Liraglutide Increases the Catabolism of Apolipoprotein B100–Containing Lipoproteins in Patients With Type 2 Diabetes and Reduces Proprotein Convertase Subtilisin/Kexin Type 9 Expression

Diabetic dyslipidemia is a major contributor to the increased cardiovascular risk associated with type 2 diabetes (T2D) (1). Diabetic dyslipidemia includes abnormal metabolism of apolipoprotein B100 (apoB100)–containing lipoproteins (2,3). Indeed, patients with T2D show increased plasma triglycerides (TGs) as a result of both increased hepatic production of VLDLs and decreased catabolism of TG-rich lipoproteins (VLDL, IDL) (3,4). More precisely, overproduction of VLDL particles in T2D has been shown to predominate on the formation of larger TG-rich VLDL particles (VLDL₁) (5). In addition, decreased catabolism of LDL-apoB100 is observed in patients with T2D (4). Thus, abnormal metabolism of apoB100-containing lipoproteins is an important feature of diabetic dyslipidemia characterized by reduced fractional catabolism of apoB100-containing lipoproteins, leading to an increase in their plasma residence time, which is potentially harmful for the vascular wall. These metabolic abnormalities of apoB100-containing lipoproteins should be considered as a therapeutic target in patients with T2D to reduce their cardiovascular risk.

Treatment with liraglutide, a glucagon-like peptide 1 (GLP-1) receptor agonist, has been shown to significantly modify plasma lipids. Compared with placebo, 3 weeks of treatment with liraglutide (1.8 mg/day) in patients with T2D significantly reduced postprandial excursions of triglycerides and apoB48 after a high-fat meal and independently of gastric emptying (6). Our group has shown that liraglutide reduces postprandial hyperlipidemia by increasing apoB48 fractional catabolism and by reducing apoB48 production in patients with T2D (7). Liraglutide has also been shown to decrease significantly plasma levels of fasting TGs and LDL cholesterol (LDL-C) in several clinical trials (8,9). For instance, in the Liraglutide on Blood Glucose Control in Subjects With Type 2 Diabetes 4 (LEAD-4) study, liraglutide 1.2 mg/day reduced TGs significantly by 16% and LDL-C by 10% (9). Furthermore, a significant reduction of plasma apoB level has been noted in the LEAD-2 trial (8). In a meta-analysis of six clinical studies, liraglutide significantly improved lipid parameters (total cholesterol −0.13 mmol/L, TGs −0.20 mmol/L, LDL-C −0.20 mmol/L) from baseline (10).

Thus, clinical data indicate a significant drop of apoB100-containing lipoproteins on liraglutide treatment. However, so far, the mechanisms responsible for the reduction in apoB100-containing lipoproteins on liraglutide remain unknown. Because apoB100-containing lipoproteins are atherogenic lipoproteins, it seems important to understand how liraglutide modifies their metabolism. This prompted us to perform an in vivo kinetic study of apoB100-containing lipoproteins (VLDL₁, VLDL₂, IDL, and LDL), with [¹³C]leucine in patients with T2D to assess the effect of a 6-month treatment with liraglutide (1.2 mg/day) on their production and catabolism. In addition, we performed animal and in vitro studies to analyze the direct effect of liraglutide on apoB100-containing lipoprotein metabolism.

This prospective, single-center study was approved by our regional ethics committee (Dijon, France), and written informed consent was obtained from all patients before study inclusion.

For this study, we recruited 10 patients with T2D and typical diabetic dyslipidemia (defined by TGs >1.70 mmol/L and/or HDL cholesterol [HDL-C] <1.03 mmol/L in men and <1.29 mmol/L in women) for whom treatment with liraglutide was indicated because of poorly controlled diabetes (glycated hemoglobin A_1c [HbA_1c] >7%). These patients were treated with oral glucose-lowering agents (metformin alone in five patients, metformin + sulfonylureas in four patients, metformin + acarbose in one patient) for at least 6 months and had stable HbA_1c during the previous 6 months. Patients with LDL-C >4.90 mmol/L, kidney failure (creatinine clearance <30 mL/min), liver failure (international normalized ratio ≥1.5), hyper- or hypothyroidism, use of drugs known to affect lipid metabolism (corticosteroids, retinoids, antiproteases, estrogen, cyclosporin, glitazones, statins, fibrates, cholestyramine, ezetimibe, nicotinic acid, n-3, or phytosterols), treatment with dipeptidyl peptidase 4 inhibitors during the 3 previous months, or previous treatment with thiazolidinediones or any GLP-1 agonists were excluded.

Two kinetic studies were performed in each patient with T2D: the first one before initiation of the treatment with liraglutide and the second one after 6 months of treatment with liraglutide. On the day following the baseline kinetic study, treatment with liraglutide was started at 8:00 a.m. at an initial dose of 0.6 mg/day, which was uptitrated to 1.2 mg/day after 1 week. The 1.2 mg/day dose was maintained throughout the remainder of the study.

On the day before the kinetic study, after a 12-h fast, each patient was admitted to the diabetes ward in the morning to undergo a physical examination and blood sampling. The following day, a kinetic study was performed in the fed state. Food intake (1,700 kcal/day, 55% carbohydrate, 39% fat, and 7% protein) was fractionated into small portions, which were provided every 2 h starting 6 h before the tracer infusion up to the end of the study, to avoid variations in apolipoprotein plasma concentration, similar to the protocol previously implemented by our group (11,12) and others (13). The endogenous labeling of apoB100 was carried out by administration of L-[1-¹³C]leucine (99 atom %; Eurisotop, Saint Aubin, France) dissolved in 0.9% NaCl solution. At 8:00 a.m., each patient received intravenously a primed infusion of 0.7 mg/kg of tracer immediately followed by a 16-h constant infusion of 0.7 mg/kg/h. Blood samples were collected at 0, 0.5, 1, 1.5, 2, 3, 4, 6, 8, 10, 12, 14, 15, and 16 h after the primed infusion. Serum was separated by centrifugation at 3,000g for 10 min at 4°C. To avoid the influence of acute exercise on lipid metabolism, all patients were instructed to refrain from strenuous exercise 3 days before the kinetic study.

VLDL₁, VLDL₂, IDL, and LDL were isolated from plasma by gradient ultracentrifugation using an SW 41 rotor in an L90 apparatus (Beckman Instruments, Palo Alto, CA). VLDL₁ fractions were dialyzed against a 10 mmol/L ammonium bicarbonate buffer (pH 8.2) containing 0.01% EDTA and 0.013% sodium azide and then delipidated for 1 h at –20°C using 10 volumes of diethyl ether ethanol 3:1. ApoB100 from VLDL₁ was isolated by preparative SDS-PAGE: The delipidated apoB100-containing material was solubilized in 0.05 mol/L Tris buffer (pH 8.6) containing 3% SDS, 3% mercaptoethanol, and 10% glycerol and applied to a 3-mm-thick vertical slab gel (3% acrylamide). After staining with Coomassie blue R-250, apoB100 was cut from the gel and hydrolyzed in HCl 6 mol/L for 16 h at 110°C under nitrogen vacuum. Samples were then centrifuged to remove polyacrylamide. Supernatants were lyophilized in a Speed Vac (Savant Instruments, Farmingdale, NY). Lyophilized samples were dissolved in 50% acetic acid and applied to an AG-50W-X8 200–400-mesh cation exchange column (Bio-Rad, Richmond, CA), and amino acids were recovered by elution with NH₄OH 4 mol/L and lyophilized. VLDL₂, IDL, and LDL apoB100 were isolated by selective precipitation with butanol-isopropyl ether as previously described (12).

Analyses were performed on a TRACE 1300 gas chromatograph connected to an ISQ LT single-quadrupole mass detector (Thermo Fisher Scientific) equipped with a chemical ionization source operating in positive mode. Derivatized samples, as N-acetyl O-propyl esters, were dissolved in ethyl acetate, and 1 μL was injected at 250°C using the split mode (ratio 5:1, 6 mL/min). Separation was carried out on an HP-5MS 30-m × 250-μm × 0.25-μm column (Agilent Technologies) using helium (1.2 mL/min) as a carrier gas. The elution program was set up from 100°C to 300°C at 20°C/min. Temperatures of the transfer line and the ion source were fixed at 280°C and 275°C, respectively. Methane (1.5 mL/min) was used as the reacting reagent. Data were acquired in selected ion monitoring mode. N-acetyl n-propyl-[¹³C]leucine (charge/mass ratio 217.2) to N-acetyl n-propyl-leucine (charge/mass ratio 216.2) response ratios were calculated using Chromeleon 7.2.9 software (Thermo Fisher Scientific). [¹³C]Leucine enrichment was initially expressed in δ per 1,000 and converted to a tracer-to-tracee ratio before modeling (12).

The data were analyzed with the Simulation Analysis and Modeling II program (SAAM Institute, Inc.) by using a multicompartmental model (Supplementary Fig. 1). This model has already been used by our group and others for apoB100 kinetic studies performed with stable isotope constant infusion (12,14). A forcing function, corresponding to the VLDL₁-apoB100 plateau enrichment, was used to drive the appearance of leucine tracer into the various lipoprotein fractions (12,14). The delay compartment accounted for the time required for the synthesis and secretion of apoB100 into the plasma. Compartments 1 and 2 represent plasma VLDL₁-apoB100 and VLDL₁ remnants-apoB100, respectively. Compartments 11 and 12 represent plasma VLDL₂-apoB100 and VLDL₂ remnants-apoB100, respectively. Compartments 21 and 22 represent plasma IDL-apoB100 and IDL remnants-apoB100, respectively. Compartment 31 represents plasma LDL-apoB100.

Because the experiment was performed in the steady state, the fractional synthetic rate equaled the fractional catabolic rate (FCR). For VLDL₁-apoB100, VLDL₂-apoB100, and IDL-apoB100, direct FCR reflects uptake by the liver when indirect FCR reflects catabolism through the VLDL-IDL-LDL cascade. Production rates (PRs) of apoB100 in each lipoprotein fraction were normalized to body weight and calculated as follows: PR = apoB100 FCR(each lipoprotein) × apoB100 pool size / body weight, where apoB100 pool size was calculated by multiplying the apoB100 concentration in the lipoprotein fraction (VLDL₁,VLDL₂, IDL, or LDL) by the estimated plasma volume (4.5% of body weight). In obese patients (BMI ≥30 kg/m²), plasma volume was corrected as previously reported (12).

The Akaike information criterion was used to compare different models, and the model with the lowest Akaike information criterion value was chosen. The goodness of fit of the model was assessed by the analysis of the residuals with the runs test. In each patient, the criteria of convergence for the objective function have been met, assessing identifiability of the model.

Glycemia, total cholesterol, HDL-C, TGs, apoA-I, and serum apoB100 were quantitated on a Vista analyzer with dedicated reagents (Siemens Healthcare Diagnostics, Deerfield, IL). LDL-C was calculated using the Friedewald formula since serum TG levels were <3.8 mmol/L. HbA_1c was measured by high-performance liquid chromatography with a G8 HPLC Analyzer (Tosoh Bioscience, Tokyo, Japan). Free cholesterol, phospholipids, total proteins, lipoprotein fraction apoB100, and apoC-III concentrations were measured on the Vista analyzer using reagents from Diasys (Condom, France). Esterified cholesterol was calculated as the difference between total and free cholesterol expressed in grams per liter and multiplied by 1.68. Plasma proprotein convertase subtilisin/kexin type 9 (PCSK9) concentrations were measured in patients by a high-sensitivity quantitative sandwich enzyme immunoassay (Quantikine ELISA; R&D Systems Europe Ltd.). Plasma lipoprotein lipase (LPL) mass was measured in patients by ELISA (ALPCO, Salem, NH).

Insulin was quantified with a chemiluminescent method on an IMMULITE 2000 XPi analyzer (Siemens Healthineers) with dedicated reagents. The HOMA of insulin resistance (HOMA-IR) score, which was used to estimate the degree of insulin resistance, was calculated using the following equation: fasting serum glucose (mmol/L) × fasting serum insulin (μU/mL) / 22.5.

The experimental protocol was approved by the local ethics committee for animal experimentation, and official French regulations (#87848) for the care and use of laboratory animals were followed throughout the experimental period. For adipose tissue and liver gene expression, 3-week-old C57BL/6J male mice were purchased from Charles River Laboratories (Saint-Germain-Nuelles, France). Mice were randomly separated in two series of 10 animals receiving a high-fat diet (60% w/w total lipids ssniff diet #E15741-34). Animals were housed on a 12-h/12-h light/dark schedule at 22–23°C with ad libitum access to water and food. After 12 weeks, mice were housed individually and treated for 10 days with either saline or liraglutide (Victoza; Novo Nordisk A/S) at 1 mg/kg body weight. At the end the treatment, mice were fasted for 12 h and then anesthetized with 2% isoflurane gas. Plasma, liver, and periepididymal adipose tissue were collected, and mice were finally sacrificed by cervical dislocation.

For liver slices gene expression, 12-week-old ob/ob mice were purchased from Charles River Laboratories. They were housed on a 12-h/12-h light/dark schedule at 22–23°C with ad libitum access to water and chow diet. For liver slices preparation, animals were fasted for 12 h.

Ob/ob mice (14 weeks old) were anesthetized with intraperitoneal injection of ketamine/xylazine (7.5 mg/1 mg for 100 g body weight). The liver was perfused with cold and oxygenated Hanks’ balanced salt solution to clear the organ of blood before slicing using a Brendel/Vitron tissue slicer (Vitron, Tucson, AZ) in the same medium. Thin slices (∼200 μm) from each liver were rinsed and preincubated for 30 min at 37°C in Hanks’ balanced salt solution before being randomly distributed in six-well plates containing 3 mL of oxygenated William’s Medium E supplemented with heat‐inactivated calf serum (10%), antibiotic-antifungal cocktail (1%), and 15 mmol/L glutamine and containing either 250 μmol/L liraglutide or saline. The six-well plates were then installed on a rocking shaker and incubated for 21 h in a 5% CO₂ and 37°C atmosphere under slight agitation. At the end of the incubation period, slices were collected, and total RNA was extracted for gene expression.

Periepididymal adipose tissue samples were collected from control and liraglutide-treated mice and were immediately homogenized in 20 mmol/L Tris (pH 7.5) and 150 mmol/L NaCl and centrifuged at 2,000g for 10 min at 4°C to separate fat. Assays were performed immediately using a Fluorometric LPL Activity Assay Kit (Cell Biolabs, Inc., San Diego, CA) according to manufacturer instructions.

PCSK9 protein level was assessed using the Mouse Proprotein Convertase 9/PCSK9 Quantikine ELISA Kit (R&D Systems, Minneapolis, MN). Briefly, plasma was diluted and assayed by the ELISA sandwich technique. Plasma concentrations were estimated by calculation using a standard curve containing recombinant PCSK9.

Total mRNAs were extracted from liver and adipose tissue with TRI-Reagent (Euromedex, Souffelweyersheim, France) and were reverse transcribed using the iScript cDNA kit (Bio-Rad, Hercules, CA). Real-time RT-PCR was performed, as previously described (15), in a 96-well plate using a Bio-Rad iCycler iQ. The sequences of forward and reverse primers used for amplification are represented in Supplementary Table 1. For each gene, a standard curve was established from four cDNA dilutions (1/10–1/10,000) and used to determine relative gene expression variation after normalization with a geometric average of β-actin and ATPase expression.

For immunoblotting, liver was homogenized in radioimmunoprecipitation assay lysis buffer (NaCl 150 mmol/L, Triton 1%, sodium deoxycholate 0.5%, SDS 1%, Tris-HCl 50 mmol/L [pH 8]) containing a phosphatase and protease inhibitor cocktail. Protein concentration was determined with the BCA Protein Assay Kit (Sigma): 10 μg of proteins were resolved in gradient (4–20%) SDS-PAGE gels (Bio-Rad, Marnes-la-Coquette, France) and blotted onto nitrocellulose membranes with the Trans-Blot system (Bio-Rad). After 1 h in blocking buffer (BSA 5%), the membranes were incubated with primary antibodies recognizing LDL receptor (LDLr) and β-actin. Appropriate secondary antibodies conjugated to horseradish peroxidase were used for detection. Finally, proteins of interest were visualized by enhanced chemiluminescence with the ChemiDoc MP Imaging System after substrate incubation (Clarity; Bio-Rad). Bands were quantified by densitometric analysis performed by ImageLab software. LDLr antibody was purchased from Cell Signaling (Leiden, the Netherlands) and β-actin from Merck-Millipore (Darmstadt, Germany).

Data are reported as mean ± SD. Statistical calculations were performed using the SPSS software package (IBM Corporation, Chicago, IL). Continuous posttreatment data were compared with baseline data using the nonparametric Wilcoxon matched pair signed rank test. Continuous data between two groups were compared using the nonparametric Mann-Whitney U test. Correlations between quantitative parameters were calculated by the nonparametric Spearman test. A two-tailed probability level of 0.05 was accepted as statistically significant.

The clinical and biological characteristics of the patients at baseline and after 3 months of treatment with liraglutide are shown in Table 1. Treatment with liraglutide 1.2 mg/day was well tolerated in all patients and induced significant decreases in body weight, fasting glucose, HbA_1c, HOMA-IR, fasting TGs, total cholesterol, and apoB100. A slight decrease in LDL-C was also observed but did not reach statistical significance (P = 0.09). A significant increase in LPL mass as well as a significant decrease in plasma PCSK9 was observed after liraglutide treatment.

Composition of VLDL₁, VLDL₂, IDL, and LDL before and after 6 months of treatment with liraglutide is shown in Supplementary Table 2. After liraglutide treatment, we observed a significant decrease in free cholesterol in VLDL₁ and a significant increase in the esterified cholesterol/TG ratio in VLDL₂ and IDL and in phospholipids in IDL and LDL.

Kinetic parameters of VLDL₁-apoB100, VLDL₂-apoB100, IDL-apoB100, and LDL-apoB100 in the patients at baseline and after 6 months of treatment with liraglutide are shown in Table 2. VLDL₁-apoB100 and LDL-apoB100 pools were significantly reduced by 22% and 12%, respectively (Fig. 1). The isotopic enrichment curves of apoB100 in each lipoprotein (VLDL₁, VLDL₂, IDL, and LDL) expressed as percentage of plateau, are presented in Fig. 2 and show a faster increase in isotopic enrichment of apoB100 after liraglutide treatment, indicating faster FCRs (Table 2). We observed a significant increase in the total FCRs of VLDL₁-apoB100 (44%, P = 0.013), VLDL₂-apoB100 (80%, P = 0.013), and LDL-apoB100 (29%, P = 0.005) when the 25% increase in the total FCR of the IDL-apoB100 did not reach statistical significance (Fig. 1). The indirect apoB100 FCRs were significantly increased for all lipoproteins: VLDL₁-apoB100 (39%, P = 0.005), VLDL₂-apoB100 (82%, P = 0.008), and IDL-apoB100 (41%, P = 0.017). Direct FCRs of VLDL₁-apoB100, VLDL₂-apoB100, and IDL-apoB100 as PRs of VLDL₁-apoB100, IDL-apoB100, and LDL-apoB100 were not significantly modified by liraglutide treatment (Table 2). The increase in total FCRs and indirect FCRs of VLDL₁-apoB100, VLDL₂-apoB100, IDL-apoB100, and LDL-apoB100 were not correlated with age or with the reduction in body weight or HbA_1c. In addition, the changes in total FCRs and indirect FCRs of VLDL₁-apoB100, VLDL₂-apoB100, IDL-apoB100, and LDL-apoB100 were not different between men and women.

Adipose tissue LPL activity was significantly increased in liraglutide-treated mice compared with control mice (0.724 ± 0.207 vs. 0.454 ± 0.159 mU/min/mg of tissue, P = 0.024).

Mean plasma PCSK9 level was significantly lower in liraglutide-treated mice compared with control mice (278 ± 145 vs. 626 ± 383 ng/mL, P = 0.0003) (Fig. 3A).

To estimate whether the effects of liraglutide are mediated by mechanisms directly targeting lipoprotein metabolism, we studied the impact of liraglutide on the expression of several genes influencing lipoprotein metabolism. As shown in Fig. 4A, liraglutide significantly increased in diet-induced obese mice the expression of LPL gene and significantly reduced PCSK9, retinol-binding protein 4 (RBP4), and tumor necrosis factor-α (TNF-α) gene expression in the adipose tissue. A not statistically significant decrease of interleukin 1β (IL-1β) gene expression (P = 0.19) was also observed.

Because PCSK9 is predominantly expressed in the liver, we also studied, in vitro, the direct effect of liraglutide on the expression of the PCSK9 gene in the liver of ob/ob mice. Similarly to what we observed in the adipose tissue, liraglutide significantly reduced PCSK9 gene expression in the liver (Fig. 4B).

In the liver of diet-induced obese mice, liraglutide treatment significantly increased the expression of LDLr mRNA (Fig. 3B1) and reduced the expression of PCSK9 mRNA (Fig. 3B2). The liver protein expression of LDLr was significantly increased by liraglutide as shown in Fig. 3C1 and 3C2.

The present study provides new information on the effects of liraglutide on the metabolism of apoB100-containing lipoproteins in patients with T2D and diabetic dyslipidemia. We show that treatment with liraglutide induces a significant increase in the fractional catabolism of apoB100-containing lipoproteins. Moreover, we demonstrate in mice that a short-term treatment with liraglutide increases activity and gene expression of LPL, the major enzyme involved in the catabolism of TG-rich lipoproteins and reduces gene expression of molecules inhibiting LPL activity, such as TNF-α and RBP4. In addition, we show that liraglutide decreases the expression of PCSK9 mRNA and increases the protein expression of LDLr in the liver. Furthermore, our data indicate that liraglutide directly reduces gene expression of PCSK9, a major inhibitor of LDLr expression at the membrane surface.

Our group has previously shown that liraglutide reduces postprandial hyperlipidemia by increasing apoB48 catabolism and by reducing apoB48 production in patients with T2D (7). However, the effect of liraglutide on the metabolism of apoB100-containing lipoproteins remained unknown. Because abnormal metabolism of apoB100-containing lipoproteins is an important feature of diabetic dyslipidemia, it is important to know how it may be affected by liraglutide treatment.

Liraglutide has been shown to decrease significantly plasma levels of fasting TGs and LDL-C in several clinical trials (8,9). In the present study, we show that this effect is due to increased fractional catabolism of apoB100-containing lipoproteins. More precisely, we show that liraglutide induces a significant increase in the indirect fractional catabolism of VLDL₁-apoB100, VLDL₂-apoB100, and IDL-apoB100 as well as a significant increase in LDL-apoB100 fractional catabolism. This indicates that liraglutide treatment accelerates the catabolism of the whole VLDL-IDL-LDL cascade. This increased catabolism leads to significantly reduced pools of VLDL₁-apoB100 by 22% and LDL-apoB100 by 12%. Because both VLDL₁ and LDL are atherogenic, the significant reduction of VLDL₁ and LDL pools are of interest. The increased fractional catabolism of apoB100-containing lipoproteins is likely to be beneficial because it decreases their plasma residence time. Indeed, many data indicate that prolonged residence time of lipoproteins is likely to be harmful because it may promote lipid deposition in artery walls and increases the chances for oxidation and glycation of lipoproteins that are involved in atherogenesis in patients with T2D (3,16).

The observed increase in the fractional catabolism of apoB100 lipoproteins with liraglutide is in line with the augmented fractional catabolism of apoB48 lipoproteins reported after liraglutide treatment (7). The overall increased catabolism of apoB48-containing lipoproteins and apoB100-containing lipoproteins, such as VLDLs and IDLs, suggests an increase in the activity of LPL, the main enzyme involved in TG catabolism within TG-rich lipoproteins. This is supported by the increased activity and gene expression of LPL that we observed in mice after a short period of treatment with liraglutide as well as by the increase in LPL mass in our patients after liraglutide treatment. ApoC-III is associated with TG-rich lipoprotein metabolism because it decreases the catabolism of TG-rich lipoproteins by inhibiting the activity of LPL and by reducing their direct hepatic uptake and because it increases the assembly and secretion of VLDL in the liver (17). In the present study, apoC-III was not significantly decreased in our patients treated with liraglutide, suggesting that the increase in apoB100-containing lipoprotein catabolism is not likely to be induced by a decrease in apoC-III. Our kinetic data showing increased indirect FCRs of TG-rich lipoproteins (VLDL₁, VLDL₂, and IDL) suggest an increase in the activity of LPL that is likely to be due to other factors than apoC-III. Such dissociation between TG metabolism and apoC-III has been reported in other situations. For instance, it has been shown in a meta-analysis of randomized placebo-controlled trials with fenofibrate, which is known to accelerate TG catabolism, that there is no association between the apoC-III–lowering and the TG-lowering effects of fenofibrate, suggesting the influence of other factors than apoC-III on the catabolism of TG-rich lipoproteins (18). The reasons for the increased LPL activity on liraglutide are not totally clear. It has been shown that weight loss increases adipose tissue LPL activity in obese subjects (19). However, in our study, the increase in the catabolism of VLDL₁-apoB100 and VLDL₂-apoB100, which depends on LPL activity, was not correlated with body weight reduction, suggesting that weight loss is not likely to play a major role. LPL is also modulated by inflammation. For instance, TNF-α and IL-6 have been shown to reduce the expression of LPL at the transcriptional level and to decrease LPL activity in plasma (20). GLP-1 and GLP-1 agonists have anti-inflammatory properties. In vitro, exenatide reduces the secretion of TNF-α and IL-1β by stimulated monocytes/macrophages (21), and in ob/ob mice, administration of a recombinant adenovirus-producing GLP-1 reduces macrophage infiltration in the adipose tissue and the production of TNF-α, IL-6, and MCP-1 by the adipose tissue (22). In vivo, a significant drop of plasma TNF-α level has been observed after liraglutide treatment in patients with T2D (23). Moreover, in the present study, we show that liraglutide significantly reduced TNF-α gene expression in the adipose tissue. Thus, we may think that reduction of inflammation induced by liraglutide treatment could increase LPL activity. Furthermore, RBP4, a cytokine produced by the adipose tissue whose plasma level is increased in patients with T2D, has been shown to be an independent factor associated negatively with VLDL-apoB100 catabolism (24). Some data have suggested that RBP4 could reduce LPL activity (25). Thus, reduction of RBP4 expression induced by liraglutide could also partly explain increased catabolism of VLDL particles. Furthermore, we show in the current study that liraglutide treatment significantly increased plasma LPL mass in our patients with T2D. Therefore, all our data indicate that liraglutide enhances LPL expression and activity, resulting in the increased catabolism of VLDL₁, VLDL₂, and IDL observed in our patients with T2D.

Our study also shows that liraglutide increases LDL-apoB100 FCR, indicating enhanced catabolism of LDL lipoproteins. This could be an explanation for the decrease in plasma LDL-C observed in patients treated with liraglutide in several studies (8,9). Because LDL particles are importantly involved in atherogenesis, this increased LDL catabolism induced by liraglutide is likely to be beneficial. The mechanisms involved in the liraglutide-induced LDL catabolism are not clarified. However, it has been shown in vitro that liraglutide downregulates the expression of PCSK9, an inhibitor of LDLr, in HepG2 cells (26). Furthermore, liraglutide reduced the expression of PCSK9 protein in hepatocytes and decreased significantly plasma PCSK9 levels by 32% in db/db mice but not in control mice (26). The reasons why the effect of liraglutide on plasma PCSK9 levels is observed mainly in obese mice are unclear. In vitro studies showed that TNF-α increases PCSK9 levels (27). In mice, systemic inflammation increased expression of PCSK9 (28). Because obesity and diabetes are associated with increased chronic inflammation, we might suppose that the effect of liraglutide on plasma PCSK9 could be more visible in mice in the situation of obesity, particularly as a result of more elevated PCSK9 at baseline. In the current study, mean plasma PCSK9 level was significantly reduced by liraglutide in both patients with T2D and diet-induced obese mice. We show that liraglutide significantly reduces PCSK9 gene expression in the adipose tissue. In addition, we show in mice livers that liraglutide treatment reduces the expression of PCSK9 mRNA and increases the protein expression of LDLr. Furthermore, we show in liver tissue from ob/ob mice that liraglutide directly reduces the expression of the PCSK9 gene. Thus, liraglutide could increase the catabolism of LDL by reducing PCSK9 expression. Many studies have provided evidence of effects of GLP-1 agonists on the liver, including direct effects on hepatocytes, leading to significant modifications of the expression of several proteins (29,30). However, whether hepatocytes express the GLP-1 receptor is controversial. GLP-1 receptor was not detected on hepatocytes in some studies (31,32), whereas other studies reported the presence of GLP-1 receptor in hepatocytes (33,34). It has been suggested that in the liver, GLP-1 could act by receptor-independent mechanisms through GLP-1 degradation products (GLP-19–36, GLP-128–36, or GLP-132–36), which may be transported through the plasma membrane without involvement of a receptor and activate intracellular signaling (35). In our study, we used liver slices to gain an idea of the effect of liraglutide on the whole organ. Thus, we cannot give precise information on which cells liraglutide was acting. However, because PCSK9 is mostly expressed in hepatocytes, we may think that the decrease in PCSK9 expression with liraglutide is due to an action on hepatocytes.

In our study, the changes in total FCR and indirect FCR of VLDL₁-apoB100, VLDL₂-apoB100, IDL-apoB100, and LDL-apoB100 were not correlated with the reduction in body weight or HbA_1c. This does not totally rule out a possible influence of body weight or HbA_1c reductions on the kinetics of apoB100 but suggests that they are not likely to play a major role.

The effect of liraglutide on VLDL production remains unknown. Some animal studies have indicated that GLP-1 agonists could reduce hepatic VLDL production. In high-fat–fed apoE*3-Leiden mice, a high dose of exendin-4 (50 μg/kg/day) administered subcutaneously during 4 weeks induced a significant reduction of VLDL-apoB production (36). In fructose-fed Syrian golden hamsters, injections of exendin-4 (20 μg/kg) twice daily reduced the production of VLDL-TGs and VLDL-apoB (37). In our human study, we did not observe any significant decrease of VLDL₁-apoB100 or VLDL₂-apoB100 production after liraglutide treatment. The discrepancy between our results and animal data may be due to the very high dose of exendin-4 used in the animal experiments, which is much higher than the dose of exenatide, the synthetic exendin-4 analog, used in clinical practice.

Our kinetic study reports a significant increase in the catabolism of apoB100-containing lipoproteins, which may reduce atherosclerosis. This positive effect on the metabolism of apoB100-containing lipoproteins could partially explain the cardiovascular benefit of liraglutide observed in the Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results (LEADER) trial in which 3.5 years of treatment with liraglutide was associated with a significant 13% reduction in the primary outcome (time to first major cardiovascular event of cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke) (P = 0.01), a 22% reduction in cardiovascular death (P = 0.007), and a 15% reduction in total mortality (P = 0.02) (38).

As a limitation of the study, we acknowledge that we performed an uncontrolled study without inclusion of a placebo arm. However, because human in vivo kinetic studies are demanding and time consuming, the absence of a placebo arm is frequent in those that have analyzed the effect a therapeutic intervention on lipid metabolism (11,39,40). We do not believe that our study design induced an important bias in our results. Indeed, the evolution of the body weight and the lipid parameters observed after 6 months of treatment with liraglutide was similar to what is usually observed in clinical practice. As far as body weight reduction is concerned, the weight loss observed in our study was similar to the mean body weight reduction reported in other clinical studies with the same dose of liraglutide during a comparable period of treatment. We do not believe that reduction in body weight induced by liraglutide played a significant role in the changes in apoB100 kinetics. Indeed, in a lipoprotein kinetic study, where a body weight reduction of 4 kg (similar to what we observed in our study) was induced by a high monounsaturated fatty acid diet, no modification of VLDL-apoB100 FCR was observed (41). Furthermore, in our study, the modifications of apoB100 kinetic parameters were not correlated with body weight reduction. Thus, body weight reduction is not likely to have played a major role in the apoB100 kinetic modifications observed on liraglutide treatment in our study. We also do not believe that reduction in HbA_1c induced by liraglutide played a significant role in the changes in apoB100 kinetics. Indeed, in other lipoprotein kinetic studies performed in patients with T2D, treatment with sitagliptin (42) or pioglitazone (43) did not modify significantly VLDL-apoB100 FCR, whereas a significant drop of HbA_1c was observed. Furthermore, in our study, the modifications in apoB100 kinetic parameters were not correlated with HbA_1c changes, suggesting that HbA_1c reduction is not likely to play a major role in the apoB100 kinetic modifications observed on liraglutide treatment. Another limitation of our study is that we studied only apoB100 kinetics and did not include a study of TG kinetics. A combined kinetic study of apoB100 and TGs would have given additional information. However, we do not believe that it would have significantly modified the main information obtained with the present study. We also acknowledge that a single-blind study design including a placebo period followed by treatment period would have reduced a potential study effect.

In summary, the current study provides evidence that treatment with liraglutide significantly reduces the fractional catabolism of VLDL₁-apoB100, VLDL₂-apoB100, IDL-apoB100, and LDL-apoB100 in patients with T2D and typical diabetic dyslipidemia. Additional human or animal data indicate that liraglutide increases LPL mass, activity, and gene expression and reduces gene expression of the LPL inhibitory cytokines TNF-α and RBP4. Moreover, in the liver, liraglutide decreases PCSK9 mRNA and increases LDLr protein expression and directly reduces gene expression of PCSK9, inhibitor of LDLr expression at the cell membrane. These overall effects of liraglutide on lipoprotein metabolism may reduce cardiovascular risk in T2D.

Acknowledgments. The authors are indebted to Véronique Grivet and Isabelle Simoneau (Department of Endocrinology-Diabetology,CHU Dijon, Dijon, France) for help during the recruitment of patients, Amélie Cransac and Véronique Jost (Department of Pharmacy, CHU Dijon, Dijon, France) for the preparation of L-[1-¹³C]leucine, Cécile Gibassier (Department of Endocrinology-Diabetology, CHU Dijon, Dijon, France) for dietary assistance, and Laurence Loiodice (INSERM LNC UMR1231, University of Burgundy, Dijon, France) for invaluable technical assistance.

Funding and Duality of Interest. The authors acknowledge grants from Novo Nordisk and Agence Nationale de la Recherche under the program “Investissements d’Avenir” with reference ANR-11-LABX-0021 (LipSTIC Labex). They also acknowledge financial support from the University of Burgundy-Franche-Comté, the National Institute of Health and Medical Research (INSERM), the Region Burgundy-Franche Comté, and the Fonds Européens de Développement Régional. The authors’ employer (University Hospital, Dijon, France) received funding from Novo Nordisk for this study. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. B.V. wrote the manuscript. B.V., B.B., S.B.-R., A.R., and J.M.P. participated in the in vivo kinetic study. B.V. and L.De. designed the study. L.Du., J.M.P., and L.De. reviewed the manuscript. L.Du., P.D., and L.De. performed the biochemical analyses and animal and in vitro studies. J.P.P.d.B. performed the gas chromatography–mass spectrometry measures. B.V. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.