Advertisement

Control of Energy Expenditure in Humans
Chapter 4A - John Speakman

Chair of Zoology, Institute of Biological and Enviornmental Sciences
University of Aberdeen, Aberdeen, Scotland, UK

J.Speakman@abdn.ac.uk

01 July 2008

TO OBTAIN A DOWNLOAD OF THIS CHAPTER IN WORD OR PDF FORMAT, CLICK HERE

Contents
Contributors
Search

Energy balance

All living things use energy continuously to fuel a number of different physiological and behavioural functions. This energy is provided predominantly by food intake (we can also use energy directly as heat, for example solar radiation, to offset thermoregulation requirements, but this is generally a minor contributor in most circumstances). There is consequently a continuous flow of energy that is being supplied by food intake, and is being utilised in cells throughout our bodies, and ultimately dissipated as heat. In the short term these two flows of energy are seldom in balance. We eat food in discontinuous meals, although the process of digestion evens out the supply somewhat, and we expend energy at different rates over time. When resting we expend considerably less than we do when exercising for example. Because animals cannot match their intake to their immediate metabolic needs there is a requirement to store energy. Energy stores can then take up excess energy following periods of excess intake, and they can be drawn upon during periods of excess requirement. The gut itself acts as a storage organ to even out the discontinuities in food intake. Food that is ingested passes immediately into a storage organ: the stomach in mammals and the crop in birds. This storage allows individuals to feed rapidly without being limited by the capacity to digest the ingested food items. Once absorbed, energy may be stored as glycogen in the liver and in muscles. The glycogen reserve fluctuates widely in time and acts as the immediate reservoir of stored energy that supplies most transient changes in energy expenditure. Any excess energy intake or requirement beyond the glycogen storage is generally stored as fat (Figure one)


Figure 1. Interactions between energy intake, energy expenditure and energy storage as glycogen and fat.

To understand energy balance a useful analogy is household financial budgeting. Our finance comes in discontinuously in a lump sum generally weekly or monthly, yet our expenditure is more continuous. To accommodate this mismatch we open a current bank account into which our salaries can be paid and from which we can readily draw money when we need it. The current account is like our glycogen reserve. If we start to accumulate too much money in our current account we can transfer it to a savings account. Similarly, if we have particularly high spend we can transfer money out of our savings into the current account. The savings account is like fat storage in our bodies. The system of energy use (and domestic financing) is slightly more complex than this. We can use ingested energy directly to fuel metabolism without storage as glycogen first, in the same way that we can spend money (taxes) without it entering our current account. Recent examples demonstrating this direct use of energy comes from humming birds and nectar feeding bats which appear to take up sugars from nectar and use this immediately to fuel their expensive flight behaviour (Welch et al. 2006; 2008; Voigt et al. 2007). We can also deposit ingested energy directly into fat storage without the need to first convert it to glycogen, in the same way one could pay money directly into a savings account. Similarly we can withdraw energy for expenditure directly from fat storage without conversion to glycogen in the same way that we can spend directly from a savings account. These flows of energy between intake, expenditure and storage are illustrated in Figure one.


Energy expenditure –subdivisions of expenditure

Basal Energy expenditure/Resting Energy expenditure

Humans and animals use energy to support diverse physiological and behavioural processes. Even when an individual is completely at rest they still burn up a substantial amount of energy. This energy expenditure is generally called the resting energy expenditure (REE) or the resting metabolic rate (RMR). There are several major factors that influence the level of RMR. One is posture. RMR is lower if we are supine than if we are sitting, and both these are lower than if we are standing. A second factor is how long it has been since we have eaten. There is a surge of energy expenditure following a meal. This can last for several hours. It was initially thought that the energy demands following feeding were mostly due to the metabolism of the digestive organs and the process of peristalsis. However, direct perenteral infusion of nutrients generates a similar effect on metabolism so the contribution of these digestive effects appears relatively minor. In fish, where this effect is a much greater contribution to the overall energy budget, studies indicate there is a surge of protein synthesis (which consumes much energy: Rolfe and Brown, 1997) following feeding that seems to comprise the bulk of the effect. Perhaps the largest impact on RMR however is from the ambient temperature. As it gets cold a body needs to generate heat to balance the increasing loss of heat from its surface if the body temperature is to be kept constant. The energy demand for this thermoregulatory heat is a linear function of the difference between ambient temperature and body temperature (Figure 2). Naturally if the ambient temperature is the same as body temperature there is no heat requirement to keep warm. However, animals and humans cannot switch off their RMR completely so in fact under these conditions too much heat is being generated and subjects need to dissipate this excess heat. Paradoxically the very process of trying to get rid of heat may generate more heat and so RMR may actually increase as temperatures rise so high.


Figure 2. The relationship between resting energy expenditure and ambient temperature in a typical endotherm (in red). The lower critical temperature is the point at which energy demands for thermoregulation (black dotted line) are exactly balanced by minimal metabolic rate (black line). Lower temperatures require specific production of energy to thermoregulate while higher temperatures require heat dissipation.


As it gets colder there is a point at which the heat generated by all the metabolic processes the animal is doing balances exactly the thermoregulation requirement. This temperature is called the lower critical temperature, and in lightly clothed humans is about 21-22 oC. We regulate our homes and working environments around this temperature as it is the most comfortable. Any colder than this lower critical temperature and we have to expend energy specifically to thermoregulate and any warmer we start to become uncomfortable because we have excess heat to dissipate. An interesting point to note here is that the lower critical temperature in mice is about 30-32 oC (Speakman and Rossi, 1991). So when we keep mice in routine ‘room temperature’ conditions they are already about 10 oC below their lower critical. Mice can compensate for being below their lower critical temperature by huddling together to keep warm. Group housing mice may therefore have very profound effects on not only their social behaviour but also on their thermoregulatory energy demands. Studies which have attributed social isolation as a factor that influences susceptibility to obesity (e.g. Nonogaki et al. 2007) may equally reflect changes in the thermoregulatory environment that is dependent on group size, which was not controlled. The difference between human and mouse lower critical temperatures and the effects of group housing on thermoregulatory demands are seldom considered or controlled in studies of mice as models of human energy regulation.

Because so many things affect RMR it was decided early in the study of energy metabolism that it would be useful to define a more rigorous measurement where all these factors were controlled. This would provide a better basis for comparison. Hence around the turn of the last century Basal Energy Expendiuture (BEE) or Basal metabolic rate (BMR) was defined as the energy expenditure of an individual animal or human at rest (i.e. not physically active), in a supine posture, in a post-absoprtive state (ie sufficiently long after feeding for the thermic effect of food to have disappeared), and at a thermoneutral temperature – i.e. at or slightly above the lower critical temperature. BMR is known to vary between the active and inactive phases of the daily cycle, independent of the activity being performed. Hence, a BMR measure in a mouse during the day, which is part of the inactive phase of the daily cycle for a mouse, is not strictly equivalent to BMR measured in a human during the day, which is the active part of the daily cycle for a human. This cycle in BMR mirrors the daily cycle in body temperature.

RMR measured in the thermoneutral zone, differs slightly from BMR normally being about 5-10% higher reflecting the thermic effect of food. At these temperatures they are almost equivalent. At lower ambient temperatures RMR can be two or even three times higher than BMR, reflecting energy expended specifically on thermoregulation. This extra thermogenesis heat can be generated in three different ways – as a by-product of physical activity, through shivering and via non-shivering mechanisms, in mammals particularly in brown adipose tissue (BAT). There is a suggestion that this energy expenditure can also be switched on deliberately in the absence of any thermoregulatory need to ‘burn off’ excess energy rather than storing it as fat – the so-called luxusconsumption (see below).

There is large individual variation in BEE between individuals. This is true in both animals and in humans (see Figure 3a and b). The major factor that influences BEE is the mass of lean tissue that is metabolising (Weinsier et al. 1992; Cunningham, 1991; Fukagawa et al. 1990; Johnstone et al. 2005). In humans, variation in lean tissue mass accounts for about 40-50% of the inter-individual variation. In vitro fat cells have very low metabolism, but in vivo fat tissue mass contributes the second largest proportion of the explained variation in BEE (Nelson et al. 1992; Svedsen et al. 1993). The effect of fat tissue is much greater than anticipated from its metabolism in vitro, suggesting secreted factors derived from fat may influence the energy demands of the lean tissue compartment. Females tend to have lower BEE than males, but this effect disappears when the adjustment is made for body composition, suggesting females only have lower BEE at a given body weight because of their greater proportional contribution of fat tissue to total body weight. BEE also declines with age (Poehlman, 1990; Roberts, 1995; Piers et al. 1998; Klausen et al. 1997; Weyer et al. 1999b; Johnstone et al. 2005). The effect is small but persists even when the changes in body composition are accounted for. The residual variation once body composition effects have been accounted for is still large. Two individuals with identical levels of lean tissue mass and similar fat mass may still differ by over 2 MJ per day in their BEE (e.g. Figure 3a). Considerable effort has been directed at trying to explain this residual variation.



Figure 3. Variation in basal and resting energy expenditure in (a) humans (b) MF1 mice as a function of fat-free mass and body mass respectively. Although fat-free mass and body mass explain a large amount of the individual variations in energy expenditure there is still considerable residual variation. (a: from Weyer et al. 2002, b: from Krol et al. 2003). In (a) individuals A and B have the same lean mass but differ in RMR by over 2 MJ/day.

One school of thought is that much of this variation can be traced to individual differences in the sizes of different components of the lean tissue mass (Heymsfield et al. 2002; Gallagher et al. 1998). Lean tissue is clearly not homogenous in its energy demands. Tissues like the brain, liver, heart and kidneys have high levels of energy demand, while skeletal muscle has a relatively low rate of energy use when a person is at rest (Elia, 1992; Krebs et al. 1950). By estimating the sizes of different tissue sizes using MRI and combining these with estimates of energy demands of the different tissues it is suggested a large portion of the residual variation might be explained in humans (Gallagher et al. 1998) and in animals (Even et al. 2001; Greenberg and Boozer, 2000). However, an opposing school of thought is that much of this variation in BEE may actually reside in tissue level variation in metabolism between individuals which is assumed to be constant in the tissue size modelling work. This is supported by the fact that some of the residual variation can be traced to variation in circulating hormone levels – such as thyroxine (e.g Johnstone et al. 2005). Thyroid status has a strong effect on RMR (Astrup et al. 1992; Toubro et al. 1996) an effect that has been established for over 100 years. Patients that were hyperthyroid had RMR levels approximately 15% high than expected when corrected for body composition (Jacobsen et al. 2005). These elevated levels were reduced when the hyperthyroidism was treated. Thyroid hormones exert their effects on BEE via effects on gene transcription acting via thyroid responsive elements (Silva, 2003; Lanni et al. 2001; Silvestri et al. 2005). These upregulated factors include nuclear respiratory factors 1 and 2 (NRF1 and 2) and uncoupling proteins 1 and 3 (UCP1 and UCP3 see below).

Studies in rodents have implicated circulating leptin as a key signal related to levels of resting metabolism (Schmidt et al. 1997; Scapace et al. 1997). For example, the mutant ob/ob mouse that lacks functional leptin has lowered resting metabolic rate and lowered body temperatures (Zhang et al. 1994). In humans many studies have sought similar associations (Jorgensen, et al. 1998; Nicklas et al. 1997; Toth et al. 1997; Niskanen et al. 1997; Bobbioni-Harsch, et al. 1999; Roberts et al. 1997; Kennedy et al. 1997; Nagy et al. 1997; Compostano et al. 1998) however some of these studies report a positive link, while others suggest the effect is negative and yet others have failed to find any relationship. This variability may reflect the manner in which the shared effects of fat mass on both traits are treated (Neuhauser-Bethold, et al. 2000) and when this shared factor is removed statistically the association between leptin and RMR disappears (Neuhauser-Bethold, 2000; Johnstone et al. 2005). The absence of a link between leptin levels and RMR is consistent with the observation that humans lacking functional leptin production (Montague et al. 2000) did not have significantly altered BMR. The differences between effects of leptin in humans and rodents may in part be because of the temperatures at which laboratory rodents are routinely kept (about 10 oC below lower critical) compared to humans which are normally at thermoneutral (as described above). Hence the effects of absence of leptin on BMR of the ob/ob mouse, and its lowered body temperature, may reflect the absence of stimulation of brown adipose tissue thermogenesis (Scapace et al. 1997). In humans, that have their BMR measured at thermoneutral temperatures, where BAT is inactive, such effects of leptin would not be anticipated. This striking difference illustrates the consequences of neglecting the thermoregulatory status differences between mice and humans when both are housed at 21 oC.

There are also several genetic factors likely to impact on tissue level metabolic rates. For example, the entire coding sequence of the alpha(2B)-adrenoceptor gene was screened in a sample of obese people from Finland. A polymorphism that leads to a deletion of 3 glutamic acids from a glutamic acid repeat element (Glu x 12, amino acids 297-309) present in the third intracellular loop of the receptor protein was discovered. This was important because this repeat element has previously been shown to influence agonist-dependent receptor desensitization. Of 166 genotyped subjects, 47 (28%) had 2 long alleles, 90 (54%) were heterozygous, and 29 (17%) were homozygous for the short form. The BEE adjusted for fat-free body mass, fat mass, sex, and age, was about 0.4 MJ/day (5.6%) lower in subjects homozygous for the short allele than in subjects with two long alleles (Heinonen et al. 1999). Similarly a variant in the UCP2 gene has also been linked to 24h energy expenditure (Walder et al. 1998; Astrup et al 1999; Kovacs et al. 2005).

Until we are able to combine in vivo PET scanning measures of tissue metabolism with simultaneous MRI estimates of tissue size and whole body gas exchange the relative contributions of individual variations in tissue size and metabolism will remain uncertain. It seems likely that both will turn out to be important. Error in measurement accounts for about 2-3% of the variation.

Activity energy expenditure

The next major category of energy expenditure is physical activity – or activity energy expenditure (AEE). In humans this has often been divided into two components – physical exercise, that is energy spent on activities such as running and swimming, and non-exercise activity thermogenesis or NEAT, which is the energy spent on all other physical activities – like walking around and fidgeting (Kotz et al. 2008). Since animals do not often perform physical exercise simply for the purpose of exercise this division is not generally applicable to animals. However, wheel running in laboratory mice has some elements of similarity to physical activity in humans and could perhaps be considered in this way. Energy demands on physical activity are often expressed in multiples of BMR. Hence it might be stated that running involves expenditure of energy at 12x BMR. This is sometimes called 12 Mets. The Met is a pseudo-unit because the expression is actually a ratio of two energy expenditure measures and therefore dimensionless.

A measure of total physical activity energy expenditure that has been popular is to express the total daily energy expenditure as a multiple of the basal energy expenditure (ie DEE/BEE). This ratio expresses the level of total energy demand as a multiple of BEE. Ignoring the thermic effect of food this is roughly equivalent to the average level of energy expended on physical activity in Mets averaged over 24 hours. The ratio DEE/BEE is called the Physical Activity level or PAL. Human PAL averages around 1.65 to 1.75 in different human populations (Black et al. 1996; Westerterp and Speakman, 2008).


Daily energy expenditure

Daily energy expenditure is the sum of all the component energy demands and their relative magnitudes in a human and in a mouse both at 21 oC are illustrated in Figure 4. These components are not strictly additively related. So for example heat generated by physical activity may substitute for the energy demands of thermoregulation. At thermoneutral however the different components are more likely to be independent and additive in nature. The interrelationships between the different components in this figure show that the concept of PAL (= DEE/BEE) which roughly equates to the relative expenditure of energy on physical activity in humans breaks down in small animals because of the component of the energy budget expended on thermoregulation. Nevertheless this ratio of DEE/BEE has been popular in studies of small mammals as a relative measure of how great their energy demands are. In addition to lack of additivity there are also complex interactions between the different components. For example, engaging in physical activity can cause a post exercise elevation in resting metabolism that may persist for many hours, yet engaging in excessive physical activity may actually cause a reduction in RMR (reviewed in Speakman and Selman, 2001).

 


Figure 4. Components of the energy budget of a human and a mouse at 21 oC. At this temperature the human has no thermoregulatory requirement so the budget consists only of the basal energy expenditure (BEE), the thermic effect of food (TEF), and the Activity energy expenditure (AEE) comprising Non-exercise activity thermogenesis (NEAT) and physical exercise. In the mouse there is additionally the thermoregulatory demand (Treg) because it is 10 oC below the lower critical temperature. Physical activity heat production may substitute for some of the thermoregulatory demands (shaded area so the components are no longer additive).



Measuring energy expenditure.

Calorimetry

The most direct way to measure energy expenditure is to measure the heat coming out of the body. There are several different pieces of equipment that permit such measurements collectively called ‘direct calorimeters’. Direct calorimetry works well for large animals, like humans, but for small animals the amounts of heat they produce are quite small and difficult to quantify. Fortunately there is another method available. When animals ingest food it consists of a wide range of different macronutrients. Cells however can only use one type of fuel – ATP. Cells don’t store much ATP – it is produced on demand. The oxidative phosphorylation process that converts macronutrients into ATP (see below) uses oxygen and generates CO2. By measuring the rate at which animals use up oxygen, or produce CO2, we can get a very accurate, but indirect, measure of their metabolism – hence this is called ‘indirect calorimetry’.

Converting oxygen consumption into energy requires knowledge of the substrate that is being metabolised. We can obtain a reasonable estimate of this from the balance of the consumption of oxygen and production of CO2. This ratio, called the respiratory exchange ratio (RER) or respiratory quotient (RQ), is approximately 1.0 for complete combustion of carbohydrates, and about 0.70 for combustion of fats. The RQ for oxidation of protein depends more on its composition but is around 0.83. If it is assumed that protein oxidation is negligible the RQ can be used to estimate the contribution of carbohydrate and fat oxidation to oxygen consumption, and this can be used to convert the oxygen consumption into energy expenditure. The error introduced by assuming negligible protein oxidation is generally less than 2%.

There are two different types of system that measure metabolism by indirect calorimetry. These are called closed and open-flow systems. In closed systems the animal is sealed into a closed chamber. Slowly over time it consumes oxygen in the chamber. By including a substance into the chamber that absorbs the expired CO2 there will be a decrease in pressure inside the chamber. The declining pressure can be converted into a volume change which allows a direct estimate of oxygen being consumed. The major problem with these systems is they have rather poor time resolution, so it is hard to separate periods of activity from inactivity. Moreover, it is not possible to estimate the CO2 production at the same time as O2 consumption, so it is difficult to estimate RQ and hence get an accurate conversion from oxygen consumption to energy.

Open flow systems are more sophisticated and involve a constant flow of fresh air coming into a chamber containing the subject. The animal (or human) consumes some of the oxygen in this air and replaces it with CO2. Hence by using a gas analyzer to quantify the change in gas composition between that coming in and that leaving it is possible to estimate oxygen consumption and CO2 production simultaneously. This allows a real time evaluation of RQ and hence substrate utilisation and conversion to energy expenditure.

Doubly-labelled water

Direct and indirect calorimetry are very accurate methods for evaluating the energy demands of people and animals when they are confined in small chambers. However, researchers are often much more interested in how much energy individuals expend when they are living in the real world. There are two basic approaches to this problem: the heart rate method and the doubly-labeled water technique (Butler et al. 2005). The heart rate method relies on the Fick equation for the cardiovascular system

 

VO2 = ƒH x Vs (CaO2 - CvO2)

where Vs is cardiac stroke volume; CaO2 is oxygen content of arterial blood and CvO2 is oxygen content of mixed venous blood, VO2 is the oxygen consumption and ƒH is the heart rate. This equation predicts VO2 will be a function of heart rate. By measuring heart rate in a free living individual one could therefore potentially directly estimate oxygen consumption. The usefulness of the method depends on how good that relation is between VO2 and ƒH. This relationship is often generated by study of individuals in the laboratory engaged in various activities. Whether or not these relationships are similar to those that would be obtained during free-living activities is unknown. Validation studies of this method suggest that individual estimates have wide errors but the mean oxygen consumption estimated for a number of individuals is, on average, within a few percent (range, 3.7% to – 2.13%) of the mean oxygen consumption measured for the same individuals.

A major advantage of the heart rate method is that it is able to provide estimates of the metabolic costs of specific activities over long periods. The current capacity of modern ƒH data loggers allows us to obtain minute by minute “pictures” of energy demands over extended periods provides an unparalleled insight into the detailed aspects of free-living energy demands. The most significant problem with the ƒH method is the fact that it is necessary to derive a calibration equation for the individuals under study. The logistical difficulties of performing such calibration work are probably the most significant constraints on current applications.

The doubly-labelled water method for estimating total CO2 production is a completely different approach. Using this method one attempts to measure the integrated energy demands over a period of time without trying to break this down into the energy allocated to different activities. The method is based on the observation that the oxygen in respiratory carbon dioxide is in isotopic exchange equilibrium with the oxygen in body water. This isotope exchange is catalysed by carbonic anhydrase. The speed of the reaction is such that within a few seconds of being produced, any carbon dioxide has the same oxygen isotopic composition as the oxygen in body water. If an isotopic label of oxygen (such as 18oxygen) is introduced into body water (either by injection or by drinking a dose of isotopically enriched water) that dose of isotope will be gradually eliminated from the body over time. This is because materials such as urine, evaporation and eliminated respiratory carbon dioxide continuously carry the isotopic dose away from the body, while respiratory water, pre-formed water in food, drinking water and respiratory oxygen continuously replenish the system with unlabelled oxygen. The rate at which all these materials flow through the body determines how fast the isotope is washed out, but this relationship also depends on the size of the body water pool as well. Larger body water pools flush out more slowly than smaller pools, everything else being equal. In fact if the logged isotope enrichment above background is plotted against time, the gradient of this linear relationship (called ko) multiplied by the body water pool size gives a quantitative estimate of the total flow of materials through the body that carry the isotopic oxygen signal. If we know the actual amount of isotope introduced into the body, then the body water pool size (No) can be estimated from the same curve by the dilution principle.

Dosing an animal or human with labelled oxygen and tracking the decline in subsequent isotope enrichment can therefore tell us the total flux of oxygen through the body, but being a combination of the CO2 and water flux this is a fairly useless measure. However, because water contains hydrogen, and carbon dioxide does not, introducing a label of hydrogen (such as deuterium or tritium) into the body water and tracking its elimination would enable a measure of the water flux alone. Consequently if one were to introduce both isotopes at the same time (hence doubly-labelled water) and measure the elimination of both isotopes, an estimate of CO2 production could be determined by difference. Validation experiments give an indication of the overall accuracy of the method by comparing estimates of MR with simultaneously-obtained measurements using, for example, indirect calorimetry. Validations have been performed on a wide range of animals weighing between 300 mg (bumble bees Bombus terrestris) and humans (80 kg). In almost all the validations performed to date using these sample sizes, the average discrepancy between the DLW and the reference method of indirect calorimetry has been less than 10%. Across all validations the mean discrepancy was 3.1% in mammals, 2.4% in birds and 0.5% in reptiles (Speakman, 1997). For individual estimates however, the estimates from DLW can be substantially more deviant from the reference method. It is not unusual within a group of 10 individuals under validation to find some individuals with estimated CO2 productions that differ by more than 20% from the reference measure. The primary advantage of the DLW method is that it provides a direct estimate of CO2 production (and hence daily energy expenditure) that is independent of assumptions concerning the mode in which the utilisation of energy has been made. A second advantage is the ease with which the method can be applied in the field. In its simplest form, the technique requires only that a subject be administered isotopes (injected or drunk) and then provide two blood or urine samples.

Molecular basis of energy expenditure

Energy consuming processes in cells, such as protein synthesis and the maintenance of ion gradients (Rolfe and Brown, 1997) are supplied only by the conversion of ATP to ADP and inorganic phosphate. Consequently, macronutrients ingested as food must be converted into ATP. This process involves several different stages which happen in different sub-cellular compartments. First, the diverse substrates (proteins, fats and carbohydrates) are processed into precursor molecules –the most important of which are Pyruvate and Acetyl-Coenzyme A, which is a common point in both glucose and fatty acid degradation pathways. Second, these substrates feed into the tricarboxylic acid cycle or TCA cycle, which utilizes the energy in the molecules to convert NAD+ to NADH by trapping electrons. NADH carries the energy rich electron and donates it to a series of proteins embedded on the inner mitochondrial membrane called the electron transport chain (ETC). The ETC functions to slowly capture the potential energy in electrons and use it to pump proteins from the mitochondrial matrix into the mitochondrial inter-membrane space. These protons then pass back through the membrane via ATP synthase and their potential energy is trapped to generate ATP. At the end of the ETC the electrons are combined with oxygen and protons to form ‘metabolic’ water, which explains the need for oxygen to generate energy.

For carbohydrates the first stage of breakdown (glycolysis) happens in the cytosol, and generates a small amount of ATP independent of the ETC. Glycolysis involves two main phases. In the first phase the process consumes energy (two ATP molecules degraded to ADP + P) to convert the six carbon glucose molecule into two three carbon molecules (glyceraldehyde 3-phosphate and dihydroxyacetone phosphate). This is followed by a phase where the two 3C molecules are converted to pyruvate. This second part of the pathway generates 4 ATPs. The end products of glycolysis are 2 Pyruvate, 2 ATPs plus 2 NADH molecules. Cells can therefore generate some stored energy (ATP) without the need for oxygen, although the yield of energy is very low relative to the energy contained in the original glucose. Each generated ATP stores about 10 kcal, making 20 kcal of energy stored while the original glucose molecule contains 686 kcals of energy. Glycolysis therefore only traps about 3% of the available energy. It is exceedingly inefficient. In addition, glycolysis could not be sustained for very long because the cell would rapidly run out of NAD+ and accumulate pyruvate. For glycolysis to be sustained organisms need to regenerate the original NAD+ and do something with the pyruvate. To regenerate the NAD+ they use various compounds to accept the hydrogen from the NADH. There are several alternative pathways and various compounds are used to take away the hydrogen from NADH to reform NAD+. These different anerobic processes generate different end products. For example, yeast convert pyruvate to ethanol and CO2 (without which we could have no bread or alcoholic drinks), while most eukaryotic cells and some bacteria generate lactic acid from the pyruvate. Many bacteria use inorganic compounds to accept the hydrogen from NADH and regenerate the NAD+. While prokaryotes appear able to live indefinitely using anerobic metabolism, most eukaryotic cells are unable to generate sufficient energy by this route to remain viable for very long.

Fat is generally stored as triglycerides. These consist of three fatty acids that are fused to a glycerol molecule. The process of breakdown of these fatty acids into precursor substrates is called beta oxidation. The first stage of the reaction is to create a fatty acyl-CoA from the free fatty acid. This involves two reaction steps. First the fatty acid reacts with ATP to give a fatty acyl adenylate, plus inorganic pyrophosphate. The acyl adenylate then reacts with free coenzyme A to give a fatty acyl-CoA ester plus AMP. This process happens in the cytosol. The next step in the process happens inside the mitochondria. The mitochondrial membranes are impermeable to long chain acyl CoA molecules so they need to be actively transported into across the membranes. This involves combining the acyl co-A with carnitine and splitting off the coenzyme-A. The acylcarnitine is then transported across the inner mitochondrial membrane by carnitine-acylcarnitine translocase. Once it crosses the inner membrane the acylcarnitine is converted back acyl CoA by carnitine acyltransferase II. This is the reverse reaction to that which formed the acylcarnitine and involves splitting off the carnitine and recombining the molecule with the coenzyme-A. The carnitine that is released can then be recycled to react with another molecule of acyl-CoA on the outer membrane.

Once the acyl-CoA molecule is in the mitochondrial matrix it is systematically cleaved at the carboxyl end to release molecules of acetyl-coA. Carbon atoms in the fatty acid molecule are labeled using greek letters from the carboxyl end of the molecule (alpha, beta, gamma, delta… etc). The molecules of acetyl-CoA are split off by breaking the bond between beta and gamma carbons. The process is therefore called beta oxidation.

The end products of glycolysis, and lipolysis may be utilised to generate ATP via aerobic metabolism which occurs inside the mitochondrial matrix. The end product of glycolysis (pyruvate) is produced in the cytosol outside the mitochondrion. Acetyl-coA from beta oxidation by contrast is generated already inside the mitochondrion. The pyruvate from glycolysis needs to be transported into the mitochondrion like the acyl-coA was. Once inside the matrix pyruvate is also converted to acetyl co-A which is the common entry point into the tricarboxylic acid cycle (TCA cycle). The conversion of pyruvate to acetyl-coA is performed by the enzyme pyruvate dehydrogenase. Pyruvate has 3 carbons and Acetyl-coA only 2, so the spare carbon in this case is released as CO2. This extra release of CO2 is why metabolism of carbohydrate generates relatively more CO2 than metabolism of fat (see RQ above).

Acetyl co-A in the mitochondrial matrix containing 2 carbons combines with oxaloacetate containing 4 carbons to form citrate, which is the first six carbon substrate in the TCA cycle. In the TCA cycle citrate is progressively converted via other six and five carbon intermediates to oxaloacetate again, with the two carbon molecules that are lost in the process appearing as carbon dioxide. Successive conversions in the cycle also generate protons, which reduce the substrates nicotine adenine dinucleotide NAD+ and flavin adenine dinucleotide FAD+ to generate NADH and FADH2. The primary substrate generated by the TCA cycle is NADH (four times as much is produced as FADH2). Both NADH and FADH2 act as electron carriers. They transfer electrons from the TCA cycle to the electron transport chain, the proteins embedded in the inner mitochondrial membrane, and then they recycle back to the TCA cycle to pick up more electrons. The ETC proteins use the energy in the electrons to pump protons from inside the mitochondrion into the space between the mitochondrial inner and outer membranes. The electrons are then ultimately combined with hydrogen ions and oxygen to form water. Oxygen is required for this process as the final acceptor of the electrons that are transferred along the chain of proteins.

Protons in the intermembrane space generate a pH gradient and a transmembrane electric potential (ΔΨ). It is the energy in this protonmotive force that is used to generate ATP. The mechanism is that protons pass back through the membrane via ATP synthase. ATP synthase involves two types of protein subunit f1 and f0, with the f0 complex located across the inner membrane which acts as a proton pore, and the f1 units projecting into the matrix, which are the catalytic units for conversion of ADP to ATP. Movement of protons across the membrane via ATP synthase results in conversion of ADP +P to ATP. The details of the mechanism of ATP synthesis at complex V have been worked out in considerable detail. ATP synthase is actually like a small molecular rotar.

We have seen above that glycolysis only traps about 3% of the energy in a glucose molecule. When animals metabolise glucose via glycolysis followed by metabolism of pyruvate in the TCA cycle the overall balance of the process results in consumption of 6 mols of oxygen and the simultaneous production of 6 mols of CO2. and 30 ATP molecules. In older text books this figure is often quoted as 36 ATPs but that estimate is erroneous. One mol of glucose contains 686 kJ. The energy captured by converting 30 ADP molecules to 30 ATP molecules is about 300 kJ. The efficiency of the process is about 40%.



Molecular regulation of metabolic activity

Activity of the ATP generating machinery of the cell is regulated by sensing the ratio of ATP to AMP. When ATP is utilized it is converted to ADP and inorganic phosphate. To reform ATP, ADP donates a phosphate to a second ADP molecule generating one ATP and one AMP. The ratio of ATP to AMP is therefore extremely sensitive to the utilization of ATP, and is called the cellular metabolic or energetic ‘charge’. When the ratio of ATP:AMP falls the cell is clearly consuming more energy than it is producing. The increasing AMP to ATP ratio activates a protein called AMP sensitive Kinase (AMPK). AMPK modulates many cellular functions (Hardie, 2003; Carling, 2004) but a major role it plays is to stimulate fat oxidation. It does this by phosphorylating and hence inhibiting Acetyl co-A carboxylase 2 (ACC-2). Acetyl-CoA carboxylase (ACC-2) converts acetyl-CoA to malonyl-CoA, which inhibits the carnitine parmitoyltransferase (CPT1). CPT-1, as we have noted above, is one of the primary transport proteins involved in moving long chain fatty acids from the cytosol into the mitochondrial matrix, where they undergo beta-oxidation. By inhibiting ACC2, AMPK reduces production of Malonyl -CoA therefore removing inhibition of CPT1, and allowing transport of fatty acids into the mitochondria to stimulate the TCA cycle and increasing ATP production. An additional route by which AMPK may reduce levels of malonyl-CoA is by increasing its decarboxylation to acetyl coA via stimulation of malonyl-CoA decarboxylase (MCD). Activated AMPK also seems to stimulate glycolysis in muscle cells via promoting translocation of the glucose transporter GLUT-4 to the cell membrane.

The calculation of the efficiency of substrate oxidation assumes that all the protons that exit the mitochondrial matrix come back via ATP synthase. However, not all protons pass back through the membrane via ATP synthase. Because they are charged, protons cannot move back through the inner membrane unless they go via a specific carrier. There are several known carrier proteins that allow protons back across the membrane without any generation of ATP. Some of these proton carrier proteins appear to serve to specifically generate heat. The first carrier protein known to perform this function was called uncoupling protein 1 because it uncouples the movement of protons across the mitochondrial inner membrane from the generation of ATP. UCP-1 is a 305-310 amino acid protein found exclusively in brown adipose tissue (BAT) and is specific to mammals. In structure UCP-1 has 6 transmembrane segments with both the C and N termini ending within the membrane. The protein forms a pore that can be gated to prevent flow of protons (Arechaga et al. 2001). Uncoupling activity is sensitive to fatty acids (positive) (Hagen and Lowell, 2000) and purine nucleotides (negative)

In the late 1990s several other uncoupling proteins were discovered with much broader tissue distributions – these consist of UCP-2 found ubiquitously , UCP-3 found mostly in muscle , UCP-4 found in the brain and BMCP (Brain Mitochondrial Carrier Protein) or UCP-5 also found in the brain . Sequence homology (bases) of these uncoupling proteins to UCP-1 is in the region of 55-75% . Whether these structurally similar proteins actually have uncoupling activity has been a matter of considerable debate (Ricquier, 2005; 2006). In mice however it seems that the level of uncoupling that these additional UCPs produce is insufficient to compensate for the knock out of UCP1. Hence UCP1 KO mice are unable to significantly upregulate heat production during acute cold exposure (Golobouzova et al. 2001; Nedergaard et al. 2001), and are able to survive chronic progressive cold only by continuously shivering to generate heat. This is consistent with the fact that proton transport by UCP1 is dependent on a pair of histidine residues that are absent in UCP2 and UCP3 (Bienengraeber et al 1998). In contrast transgenic overexpression of UCP3 leads to a large increase in oxygen consumption showing that super-physiological levels of this protein do generate an uncoupling effect (Clapham et al. 2000). Whether this is a native effect of the UCP3 protein or not has been questioned (Cadenas et al. 2002).

At thermoneutral temperatures where there is no specific thermoregulatory requirement it appears that UCP1 is inhibited by nucleotides and is virtually inactive (Cannon and Nedergaard, 2004). As we noted above this is the condition that most adult humans live in most of the time. Yet mice in the laboratory at 21oC will have their UCP1 activated most of the time to provide the extra heat required for thermoregulation. For a long time it was believed that while infants have large quantities of brown adipose tissue adult humans might only have occasional brown adipocytes hidden among white adipocytes in regular white fat depots. Recent work using PET scanning has revealed that there are depots of brown adipose tissue also in adult humans (Nedergaard et al. 2007) (Figure 5). As might be expected these reserves are most obvious when the subjects are scanned in mildly cool conditions below 21oC where there is a thermoregulatory requirement. Whether human brown adipocytes can be recruited to burn off excess calorie intake has been an issue of debate for almost 30 years.


Figure 5. Sites of flourinated glucose uptake corresponding to brown adipose tissue in adult humans. The black areas are those that are most frequently described; the gray areas are not always found, even in humans positive in the black areas. After Nedergaard et al. 2007.



The activity of UCP1 in brown adipose tissue is under central control. The primary pathway for control is the sympathetic nervous system (Wirsen and Hamberger, 1967; Ballard et al. 1974). Retrograde trans-synaptic tracing from brown adipose tissue back to the brain using attenuated rabies virus strongly implicates the origin of this control among the melanocortin circuitry of the brain (Voss-Andreae et al. 2007). This role for MC4 receptors is supported by data showing MC4R knock out animals have defective capacity to upregulate thermogenesis when exposed to the cold (Voss-Andreae et al. 2007). Sympathetic nerves innervate BAT and brown adipocytes have three beta adrenergic receptors (b1-AR, b2-AR and b3-AR) (Galizsky et al. 1993). The importance of the sympathetic system is demonstrated by the effects of knocking out dopamine beta hydroxylase which is involved in the synthesis of noradrenaline. These animals are also unable to upregulate their heat production when exposed to the cold (Thomas and Palmiter, 1997). A standard assay to assess the thermogenic capacity of brown adipose tissue is to inject an animal with Noradrenaline. This generates a large increase in heat production and oxygen consumption as the UCP-1 is activated and mitochondria are uncoupled (see figure 6). The beta-adrenergic receptors are members of the large family of G protein-coupled receptors, each of which is coupled to GalphaS and signals via increases in intracellular cAMP levels and the cAMP dependent protein kinase (PKA). There is some redundancy in the roles of the three receptor sub-types (Rholfs et al. 1995). Simulaneously knocking out all three beta-receptors in mice results in the inability to upregulate thermogenesis and the resultant mice are also obese (Jimenez et al. 2002; Bachman et al. 2003). Interestingly lipolysis in these mice remains effectively intact, suggesting alternative regulatory mechanisms – perhaps involving hormone sensitive lipase. In humans similar redundancy in the beta-receptor system is also apparent. Infusion for 3h with dobutamine (a beta-1 specific agonist) and with salbutamol (a beta 2 specific agonist) both elevated energy expenditure at rest by about 13%. Substrate ultilisation however did differ between the treatments with dobutramine increasing it by 47 +/- 7% and salbutamol increasing fatty adid oxidation by 19 +/- 7% (Hoeks et al. 2003).

 


Figure 6. Response of animal (mouse) oxygen consumption to injection of NA (indicated by arrow). Following injection there is a profound increase in metabolism peaking at over 3x the basal rate and which remains elevated for about 40-50 minutes from Arch et al. 2006).

Regulation of UCP1 activity occurs in two ways. First, the opening of the UCP1 pore to allow flux of protons across the inner membrane is stimulated by fatty acids. This overcomes the inhibitory effect of nucleotides which normally maintains the pore closed (Ricquier, 2006). Hence lipolysis stimulated by the PKA generates free-fatty acids that directly elevate UCP1 activity. Second there is transcriptional regulation of UCP1 production via a 220 bp enhancer region approximately 2.4 kb upstream of the UCP1 gene. This enhancer region binds several nuclear hormone receptors including the thyroid hormone receptor, the retinoic acid receptor and most importantly in the context of beta-adrenergic stimulation, the Peroxisome proliferator activator receptor gamma (PPARgamma). Following beta adrenergic stimulation, downstream of PKA the p38 Mitogen activated kinase (p38 MAPK) pathway is implicated (Puigserver et al. 2001; Cao et al. 2001). p38alpha upregulates the transcription factor ATF2, leading to an increase in PPARgamma co-activator 1 alpha (PGC1a) levels. The newly synthesised PCG1a is phosphorylated by p38alpha and this binds to and activates binding of PPARgamma to the UCP-1 promotor (Collins et al. 2004). Surprisingly given this mechanism animals with PPARgamma knocked out have a very mild phenotype, suggesting additional compensatory factors are involved in transcriptional regulation of UCP1. Such an effect may include the impact of cAMP on diodinase activity which converts intracellular T4 to T3 which binds to the thyroid hormone receptor beta in the control region also stimulating UCP1 gene expression (Ribiero et al. 2001; de Jesus et al. 2001) Beta-adrenergic, cAMP mediated transcriptional regulation of UCP can be directly inhibited by the liver X receptor alpha (LXRalpha). This involves LXRalpha recruiting the co-repressor RIP140 and binding to a LXR binding site in the UCP1 promoter enhancer which overlaps with a response element for PPARgamma/PGC1a and hence displaces PPARgamma (Wang et al 2008). Over the even longer term PGC1a and PPARgamma are also involved in differentiation and proliferation of brown adipocytes. Many other factors are involved in the regulation of thermogenic capacity. For example knocking out ACC2 results in elevated levels of UCP1 in BAT, UCP2 in heart and UCP3 in skeletal muscle (Abu-Elheiga et al. 2001).



Contribution of variation in energy expenditure to obesity

AEE

Changes in the amount of energy that we spend on exercise have often been suggested to be a primary factor that underpins the obesity epidemic. It has been noted that in parallel with the expansion in the levels of obesity there has been a gradual shift in our lifestyles which have entailed reductions in the levels of physical activity (Prentice and Jebb, 1991; Browson et al. 2005). This is often exemplified by increasing trends in the levels of car ownership and the time spent watching television. In addition more direct comparisons show that obesity is positively linked at the individual level to the time spent watching TV (Robinson 2001; Hu et al 2003, Jeffery and French, 2004; Marshall et al 2004). A randomised controlled trial reducing TV viewing reduced body fatness in children (Robinson 1999). By contrast reported levels of food intake have remained stable or have even decreased over the same interval (Prentice and Jebb, 1991; Brinkley et al. 2000; Cavaldini et al. 2000; Willet and Leibel, 2002). Direct observations of the obese suggest that they are less physically active than lean subjects (Ekelund et al. 2002) – although whether this is a cause of, or a consequence of their obesity is unclear. In rats differences in spontaneous physical activity predict suspectability to obesity (Kotz et al. 2008), but the same is not necessarily true in humans where individual variation in AEE does not predict future weight gain (Goran et al. 1998; Tatranni et al 2003).

One problem however is that these changes in activity are only indirect measures of activity energy expenditure. When direct measures of activity energy expenditure made using the DLW method are analysed a different picture emerges. First, the adjusted levels of activity expenditure (DEE minus BEE) do not show any significant trend between the 1980s and 2005 (Westerterp and Speakman, 2008). Moreover, the levels of energy demand on activity in modern western societies that are obese differ little from the levels of activity expenditure in the third world where subjects are considerably leaner and engaged in lifestyles that we would consider as energetically demanding. An excellent comparative study of energy demands in the third and first world involved a comparison of Nigerians living a rural existence in a village in Nigeria and African Americans living in Chicago. Despite the African Americans weighing on average 30 kg more than their Nigerian counterparts, there were no differences in AEE between the two groups (paper in press ‘Obesity’ : Amy Luke pers. Comm.)

Hayes et al. (2005) compared the levels of activity energy expenditure (PAL) in modern humans with those reported in wild animals and concluded that modern humans with a PAL level around 1.6 to 1.7 have substantially lower PAL levels than wild animals which average around 2.8 to 3.0. This analysis however is spurious because most of the wild animals that have been measured by DLW are relatively small (less than 10 kg) and it fails to account for the thermoregulatory component in the energy budgets of these small mammals (Figure 4). In fact the PAL declines as a function of body size in wild mammals and when this factor is taken into account the levels in modern humans do not differ significantly from the expectation based on their body weight (Westerterp and Speakman, 2008) (Figure 7).

 


Figure 7. Effect of body mass across different species of terrestrial mammal on physical Activity level (PAL). Data for humans in modern western society are shown as the black square. PAL declines with increasing body size as the demands for thermoregulation become lower. Modern human expenditure sits exactly where one might predict for an animal of our body size (after Westerterp and Speakman 2008).



Despite the attractive nature of the correlational evidence between trends in our lifestyles and the obesity epidemic actual measures of energy expenditure do not support a strong link. This is probably because the changes in our lifestyle while making a large impact on our lives may have little impact on changes in expenditure. Take TV viewing as an example. We have certainly dramatically increased our levels of TV viewing since the 1960s. However, TV viewing occurs predominantly in a time slot in the evening when we have probably always been sedentary. Prior to TV we probably spent this time listening to radio, prior to that reading, and prior to that either asleep or sitting around a fire. Hence a large change in TV viewing does not necessarily have a large impact on expenditure if it is replacing other sedentary pursuits. In fact the evidence increasingly shows that the individual links between TV viewing and fatness arise because of elevated food consumption during the time we spend watching TV (Matheson et al 2004; Coon et al 2001; Temple et al 2007).

The fact that changes in our habits and hence activity energy expenditure over the past 50 years have probably not been sufficient to drive the obesity epidemic does not mean that exercise interventions will be ineffective. In fact carefully controlled studies have shown that making subjects perform 1MJ of exercise does not result in a compensatory increase in intake at least in the short term (Blundell et al. 2003). However, the effects of more prolonged exercise are less clear and generally exercise interventions are difficult to adhere to and the levels that can generally be tolerated do not result in dramatic reductions in body mass or fat mass.

BEE/REE

The effects of variation in REE and BEE on susceptibility to obesity are unclear. The variability in BEE that can be attributed to differences in lean tissue mass do not appear to predispose people with low lean tissue mass to energy imbalance and fat deposition. If they did then shorter people would be much more prone to obesity than larger people, yet they do not appear to be so. This suggests that food intake levels are regulated in some manner to match the levels of demand that result from variance in lean tissue mass. How this happens is uncertain. Models of long term food intake regulation have focused primarily on the roles that may be played by leptin and insulin interacting with receptor populations in the hind brain and hypothalamus and how these long term signals interact with short-term regulatory signals derived from the alimentary tract such as Ghrelin, PYY and CCK (Schwartz et al 2000; Mercer and Speakman, 2001). Leptin and insulin both circulate in proportion to fat tissue mass, and are hence presumed to provide the architecture of a lipostatic regulatory system. However, it seems likely that there must also be signals involved in regulation of intake that sense the size and hence metabolic requirements of the lean tissue compartment. The nature of such signals is presently unknown, but the discovery of so-called myokines derived from skeletal muscle clearly show that the lean tissue compartment can also have endocrine functions in the same way that fat tissue does. Direct measurements in mice and humans show that individual variation in BEE do not predispose the mice to differential fat gain if they are then exposed to a high fat diet (mice: Johnston et al. 2007 Figure 8) because of compensatory variation in energy intake.






Figure 8. Weight gain of C57Bl/6 mice when exposed to medium or high fat diet as a function of their RMR measured prior to dietary exposure. Variation in RMR was not a predisposing factor for variation in weight gain (From Johnston et al. 2007). Note that despite being an inbred strain the response to the high fat diet in weight gain was highly variable across individuals.

 

Although the variation in BEE that is associated with lean tissue mass does not translate to variation in energy balance because the variability is compensated for by variation in intake, the effects of the component that is not linked to lean tissue mass remains less clear. Some studies have failed to find any link between this variance and levels of obesity (Goran et al. 1998; Katzmarzyk et al. 2000). However, many other studies have suggested that variation in the adjusted levels of RMR IS a risk factor for subsequent weight gain (Roberts et al. 1988; Ravussin et al. 1988; Siedell et al. 1992; Weyer et al. 1999; Tataranni et al. 2003; Buscemi et al. 2005). Moreover, some evidence suggests that common polymorphic variants in the beta- and alpha-adrenoceptors are linked to variation in BMI. For example, the ADRA2A GI 780A SNP was associated with BMI and percentage of body fat in African Americans (P =.05). Interactions were detected between ADRA2A C-1291G and ADRB2 Gln27 -> Glu variants for obesitv in African Americans and between ADRA2A C-1291G SNP and ADBRI haplotype for obesity in whites. However whether such polymorphic variation in the receptors is linked to obesity via effects on energy expenditure is uncertain. Adrenoceptor stimulation also mediates the switch from glucose to fat oxidation and hence polymorphisms in these receptors may mediate their effects on obesity not only via effects on BEE but also on substrate oxidation and partitioning. The individual variation in weight gain in Pima Indians reported by Tatranni et al. (2003) was also associated with variation in RQ. Differences in thyroid status also appear to not be completely compensated by differences in intake. Hence hyperthyroid subjects tend to have elevated RMR and lose weight (Hoogwerf and Nuttall, 1984; Abid et al. 1999) and when this is treated the metabolism declines, and often subjects gain weight (Hoogwerf and Nuttall, 1984; Lonn et al. 1998; Jannson et al. 1993; Dale et al. 2001; Jacobsen et al. 2005). However, the effect is not uniformly observed (e.g. de la Rosa et al. 1997) and there may be complex interactions between thyroid status and hormones linked to appetite control (Riis et al. 2003; Pijl et al. 2001).

The existence of a component of BEE that may not be well linked into the energy balance sensing apparatus has driven the search for compounds that will increase energy expenditure but not simultaneously stimulate appetite (Clapham and Arch, 2007). In fact compounds that increase metabolic rate have a long history of being used to increase BEE and produce weight loss. For example, thyroid hormones were popular in the 1890s, and the mitcohondrial uncoupler di-nitrophenol was widely used in the 1930s, before being banned because of dangerous side effects including hypoxia. Targeting increases in BEE in the periphery is attractive because it may avoid some of the complex issues and problems that have been encountered when attempting to manipulate energy intake via centrally acting anorectic drugs. These problems stem from the fact that the circuitry in the brain which senses and controls energy intake is closely linked to may other systems – such as the reproductive axis, and disentangling these multiple effects has proved challenging. Attempts to pharmaceutically increase BEE have principally involved stimulating the beta-adrenergic system (e.g. Arch et al. 1984). In particular selective agonists for the beta-3 AR have been able to prevent or reverse obesity and accompanying insulin resistance in animal models (Arch et al. 1984; Collins et al. 1997). Whether this is a viable therapeutic option in human obesity however has caused much debate (Arch, 2002), because the b3 receptor in humans appears different from that in rodents, and the effects of agonists have been disappointing (e.g. Larsen et al. 2002).

A second target is the thyroid axis (Silva, 2003). Stimulation of the metabolism with thyroxine has been used to treat obesity for over 100 years (Clapham and Arch, 2007), but such treatments generate many unwanted side effects such as muscle loss, bone loss, fatigue and cardiac stimulation. Attempts to get around the latter effect have utilised compounds that stimulate only the beta thyroid receptor which do not appear to simulate the heart (e.g. Grover et al. 2003). Growth hormone receptors are another target (e.g. Heffernan et al. 2001), or modulating transcription factors such as PPARdelta that stimulate mitochondrial proliferation, or others such as PPARalpha that promote fatty acid oxidation.

Many other compounds are known to cause increases in metabolic rate, perhaps the best known and most widely used of which is caffeine. Some, such as 3,4 methylene dioxy-methamphetamine (more widely known as the drug ecstacy) (Mills et al. 2003), capsicin (Kawada et al. 1991), and the sweet paper extract capsiate (c19) (Masuda et al. 2003) specifically cause increases in energy expenditure by stimulation of uncoupling proteins. In some cases their efficacy as weight loss agents is under investigation and initial results look promising.

Responses of energy balance to over and undernutrition

Caloric restriction

During caloric restriction BEE declines. This decline opposes the lowered calorie intake, making the rate of weight loss lower. The cause of the decline is two-fold. First, the restriction causes a decline in body weight (McNeil et al. 1987; Soares and Shetty, 1991; Shetty, 1993; Weinsier et al. 2000). Second, there may be an additional effect on metabolism per gram of metabolising tissue. There has been considerable debate over the extent to which these two effects contribute to the lowered metabolic rate. The general consensus from animal models however is that any ‘adaptive’ reduction in tissue specific metabolic rate is only transient, and that almost all the altered metabolism can be explained by changes in the body composition (McCarter et al. 1985; McArter and Palmer 1992; Ballor, 1991; Dulloo and Giradier, 1993; Greenberg and Boozer, 2000; Hambly and Speakman, 2005). Indeed it has been suggested that tissue specific metabolism may even increase in some circumstances (Selman et al. 2007). In humans there is more evidence pointing to an effect of reduced metabolism above and beyond that induced by changes in body composition alone (e.g. Weyer et al. 2002; Leibel et al. 2001; Heilbonn et al. 2006). However, the studies in humans are generally of much shorter duration relative to lifespan compared with the studies in rodents, and so any difference may actually be reflective of these different experimental durations rather than a fundamental difference between rodents and humans. Nevertheless studies of non-human primates over multiple years do suggest sustained reductions in metabolism beyond that expected from body composition change (Blanc et al. 2003; Delany et al. 1999).

Rodents have the flexibility to respond to undernutrition by modulating their thermoregulatory energy demands by entering periods of torpor. Torpor is a regulated state of very low energy demand where the heat production from metabolism is barely sufficient to elevate body temperature more than a few degrees above ambient. Rikke and Johnson (2004) showed that the extent of torpor use in mice under restriction varied enormously between strains and was most evident in the smallest strains. Finally animals may reduce their activity levels to compensate for the energy shortage. Hambly and Speakman (2005) showed that this was the largest component of energy compensation for restriction in the MF1 mouse strain accounting for about 75% of the energy saved – the balance coming from reductions in RMR due to body composition change. This is a relatively large strain (adults weigh around 35-40g) that does not use torpor to compensate its energy budget. In humans it is suggested that activity may become more efficient (Rosenbaum et al. 2003) and similar changes in efficiency are also inferred to occur in some animal studies (Dulloo and Giradier, 1992; MacLean et al 2004).

Overfeeding

The existence of an adaptive upregulation of metabolic rate following caloric overconsumption of energy or luxusconsumption is much more controversial. Certainly weight gain often does not match expectations based on simple balance measures in the same way as losses during restriction are lower than anticipated. However, in a review of studies of overfeeding Joosen and Westerterp (2006) highlighted that in many cases the weight gained by subjects did match the extent of excess energy consumption – suggesting that compensatory upregulation of expenditure did not normally occur. Where such discrepancies have been reported they are highly correlated across twins (Bouchard et al 1990) suggesting there is a strong genetic component to the compensatory response. However, whether the compensatory upregulation is in BEE, or a component of AEE remains controversial as does the magnitude of the effect observed in different studies (Joosen and Westerterp, 2006). Perhaps the most well studied aspect is the response in non-exercise activity thermogenesis (NEAT). Several studies have indicated that variation in upregulation of NEAT might contribute to differences in weight gain (for example Freymond et al. 1989 and Zurlo et al. 1992 in Pima Indian children and adults respectively). Perhaps the best known study however is that of Levine et al. (1999) who exposed subjects to excess energy intake (1000 kcal = 4.2 MJ) above baseline for a period of xx weeks. In common with other over-feeding studies there was substantial individual variation in the weight gain. Levine et al. (1999) suggested that the cause of this differential weight gain was differences in the levels of increase in non-exercise adaptive thermogenesis (Figure 9) quantified directly and indirectly from their activity patterns using acceleromters.


Figure 9. Changes in non-exercise activity energy expenditure in response to a period of over-feeding and the change in fat mass. Individuals that had greater stimulation of NEAT gained less weight (after Levine et al. 1999).

Spontaneous physical activity levels vary significantly between strains of mice (refs) and studies of humans similarly indicate a strong genetic component (refs). Evidence points to central factors controlling levels of NEAT. In particular Orexin A has a profound effect on levels of NEAT, and effect which varies when it is injected into different areas of the brain (Kotz et al. 2007: Table 1). This effect of Orexin A on NEAT can be blocked by pre-administration of the orexin 1 receptor antagonist SB334867 (Kiwaki et al 2004; Kotz et al 2002; 2006; Teske et al 2006), suggesting that the orexin 1 receptor is important in mediating the effect. Another neuropeptide stimulating physical activity when supplied ICV is Neuromedin U (reviewed in Kotz et al 2008). Initial reports also suggested that ICV administered Apelin was also associated with elevated activity and body temperature (Jazberenyi et al 2004). However more recent studies have indicated that the primary effect of Apelin is to increase food intake and the effects on activity and body temperature are secondary to this elevated food intake (Valle et al. 2008). In contrast to these stimulatory effects central administration of Agouti related peptide (AgRP) seems to produce a reduction in physical activity. Understanding how these various neuropeptides interact to regulate NEAT still remains a mystery.


Analysing the effect of genetic manipulations on energy expenditure

In the last few years there has been an explosion of studies of mice that have been genetically manipulated to generate obese or lean phenotypes. Naturally there is a desire to understand if the cause of these phenotypes stems from variability in energy intake or energy expenditure and many studies have made evaluations of these effects. Many researchers are aware that energy demands are impacted by lean tissue mass, and that it might be appropriate to make some correction for this effect in comparing different genotypes of mice, as often the genetic manipulation generates a body mass phenotype. However, while there is an appreciation that some form of correction for body or lean tissue mass might be necessary, there is an apparent widespread ignorance about the most appropriate procedure to make this analysis. This appears in part to be a consequence of large numbers of measurements being made using the Comprehensive Laboratory Animal Monitoring System (CLAMS) manufactured by Columbus instruments. While this is an excellent piece of indirect calorimetric equipment, the output of this machine automatically divides the measured oxygen consumption by body weight of the animal. It is important to recognise that simply dividing oxygen consumption by body weight (or fat free weight) is NOT an appropriate method for normalising body size effects on energy metabolism. The correction made in this manner generates a spuriously high metabolic rate in animals with low body weight. Many papers have been published in the last few years claiming to show high levels of energy expenditure in genetically manipulated animals that have also a low body mass, and almost all these studies have reached this conclusion by making inappropriate corrections for size effects. Other studies have not used the simple division of oxygen consumption by body mass, but have instead divided by body mass raised to the 0.75 exponent (so called metabolic mass). This approach is equally invalid because it utilises a scaling exponent derived from interspecific comparisons to normalise within species effects. It would be invidious to pick out examples of this bad practice, but they are unfortunately very numerous, and worse, are often published in high profile journals.

To overcome these problems the best and simplest statistical approach is to use analysis of covariance (ANCOVA) (see Arch et al. 2006) This approach is useful because it makes no prior assumptions about the nature of the scaling relationships between different body compartments and metabolism, but rather derives these empirically using the actual data. The generalised linear model for ANCOVA includes the assumption that the effects of body mass or fat-free mass on metabolic rate are linear and the traits are normally distributed. It is best to check these assumptions first by plotting the data for individual groups, and using a standard normality test. If the data are non-linear or not normally distributed, they can generally be corrected by using a normalisation procedure. Having convinced oneself that the data conforms to the test assumptions, the analysis assesses the effects of body weight, the effects of the group allocation (e.g.genotype) and the interaction of these effects.

The interaction evaluates whether the gradients of the effects of the independent variable (weight or fat-free mass) on metabolism are different between treatments (genotypes). If they are, then a formal comparison of a group effect is not possible. This is because at some point two non-parallel lines (i.e. that differ in their gradients) must cross. However, in practical terms one is not interested in whether the lines may eventually cross, but whether they differ in the region where one has data. The problem of an interaction can be overcome by using the regression line for each group to predict what the EE of each animal in that group would have been if it had had the average body weight for the two groups combined. The normalised data may then be compared using a standard t-test. Alternatively more sophisticated approaches such as the Johnson-Neyman technique may be applied, which defines the region of body weights over which no significant differences in metabolism are detectable. If, on the other hand, the interaction is not significant, this means the gradients of the effects of body weight (or fat-free mass) are not different in the two groups (i.e. they are parallel). Re-running the ANCOVA excluding the interaction term from the model can then formally test for a group effect.

The strength of this approach is that it makes no a priori assumptions about scaling relationships or differences in body composition, and can be run using either body weight or fat-free mass as the independent variable. Indeed if fat-free and fat mass are known, they can both be included in the analysis as two independent covariable factors. Examples good practice in the use of ANCOVA to analyse energy expenditure data are a recent study of mice with a mis-sense mutation in growth hormone (Meyer et al. 2001).