Obesity, energy balance and fat balance

By definition, the development of obesity and overweight is characterized by a positive energy balance. Numerous investigations (Schutz et al. Reference Schutz, Flatt and Jequier1989; Bennett et al. Reference Bennett, Reed, Peters, Abumrad, Sun and Hill1992) have shown that in the long term an imbalance between energy intake and energy expenditure is reflected in a positive fat balance. In the last 20–30 years many food products have become available that are cheap, palatable and high in fat content. Since dietary fat is the most energy-dense macronutrient, about 38 kJ/g (in comparison, carbohydrate and protein only provide about 17 kJ/g), an increase in dietary fat intake profoundly increases dietary energy intake and thus results in a positive energy balance. In addition, in human subjects there is evidence for a clear substrate hierarchy for the utilization of macronutrients, in which fat balance is least regulated. For example, the human body responds only very slowly by increasing fat oxidation when fat intake is increased (Thomas et al. Reference Thomas, Peters, Reed, Abumrad, Sun and Hill1992; Schrauwen et al. Reference Schrauwen, van Marken Lichtenbelt, Saris and Westerterp1997a), leading to a deposition of dietary fat in the fat stores. On the other hand, the storage capacity for carbohydrate and protein in the human body is limited and therefore carbohydrate and protein oxidation are very well and rapidly adjusted to their respective intake (Abbott et al. Reference Abbott, Howard, Christin, Freymond, Lillioja, Boyce, Anderson, Bogardus and Ravussin1988). As a consequence, a positive energy balance will be reflected in a positive fat balance.

Fat oxidation on a high-fat diet

Although there is ample evidence that the adaptation of fat oxidation to increased fat intake is slow in man, the reason for this slow adaptation is relatively unknown. According to the two-compartment model of Flatt (Reference Flatt1987), whole-body fat oxidation can be increased via an expansion of fat mass, leading to increased plasma NEFA levels available for oxidation. In this model the body is divided into two compartments, fat mass and glycogen stores, and the oxidation mixture of fatty acids and glucose depends on the size of these two compartments. As the glycogen stores are very limited in size, small changes in the size of the glycogen stores will affect glucose oxidation. In contrast, as the fat mass can be relatively unlimited in size, a large expansion of fat mass is needed before changes in fat oxidation will occur. This model can explain why the addition of a surplus of fat to a single meal, which will not result in a change in fat mass, does not affect fatty acid oxidation rates, and why obese subjects have relatively high fat oxidation. In fact, expansion of fat mass (obesity) could be considered as an adaptation of the body to increase fat oxidation to a level that matches a high dietary fat intake. However, it has been shown (Schrauwen et al. Reference Schrauwen, van Marken Lichtenbelt, Saris and Westerterp1997a) that healthy human volunteers who consume a high-fat diet for 7 d, while being in energy balance, are able to slowly increase their fat oxidation to a level that equals the high-fat intake. As no substantial expansion of fat mass can be expected after 7 d of a high-fat diet, these results seem to contradict Flatt's (Reference Flatt1987) model. These findings have, however, been explained in terms of the changes in glycogen stores that may have occurred (Schrauwen et al. Reference Schrauwen, van Marken Lichtenbelt, Saris and Westerterp1997b, Reference Schrauwen, Lichtenbelt, Saris and Westerterp1998). During the first days on a high-fat diet, when fat oxidation does not equal fat intake, subjects are in negative carbohydrate balance (carbohydrate oxidation>carbohydrate intake), leading to a decrease in the body's glycogen stores. According to Flatt's (Reference Flatt1987) two-compartment model, a decrease in glycogen stores would result in a decrease in glucose oxidation and would therefore be another way to increase fat oxidation. Indeed, it has been shown (Schrauwen et al. Reference Schrauwen, van Marken Lichtenbelt, Saris and Westerterp1997b, Reference Schrauwen, Lichtenbelt, Saris and Westerterp1998) that lowering glycogen stores by exhaustive exercise markedly improves the rate at which subjects are able to adapt their fat oxidation to an increased fat intake.

In order to further investigate the mechanisms by which fat oxidation increases on a high-fat diet, a more detailed determination of fatty acid oxidation has been conducted in subjects consuming high-fat diets (Schrauwen et al. Reference Schrauwen, Wagenmakers, van Marken Lichtenbelt, Saris and Westerterp2000). Interestingly, it was observed that the increase in fat oxidation after 7 d of a high-fat diet is completely accounted for by an increase in TAG-derived fatty acid oxidation. Thus, using labelled fatty acids plasma-derived fatty acid oxidation can be distinguished from the oxidation of TAG-derived fatty acids (mainly intramyocellular lipids (IMCL)). IMCL are small lipid droplets that are located in the sarcoplasm and predominantly found in the vicinity of mitochondria, suggesting that they may serve as a rapidly-available energy source for the muscle. Using proton magnetic resonance spectroscopy to quantify these IMCL non-invasively, it has recently been shown (Schrauwen-Hinderling et al. Reference Schrauwen-Hinderling, Kooi, Hesselink, Moonen-Kornips, Schaart, Mustard, Hardie, Saris, Nicolay and Schrauwen2005) that the amount of IMCL is already markedly increased after 7 d of a high-fat diet in healthy lean subjects. In analogy with the model of Flatt (Reference Flatt1987), which predicts that an increase in body fat mass is needed to increase whole-body fat oxidation, these combined observations of an increased IMCL mass and increased IMCL oxidation indicate that increases in IMCL content are also needed to drive increased IMCL oxidation. Thus, the slow rate at which fat oxidation adapts to increased fat intake when a high-fat diet is consumed may also be attributed to the time needed to increase IMCL content. Interestingly, it has been found (Schrauwen et al. Reference Schrauwen, Van Aggel-Leijssen, Hul, Wagenmakers, Vidal, Saris and Van Baak2002c) that when sedentary middle-aged subjects follow an endurance training programme for 3 months whole-body fat oxidation increases, and again this increase is completely accounted for by an increase in TAG-derived fatty acid oxidation. In accordance with the earlier mentioned hypothesis, endurance training is also known to increase IMCL content (Goodpaster et al. Reference Goodpaster, He, Watkins and Kelley2001; Schrauwen-Hinderling et al. Reference Schrauwen-Hinderling, Hesselink, Moonen-Kornips, Schaart, Kooi, Saris and Schrauwen2006a), suggesting that similar mechanisms may be involved in the training- and diet-induced increase in fat oxidation.

Intramyocellular lipid and insulin sensitivity

As mentioned earlier, the steadily-increasing prevalence of obesity is accompanied by a steady increase in the prevalence of type 2 diabetes mellitus. Central to the aetiology of type 2 diabetes is insulin resistance, an early detectable characteristic preceding the development of full-blown and clinically-overt type 2 diabetes. Insulin resistance indicates a state of reduced responsiveness of the insulin-sensitive tissues to circulatory levels of insulin, i.e. a reduced glucose disposal and impaired inhibition of hepatic glucose output at a given concentration of insulin (McGarry, Reference McGarry2002). A major contribution to whole-body insulin resistance comes from the skeletal muscle, as revealed by classical work by DeFronzo et al. (Reference DeFronzo, Jacot, Jequier, Maeder, Wahren and Felber1981), who have shown a 45% reduction in leg glucose uptake following a hyperinsulinaemic euglycaemic clamp in patients with type 2 diabetes and a strong positive correlation (r 0·70, P<0·001) between leg and total body glucose uptake.

Evidence gathered in recent decades has pointed towards an important causal role of disturbed fatty acid metabolism in the development of type 2 diabetes mellitus. Not only are plasma glucose levels increased in uncontrolled type 2 diabetes, but also plasma NEFA, and the storage of fatty acids in non-adipose tissues such as pancreas, liver and muscle is elevated in patients with type 2 diabetes (Schalch & Kipnis, Reference Schalch and Kipnis1965). Moreover, a strong negative correlation has been found between the level of IMCL and insulin sensitivity in non-trained subjects (Perseghin et al. Reference Perseghin, Scifo, De Cobelli, Pagliato, Battezzati, Arcelloni, Vanzulli, Testolin, Pozza, Del Maschio and Luzi1999), and levels of IMCL are increased in first-degree relatives of patients with type 2 diabetes who are insulin resistant, but not diabetic (Jacob et al. Reference Jacob, Machann, Rett, Brechtel, Volk, Renn, Maerker, Matthaei, Schick, Claussen and Haring1999). These data suggest that IMCL accumulation may be a primary factor in the development of type 2 diabetes. Remarkably, this suggested role differs from the putative physiological function of IMCL as an important energy source that drives muscular fatty acid oxidation. On the other hand, the high levels of IMCL in the endurance-trained state contradict the suggestion that IMCL may have detrimental effects on insulin sensitivity, as endurance training is well known to improve whole-body and muscular insulin sensitivity. To unravel this apparent paradox the molecular adaptations that occur in skeletal muscle have been examined in subjects who were either trained for 2 weeks or consumed a high-fat diet for 7 d. As expected, both interventions increased IMCL levels to a similar extent (Schrauwen-Hinderling et al. Reference Schrauwen-Hinderling, Hesselink, Moonen-Kornips, Schaart, Kooi, Saris and Schrauwen2006a, Reference Schrauwen-Hinderling, Hesselink, Schrauwen and Kooib). Furthermore, consistent with the hypothesis that elevated IMCL levels may serve as an important energy source, it has been found (Schrauwen-Hinderling et al. Reference Schrauwen-Hinderling, Hesselink, Moonen-Kornips, Schaart, Kooi, Saris and Schrauwen2006a) that after 2 weeks of endurance training in previously untrained subjects there are already molecular adaptations in skeletal muscle that point towards an improved fat oxidative capacity. In contrast, however, the consumption of a high-fat diet for 1 week is accompanied by molecular changes that would favour fat storage in muscle rather than oxidation (Schrauwen-Hinderling et al. Reference Schrauwen-Hinderling, Kooi, Hesselink, Moonen-Kornips, Schaart, Mustard, Hardie, Saris, Nicolay and Schrauwen2005). These findings suggest that IMCL may be increased for two different reasons: (1) a functional increase, whereby IMCL serves as a rapidly-available energy source; in these conditions (e.g. as with endurance training) fat oxidative capacity improves, followed by the storage of more muscular lipids that can serve as rapidly-available energy source; (2) a pathophysiological increase, whereby the increase in IMCL is merely the result of a continuous over-supply of fat to the muscle and is not accompanied by an increased fat oxidative capacity; in this condition muscular fat accumulation results from passive uptake of excessive fatty acids, and the increased IMCL content may ultimately drive IMCL oxidation. Thus, in the latter condition the increase in IMCL can be seen as an ‘adaptation’ to allow increased muscular fat oxidation when fat intake is high, similar to the concept of obesity being an adaptation to increase whole-body fat oxidation. However, as with obesity, the ‘adaptive’ increase in IMCL in the latter condition is accompanied by negative side effects, which is indicated by the relationship between IMCL and insulin resistance.

Thus, it can be deduced that high IMCL content is only harmful in the absence of adaptations in oxidative capacity, such as with high-fat diet consumption and/or type 2 diabetes mellitus. To examine this hypothesis a comparison has been made of the IMCL content and mitochondrial oxidative capacity in patients with type 2 diabetes and in normoglycaemic controls matched for BMI (Schrauwen-Hinderling et al. Reference Schrauwen-Hinderling, Kooi, Hesselink, Jeneson, Backes, van Echteld, van Engelshoven, Mensink and Schrauwen2007). Thus, a ³¹P magnetic resonance spectroscopy method was used to investigate in vivo mitochondrial function by measuring the phosphocreatine kinetics during recovery from exercise (Meyer, Reference Meyer1988; Kemp & Radda, Reference Kemp and Radda1994). During exercise phosphocreatine content decreases transiently but recovers rapidly after exercise. In the post-exercise state phosphocreatine resynthesis is driven almost purely oxidatively (Sahlin et al. Reference Sahlin, Harris and Hultman1979) and the resynthesis rate reflects in vivo mitochondrial function (Kemp & Radda, Reference Kemp and Radda1994). It was found that despite similar IMCL levels, phosphocreatine resynthesis rate is reduced by approximately 50% in patients with type 2 diabetes, indicating mitochondrial dysfunction. Thus, the combination of high IMCL and low oxidative capacity is important in the aetiology of type 2 diabetes mellitus. Along the same line of reasoning Goodpaster et al. (Reference Goodpaster, He, Watkins and Kelley2001) have suggested that fat oxidative capacity might be more important than IMCL content in determining insulin sensitivity, whereas Petersen et al. (Reference Petersen, Dufour, Befroy, Garcia and Shulman2004) have reported decreased mitochondrial function (oxidative capacity) in the insulin-resistant offspring of patients with type 2 diabetes when compared with insulin-sensitive controls. Together, these findings indeed point towards the importance of mitochondrial oxidative capacity in the aetiology of type 2 diabetes. Although beyond the scope of the present review, a functional explanation is provided by the observation that lipid metabolites such as fatty acyl-CoA or diacylglycerol are more likely to intervene with insulin signalling than IMCL per se. Diacylglycerol is the intermediate between fatty acyl-CoA, which can enter the mitochondria for oxidation, and TAG, the form in which lipids are stored in muscle. If an increase in IMCL is not accompanied by a sufficiently increased fat oxidative capacity (i.e. a low mitochondrial oxidative capacity), levels of lipid metabolites will rise and interfere with insulin signalling (for review, see Shulman, Reference Shulman2000).

What causes mitochondrial dysfunction?

Although the finding of a reduced mitochondrial oxidative capacity in both first-degree relatives and patients with type 2 diabetes is intriguing, the mechanism underlying this reduction remains to be determined. One likely explanation for the mitochondrial dysfunction observed in the (pre)diabetic state is a reduced expression of genes involved in mitochondrial oxidative metabolism, as has been observed in type 2 diabetes (Mootha et al. Reference Mootha, Lindgren, Eriksson, Subramanian, Sihag and Lehar2003; Patti et al. Reference Patti, Butte, Crunkhorn, Cusi, Berria and Kashyap2003). Many of the oxidative genes are under transcriptional control of PPARγ coactivator-1α (PGC-1α), which stimulates oxidative phosphorylation, mitochondrial biogenesis and the generation of oxidative type 1 muscle fibres (Lin et al. Reference Lin, Wu, Tarr, Zhang, Wu and Boss2002). In addition to the down-regulation of the so-called OXPHOS gene set, it has been shown that the expression of PGC-1α itself is also reduced in patients with type 2 diabetes (Mootha et al. Reference Mootha, Lindgren, Eriksson, Subramanian, Sihag and Lehar2003; Patti et al. Reference Patti, Butte, Crunkhorn, Cusi, Berria and Kashyap2003). Strikingly, the same observations have been reported in subjects who do not have type 2 diabetes, but have a family history of type 2 diabetes (Patti et al. Reference Patti, Butte, Crunkhorn, Cusi, Berria and Kashyap2003). Recently, these findings have been confirmed (M Mensink, MKC Hesselink, AP Russell, G Schaart, J-P Sels and P Schrauwen, unpublished results) in patients with type 2 diabetes, showing an approximately 60% reduction in PGC-1α gene expression compared with BMI and age-matched controls. Interestingly, it has been found (M Mensink, MKC Hesselink, AP Russell, G Schaart, J-P Sels and P Schrauwen, unpublished results) that treating these patients with the anti-diabetic agent Rosiglitazone restores muscular PGC-1α levels, along with an improvement in muscular insulin sensitivity. In addition, it has been shown (Russell et al. Reference Russell, Hesselink, Lo and Schrauwen2005) that acute exercise, which is well known to improve muscular oxidative capacity and insulin sensitivity, up regulates the expression of PGC-1α in healthy subjects. Also, prolonged endurance training increases PGC-1α protein content (Russell et al. Reference Russell, Feilchenfeldt, Schreiber, Praz, Crettenand, Gobelet, Meier, Bell, Kralli, Giacobino and Deriaz2003a). These data indicate that PGC-1α may indeed play a crucial role in the development of mitochondrial dysfunction, insulin resistance and type 2 diabetes mellitus.

If PGC-1α is indeed a crucial factor in relation to mitochondrial function, the question arises of why PGC-1α is reduced in the prediabetic state?

Although it is tempting to speculate that genetic factors underlie the reduced expression of PGC-1α in the prediabetic state, the impact of environmental factors should not be ruled out. As mentioned earlier, acute exercise (Russell et al. Reference Russell, Hesselink, Lo and Schrauwen2005), as well as endurance training (Russell et al. Reference Russell, Feilchenfeldt, Schreiber, Praz, Crettenand, Gobelet, Meier, Bell, Kralli, Giacobino and Deriaz2003a), can increase the expression and protein content of PGC-1α. Whether the opposite effect, a reduction in PGC-1α as a result of physical inactivity, also occurs is presently unknown, but it is tempting to suggest that the low level of physical activity that characterizes Westernized society is partly responsible for the reduction in PGC-1α levels. In that context, the study of Mootha et al. (Reference Mootha, Lindgren, Eriksson, Subramanian, Sihag and Lehar2003) has observed that the reduction in OXPHOS genes is strongly correlated with whole-body aerobic capacity. Although this result can be interpreted as PGC-1α being a determinant of oxidative capacity, it is also known that whole-body aerobic capacity is mainly determined by an individual's physical activity history (Katzel et al. Reference Katzel, Sorkin and Fleg2001; Nagaya et al. Reference Nagaya, Kondo and Shibata2001). Thus, the strong correlation between OXPHOS genes and aerobic capacity may also support the suggestion that physical (in)activity underlies the reduction in PGC-1α and oxidative gene expression.

Very interestingly, consumption of a high-fat diet for 3 d reduces PGC-1α protein levels by approximately 20% in human volunteers, whereas there is a 40% reduction in c57Bl/6J mice after 3 weeks on a high-fat diet (Sparks et al. Reference Sparks, Xie, Koza, Mynatt, Hulver, Bray and Smith2005). Again, this finding suggests that the reduction of PGC-1α in the prediabetic state may very well be associated with Westernized lifestyle factors. To investigate whether a high fatty acid supply is indeed able to down regulate PGC-1α a study has been conducted to establish whether the acute elevation of plasma NEFA during a hyperinsulinaemic euglycaemic clamp is associated with a down-regulation in the expression of skeletal muscle genes involved in oxidative metabolism (Hoeks et al. Reference Hoeks, Hesselink, Russell, Mensink, Saris, Mensink and Schrauwen2006a). Thus, when healthy volunteers receive an infusion of Intralipid (a fat emulsion; Fresenius Kabi, Bad Homburg, Germany) over a period of 6 h there is an approximately 2·7-fold increase in plasma NEFA levels, a 31% decrease in insulin sensitivity and an approximately 1·6-fold increase in muscular fat content. Under these conditions of high circulating and muscular fat content, PGC-1α is acutely down regulated. Again, this finding suggests that the decrease in PGC-1α expression observed in the (pre)diabetic state may be the result of diabetogenic environmental factors, and PGC-1α may therefore provide a link between an unhealthy lifestyle and the development of mitochondrial dysfunction and ultimately type 2 diabetes mellitus.

A role for oxidative stress in the development of insulin resistance

Since the finding that mitochondrial function may be impaired in the (pre)diabetic state, most studies have focused on PGC-1α and mitochondrial biogenesis. However, mitochondrial function is not only determined by mitochondrial biogenesis, but also by mitochondrial quality. In the latter context, it has been shown (Kelley et al. Reference Kelley, He, Menshikova and Ritov2002) that mitochondria from patients with type 2 diabetes are smaller and show morphological abberations when compared with controls, and mitochondrial area correlates positively with insulin sensitivity. The smaller and damaged mitochondria in skeletal muscle of patients with diabetes also result in an impaired functional capacity (Kelley et al. Reference Kelley, He, Menshikova and Ritov2002). Thus, mitochondria of patients with type 2 diabetes have a reduced electron transport chain capacity, as measured by NADH:O₂ oxidoreductase activity, as well as a reduced citrate synthase activity (Kelley et al. Reference Kelley, He, Menshikova and Ritov2002).

To explain the observed mitochondrial damage, elevated production of reactive oxygen species (ROS) and its by-products (e.g. lipid peroxides) has been suggested. In addition to the production of ATP, mitochondria are also the major contributor to the production of ROS. Mitochondrial ROS can react rapidly with DNA, protein and lipids, thereby leading to so-called oxidative damage. Recent evidence points towards a causal role for ROS in the development of insulin resistance. Thus, Houstis et al. (Reference Houstis, Rosen and Lander2006) have found that cellular ROS levels are elevated in two cell models of insulin resistance in which insulin resistance is either induced by treatment with the cytokine TNFα or by the glucocorticoid dexamethasone. To provide evidence for a causal relationship between the elevated ROS levels and insulin resistance, the authors over-expressed a variety of antioxidant genes in their model, and it was found that the over-expression of each of these genes could reverse the insulin resistance induced by both TNFα and dexamethasone (Houstis et al. Reference Houstis, Rosen and Lander2006).

Fatty acids are especially very prone to ROS-induced oxidative damage, resulting in the formation of lipid peroxides, which in turn can induce damage to proteins and DNA. Thus, accumulation of fatty acids in the vicinity of the mitochondrial matrix, where ROS are formed, increases the likelihood of lipid peroxidation. As discussed earlier, patients with type 2 diabetes are characterized by the accumulation of IMCL and these lipid droplets are located close to the mitochondria. To prevent simple diffusion of fatty acids into the mitochondria their entry is regulated by the enzyme carnitine palmitoyltransferase 1, which serves as an inward transporter of oxidizable fatty acids. However, this system cannot completely prevent the diffusion of fatty acid into the mitochondria. Mitochondrial membranes consist of lipid bilayers and therefore fatty acids can still enter the mitochondria matrix via a so-called ‘flip–flop’ mechanism over the membrane (Ho et al. Reference Ho, Duclos and Hamilton2002). It can easily be imagined that this ‘passive diffusion’ is more likely to occur under conditions of a high IMCL concentration, as is the case with insulin resistance and type 2 diabetes mellitus. Consistent with this notion, skeletal muscle of subjects who are obese and insulin resistant not only contains a higher amount of IMCL, but these lipids also show a higher extent of lipid peroxidation (Russell et al. Reference Russell, Gastaldi, Bobbioni-Harsch, Arboit, Gobelet, Deriaz, Golay, Witztum and Giacobino2003b). Potentially, these lipid peroxides could lead to oxidative damage to mitochondrial structures and explain the increased mitochondrial damage observed in patients with type 2 diabetes (Kelley et al. Reference Kelley, He, Menshikova and Ritov2002).

Mitochondrial uncoupling protein-3 to prevent lipid-induced oxidative damage

If mitochondria are vulnerable to their own ROS and the resulting lipid peroxides, these organelles should be equipped with a protective mechanism. Recently, it has been proposed (Schrauwen et al. Reference Schrauwen, Hoeks and Hesselink2006) that the mitochondrial uncoupling protein (UCP) 3 fulfils such a protective role against lipid-induced oxidative damage. UCP3 was discovered in 1997 as a human homologue of the brown adipose tissue-specific UCP1. However, after several years of research it became evident that in contrast to UCP1, UCP3 has no major role in the regulation of energy metabolism, but is rather involved in fatty acid metabolism. In several physiological experiments it has been observed that UCP3 is up regulated under conditions in which the supply of fatty acids to the muscle exceeds the fat oxidative capacity, such as fasting, high-fat feeding (Hesselink et al. Reference Hesselink, Greenhaff, Constantin-Teodosiu, Hultman, Saris, Nieuwlaat, Schaart, Kornips and Schrauwen2003) and acute exercise (Schrauwen et al. Reference Schrauwen, Hesselink, Vaartjes, Kornips, Saris, Giacobino and Russell2002a), and in type 2b muscle fibres (Hesselink et al. Reference Hesselink, Keizer, Borghouts, Schaart, Kornips, Slieker, Sloop, Saris and Schrauwen2001). On the other hand, UCP3 is down regulated when the fat oxidative capacity is high, as in type 1 muscle fibres (Hesselink et al. Reference Hesselink, Keizer, Borghouts, Schaart, Kornips, Slieker, Sloop, Saris and Schrauwen2001) and with endurance training (Schrauwen et al. Reference Schrauwen, Russell, Moonen-Kornips, Boon and Hesselink2005). When fatty acid supply does not match fatty acid oxidation, fatty acids will accumulate in the cytosol and the possibility of fatty acids entering the mitochondrial matrix via ‘flip–flop’ will increase. Based on these findings, it has been postulated (Schrauwen et al. Reference Schrauwen, Saris and Hesselink2001) that UCP3 could function as a fatty acid anion exporter. Fatty acids that ‘flip–flop’ over the inner mitochondrial membrane will become deprotonated as a result of a proton gradient over the inner mitochondrial membrane, resulting in fatty acid anions. Importantly, fatty acid anions are not able to ‘flip-flop’ back across the membrane (Ho et al. Reference Ho, Duclos and Hamilton2002), nor can they undergo β-oxidation because the enzyme that converts long-chain fatty acids to fatty acyl-CoA is absent in the mitochondrial matrix (Tol, Reference Tol1975). Thus, the inner mitochondrial membrane would accumulate fatty acid anions on the side of the mitochondrial matrix. Moreover, these fatty acid anions could potentially leave the inner mitochondrial membrane and accumulate inside the mitochondrial matrix; these anions could accumulate either dissolved in the matrix at very low concentrations or bound to matrix-associated fatty acid-binding proteins (Unterberg et al. Reference Unterberg, Borchers, Hojrup, Roepstorff, Knudsen and Spener1990). In the matrix (or at the inner side of the inner mitochondrial membrane), these fatty acid anions would be prone to peroxidization by ROS. As peroxidized fatty acids are highly reactive (Yagi, Reference Yagi1987), they could damage DNA, RNA and proteins of the mitochondrial machinery, which are mainly present in the mitochondrial matrix. It has been postulated (Schrauwen et al. Reference Schrauwen, Saris and Hesselink2001) that UCP3 may serve as an exporter of non-metabolizable fatty acid anions, protecting mitochondria from oxidative damage. This hypothesis is consistent with those postulated by other researchers, which also predict that UCP3 is a fatty acid anion exporter (Skulachev, Reference Skulachev1991; Garlid et al. Reference Garlid, Jaburek and Jezek1998; Jezek et al. Reference Jezek, Engstová, Zácková, Vercesi, Costa, Arruda and Garlid1998).

Subsequent to postulating this hypothesis several experiments were carried out to test its validity. First, use was made of the potential of Etomoxir to inhibit carnitine palmitoyltransferase 1, the rate-limiting step in the mitochondrial uptake of fatty acyl-CoA. Healthy human subjects were given Etomoxir for 36 h, while consuming a high-fat diet to increase fatty acid supply (Schrauwen et al. Reference Schrauwen, Hinderling, Hesselink, Schaart, Kornips, Saris, Westerterp-Plantega and Langhans2002b). According to the hypothesis, this condition would be one of the most formidable in which UCP3 is needed, as both the oxidative capacity is low and the supply of fat is high. As anticipated, there is a 67% induction of UCP3 at the protein level after only 36 h of Etomoxir treatment. These findings have been confirmed in rats (Schrauwen et al. Reference Schrauwen, Hoeks, Schaart, Kornips, Binas and Van De Vusse2003), where it was found that Etomoxir treatment for 8 d results in an approximately 2-fold decrease in carnitine palmitoyltransferase 1 activity and an approximately 2-fold increase in UCP3 protein levels.

A second study (Hoeks et al. Reference Hoeks, Hesselink, Van Bilsen, Schaart, van der Vusse, Saris and Schrauwen2003) has utilized the unique ability of medium-chain fatty acids to pass the mitochondrial bilayer independently of carnitine palmitoyltransferase 1. An important aspect of the hypothesis is that neutral fatty acids that enter the mitochondrial matrix cannot be oxidized because they cannot be converted to fatty acyl-CoA inside the matrix (Kerner & Hoppel, Reference Kerner and Hoppel2000). However, this premise is not true for medium-chain fatty acids; these fatty acids can be activated inside the mitochondrial matrix and do not rely on carnitine palmitoyltransferase 1 for mitochondrial uptake (Tol, Reference Tol1975; Rasmussen et al. Reference Rasmussen, Holmback, Volpi, Morio-Liondore, Paddon-Jones and Wolfe2002). Thus, even when neutral medium-chain fatty acids reach the mitochondrial matrix they can be activated to fatty acyl-CoA and directed to β-oxidation (Fujino et al. Reference Fujino, Takei, Sone, Ioka, Kamataki, Magoori, Takahashi, Sakai and Yamamoto2001). Obviously, these fatty acids do not need to (and should not) be exported from the matrix, eliminating a role for UCP3 in the handling of these specific medium-chain fatty acids. Thus, there is no need for UCP3 to control increased levels of medium-chain fatty acids inside the matrix, in contrast to the anticipated increase in UCP3 with a high-fat diet containing predominantly long-chain fatty acids. To test this notion, rats were fed high-fat diets containing either long-chain fatty acids or medium-chain fatty acids for 14 d (Hoeks et al. Reference Hoeks, Hesselink, Van Bilsen, Schaart, van der Vusse, Saris and Schrauwen2003). UCP3 protein content was found to increase 2-fold on feeding the long-chain fatty acid high-fat diet, but to be unaffected by feeding the medium-chain fatty acid high-fat diet. Furthermore, this differential UCP3 response is accompanied by a similar rise in plasma NEFA levels and also cannot be explained by a PPAR effect, as other PPAR-responsive genes are similarly affected by the diets (Hoeks et al. Reference Hoeks, Hesselink, Van Bilsen, Schaart, van der Vusse, Saris and Schrauwen2003). Again, these data support the hypothesis that UCP3 is involved in the export from the mitochondrial matrix of fatty acids that cannot be metabolized.

According to the hypothesis UCP3 prevents the accumulation of fatty acid anions inside the mitochondrial matrix and membranes, which otherwise would be vulnerable to ROS-induced oxidative damage, resulting in lipid peroxides. It can therefore be predicted that lack of UCP3 would lead to increased lipid peroxide levels and/or oxidative damage. Indeed, mice lacking UCP3 are characterized by increased levels of lipid peroxides (Hoeks et al. Reference Hoeks, Hesselink, Sluiter, Schaart, Willems, Morrisson, Clapham, Saris and Schrauwen2006b) and damage to mitochondrial proteins (Brand et al. Reference Brand, Pamplona, Portero-Otin, Requena, Roebuck, Buckingham, Clapham and Cadenas2002). Very interestingly, it has recently been shown (Echtay et al. Reference Echtay, Esteves, Pakay, Jekabsons, Lambert, Portero-Otin, Pamplona, Vidal-Puig, Wang, Roebuck and Brand2003) that UCP3 can be activated by 4-hydroxy-2-nonenal, an important byproduct of lipid peroxidation, suggesting that a negative feedback loop is involved in the regulation of mitochondrial lipid peroxide production, whereby elevated lipid peroxide levels activate UCP3, thereby limiting further accumulation of these toxic lipid products.

Uncoupling protein 3 and type 2 diabetes mellitus

The increased content of muscular fat, the higher extent of lipid peroxidation of these fats and the observed mitochondrial damage in patients who are insulin resistant and/or patients with type 2 diabetes could be linked if the previously mentioned negative feedback loop is disturbed. Although the possibility of a link clearly needs further investigation, it is intriguing to note that patients with type 2 diabetes, as well as subjects with impaired glucose tolerance who are often considered as being ‘prediabetic’, have a 50% reduction in UCP3 protein levels (Schrauwen et al. Reference Schrauwen, Hoeks and Hesselink2006). It is tempting to speculate that in patients with type 2 diabetes the reduction in UCP3 is a pathological condition in which low levels of UCP3 fail to sufficiently protect mitochondria against lipid-induced mitochondrial damage, indicating a defective feedback mechanism between lipid peroxides and mitochondrial uncoupling. This notion is further supported by the finding (Mensink et al. Reference Mensink, Hesselink, Borghouts, Keizer, Moonen-Kornips, Schaart, Blaak and Schrauwen2006) that in subjects who are prediabetic (i.e. impaired glucose tolerant) a lifestyle intervention programme results in a 2-fold increase in UCP3 protein after 1 year, thereby restoring UCP3 to normal physiological levels. In addition, UCP3 protein content is also restored in patients with type 2 diabetes after 1 year of an endurance training programme, in parallel with improvement in insulin sensitivity (Mensink et al. Reference Mensink, Hesselink, Borghouts, Keizer, Moonen-Kornips, Schaart, Blaak and Schrauwen2006). Finally, it has recently been shown (Schrauwen et al. Reference Schrauwen, Hoeks and Hesselink2006) that the insulin-sensitizing agent Rosiglitazone up regulates the protein content of UCP3 in skeletal muscle of patients with type 2 diabetes, in parallel with an improved oxidative capacity, mitochondrial succinate dehydrogenase activity and muscular insulin sensitivity. In addition, evidence for an important role for UCP3 in the aetiology of type 2 diabetes also comes from the recent finding that genetic variation in the UCP3 gene is associated with the 10- and 15-year risk for developing type 2 diabetes mellitus (Gable et al. Reference Gable, Stephens, Cooper, Miller and Humphries2006). Obviously, these findings in human subjects do not yet prove that reduced levels of UCP3 are causally related to the development of lipid-induced oxidative damage and insulin resistance, and further studies are needed to prove a functional role for UCP3 in type 2 diabetes.