*1.4. From the Behavioral Neuroscientist's Point of View: Is There a "Food Addict" Brain?*

Before answering the question "is there a 'food addict' brain?", it is necessary to remember how drug addiction is described in light of its brain phenotype. As stated earlier, the DSM-5 [118] does not recognize FA in itself, but it identifies different forms of substance-related and addictive disorders (including gambling) that can be used as a reference framework for our discussion. In this context, if we accept the existence of FA, then the neurobiological characteristics of substance-related and addictive disorders should reveal common patterns between food and drug abuse. Valuable recent review papers were aimed at describing common underlying neurobiological mechanisms contributing to drug and FA [119,120]. One of the main pitfalls of these overviews, which is honestly highlighted by the authors but often bypassed in general, is the fact that most human studies taken as support were performed in obese subjects and/or patients suffering from eating disorders (ED), especially bingeing ED-subtype patients. This bias is usually accepted because there are very few studies that aimed at characterizing the brain phenotype/responses of human patients who have been specifically diagnosed with FA. There is consequently a significant risk for circular reasoning: because we make the assumption that FA should resemble drug addiction and its associated brain phenotype, then the observation of this specific brain phenotype in obese and/or bingeing patients should be sufficient to defend the FA hypothesis. The point is that the FA construct must be supported by precise definitions, as well as dedicated neurobiological and neuroimaging studies. These definitions must be supported by concrete data and not only by shortcuts based on analogies with obesity or food abuse. Even though we must assume that substance addiction always starts with substance use, not all obese and/or bingeing ED-subtype patients have FA, and not all "food addicts" are obese.

The diagnosis of an SUD is based on a pathological pattern of behaviors related to use of the substance, and the DSM-5 assists this diagnosis with eleven criteria categorized under four groupings (Table 1). It is important to mention that only one criterion, related to craving (Criterion 4), refers to a specific brain pattern associated with this condition. Craving corresponds to an intense desire or urge for the substance and is described as being associated with the activation of specific reward structures in the brain.

The literature on the neurobiology of addiction provides a consensus on the fact that drugs of abuse, as well as particular excessive behavioral patterns (e.g., gambling), exert a direct activation of the brain reward system. On a chronic basis, they also induce profound neuronal plasticity changes in the corticostriatal and limbic systems. Initially, drugs of abuse trigger abnormal surges of dopamine in the nucleus accumbens, which promotes the direct striatal pathway and inhibits the indirect striato-cortical pathway [121]. Repeated drug consumption and/or administration induce mesolimbic sensitization [122], as well as neuroplasticity changes in the glutamatergic inputs to the striatum and midbrain dopamine neurons. These changes enhance the brain's reactivity to drugs and their associated cues that gain incentive salience, i.e., incentive sensitization [123], reduce the sensitivity to other types of reward, decrease cognitive control mechanisms, and increase the susceptibility to stress and emotional dysregulation [121]. Eventually, there is a transition between controlled to habitual and compulsive use or intake [124]. The precise neuropharmacological mechanisms involved in this transition may depend on the type of drugs used, but a recurrent feature of repeated exposure to substances of abuse is the downregulation of the dopaminergic system, especially the dopamine type-2 receptor (D2R) in the ventral and dorsal striatum. Similar observations have been made in humans [125] and animal models [126], but this is not the scope of our review. Here, we are rather interested in the very few studies that tried to describe the brain phenotype of patients who were specifically diagnosed with FA.

As stated by Fletcher and Kenny [18], information will be lost if we begin with the assumption that drug addiction processes explain food overconsumption and schedule our empirical endeavors exclusively toward a survey of similarities, some of which are superficial and imprecise. In light of the DSM-5 substance use nosology, many authors consider FA as a true addiction [13,127]. As previously stated, the most widely used and accepted tool to measure FA to date is the Yale Food Addiction Scale (YFAS), of which Version 2.0 has been validated in different languages in addition to English [30], including French, Spanish, and Japanese [101,128,129]. As a consequence, we decided to gather information on brain imaging studies that were aimed at describing anatomical and functional features that are characteristic of patients fitting the YFAS criteria for FA (Table 2).

**Table 2.** Studies investigating the brain anatomical or functional specificities associated with YFAS-diagnosed food addiction (FA) in the human.



**Table 2.** *Cont*.

aCC, anterior cingulate cortex; BMI, body mass index; BOLD, blood-oxygen-level-dependent; CAU, caudate; DLPFC, dorsolateral prefrontal cortex; EEG, electroencephalography; ERN, error-related negativity; FA, food addiction; FDG, F-2-fluoro-2-deoxy-glucose; fMRI, functional magnetic resonance imaging; iFG, inferior frontal gyrus; INS, insula; iPL, inferior parietal lobe; lOFC, lateral orbitofrontal cortex; OFC, orbitofrontal cortex; Pe, error positivity; PET, positron emission tomography; TFEQ, three-factor eating questionnaire; YFAS, Yale Food Addiction Scale; yo, years old.

The first study of this kind was performed by Gearhardt herself in collaboration with American colleagues from different teams investigating the neural correlates of eating behavior [136]. These authors used the well-known "milkshake paradigm" of blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) in response to receipt and anticipated receipt of palatable food (i.e., chocolate milkshake). With this paradigm, it has been well described that obese compared to lean individuals show greater activation of the gustatory cortex and oral somatosensory regions in response to anticipated intake and consumption of palatable foods. Obese individuals also show increased activation in the orbitofrontal cortex and putamen in response to palatable food pictures (i.e., reward anticipation), as well as decreased activation in the caudate nucleus in response to consumption of milkshake vs. a tasteless solution (i.e., reward receipt) [137,138]. Gearhardt et al. [136] demonstrated that YFAS scores correlated with greater activation in the anterior cingulate cortex, orbitofrontal cortex, and amygdala in response to anticipated receipt of food. Subjects with higher vs. lower YFAS showed greater activation of the dorsolateral prefrontal cortex and caudate, in response to anticipated receipt of food, but less activation in the lateral orbitofrontal cortex in response to receipt of food. This enhanced anticipation of the rewarding properties of food resembles the reward surfeit theory of obesity, suggesting that individuals at risk for obesity initially show hyper-responsivity of reward circuitry to high-calorie food cues, which would further increase intake of such foods. The fact that some of the reward circuit responses are decreased after consumption of the palatable food illustrates a reward deficit that may drive further intake to fulfill the need for food pleasure. The loss of control over food intake is an important criterion for FA. This inhibitory control was specifically investigated in youth with symptoms of FA by Hardee et al. [130] using a dedicated go/no go task. They demonstrated that YFAS-positive subjects showed deactivation in three clusters of brain regions including the middle temporal gyrus/occipital gyrus, precuneus/calcarine sulcus, and inferior frontal gyrus. The inferior frontal gyrus, notably, has been regularly described as being involved in executive and motor control, and decreased activation of this structure is usually interpreted as a lack of inhibitory control during a go/no go task. The decreased activity in the other clusters was perhaps related to decreased sustained attention during the task; a lack of attention, notably towards interoceptive perceptions, might contribute to the difficulty of obese people to regulate calorie intake, as postulated by Volkow et al. [139]. These results are somewhat corroborated by another study demonstrating that people who meet the YFAS criteria for FA have impaired performance monitoring,

both on the behavioral and neural levels, consequently sharing some neurocognitive characteristics with patients diagnosed with substance use disorder [135].

Surprisingly, Gearhardt et al. [136] found no correlation between YFAS scores and BMI, which suggests that FA can occur in subjects within different body weight categories (as confirmed by the prevalence data previously cited). Even though the authors showed limited differences in reward circuitry activation between high- and low-YFAS subjects during food intake, high-YFAS individuals exhibited patterns of neural activation associated with reduced inhibitory control, which might explain their difficulty to resist food craving. In our opinion, two main questions arise from this work, and they still require additional retrospective and prospective studies to provide answers. First, does a FA profile in normal-weight individuals increase the risk to further declare obesity or other nutritional disorders, as postulated by the reward surfeit theory? Second, since almost all imaging studies performed in obese subjects did not include the YFAS score as a factor in their analysis, what is the probability that the brain pattern associated with YFAS in some "undiagnosed" individuals had influenced the general patterns observed in obese people? Considering the high prevalence of FA in obese patients, there is a significant bias in most studies describing the brain patterns characteristic of obesity, simply because there are many forms and behavioral phenotypes of obesity. Most studies probably characterized brain responses in obese subjects with different clinical profiles, since YFAS was not part of their routine checking, and a fair proportion of their obese subjects might very well have been YFAS-positive. Consequently, this percentage of "undiagnosed" YFAS patients might have influenced and biased our knowledge of the "obesity brain phenotype".

Interestingly, Beyer et al. [131] showed that symptoms of FA were not associated with the major structural brain differences correlated with BMI in the general population, but they might rather explain additional variance towards a lower right lateral orbitofrontal cortex thickness. Whether this anatomical specificity in the orbitofrontal cortex is responsible for the functional differences observed in this particular structure after food reward [136] necessitates further validation. However, the criteria for manifest FA were met by only 6% of the general population in this study [131], which does not indicate what the effects of YFAS symptoms on brain anatomy would be in a large cohort exclusively composed of obese subjects with higher prevalence of FA.

Using (18) F-2-fluoro-2-deoxyglusose (18FDG) positron emission tomography (TEP) instead of BOLD fMRI, Guzzardi et al. [133] investigated in overweight women the brain responses to high-calorie sweet food pictures and found greater activation in the thalamus, hypothalamus, midbrain, putamen, and occipital cortex, but not in the prefrontal and orbitofrontal cortices, in high-YFAS compared to low-YFAS subjects. Interestingly, in high-YFAS women, metabolic responsiveness in the orbitofrontal cortex was progressively lower with increasing YFAS severity and hunger subjective ratings. The authors' conclusions were that inadequate activation in response to the rewarding food in brain regions involved in inhibitory control and reward processing, in spite of greater activation in brain areas involved in somatosensory stimuli processing, reward and memory of hedonic behavior, distinguishes overweight women with FA from women with similar overweight but not FA [133]. It is also very interesting to highlight that the same authors demonstrated that a 3-month low-calorie diet was sufficient to reverse these specific brain activation patterns, which suggests that weight loss (3.8 kg or 4.1% of initial body weight in high YFAS) can help in correcting the neurocognitive anomalies associated with FA [133], exactly as demonstrated for the brain anomalies associated with obesity in formerly obese women who have successfully lost weight [140]. However, restrictive diets are usually ineffective in the long term and should not be advocated alone as obesity treatment.

The nutritional environment consequently has a major role in sustaining or correcting brain anomalies related to FA. The regular consumption of palatable high-calorie foods profoundly modifies many cognitive processes related to food perception, valuation, and motivation. The incentive sensitization theory postulates an excessive amplification of the psychological "wanting" of food, but it also highlights the particular role of external triggering cues and specific attention to these cues in maintaining a vicious circle. Food cue reactivity was found to be modified in overweight or obese women with YFAS-diagnosed FA, towards modest, elevated responses in the superior frontal gyrus for highly processed food pictures and more robust, decreased activations for minimally processed food cues, these responses being opposite in control subjects with similar overweight or obesity [132]. Exteroceptive stimuli such as visual cues are therefore very important in triggering and maintaining the neurocognitive patterns of food addiction. As reminded by Gearhardt et al. [136], activation in the nucleus accumbens is associated with craving in SUD and the amygdala is commonly implicated in drug cue reactivity and craving. Interestingly, Osadchiy et al. [134] demonstrated in healthy subjects with or without elevated BMI that YFAS scores had positive associations with functional connectivity between the amygdala and nucleus accumbens. In the same study, gut microbiota-derived indole metabolites were found to have a direct positive association with BMI and an indirect positive association with YFAS through functional connectivity of the nucleus accumbens [134], which might suggest a role of the gut microbiota in hedonic food intake in the context of FA. Both exteroceptive (e.g., related to food and environment) and interoceptive cues (e.g., related to the internal state and gut microbiota metabolites) are consequently important to understand how the neurocognitive patterns of FA emerge and establish in the long term, with the possibility to increase the risk for further psychological and metabolic disorders.

All these data support the existence of a specific FA brain phenotype that can be detected in normal-weight, overweight, or obese individuals and that is characterized by anomalies in the reward and inhibitory control processes, with likely corollary consequences in the limbic/emotional and cognitive/attentional spheres (Figure 2). Even though a recent meta-analysis of fMRI studies defends an addiction model of obesity, characterized by reduced cognitive control and interoceptive brain responses [141], this vision is probably restricted to part of the obesity spectrum and cannot be generalized to all forms of obesity. Further research is needed to better phenotype the neurobehavioral patterns of YFAS-positive subjects and disentangle their complex relationships and overlap with other diseases including obesity and other forms of addiction. Such work is mandatory to improve medical care because a better understanding of the patients' specificities leads to better treatment. As reminded by Ho et al. [142], post-obesity surgery patients are at increasing risk for developing alcohol and SUD, which likely represents an "addiction transfer" from food to other means of fulfilling the individuals' drives for pleasure or comfort. This risk could be especially increased if the presence of an FA profile has not been diagnosed and treated beforehand. The YFAS 2.0 questionnaire is a useful tool to predict continued emotional and binge eating behavior following obesity surgery [143] and might be used to identify subpopulations of patients with higher risk for unsuccessful obesity surgery. However, as a questionnaire, this method remains limited by the usual constraints and uncertainties of declarative diagnostic methods, which necessitates the development of additional diagnostic tools and markers, derived from brain imaging or biological measurements at the gut–microbiota–brain level, for example.

**Figure 2.** Neurocognitive functions and brain areas that are impacted by food addiction and for which people who meet the YFAS criteria for food addiction have different brain activity, metabolism, or functional connectivity compared to normal subjects. Please refer to Table 2 for details on results and imaging modalities used. Brain schematic representations were collected from Servier Medical Art (Suresnes, France; http://www.servier.fr).
