Translational potential of mouse models of human metabolic disease

SUMMARY


INTRODUCTION
Obesity is defined as an increase in fat mass (adiposity) that is sufficient to adversely affect health.Obesity is a major driver of type 2 diabetes, hypertension, and cardiovascular disease. 1,2lthough obesity is well-known to increase the risk of colorectal, esophageal, endometrial, breast, and liver cancer, recent studies have shown that increased adiposity increases the risk of many more cancer types. 3Although globally, the rise in prevalence of obesity in children and adults is driven by changes in environmental factors, there is considerable heterogeneity within and among populations. 4For example, some individuals are more susceptible to developing obesity, while others remain slim in highly obesogenic environments.To reduce the impact of obesity on health, economies, and society, there is a pressing need to understand how environmental and biological factors contribute to weight regulation and how their disruption causes obesity.
Following pioneering parabiosis experiments in severely obese inbred strains of mice (ob/ob and db/db) and in rats, 5,6 a breakthrough in our understanding of the fundamental molecular mechanisms that regulate body weight came in 1994, with the positional cloning of the ob gene by Friedman and colleagues. 7The discovery that severely obese ob/ob mice were lacking a circulating hormone derived from adipose tissue (leptin), and the reversal of severe obesity in these mice following leptin administration, [8][9][10] provided proof of principle for leptin's critical role in the regulation of food intake and body weight.Subsequent studies in humans with congenital leptin deficiency established the relevance of these findings to human physiology and eating behavior. 11,12Ahima and Flier showed that starvation is associated with a fall in leptin levels, which triggers increased food intake and adaptive responses, which form part of the defense against starvation. 138][19][20] Our current understanding of the mechanisms that regulate energy homeostasis has been shaped by these discoveries and others, which collectively have the potential to inform interventions to improve human health and tackle metabolic disease.The aim of this perspective is to discuss the limitations and benefits of studies in mice and humans when investigating mechanisms involved in energy homeostasis.In this perspective, we will not highlight all areas nor attempt an exhaustive review of each area.Instead, we will offer our view as to how studies in mice and humans can inform the study of metabolic disease and discuss current and future challenges and opportunities.

ENERGY HOMEOSTASIS AND SET-POINTS
Fundamentally, obesity arises when an imbalance between energy intake and energy expenditure results in sustained positive energy balance over time.The concept of energy balance or homeostasis was first described by Kennedy who showed that rats respond to short-term caloric restriction by rapidly increasing food intake to restore their weight trajectory, allowing him to propose that circulating signals inform the brain regarding the amount of body fuel stored in the form of fat. 21These seminal studies led to the concept of the set-point, whereby in response to changes in energy balance, the brain initiates compensatory adjustments to energy intake and expenditure to maintain a given body weight.
The first evidence for the role of the hypothalamus in maintaining energy homeostasis came from Brobeck and Frohlich who reported children with tumors involving hypothalamo-pituitary structures associated with hyperphagia (increased drive to eat) and obesity.Chemical and electrolytic lesioning experiments in rodents and cats in the 1930s and 1940s established that specific clusters of neurons (nuclei) within the hypothalamus play a critical role in the regulation of energy homeostasis. 22,23ypothalamic neuronal circuits are also responsible for the homeostatic regulation of body temperature, electrolyte balance, sleep-wake cycle, and circadian rhythms (Figure 1).This homeostatic function consists of several key components: (1) the detection and integration of internal and external sensory inputs; (2) the comparison of those inputs to physiological setpoints; and (3) the initiation of adaptive responses to maintain these set-points by modulating autonomic, neuroendocrine, and behavioral outputs.Lean and obese animals have different set-points but still respond to perturbations of positive and negative energy balance. 24umans maintain their body weight over time despite day-today variation in calories consumed and energy expenditure.Physiological studies examining the adaptive responses of normal-weight humans to changes in energy balance support the concept of a system that maintains energy homeostasis in humans, and the structural components of this system appear to be highly conserved across species. 26Weight loss induced by caloric restriction, for example, results in both an increased drive to eat and a reduction in energy expenditure.These adaptive responses have been reported in lean and obese individuals, suggesting the physiological defense of a higher level of adiposity, a higher set-point.Insights from overfeeding studies in humans are less clear and discussed more fully by Ravussin, Ferrante, and colleagues. 26,27As elucidated by Speakman and colleagues, rather than a single value for a set-point, a set-point range within which people and animals oscillate depending on their environment (Figure 1) may be a more appropriate construct. 25hy might some people have a higher body weight set-point range than others?It is likely that the multiple genetic influences on body weight act in part by determining a person's set-point range, an assertion that is consistent with the defense of a higher set-point in some inbred strains of mice.The ''thrifty gene hypothesis'' proposes that genetic variants that predispose to a higher set-point may have evolved due to natural selection as an adaptation to promote resistance to starvation during periods of famine. 28,29Alternatively, the risk of predation may have been a key factor defining the upper point of a set-point range in animals.The loss of predation risk in humans during evolution may have removed this restraint, allowing genetic drift because of an absence of selection. 30To date, there is no clear evidence of selection around common genetic loci that predispose to obesity. 31hese studies will need to be further expanded to capture the increasing number of genetic susceptibility loci being identified in larger population-based cohort studies and studies in populations of different ancestries.There is some evidence that a subset of genes in which variants cause severe obesity in mice and humans do offer a survival advantage.For example, heterozygous ob mice exhibit prolonged survival in response to a fast. 32Another example comes from Mexican cavefish (Astyanax mexicanus), which have two forms: a surface form that lives in mid-level waters and a cave form that relies on a sporadic food supply from outside the nutrient-poor environment in underwater caves.Interestingly, cave-dwelling fish carry loss-of-function mutations in the melanocortin 4 receptor (MC4R) gene and exhibit an increase in food (B) Schematic depicting how body weight or adiposity may fluctuate within a set-point range that has upper and lower intervention points (as described 25 ).Challenges such as overeating or undereating at times of stress (grey arrows) may increase or decrease adiposity within this range.The black curved arrow indicates physiological mechanisms that act to maintain/defend an upper limit to this range.intake and starvation resistance, 33 suggesting that they have adapted to the limited nutrient availability in their environment.
In humans, the defense of a set-point/set-point range is supported by the response to weight loss interventions. 34In a series of studies, Leibel and Rosenbaum showed that a reduced body weight (10% of baseline) is associated with increased hunger and food reward and reduced skeletal muscle efficiency due to reduced sympathetic tone and thyroid hormone levels, features that cumulatively drive weight regain. 35,36As the mechanisms that drive weight loss differ from those that drive weight maintenance after weight is lost, it is likely that a range of treatments will be needed for the effective chronic treatment of people with obesity.

GENES AND ENVIRONMENT
Human adiposity is influenced by complex interactions between genetic, social, and environmental influences (Figure 2).Evidence for a substantial genetic contribution to body weight regulation comes from studies of families, twins, and adopted children, which have consistently shown that the heritability (fraction of phenotypic variance of a quantitative trait attributable to genetic variation) of BMI (Body Mass Index; weight in kg/ height in meters squared) is between 40%-70%. 37,38Similarly, body fat distribution is also highly heritable with many of the genes emerging from genome-wide association studies (GWASs) suggesting effects on adipose tissue biology. 39Landmark studies by Bouchard and colleagues showed that weight gain induced by the overfeeding of twins was highly correlated within twin pairs but varied widely among pairs of twins, 40 demonstrating that genetic factors influence the response to changes in energy balance.
GWASs have identified hundreds of common genetic variants associated with increased risk of obesity/higher BMI in populations of differing ethnic ancestries (Figure 2).3][44][45] As many GWAS association signals are linked to variants that are non-coding and of low penetrance, establishing their functional importance and the underlying mechanism of action has been challenging.These challenges are exemplified by studies on the first and most reproducible GWAS locus for BMI, FTO (fat mass and obesity-associated; an a-ketoglutarate-dependent dioxygenase [46][47][48] ).Knockout of Fto in mice results in a complex phenotype with perinatal lethality, stunted growth, and low-fat mass, not obesity. 49Humans with homozygous FTO mutations have cardiac malformations and a complex and severe neurodevelopmental phenotype. 50Studies of other genes in this locus suggested a role for a neighboring gene, RPGRIP1L (Retinitis Pigmentosa GTPase Regulator Interacting Protein 1 Like), 51 in the trafficking of leptin receptors to primary cilia on hypothalamic neurons.More recently, FTO has been implicated in the maintenance of lipid content of mature adipocytes. 52How these divergent findings explain an association between common variants and human obesity remains incompletely understood after 15 years of experimental work.
GWASs in people who are thin but healthy have shown they have a reduced burden of obesity-susceptibility alleles 53 (Figure 2).The discovery of additional genetic variants that may protect against obesity 54,55 or promote thinness has the potential to identify new targets for weight loss therapy.Studies of monogenic obesity syndromes identified in children with severe obesity 12,[17][18][19][20] have shown that disruption of the development, Environmental and social factors influence changes in energy intake and expenditure, which over time can lead to changes in body weight.However, there is variation in response to these factors, which is largely explained by genetic factors.People with obesity have a higher burden of obesity susceptibility alleles, whereas thin people have a much lower burden of these risk alleles, explaining their protection from obesity.function, and maintenance of the hypothalamic leptin-melanocortin pathway involved in weight regulation can cause human obesity. 56These studies have established that changes in appetite and eating behavior (rather than energy expenditure) are the major drivers of obesity in humans. 57 major challenge for assessing the contribution of biological and genetic contributors to obesity is that many humans live in environments that are, or are becoming, highly obesogenic (Figure 2).It is clear that the consumption of cheap, palatable, high-calorie food has a major impact on energy intake at a time when factors at home (television watching, leisure activities) and at work (transport, prolonged desk-sitting 58 ) reduce energy expenditure, particularly in socio-economically disadvantaged groups. 59,60There has been considerable debate around the consumption of sugar, the harmful effects of carbohydrate consumption, and its possible causal role in the development of obesity; the arguments for and against have been summarized elsewhere. 613][64] The contribution of these environmental factors is often challenging to quantify, but more research is needed given the potential to modify these risk factors and reduce the prevalence of obesity and its complications.

INTRAUTERINE ENVIRONMENT, NUTRITIONAL PROGRAMMING, AND EPIGENETIC MODIFICATIONS
Epidemiological studies by Barker and Hales demonstrated that low birth weight was associated with increased risk of developing type 2 diabetes and cardiovascular disease in adulthood. 65,668][69][70] The resulting gene expression changes in organs such as the brain and liver modulate growth, glucose homeostasis, and body weight of the offspring.Some of these nutritional effects are associated with long-term changes in DNA methylation, histone modifications, and changes in microRNA expression. 71In animals, there is evidence that epigenetic marks introduced by nutritional interventions in the peri-and postnatal period can persist into adulthood and can alter the development and function of leptin-responsive hypothalamic feeding circuits. 72When considering how far these findings may translate to humans, an important consideration is that hypothalamic circuits that regulate food intake develop at different times in rodents and humans (postnatally versus final trimester of pregnancy 73 ).
Studies of nutritional programming in humans have been challenging to conduct as trans-generational cohorts are often required, information on maternal diet during pregnancy, and access to the most informative tissues/organs (such as the brain) is often limited. 74Notably, many human imprinting disorders (e.g., Prader Willi syndrome, Beckwith-Wiederman syndrome) are characterized by low birthweight and accelerated forms of metabolic disease.Studies of epigenetic modifications in blood cells have shown evidence of hypomethylation of imprinted genes (Insulin-like Growth Factor 2; IGF2) and metastable epialleles (Pro-opiomelanocortin; POMC) when the mother is undernourished in early gestation. 75Some human studies have suggested associations between methylation marks identified in cord blood or peripheral blood and future risk of obesity. 76,77If changes in methylation are established very early (i.e., metastable epialleles) then methylation status in blood may be reflective of changes in metabolically relevant tissues; however, tissue-specific analyses are likely to be needed to detect later epigenetic modifications.Further studies in countries undergoing nutritional transition (i.e., where the diet of mothers is changing) are needed to test the impact of maternal nutritional state (both under-and overnutrition/obesity) on metabolic health and inform interventions that may reduce the risk of metabolic disease.

ENERGY INTAKE AND EATING BEHAVIOR
The identification of neural and molecular mechanisms that integrate short-term and long-term control of eating will almost certainly enable better preventive and therapeutic approaches to obesity.Our framework for considering the regulation of eating behavior comes from experiments in mice but appears to be highly conserved in humans as demonstrated by the finding of children with severe obesity due to disruption of genes that encode molecular components of this system. 17,19,78The adipocyte-derived hormone leptin is fundamental for survival, and low levels in the fasted state signal to hypothalamic neurons expressing the leptin receptor to drive food intake, effects that are mediated by neurons expressing agouti-related protein (AgRP) (an antagonist of melanocortin signaling 79 ).In the fed state, leptin and other factors stimulate the expression of POMC; POMC-derived melanocortin peptides activate the MC4R on downstream neurons.Genetic disruption of leptin, its receptor, POMC, PCSK1 (Proprotein Convertase Subtilisin/ Kexin type 1, which cleaves POMC), and MC4R all cause human obesity.The neural circuits that mediate the homeostatic regulation of energy balance project to and receive inputs from midbrain, limbic, and cortical areas involved in mood, anxiety, sleep, and arousal. 80,81][84][85] These findings indicate the challenges associated with targeting some of these molecules for weight loss therapy while avoiding on/off target effects on mood, pleasure, and anxiety.
Alongside these homeostatic drives, food intake in mice and humans is affected by the rewarding (hedonic) properties of food 86 with leptin regulating the electrical activity of dopaminergic neurons in the ventral tegmental area. 87The role of the melanocortin pathway in fat and sucrose preference has also been shown to be remarkably aligned in mice and humans, 88 in keeping with the fundamental role of this pathway in the defense against starvation.The contribution of oral nutrient sensing to mouthfeel and preference for fat is an area that needs further study with human neuroimaging studies suggesting that there are distinct midbrain and cortical regions where these sensations are detected. 89][95] Gut-derived neural and hormonal signals, which regulate meal initiation, termination, and meal size in rodents and confer the sensation of satiety and satiation in humans, appear to be highly conserved across species.Meal onset is strongly influenced by both hunger and external cues relating to food availability, smell, appearance, and emotions (Figure 3), whereas meal termination (satiety) appears to be regulated by vagal and hormonal inputs to brainstem neurons, which project to the hypothalamus. 96,97In humans, infusion of peptide YY (PYY) can modulate activation of the brainstem in keeping with its role in mediating satiety. 98he interaction between satiety signals arising from the gut and long-term signals of nutritional state/fat storage underpins the maintenance of energy homeostasis.The mechanisms underlying this integration of signals are incompletely understood.
With regard to obesity risk, the number of calories consumed appears to be more critical than variation in the composition of the diet, although there has been some debate in this area (reviewed here 99 ).When modeling obesity, high-fat diet (HFD)fed mice are referred to as diet-induced obese (DIO) animals.Chow diets typically contain 4%-6% fat, whereas HFDs are commonly 45%-60% fat, and so DIO rodents are also models of high-fat intake. 100Thus, interpretation of such studies must always be mindful of the possibility that the excessive fat intake or composition, rather than the underlying obesity, may be responsible for some of the observed effects.When multiple diets are compared in rodents, dietary fat content (not protein or sucrose) has been shown to promote eating and adiposity. 101Unlike laboratory mice, which eat the same artificially prepared diets day to day, humans consume a range of diets; diets with a mixture of high fat and high sucrose appear to be the most rewarding and promote overconsumption. 89,102hile studies in rodents (see Review from Carmody et al. in Cell 103 ) are generating new hypotheses about the role of the microbiome, their relevance to human physiology and pathophysiology remains unclear. 104Although it is plausible that differences in diet may explain differences in the microbiome and potentially even in nutrient absorption, the link to diseases such as obesity and type 2 diabetes will require challenge experiments/long-term intervention studies in humans.

ENERGY EXPENDITURE AND THERMOGENESIS
In humans, the main contributors to total energy expenditure are basal metabolic rate (BMR) (60%-70%), diet-induced thermogenesis (5%-10%), and physical activity (15%-20%); energy loss also occurs through the feces and urine (<5% 105 ).Their absolute and relative contribution to daily energy expenditure varies throughout the life course. 106lthough brown adipose tissue (BAT) is detectable in some (mostly normal weight) people at room temperature and may contribute to resting energy expenditure, its activation by cold suggests that its dominant physiological role in humans may be in adapting to changes in environmental temperature alongside skeletal muscle contraction (shivering). 107There is considerable interest in whether hormones/cytokines released by BAT may protect against cardiometabolic disease by exerting beneficial effects on insulin secretion, the liver, and other metabolically active tissues and organs. 108MR correlates with the amount of fat-free mass, is influenced by fat mass, age, sex, and ethnic variation in BMR, and is often increased in people with obesity.Additionally, BMR is highly heritable, which may in part explain the variability in this parameter at any given BMI.109 Size-adjusted BMR estimates across cohorts suggest this component of size-dependent energy expenditure may have declined over recent decades.110 Some studies have suggested that people with a relatively low BMR for their body composition are more likely to gain weight over time.111 Physical activity is the most variable component of total energy expenditure and consists of energy expended during sedentary behavior, non-exercise activity thermogenesis (NEAT) during fidgeting, and voluntary exercise.112 Further studies are needed to understand whether variation in levels of NEAT contribute significantly to the variation in body weight in the population.Increasing levels of exercise is beneficial for health and cardiometabolic fitness but often does not lead to weight loss in people with obesity, likely due to compensatory increases in energy intake.113 However, in the weight reduced state, there is evidence that high levels of physical activity (60 min/day) aid weight loss maintenance, possibly due to increased muscle mass and/or efficiency.114 In mice, total energy expenditure is comprised of BMR (about 30%), spontaneous physical activity (up to 40% for a sedentary mouse, increased if a running wheel is provided), and the thermic effects of food (5%-10%).115,116 However, BMR varies significantly with age, body size, temperature, and environment.
Since laboratory mice are usually housed at 20 C-23 C, which is lower than thermoneutrality (around 30 C), about 20% of energy is used to maintain body temperature (thermogenesis).Thermogenesis is predominantly mediated by the sympathetic nervous system (SNS) 117 and thyroid hormones 118 through their actions in the brain, BAT, and/or skeletal muscle.Whereas rodents rely heavily on BAT to stimulate thermogenesis due to a high surface-to-volume ratio, humans rely much more on skeletal muscle to generate heat. 119Studies using positron emission tomography (PET) scans have shown that BAT exists in adult humans [120][121][122][123] and displays molecular and cellular characteristics that are more akin to classical BAT in mice, rather than the beige adipocytes (adipocytes in white adipose tissue [WAT] with thermogenic properties) typically observed in the inguinal 'white' adipose tissue depot of mice. 124In humans, thermogenesis induced by overfeeding is highly correlated with mild cold-induced adaptive thermogenesis. 125However, BAT contributes a small and highly variable amount to overall energy expenditure in humans, and currently there is limited evidence to support the view that cold-induced BAT activation can produce weight loss. 119Notably, recent evidence indicates that individuals with more BAT have a lower risk of cardiometabolic diseases, 126 findings that require further investigation and may shed light on whether and how BAT activation, or hormones produced by BAT, may be harnessed to treat people with cardiometabolic disease.
There has been considerable debate on the optimal housing temperature for mice to simulate human metabolism and living conditions.Thermoneutrality (about 30 C for mice) was recommended as an optimal housing condition.However, Speakman suggested that 25.5 C -27.6 C is the best housing temperature for singly housed adult C57Bl6 mice because at these conditions, their energy expenditure is 1.7-1.8times the BMR, a ratio similarly observed in humans. 127,128Cannon and Nedergaard demonstrated that at 30 C, energy expenditure is about 1.6 times the BMR and suggested that 30 C is the condition of choice. 129More recently, Reitman and colleagues proposed the thermoneutral point (TNP) concept and showed that energy expenditure increases below the TNP and body temperature increases above the TNP. 130This team further demonstrated that mouse TNP is 29 C in the light phase and 33 C in the dark phase 130 suggesting that measurement of energy expenditure should be performed at various ambient temperatures.While warm temperatures can minimize the contribution of muscle shivering, they also affect lipolysis and mouse behavior.As such, care is needed when interpreting mouse energy expenditure data and assessing its relevance to human physiology.While several genetic mouse models of obesity such as ob/ob exhibit reduced BMR, 131 BMR and total energy expenditure (measured using indirect calorimetry and the doubly labeled water method) are appropriate for age, sex, and body composition in humans with congenital leptin deficiency. 132Species differences and differences in the sensitivity of measurements made in humans are possible explanations for these findings.Additionally, studies by Ravussin and Licinio in adults with leptin deficiency pre-and post-weight loss induced by recombinant leptin administration, 133 showed that the adaptive response to weight loss (which typically results in a fall in total energy expenditure) is attenuated in human congenital leptin deficiency.There are subtypes of human obesity where deficits in BMR are observed in mice and humans (e.g., loss-of-function mutations in Kinase Suppressor of Ras-2, KSR2) with a defect in thermogenesis being unmasked by the failure of KSR2 null mice to adapt to a drop in environmental temperature. 134he most commonly used method to measure energy expenditure in mice is indirect calorimetry, which calculates relative values of oxygen consumption and carbon dioxide production. 135However, there has been debate in the metabolism field about how energy expenditure should be normalized to account for body size or mass. 136,137Many early studies divided energy expenditure by body weight, but this normalization is problematic because metabolic rate is not linearly proportional to body weight, and WAT, which increases in obese animals, has lower metabolic activity than fat-free tissue. 138][141][142] ANCOVA can statistically account for the influence of a continuous variable (e.g., body mass) from group comparisons of a dependent variable (e.g., energy expenditure).A web-based analysis tool, CalR, provides an easy-to-use interface with a built-in ANCOVA function, 143 which has become a standard tool for investigators to analyze energy expenditure data in rodents.However, as recently reviewed, 144 many researchers studying rodents still divide energy expenditure measurements by weight, which may result in the spurious interpretation of their data.
Other considerations are that measurement of energy expenditure can only be conducted in singly housed animals.There is evidence that the impact of single versus group housing on locomotor activity 145 and social behaviors 146 can influence energy balance. 147Energy loss through feces and urine, directly measured by bomb calorimetric combustion of dried fecal and urine samples should also be considered, although sample collection may be difficult. 148hen investigating the relative contribution of changes in energy intake and/or expenditure to obesity or altered body composition, another approach is to measure these parameters early in life, before the divergence in body weight or fat/lean mass.The pair-feeding strategy has also been often used to tease out the contributing effects of feeding, where one group of animals (with hyperphagia and often obesity) is given a reduced amount of food that matches the other group.However, one needs to be cautious when interpreting the outcome of such experiments, where one group is being calorically restricted while the other group is not.Caloric restriction is wellknown to cause numerous changes in metabolic, endocrine, and neurocognitive systems, 149,150 which may confound the results.Fundamentally, energy intake and energy expenditure interact 105 ; interventions in mice and humans need to quantify changes in both components to capture the consequences of any perturbation on energy homeostasis.

NUTRIENT PARTITIONING
There are key species differences in how dietary macronutrients (fat, carbohydrate, and protein) are digested, absorbed, and metabolized (substrate utilization) and in how excess energy is stored in rodents and humans.For example, lipids are predominantly transported in the blood as lipoproteins.However, nonhuman primates, mice, and dogs transport cholesterol primarily within high-density lipoprotein (HDL) particles, whereas humans (and pigs) predominantly rely on low-density lipoprotein (LDL).Bile acid species, relative quantities of sphingolipids and ceramides (associated with cardiovascular disease), intestinal fat absorption, and reverse cholesterol transport differ between species, 151 factors that are sometimes overlooked when modeling cardiometabolic disease in animals.Important species differences in postprandial glucose uptake by skeletal muscle, liver, and adipose tissue become relevant when studying mechanisms underpinning type 2 diabetes. 152,153In healthy humans, about one-third of glucose is taken up by the liver with the remainder utilized by skeletal muscle, where insulinstimulated glucose uptake is regulated by the high-affinity transporter GLUT4 (glucose transporter type 4).By contrast, mice have a reduced requirement for insulin to restore glucose levels after challenges, and the liver, rather than skeletal muscle, plays the predominant role in glucose disposal. 154Quantification of the contribution of different organs to overall glucose homeostasis, using hyperinsulinemic/euglycemic clamps, can be particularly informative.
In humans, energy is predominantly stored as triglycerides in subcutaneous WAT and, to a more limited extent, as glycogen in the liver.Due to its far greater mass in humans, WAT stores 100 fold more energy than is stored in the liver and muscle combined; similarly, glycogen stores in skeletal muscle are much larger than in the liver.In mice, inguinal and gonadal fat pads display similar cellular characteristics and hormone responsiveness to human WAT, 155 but there are particular challenges in modeling human visceral fat 156 and WAT single-cell RNA sequencing datasets are revealing the extent of differences in gene expression and gene regulation between species. 157n epidemiological studies, weight gain is associated with the development of insulin resistance and increased risk of type 2 diabetes.Studies in people with congenital lipodystrophy syndromes (characterized by the impaired development of adipose tissue 158 ) and in mouse models of these diseases 159 have shown that when the buffering capacity of subcutaneous WAT is exceeded, ectopic lipid accumulation in the liver and skeletal muscle causes peripheral insulin resistance and type 2 diabetes.Body fat distribution is important; increased ratio of visceral to subcutaneous fat (inferred by waist-to-hip ratio) increases the risk of type 2 diabetes, particularly in men compared with pre-menopausal women.Additionally, genetic variation in the amount of gluteo-femoral (hip) fat significantly increases metabolic risk, 160 supporting the concept that when the amount of fuel entering WAT exceeds its oxidative or storage capacity, metabolites (e.g., ceramides, diacylglycerol, reactive oxygen species) that inhibit insulin signaling are formed, contributing to insulin resistance and increased risk of type 2 diabetes and associated metabolic complications (e.g., hepatic steatosis).
Insulin resistance drives hyperplasia, dysfunction, and then failure of pancreatic b-cells, and common genetic variants affecting pancreatic insulin secretion (studied in human islets) play a major role in mediating the risk of type 2 diabetes. 161While many rodent models have been used to study the pathophysiology of type 2 diabetes, the glucose threshold that equates to human diabetes is unclear; background strain, sex, and exposure to stressful stimuli at the time of experimentation can affect the onset and severity of hyperglycemia. 162ome of the challenges of investigating mechanisms and validating drug targets for type 2 diabetes in mice are illustrated by research on the nuclear receptor peroxisome proliferator activated receptor gamma (PPARg), which is required for adipogenesis and insulin action and is the target of insulin-sensitizing drugs, the thiazolidinediones. 163,164In humans, heterozygous dominant negative mutations in PPARg cause severe insulin resistance, partial lipodystrophy, hypertension, and hypertriglyceridemia. 165 However, several mouse models of PPARg disruption did not replicate these features.PPARg has two isoforms: PPARg1 is widely expressed, whereas PPARg2 is expressed predominantly in adipose tissue, where its expression is nutritionally regulated.Its essential role in adipose tissue function only became apparent when PPARg2 null mice were studied in response to overnutrition (generated on the ob/ob genetic background) and developed severe insulin resistance, early type 2 diabetes, and hyperlipidemia. 166Hypertriglyceridemia in humans with PPARg mutations (and other forms of lipodystrophy) may similarly only become apparent after sustained positive energy balance.These studies demonstrate the importance of precise physiological challenges in unmasking mechanisms underlying metabolic disease in both mice and humans.

SYSTEMIC INFLUENCES ON METABOLISM AND METABOLIC DISEASE
Metabolic diseases arise when the systems that regulate energy homeostasis are disrupted by external or internal factors.For example, in humans, night-shift workers are more likely to develop obesity and type 2 diabetes, 167 highlighting the impact of biological clocks/circadian rhythms that maintain hormone secretion, eating patterns, metabolic flexibility, and other parameters over 24 h.While humans eat and are active during the day, rodents are more active during the night. 168Disruption of circadian rhythm in human experimental medicine studies leads to metabolically adverse effects. 169Studies in rodents and other model organisms have identified key molecular regulators of circadian rhythm, such as the Clock genes, which maintain an internal timing system regulating gene expression networks that oscillate over a 24 h cycle. 170While repeated access to multiple human tissues/organs during a 24 h period is often not possible, analysis of diurnal variation in the human blood transcriptome has revealed clusters of gene expression networks specific to the day and night that suggest possible shared mechanisms of gene regulation. 171Stimulated by studies in rodents, there is considerable interest in measuring the impact of time-restricted eating on metabolic health in humans. 172,173dditionally, studies in rodents have shown that hypothalamic circuits regulating the sleep/wake cycle are directly connected to those that regulate energy homeostasis.For example, information about the time of day is integrated with current metabolic status by AgRP neurons in the arcuate nucleus of the hypothalamus, which connect to neurons in the suprachiasmatic nucleus. 174,175leep deprivation markedly alters food intake and insulin sensitivity independent of effects on adiposity, 176 while acute caloric restriction increases the time spent in deep (stage 4) sleep in humans. 177These experimental studies in mice and humans provide a framework for understanding the impact of sleep duration and quality on the risk of obesity and metabolic disease observed repeatedly in epidemiological studies; however, further studies are needed to more fully understand underlying mechanisms at the cellular, tissue, and systemic level and how they might be targeted by interventions to improve health.
In mice and humans, obesity is associated with the recruitment of macrophages and increased expression of pro-inflam-matory cytokines in many tissues and organs. 178,179Studies in rodents have shown that in WAT, obesity is characterized by adipocyte cell death and the invasion of adipose tissue macrophages, which are polarized into a pro-inflammatory M1-like state, secrete cytokines including tumor necrosis factor-a (TNF-a), and contribute to tissue remodeling.Sustained inflammation in WAT can lead to fibrosis.Multiple studies in animals have shown that TNF-a administration impairs insulin receptormediated signaling and can cause insulin resistance. 180Binding of TNF to its cognate receptor activates nuclear factor kB (NF-kB), multiple downstream complexes referred to as the ''inflammasome,'' and leads to the transcriptional activation of chemokines and cytokines (in particular interleukin [IL]-6 and IL-1b), which can directly impair insulin signaling and may drive the development of type 2 diabetes (this topic has been extensively reviewed by others [181][182][183] ).Human studies suggest that inflammation (and fibrosis) in WAT may be a consequence of obesity. 184However, to date, there is no evidence from GWAS studies that variation in inflammatory genes is associated with increased susceptibility to type 2 diabetes.Circulating levels of TNF-a in people with obesity are below concentrations shown to impair insulin signaling in cells and animals.Clinical trials of anti-inflammatory drugs such as inhibitors of IL-1b or TNF-a have not consistently demonstrated improvements in insulin sensitivity. 185However, absence of evidence (in humans) is not evidence of absence.Given the challenges of establishing causality in humans, critical information about the translatability of rodent data demonstrating inflammation as a driver of metabolic and cardiovascular disease may come from clinical trials of drugs that target the inflammasome, which are showing considerable promise in pre-clinical studies. 186ne key question for the field is whether inflammation in the hypothalamus contributes to impaired function of circuits involved in energy homeostasis and thereby contributes to weight gain, resistance to weight loss, or the variable response to pharmacotherapy or bariatric surgery in some people.Inflammatory changes in the hypothalamus, typically reactive gliosis, are seen in mice placed on a high-fat diet within 2 days (i.e., before obesity develops 187 ).These changes involve activation of microglia, astrocytes, and host immune cells, a process that can directly or indirectly inhibit molecular components of leptin signaling.Currently, data in humans is lacking, but MRI studies have reported gliosis in the hypothalamus of people with obesity compared with normal-weight people. 188Animal models of weight loss and regain 189 have suggested that the immune cell composition of WAT may also change with these perturbations.These are some examples, but many other factors impact energy homeostasis.Stress and hormones that mediate the cellular and systemic stress response can impact energy homeostasis.Energy demands and body composition change during the life cycle from gestation, infancy, puberty, and pregnancy through to old age, requiring adaptative changes that remain incompletely understood across species.

FUTURE PERSPECTIVES
Studies in mice and humans have transformed our understanding of the system that regulates energy homeostasis and its disruption in obesity and metabolic disease.These discoveries have paved the way for the development of pharmacotherapies that are reducing the morbidity and mortality of people with obesity (see Review by Kusminski et al. in Cell 190 ).However, there are many outstanding questions, including how genetic risk alleles and other factors influence a person's set-point/set-point range, how internal and external cues are integrated to drive adaptive changes in eating behavior in the free-living (rather than controlled experimental) environment, how dietary/nutritional components are absorbed and sensed, and whether this process is modified by the brain/other organs.Some of these questions can be addressed in mice and other model organisms (including flies and zebrafish, which share up to 70% of human disease genes); indeed, genetic and other manipulations in tightly controlled conditions will continue to have an important role to play in the discovery of causal mechanisms.However, whether findings in mice translate fully to humans should not be assumed.We have discussed some of the key challenges here, though there are others.Even where molecular mechanisms are preserved between rodents and humans, the magnitude of effect or timing of the effect of disruption may differ substantially between species, as illustrated by the phenotype of complete MC4R deficiency, which results in late-onset obesity in mice, 15 but childhood-onset obesity in humans. 78Alongside differences in free-living eating behavior, access to food, environmental temperature, and other factors mentioned in this perspective, the contribution of genetic background, gene-gene, and gene-environment interactions is still poorly understood.
The impact of overlooking differences between mice and humans can be substantial particularly in drug discovery if success in pre-clinical studies is not replicated in clinical trials.For several decades, there were major challenges to the development of safe and effective anti-obesity medicines, in part because drugs that were effective in pre-clinical models were less effective when administered in clinical trials, caused clinical side effects that were not evident in animals (e.g., depression, suicidal ideation with Rimonabant), or were associated with effects that were not anticipated based on their known mechanism of action (e.g., Lorcaserin).The checkered history of anti-obesity drug development (reviewed by Muller, Bluher, Tschop, and DiMarchi 191 ) led to the withdrawal of many large pharmaceutical companies from this space.However, building on a substantial body of work in rodents and clinical trials in people with type 2 diabetes, the licensing of agonists targeting the receptors for GLP-1, glucose-dependent insulinotropic polypeptide (GIP), and glucagon has changed the landscape of drug development in metabolic disease (see Review by Kusminski et al. 190 ).It is clear that improved mechanistic understanding can inform the development of effective medicines for weight loss; however, there is much more to be learned.For example, single cell and spatial transcriptomics of the human heart revealed GLP-1 receptor expression on pacemaker cells, which control heart rate 192 ; a finding not observed in mice, but one that is clinically relevant as people treated with GLP-1 receptor agonists experience an increase in heart rate, which may pose a risk in those with cardiovascular disease.It is likely that studies in cells (including human stem-cell-derived cell lines/organoids), mice, and, increasingly, in humans will be important for the discovery of new drug targets and their validation.Early experimental medicine studies in targeted groups of patients have been underutilized to date but can aid decision-making and provide insights into the mechanism of action in humans.
While there are significant limitations associated with research in humans (e.g., access to relevant tissues, especially the brain), insights from human genetics and experimental medicine studies are poised to play an increasingly important role in our understanding of disease mechanisms.Technological developments have the potential to deliver an unprecedented advance in our understanding of disease mechanisms through the characterization of genomic, transcriptomic, proteomic, and metabolomic variation in samples from clinical cohorts and populations, which can be compared with large datasets derived from mice. 193Newer imaging techniques (such as 7 Tesla Magnetic Resonance Imaging [MRI]) can characterize the anatomy of the brain, heart, and liver at high resolution and can be coupled to studies of metabolic function (magnetic resonance spectroscopy).Just as in mice, in humans, experiments that test the adaptive response to a challenge will often be needed, and we need to train the next generation of clinician scientists who have the skills to ask these questions, undertake, and interpret these studies.
Currently, due to concerns about ethics and/or logistics, some research databases capture only minimal clinical data, which can limit the interpretation of molecular findings and obscure potential explanations for the heterogeneity of molecular results.To maximize the utility of these datasets and resources, permission to link to electronic health records and approach people with specific variants/traits for detailed phenotyping needs to become the rule rather than the exception.Appropriate ethical and data security safeguards will be needed.While some researchers have successfully navigated these challenges, more needs to be done to systematically allow for such translational and clinical research.Importantly, processes for clinical research could be simplified and should be proportionate to the risks involved (often minimal) while recognizing the potential benefits to patients.
After 30 years in which many fundamental discoveries have transformed our understanding of human metabolic disease, we are at an exciting crossroads.Ultimately, we will need to leverage data from multiple species, obtained using orthogonal approaches, to dissect the complex and heterogeneous drivers of human metabolic disease and inform approaches to the prevention and treatment of conditions that cumulatively have a substantial impact on global public health.

Figure 1 .
Figure 1.Hypothalamic regulation of energy homeostasis(A) Neurons in the hypothalamus (green) maintain homeostasis for many parameters by integrating internal and external inputs, then changing outputs to main physiological set-points.(B) Schematic depicting how body weight or adiposity may fluctuate within a set-point range that has upper and lower intervention points (as described25 ).Challenges such as overeating or undereating at times of stress (grey arrows) may increase or decrease adiposity within this range.The black curved arrow indicates physiological mechanisms that act to maintain/defend an upper limit to this range.

Figure 2 .
Figure 2. Environmental, social, and genetic influences on body weight

Figure 3 .
Figure 3. Energy homeostasis system conserved in mice and humans Neurons in the hypothalamus and brainstem receive hormonal signals from adipose tissue and neural and hormonal signals from the gut and pancreas.Within the hypothalamus, there are several distinct nuclei involved in weight regulation: ARC (arcuate), VMH (ventromedial hypothalamus), LHA (lateral hypothalamic area), and PVH (paraventricular hypothalamus).Leptin-responsive neurons expressing agouti-related protein (AgRP) and pro-opiomelanocortin (POMC) project to second-order neurons expressing the melanocortin 4 receptor (MC4R), forming a key circuit whose disruption causes obesity in mice and humans.CCK, cholecystokinin; PYY, peptide YY; OXM, oxyntomodulin; GLP-1, glucagon-like peptide-1.