Fetal Programming of Human Energy Homeostasis Brain Networks: Issues and Considerations

Abstract

In this paper we present a transdisciplinary framework and testable hypotheses regarding the process of fetal programming of energy homeostasis brain circuitry. Our model proposes that key aspects of energy homeostasis brain circuitry already are functional by the time of birth (with substantial inter-individual variation); that this phenotypic variation at birth is an important determinant of subsequent susceptibility for energy imbalance and childhood obesity risk; and that this brain circuitry exhibits developmental plasticity, in that it is influenced by conditions during intrauterine life, particularly maternal-placental-fetal endocrine, immune/inflammatory and metabolic processes and their upstream determinants. We review evidence that supports the scientific premise for each element of this formulation, identify future research directions, particularly recent advances that may facilitate a better quantification of the ontogeny of energy homeostasis brain networks, highlight animal and in-vitro based approaches that may better address the determinants of inter-individual variation in energy homeostasis brain networks, and discuss the implications of this formulation for the development of strategies targeted towards the primary prevention of childhood obesity.

1. Introduction and Overview

Obesity represents among the most urgent national and global health challenges due to its high prevalence and established health risks^1–4. Childhood obesity is a particularly grave concern as obese children are substantially more likely to have obesity as adults^5,6 and to develop obesity-related disorders at earlier ages^7,8 and of greater severity^9–11. Understanding the causal origins of obesity requires not only the consideration of the multi-factorial interplay between genetic and environmental risk factors, but also the temporal sequence of conditional probabilities underlying its emergence. That is, the effects of any given risk factor at any given stage of the life span are conditioned upon the enduring effects of prior exposures at prior life stages.

Over the past several years there has been substantial interest in the role of developmental processes, over and above those of genetic and lifestyle factors, in shaping individual differences in susceptibility for developing obesity^12–16. Indeed, a rapidly growing and convergent body of epidemiological, clinical and experimental evidence in humans and animals suggests that the origins of obesity can be traced, in part, back to developmental processes occurring during the intrauterine period of life, at which time the developing embryo/fetus responds to ‘suboptimal’ conditions by producing structural and functional changes in cells, tissues and organ systems that persist across the life span and modulate susceptibility for many complex common disorders (i.e., the concept of fetal, or developmental, programming of health and disease risk). To date, research on fetal programming of obesity has focused largely on processes and mechanisms within peripheral cells, tissues and organ systems, such as adipocyte¹⁷, pancreas¹⁸, liver^19–21 and muscle^22,23 biology. While this focus is entirely justified, we suggest there is yet another system of critical importance that also warrants attention in the context of fetal programming: the brain circuitry that underlies energy balance homeostasis.

Energy balance homeostasis refers to the phenomenon that governs homeostatic regulation, or synchronization, between energy intake (EI) and energy expenditure (EE). The importance of the brain, and specifically hypothalamic-limbic-cortical brain circuitry, in the regulation of energy homeostasis is well-established. Differences have been described in this brain circuitry between normal-weight individuals and those diagnosed with obesity, however, it is not yet clear whether these differences are a cause, consequence, or both, of the obese state. In light of this fundamental gap in our understanding, and based on the growing evidence that supports the phenomenon of fetal programming of the brain (but has not yet focused specifically on energy balance homeostasis circuitry, particularly beyond the hypothalamus), we propose here a conceptual, trans-disciplinary framework and articulate testable hypotheses regarding the process of fetal programming of energy homeostasis brain circuitry along the developmental trajectory to obesity (excess adiposity) (see Figure 1). This framework is built on converging evidence that overlapping mechanisms of fetal programming of the brain and peripheral systems may play a substantial role in explaining inter-individual variation in the structure, function, and connectivity of energy homeostatic-relevant brain circuitry at birth, and its subsequent implications for the trajectory and magnitude of fat gain. Our conceptual model proposes that a) key aspects of energy homeostasis brain circuitry already are established and functional by the time of birth; b) there is substantial structural and functional inter-individual variation in this circuitry by the time of birth; c) this phenotypic variation at birth is a determinant of subsequent susceptibility for energy imbalance and is hypothesized to be prospectively associated with markers of childhood and adult obesity risk (e.g., change in adiposity over the early postnatal period and persistence of the effect into childhood); and d) this newborn brain circuitry exhibits developmental plasticity, in that it is influenced by conditions during the intrauterine life, particularly key maternal-placental-fetal endocrine, immune/inflammatory, metabolic and oxidative processes across gestation and their upstream determinants.

Figure 1. Conceptual Model.

The conceptual model incorporates the role of fetal programming of the setpoint of energy homeostasis-related brain circuitry at birth into the process of fetal programming of offspring body composition.

In this perspectives paper, we first review the concept of energy balance and provide a brief description of the role of the brain in regulating energy balance. We then review and synthesize evidence that examines the premise that energy homeostasis-related brain circuitry is already established and functional by the time of birth, with meaningful inter-individual variation. Next, we review and synthesize evidence underlying the fetal programming of energy homeostasis-related brain circuitry hypothesis. Finally, we identify questions, issues and research directions aimed at testing key elements of our model and discuss their potential implications for the development of strategies ultimately targeted towards the primary prevention of childhood obesity.

2. The Concept of Energy Balance Homeostasis

Human physiology conforms to the first law of thermodynamics, which states that energy cannot be created or destroyed; it can only be transformed from one form to another. It then follows that the rate of change in the body’s energy stores (ES) is equal to the difference between the effective rates of energy intake (EI) and energy expenditure (EE). EI occurs primarily via the consumption of the three major macronutrient food groups - carbohydrate, protein, and fat²⁴. Absorbed carbohydrates, proteins, and fats are then transformed in vivo to substrates that ultimately are either oxidized to produce metabolically useful energy that drives biological processes such as growth, maintenance, physical activity, etc., (EE) or they are stored (ES). Any imbalance between the intake (EI) and utilization (EE) of these macronutrients results in a non-zero balance (ES) that will lead to an alteration in body composition in the form of tissue contraction/expansion via triglycerides in adipocytes (dominant form), and protein and intracellular glycogen in skeletal muscle. Thus, the long-term stability of body weight and composition is considered a marker of being in a state of energy balance. The development of obesity necessitates positive energy imbalance over and above that required for normal growth and development from conception through adolescence. In lean individuals and those diagnosed with obesity, a state of energy balance ultimately occurs over the long term, but in individuals with obesity this steady state is simply achieved with a higher amount of body fat.

In the context of an acute reduction in EI relative to EE it also is clear that evolutionary adaptations over time have favored a number of physiological responses acting against weight loss to promote weight regain towards a homeostatic set point²⁵. Such weight-preserving mechanisms are thought to be equally present in lean individuals and those with obesity despite the over-abundance of energy stores present in obesity. Thus, once established, the condition of excess adiposity is difficult to reverse. In contrast to the oversimplified “eat less move more” solution to addressing the problem of unwanted weight gain and obesity, the U.S. Endocrine Society has taken the position, based on the convergence of evidence, that obesity should now be conceptualized as a disorder of the energy homeostasis system, rather than simply arising from the accumulation of excess weight. Moreover, the Society’s position paper on this issue has emphasized the need to elucidate underlying mechanisms, with a major focus on the influence of developmental processes²⁶.

3. Components of the Energy Balance Homeostasis System

Energy homeostasis is centrally regulated within the central nervous system (CNS), with peripheral inputs and outputs (see Figure 2)²⁷. Communication within and between the central and peripheral systems occurs via afferent and efferent signaling involving endocrine and other effectors. The CNS relies on long- (tonic) and short-term (episodic) peripheral inputs to obtain information regarding the state of energy intake (nutrients) and energy stores (body composition). Outputs from the CNS guide the adjustments to energy expenditure and caloric intake that are needed to achieve energy balance. Ultimately, the majority of peripheral signals converge on the hypothalamus where, in concert with other cortical and subcortical regions, these signals are integrated with gustatory (e.g., olfaction), interoceptive (e.g., gastric expansion), behavioral (e.g., stress) and social (e.g., peer pressure) contextual information. Below, we briefly review the role of the brain in maintaining energy homeostasis, beginning with the hypothalamus and then vertically expanding beyond the hypothalamus to include higher-order brain structures critical to feedback and feed-forward energy homeostasis signaling, including the ventral striatum, limbic, insula and anterior cingulate. Because these individual structures do not operate in isolation, we organize the discussion into four canonical energy homeostasis brain networks: satiety, visceral, reward, and salience networks. Finally, we frame the development of these systems in the context of the effects of the fetal programming process, with a focus on those elements of this circuitry that exhibit developmental plasticity (wherein maternal-placental-fetal biological processes during fetal life forecast its observed variation at the time of birth).

Figure 2. Hormonal Inputs to the Central Regulation of Energy Homeostasis.

Hormonal orexegenic input to the CNS originates from stomach contraction, specific nutrient sensing in the gut, and peripheral energy substrate sensing. Anorexigenic signaling arises from stomach distension, nutrient sensing in the gut, and adipocytes. The CNS integrates these signals in order to regulate energy homeostasis through modulation of energy intake and expenditure.

3.1. The satiety network

Hypothalamic nuclei constitute the principal CNS regulators of energy balance homeostasis^28–30. Early lesion studies were instrumental in establishing the key satiety network nuclei underlying hyperphagic (i.e., paraventricular, dorsomedial and ventromedial nuclei) and hypophagic (i.e., lateral nuclei) behaviors. The melanocortin system of peptide hormones functioning within the hypothalamus is the key regulator of energy intake (i.e., promotion or suppression of appetite signaling) via the sensing of current energy status in peripheral organs³¹. Hormonal signaling from adipose tissue (leptin, representing satiety) and the gastrointestinal tract (ghrelin, representing hunger) confers homeostatic status to melanocortin receptors in orexegenic (agouti-related peptide/neuropeptide Y [AgRP/NPY]) and anorexigenic (proopiomelanocortin [POMC]) neurons residing in the arcuate nucleus of the hypothalamus. Ghrelin promotes AgRP/NPY activation and inhibits POMC activation, providing higher order brain systems with an increased appetite signal. Conversely, leptin inhibits AgRP/NPY activity and promotes POMC activation, decreasing the CNS’s motivation for energy intake while increasing signaling to upregulate energy expenditure. In coordination with the hypothalamus, the brainstem (e.g., nucleus of the solitary tract [NTS]) acts as a co-regulator of energy balance through the peripheral sensing of nutrients³², descending projections from the hypothalamus³³, and interoceptive (see Section 3.3 Visceral network below) and nutrient signaling ascending from the gastrointestinal tract³⁴.

The importance of POMC (appetite suppressing, expenditure promoting) and AgRP (appetite promoting, expenditure suppressing) neurons in the modulation and regulation of energy balance is well established. However, the full relevance of direct communication between these cell types is not yet well understood, as AgRP neurons do not appear to contribute to spontaneous input to POMC neurons³⁵, and while optogenetic stimulation of AgRP does inhibit POMC neurons, such inhibition is not necessary for the induction of feeding behaviors³⁶. It is clear, however, that other intrahypothalamic signaling pathways are of relevance to energy homeostasis as postprandial POMC signaling to the paraventricular nucleus (PVN) via α-melanocyte-stimulating hormone and melanocortin 3 and 4 receptors results in decreased appetite and increased energy expenditure³⁷, and AgRP signaling to the PVN promotes appetite and decreases energy expenditure³⁸.

Central regulation of energy expenditure to meet the demands of thermoregulation, basal metabolism and physical activity also is accomplished, in part, via hormonal signaling to the hypothalamus. For example, thermoregulation in response to exogenous cold exposure is achieved through the activation of the hypothalamic-thyroid-adrenal axis and the downstream effects of thyroid (T3/T4) hormones in upregulating cellular metabolism and heat production in peripheral tissues³⁹. The hypothalamus also modulates cellular metabolism in response to endogenous conditions reflecting over-nutrition. For example, increased diet induced thermogenesis occurs in response to high-caloric intake, whereas fasting produces a reduction in energy expenditure. As noted previously, these energy balance mechanisms are asymmetric, in that they are more sensitive to weight loss than weight gain, thus emphasizing the importance of higher cognitive regulation of energy intake in the context of the abundance of western diets (whose influence is believed to further mask or reduce the effectiveness of energy balance set points^40,41).

3.2. Reward network

Over and beyond the hypothalamus, the ultimate ‘decision’ of what, when, and how much to eat implicates the recruitment of several inter-linked cortical and subcortical structures (see Figure 3), of which the mesolimbic dopaminergic (reward) pathway plays a crucial role^42–46. The reward network is the catalyst for motivated behavior in the context of feeding, in that learned associations with foods that are high in energy content provide positive reinforcement through the release of dopamine⁴⁷. Direct evidence comes from lesion studies demonstrating a decreased desire for food after experimental damage to reward areas⁴⁸. The reward network relies on projections between dopaminergic neurons in the ventral tegmentum area (VTA) and the nucleus accumbens, but also relies on modulatory connections between the amygdala⁴⁹, hippocampus^50,51, insula⁵², prefrontal⁴⁹/orbitofrontal cortex, caudate⁵³, putamen, thalamus and pituitary gland⁵⁴. Upon novel exposure to food rewards, dopamine is released in the nucleus accumbens. Prolonged exposure habituates such a response, ultimately transferring the reward response on to food-related stimuli (e.g., images, smells), and thus becoming a potent predictor of reward^55–58. From an evolutionary perspective, environmental stimuli motivating food seeking behaviors clearly are adaptive. Yet, in the modern obesogenic environment, in which such stimuli are ubiquitous, they have often become a liability. Because dopamine and the reward network play such a prominent role in motivating food intake behaviors⁵⁹, it has been suggested in the context of the problem of obesity that this feature may now constitute an affliction similar to other addictive behaviors including narcotics abuse⁶⁰ and reflect an inherent obstacle to weight loss.

Figure 3. Energy Homeostasis Relevant Brain Regions.

Highlighted brain regions are organized based on their network membership (color outline).

3.3. Visceral network

The visceral/gustatory/interoception network represents another key component⁶¹ of the brain circuitry implicated in caloric intake. This network integrates sensory information about internal bodily states (e.g., insula activation via gastric distension⁶²), food-related cues (e.g., olfactory/gustatory cortex activation via olfaction/taste), and their interaction⁶³, enabling feedback (setpoint error detection and correction) and feedforward (anticipation of future energy availability) input⁶⁴. The olfactory system consists of signaling from smell receptors (chemosensation) in the nasal cavity to the primary/secondary olfactory cortex via the olfactory bulb. Conversely, the gustatory system consists of an ascending pathway originating from taste receptors in the tongue through to the rostral division of the NTS, the ventroposterior medial nucleus of the thalamus (VPMpc), and terminating in the gustatory cortex located within the insula. The role of the insula in modulating complex feeding behaviors in the context of obesity is well established^65–68. In addition to gustation, the insula plays an essential role in somatosensation, interoception, reward/addiction and emotion⁶⁹. Thus, the insula is believed to integrate multisensory input^66,70. Further, the insula has dense, reciprocal structural connections to the amygdala⁷¹, a brain structure involved in taste and textural coding⁷², and therefore hypothesized to mediate the processing of taste and reward^73,74.

3.4. Salience network

The salience network (anterior cingulate cortex [ACC], anterior insula⁷⁵), in concert with satiety-related input from the hypothalamus, is believed to underlie the drive for the prioritization of sustenance and the de-prioritization of other functional tasks of less importance⁷⁶. The salience network is centered on bilateral orbital frontoinsular and dorsal anterior cingulate cortices⁷⁷, with connectivity to the VTA, thalamus, hypothalamus and amygdala. This network is also likely important for switching between internal (default mode network) and external (central executive network) modes of thought^78,79, a property consistent with and necessary in the context of feeding behaviors⁸⁰. Critically, structural and functional variation in the salience network has been shown to predict food intake under stressful conditions, has been associated with familial obesity risk, and differences have been described between normal-weight and individuals with obesity⁸¹. This then suggests that the integrity of the salience network is functionally relevant in the context of energy homeostasis.

4. The Ontogeny of Inter-individual Variation in Energy Homeostasis Brain Networks

Our framework postulates that energy homeostasis brain networks are already established at the time of birth, albeit further maturation and finetuning occurs postnatally, and that the inter-individual variation in their functional variation is prospectively associated with key markers of subsequent obesity risk (e.g., postnatal feeding behaviors and infant weight/fat gain). In this section we describe animal and, where available, human evidence that supports the scientific premise of these postulates (see Table 1 for summary).

Table 1. Summary of Evidence Supporting Structural and Functional Inter-individual Variation in Newborn Energy Homeostasis Brain Circuitry.

Energy Homeostasis Relevant Brain Networks and Structures	Model species	Evidence supporting notion that function is present by the time of birth?	Evidence supporting notion that variation at birth is prospectively associated with obesity risk?
Satiety: Hypothalamus	Rodents, non-human primates	Yes See: Baquero 2014; Bouret 2012; Caron 2010; Gali Ramamoorthy 2015; Greyson 2006; Steculorum 2011	Yes See: Barrand 2017; Gali Ramamoorthy 2015; Plagemann 2017
Reward: VTA, NA, Amygdala, Hippocampus, Insula, Prefrontal/Orbitofrontal Cortex, Caudate, Putamen, Thalamus, Pituitary Gland	Rodents, humans	Yes See: Antonopoulos 2002; Van den Heuvel 2008; Meng 1999; Contreras 2013; Loos 2019	Yes See: Ayres 2012; Dalle Molle 2015; Dalle Molle 2016; Laureano 2016; Silveira 2018
Visceral: Insula, Olfactory and Gustatory Cortices, NTS, VPMpc	Rodents, humans	Yes See: Rinaman 1999; Swithers 1989	Yes See: Rasmussen 2017
Salience: ACC, anterior Insula	Humans	Yes See: Gao 2016; Li 2016; Rudolph 2018;	Unknown

4.1. Ontogeny of the Satiety Network

As reviewed above in Section 3.1, the hypothalamus represents the hub of the primary appetite regulatory network. A considerable body of evidence, primarily in rodents, suggests that the hypothalamus takes shape beginning with the formation of neuronal progenitor cells appearing shortly after the closure of the neural tube, and this is followed within days by cellular specification into hypothalamic neurons (e.g., POMC, VMH neurons). Axonal growth occurs throughout gestation from differentiated cell types to various target regions. Importantly, inter-individual variation is evident in hypothalamic expression of genes encoding extracellular matrix (ECM) protein production (e.g., Col1a1, Col3a1) at postnatal day 10 in rodents (roughly equivalent to the human newborn period), accompanied by subsequent variation in postnatal body fat accrual, thus indicating functionally meaningful inter-individual phenotypic variation at this early life^82,83 stage. Interestingly, while the rodent hypothalamus responds functionally at birth^84–87 to metabolic stimuli, appetite is not yet modulated by acute injections of leptin or ghrelin^84,88. Further, a functional response to food deprivation (i.e., NPY expression) is not observed until between 10 to 20 days postnatal age. However, in rodents structural connectivity originating within and between the hypothalamus is largely incomplete at birth, whereas in precocial species such as non-human primates⁸⁹ and humans these connections are comparatively mature at birth⁹⁰. In contrast to the ontogeny of the rodent satiety network, fetal hypothalamic NPY and POMC have been shown to be well expressed in baboons in late gestation, even prior to birth. Moreover, an appetite-promoting functional phenotype (i.e., increased NPY and decreased POMC expression in the hypothalamus) is established by late gestation in fetuses exposed to in utero caloric restriction. Thus, collectively, this evidence suggests that in precocial species the satiety network exhibits structural and functional maturity by the time of birth.

4.2. Ontogeny of the Reward Network

In rodents, dopaminergic projections from the VTA to the Nucleus Accumbens are in place by the time of birth and rapidly expand over the first few weeks of postnatal life^91–93. Recent evidence in rodents also supports substantial inter-individual variation in Dopamine Receptor D2 expression within the VTA and the Nucleus Accumbens as early as postnatal day 10⁹⁴. In humans, the connections of the reward network are largely in place by the end of the gestational period, with dopamine transporter immunoreactivity reaching a functional state by 40-weeks gestation⁹⁵. And based on the presence of a diverse set of human infant behaviors in response to gustatory and olfactory stimuli, it appears that key aspects of learned behaviors and reward-based anticipatory responses to food cues are established and functional by birth. For example, it is well established that breast milk odor produces an analgesic effect on term and preterm newborns⁹⁶. Based on the consideration that the aversiveness of pain is encoded by reward mesocorticolimbic circuitry⁹⁷, it seems reasonable to suggest that aspects of this circuitry, particularly in the context of feeding, are established and functional in humans by late gestation. Moreover, the olfaction of fatty acids from maternal amniotic fluid is sufficient to elicit appetitive responses within 24 hours of birth^98,99. Here, mesocorticolimbic circuitry appears to underlie such responses, as maternal colostrum and novel odorant¹⁰⁰ olfaction is sufficient for the activation of orbitofrontal cortical regions in newborns^101–104. Thus, these findings support the premise that key aspects of conditioned learned associations⁵⁰ with stimuli are present during gestation and are functional at the time of birth. Moreover, it is clear that there is inter-individual variation in newborns in behavioral preferences for flavors (e.g., garlic, anise) consumed by the mother with volatile compounds present in the amniotic fluid during pregnancy^105–110, further supporting the premise of functional maturity in its underlying neurocircuitry. In addition to flavor preference, recent evidence suggests fetal growth conditions (small for gestational age and intrauterine growth restriction) are associated with variation in early life dopamine signaling in the nucleus accumbens^111,112 and newborn hedonic responses to sweet stimuli¹¹³ with long term consequences on the preference for highly palatable foods (e.g., high fat and sugar content) ^114,115. While more complex reward-based feeding behaviors begin to emerge around the weaning period^116,117, we suggest that reward circuitry in humans in response to gustatory and olfactory cues is functional by birth and acts in coordination with limbic regions through dopamine release.

4.3. Ontogeny of the Visceral Network

In rodents, visceral network connectivity between the gastrointestinal tract and the gustatory pathway are structurally in place as early as the first postnatal day of life^118–120, with evidence of the presence of direct control of ingestive behavior (e.g., postingestive but preabsorptive signaling). The visceral network continues to mature postnatally through changing biochemical signaling¹²¹ and further synaptic organization¹²² and appears to play a clear role in energy homeostasis. In a recent report in human newborns, we demonstrated a prospective, inverse association between the volume of the newborn insula (a key component for sensing the physiological state of visceral organs) and fat gain during early life (a key marker of childhood obesity risk) ¹²³. We note this finding is consistent with those evident in human adults linking insula gray matter volume and activity with gastric distension, the obese state^124–127 or future risk of developing obesity¹²⁸, and metabolic factors associated with obesity, including blood leptin concentration¹²⁹.

4.4. Ontogeny of the Salience Network

Although the developmental ontogeny in prenatal life of the salience network in the context of feeding and childhood obesity risk has yet to be described, we suggest, based on three lines of evidence, that it likely is structurally and functionally intact by the time of birth, with meaningful inter-individual variation in the context of obesity risk. First, as discussed previously in Section 3.4, inter-individual variation in adults in the functional connectivity of the salience network has consistently been shown to be associated with several obesity-related phenotypes¹³⁰. Second, the salience network is structurally supported by multiple white matter pathways¹³¹, (e.g., Uncinate Fasciculus¹³², corpus callosum¹³¹) that develop in humans during the mid-gestation period. And third, the functional connectivity of the salience network is evident by birth¹³³, and the strength of its connectivity is associated with maternal BMI¹³⁴ and inflammation¹³⁵ during pregnancy, supporting the premise that it exhibits meaningful inter-individual variation at birth.

Table 1 provides a summary of the above evidence supporting the presence of structural and functional inter-individual variation in brain circuitry at birth and its prospective associations with offspring obesity risk. It is organized by network and depicts the model species studied and the presence or absence of evidence.

5. Fetal Programming of Energy Homeostasis Brain Networks

A rapidly growing and convergent body of epidemiological, clinical and experimental evidence in humans and animals suggests that the origins of obesity can be traced, in part, back to developmental processes occurring during the intrauterine period of life, at which time the developing embryo/fetus responds to ‘suboptimal’ conditions by producing structural and functional changes in cells, tissues and organ systems that persist across the life span and modulate susceptibility for many complex common disorders. This process, commonly referred to as fetal programming, is adaptive from an evolutionary perspective, but at the individual level may confer a trade-off favoring short-term survival/reproductive fitness at the long-term cost of susceptibility to complex common disorders. The importance of fetal programming is supported by evidence that exposures to suboptimal maternal nutrition, suboptimal maternal metabolic health, excess maternal stress, and excess maternal inflammation during intrauterine life have the potential to alter neurodevelopmental outcomes and obesity risk throughout the life course. Specifically, several convergent lines of evidence suggest that key brain structures involved in energy homeostasis (e.g., hypothalamus, amygdala and hippocampus) exhibit plasticity in response to maternal conditions during pregnancy. However, with the exception of the hypothalamus, this literature has, to date, largely focused on observations in the context of neuropsychiatric disorder risk¹³⁶. For the remainder of this section, we highlight standing and recent developments in which maternal states and conditions influence offspring body composition and energy homeostasis relevant brain circuitry in a manner consistent with one another. These findings support our premise that fetal programming of energy homeostasis brain circuitry may, in part, mediate the influence of maternal conditions on offspring obesity risk.

5.1.1. Sociodemographic factors

In humans, offspring adiposity risk is associated with several sociodemographic factors, including maternal age at conception, race/ethnicity, and socioeconomic status. Younger maternal age at conception has been linked with child outcomes including low birthweight¹³⁷, an increased obesity risk¹³⁸, and increased risk for central adiposity¹³⁹. Race and ethnicity are also associated with birthweight¹⁴⁰, newborn adiposity¹⁴¹, and early life feeding behaviors. It is likely that economic deprivation may explain some of the racial and ethnic disparities in infant birth outcomes (e.g., birthweight), feeding behaviors, and subsequent childhood obesity^142–145 prevalence.

In addition to obesity risk, sociodemographic states and conditions have been associated with energy homeostasis-relevant brain circuitry and development. For example, it is well established that children from high socioeconomic status (SES) families perform better on a number of cognitive tasks¹⁴⁶, including executive functioning tasks believed to underlie aspects of childhood obesity¹⁴⁷. In addition, current evidence suggests that structural brain development (e.g., insula, frontal cortex) is SES-dependent and mediates the association between SES and cognitive performance, particularly in executive function domains¹⁴⁸. Further, the effects of SES on brain development can be traced back to well before childhood because SES has been shown to influence growth and functional connectivity of the frontal cortex already in human infancy^149,150. In fact, recent evidence has emerged that the effects of SES can be traced back to fetal brain development¹⁵¹, particularly in the amygdala, a structure highly integrated into the central control of energy homeostasis. Certainly, because SES could not exert a direct effect on the fetal brain, these findings motivate the needs for a mechanistic understanding of the role of putative mediators such as gestational biology on programming homeostasis-relevant brain circuitry throughout the life course.

5.1.2. Biophysical/clinical factors

Maternal biophysical and clinical factors such as obesity, excess gestational weight gain and gestational diabetes are associated with greater offspring birth weight and fat mass. The metabolic consequences of over-nutrition during the fetal period are also believed to track beyond infancy, as children born to mothers with obesity or diabetes continue to have increased obesity risk^152–154. Additionally, not only are offspring of over-nutrition pregnancies larger at birth and in later life¹⁵⁵, they are also more likely to suffer from adverse metabolic conditions such as hypertension and diabetes^156,157 as adults. While genetic variation (DNA base pair variation) certainly plays a role in the intergenerational transmission of obesity and obesity-related outcomes, it is clear that the intrauterine milieu plays an important role above and beyond genetics. For example, one Swedish national cohort study of pregnancy following bariatric surgery found a marked decrease in pre-pregnancy BMI (from roughly 44 to 30, on average) accompanied by a reduced risk of gestational diabetes and excessive fetal growth¹⁵⁸. Importantly though, the Swedish study and others^159,160 have found an increased risk for small-for-gestational-age offspring following gastric bypass, suggesting that such restrictive surgery may result in a milieu of nutritional deficiency during pregnancy. In a recent meta-analysis of natural weight gain between successive pregnancies¹⁶¹, it was estimated that a BMI increase of greater than three units between pregnancies was associated with a 63% higher risk (adjusted odds ratio) of large-for-gestational age offspring.

In addition to offspring obesity outcomes across the lifespan, there is emerging evidence that maternal obesity is associated with offspring cognitive^{162 163 164,165} and neurodevelopmental disorder¹⁶⁶ outcomes, and plays a specific role in shaping newborn functional and structural brain connectivity patterns within key aspects of the satiety, salience, and reward networks. Indeed, recent neuroimaging evidence supports a direct role for the hypothalamus as a mediator between maternal nutrition status and susceptibility to weight gain in childhood¹⁶⁷. Within interoception, salience and reward systems^134,168,169, maternal obesity has also been linked with fetal and newborn human offspring functional and structural connectivity. Because salience and reward networks play a requisite role in assigning motivation and importance based on integrated input from the hypothalamus by de-prioritizing other⁷⁶ functional tasks of less importance, it is plausible that maternal obesity (through associated gestational biological factors) also plays a meaningful role in programming aspects of central feeding behaviors that then contribute to obesity throughout the human life-course.

5.1.3. Maternal Diet/Nutrition

The influence of maternal nutrition on offspring brain function in the context of obesity is well established. In rodents, a maternal high fat diet alters offspring feeding behavior and predisposes to an obese phenotype^170–172. This effect is partially mediated by POMC expression in the hypothalamus and likely programmed via CpG methylation in a POMC promoter region. Other recent findings support endoplasmic reticulum stress brought on by fetal exposure to maternal free fatty acid concentrations as a cause of altered hypothalamic development, with metabolic consequences in postnatal life¹⁷³. Beyond the hypothalamus, maternal diet-induced obesity models have demonstrated altered feeding behavior in rodent offspring, favoring highly palatable foods (e.g., high-fat, high-sugar) accompanied by altered aspects of limbic¹⁷⁴ and reward^175–179 circuitry. Strikingly, maternal under-nutrition during pregnancy also promotes an increased appetite for high fat foods in offspring¹⁸⁰. Because preference for high fat foods is commonly considered to be associated with reward circuitry function¹⁸¹, it is likely that under-/over-nutrition models share a common programming target in reward circuitry, albeit through a different mechanism (e.g., differential epigenetic modification)¹⁸² of nutrition status. Potentially, this may also be attributed to separate motivational mechanisms where exposure to excess nutrition could influence offspring hedonic liking (opioid and cannabinoid) via in utero priming, whereas exposure to undernutrition could influence the fetal development of circuitry underlying homeostatic wanting (dopamine)¹⁸³ in preparation for an ex utero environment limited in food resources.

In humans, higher maternal protein intake (relative to carbohydrate or fat intake) in mid-pregnancy is associated with lower visceral adiposity in neonates¹⁸⁴. In the same cohort, maternal sugar intake was positively associated with peak BMI in infancy, and maternal carbohydrate and sugar intakes were positively associated with BMI between two to four years of age¹⁸⁵. In a separate study, an imbalance in maternal micronutrient intake (excess folic acid in a B₁₂ deficient diet) during pregnancy was associated with whole-brain oxidative stress. In addition, in a large prospective mother-child cohort study, researchers documented a negative association between maternal depression during pregnancy and offspring IQ, mediated by the maternal diet at 32-weeks gestation¹⁸⁶. Collectively, these findings support the plausibility that aspects of maternal behavior (e.g., diet and nutrition) influence early life adiposity outcomes through changes to appetite regulating networks (e.g., satiety, limbic and reward circuitry) during fetal brain development.

5.1.4. Maternal-Placental-Fetal Metabolic, Endocrine and Immune/Inflammatory Biology

The effects of maternal states and conditions during pregnancy on fetal programming of the brain and peripheral systems are ultimately mediated via gestational biology. Over and above the role of gestational biology in mediating the effects of maternal diet/nutrition during pregnancy (as discussed above) it is evident that aspects of gestational biology also may mediate the effects of other key prenatal states and conditions on fetal brain energy homeostasis-related phenotypes. Here, we discuss these offspring outcomes in relation to fetal exposures to three key aspects of maternal-placental-fetal biology: metabolic, endocrine, and immune/inflammatory gestational state.

5.1.4.1. Maternal Metabolic State

Glucose and insulin are important markers of maternal metabolic state, and of consequence to offspring health and development, including the brain. In humans, recent evidence supports the notion that maternal gestational diabetes status and insulin sensitivity¹⁸⁷ are associated with fetal brain function (auditory evoked response latency) following a maternal oral-glucose stimulus¹⁸⁸. While maternal insulin does not cross the placental barrier, glucose constitutes a predominant form of energy substrate provided to the developing fetus by crossing the placenta. The mechanisms that regulate the differential effect of maternal insulin sensitivity on the fetal brain response to glucose administration are not well established. However, Linder et al observed that insulin-resistant mothers have higher glucose levels accompanied by increased insulin levels in the postprandial state and go on to suggest that as glucose transfer across the placenta increases, hyperinsulinemia (e.g., increased pancreatic insulin production) in the fetus may be induced (Pedersen hypothesis) ¹⁸⁹. As a result, increased insulin levels in the mother correspond with increased insulin levels in the fetal brain compartment. Importantly, insulin within the brain compartment appears sufficient for modulating energy homeostasis brain networks both in utero and postnatally, with ramifications on food intake, body weight, and ultimately obesity^190–193. For example, intravenous insulin elicits changes in brain activity measured using magnetoencephalography¹⁹⁴, and insulin reactivity is associated with limbic activation as measured by fMRI in response to visual food cues¹⁹⁵. Further, recent evidence in animals supports a specific role for brain-based control of glucose homeostasis in Type-2 diabetes¹⁹⁶, further supporting the role of the brain as both a sensor and effector of energy homeostasis. To summarize, the fetus experiences variation in glucose concentration exposure that is dependent on variation in maternal metabolism, which in turn is sufficient to elicit differential brain function in utero¹⁹⁷ likely via fetal insulin production. Thus, because the brain is a regulator of energy homeostasis, the above collectively suggests a pathway through which maternal insulin sensitivity status programs offspring metabolic health, mediated through fetal brain development and function.

5.1.4.2. Maternal endocrine state

Stress plays an important role in obesity; it alters metabolism, affects appetite, and influences dietary intake^198,199. During pregnancy, excess stress has been associated with adverse developmental outcomes²⁰⁰ including obesity risk²⁰¹ and atypical brain development^202–205. The stress hormone cortisol is perhaps the most robust biological indicator of response to a diverse range of stressors, including low socioeconomic status, psychosocial distress, inter-personal conflict, performance pressure, and lack of social support^206–209. Indeed, cortisol’s importance in effecting appetite and metabolism is well established. For example, differentiation of human adipocyte precursor cells in the presence of insulin is promoted by cortisol in a dose-dependent manner and occurs at physiological concentrations in vitro^210,211. Consistent with this observation, prenatal maternal cortisol levels have been positively associated with early life adipose tissue accumulation in human newborns²¹².

With respect to the brain and appetite, several studies have reported associations between fetal cortisol exposure and alterations in brain development in energy homeostasis-relevant brain regions including the amygdala²¹³, hippocampus, insula²¹⁴, frontal cortex²¹⁵, and anterior cingulate²¹⁶. For example, in rodents, dexamethasone treatment during pregnancy has been shown to affect offspring birth weight, and glucocorticoid receptor mRNA expression in the amygdala and hippocampus in a time-dependent manner²¹⁷. More specifically to appetite, induction of corticotropin releasing factor (a hypothalamic precursor of cortisol production) in the maternal brain during late pregnancy predisposes rodent offspring, in a sex-specific manner, to episodes of binge eating²¹⁸ when subjected to a limited-access food paradigm²¹⁹. Such binge eating behavior is accompanied by hypothalamic adaptation in the offspring, as evidenced by hypomethylation of miR1-a and downstream dysregulation of the melanocortin system. Collectively these findings suggest that stress during pregnancy may play a meaningful role in setting up obesity susceptibility by influencing fetal brain development of energy homeostasis-relevant networks.

In addition to nutrition (see above Section 5.1.3), maternal stress during pregnancy is one of the most commonly-studied factors in the context of the fetal programming of brain development. However, the effects of maternal nutrition and stress on offspring brain development are often studied separately despite evidence of a bi-directional relationship between them^220,221. The limited literature that does consider the combined effect of diet and stress seems to point to a high fat diet, or higher dietary intake of targeted nutrients (e.g., antioxidants), as being neurodevelopmentally protective with respect to high levels of stress exposure during pregnancy²²². However, at least one recent study suggests synergistic effects between maternal stress and diet on offspring body weight, food-motivated behaviors, and reward center functioning²²³. Importantly, the interaction between diet and stress during pregnancy appears to consistently affect regions in the brain that are of high relevance to energy homeostasis, including hippocampal morphology²²⁴, nucleus accumbens dopamine release²²³, and the morphology of the prefrontal cortex²²⁵.

In humans, psychosocial stress is positively associated with oxidative stress (OS) ^226,227. Further, correlative studies during pregnancy have demonstrated positive associations of maternal and umbilical cord blood oxidative stress markers²²⁸ with maternal and neonatal oxidative status in small-for-gestational age newborns²²⁹. However, the precise mechanisms are not yet fully understood^230,231. While a moderate increase in OS is an expected physiological component of pregnancy^232,233, excessive maternal OS during pregnancy has been linked with a number of adverse pregnancy outcomes (e.g., growth restriction²³⁴ and gestational diabetes²³⁵) that have a demonstrable influence on offspring metabolic health in later life. In support of the role of OS in fetal programming of energy homeostasis relevant brain regions, several rodent studies have established negative effects of maternal OS on offspring neurogenesis, hippocampal glucocorticoid receptor density, and cognitive functioning, that can be ameliorated by antioxidant administration to the pregnant dam^236–238. Interestingly, in humans, excess maternal adiposity is associated with increased OS^239,240 and has been shown to alter DNA methylation sites relevant for CNS development and appears to reduce limbic white matter microstructure in offspring¹⁶⁹. Thus, OS represents an additional pathway between maternal stressors and alterations to offspring energy homeostasis-relevant neurocircuitry. It should be noted, however, that because direct maternal-fetal transfer of reactive oxygen species is unlikely, OS should be considered as part of a broader pathway involving physiological mediating correlates (e.g., inflammatory processes, elevated glucocorticoid concentrations) ²⁴¹.

5.1.4.3. Maternal immune/inflammatory state

Maternal inflammation is another potentially critical factor in programming of newborn energy homeostasis-relevant brain circuitry (e.g., salience, fronto-limbic). Animal studies have established that prenatal exposure to pro-inflammatory cytokines results in increased fat depots in offspring postnatally²⁴². Further, adult male offspring of lipopolysaccharide-exposed dams have been shown to have higher caloric intake, fat mass, circulating leptin and up-regulation of hippocampal glucocorticoid receptor proteins²⁴³. While healthy fetal brain development and metabolic programming requires a rich set of cell-signaling mechanisms, the literature suggests that both are susceptible to excessively abundant inflammatory cytokines. Animal models demonstrate maternal immune activation (MIA) alone is sufficient to alter early neurodevelopment with long term consequences on the susceptibility to developing neuropsychiatric disorder (e.g., autism, schizophrenia²⁴⁴). In humans, recently emerging evidence points to the importance of interleukin-6 (IL-6) in programming energy homeostasis relevant brain circuitry and behavior. For example, maternal IL-6 concentrations during pregnancy are associated with newborn amygdala volume and connectivity²⁴⁵, fronto-limbic white matter structural connectivity²⁴⁶, and the functional connectome¹³⁵. Importantly, each of these studies used inter-individual variation in the newborn brain circuitry to demonstrate prospective associations with cognitive outcomes that are relevant for feeding behavior including impulse control, cognition, and working memory (a cognitive domain associated with obesity²⁴⁷).

5.1.5. Summary of Fetal Programming of Energy Homeostasis Brain Networks

Based on the above-discussed findings, it is apparent that variation in metabolic function and in brain development share considerable overlap in terms of their developmental origins. Yet, a comprehensive understanding of the relationship between these two domains has not yet been mapped and is potentially of high value to the field of fetal programming of obesity. The recent explosion of research on the fetal origins of obesity has focused primarily on peripheral cells, tissues and organ systems. It is plausible that fetal programming of energy homeostasis brain circuitry likely and additionally plays an important role in mediating the influence of maternal states and conditions on offspring obesity risk.

Table 2 summarizes the above-discussed evidence supporting the developmental plasticity of newborn brain energy homeostasis regulating networks and structures in humans and rodents. It is organized by maternal risk factors that, on one hand, have been prospectively associated with offspring obesity risk, and, on the other hand, have separately been associated with the development of key offspring brain phenotypes that are known to underlie energy homeostasis.

Table 2. Summary of Evidence supporting the presence of developmental plasticity in Newborn Brain Energy Homeostasis Regulating Structures and Networks.

The presented evidence relates to human studies unless otherwise indicated by an asterisk (*) in which the evidence predominantly exists in rodent models.

Maternal Risk Factors that are prospectively associated with Obesity Risk	Brain Networks/Structures/Regions that exhibit Developmental Plasticity
Sociodemographic (e.g., SES)	Executive function (e.g., Insula, Frontal Cortex)148, Amygdala151
Maternal Biophysical (e.g., Obesity, Gestational Weight Gain)	Prefrontal Cortex, Anterior Cingulate Cortex134, 169, Hypothalamus248
Maternal Nutrition*	Indirect evidence for appetite regulating networks (e.g., satiety170, 172, limbic/reward circuitry 171, 174–179), hypothalamus173
Maternal Metabolic State	Non-specific to date. Literature suggests insulin sensitivity-dependent brain function
Maternal Endocrine Stress Biology	Amygdala213, 217*, Hippocampus217, Insula214, Frontal Cortex215, Anterior Cingulate216
Maternal Inflammation	Limbic (Hippocampus, Amygdala)/Frontal Cortex135,245,246

6. Questions, Issues, Considerations, and Future Research Directions

We outline below issues, considerations, and opportunities for future research directions. We begin by discussing recent advances that may facilitate a better quantification in the human newborn brain of the ontogeny of energy homeostasis brain networks at a systems level. Second, we discuss state-of-the-art approaches (recent and proposed) for phenotyping energy expenditure and energy intake in the human newborn. Third, we consider optimal approaches to phenotyping the consequences of variation in newborn energy homeostasis brain networks. Finally, we highlight several novel animal and in-vitro-based approaches that may advance our understanding of the determinants of inter-individual variation in energy homeostasis brain networks (through experimental models that trade off translational generalizability for mechanistic specificity).

6.1.1. Advances in neonatal brain imaging

The key variables of interest in neonatal brain imaging in the context of fetal programming are resolving the structural morphometry (size/shape), structural connectivity (integrity of the physical connections between brain structures/regions/networks), and functional connectivity (the degree to which structures/regions/networks covary in activity) within and between brain networks.

Over the course of the last decade, there has been a dramatic increase in the ability to non-invasively image the developing human brain using non-invasive MRI-based techniques^133,249,250. This methodological advancement holds great promise in increasing our understanding of the developmental ontogeny of energy homeostasis-relevant brain circuitry and its prospective role in shaping propensity for childhood obesity. Specifically, differences are well established between normal weight adults and those with obesity in MRI-based (gray matter (GM) structure, white matter (WM) fiber integrity, and functional connectivity) measures of the hypothalamic-limbic-cortical brain circuitry that regulate energy homeostasis. It is, however, unclear whether the observed difference in this brain circuitry between normal-weight and individuals with obesity is a cause or consequence of the obese state. Identifying neural correlates of childhood obesity risk that predate the influence of the postnatal obesogenic environment and their fetal determinants would advance scientific understanding of the neurobiological underpinnings of obesity. In this section, we summarize and identify potential MRI-based biomarkers of human energy homeostasis brain circuitry in infancy, including structural morphometry, structural connectivity and functional connectivity. Finally, we discuss the importance of and the opportunities in longitudinal imaging of early life neurodevelopment with MRI.

6.1.1.1. Structural Morphometry

GM/WM volume is an indirect measure of the amount and size of neurons, glial cells, and dendritic processes. Decreased GM and WM volume has been regionally associated with excess weight/obesity in adolescents and adults^124,126. Adiposity effects neurotrophic factors such as leptin, insulin and pro-inflammatory cytokines, whose long-term influence may lead to excessive burden on the brain, as evident in elderly populations²⁵¹. Alternatively, it is plausible that volume reductions in areas within energy homeostasis networks may also effect feeding behaviors. Insula volume serves as an ideal example because its importance in obesity is well established^65–67, and it is a brain region known to independently modulate complex-feeding behaviors^64,68. Furthermore, reduced insula GM in adult individuals identified as obese-prone, relative to those categorized as obese-resistant, supports the notion that reduced insula volume is itself a risk factor for future weight gain¹²⁸. This is further supported by data in newborns demonstrating a prospective association between newborn insula volume and increased gains in body fat percentage in the first six-months of postnatal life¹²³. Collectively, these findings support the premise that structural morphometry of the newborn brain represents a promising neurophenotype underlying the functional consequences of variation in energy homeostasis brain circuitry.

6.1.1.2. Structural Connectivity

Diffusion Tensor Imaging (DTI) directly measures the isotropy of water diffusion in the brain and, indirectly, the capacity to transmit efferent and afferent signals between cortical nodes in the brain. In adults, hypothalamic mean diffusion²⁵², and WM anisotropy and volume²⁵³ are associated with obesity. However, the directions of these WM associations are inconsistent and brain-region/tract-dependent²⁵⁴. In normal-weight children relative to those with obesity, Ou et al demonstrated increased fractional anisotropy in tracts related to food intake (e.g., forceps minor, inferior fronto-occipital fasciculus)²⁵⁵, and they suggest fatty acid presence in the extracellular space surrounding the axons as a cause. In contrast to their findings in children, Ou et al have also demonstrated reduced anisotropy in neonatal offspring of mothers with obesity¹⁶⁹. Analogous with obesity-related GM changes, it is possible that the WM deficits observed in adult individuals with obesity are a consequence, as opposed to a cause, of long-term exposures to obesity-related factors. To date, there are no studies characterizing the association between newborn WM phenotypes and postnatal obesity -related outcomes.

6.1.1.3. Functional Connectivity

Resting state functional MRI provides information about how the brain is organized, with “connected” brain regions working more closely in synchrony. Alterations in functional connectivity in and between energy homeostasis brain networks have been observed in overweight and adults with obesity^256–258. In one early study of MR-based hypothalamic connectivity, Wijngaarden et al observed hypothalamic-insula connectivity differences in adult subjects with obesity²⁵⁹. A second study of hypothalamic connectivity identified deficits in overweight/obese adults between both the orbitofrontal cortex and ventral striatum and the medial hypothalamus^260,261. Since these two early studies, several others support nutritional status-dependent connectivity between the hypothalamus and the insula, striatum, and hippocampus^262–265. In the salience network, adult obesity^261,266 and fat intake after sleep deprivation²⁶⁷ have been associated with deficits in ACC-putamen connectivity. Importantly, similar circuitry (ACC-prefrontal) in the human newborn appears to be associated with maternal obesity state (weight¹³⁴). One recent report has further identified effects of maternal weight status on newborn offspring functional brain connectivity strength within sensory, reward, executive, and motor control networks²⁶⁸. Collectively, these results suggest altered connectivity patterns across multiple homeostasis-regulating networks in the obese relative to lean state. However, little is known about the developmental ontogeny or integrity of these specific networks in obesity prone newborns.

6.1.1.4. Longitudinal Imaging

In both humans and rodents, the mechanisms motivating the initiation and cessation of feeding transition from visceral/gustatory/interoceptive dominated in early life, to satiety/reward-dominated control in post-pubertal life¹¹⁶. However, these dominant networks act in concert with other cortical and subcortical networks that regulate feeding decisions, justifying a more systems level approach that includes and integrates feedback and feed-forward input in order to understand energy homeostasis relevant neurodevelopment. Yet, despite acknowledgement of the likelihood of its developmental importance²⁶⁹, little attention has been paid to the ontogeny of the system of networks as a whole, the maternal influences during pregnancy on this ontogeny, or the postnatal consequences of variation in this ontogeny for later obesity. MRI-based techniques present a unique and timely opportunity to address this current knowledge gap. Specifically, MRI provides a maximally non-invasive approach suitable for longitudinal observation of systems-based neurodevelopment across species. In neonatal/infant human research that likely requires natural sleep during scanning, this is largely limited to passive observation of anatomical morphology, and functional/structural connectivity within and between systems. However, recent evidence supports the plausibility of odorant delivery and functional response recorded using MRI in infants¹⁰⁰. In contrast to human research, stimuli in early rodent life can likely be extended to include nutritive gastric distension and hormonal inputs (e.g., leptin administration), though care should be taken in experimental design as most hormonal stimuli-of-interest are known effectors of development. In later stages of development, when human children are old enough for compliance, the opportunity for more complex stimuli (e.g., tasteants) and challenge protocols (e.g., stressors) would allow gaining a greater understanding of the gestational influence on specific systems including reward, satiety, visceral and salience networks.

6.2.1. Advances in the quantification of energy expenditure during the early postnatal life period/phase.

The key variables of interest in the quantification of energy expenditure during early postnatal life are focused on the constituents of total energy expenditure (TEE): basal metabolic rate (BMR), diet induced thermogenesis (DIT) and physical activity levels (PAL).

Doubly labeled water (DLW) is the most widely-used method for measuring infant TEE^270–273 owing to its minimal invasiveness, high accuracy, and repeatability²⁷⁴. Indirect calorimetry is a complimentary measure capable of further delineating the two components of TEE that compromise BMR: sleeping metabolic rate (SMR) and wakeful resting energy (WEE) ^275,276. DIT is also typically measured via indirect calorimetry, however, is measured using pre- and postprandial conditions²⁷⁷. DIT is then defined as the post-ingestion increase in BMR divided by the energy content of the food. Thus, there exists a challenge and opportunity with respect to refined measures of intake for normalization (see Section 6.2.2). With respect to PAL, actigraphy devices are a common method for determination in pediatric samples as they are relatively inexpensive, reliable, and convenient, making them highly accessible in a research context. Recent improvements in wearable devices for infants have also begun to provide insights into the role of early life PAL in typical growth trajectories^278,279. As a proxy, when included with TEE, SMR can also be used to objectively estimate PAL^280,281 in early life by assuming that variation in WEE is principally driven by variation in PAL. In ideal study conditions, frequent measures of energy intake quantity and quality (see Section 6.2.2) would be coupled with repeated measures of energy expenditure (e.g., TEE, DIT, PAL, and BMR), and longitudinal measures of body composition (see Section 6.3.1) in order to provide high temporal resolution measures of early life energy balance.

6.2.2. Advances in quantification of energy intake during the early postnatal life period/phase.

The key variables of interest in the quantification of energy intake during early postnatal life can be categorized as indirect (frequency of intake, nutritional biomarkers) and direct (breastmilk/formula composition and volume) measures.

In children and adults, self-reported measures of dietary intake are commonly used despite their susceptibility to day-to-day variation in diet (e.g., 24-hour recall) and recall error^282,283. Recent advances in smartphone applications have helped address this by allowing for high frequency intake assessments over long periods of time at minimal cost and subject burden^284,285 yet remain susceptible to rater bias and perception. Alternatively, nutritional biomarkers (e.g., plasma carotenoid²⁸⁶, lipid profiles²⁸⁷, urinary sucrose²⁸⁸, plasma alkylresorcinol²⁸⁹) hold the potential to reduce error from day-to-day diet variation and rater bias by providing objective measures of energy intake over varying time scales dependent on the tissue being sampled²⁹⁰ (e.g., blood, hair, adipose tissue). Further towards this aim, recent biosensor advances are enabling sweat-based continuous monitoring of metabolites and protein concentrations using wearable micropatch arrays^291–293. While the majority of developments in the electronic recording of feeding behaviors²⁹⁴ and nutritional biomarkers have thus far focused on older populations, emerging evidence suggests they provide similar advantages in studies focused on early life²⁹⁵ behaviors designed for the detailed characterization of energy intake.

Early life assessments of energy intake reflect a relatively stable and simple dietary intake (breast milk/formula). However, objective and accurate quantification of intake volume and composition during breastfeeding in an accessible and minimally invasive manner remains challenging²⁹⁶. In addition to the challenges of accurate volume measures, there exists variability in the composition of breast milk temporally across a single feed. For example, hind milk has higher lipid content relative to fore milk, and such a difference in composition appears to be of consequence to early life growth in very premature infants²⁹⁷. For this reason, assessment of breast milk composition (e.g., lipid content) across feeding duration would advance current understanding of the influence of intake on energy intake/imbalance in early life. Towards this, limiting study recruitment to maternal cohorts identified as suboptimal infant breastfeeding behavior²⁹⁸ (SIBB) would enrich for exclusively bottle fed infants, thus allowing for easy, frequent and accurate quantification of homogeneous intake by objectively and precisely weighing bottles before and after feeding. While this would result in a comprehensive, precise and well-controlled measure of intake, it would generalize poorly (e.g., to breastfed individuals) and limit understanding with respect to heterogeneity in breast milk/formula composition. Nipple shields²⁹⁹, while controversial with respect to its level of invasiveness, are instrumental in SIBB populations and could address such a shortcoming by leveraging negative pressure^300–302 to proportionally sample volume. Minimal fluidic mixing is a feature of microfluidic devices³⁰³, and if incorporated into a nipple shield design, could allow for the retention of temporal composition profiles, providing valuable information about the interaction between breast milk temporal composition dynamics, feeding behaviors and subsequent growth. Accurate and precise characterization of both breast and bottle feeding is particularly relevant to understanding early life feeding behaviors including cessation, duration and intensity of feeding. Because variation in these feeding behaviors^304,305 is likely driven by variation in energy homeostasis brain circuitry, they are of high relevance to the context of the current framework.

6.3.1. Advances in quantification of the biophysical, metabolic and clinical consequences (outcomes) of variation in newborn energy homeostasis brain circuitry

The key variable of interest in furthering our understanding of the clinical consequences of variation in newborn energy homeostasis brain circuitry is the magnitude and trajectory of adiposity and its regional distribution.

The consequence of positive imbalance in energy homeostasis above and beyond normal growth is excess tissue deposition. Mounting evidence suggests child size and growth velocity during early infancy represent the most reliable, valid, and strongest predictors of childhood obesity risk. Rapid size increase, and weight or fat gain during early life is associated with increased infant cardiovascular risk factors³⁰⁶, increased childhood (and adult) obesity risk³⁰⁷ and related outcomes^308–314 including type 1 diabetes³¹⁵, metabolic disorders³¹⁶, hypertension³¹⁷, and asthma³¹⁸ in later life. Recent evidence supports the notion that change in infant fat mass during early infancy is a much stronger predictor of childhood obesity risk than weight gain alone³⁰⁸. More specifically, the phase of rapid fat accrual in the first three months of life has been identified as a critical window for adiposity growth³¹⁹.

However, it should be noted that early life growth is highly non-linear, suggesting that two single-measure time points insufficiently characterize growth features including tempo, peak magnitude/velocity^320,321, and age at peak magnitude/velocity, particularly in relation to later health risks³²². It therefore seems imperative that studies using growth as an outcome ought to record frequent measures (e.g., skinfold, length, weight) and, when possible, anchor them to more informative measures of body-composition including MRI, DXA, or air displacement plethysmography. One emerging technology that promises to deliver frequent, precise and, most importantly, convenient quantification of body composition is 3d structured light imaging^323,324. Commercially available structured light scanners are now affordable and can be packaged with user-friendly software, enabling research participants to obtain in-home surface and volume measures of their growing child, independent of support staff. Coordinated measurements of volume and weight allow for body composition (e.g., body fat percentage) estimates in a manner consistent with those derived from air displacement plethysmography at a fraction of the cost and inconvenience. Serial measures utilizing structured light imaging would also reduce the effect of measurement bias (via repeated measures) while simultaneously enabling true non-linear modeling of early life growth patterns.

While frequent serial longitudinal measures will provide a greater understanding of growth, it is likely that proxy measures of total body fat will only provide a partial picture of metabolic health and disease risk. For example, it has recently been argued that it is the spatial and class distribution of lipids in fat deposits that determines metabolic risk^325,326. Importantly, emerging MRI-based techniques allow for the precise quantification of fat composition throughout the whole body under free breathing conditions^327,328. The importance of this development lies in the lack of compliance observed during the traditional breath-holding protocols in pediatric populations. The application of these emerging techniques would provide whole-body proton density fat fraction maps, thus allowing for the precise quantification of lipid distribution and density throughout the body. Important deposits of interest include the liver and classical Brown Adipose Tissue sites^329,330. In addition to such detailed measures, combining them with frequent longitudinal (serial) assessments would not only provide a more complete picture of body composition but could provide important anchor points in longitudinal curve fitting, as they would represent a more accurate and precise measurement of total body fat.

Collectively, these findings and considerations support the critical importance of longitudinal and detailed cross-sectional assessments of body composition during the period of rapid tissue expansion in early life and the validity of their use as outcomes in the context of research pertaining to fetal programming of energy homeostasis brain circuitry.

6.3.2. The role of postnatal conditions in assessing the consequences of altered newborn energy homeostasis brain circuitry

Prenatal states, conditions, and biological factors that influence newborn energy homeostasis brain networks can also influence brain networks postnatally. For example, as suggested above, maternal stress may influence the intrauterine development of limbic brain circuitry implicated in several aspects of food intake. Moreover, it is evident that sequelae of maternal stress, including depression³³¹, anxiety³³², and obesity³³³, are associated with reduced breastfeeding initiation and/or duration. This then creates a scenario in which susceptibility to stress exposure may be established in utero (altered brain development), only to then encounter adverse postnatal exposures that are themselves conditioned on the same stress exposure. The infant microbiome is an additional postnatal factor increasingly acknowledged to play a role in postnatal brain development^334,335. Further, infant microbiome composition and activity are believed to be susceptible to maternal conditions (e.g., inflammation) ³³⁶ that may also alter fetal brain development. Because most maternal conditions with the potential to alter fetal brain development during pregnancy are likely to persist through the postnatal period, they then may potentiate the adverse effects of prenatal exposures. Indeed, our model recognizes that the effects of prenatal and postnatal states and conditions may not be mutually exclusive, and that in many instances the effects of postnatal exposures may, in part, be conditioned upon the effects of prenatal exposures and that these effects may be additive or multiplicative in nature.

6.4. Advances in the use of animal models and in vitro approaches for studies of the determinants and consequences of variation in newborn energy homeostasis brain circuitry

Till this point we have focused on currently-available opportunities for human research on fetal programing of energy homeostasis brain circuitry. Now, we briefly provide a perspective on animal and in vitro research opportunities, with a specific focus on how they can complement and inform human research.

First, recent advances in optogenetics³³⁷ allows for high temporal and spatial resolution control of living neural tissue. Thus, this method provides an in vivo basis for experimentally validating the behavioral/physiological consequences of activating specific brain regions that have been identified in human cohorts as susceptible to developmental programming and prospectively relevant to early life energy imbalance. Further, combining optogenetic methods with fMRI^338–340 could facilitate assessment of the functional connectivity of precise neural circuitry, particularly in the context of variation in fetal exposures (e.g., stress/inflammation). For instance, one could vary maternal exposures like diet (e.g., saturated fats) during pregnancy and then characterize the inter-individual variation in offspring functional connectivity to the arcuate nucleus by optically exciting virally transfected cells during fMRI acquisition. The advantage of this approach combines features of optogenetics not available using conventional MRI techniques (precise neural control) while obtaining whole brain function on a scale translatable to that observed in human studies. Yet several challenges remain in establishing its utility in a developmental context when considering the time necessary to establish the virus necessary for light-based excitation relative to the pace of early life development, and its invasive nature relative to the plasticity of the developing brain. In contrast, manganese enhanced MRI (MEMRI) has recently emerged as a powerful but minimally invasive (administration of MnCl₂ through the tail) tool for understanding neural recruitment in rodents. MEMRI is considered non-toxic at viable doses, validated with histology and gene expression, has been used to map pain circuitry in newborn rodents³⁴¹, and most relevantly, has been instrumental in recent gains in the fundamental understanding of early life neurodevelopment at a systems level^342–344. Further, it is compatible with more conventional MR imaging (DTI and resting state functional connectivity) techniques allowing for its use as a complimentary measure^345,346. Finally, MEMRI has the demonstrated ability to map neural dynamics in the hypothalamus during administration of leptin. Because MEMRI and optogenetics provide precise neural control and inference observable at a macroscopic scale, these tools promise maximal translation from rodent to human research in a developmental context. In particular, they present a unique opportunity to gain insight into the early life transition from visceral-based control of feeding to reward-based feeding, as well as its gestational determinants and later consequences on the obese state.

Second, the placenta (and placental biology) represents the key interface between the maternal and fetal compartments. A number of questions remain regarding when, what, and how circulating maternal factors (e.g., lipids, cytokines) are able to cross the placental barrier. Recently, MRI techniques have been developed to non-invasively detect complement activation in the placenta using C-3 targeted ultrasmall paramagnetic iron oxide (USPIO) nanoparticles to bind within inflamed placenta and fetal cortex^347,348. USPIO is an MR contrast agent that locally increases field inhomogeneity and thereby decreases T2* relaxation time (signal void). Ultimately this technique could be leveraged to gain a greater understanding of the impact of maternal inflammation (e.g., resulting from maternal obesity) on the placenta and fetal brain. While this technique is potentially viable in humans (as Ferumoxytol is FDA-approved for human use), it is unlikely to see applications in pregnancy research in the near future and therefore should be considered for rodent models only.

Finally, tissue engineering methods represent an additional opportunity to gain a greater understanding of the role of the gestational milieu on neurodevelopment. Specifically, lab-on-a-chip and organoid models allow for well-controlled environments in which to perturb living tissues with metabolic, inflammatory or endocrine challenges. One recently micro-engineered placenta-on-a-chip³⁴⁹ provided an in vitro testbed for assessing the structural and functional complexities of the placental barrier under varying conditions. Importantly, this placenta-on-a-chip was able to recapitulate maternal-to-fetal glucose transfer properties found in ex-vivo perfused human placentas. Similarly, brain organoids are continually improving in both complexity and function with respect to their respective target tissues^350,351. Specifically, four recent advances have allowed for the development of highly complex functions in brain organoids including: 1) fine scale transcriptional mappings³⁵², 2) photo-responsive retinal networks³⁵³, 3) inter-neuronal migration³⁵⁴, and 4) enhancement of forebrain and cortical plate formation³⁵⁵. By perturbing the microenvironment with physiologically relevant levels of inflammatory and stress-related factors during development of lab-on-a-chip devices and/or brain organoids, one could gain a highly detailed understanding of the impact of the gestational milieu on tissue development at the cellular level.

6.5. The scientific and clinical relevance of characterizing variation in newborn energy homeostasis brain circuitry

The scientific relevance of characterizing variation in the newborn brain derives from the logic that energy homeostasis brain circuitry near birth represents a relatively (to children/adults) simple model of the neurobiology of energy homeostasis, and that brain circuitry at this time largely precedes influence by the postnatal obesogenic environment. The conceptual framework discussed above includes several brain networks that have downstream behavioral analogues to feeding in later life (e.g., reward/motivation, learned associations), however in infancy, there are also several external pressures and networks found in adulthood that are not yet relevant. For example, brain networks associated with conscious inhibition³⁵⁶, food-related monetary economic decision making³⁵⁷ (i.e., the balance between food costs and nutrition), and susceptibility to food advertising³⁵⁸ are not of direct relevance to infant feeding. In addition, given the extreme nutritional requirements during infancy (several times the caloric requirements, per kg, in adulthood) it is likely that feeding behaviors are especially salient during infancy, and are thus more readily measurable. Finally, because the state of the brain at birth reflects the cumulative sum of influences by in utero exposures during fetal brain development, yet precedes influence by the postnatal environment under which the brain further adapts, the inter-individual variation present at birth provides the most unconfounded characterization of in utero influences possible. Therefore, we assert that the time of birth is of scientific importance as it enables a more complete understanding of the temporality and sequence of effects in fetal programming than observations made in later life alone (e.g., early childhood).

The clinical importance (i.e., effect size) of inter-individual variation in human energy homeostasis brain networks at birth (i.e., initial setting) is largely unknown. While it is unlikely that any single conceptualization of the causal origins of obesity will be of major consequence to the global obesity crisis, obesity has been labeled a “heritable neurobehavioral disorder that is highly sensitive to environmental conditions” ³⁵⁹. This then suggests that pre- and postnatal environmental exposures play a critically important role in expressing inter-individual variation in the developing brain. Further, because fetal brain development happens at such a rapid pace culminating with the formation, segregation (differentiation/specialization) and integration (forming a network of networks) of brain networks by the time of birth, it is likely that the foundational setting of these brain networks plays a meaningful role in the ontogeny of downstream feeding behaviors in later life. However, because brain development in humans is especially protracted in order to further adapt to the postnatal environment, we assert that the proposed framework should largely be viewed in the context of susceptibility. That is, future efforts also focusing on the postnatal ontogeny (e.g., cascading effects, reversibility) of these brain networks under the influence of the postnatal obesogenic environment, as well as potential interventions and positive influences³⁶⁰, are critical in further contributing to the current understanding of the clinical importance of fetal programming of human energy homeostasis brain networks.

7. . Summary/Conclusion

Here, we have provided an overview of the concept of energy balance and the role of the brain in regulating energy balance. We have described evidence supporting the premise that energy homeostasis relevant brain circuitry is established and functional by the time of birth, with contextually-meaningful variation across individuals. We have advanced a novel conceptual framework placing fetal brain development on the putative pathway between maternal conditions during pregnancy and offspring susceptibility for obesity. Specifically, we describe a model by which classes of exposure in utero (sociodemographic, maternal biophysical, metabolic, endocrine, and inflammation) may influence the development of brain networks (satiety, reward, visceral, salience) that are ultimately for responsible for regulating downstream feeding behaviors. Further, we have identified questions, issues, considerations, and research directions relevant to our proposed model. By identifying determinants of childhood obesity that are already present at birth (i.e., predating the influence of postnatal conditions), this conceptual framework seeks to lay the groundwork for primary prevention strategies. In addition, through the identification of neural correlates of childhood obesity risk that predate the influence of the postnatal obesogenic environment and their fetal determinants, this framework aims to advance the current scientific understanding of the neurobiological underpinnings of obesity in humans.