HDL flux is higher in patients with nonalcoholic fatty liver disease

Abstract

Altered lipid metabolism and inflammation are involved in the pathogenesis of both nonalcoholic fatty liver disease (NAFLD) and cardiovascular disease (CVD). Even though high-density lipoprotein (HDL), a CVD protective marker, is decreased, whether HDL metabolism and function are perturbed in NAFLD are currently unknown. We examined the effect of NAFLD and disease severity on HDL metabolism and function in patients with biopsy-proven simple steatosis (SS), nonalcoholic steatohepatitis (NASH), and healthy controls. HDL turnover and HDL protein dynamics in SS (n = 7), NASH (n = 8), and healthy controls (n = 9) were studied in vivo. HDL maturation and remodeling, antioxidant, cholesterol efflux properties, and activities of lecithin-cholesterol ester acyltransferase and cholesterol ester transfer protein (CETP) were quantified using in vitro assays. All patients with NAFLD had increased turnover of both HDL cholesterol (HDLc; 0.16 ± 0.09 vs. 0.34 ± 0.18 days, P < 0.05) and apolipoprotein A1 (ApoAI) (0.26 ± 0.04 vs. 0.34 ± 0.06 days, P < 0.005) compared with healthy controls. The fractional catabolic rates of other HDL proteins, including ApoAII (and ApoAIV) were higher (P < 0.05) in patients with NAFLD who also had higher CETP activity, ApoAI/HDLc ratio (P < 0.05). NAFLD-induced alterations were associated with lower antioxidant (114.2 ± 46.6 vs. 220.5 ± 48.2 nmol·mL⁻¹·min⁻¹) but higher total efflux properties of HDL (23.4 ± 1.3% vs. 25.5 ± 2.3%) (both P < 0.05), which was more pronounced in individuals with NASH. However, no differences were observed in either HDL turnover, antioxidant, and cholesterol efflux functions of HDL or HDL proteins’ turnover between subjects with SS and subjects with NASH. Thus, HDL metabolism and function are altered in NAFLD without any significant differences between SS and NASH.

INTRODUCTION

The spectrum of nonalcoholic fatty liver disease (NAFLD), the most common chronic liver disease worldwide, includes simple steatosis (SS) and the more severe nonalcoholic steatohepatitis (NASH), which is characterized by necroinflammation and fibrosis. Despite liver-related morbidity and mortality, cardiovascular disease (CVD) is the major cause of death in NAFLD (49, 53, 55). As the hepatic manifestation of metabolic syndrome, NAFLD is associated with insulin resistance and dyslipidemia that may lead to type 2 diabetes mellitus, a common CVD risk factor (4). Multivariate analysis indicates that NAFLD, particularly its severe form NASH, predicts CVD risk independent of the metabolic syndrome and diabetes (18, 52, 53). A cross-sectional study demonstrated that patients with biopsy-proven NAFLD have an increased CVD risk compared with matched healthy controls, and the risk is higher in patients with NASH relative to those with SS (55). Based on these findings, it has been proposed that NAFLD is not a merely marker of CVD but may also be mechanistically involved in CVD development (53). However, adjustment for confounders is often incomplete, and recent studies show that even after resolution of NAFLD, patients continue to have an increased risk of CVD (8). It is also unclear whether CVD risk is due to perturbations in lipid metabolism that also result in hepatic fat accumulation, or if hepatic inflammation contributes to atherogenesis.

Patients with NAFLD exhibit low levels of HDL cholesterol (HDLc), a known risk marker of CVD (7, 9). However, the role of HDL in CVD development has been questioned because treatment with cholesterol ester (CE) transfer protein (CETP) inhibitors increased HDLc levels but failed to improve cardiovascular outcomes (1, 36). In addition, genetic studies demonstrated the lack of a causal relationship between low plasma HDLc and higher risk of myocardial infarction (58). In contrast, high levels of dysfunctional HDL are associated with increased CVD risk (39, 42). These studies highlight the need for HDL functionality studies. HDL has multiple functions, including preventing inflammation, oxidation, platelet activation, and most importantly transport of cholesterol from peripheral tissues to the liver for biliary excretion through reverse cholesterol transport (RCT) pathway (12). Conditions associated with insulin resistance are characterized by altered HDL functions (22). Systematic inflammation further impairs HDL functions, including RCT, and converts HDL into proinflammatory particles (19). The reasons for these changes are not fully understood but have been attributed to alterations in HDL particle composition. In addition to ApoAI and ApoAII, more than 80 less-abundant HDL proteins have been identified (46). HDL proteins are involved in lipid metabolism, acute-phase response, and innate immunity, which determine both HDL’s anti-inflammatory and anti-atherogenic properties (51).

Recently, we demonstrated proinflammatory and pro-oxidant remodeling of the HDL proteome in a mouse model of NAFLD (29). These changes were associated with impaired HDL functions in vitro. Although very little is known about the HDL composition and function in patients with biopsy-proven SS and NASH, it has been shown that NAFLD is associated with altered distribution of HDL proteins (40) and smaller HDL particles (5). Changes in HDL composition may alter cardioprotective functions of HDL in NAFLD. Although lipid-free ApoAI and small-density HDL particles are the most efficient mediators of ATP-binding cassette family transporter ABCA1-dependent cholesterol efflux (11), it is unknown how HDL’s particle distribution affects HDL flux. Systemic inflammation in several conditions, including coronary artery disease, is characterized by impaired HDL metabolism and functionality (37). However, it is unknown whether hepatic inflammation in NASH impacts HDL composition and function and if these changes contribute to increased CVD risk in NASH.

Although cholesterol efflux and other ex vivo assays measure different functions of HDL, currently it is challenging to measure HDL functions in vivo in humans. It has been recognized that the efficiency of RCT depends on HDL flux (50). Recently, we developed a ²H₂O-metabolic labeling method to measure integrated HDL flux in mice in vivo and established its utility in humans (17). In the present studies, we used the ²H₂O method to simultaneously quantify HDL (ApoAI and HDLc) turnover and the dynamics of several other HDL proteins as surrogate indices of dynamic HDL functions in patients with NAFLD. We also used this approach to assess the effect of the disease severity on HDL proteome dynamics and functions in vivo. We hypothesized that NAFLD would alter HDL metabolism and functions, which will be more noticeable in NASH.

METHODS

Details regarding materials and methods are provided in the Supplemental Materials (all Supplemental material for this article is available online at https://doi.org/10.6084/m9.figshare.8114186).

Clinical Studies

The Institutional Review Board at both the Cleveland Clinic and NEOMED reviewed and approved these protocols. The healthy control subjects were recruited from the general population of the Greater Cleveland area, Ohio, by means of flyers. They underwent a medical history and physical examination. Real-time sonography was performed by hepatologists (S. Dasarathy and A. McCullough), and based on that, eligible healthy subjects with liver fat ≤5% (n = 9) were enrolled to study. Patients with biopsy-proven NAFLD (7 steatosis and 8 NASH) were recruited from outpatient clinics of the Cleveland Clinic and Metro Health Medical Center, Cleveland, Ohio. Participants with established CVD, i.e., coronary stenting, coronary artery bypass graft surgery, transient ischemic attack, cerebrovascular events, myocardial infarction, atrial fibrillation, or congestive heart disease, were excluded. Subjects were also excluded if they had a history of alcohol abuse (≥30 g/day for men and ≥20 g/day for women), diabetes, or any liver disease other than NAFLD (i.e., Wilson’s disease, hepatitis B and C, cirrhosis, or drug-induced hepatitis), history of bariatric surgery, pregnancy, lactation, inflammatory diseases associated with HDL dysfunction (AIDS, HIV-positive, sepsis, rheumatoid arthritis, renal failure, etc.). In addition, we excluded all subjects on any lipid-lowering drugs (statins, fibrates). Patients with NAFLD with hypertension were on antihypertension medications (lisinopril, amlodipine, nadolol, metoprolol, and hydrochlorthiazide). All subjects were advised to avoid strenuous exercise and to consume an isocaloric diet to prevent any diet- and exercise-induced changes in HDL metabolism.

Written informed consent was obtained from each participant. All studies adhered to the requirements of the United States Federal Policy for the protection of human subjects (45 Code of Federal Regulations, part 46) and the general ethical principles of the Declaration of Helsinki.

Kinetic Study

Studies were conducted in the clinical research unit at the Cleveland Clinic. On the study day, each subject was admitted to the clinical research unit at 7:00 AM after a 12-h fast. An intravenous catheter was placed into their dorsal hand vein for blood sampling, and a baseline sample was drawn. At 8:00 AM, each subject consumed 70% ²H₂O in their drinking water (4.0 mL ²H₂O/kg body mass) in 5 doses over a 4-h period. Arterialized blood samples (10 mL) were obtained at 0, 2, 4, 5, 6, 7, 8, and 10 h after the initial dose of ²H₂O for serum and plasma isolation as previously described (10). Study subjects continued to take daily doses of ²H₂O (10% of loading dose/day) and reported for the scheduled overnight fasted blood draws on days 1, 2, 3, 4, and 7 after the initiation of the kinetic study. Serum samples were immediately processed for the analyses of HDL kinetics, and the remaining serum and plasma were stored at −80°C for further analysis.

Analytical Procedures and enzyme activity assays

Fasting plasma glucose, triglycerides (TGs), total cholesterol, and HDLc were measured by enzymatic analysis on an automated platform (Roche Modular Diagnostics, Indianapolis, IN). HDL-CE was quantified in serum using the cholesterol assay kit (Invitrogen Corporation, Carlsbad, CA). Plasma insulin concentration was determined by radioimmunoassay (Diagnostic Products, Los Angeles, CA). ApoAI and apolipoprotein B 100 (ApoB 100) were quantified by immunoassay methods on an Integrated Analyzer System (Abbott Laboratories, Abbott Park, IL). The serum levels of high-sensitivity C-reactive protein (hsCRP) were measured in the clinical chemistry laboratory of the Cleveland Clinic. IL-6 and TNF-α levels were quantified with plasma samples using a high-sensitive sandwich ELISA kit from R&D System Inc. (Minneapolis, MN) per manufacturer’s instructions. The acyltransferase activity of paraoxonase 1 (PON1), an HDL-associated antioxidant protein involved in the prevention of LDL oxidation, was assayed spectrophotometrically as previously described (22). Lecithin-cholesterol acyltransferase (LCAT) activity was measured using an assay kit (Roar Biomedical, New York, NY) per manufacturer’s protocol (22). CETP activity was determined using a fluorometric assay kit II (BioVision Incorporated, Milpitas, CA). Lipid peroxidation in plasma and HDL, thiobarbituric acid reactive substances (TBARS), including malonyl dialdehyde, were quantified in EDTA plasma and isolated HDL samples, respectively, using the TBARS Assay Kit (Cayman Chemical Co., Ann Arbor, MI) as previously described (17). All assays were performed in duplicate.

Cholesterol Efflux

Murine RAW264.7 cells (American Type Culture Collection, Manassas, VA) were incubated in 48-well plates and maintained in DMEM containing 1% FBS. Cells were loaded with ³H-cholesterol (0.50 μCi/mL; American Radiolabeled Chemicals, St. Louis, MO) in media for 24 h at 37°C and subsequently chased with media containing patient-derived ApoB-depleted serum (2%, vol/vol) for 4 h at 37°C (31). In parallel, a second set of cells was pretreated with 8-Br-cAMP (0.3 mM final, Sigma-Aldrich) to induce ABCA1-dependent efflux. Radioactivity in the media, after spinning down cells, was measured directly. Radioactivity in the cells was determined after extraction with hexane-isopropanol (3:2). Cholesterol efflux was determined as previously described using the equation: Efflux = ³H disintegrations/min in medium/(³H disintegrations/min in medium + ³H disintegrations/min in cells) (31). The ABCA1-dependent efflux was calculated as the difference between efflux in the presence of 8-Br-cAMP (total efflux) and the absence of 8-Br-cAMP (ABCA1-independent efflux).

HDL Isolation, Particle Size, and Proteomics

Anti-HDL polyclonal IgY resin column (GenWay BioTech, San Diego, CA) was used to isolate HDL (22). The protein concentration of the immunocaptured HDL fraction was measured by the bicinchoninic acid protein assay method. To analyze HDL proteins, after extracting the lipids with cold acetone, protein pellets (20 μg) were resuspended, disulfide bonds were reduced with 2.5 μL of 0.1 M DTT at 60°C for 30 min. Free thiol groups were alkylated with 2.5 μL of 0.2 M iodoacetamide at room temperature for 30 min. Proteins were digested by addition of 10 μL of 0.1 μg/μL Promega sequencing grade modified porcine trypsin (20 μg lyophilized trypsin in 200 μL 50 mM ammonium bicarbonate solution) at 37°C overnight. Protein digests were cleaned up through solid phase extraction using the Pierce Pepclean C18 spin columns, and tryptic peptides were analyzed by LC-MS/MS (30). HDL proteins were cross-referenced against the HDL Proteome Watch Initiative database (available online at http://homepages.uc.edu/~davidswm/HDLproteome.html).

HDL particles were separated by size using nondenaturing polyacrylamide gel electrophoresis on a 4%–20% Mini-Protean TGX Precast gel (BioRad) to fractionate pre-β, light HDL₂, and small-density HDL₃ particles, with distinct compositions as previously described (22). For Western blotting, gels were electroblotted onto the PVDF membrane and incubated with an anti-ApoA1 antibody.

HDL Turnover

HDL turnover was quantified in ApoB-depleted serum using the ²H₂O-metabolic labeling approach as described earlier (24). Briefly, after removal of VLDL by ultracentrifugation, the remaining ApoB-containing particles were precipitated with a magnesium chloride/dextran sulfate reagent (Stanbio Laboratory, Boerne, TX). The supernatant containing ApoB-depleted serum was recovered and used for the analysis of both HDLc and HDL proteins. After extraction of lipids, HDLc was analyzed by GC-MS (24), and proteins were analyzed using LC-MS/MS (22).

Data Analysis

Database searching.

For protein identification, higher-energy collisional dissociation-generated MS/MS spectra were searched by Mascot software (Matrix Science, London, UK, version 2.5.1) against the National Center for Biotechnology Information Human reference sequence database (ftp://ftp.ncbi.nih.gov/refseq/). The mass tolerances for the precursor and product ions were 10 ppm and 0.04 Da, respectively. Proteins were characterized based on multiple unique peptides at 99% confidence and false discovery rate of 1%.

Isotope incorporation was assessed based on mass isotope distribution analysis of the high-resolution full-scan spectra (23). The kinetics of a protein were analyzed using the isotopic distribution of its tryptic unique peptides. The proteomics data have been deposited to the ProteomeXchange Consortium (57) via the PRIDE partner repository (56) with the dataset identifier PXD 010199.

Calculations.

The fractional catabolic rates (FCRs) of HDLc, ApoAI, and other HDL proteins were determined based on a single compartmental model by fitting the time course ²H-labeling of analytes to an exponential growth curve equation (22). This model assumes a steady-state HDL metabolism during the study as shown previously (24) and allows determination of the rate constant (k) and the half-life (t_1/2 = ln2/k). Under these conditions, the FCR is equal to the fractional synthesis rate. The production rates (PRs) of HDLc and ApoAI were calculated as the product of their FCR and pool size:

$$\text{PR}\left(\text{g}\times {\text{kg}}^{-1}\times {\text{h}}^{-1}\right)=\text{pool\hspace{0.17em}size}\times \text{FCR}.$$

The pool size (absolute content) in circulation is the product of HDLc or ApoAI concentrations and plasma volume, estimated as 4.5% of the body weight.

Data presentation and statistical analysis.

The average of duplicate GC-MS and LC-MS injections, which differed less than 2%, was used for mass-spectrometric analyses. Proteomics data are shown as the excess labeling of multiple unique peptides (n = 2–11) of each protein in a given human experiment. Error bars represent standard deviations based on biological variabilities between subjects. Continuous variables were evaluated for normality graphically and then using the Shapiro–Wilk test. Categorical factors were compared between groups using the Fisher exact test. Parameters showing strong departures from normality were compared between groups defined by NAFLD status using nonparametric Wilcoxon rank sum tests and were summarized with medians. Normally distributed continuous measures were summarized using means and standard deviations and were compared between control and NAFLD groups using two-sample t tests. Multiple group comparisons using ANOVA with Bonferroni post hoc testing was used to compare the data from control subjects and subjects with SS and NASH. Mean differences between groups with 95% confidence intervals were presented.

RESULTS

Study Population

The clinical characteristics of control subjects and subjects with NAFLD are shown in Table 1. The adult healthy control subjects were matched to the patients with NAFLD by sex, but they were relatively younger [50 (35, 66) vs. 58 (43, 65), P = 0.088). Patients with NAFLD had higher body mass index (P = 0.016) and higher systolic blood pressure (P < 0.005), and 75% of them had hypertension (P < 0.001). As expected, patients with NAFLD also had higher serum transaminases. Control subjects and subjects with NAFLD did not have diabetes and had similar fasting blood glucose (P = 0.19). However, patients with NAFLD compared with healthy controls had significantly higher glycated hemoglobin (5.5 ± 0.3 vs. 6.0 ± 0.4, P = 0.007), ~4-fold higher mean plasma insulin levels (P < 0.001), and greater insulin resistance [homeostatic model assessment 1.8 (1.0, 4.1) vs. 7.6 (3.2, 22.1), P < 0.001]. There were no significant differences in fasting lipid values, including HDLc (P > 0.05). Subjects with NAFLD and control subjects had similar ApoAI levels; however, ApoB100 levels and ApoB/ApoAI ratio were significantly higher in patients with NAFLD (P < 0.05). Subjects with NAFLD had lower levels of adiponectin but higher levels of inflammatory cytokine (IL-6) and hsCRP, which was associated with a greater Framingham Risk Score, a gender-specific estimate of 10-yr cardiovascular risk (all P < 0.05, Table 1).

Table 1.

Clinical and biochemical characteristics of study participants

Characteristics	Control, n = 9	NAFLD, n = 15	P Value
Demographics
Age, yr	50.0 (35.0, 66.0)	58.0 (43.0, 65.0)	0.088a
BMI, kg/m2	30.8 (23.7, 35.4)	34.7 (26.9, 48.9)	0.016a
Male sex, %	44.4	46.7	>0.99c
Systolic blood pressure, mmHg	120.3 ± 1.9	138.0 ± 11.2	0.025b
Diastolic blood pressure, mmHg	75.9 ± 8.4	80.9 ± 8.6	0.186b
Hypertension			<0.001c
No	8 (89)	4 (25)
Yes	1 (11)	12 (75)
AST, IU/mL†	17.0 (11.0, 43.0)	33.0 (16.0, 87.0)	0.04a
ALT, IU/mL†	17.2 ± 8.3	45.8 ± 25.6	0.001b
Total bilirubin, mg/dL†	0.4 ± 0.2	0.6 ± 0.3	0.247b
Insulin resistance
Glucose, mg/dL	103.2 ± 14.7	104.4 ± 15.0	0.189b
Insulin, mIU/dL	7.6 (4.1, 16.5)	34.0 (13.0, 75.0)	<0.001a
HOMA index	1.8 (1.0, 4.1)	7.6 (3.2, 22.1)	<0.001a
HbA1c, %	5.5 ± 0.3	6.0 ± 0.4	0.007b
Lipids and lipoproteins
Cholesterol, mg/dL	168.8 ± 43.3	176.7 ± 30.9	0.629b
TGs, mg/dL	91.0 (50.0, 173.0)	127.5 (62.0, 252.0)	0.138a
HDLc, mg/dL	49.7 ± 15.5	45.1 ± 12.0	0.463b
HDL-CE, mg/dL	38.9.3 ± 13.0	41.8 ± 15.2	0.431b
HDL-TG, mg/dL	12.4 (6.3, 20.1)	15.3 (8.9, 38.7)	0.245a
HDL-TG/HDL-CE	0.2 (0.1, 0.5)	0.4 (0.2, 0.8)	0.024
LDL cholesterol, mg/dL	103.9 ± 26.6	102.3 ± 28.7	0.896b
ApoB100, mg/dL*	76.5 ± 12.5	96.7 ± 24.9	0.024b
ApoAI, mg/dL	128.3 ± 20.4	129.2 ± 13.3	0.906b
ApoB100/ApoAI	0.60 ± 0.09	0.75 ± 0.19	0.022b
ApoAI/HDLc	2.5 ± 0.4	3.2 ± 0.7	0.040b
LCAT, RFU at 470 nm	37,080.1 ± 1,674.0	36,666.3 ± 1,690.0	0.111b
Inflammatory markers
Adiponectin, µg/mL	9.5 (2.7, 12.9)	2.6 (1.4, 3.4)	<0.001a
IL-6, pg/mL*	1.0 ± 0.5	2.2 ± 1.0	0.001b
TNF-α, pg/mL*	1.4 ± 0.2	1.6 ± 0.6	0.18b
hsCRP, mg/L	1.4 (0.2, 3.6)	5.0 (0.6, 33.0)	0.045a
Framingham Risk Score, %	0.4 (0.0, 6.7)	2.9 (0.6, 15.7)	0.003a

Statistics are presented as means ± SD, median (P25, P75), or n (column %). The bold P values indicate P < 0.05. ALT, alanine transaminase; AST, aspartate aminotransferase; BMI, body mass index; GGT, gamma-glutamyl transpeptidase; HDL-CE, HDL cholesterol ester; HDL-TG, HDL triglyceride; HOMA, homeostatic model assessment; hsCRP, high-sensitivity C-reactive protein; LCAT, lecithin-cholesterol acyltransferase; NAFLD, nonalcoholic fatty liver disease; RFU, relative fluorescence unit.

^aKruskall–Wallis test,

^b2-tailed t test,

^cChi-squared test,

^*Measured in 8 healthy controls and 14 patients with NAFLD,

^†Measured in 6 healthy controls and 15 patients with NAFLD.

HDL Proteome and Particle Sizes

Proteomics analysis identified 78 proteins in all healthy control subjects and patients with NAFLD (Supplemental Table S1). The majority of these proteins were identified as the HDL proteins using other isolation methods (45). To assess the effect of NAFLD on HDL particle size, we separated HDL by size using nondenaturing polyacrylamide gel electrophoresis, which allowed separation of small pre-β HDL, mid-sized HDL_3, and large HDL₂ particles (Supplemental Fig. S1). Mean HDL particle size analysis demonstrated that HDL from patients with NAFLD was significantly smaller compared with HDL from control subjects. In particular, patients with NAFLD displayed increased levels of small-density HDL₃ particles. These results are consistent with recent reports showing reduced HDL_2β in patients with NAFLD compared with healthy controls (4, 9, 47).

HDL Turnover

To evaluate the effect of NAFLD on HDL turnover, we used a 4.0-mL/kg ²H₂O bolus followed by a 10% maintenance dose for 1 wk. This dose resulted in 0.79 ± 0.07% and 0.91 ± 0.08% ²H₂O body water enrichments in control subjects and subjects with NAFLD, respectively (Supplemental Fig. S2). ²H-labeling of HDLc and ApoAI in subjects with NAFLD increased at a higher rate than that in controls, indicating increased turnover (Fig. 1 and Supplemental Fig. S3). Indeed, patients with NAFLD had a significantly higher FCR of HDLc (0.16 ± 0.01 vs. 0.34 ± 0.08 pool/day, P < 0.05) and ApoAI (0.26 ± 0.04 vs. 0.33 ± 0.06 pool/day, P < 0.001) compared with the controls (Table 2). Despite their greater FCR, HDLc, and ApoAI, pool sizes were not altered in patients with NAFLD. However, the production rates of ApoAI and HDLc were higher in patients with NAFLD compared with the healthy controls, suggesting that the increase in ApoAI (P < 0.05) and HDLc (P = 0.053) production rates had compensated for their increased clearance.

Fig. 1.

HDL turnover in patients with nonalcoholic fatty liver disease (NAFLD; n = 14) and healthy controls (n = 8) was determined using a ²H₂O-metabolic labeling approach. Administration of a bolus dose (4 mL/kg) followed by maintenance doses (10% of bolus/day) of ²H₂O in drinking water resulted in a steady-state body water enrichment of ~0.8%–0.9%. Time course labeling of total HDL cholesterol (HDLc) (A) and tryptic ApoAI peptide VSFLALEEYTK (B). Data represent mean ± SD.

Table 2.

HDL cholesterol and ApoAI kinetic parameters in NAFLD patients (n = 15) and healthy controls (n = 9)

Parameters	Control	NAFLD	P Value
HDLc pool, mg/kg	23.5 ± 6.1	19.1 ± 4.2	0.076
HDLc FCR, pool/day	0.16 ± 0.09	0.34 ± 0.18	<0.05
HDLc PR, mg·kg−1·day−1	3.5 ± 1.8	6.4 ± 3.5	0.053
ApoAI pool, mg/kg	57.7 ± 8.3	64.4 ± 17.3	0.347
ApoAI FCR, day	0.26 ± 0.04	0.33 ± 0.06	<0.005
ApoAI PR, mg·kg−1·day−1	14.7 ± 3.4	19.4 ± 3.9	<0.05

Data are means ± SD. HDLc turnover was calculated in 8 healthy controls and 14 patients with NAFLD. The bold P values indicate P < 0.05. P values are based on 2-tailed t tests. FCR, fractional catabolic rate; HDLc, HDL cholesterol; NAFLD, nonalcoholic fatty liver disease; PR, production rate.

HDL Proteome Dynamics

Like ApoAI, higher FCR was observed for several other proteins in patients with NAFLD compared with control subjects (Table 3). The FCR of proteins involved in lipid metabolism, ApoAII, ApoAIV, and vitamin D-binding protein (P < 0.05) were significantly higher in subjects with NAFLD (all P < 0.05), which resulted in the reduced stability (shorter half-life) of these proteins. In contrast, the half-lives of the acute-phase response protein complement C3 and the antioxidant PON1 were significantly higher (Table 3), whereas the half-lives of transthyretin, hemopexin, haptoglobin, and several other proteins were not affected because of NAFLD.

Table 3.

Half-lives of HDL proteins in healthy controls (n = 9) and patients with NAFLD (n = 15)

Accession Number	Protein	GO-Molecular Function	Control t½, h ± SD	NAFLD t½, h ± SD
Immune response
P01024	Complement C3	Endopeptidase inhibitor	59.6 ± 20.7	247.3 ± 96.8‡
P10909	Apoprotein J (Clusterin)	Misfolded protein binding	26.6 ± 9.4	24.0 ± 8.8
Antioxidant
P27169	Paraoxonase 1	Antioxidant	260.8 ± 65.6	385.7 ± 122.6*
Lipid metabolism
P02647	Apolipoprotein A-I	Cholesterol transporter	66.7 ± 11.5	51.6 ± 8.9†
P02652	Apolipoprotein A-II	Cholesterol transporter	87.8 ± 24.8	46.4 ± 11.7‡
P06727	Apolipoprotein A-IV	Antioxidant, cholesterol transporter	37.7 ± 7.6	19.9 ± 4.4‡
P02766	Transthyretin	Thyroid hormone-binding protein	39.6 ± 14.4	30.1 ± 9.8
Vitamin transport
P02774	Vitamin D-binding protein	Actin-binding, vitamin transporter activity	45.8 ± 12.12	34.4 ± 5.9*
Acute-phase response
P01009	Alpha-1-antitrypsin	Serine-type endopeptidase inhibitor	40.7 ± 13.4	38.5 ± 11.5
P02765	Alpha-2-HS-glycoprotein	Cysteine-type endopeptidase inhibitor	59.7 ± 30.5	48.4 ± 10.5
P02790	Hemopexin	Metal ion-binding	49.4 ± 12.2	47.9 ± 9.0

Data are means ± SD. NAFLD, nonalcoholic fatty liver disease. Different from controls:

^*P < 0.05,

^†P < 0.005,

^‡P < 0.001 (t test).

HDL Functions

Insulin resistance and inflammation are associated with altered HDL functions (43, 48). We therefore examined the effect of NAFLD on cholesterol efflux and antioxidant properties of HDL. Surprisingly, total cholesterol efflux capacity from macrophages to ApoB-depleted serum of patients with NAFLD was slightly higher compared with healthy controls (Fig. 2A).

Fig. 2.

Effect of nonalcoholic fatty liver disease (NAFLD) on HDL metabolism and functions. Total cholesterol efflux and oxidative stress are increased, but antioxidant activity of HDL is reduced in patients with NAFLD. A: ApoB-depleted sera from patients with NAFLD and healthy controls were investigated for their ability to promote [³H]cholesterol efflux from RAW264.7 macrophages. B: serum paraoxonase 1 (PON1) activity was measured spectrophotometrically using paraoxon as a substrate. C: thiobarbituric acid reactive substances (TBARS) in EDTA plasma and ApoB-depleted plasma were measured to quantify lipid peroxidation in total plasma and HDL. D: cholesterol ester transfer protein (CETP) activity was measured in ApoB-depleted plasma. Data represent mean ± SD (n = 8 for control group; n = 14 for NAFLD group). *P < 0.05 and **P < 0.005.

Serum PON1 activity, a marker of the antioxidant property of HDL, in subjects with NAFLD were significantly lower compared with healthy individuals (P < 0.05) (Fig. 2B). These changes were associated with higher levels of lipid peroxidation products (TBARS) in both total plasma and HDL (Fig. 2C).

To determine if altered lipidation is involved in higher HDL turnover and functions in NAFLD, we also measured LCAT and CETP activities. LCAT is a critical enzyme involved in free cholesterol esterification that facilitates HDL maturation. The enzyme CETP transfers CE from HDL to the ApoB-containing particles in exchange with TGs. Although no differences were observed in serum LCAT activities between subjects with NAFLD and controls, CEPT activity was significantly higher (P < 0.05) in patients with NAFLD (Fig. 2D) and was associated with greater ApoAI/HDLc and HDL-TG/HDL-CE ratios (P < 0.05; Table 1), indicating alterations in lipid and protein composition of HDL in patients with NAFLD. Furthermore, subjects with NAFLD had reduced levels of adiponectin but increased levels of inflammatory cytokine IL-6 and hsCRP, which was associated with an increased Framingham Risk Score, a gender-specific estimate of 10-yr cardiovascular risk (all P < 0.05, Table 1).

Role of the Disease Severity in HDL Metabolism

To assess the effect of the disease severity on HDL metabolism, patients with NAFLD were stratified into SS and NASH groups (Table 4) according to their liver histology (Table 5, Supplemental Table S2). The SS and NASH groups were matched for body mass, body mass index, and age, with no significant differences in blood pressure, fasting blood glucose, HbA1c, insulin, and insulin sensitivity (P > 0.05). Compared with patients with SS, those with NASH had higher plasma aspartate aminotransferase, alanine transaminase, gamma-glutamyl transpeptidase, and TNF-α levels. Interestingly, patients with NASH had lower IL-6 levels. Similarly, lower IL-6 levels have been reported in patients with NASH with hepatic iron overload (20). Surprisingly, patients with NASH had significantly higher plasma adiponectin levels compared with those with SS (2.1 ± 0.5 vs. 2.9 ± 0.4, P = 0.005, Table 4). Although TG levels were significantly higher in patients with NASH (P < 0.05), there were no differences observed in other plasma lipids (total cholesterol, LDL cholesterol, and HDL-TG) and lipoproteins (ApoAI and ApoB) between NASH and SS. Of note, we observed a trend toward lower HDLc and HDL-CE levels in patients with NASH, which did not reach the significance, possibly because of the sample size. Although the median Framingham Risk Score was higher in patients with NASH compared with SS, the differences between groups were not significant (P = 0.124).

Table 4.

Clinical and biochemical characteristics of patients based on the presence of NASH

Characteristics	SS, n = 7	NASH, n = 8	P Value
Demographics
Age, yr	56.2 ± 6.7	60.0 ± 4.5	0.314†
BMI, kg/m2	37.7 ± 7.9	33.7 ± 2.5	0.222†
Male, %	2 (29)	5 (63)	0.315§
Insulin resistance
Glucose, mg/dL	104.6 ± 11.1	106.8 ± 12.7	0.555†
Insulin, mIU/dL	33.7 ± 17.9	36.0 ± 19.2	0.831†
HOMA index	8.5 (3.2, 22.1)	7.1 (3.3, 21.7)	0.924‡
HbA1c, %	6.0 ± 0.4	5.9 ± 0.3	0.682†
Lipids and lipoproteins
Triglycerides, mg/dL	99.3 ± 32.1	168.1 ± 63.8	0.035†
Cholesterol, mg/dL	179.9 ± 19.1	174.4 ± 39.2	0.803†
LDL cholesterol, mg/dL	104.9 ± 17.4	103.3 ± 35.5	0.928†
HDL cholesterol, mg/dL	51.2 ± 14.4	40.5 ± 5.5	0.098†
HDL cholesterol ester, mg/dL	41.8 ± 12.3	32.9 ± 6.9	0.127†
HDL triglyceride, mg/dL	15.5 (8.9, 38.7)	15.3 (9.1, 30.4)	0.954†
ApoB100, mg/dL*	100.3 ± 20.7	93.6 ± 26.1	0.647†
ApoAI, mg/dL	133.5 ± 13.7	125.6 ± 10.6	0.304†
ApoB100/ApoAI*	0.76 ± 0.19	0.74 ± 0.17	0.823†
ApoAI/HDLc	3.20 ± 1.04	3.17 ± 0.37	0.962
LCAT, RFU at 470 nm	3,6202 ± 1,670	3,7071 ± 1,708	0.338
Liver enzymes
AST, IU/L	27.0 (16.0, 80.0)	40.0 (31.0, 87.0)	0.02‡
ALT, IU/L	32.1 ± 26.6	57.8 ± 14.8	0.048†
GGT, IU/L	1.3 ± 0.7	4.0 ± 0.8	<0.001†
Inflammation
Adiponectin, μg/mL	2.1 ± 0.5	2.9 ± 0.4	0.005†
IL-6*	2.8 ± 0.9	1.6 ± 0.5	0.021†
TNF-α*	1.3 ± 0.4	1.9 ± 0.5	0.045†
hsCRP, mg/dL	3.7 (0.6, 33.0)	5.0 (0.9, 13.2)	0.775‡
Framingham Risk Score, %	1.4 (0.6, 11.7)	5.3 (1.4, 15.7)	0.12‡

Statistics are presented as means ± SD, median (P25, P75), or n (column %). The bold P values indicate P < 0.05. ALT, alanine transaminase; AST, aspartate aminotransferase; BMI, body mass index; GGT, gamma-glutamyl transpeptidase; HOMA, homeostatic model assessment; hsCRP, high-sensitivity C-reactive protein; LCAT, lecithin-cholesterol acyltransferase; NASH, nonalcoholic steatohepatitis; RFU, relative fluorescence unit.

^*Measured in 7 patients with SS and 7 patients with NASH,

^†2-tailed t test,

^‡Kruskall–Wallis test,

^§Chi-squared test.

Table 5.

Histological characteristics of liver biopsies

Characteristics	Steatosis, n = 7	NASH, n = 8	P-Value
Steatosis			0.005
0	0 (0.0)	0 (0.0)
1 (<33% liver fat)	7 (100)	2 (25.0)
2 (33%–66% liver fat)	0 (0.0)	4 (50.0)
3 (>66% liver fat)	0 (0.0)	2 (25.0)
Inflammation			0.003
0 (%)	5 (71.4)	0 (0.0)
1 (%)	2 (28.6)	4 (50.0)
2 (%)	0 (0.0)	4 (50.0)
Ballooning			0.055
0 (%)	6 (85.7)	3 (37.5)
1 (%)	1 (14.3)	3 (37.5)
2 (%)	0 (0.0)	2 (25.0)
Fibrosis			<0.001
0 (%)	0 (0.0)	1 (12.5)
1, zona 3 perivenular, perisinusoidal or pericellular fibrosis, (%)	0 (0.0)	4 (50.0)
2, as grade 1 plus periportal fibrosis (%)	0 (0.0)	1 (12.5)
3, bridging fibrosis (%)	0 (0.0)	2 (25.0)

NASH, nonalcoholic steatohepatitis.

Despite higher turnover of HDL in both SS and NASH relative to controls (not shown), the disease severity did not have any effect on the kinetics of HDLc, ApoAI (Supplemental Table S3), and other quantified HDL proteins (data not shown). However, patients with NASH had slightly higher total and ABCA1-dependent cholesterol efflux capacities compared with both control subjects and subjects with SS, respectively (Fig. 3A, P < 0.05). The PON1 activity was lower in both patients with SS and NASH compared with the control subjects without any significant differences between SS and NASH groups (Fig. 3B). These changes were consistent with measures of systemic oxidative stress quantified by the levels of TBARS, lipid peroxidation products in total plasma and HDL (Fig. 3C). Although no differences were observed in LCAT activity between subjects with SS and NASH (Table 4), surprisingly, CETP activity was significantly higher only in the SS group compared with controls (Fig. 3D, P < 0.05). Size distribution analysis revealed a marked shift toward the medium-sized HDL₃ particles in patients with SS and relatively higher levels of pre-β HDL particles in individuals with NASH that was associated with lower expressions of HDL₂ particles (Supplemental Fig. S1).

Fig. 3.

Impact of the disease severity in nonalcoholic fatty liver disease (NAFLD) on HDL metabolism and functions. The data from subjects with NAFLD presented in Fig. 2 was divided into steatosis and nonalcoholic steatohepatitis (NASH) groups based on the liver histology data. A: cholesterol efflux. B: paraoxonase 1 (PON1) activity. C: thiobarbituric acid reactive substances (TBARS) measured to quantify lipid peroxidation in total plasma and HDL. D: cholesterol ester transfer protein (CETP) activity. Data represent mean ± SD (n = 8 for control group; n = 14 for NAFLD group). *P < 0.05, **P < 0.005 vs. controls. $$P < 0.005 vs. controls and steatosis.

However, no significant differences in HDL particle distribution were observed between NASH and steatosis cohorts. Overall, these results show that in patients with NAFLD, steatosis but not inflammation is the main determinant of altered HDL metabolism and functions.

DISCUSSION

We investigated the role of NAFLD and its severity in HDL metabolism and functions in patients with biopsy-proven SS and NASH in vivo. Our results show that the turnover rates of HDL proteins, including ApoAI and ApoAII, were significantly altered in patients with NAFLD compared with controls, with reduced antioxidant function of HDL. However, a paradoxical increase in cholesterol efflux function of HDL was observed in NAFLD that was perturbed more in subjects with NASH than those with SS. These alterations were associated with insulin resistance, changes in adiponectin levels, CETP activity, and remodeling of HDL composition and particle sizes, but their relative contributions to HDL dysfunction need further study. ApoAI turnover and cholesterol efflux in all subjects have been associated with the 10-yr risk of CVD, suggesting that insulin resistance-induced and steatosis-induced degradation ApoAI may contribute to CVD risk in NAFLD. Despite these perturbations in HDL in NAFLD, no differences were observed either in the kinetics of HDLc and HDL proteins or in the antioxidant function of HDL between subjects with SS and NASH.

Although ApoAI turnover analysis is widely used to study HDL metabolism, it has recently been shown that ApoAI and HDLc have distinct uptake mechanisms in different tissues (16), suggesting that combined ApoAI and HDLc kinetic studies are necessary to investigate the HDL flux relevant to RCT efficiency. Previous human studies of HDL turnover with stable isotopes utilized two different tracers, i.e., a labeled amino acid for ApoAI and acetate or cholesterol for HDLc turnover studies (38, 54). In this study, we measured both HDLc and ApoAI turnover in humans using a single ²H₂O tracer. In contrast to other tracers, ²H₂O rapidly equilibrates with the total body water and the intracellular precursors of cholesterol (acetyl-CoA and NADPH) and protein synthesis (amino acids). Since body water is easily accessible, this method avoids the need for assessment of the true intracellular precursor enrichment, which is a critical step in tracer studies (34). Furthermore, since ²H₂O is administered in drinking water, it is possible to study HDL metabolism in free living subjects without the need for in-patient intravenous tracer infusion. Direct comparison of the ²H₂O method against the commonly used amino acid tracer method demonstrated that ²H₂O provides an effective alternative to quantify protein turnover in rodents and humans (15, 59).

The ²H₂O metabolic labeling approach revealed that in addition to increased FCR of HDLc and ApoAI, their production rates were also increased, highlighting an overall increase in HDL flux in patients with NAFLD. Thus, an increased production of HDLc and ApoAI compensated for the accelerated clearance that resulted in their stable levels in NAFLD.

Although enhanced ApoAI catabolism is considered the major cause of lower HDLc levels in conditions associated with insulin resistance, the mechanisms of these processes are not fully understood. HDL metabolism is regulated by multiple factors, including LCAT- and CETP-catalyzed HDL remodeling, SR-B1-dependent selective uptake of HDLc by the liver, and renal clearance of lipid-poor ApoAI. Genetic deficiency of LCAT is associated with reduced ApoAI plasma levels because of rapid clearance of lipid-poor ApoAI from circulation (3). In this study, we observed no change in LCAT activity in subjects with NAFLD. Increased degradation of ApoAI in conditions associated with insulin resistance has been ascribed to TG enrichment of HDL in insulin-resistant states (28). CETP-induced exchange of core CEs in HDL with the core TGs in VLDL (21, 41) would mediate increased clearance of HDL. Insulin-sensitizing adiponectin may affect HDL metabolism via lipoprotein lipase-mediated HDL maturations from small to large particles, which would lead to decreased clearance of HDL (6, 14) and an increase in HDL synthesis (33). We also noted that greater HDL flux was associated with higher CETP activity but lower adiponectin levels. Importantly, these changes were also associated with higher CETP activity in patients with NAFLD (Fig. 2D), which resulted in higher HDL-TG/HDL-CE ratio (Table 1), suggesting that and CETP-induced remodeling of HDL, a key step in HDL maturation, was involved in enhanced degradation of ApoAI. It is known that transgenic expression of CETP in mice leads to formation of TG-rich HDL prone to lipolysis by serum lipases, which results in dissipation of ApoAI from particles (21), followed by its accelerated renal clearance (27). In NAFLD, adiponectin depletion and CETP-induced remodeling of HDL with consequent dissociation of ApoAI from HDL particles may potentially facilitate its degradation. We have shown that higher clearance of lipid-poor HDL in ApoE^−/− mice was associated with greater expression of hepatic SR-B1, suggesting that SR-B1-assisted uptake facilitates HDL catabolism (24). However, it is unclear whether hepatic-specific uptake of HDL is involved in its enhanced flux in humans.

In this study, higher turnover of HDL was associated with an unexpected increase in cholesterol efflux function of HDL in NAFLD and was more pronounced in patients with NASH. Cholesterol efflux capacity from macrophages to nascent HDL particles is the first and rate-limiting step in RCT (13). The seminal study by Khera et al. (25) demonstrated that lower cholesterol efflux capacity from free cholesterol-enriched macrophages is correlated with the history of CVD independent of HDLc levels, suggesting that efflux capacity measurements may have prognostic value. However, increases in cholesterol efflux capacity were observed in subjects who were overweight and insulin resistant (32, 35, 44) and in patients who were hypertriglyceridemic with type 2 diabetes mellitus (32, 60).

The reason for these conflicting findings is not clear, but they highlight the importance of future mechanistic studies on the regulation of cholesterol efflux. HDL is a heterogenous population of particles with different sizes and compositions. Differences in functional activities of distinct HDL particles could be related to their compositions. Indeed, using reconstituted HDL particles, it has been elegantly shown that small pre-β-HDL and mid-sized HDL₃ particles are the most efficient acceptors of cholesterol (11). Consistent with recent reports on alterations in HDL particles in NAFLD (5), we observed significantly higher expression of HDL₃ particles in patients with NAFLD and relatively higher levels of pre-β-HDL particles in patients with NASH. These changes were also associated with enhanced degradation of several HDL proteins, including ApoAI, that resulted in altered HDL particle composition. The changes were accompanied with reduced antioxidant activity of HDL and increased HDL lipid peroxidation. Thus, it is possible that these detrimental consequences of altered HDL particle composition with higher HDL catabolism countered the benefits of enhanced cholesterol efflux with an overall increase of CVD risk in patients with NAFLD. Despite the potential limitation of a relatively small sample size, these data lay the foundation for clinical translation of our observations, replacing HDL levels with functional assays.

Although the role of obesity, insulin resistance, and systemic inflammation on the increased clearance of ApoAI has been well studied (32, 35, 44), little is known about the effect of hepatic inflammation and fibrosis on HDL metabolism and CVD risk. Here, we studied the role of steatohepatitis on HDL metabolism and functions compared with hepatic fat accumulation alone in patients with biopsy-proven NASH and SS, respectively. Despite overall NAFLD-induced alterations in HDL metabolism, no significant differences were observed between SS and NASH. Although our patients with NASH only had a modest increase in hepatic inflammation compared with those with steatosis, our proof of concept study shows that NAFLD progression from steatosis to NASH does not further worsen obesity and insulin resistance-induced alterations in ApoAI stability and HDL flux in NAFLD. HDL metabolism is impaired by systematic inflammation (26), another systemic condition characterized by obesity and insulin resistance present in patients with NAFLD. However, in our study, we did not observe any significant impairment in systemic inflammation in patients with NASH compared with patients with SS. In contrast, patients with NASH in this study displayed decreased IL-6 but increased adiponectin levels, two adipokines with opposing effects on ApoAI and HDL metabolism (6, 19). Although the mechanisms of these reciprocal changes are not clear, it is possible that these changes combined with mild hepatic inflammation masked the NASH-induced perturbations in HDL kinetics and function. Our results are also consistent with a recent report that atherogenic dyslipidemia in NAFLD is driven by hepatic steatosis and insulin resistance, but not steatohepatitis (5).

Consistent with previous studies, demonstrating that serum HDL levels are neither representative of the exchanging pool of body cholesterol nor a marker of the RCT (2), our data show that HDLc or ApoAI levels do not reflect HDL functions. Our studies on HDL turnover quantifies the flow of cholesterol through RCT and determines the dynamic HDL functionality in vivo. It is important to recognize that the HDL turnover assay provides an integrated measure of total HDLc flux that cannot differentiate various steps of HDL biogenesis, remodeling, and clearance in the RCT pathway. Furthermore, HDLc freely exchanges between different HDL particles with distinct sizes and ApoB-containing lipoproteins. The FCR of HDLc measured in this study represents the integrated flux of HDLc and cannot discriminate specific fluxes between diverse particles. Furthermore, although direct comparison of the HDL turnover method against macrophage-RCT assay shows compatibility of these two methods (24), in contrast to RCT assay that measures macrophage-specific HDL flux, turnover rates of total HDLc measured in this study cannot distinguish specific fluxes from different organs, including macrophages. Nevertheless, the method presented in this study provides a useful and robust tool for the measurement of global HDL flux, which could be used to assess the effect of the disease and therapy on HDL functions in vivo.

Our study has several limitations. We suspect that the small number of subjects limited our ability to detect the differences in HDL turnover between SS and NASH groups. It is also possible that we missed the differences between groups because of the low-grade hepatic inflammation in our NASH cohort. Furthermore, the observed differences between groups need to be confirmed in a future large-scale study.

In conclusion, the ²H₂O metabolic labeling allows the detection of changes in HDL metabolism and HDL proteome in patients with NAFLD. HDL metabolism and function are altered in NAFLD without any significant differences between SS and NASH, suggesting that steatosis is a more important contributor of altered HDL metabolism and functions than hepatic inflammation in NAFLD. Finally, the ²H₂O-based metabolic labeling method is a simple and safe strategy to study HDL metabolism and kinetics in humans and can potentially provide insights into in vivo HDL functionality in other diseases also.

GRANTS

This study was supported by American Heart Association Grant 15GRNT25500004, American Diabetes Association Grant 1-15IN-31, and National Institutes of Health Grants 5-R01-HL-129120-03 and RO1-GM-112044.

DISCLAIMERS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

DISCLOSURES

S. F. Previs is employee at Merck & Co., Inc., and may own stock in the company. All other authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

T.K. conceived and designed research; J.D., K.L., S.D., and T.K. performed experiments; K.L., A.O., C.K., S.I., S.W.L., J.D.S., and T.K. analyzed data; A.M., S.F.P., S.D., and T.K. interpreted results of experiments; K.L., A.O., C.K., and T.K. prepared figures; T.K. drafted manuscript. A.M., J.D.S., S.D., and T.K. edited and revised manuscript; A.M., S.F.P., J.D., K.L., A.O., C.K., S.I., S.W.L., J.D.S., S.D., and T.K. approved final version of manuscript.