Nucleus Accumbens Functional Connectivity with the Frontoparietal Network Predicts Subsequent Change in Body Mass Index for American Children

Background: Nucleus accumbens (NAc) is a brain structure with a well-established role in the brain reward processing system. Altered function of the NAc is shown to have a role in the development of food addiction and obesity. However, less is known about sex differences in the role of NAc function as a predictor of children’s change in body mass index (BMI) over time. Aim: We used the Adolescent Brain Cognitive Development data (version 2.01) to investigate sex differences in the predictive role of the NAc functional connectivity with the frontoparietal network on children’s BMI change over a one-year follow-up period. Methods: This 1-year longitudinal study successfully followed 3784 9–10-year-old children. Regression models were used to analyze the data. The predictor variable was NAc functional connectivity with the frontoparietal network measured using resting-state functional magnetic resonance imaging (fMRI). The primary outcome was BMI at the end of the 1-year follow up. Covariates included race, ethnicity, age, socioeconomic factors, and baseline BMI. Sex was the effect modifier. Results: NAc functional connectivity with the frontoparietal network was predictive of BMI changes over time. This association remained significant above and beyond all covariates. The above association, however, was only significant in female, not male children. Conclusion: The epidemiological observation that NAc functional connectivity is associated with BMI changes in children is an extension of well-controlled laboratory studies that have established the role of the NAc in the brain reward processing. More research is needed on sex differences in the brain regions that contribute to childhood obesity.

The NAc mediates cue-triggered reward-seeking behaviors [9,18,[22][23][24]. In animal models [25] and humans [26][27][28][29][30][31], the obesity-prone differ from the obesity-resistant in NAc activity. Due to its central role in incentive motivation, the NAc is involved in the expression of behaviors that contribute to the development of obesity [25]. Food cues are shown to elicit a robust NAc dopamine response at the time frontoparietal network and the automatic subcortical segmentation (ASEG) region of interest (ROI) right-accumbens area was calculated. Tesla 1 (T1) and T2 weighted fMRI and sMRI images were taken using a 3 tesla (T) Siemens Prisma (Erlangen, Germany), General Electric 750 (Chicago, IL, USA), and Phillips (Amsterdam, The Netherlands) multi-channel coiled scanners, all capable of multiband echo-planar imaging (EPI) acquisitions. A localizer was implemented at the beginning of each scan, followed by a T1 weighted fMRI image acquisition. Functional T2 weighted scans were then acquired at rest and then throughout three psychological tasks. A large-scale multimodal data acquisition allowed the ABCD study to collect an unprecedented set of images on a large number of adolescents that were selected and enrolled via 21 data acquisition sites across various U.S. states. As mentioned above, both ABCD structural and functional MRI data are processed and the tabulated regions of interest (ROI) data were downloaded from the National Institute of Health (NIH) Data Archive (NDA). ABCD imaging data were limited to resting-state fMRI. More information on the imaging protocols are available here [52][53][54].

Outcome
Body Mass Index (BMI). The children's BMI at baseline and at the end of the 1-year follow-up was calculated based on participants' measured height and weight. Height was measured three times in inches. The weight of the child was measured up to three times in pounds. BMI was treated as a continuous measure. Although some research has used percentile BMI related to age and sex norms, we used raw BMI because the analyses broke the sample by sex and the age range of the participants was very narrow (only 9-and 10-year-old children). Other research has also used BMI rather than percentile BMI related to age and sex norm.

Independent Variable
NAc. Using resting-state fMRI, NAc was defined as the average correlation between the frontoparietal network and ASEG ROI right-accumbens area. This is functional connectivity between the frontoparietal network and the NAc. The resting-state fMRI was measured at baseline at the same time that socioeconomic status (SES) indicators and baseline BMI were measured.

Confounders
Race. Race, a self-identified variable, was two binary variables: Blacks, other races, and Whites (reference category). Ethnicity. Ethnicity, a self-identified variable, was 1 for Hispanics and 0 for non-Hispanics (reference category).
Sex. A dichotomous variable, sex was coded as below: males = 1, females = 0. Age. Age (months), calculated as the difference between birth and the time of enrollment to the study, measured in months, was reported by parents.
Parent education (y). Parents reported their years of schooling. This variable was operationalized as a continuous (interval) variable ranging from 0 for no formal education to 21 for doctoral degrees.
Parent employment. Parents reported their employment. This variable was operationalized as a categorical variable with 0 for non-employed and 1 for employed.

Main Data Analysis
The statistical package, SPSS 22.0 (IBM, Armonk, NY, USA), was applied for data analysis. Mean (standard deviation; SD) and frequency (relative frequency; %) was used to describe all the study variables. We also used the Pearson correlation test for bivariate analysis of the associations between the study variables. For multivariable modeling, we ran four regression models. To do so, first we tested the assumptions that are required for running a regression model. BMI time 1 and time 2 had a near to normal distribution. There is no collinearity between the study variables. In our model, the NAc functional connectivity with the frontoparietal network was used as the independent variable (predictor), baseline BMI and demographic and SES indicators as the covariates and follow-up BMI as the outcome. Our 1st two models were performed in the pooled sample. Our 1st model was without (Model 1), and our 2nd model was with (Model 2) an interaction term between sex and NAc functional connectivity with the frontoparietal network. Our next models were run for females (Model 3) and males (Model 4). Models were identical for right and left NAc. All models were statistically significant at a 0.001 and explained more than 20% of the variance of the outcome, mainly because the outcome and independent variable (predictor) had a strong correlation (baseline and time 2 BMI). This information is added to the statistical analysis section of the paper. We did not adjust the BMIs based on the age and sex of the participants because the BMI values of each individual were being compared at two time points (baseline and follow up BMI). As every individual is compared to his/herself, there is no need to adjust based on the distribution of BMI in the population. There are several studies that have used the same approach and have not adjusted based on norms. Unstandardized coefficient (b), standard error (SE), 95% confidence interval (95% CI), and p-value were reported for our model. A p-value equal or less 0.05 was significant.

Sensitivity Analysis
In our main analysis, the outcome was BMI time 2, while BMI time 1 and all covariates were controlled. We ran a mixed effects model with BMI time 1 as the outcome. The argument behind this approach if the result of our main analysis with BMI time 2 is robust is that we should test the same interaction in our replication model, despite the sample size and the temporal aspects of the measures having changed. As our findings show, our sensitivity analysis replicated the same interaction between sex and NAc functional connectivity with the frontoparietal network on BMI (Appendix A Tables and Figures).

Ethics
The ABCD study protocol received Institutional Review Board (IRB) approval from several institutions, including but not limited to the University of California, San Diego (UCSD). All participating children provided assent. All participating parents signed informed consent [50]. As we only performed a secondary analysis of fully de-identified data, our study was non-human subject research. Thus, our analysis and report did not require an IRB review (exempt from a full IRB review).

Descriptives
A total number of 3784, 9-10-year-old children were analyzed. Participants were 1953 male and 1831 female children. Table 1 presents a summary of the descriptive statistics for the children overall and by sex. Nucleus accumbens (NAc): average correlation between frontoparietal network and right-and left-accumbens area; SD = standard deviation, BMI = body mass index. * p < 0.05 for a comparison of males and females. Table 2 shows a summary of bivariate correlations. Functional connectivity between the frontoparietal network and the NAc was inversely correlated with follow-up BMI. The bivariate correlation between right and left functional connectivity between NAc and frontoparietal network was almost zero. Although baseline BMI was not associated with right and left functional connectivity between the NAc and frontoparietal network, follow-up BMI was correlated with right and left functional connectivity between the NAc and frontoparietal network. Baseline and follow-up BMI were positively correlated; however, this correlation was not very strong. From various SES indicators, parental marital status and parental education were correlated with right but not left functional connectivity between the NAc and frontoparietal network. Table 3 shows a summary of the results of two regression models in the pooled sample. These models show a significant and negative effect of NAc functional connectivity with the frontoparietal network on follow-up BMI in children. This means a higher beta coefficient that reflected higher NAc functional connectivity with the frontoparietal network was associated with lower BMI gain over time. Results were identical for right and left NAc.  NAc: average correlation between the frontoparietal network and ASEG ROI right-and left-accumbens area; BMI = body mass index; * p < 0.05; ** p < 0.01.  Table 4 shows a summary of the results of two regression models in female and male children.

Sensitivity Analysis
As the appendix shows, we ran a full series of sensitivity analyses to test if we could successfully replicate the results of our main analysis. As shown above, our main analysis shows that the effect of NAc functional connectivity with the frontoparietal network on BMI time 2, while covariates and BMI time 1 was controlled. We ran a mixed effects model with BMI time 1 as the outcome. The argument behind this approach is if the result of our main analysis with BMI time 2 is robust, then we should test the same interaction in our replication model, despite the sample size and the temporal aspects of the measures having changed. As our findings show, our sensitivity analysis replicated the same interaction between sex and NAc functional connectivity with the frontoparietal network on BMI (Appendix A Tables and Figures). That means that not only does the effect of NAc functional connectivity with the frontoparietal network on future BMI depend on sex, but the cross-sectional association between NAc functional connectivity with the frontoparietal network and BMI also statistically differs between male and female children. NAc: average correlation between the frontoparietal network and right-or left-accumbens area (rsfmri_cor_ngd_fopa_scs_aarh or rsfmri_cor_ngd_fopa_scs_aalh); Outcome: BMI at the end of follow up; unstandardized regression coefficient: b; statistic: t; standard error: SE; confidence interval: CI.

Discussion
Our first main finding suggested that the NAc functional connectivity with the frontoparietal network was associated with 1-year BMI among 9-10-year-old American children. According to our second main finding, this predictive role was true for female children, but not male children. We also replicated these main results by cross-sectional association between NAc functional connectivity with the frontoparietal network and BMI, which also differed by sex. Another source of confidence in our results was that the findings were identical for right and left NAc.
Our first main finding is supported by the lab-based observations that the basal ganglia (striatum and NAc) and frontoparietal network function together [35], and that their functional connectivity is also linked to food preference [36], obesity [37][38][39][40][41], and eating disorders [42,43]. An extensive body of laboratory research has shown that the NAc shell receives dopaminergic inputs from various structures due to various sensory inputs including those related to food [26,[55][56][57][58][59][60]. The NAc is involved in the regulation of feeding, eating, and food seeking behaviors [1,7,8,20,24,55,61]. Gamma-aminobutyric acid-A (GABA A ) receptors (a receptor for the GABA hormone released by the brain to regulate dopamine levels in its reward pathways) in the NAc shell mediate hyperphagia, overeating, and associated weight gain [62]. NAc controls appetite as well [63]. Motivational responses to food are mediated in part by NAc α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor transmission [64]. NAc medium spiny neurons (MSNs) are hyper-responsive in obesity-prone individuals [34]. The NAc's interaction with predisposition factors and obesogenic environment may alter neurobehavioral plasticity in the NAc, which promotes weight gain and reduces weight loss in obesity-susceptible animals and people [34]. Recent work shows that cue-triggered motivation is enhanced in obesity-susceptible individuals after consuming "junk-food" [64]. Chronic and repeated exposure to food cues and specific diets may result in changes within the NAc, a part of the mesolimbic pathway that regulates food seeking behaviors [65]. The NAc's interaction with predisposition factors and an obesogenic environment may alter neurobehavioral plasticity in the NAc which may promote weight gain and reduce weight loss among obesity-susceptible animals and people [34].
Our second main finding, the predictive role of the frontoparietal network and NAc functional connectivity on BMI time 2 of female but not male children is very novel. An extensive body of research has previously shown that sex alters predictors of obesity and BMI change over time [66][67][68][69][70]. Socioeconomic status, mood regulation, and stress all differently correlate with BMI of males and females [66][67][68][69][70]. In many studies, obesity and BMI change have shown more associations in females than males [66][67][68][69][70]. This might be because females are more vulnerable to the risk factors of obesity, given their environmental situation, societal status, and coping mechanisms [66,[68][69][70][71][72][73][74][75][76], or simply because obesity has a larger variance in females so our existing models can better explain BMI variation in females than males.
In our main analysis, we had follow-up BMI as the outcome and baseline BMI as a covariate. This is one of the two main approaches to analyze the repeated measure outcome only with two observations. Maria Glymour and others have discussed how controlling for baseline status of an outcome alters the interpretation of the results. We decided to control for baseline BMI when the outcome was time 2 BMI based on the causal diagram [77]. Still, to replicate the results in the same study, we ran a series of sensitivity analysis with BMI time 1 as the outcome. The interaction term between sex and NAc functional connectivity with the frontoparietal network on baseline BMI suggested that our results are robust.

Implications
Function of the NAc is believed to be modifiable through neuro-modulation [78]. NAc function has been modulated in both research and clinical settings [20,[79][80][81][82]. Smoking cessation and weight loss programs, for example, have used modalities that leverage NAc modification [78]. Some programs that have stimulated the NAc have generated promising results in reducing food craving, as well as treatment of obesity-induced food addiction and other cue-related disorders [78]. Deep brain stimulation of the NAc is also a technique that can be implemented in therapies [20,55,[79][80][81][82][83][84][85].

Limitations
All studies, particularly those who are secondary analysis of some existing data, are limited in their design, methods, and measurements. NAc activity and high BMI may have bidirectional effects. However, we used a longitudinal design and established the longitudinal effect of baseline NAc function on children's BMI change over time. We only had a few confounders, and we did not have data on diet, food cues, obesogenic environment, mental health, emotion regulation, and food addiction. Finally, we should emphasize that in this study, our outcome was not a dichotomous variable of obesity but the continuous measure of BMI. As such, the result should not be interpreted as NAc function predicts obesity, but BMI value at time 2 (while BMI value at time 1 is controlled). That means, regardless of passing or not passing the threshold that we use to define obesity or overweight, NAc function predicts who develops a higher BMI over time.

Future Research
As the duration of follow up in our study was short, long-term longitudinal data are needed. Imaging studies on longitudinal association between the NAc, and behavioral development are needed. We only described a link between the NAc and BMI change. Future research should study the mechanism of the effects of the NAc on BMI change. Future work is needed on various social, behaviors, and brain mechanisms that can explain how baseline NAc function predicts future obesity, and how Brain Sci. 2020, 10, 703 9 of 16 these paths may differ for various groups of American children. As the NAc is not only responsive to food but also other categories of rewards such as alcohol [65] and sex [86,87], future research should replicate this result for other cue-induced behaviors. Research should also investigate how food environment and past behaviors modulate the food cue-induced behaviors that regulate eating through a change in the NAc activity. The same can be relevant to tobacco, alcohol, illicit drugs, and sex. Changes in the NAc may regulate, promote, or inhibit tobacco, food, alcohol, and drug seeking behaviors [88][89][90][91][92]. At a public health level, knowledge regarding the role of the NAc on BMI and obesity may help us undo social, economic, and environmental inequalities in childhood obesity. It is important to study how social distribution of food cues generate inequalities and disparities in the burden of obesity in children. We need to study environmental risk factors that impact NAc function across populations of children. We need to know the societal and behavioral conditions that may alter the function of the NAc in children.

Conclusions
In this longitudinal epidemiological study, the NAc functional connectivity with the frontoparietal network was predictive of future BMI increase among female but not male American children. The NAc, the frontoparietal network, and their functional connectivity may be a part of the brain circuits involved in the development of obesity among female children. Acknowledgments: Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Development (ABCD) Study (https://abcdstudy.org), held in the NIMH Data Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 10,000 children age 9-10 and follow them over 10 years into early adulthood. A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/principal-investigators.html. ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data repository grows and changes over time. The ABCD data used in this report came from DOI: 10.15154/1504041 which can be found at: https://nda.nih.gov/study.html?id=721.    Model also controlled for the nested data. * p < 0.05; **p < 0.01; *** p < 0.001.