Background

I am greatly honoured to receive the Gowland Hopkins Award from the Nutrition Society, and indeed to be the first recipient. I never met Sir Frederick Gowland Hopkins OM PRS, who received the Nobel Prize for Physiology or Medicine in 1929 for the discovery of vitamins, being born a year after he died. There is, however, a tangential link in that my entry into nutrition from a basic science background in physiology and biochemistry came through joining the MRC Dunn Nutritional Laboratory in Cambridge in 1975. The Dunn had been founded in 1927 when Gowland Hopkins was Professor of Biochemistry in the University, and the original intent was that he should be directly involved in its research. In practice, because of extensive other commitments, including as President of the Royal Society, his primary role with respect to the Dunn was as an advisor and member of the Management Committee. Gowland Hopkins has, of course, a close association with the Nutrition Society as one of the principal figures behind its foundation.

My research over the past 40+ years since entering nutrition has centred on the adipose tissues; first brown and then white. This began following my initial studies at the Dunn where I had been recruited by Philip James as a member of the then newly formed Energy Group (Fig. 1). The group had been established in recognition of obesity beginning to emerge as a public health problem. At the time, the incidence of obesity was considerably less than now; in 1980, for example, 6 % of adult males and 8 % of adult females in the UK were classified as obese^{(Reference Prentice and Jebb1)} on the basis of a BMI ≥30, while by 2017 the figure was approximately 30 % of all adults (https://www.worldobesitydata.org/country-profiles/). Obesity is, of course, important primarily because of the increased risk of several associated diseases, particularly type-2 diabetes, hypertension, CHD and certain cancers^{(Reference Kopelman2,Reference Blüher3)} .

Fig. 1. (Colour online) ‘Engaging’ with nutrition at the Dunn: Friday morning group ‘coffee and cake’. Eating and drinking in the laboratory is, of course, prohibited now, but was normal in my early years as a scientist.

The ethos prevailing when we began in the mid-1970s was that obesity is the product of ‘gluttony and sloth’, but our focus was on exploring whether there are important metabolic factors which underpin the development of the disorder. My remit was to investigate the fundamentals of the regulation of energy balance using animal models. The animal of choice was the genetically obese ob/ob (Lep^ob/Lep^ob) mouse, and a colony of the Aston strain of these mutants was set-up. The attraction of ob/ob mice, which at the time were the most widely used animal model in obesity research, was that not only is the obese state extreme with body weight being up to three times that of lean siblings but that it is reducible to a mutation in a single recessively inherited gene^{(Reference Trayhurn4)}. The link to a mutant gene meant that the obesity of ob/ob mice results from a change in just one protein, and that protein must play a critical role in the regulation of energy balance.

The ob/ob mouse is not, of course, the only rodent in which obesity is the result of a single gene mutation, and we subsequently established colonies of the other major obese mutants: the Zucker fa/fa (Lepr^fa/Lepr^fa) rat and the diabetic-obese db/db (Lepr^db/Lepr^db) mouse^{(Reference Trayhurn4)}. We were later able to house colonies of the adipose mouse (Ad), golden hamsters (Mesocricetus auratus) and Djungarian hamsters (Phodopus sungorus), each as a specific model within our energy regulation studies. This was based on the August Krogh Principle that ‘for a large number of problems there will be some animal of choice or a few such animals on which it can be most conveniently studied’^{(Reference Krogh5)}. As I have noted previously, the ability to maintain multiple colonies of experimental animals at the Dunn without direct cost to the investigator was remarkable^{(Reference Trayhurn6)}.

Energy balance and thermogenesis

Hyperphagia is part of the basis for the obesity of the ob/ob mouse, and indeed that of the other obese mutants, food intake being greater than in lean siblings^{(Reference Trayhurn4)}. However, studies where young ob/ob mice were either directly pair-fed to the ad libitum intake of their lean siblings, or otherwise given restricted amounts of food, indicated that obesity still develops without hyperphagia^{(Reference Thurlby and Trayhurn7,Reference Romsos8)} . Our own work, in which full energy balance studies were performed, clearly illustrates the point; young ob/ob mice pair-fed to the ad libitum intake of lean siblings at room temperature (23°C) exhibited a rate of energy deposition 2⋅3 times that of the lean^{(Reference Thurlby and Trayhurn7)}. The study was conducted at four different environmental temperatures: 33 (thermoneutral for the mouse), 28, 23 and 17°C. At each temperature, the energy gain of the obese animals was greater than the lean, but the lower the temperature the higher the excess gain^{(Reference Thurlby and Trayhurn7)}.

The capacity for excess energy deposition in the absence of hyperphagia indicated that one or more components of energy expenditure is reduced in the ob/ob mutants. Of the main components of expenditure, facultative (or adaptive) non-shivering thermogenesis (NST) was particularly attractive as the key element. Not only is thermoregulatory thermogenesis a major part of total expenditure in small mammals in order to maintain body temperature, but reduced expenditure on thermogenesis was also being advocated by Miller and Stock as a causal factor in the development of obesity^{(Reference Miller and Mumford9,Reference Miller, Mumford and Stock10)} . Furthermore, some 25 years earlier impaired homoeothermy had been noted in ob/ob mice^{(Reference Davis and Mayer11)}. This appeared counter-intuitive given the improved insulation provided by the additional body fat of the obese animals, and was suggestive of a reduced capacity to generate heat. In our own studies, core temperature fell rapidly, to as low as 15°C, just 3 h after exposure of ob/ob mice to 4°C, whereas lean siblings maintained their temperature above 35°C^{(Reference Trayhurn and James12)}.

Direct measurements of NST from the peak increase in the RMR at thermoneutrality following the administration of noradrenaline indicated that the capacity for this form of heat production was 2-fold lower in ob/ob mice than that in lean siblings^{(Reference Trayhurn and James12)}. Furthermore, RMR expressed ‘per animal’, as should be done in energetic studies, was reduced in the obese mice relative to the lean at every temperature examined below thermoneutrality, indicating a lower expenditure on NST^{(Reference Trayhurn and James12)}.

Brown adipose tissue

An immediate question raised by these physiological studies was the nature of the molecular and cellular mechanisms of NST. Several possibilities were under consideration at the time, including protein turnover, the α-glycerophosphate shuttle, Na⁺ transport across the plasma membrane mediated by Na⁺-K⁺ATPase, and futile/substrate cycles such as that between fructose-6-phosphate and fructose-1,6-bisphosphate^{(Reference Stirling and Stock13–Reference Trayhurn, Goodbody and James19)}; in practice, most are in effect a form of energy-consuming substrate cycle. There were, however, substantial reservations with each of these mechanisms^{(Reference Trayhurn20)} in that it seemed unlikely that they had the potential to generate sufficient quantities of heat and to do so acutely without disrupting normal metabolic control. In addition, there was a central issue of tissue localisation; several of the mechanisms, particularly protein turnover and Na⁺ transport, being essentially universal rather than restricted to a specific tissue site.

The question of the tissue basis for NST subsequently centred on brown adipose tissue (BAT, or brown fat), which had first been described by Conrad Gessner in 1551. Although different roles had been proposed for this tissue, including as an endocrine organ, the principal function was resolved in the early 1960s, as a thermogenic organ with heat as the primary product^{(Reference Smith and Horwitz21)}. The tissue is prominent in hibernating species, in the newborn of many mammals (including human subjects) and in rodents acclimated to the cold^{(Reference Nicholls and Locke22,Reference Cannon and Nedergaard23)} . The quantitative importance of BAT in adaptive NST in rodents was demonstrated in influential studies by Foster and Frydman^{(Reference Foster and Frydman24–Reference Foster and Frydman26)}. These authors mapped regional blood flow to different tissues using radioactively-labelled microspheres in rats in which NST was maximally stimulated following either cold-acclimation or the administration of noradrenaline. From the measurements of regional blood flow, together with the cardiac output and the oxygen extraction across the interscapular depot, BAT was estimated to account for 60 % of NST in cold-acclimated rats^{(Reference Foster and Frydman25)}. Our own studies on mice using the same approach suggested a broadly similar figure^{(Reference Thurlby and Trayhurn7)}.

In parallel with the identification of BAT as the principal locus for NST, the unique bioenergetic properties of the tissue's mitochondria were being elucidated. Heat was shown by Nicholls to be generated by a regulated uncoupling of oxidative phosphorylation, the energy inherent in the proton gradient across the inner mitochondrial membrane being dissipated as heat rather than coupled to ATP synthesis^{(Reference Nicholls and Locke22)}. This process is controlled by the 32 000-M_r mitochondrial uncoupling protein-1 (UCP1) discovered by Ricquier^{(Reference Nicholls and Locke22,Reference Ricquier27)} . Acute stimulation of BAT thermogenesis leads to an activation of UCP1, while chronic stimulation results in an increase in the amount of the protein, through a combination of a higher concentration in the mitochondria and through mitochondriogenesis^{(Reference Cannon and Nedergaard23)}. These changes, both acute and chronic, are primarily driven by the release of noradrenaline from the extensive sympathetic innervation of BAT, acting mainly via β₃-adrenoceptors^{(Reference Himms-Hagen28,Reference Arch29)} .

My group at the Dunn, in parallel with several other groups, began to explore the potential role of BAT in energy balance and the development of obesity. Two seminal observations were pivotal; in the first, Himms-Hagen and Desaultels in Ottawa demonstrated reduced GDP binding to BAT mitochondria in ob/ob mice relative to lean siblings, this reflecting a reduction in the thermogenic proton conductance pathway of the tissue^{(Reference Himms-Hagen and Desautels30)}. Our own blood flow studies indicated that the reduced NST and consequent lower energy expenditure of the obese mutant is entirely due to decreased metabolic activity in BAT^{(Reference Thurlby and Trayhurn31)}. In the second key study, Rothwell and Stock in London proposed that BAT is the locus of the diet-induced thermogenesis that they were observing in rats overfed through the provision of a cafeteria diet^{(Reference Rothwell and Stock32)}. In follow-up studies with our group, key molecular indices of BAT thermogenic activity were demonstrated in the cafeteria-fed animals; increased mitochondrial mass and GDP binding, as well as GDP-sensitive respiration^{(Reference Brooks, Rothwell and Stock33)}.

These initial reports were followed by a series of studies in which the thermogenic activity of BAT was shown to be reduced in a variety of obese rodents. They included other single gene mutants, fa/fa rat and db/db mouse, and rodents with experimentally-induced obesity such as that following lesioning of the ventromedial hypothalamus, the administration of gold thioglucose, and treatment with corticosteroids^{(Reference Trayhurn20,Reference Himms-Hagen34,Reference Trayhurn, Trayhurn and Nicholls35)} . Along with the studies on obese animals, the role of BAT in nutritional energetics was further explored in a range of physiological and pathophysiological situations in which body fat and energy flux change (Fig. 2). These included the reproductive cycle (pregnancy and lactation) hibernation, photoperiod, cancer cachexia and nutritional manipulations such as fasting/refeeding and the provision of a low protein diet^{(Reference Himms-Hagen34,Reference Trayhurn, Trayhurn and Nicholls35)} .

Fig. 2. (Colour online) Schematic of different physiological and pathological conditions in experimental animals in which energy flux and/or balance are altered where increases, or decreases, in brown adipose tissue thermogenesis have been demonstrated. Examples of key situations in which brown fat thermogenesis changes are shown. DIT, diet-induced thermogenesis; VMH, ventromedial hypothalamus.

Lactation was a physiological stress of particular interest to us at the Dunn. The energy cost of lactation is high in small mammals, and energy intake is increased approximately 3-fold in lactating rats, for example, compared to virgin animals^{(Reference Fell, Smith and Campbell36,Reference Williamson37)} . Our studies in mice showed that BAT thermogenesis is suppressed in lactation, the suppression being maximal in late lactation when milk production peaks^{(Reference Trayhurn, Douglas and McGuckin38–Reference Krol and Speakman40)}. Mitochondrial mass and GDP binding are both markedly reduced in BAT of lactating mice, the latter to the same level as virgin animals at thermoneutrality^{(Reference Trayhurn, Douglas and McGuckin38,Reference Trayhurn and Jennings39)} . The concentration of UCP1 in the mitochondria is also reduced relative to that of virgin mice, with the total UCP1 content of the interscapular pad at late lactation being <10 % of the virgin animals (Fig. 3)^{(Reference Trayhurn and Jennings39)}. These changes in BAT activity effectively lead to a substantial energy saving, helping to meet the high energy cost of milk production. However, this adaptation essentially reflects the limited scope for heat dissipation in the face of the high metabolic heat generation associated with milk synthesis rather than a specific energy saving mechanism as such, as convincingly argued in a recent review on BAT in lactation^{(Reference Krol and Speakman40)}.

Fig. 3. (Colour online) The thermogenic activity and capacity of brown adipose tissue (BAT) is decreased in lactation (Lact). The changes in mitochondrial GDP binding, the mitochondrial concentration of uncoupling protein-1 (UCP1; UCP1 conc) and the total UCP1 content of the interscapular BAT depot are shown for mice at late lactation (when milk production is close to maximal) relative to virgin mice (virgin = 1)^{(Reference Trayhurn and Jennings39)}.

Perhaps the strongest illustration of the link between energy expenditure and BAT thermogenesis comes from the changes in the tissue that occur when small rodents are acclimated to the cold. In mice, energy expenditure and food intake are increased 3-fold between thermoneutrality and 4°C, reflecting the energy cost of generating heat for homoeothermy^{(Reference Trayhurn and James12,Reference Trayhurn41)} . In rats acclimated at 4°C, the mitochondrial content, mitochondrial GDP binding and UCP1 concentration were each substantially higher than in rats acclimated to thermoneutrality (29°C), while the total UCP1 content of the interscapular BAT depot was increased >100-fold^{(Reference Trayhurn, Ashwell and Jennings42)}.

Brown adipose tissue in human subjects

By the beginning of the 1990s, the importance of BAT in nutritional energetics had been firmly established across a range of obesity models and under other conditions in experimental animals in which energy flux is altered. In the case of human subjects, interest in the tissue had been driven to a considerable extent by the concept that reduced thermogenesis is a key factor in the development of obesity in human subjects and that BAT is a potential therapeutic target for the treatment of the disorder.

Although brown fat was widely recognised to be an important locus of heat production in the human neonate, the tissue appeared, on the basis of histological appearance, to be absent after the first few years of life. The presence of BAT in adult human subjects was confirmed, however, by immunological studies identifying UCP1 in fat depots, including in some elderly subjects^{(Reference Bouillaud, Combes and Ricquier43–Reference Lean45)}. In addition, expression of the UCP1 gene was evident through detection of the encoded mRNA^{(Reference Bouillaud, Villarroya and Hentz46)}. Activation of the tissue in patients with phaeochromocytoma was also demonstrated^{(Reference Ricquier, Néchad and Mory47,Reference Lean, James and Jennings48)} .

Despite the clear evidence for the presence of BAT in adults, with the capacity for adaptive changes, the prevailing view was that the tissue was of little, or no, significance in human energetics other than in neonates and during the first years of life. Interest in BAT then declined markedly, with the notable exception of those groups (particularly that of Cannon and Nedergaard in Stockholm^{(Reference Cannon and Nedergaard23)}) whose principal focus was on understanding the fundamental biology of the tissue. Since 2009 there has, however, been renewed interest in BAT in human subjects following the application of fluorodeoxyglucose positron emission tomography^{(Reference Cypess, Lehman and Williams49–Reference van Marken Lichtenbelt, Vanhommerig and Smulders51)}. This has firmly demonstrated active BAT in adults, activity being reduced in obesity and with ageing, for example, while being stimulated on cold exposure and by the administration of a selective β₃-adrenoceptor agonist^{(Reference Cypess, Lehman and Williams49–Reference Moonen, Nascimento and van Marken Lichtenbelt55)}.

White adipose tissue: the discovery of leptin

As interest in BAT declined, my own research focus changed and abruptly so following the cloning of the Ob (Lep^ob) gene and the identification of the encoded protein^{(Reference Zhang, Proenca and Maffei56)}. Within days of the report in Nature on 1 December 1994, my group at the Rowett in Aberdeen (where I had relocated in 1988) had designed and validated oligonucleotide probes to examine Ob gene expression. This move reflected the fact that some 2 years earlier a consortium of us in the UK, which included Michael Stock and John Stirling, had sought funding to identify the defective genes in the obese mouse mutants. We were, however, unsuccessful since it was argued (correctly) that at least one group in the United States was well-advanced in the goal and that it was unlikely that we could be competitive. My response was that once the Ob gene had been identified, our strategy would be to explore the physiology of the protein product.

The Ob gene was reported to be expressed in white adipose tissue (WAT) and the protein, initially termed ‘OB’, and then leptin, to act as a lipostatic signal^{(Reference Zhang, Proenca and Maffei56–Reference Friedman58)}. Subsequently, the hormone was found to be produced by several tissues, including BAT^{(Reference Dessolin, Schalling and Champigny59)} and the placenta^{(Reference Hoggard, Hunter and Duncan60)}, although WAT is the major source. Similarly, the functions attributed to leptin quickly expanded and it became regarded as a pleiotropic factor^{(Reference Trayhurn61)}. The early studies of my group at the Rowett demonstrated that expression of the Ob gene is nutritionally regulated, the mRNA level in WAT of lean rodents rapidly decreasing on fasting with a restoration on refeeding^{(Reference Trayhurn, Thomas and Duncan62)}. The circulating levels of the hormone change in parallel with the alterations in gene expression^{(Reference Hardie, Rayner and Holmes63)}.

We then showed that acute exposure of mice to cold led to a strong inhibition of Ob expression, and a fall in the circulating leptin level, both of which are rapidly reversed on return to a warm environment^{(Reference Trayhurn, Duncan and Rayner64,Reference Trayhurn, Duncan and Rayner65)} . The cold-induced reduction in the Ob mRNA level was mimicked by the administration of noradrenaline and by the β-adrenoceptor agonist isoprenaline. From these observations we proposed that the sympathetic system plays a key role in the regulation of Ob gene expression^{(Reference Trayhurn, Duncan and Rayner64)}. Subsequent observations indicated that this operates primarily through β₃-adrenoceptors^{(Reference Trayhurn, Duncan and Rayner65,Reference Mantzoros, Qu and Frederich66)} . Further studies on leptin at the Rowett included the demonstration by in situ hybridisation that the receptor, and particularly the long form responsible for signalling, is strongly expressed in the regions of the hypothalamus, consistent with being an adipocyte-derived signal for appetite^{(Reference Mercer, Hoggard and Williams67)}.

Adipokines and the secretory function of white adipocytes

Leptin quickly became a major area in research on obesity and its associated disorders. One of the key outcomes of the discovery of the hormone was a radical change in perspective on the functions of white adipocytes and therefore of WAT itself. Adipocytes were recognised as endocrine cells with WAT as a major signalling organ^{(Reference Frühbeck, Gómez-Ambrosi and Muruzabal68–Reference Trayhurn71)}. Although secreted protein factors had been identified previously, this had not led to the conceptualisation of white adipocytes as endocrine and signalling cells. The secreted proteins known prior to leptin were adipsin (complement factor D)^{(Reference Cook, Min and Johnson72)}, which is a serine protease, the cytokine TNFα^{(Reference Hotamisligil, Shargill and Spiegelman73)}, and lipoprotein lipase. Lipoprotein lipase is, of course, released from adipocytes to catalyse the breakdown of circulating TAG to enable the uptake of fatty acids into adipocytes; it was not, however, regarded as a fat cell secretory protein as such.

The secretome of adipocytes, and of WAT as a whole, is extensive (Fig. 4). Quantitatively, fatty acids are the largest secretory product, but there are several other lipid groups released from the cells. Some, such as specific prostaglandins and the endocannabinoid anandamide, are synthesised de novo within adipocytes, while others, including cholesterol and vitamin A, are taken up, stored and subsequently released^{(Reference Trayhurn61)}. A question raised by the discovery of leptin was whether there are a range of protein hormones and signals synthesised and secreted by fat cells. The answer is very much in the affirmative and one of the earliest of these adipokines, as they are termed, identified was another major adipocyte hormone, adiponectin, whose functions encompass insulin sensitising, angiogenic and anti-inflammatory actions (Fig. 3)^{(Reference Ouchi, Kihara and Arita74–Reference Brakenhielm, Veitonmaki and Cao78)}.

Fig. 4. (Colour online) The secretome of white adipocytes. Fatty acids and other lipids are secreted, together with a multiplicity of adipokines (proteins); examples of some of the lipids and key adipokines are shown. The major adipocyte hormones, leptin and adiponectin, are highlighted. angptl4, angiopoietin-like protein-4; CETP, cholesteryl ester transfer protein; DPP4, dipeptidyl peptidase-4; IGF, insulin-like growth factor-1; LPL, lipoprotein lipase; MCP-1, monocyte chemoattractant protein-1; MIC-1, macrophage inhibitory cytokine-1; MIF, macrophage migration inhibitory factor; MMP, matrix metalloproteinase; NGF, nerve growth factor; PAI-1, plasminogen activator inhibitor-1; RBP4, retinol binding protein-4; TGFβ, transforming growth factor-β; VEGF, vascular endothelial growth factor; ZAG, zinc-α₂-glycoprotein.

The search for novel adipokines became a core focus of my group, both at the Rowett and later at the University of Liverpool to where I moved in 2002. Among the several adipokines that we discovered were: the neurotrophic signal nerve growth factor^{(Reference Peeraully, Jenkins and Trayhurn79)}; specific metallothioneins^{(Reference Trayhurn, Duncan and Wood80)}, these having metal binding actions, and the lipolytic/cachectic factor zinc-α₂-glycoprotein^{(Reference Bing, Bao and Jenkins81,Reference Bao, Bing and Hunter82)} . Nerve growth factor was found to be linked to the inflammatory response in WAT, secretion of the protein being strongly stimulated by TNFα^{(Reference Peeraully, Jenkins and Trayhurn79)}. As in mice, zinc-α₂-glycoprotein expression increased substantially in WAT of patients with cachexia associated with gastrointestinal cancer^{(Reference Mracek, Stephens and Gao83)}.

An extensive range of individual adipokines has now been identified, and proteomic studies and in silico analysis suggest that there are several hundred in total^{(Reference Lehr, Hartwig and Lamers84–Reference Deshmukh, Peijs and Beaudry86)}. The wide-ranging secretory function of white adipocytes established over the past two decades has in part served as a model for other cell types which were not previously regarded as having a significant endocrine or signalling function. Myocytes, for example, are now known to release a range of protein signals, myokines^{(Reference Pedersen and Febbraio87,Reference Lee and Jun88)} , while another example is hepatocytes which secrete multiple hepatokines^{(Reference Stefan and Haring89)}.

The identification of a multiplicity of protein signals and factors from adipocytes indicated that WAT is involved in a range of physiological and regulatory processes^{(Reference Trayhurn61,Reference Rajala and Scherer70,Reference Trayhurn and Wood90–Reference Scherer92)} . While some adipokines are endocrine in function, signalling to tissues and organs distant to the adipose depots, others have local paracrine and/or autocrine actions. The processes in which various adipokines play a role include appetite and energy balance, lipid metabolism, vascular haemostasis, blood pressure, angiogenesis and insulin sensitivity (see^{(Reference Trayhurn61,Reference Trayhurn91)} ). A number of adipokines are linked to immunity and inflammation, these including classical cytokines and chemokines such as IL-1β, IL-6, IL-10 and monocyte chemoattractant protein-1; they also include inflammation-related factors, examples being vascular endothelial growth factor (VEGF), serum amyloid A and adiponectin (see^{(Reference Trayhurn61,Reference Rajala and Scherer70,Reference Trayhurn91} ).

In obesity, WAT exhibits chronic mild inflammation with increased production and release of inflammatory adipokines. There is a notable exception to this in that the synthesis and release of adiponectin, with its anti-inflammatory action, falls^{(Reference Arita, Kihara and Ouchi93,Reference Hotta, Funahashi and Arita94)} . Inflammation in expanded WAT is augmented by the infiltration and activation of macrophages in particular, but also of other immune cells^{(Reference Scherer92,Reference Weisberg, McCann and Desai95–Reference Mraz and Haluzik97)} .

Hypoxia and the metabolic response to oxygen deprivation in adipocytes

Inflammation in WAT has been considered a key factor in the development of the major obesity-associated disorders, particularly insulin resistance and the other components of the metabolic syndrome^{(Reference Blüher3,Reference Rajala and Scherer70,Reference Yudkin98,Reference Hotamisligil99)} . The question that intrigued me in the early 2000s was why does inflammation develop as adipose tissue mass expands? A ‘News’ article in Science on how cells endure low oxygen^{(Reference Marx100)} encouraged me to consider the possibility that hypoxia might be a key. This was presented as a hypothesis in a Horizons article in the British Journal of Nutrition in 2004^{(Reference Trayhurn and Wood90)}. I am particularly proud of this paper: not only does it describe the hypoxia hypothesis, but it is my most highly cited publication (>1400 citations in the Web of Science; >2450 citations in Google Scholar) as well as being the fourth most highly cited article in the Nutrition Society's flagship journal (or indeed in all of its journals).

The hypothesis proposed that as adipose tissue mass expands with the development of obesity, areas within the tissue become relatively hypoxic as the enlarging adipocytes become more distant from the vasculature, this leading to major adaptive changes involving the hypoxia-inducible transcription factor-1 (HIF-1). The recruitment of HIF-1 was hypothesised to lead to increased expression of a series of hypoxia-sensitive genes linked to inflammation and the inflammatory response in WAT. The proposition was based on the following: (i) hypoxia occurs in situations such as ischaemic injury, wound healing and solid tumours leading to extensive metabolic changes^{(Reference Trayhurn and Wood90)}, (ii) blood flow to WAT is not increased in obese subjects, despite the higher mass of the tissue^{(Reference Blaak, van Baak and Kemerink101–Reference Frayn and Karpe104)}, (iii) in contrast to lean subjects, blood flow to WAT does not increase post-prandially in the obese^{(Reference Frayn and Karpe104–Reference Goossens, Bizzarri and Venteclef106)}, (iv) large adipocytes (which may be up to 200 μm diameter^{(Reference Skurk, Alberti-Huber and Herder107)}) are further from the vasculature than the normal diffusion distance for O₂ (100 μm)^{(Reference Brahimi-Horn and Pouysségur108)}. These observations refer to local hypoxia, but the provision of O₂ on a whole-body level is reduced in specific environmental and pathological situations, such as high altitude, deep sea dives, lung diseases and obstructive sleep apnoea^{(Reference Trayhurn61,Reference Brahimi-Horn and Pouysségur108)} .

In 2007, studies using two separate techniques reported that WAT depots in different types of obese mice are hypoxic, with the O₂ tension being 2- to 3-fold lower than in lean mice^{(Reference Hosogai, Fukuhara and Oshima109,Reference Ye, Gao and Yin110)} . Subsequent studies on mice were consistent with these observations^{(Reference Rausch, Weisberg and Vardhana111)}. In contrast, although some human studies have indicated that WAT depots are relatively hypoxic^{(Reference Kabon, Nagele and Reddy103,Reference Pasarica, Sereda and Redman112)} others have reported either the same or an increase in pO₂ (partial pressure of oxygen)^{(Reference Goossens, Bizzarri and Venteclef106,Reference Hodson, Humphreys and Karpe113)} . The issue remains unresolved, but there is evidence that differences in the way in which O₂ is delivered in terms of vascularisation and utilisation may occur^{(Reference Goossens, Bizzarri and Venteclef106,Reference Hodson, Humphreys and Karpe113,Reference Lempesis, van Meijel and Manolopoulos114)} .

From 2004 the focus of my group in Liverpool was on examining the direct effects of hypoxia on gene expression and cellular function in adipocytes. Almost all of our studies were conducted on human adipocytes, differentiated in culture from fibroblastic preadipocytes. The initial priority was to examine whether incubation under a low O₂ tension leads to increased expression and release of inflammation-related adipokines consistent with our initial hypothesis. A candidate gene approach was employed and increased production of several adipokines was observed, including IL-6, VEGF and leptin^{(Reference Wang, Wood and Trayhurn115)}. Raised production of VEGF and leptin, as well as of specific matrix metalloproteinases, had been reported earlier in 3T3-F442A adipocytes (a mouse cell line), reflecting a pro-angiogenic response to hypoxia^{(Reference Lolmède, Durand de Saint Front and Galitzky116)}.

The key cellular adaptation to O₂ deficiency is a switch from aerobic to anaerobic metabolism. Mitochondrial oxidative phosphorylation cannot, of course, continue when O₂ is severely limited, and there is instead increased anaerobic glycolysis. As expected, adipocytes exhibit greater glucose uptake under hypoxic conditions, as demonstrated by 2-deoxy-d-glucose uptake studies^{(Reference Wood, Wang and Lorente-Cebrián117)} and by measurement of glucose in the culture medium^{(Reference Lolmède, Durand de Saint Front and Galitzky116)}. This is mediated through increased synthesis of the GLUT1 facilitative glucose transporter, driven by a marked stimulation of GLUT1 gene expression^{(Reference Wood, Wang and Lorente-Cebrián117)}. The expression of several genes encoding glycolytic enzymes is also raised, glucose-6-phosphate isomerase and phosphofructokinase, for example^{(Reference Geiger, Leiherer and Muendlein118,Reference Mazzatti, Lim and O'Hara119)} . Lactate release is augmented in hypoxic adipocytes^{(Reference Lolmède, Durand de Saint Front and Galitzky116,Reference Pérez de Heredia and Wood120)} , reflecting the increased glucose utilisation and glycolytic flux, this being mediated by increases in the synthesis of the monocarboxylate transporter, MCT1^{(Reference Pérez de Heredia and Wood120)}.

While our initial exploration of the effects of hypoxia on gene expression in human adipocytes probed selective candidate genes, in subsequent studies more comprehensive approaches were taken. In the first of these, PCR arrays for eighty-four genes linked to the hypoxia-signalling pathway were employed. The expression of a number of the genes changed, with one particular gene exhibiting dramatically increased expression^{(Reference Wang, Wood and Trayhurn121)}. The gene in question was MT3, which encodes a member of the metallothionein family, metallothionein-3 (also known as growth inhibitory factor). This protein binds zinc and copper, and linked to its marked induction by O₂-deprivation has been implicated as an angiogenic factor and to protect against hypoxic damage^{(Reference Penkowa, Carrasco and Giralt122,Reference Zbinden, Wang and Adenika123)} .

PCR arrays are themselves limited in terms of the number of genes whose expression can be screened and a specific pathway or metabolic system needs to be selected. DNA microarrays offer an unbiased approach in which all, or almost all, the genes expressed in a tissue or cell can be probed simultaneously. Our microarray studies at Liverpool, in collaboration with colleagues at Unilever, indicated that the expression of >1300 genes was altered in human adipocytes cultured under hypoxic conditions, stringent criteria being used to evaluate changes^{(Reference Mazzatti, Lim and O'Hara119)}. Of these genes, the expressions of approximately half were up-regulated and half down-regulated under low pO₂. Bioinformatic analysis showed that a number of metabolic pathways and functions are altered in human adipocytes by hypoxia, these including lipolysis, lipid oxidation, glucose utilisation, cell to cell signalling and cell death^{(Reference Mazzatti, Lim and O'Hara119)}.

It is evident from these and other studies that hypoxia results in extensive changes in gene expression in adipocytes. Several important functional changes have been described, in addition to increased anaerobic glycolysis (Fig. 5). These include the rapid induction of insulin resistance through the direct inhibition of insulin signalling^{(Reference Regazzetti, Peraldi and Gremeaux124,Reference Yin, Gao and He125)} , and the disruption of the extracellular matrix within WAT that characterises fibrosis^{(Reference Halberg, Khan and Trujillo126,Reference Sun, Tordjman and Clément127)} . With respect to fibrosis, hypoxia leads to changes in the expression of collagens released as components of the extracellular matrix, as well as matrix metalloproteinases^{(Reference Lolmède, Durand de Saint Front and Galitzky116,Reference Halberg, Khan and Trujillo126)} involved in tissue remodelling. The overall cellular response to low O₂ is regulated by a series of hypoxia-responsive transcription factors of which HIF-1, consisting of two subunits (HIF-1α and HIF-1β), is the best characterised^{(Reference Brahimi-Horn and Pouysségur108,Reference Semenza128,Reference Coleman and Ratcliffe129)} .

Fig. 5. Schematic representation of the central cellular responses to hypoxia in white adipocytes. The effect of low partial pressure of O₂ on gene expression, glucose uptake and utilisation, and the production of selected key adipokines is shown. angptl4, angiopoietin-like protein-4; FA, fatty acid; GLUT1, facilitative glucose transporter 1; HIF-1, hypoxia-inducible factor-1; MCT1, monocarboxylate transporter-1; MIF, macrophage migration inhibitory factor; MMP, matrix metalloproteinase; MT-3, metallothionein-3; PAI-1, plasminogen activator inhibitor-1; TF, transcription factors (other than HIF-1); VEGF, vascular endothelial growth factor. Modified from^{(Reference Trayhurn131)}.

Our studies, like most that examine the response of cells to hypoxia, were undertaken by comparing 1 to 20 % O₂ (95 % air/5 % CO₂; ‘normoxia’). However, this is an extreme, effectively representing a comparison between ambient air (higher than arterial pO₂) and marked hypoxia. A question that interested me was whether there is a critical point at which adaptation to reduced O₂ is initiated in adipocytes, or if there is a gradual response to falling O₂ tension. Experiments in which human adipocytes were incubated with a range of O₂ levels between 20 and 1 % clearly demonstrated a dose-dependent response to lowering O₂ tension with differences being observed between 21 and 15 % and between 15 and 10 % O₂^{(Reference Wood, Stezhka and Trayhurn130)}. This was true for the expression and secretion of several adipokines, including leptin and VEGF, as well as 2-deoxy-d-glucose uptake and GLUT1 gene expression. Nevertheless, changes tended to be more marked between 10 and 5 % O₂. Since the pO₂ in WAT of lean mice is equivalent to about 7 % O₂ while in obese mice it is about 2 %, it is evident that there are responses to O₂-deprivation over physiologically relevant differences in tissue oxygenation between the phenotypes^{(Reference Trayhurn91)}.

These experiments on the effects of a range of O₂ levels demonstrate that while the customarily employed protocols in hypoxia studies offer proof of principle, they lead to an exaggerated view of the scale of the cellular response to relative O₂ lack under normal physiological conditions. This raises a question of the extent to which our understanding of cellular processes has been conditioned, or even distorted, by the routine use of 20 % O₂ as the gas phase in cell culture and other in vitro experiments. It is intriguing that careful attention is paid to the pH (7⋅4), temperature (37°C) and the concentration of glucose and other nutrients in cell culture to ensure physiological conditions' (except for when they are the parameters under investigation), but the O₂ tension employed is quite unphysiological and indeed reflects overt hyperoxia.

Oxygen: an overlooked macronutrient

A corollary of our studies on hypoxia is that they underscore that O₂ is a key nutrient at the cellular level. Indeed, investigation of hypoxia is in effect exploration of the molecular and metabolic consequences of the deficiency of a nutrient. However, O₂ is not considered as a nutrient as such in the context of nutritional science. Textbooks of nutrition do not contain sections on O₂, and reference to it is generally restricted to discussion of metabolic rate and respiratory quotient. I have argued recently that O₂ should be included alongside the other elements/molecules/macromolecules that are defined as nutrients^{(Reference Trayhurn131,Reference Trayhurn132)} .

O₂ undoubtedly meets dictionary definitions of a nutrient; for example, ‘as a substance that provides nourishment for the maintenance of life and for growth’ (Oxford English Dictionary). The central reason why O₂ is not considered as part of nutritional science is because of the route of entry; the nose/lungs in higher terrestrial animals, rather than the mouth/gastrointestinal tract. However, I argue that the route of entry should not be the critical determinant of whether O₂ is, or is not, considered a nutrient, but rather its function and essentiality^{(Reference Trayhurn132)}. O₂ is, of course, critical to all aerobic species without which mitochondrial oxidative phosphorylation cannot take place.

Early organisms developed under anoxic conditions, the level of O₂ in the atmosphere being just one part in a million soon after the Earth was formed some 4⋅54 billion years ago^{(Reference Kerr133,Reference Lyons, Reinhard and Planavsky134)} . It was only after the initiation of the ‘Great Oxidation Event’ some 2⋅45 billion years ago that considerable amounts of O₂ began to appear in the atmosphere^{(Reference Lyons, Reinhard and Planavsky134–Reference Gumsley, Chamberlain and Bleeker136)}, the present level of 21 % being essentially stable over the past 600 million years^{(Reference Lyons, Reinhard and Planavsky134)}. The availability of O₂ in abundance in the atmosphere was critical to the evolution of life as we know it.