Sensitivity to Acute Insulin-Mediated Suppression of Plasma Free Fatty Acids Is Not a Determinant of Fasting VLDL Triglyceride Secretion in Healthy Humans

Eighteen healthy subjects (10 women, 8 men), aged 18–60 years with a BMI between 18 and 35 kg/m², participated in the studies. None were diabetic based on repeated assessment of fasting glucose concentration (14). One subject had mildly impaired fasting blood glucose. None of the study participants were taking any medication, had any current medical condition known to affect lipid levels or insulin sensitivity, had any known cardiovascular disease, or had any surgical intervention within 6 months before the studies. All the women who participated were premenopausal, and the studies were conducted during the follicular phase of their menstrual cycle. Informed written consent was obtained from all participants in accordance with the guidelines of the Human Subjects Review Committee of the University Health Network.

All of the subjects participated in three studies (as described below) 3–4 weeks apart, in random order. They were on a stable diet, had no change in their degree of physical activity, and had stable weight for at least 3 months before and throughout the study period. All studies were performed in the Metabolic Investigation Unit of the Toronto General Hospital, where the patients were admitted on each occasion between 7:30 and 8:30 a.m., following a 12-h overnight fast. They remained fasting during the studies, but had access to water ad libitum. For each study, an intravenous catheter was placed in each forearm, one for infusion and one for blood sampling. The arm containing the sampling catheter was maintained in a heating blanket (∼65°C) to arterialize venous blood. Body weight and height were assessed during fasting on the morning of the three studies; waist circumference was measured at midlevel between the costal and iliac ridge while standing.

All of the study participants received detailed instructions to maintain an isocaloric diet containing 30% fat, 20% protein, and 50% carbohydrate for 2 days before these studies. After a 12-h overnight fast, the participants received an intravenous bolus injection of [1,1,2,3,3-²H₅] glycerol (100 μmol/kg) (Cambridge Isotopes Laboratories, Andover, MA) to determine VLDL-TG kinetics, as previously described (15). The tracer was dissolved in sterile normal saline and was tested for sterility and pyrogenicity before injection. Blood samples were taken at 0, 10, 20, 30, 60, 90, 120, 150, 180, 240, 300, 420, 600, 660, and 720 min for separation of the VLDL fraction (Svedberg flotation 20-400) by ultracentrifugation and determination of VLDL-TG [1,1,2,3,3-²H₅]glycerol enrichment. The samples were collected into Vacutainer tubes (Becton Dickinson Vacutainer Systems, Rutherford, NJ) containing Na₂EDTA and placed immediately on ice, and the plasma was separated within 2 hours. VLDL apolipoprotein B (apoB) and VLDL-TG concentrations were also measured at 0, 120, 240, 360, 600, and 720 min to verify that a steady-state level was maintained. The averages of these measurements were taken as the VLDL-apoB and VLDL-TG levels.

An insulin-modified frequently-sampled intravenous glucose tolerance test (mFSIVGTT) was performed after a 12-h overnight fast to determine the insulin sensitivity index (S_I), using the minimal model approach (16,17). Briefly, after three baseline blood samples, an IV bolus of glucose (0.3 g/kg; dextrose 50%) was administered at time 0, and an IV bolus of biosynthetic human insulin (0.03 U/kg) (Humulin R; Eli Lilly, Toronto, Canada) was injected at 20 min. Blood samples were taken at 2, 3, 4, 6, 8, 10, 12, 14, 16, 19, 22, 23, 24, 25, 27, 30, 40, 50, 60, 70, 80, 100, 120, 140, 160, 180, 200, 220, and 240 min for determination of plasma glucose, insulin, and C-peptide levels.

This protocol is a modification of our previously published protocol (18) that allows the determination of insulin-mediated suppression of fasting plasma FFAs while keeping constant the plasma glucose level. Essentially, this protocol was designed to provide a single integrated measure of the suppressibility of plasma FFA concentrations in response to a low-dose, stepwise increment in plasma insulin concentration during euglycemia. This protocol also allowed us to calculate insulin clearance for each individual (see below). After a 30-min baseline period, a programmed IV insulin (Humulin R; Eli Lilly) infusion was started at a rate of 1 mU · m⁻² · min⁻¹ and was increased in a stepwise fashion to 2, 4, 8, 16, and 32 mU · m⁻² · min⁻¹ every 40 min. An IV glucose infusion (dextrose 20%), adjusted every 5 min according to the plasma glucose level, was used to maintain plasma glucose at the fasting level. Glucose levels did not change from baseline (P > 0.20) (not shown). This protocol resulted in small stepwise increments of plasma insulin levels from 0 to ∼35 mU/l above basal levels, which allowed for enough discrimination of the suppression of plasma FFAs by insulin in healthy subjects (19). Blood samples were taken at 10-min intervals throughout the study for measurement of glucose, insulin, C-peptide, and FFAs. Blood samples were collected on ice into tubes containing Na₂EDTA and 30 μg of the lipase inhibitor tetrahydrolipstatin (Orlistat; Hoffman La Roche, Mississauga, ON, Canada) per milliliter blood to prevent ongoing in vitro lipolysis of the samples (20). Figure 1 shows insulin infusion rate (A), plasma glucose (B), insulin (C), FFAs (D), insulin appearance rate versus change in insulin (E) (used to calculate insulin clearance; see below), and FFAs versus change in insulin (F) (used to calculate insulin-mediated suppression of plasma FFAs, k; see below) in a representative subject.

Glucose was assayed enzymatically at the bedside using a Glucose Analyzer II (Beckman Instruments, Fullerton, CA). Insulin was measured by radioimmunoassay using a double-antibody separation method (kit supplied by Pharmacia Diagnostic, Uppsala, Sweden) (intra- and interassay coefficients of variation [CVs], 5.8 and 11.2%, respectively). C-peptide was measured by a double-antibody C-peptide radioimmunoassay (kit supplied by Diagnostic Products Corporation, Los Angeles) (intra- and interassay CVs, 4.1 and 15.1%, respectively). FFAs were measured by a colorimetric method (kit supplied by Wako Industrials, Osaka, Japan) (CV < 3%). Cholesterol and TG were measured for each lipoprotein fraction using colorimetric assays (Boehringer Mannheim GmbH Diagnostica) (CV <3%). ApoB was measured in each fraction using an electroimmunoassay, as previously described (21) (CV <10%). VLDL was isolated from the plasma by ultracentrifugation at density 1.006 g/ml using a type 50.3 Ti rotor (Ultracentrifuge model L870; Beckman, Palo Alto, CA) at 39,000 rpm and 16°C for 16 h.

The VLDL fraction was delipidated with diethyl ether/methanol in acid-washed glass tubes to separate VLDL lipids from VLDL proteins. The supernatant containing VLDL lipids (TG, free cholesterol, cholesteryl ester, and phospholipids) was separated and dried under nitrogen and VLDL-TG and then isolated by thin-layer chromatography on 20 × 20 cm silica gel plates (06-600A Gel G TLC; Fisher) using heptane:isopropyl ether:acetic acid 80:20:2 (22). VLDL-TG glycerol was recovered and converted to heptafluorobutyryl derivative as previously described (15), and enrichment of [1,1,2,3,3-²H₅]glycerol was measured by electron impact ionization gas chromatography/mass spectrometry (GC/MS) using a 30 m × 0.25 mm, 0.25 m film DB-17 capillary column (Supelco) on a model 5973 quadrupole GC/MS (Hewlett-Packard). Glycerol (m/z 467) and its m+5 isotopomer (m/z 472) were measured, and the glycerol tracer:tracee ratio was determined by calibration of measured m+5/m+0 ratios for standards of known isotopic enrichment (CV <2%).

The fractional clearance rate (FCR) of VLDL-TG was estimated from the monoexponential slope of VLDL-TG [²H₅]glycerol isotopic enrichment immediately after the maximum enrichment, as previously described (15). We have recently shown that the monoexponential slope captures the major feature of the VLDL-TG time course curve, which is the turnover of VLDL-TG (tracer recycling is a minor feature) and is an index that correlates nonlinearly with the FCR (23). The VLDL-apoB and VLDL-TG levels did not change significantly over time during the 12-h turnover study (P > 0.20); therefore, steady state was assumed. VLDL-TG secretion rates (VLDL-TG ASR) were calculated as follows (15):

Where [VLDL-TG] is the mean concentration of VLDL-TG (in millimoles per liter) over the entire 12 h of the study.

S_I (10⁴ · min⁻¹ · μU⁻¹ · ml⁻¹) was calculated using MINMOD software (version 3.0; provided by Dr. R.N. Bergman) (24). The acute insulin response (AIR) and acute C-peptide response (ACR) were determined as surrogates of the acute glucose-stimulated insulin secretory response by calculating the average of the incremental response from baseline of plasma insulin and C-peptide, respectively, at 2, 3, 4, and 5 min after the IV glucose bolus (25).

To calculate the plasma insulin clearance rate (Cl_ins), the endogenous insulin secretion rate (ISR) during the graded insulin infusion study was first assessed by deconvolution of plasma C-peptide (26) using standard kinetic parameters for C-peptide clearance, as previously validated by others (27). ISR during the last 20 min of each insulin infusion step was then added to the exogenous insulin infusion rate to obtain the total insulin appearance rate (R_a) for each insulin infusion step. R_a was plotted against the average insulin levels during the last 20 min of each insulin infusion step, and Cl_ins was obtained by determining the slope of the relationship by linear regression (Fig. 1E). The average R² of the fit was 0.87 ± 0.05, suggesting a good fit of the data.

The natural logarithms of the mean plasma FFA levels during the last 20 min of each graded intravenous insulin infusion were plotted against the corresponding change from baseline of plasma insulin, and the data were fitted by linear regression for each subject (Fig. 1F). The absolute value of the slope of this relationship, k [natural log(millimoles of FFA suppression)/picomoles of increase in plasma insulin], is proportional to the suppression of plasma FFA levels by insulin over small, stepwise elevations of plasma insulin concentration. The mean R² of the fit was 0.80 ± 0.05, suggesting a good fit of the data.

The data are expressed as means ± SE. When not normally distributed, the variables were transformed using the mathematical function leading to the best normalization of the data before linear regression analysis. Linear regression analyses were performed with total TG levels, VLDL-apoB levels, VLDL-TG levels, and VLDL-TG kinetic parameters as independent variables and the other metabolic, clinical, and anthropometric variables as dependent variables. For variables available from more than one of the studies (BMI, waist circumference, and fasting glucose, insulin, C-peptide, FFA, and total plasma TG levels), the average values from the three studies are reported and were used for the analyses. Forward-step multivariate linear regression analyses including only variables with at least a trend toward an association (P < 0.20) in bivariate analyses were performed to define the best model (two of the following three criteria: maximum adjusted R², minimum residual sum of squares, and optimal Mallows C_p) to predict total TG levels, VLDL-TG levels, total plasma apoB levels, VLDL-apoB levels, and VLDL-TG kinetic parameters. For all the analyses, a P value <0.05 was considered significant. All the analyses were performed with SAS software for Windows (version 7; SAS Institute, Cary, NC).

The clinical characteristics and the fasting metabolite and hormone levels of the 18 participants are shown in Table 1. VLDL-TG kinetic parameters and insulin secretion, clearance, and sensitivity parameters are shown in Table 2. Plasma TG, VLDL-TG, and VLDL-apoB levels did not change significantly over time during the VLDL stable isotope administration study and were therefore in steady state (P > 0.98).

Table 3 shows the bivariate correlation between both total TG levels and VLDL-TG levels (log transformed) and the other metabolic and clinical variables. The best multivariate model to predict total TG levels included BMI (log) and Cl_ins (square root) (R² = 0.73, P < 0.0001; P = 0.005 and P = 0.03, respectively). No other variable independently predicted total TG levels in our subjects, and BMI remained independently associated despite the addition of any other variable in the model. The best multivariate model to predict VLDL-TG levels included BMI (log), fasting C-peptide levels, and S_I (R² = 0.76, P = 0.0001; P = 0.001, P = 0.05, and P = 0.17, respectively). BMI was the only variable that remained significantly correlated with VLDL-TG levels when any other variable was included in the model.

Table 4 shows the bivariate correlation between both total apoB levels and VLDL-apoB levels (log transformed) and the other metabolic and clinical variables. The best multivariate model to predict total plasma apoB levels included BMI, age (1/age), and ACR (R² = 0.56, P = 0.008; P = 0.05, P = 0.11, and P = 0.21, respectively). BMI (log) was again the only variable that remained independently correlated with total apoB levels irrespective of the inclusion of any other variable in the model. The best multivariate model to predict VLDL-apoB levels included BMI (log), Cl_ins (square root), and age (1/age) (R² = 0.63, P = 0.002; P = 0.11, P = 0.13, and P = 0.10, respectively).

In this healthy population, VLDL-TG ASR (log) and VLDL-TG FCR were both contributing significantly to total plasma TG (log) (R² = 0.71, P < 0.0001, and R² = 0.27, P = 0.03, respectively), although only VLDL-TG ASR was significantly associated when both variables were included in the model (R² = 0.77, P < 0.0001; P < 0.0001 and P = 0.07, respectively.

Table 5 shows the correlation between VLDL-TG FCR and ASR and the various clinical, anthropometric, and metabolic variables assessed in the present study. BMI (log) (Fig. 2A), fasting C-peptide levels (Fig. 2B), Cl_ins (square root), ACR, and k (square root) (Fig. 2C) were all significantly associated with VLDL-TG FCR (log). S_I (Fig. 2D), sex, AIR (log), and fasting FFAs did not correlate with VLDL-TG FCR. The best multivariate model to predict VLDL-TG FCR (log) included fasting C-peptide levels, Cl_ins (square root), and ACR (R² = 0.79, P < 0.0001; P = 0.0001, P = 0.06, and P = 0.07, respectively). Fasting C-peptide level was the only variable that remained significantly correlated with VLDL-TG FCR when any other variable was included in the model.

BMI (log) (Fig. 3A), fasting insulin (log) (Fig. 3B), and AIR (log) were significantly positively correlated with VLDL-TG ASR (log). Fasting FFAs (Fig. 3C), and k (square root) (Fig. 3D) did not correlate with VLDL-TG ASR. The best multivariate model to predict VLDL-TG ASR included BMI (log) and S_I (R² = 0.61, P = 0.0008; P = 0.0008 and P = 0.12, respectively). Only BMI remained significantly correlated with VLDL-TG ASR when any other variable was included in the model. Correction of VLDL-TG secretion rate for body weight did not affect any of the above associations, and BMI still remained significantly and independently associated with VLDL-TG secretion (not shown).

There were no significant differences in VLDL-apoB, VLDL-TG, VLDL secretion, FCR, S_I, insulin levels, C-peptide levels, FFA levels, or insulin-mediated suppression of plasma FFAs between men and women. Furthermore, forcing sex into the various models did not alter the relationship between BMI, FFAs, S_I, and insulin-mediated suppression of plasma FFA and VLDL levels or secretion.

In the present study, BMI alone accounted for 40–65% of the variance of total plasma apoB, plasma TG, VLDL-TG, and VLDL-apoB levels and VLDL-TG secretion rate in our group of participants and was the only factor that remained significantly associated when the effect of any other variable was taken into account. This strong and independent association between BMI and VLDL-TG secretion rate remained even when the latter was normalized for body weight (accounted variance, 37%; data not shown). BMI also accounted for 47% of the variance of VLDL-TG fractional clearance rate, although it was not independent from the effect of fasting C-peptide levels, the main determinant of VLDL-TG FCR in our study. An association between VLDL metabolism and plasma C-peptide levels is in agreement with findings by others (28–30).

Although the magnitude of these associations with BMI is remarkable, these findings are not surprising in view of the well-established association between excess body fat, especially visceral fat, and VLDL levels and secretion rate (31–36). It is generally thought that high FFA flux from visceral adipose tissue, which would not necessarily affect or be reflected by systemic plasma FFA levels, could explain the association between elevated VLDL secretion rate and visceral obesity. In the present study, we used waist circumference as a surrogate marker of visceral adipose tissue mass (37) and found that it was not as well correlated with either VLDL levels or secretion rate as BMI. However, this may simply mean that waist circumference does not have enough accuracy to predict visceral fat mass and that more direct assessment of visceral fat would have been more appropriate to use with our relatively small sample size. Furthermore, it is important to note that we did not measure total body fatty acid flux or the direct influx of FFAs from visceral adipose tissue. The present study measured the sensitivity of FFA output (presumably predominantly from peripheral adipose cells) to insulin suppression. Therefore, we believe that visceral fat mass could still explain the association between BMI and VLDL-TG levels and secretion rate in the present study.

The most important and novel finding of the present study is that the sensitivity of insulin-mediated suppression of fasting plasma FFAs was not associated with the rate of postabsorptive VLDL-TG production in healthy individuals. Because we assessed the sensitivity to insulin-mediated suppression of plasma FFAs and not the effect of FFA suppression itself on VLDL secretion, this finding is not in conflict with the generally accepted view that VLDL-TG production is predominantly a substrate-driven process (10). We are in agreement with the observation that an increased flux of FFAs from extrahepatic tissues to the liver plays an important role in the causation of hypertriglyceridemia associated with insulin-resistant states (1,6,10,13,38–43). In addition, our study did not measure the overall channelling of fatty acids toward the liver over a 24-h period. Intracellular lipolysis of adipose tissue triglyceride is much more sensitive to the suppressive effects of insulin than stimulation of peripheral tissue glucose uptake (44), with a half-maximal suppressive dose of insulin on plasma FFA concentration found to be <20 μU/ml (19). Insulin-mediated suppression of plasma FFAs in the postabsorptive state is believed to reflect predominantly inhibition of adipose tissue lipolysis of stored triglycerides, with consequent reduction in the release of FFAs, due to suppression of hormone-sensitive lipase and stimulation of adipose tissue triglyceride reesterification (45). Insulin-mediated suppression of plasma FFAs in the postabsorptive state would not accurately reflect the total 24-h flux of fatty acids to the liver in free-living subjects, where a major contribution occurs from fatty acids derived from ingested fat (46). Fasting plasma FFA levels were not significantly correlated with VLDL levels and secretion in the present study, in accordance with findings by others (28,33,47,48). In contrast, others have observed a significant correlation between postprandial plasma FFA levels and VLDL-TG levels and production rate (47,49). Our findings, therefore, do not preclude an important role for elevated FFA flux to the liver in causing VLDL overproduction in insulin-resistant states, but do suggest that VLDL overproduction is not predicted by impaired insulin-mediated suppression of plasma FFAs in the low insulin-dose range in the postabsorptive state. In other words, impairment of insulin-stimulated plasma FFA esterification in adipose tissue during the postprandial period or impaired suppression of visceral tissue lipolysis may have a greater impact on chronic liver triglyceride secretion than impairment of insulin’s antilipolytic effect at the peripheral adipose tissue during fasting. Whether this mechanism could perhaps explain some of the association between increased adiposity and VLDL secretion is an unresolved issue. It is important to note that the present study was not designed to explore the mechanism of the association (or lack of association) between insulin-mediated suppression of fasting plasma FFAs and VLDL secretion. More mechanistic in vivo studies are clearly needed to clarify this important issue.

Insulin sensitivity with regard to glucose metabolism (S_I) accounted for only ≤15% of the variance of VLDL levels and kinetics in our study participants, but S_I entered the best model to predict VLDL-TG secretion. However, this association was much less robust than the one with BMI: the significance of the association between S_I or plasma insulin level, but not BMI, and VLDL secretion was lost when we excluded the two subjects with the lowest S_I. Some (28,33,50,51), but not all (35,47), previous kinetic studies have shown a similar degree of association (10–20% of the variance) between fasting insulin levels (surrogate of S_I) or insulin sensitivity per se and VLDL secretion. The findings of the present study further support the concept that insulin resistance to suppression of fasting plasma FFAs and to the stimulation of peripheral tissue glucose uptake may not be the main pathogenic factors causing chronic fasting VLDL oversecretion in the insulin resistance syndrome, but that other associated factors may largely determine the fate of the excess FFA flux to the liver. Previous findings from our group (11,18) and others (52) point to hepatic insulin resistance/ hyperinsulinemia as an important determinant of VLDL overproduction. Therefore, while peripheral insulin resistance results in high FFA flux to the liver, liver insulin resistance and/or overinsulinization is likely to be required for the channeling of this excess supply toward VLDL synthesis and secretion and thus to establish clinically significant VLDL overproduction.

In conclusion, neither fasting plasma FFA levels nor the sensitivity of insulin-mediated suppression of plasma FFAs are predictors of fasting VLDL-TG secretion rate in healthy humans. BMI—and to a lesser extent, insulin sensitivity assessed by FSIVGTT with Bergman’s minimal model—are the main predictors of fasting VLDL-TG secretion rate in humans. We speculate that obesity may exert its major effects on VLDL production through mechanisms other than its association with extrahepatic insulin resistance, although our studies in no way preclude an important role of elevated postprandial and/or visceral adipose tissue fatty acid flux to the liver in insulin resistance.

This work was funded by a research grant from the Canadian Diabetes Association, and the National Institutes of Health (NIH) Biomedical Mass Spectrometry Resource (NIH RR-00954) and Clinical Nutrition Research Unit (NIH P30 DK56341). A.C. was the recipient of a Medical Research Council/Heart and Stroke Foundation of Canada fellowship and is currently a new investigator of the Canadian Institutes of Health Research. G.F.L. is a Career Investigator of the Heart and Stroke Foundation of Canada and Canada Research Chair in Diabetes (www.chairs.gc.ca).