Functional Connectivity of the Nucleus Accumbens and Changes in Appetite in Patients With Depression

Abstract

Importance: Major depressive disorder (MDD) is characterized by a substantial burden on health, including changes in appetite and body weight. Heterogeneity of depressive symptoms has hampered the identification of biomarkers that robustly generalize to most patients, thus calling for symptom-based mapping.

Objective: To define the functional architecture of the reward circuit subserving increases vs decreases in appetite and body weight in patients with MDD by specifying their contributions and influence on disease biomarkers using resting-state functional connectivity (FC).

Design, setting, and participants: In this case-control study, functional magnetic resonance imaging (fMRI) data were taken from the Marburg-Münster FOR 2107 Affective Disorder Cohort Study (MACS), collected between September 2014 and November 2016. Cross-sectional data of patients with MDD (n = 407) and healthy control participants (n = 400) were analyzed from March 2018 to June 2022.

Main outcomes and measures: Changes in appetite during the depressive episode and their association with FC were examined using fMRI. By taking the nucleus accumbens (NAcc) as seed of the reward circuit, associations with opposing changes in appetite were mapped, and a sparse symptom-specific elastic-net model was built with 10-fold cross-validation.

Results: Among 407 patients with MDD, 249 (61.2%) were women, and the mean (SD) age was 36.79 (13.4) years. Reduced NAcc-based FC to the ventromedial prefrontal cortex (vmPFC) and the hippocampus was associated with reduced appetite (vmPFC: bootstrap r = 0.13; 95% CI, 0.02-0.23; hippocampus: bootstrap r = 0.15; 95% CI, 0.05-0.26). In contrast, reduced NAcc-based FC to the insular ingestive cortex was associated with increased appetite (bootstrap r = -0.14; 95% CI, -0.24 to -0.04). Critically, the cross-validated elastic-net model reflected changes in appetite based on NAcc FC and explained variance increased with increasing symptom severity (all patients: bootstrap r = 0.24; 95% CI, 0.16-0.31; patients with Beck Depression Inventory score of 28 or greater: bootstrap r = 0.42; 95% CI, 0.25-0.58). In contrast, NAcc FC did not classify diagnosis (MDD vs healthy control).

Conclusions and relevance: In this study, NAcc-based FC reflected important individual differences in appetite and body weight in patients with depression that can be leveraged for personalized prediction. However, classification of diagnosis using NAcc-based FC did not exceed chance levels. Such symptom-specific associations emphasize the need to map biomarkers onto more confined facets of psychopathology to improve the classification and treatment of MDD.

Figure 1.. Developed Scoring of Opposite Changes in Appetite and Body Weight in Patients With Major Depressive Disorder (MDD)

A, Heat map depicting the scores of the 4 positive items taken from the Structured Interview Guide for the Hamilton Depression Rating Scale With Atypical Depression Supplement interview (atypical Hamilton Depression scale [aHAMD]) and the 4 negative items taken from the regular set of items from HAMD and Beck Depression Inventory (BDI) as well as scores for appetite score, HAMD, and BDI (each z score–transformed for display). B, Distribution of the composite appetite score reflecting changes in appetite across severity categories according to BDI. The spread of the appetite score increased with increasing severity, but there was little shift in the average score. C, Body mass index (BMI) was positively associated with appetite score, albeit weakly. The hot color indicates higher density of points. D, Higher symptom severity according to BDI was associated with stronger changes in appetite. However, although the negative items contributed to the total BDI score, the linear association with the appetite score was weak, indicating that changes in appetite can be statistically dissociated from BMI and symptom severity.

Figure 2.. Association of Body Mass Index (BMI) With Functional Connectivity (FC) Between the Nucleus Accumbens (NAcc) and the Hypothalamus

A, Representative sections showing decreased NAcc-based FC with increasing BMI. Colors indicate t values of the contrast, and only clusters exceeding the extent threshold k ≥ 100 are depicted. Slice numbers indicate Montreal Neurological Institute (MNI) coordinates; for details, see the eTable in the Supplement. B, The association was equally strong in both groups. C, Bootstrapping indicated robust and largely overlapping correlation coefficients. HC indicates healthy control; MDD, major depressive disorder. ^aP < .001. ^bP < .01.

Figure 3.. Association Between Nucleus Accumbens (NAcc)–Based Functional Connectivity (FC) With Appetite

A, Sagittal sections showing increased and decreased NAcc-based FC with colors reflecting t values of the contrast. Colors indicate t values of the contrast, and only clusters exceeding the extent threshold k ≥ 100 are depicted. Slice numbers indicate Montreal Neurological Institute (MNI) coordinates; for details, see the eTable in the Supplement. B, Bar plots depicting average unsmoothed FC to corresponding regions of interest (ROIs) from the anatomical atlas in all groups. Error bars depict 95% CIs. Relative to healthy controls (HCs), patients with major depressive disorder (MDD) with decreased appetite (appetite score less than 0, in blue) showed lower FC to the vmPFC and hippocampus, whereas patients with MDD with increased appetite (appetite score greater than 0, in light blue) had similar average FC in these regions. In contrast, patients with MDD with increased appetite showed a reduced NAcc-based FC with the frontal operculum. Density plots show the distribution of bootstrapped correlation coefficients, with the color corresponding to the color of the ROI.

Figure 4.. Association of Functional Connectivity Profiles of the Nucleus Accumbens (NAcc) With Changes in Appetite in Patients With Major Depressive Disorder (MDD)

A, Schematic summary of the elastic net model. We obtained individual seed-based functional connectivity profiles and used an elastic net to select the best features for cross-validated associations across the whole sample of patients. B, The trained model reflected changes in appetite. However, because of the restricted variance in appetite changes in patients with low symptom severity, the elastic net model’s predicted score was more accurate for severe cases. C, To map the improved association with changes in appetite as severity increased, we bootstrapped the correlation of the elastic net model predicted score and observed appetite scores while gradually increasing the symptom severity threshold for inclusion. Although the model was trained on the whole sample, restricting the elastic net model to severe cases gradually increased accuracy. D, Increasing the Beck Depression Inventory (BDI) inclusion threshold leads to gradually increasing associations exceeding models trained on permuted appetite scores (orange line) for moderate levels of severity (green diamonds above the orange line). E, The slope of increased accuracy with increasing BDI threshold (vertical dotted line) was compared with the permuted models (gray distribution). ^aP < .01.