Brain imaging research in to disordered eating behaviour
This section summarizes conclusions drawn from research involving functional brain imaging methods, that provide some evidence to support a neural model of eating disorders across a spectrum of restricted versus impulsive eating behavior. It is not meant to be an exhaustive meta-analytical review, which has been done by expert researchers in the field previously (see below) but a summary of the main views arising from contemporary brain imaging findings, in an attempt to spark a debate about a model of disordered eating behaviour.
As has been eloquently reviewed previously [21
], functional brain imaging studies of females with AN (in comparison to healthy, age and gender matched controls), particularly when viewing rewarding stimuli, largely report aberrant, often reduced activation in ‘bottom-up’ mesolimbic regions associated with somatic states (e.g. appetite) such as in the striatum, hippocampus, amygdala, hypothalamus and cerebellum. This is often in conjunction with increased activation in ‘top-down’ prefrontal cortical (PFC) regions linked to cognitive evaluation, attention and executive functioning (e.g. working memory, goal-orientation, self-reference, evaluation of salience) such as the dorsolateral prefrontal cortex (DLPFC), medial prefrontal cortex (mPFC), orbitofrontal cortex (OFC), and anterior cingulate cortex (ACC). Furthermore, there appears to be a strong indication from various neuroimaging modalities (e.g. Positron Emission Tomography, Single Photon Emission Tomography) that the subtypes of AN (e.g. restricting and binge purging) can be distinguished by relatively higher or lower activations to appetitive stimuli in these regions [25
]. Indeed other fMRI studies show that the DLPFC, which appears to be highly activated when thinking about eating food shown in images in people with AN [26
], is strongly implicated in both appetite suppression and working memory function [28
]. Also, a recent neuropsychological study showed that automatic appetite activation (using subliminal images of food) interfered with a DLPFC-dependent working memory task [31
]. Thus, the excessive ruminations about weight, shape and eating in those who pathologically restrain their appetite is reminiscent of classic, yet excessive, working memory function. In turn, the ability to restrain appetite in those with anorexia might be associated with excessive top-down activation (mainly DLPFC) to varying degrees (e.g. different levels of restraint between the subtypes of AN), combined with varying levels of desire to restrain an over or underactive appetitive system. Furthermore, recent evidence suggests that the neural response during working memory performance is modulated by genetic polymorphisms in COMT and BDNF described below [32
]. However, there is one caveat to linking hyperactivation in the DLPFC to excessive working memory function in AN: it is currently unclear whether increased DLPFC working memory-related activation reflects a deficiency in working memory and a need to work harder.
Conversely, in those who exhibit binge eating behavior (e.g. BN, BED), previous research summarises a neural pattern to rewarding stimuli that is somewhat reversed, with a reduced or sporadic response in PFC regions, combined with hyperactivation in mesolimbic areas. In further support, a recent fMRI study that directly compared neural activation to food images in women with AN and BN found a clear PFC vs. striatum response to images of food [33
]. However, it is also plausible that binge eating might coincide with hypofunctioning of the reward system [34
], prompting the hypothesis that during binge eating, like an addict, a person must consume greater quantities of food to achieve the feeling of satiation. Furthermore, there appears to be a differential pattern of activation in binge eaters, when consuming versus anticipating food consumption during fMRI (e.g. [35
]), suggesting that the model proposed here might be more applicable to anticipation of food intake, and cognitive biases towards food stimuli. It is also often observed that people who binge (e.g. those with BN) have a strong desire to restrain their intake, and recent evidence suggests that restraining one’s appetite increases the chances of binge eating in those with a genetic risk for binge eating [36
]. Moreover, recent evidence suggests that artificial stimulation of the DLPFC using repetitive Transcranial Magnetic Stimuluation (rTMS) reduces craving in those prone to binge-eat [37
]. Thus, the neurobiological interactions underpinning restrained and binge eating behaviour are complex. It is proposed here that, in response to appetitive food stimuli that may underlie responses to food, a specific neural signature, with sporadic dominance of restraint versus binge eating can be mapped onto a single model with the support of recent neuroimaging findings. See Figure .
Impulse-control spectrum model of eating disorders.
A commonality between all ED subtypes is seemingly dysregulated neural activation in response to food stimuli in the parietal cortex, and the insular cortex in the temporal lobe, regions associated with somatosensory perception and interoceptive awareness [25
]. One might argue that somatosensory perception occurs via a balance between basic bottom-up arousal systems, and higher order top-down cognitive evaluations. Against this background, it is proposed that an excessive activation at one or other extremes (e.g. restraint versus binge eating systems) in those with eating disorders causes an imbalanced convergence on these somatosensory brain regions, associated with dysfunctional processing of the body state (inexorably linked to body image distortions). See Figure . Additionally, hyperactivation of the amygdala is often seen across ED diagnoses [39
] and likely leads to anxiety experienced by most ED sufferers regardless of subtype. Emotional experiences that derive from somatosensory states can be positively or negatively arousing (e.g. pleasant feelings during appetite satisfaction versus unpleasant anxiety and anger) and are linked to areas of the ‘temporoparietal junction’ that are involved in creating a sense of self-control over one’s actions [41
]. It has long been established that people with ED have problems identifying and expressing their emotions [42
], a core symptom known as alexithymia
, which has been further demonstrated in recent neuropsychological studies [43
]. An imbalanced regulatory neural circuit between the PFC and mesolimbic regions likely hampers a healthy orchestration of these systems, leading to a ‘rate-limiting’ defect in the insula [45
] and other somatosensory regions that normally provide a solid sense of ‘emotional self’ (See Figure ).
Imbalanced PFC-mesolimbic convergence on the insular cortex in those with eating disorders.
Findings from fMRI research into obesity, which could be regarded as representing the excessive appetitive extreme of the proposed spectrum model of eating disorders, fit well with the notions proposed here. One might predict, against the background of the present taxonomy, that people who are obese have deficits in prefrontal cortical activation, particularly the DLPFC, combined with hyperactivation of brain reward systems. Such a neural substrate does appear to exist in obesity, in that a hypersensitive reward system is demonstrated, in response to various types of food stimuli [46
], but particularly when overweight people think about the taste of food [47
]. Hyperactivation of reward systems when thinking of and/or viewing food stimuli may well impinge on cognitive systems, particularly memory and cognitive control, in those who are obese [46
]. However, causality is difficult to determine, and it could also be that deficits in cognitive processing prevent an adequate top-down control of one’s impulsive responses, e.g. over-eating in those who are obese. Thus, fMRI studies of obesity do seem to support the notion that specific eating disorder symptoms, ranging from restraint to impulsivity, can be mapped on to a single spectrum model, as proposed here.
Eating disorder cases that do not easily fit in to the proposed model must also be considered in this debate. For example, prima facie
it is not easy to fit binge-purge anorexia nervosa to the model, because such cases often consume excessive amounts of food in one sitting, while subsequently adopting compensatory measures (e.g. vomiting, laxative abuse) to avoid weight gain, and so might not easily fit the restrictive extreme. Furthermore, profile analyses of individual diagnoses of eating disorders often exhibit an inhibited and non-inhibited style based on personality traits [48
], suggesting that disordered eating behaviour is based not on appetitive and restricting temperaments, but more on personality dimensions. However, we argue that the model can also explain these cases, because we do not suggest that a person is exclusively restrictive
. Rather, we purport that, depending on one’s response to fluctuating social/environmental challenges, at any given time a person can present as restrictive or impulsive. Against this background, we further suggest that it is temperamental dominance
that dictates which behaviour is more frequently observed (restrictive vs. impulsive), for example, a person with AN who is generally restrictive, but with bouts of binging and purging. Temperamental dominance might also underlie personality characteristics, e.g. a person with a restrained eating disorder might also be introverted and mildly psychotic, a person with an impulsive eating disorder might present features of extroversion and neuroticism. Temperamental dominance, we purport, is determined by inherent neurobiological factors that are acquired genetically, but modified epigenetically via learned interactions with new environmental challenges (e.g. during puberty), particularly in those who develop eating disorders [50
It is of note that despite its apparent appeal, body weight may not be the most accurate measure to indicate one’s position on the spectrum, because for example, some people with bulimia are able to excessively exercise to compensate for high levels of food consumption - a factor that makes bulimia nervosa difficult to detect and treat. Furthermore, cognitive and personality traits often precede starvation and are sometimes present following weight restoration. It could be that questionnaire measures of temperament (e.g. Restraint Scale [51
], Eysenck Impulsivity Questionnaire [52
]) will prove more appropriate for measuring susceptibility or current diagnosis of disordered eating behaviour at a given timepoint. Thus, in the proposed spectrum model, a fluctuating temperament linked to eating could present (e.g. restraint, binge-purge, extroversion-introversion), but it is the temperamental dominance, persisting over time (e.g. excessive restriction, excessive food consumption, regardless of compensatory measures) that determines ones true place on the model.
Another major caveat to the proposed model are the recent findings that self-starvation may evoke reward responses in the brain [53
]. In line with animal models of AN, this suggests that the stress induced by self-starvation becomes a rewarding and/or motivating factor. However, despite fMRI studies showing increased striatal responses in females with AN, it is to images of thin bodies (which are probably rewarding for a person with AN to look at, relative to larger bodies). Previous fMRI studies using images of bodies (e.g. [55
] show different neural responses to food stimuli, and appetitive activation is the main focus of the model proposed here. Furthermore, it could be that the striatal response to self-starvation does not reflect reward, but rather an increased motivational state (which could explain a greater propensity to exercise in AN, despite energy deficits), akin to the evolutionary homeostatic drive to seek food [58
]. Given that the motivation neural circuity, although shown to be separate, is closely tied to reward responses in the mesolimbic pathway (e.g. [59
]), it might be difficult to separate food seeking behaviour from pleasurable responses to food in those with AN. The problem of the model therein lies: how to differentiate between striatal responses associated with seeking, and striatal responses associated with impulsive and excessive binge eating. In an attempt to account for this discrepancy, we again emphasise temperamental dominance: a person with AN who has excessive striatal-induced seeking behaviour, provoked by prolonged restraint of appetite, would still have (over time) a dominant PFC-related restraint phenotype. We do not propose that one cannot have striatal (or other reward-related responses) in conjunction with top-down restraint, rather, we suggest that it is one of the extremes of the model that will ‘win out’ to highlight the person’s temperamental dominance during a given time period. Finally, according to the model we propose, it is not only striatal responses that predict impulsive eating behaviour, but also responses in the amygdala, cerebellum and hypothalamus.
In summary, brain imaging studies support a taxonomy for a spectrum model of ED, whereby increased PFC responses to food stimuli correspond to reduced consumption of food, cognitive restraint and rigidity, obsessive ruminations about food, perfectionism and high attention to detail - symptoms that are typically observed in people with AN. Conversely, increased mesolimbic reward responses to food stimuli may correspond to greater consumption of food, impulsivity and high arousal, lack of control and aversive social situations (e.g. problems in relating positively to others), promiscuity and alcoholism – symptoms that are sometimes observed in people with BN and also those with obesity. This pattern of neural activation can easily be mapped on to the previously presented spectrum model (see Figure ). In both extremes, an imbalanced convergence on the insula may lead to a rate-limiting defect [45
] that heightens the experience of anxiety and fear, particularly for the way the body feels
(a phobia that could perhaps be termed, pathodysmorphia
). Finally, it must be noted that we suggest that it is the temperamental dominance
of a particular extreme (e.g. restriction versus impulsivity) that produces phenotypic behavior, and that reduction in one dominant extreme could present a new dominance at the opposing extreme, if the secondary dominance is sufficiently activated (e.g. a new presentation of impulsivity that was previously suppressed by excessive restriction).
Potential genetic markers underlying disordered eating behaviour
In the next section, we discuss contemporary genetic findings that may help to explain how these patterns of brain activation occur in people with ED. It must be noted that the proposed candidate genes might in part contribute to the etiology of ED, but their effects are quite small and exploratory genetic studies of ED in their infancy. Furthermore, it is most likely that complex gene-environment interactions and epigenetic effects play a greater part in the pathophysiology of ED [50
]. Nevertheless, we introduce some exciting candidate genes to provoke additional debate about the role of genes in the proposed impulse control model of ED.
One of the most interesting findings in genetic studies of ED comes from a genome wide linkage test by Grice and colleagues in 2002 [60
]. By screening 386 genomic markers in 192 families of ED affected patients, they were able to identify an association of a region on chromosome 1 linked to AN. This region contains, among others, the genes encoding the delta opioid receptor 1 (OPRD1) and serotonin receptor 1D (HTR1D). Single nucleotide polymorphisms (SNPs) in these genes were confirmed to be associated to AN in subsequent association studies [61
]. Further, the candidate gene approach (i.e. an a priori selection of genetic variations related to genes proposed to be involved in EDs, based on functional specification of the gene) has produced some potential linkage to ED pathology [63
]. Studies on the Val66Met polymorphism in BDNF, which is largely expressed in the mesolimbic reward region of the brain, are the most robust results in association studies on EDs to date. Through collaboration of several eating disorder clinics across Europe, large ED patient cohorts of about 1,000 patients were assembled for association studies and subsequent replication studies [64
]. The 66Met variant leads to an amino acid exchange in the functional BDNF protein, from valine to methionine at position 66, which is denoted Val66Met. This variant is one of few for which the functional implications are known: 66Met-BDNF localizes poorly to intracellular granula compared to 66Val-BDNF. This negatively affects secretion of BDNF. Despite the large cohorts used in these studies, replication studies from other groups are somewhat more inconsistent as to this variant’s association to EDs [68
] (Reviewed in 48). Nonetheless, the BDNF polymorphism may be linked to patterns of mesolimbic activation observed in neuroimaging studies, but the link is complex and undoubtedly incorporates many other intra- and inter-cellular, as well as environmental interactions.
The Val158Met polymorphism in catechol-o-methyl transferase (COMT), which was originally observed to be associated with AN in a Transmission Disequilibrium Test (TDT) performed on 51 Israeli family trios [74
], and later observed to also be associated to BN [75
], is of particular interest as it leads to a thermodynamically more unstable protein, which in turn leads to a lower rate of dopamine (DA) turnover in the synaptic cleft, particularly in the PFC. Subsequently, a higher DA activity in COMT-regulated neurons could be a predisposing risk factor for developing EDs. Alterations in COMT activity due to the Val158Met polymorphism is likely to have implications for DA signaling in the PFC, as clearance of DA from the synaptic cleft has been shown to be regulated via COMT in the PFC [76
]. However, a recent meta-analysis of the COMT polymorphism alone and its link to AN has not shown significant results [77
]. Additionally, despite set-shifting being potentially the most robust endophenotype for ED pathology [78
] the COMT polymorphism alone does not relate to set-shifting impairments observed in AN and BN [79
]. Both these findings suggest, as we imply in our model, that it is an interaction between genotypes (e.g. COMT vs. BDNF) that might lead to a specific ED cognitive and appetitive phenotypes (See Figure ).
Neurobiological impulse-control model of temperamental dominance in ED.
Variants related to agouti related protein (AGRP) have been associated with EDs, which is a central mediator of appetite via feeding regulating hypothalamic neuronal circuits; and SK3, a protein which regulates ion flow through the NMDA receptor [68
]. AGRP acts as an antagonist on the melanocortin 4 receptor and blocks the anorexigenic effect of α-melanocyte stimulating hormone (α-MSH). Melanocortin 4 receptor antagonists are very effective in reversing different types of anorectic conditions in animal models, for example, selective melanocortin MC4 receptor blockage reduces immobilization stress-induced anorexia in rats [84
The monoaminergic system has long been believed to be involved in the development of EDs. This is partly due to the psychiatric co-morbidities such as depression and anxiety disorders such as obsessive compulsive disorder which are commonly observed in patients diagnosed with AN. Also, clinical findings show increased CSF monoamine metabolite concentrations detected in AN patients [85
]. Early findings of an association of a SNP in the promoter region of the gene encoding the 5-HT receptor 2a (5HT2A) to development of AN by Collier et al. [86
], also led to this SNP being the single most studied genetic variant in genetic research in EDs. 14 studies have to date been published on this one SNP alone, 28 papers in total on association studies on SNPs related to serotonin function [63
]. Despite this intense focus on variations in the gene encoding 5HT2A replication studies have been highly inconsistent in their results [63
]. Some promising preliminary results have been published on various other SNPs associated with genes related to serotonin function such as 5HT1D [61
], 5HT2C [87
], the norepinephrine transporter (NET) and monoamineoxidase A (MAOA) [91
]. However, these results have yet to be robustly replicated in larger cohorts.
In summary, despite there being indications that certain cognitive deficits and appetitive behaviours in ED are endophenotypic, it is not currently clear which, and how genetic polymorphisms contribute to these behaviours. Neuroimaging studies in to EDs are beginning to show consistent patterns of differential neural activation between those with and without EDs, and between the ED subtypes. However, it is not presently possible to combine genetic research with neuroimaging, because further clarification is needed about which genetic interactions (not single polymorphisms) are associated with ED pathology. Nevertheless, the proposed genes presented here are meant to provide food for thought, to encourage debate and to prompt more clarifying genetic research into EDs.