Neurodevelopmental Programming of Adiposity: Contributions to Obesity Risk

Abstract

This review analyzes the published evidence regarding maternal factors that influence the developmental programming of long-term adiposity in humans and animals via the central nervous system (CNS). We describe the physiological outcomes of perinatal underfeeding and overfeeding and explore potential mechanisms that may mediate the impact of such exposures on the development of feeding circuits within the CNS—including the influences of metabolic hormones and epigenetic changes. The perinatal environment, reflective of maternal nutritional status, contributes to the programming of offspring adiposity. The in utero and early postnatal periods represent critically sensitive developmental windows during which the hormonal and metabolic milieu affects the maturation of the hypothalamus. Maternal hyperglycemia is associated with increased transfer of glucose to the fetus driving fetal hyperinsulinemia. Elevated fetal insulin causes increased adiposity and consequently higher fetal circulating leptin concentration. Mechanistic studies in animal models indicate important roles of leptin and insulin in central and peripheral programming of adiposity, and suggest that optimal concentrations of these hormones are critical during early life. Additionally, the environmental milieu during development may be conveyed to progeny through epigenetic marks and these can potentially be vertically transmitted to subsequent generations. Thus, nutritional and metabolic/endocrine signals during perinatal development can have lifelong (and possibly multigenerational) impacts on offspring body weight regulation.

Graphical Abstract

Graphical Abstract.

Essential Points.

• Maternal obesity has become more prevalent in recent decades and can alter lifetime obesity risk in successive generations of progeny

• Mouse studies indicate that overnutrition of pups can influence the development of hypothalamic, brain stem, and hedonic circuits critical in energy homeostasis

• Various hormones—including leptin, insulin, ghrelin, amylin, and glucagon-like peptide-1—may have direct effects on adult adiposity via the effects on developing feeding circuits

• The nutritional environment during development may be conveyed to progeny through epigenetic marks which can be vertically transmitted to subsequent generations

• Experiments investigating the impact of nutritional environment during early postnatal brain development in mice are likely relevant to late gestation in humans

The prevalence of childhood obesity has increased dramatically over the past 4 decades; its prevalence is currently 18.5% for children between 2 and 19 years of age (1). The long-term health consequences of obesity—increased risk of developing cardiovascular disease, type 2 diabetes (T2D), stroke, and some cancers—represent staggering current and projected burdens on the economy and society (2-4). A newly revealed complication of obesity and its negative effects on the global health care system is that obesity is a significant risk factor for severe disease and complications resulting from SARS-CoV-2 infection (5). In addition, the implementation of lockdown measures during the global COVID-19 pandemic (ie, school and businesses closures) aimed at halting the spread of SARS-CoV-2 has led to increased weight gain in adults (6-8) and children (9, 10) by increasing sedentary behaviors and changing eating habits (11, 12). Increased adiposity in children puts them at higher risk of obesity in adulthood. Since maternal obesity per se (as distinct from shared genotype) increases the risk of obesity in offspring, the Covid-19 pandemic will likely affect obesity prevalence in progeny of these gestations, and possibly in subsequent generations. In addition to imposed economic burdens, obesity has profound effects on quality of life as well as life expectancy (13), especially in younger adults (14, 15). Individuals with severe obesity have an estimated 5- to 20-year reduction in life expectancy (14). Obese adolescents are 22-fold more likely to be obese as adults and efficacious obesity treatments are limited; therefore, prevention of early obesity is critical to reduce the disease burden of this epidemic (16, 17).

Perinatal Programming of Body Weight: Evidence From Humans

Concomitant with large increases in obesity prevalence generally, maternal obesity has become common and could have long-lasting impacts on future generations. Genetic predisposition is an important contributor to obesity risk. Heritability studies suggest that ∼40% to 70% of the variation in adiposity among individuals within a shared environment reflects genetic factors (18-20). Body mass index (BMI)–associated genetic signals are highly enriched for variants that are in or near genes expressed in the central nervous system (CNS) (21, 22). With the advent of ’ ’omics level genetic susceptibility scoring, relative obesity propensity can be accurately scored using 2.1 million single nucleotide polymorphisms (23). These polygenic risk scores have been used to stratify populations to isolate the effects of single gene mutations on body weight (24), coronary heart disease, and breast cancer (25), as well as to identify genetic indicators of response to pharmaceutical interventions (26).

The rapid increase in obesity during the past 50 years cannot be explained by gene sequence changes. Transmission of altered methylation/acetylation gene signatures—in response to environmental influences—is a possibility discussed below. In many countries, the current environment is highly obesogenic by virtue of a combination of readily available highly palatable and energy-dense foods and sedentary lifestyles. While there is a lack of support for significant declines in physical activity in the last 4 decades (27) and it is difficult to accurately measure energy intake, 1 or both of these factors must be driving secular increases in obesity. Epidemiological studies as well as mouse models suggest that obesity and metabolic disease originate not only from genetic and environmental factors, but also from their interplay with the developmental milieu. Identifying mechanisms possibly linking the perinatal environment to future health outcomes is important to the prevention of obesity and obesity-related diseases.

The Dutch Hunger Winter of 1944-45 provided a tragic opportunity to investigate the impact in adulthood of gestational undernutrition on disease risk. Fetuses exposed to famine during the first 2 trimesters of gestation had a higher incidence of obesity and cardiovascular disease in adulthood (28). Undernutrition during late gestation resulted in lower birth weight and those individuals continued to have lower rates of obesity (28, 29). Subsequently, David Barker proposed the “Fetal Origins of Adult Disease” hypothesis in which he postulated that adverse or suboptimal perinatal developmental environment leads to increased risk of disease in adult life. Barker emphasized a relationship between low birth weight and hypertension and coronary heart disease risk in adulthood (30); subsequently, numerous epidemiological studies have demonstrated that low birth weight is associated with increased risk of metabolic disease (reviewed in (31)). As a mechanism, Barker's “thrifty phenotype” hypothesis proposed that developmental programing by undernutrition primes the organism for low energy intake through metabolic adaptations that maximize survival in food-scarce environments (32). However, in an environment in which access to high-energy food is plentiful and easily accessible, the anticipatory adaptations can become detrimental to the individual's health.

At the time of Barker's observations, investigators were focused on the influence of poor nutritional state, stress, or disease of the mother on the outcomes in progeny as adults. However, over the intervening 3 decades, as populations have become more obese, focus has shifted towards developmental programming resulting from maternal overnutrition (33-35). There is currently more interest in the effects of excess materno-fetal nutrients on the long-term consequences in the offspring. Such developmental programming is the focus of “Developmental origins of health and disease” (36), predicated on the idea that environmental influences during critical periods of development and growth (which includes the gestational period through early postnatal development) may have adverse consequences for an individual's future health. These consequences presumably reflect both static structural changes in the brain, islets, adipose tissue and other organs as well as “imprinting” of molecular/biochemical pathways mediating a range of physiological processes such as blood pressure and lipid homeostasis. But the precise nature and mechanism of these putative effects remain poorly understood.

Prepregnancy obesity in US women increased from 26.1% to 29% between 2016 and 2019 (37). Excess adiposity leads not only to adverse pregnancy outcomes (38) (eg, increases in maternal mortality (39), increased risk of miscarriage (40), gestational hypertension (41), preeclampsia (41-43), gestational diabetes (GDM) (41, 44), stillbirth (43), and cholecystitis (45)), but also imperils the developing child (including large for gestational age [LGA] babies, fetal defects, congenital anomalies and perinatal death) (46). GDM, hypertension, and preeclampsia contribute to a suboptimal intrauterine environment. Maternal obesity (“controlling” for shared genotype) predisposes offspring to obesity, diabetes, hypertension, and cardiovascular disease (47-49). Infants born to obese mothers are more likely to be LGA (birth weight >90th percentile for gestational age and sex) or macrosomic (birth weight >4000 g) (43, 50, 51), which puts them at a higher risk of obesity later in life (52, 53). Obesity of both parents increases the risk of obesity in the offspring (54), but maternal pregravid BMI is the strongest predictor of childhood obesity (55); maternal obesity is a stronger determinant than paternal obesity of progeny neonatal BMI (56), as well as offspring BMI at 2-3 (56), 5-7 (57), and 14 (33) years of age. Excessive weight gain (based on Institute of Medicine gestational weight gain recommendations) during pregnancy is associated with higher infant adiposity (58, 59). Infants of prepregnancy obese women have the highest fat mass regardless of degree of maternal weight gain during gestation (59).

LGA babies have increased risk of obesity in adulthood (60, 61), and rapid weight gain in infancy is associated with later risks of obesity and metabolic disease, particularly insulin resistance (IR) (62, 63). A meta-analysis of 66 studies including 643 902 individuals aged between 1 and 75 years found that infants born at high birth weight (>4000 g) are at higher risk of overweight in adulthood (odds ratio [OR] 1.66). High rates of postnatal weight gain, termed “catch-up” growth, often occur in small for gestational age (SGA) infants. A systemic review of data reported from 1927 to 1994 found that birth size (infants at the highest end of the distribution for weight or BMI) and growth rate during infancy (babies who grow rapidly) are related to risk of obesity in children and adults (64). Rapid infant growth correlated with higher rates of obesity in childhood, adolescence, and early adulthood (64). A subsequent Finnish study that correlated postnatal weight and height velocity from birth to 2 years of age with metabolic outcomes at 31 years of age found that higher peak weight velocity was associated with increased incidence of adult obesity (by BMI and weight circumference) and hypertension, while peak height velocity correlated with blood pressure and weight circumference (65). A large-scale population-based study in Germany found that the most rapid BMI acceleration in obese adolescents occurred between 2 and 6 years of age and that 90% of obese 3 year olds remained overweight or obese in adolescence (66). These longitudinal studies clearly point to obesity in early childhood being an important predictor of obesity in adolescence or adulthood (67). A systematic review of over 200 000 children followed to adulthood showed that obese children and adolescents were around 5 times more likely to be obese in adulthood than those who were not obese (68). A Mendelian randomization study found that genetic predisposition to higher childhood BMI (study used 15 single nucleotide polymorphisms (associated specifically with childhood rather than adult obesity to test effects that are specific to early-life BMI) was associated with increased adult BMI (the effect size: 1 SD increase in childhood BMI associated with a 0.587 SD increase in adulthood BMI) (69). Importantly, Mendelian randomization studies demonstrate that childhood obesity (controlled to some extent for genetic risk) has a causal association with adult T2D (OR up to 1.83) and adult coronary heart disease (OR 1.38) (69, 70), suggesting that adult disease is driven by childhood obesity independently of genetic obesity risk.

Diabetes (type 1 diabetes [T1D], T2D, and gestational [GDM] diabetes) and impaired glucose tolerance during pregnancy have been linked to increased infant birth weight and more specifically, infant adiposity (71-74). Infants of diabetic mothers are born LGA and show a higher BMI throughout childhood (75) and adulthood (76). Intense treatment of GDM reduces the likelihood of LGA infants (77, 78) and tight glycemic control in gravidas with T1D was shown to reduce the incidence of LGA in some studies, but others indicate that even with well controlled maternal glucose, infants of mothers with T1D are at increased risk of macrosomia (79, 80). The precise glycemic targets required to mitigate LGA are debated, and may differ depending on the type of diabetes (81). In women with GDM, pregravid obese or overweight BMI increased risk of macrosomic baby even when hyperglycemia was well controlled with diet (82). In overweight or obese women with well-controlled GDM treated with insulin, the macrosomia risk was comparable to normal weight gravida (82). It is unclear by what mechanisms maternal insulin treatment may decrease neonatal macrosomia rates other than its effects on glycemia. In Pima Indian women who gave birth to children before and after a T2DM diagnosis the sibling who was exposed to a diabetic intrauterine environment had a greater risk of obesity and diabetes than offspring born before the mother was diagnosed with diabetes (83). While this study found no increase in diabetes propensity associated with paternal diabetes status (83) there have been other studies reporting that paternal obesity and diabetes status can influence susceptibility to future metabolic disorders in offspring (84). These data indicate that in addition to the genetic predisposition, intrauterine exposure to hyperglycemia (and attendant metabolic changes) per se contributes to the risk of obesity in the offspring.

GDM is often associated with maternal obesity and characterized by hyperglycemia and hyperinsulinemia. While insulin does not cross the placenta (85, 86), women with diabetes have increased circulating glucose concentrations leading to higher transplacental glucose transfer. Increased fetal glucose concentration stimulates fetal insulin production leading to hyperinsulinemia and fetal overgrowth (87). The Hyperglycemia and Adverse Pregnancy Outcome (HAPO) study of 23 316 mother–newborn pairs found a linear relationship between increased maternal blood glucose measured at 24 to 32 weeks of gestation and cord blood C-peptide and an association of each of these parameters with birth weight (88).

Progeny of women with prepregnancy diabetes are at an increased risk for congenital malformations (organ teratogenesis) with some studies reporting up to 10-fold greater risk (89-91). The highest occurrence of birth defects associated with prepregnancy diabetes is congenital heart disease but other, noncardiac malformations include nervous system (ie, neural tube defects), respiratory, skeletal tissue, and urinary tract abnormalities (91, 92). Compared with pre-existing diabetes, GDM (usually diagnosed in the third trimester) is associated with fewer birth defects with weaker associations since most birth defects develop in the first trimester. Freinkel hypothesized that abnormal fuel levels (not limited to glucose) during gestation can not only induce congenital malformations but also have far-reaching implications on offspring physical and behavioral development (93). He coined the term “fuel-mediated teratogenesis,” proposing that the maternal fuels (in diabetes: elevated glucose, amino acids, ketones, and lipids) influence the development of metabolic tissues including pancreatic beta-cells, adipose tissue, and CNS cells undergoing differentiation, proliferation, and maturation in utero. It is possible that structural changes in the CNS resulting from gestational exposure to these abnormal fuel levels cause increased body weight throughout life, while higher fat mass at birth does not mediate adult obesity but could serve as an early marker of future obesity.

Direct administration of insulin via minipumps to fetal rhesus monkeys for 3 weeks prior to birth increased birth weight (by 23-27%), and increased weights of the placenta, heart, liver, and spleen, while not affecting the mass of other organs including brain and skeletal muscle (94). Similarly, a single insulin injection into a fetal rat at embryonic day 18.5 was sufficient to increase in body weight by 7% to 12% (95, 96); this increase persisted for the first 2 weeks of life (95). However, supplementing insulin to the pregnant rat dam (also through subcutaneous minipump from day 14 of gestation until parturition), resulted in intrauterine growth retardation (IUGR) of the fetuses, presumably due to decreased availability of metabolic substrates for growth (dams were hypoglycemic due to hyperinsulinemia with no difference in weight gain compared to control dams from baseline). The pups were hypoglycemic and hypoinsulinemic during gestation and after birth (97). These data suggest that fetal hyperinsulinemia is the primary driver of fetal overgrowth. Consistently, infants of diabetic or hyperglycemic mothers, display elevated cord blood insulin (88), and this fetal hyperinsulinemia is strongly associated with macrosomia at birth (88, 98, 99). A heterozygous loss of function mutation in glucokinase (GCK)—the most common cause of maturity-onset diabetes of the young (MODY)—is characterized by mild hyperglycemia resulting from a defect in glucose sensing by beta cells. Individuals with GCK-MODY have reduced concentration of insulin at any given glucose concentration (100). Gravidas with GCK-MODY have mildly elevated glucose but the body weight outcome in the offspring is dependent on the offspring genotype: the sibling who does not inherit the GCK mutation has a higher birthweight than a sibling with the mutation presumably due to decreased fetal insulin secretion in utero in the sibling with the GCK mutation. Offspring with paternally transmitted GCK mutations born to an unaffected mother have reduced birthweight due to intrauterine hypoinsulinemia (101).

Hyperinsulinemia and IR in gravidas with GDM has been linked to increased placental growth during early pregnancy (102, 103), while fetal hyperinsulinemia is thought to drive the increased placental growth (104) in later pregnancy which is consistent with higher expression of insulin receptors (InsRs) on the maternal side of the placenta early on and a shift of InsR expression to fetal side in late gestation (105). In offspring born to mothers with GCK-MODY, those with GCK mutation had reduced placental weight compared with those offspring without the mutation. These results implicate a role for insulin in placental growth. Fetal hyperinsulinemia increases placental vascularization (106), which is a mechanism of placental adaptation to metabolic stress.

Animal studies of in utero insulin administration indicate insulin's direct role in fetal growth. In humans, since obesity in gravidas is associated with LGA infants, and maternal obesity is associated with T2D and GDM in the gravida, it is challenging to determine the degree to which the intrauterine hyperglycemia per se drives increased fetal growth independently of other aspects associated with the maternal obese state. Studies controlling for maternal prepregnancy BMI found that the offspring of mothers with GDM had an increased incidence of overweight throughout childhood (2, 8, and 11 years of age (107)) compared with offspring of mothers with T1D or controls without diabetes; maternal obesity was a strong predictor of overweight in children (107). These studies highlight the importance of maternal and fetal hyperinsulinemia in predisposing to later obesity in progeny.

While disentangling genetic from development–environmental contributions to obesity outcomes is challenging—particularly in humans—weight gain between successive pregnancies is associated with an increased risk of LGA neonates, while weight loss after an LGA birth is correlated with a lower risk LGA in the next pregnancy (108-111). Additional evidence implicating developmental environment rather than shared genetic factors in programming of obesity comes from women who had children both before and after bariatric surgery. Obesity is reduced in offspring (aged 2-26 years) born after maternal bariatric surgery relative to siblings born prior to surgery (112-114), and the latter progeny show improved cardiometabolic markers into adolescence (113). A retrospective, Australian population-based study compared birthweight outcomes in women who had bariatric surgery between first and second pregnancy and showed reduced rates of LGA infants (OR 0.63) in the second pregnancy (115). These data suggest that the developmental environment contributes directly to increased offspring obesity. Differential gene methylation is a potential mechanism mediating these obesity differences (116).

The underlying cellular and molecular mechanism(s) for effects of maternal “programming” are unclear. Because human data cannot directly address these questions—all such influences are implied by epidemiological evidence and this evidence is confounded by heterogeneous genetics, diet, and lifestyle—animal models of identical genetic background and environmental exposure are relied on. Genetically modified animals have been indispensable in broadening our understanding of molecular mechanisms underpinning the developmental programming of body weight.

Rodent Models of Developmental Programming

Numerous investigations of the effects of nutritional status during early development on energy homeostasis in rodents have confirmed that gross physiological phenotypes are generally consistent—allowing for differences in the timing of neurodevelopment—between humans and rodent models. This correspondence is, of course, critical to the utility of rodent models use in the investigation of the molecular mechanisms by which early development programs adult body weight.

Perinatal Underfeeding

Rodent models of perinatal undernutrition primarily include maternal total calorie restriction (caloric intake usually decreased by 30-70%) or protein restriction (isocaloric diet containing 8-10% kcals from protein vs 20% in normal chow) during gestation and/or lactation (117-122), and increasing litter size to reduce breast milk availability per offspring (123, 124) (Fig. 1). Maternal calorie or protein restriction during gestation (durations differ from 1 to 2 weeks to the entire 21-day pregnancy) leads to significantly lower body weight of fetuses at birth (IUGR) than rat dams fed ad libitum (125). IUGR is associated with metabolic phenotypes related to glucose-insulin homeostasis including beta cell dysfunction, IR, and glucose intolerance in adult rodent offspring (126-130) (Fig. 1).

Figure 1.

Effects of diet composition and caloric underfeeding and overfeeding during gestation and/or lactation on the programming of body weight in the offspring. Note that different conditions may lead to similar outcomes as indicated by the arrows pointing to the effect boxes. Abbreviations: IUGR, in utero growth restriction; BW, body weight; DIO, diet-induced obesity; HFD, high-fat diet. Figure created with BioRender.com.

The timing of maternal calorie restriction is critical, as it exerts differential effects on the offspring, and in some models the effects are sexually dimorphic in rats (117, 131) and mice (132). Both male and female rat offspring born to dams undernourished only during pregnancy are smaller at birth than controls. When these rat pups are cross-fostered to dams fed ad libitum they gain weight rapidly (“catch-up” growth) and most studies report increased body weight and fat percentage vs control offspring in adulthood when maintained on a regular chow (117, 118). This catch-up growth is associated with increased susceptibility to diet-induced obesity reported in both mice (122) and rats (133). When maternal undernutrition spans from gestation through the lactation period, the accelerated postnatal growth is prevented resulting in ad libitum-fed offspring that have persistently lower body weight than control animals found in both mice (132) and rats (117, 118, 134). Howie et al reported that under these circumstances female, but not male, rat offspring have a reduction in percentage of body fat at 5 months of age (117). When maternal undernutrition is restricted to the lactation period only, mouse (132) and rat (134) offspring have persistently reduced body weight and fat when subsequently fed regular chow ad libitum. Howie et al found such differences only in female rat offspring (117). Desai et al exposed the progeny nursed by undernourished rat dams to a Western-style diet in adulthood. There was no difference in body weight of the rats compared with controls but males showed a small increase in body fat percent (difference not seen in females) (135).

Similar to maternal calorie restriction, restricting the amount of protein in maternal diet in rats during gestation is sufficient to induce fetal growth retardation (120). Providing these pups with proper postnatal nutrition, either by cross-fostering them with dams fed control diet and/or placing them in smaller litters (for increased milk availability), results in “catch-up” growth, but the resulting pups are more susceptible to diet-induced obesity (reports of consistent effects in both rats and mice) (133, 136-138). When a low-protein diet is enforced throughout rat dam pregnancy and lactation, the offspring remain significantly lower in weight throughout life compared to controls (133, 139, 140). In fact, nursing by dams fed a reduced protein diet only in postnatal life results in reduced preweaning body weight that persists into adulthood found in mice (136) and rats (120, 140, 141).

While there are some inconsistencies among studies assessing the effects of gestational and/or postnatal calorie and protein restriction on body weight in the adult offspring, this is not unexpected given variations in the experimental conditions (including the exact timing) and differences in rodent strains and whether mice or rats were studied. The most consistent phenotype is evident in the offspring that experience intrauterine growth restriction (via caloric or protein restriction of the dam) followed by “catch up” growth during lactation (via cross-fostering to ad libitum-fed dams) constituting a mismatch between the prenatal and postnatal environments. These rats are more prone to weight gain when exposed to hypercaloric (32% fat and 10% sucrose liquid) diet as adults (133). Some reported consequences of prenatal undernutrition that may contribute to subsequent somatic phenotypes include alterations in hypothalamic expression of energy homeostasis genes (142), persistent changes in the plasma metabolome (143), decreased sensitivity to leptin (144, 145), and disturbances in the hypothalamic “wiring” of the proopiomelanocortin (POMC) neurons (121), all reported in rat offspring. Interestingly, the offspring of rat dams that were calorie restricted during gestation and lactation experienced a delay in postnatal leptin surge (more details on the surge in the following sections) (121, 146), whereas those that were born to gestationally calorie-restricted mouse dams showed an augmented and earlier leptin surge, likely a result of accelerated postnatal growth of fat mass (122).

To define the effects of undernutrition on developmental programming of body weight specifically during the lactation period, the amount of milk available per pup is experimentally reduced by expansion of litter size. Offspring raised in large litters experience slower growth and are significantly lower in body weight (up to 35% by weaning) during nursing period compared to offspring reared in normal size litter, a consistent phenotype reported in both mice and rats (124, 147-150). Most studies report that the mouse (123, 148) and rat (147, 151) offspring remain smaller in adulthood, while some find that the offspring display catch-up growth after weaning and reach higher adiposity than offspring reared in normal size litters. Undernutrition (by increasing pups in a litter) results in a delayed postnatal leptin surge of decreased magnitude reported in mice and rats (121, 124, 149) and altered expression patterns of hypothalamic genes involved in the melanocortin axis found in mice (147).

Perinatal Overfeeding

Rodent models of perinatal overfeeding include maternal obesity induced by variations of high-fat diet (HFD) (some of which include sugar) during gestation and/or lactation as well as overnutrition during nursing by litter size reduction. These studies generally suggest that perinatal overnutrition results in obesity and metabolic abnormalities (such as IR or hepatic steatosis) in offspring as adults (Fig. 1).

Increased body weight/adiposity in rodent adult offspring is one of the most consistently associated phenotypes in experiments of imposed maternal obesity during perinatal period. Not surprisingly, characteristic metabolic phenotypes are often associated with obesity in these models, including disturbances in glucose tolerance (in mice and rats) (152, 153), IR (in rats) (153), pancreatic b-cell dysfunction, fatty liver (in mice) (154, 155), increased in systolic (in mice) (155, 156) and diastolic blood pressure (in mice) (152), and decreased hypothalamic leptin and insulin signaling (in rats) (153, 157). Rodent studies suggest that maternal obesity/overnutrition during both prenatal and postnatal (lactation) periods are important in programming body weight in the offspring. Some mouse studies report that maternal HFD-feeding during gestation or lactation each contribute independently to increased adiposity of the offspring, and that exposure during both periods produces even greater increase in adiposity in the adult offspring (152, 156). Others report that high-energy density diets (this study used a “junk” also known as “cafeteria” diets biscuits, marshmallows, cheese, jam doughnuts, chocolate chip muffins, butter flapjacks, potato crisps, and caramel/chocolate bars) during both gestation and lactation periods are necessary to program increased weight and propensity for high-density foods in the adult rat offspring (158). However, limiting maternal HFD exposure to lactation alone may be sufficient and has a greater influence on weight of the adult offspring than gestational maternal HFD found in rats and mice (159, 160). A recent study investigating the impact of increasing levels of fat content in maternal diet (during lactation only) on mouse offspring adiposity and found that ≥41.7% dietary fat compared with 8.3% or 25% resulted in persistent increases in offspring body weight at weaning and at 14 weeks of age when maintained on regular chow (161, 162). HFD (41.7% fat) exposure of 14-week-old mouse offspring nursed by dams fed HFD with fat content ≥41.7% resulted in an increased susceptibility to weight gain which was more pronounced in males than females; interestingly, increasing fat in the maternal diet to 58% or 66.1% had little or no additional impact (161, 162), suggesting that 41.7% of fat in the maternal diet is sufficient to induce the body weight changes in the offspring (Fig. 1).

In the rodent model of postnatal overfeeding via litter size reduction, the remaining pups (eg, n = 3) have greater caloric availability during the suckling period causing an increased growth rate compared with pups reared in normal size (eg, n = 8) litter. The inverse relationship between litter size and pup body weight throughout suckling suggests that in the early postnatal period, immature pups lack satiety signals and do not regulate food intake but rather the amount of ingested milk is solely dependent on availability (163). In fact, a major regulator of energy homeostasis, leptin, does not suppress food intake or body weight in neonatal mice and rats (164-166). In addition to increased weight gain during the lactation period, mouse pups raised in small litters have persistently higher body weights throughout life (up to 23 weeks of age on chow diet) compared to controls and gain more weight when exposed to HFD at 6 weeks of age (167, 168).

While the models of reduced litter size and maternal obesity/HFD feeding differ, primarily in metabolic status of the nursing dams, both of them result in chemical compositional changes to the breastmilk. Rat dams that are fed a high-energy density diet show an increase in caloric density of the breastmilk (∼20% increase on P21 based on calculation from proportions of macronutrients (169)) and elevation of macronutrient concentrations including sugar, protein, triglycerides and cholesterol (153, 157, 169, 170). The data on effects of maternal HFD feeding on milk production are conflicting. Del Prado et al found the volume of milk production to be higher (171) while Bautista et al reported less milk volume per day (170) in HFD-fed compared to control rat dams. Maternal cafeteria feeding (consisting of salami, cheese crackers, and chocolate chip cookies) during different time combinations, including before conception, and during gestation and lactation, alters the composition of milk produced. The largest effect of maternal diet on milk composition in rats was during lactation (172). In cafeteria-fed rat dams, the milk fat content was increased and consequently the caloric density calculated from macronutrient composition was higher (172). Interestingly, even when the diet of dams nursing normal or reduced number of rat pups is the same, in dams with reduced size litter the fat content of milk is elevated while protein and sugar concentrations are largely unchanged (173-175). Caloric density of milk from rat dams nursing small litters was higher by up to 30% (174, 176), indicating that in addition to increased milk volume consumption the increased weight/fat gain of the rat pups reared in reduced-size litters is a result of the increased fat content and caloric density of the breastmilk (174).

Rat and mouse pups experience a leptin surge during which concentrations of circulating leptin are elevated 5- to 10-fold with peak concentrations occurring between postnatal day 7 and 10; these concentrations are independent of fat mass (177). We have since shown that the leptin concentrations in neonatal mice are higher than what would be predicted by their fat mass up to postnatal day 27 (149). While it is unclear by what mechanism this postnatal leptin surge is induced, the nutritional status of the pups during this period affects the timing and magnitude of the leptin surge. In both models of overnutrition described above—either the reduced litter size or rearing by dams fed HFD during gestation and/or lactation in both mice and rats—the leptin surge in the pups starts earlier with an increased magnitude (124, 149, 156, 178) even when plasma leptin concentrations are adjusted for fat mass (149).

The reported effect sizes on body weight in adult offspring of obese dams and small litter size overfeeding are variable. The phenotypic inconsistencies likely result from different experimental design including the timing of HFD feeding or overnutrition. Similar to perinatal undernutrition, the design of overnutrition experiments varies widely across reported studies. These variables include the timing of hypercaloric diet exposure (pregestation, gestation, lactation, or all 3), the diet used (different % of calories coming from fat in HFDs, and different proportions of sucrose), sex of studied offspring, the strain of mice or rats used, and whether the offspring were cross-fostered (this is especially important when the dam was switched from HFD during pregnancy to chow at parturition). Cross-fostering is critical to avoid the effects of postnatal undernutrition which would occur as a result of food intake decrease following a HFD to chow diet switch [179).

Mechanisms Mediating the Developmental Programming of Adiposity

Hypothalamic Development

The arcuate nucleus of the hypothalamus (ARH) is adjacent to the third ventricle and median eminence, a circumventricular organ, all located within the mediobasal hypothalamus with a “permissive” blood–brain barrier access. The 2 best characterized populations of leptin-responsive neurons—POMC and neuropeptide Y (NPY)/agouti-related protein (AgRP)—reside in the ARH and are therefore well positioned to relay circulating peripheral signals of energy homeostasis to other parts of the hypothalamus and CNS. Hypothalamic cells and circuits are sensitive to the impact of obesity and HFD-feeding during development. In rodent models, these effects are conveyed in part by circulating molecules such as leptin, insulin, ghrelin, and amylin. Perturbations in the concentrations of these molecules affect neurogenesis of feeding-critical neurons generated during gestation, axonal outgrowth of neurons participating in feeding circuits during early postnatal period, and/or epigenetic modifications to genes involved in energy homeostasis (180, 181). The pathways predicted to be primarily affected by such perinatal programming are those critical in energy homeostasis, such as those involving leptin- and insulin-responsive POMC and AgRP neurons (182, 183).

With the important caveat that mice are born earlier in neural development than humans, hypothalamic development in rodents begins during gestation and extends well into the postnatal period (Fig. 2). Rodent timing of neurogenesis of hypothalamic neurons, using ³H thymidine labeling, suggests that the majority of neurons are created between E11 and E14 with some differences among nuclei (187, 188). In a more recent study, Ishii and Bouret used bromodeoxyuridine to systematically identify neurogenesis in the hypothalamus, specifically those neurons that are a part of leptin-sensitive circuits. They found that there is a sharp peak of neurogenesis for leptin-responsive hypothalamic neurons that occurs at E12 and that the majority of these neurons are generated between E12 and E16 (189). Expression of Lepr (leptin receptor) is detected in the developing rodent brain at E10.5 (measured in whole brain with reverse transcription polymerase chain reaction); using in situ hybridization, Lepr mRNA expression is observed in the premamillary hypothalamic nucleus from E16.5 and in the ARH and ventromedial hypothalamic nucleus at E18.5 (190). Maternal HFD feeding during pregnancy increases expression of POMC and NPY neuropeptides in E18 fetal whole brains (191). Leptin influences neurogenesis during gestation (192) and lactation (193) in mice. Congenitally leptin-deficient, Lep^ob/ob, mice have smaller brains (194) with fewer neurons at E16 and E18, and fewer recently created neuroepithelial cells at E14 and E16 (192). The number of these neuroepithelial cells in E16 Lep^ob/ob embryos is normalized by intracerebroventricular leptin injection at E14 (192). Leptin induces neuronal proliferation and differentiation in ex vivo hypothalamic neurospheres (derived from the embryonic E20 rats) in a dose-dependent response (195). Neurogenesis of the feeding circuit neurons in mice is thought to be confined primarily to E12-E16. However, in Lep^ob/ob mice, systemic leptin injections for 2 weeks starting at week 4 of age (just after weaning) increases brain mass and total brain DNA content at 6 weeks of age. In 9 weeks old Lep^ob/ob mice, such injections do not affect brain mass or DNA content at 11 weeks of age (193). Clearly, there are important developmental windows/critical periods for effects of “environmental” influences on virtually all aspects of brain development.

Figure 2.

Comparison of brain development in humans (top) and rodents (bottom) from gestation to maturity. Humans and rodents experience similarly sequenced brain maturation but the timing is shifted to the right in rodents. At birth, humans have more mature brains than mice (by timing of neural events such as neurogenesis, synaptogenesis, and myelination (184), morphological observations (185), and transcriptional analysis (186)). In humans the peak brain growth rate occurs in the last 4 weeks of gestation but in mice this peak occurs in the second postnatal week. Figure created with BioRender.com.

In mice in early development, between E11.5 and 13.5, Pomc is expressed broadly across the hypothalamus but, by adulthood, only 50% of these neurons remain Pomc positive while the remainder switch fates, including differentiating into NPY/AgRP neurons (196). Offspring of mice fed HFD, in addition to maternal IR (IR dams were heterozygous for a null allele of Insr and displayed hyperinsulinemia with normal glucose concentrations), showed a transient increase in body weight between 4 and 8 weeks of age when fed HFD after weaning. The offspring of HFD-fed IR dams had increased numbers of POMC-expressing cells in the ARH at P9 indicating that ambient insulin concentration in the pup gestational environment can alter the switch from POMC to NPY/AgRP neurons (197). These data suggest that midgestation is an important developmental period during which the hypothalamus may be particularly sensitive to aberrant intrauterine environmental factors which could affect neurogenesis.

In rodents, hypothalamic neurons created in midgestation do not complete neuronal circuits or send axonal projections to their terminal target sites until the postnatal period (198, 199). Immaturity of feeding circuits during the first 1-2 weeks of postnatal life accounts for the absence of a difference in adiposity of leptin deficient or leptin overexpressing mice prior to P10. At birth, neural projections originating from the ARH are immature and are established primarily within the first 2 weeks of postnatal life (199), processes that are concurrent with the naturally occurring leptin surge (149, 177). In rodents, plasma leptin concentrations increase 5- to 10-fold in the second week of life, and peak at P10; this surge persists for the first 4 postnatal weeks (149, 177). During this period, leptin acts as a neurotrophic factor to stimulate growth of axonal projections of the AgRP/NPY neurons (198-200). Axonal innervation patterns are exquisitely timed in the mouse. Mature innervation from the ARH to the dorsomedial hypothalamic nucleus (DMH) is achieved by P6, then to the periventricular hypothalamic nucleus (PVH) between P8-10, followed by projections from ARH to lateral hypohtalamic area (LHA) by P12 (199). Congenitally leptin-deficient, Lep^ob/ob, mice display severe reduction in axonal projection density from AgRP neurons in the ARH to the PVH, DMH, and LHA, which persists through adulthood (198). These diminished ARH projections in Lep^ob/ob mice can be largely rescued with exogenous leptin if it is administered when the leptin surge would naturally occur (P4-P14). The temporal window for this effect is tightly restricted (198). Administration of leptin to adult Lep^ob/ob mice fails to restore hypothalamic projection densities. Early exposure of Lep^ob/ob mice to exogenous leptin rescues AgRP inputs to preautonomic but not the neuroendocrine neurons within the PVH indicating that, in addition to its role in formation of hypothalamic projections, leptin is also important in the targeting of axons originating from the ARH to specific cell types within the hypothalamus (200).

In mice, the number of Lepr-positive neurons in the ARH increases by roughly a factor of 3 from P5 to P25, at which time adult levels are achieved (201). Lepr is expressed in ∼40% to 50% of NPY- and in less than 20% of POMC-expressing neurons within the first 2 postnatal weeks of life; as the neurons mature, Lepr positivity increases to 80% in these neurons by the third postnatal week. In P13-15 brain slices leptin exposure depolarizes ARH NPY/AgRP/GABA (“NAG”) neurons. By P27, the electrophysiological response to leptin in NAG neurons is mixed with some depolarizing and others hyperpolarizing. Throughout the postnatal, period, hyperpolarizing neurons increase in number. At terminal maturity, by P30, all NAG neurons exhibit the classic inhibitory (hyperpolarization) response to leptin (201). This electrophysiological change in response to leptin in NAG neurons is consistent with leptin's dual roles. During the immediate postnatal period, leptin exposure promotes axonal outgrowth and, after mature feeding circuits are formed, leptin has an inhibitory role to regulate energy homeostasis.

Brain Stem Neurocircuits

Nuclei within the hindbrain—including the nucleus of solitary tract (NTS) and the area postrema (AP)—are involved in feeding behavior but until recently were considered to primarily regulate short-term feeding by responding to visceral signals to control meal termination. However, the NTS also contributes to long-term energy balance. The NTS integrates hormonal and viscerosensory information conveyed through the vagus nerve to coordinate metabolic regulatory processes. A subset of NTS neurons express LepRb; leptin signaling through these neurons affects feeding and energy homeostasis (202). Knockdown of LepRb expression in NTS neurons via AAV-shRNAi injection results in increased food intake and body weight of rats on both chow and HFD (203, 204), while direct leptin injections into the NTS reduce food intake and body weight (205). The NTS is the only site within the CNS that contains neurons expressing Gcg which encodes preproglucagon (206). Preproglucagon is processed into glucagon-like peptide-1 (GLP-1), a peptide that suppresses food intake. These GLP-1 neurons are glutamatergic and project primarily to the PVH. Approximately 60%, 40%, and 26% of the NTS GLP-1-, CCK-, and POMC-expressing neurons, respectively, coexpress LEPR (207), suggesting that leptin signaling may modify the activity of these neurons. In support of leptin importance in GLP-1 and CCK neurons, lack of leptin signaling in Lep^Ob/Ob mice results in increased expression of both of these neuropeptides in the NTS, whereas POMC expression remains unchanged (207). When stimulated chemogenetically using DREADDs, GLP-1/LEPR coexpressing neurons in the NTS only moderately reduce food intake. However, when all NTS LEPR neurons are activated the effect on food intake reduction is significantly stronger—suggesting that NTS LEPR neurons can, independently of GLP-1, suppress food intake (208).

As noted under “Hypothalamic development,” leptin impacts the innervation of hypothalamic neurons from the ARH to other hypothalamic targets during the postnatal period. Biddinger et al showed that axonal contact of GLP-1-expressing neurons in the NTS with the PVH are influenced by leptin as well (209). Projections from the NTS GLP-1/LEPR neurons to the PVH are elevated in Lep^ob/ob mice at P16, P24, and P60. Restoring leptin signaling to just these neurons (using LepRb^TB/TB mice with a transcriptional block flanked by LoxP sites upstream of LepRb which was excised in Cre-expressing GLP-1 neurons (210)) normalized the density of the GLP-1 projections in the PVH, suggesting that leptin reduces targeting of GLP-1 projections to the PVH during development (209). However, this leptin rescue of neuronal projections did not result in any detectable physiological or behavioral feeding improvements in mice fed regular chow. The phenotype was not assessed under conditions of HFD feeding (209). The discovery of increased GLP-1 neurite density from NTS to the PVH in Lep^ob/ob mice was orthogonal to an earlier-reported decrease in AgRP neurite density from ARH to PVH, DMH, and LHA in the absence of leptin. The former suggests that leptin suppresses axonal outgrowth while the latter indicates a neurotrophic role. In addition to the apparent brain region specificity of leptin's developmental actions, cell specificity was detected. Leptin's effects in the NTS were not uniform on all neuronal inputs into the PVH. Increased GLP-1 inputs onto CRH neurons were noted with no change in the oxytocin neurons (209). This specificity of leptin cellular effects was observed earlier within the hypothalamus: Postnatal leptin supplementation to Lep^ob/ob mice restored AgRP innervation specifically in the preautonomic neurons of the PVH but not the inputs in the neuroendocrine compartment of PVH (200). It is important to note that leptin exerts diverse and divergent effects on AgRP neurons during both developmental and adult stages. In early development, leptin enhances neurite outgrowth from AgRP neurons while in adulthood, its role shifts to regulating adiposity and inhibiting AgRP neuron activity. In contrast, leptin apparently inhibits GLP-1 axon growth during immediate postnatal period but in adulthood, leptin activates the GLP-1 neurons in the NTS. These data indicate that leptin is necessary for the proper neuro-organization of the hypothalamic and brain stem feeding circuits but its developmental actions are brain region– and cell type–specific, promoting the axonal growth of AgRP neurons and suppressing the GLP-1 axonal outgrowth.

Hedonic Neurocircuits

Leptin mediates the feeding behavior in adult animals in part through its actions on the mesolimbic dopamine (DA) system—one of the critical brain reward pathways originating in the ventral tegmental area (VTA) and projecting to the nucleus accumbens (NAcc), prefrontal cortex, and the amygdala (211). The DA neuron signaling in the VTA to its targets promotes reward-related activities. Food can act as a reward and stimulate the release of DA into the NAcc. Diet-induced obese rats have a reduced basal extracellular DA in the NAcc and release of DA in the NAcc is induced by highly palatable food but not regular chow (212). Human imaging studies suggest that heightened reward responsivity in the NAcc to food cues is associated with indulgence in overeating (213). Functional magnetic resonance imaging in 4- to 6-year-old children found a positive association of eating in the absence of hunger with activity of the NAcc (214).

Dopaminergic and GABA neurons in the VTA express LepRb (215-218) and direct injections of leptin into this area in adult rats cause phosphorylation of STAT3 primarily in DA neurons, a decrease in their firing rate, and a decrease in 24-hour regular chow intake and body weight of the rats (216, 219, 220). Reduction of VTA LepRb with shRNA increases acute HFD food intake and response to progressive ratio for chow (which measures effort expended for food rewards) but had no effect on long-term (4 weeks of HFD feeding) body weight (216, 219). A subset of GABAergic/inhibitory neurons within the LHA also express LepRb and innervate the VTA (221). Direct LHA leptin injection decreases food intake and body weight over a period of 24 hours (221). Decreasing LHA LepRb via direct shRNA injections increased food intake and body weight on HFD over 4 weeks (219). However, since the LHA participates in both homeostatic and hedonic systems the effects of reduced LepRb may be mediated by both. Recently Omrani et al found that leptin activates the LepRb-expressing GABA neurons in the VTA which in turn inhibit DA neurons that project to NAcc to reduce food reward seeking (222) This result suggests that leptin's effects on DA neurons are indirect. They found that optogenetic stimulation of LHA LepR neurons projecting to VTA activates DA neurons via disinhibition of the VTA GABA neurons. Chemogenetic activation of these neurons increases motivation to work for a food reward only in fed (not food-restricted hungry) mice. These results are consistent with leptin's inhibitory effects on VTA-projecting LepRb-expressing LHA neurons (222). In adult animals leptin modulates the DA activity in the reward centers through multiple pathways.

While the development of hypothalamic circuits involved in homeostatic feeding are well described, the data on neurocircuit-formation subserving hedonic systems are limited. Neuronal connections from leptin-responsive neurons in LHA to the VTA were detected within the first week of life in rats (223). The responsiveness of neurons in the VTA and LHA to leptin—defined by activation of pSTAT3 and pERK1/2—is sparse on P10 and more pronounced by P16 (223, 224). Maternal “junk food” diets (consisting of peanut butter, hazelnut spread, cookies, savory snacks, breakfast cereal, processed meat, and a mixture of lard and standard rat chow) fed during lactation are associated with increased preference for fat in the offspring at weaning and at 3 months of age. The offspring of “junk food” fed dams had increased expression of μ-opioid receptor (Mu) and decreased DA active transporter expression in the VTA at 6 weeks of age, but these phenotypes were reversed (decreased Mu and increased DA active transporter expression) by 3 months (225). Another group used RNAseq to compare expression in the VTA and substantia nigra of 6-month-old offspring reared by dams fed HFD during lactation. Expression signatures in the VTA and substantia nigra were altered and the most affected biological functions included neuron development, ion transport, regulation of membrane potential, DA metabolism, locomotion, and behavior. Additionally, the adult offspring of HFD-fed dams had impaired development of the DA neurons and their projections (decreased density of tyrosine hydroxylase) in the substantia nigra and nigrostriatal tract but no changes in the density of DA neuronal fibers in the VTA, no alterations in DA neuronal numbers, and reduced stimulus-evoked DA release in the striatum, a midbrain dopaminergic neuron target site (226). These data suggest that maternal HFD exposure during lactation directly affects the development and function of midbrain dopaminergic circuits in progeny.

We and others have demonstrated that elevating leptin during the first 3 postnatal weeks—a critical time window for brain development—predisposes rodents to subsequent increased weight gain when fed a highly palatable energy-dense diet as adults. Compared with controls, these postnatally hyperleptinemic animals are at the same weight and body composition when maintained on regular chow but gain more weight as soon as they are offered a HFD (more details under “Leptin” within the “Hormones During the Perinatal Period” section below) (149, 227, 228), suggesting that hedonic circuit development (eg, midbrain dopaminergic circuits) may be modulated by early exposure to leptin. More research is warranted to assess the molecular mechanisms that drive changes to the reward system under conditions of maternal HFD feeding during lactation: Does the leptin surge that occurs during this interval influence the development of reward system neurocircuits and subsequent neuronal responsiveness to leptin?

Leptin Effects on Glia

Neurons are not the only cell type affected by leptin during the postnatal brain development. In rodents, during postnatal weeks 2 and 3, glia cell number markedly increases (229, 230), coinciding with the leptin surge (177). The most abundant glial cells in the brain are astrocytes which provide physical neuronal support but also participate in neurogenesis, neuronal development, synaptogenesis, synaptic plasticity, and synaptic transmission (231, 232). Importantly, astrocytes mediate metabolic sensing of nutrients, including glucose (233) and lipid, and energy homeostasis. Astrocytes within the hypothalamus express receptors for and respond to hormones implicated in energy balance, such as leptin (229, 234, 235) ghrelin (236), and insulin (237). Systemic leptin administration between P8-P12 (during the physiological leptin surge) increases proliferation of astrocytes in the hypothalamus; astrocyte-specific deletion of LepRb results in decreased astrogenesis (238). Conditional deletion of LepRb in adult mouse astrocytes leads to morphological changes in astrocytes and increased synaptic inputs onto hypothalamic POMC and AgRP neurons (229). Acknowledging the potential issue of cre drivers being expressed in unintended cell types, these mice display decreased leptin-induced suppression of food intake (229). These data suggest a direct impact of leptin on astrocyte development and function in adult mice.

Early Overnutrition Effects on Feeding Circuits

Offspring of dams fed HFD during gestation and lactation (178) or only during lactation (160) gain more weight than controls during adulthood. The density of AgRP immunoreactive fibers in the PVH is decreased in the offspring of HFD-fed compared with those of chow-fed dams; these progeny display decreased hypothalamic leptin sensitivity (determined by leptin-induced pSTAT3 in the ARH and ventromedial nucleus of the hypothalamus [VMH]) at P30 and P90 (178). Consistent with this finding, Vogt et al reported reduced densities of AgRP and α-MSH (alpha-melanocyte stimulating hormone) fibers from the ARH to functionally downstream hypothalamic areas (PVH, DMH, and LHA) in the offspring of dams fed HFD during lactation [160) (Fig. 3). Maternal HFD feeding increases the projection density of SF-1 and BDNF-positive neurons from the VMH to the LHA (239). Since postnatal overfeeding causes an augmented leptin surge (149), it is surprising that the alterations in hypothalamic circuits reported in postnatally overfed (and therefore postnatally hyperleptinemic) offspring are similar to alterations seen in leptin-deficient mice (198), suggesting that developing hypothalamic feeding circuits are critically sensitive to ambient leptin concentrations. Kirk et al hypothesized that postnatal hyperleptinemia induces leptin resistance thereby attenuating leptin signaling and impairing the development of hypothalamic projections (178). Park et al showed that maternal obesity induces endoplasmic reticulum (ER) stress in tissues that are critical for energy metabolism, including the arcuate nucleus of hypothalamus, during the early postnatal period (Fig. 3). At P14 the POMC and a-MSH axonal densities in the PVH were decreased in pups born to HFD-fed dams compared to control-fed dams, while the densities of AgRP-immunoreactive fibers were not affected. Additionally, leptin signaling (measured by STAT3 phosphorylation) in the offspring of HFD-fed dams was reduced and relieving ER stress in neonatal pups with chemical chaperones such as TUDCA normalized neurostructural changes in the hypothalamus, possibly via improvement in leptin sensitivity (240). Other disturbances to early hypothalamic development induced by maternal obesity include a decrease in proliferation of neural progenitor cells and IR in fetal hypothalamus (241) (Fig. 3). Such fetal IR could be another driver of long-term dysregulation of energy homeostasis in offspring exposed to maternal obesity.

Figure 3.

Developmental overnutrition effects on feeding circuits. Under conditions of maternal HFD feeding or overfeeding of offspring via reduced litter size during the early postnatal period, the offspring are more susceptible to obesity. This association could be mediated by mechanisms that include an augmented postnatal leptin surge, changes in the quality or quantity of neuronal connections within the feeding circuits, decrease in central leptin and insulin signaling, increase in ER stress within the ARH, decrease length of the hypothalamic primary cilia, alterations in synapses of POMC neurons, and decreased neuronal progenitor cell proliferation. Some or all of these mechanisms are likely affected by early overnutrition and contribute to subsequent susceptibility to obesity. Figure created with BioRender.com.

In addition to quantitative changes of neuronal processes and targets within the feeding circuits, overnutrition during development alters synaptic connections per se (Fig. 3). Two-week-old offspring raised in small litters had an increase in postsynaptic inhibitory currents onto POMC neurons compared with pups from normal-size litters, but by week 3 this phenotype was absent (242). In control mice, leptin effects on the membrane potential of the POMC neurons were heterogenous (stimulatory and inhibitory) in the first 2 weeks of life but transitioned to primarily stimulatory actions in adult mice (242). Leptin did not change inhibitory postsynaptic currents onto POMC neurons from postnatally overfed (via rearing in small litters) mice during the early postnatal period or in adulthood, suggesting that the ability of leptin to reduce the strength of inhibitory inputs onto POMC neurons is blunted by early overnutrition (242) and may be influencing the development of obesity in the offspring.

Maternal HFD feeding during gestation and lactation significantly reduces the length of hypothalamic primary cilia of neonatal pups (243) (Fig. 3). Defects in proteins required for ciliary formation and function cause severe obesity in humans and mice (244, 245). Disruptions in proper development of cilia resulting from maternal nutrition could contribute to the obese phenotype of the offspring. Microglia may be the mediator of hypothalamic feeding circuit programming by maternal nutrition. Microglial activation (measured by morphological changes in Iba1⁺ cells) was increased in the chow-fed offspring of dams maintained on HFD throughout pregnancy and lactation (246). Depleting microglia from the developing fetuses during gestation by supplementing the diet of pregnant females with colony-stimulating factor 1 receptor inhibitor (PLX5622) causes a reduction in the number of hypothalamic POMC neurons and an increase in offspring weight gain starting on P5 (247).

The primary issue with the maternal HFD model is that it is noisy by virtue of multiple metabolic consequences, complicating identification of primary molecular mechanisms driving the programming of body weight. For that reason, many groups have tested isolated characteristics of the HFD and related obesity models. Concentrations of 4 hormones, leptin, insulin, ghrelin, and amylin, have been shown to be influenced by postnatal overnutrition, and it is plausible that they drive—at least in part—the adult phenotype.

Hormones During the Perinatal Period: Effects on Feeding Circuits and Body Weight in Adulthood

Leptin

Perturbation of circulating leptin concentrations during the postnatal period—specifically, altering the magnitude of the naturally occurring leptin surge between P0 and P22—has long term physiological consequences. Both reduced and augmented leptin surges result in subsequent increased body weight gain when fed high-energy diet or HFD (166, 248). The timing and route of leptin administration has varied in studies of the physiological consequences of postnatal leptin supplementation on adult body weight of the offspring in rodents. Elevating plasma leptin concentration using a TET-On (doxycycline-dependent) leptin transgenic mouse (LepOE) during the period of the natural leptin surge (first 3 weeks of life) increased weight gain when mice were subsequently exposed to HFD at 10 weeks of age (166) (Fig. 4). In this study the LepOE and control pups were littermates, hence prenatal or postnatal environment were the same; the transgenic construct allowed for noninvasive (doxycycline in drinking water) manipulation of circulating leptin in LepOE mice at any point. Importantly, in the same study, exposure of “adolescent” mice (between 3 and 8 weeks old) or adult mice (9-29 weeks old) to doxycycline-induced hyperleptinemia did not cause long-term changes in energy homeostasis of the mice (Fig. 4). The, long-term consequences of transient hyperleptinemia were entirely dependent on the timing of exposure (166). Since overnutrition in mice (either by reducing litter size or maternal HFD feeding) augments the leptin surge, inducing a state of elevated leptin constitutes a physiological and mechanistic surrogate for overnutrition, supporting the inference that the increased adiposity associated with postnatal overnutrition is driven centrally by this augmented postnatal leptin surge. Consistent with this inference, others have reported that intraperitoneal administration of leptin to nursing pups (P0-10, P10-20, or P3-13) increases weight gain in adulthood when the animals are fed a HFD (Table 1) (227, 228).

Figure 4.

Effects on body weight of transient hyperleptinemia. Transgenic mice that transiently overexpress leptin during the immediate postbirth period have no obvious phenotype until exposed to high-fat chow as adults. This exposure reveals a developmentally programmed susceptibility to increased weight gain on HFD compared with control littermates. When exposure to elevated leptin occurs in adults with a slow and gradual increase in leptin concentrations (to mitigate weight loss), the mice do not show any difference in body weight compared to controls even after HFD challenge. Figure created with BioRender.com.

Table 1.

Effects of hormone exposure during development on the programming of body weight in the offspring

Hormone	Effects on feeding circuits and body weight in adulthood
Leptin	Augmented postnatal leptin surge → Increased body weight gain in adult mice on HFD (166, 248). Oral leptin administration during early postnatal development → Protection from DIO in adulthood (249). Reduced postnatal leptin surge → Increased weight gain in adult mice on HFD (248, 250). Transient hyperleptinemia effects depend on timing of exposure (166).
Insulin	Hypothalamic insulin administration in early postnatal development → Increased body weight gain and glucose impairment (251). Elevated maternal insulin → Altered hypothalamic circuits and glucose homeostasis (197).
Ghrelin	Ghrelin signaling modulation in early postnatal development → Altered hypothalamic neurocircuits, increased body weight and fat (252).
Amylin	Amylin administration in early postnatal development to DIO rats → Enhanced leptin signaling, restored hypothalamic neurocircuits; no protection from weight gain in adulthood (253). Maternal HFD exposure in gestation → Amylin resistance and impaired POMC neuron differentiation (254).
GLP-1	GLP-1R agonist (Exendin-4) administration in early postnatal development to IUGR offspring→ Prevented diabetes, attenuated weight gain in adulthood, and altered neuroarchitecture (255, 256). Early postnatal Exendin-4 administration in wild type → Reduced body weight and adiposity, and altered neuroarchitecture (256).

Abbreviations: DIO, diet-induced obesity; GLP-1, glucagon-like peptide-1; HFD, high-fat diet; IUGR, intrauterine growth restriction; POMC, proopiomelanocortin.

The concentrations of leptin in breast milk differ between mice and rats, with reported values up to 20 ng/mL in mice (257, 258) and 3 ng/mL in rats (170, 259, 260). Leptin concentration is higher in milk from HFD-fed dams (at P10) than in control diet–fed dams (258, 261). Administration of leptin orally is not equivalent to elevating circulating leptin. Casabiell et al reported that in mice, ingested leptin reaches the circulation (262). Oral administration of leptin from P1 to P20 (estimated to be 5 times the mean daily leptin intake from the mother's milk, by increasing supplementation from 1 to 44 ng per day in each animal over the first 20 postnatal days (263)) resulted in protection from diet-induced obesity (DIO) in adulthood. When postnatally leptin-fed rats (P1-P20) were exposed to HFD from weaning, they had ∼7% lower body weight than controls at 6 months of age (249) and improved insulin and leptin sensitivity (although these results were confounded by significant difference in body weight at the time of measurement (264)). Ingested leptin could drive physiological changes via the intestinal K and/or L cells, which secrete glucose-dependent insulinotropic polypeptide and GLP-1, respectively, which could then influence the sympathetic and/or sensory innervation of the gut resulting in lasting effects on energy homeostasis.

While elevated circulating leptin during the immediate postnatal period is associated with an increase in adult body weight, insufficient postnatal leptin signaling promotes a similar phenotype. Newborn rats administered a leptin receptor antagonist from P2-P13 are more leptin resistant in adulthood—they do not reduce food intake or weight after 1 week of exogenous leptin administration—and gain more weight than controls when fed a high-density diet in adulthood (248, 250). These data suggest that the relationship between circulating leptin concentrations (or the amount of CNS leptin signaling) during the postnatal period and adverse energy homeostatic outcomes in adulthood are U-shaped; too much or too little leptin during the critical developmental time window can predispose to diet-induced obesity in the offspring. The mechanisms by which postnatal leptin is driving this phenotype is not clear but may involve the changes in neuroanatomic architecture of the feeding circuits or alterations to the early neuronal specification that could ultimately result in decreased cellular responsiveness to increasing leptin concentrations. The effects of postnatal leptin elevation could also be conveyed via epigenetic (ie, DNA methylation or histone modifications) changes within the CNS, specifically in the leptin-dependent circuits (more on epigenetics in the following sections).

Insulin

Insulin is implicated as a neurotrophic factor during development. Plagemann et al implanted insulin directly into the hypothalamus of P2 or P8 rats, causing increased body weight gain and impaired glucose homeostasis in adulthood (251). Hypothalamic insulin administration at P8 caused morphological changes within hypothalamic nuclei—including hypotrophy of neuronal nuclei within the VMH and hypertrophy in the DMH—suggesting that postnatal hyperinsulinemia may disturb the neuro-organization of the hypothalamus (265). Subcutaneous injections of long-acting insulin on P8-P11 increased body weight of the rats as early as P10 and throughout lifetime (Table 1). At 7 months of age these rats showed hyperinsulinemia, impaired glucose tolerance, and increased systolic blood pressure; at 30 months of age neuronal number was decreased in the VMH (266). In a study in which wild type offspring born to dams who were either wild type or heterozygous for a null allele of the insulin receptor (hence hyperinsulinemic, insulin resistant), Carmody et al found that maternal IR increased the number of POMC-expressing cells in the ARH of P9 wild-type offspring of the hyperinsulinemic dams (197). These results suggest that maternal insulin influences the specification of POMC neurons (197). Vogt et al showed that elevated maternal insulin resulting from maternal HFD exposure during lactation is associated with decreased α-MSH and AgRP fiber density in the PVH, DMH, and LHA in 8-week-old offspring. They then blocked insulin signaling specifically in POMC neurons of the offspring only (not the dams, using insulin receptor flox/flox [IR^fl/fl] dam crossed with IR^fl/fl and POMC-Cre sire), under conditions of maternal HFD feeding (causing maternal hyperinsulinemia), and found that lacking insulin signaling in POMC neurons rescues the projections in this circuit and corrects glucose intolerance but does not prevent increased body weight (160). These results implicate a role of POMC insulin signaling in the development of hypothalamic circuits involved in glucose homeostasis.

Ghrelin

Ghrelin—an appetite-stimulating circulating factor—is primarily produced in the stomach and signals through the growth hormone secretagogue receptors located across the CNS (267-269). Neural expression for growth hormone secretagogue receptors is highest in the hypothalamus (including ARH, VMH, and lateral mamillary nuclei) (269); administration of ghrelin activates neurons in brain areas known to control feeding (such as ARH, VMH, and PVH) (270). In wild-type mice circulating ghrelin levels rise from P6 and reach adult concentrations by P14 (252). To evaluate the physiological and neurobiological roles of ghrelin during postnatal development, ghrelin signaling was either blocked (by daily intraperitoneal injections of an antighrelin compound that specifically binds and inhibits the bioactive acylated form of ghrelin) from P4 to P22 or circulating ghrelin concentrations were elevated (via daily intraperitoneal injections resulting in at least 50-fold increase in plasma concentration; ghrelin concentrations were increased within 15 minutes of injection and returned to control concentration within 12 hours) from P4 to P12. Blocking ghrelin action enhanced overall PVH fiber density, including AgRP and αMSH immunoreactive fibers; administration of ghrelin had the opposite effect—PVH innervation was reduced. Surprisingly, both interventions resulted in increased body weight, adiposity, and circulating blood glucose. While the overall density of AgRP and αMSH fibers in the PVH are oppositely affected by ghrelin and antighrelin administration, the ratio of orexigenic AgRP to anorexigenic αMSH fibers is increased in both conditions potentially explaining the increased body weight seen with both manipulations (271). These data emphasize that both the developmental timing and the magnitude of ghrelin signaling are important in the proper formation of hypothalamic neurocircuits and are critical in the control of body weight later in adulthood (Table 1) (252). Postnatal overnutrition induces central ghrelin resistance (demonstrated by an attenuated induction of c-Fos immunoreactivity in the ARH following peripheral ghrelin injection) and supports the evidence that hypothalamic development is influenced by ghrelin during the postnatal period (272).

Amylin

Amylin, or islet amyloid polypeptide, is coreleased with insulin from the pancreatic beta cells (273) in molar ratio of 1: up to 100 (274-276). Amylin complements insulin's actions to regulate postprandial glycemia via suppression of glucagon secretion (277). Amylin inhibits food intake and delays gastric emptying via heterodimeric receptors (278) composed of the calcitonin receptor (CTR a or b isoform) and 1 of the 3 receptor activity modifying proteins (RAMP 1-3) in the AP which project to the NTS and lateral parabrachial nucleus. In the hypothalamus, amylin enhances leptin signaling in the VMH and ARH by elevating microglial interleukin-6 secretion which amplifies pSTAT3 expression (253, 279-281).

During gestation, amylin is initially detected in pancreatic islet β-cells at E12 and peaks at E17 (282, 283); the amylin receptor, CTR, is expressed in the brain starting at E12-13. Amylin knockout mice have fewer microglial cells in the ARH and AP at E12; neuronal numbers are unchanged (284). Postnatal neurotrophic effects of amylin have also been reported using RAMP 1/3 knockout mice and knock down of CTR in brains of P4 rats (both of which result in reduced amylin receptor signaling) and amylin KO mice, all showing reduced α-MSH fiber projections from ARH to PVH. The density of AgRP/NPY fibers was higher in amylin KO mice, lower in RAMP 1/3 KO mice and not detectably different in the rats with CTR knocked down specifically in the VMH at P4 (281). Amylin induces p-ERK specifically in arcuate POMC neurons but not AgRP/NPY neurons, suggesting that the effects of amylin to enhance outgrowth of POMC neurons may be direct while those on AgRP/NPY may be indirectly mediated (eg, via a microglia-secreted interleukin-6–related mechanism) (281). DIO rats selectively bred to gain excess weight on high-energy diet are obese and leptin resistant (reduced pSTAT3 activation in the ARH following leptin administration) in adulthood when fed HFD. At P12 these animals have decreased fiber densities from ARH to the PVH (by 1,1′-Dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI), a fluorescent lipophilic tracer that labels axonal projections labeling) (285). Administering amylin to these DIO rats during the postnatal period (P0-P16) restores a-MSH and AgRP/NPY projection densities in the PVH and improves leptin signaling; however, amylin administration does not protect these animals from excessive weight gain in adulthood (253). Maternal HFD during gestation leads to increased serum amylin and reduced hypothalamic Ramp3 mRNA and protein in male fetuses at E16.5 (254). Decreased hypothalamic Ramp3 causes amylin resistance (decreased amylin-induced phosphorylation of STAT3 and ERK used as surrogates) which impairs differentiation of POMC (but not AgRP/NPY) neurons in hypothalamic neural stem cells (254). Gestational elevation of fetal serum amylin via subcutaneous injections into pregnant dam resulted in decreased body weight and increased hypothalamic Pomc expression and protein level in male offspring compared with controls. These data suggest that amylin plays an important role in differentiation of POMC neurons. Amylin influences the establishment of neurocircuits within, not only the hypothalamus, but also the AP projections to the NTS within the brainstem. Neurite outgrowth (measured with DiI inserted in the AP) from AP to the NTS in amylin KO mice at P10 and P60 showed reductions relative to wildtype mice in neuronal fiber densities from AP to the NTS (286) confirming amylin's role in development of brainstem circuits (Table 1). Similar to insulin and leptin, in human placenta perfused with pramlintide (an analog of human amylin) ex vivo, only a negligible amount of pramlintide crossed the placenta from maternal to fetal circulation (287). Therefore, circulating maternal amylin is unlikely to contribute meaningfully to the increased fetal serum concentration under maternal HFD feeding conditions (254). Possibly, maternal obesity or nutritional status (such as specific macronutrients, hormones, inflammatory molecules or other factors) may directly induce placental or fetal amylin expression. Subcutaneous amylin injections to pregnant mouse dams caused a 2-fold amylin increase in E16.5 fetal serum, suggesting that maternal amylin induces an increase in fetal amylin concentrations (254).

GLP-1

GLP-1, expressed in the intestinal L-cells and NTS neurons, acts in the periphery to slow gastric emptying as well as centrally to reduce food intake and stimulate glucose-induced insulin secretion (288). In a model of IUGR induced by uterine artery ligation in pregnant rats, the offspring have low birth weight and develop obesity and diabetes in adulthood (255, 289). Subcutaneous administration of Exendin-4—a GLP1R agonist—during early postnatal period (P1-P6) to these IUGR offspring prevents the development of diabetes and attenuates weight gain in the adulthood (255). Interestingly, reduced weight gain was also observed in control animals given Exendin-4 on P1-6 (255, 256). Neonatal (P1-P6) administration of Exendin-4 to wild type mice resulted in lower body weight (as early as at P28) and adiposity when maintained on low fat chow (Table 1). Females (but not males) gained less weight when fed 45% HFD, driven by increased energy expenditure, compared with controls. In addition to physiological changes, neonatal Exendin-4 administration reduced NPY neuronal fiber density in the PVH of 4-week-old female mice, an effect that was reversed in animals with Glp1r knocked out in PVH Sim1 neurons (256). This study suggests that manipulation of GLP-1 concentrations early in the postnatal period may affect the neuroarchitecture of the hypothalamus with physiological consequences on body weight in adulthood (in a sex-specific manner) and specifically implicates the GLP1R signaling pathway within the PVH (Table 1).

In addition to the hormones described above, several others, including glucagon, glucose-dependent insulinotropic peptide, and peptide YY, are recognized as important players in regulating energy homeostasis. These hormones may contribute to the developmental programming of adiposity, although direct investigations in this regard are currently lacking. Further research is needed to elucidate their specific roles in shaping adiposity and overall metabolic health during development.

Epigenetic Changes

Epigenetic changes during gestation or the postnatal period are another potential mechanism mediating developmental programming of body weight (290). Epigenetic modifications can alter gene expression without changing the DNA sequence; instead, expression is altered by DNA methylation, histone modifications, and microRNAs (miRNAs). Patterns of DNA methylation have been studied in the context of developmental programming of weight in humans and rodents. Alterations of DNA methylation incurred during development can persist into adulthood, making such alterations a plausible mechanism for developmental programming of body weight by virtue of their “memory” of the nutritional, metabolic and other environmental characteristics. DNA methylation patterns of genes critical in energy and glucose homeostasis, metabolism, and insulin signaling can be altered by exposure of progeny to maternal obesity, HFD-feeding, or overnutrition during gestation and/or lactation (291-297). Overnutrition of weanling rats (by decreasing litter size) during the postnatal period results in hypermethylation of the hypothalamic POMC promoter (293) which is negatively correlated with hypothalamic POMC expression adjusted for circulating leptin. Similarly, offspring of dams fed HFD during pregnancy and lactation show increased methylation of the POMC promoter even when fed regular chow after weaning (291, 298). Maternal HFD feeding results in hypermethylation of the ARH InsR promoter and an associated decrease in InsR expression in the ARH of male offspring only (299). It is unclear by what molecular physiological mechanisms gestational and early postnatal states are conveyed directly to the genome.

Investigating the effects of perinatal nutrition on epigenetic changes in humans is challenging since the tissues of interest (ie, specific brain regions) cannot be directly assessed. As a proxy for overall methylation status some studies assess methylation in cord blood DNA. Even though a number of studies report associations of DNA methylation patterns in cord or newborn blood with maternal BMI or gestational weight gain, the individual genes identified are not consistent across the studies (300-303). In a meta-analysis (integrating epigenome-wide association studies from 19 cohorts), prepregnancy BMI was associated with small variations in DNA methylation of newborn blood at 86 CpGs sites; the effect size for each of the sites was less than a 0.15% change in methylation per unit of maternal BMI increase (304). Following further analyses with causal inference strategies, a support for causal intrauterine effect was found in only 8 out of 86 CpG sites (304). A major limitation of studies associating DNA methylation with maternal BMI or other variables is that they use circulating blood cells as a surrogate for specialized cells in brain, adipose tissue, liver, and other organs. Another meta-analysis of epigenome-wide association studies found that there are only minimal associations between the DNA methylation of specific CpG sites in cord or whole blood from children and adolescents and the BMI in children 2-18 years of age (305). While the authors conclude that differences in DNA methylation are likely a consequence rather than a cause of obesity, it is equally likely that the methylation states of cord or peripheral blood does not reflect the patterns of methylation in tissues that are more relevant to energy homeostasis. One study compared the whole blood methylome of siblings born before and after maternal bariatric surgery to investigate the effects of maternal weight loss and found changes in methylation of genes involved in leptin and insulin receptor signaling, and genes associated with type 2 diabetes (306).

Effects of maternal obesity on placental DNA methylation could more directly reflect intrauterine environment and placental function. DNA methylation in the placenta is increased overall by 21% in obese compared with lean gravidas (307). In genome-scale DNA methylation analysis, altered placental methylation of 56 CpGs was associated with maternal BMI. The LEP promoter is hypermethylated, and the adiponectin promoter hypomethylated on fetal and maternal sides of the placenta, respectively, in obese compared with lean gravidas (308, 309). Similar to maternal obesity effects, Lesseur et al reported increased placental (average of maternal and fetal sides) leptin promoter methylation in infants exposed to GDM (308). In contrast, Bouchard et al observed a negative and positive correlation between glucose levels (at 24-28 weeks) and LEP methylation on the fetal and maternal side, respectively (310). Since hypermethylation is associated with reduced gene expression, maternal obesity and diabetes may influence leptin production by the placenta, and the intrauterine milieu affecting fetal development. This is a complicated interplay since it is also possible that leptin (or other metabolites/ hormones) during development influence the patterns of DNA methylation.

Animal studies on ancestral exposure to HFD, malnutrition, and environmental toxicants have demonstrated that obesity risk can be conveyed transgenerationally through epigenetic modifications (311). One study found that maternal HFD exposure results in increased body mass in third-generation (F3) female (only) offspring that is transmitted only via the paternal lineage (312). A recent study found that maintaining male mice on a Western diet (high in fat and sugar) for 5 consecutive generations resulted in a progressively amplified obesity phenotype and related metabolic diseases in the offspring (313). Furthermore, even when the progeny of these multigenerational Western diet–fed males were fed a standard diet, both males and females remained heavier than controls for 4 subsequent generations (313). This transgenerational obesity may be mediated by epigenetic marks. The Agouti viable yellow (A^vy) mouse has been investigated as a model of maternal obesity. The A^vy is a metastable epiallele—an allele of the same gene that differs in DNA methylation status and is susceptible to methylation changes by environmental factors—allowing for isogenic mice to exhibit different body weight and fur color phenotypes (314). Undermethylation of the A^vy allele results in increased expression of ASP (which blocks MSH activity at melanocortin receptors) and obese mice with yellow coats (A^vy/a), while mice with a silenced allele are lean with agouti fur. In successive generations of obese A^vy/a female mice Waterland et al found that with every successive generation maternal obesity increased and body weight in adult progeny increased incrementally (by generation) while maintained on standard chow (315). A contemporaneous cohort of mice fed a methyl-group supplemented diet (supplemental folic acid, vitamin B12, betaine, and choline), inducing developmental DNA hypermethylation, did not experience the same incremental increase in offspring obesity with subsequent generations. The coat color phenotype—used as a proxy for A^vy methylation—was not associated with body weight in either group, suggesting that maternal obesity affects the developmental establishment of epigenetic modifications at other genomic loci critical in body weight regulation (315). Consistent with the idea that obesity inhibits methylation and this reduced methylation status confers offspring obesity risk, supplementation of methyl donors to HFD-fed C57BL/6J dams during gestation and lactation mitigates the increase in body weight found in adult offspring born to dams fed HFD (316). Promoter region methylation of obesity-related genes, Pparg, Fas, Lep, and Adipoq was reduced in the adipose tissue of offspring reared by HFD-fed mothers but was mostly rescued by dietary supplementation of methyl donors in the HFD. With the exception of Lep, gene expression was inversely related to methylation status of these genes (316). Another group carried out a similar experiment and found that maternal consumption of HFD during pregnancy and lactation results in global DNA hypomethylation in the offspring brain and this effect can be reversed by supplementing HFD with methyl donors (317, 318). Body weight gain in the offspring of HFD-fed mothers was attenuated with dietary methyl group supplementation at 12 weeks of age, however this effect was transient as no body weight difference was detected among the groups (including offspring born to chow vs HFD-fed dams) at 20 weeks of age and beyond. These data support the role of epigenetic modifications in vertical transmission of obesity. The role—if any—of maternal folate status on progeny birth weight in humans is unclear.

It is important to point out that the epigenetic data supporting transgenerational effects of HFD are not consistent across the various studies which may be partially attributed to heterogeneity in epigenetic status by cell type and by some studies being underpowered. These differences underscore the need for further research with improved study designs as suggested by others (319, 320). Such improvements would include interrogating epigenetic status of single cells, having a detailed a priori analysis plan, and validation of results in independent experiments.

Relevance to Humans

Comparison of Brain Development in Human vs Rodent

The timing of brain development between humans and rodents differs considerably; but the sequence of events occurring during brain maturation is largely the same (321) (Fig. 2). Rodents at parturition have less mature brains than humans at birth and experience an accelerated postnatal development (321). Neurogenesis in rodents starts at E9.5 days and lasts through the second postnatal week, while in humans it is estimated to begin at E35 weeks and lasts generally until midgestation but may continue through gestation and postnatally to a limited degree in certain regions such as hippocampus (185). Neurogenesis has been reported in dentate gyrus of adult rodents (322, 323) and humans (324). Neurogenesis specifically in the hypothalamus of mice takes place between E10 and 16 (188, 325) while in humans between gestational weeks 7 and 10 (326), but may also continue in the second trimester (327). Synaptogenesis in humans starts around week 20 of gestation, peaks around 2 years of age, and persists until adolescence. In rodents, synaptogenesis starts at E16 and is completed roughly by weaning but hypothalamic synaptogenesis during which the feeding circuitry matures occurs postnatally, between approximately P6 and 16 (199). The process of myelination also varies among brain regions and species. In humans, myelination begins in the third trimester of gestation and continues postnatally throughout childhood and adolescence and into adulthood (328, 329). In rodents, myelination begins after birth and continues until 6-7 weeks of age (330, 331). Myelination of the hypothalamus in rodents occurs around week 4 and 5 of life (330). Brain growth rate (defined by mass) peaks at the end of third trimester (∼36-40 weeks of gestation) in humans and at P7-P10 in rodents (332-334). The third trimester of human pregnancy is analogous to P1-P10 in rodents; the human neonate's brain development, by mapping cellular transcriptional profiles, corresponds to that of a P7-P10 mouse (186). Based on these estimates, experiments investigating the nutritional environment during early postnatal brain development in mice are relevant to late gestation in humans.

Adiposity in Fetus and Neonates

In the human fetus, accrual of body fat begins at ∼25 weeks of gestation. Based on body composition estimates measured with air displacement plethysmography (ADP, peapod) in healthy preterm infants, the mean fat mass at 30 weeks of gestation is 0.1 kg (7.4% total body mass) and increases to 0.28 kg (10.1%) by 36 weeks (335). These estimates are compatible with fetal chemical analyses (336). In normal term infants, fat mass increases to 0.4 kg (11.2%) by 40-41 weeks of gestation (ADP) (337) and to ∼2 kg (∼26%) by 6 months postnatal (ADP). By 12 months of age, fat mass ranges from 23% to 28% of body mass, and by 3-5 years of age, fat mass percentage drops and ranges from 14% to 24% (depending on gender and ethnicity (338);).

Leptin in Pregnancy and in the Human Fetus

In adult humans, leptin is secreted from adipose tissue in proportion to adipose tissue mass and fat cell volume. During pregnancy, the placenta is an additional source of leptin (339, 340). In nonhuman primates (341) and humans (342-345), maternal circulating leptin increases throughout pregnancy. This increase occurs even during the first trimester, prior to any substantial maternal fat mass increase (344), suggesting that the placenta may be a contributor to the maternal circulating leptin. Most studies report a 1.5- to 3-fold increase in circulating leptin in the second and third trimesters (342-346) with a significant decline in the early postpartum period, as early as 24 hours after delivery (347, 348). Since circulating leptin concentrations are proportional to fat mass prior to pregnancy, overweight or obese pregnant women have higher circulating leptin at baseline. However, the rate at which circulating leptin increases during gestation is significantly lower in overweight/obese than in normal-weight gravida (349). Circulating leptin increases as little as 10% from the first to third trimester in overweight/obese pregnant women compared with a 54% increase in normal weight gravida [349-351). In lean gravidas, Misra et al observed a significant increase in plasma leptin concentrations per unit of body weight with pregnancy progression while the trend was opposite in overweight/obese group suggesting a lower production of leptin (either per fat or placental mass) in overweight/obese gravida in advanced pregnancy (350).

While leptin is detectable by 19 weeks of gestation in human fetal circulation, it is unclear which tissue is its primary source. Fetal plasma leptin concentration correlates with gestational age, fetal size, and birthweight (352). All 45 studies included in a meta-analysis reported a positive correlation between plasma leptin concentration and birth weight (353). Fetal plasma leptin correlates with fat mass, and the slope of the linear regression of gestational week vs circulating leptin during weeks 21 to 32 of gestation is low but increases between weeks 33-41. This rapid increase in slope is associated with a more rapid increase in fetal fat mass (342). Infants born prior to 30 weeks of gestation had ∼6-fold lower circulating leptin concentration than those born between 30 and 36 weeks of gestation (354). Compared to appropriate for gestational age (AGA) controls, IUGR fetuses have lower circulating leptin concentrations (355, 356) likely due to decreased adiposity since leptin per kilogram of mass is unchanged. Fetuses from gestational diabetic (GDM) gravidas have higher leptin than AGA fetuses, even when leptin is corrected for fetal weight (352, 356). Infants born to women with GDM have increased fat mass compared with infants of mothers with normal glucose tolerance (71, 72, 74), suggesting that elevated circulating leptin in these infants is likely due to higher fat mass. Cord blood leptin is also significantly higher in the offspring of women with T1D, and is likely mediated via an increase in fat mass [75). As mentioned under “Perinatal programming of body weight: evidence from humans,” infants born to diabetic mothers have elevated insulin concentration which are likely driving the increase in fat mass accrual during gestation, ultimately resulting in elevated fetal plasma leptin concentrations.

Also, LGA babies have higher, while SGA babies have lower circulating leptin concentrations (357, 358) than AGA, with concentrations that are directly proportional to fat mass (359). This dependence of plasma leptin on fetal size suggests that fetal adipose tissue is an important leptin source in fetal circulation. This interpretation is somewhat confounded by the fact that mothers with GDM are often obese and therefore have higher plasma leptin concentrations; in addition, placental weights are higher in GDM pregnancies (360). However, as mentioned above, infants of gravida with diabetes accrue more fat and hence more leptin. There are also multiple reports of leptin concentrations being significantly higher in female vs male newborns (355, 361-364). Based on body composition of the reference infants, there is no significant difference in fat mass between male and female at birth (365, 366), suggesting that this sexual dimorphism is independent of fat mass. There are mixed reports regarding differences in plasma leptin concentrations in male vs female prepubertal children, but the data are consistent in that circulating leptin is higher in females than males starting from late puberty (367-372) and persists into adulthood even when adjusted for fat mass (373-375). The production of leptin is higher in subcutaneous than in visceral adipocytes. As females tend to have more subcutaneous fat than males, this difference in leptin production accounts in part for the higher concentrations of circulating leptin (adjusted for body fat) in females than males (373). However, it is not entirely clear at which point during development this difference in leptin production emerges.

To determine the extent of placental syncytiotrophoblast contribution to circulating leptin in the human fetus, 3 groups assessed leptin secretion in human ex-vivo placentas (376-378); 2 groups found that only a small fraction (1-5%) of placental leptin is secreted into the fetal side of the placenta with the remainder secreted into maternal circulation, suggesting that the placenta only supplies limited amounts of leptin to the fetus. However, 1 study showed that 46% of placental leptin is secreted to the fetus (378). Since maternal plasma leptin concentrations are ∼2 to 3 times higher than fetal and the fetal blood volume is much smaller than maternal; even ∼5% of placental leptin secreted to the fetal side could be a substantial contributor to circulating fetal leptin. Additionally, as gestation progresses, the weight of the placenta increases. The weight of the placenta between 23 and 27 weeks of gestation increases by 44 g (379) and those delivered between 37 and 41 weeks increases by 59 g (380), indicating that the placenta—whose weight gain parallels the increase in fetal mass and fetal plasma concentrations—could be an additional source of leptin. Most clinical studies do not find any correlation between maternal and fetal/cord plasma leptin concentrations (multiple references (381)). In a study of healthy pregnant women, leptin concentrations in infants classified as small, appropriate, or LGA correlated with their ponderal indices (kg/length m³) (358) but were not related to maternal or placental leptin. Cord blood leptin concentration correlated with fetal size and fat mass suggesting again that fetal adipose tissue may be a more important source of circulating fetal leptin.

As in humans, circulating leptin concentration increases during mouse pregnancy but to a much greater degree (20- to 40-fold). In contrast to humans, the maternal hyperleptinemia in mice is primarily a result of increased placental production of soluble leptin receptor (LepRe isoform) which binds leptin in circulation (382). Similar to humans, secretion of leptin from placenta into fetal circulation in rodents is low (383). ¹²⁵I-leptin is transported from maternal circulation to the placenta and fetus. This transplacental leptin transfer increases by 10-fold between day 16 and day 22 of gestation in rats (383). The magnitude of molecular leptin transfer from maternal to fetal circulation is unclear and was not addressed by this study. In a study of pregnant leptin deficient (Lep^ob/ob) females carrying ob/+ fetuses, no detectable leptin was found in maternal circulation, suggesting that leptin does not cross back into maternal circulation from the fetus (382, 384).

Plasma leptin concentrations in nonhuman primates during the third trimester of gestation were 46-fold higher in the maternal vs fetal circulation under regular diet conditions but in mothers fed a 30% HFD, leptin in maternal circulation was doubled while the fetal leptin remained the same. The low fetal compared with maternal leptin concentrations and the lack of increase in fetuses of HFD- vs regular diet-fed mothers suggest that little or no maternal leptin is transported across the placenta (385).

Leptin concentrations in pregnant women increase throughout gestation and drop significantly shortly after parturition, before maternal body fat changes significantly (344). Leptin concentrations in the progeny decline immediately after birth relative to cord blood concentrations. Schubring et al found that leptin in venous cord blood at term of neonates was 3.47 ng/mL, dropped to 0.26 ng/mL by 56 to 79 hours after birth, and remained low (0.59 ng/mL) at 99 to 128 hours after birth (386). Hytinantti et al reported a decrease in cord blood leptin from 9.7 ng/mL at birth to 2 ng/mL by postnatal day 3 (387). Ertl et al reported cord blood leptin at birth of 4.5 ng/mL, decreasing to 0.6 ng/mL by P7 (361). Many other studies have reported 4 times higher values at birth compared with plasma leptin concentrations within a few days after birth (357). Even though there is a decline at birth, leptin concentrations after birth still correlate with body mass. As discussed earlier, leptin concentrations in human fetuses increase during gestation peaking near the end of gestation. Similar to rodents—which have a leptin surge in the immediate postnatal period when the hypothalamic circuits are being formed—the timing of this “leptin surge” in humans is contemporaneous with the development and maturation of the feeding neurocircuits within the hypothalamus.

Rodents are not the only species that experience a postnatal leptin surge. Lambs increase leptin concentration in the early postnatal period. Plasma leptin increases from 1 to 4 ng/mL in lambs between P6 and P10 (388-390) born to ewes fed standard chow. In contrast to rodents, gestational undernutrition or overnutrition eliminates the leptin surge in lambs (at least in the first 10 postnatal days, when the measurements were made); both nutritional manipulations lead to increased food intake and adiposity in the offspring compared with controls (391, 392). In postnatal cows, plasma leptin increases from birth to P2 and then decreases until P16 of age. One study reported stable circulating leptin concentrations in P3 to P8 postnatal pigs while another group found that plasma leptin during the first measurement at P5 was higher than that at P9, 16 or 23. In the aggregate, there is some evidence supporting the presence of postnatal leptin surge in neonatal lambs, cows, and pigs, but it is unclear whether the equivalent may be occurring in utero in nonhuman primates. While longitudinal leptin concentrations in fetal and neonatal nonhuman primates have not been reported to an extent that would allow identification of a postnatal leptin surge per se, similar to humans, plasma concentrations of leptin increase in fetuses of these animals with the progression of gestation (393). As in humans, maternal plasma leptin concentrations during gestation increase in nonhuman primates with the progression of pregnancy and decline after parturition. The magnitude of the pregnancy-associated leptin increase in nonhuman primates is higher than in humans in both absolute concentration and relative to nonpregnant state.

The precipitous decrease in leptin at birth may function as a signal to promote feeding in the neonate. This hypothesis is consistent with the observation that in mice during the postnatal leptin surge, leptin does not elicit anorexic response; mice do not decrease food intake in response to exogenous leptin prior to ∼P15 (166). Immediately after leptin surge (coincidental with maturing of neuronal feeding circuits), mice start to respond to leptin by reducing energy intake. As discussed above, leptin has a dual role as a neurotrophic factor and a regulator of energy homeostasis; the leptin surge may be the inflection point between the 2 functions; until the leptin surge (including the period of elevated leptin), leptin plays a neurotrophic role and, immediately post leptin surge, after plasma leptin concentrations drop to the levels proportional to fat mass, leptin switches primarily to its role as fat mass regulator.

In humans, circulating fetal leptin concentration correlates with birth weight but not with maternal circulating leptin concentrations, and maternal circulating leptin concentrations peak midgestation decreasing towards term, while fetal leptin is highest at the end of gestation. This divergence, coupled with the fact that leptin concentrations decrease dramatically after birth, suggests a contribution of placental leptin production to fetal circulating leptin. It is also possible that there are in utero signals stimulating fetal leptin secretion either from fat or placental tissue and that this stimulation ceases abruptly at birth. Leptin acts as a predictive early starvation warning signal, dropping rapidly at the onset of negative energy balance in adults. The decline in leptin at birth could act as an anticipatory starvation signal for the weight loss that occurs within the first week after birth.

Important Considerations

Critical maternal hormones, including insulin and leptin, do not cross the placenta to any significant degree, therefore, direct effects of these hormones on the developing fetus are unlikely. One possibility is that developmental programming is mediated by increased placental insulin and leptin signaling, inducing epigenetic changes to the placenta directly altering placental function. Potential functional changes include transferring nutrients across the placenta, increasing placental oxidative stress and inflammation. Under maternal hyperglycemic conditions, increased glucose transfer from maternal to fetal circulation leads to fetal hyperinsulinemia which contributes to increased fat accrual and this increased adiposity results in elevated fetal leptin concentrations. It is possible that elevated endogenous fetal insulin and leptin are the primary effectors causing altered CNS development. Alternatively, maternal circulating hormones such as insulin and leptin may be acting as biomarkers for some other unknown factors that do cross the placenta and confer effects on the hypothalamic development and ultimate programming of adiposity.

Understanding the neuromolecular pathways for the developmental programming of adiposity maybe be helpful in better informing pregnant women about gestational obesity and diabetes consequences as well as for the discovery of new therapeutics to treat transgenerational obesity. Numerous studies indicate that pregnancy is the 1 time during which people are more motivated to modify difficult behavior for the betterment of their offspring. Current interventions are limited primarily to lifestyle changes (dietary and exercise) during pregnancy. However, these interventions are minimally effective at improving progeny outcomes such as decreasing the risk of high birth weight (394) and decreasing the rate of fetal adipose deposition during pregnancy (395). Other studies have not found improvement in the infant birth weight following maternal lifestyle interventions (396). Reducing body weight and improving metabolic health in women prior to conception could have the most profound impact, not only on the maternal and progeny outcomes, but also on the health of future generations.

Conclusions

In this review we discuss neuromolecular mechanisms by which aspects of nutritional status during early life can induce persistent changes within the CNS which influence subsequent weight homeostasis. Developmental overnutrition in rodents via maternal HFD feeding or postnatal overfeeding by litter size reduction leads to obesity in adulthood. The mechanisms by which this programming of body weight occurs are complex. The data currently available suggest that attention to quantitative and qualitative aspects of maternal body mass/composition and nutrition during the third trimester and to neonatal nutrition could impact long-term susceptibility to obesity by effects on the fine structure of specific regions of the brain. Large clinical studies to define and quantify these factors may not be feasible or worthwhile, but animal models could point to specific mechanisms that could inform focused hypotheses that would be testable in limited numbers of human subjects.

Abbreviations

ADP

air displacement plethysmography

AGA

appropriate for gestational age

AgRP

agouti-related protein

AP

area postrema

ARH

arcuate nucleus of the hypothalamus

BMI

body mass index

CNS

central nervous system

DA

dopamine

DIO

diet-induced obesity

GCK

glucokinase

GDM

gestational diabetes

GLP-1

glucagon-like peptide-1

HFD

high-fat diet

InsR

insulin receptor

IR

insulin resistance

IUGR

intrauterine growth restriction

Lepr

leptin receptor

LGA

large for gestational age

MODY

maturity-onset diabetes of the young

NAcc

nucleus accumbens

NPY

neuropeptide Y

NTS

nucleus of solitary tract

POMC

proopiomelanocortin

SGA

small for gestational age

T1D

type 1 diabetes

T2D

type 2 diabetes

VTA

ventral tegmental area

Funding

This work was supported by research grants from the National Institutes of Health: R01 DK052431 to R.L.L., and the NY Nutrition and Obesity Research Center: P30 DK026687 to R.L.L.; and Pfizer in the form of an Pfizer Obesity ASPIRE grant to C.A.L.

Disclosures

The authors declare no competing financial interests.