes in cell-based and predicted measures of HDL function. This demonstrates that in our data, these measures are likely not capturing the same aspects of HDL functionality, and the biological mechanisms underlying their responses to exercise training also likely differ.

We previously found that HDL-SPE was inversely associated with CAD in three unique cohorts, including two case-control studies (individuals with and without CAD) as well as a nested case-control general population cohort (incident CVD events), independent of HDL-C and apoA-I (14). Moreover, HDL-SPE showed stronger associations with CAD than cell-based CEC. Specifically, CEC was not associated with CAD in case-control models, whereas the association of CEC with incident CVD events (odds ratio [OR] = 0.34; 95% confidence interval = 0.16–0.74) was weaker than that of HDL-SPE (OR = 0.09, 95% confidence interval = 0.02–0.33), particularly in models that included adjustment for the other measure (14). In these cohorts, HDL-SPE was correlated with multiple HDL-related traits, including HDL-C, HDL size, and apoA-I. Our current analysis yielded similar results, as the relationship of HDL-SPE with HDL-C, HDL particle concentration, HDL size, apoA-I, and apoB was similar in strength and direction between studies. The major difference between studies was that the original study found a weak but significant relationship between cell-based global ³H-CEC and HDL-SPE (Sato et al. (14): r = 0.22, P = 0.04), whereas we found no correlation at baseline in the present report. The clinical cohorts used in the study by Sato et al. (14) were older, primarily male (≥69%), had clinical indications for coronary computed tomography angiographies, a 20% prevalence of type 2 diabetes, and were obese on average. Conversely, the HERITAGE cohort was younger, majority female, and overweight on average, with no overt CVD or diabetes. These differences are reflected in the noticeably lower HDL-SPE values in the Sato et al. groups (mean values of 0.93 and 0.94 in CAD patients and 1.04 and 1.11 in non-CAD patients) compared with HERITAGE (mean = 1.40). Accordingly, subgroup analysis in the HERITAGE study demonstrated that HDL-SPE was nominally correlated with CEC at a similar strength as Sato et al. in males and participants from the oldest age quartile (r = 0.20, P < 0.05 for both). Furthermore, the HDL-SPE and CEC assays reflect different underlying mechanisms of HDL-mediated lipid efflux. Unlike phospholipids, cholesterol rapidly exchanges between lipid surfaces. The CEC assay reflects the net movement of cholesterol from cultured mouse J774, RAW264.7, or human THP-1 macrophages to plasma HDL. Cellular cholesterol efflux occurs by four different mechanisms involving passive aqueous diffusion, SR-B1-mediated facilitated diffusion, and active transport mediated by ABCA1 and ABCG1 (34). Moreover, LCAT-mediated cholesterol esterification may facilitate HDL acquisition of cellular cholesterol. The HDL-SPE assay, on the other hand, is based on the active solubilization of donor particle lipids, including the nonexchangeable marker fluorescent phosphatidylethanolamine (PE), by HDL-derived apoA-I, and to a lesser extent, other HDL exchangeable apolipoproteins (14). We have previously shown that apoA-I and other HDL exchangeable lipoproteins support both cellular cholesterol and phospholipid ABCA1-mediated efflux (35). Unlike cholesterol, the donor particle nonexchangeable fluorescent PE efflux marker can only be transferred to HDL in the HDL-SPE assay by active exchangeable apolipoprotein solubilization of donor particle lipids. We infer that the lipid coat on the particles likely mimics the ABCA1-generated lipid domains on cells (14), which have been shown to support apolipoprotein-mediated solubilization of phospholipids, including plasma membrane PE (36). Thus, the differences in the correlation of HDL-SPE with CEC may be due to a combination of differences in demographics between studies and underlying efflux mechanisms between assays.

The current report is the first to examine whether HDL-SPE is responsive to regular exercise. Although the absolute exercise training-induced increase (∼2.4%) in HDL-SPE was relatively small, we can use our findings from a general population cohort (14) to extrapolate the potential clinical significance of these changes. Specifically, in the PREVEND cohort, we found that the OR of incident CVD events per standard deviation increase in HDL-SPE was 0.09 in fully adjusted models (14). Thus, the observed exercise-induced increase in HDL-SPE would correspond to a ∼10% reduction in incident CVD events from just 20 weeks of regular exercise. As our participants were relatively young and healthy, we would expect larger, and thus potentially more clinically meaningful, increases in HDL-SPE with exercise training in older and/or higher risk populations.

Previous studies examining the effects of exercise interventions on CEC had mixed results (37). Findings from our group demonstrated that only a high volume of vigorous endurance exercise (16 kcal/kg of body weight/week [KKW] at 75% intensity) increased cell-based global ³H-CEC in participants with prediabetes (STRRIDE-PD [Studies of a Targeted Risk Reduction Intervention through Defined Exercise—Prediabetes] study), wheereas non-ABCA1 ³H-CEC only increased in the high volume endurance exercise group (20 KKW at 65–85% intensity) of the E-MECHANIC study (20). The current analysis showed that 20 weeks of endurance exercise training significantly increased HDL-SPE, despite large interindividual variations in response. However, it is important to note that the HERITAGE exercise program only prescribed one dose; thus, it is possible that higher doses may be needed to appreciably increase HDL-SPE with regular endurance exercise. The volume (∼12–14 KKW) but not intensity level of exercise in the HERITAGE study was also lower than that of the exercise groups included in our aforementioned previous report that suggested an exercise dose threshold was needed to improve CEC (20). Therefore, more dose-response studies are needed to identify the optimal amount and/or intensity of exercise that increases HDL-SPE.

Both baseline and change in HDL-SPE were inversely associated with predictive metrics of CVD risk. Although the observed absolute increase in HDL-SPE with training in the HERITAGE study was small, we found that individuals with the lowest baseline HDL-SPE levels experienced significant increases in HDL-SPE, demonstrating that those most at risk for CAD development may receive the greatest benefits from regular exercise. Given the known effects of exercise training on CVD risk reduction, the exercise-induced increase in HDL-SPE and subsequent inverse associations with CVD risk score provide promising indirect evidence that increases in HDL function may be one mechanism involved in the cardioprotective benefits of regular exercise.

We found large interindividual variations in HDL-SPE response to exercise training across all subgroups (sex, race, and age), which has also been observed in our previous exercise training studies of other measures of HDL function (i.e., CEC, HDL-apoA-I-exchange) (20). However, the factors underlying the variation in exercise response are unknown. Potential factors include concomitant changes in HDL particle concentration, subclass size distribution (i.e., a greater number of small or large HDL particles following exercise training), or particle composition. As the HDL-SPE assay measures the ability of the HDL particle to pick up fluorescent-tagged PE, we hypothesize that changes in HDL particle composition are likely underlying the changes in HDL-SPE. Specifically, as HDL-SPE, cell-based CEC (38), and pCEC (23, 24) are all associated with and/or mediated by exchangeable apolipoproteins, differences in HDL apolipoprotein responses to exercise training may be underlying the heterogeneity. However, we cannot discount the potential for multifactorial drivers of change. Regardless, further research is needed to clarify the clinical and molecular factors underlying the variations in HDL-SPE exercise response.

We observed differences in HDL-SPE between males and females. However, the inclusion of HDL-C and larger HDL subclass traits attenuated this sex difference but not indicators of HDL particle number. Furthermore, the baseline and delta HDL-SPE predictors were largely similar between males and females and explained similar levels of variance. Sex-stratified cardiometabolic trait associations with HDL-SPE were also mostly similar both at baseline and in response to exercise training. Given these findings, the observed differences in HDL-SPE between males and females may be due to known sex differences in HDL-C and subclass distribution rather than true differences in HDL-SPE (39, 40, 41, 42). Specifically, females tend to have a higher proportion of large HDL particles than males, which tend to have a higher proportion of lipids and thus greater exchangeability and surface area for the PE to bind to. However, further research into the interactions between sex, HDL composition, and HDL function is needed.

This study has several strengths, including a large sample size of diverse participants undergoing a supervised exercise training intervention, with high adherence (≥95%). Thus, the interindividual differences in HDL-SPE response were likely because of factors other than differences in the exercise intervention. The extensive phenotyping of HERITAGE participants allowed for a thorough characterization of HDL-SPE, including identifying relationships with HDL- and lipid-related traits, other measures of HDL function, and CVD risk factors. Finally, HDL-SPE is a highly reproducible assay, ensuring differences observed are not because of variations in the assay procedure itself. The limitations of this study include a lack of hard CVD outcome measures from participants in the HERITAGE Family Study; thus, comparisons to our previous findings with actual CAD risk were not possible. Blood collection occurred at only two timepoints (pre- and postexercise intervention), and the same progressive relative dose of exercise was prescribed to all participants. Therefore, we are unable to analyze dose-response effects or the time course of changes in HDL-SPE over the duration of the exercise program. The lack of a control group also limits the conclusions from this study, as we cannot exclude that other factors contributed to the changes in HDL-SPE. However, other HERITAGE findings on the effects of endurance exercise training on multiple lipid and lipoprotein traits were similar to other studies with control groups (43, 44). While the health of the HERITAGE Family Study cohort was beneficial in limiting confounding factors associated with chronic disease, it is possible that the lack of improvement in HDL-SPE in participants with higher baseline levels demonstrates a ceiling effect, given the younger age and general health of the cohort. Future studies on the effect of exercise training on HDL-SPE in older and/or diseased populations, as well as interventions with different durations and/or doses of exercise training, would be helpful in elucidating this possibility. Finally, the measures of HDL function in this study were not separated by sized-based HDL subclasses (e.g., small-sized HDL-SPE vs. medium-sized HDL-SPE), which may be able to better elucidate the biological mechanisms underlying changes in HDL functionality with regular exercise and specifically differences in the exercise response of HDL-SPE compared with other (cell-based) measures.

Conclusion

HDL-SPE was generally distinct from cell-based measures of CEC and was associated with the number and size of HDL particles and apoA-I concentration. HDL-SPE significantly increased with exercise training, though subgroup analyses found these increases, on average, only occurred in males or those with the lowest baseline HDL-SPE. These findings indicate that individuals most at-risk for CAD may experience the largest benefits from exercise training as related to this novel biomarker of HDL function.

Data Availability

The data used in this study are available upon request through collaborations from the corresponding author, Mark A. Sarzynski (sarz@mailbox.sc.edu).

Supplemental Data

This article contains supplemental data.

Conflict of Interest

T. S. C. reports a relationship with Quest Diagnostics that includes employment and equity or stocks. R. J. K. reports a relationship with Eli Lilly and Company that includes employment and equity or stocks. M. S. reports a relationship with Eiken Chemical Co, Ltd that includes employment and equity or stocks. All other authors declare that they have no conflicts of interest with the contents of this article.