Training low is a general term to describe training with low-carbohydrate availability. This low-carbohydrate availability could be low muscle glycogen, low liver glycogen, low-carbohydrate intake during or after exercise, or combinations thereof. The rationale for reducing carbohydrate availability is derived from early studies that observed links between carbohydrate availability (muscle glycogen) and gene expression [
16] because it is generally believed that training adaptations are the result of accumulated small changes in protein synthesis that result in an altered phenotype and improved performance. For this protein synthesis to occur, it is important that there is a stress signal, transcription, and translation, that messenger RNA remains stable, and that sufficient amino acids are available for protein synthesis. Many of these factors are influenced by nutrition. For example, the metabolic changes that occur as a result of muscle contraction, including a rise in AMP-activated protein kinase (AMPK), are important factors in regulating gene transcription. A single bout of endurance exercise will increase AMPK and transcription and/or messenger RNA content for various metabolic and stress-related genes. Typically, transcriptional activity peaks within the first few hours of recovery, returning to baseline within 24 h. These findings have led to the overall hypothesis that training adaptations in skeletal muscle may be generated by the cumulative effects of transient increases in gene transcription during recovery from repeated bouts of exercise [
17]. Although it is clear that gene transcription alone is not a guarantee that protein synthesis will occur, it is a necessary step for protein synthesis to occur. Studies have also demonstrated a link between muscle glycogen and AMPK expression, as lower muscle glycogen results in greater AMPK expression [
18]. It is likely that muscle glycogen directly influences AMPK because a subunit of AMPK binds to specific glycogen-binding sites, which prevents it from being phosphorylated by upstream kinases [
19]. However, when glycogen is broken down, this AMPK becomes more active [
19] and with low concentrations of glycogen, high AMPK activity is observed [
18,
20]. Other signaling molecules such as p38 mitogen-activated protein kinase [
21] and p53 [
22], as well as the expression of peroxisome proliferator-activated receptor-γ coactivator 1-alpha [
23] may be enhanced to a greater extent when exercise is performed under conditions of carbohydrate restriction. It has also been demonstrated in rats that peroxisome proliferator-activated receptor-gamma transcriptional activity is sensitive to the combined effect of skeletal muscle contraction and glycogen depletion [
24]. Glycogen thus plays an important role in regulating gene transcription in the muscle, which can alter protein synthesis and ultimately the training adaptation. Manipulating glycogen stores may therefore be a tool to optimize training adaptation. Training low has received considerable attention in the last few years. Here, I summarize the principles of the different methods, but for a more detailed discussion the reader is referred to several excellent recent review papers [
4,
6,
7,
25‐
29].
4.1.1 Training Twice a Day
The first study to use this principle was a study by Hansen et al. [
30] who used a one-legged kicking model to compare training daily, once a day, vs. training twice a day, every other day. The second exercise bout was performed with low muscle glycogen and essentially, therefore, subjects trained 50% of the time with low muscle glycogen. This produced marked improvements in the markers of oxidative capacity [activity of the mitochondrial enzymes 3-hydroxyacyl-CoA dehydrogenase (HAD) and citrate synthase (CS)] and increased glycogen levels compared with training in a glycogen-loaded state all the time [
30]. This sparked a reaction by various researchers and coaches who argued that a single-legged kicking model did not reflect a real-life situation. In addition, the study used untrained individuals, thus the real-life relevance for athletes was still unknown. Studies with the same design and performed in parallel in the UK and Australia by Hulston et al. [
31] and Yeo et al. [
20] investigated the effects in a more realistic athletic setting. In both studies, cyclists trained twice a day every other day or once every day. Both studies produced similar results. The first observation in both studies was that the cyclists who trained twice a day (train low) could not maintain the same intensity as the cyclists who trained once a day. Despite the fact that the former performed less work, some of the adaptations were greater. For example, Hulston et al. [
31] reported that HAD and fatty acid translocase (FAT/CD36) protein content was increased more when ‘training low’ and the ability to use fat as a fuel was improved [
20,
31]. Morton et al. [
32] also observed beneficial adaptations (increased succinate dehydrogenase activity) when training with low muscle glycogen. However, there were no differences in performance after 3 weeks of training low compared with the control [
20,
31], but perhaps the relatively short training period in these studies was insufficient to demonstrate changes in performance. It appears that training twice a day may result in adaptations that favor fat metabolism, but it is too early to definitely conclude that this training method will also result in long-term performance benefits.
4.1.2 Training Fasted
Perhaps the most common way to ‘train low’ is training in an overnight fasted state. Typically, the last meal is consumed between 8 and 10 P.M. the night before, and exercise is performed in the morning before breakfast is consumed. This situation is different from the previous methods, where muscle glycogen was reduced by prior exercise. When training fasted, muscle glycogen should be unaffected by the overnight fast, but liver glycogen will be very low [
33].
Studies by Hespel and coworkers [
34,
35] demonstrated that training in the fasted state may induce more profound adaptations then training in the fed state (a carbohydrate-containing breakfast and consuming carbohydrates during exercise). For example, in one study it was demonstrated that oxidative enzymes such as CS and HAD were upregulated to a greater degree (47 and 34%, respectively) when fasted was compared with fed after 6 weeks of training (4 × per week, 1–1.5 h at 75% maximal oxygen uptake) [
34]. The authors concluded that training in the fasted state was more effective to increase muscle oxidative capacity than training in the fed state. They also observed that intramuscular fat utilization was increased with fasted training and noted improvements in the regulation of blood glucose levels. The mechanisms are likely to be different from training with low glycogen. Van Proeyen et al. [
36] found no differences in AMPK in subjects training in the fasted vs. fed state, but did observe differences in post-exercise eukaryotic elongation factor 2 phosphorylation (elevated after carbohydrate feeding but not after fasting). De Bock et al. [
35] showed that exercise in the fasted state facilitated intramuscular fat use during exercise and improved glycogen resynthesis [
35]. It was also demonstrated that carbohydrate ingestion blunted uncoupling protein 3 gene expression, whereas training in the fasted state resulted in a marked increase in uncoupling protein 3 gene expression [
35]. Another study by the same research group did not result in any marked improvements with training in the fasted state [
37]. In this study, small changes were observed in proteins involved in the regulation of fat metabolism but this did not result in measurable changes in fat oxidation. The results of these studies are promising and there appear to be potential benefits of training in the fasted state. However, there are still a number of practical questions that need to be answered such as how many days of training per week are needed? What is the type of training (intensity and duration) that is most suitable for fasted training? How many weeks should this training be performed to see meaningful effects? In addition, studies to date have focused on metabolic adaptations and few have addressed the potential effects on exercise performance, such as whether fasted training results in performance improvements over time?
4.1.3 Training Adaptation with Low Exogenous Carbohydrate Availability
Although the benefits of carbohydrate ingestion during exercise are generally recognized [
38‐
41], carbohydrate supplementation during exercise may not have only positive effects. The positive effects may refer to the acute situation, but it has been suggested that chronic use of carbohydrate during exercise may limit training adaptations. This idea stems from observations that muscle glycogen stores are related to the expression of genes relevant to the adaptation to training. It is generally thought that training adaptations are the result of recurrent changes in gene expression, which occur with every bout of exercise, leading to a change in phenotype such as increases in fatty acid transport and oxidation. Long-term glucose ingestion might negatively affect the expression of relevant genes. Glucose ingestion can attenuate the rise in AMPK [
42], and long-term suppression of AMPK in turn could reduce the increase in CS activity [
43] and reduce muscle glycogen accumulation [
44], two common markers of training adaptation. Glucose ingestion will suppress lipolysis and reduce the concentration of fatty acids in the plasma, and this possibly attenuates some of the training-induced adaptations. It has been shown that glucose ingestion during exercise may suppress the expression of carnitine palmitoyl transferase mRNA, mitochondrial uncoupling protein 3, and FAT/CD36 [
45]. However, in a carefully conducted study by Akerstrom et al. [
46] in which a 10-week leg extension training program was followed by the subjects, glucose ingestion did not alter training adaptations related to substrate metabolism, mitochondrial enzyme activity, glycogen content, or performance. Significant increases were observed in CS and HAD activities after the 10-week training program but there was no effect of carbohydrate supplementation on these changes. It appears that the effects of glucose ingestion during exercise were distinctly different from those induced by exercising with low muscle glycogen. It is interesting to note that Morton et al. [
32] observed improvements in succinate dehydrogenase activity with low-glycogen training in the presence or absence of exogenous carbohydrate feeding.
4.1.4 Low-Carbohydrate High-Fat or Ketogenic Diets
Another way to train low would be to remove carbohydrate from the diet and have a long-term, low-carbohydrate, high-fat diet. It was demonstrated in the 1920s that reducing carbohydrate intake and increasing fat intake will result in higher rates of fat oxidation [
47]. However, it was also observed that subjects felt more fatigued [
47] and exercise capacity was reduced with this practice [
48]. Burke and colleagues [
49‐
52] performed a series of short-term, low-carbohydrate, high-fat diet studies and one of their observations was that 5 days on a low-carbohydrate, high-fat diet already showed some adaptations to that diet that could not be reversed completely by refilling muscle glycogen stores. Enzymes involved in fat oxidation were upregulated and fat oxidation was increased [
49]. In none of the studies, however, were any improved performance effects observed [
50‐
52]. When athletes were training over a longer period of time (7 weeks) with either a high-fat (62% fat, 21% carbohydrate) or a carbohydrate-rich (20% fat, 65% carbohydrate) diet, it was observed that both groups improved with training, but training effects were more profound in the high-carbohydrate group.
There is one study that is always referred to as evidence for the benefits of a ketogenic diet. In the 1980s, a study with five subjects showed that a ketogenic diet, containing less than 20 g of carbohydrate per day, for a prolonged period of time (4 weeks) resulted in hyperketonemia and increases in fat oxidation [
53]. In this study, exercise capacity was only tested at a low intensity and showed a large degree of variation both between subjects and within subjects. On average, there was no difference in exercise capacity before and after the ketogenic diet. As expected, fat oxidation was increased and some adaptations occurred in the muscle.
A study by Stellingwerff et al. [
54] demonstrated that although a high-fat diet will increase fat oxidation, perhaps by increasing enzyme activity related to fat metabolism, it can reduce enzyme activities related to carbohydrate metabolism. Thus, whilst many studies observed improvements in HAD, for example, Stellingwerff et al. [
54] demonstrated compromised pyruvate dehydrogenase activity. It may therefore be that fat oxidation is increased, at least partly, as a result of an inability to use carbohydrates. Because carbohydrates are important substrates for high-intensity exercise, such adaptations would be unwanted. In fact, a carefully controlled study by Burke et al. [
55] demonstrated that there were no benefits of a ketogenic vs. a high-carbohydrate diet, or a mixed approach (higher or lower carbohydrates depending on training) in elite endurance athletes. In fact, performance of high-intensity exercise was not improved by 3 weeks of intensified training in the ketogenic diet group (−1.6%), while athletes consuming the other diets made substantial performance improvements (6.6% in the high-carbohydrate group and 5.5% in the mixed group).
The ketogenic diet has received considerable attention in the popular press and many claims have been made recently. However, it is important to realize that, to date, not a single study has demonstrated performance benefits of a ketogenic diet, including the early study that is often referred to [
53]. Thus, at present, there are no data on ketogenic diets in athletes on which to base performance claims.
4.1.5 Carbohydrate Restriction During Recovery
Another concept is restricting carbohydrate intake in the first hours after exercise. The time course of transcriptional activation for many exercise-induced genes stretches across the first few hours of recovery and usually returns to baseline within 24 h [
56]. Traditionally, it was recommended to consume carbohydrate immediately after exercise as this results in the highest rates of glycogen synthesis [
56]. When studying the effects on gene expression post-exercise with or without carbohydrates, interesting observations were made by Pilegaard et al. [
57]: activation of metabolic genes was augmented following 75 min of cycling exercise when carbohydrate intake was restricted for over 5 h compared with controls [
57]. Cochran et al. [
21] also showed that carbohydrate ingestion post-exercise, and not necessarily changes in muscle glycogen content per se, altered the metabolic response to repeated sessions of high-intensity interval exercise. Specifically, it was observed that p38 MAPK was activated more when carbohydrate intake was restricted and this has been linked to enhanced expression of peroxisome proliferator-activated receptor-γ coactivator 1 and an improved metabolic adaptation to exercise. Other studies, however, could not confirm this and found no differences between a high- and a low-carbohydrate intake during recovery [
58‐
60]. A recent study [
60] investigated the effects of post-exercise carbohydrate intake vs. carbohydrate restriction on glycogen and gene expression. The carbohydrate intake resulted in partial glycogen replenishment but gene expression was not different between the two groups. After 24 h, glycogen replenishment was similar in the two groups (a finding that seems consistent in well-trained individuals) and gene expression levels returned to baseline levels in both groups. It is not impossible that changes in gene expression levels went unnoted because the timing of the relatively small number of sampling points may not have been perfect. Thus, the effect of carbohydrate manipulation in the recovery phase is still uncertain.
4.1.6 Sleep Low
The concept of carbohydrate restriction during recovery was extended by Lane et al. [
61]. They performed the first study into the concept of ‘train high-sleep low’, which refers to a hard workout in the evening resulting in lowered carbohydrate availability (muscle and liver glycogen) followed by sleep. This practice goes against the typical advice to athletes to consume carbohydrates post-exercise (and before going to sleep) to speed up recovery. Training high and sleeping low, however, resulted in a greater upregulation of several exercise responsive signaling markers with roles in lipid oxidation the following morning compared with when an evening meal was consumed [
61]. In this study, ‘train high-sleep low’ did not elicit a greater upregulation of cellular markers of mitochondrial biogenesis. The study only addressed the acute changes and did not intend to study the long-term effects on metabolism or performance.
A follow-up study performed in France studied the longer term effects of the ‘train high-sleep low’ approach. Two groups of triathletes undertook the same endurance training program for 3 consecutive weeks [
62]. One group (
n = 11) followed a ‘sleep low’ strategy for manipulating carbohydrate availability (high-intensity workout with high-carbohydrate availability followed by a carbohydrate-restricted recovery plus an overnight fast; then a prolonged submaximal workout the following morning commenced with low-carbohydrate availability) in the training schedule. The control group (
n = 10) maintained regular carbohydrate intake throughout the day and undertook each training session with normal/high carbohydrate availability. The triathletes performed a simulated triathlon race at the start and end of the intervention period. The authors found a small but significant effect on performance as 10-km running performance increased after 3 weeks in the ‘sleep-low’ group but not in the control group. Interestingly, the authors also reported improvements in supramaximal cycling time to exhaustion with sleep low but not control.
These are the only two studies on the concept of ‘train high-sleep low’ and it may be too early to draw firm conclusions. However, the studies do provide promising results. From a practical point of view, it is important to be aware of other potential side effects and unknowns of this approach: what are the effects on recovery when applied frequently, the effects on immune function, and perhaps most importantly, the effects on quantity and quality of sleep?