|Home | About | Journals | Submit | Contact Us | Français|
Over the last decade there has been a striking increase in the early onset of metabolic disease, including obesity and diabetes. The regulation of energy homeostasis is complex and involves the intricate integration of peripheral and central systems, including the hypothalamus. This review provides an overview of the development of brain circuitry involved in the regulation of energy homeostasis as well as recent findings related to the impact of both prenatal and postnatal maternal environment on the development of these circuits. There is suprising evidence that both overnutrition and undernutrition impact the development of these circuits in a similar manner as well as having similar consequences of increased obesity and diabetes later in life. There is also a special focus on relevant species differences in the development of hypothalamic circuits. A deeper understanding of the mechanisms involved in the development of brain circuitry is needed to fully understand how the nutritional and/or maternal environments impacts the functional circuitry as well as the behavior and physiological outcomes.
In recent decades, obesity during infancy and childhood has become a devastating consequence of the “obesity epidemic” in our country and around the world. According to the recent National Health Examination Survey (NHANES), 33.6% of children within the U.S. are considered overweight, as are 12% of infants . Similar to that in adults, the increased prevalence of childhood obesity has led to an increased risk of hypertension, type II diabetes, dyslipidemia, left ventricular hypertrophy, nonalcoholic With childhood obesity, the resulting metabolic disturbance may permanently disrupt normal development and increase susceptibility to these diseases independently of genetic factors and nutrition in adulthood.
In the United States, it is estimated that nearly two-thirds of adults are either overweight or obese [2; 3; 4; 5]. According to the Center for Disease Control and Prevention (CDC), for adults aged 20-74 years the prevalence of obesity increased from 15.0% in the 1980’s to 32.9% in the 2006. However, since 2003 the prevalence of obesity has not increased, but the degree of obesity has. Worldwide, excess body-weight affects 1.1 billion adults . In adults, obesity is diagnosed as a body mass index (BMI) of 30 kg/m2 or higher (calculated by dividing body-weight by height squared). Obesity is part of a larger disease process termed “metabolic syndrome” which generally includes insulin resistance, impaired glucose tolerance, hypertension, elevated triglycerides and/or cholesterol, and elevated urinary albumin excretion rate, an indicator of renal failure [7; 8; 9]. If sustained for long periods of time, these conditions expand to to diabetes, cardiovascular disease and neuropathies.
Considering the sheer numbers of individuals dealing with issues of body-weight regulation, only about 5% of obesity is linked to known single-locus gene mutations. Taking into account the dramatic rise in the incidence of obesity in recent decades, Mendelian genetic shifts must be ruled out as the direct causative factor of this incidence. However, a more likely explanation is that the majority of humans are genetically susceptible to obesity when consuming a calorically dense diet. The growing population of obese individuals indicates that the majority of cases of obesity are likely a combination of genetics, epigenetics and environment [10; 11].
In children, metabolic disease has increased concomitantly with increases in adult populations. Childhood overweight/obesity is defined using BMI in the same calculation as previously described for adults. However, for children the BMI is then plotted on a growth chart - with greater than the 95th percentile considered overweight or obese. A BMI between the 85th percentile and 95th percentile is considered at risk for becoming overweight. Among U.S. children the overall prevalence of risk for overweight is estimated to be 33.6% for the years 2003-2004 . Between studies analyzed in 1999-2000 and 2001-2002, the prevalence of overweight/obese children aged 2-5 years increased from 5.0% to 13.9%; for those aged 6-11 years, prevalence increased from 6.5% to 18.8%; and for those aged 12-19 years, prevalence increased from 5.0% to 17.4%. The serious issue in our modern times is that nearly 15% of the US child population is in the 95% percentile. Many risk factors have been identified for childhood obesity. These include parental obesity, higher birth weight, higher weight gain during the first year and short sleep duration at age 3, neonatal adiposity rebound due to low weight or weight gain at birth, more than 8 hours of television per week at age three, and presence of catch-up growth phases [12; 13; 14]. Maternal undernutrition during pregnancy, maternal obesity during pregnancy, low birth weight and high birth weight all appear to predispose individuals to obesity and type 2 diabetes later in life . In addition, milk formula feeding as opposed to breastfeeding has a higher association with long-term obesity [16; 17]. Smoking during pregnancy is also linked with obesity in offspring . Importantly, population-based human studies reveal correlations between fetal high-fat diet exposure and adult weight regulation [19; 20; 21]. With so many factors appearing to have a significant impact on long-term body-weight maintenance, it is difficult to estimate the relative contribution of each to the childhood obesity epidemic.
Obesity during childhood negatively impacts many facets of a child’s well-being encompassing cardiovascular health, musculoskeletal development, reproduction (including puberty) and psychological health. Though these problems affect the peripheral organs and functions, they are all controlled functions of the central nervous system. In the Bogulusa Heart Study, a study of the long-term cardiovascular health of over 16,000 children and young adults, overweight children were reported to be 4.5 and 2.4 times more likely to have elevated systolic and diastolic blood pressures, respectively . Hypertension was found to be 22% more likely in obese than nonobese children . Overweight children report a greater prevalence of fractures and musculoskeletal discomfort as well as greater impairment of mobility . Various parameters of childhood metabolic syndrome such as elevated cholesterol, glucose and triglyceride levels are associated with pre-pubertal sleep disordered breathing [24; 25]. Children with high body-weight, either at birth or later in childhood, are at increased risk for future asthma .
The most immediate yet underemphasized consequences of obesity during childhood are psychosocial, resulting in discrimination and negative self-image during adolescence [27; 28]. Obese adolescent females are at especially high risk for major depressive and anxiety disorders . Together, these findings indicate that childhood obesity has damaging implications in the immediate as well as long-term.
The fetal origins of disease hypothesis or Barker Hypothesis purports that adverse intrauterine conditions can cause permanent changes in the embryo/fetus, predisposing it to chronic disease later in adult life [30; 31; 32]. Cohort studies have shown that disturbed intrauterine growth has a negative influence on the development of the cardiovascular system and favors the occurrence of hypertension, insulin resistance, hypercholesterolemia, and hyperuricemia in adult life [33; 34; 35; 36]. The most well documented examples are from the Dutch Famine Studies. Evidence from these cohorts revealed that the early nutritional deprivation in pregnant mothers, as a result of famine, produced small-for-gestational-age (SGA) babies. The consequences were irreversible health issues in the offspring. Through longitudinal studies of this population, it came to light that exposure to famine during any period of gestation can lead to glucose intolerance during mid-life (50-60 years) . The worst consequences exist for individuals who experienced famine during early gestation.
These early studies linking low-birth weight and late-onset obesity and diabetes laid the groundwork for the thrifty phenotype hypothesis, which proposes that poor nutritional conditions during gestation and early infancy can modify metabolic systems in the offspring to adapt to expectations of chronic undernutrition. With a subsequent calorically dense environment, these offspring are poorly equipped to cope with energy-dense diets and are possibly wired to store as much energy reserves as possible, leading to late-onset metabolic syndrome . Insults such as poor nutrition and decreased maternal calorie or protein consumption are known to cause intrauterine growth retardation in various mammals [38; 39; 40; 41; 42]. However, in today’s world of overweight and obese adults, the more likely insult during pregnancy is overnutrition and diabetes. Yet overall poor nutrition, such as poor protein sources, definitely contribute to the situation.
In support of the Barker Hypothesis, it is well established that abnormal fetal growth affects long-term metabolic health [43; 44]. High maternal BMI, even in the absence of maternal gestational diabetes, is a risk factor for large-for-gestational-age (LGA) infants who are highly susceptible to lifelong disease [45; 46]. Fetal exposure to maternal obesity increases the risk of childhood metabolic syndrome by approximately two-fold [15; 45]. In addition, obesity during pregnancy has implications for morbidity and mortality for the mother and the baby. Maternal obesity more than doubles the risk of stillbirth and neonatal death [47; 48]. Overweight mothers also have a greater risk of hypertensive complications and preeclampsia . Finally, women with poor glycemic control and/or gestational diabetes have a high risk of giving birth to a macrosomic baby that is more likely to develop obesity and insulin resistance later in life [50; 51; 52]. Though much work has been done to document the trends and associations between maternal nutrition and health during gestation and long-term metabolic health of offspring, it is still not known which mechanisms are responsible for the increased risk for obesity in adulthood.
Critical periods exist in child development during which nutrition and health status may predispose the individual to life-long body-weight regulation difficulties. One key period is the in utero/early infancy period . Hypothalamic appetite regulatory systems develop and mature during the in utero/early infancy period. The primary focus of this review is to describe the normal development of hypothalamic appetitive systems as well as their development in the context of genetic, nutritional and environmental manipulations to further understand their impact on body-weight regulation.
The neuronal circuitry that controls food intake and energy homeostasis has been extensively studied over the last two decades. Numerous excellent reviews have been published covering this literature [53; 54; 55; 56; 57; 58; 59; 60; 61; 62; 63; 64]. Briefly, appetite regulation is a complex system with both homeostatic and hedonic aspects as well as central and peripheral components integrated at the level of the hypothalamus and brainstem  (Figure 1). At the crossroads of the central nervous system and peripheral components is the arcuate nucleus of the hypothalamus (ARH). A critical aspect of ARH neurocircuitry is the interplay between the anorexigenic peptide-containing neurons such as α melanocyte-stimulating hormone (αMSH) (a cleavage product of the pro-opiomelanocortin, or POMC, gene) and cocaine and amphetamine-related transcript (CART), and orexigenic peptide-containing neurons, namely neuropeptide Y(NPY) and agouti-related peptide (AgRP) (Figure 1). These neurons are sensitive to blood-borne indicators of energy status (leptin, insulin, glucose, fatty acids, amino acids and gut hormones) and project to other feeding centers in the hypothalamus such as the dorsomedial nucleus of the hypothalamus (DMH) and lateral hypothalamic area (LHA) where they recruit or modulate orexin and melanin concentrating hormone (MCH) neurons (Figure 1). Projections from these nuclei re-converge on the paraventricular nucleus (PVH), which is a site of integration of metabolic signals from many parts of the brain (Figure 1). The efferent outputs of the PVH involved in the regulation of food intake are poorly understood, but is thought to involve inputs onto corticotrophin releasing hormone (CRH) and/or thyrotropin releasing hormone (TRH). The efferent outputs for the regulation of energy expenditure are better understood and involve CRH and oxytocin neurons in the preautonomic area of the parvocellular region of the PVH.
In addition to the critical role of the hypothalamus, various nuclei in the hindbrain contribute to the control of food intake particularly through motor and gustatory mechanisms (Figure 1). Brainstem control of food intake is largely mediated through the dorsal vagal complex (DVC) which includes the dorsal motor nucleus of the vagus (DMV), nucleus of the solitary tract (NTS) and area postrema (AP) (see reviews [66; 67]). For example, mastication, swallowing, and salivation are controlled by the caudal brainstem. The DVC acts as a relay site for short-acting gastrointestinal signals, mediated predominantly through the vagus nerve. In addition, gut peptides communicate with other hindbrain nuclei at the AP, a circumventricular organ with decreased blood brain barrier. These gut-hindbrain connections are sufficient to produce satiety on their own, but can be overridden by more cognitive inputs of body-weight regulation.
Superimposed on the homeostatic neurocircuitry of body-weight regulation are the hedonic aspects of food intake (see review ). This behavioral aspect, based on the palatability and reward of food, can override the normal requirements of daily energy needs to meet normal body weight homeostasis. These hedonic neurocircuits span brain areas such as the nucleus accumbens (NAc), ventral pallidus (VP) ventral tegmental area (VTA) and higher cortical areas such as the prefrontal cortex and converge on the LHA. These pathways utilize biological substrates such as glutamate, opioids, endocannabinoids and dopamine. While these pathways can over-ride homeostatic regulation they are still responsive to peripheral metabolic signals, most notably, leptin [68; 69; 70; 71; 72; 73] and ghrelin [74; 75]. The well-known effects of these compounds on mood likely contribute significantly to the powerful and overwhelming impact of the hedonic pathways on food intake.
Though many of the neuropeptide systems which control food intake are “born” in utero in most species, the timing for the development of their projections is dependent on the neuronal phenotype as well as there being a large species difference. The most well studied nucleus in this regards is the ARH. In the rat and mouse ARH neurons have a birth date around E11.5-12.5 [76; 77; 78] (Dr. L Zeltzer, personal communication). However, ARH neurons do not start to develop their efferent projections until near the end of the first postnatal week, and they are not completed until the end of the 3rd postnatal week [79; 80]. The physiological significance of this delayed postnatal development is the lack of homeostatic regulation of energy balance through the ARH. An example of this is the inability of leptin to stimulate energy expenditure and inhibit food intake until after the 3rd postnatal week [81; 82; 83; 84]. This lack of homeostatic feedback on body weight and food intake is not due to a lack of function receptors within the ARH since leptin can stimulate pSTAT3 immunoreactivity and regulate ARH gene expression soon after birth [80; 82]. Furthermore, fasting (or rather maternal deprivation) increases ARH-NPY expression within the first postnatal week [85; 86]. However, these infant pups lack the ability to generate a rebound hyperphagia to normalize body weight . Models that result in the abnormal development of the ARH circuits lead to obesity. This includes 1) the neonatal MSG-treated animal that leads to an early lesioning of the ARH, resulting in stunted growth and increased adiposity ; 2) the obese leptin deficient ob/ob mouse which lacks normal development of the ARH circuits ; and 3) the Levin high fat diet obese (DIO) Sprague-Dawley rat . In this last model, Bouret et al. demonstrated the DIO sensitive rat, which has a polygenic mode of inheritance for diet induced obesity, has a significantly reduced development of ARH circuits compared to the diet resistant animals. The development of specific neuropeptide circuits and the key signals that may be involved in the development of ARH circuits will be discussed below. Overall, this highlights the importance of the ARH circuits in the regulation of energy homeostasis. So while it is recognized that there are many brain regions involved in the regulation of body weight and food intake, disruption of the normal development of the ARH leads to obesity ... at least in rodents.
There is an overwhelming deficiency in our understanding of the development of brain circuits that are involved in the regulation of energy homeostasis in higher species. In humans and nonhuman primates (NHP), the neurons in the ARH can be morphologically distinguished during the 2nd trimester [90; 91], surprisingly, not that different from the rodent; however, the development of their projections occurs during the 3rd trimester of pregnancy . This shift in timing between species has many potential implications for the impact of maternal health and nutritional status versus milk quality and availability on the developing offspring. This has been most extensively studied in the rodent. The subsequent portions of this review will focus on studies in rodent models but will include additional discussion on studies performed in higher species, such as in primates and sheep, when available. This review will cover 1) the natural development of the specific neuropeptide circuits involved in the regulation of energy homeostasis, 2) the evidence for the role of peripheral metabolic signals in this development and 3) the effects of maternal and early postnatal nutrition and metabolic health on the development of these circuits.
In considering the nutrient demands to meet the rapid growth of a neonate, the early appearance of NPY within development is not surprising. In the rodent, NPY mRNA is present in the brainstem and forebrain as early as embryonic day 14 (E14) [92; 93]. There is a relatively abundant level of NPY expression in the ARH throughout life in the rodent, starting at birth. However, NPY mRNA expression does increase from birth, to peak around P15-16, and subsequently slowly declines to adult levels by P30 . While in the normal adult mouse, the expression of the orexigenic neuropeptide within the hypothalamus is restricted to the ARH, during early postnatal development NPY mRNA is also transiently expressed in the DMH, PVH, LHA, perifornical region (PFR), anteroventral periventricular (AVPV) and suprachiasmatic (SCN) nuclei as well as the medial preoptic area (MPO) [85; 94; 95; 96]. The significance of this largely developmental-specific expression of NPY in multiple hypothalamic nuclei has not been elucidated. Yet it is difficult to overlook the fact that many of these regions are key sites for the regulation of energy homeostasis.
While NPY is expressed in neurons in the ARH prior to birth in the rodent, the efferent projection of these neurons develop during the postnatal period [85; 94]. In spite of this delayed development of ARH-NPY projections, NPY immunoreactive fibers are abundant throughout the hypothalamus, suggesting that these other transient populations of NPY neurons are significant contributors to the NPY tone early in development. However, the efferent projections for most of these other transient populations of NPY neurons are unknown. The brainstem also contains populations of NPY neurons that project to the hypothalamus. Our group has demonstrated that brainstem NPY projections to the PVH are present as early as P2, but likely develop prior to that [87; 95]. It might be assumed that since there is an overwhelming abundance of neurons expressing this orexigenic peptide during early development, that NPY may be a critical factor for maintaining sufficient energy intake during this rapid growth phase. It might be expected that a mutation resulting in the lack of NPY during critical developmental stages would result in stunted growth and development. However, this is not the case; rather mice with a genetic knockout of the NPY gene appear to grow normally and maintain a normal body weight phenotype, yet have an increased susceptibility to seizures [97; 98]. Furthermore, loss of NPY receptors also does not cause stunted growth and development, even though some of the receptors are highly expressed in the hypothalamus from birth [87; 95; 96]. This lack of effect has been attributed to many things, including developmental compensation. Indeed, ARH-NPY neurons also coexpress another potent orexigenic neuropeptide, AgRP; it is possible that AgRP is sufficient to sustain normal food intake necessary for normal growth and development. However, these projections do not develop until later in the postnatal period, and the NPY/AgRP double knockout mouse shows near normal growth and development . So where does this leave us? Does this mean that NPY neurons are not critical for normal growth and development? This depends on how one define “critical”. It is clear that mice can develop to a normal stature without NPY and/or AgRP; however, NPY and NPY/AgRP knockout mice are not completely normal, as they lack the ability to adapt to metabolic challenges, such as fasting and diet induced obesity [100; 101]. But the evidence of developmental compensation was highlighted in a study by Richard Palmiter’s group in which they demonstrated that ablation of ARH-NPY/AgRP neurons, which would include the GABA signal from these neurons, in adulthood results in starvation and wasting , a phenotype that reinforces the unique importance of these neurons in the regulation of energy homeostasis. Interestingly, ablation of these neurons soon after birth results in a near normal growth and development. What signals allow for this developmental compensation, or exactly what systems are responsible for the compensation remain a mystery. But this adaptation may not be surprising since there are numerous systems that have orexigenic effects.
In contrast to rodents, in both the human and NHP, NPY neurons are born during the middle of the second trimester [90; 91]. Like the rodents during early development, NPY expressing neurons are evident throughout the hypothalamus; such as in the SON, PVH, and DMH. However, this expression is not limited to early development in the NHP, but rather these neurons continue to express NPY in the normal adult . In the NHP, ARH-NPY projections begin to develop in the late second trimester, and they continue to develop throughout the third trimester and into the early postnatal period . While the effects of maternal nutrition on the fetal NPY system have been studied in the sheep, the development of their projections has not been characterized [104; 105; 106]; however, it would be assumed that it would be similar to the NHP and human. The in utero development of the NPY system in higher species again raises the question as to the physiological relevance of the development of an orexigenic system during the prenatal period, prior to the need for independent ingestion. The only evidence of a significant role comes from the sheep model in which icv NPY injections were shown to stimulate swallowing of amnionic fluid in the fetal sheep , a function that is thought to be important for fluid balance and for the development of the gastrointestinal tract.
In adult animals (including most species), AgRP is exclusively colocalized in in ARH-NPY neurons; this is also the case throughout development. Surprisingly though, AgRP mRNA levels appear to be first detectible during the first few postnatal days whereas NPY mRNA levels are detected much earlier in gestation [95; 108]. This suggests that NPY and AgRP may be differentially regulated within the same cell, possibly by different transcriptional regulators and/or different afferent inputs. Indeed, in the adult rodent, NPY and AgRP are not always regulated in parallel [109; 110; 111]. However, as expected, the development of AgRP projections follows the pattern of innervation by NPY .
In a like manner to NPY, POMC neurons are present within the hypothalamus in the developing rat at E12-13, several days before their appearance in the pituitary [112; 113]. At birth, and unlike AgRP, αMSH fibers are evident throughout the hypothalamus, including the PVH [87; 112]. At first glance this may suggest that ARH-POMC (αMSH) projections develop sooner than ARH-NPY/AgRP projections. However, we know from studies by Bouret and Simerly [79; 80] that this is not likely that case, since they used a nondiscrimanent anterograde tracer to demonstrate that all ARH projections are not present at birth. It is more likely that ARH-POMC neurons from the solitary tract nucleus in the brainstem [114; 115; 116; 117] are contributing to the PVH αMSH immunoreactivity at this early age . Though the projections of POMC neurons become evident towards the end of gestation they do not reach their full maturity until weaning.
It may seem counterproductive to have a satiety signal intact at such an early developmental stage, but early work has shown that endogenous αMSH functions to stimulate fetal growth [118; 119; 120]. When antibodies to αMSH are administered in late gestation, fetal growth retardation results . These data reveal substantial differences in αMSH actions during early development and adulthood, the former being growth promoting and the latter resulting in satiation. These divergent actions have not been further pursued. The growth promoting effect does conflict with evidence that POMC or MC4 receptor knockout mice are larger in both lean and fat mass [121; 122]. However, this may just be highlighting the critical role that melanocortin agonists play in the regulation of energy homeostasis later in life.
In the NHP, POMC neurons are readily detectable at gestational day 100 (early 2nd trimester) . Furthermore, the development of ARH-POMC projections, like the NPY/AgRP projections, occurs during the 3rd trimester; however, there is a delay in their development. While ARH-NPY/AgRP projects are readily detectable by gestational day 130 (early 3rd trimester), αMSH immunoreactive fibers are sparse until later in the 3rd trimester when αMSH projections become abundant. This is not simply due to preferential conversion to other POMC derived peptides since ACTH and β-endorphin immunoreactivity are also sparse in the early 3rd trimester (Grayson and Grove, unpublished observations).
Highlighting the importance of the melanocortin system, and especially the MC4 receptor, humans with mutations in this system develop obesity early in childhood. Like the mouse models, these individuals also have increased fat and lean mass and are generally taller [10; 11; 123; 124; 125; 126; 127; 128]. However, while MC4 and POMC mutant mice do develop obesity this does not occur until after or near weaning, paralleling the development of this circuitry.
CART is another neuropeptide implicated in diverse physiological functions, which include stress response, feeding, reward and autonomic function. CART is abundantly expressed in the brain, including hypothalamic regions such as the VMH, PVH, ARH, but also the nucleus accumbens, VTA, thalamic nuclei, dorsal raphe, amygdala, olfactory bulbs, spinal cord and NTS [129; 130]. In the rodent, CART has been found to co-localize with other prominent neuropeptides and neurotransmitters such as POMC, tyrosine hydroxylase and MCH in the hypothalamus, GABA in the nucleus accumbens and acetylcholine in the gut myenteric system [131; 132; 133; 134]. While CART likely has many physiological roles it has primarily been studied as a feeding-related peptide. The primary limiting factor in studying the CART systems is that its receptor remains unidentified.
CART neurons appear in the hypothalamus as early as E11 in the rodent . During the postnatal period CART immunoreactive neurons are detectable within the LHA as early as P5; however, it is not until between P11 and P15 that neurons in ARH become detectable by this method (Figure 2). This relatively late onset of expression in the ARH correlates with the development of these projections. Because there are so many populations of CART neurons it is difficult to know when each population develops its projections; however within the ARH it seems likely that it coordinates with ARH-POMC projections and within the LHA with MCH projections (see below). Very little is known about the potential role of CART during development, except for a potential role in the neuronal migration of dopamine (DA) neurons . This was evidenced by the presence of DA (tyrosine hydroxylase-positive) neurons migrating laterally from the neuroepithelium through the area containing the CART-immunoreactive neurons. These DA cells exhibited close dendritic appositions with CART perikarya. Mice with a deletion of the CART gene; however, display no abnormal developmental phenotype, although this has not been extensively studied. They do develop obesity later in life, but this is likely secondary to dysfunction in the pancreatic beta cell, which also expresses CART .
In the adult rodent, CART is known to colocalize with αMSH in the ARH and MCH in the LHA ; however, this does not appear to be phylogenetically conserved. In the human, CART is colocalized with NPY and AgRP and does not colocalize with αMSH . In the NHP, however, CART forms an independent population of neurons within the ARH that does not express either αMSH or AgRP at any stage of development  (Figure 3). This significant species difference accents the yet undescribed role of CART in the primate regarding food intake. Because, like the rodent, the NHP contains many populations of CART neurons it is difficult to characterize the development of their projections using histochemistry techniques.
Orexin (hypocretin) and MCH neurons within the LHA also play key roles in the regulation of energy homeostasis. These two independent populations of neurons are exclusively localized in LHA area, but have a broad distribution throughout this region including cells scattered as medially as the DMH. Furthermore, both populations of neurons have an impressive distribution of efferent projectsion that extend to most regiosn of the brain. Orexin neurons are evident in the LHA of the rodent as early as E19 . They have robust synaptic activity around parturition , suggesting active afferent inputs, but the projections do not fully develop until the 3rd postnatal week . In contrast, MCH mRNA expression is detectable as early as E13  with immunoreactive fibers being detectable as early as E18; however, MCH fiber projections are not readily detectable throughout the brain until P5 .
Both orexin and MCH immunoreactive neurons are readily evident in the LHA in the NHP as early as G100 (late 2nd trimester; Figure 4). However, it seems likely that at least the orexin neurons must exist even earlier since their projections to the ARH (Figure 5), as well as to other brain regions, are apparent as early as G100, even though they are relatively sparse at this age. In contrast, the MCH projections are delayed in their development and don’t become prominent until just prior to parturition (Figure 5). It has always been intriguing as to the physiological relevance of such an early development of the orexin system. Why does the 2nd trimester fetal primate brain need orexin projections to such far reaches as the cortex? This implicates the orexin system as an important developmental signal in the primate brain. In the adult NHP, CART and orexin immunoreactive fibers have the highest density projections of all neuropeptide systems we have studied.
In direct contrast to hypothalamic circuit maturation in the rodent, the brainstem neurocircuitry matures prenatally and with further refining and direct input from hypothalamic nuclei postnatally. Therefore, in the rodent, these circuits are conceivably far more sensitive to the maternal environment in comparison to the hypothalamic development which appears to be influenced more by the postnatal nutrient environment. This section will only discuss the development of the serotonin, dopamine and norepinephrine systems because of their significant roles in the regulation of energy and glucoses homeostasis, [71; 142; 143; 144; 145; 146; 147; 148; 149; 150; 151; 152; 153; 154]. However, it should be recognized that these systems are extensive, with many different populations of neurons, making it difficult to distinguish the development of one population versus the next. Also, very little is known about the development of these systems, with regards to energy homeostasis, in higher species.
In the rodent, serotonin-ir appears in the raphe at E12  and subsequent expression of the serotonin transporter (5HTT) ensues shortly thereafter [156; 157]. By E18-20, serotonin immunoreactivity can be visualized throughout the brain, including the hypothalamus . Because of the ubiquitous expression of serotonin in the brain, nearly every neuropeptide and neurotransmitter system is impacted by serotonin terminal differentiation . However, the exact timing of the development of serotonin projections to hypothalamic circuits involved in the regulatin of energy homeostasis has not been directly studied. The melanocortin neurons are a primary efferent target by which serotonin regulates energy homeostasis [142; 143; 144; 145; 146; 159; 160]. Since the serotonin system develops relatively early it could be a key signal for the normal development of the melanocortin system.
Similar to serotonin, noradrenergic neurons appear on E12 in the rat . In general immunoreacitivity for the different components of the noradrenergic system (i.e., the density of immunoreactive projections) is surprisingly more robust earlier in development in comparison to the serotonin system even though both types of neurons have similar birth dates . In the rat, a wave of proliferation in fiber density and intensity occurs during the first postnatal week, with adult levels being reached by the third postnatal week . Noradrenergic projections to the hypothalamus, and especially the PVH, have been strongly implicated in glucoprivic feeding - hyperphagia induced by hypoglycemia. Glucoprivic feeding is induced by administration of 2-deoxy-D-glucose, a nonmetabolizable form of glucose. These brainstem catecholaminergic projections, which also contain NPY, develop during the perinatal period in the rodent [95; 163; 164]. However, glucoprivic feeding responses do not appear to develop in rodents until weaning at approximately 21 days postnatal [165; 166]. In general, noradrenergic innervation appears to be crucial to the proliferation of brain astrocytes and glia and therefore neuronal migration is affected throughout the brain when noradrenergic development is affected .
The dopamine system within the VTA is intricately involved in the rewarding and/or addictive properties of food [167; 168; 169; 170; 171]. The dopamine neurons within the VTA have been shown to be directly regulated by important peripheral metabolic signals such as leptin and ghrelin [71; 154; 167]. Like the above noradrenergic and serotonin monoamine transmitter systems, the VTA dopamine projections appear to develop early in the postnatal period in the rodent, as indicated by the presence of dopamine terminals within the nucleus accumbens, a major target site for VTA-dopamine neurons . However, synaptic maturation within the nucleus accumbens does not fully develop until the 3rd postnatal week , indicating that both maternal and neonatal nutrition and environment may impact their development. While little is known about the signals that are important for the normal development of this system, the implications for abnormal development are very broad, including changes in food preferences and increased drive to seek palatable or more “rewarding” or “comforting” foods, as is observed with drug and alcohol addictions . Maternal obesity in rodents, induced by high fat diet (HFD) feeding, has recently been shown to have long term effects on the VTA-dopamine system . In these studies HFD offspring, that were weaned and maintained on a standard chow diet, had increased tyrosine hydroxylase expression in the VTA and increased DA content within the nucleus accumbens. Furthermore, the HFD offspring were relatively and significantly insensitive to amphetamine induced activity, compared to control offspring. What components of the HFD exposure or of the maternal phenotype (obesity, hyperinsulinemia, hyperleptinemia, etc) responsible for these effects on the DA system were not determined; however the changes the HFD offspring point to potential life-long changes in behavior.
Though leptin is largely considered an anorectic hormone, there are dynamic changes in leptin expression and levels during the pre- and postnatal period in rodents that suggest a more critical role in early developmental processes. As mentioned above, although neurons within the ARH can respond to leptin early in the postnatal period, as indicated by pSTAT3 activation, these signals are not transmitted to downstream sites . Therefore, leptin can not modulate food intake or energy expenditure through actions in the ARH until afte the 3rd postnatal week . However, there is a sizable (more than 5 fold) surge in leptin production and secretion during the 2nd postnatal week that is independent of the amount of fat present , suggesting that leptin must be playing some development role. The most striking demonstration of a functional role of leptin as a neurotrophic factor is its importance for the normal development of ARH projections. The obese leptin deficient Lepob/Lepob mouse has a significant impairment in the development of the both ARH-NPY/AgRP and ARH-POMC projections . While it has not been systemimatically investigated, this deficiency is not apparent in some other hypothalamic projections (such as the DMH). Furthermore, treatment of Lepob/Lepob mice with exogenous leptin specifically during the early postnatal period, but not in adulthood, normalizes the development of these projections, indicating that there is a critical window for leptin actions . Further demonstration that the postnatal leptin surge is important for “programming” metabolic systems, recent work using a leptin antagonist specifically during the time of the postnatal leptin surge results in adult-onset leptin resistance as well as increased overall adiposity and hyperleptinemia, in addition to greater propensity to obesity and diet-induced weight gain . This study did not investigate the effect on the development of the hypothalamic circuitry. Unfortunately, the story is not as simple as the leptin surge being important for the normal development of metabolic systems. The studies in the Lepob/Lepob mouse indicate that no leptin is definetly “bad” for the development of ARH-melanocortin cirtcuits; however it is now apparent that too much or too little and the timing of the leptin surge can all affect the development of these circuits. For example, our group and others have demonstrated that early postnatal overfeeding, using a “small litter” manipulation, can result in an exaggerated leptin surge that results in early leptin resistance in ARH neurons (unpublished observations)[176; 177] (This model is discussed in greater detail below). Importantly, this leptin resistance persists past the point that endogenous leptin levels normalize. Bouret and Simerly have recently demonstrated that this rather subtle overfeeding manipulation early in life also results in reduced development of ARH projections, similar to that observed in the Lepob/Lepob mouse (Bouret and Simerly, personal communication). Furthermore, Yura et al. demonstrated that offspring from dams with food restriction during pregnancy have a premature leptin surge during the early postnatal period . These offspring develop a hypersensitivity to a HFD displaying increased weight gain and adiposity. Surprisingly, mimicking the premature leptin surge with exogenenous leptin administration resulted in a similar hypersensitivity to the HFD. Both groups of offspring showed abnormalities as adults in neuropeptide terminal densitites in the PVH; thus, simply shifting the timing of the leptin surge is sufficient to cause developmental abnormalities in the hypothalamic circuitry.
To date, there is no evidence supporting the importance of leptin for the development of ARH circuitry in higher species. In the NHP, leptin levels are very low to undetectable until the middle of the 3rd trimester, which correlates with the late development of adipose tissue. This is the case even in fetal offspring of pregnant animals that are obese and hyperleptinemic . Thus, significant leptin in the developing fetus is not evident until after the initiation of the ARH-NPY/AgRP projections . However, ARH-POMC projections are more delayed in their development; thus leptin could have more of an impact on this specific projection. Furthermore, leptin still may play a role in the refinement or arborization of the projections [180; 181; 182], which occurs closer to parturition.
In addition to its other roles, insulin is also an important neurotrophic factor during development along with insulin-like growth factor (IGF) and brain derived nerve factor (BDNF) . The local production of insulin within the hypothalamus and high level of receptor expression confirms the importance of insulin signaling for neuronal maturation in addition to its neuromodulatory functions. Insulin immunoreactivity and insulin receptors are present very early during fetal life . However, the function of insulin within the hypothalamus has not been emphasized in the literature predominately because of the difficulties in separating central actions versus the robust peripheral expression and activity of insulin.
Many animal models have been developed to investigate the effects of dietary manipulations on fetal and postnatal development and growth. Models resulting in an adult obese phenotype have relied on maternal overnutrition and undernutrition during gestation. These models change the hormonal and nutrient status of the mother that are then reflected in fetal growth and development. Conversely, models of placental insufficiency result in suppressed blood flow and hence nutrients to the growing fetus in the absence of maternal health status being altered. This also leads to maturity-onset obesity due to a series of catch-up growth periods. Alternatively, obesity can also result from neonatal overnutrition or undernutrition. This can be accomplished by creating either very small litters of several pups where an abundance of milk is available, or very large litters whereby milk availability is spread thinly among litter mates, thus limiting developmental growth but again spurring catch-up growth phases. Other groups have formula fed pups with milk of varying composition, thus controlling the nutritional quality of the milk. These are less natural models but do mimic the altered nutrition that might be received in formula-fed infants. In the following sections, each of these types of manipulations will be described in light of their impact on the origin of developmental obesity; however, we will focus primarily on maternal and neonatal overnutrition because of their relevance to the current epidemic of childhood obesity. Furthermore, because of the extensive number of such studies in some of these areas is not possible to discuss each one in detail, thus we will focus on recent literature in the area.
Models in which overall content and quality of the maternal diet is altered prior to and during gestation demonstrate long-lasting effects on energy balance in the offspring. While the results from many different groups all support an effect by maternal HFD, there is often varying effects on brain neurochemistry, hormone levels and the extent of obesity as well as sex differences. This is likely attributed to the different types of diets (either different types or percent of fat in the diet), different exposure periods (i.e., chronic vs acute) as well as different rodent models (mouse vs rat). A recent study by Patel’s group using rats has shown that fetuses exposed to maternal HFD have elevated serum leptin and insulin levels as well as increased expression of NPY, AgRP, POMC and MC4R in the hypothalamus . The reason for elevations in both orexigenic and anorexigenic neuropeptides is unknown, but suggests that this nutritional manipulation is driving some common excitatory signal within the hypothalamus, possibly increased cytokines and/or corticosterone, that are known to be elevated in response to obesity. It should be recognized that while we understand much of how different hormonal (i.e., leptin, insulin, glucocorticoids and cytokines) and nutrient (i.e., glucose and fatty acids) signals modulate the activity of the ARH circuit in the adult animal, very little is known about how these signals impact neurons during fetal and neonatal life. It is entirely possible that receptor signaling within these different neurons changes throughout development. While the long-term implications on the development of the hypothalamic signal was not investigated the Patel study  it is reasonable to speculate that this chronic abnormal drive on these neurons during critical periods of development will impact the manner in which this circuit is programmed. Indeed, others have shown that offspring in this type of model have higher insulin/glucose ratios, higher body fat percentage, and greater triglyceride levels at birth , and go on to develop obesity [173; 187]. Others have also used a “junk food diet” which is high not only in fat but preferentially higher in sugars and salt. This diet results in enhanced adiposity specifically in female offspring as well as a host of changes in metabolic parameters . The most striking results and strongest evidence of maternal HFD causing long-term reprogramming within the hypothalamus comes from studies performed by Leibowitz and colleagues in which they deomonstrated that maternal HFD consumption stimulated neuroepithelium to proliferate and preferentially become orexigenic peptide-producing neurons of the LHA in the offspring . Furthermore, there was increased expression and increased cell numbers of several neuropeptide systems including galanin, dynorphin and enkephalin in the PVH and MCH and orexin in LHA . From these studies it is suggested that the increased neurogenesis in the developing fetus, stimulated by the maternal HFD, would be predicted to cause a broad range of behavioral and physiological complications. Supporting this conclusion, Walker and colleagues , as highlighted earlier, have demonstrated that maternal HFD consumption has long-term effects on the VTA-dopamine system, causing desensitization to amphetamines.
Most studies in this field focus on one type of manipulation (i.e., maternal HFD); however, in real situatins individuals are faced with multiple types of stressors. Recent studies by Tamashiro’s group have investigated the interaction of different types of insults during pregnancy. In these studies, pregnant animals were either exposed to a HFD, repeated novel variable stress or both . Compared to controls, offspring from all three experimental groups appeared normal if they were raised (after weaning) on standard chow diet; however, offspring from all three groups displayed a hypersensitivity to a HFD, in that they exhibited exacerbated weight gain and insulin resistance compared to control offspring. Although predicted, it was surprising from these studies that there was not a significant interaction between the maternal variable stress and the HFD; however, it is possible that the levels of insult used in each model (60% HFD or 3 weeks of repeated variable stress) elicited maximal effects individually. It is more likely that an interaction would have been observed if a diet lower in fat or a more modest stress regime had been used (K. Tamashiro, personal communication).
The effect of maternal overnutrition has also been studied in higher species. In the sheep, infusions of glucose directly into the fetus for 10 consequtive days, mimicking maternal hyperglycemia, or overfeeding the pregnant sheep results in increased insulin levels, adiposity and increased POMC gene expression in the fetal offspirng [105; 106; 191]. Interestingly, the effect of both fetal glucose infusion and maternal overnutrition appeared to have a relatively specific effect on the POMC system, since NPY, AgRP and CART were unaffected by these treatments. The long-term implilcations of these fetal changes have not been reported.
Our group has used a NHP model of maternal overnutrition, using a chronic maintenance on a diet high in fat and calories . Like humans, some adult animals became obese and insulin resistant on the HFD, while others maintained a normal body weight, adiposity and insulin sensitivity, even after 4-6 years on the HFD. The primary findings from these initial studies were the observation that 100% of the fetal and neonatal offspring developed signs of fatty liver disease, as evidenced by increased liver triglycerides, oxidative damage and activation of the gluconeogenic pathway. Furthermore, we demonstrated that the fetal offspring also have elevations in a broad range of circulating cytokines. The offspring from the HFD animals also displayed increased adiposity and liver triglycerides through the early postnatal period. What was concerning and surprising from these studies was that the fetal lipotoxicity was evident whether the mother was obese or lean, suggesting that the high fat/high caloric diet alone was significantly contributing to this effect. It remains to be deteremined to what extent fetal lipotoxicity leads to abnormalities in the development of hypothalamic circuits; although unpublished data from our group indicates that there are abnormalities in the hypothalamic melanocortin system as well as the serotonin system in the dorsal raphe (Grayson and Grove, unpublished observations). To date, no long-term studies have been done in any higher species; however, we can speculate with some confidence that maternal overnutrition or HFD consumption, as has been observed in the rodent models, can and will have long-term effects on the function and efficiency of metabolic systems in the offspring.
There are numerous studies investigating the effects of nutritional manipulations during the early postnatal period in rodents, primarily by groups led by Andres Plageman [192; 193; 194; 195] and by Mulchand Patel [196; 197; 198; 199; 200]. Plagemann’s group has performed extensive studies on the metabolic characterics and hypothalamic programming of pups reared in small vs normal litters. The small litter model (i.e. 3 pups per litter) results in reduced competition for nursing time and, thus, increased milk consumption and increased weight gain during the early postnatal period [201; 202]. These animals are hyperphagic and hyperleptinemic and hyperinsulinemic as early as P10 . These offspring maintain their increased body weight throughout life even when weanded on to a standard chow diet; however, the adult phenotype is rather mild, with the animals being only about 10% heavier and having normal leptin levels . In spite of this mild phenotype young adult offspring from small litters have been shown by Plagemann’s group to have a shift in the hypothalamic signaling to insulin and leptin as well as numerous neuropeptide and neurotransmitter systems [176; 177; 205; 206; 207; 208; 209; 210; 211]. Furthermore, our group has demonstrated that these postnatally overfed animals have reduced adaptive thermogenesis that is likely secondary to abnormalities in regulation of sympathetic outflow to brown adipose tissue . As mentioned earlier, recent studies from our group have shown that the postnatally overfed animals develop leptin resistance in the ARH, but not the DMH, during the postnatal period (unpublished observations). This may not be surprising since these animals are hyperleptinemic during this period; however, the leptin resistance in the ARH persists into adulthood, even though leptin levels normalize. While these animals have a relatively mild metabolic phenotype as adults, they do display an exacerbated phenotype when given a HFD. The increased sensitivity to the HFD appears to be partially due a prolonged initial hyperphagic response when placed on the HFD. Most animals normalize their caloric intake within two days of being placed on a HFD; however, the postnatally overfed animals take 7-9 days. Furthermore, they fail to increase their activity when placed on the HFD and have a slightly lower BAT temperature, indicating that early life overnutrition has functionally altered the tone of the neurons within the ARH to overall sensitivity to leptin and insulin and perhaps other nutrients such as fatty acids and glucose. Their altered ability to respond to incoming signals is consistent with abnormalities in the development of the ARH circuits, as mentioned earlier (Bouret and Simerly, personal cummication).
Patel’s groups, using the “pup in a cup” model, have investigated the effects of milk diets either high in lipids (mimicking natural milk) or high in carbohydrates [196; 197; 198; 199; 200]. While this artificially raised animal model has its limitations, the importance of these studies is that the investigators are simply altering fat and carbohydrate levels, while the caloric intake and other nutrients are kept constant. In this model, pups that are reared on a high carbohydrate milk throughout the postnatal period are relatively hyperphagia and develop obesity even when weaned onto a standard chow diet . The hyperphagia and obesity have been attributed to the increased expression of several orexigenic neuropeptides and decreased expression of anorexigenic systems; these changes are observed as early as P12, and persist into adulthood. It is worth noting that this group has compared pups raised artificially on a milk diet high in fat with pups naturally raised and reported no significant differences in the hypothalamic gene expression or the metabolic phenotype. These studies indicate that simply changing nutrient content, without changing caloric intake, can cause significant reprogramming within the hypothalamus, increasing the susceptibility to obesity and diabetes. However, it still remains to be determined to what extent the hypothalamic circuits or wiring are altered. This group has also shown that high carbohydrate milk reared female offspring, when raised to adulthood and bred, pass on their metabolic abnormalities to their offspring . However, this may not be surprising since the adult pregnant females display characteristics consistent with gestational diabetes, which itself is well known to cause complications in the developing fetus.
In studies in which maternal nutrient restriction is imposed during early and mid-gestation, obesity and energy balance dysregulation also results. This in many ways is paradoxical considering that undernutrition also results in obesity. However, though the result is similar, the mechanism may indeed be different. Offspring of dams placed on a 50% food restriction during gestation have a dampened leptin surge during the first two postnatal weeks in comparison to control animals and have suppressed POMC as well as a reduction in endorphin-ir at weaning [215; 216]. These animals show decreased body-weight during early postnatal life. Since leptin has been shown to be crucial to the maturation of ARH circuits, the dampened leptin surge likely results in disruptions in the development of the hypothalamic circuits. Pregnant dams exposed to low protein diets during pregnancy give rise to offspring who have a preference for fatty foods as opposed to high protein and carbohydrate . By reducing caloric intake by the pregnant dam, the sparing that occurs to support fetal growth appears to result in wiring that favors fuel storage, or increased metabolic efficiency. This is consistent with the thrifty phenotype hypothesis.
One of the most widely used models of neonatal undernutrition is to alter pup milk availability by placing an extraordinary demand on maternal milk production through maintenance of very large litter (14-24 pups) [218; 219; 220]. This model describes early stunted growth and overall lower body-weight early in life with catch-up growth and late-onset insulin resistance and glucose intolerance [221; 222]. These animals are thus exposed to inadequate nutrition during the time of hypothalamic development and have dampened levels of leptin during this critical time period.
Intrauterine growth retardation (IUGR) is a common gestational condition that results in increased risk of infant morbidity and mortality. In animal models IUGR is accomplished through bilateral uterine artery ligation resulting in reduced blood flow to the placenta and placental insufficiency. This model has been characterized in the rodent as well as the sheep. IUGR rodents become glucose intolerant and insulin resistant at an early age which eventually results in diabetes . In the sheep, IUGR results in variety of metabolic alterations in the offspring which include accelerated “catch-up” growth after birth, reduction in lean body mass and adult obesity [224; 225; 226].
From the studies discussed in this review it is clear that early life events (prenatal and postnatal) can significantly impact the development of metabolic systems, including hypothalamic circuitry, causing long-term consequences on body weight and energy homeostasis in adult life. It seems logical that the maternal environment should be a key signal for the developing fetus to provide programming for the adaptations to shifting nutritional environments. However, when developing fetus is impacted by a strong acute signal this can lead to increased health risks later in life; this is the basis of the “Barker Hypothesis”. This part of the equation is clear, but there are many questions that remain to be answered. For example, what are the mechanisms and signals for the development of the hypothalamic circuitry that control food intake and energy homeostasis? Is it simply changes in leptin levels signaling the development of the ARH circuits? The leptin signal, at least for the rodent, appears to be somewhat selective for the ARH circuits. Are there yet unidentified metabolic signals that are critical for the development of other hypothalamic or brainstem circuits involved in energy homeostasis? Undoubtedly. Much focus has been applied to the function and development of the ARH circuits; more effort needs to be spent on other areas of the brain as well as how the complete integrated circuit develops, including peripheral metabolic systems.
Why does overnutrition and undernutrition during the prenatal and postnatal periods result in similar adult phenotypes - overweight/obese and diabetic? Are similar mechanisms involved? While the thrifty phenotype hypothesis can explain undernutrition early programming adaptations, it is difficult to apply this to the situations of overnutrition. This paradox has been shown for many species including rodents, sheep, NHPs and humans. Our studies in the NHP have indicated that maternal HFD results in increased circulating cytokines in the developing fetus (termed fetal lipotoxicity) which can cause broad oxidative damage. Furthermore, we have demonstrated that this model has reduced placental blood flow (unpublished observations), which is similar to the placental insufficiency models in sheep and rodents. Does both undernutrition and overnutrition cause similar abnormalities in placental function? This is a definite possibility that warrants further investigation.
A final consideration; while we have and will continue to gain important insight using rodent species, it is important to consider the significant species differences in the development of metabolic systems. It still remains to be determined if we can completely model the complex circuitry and function of the human using rodents. This may especially be true for complex behaviors. Ultimately the goal is to understand the underlying cause and consequences of the dramatic increase in childhood obesity.