Altered hippocampal effective connectivity predicts BMI and food approach behavior in children with obesity

Abstract

Objective: The vicious circle model of obesity proposes that the hippocampus plays a crucial role in food reward processing and obesity. However, few studies focused on whether and how pediatric obesity influences the potential direction of information exchange between the hippocampus and key regions, as well as whether these alterations in neural interaction could predict future BMI and eating behaviors.

Methods: In this longitudinal study, a total of 39 children with excess weight (overweight/obesity) and 51 children with normal weight, aged 8 to 12, underwent resting-state fMRI. One year later, we conducted follow-up assessments of eating behaviors and BMI. Resting-state functional connectivity and spectral dynamic casual modeling (spDCM) technique were used to examine altered functional and effective connectivity (EC) of the hippocampus in children with overweight/obesity. Linear support vector regression, a machine learning method, was employed to further investigate whether these sensitive hippocampal connections at baseline could predict future BMI and eating behaviors.

Results: Compared to controls, children with excess weight displayed abnormal bidirectional inhibitory effects between the right hippocampus and left postcentral gyrus (PoCG), that is, stronger inhibitory hippocampus→PoCG EC but weaker inhibitory PoCG→hippocampus EC, which further predicted BMI and food approach behavior one year later.

Conclusion: These findings point to a particularly important role of abnormal information exchange between the hippocampus and somatosensory cortex in pediatric obesity and future food approach behavior, which provide novel insights into the neural hierarchical mechanisms underlying childhood obesity and further expand the spDCM model of adult obesity by identifying the directionality of abnormal influences between crucial circuits associated with appetitive regulation.

Keywords: Childhood obesity; Eating behaviors; Effective connectivity; Hippocampus; Spectral dynamic casual modeling.

Fig. 1

Different effective connectivity between groups. In panel A, the bar graphs show the group-level values of the HIP→PoCG connection. In panel B, the green lines represent inhibitory effects between regions. The dark plus (+) and light minus (–) signs indicate the stronger and weaker connectivity effect (strength) in the EW group (compared to NW group), respectively. In panel C, the bar graphs show the group-level values of the PoCG→HIP connection. All displayed connections are for posterior probability > 0.95. Abbreviation: HIP = hippocampus; PoCG = postcentral gyrus; EW = excess weight group; NW = normal weight group.

Fig. 2

Results of effective connectivity predicting longitudinal variables in the excess weight group. In panel A, the plot showed that the baseline HIP→PoCG connection significantly predicted BMI z-score, food approach, and food as reward one year later, using support vector regression with LOOCV. In panel B, the baseline PoCG→HIP connection significantly predicted food approach one year later, using support vector regression with LOOCV. The p_perm values were based on 1000 times permutation tests. Abbreviation: HIP = hippocampus; PoCG = postcentral gyrus; LOOCV = leave-one-out cross-validation.

Fig. 3

Opposite patterns of the relationship of effective connectivity with eating behaviors and BMI one year later. In panel A, the HIP→PoCG connection was inversely associated with food responsiveness and BMI z-score, while in panel B, the PoCG→HIP connection was positively associated with food responsiveness and BMI z-score. * p < 0.05. Abbreviation: HIP = hippocampus; PoCG = postcentral gyrus; CEBQ = children's eating behaviour questionnaire; CFQ = child feeding qestionnaire; EDI-C = eating disorder inventory-child; FCT = food choice task; BMI = body mass index.

Fig. 4

The extended neural model of childhood obesity. Compared to the NW group, the EW group exhibited stronger inhibitory effect from RN to ECN, stronger excitatory effect from RN to SN, and weaker inhibitory effect from RN and SN to EHN. The green and yellow lines refer to inhibitory and excitatory effect between regions, respectively. The plus (+) and minus (–) signs indicate the stronger and weaker connectivity effect (strength) in the EW group (compared to NW group), respectively. All displayed connections are for posterior probability > 0.95. Abbreviation: EHN = energy homeostatic network; SN = somatosensory network; RN = reward network; ECN = executive control network; EW = excess weight group; NW = normal weight group.