The study of the brain structures associated with these motivational, personality and neuropsychological variables in obese adolescents could provide more sensitive information about excess weight during adolescence, since regional brain anatomy indices may be considered a more stable measurement ultimately linked to both personality and cognitive modulators associated to the development of particular disorders [7]. Previous evidence from structural imaging studies have revealed that obese adolescents have lower total gray matter (GM) volumes and reduced regional GM volumes in the orbitofrontal cortex compared to lean controls [8], [9]. Moreover, in the Yokun et al. [8] study, higher body mass indices (BMIs) were correlated to volume changes in brain regions involved in reward processing (striatum), memory (middle temporal/parahippocampal gyri), and somatosensory processing (rolandic operculum), whereas reduced regional GM volumes in the prefrontal cortex correlated with steeper rates of BMI increase at 1-year follow-up. Furthermore, Maayan et al. [9] found that obese adolescents were characterized by increased trait disinhibition scores and poorer cognitive control, and that both features correlated with the reduced GM volumes in the orbitofrontal cortex. These findings indicate that volumetric brain measures are useful to characterize the neurobiological underpinnings of adolescent obesity, and that brain structural volumes are associated with both disease-specific features (e.g., BMI) and impulsive personality and cognitive control functions.

Such findings are broadly in agreement with the results from functional imaging studies in obese adolescents and adults, in which these regions seem to play different roles. For example, during the processing of food rewards striatal activation is decreased whereas activations of prefrontal and somatosensory regions are increased in obese adolescents [10], [11]. There is also evidence of increased resting activity in the somatosensory cortices of obese adults [12]. Moreover, the hippocampus is selectively engaged during gastric stimulation and this activation correlates with emotional eating and lack of control in obese adults [13]. Such results have led to hypothesize that decreased striatal functioning and increased somatosensory functioning may be associated with increased reward sensitivity in obese individuals, whereas increased hippocampal and prefrontal reactivity may relate to the balance between the emotional appeal of food and the cognitive control of eating behaviour [10], [12], [13].

The finding of an increased right hippocampal volume in excess weight adolescents is in fitting with the role of this region in the processing of motivational signals associated with appetite [26]. For example, functional imaging studies have shown that right hippocampal activation is significantly associated with food cues-induced insulin release in obese adolescents [27] and with direct gastric stimulation in obese adults [13]. Furthermore, the gastric stimulation-induced increases of hippocampal activity were associated with scores of emotional eating and lack of control [13], supporting the role of this region in the incentive motivation and cognitive control of eating behavior in obesity.

Correlation analyses showed that the regional volume of SII was associated with reward sensitivity and positive urgency in lean controls but not in excess weight adolescents. Within SII, the specific region of correlation with reward sensitivity and positive urgency was the subcentral gyrus, or Brodmann area 43, also known as area OP4 [28]. This area occupies the most lateral aspect of SII, adjacent to the representation of the oral cavity within the primary somatosensory cortex, and thus it is mainly involved in the processing of somatosensory information, including the sensory input relevant for gustatory awareness [29], [30]. Interestingly, somatosensory processing regions have been associated with reward sensitivity in healthy individuals with high scores in this personality trait [31]. Moreover, somatosensory regions consistently show increased activations towards food cues in both adolescents at risk of developing obesity [10] and in obese adolescents [10]. The fact that the negative associations of personality measures with SII volume were only observed within healthy controls would suggest that in excess weight subjects the normal function of somatosensory regions in relation to reward sensitivity and impulsivity is missed or hijacked by disease-specific mechanisms. The latter notion would be similar to what is found in addiction, in which drug craving rewires the function of stimulus-valuation and response control brain regions [32], putatively modifying the link between trait impulsivity and brain structure [33]. In this case, the function of SII may be rewired by the persistent activation of somatosensory regions during anticipation or encoding of sensory and hedonic aspects of palatable food, as shown by fMRI studies [10], [11], [34].

Unlike previous studies [8], [9] we did not find significant structural abnormalities in the prefrontal cortex of excess weight adolescents. However, we found a positive association between cognitive inhibitory control (Stroop performance) and a cluster located in the left dorsolateral prefrontal cortex of normal weight subjects. This region has been shown to mediate the link between aerobic fitness and response inhibition in ageing adults, suggesting a link between physical fitness, production of neurotrophic agents (including insulin-like growth factor-1) and protection of higher-order executive skills [35]. Such region may play a similar role in the developing adolescent brain, and thus in terms of individual differences in response inhibition in normal weight adolescents, which is once again absent in the excess weight group. In agreement with such a notion, over-activity of this region during response inhibition has previously been observed in adolescents compared to healthy adult groups [36]. More research is needed to understand why this link is altered in excess weight adolescents, but the impact of adiposity on vascular health and insulin production may particularly impact frontal brain regions and executive functions [37].

The potential limitations of our study include the decision to merge the overweight and obese subgroups, the lack of significant behavioral performance differences, and the lack of significant volumetric differences in a priori regions of interest such as the orbitofrontal cortex and the striatum. The first decision was based on the observation that comparisons between obese and overweight subgroups failed to yield any significant findings. In addition, the study of dimensional measures of adiposity (BMI) did not either add significant results beyond the categorical diagnosis comparison (normal vs. non normal BMIs). Therefore, we consider that these findings actually reflect that the association between BMI and brain anatomy is better captured by a qualitative analysis comparing participants with vs. without clinical problems related to excess weight. With regard to the lack of behavioral differences and of GM differences in the prefrontal cortex and the striatum, we acknowledge that these negative results are somehow opposed to previous findings, and may reflect the fact that our sample was composed of less severe individuals than those of previous studies including higher BMIs and individuals with other comorbidities [9], [38]. In addition, it might be also argued that the unequal number of voxels included in the different ROIs assessed might have favored the detection of significant differences in smaller regions, such as the medial temporal lobe (in opposition to orbitofrontal or dorsolateral prefrontal cortices, for instance). In any case, we also performed a whole-brain analysis, and, even at an uncorrected significances threshold, we only observed a volume decrease in the left precentral region of excess weight participants, but no findings were observed in the prefrontal cortex or the striatum.

Figure S1