Overproduction of large VLDL particles is driven by increased liver fat content in man
We determined whether hepatic fat content and plasma adiponectin concentration regulate VLDL1 production.
A multicompartment model was used to simultaneously determine the kinetic parameters of triglycerides (TGs) and apolipoprotein B (ApoB) in VLDL1 and VLDL2 after a bolus of [2H3]leucine and [2H5]glycerol in ten men with type 2 diabetes and in 18 non-diabetic men. Liver fat content was determined by proton spectroscopy and intra-abdominal fat content by MRI.
Univariate regression analysis showed that liver fat content, intra-abdominal fat volume, plasma glucose, insulin and HOMA-IR (homeostasis model assessment of insulin resistance) correlated with VLDL1 TG and ApoB production. However, only liver fat and plasma glucose were significant in multiple regression models, emphasising the critical role of substrate fluxes and lipid availability in the liver as the driving force for overproduction of VLDL1 in subjects with type 2 diabetes. Despite negative correlations with fasting TG levels, liver fat content, and VLDL1 TG and ApoB pool sizes, adiponectin was not linked to VLDL1 TG or ApoB production and thus was not a predictor of VLDL1 production. However, adiponectin correlated negatively with the removal rates of VLDL1 TG and ApoB.
We propose that the metabolic effect of insulin resistance, partly mediated by depressed plasma adiponectin levels, increases fatty acid flux from adipose tissue to the liver and induces the accumulation of fat in the liver. Elevated plasma glucose can further increase hepatic fat content through multiple pathways, resulting in overproduction of VLDL1 particles and leading to the characteristic dyslipidaemia associated with type 2 diabetes.
KeywordsApolipoprotein B Compartmental modelling Diabetes Dyslipidaemia Kinetics Stable isotope Triglycerides VLDL
carbohydrate responsive-element binding protein
fractional catabolic rate
fractional direct catabolic rate
fractional transfer rate
homeostasis model assessment of insulin resistance
peroxisome proliferator activated receptor
sterol regulatory element binding protein 1c
Insulin resistance is associated with atherogenic dyslipidaemia characterised by high levels of triglyceride (TG)-rich lipoproteins and their cholesteryl ester-rich remnant particles, low levels of HDL-cholesterol, and small dense LDL . Previously, we reported that elevation of large VLDL particles is the major determinant of plasma TG levels in subjects with and without type 2 diabetes mellitus  and that the production rate of large VLDL1 particles is increased in type 2 diabetic patients . Thus, in diabetic dyslipidaemia, overproduction of large VLDL particles seems to initiate a sequence of lipoprotein changes that increase the risk of atherosclerosis, including an increase in smaller LDL and a decrease in HDL particles. Importantly, these lipid abnormalities precede the clinical diagnosis of type 2 diabetes and may explain the ‘ticking clock’ hypothesis for macrovascular disease that develops long before diabetes. Thus, it is of great importance to elucidate the mechanisms behind the overproduction of large VLDL particles in diabetic dyslipidaemia.
Non-alcoholic fatty liver disease has emerged as a new component of the metabolic syndrome  and a predictor of type 2 diabetes [5, 6, 7]. In observational studies, liver fat content was associated with common features of insulin resistance (e.g. increased fasting insulin levels, high C-peptide levels, glucose intolerance and intra-abdominal fat) both in patients with type 2 diabetes and in non-obese and moderately obese subjects without diabetes [4, 5, 6, 8, 9]. Therefore, the strong relationship between the hepatic fat content and hepatic insulin sensitivity is not surprising [8, 10, 11]. Liver fat content correlates closely with serum TGs, small dense LDL, and HDL-cholesterol [4, 8, 9]. Intriguingly, the fatty liver seems to exacerbate the dyslipidaemia in type 2 diabetic patients . The mechanisms underlying the relationship between dyslipidaemia and liver fat have not been fully defined.
Excessive visceral fat appears to be a key factor in the development of the disturbed metabolism in type 2 diabetes. Intra-abdominal fat is considered the major source of hepatic NEFA released directly into the portal vein . Thus, hepatic delivery of NEFA increases as the visceral fat compartment expands, and visceral adiposity has been reported to predict liver fat content [9, 10]. Recent evidence suggests that dyslipidaemia is mainly attributable to increased intra-abdominal fat .
Adiponectin is an adipocyte-secreted hormone that acts as an anti-diabetic and anti-atherogenic adipocytokine [14, 15]. Decreased plasma concentrations of adiponectin, unlike all other adipocytokines, correlate negatively with liver fat content and liver insulin sensitivity [7, 9, 16]. Thus, the plasma adiponectin concentration is lower in human obesity and in other insulin-resistant states , suggesting that adiponectin may be an important marker of the metabolic syndrome. Emerging evidence suggests that adiponectin also influences lipid metabolism. Adiponectin levels correlate inversely with plasma TGs but positively with HDL-cholesterol levels and LDL size [4, 8, 9, 17, 18, 19, 20]. Importantly, these relations persist even after adjustment for measures of overall obesity and visceral fat, consistent with a direct action of adiponectin on lipoprotein metabolism [19, 20]. Thus, adiponectin may be a mediator between adipose tissue and the liver, influencing both glucose and lipid metabolism.
To determine whether hepatic fat content and plasma adiponectin concentration are key regulators of VLDL1 production, we used a novel multicompartment model that allows the kinetic parameters of TG and ApoB in VLDL1 and VLDL2 to be simultaneously determined after a bolus of [2H3]leucine and [2H5]glycerol . Magnetic resonance proton spectroscopy was used to determine hepatic fat content and MRI to determine intra-abdominal and s.c. fat volumes. We studied both non-diabetic and type 2 diabetic subjects with different levels of insulin resistance, which permitted us to explore the effects of disturbances in lipid and glucose metabolism on VLDL kinetics over a wide range of liver fat and intra-abdominal fat content. Here we present the first human in vivo study on the relationship between liver and abdominal fat content and the production rate of VLDL1 TG and ApoB. From the results we hypothesise that liver fat and plasma glucose are the major determinants of VLDL1 TG and ApoB production rates.
Subjects and methods
We studied ten subjects with type 2 diabetes and 18 non-diabetic subjects, all of whom were men of white European origin. The subjects were recruited by advertisement in local newspapers and from a Finnish database, Health 2000, which is a population-based cohort. All subjects underwent a physical examination and laboratory tests to exclude hepatic abnormalities (other than hepatic steatosis) and renal, thyroid and haematological abnormalities. Subjects with CHD, diabetic retinopathy or microalbuminuria, total cholesterol greater than 7.0 mmol/l, TG greater than 5.0 mmol/l, or BMI greater than 40 kg/m2 were excluded. All non-diabetic subjects had lipid values representative of a Western population; three had IGT according to WHO criteria . The diagnosis of type 2 diabetes was based on glucose tolerance tests according to WHO criteria  or on use of oral anti-diabetic medication. The age at onset of type 2 diabetes was at least 35 years. None of the subjects was taking lipid-lowering therapy or insulin. Four patients with type 2 diabetes and one non-diabetic subject were taking acetylsalicylic acid. Two patients with type 2 diabetes were treated with diet alone, one with diet and metformin, two with diet and sulfonylurea, and five with a combination therapy of sulfonylurea, metformin and diet. One patient with type 2 diabetes was using antihypertensive therapy (felodipine). The medications were continued throughout the study. The mean duration of diabetes was 7.0 years (range 0–13 years). The study design was approved by the ethics committee of Helsinki University Central Hospital, and each subject gave written informed consent. All samples were collected in accordance with the Declaration of Helsinki. One subject did not have intra-abdominal or s.c. fat measured. Fifteen of the control subjects and six of the type 2 diabetic subjects were presented in a previous paper .
ApoB and TG turnover
Turnover of ApoB and TG was measured after an overnight fast. Particle composition and ApoB mass of the VLDL1 and VLDL2 fractions were determined 30 min before and 0, 4, and 8 h after a bolus injection of [2H3]leucine and [2H5]glycerol . The subjects fasted until 17.00 h, when the last blood sample was taken.
Measured ApoB and TG pool sizes in VLDL1 and VLDL2 were converted into leucine and glycerol equivalents. The injected amount of [2H3]leucine and [2H5]glycerol, the leucine and glycerol pool sizes in VLDL1 and VLDL2, and the enrichment curves of leucine in plasma and leucine and glycerol in VLDL1 and VLDL2 were used as the data set for a multicompartment model that allowed simultaneous modelling of ApoB and TG kinetics as described . The outputs from the model are the production rates and fractional catabolic rate (FCR) of VLDL1 and VLDL2 (the fractional loss of mass due to transfer and direct catabolism), the fractional direct catabolic rate (FDCR; the fractional loss of mass due to direct catabolism from VLDL1) and the fractional transfer rate (FTR; the fractional transfer from VLDL1 to VLDL2). In this model, we cannot distinguish between transfer from VLDL2 to intermediate-density lipoprotein and direct catabolism of VLDL2 particles; hence they are represented by the FCR. The ApoB and TG production rates were calculated as mg day−1 kg body weight−1. FCR, FDCR and FTR were calculated for both ApoB and TG as pools per day.
Isolation of lipoproteins and biochemical analyses
VLDL1 and VLDL2 were isolated from 8.4 ml of plasma . ApoB and TG concentrations in these lipoprotein fractions were measured in samples obtained at 0, 4 and 8 h. ApoB and TG pool sizes were calculated as the product of plasma volume (4.5% of body weight) and the plasma concentration of ApoB and TG in VLDL1 and VLDL2. The leucine and glycerol contents of each pool were determined , and the TG, cholesterol, ApoB, glucose, insulin, NEFA, HbA1c and protein concentrations in lipoprotein fractions were analysed as described . LDL peak particle diameter (LDL size) was measured with 2–10% gradient polyacrylamide gel electrophoresis . Serum adiponectin concentrations were measured with an ELISA kit (B-Bridge International, San Jose, CA, USA) that detected full-length peptides. The HOMA-IR (homeostasis model assessment for insulin resistance) was calculated from the fasting glucose and serum insulin concentrations as fasting insulin (mU/l)×fasting glucose (mmol/l)/22.5 .
Liver fat analysis
Image-guided proton magnetic resonance spectroscopy was performed with a 1.5 T whole-body device (Magnetom Vision; Siemens, Erlangen, Germany) using a combination of whole-body and loop surface coils for the radio frequency transmitting and receiving of signals [9, 11]. Volumes of interest (8 cm3) in the liver were placed, avoiding vascular structures and s.c. fat tissue. These areas were typically within the right lobe. Chemical shifts were measured relative to water signal intensity at 4.8 ppm (Swater). Methylene signal intensity, which represents intracellular TGs in the liver , was measured at 1.4 ppm (Sfat). The measurement of the percentage of hepatic fat by proton spectroscopy has been validated against the lipid content of liver biopsies in humans  and against liver density measurements performed by computed tomography . The percentage of liver fat was calculated as 100×Sfat/(Sfat+Swater). Regarding the error introduced by the use of a single rather than multiple voxels, we have shown in 24 normal subjects that the Spearman’s correlation coefficient for liver fat between the left and right lobes is 0.95 (95% CI 0.89–0.98, p<0.001) and the CV is 6%.
Determination of intra-abdominal and s.c. fat volumes
Intra-abdominal and s.c. fat volumes were determined by MRI. A series of 16 T1-weighted transaxial scans were acquired from a region extending from 8 cm above to 8 cm below the fourth and fifth lumbar interspaces . Intra-abdominal and s.c. fat areas were measured with an image analysis program (Alice 3.0; Parexel, Waltham, MA, USA) .
Differences between non-diabetic and type 2 diabetic subjects were evaluated with unpaired, two-sided t tests. Correlations were evaluated by linear regression and multiple interactions by forward stepwise multivariate linear regression and by exhaustive check of all possible multivariate models. Variables with univariate correlation with p<0.05 were included in the multivariate regression but HOMA-IR was left out of the multivariate when both plasma glucose and fasting insulin were included. Also, we included all fat compartment variables in the multivariate analysis, even though we did not find a univariate correlation. The correlations were calculated for all subjects as well as for the type 2 diabetic subjects and the non-diabetic subjects. For non-normally distributed variables, values were log-transformed before correlation analysis. Values of p<0.05 were considered significant.
Statistical analysis was performed with SPSS, Microsoft Excel and Matlab (MathWorks, Natick, MA, USA).
Basic characteristics of the subjects
Non-diabetic subjects (n=18)
Type 2 diabetic subjects (n=10)
Liver fat (%)
Total abdominal fat (cm3)
S.c. fat (cm3)
Intra-abdominal fat (cm3)
Fasting serum glucose (mmol/l)
Fasting serum insulin (mU/l)
Serum adiponectin (mg/l)
Serum ALT (U/l)
Serum NEFA (μmol/l)
Plasma TG (mmol/l)
LDL size (nm)
Production and clearance rates
Non-diabetic subjects (n=18)
Type 2 diabetic subjects (n=10)
VLDL TG total production (mg kg−1 day−1)
VLDL1 TG production rate (mg kg−1 day−1)
VLDL1 TG FCR (pools/day)
VLDL1 TG FDCR (pools/day)
VLDL1 TG FTR (pools/day)
VLDL1 TG to VLDL2 transfer (mg kg−1 day−1)
VLDL2 TG direct production rate (mg kg−1 day−1)
VLDL2 TG FCR (pools/day)
VLDL ApoB total production (mg kg−1 day−1)
VLDL1 ApoB production rate (mg kg−1 day−1)
VLDL1 ApoB FCR (pools/day)
VLDL1 ApoB FDCR (pools/day)
VLDL1 ApoB FTR (pools/day)
VLDL1 ApoB to VLDL2 transfer (mg kg−1 day−1)
VLDL2 ApoB direct production rate (mg kg−1 day−1)
VLDL2 ApoB FCR (pools/day)
Hepatic lipid and intra-abdominal fat predictors
Univariate regressions of plasma concentrations and adiposity measures
Total abdominal fat
Multivariate regression models searching for independent correlates of liver fat content, VLDL1 TG and ApoB production, and VLDL1 TG and ApoB FTR: results of stepwise multivariate regression
VLDL1 TG production rate
VLDL1 apoB production rate
VLDL1 TG FTR
VLDL1 ApoB FTR
Production and clearance rates
In a stepwise multivariate regression analysis of VLDL1 TG or apoB production rate as the dependent variable and plasma glucose, insulin, liver fat, intra-abdominal fat, subcutaneous fat and adiponectin as independent variables, only plasma glucose and liver fat remained significant (Table 4). When VLDL1 TG or ApoB FTR served as the independent variable, insulin was the best predictor of VLDL1 TG and ApoB (Table 4).
This study shows that liver fat, insulin and plasma glucose are independent determinants of VLDL1 TG and ApoB production rates. The key metabolic determinants of liver fat content were intra-abdominal fat, adiponectin and plasma glucose. These findings emphasise the critical role of substrate fluxes and lipid availability in the liver as the driving force for the overproduction of VLDL1 in subjects with type 2 diabetes .
The liver fat content reflects the balance between NEFA flux, fatty acid oxidation, de novo lipogenesis and VLDL secretion. Increased NEFA flux is a major contributor to the availability of TG for VLDL assembly [27, 28]. Indeed, increased delivery of NEFA to the liver in humans is associated with fatty liver [29, 30] and has been linked to hepatic insulin resistance [31, 32, 33]. These findings are consistent with the strong relationship we observed between intra-abdominal fat and liver fat content.
Both hepatic and intra-abdominal fat content correlated with VLDL1 TG and ApoB production rates in the univariate analysis, but only the hepatic fat content was significant in multiple regression models. Our data suggest that hepatic steatosis is the proximate cause of the characteristic atherogenic dyslipidaemia seen in type 2 diabetes [1, 10]. This conclusion is consistent with the fact that VLDL is produced by liver fat, not visceral adipose tissue. VLDL1 particles are converted in the circulation by cholesteryl ester transport protein and hepatic lipase – which are commonly increased in type 2 diabetes – to small atherogenic LDL particles . This explains the strong inverse correlation between VLDL1 TG production rate and LDL particle size. In a parallel process, excess of VLDL1 particles alters the lipid composition of HDL particles, leading to enhanced catabolism of these particles ; this explains the strong inverse correlation between HDL-cholesterol and liver fat in the present study (p<0.001, r=−0.65). Thus, overexpression of VLDL1 leads to the generation of both small dense LDL and low HDL, the two hallmarks of diabetic dyslipidaemia. Our data highlight liver fat as a major driving force for diabetic dyslipidaemia. However, we recognise that regulatory factors not measurable in man (i.e. enzymatic activities in the liver, transcription factors and substrate fluxes) can influence hepatic lipid metabolism and influence VLDL production.
Univariate regression analyses of plasma concentrations and adiposity measures vs VLDL1 and VLDL2 production and clearance rates
VLDL1 production rate
VLDL1 production rate
Total abdominal fat
Insulin appears to be pivotal in regulating VLDL assembly in the liver [35, 36]. Insulin acts on two key factors that regulate the initiation of VLDL assembly and the maturation of VLDL particles: microsomal triglyceride transfer protein and phosphatidylinositol 3-kinase [37, 38, 39, 40]. Consequently, impaired insulin signalling results in increased VLDL production. Consistent with previous studies [3, 41], we observed significant relationships between fasting insulin and HOMA-IR and VLDL1 TG and apoB production rates. However, these relationships were not significant in regression models that included liver fat. HOMA-IR can be considered at best a surrogate marker for type 2 diabetes, and, notably, is influenced more strongly by fasting glucose than by insulin, since fasting glucose closely reflects hepatic glucose production . Thus, liver fat content may influence hepatic insulin resistance by blunting the insulin signalling pathways in the liver. This would explain why fasting glucose predicted VLDL1 production, especially in subjects with type 2 diabetes (Fig. 2d).
Insulin appears to be pivotal in regulating VLDL assembly in the liver [35, 36]. Insulin acts on two key factors that regulate the initiation of VLDL assembly and the maturation of VLDL particles: microsomal triglyceride transfer protein and phosphatidylinositol 3-kinase [37, 38, 39, 40]. Consequently, impaired insulin signalling results in increased VLDL production. Consistent with previous studies [3, 41], we observed significant relationships between fasting insulin and HOMA-IR and VLDL1 TG and ApoB production rates. However, these relationships were not significant in regression models that included liver fat. HOMA-IR can be considered at best a surrogate marker for type 2 diabetes, and, notably, is influenced more strongly by fasting glucose than by insulin, since fasting glucose closely reflects hepatic glucose production . Thus, liver fat content may influence hepatic insulin resistance by blunting the insulin signalling pathways in the liver. This would explain why fasting glucose predicted VLDL1 production, especially in subjects with type 2 diabetes (Fig. 2d).
The strong relation between fasting glucose and VLDL1 production probably reflects disturbed regulation of VLDL1 production. The relation between plasma glucose and VLDL1 TG production could be explained by increased expression of sterol regulatory element binding protein 1c (SREBP1c) because of the increased levels of insulin combined with increased production of glycerol 3-phosphate. Glycerol 3-phosphate could either be formed during the glycolysis, but more probably as a result of the increased gluconeogenesis present in insulin resistance. The observation that fasting glucose in insulin resistance is closely linked to the production of glucose in the liver would support such a mechanism.
Elevated plasma glucose concentrations may also contribute directly to increased VLDL1 production by providing the carbon skeleton for glycerol utilised in the re-esterification of NEFA. Therefore, in the setting of increased NEFA flux, hyperglycaemia may enhance TG synthesis. Along these lines, it has been indicated that glucose could influence the bulk lipid addition step of VLDL assembly . Another potential source of liver TG is de novo lipogenesis [29, 30]. In liver insulin resistance, hyperinsulinaemia upregulates SREBP1c expression and leads to activation of key enzymes for lipogenesis [44, 45]. Carbohydrate responsive-element binding protein (ChREBP) is also stimulated by hyperglycaemia and activates key lipogenic enzymes . Thus, concomitant up regulation of SREBP1c and ChREBP by insulin and glucose in type 2 diabetes can increase hepatic lipogenesis. Production of malonyl-CoA is increased by fatty acid synthesis, but is also increased directly by acetyl-CoA carboxylase-2, which is activated by SREBP1c. Malonyl-CoA inhibits carnitine palmitoyl transferase-1, which results in decreased fatty acid oxidation [47, 48]. Thus, elevated plasma glucose acts via multiple pathways to increase liver fat content. Not surprising, plasma glucose strongly correlated with both liver fat content and VLDL1 production rate.
Adiponectin also appears to contribute to the metabolic dysregulation in type 2 diabetes. Adiponectin strongly predicted liver fat content, as in previous studies . By activating AMP-activated protein kinase in hepatocytes , which stimulates β-oxidation and simultaneously inhibits lipogenesis , adiponectin may regulate fatty acid oxidation. By activating peroxisome proliferator activated receptor (PPAR)α, adiponectin may be pivotal in upregulating the expression of key enzymes of fatty acid oxidation . Thus, hypoadiponectinaemia in type 2 diabetes may impair fatty acid oxidation through at least two mechanisms. In the setting of increased NEFA flux, this would direct NEFA for re-esterification and enhance TG accumulation in the liver.
Several studies have reported a strong association between dyslipidaemia and adiponectin. Despite the close relationship between adiponectin and the liver fat content, adiponectin did not predict VLDL1 production and therefore must influence plasma TGs by mechanisms other than effects on VLDL1 production. Interestingly, adiponectin regulates plasma lipoprotein lipase activity independently of insulin sensitivity and BMI . Thus, adiponectin can influence the intravascular catabolism of VLDL particles by several mechanisms. This hypothesis is consistent with our finding that adiponectin correlated with the FCR of VLDL1 TG and ApoB and the FTR of VLDL1 TG, and is consistent with recent data by Ng et al. . Notably, adiponectin stimulates (via its action on PPARα) the expression of ApoCIII, which regulates the metabolism of TGs .
A potential weakness of our study is that the data are derived from two populations with different characteristics. Thus, the reported findings could be a function of the predictors or of the diabetic state in itself. However, the two groups are not entirely separate but represent an overlapping continuum of VLDL TG and ApoB pools.
In summary, we propose that the metabolic effect of insulin resistance, partly mediated by depressed plasma adiponectin levels, increases fatty acid flux from adipose tissue to the liver and induces the accumulation of fat in the liver. Elevated plasma glucose can further increase hepatic fat content though multiple pathways, resulting in overproduction of VLDL1 particles and leading to the characteristic dyslipidaemia associated with type 2 diabetes.
We thank H. Hilden, H. Perttunen-Nio, K. Skålén and D. Bedford for expert technical assistance and S. Ordway for editing the manuscript. This work was supported by The Swedish Research Council, The Swedish Heart-Lung Foundation, The Göran Gustafsson Foundation, The Sigrid Juselius Foundation and the Helsinki University Central Hospital Research Foundation.