Exercise Prevents Fructose-Induced Hypertriglyceridemia in Healthy Young Subjects

Eight healthy, nonobese, male volunteers aged 21.5 ± 2.7 years, with mean ± SD body weight 68.5 ± 7.0 kg, height 1.76 ± 0.03 m, and BMI 22.1 ± 1.9 kg/m², were studied. The subjects were sedentary, defined as undergoing <60 min exercise per week, and nonsmokers, who were not taking medication and had no history of diabetes. The experimental protocol was approved by the ethics committee of Lausanne University School of Medicine. All participants provided informed, written consent. The experimental, randomized crossover design is illustrated in Fig. 1. Each of the eight volunteers was studied on three different occasions after having followed three different diet and physical activity programs during 4 days as follows:

• • Control (C): Subjects received a low-fructose, weight-maintenance diet containing 1.5 times their basal energy requirements calculated using the Harris-Benedict equation (31), which was composed of 50% complex carbohydrates, 5% sugars (mainly lactose), 30% fat, and 15% protein, and performed minimal physical activity (<30 min walking/day and no other exercise).

• • High-fructose diet (HFr): Subjects received a weight-maintenance, high-fructose diet containing 1.5 times their basal energy requirements composed of 20% complex carbohydrates, 5% nonfructose sugars, 30% fructose provided as lemon-flavored drinks, 30% fat, and 15% protein and performed minimal physical activity.

• • High-fructose diet plus exercise (HFrEx): Subjects received a high-fructose diet containing 1.7 times their basal energy requirements and completed two 30-min cycling exercise sessions at a power output of 125 W—one at noon and one at 5:00 p.m. on each day during 4 days. The physical activity factor of 1.7 was calculated at the time of protocol submission for a 22-year-old male weighing 75 kg and measuring 175 cm, with a predicted basal metabolic rate of 1.2845 kcal/min calculated using the Harris-Benedict equation (31). The predicted 24-h sedentary energy requirement was calculated assuming a nighttime energy expenditure of 8 h at a physical activity level of 1.0 and a daytime energy expenditure of 16 h at a physical activity level of 1.75, totaling 1.5 times basal energy requirements over 24 h or 2,775 kcal/day. The predicted 24-h energy requirements for HFrEx was estimated to be 3,205 kcal/day by assuming an exercise energy efficiency of 25% and equivalent to an extra 430 kcal/day, which corresponds to 1.7 times basal energy requirements.

Diets were calculated individually for each subject by a research nutritionist and were distributed each day as prepacked food items and drinks to be consumed at home. Subjects were instructed to consume all the food and beverages that were provided according to detailed instructions and not to consume anything else except unsweetened tea. All exercise sessions were performed at the Department of Physiology under supervision of one of the investigators.

On the fifth day of each condition, subjects reported to the Clinical Research Centre at 7:00 a.m. after an 11-h fast. Upon their arrival, subjects were weighed and their body composition was measured using bioelectrical impedance plethysmography (Imp DF50; ImpediMed, Eight Mile Plains, Australia). After subjects were lying in a bed, a catheter was inserted into a vein of the right arm for blood sampling and was maintained patent by a slow infusion of normosaline. Another catheter was inserted into a vein of the left arm and was used for the administration of a primed continuous infusion of [6,6-²H₂]glucose (bolus, 2 mg/kg body wt, and continuous infusion, 0.02 mg/kg/min; Cambridge Isotope Laboratories, Andover, MA) throughout the test. Oral loads of fructose (0.2 g/kg fat-free mass, enriched with 0.1% [U-¹³C₆]fructose; Cambridge Isotope Laboratories) were given every hour for 9 h. No other food was provided, and fructose intake corresponded to 78 ± 5% total energy expenditure during the experimental period. A bolus of 100 μmol/kg [1,1,2,3,3-²H₅]glycerol (Cambridge Isotope Laboratories) was administered, and TRL-TG kinetics was calculated from the modeling of the decrease in [²H₅]glycerol enrichment in TRL-TG over time. This method provides a single value over several hours. For this reason, the bolus of [²H₅]glycerol was administered after the administration of the third dose of fructose (time 120 min) to allow 2 h for stimulation of hepatic fructose uptake and metabolism. Blood samples were collected at baseline (t = 0) and after 60, 120, 140, 160, 180, 240, 300, 360, 420, 480, and 540 min. Blood pressure was measured at baseline using an automatic blood pressure device (Omron 907; Omron, Hoofddorp, the Netherlands). Energy expenditure and net substrate oxidation rates were monitored over the last 3 h of the test by open-circuit indirect calorimetry (Deltatrac II; Datex Instrument, Helsinki, Finland). However, owing to technical problems with the indirect calorimeter, results could be obtained for only six of the eight participants. Urine was collected throughout the day to determine urea nitrogen excretion rate.

Plasma was immediately separated from blood cells by centrifugation at 3,600g for 10 min at 4°C, and aliquots were stored at −20°C until assayed. Plasma metabolites (glucose, TG, nonesterified fatty acids [NEFAs], cholesterol, and lactate) (Randox Laboratories, Crumlin, U.K.), β-hydroxybutyric acid (BHB) (Roche Diagnostics Hitachi, Rotkreuz, Switzerland), and urinary urea (Randox Laboratories, Crumlin, U.K.) were measured by enzymatic methods; insulin and glucagon (Millipore, Billerica, MA) were assessed by radioimmunoassay. Apolipoprotein (apo)B48 was measured by ELISA using a kit from Shibayagi, Shibukawa, Japan.

The TRL fraction (Svedberg flotation unit [Sf] >20) was separated by ultracentrifugation (17 h at 45,000 rpm at 4°C) in an Optima L-90 K ultracentrifuge (Beckman Coulter, Brea, CA) in a fixed-angle rotor (50.3 Ti; Beckman Coulter). After plasma deproteinization and partial purification over anion- and cation-exchange resins, plasma [²H₅]glycerol and [6,6-²H₂]glucose were acetylated in the presence of acetic anhydride and pyridine and their enrichments were measured by GC-MS (Agilent Technologies, Santa Clara, CA) in chemical ionization mode, with selective monitoring of m/z 331 and 333 for glucose and 159 and 164 for glycerol. For [¹³C]palmitate enrichment and concentration, total lipids were extracted from plasma and fatty acid methyl esters (FAMEs) were prepared from TG fractions. The ratio of ¹³C to ¹²C in the FAME derivatives was ascertained by using Δ Plus XP GC-combustion isotope ratio MS (Thermo Electron, Bremen, Germany). Tricosanoic acid methyl ester was used as an isotopic enrichment standard, and a quality-control sample (certified standard of eicosanoic acid FAME; Department of Geological Sciences, Indiana University, Bloomington, IN) was run with each set of samples.

LDL size and subclasses were determined in frozen samples from baseline measurements. For analysis of LDL size and subclasses, nondenaturing PAGE of plasma was performed and analyzed and LDL subclass distribution (Class I-IVb) was calculated as previously described (10).

TRL-TG production was calculated as follows: the fractional turnover rate (FTR) of TRL-TG was determined by using compartmental modeling analysis as previously described (32). The rate of TRL-TG secretion (in millimoles per hour), which represents the amount of TRL-TG entering the bloodstream, was calculated by multiplying the FTR of TRL-TG (in pools/hour) by the pool of TRL-TG in plasma (in millimoles). The clearance rate of TRL-TG (in millimeters plasma per minute), which is an index of the removal efficiency of TRL-TG from the systemic circulation, was calculated by dividing the TRL-TG secretion rate (in millimoles per minute) by the TRL-TG concentration (in millimoles per milliliter).

Plasma [¹³C]TRL-palmitate (nmol/L) was calculated as follows: [¹³C]TRL-TG-palmitate isotopic enrichment [atom percent excess] × weight % palmitate × [TRL-TG (nmol/L)]. [¹³C]TRL-palmitate production (nanomoles per hour) was calculated by multiplying the FTR of TRL-TG (pools/hour) by the pool of [¹³C]TRL-palmitate (nanomoles).

The contribution of gluconeogenesis from fructose to endogenous glucose production [EGP_(F) in grams per kilogram per hour] was calculated as follows:

where EGP is the endogenous glucose production (grams per kilograms per hour) calculated with [6,6-²H₂]glucose (20), ¹³CG is the isotopic enrichment of plasma glucose (atom percent excess), G is the glucose concentration (grams/liter), ¹³CF is the isotopic enrichment of oral fructose, p is the pool fraction (set at 0.75), V is the glucose distribution space (set at 0.2 times body weight), and T is the time (hours).

Glycogen synthesis (GS) was estimated as:

where net CHO_ox (oxidized carbohydrates) was calculated using standard indirect calorimetry equations (20).

All values are expressed as means ± SEM. Log or box-cox transformation was applied to skewed data before statistical analysis. Values obtained at time 0 were used as fasting values. Changes in body weight, body fat content, blood pressure, and fasting parameters were assessed by using ANOVA for repeated measures, followed by Tukey post hoc tests for multiple comparisons.

Metabolic effects of fructose loading in the C condition were expressed as the percent change from T = 0 to T = 540 min. The effects of fructose loading over time, dietary condition with or without exercise, and their interaction were evaluated by a two-way ANOVA for repeated measures with interaction, followed by Tukey post hoc tests for multiple comparisons.

TRL-TG production and clearance, [¹³C]TRL-palmitate production, and averaged postfructose gluconeogenesis from fructose and GS were compared using ANOVA for repeated measures, followed by Tukey post hoc tests.

The prespecified primary outcome was postfructose [¹³C]TRL-palmitate concentrations. Power analysis was based on our previous study comparing the effects of oral fructose in male and females (33) and indicated that a sample size of six was required to detect an effect size of 0.5 with a power of 80%.

All statistical calculations were performed with Stata 10 (Stata, College Station, TX). P < 0.05 was considered statistically significant.

There was no difference in body weight (C 68.7 ± 2.6 kg, HFr 68.8 ± 2.7 kg, and HFrEx 68.9 ± 2.7 kg), body fat content (C 17.9 ± 2.3%, HFr 17.8 ± 2.3%, and HFrEx 17.7 ± 2.2%), or blood pressure (C 118/63 ± 3/2 mmHg, HFr 115/65 ± 2/2 mmHg, and HFrEx 119/61 ± 3/2 mmHg) after the three different dietary conditions.

HFr significantly increased fasting plasma TG, TRL-TG, apoB48, and glucose concentrations (all P < 0.05 [Fig. 2, Fig. 3, and Table 1]). However, it did not significantly change fasting cholesterol, HDL cholesterol, insulin, glucagon, NEFA, glycerol, BHB, or lactate concentrations (NS [Fig. 2, Fig. 3, and Table 1]). LDL size and LDL subclass distribution were not different after the HFr diet compared with the C diet, except for a small increase in LDL IIa (P < 0.05) (Table 1).

HFrEx significantly blunted the effects of HFr on plasma TG and normalized fasting total TG, TRL-TG, and apoB48 concentrations. It also significantly increased LDL-I particles (P < 0.05 vs. HFr) (Table 1) and decreased small LDL particles (III + IV) compared with C (P < 0.05 vs. C) (Table 1). Fasting plasma glucose concentrations, however, remained unchanged compared with HFr. HFrEx also increased fasting BHB concentrations (P < 0.05 vs. HFr). All other parameters were unchanged (Fig. 2, Fig. 3, and Table 1).

To further focus on the postprandial metabolic effects of fructose, we measured several metabolic parameters relevant to lipid and glucose metabolism during a 9-h fructose loading test. In C, fructose ingestion caused a modest increase in total TG and TRL-TG (P < 0.05) (Fig. 2A and B) and a marked suppression of NEFA (P < 0.005) (Fig. 2C), glycerol (P < 0.01, Fig. 2D), and BHB concentrations (P < 0.0005) (Fig. 2E). [¹³C]TRL-palmitate concentrations increased (P < 0.05), indicating conversion of fructose into lipids (Fig. 2F). Although a substantial amount of ingested fructose was converted into glucose and released in the systemic circulation, plasma glucose (Fig. 3A) and insulin concentrations were not increased (Fig. 3B) compared with baseline. There was also a significant increase in lactate concentration, most likely from splanchnic fructose metabolism (P < 0.0005) (Fig. 3C).

Compared with C, HFr enhanced postfructose total TG and TRL-TG concentrations (P < 0.0001) (Fig. 2A and B) and [¹³C]TRL-palmitate concentration (P < 0.0001) (Fig. 2F) and decreased postfructose NEFA concentrations (P < 0.05) (Fig. 2C). TRL-TG secretion was not significantly increased, but TRL-TG clearance tended to decrease (Table 2). [¹³C]TRL-palmitate production was increased 2.9-fold (Table 2) compared with C, but the difference did not reach statistical significance owing to unexpected, very high interindividual variations. Gluconeogenesis from fructose was similar to that with C (Table 3), but glucose concentrations were significantly increased (P < 0.0001) (Fig. 3A).

In the HFrEx condition, postfructose total TG, TRL-TG, and [¹³C]TRL-palmitate concentrations were completely normalized (all NS compared with C) (Fig. 2A, B, and F). TRL-TG clearance was increased by 66% compared with HFr (P < 0.05 vs. HFr) (Table 2). TRL-TG secretion was not significantly changed, but [¹³C]TRL-palmitate production was markedly decreased compared with HFr (P = 0.074 vs. HFr) (Table 2). Furthermore, NEFA and glycerol concentrations were increased compared with both HFr and C (both P < 0.0001 vs. HFr) (Fig. 2C and D). Gluconeogenesis from fructose and plasma glucose concentrations remained increased to the same extent as HFr (NS) (Table 3 and Fig. 3A), while insulin concentrations were slightly decreased (P < 0.05 vs. HFr) (Fig. 3B). Lactate concentrations were significantly decreased compared with those observed with C (P < 0.005) (Fig. 3C). Net lipid oxidation was increased (P < 0.0005 vs. HFr) (Table 3) while net carbohydrate oxidation was decreased (P < 0.001 vs. HFr) (Table 3), and net GS was nearly doubled compared with HFr and C (P < 0.05 vs. HFr) (Table 3). Energy expenditure was increased with HFrEx but only compared with C (P < 0.05) (Table 3). Mean LDL size was not significantly different from HFr or C, but the amount of large, less atherogenic LDL subclass I particles was significantly increased after HFrEx compared with HFr (P < 0.05) (Table 1), and small dense LDL particles (III + IV) were significantly decreased compared with C (P < 0.05) (Table 1).

The current study was specifically designed to evaluate whether fructose, when administered together with a weight-maintenance diet, causes significant alterations of blood lipids and whether these effects can be modulated by exercise. For this purpose, we selected a design in which fructose was administered to the same subjects on three occasions, i.e., after having consumed a low-fructose diet without exercise and after having received a high-fructose diet with and without exercise. Total energy intake was adapted to meet energy expenditure on all three occasions. Fructose intake corresponded to 30% total energy intake on both occasions but was higher in absolute values when subjects exercised.

With low physical activity, an HFr diet over 4 days led to an increase in total plasma TG and TRL-TG concentrations both after an overnight fast and after fructose loading. There was no significant change in TRL-TG kinetics after fructose loading, most likely due to the small number of subjects included and to a relatively large interindividual variability. It appears likely, however, that decreased TRL-TG clearance contributed to the rise in TRL-TG, since this parameter was previously reported to decrease after fructose ingestion (34) and showed a strong trend toward a 22% reduction. Increased TRL-TG production has also been reported after high-carbohydrate diets (35) and may have contributed as well.

During the fructose-loading experiments, oral fructose was labeled with ¹³C, and incorporation of ¹³C into TRL-TG palmitate could be documented. Furthermore, the increase in [¹³C]TRL-palmitate concentration was markedly enhanced when fructose was ingested after consumption of a high-fructose diet versus a low-fructose diet without exercise. This suggests that chronic fructose intake upregulated hepatic de novo lipogenesis. However, it is also possible that an increase in gut fructose absorption, induced by chronic fructose intake, contributed to increase the total fructose load delivered to the liver (36).

In humans, carbohydrate-induced de novo lipogenesis occurs essentially in the liver, although massive carbohydrate overfeeding may also stimulate this pathway in adipose tissue (37,38). Hepatic de novo lipogenesis has further been suggested to make a significant contribution to fructose- or sucrose-induced hypertriglyceridemia through enhanced secretion of hepatic VLDL (21,39). Interestingly, fasting apoB48 concentrations nearly doubled with HFr. Since this apolipoprotein is exclusively synthesized in enterocytes, this observation indicates that a high-fructose diet stimulates the secretion of intestinal, chylomicron-like particles even after an overnight fast. Activation of this pathway has indeed been documented in high fructose–fed hamsters (40). However, the contribution of intestinal lipogenesis to the fructose-induced increase in TRL-TG cannot be estimated from the mere increase in apoB48 concentration.

When the high-fructose diet was associated with exercise, all the effects of fructose on lipoprotein metabolism were totally prevented. There was indeed complete normalization of fasting and postprandial TRL-TG concentration, which was essentially due to a 65% increase in TRL-TG clearance and suggests that exercise-enhanced lipoprotein lipase activity facilitates the disposal of lipids in adipose cells or skeletal muscle fibers. In addition to this accelerated TRL-TG removal from the circulation, exercise also very dramatically decreased [¹³C]TRL-palmitate concentrations and secretion. This indicates that exercise inhibited de novo lipogenesis and that this may also have contributed to the TG-lowering effects of exercise. Consumption of high-fructose or high-sucrose diets has also been shown to decrease the concentration of LDL subclass I particles, which have a lower atherogenic potential than other LDL subclasses (10,41). In the current study, HFr failed to significantly alter LDL particle size and the proportion of LDL subclass I, possibly due to the short duration of diet administration. However, LDL subclass I was significantly increased with HFrEx. This is consistent with previous reports showing that exercise has beneficial effects on lipoprotein profiles in obese dyslipidemic patients (42) and in healthy subjects fed a high-carbohydrate diet (30) and adds to the evidence that it can efficiently prevent the adverse consequences of a high-fructose diet.

Unexpectedly, exercise restored normal fasting apoB48 concentrations, indicating that exercise regulated not only hepatic but also intestinal fructose metabolism. However, the functional significance of gut fructose metabolism remains to be more fully evaluated.

All of these effects of exercise were observed even though the additional energy expended during physical activity (∼430 kcal/day) was compensated for by increased total energy and fructose intakes indicating that exercise potently impacts on fructose metabolic pathways independently of changes in overall energy balance.

We can only speculate on the mechanisms leading to these effects. Exercise may indeed cause multiple metabolic alterations, which converge to reduce plasma TG concentrations. First, acute exercise enhances fructose conversion into glucose and lactate and their use as energy providing substrates by the working muscle (43), which may have decreased the availability of fructose carbons for de novo lipogenesis during the days preceding the oral fructose test. Second, exercise-induced hepatic and muscle glycogen depletion can be expected to result in an enhanced conversion of fructose into hepatic glycogen when fructose is subsequently ingested, thus diverting fructose away from hepatic de novo lipogenesis. Our observation that exercise increased net GS after fructose loading is entirely consistent with this hypothesis. Third, acute exercise increases LPL activity in skeletal muscle (44,45), resulting in enhanced TRL-TG clearance. Finally, exercise stimulates lipolysis, and the NEFA, released into the systemic circulation, may activate the nuclear receptors peroxisome proliferator–activated receptor-α in the liver and possibly in enterocytes as well to stimulate fat oxidation and reduce hepatic and intestinal de novo lipogenesis (46).

Besides its effect on plasma TG, HFr also increased slightly, but significantly, fasting and postfructose plasma glucose concentrations. Other studies have also documented that short-term fructose overfeeding increases basal hepatic glucose output and impairs the ability of low-dose insulin to suppress hepatic glucose output (21,22). These alterations of glucose homeostasis can be observed within a few days of fructose overfeeding and may be due to fructose increasing hepatic glycogen stores (47). It may therefore correspond to a physiological adaptation to important changes in sugar intake rather than to pathological adverse effects of fructose.

Our study has several limitations that need be addressed. First, energy intake was calculated to match energy requirements based on calculated energy expenditure. This was done using a well-accepted, reliable equation to predict basal energy expenditure (31) and a physical activity level of 1.5. However, since the 24-h energy expenditure was not actually measured, it is possible that energy balance was not reached for every subject. This may have affected our results to some extent, since the effect of exercise to lower plasma TG is known to be significantly blunted when energy expended during exercise is fully replaced compared with negative energy balance (48). Second, incorporation of ¹³C administered as [¹³C₆]fructose into TRL-TG-palmitate was used as an indirect estimate of de novo lipogenesis. This is a relatively crude method that does not allow a quantitative estimate of fatty acid synthesis owing to the fact that the isotopic enrichment of intrahepatic acetyl-CoA, the actual precursor for fatty acid synthesis, was not measured. There may also be an underestimation of hepatic fatty acid synthesis due to the short period of tracer administration, which may have been insufficient for equilibration of newly synthesized fatty acids and delayed secretion of TG from the intrahepatic pool into VLDL (49). Third, this method provides an estimate of the contribution of de novo lipogenesis to TRL-palmitate secretion but not to intrahepatic TG storage, and we cannot directly assess the effect of fructose or exercise on the latter pathway. Fourth, we cannot discard the hypothesis that exercise may alter plasma lipoprotein metabolism irrespective of dietary fructose content. Finally, this study evaluated the effects of a very high fructose intake (30% total energy, which far exceeds the average U.S. per capita daily fructose consumption [~10% total energy] [50]) over a 4-day period and, hence, does not provide information on the effects of fructose in the general population or over longer periods of time. It does, however, demonstrate that exercise can efficiently reduce some potentially adverse effects of fructose.

In summary, we observed that substitution of fructose for starch in a weight-maintenance diet increased plasma total and TRL-TG. These alterations may increase cardiometabolic risk but were observed at a very high fructose intake (∼200 g/day), largely exceeding the average fructose intake observed in the population (50). Even with such a high intake, exercise completely prevented fructose-induced alterations of lipid metabolism.

This study was supported by grant 320030_138428 from the Swiss National Foundation for Science.

L.T. has received research support from Nestle SA and Ajinomoto Co., Inc., for research unrelated to this article. No other potential conflicts of interest relevant to this article were reported.

L.E. designed the study, recruited participants, performed tests, analyzed data, performed statistical analysis, drafted the manuscript, and revised the manuscript. V.L. analyzed data, performed statistical analysis, and revised the manuscript. F.T. and V.C. recruited participants, performed tests, and revised the manuscript. L.H. analyzed the isotopic enrichment of plasma lipids and revised the manuscript. P.S. designed the study and revised the manuscript. B.M. and B.W.P. calculated VLDL-TG kinetics and revised the manuscript. B.A.F. analyzed the isotopic enrichment of plasma lipids and revised the manuscript. P.A.G. analyzed LDL subfractions and revised the manuscript. V.G. revised the manuscript. K.B. analyzed LDL subfractions and revised the manuscript. L.T. designed the study, analyzed LDL subfractions, drafted the manuscript, and revised the manuscript. L.T. 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.

Parts of this study were presented in abstract form at the 76th and 77th Scientific Sessions and Annual Meetings of the American Society of Nutrition in conjunction with Experimental Biology, San Diego, California, 21–25 April 2012, and Boston, Massachusetts, 19–24 April 2013.

The authors thank the staff of the Department of Physiology of Lausanne and of the Clinical Research Centre for their outstanding assistance and all the volunteers for their participation.