Small Amounts of Dietary Medium-Chain Fatty Acids Protect Against Insulin Resistance During Caloric Excess in Humans

For decades, saturated fatty acids (FAs) have been considered detrimental to insulin sensitivity and cardiovascular health, as reflected by dietary guidelines. However, this consensus has recently been challenged (1,2), and it has been suggested to be of importance to consider the specific type of saturated FA (1). Saturated FAs are specified by their number of carbon atoms, and those with medium-chain length (MCFAs), like hexanoic (C6:0), octanoic (C8:0), capric (C10:0), and lauric (C12:0) acid, have received particular attention in metabolic research (3). Besides their presence in coconut oil and synthetic medium-chain triacylglycerol (MCT) oils, MCFAs make up 15–28% of FAs in bovine and human milk (4,5) and are enriched in palm kernel oil. Saturated long-chain FAs (LCFAs), like palmitic acid (C16:0) and stearic acid (C18:0), are typically enriched in animal fat but are also present in dairy products. Dairy products are thus a source of both saturated MCFAs and LCFAs.

Due to their lower lipophilicity, MCFAs primarily enter the portal vein in their nonesterified form after intestinal absorption (6,7). This is in contrast to the chylomicron incorporation and lymphatic absorption route of LCFAs. Therefore, intake of triacylglycerols (TGs) containing MCFAs leads to a blunted elevation in plasma TG concentration compared with intake of LCFA-rich TGs in humans (8). In rats, ingested MCFAs are predominately oxidized in the liver (9), suggesting that most MCFAs are metabolized there. Still, following intake of a diet that is very enriched in MCFAs (40% energy [40E%]), up to 17% of ingested MCFAs were found to reach the circulation in humans (10).

MCFAs are thought to enter mitochondria independently of the carnitine transport system, which facilitates the mitochondrial import and, hence, oxidation of LCFAs. Hence, when CPT was inhibited in isolated rat muscle (11) and in vivo in mice (12), oxidation of LCFA was inhibited, whereas oxidation of MCFAs was not. Moreover, carnitine independence by MCFAs has been demonstrated in isolated liver mitochondria (13,14).

In humans, MCFAs are shown to be oxidized to a greater extent than LCFAs (35–58% of ingested dose vs. 15–25%) following intake, as shown in both lean and obese individuals by use of FA tracers (15–17). Feeding mice or rats diets with 42–60E% fat comprised of coconut oil or MCT oil for 4–12 weeks maintained insulin and glucose tolerance at the level of chow diet in contrast to diets with saturated or unsaturated LCFAs (12,18–20), which induced insulin resistance. These studies in rodents imply that MCFAs have protective effects on glucose homeostasis during conditions of high energy and FA availability. Of note, although MCFAs are suggested to mainly be taken up and metabolized in the liver, increased skeletal muscle mitochondrial oxidative capacity has been reported with MCFA feeding in mice (21).

In a 14-year prospective study, intake of C6:0-C12:0 MCFAs from full-fat dairy products was associated with lower risk of developing type 2 diabetes (22). Dietary intervention studies investigating the effect of MCFAs on glucose homeostasis in humans are scarce and have not investigated the underpinning physiological and molecular mechanisms (23–25). The purpose of the present human study was thus to delineate whether inclusion of food sources with MCFAs has protective effects on whole-body, hepatic, and skeletal muscle insulin action and glucose metabolism in skeletal muscle during overfeeding. For this purpose, volunteers consumed a diet with caloric excess high in FAs. Healthy, lean men ingested, in a randomized order, a eucaloric control diet and a 3-day hypercaloric high-fat diet with 81–83E% from fat, corresponding to ∼450 g fat per day. For one group, the high-fat diet was enriched in saturated LCFAs, while the other group ingested a similar diet with a small fraction of the saturated LCFAs substituted with ∼30 g (5E%) MCFAs, derived from coconut oil, palm kernel oil, and dairy products.

Nine and eight young (mean ± SD age 23 ± 3 and 23 ± 3 years), healthy men were randomized to the MCSFA-HFD and LCSFA-HFD intervention groups in a parallel design (Fig. 2A). Subjects were normal weight (BMI 23.9 ± 2.1 [MCSFA-HFD] and 23.8 ± 1.9 [LCSFA-HFD] kg ⋅ m⁻²) and moderately exercise trained (maximal oxygen uptake 52.0 ± 2.7 and 52.2 ± 3.7 mL ⋅ kg LBM⁻¹ ⋅ min⁻¹). The study was approved by the Copenhagen Ethics Committee (KF 01 261127). A few data from the LCSFA-HFD intervention (described below) have previously been published (26), as specified in the legend of Table 1 and Figs. 2 and 3.

Habitual energy and macronutrient intake was registered by weighing of all ingested foods and drinks for 3 days and subsequently quantified by Dankost 2000 (Herlev, Denmark). Subjects consumed in randomized order a eucaloric control diet (CON) (63E% carbohydrate, 14E% protein, 24E% fat) and a hypercaloric (75% more energy) high-fat diet, enriched in either saturated LCFAs (LCSFA-HFD) or saturated LCFAs of which ∼30 g (5E%) was substituted by MCFAs (MCSFA-HFD) of C8:0, C10:0, and C12:0 origin (coconut oil, palm kernel oil, and dairy products). Both high-fat diets were matched in total fat content (∼82E%), equivalent to ∼450 g (Fig. 1 and Supplementary Table 1. The FA composition was calculated from a food product database (Dankost 2000, Herlev, Denmark). CON and high-fat diets were separated by 3 weeks. The hypercaloric diets were ingested for 3 days, preceded by 5 days of CON diet. In the CON trial, subjects ingested the eucaloric CON diet for 8 days (Fig. 2A). All food items were delivered, and subjects were supervised to ensure compliance.

Each subject completed two identical experimental days. For the 72 h before the experiment subjects abstained from strenuous physical activity, and on the day of the experiment the subjects arrived by passive transport under postabsorptive conditions, following a small meal (1.6 MJ) ingested at 5:00 a.m. (26), i.e., 7 h before the clamp. After 30 min rest, a venous catheter was inserted into an antecubital vein and teflon catheters were inserted in one femoral artery and one femoral vein under local anesthesia. After basal blood sampling, a bolus of 6,6-²H₂ glucose (3.203 mg · kg⁻¹) was administrated in the cubital vein, followed by constant infusion (0.055 mg · kg⁻¹ ⋅ min⁻¹) for 120 min during basal conditions. Subjects then underwent a 120-min hyperinsulinemic-euglycemic clamp (1.4 mU insulin ⋅ kg⁻¹ · min⁻¹), initiated with an insulin bolus (9.0 mU · kg⁻¹) (Actrapid; Novo Nordisk, Bagsværd, Denmark). During the clamp, 6,6-²H₂ glucose tracer was added to the nonlabeled glucose infusate, so tracer infusion followed the variable glucose infusion rate (GIR), which was continuously adjusted for maintenance of euglycemia. Femoral arterial blood flow was determined by ultrasound Doppler (Philips Ultrasound, Bothell, WA) every 20 min, concomitant with femoral arteriovenous blood sampling. Before and at the end of the clamp, indirect calorimetry was performed and vastus lateralis muscle biopsies were obtained under local anesthesia with 2–3 mL lidocaine.

Plasma glucose and blood lactate concentrations were determined with an ABL615 analyzer (Radiometer Medical, Brønshøj, Denmark). Plasma insulin was measured by ELISA (ALPCO, Salem, NH). The concentrations of FA (NEFA C kit; Wako Chemicals GmbH, Neuss, Germany), TG (GPO-PAP kit; Roche Diagnostics, Rotkreuz, Switzerland), HDL, LDL, and total cholesterol (HORIBA Medical, Kyoto, Japan) were measured by colorimetric kits according to the manufacturers’ instructions. Plasma epinephrine and norepinephrine concentrations were determined by radioimmunoassay (2-CAT ¹²⁵I RIA kit; Labor Diagnostika, Nordhorn, Germany. Plasma tumor necrosis factor α (TNFα) and interleukin 6 (IL-6) were measured on an AutoDELFIA (PerkinElmer, Waltham, MA) analyzer. Plasma enrichment of ²H glucose was measured by liquid chromatography–mass spectrometry (ThermoQuest Finnegan AQA, Thermoquest, Austin, TX).

Glycogen was measured in basal samples by a fluorometric method that detects glycosyl units after acid hydrolysis of freeze-dried muscle (27).

Glycogen synthase (GS) activity was measured in homogenates of basal and insulin-stimulated samples in the presence of 0.25 and 12 mmol ⋅ L⁻¹ glucose-6-phosphate (G6P), using a UNIFILTER 350 microtiter plate (Whatman, Headstone, U.K.) assay (28). GS activity was expressed as % of fractional velocity, calculated as 100 × activity in presence of 0.25 mmol · L⁻¹ divided by activity in the presence of 12 mmol/L G6P (saturated conditions).

Muscle lysis and Western blotting were conducted as previously described (29). The primary antibodies used were anti–pyruvate dehydrogenase-E1α (PDH-E1α) Ser³⁰⁰ and anti–PDH-E1α (Graham Hardie, University of Dundee, Dundee, Scotland, U.K.) and anti-AKT Ser⁴⁷³ and anti-AKT2 (cat. nos. 9271 and 3063; Cell Signaling Technology, Daners, MA). After 45 min incubation with horseradish peroxidase–conjugated secondary antibodies (Jackson ImmunoResearch, Westgrove, PA), signals were visualized (Bio-Rad ChemiDoc MP Imaging System, Hercules, CA) and quantified (Image Lab, version 4.0, Bio-Rad, Hercules, CA).

HOMA of insulin resistance (HOMA-IR) index was calculated as [basal insulin] · [basal glucose]/22.5. Insulin-stimulated leg glucose uptake was calculated as the arteriovenous blood glucose difference multiplied by blood flow. Endogenous glucose production was calculated from triplicate measures during the last 20 min of the basal conditions and during the last 20 min of the clamp period using the Steele equation. Oxidative glucose disposal was calculated from Vo₂ and Vco₂ values (4.55 · Vco₂ − 3.21 · Vo₂) ⋅ 1,000. Nonoxidative glucose disposal during the clamp was calculated as follows: (GIR + glucose R_a) − oxidative glucose disposal.

Data in Table 1 are means ± SE, while the subject characteristics described in the research design and methods are means ± SD. In Figs. 2 and 3., data are shown as univariate scatterplots where the bars are showing mean ± SE. The Shapiro-Wilk test was performed to test for normal distribution and Brown-Forsythe to test for variance homogeneity. Two-way repeated-measures ANOVAs were applied to test for effect of diet type (LCSFA-HFD or MCSFA-HFD) and effect of intervention (fat surplus) or effect of insulin and intervention within each group. When ANOVA revealed an interaction, Tukey post hoc test was used. A significance level of P < 0.05 was chosen. Statistical analyses were performed in GraphPad PRISM 8.

The data generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. No resources were generated.

Subjects ingested 12.2 ± 0.3 MJ/day during CON and 21.3 ± 0.3 MJ/day during the hypercaloric high-fat diets. Postabsorptive plasma glucose and insulin concentrations increased following LCSFA-HFD (Table 1) but were not altered following MCSFA-HFD (Table 1). HOMA-IR index increased by 90% following LCSFA-HFD (Fig. 2B). The basal endogenous glucose production, measured during basal conditions, was not affected by either type of hypercaloric high-fat diet (Fig. 2C). Postabsorptive plasma TG levels decreased by 26–31% with both high-fat interventions (Table 1). Plasma LDL and total cholesterol concentrations remained unchanged, while HDL cholesterol concentration increased 9–24% with both high-fat interventions (Table 1). The plasma level of the cytokine TNFα was increased 17% in LCSFA-HFD but remained unchanged in MCSFA-HFD (Table 1).

GIR was 17% lower following 3 days of LCSFA-HFD, while a small amount of MCSFA substitution prevented this reduction in insulin sensitivity (Fig. 2D and E). The endogenous glucose production, measured during the clamp, was not altered by either type of hypercaloric high-fat diet (Fig. 2F). In agreement, insulin-mediated suppression of endogenous glucose production relative to basal values (suppression of ∼60%) was not affected by either high-fat diet intervention (Fig. 2C and F). Whole-body glucose R_d was decreased by 21% in LCSFA-HFD, with no change in MCSFA-HFD (Fig. 2G). Insulin-stimulated leg glucose uptake was decreased by 26% in LCSFA-HFD but remained unchanged in MCSFA-HFD (Fig. 2H and I). Respiratory exchange ratio (RER) was lower for LCSFA-HFD during the clamp compared with CON (0.82 ± 0.02 vs. 0.90 ± 0.01) (Fig. 2J). RER was lower for MCSFA-HFD during both basal and insulin-stimulated conditions compared with CON (Fig. 2K). The increase in RER in response to insulin was reduced in LCSFA-HFD but not in MCSFA-HFD (Fig. 2L). The oxidative glucose disposal during the clamp was lower following both high-fat diet interventions compared with CON (Fig. 2M). Nonoxidative glucose disposal tended to be decreased in LCSFA-HFD (−20%, P = 0.06). In contrast, nonoxidative glucose disposal was increased by 20% in MCSFA-HFD compared with CON (Fig. 2N), which likely contributed to the preserved glucose disposal in MCSFA-HFD.

The resting basal muscle glycogen content was unchanged in LCSFA-HFD but reduced by 27% after MCSFA-HFD (Fig. 3A), despite similar energy and carbohydrate (9E%) intake during the interventions. The lower muscle glycogen content after MCSFA-HFD was accompanied by increased basal and insulin-stimulated muscle GS activity compared with CON (Fig. 3C). This was not the case after intake of LCSFA (Fig. 3B). These observations support the idea of an improved nonoxidative glucose disposal following MCSFA-HFD. During the clamp, the inhibitory phosphorylation of PDH-E1α at Ser³⁰⁰ in skeletal muscle was higher, and PDH-E1α protein content was lower following both interventions (Fig. 3D and E). High-fat intake thus increased the potential for β-oxidation–derived acetyl-CoA influx into the tricarboxylic acid cycle, potentially via inhibition of PDH and attenuated availability of acetyl-CoA from glycolysis, supporting the lower oxidative glucose disposal during the clamp. In MCSFA-HFD, a 42% increase compared with CON in leg lactate venous-arterial difference was observed (Fig. 3F), indicating maintained glycolytic flux despite PDH inhibition. There were no changes in insulin-stimulated AKT Ser⁴⁷³ phosphorylation with the interventions (Fig. 3G and H), indicating that proximal insulin signaling fails to explain the reduced glucose disposal in LCSFA-HFD.

Here, we demonstrate that substitution of only ∼30 g (5E%) saturated LCFAs with saturated MCFAs prevents both whole-body insulin resistance and impaired insulin-stimulated muscle glucose uptake induced by a hypercaloric saturated LCFA-rich diet in humans. Remarkably, both in the basal postabsorptive state and during the clamp, glucose metabolism and insulin sensitivity were preserved at the level of the eucaloric control diet, despite intake of ∼82E% fat and 75% caloric excess. The basal FA oxidation was increased with MCSFA inclusion. The main difference between the hypercaloric saturated high-fat diets was a greater availability of C8:0, C10:0, and C12:0 FAs from coconut oil, palm kernel oil, and dairy products in the MCSFA-HFD group, with a corresponding greater intake of C16:0 and C18:0 FAs by the LCSFA-HFD subjects. The type of FAs ingested thus had marked impact on substrate oxidation and insulin action.

In one other human study, insulin sensitivity was evaluated following substantial MCFA intake from MCT oil. In that study, whole-body insulin sensitivity was reported to be 17 and 30% greater in lean healthy and obese individuals with type 2 diabetes, respectively, following 5 days of eucaloric MCFA-rich diet with 40E% fat and a large percentage of MCFA (77% of the fat was MCFA) compared with a eucaloric LCFA-rich diet also with 40E% fat in a crossover design (24).

In the current study, the basal FA oxidation was markedly increased in MCFA-HFD compared with both CON and LCSFA-HFD. At the same time, the increase in RER during the clamp was actually larger in MCSFA-HFD compared with CON, while in LCSFA-HFD the increase in RER was attenuated compared with CON. This suggests preservation of glucose metabolic flexibility in MCSFA-HFD, which could play a role in maintenance of glucose disposal. Due to the almost 100% FA oxidation at basal conditions following MCSFA-HFD, absolute glucose oxidation during the clamp was still lower compared with CON (as in LCSFA-HFD). However, during the clamp nonoxidative glucose disposal was higher in MCSFA-HFD than in CON, likely linked to the lower basal muscle glycogen content and greater GS activity compared with CON. The substantial contribution of glycogen storage to insulin-stimulated glucose disposal is evident from tracer and MRS studies (30,31), and an inverse relationship between basal glycogen content and insulin-stimulated glucose uptake in muscle has been described in rat muscle (32). In the context of nonoxidative glucose disposal, a 42% greater leg lactate venous-arterial difference was observed following MCSFA-HFD compared with CON. This indicates that while some G6P was directed toward glycogen, glycolysis rate was also maintained, giving rise to lactate release. Preserved muscle insulin action after MCFA intake thus appeared due to both high glycogen synthesis and glycolytic flux.

An intriguing question is why muscle glycogen content was reduced when nonoxidative glucose disposal during the clamp was not compromised. We did not obtain any change in resting metabolic rate in the morning following the interventions (data not shown), confirming another human study in which 1 week of overfeeding with as much as 40E% MCFAs did not affect the basal metabolic rate (33). However, several studies have demonstrated that MCFA intake leads to an increased postprandial metabolic rate of 5–10% following MCFA intake of 15–50 g in both lean and obese individuals compared with saturated and unsaturated LCFAs (33–37). The mechanisms for this have not been revealed. In the liver, MCFAs could increase β-oxidation and ketogenesis (38), which could contribute to an increased postprandial energy expenditure. In rodents, acute MCFA administration (C8:0) was shown to activate hypothalamic proopiomelanocortin neurons and increase metabolic rate, pointing to additional central regulation (39). A slight increase in the metabolic rate following meals during the MCSFA-HFD intervention could thus be hypothesized to have increased total substrate oxidation and thereby reduced muscle glycogen, considering the low absolute carbohydrate intake of ∼100 g/day.

We did not observe intervention effects on endogenous glucose production with either high-fat diet during basal postabsorptive conditions or during the clamp. Insulin-stimulated glucose production was evaluated in response to a high, but still physiological, insulin infusion rate, which suppressed endogenous glucose production by an average of 60%. Since the different response of the GIR during the clamp following the two high-fat diets was largely explained by differences in leg glucose uptake, and endogenous glucose production was only suppressed by 60%, it is unlikely that a lower insulin infusion rate would have revealed differences in hepatic insulin sensitivity following the two diets.

The downregulation of insulin action in LCSFA-HFD was likely related to the energy surplus, as we previously showed that 6 weeks of a 64E% high-fat diet rich in saturated LCFAs did not alter insulin sensitivity in overweight men under eucaloric conditions (2). In the current study, the downregulation of glucose disposal was not related to impaired proximal insulin signaling. This observation is confirmed by other high-fat overfeeding studies in lean men and women showing intact insulin-stimulated AKT and TBC1 domain family member 4 (TBC1D4) phosphorylation, as well as phosphoinositide 3-kinase (PI3K) activity, following 3–5 days of high-fat diet (50–78E% fat) and 40–75% energy excess (40–42). This points toward a more downstream regulation of glucose uptake potentially at the step of GLUT4 transport or membrane association that is impaired with surplus LCSFA-HFD intake. We did find that the muscle insulin resistance in LCSFA-HFD was accompanied with an impaired glucose metabolic flexibility, 40% lower absolute glucose oxidation during the clamp, and also a 20% reduced nonoxidative glucose disposal (P = 0.06).

In conclusion, the findings in this study show that substitution of only 30 g saturated LCFA with MCFAs led to full reversal of insulin resistance at the whole-body level and specifically in the skeletal muscle induced by marked energy excess provided by a diet rich in saturated LCFAs. Minor intake of saturated MCFAs had metabolic impact on skeletal muscle in human individuals. Together, the findings suggest a potential of MCFA supplementation in regulation of both lipid and glucose homeostasis and also highlight the need for nuanced dietary guidelines on saturated FAs.

Acknowledgments. The authors acknowledge the skilled technical assistance of Irene Bech Nielsen and Betina Bolmgren and the experimental contributions from Jørgen Wojtaszewski (University of Copenhagen).

Funding. B.K. and E.A.R. were funded by The University of Copenhagen Excellence Program for Interdisciplinary Research (2016) “Physical Activity and Nutrition for Improvement of Health” and the Danish Council for Independent Research/Medicine (grant 4183-00249). A.-M.L. and A.M.F. were supported by a postdoctoral research grant from the Danish Diabetes Academy, funded by the Novo Nordisk Foundation, grant NNF17SA0031406. Furthermore, A.M.F. was supported by the Alfred Benzon Foundation. M.K. was supported by postdoctoral research grants from the Danish Council for Independent Research/Medicine (grant 4004-00233) and Lundbeckfonden (grant R288-2018-78). K.A.S. was supported by a postdoctoral research grant from the Danish Council for Independent Research/Medicine, grant 4092-00309.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. B.K., E.A.R., A.-M.L., and K.A.S. designed the study and carried out the experiments. A.-M.L., A.M.F., and K.A.S. contributed to the results. A.-M.L., A.M.F., M.K., and B.K. wrote the manuscript. All authors contributed in the writing process of the manuscript and approved the final version of the manuscript. B.K. 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.