Fat brains, greedy genes, and parent power: A biobehavioural risk model of child and adult obesity

Abstract

We live in a world replete with opportunities to overeat highly calorific, palatable foods – yet not everyone becomes obese. Why? We propose that individuals show differences in appetitive traits (e.g. food cue responsiveness, satiety sensitivity) that manifest early in life and predict their eating behaviours and weight trajectories. What determines these traits? Parental feeding restriction is associated with higher child adiposity, pressure to eat with lower adiposity, and both strategies with less healthy eating behaviours, while authoritative feeding styles coincide with more positive outcomes. But, on the whole, twin and family studies argue that nature has a greater influence than nurture on adiposity and eating behaviour, and behavioural investigations of genetic variants that are robustly associated with obesity (e.g. FTO) confirm that genes influence appetite. Meanwhile, a growing body of neuroimaging studies in adults, children and high risk populations suggests that structural and functional variation in brain networks associated with reward, emotion and control might also predict appetite and obesity, and show genetic influence. Together these different strands of evidence support a biobehavioural risk model of obesity development. Parental feeding recommendations should therefore acknowledge the powerful – but modifiable – contribution of genetic and neurological influences to children's eating behaviour.

Introduction

The developed world bulges with opportunities to consume palatable high-calorie foods in vast quantities – yet not all of us become obese. One explanation for this could be biologically and environmentally influenced individual differences in appetitive characteristics that emerge early in life (Carnell & Wardle, 2007; Llewellyn et al., 2011), and persist through development (Ashcroft et al., 2008; Farrow & Blissett, 2012; Parkinson et al., 2010). Since the beginning of the child obesity epidemic, a wealth of cross-sectional evidence has accumulated to suggest that adiposity is related in a stepwise fashion to higher food cue responsiveness and/or lower sensitivity to internal satiety cues in children and infants, as assessed by either behavioural tests (Hill et al., 2008; Llewellyn et al., 2008), or parent-report questionnaire (Carnell & Wardle, 2008; Spence et al., 2011; Webber et al., 2009). Now, a number of prospective studies are beginning to extend this research to provide persuasive support for the theory that these traits predict weight trajectories in childhood (Berkowitz et al., 2010; Webber et al., 2010b) and infancy (Faith & Hittner, 2010; van Jaarsveld et al., 2011).

A host of important questions remain, though. Where do these traits originate? Are they shaped by environmental factors? Or do they have a biological basis and if so, what is it? Most importantly, what are the practical implications of this information? Can the forces that shaped obesogenic traits be harnessed to bend an individual's dispositions back towards a healthier course? Our goal in this paper is to present evidence for an environmentally, genetically and neurologically informed biobehavioural susceptibility model that provides a useful framework for understanding the sources of individual variability in body weight in childhood and later in life. To do this we give a brief overview of relevant key findings from the diverse fields of parental feeding behaviour, obesity genetics, and neuroimaging of appetite and obesity, highlighting what they teach us about the origins of appetite. We finish by discussing some potential future research directions and the implications of our model for obesity prevention and treatment.

Methods

For this narrative review, we conducted three initial PubMed searches: (1) parent feeding AND child obesity OR weight OR appetite OR eating; (2) obesity genes; (3) obesity AND fMRI OR PET OR MRI. The parent feeding literature was then whittled down to contain only studies whose abstracts indicated content pertaining to relationships between the most commonly assessed feeding behaviours/styles (restriction, pressure to eat), and appetitive traits/behaviours known to predict weight (e.g. food cue responsiveness, satiety sensitivity, disinhibited eating, external eating, dietary intake, eating in the absence of hunger in the laboratory). Other feeding strategies (e.g. emotional feeding, instrumental feeding) and more general measures of the home environment (e.g. fruit and vegetable availability, family meals, TV watching) may also have long-term effects on appetite but are beyond our scope here. For the genetic literature we focused on twin and family studies generating heritability estimates for appetitive behaviours, and studies that directly related the genetic variants emerging from recent genome-wide association studies to such behaviours (e.g. associations between FTO, MC4R or combined variants and reported eating behaviours, dietary intake, eating in the absence of hunger). For the neuroimaging literature we focused on studies of appetitive traits (e.g. food cue responsiveness, emotional eating), child and adolescent obesity, and populations at high familial obesity risk. In order to represent the current state of each field, publications since 2007 are given the most attention; however, key findings prior to this date are described where context is needed.

Parent power

Ask someone why they weigh what they do and you can be sure that most of them will begin telling you about the way their parents (usually their mother) fed them during childhood. But is there really evidence that the explicit strategies parents commonly employ to control their children's intake (e.g. restriction, pressure to eat) have long-term impacts on children's enduring appetitive traits, and therefore their risk of developing obesity later in life?

Restriction

Parental restriction may be defined as parents’ deliberate limiting of the amount and type of food accessible to the child, typically their access to high-calorie ‘junk’ foods. The strategy is generally adopted to promote a healthy diet. However, some experimental studies have suggested that restricting access to a certain food may have the paradoxical effect of increasing preferences for and intake of that food when restrictions are lifted (Fisher & Birch, 1999b; Jansen et al., 2007), leading investigators to ask whether restriction could have long-lasting negative effects on children's eating behaviour and ultimately their body weight.

Cross-sectional studies of samples of children ranging from 3 to 12 years have shown that higher restriction, usually measured via the Child Feeding Questionnaire (CFQ) (Birch et al. (2001); e.g. ‘I have to be sure my child does not eat too many sweets’), is associated with: greater daily energy intake (Birch & Fisher, 2000), higher intake of snack foods following a satiating lunch (eating in the absence of hunger test (EAH)) (Fisher & Birch, 1999a), higher self-reported disinhibited eating (eating in response to external cues) (Carper et al., 2000), increased parent-reported food responsiveness (Webber et al., 2010a), and higher weight (Lewis & Worobey, 2011; Ventura & Birch, 2008). But associations with greater dietary restraint and lower levels of emotional and external eating (van Strien & Bazelier, 2007), better parent-reported self-regulation (Tan & Holub, 2011), and self-selection of lower energy-dense foods at a lab meal (among overweight children only) (Sud et al., 2010), have also been observed. ‘Softer’ forms of restriction also seem to show more positive associations – Brann & Skinner (2005) found that fathers with high scores on the ‘Monitoring’ scale of the CFQ, which measures how much parents ‘keep track’ of children's intake of unhealthy foods (Birch et al., 2001), had 8–10-year-old sons with lower body mass indexes (BMIs).

Longitudinal studies, which have the potential to reveal more about cause and effect, have also produced mixed results. Fisher & Birch (2002) found that restriction at 5 years predicted greater EAH at 7 years, Birch et al. (2003) showed that girls who were overweight and most restricted at 5 years had the highest EAH at 9 years and the greatest EAH increases from 5 years to 9 years, and Faith et al. (2004) found that greater restriction scores in 5 year olds with heavier mothers were associated with greater increases in BMI z-scores 2 years later, even controlling for child weight at 3 years. The latter studies are both consistent with an obesogenic gene–environment interaction. However, Campbell et al. (2010) found associations between restriction and less 3 year weight gain in 5–6 year olds, and the Faith et al. (2004) study also found that among low-risk children, parental ‘monitoring’ of children's intake of less healthy foods predicted reduced child BMI z-scores at 7 years, suggesting that for some kinds of parenting strategies and some families, restriction may be protective.

Other studies have also shed light on the cause–effect issue. For example, Webber et al. (2010c) found that maternal concern about weight significantly mediated the association between the use of restrictive feeding practices and adiposity in 7–9 year olds. Further, using a discordant sibling design, which partially controls for genetic influences, Roemmich et al. (2010) found that parents exerted higher levels of restriction for their overweight 8–17-year-old boys compared with their lean siblings, and Payne et al. (2011) found that although parents of 6–12 year olds generally used similar feeding practices with both siblings, differences in parental concern for a sibling's weight (rather than perceived weight or actual weight) predicted differences in the degree of restriction used. These results support a model in which some forms of restriction may promote healthy weight maintenance, but restriction is at least sometimes a response to maternal concern about overweight, rather than a cause of child weight gain per se.

Pressure to eat

Another common parental feeding strategy is ‘pressure to eat’, which normally describes parents’ attempts to encourage their child to eat food perceived as healthy, typically at meal times. One might think that this would promote eating in the short term, and this is supported by observational results (Drucker et al., 1999). However, some evidence suggests that persistent pressure can lead to less intake and negative food-related attitudes (Galloway et al., 2006), and others have suggested that systematic pressure to eat may ultimately override children's internal satiety signals, leaving them less able to regulate their own intake.

In support of these rather pessimistic hypotheses, higher pressure shows associations with parent reports of lower child enjoyment of food, and higher satiety sensitivity, slowness in eating and food fussiness (Gregory et al., 2010b; McPhie et al., 2011; Webber et al., 2010a), greater child dietary restraint, disinhibited eating, and emotional eating (Carper et al., 2000), greater food avoidance behaviours (Powell et al., 2011), greater neophobia (fear of new foods) and intake of unhealthy snack foods (Brown et al., 2008), and lower fruit and vegetable intake (Fisher et al., 2002; Wardle et al., 2005; Wyse et al., 2011). Johnson & Birch (1994) additionally found that high scores on a general scale of parental control over feeding (e.g. ‘My child should eat everything on his/her plate’) were associated with poorer compensation for preload calories at a subsequent lunch.

In contrast, Van Strien et al. (2009) found that higher pressure as perceived by children was associated with less overweight and greater breakfast- eating and physical activity, suggesting that perceived pressure could be a marker for other health- supporting behaviours. Additionally, Sleddens et al. (2010) found that a scale measuring a ‘softer’ form of encouragement (e.g. I encourage my child to enjoy his/her food) showed negative associations with unhealthy snacking. It should also be noted that the measure of pressure used in most of these studies was the pressure to eat scale from the CFQ, which might be thought of as representing an ‘authoritarian’ form of control. Both authoritarian and indulgent feeding styles as defined by Hughes et al. (2005) (largely based on items describing encouragement to eat) have been associated with higher child BMI (Hughes et al., 2008), whereas authoritative and authoritarian feeding have been associated with higher fruit and vegetable intake (Blissett, 2011; Patrick et al., 2005). These results suggest that, on balance, authoritative forms of encouragement may have the best outcomes.

The long-term effects of pressure to eat are unclear. However, it is striking that parent-report measures of pressure to eat are associated with lower (rather than higher) weight, in both cross-sectional (Ventura & Birch, 2008) and longitudinal (Farrow & Blissett, 2008) studies. Longitudinal studies by Gregory et al. have additionally found that pressure to eat predicts lower interest in food and greater food fussiness in children (Gregory et al., 2010a), and less fruit intake, even when controlling for initial intake (Gregory et al., 2011). These studies could reflect pressure causing low weight and unhealthy eating behaviour, but even the longitudinal results do not rule out the possibility that pressure simply tracks children's eating behaviour and weight. Gene–environment interactions should also be considered: Farrow & Blissett (2006) found that if maternal control (observed during a home meal) was low/moderate, then higher weight gain in the first 6 months was related to lower weight gain in the next 6 months, but if maternal control was high, the child maintained an accelerated weight trajectory, suggesting a lack of self-regulation. Together, the results suggest long-term effects may depend crucially on the type of pressure exerted, and the growth trajectory displayed by the child. In many cases relationships may reflect parents responding to unhealthy eating preferences and low weight in children.

Greedy genes

If parents are in some cases merely responding to their children's eating dispositions and body weight, then what other influences are at work? Many sources of evidence suggest substantial genetic influences.

Twin and family studies

The heritability of body weight in adults is well-established, with two reviews of twin studies estimating that 50–90% of variance in BMI is attributable to genes (Maes et al., 1997; Schousboe et al., 2003). Heritability is also high in children (Silventoinen et al., 2010; Wardle et al., 2008), and longitudinal genetic analyses have shown that although, new genetic factors become influential from early childhood to adulthood, there is also significant genetic continuity, suggesting that the genes that influence weight early in life are the same ones that influence weight in early adulthood (Silventoinen & Kaprio, 2009).

Although people often assume that quantitative genetic effects on weight act via metabolism, there is a good deal of evidence that appetitive traits are heritable, suggesting some of this genetic effect on weight could also be mediated behaviourally. For example, although estimates vary widely, twin and family studies of adults have indicated significant heritability for scores on cognitive restraint, disinhibition and hunger (de Castro & Lilenfeld, 2005; Neale et al., 2003; Provencher et al., 2005; Steinle et al., 2002; Tholin et al., 2005). Parent-reported satiety responsiveness and food cue responsiveness have demonstrated 63% and 75% heritability respectively, in a large sample of 8–11 year olds (Carnell et al., 2008), and similar data from a large study of twin infants have suggested heritability of 84% for slowness in eating, 72% for satiety responsiveness, 59% for food responsiveness, and 53% for enjoyment of food (Llewellyn et al., 2010). Data from this cohort have also demonstrated significant shared genetic influence (22–37%) for infant appetite and infant weight, which is able to explain 41–45% of phenotypic associations, and supports the genetic mediation model outlined above (Llewellyn et al., 2012).

There is also evidence for heritability of eating behaviours as assessed in the lab or in the field. Eating rate measured during a meal eaten at home was found to be 62% heritable in 10–12-year-old twins (Llewellyn et al., 2008); a family study by Fisher et al. (2007) reported 52% heritability for energy intake during an ad libitum dinner (52%) and 51% for post-dinner EAH; and Faith et al. (2004) observed significant sibling aggregation for total energy intake and percentages of fat, carbohydrate and protein intakes at a lab meal. Interestingly, twin designs may also help to reveal dietary differences between children which are independent of genetic similarity: Rissanen et al. (2002) studied weight-discordant monozygotic twin pairs and found a three-fold preference among the obese twin for fatty foods. These studies tell us nothing about the actual genes involved, but they argue for strong genetic influence on variation in weight and appetite in children and adults, and are consistent with the action of multiple genetic variants of small effect combining to influence the umbrella phenotype of body weight.

The search for ‘obesity genes’

Several mutations in single genes, including LEP, LEPR, MC4R and POMC, are known to give rise to extreme, early-onset obesity (Farooqi & O'Rahilly, 2008). But these are estimated to account for a maximum of 5–10% of cases of obesity (Choquet & Meyre, 2011), leaving much variation unexplained. Some of this remaining variance might be explained by common variants in the genome, and the sequencing of the human genome and creation of databases of single nucleotide polymorphisms (SNPs), together with technological advances making genotyping cheaper and faster, have recently paved the way for genome-wide association studies (GWAS) to be conducted in large populations, giving power to detect very small effects (Day & Loos, 2011).

The first gene to be associated with BMI via GWAS was FTO (Frayling et al., 2007; Scuteri et al., 2007). The high-risk allele is present in 42% of white Europeans, and each additional risk allele is accompanied by a 0.39 increase in BMI (Speliotes et al., 2010). Frequency of the high-risk A allele varies with ethnicity, and there is some variation in effect size, but the association with BMI appears relatively consistent across backgrounds (Day & Loos, 2011). The latest and largest GWAS (n ≈ 250,000) confirmed 32 BMI-associated loci, including FTO and several near MC4R (Speliotes et al., 2010). However, even when combined these SNPs explained only 1.45% of population variation in BMI, significantly limiting their ability to predict obesity (Speliotes et al., 2010). Notably, many of the BMI-associated loci identified in adults are similarly associated with BMI in children. For example, in a meta-analysis of n ≈ 13,000 children and adolescents den Hoed et al. (2010) found that 13 of the variants observed in adults showed similar BMI associations in children (although only nine reached significance), and whereas the effect size of certain variants differed by age, the cumulative effect size was similar across ages.

Consistent with diet-influenced expression of FTO in the hypothalamus, a structure known for its role in energy intake (Gerken et al., 2007; Stratigopoulos et al., 2008; Tung et al., 2010), studies have revealed associations between the high-risk allele and higher fat and energy intake in both adults (Speakman et al., 2008) and children (Timpson et al., 2008). Wardle et al. (2009) found that 4–5 year olds homozygous for the low-risk T allele ate significantly less than children heterozygous or homozygous for the high-risk A allele and demonstrated that the high-risk allele was associated with lower levels of parent-reported satiety sensitivity (Wardle et al., 2008), while Tanofsky-Kraff et al. (2009) reported that children with 1 or 2 (versus 0) high-risk alleles were more likely to experience loss of control and have greater intake of fat during an eating episode. Variants near MC4R (Stutzmann et al., 2009) as well as other common variants (SH2B1, KCTD15, MTCH2, NEGR1, BDNF) (Bauer et al., 2009), may also be associated with increased intake of high-energy foods.

Of course, genes do not cause obesity in isolation and increasing evidence suggests FTO and other common variants may interact with environmental factors to determine body weight. For example, physical activity has been shown to attenuate the association between FTO and obesity in adults (Kilpelainen et al., 2011) and adolescents (Ruiz et al., 2010). Low-fat diets (Sonestedt et al., 2009) and embarking on a Mediterranean diet (Razquin et al., 2010) may also ‘protect against’ FTO-related obesity, and some have suggested that education (Corella et al., 2010) and breast-feeding may also be protective (Dedoussis et al., 2011). There is also evidence that high genetic risk in terms of common variants could mitigate against weight loss via bariatric surgery (Still et al., 2011), caloric restriction (Matsuo et al., 2011), or within a diabetes prevention programme (Delahanty et al., 2012).

In addition to the common variants identified via GWAS, growing evidence suggests that genes influencing dopamine and serotonin function may play a role in obesity and appetite. For example, Spitz et al. (2000) and others have reported greater obesity rates in those with the Taq1A A1 allele, which is linked with decreased dopamine receptor density, and Epstein et al. (2007) have observed that food reinforcement, (i.e. the willingness to work for food) was particularly high in individuals with that allele. Polymorphisms in the serotonin gene SLC6A4, and the monoamine-regulating gene MAOA, which produces an enzyme that metabolizes dopamine, serotonin and noradenaline, may be associated with BMI, and SLC6A4 may be additionally associated with changes in BMI from adolescence to young adulthood (Fuemmeler et al., 2008). The next wave of genetic studies is also likely to unearth lower frequency obesity-associated variants with larger effects (e.g. copy number variants), as well as evidence for specific epigenetic changes which might influence weight (Day & Loos, 2011).

Fat brains

As we outlined above, there is substantial evidence for genetic influence on appetitive characteristics, and many of the common obesity-associated genetic variants are highly expressed in the human brain. But what about the missing link between the brain and appetitive traits? Is there evidence that individual differences in appetite and body weight are associated with differential neurobiology? Might certain functional or structural characteristics of the brain predispose someone to eat more and gain weight? And, if so, is there evidence that the obesity and appetite genes identified thus far are related to these characteristics?

Neuroimaging studies of obesity and appetitive traits

There are now a considerable number of neuroimaging studies comparing obese and lean adults. Many of them have examined responses to food cues, which might be considered an implicit operationalization of the appetitive trait ‘food cue responsiveness’, and have revealed abnormal patterns of activation and connectivity among the obese in areas associated with reward, somatosensory processing, emotion, memory and cognitive control. Others have tested responses to food ingestion, revealing evidence for blunted responses in the striatum, hypothalamus and dorsolateral prefrontal cortex (dlPFC), which are suggestive of impaired ingestion reward and satiety/inhibition mechanisms (see Carnell et al. (2012) for a review).

In addition, a small but growing number of studies relate established appetitive phenotypes assessed by questionnaire measures to neuroimaging data. For example, in a study of lean adults, ‘external food cue sensitivity’ scores were found to be associated with greater functional connectivity between the ventral striatum and emotion/motor preparation structures (amygdala, premotor cortex), and lesser connectivity between the ventral striatum and amygdala and attention-related regions (anterior cingulate cortex (ACC)), in response to appetizing versus bland pictures (Passamonti et al., 2009). Across obese and lean subjects, scores on disinhibition (i.e. the tendency for restraint to break down when confronted by emotional or external cues to eat) has been associated with lesser pre-meal ACC responses to visual food versus non-food cues (Martin et al., 2010), and scores on food addiction (i.e. a pattern of compulsive eating typified by tolerance, withdrawal and loss of control) have been associated with greater medial orbitofrontal cortex (OFC), amygdala and ACC responses to anticipated receipt of a milkshake, and lesser activation in the lateral OFC in response to actual receipt (Gearhardt et al., 2011).

Other studies have examined phenotypes that may involve eating in response to negative emotion. For example, Volkow et al. (2003) found that higher emotional eating scores were associated with greater dopaminergic striatal responses to the taste and smell of food among lean adults, and in a study of lean and overweight adolescent girls Bohon et al. (2009) found that those high in emotional eating showed greater parahippocampal gyrus and ACC activation in response to a cue signalling receipt of a milkshake – as well as greater activation in the ventral pallidum, ACC and thalamus when actually tasting the milkshake. Meanwhile, studies of binge eating have suggested that binge eaters show heightened responses to high-palatability food cues in the frontal premotor area (Geliebter et al., 2006), and medial OFC (Schienle et al., 2009), increased responses to the sight and smell of real cooked food in the frontal and prefrontal cortex (Karhunen et al., 2000), and caudate/putamen (increased dopamine activity) (Wang et al., 2011), and greater grey matter volume in the ACC and medial OFC (Schafer et al., 2010).

Together these studies suggest that appetitive endophenotypes may be associated with subtly distinct neural traces, although since each study examined associations with a single trait it is not possible to tell whether these traces are independent. Studying them separately, however, may help refine understanding of the neural basis of appetite and possibly increase our power to detect gene–brain associations by allowing us to map specific genes to specific endophenotypes relating to their functional pathways.

Neuroimaging studies of children and high familial or genetic obesity risk

The studies above suggest that the obese state and appetitive traits have measurable neural correlates. However, they do not tell us whether these neural traits are capable of predicting weight gain. Indeed, since the majority of studies described were conducted in adults, any obesogenic neural dispositions may already have been translated into weight differences – or may have been prevented from doing so via the exertion of cognitive control – making predictive relationships harder to see. Studying children may shed light on causal mechanisms since observed abnormalities are less likely to be the result of long-term obesity or exposure to high-energy foods.

With the exception of the study by Bohon et al. (2009), no neuroimaging studies have assessed appetitive traits in children per se. However, a number of studies have compared fMRI responses to food cues in obese versus lean children, and these seem to largely replicate findings in adults. For example, Bruce et al. (2010) found greater pre-meal (after 4 h fasting) responses to food pictures in the PFC, greater post-meal (after 500 kcal standardized mixed meal) activation in the OFC, and relatively smaller post-meal (versus pre-meal) decreases in nucleus accumbens, limbic, and prefrontal activation among obese versus lean 10–17 year olds. In adolescent girls, Stice et al. (2010) reported positive associations between BMI and post-fast responses to appetizing food pictures in the putamen, OFC and frontal operculum. Most recently, Davids et al. (2010) described increased dLPFC responses, but lesser caudate and hippocampus responses to food pictures in overweight/obese versus lean 9–16 year olds. Stice et al. (2008b) have also examined responses to conditioned cues signifying imminent milkshake delivery, and have shown greater cue-associated activation in obese versus lean adolescent girls in the anterior and middle insula and somatosensory cortex. Obese versus lean adolescent girls additionally show relatively decreased caudate responses to actual delivery of 0.5 mL tastes of milkshake (Stice et al., 2008a). There may also be structural differences between obese and lean adolescents: in a large study of obese and lean 14–21 year olds, Maayan et al. (2011) reported lower OFC volume in the obese, and found that lower OFC volume was associated with higher dietary disinhibition scores – although only among the lean group.

While studies of children are informative, studying populations at high risk for obesity based on either family weight history (i.e. familial (genetic and environmental) risk) or possession of obesity-associated genes (i.e. genetic risk) presents an even better opportunity to assess ‘raw’ neural patterns before they have been behaviourally translated into an obese phenotype. Taking this approach, Stice et al. (2011) found that adolescents with two obese (versus two lean) parents showed greater caudate, frontal operculum and parietal operculum responses to milkshake tastes. No studies have yet linked commonly occurring obesity- and appetite-associated genetic variants to neural food cue responses. However, high-risk variants on FTO have been associated with volume reductions in the frontal and occipital lobes (Ho et al., 2010). There is also some evidence to suggest that genetic markers of dopamine function (e.g. Taq1A A1 allele) may exaggerate relationships between blunted neural responses to tastes of a milkshake and obesity (Stice et al., 2008a), and interact with patterns of brain responsivity to food cues to predict 1-year weight increases in adolescence (Stice et al., 2010). Together these studies suggest that certain functional and structural characteristics of the brain (e.g. abnormal reward area responses to food cues, blunted striatal responses to food ingestion, reduced volume in areas associated with reward and inhibition) may be influenced by familial and genetic factors, and may be associated with the risk of developing obesity in the future.

Discussion

We have described an array of interesting studies from the worlds of psychology, epidemiology, behavioural genetics, molecular genetics and neuroscience. But how can we make sense of these multi-disciplinary findings and what do they mean for an individual during his or her lifespan? We believe that individuals come into the world genetically and epigenetically loaded with powerful biologically influenced predispositions towards food (e.g. reduced satiety and heightened food cue responsiveness as a result of possessing high risk alleles for FTO, or Taq1A). These are captured in our biology and particularly in our brains. Environmental factors can certainly affect these dispositions (e.g. authoritative forms of parental restriction and encouragement may promote moderate appetites and eating behaviour in children), but often these influences are overwhelmed by the expression of innate tendencies, and environmental factors correlate with rather than cause appetitive behaviour (e.g. evidence for restriction and pressure to eat as responses to children's eating behaviour and weight). This model is roughly illustrated in Fig. 1.

Figure 1. Biobehavioural susceptibility model of child and adult obesity. In this putative model, individuals develop enduring, measurable appetitive traits early in life. These traits are influenced by genetic factors as well as factors in the early environment (e.g. parental feeding behaviour), and may be associated with differential patterns of neural activation in response to food and food cues. Together, these traits impact how each person responds to obesogenic influences in the modern-day food environment, which in turn influences his or her energy intake and adiposity level.

The model has empirical gaps. For example, very little is understood about how even the known obesity-associated genetic variants affect the brain to influence appetitive traits. Although many studies have looked for parental and environmental influences on appetitive traits, we do not thoroughly understand how these are represented at the neural level. We also lack large-scale prospective studies with the right measures and sufficient power to understand how genetic and environmental factors interact at a behavioural and neurological level to influence eating and predict weight trajectories. That said, new genetic and imaging technologies promise to rapidly accelerate progress over the coming decades, and continual improvements in our understanding of and ability to capture the relevant behavioural factors (e.g. interactions between parent feeding behaviours and children's eating behaviours) will allow us to make the most of these advances when they arrive. For example, it is becoming increasingly feasible to genotype large numbers of individuals and collect validated psychometric information about their eating traits. In parallel, neuroimaging methods are increasing in their spatial and temporal resolution and may one day be able to provide more information regarding pathways of information transfer, and therefore the precise causal mechanisms behind the influence of genes and environment on neural appetite networks. It should be noted that, as well as revealing potential neural precursors of overeating and obesity, brain imaging may help to establish biomarkers for appetitive responses which are independent of the social desirability effects that plague other assessments of appetite such as self-report questionnaires, food records, and public eating behaviour.

Future insights and those already gained may aid the development of targeted behavioural and biological interventions to increase healthy appetitive behaviours. For example, we may be able to behaviourally retrain children's appetitive traits (Johnson, 2000; Ford et al., 2009), and neuroimaging could provide a nuanced and unbiased way of evaluating resulting alterations in the responsiveness of reward networks. The personal desire to change behaviour might be awakened by individualized genetic feedback (Meisel et al., 2011), and personalized neurological feedback could have the same effect. Indeed, evidence that individuals can be trained to engage the dlPFC (Delgado et al., 2008) – a key control area associated with successful dieting in a number of studies in adults (see Carnell et al., 2012) – provides hope that fMRI-aided interventions, perhaps even incorporating real-time fMRI feedback (Hinds et al., 2011), could help individuals develop self-control over their reward responses to food. The identification of neural networks involved in specific obesogenic feeding phenotypes could also aid the development of novel pharmacological interventions to weight loss – although the enmeshment of appetite networks with motivation circuits involved in many essential human behaviours makes this goal a challenging one. The key message to convey to adults and children is that we all have different biologically determined predispositions towards food, and – even if these are largely immutable – we still have the power to exercise conscious control over the eating environments which elicit these predispositions. Some of us, though, will need to wield that control more than others.

Declaration of interest: Support was provided in part by US National Institutes of Health grant K99 DK 088360 (to S.C.). All authors participated sufficiently in this work to take public responsibility for the content. The authors alone are responsible for the content and writing of the paper.