If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Many successful ultra-endurance athletes have switched from a high-carbohydrate to a low-carbohydrate diet, but they have not previously been studied to determine the extent of metabolic adaptations.
Methods
Twenty elite ultra-marathoners and ironman distance triathletes performed a maximal graded exercise test and a 180 min submaximal run at 64% VO2max on a treadmill to determine metabolic responses. One group habitually consumed a traditional high-carbohydrate (HC: n=10, %carbohydrate:protein:fat = 59:14:25) diet, and the other a low-carbohydrate (LC; n=10, 10:19:70) diet for an average of 20 months (range 9 to 36 months).
Results
Peak fat oxidation was 2.3-fold higher in the LC group (1.54 ± 0.18 vs 0.67 ± 0.14 g/min; P=0.000) and it occurred at a higher percentage of VO2max (70.3 ± 6.3 vs 54.9 ± 7.8%; P=0.000). Mean fat oxidation during submaximal exercise was 59% higher in the LC group (1.21 ± 0.02 vs 0.76 ± 0.11 g/min; P=0.000) corresponding to a greater relative contribution of fat (88 ± 2 vs 56 ± 8%; P=0.000). Despite these marked differences in fuel use between LC and HC athletes, there were no significant differences in resting muscle glycogen and the level of depletion after 180 min of running (−64% from pre-exercise) and 120 min of recovery (−36% from pre-exercise).
Conclusion
Compared to highly trained ultra-endurance athletes consuming an HC diet, long-term keto-adaptation results in extraordinarily high rates of fat oxidation, whereas muscle glycogen utilization and repletion patterns during and after a 3 hour run are similar.
The concept that a diet high in carbohydrate is necessary for optimizing exercise performance gained credence in the late 1960s when it was discovered that muscle glycogen depletion was associated with fatigue, and that a high-carbohydrate diet maintained muscle glycogen and performance [
] reinforce the supremacy of the high-carbohydrate fueling paradigm for maximizing performance and recovery.
Less appreciated is the perspective that there is no essential requirement for dietary carbohydrate because humans possess a robust capacity to adapt to low carbohydrate availability. Around the same time the importance of glycogen was recognized in the 1960s, ground-breaking work on the metabolic adaptations to starvation revealed an elegant mechanism by which humans switch to using lipid-based fuels [
]. After a few weeks of starvation when glycogen levels significantly decrease, hepatic ketone production increases dramatically to displace glucose as the brain's primary energy source, while fatty acids supply the majority of energy for skeletal muscle. Glucose production from non-carbohydrate sources via gluconeogenesis supplies carbons for the few cells dependent on glycolysis. Similar metabolic adaptations favoring near-exclusive reliance on lipid-based fuels occur after several weeks of a ketogenic diet when carbohydrate is restricted to very low levels, protein is kept moderate, and dietary fat is emphasized [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
A critical mass of experimental data on ketogenic diets and ketone physiology has been generated in the last decade demonstrating the safety and therapeutic efficacy in managing a range of clinical conditions [
]. Beyond the clinical applications, it was demonstrated more than three decades ago that consumption of a ketogenic diet with less than 10 g of carbohydrate per day for 4 weeks did not compromise endurance performance in elite cyclists [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
]. These keto-adapted cyclists derived >90% of their fuel from the oxidation of fat while exercising at 64% maximal oxygen consumption, corresponding to an average fat oxidation of 1.5 g/min.
Maximal rates of fat oxidation in humans are generally thought to be much less, even in highly trained endurance athletes with very high aerobic capacity. In the largest study to date that examined peak rates of fat oxidation during exercise, the highest reported value in any single individual was 1.0 g/min [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
Fat utilization during exercise: adaptation to a fat-rich diet increases utilization of plasma fatty acids and very low density lipoprotein-triacylglycerol in humans.
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
]. This is likely due to a higher carbohydrate intake and insufficient time to allow full keto-adaptation, thus resulting in less pronounced shifts in fuel use.
There is a distinct lack of research on keto-adapted athletes who have been restricting carbohydrate for several consecutive months despite the fact that many endurance athletes have adopted a low-carbohydrate lifestyle in the past few years. Metabolic studies in this group may yield new insights into human capacities to adapt and thrive under different nutritional inputs. Therefore, we designed the FASTER (Fat Adapted Substrate use in Trained Elite Runners) study to compare metabolic differences between competitive ultra-marathoners and ironman distance triathletes consuming low-carbohydrate (LC) and high-carbohydrate (HC) diets.
2. Methods
2.1 Experimental Approach
This was a cross-sectional study comparing two groups of elite ultra-endurance athletes habitually consuming either an LC or HC diet. Athletes were carefully matched for age, physical characteristics, primary competition distance, and competition times. With one exception, all athletes lived in the United States and traveled via plane or car to our laboratory for two consecutive days of testing. On Day 1, participants performed a maximal oxygen consumption test during which peak fat oxidation was determined. Day 2 consisted of a treadmill run at 64% of maximal oxygen consumption for 3 hours to determine metabolic responses before, during, and after exercise. Subjects were informed of the purpose and possible risks of the investigation prior to signing an informed consent document approved by the institutional review board.
2.2 Participants
We targeted male ultra-endurance runners 21–45 years of age consuming an LC (n=10) or an HC (n=10) diet who were in the top 10% of finalists competing in sanctioned running events ≥50 km and/or triathlons of at least half ironman distance (113 km). Several athletes had sponsors (55%), course records (30%), national records (10%), international records (10%), and national/international-level appearance for Team USA (25%). Interested athletes completed questionnaires to assess their medical, diet, training, and running competition histories. As a pre-screening technique to determine if participants met dietary inclusion criteria, a registered dietitian had at least one phone call and email communication with each athlete about their habitual diet. Eligible participants were required to measure and weigh all foods consumed and provide a detailed description for 3 days (2 weekdays and 1 weekend day). All subjects were contacted through phone calls and emails to obtain home recipes, local restaurant menus, and nutritional information as necessary for determining accurate nutrient information. Habitual diet information was entered into commercial nutrient analysis software (Nutritionist Pro™, Axxya Systems, Stafford, TX). Subjects consuming an LC diet, defined as <20% energy from carbohydrate and >60% from fat, consistently for at least 6 months were eligible for the LC group. Subjects consistently consuming >55% energy from carbohydrate were considered for the HC group. Athletes were excluded if they did not consume the appropriate diet for the allotted amount of time or had any health issues, including, but not limited to diabetes, heart disease, kidney, liver, or other metabolic or endocrine dysfunction, current injury, anti-inflammatory medication use, anabolic drug use, or prone to excessive bleeding.
2.3 Exercise Protocols
2.3.1 Maximal Aerobic Capacity
Athletes were in racing condition but refrained from any competitions for a minimum of 7 days prior to testing. They were instructed to maintain their habitual diet leading up to testing and to record their diet for the day before and the day of travel (Testing Day 1). Subjects arrived at the laboratory between 1600 and 1900 hours following a 4 hour fast. Upon arrival, height and body weight were recorded, after which athletes were familiarized with the equipment and procedures prior to completing a maximal oxygen consumption (VO2max) test on a motorized treadmill (956i Treadmill, Precor, Woodinville, WA) using indirect calorimetry (TrueOne 2400, ParvoMedics, Sandy, UT). Subjects were fitted with a facemask and headgear (7450 Series Silicone V2™ Oro-Nasal Mask, Hans Rudolph, Shawnee, KS) with an adaptor ring to attach the two-way air chamber and gas collection hose to the mask (Adapter 7450 V2™, Hans Rudolph, Shawnee, KS). The protocol involved a 3 min walking warm-up at 3.5 mph, followed by 2 min stages that incrementally increased speed and incline linearly to elicit ~5% increase in VO2 at each stage. The test was considered complete when subjects voluntarily stopped the treadmill, which occurred within 18 min. Ratings of perceived exertion (RPE) on a 0–10 scale and heart rate were measured at each stage.
Breath-by-breath gas exchange measurements of VO2 and VCO2 were collected and recorded every 30 s and used to calculate oxygen uptake, minute ventilation, carbohydrate and fat oxidation rates, and respiratory exchange ratio (RER) with the assumption that the urinary nitrogen excretion was negligible. Peak fat oxidation was determined from the highest recorded 30 s interval. Following testing, subjects consumed a meal (dinner) consistent with the macronutrient percentages of their habitual diet. After dinner, all subjects were instructed to fast overnight, restrict caffeine and abstain from taking over-the-counter medications. To ensure hydration, liberal consumption of water was strongly encouraged for the rest of the evening.
2.3.2 Submaximal Exercise Protocol
Athletes returned to the laboratory for testing the following morning at 0600 after a 10 hour overnight fast (Fig. 1). Each subject provided a small urine sample to assess specific gravity (Model A300CL, Spartan, Japan) as a measure of hydration (all subjects had a USG <1.025). Body composition was determined via dual-energy X-ray absorptiometry (DXA) (Prodigy, Lunar Corporation, Madison, WI), height was measured to the nearest 0.1 cm and total body weight was recorded to the nearest 0.1 kg on a digital scale (OHAUS, Fordham Park, NJ). The facemask and headgear were used to determine resting energy expenditure by indirect calorimetry for 10 min after which a seated blood sample was obtained from a forearm vein. Subjects were then moved to a hospital bed for muscle biopsy preparation and acquisition.
Fig. 1Experimental protocol to determine metabolic responses to submaximal exercise.
After these baseline measures, subjects consumed a shake (5 kcal/kg body mass) with a macronutrient distribution (%carbohydrate:fat:protein) of 5:81:14 in the LC group and 50:36:14 in the HC group. On average, this translated into shakes consisting of 343 kcal, 4.3 g carbohydrate, 31.3 g fat, and 12.6 g protein for the LC group and 332 kcal, 42.7 g carbohydrate, 13.7 g fat, and 12.4 g protein in the HC group. Proportions varied, but both the LC and HC shakes were made from heavy cream, olive oil, whey protein, walnut oil, and strawberries. The HC shake also contained banana and agave syrup. Subjects rested quietly for 90 min after ingestion of the shake. Indirect calorimetry measurements were taken during the last 10 min and a second pre-exercise blood sample was taken.
Subjects then moved to the treadmill and began running at an intensity equivalent to 64% VO2max for 180 min. Slight adjustments to the pace were made during the first 10 min based on real-time measurements of oxygen uptake. During the endurance exercise test, subjects were allowed to drink water ad libitum but no other nutritional supplementation was consumed. The treadmill was stopped briefly at 60 and 120 min to obtain blood. Heart rate and RPE were recorded every 30 min, and indirect calorimetry measurements were obtained for 10 min at the following intervals: 50–60, 110–120, 140–150, and 170–180 min.
After 180 min of running, subjects moved to a wheel chair and blood was obtained immediately. Subjects were moved to a bed and a second muscle biopsy was performed ~15 min after completing the run. Subjects then consumed a shake identical to the pre-run shake. Indirect calorimetry measurements and blood were taken at 30, 60, and 120 min post-exercise. A final muscle biopsy was taken 120 min post-exercise. The entire day of testing lasted about 8 hours finishing mid-afternoon (see Fig. 1), which is the primary reason for including the shakes before and after exercise since no other caloric intake was allowed during testing.
2.4 Blood Draws and Muscle Biopsies
An indwelling Teflon cannula or a 21G butterfly needle was inserted into an antecubital vein of the subject. The intravenous line was kept patent with normal saline solution flushes. Prior to each blood draw, 3 mL of blood was extracted and discarded to avoid inadvertent saline dilution of the blood. During each draw, blood was collected into appropriate tubes to allow for whole blood, serum, and plasma analyses. Blood was centrifuged at 1500g for 15 min and 4 °C, aliquoted into storage tubes, and stored in ultra-low freezers for batch analysis.
Muscle biopsies were obtained from the superficial portion of the vastus lateralis using the percutaneous needle technique with suction. The biopsy site was prepped and the skin was anesthetized by a local subcutaneous injection of 2% lidocaine hydrochloride. A small incision (~1 cm) was made through the skin and muscle fascia and a 5 mm diameter sterile biopsy needle (Surgical Instruments Engineering, Midlothian, United Kingdom) was introduced into the muscle to a depth of ~2 cm. To ensure adequate sample sizes a double-chop method combined with suction was used. The muscle sample was removed from the needle and divided into multiple pieces of roughly equal size. The specimens were cleaned of connective tissue and blood, flash frozen in liquid nitrogen, and stored at −80 °C for later determination of glycogen. The incision was covered with sterile gauze and compression was applied to prevent bleeding. The incision was then closed with a single suture. In order to avoid possible impairment of glycogen synthesis resulting from microtrauma in the area near the biopsy [
], we performed the immediate post-exercise biopsy on the opposite leg, and the 2 hour post-exercise biopsy 3 cm apart from the first incision site.
2.5 Biochemical Analyses
Frozen samples were thawed only once before analysis. Plasma glucose and triglycerides were measured using a Cobas C 111 analyzer (Roche Diagnostics, Indianapolis IN). Serum insulin was analyzed by ELISA (Calbiotech, Spring Valley, CA). Serum glycerol (Cayman Chemical, Ann Arbor, MI), ketones and total non-esterified fatty acids (NEFA) (Wako Diagnostics, Mountain View, CA) were analyzed by enzymatic assays. Insulin, ketones, lactate, NEFA, and glycerol had intra-assay CVs of 4.31%, 6.40%, 4.65%, 4.51% and 5.31% respectively. All samples were run in duplicate except for those measured on Cobas. Glucose and insulin values were used to calculate an index of insulin resistance [HOMA-IR; calculated as glucose (mmol/L)∙insulin (μIU/mL)/22.5] from the baseline resting sample [
]. Glycogen was determined from a portion of muscle (10 mg) that was hydrolyzed in 500 μL of 2 N HCl by heating at 99 °C for 2 hours, with occasional vortexing. Any weight lost to heating was replaced using ddH2O. The solution was neutralized with 500 μL of 2 N NaOH and 50 μL of Tris buffer (pH 6.5). 1 mL of glucose hexokinase reagent (ThermoFisher), including 20 μL of supernatant of the hydrolysis product was prepared, and the free glycosyl units were then spectrophotometrically assayed in triplicate at 340 nm on a Nanodrop 2000c spectrophotometer [
Independent t-tests were used to examine differences between LC and HC groups for dietary intake and physical characteristics. For biochemical responses to exercise, we used a two-way repeated measures analysis of variance with group (HC and LC) as a between factor and time (i.e., exercise-induced responses) as a within factor. Fisher's least significant difference post hoc was used to examine pairwise comparisons when significant main or interaction effects were observed. The alpha level for significance was set at P≤0.05.
3. Results
3.1 Subject Characteristics and Habitual Diet
There were no significant differences between groups in physical characteristics or aerobic capacity (Table 1). Two athletes in each group were ironman distance triathletes while the remainder competed primarily in running events ranging from 80 to 161 km (50 to 100 miles). The main difference between groups was their habitual diet (Table 2). Average time on an LC diet was 20 months (range 9 to 36 months). In the LC group, an overwhelming majority of energy intake was derived from fat (70%), primarily saturated and monounsaturated fatty acids. Only ~10% of their energy intake was from carbohydrate. Conversely, the HC group consumed over half their energy as carbohydrate (59%). Absolute protein was not significantly different between groups, but LC athletes consumed a greater relative amount than HC athletes (19 vs 14% of total energy). Daily nutrient intake determined from food records of the 2 days prior to testing showed similar dietary patterns as the habitual diet and was not significantly different (Supplementary Table 1).
Peak fat oxidation was on average 2.3-fold higher in the LC group (1.54 ± 0.18 vs 0.67 ± 0.14 g/min; P=0.000), with every subject in the LC group (range 1.15 to 1.74 g/min) exceeding the highest value in the HC group (range 0.40 to 0.87 g/min) (Fig. 2A ). The percentage of maximal aerobic capacity where peak fat oxidation occurred was also significantly higher in the LC group (70.3 ± 6.3 vs 54.9 ± 7.8%; P=0.000) (Fig. 2B).
Fig. 2Individual peak fat oxidation rates (A) and the exercise intensity eliciting peak oxidation (B) during a maximal graded treadmill test. Mean responses between groups were significantly different (independent t-test, P=0.000). Circles indicate mean and 95% CI. LC = low-carbohydrate diet group; HC = high-carbohydrate diet group.
All 20 subjects completed 180 min of running. The average percent maximum oxygen consumption during exercise was similar in the LC (64.7 ± 0.0%) and HC (64.3 ± 0.0%) groups. Ratings of perceived exertion over the 3 hour run were also similar between groups gradually increasing from 3.0 ± 1.3 at the start of exercise to 5.1 ± 1.9 at the end of exercise in the LC group and from 2.9 ± 0.9 to 5.2 ± 2.9 in the HC group. Absolute energy expenditure during the run was not different between the LC (12.4 ± 0.1 kcal/min) and HC (12.2 ± 0.2) groups; however, substrate oxidation patterns at rest and during exercise were significantly different (Fig. 3). At rest prior to exercise, the RER was significantly (P=0.000) lower in the LC (0.72 ± 0.05) than the HC (0.86 ± 0.08) group, indicating a contribution from fat of 95 vs 47%, respectively. During 3 hours of exercise, RER fluctuated between 0.73 and 0.74 translating into relatively stable and higher fat oxidation rates of ~1.2 g/min in the LC group, whereas fat oxidation values were significantly lower in the HC group at all time points (Fig. 3A). The rate of carbohydrate oxidation in the LC group was stable during exercise and significantly (P=0.000) lower than the HC group (Fig. 3B). The average contribution of fat during exercise in the LC and HC groups were 88% and 56%, respectively.
Fig. 3Fat (A) and carbohydrate (B) oxidation rate during 180 min of running at 64% VO2max and 120 min of recovery. All time points were significantly different between groups. LC = low-carbohydrate diet group; HC = high-carbohydrate diet group.
Circulating markers of lipid metabolism indicated a significantly greater level of ketogenesis (Fig. 4A ) and lipolysis (Fig. 4B) in the LC athletes. Serum non-esterified fatty acids were higher at the start of exercise in LC athletes, but peak levels at the end of exercise were not significantly different between groups (Fig. 4C). Plasma triglycerides were not different between groups (Fig. 4D).
Fig. 4Circulating concentrations of ketones (A), glycerol (B), non-esterified fatty acids (C), and triglycerides (D). LC = low-carbohydrate diet group; HC = high-carbohydrate diet group. BOHB = beta-hydroxybutyrate. All variables showed significant main time and interaction (group × time) effects. H and L = indicates significant (P≤0.05) difference from the corresponding baseline (BL) value for the HC and LC diet group, respectively. *Indicates significant (P=0.000) difference between HC and LC values at that time point.
Plasma glucose and serum insulin were not significantly different between groups at rest and during exercise but increased during the last hour of recovery in the HC athletes, likely due to the greater amount of carbohydrate in the shake (Fig. 5A and B ). There was no significant difference between groups in insulin resistance as determined by HOMA. Serum lactate responses were variable, but were significantly higher in LC athletes during the last hour of exercise (Fig. 5C).
Fig. 5Circulating concentrations of glucose (A), insulin (B), and lactate (C). LC = low-carbohydrate diet group; HC = high-carbohydrate diet group. All variables showed significant main time and interaction (group × time) effects. H and L = indicates significant (P≤0.05) difference from the corresponding baseline (BL) value for the HC and LC diet group, respectively. *Indicates significant (P=0.000) difference between HC and LC values at that time point.
Compared to baseline, muscle glycogen was significantly decreased by 62% immediately post-exercise and 38% at 2 hours post-exercise in the HC group. The LC group exhibited a similar pattern; muscle glycogen was decreased by 66% immediately post-exercise and 34% at 2 hours post-exercise (Fig. 6A ). There were no significant differences in pre-exercise or post-exercise glycogen concentrations between groups. There was a high degree of variability in muscle glycogen concentrations pre-exercise in both groups. In contrast, the depletion and resynthesis patterns showed a more uniform response, especially the amount of glycogen synthesized during the 2 hour recovery period in LC athletes (44.8 ± 7.5; 95% CI 40.2–49.4 μmol/g w.w.), which was one-third less variable than HC athletes (34.6 ± 23.9; 95% CI 19.8–49.4 μmol/g w.w.) (Fig. 6B). Interestingly, in all ten LC athletes the total amount of carbohydrate oxidized during the 3 hour run as calculated from indirect calorimetry (mean ± SD; 64 ± 25 g) was lower than the total amount of glycogen disappearance (mean ± SD; 168 ± 65 g), assuming 10 kg of active tissue.
Fig. 6Mean (A) and individual (B) muscle glycogen concentrations at baseline (pre-exercise), immediate post-exercise (IP), and 120 post-exercise (PE-120). Subjects ran on a treadmill at 64% VO2max for 180 min. *Indicates significant (P=0.000) difference from baseline. †Indicates significant (P=0.000) difference from IP. No significant differences between groups. LC = low-carbohydrate diet group; HC = high-carbohydrate diet group.
We studied two groups of highly trained competitive ultra-endurance athletes who were well matched in regard to training status and physical characteristics. The main difference was that the LC athletes consumed 6-times less dietary carbohydrate than LC athletes (82 vs 684 g/day) for an average of 20 months. The most notable findings were that compared to HC athletes, the LC keto-adapted runners showed: (1) two-fold higher rates of peak fat oxidation during graded exercise, (2) greater capacity to oxidize fat at higher exercise intensities, (3) two-fold higher rates of fat oxidation during sustained submaximal running, and (4) no differences in pre-exercise muscle glycogen concentrations, the rate of glycogen utilization during exercise, and the rate of glycogen synthesis during recovery. Thus, we show for the first time that chronic keto-adaptation in elite ultra-endurance athletes is associated with a robust capacity to increase fat oxidation during exercise while maintaining normal skeletal muscle glycogen concentrations.
The whole body maximal fat oxidation rates in the LC athletes were similar to those observed in elite keto-adapted cyclists [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
]. During submaximal exercise, fat contributed to 88% vs 56% of energy expended in LC and HC athletes, respectively. Several studies have consistently reported that short-term consumption of a high-fat diet in highly trained endurance athletes increases fat oxidation during submaximal exercise (reviewed in Ref. [
]). However, no studies have showed the same degree of shift toward fat oxidation as observed in LC athletes reported here. In the LC ultra-endurance runners studied here, a combination of greater degree of carbohydrate restriction and longer period of adaptation likely resulted in a more robust capacity for fat oxidation during exercise.
Serum glycerol and non-esterified fatty acid concentrations increased rapidly during exercise and decreased during recovery. Whereas fatty acids were similar between groups, glycerol concentrations were ~2-fold higher in LC athletes. Phinney et al. [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
] also reported similar circulating fatty acids concentrations before and after keto-adaptation. Serum glycerol is a better indicator of adipose tissue lipolysis compared to fatty acids since adipose tissue and skeletal muscles have lower glycerol kinase activity and thus are not able to use glycerol as effectively. Still some reports suggest that muscle can use circulating glycerol for intramuscular triglyceride synthesis [
]. Previous studies have indicated a greater capacity to transport circulating fatty acids into muscle after a short-term low-carbohydrate/high-fat diet [
Fat utilization during exercise: adaptation to a fat-rich diet increases utilization of plasma fatty acids and very low density lipoprotein-triacylglycerol in humans.
]. Thus, these data indicate a higher rate of adipose tissue lipolysis in LC athletes resulting in a greater release of both glycerol and fatty acids into the circulation, with greater overall uptake of fatty acids into skeletal muscle.
Carbohydrate oxidation was significantly lower in the LC ultra-endurance runners, but unexpectedly muscle glycogen concentrations were not different between groups. It was previously reported that a 4 week ketogenic diet in elite cyclists decreased resting muscle glycogen by half and the rate of glycogen use during exercise by 4-fold [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
Fat utilization during exercise: adaptation to a fat-rich diet increases utilization of plasma fatty acids and very low density lipoprotein-triacylglycerol in humans.
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
], suggesting that complete adaptations in glycogen homeostasis and kinetics may take several months. The different glycogen responses could also be due to lower carbohydrate intake, which was <10 g/day in cyclists [
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
] versus 86 g/day in the LC runners. A short-term glycogen loading effect is unlikely since food logs were recorded for the two days leading up to testing and indicated the average carbohydrate intake was 64 g/day in LC athletes (Supplementary Table 1). The small relative contribution of carbohydrate to energy expenditure in LC athletes, but similar use of glycogen as HC athletes, indicates a decreased reliance on circulating glucose in the keto-adapted athlete.
The muscle glycogen responses in the LC athletes share some similarities to that of highly trained Alaskan sled dogs [
]. Sled dogs have an innately high endurance capacity and often perform several hours of running at submaximal intensity while consuming a high-fat/low-carbohydrate diet. Dogs running 160 km/day for 5 days showed no cumulative muscle glycogen depletion despite eating a diet consisting of only 15% carbohydrate [
], dogs ran 140 km/day for 4 days and showed a 66% reduction in muscle glycogen after the first 140 km run (similar to the 64% reduction in muscle glycogen in the LC athletes) and a progressive increase in muscle glycogen over subsequent days of running. A more recent study reported that trained Alaskan sled dogs eating a 16% carb diet showed an unexpected high rate of carbohydrate oxidation during exercise that was associated with a significant increase in gluconeogenesis from glycerol and increased lactate oxidation [
]. Thus, highly trained high-fat adapted sled dogs show very different fuel utilization patterns than would be predicted based on studies done in trained humans eating a high-carbohydrate diet.
A provocative finding was that LC athletes appeared to break down substantially more glycogen (>100 g) than the total amount of carbohydrate oxidized during the 3 hour run. This was the case in all ten LC athletes. Why would athletes with high rates of acetyl CoA generation from fatty acids bother breaking down muscle glycogen if those carbons are not terminally oxidized? Although speculative, we believe that the reason may be to provide a source of glucose for the pentose phosphate pathway (PPP) and a source of pyruvate to form oxaloacetate. The PPP generates important 5 carbon sugars and reducing power (i.e., NADPH) for biosynthetic reactions such as ribonucleic acid synthesis and maintaining glutathione levels, respectively. Glyceraldehyde-3-phosphate and fructose 1,6 bisphosphate generated from the PPP can also enter glycolysis after the energy input stage, thereby generating ATP through substrate level phosphorylation and increasing formation of pyruvate. Pyruvate may be necessary in a keto-adapted athlete for two reasons. First, pyruvate can be used as an anaplerotic substrate by pyruvate carboxylase to generate oxaloacetate. At the onset of exercise several TCA cycle intermediates increase in concentration including fumarate, citrate and malate, however oxaloacetate remains low. Thus, in the keto-adapted athlete glycogen breakdown may be necessary to ensure a constant source of oxaloacetate for optimal TCA functioning. Second, pyruvate can be converted to lactate or alanine in muscle, which may then serve as gluconeogenic substrate for the liver.
Rates of muscle glycogen synthesis in humans are highest when large amounts of carbohydrate are consumed immediately post-exercise [
], yet the LC athletes had similar rates of glycogen repletion compared to the HC athletes, despite receiving a negligible amount of carbohydrate after exercise (4 vs 43 g) and more fat (31 vs 14 g). When no carbohydrate or energy is provided after prolonged exercise, a small amount of muscle glycogen synthesis occurs [
] presumably due to hepatic gluconeogenesis providing a source of glucose for glycogen. Horses supplemented with fat after exercise showed impaired glycogen synthesis, but 3 weeks of a high-fat diet resulted in similar glycogen repletion as horses fed a high-carbohydrate diet [
]. An obvious question that arises is what is the carbon source for glycogen synthesis in the absence of carbohydrate intake post-exercise? Although speculative, lactate and/or glycerol, which were two-fold higher at the end of exercise in LC athletes and then sharply decreased during recovery, may have provided a source of carbons for glycogen synthesis during recovery [
]. Lactate conversion to glycogen could occur directly (lactate glyconeogenesis) or indirectly via the Cori cycle. Interestingly, lactate and ketones are both transported across cell membranes by monocarboxylic acid transporters, which are upregulated after a ketogenic diet [
]. It could be that lactate rapidly replenished liver glycogen and it has an ability to maintain hepatic glucose output in the face of limited exogenous carbohydrate intake. Regardless of the mechanism, these results suggest that long-term consumption of a low-carbohydrate/high-fat diet in highly trained ultra-marathoners results in adaptations in the homeostatic regulation of muscle glycogen that acts to preserve levels similar to those observed when exogenous carbohydrate availability is high.
LC athletes had ~3-fold higher levels of circulating beta-hydroxybutyrate and total ketones at rest, during exercise, and recovery. As expected, ketone levels increased progressively during exercise and peaked 30 min into recovery consistent with the previously reported post-exercise ketosis, also referred to as the Courtice–Douglas effect [
]. Since ketones are derived from fatty acids, the keto-adapted athlete has access to and an ability to utilize a stable, abundant source of fuel in the form of beta-hydroxybutyrate. Beyond its role as a metabolite, beta-hydroxybutyrate in the range of concentrations observed in the LC athletes has recently been shown to be a potent inhibitor of histone deacetylases, which in turn upregulates expression of antioxidant genes [
]. It remains to be demonstrated whether LC athletes derive benefit in terms of cognitive function or recovery from these unique metabolic and signaling effects of beta-hydroxybutyrate.
Since this was a cross-sectional study it is possible that LC athletes had naturally higher rates of fat oxidation or were somehow better suited to respond to a low-carbohydrate diet. Long-term prospective diet intervention studies would be helpful to better understand the time course and variability in adaptations to a low-carbohydrate diet. While we showed distinct differences in metabolic responses between LC and HC athletes, we did not measure performance. It will be important for future work to address a range of performance outcomes from endurance to anaerobic capacity to strength/power, as well as sport and/or military-specific tasks. All the LC athletes were initially high-carbohydrate athletes. Thus, they made the choice to switch to a very low-carbohydrate diet and had enough self-perceived benefit to continue this lifestyle. From a practical standpoint, these observations broaden the dietary options available for endurance athletes to include a very low level of carbohydrate.
In summary, these results provide the first documentation of the metabolic adaptations associated with long-term consumption of a very low-carbohydrate/high-fat diet in highly trained keto-adapted ultra-endurance athletes. The enhanced ability to oxidize fat during exercise across a range of intensities is striking, as is the ability to maintain “normal” glycogen concentrations in the context of limited carbohydrate intake. Keto-adaptation provides an alternative to the supremacy of the high-carbohydrate paradigm for endurance athletes.
The following are the supplementary data related to this article.
Daily nutrient intake during the two days prior to testing.
Funding
This work was supported by contributions from Quest Nutrition and The Robert C. and Veronica Atkins Foundation.
Conflict of Interest
Dr. Volek and Dr. Phinney receive royalties from books on nutrition and exercise.
Acknowledgments
The authors would like to thank Peter Defty for assistance in recruiting participants and the extraordinary research participants for their enthusiasm to participate in this project.
References
Bergstrom J.
Hultman E.
Muscle glycogen synthesis after exercise: an enhancing factor localized to the muscle cells in man.
The human metabolic response to chronic ketosis without caloric restriction: preservation of submaximal exercise capability with reduced carbohydrate oxidation.
Fat utilization during exercise: adaptation to a fat-rich diet increases utilization of plasma fatty acids and very low density lipoprotein-triacylglycerol in humans.
00334-0&id=gr1.jpg){kind=link}
00334-0&id=gr2.jpg){kind=link}
00334-0&id=gr3.jpg){kind=link}
00334-0&id=gr4.jpg){kind=link}
00334-0&id=gr5.jpg){kind=link}
00334-0&id=gr6.jpg){kind=link}