Discrimination exposure impacts unhealthy processing of food cues: crosstalk between the brain and gut

Racial and ethnic disparities in obesity persist, with minority groups experiencing higher rates of obesity and related morbidities. Beyond genetics, diet, physical activity and psychological factors, discrimination is a salient psychosocial stressor that may increase obesity risk by influencing ingestive behavior, increasing appetite and cravings for highly palatable foods. Stress alters food-cue reactivity in neuroimaging studies and can deactivate frontal executive control while potentiating limbic reward responses, biasing choices toward energy-dense foods. Discrimination may also perturb the brain–gut–microbiome (BGM) axis via vagal, immune-inflammatory, microbial metabolite, neurotransmitter and HPA pathways, promoting gut dysbiosis, increased gut permeability and inflammation. Dysregulation of glutamate metabolism is implicated in stress-related disorders and obesity through excitotoxicity and oxidative stress, and stress-related alterations in gut glutamate-pathway metabolites have been linked to changes in brain networks subserving cognition and emotion. This study examined whether higher discrimination exposure relates to altered neural reactivity to unhealthy versus healthy food cues, altered gut glutamate-pathway metabolites, and their brain–gut associations. The authors hypothesized that greater discrimination-related stress would be associated with heightened reactivity to unhealthy hyperpalatable foods in reward and executive control regions, altered glutamate-pathway metabolites linked to oxidative stress and inflammation, and stronger brain–gut coupling particularly for unhealthy sweet foods.

Prior work indicates that discrimination and chronic stress engage brain networks involved in reward and cognitive control, including central executive and limbic circuits, and are associated with altered functional connectivity and amygdala activity. Stress can enhance responsivity to palatable food cues and impair prefrontal executive control, contributing to obesogenic eating. Stress also affects the gut microbiome, promoting dysbiosis, leaky gut and inflammation, and unhealthy diets can reinforce dysregulated eating via microbiome-driven mechanisms. Glutamate metabolism plays key roles in CNS inflammation, excitotoxicity and oxidative stress, and is implicated in depression, anxiety and obesity. Early-life stress associates with altered gut glutamate metabolites and brain connectivity changes. Glutamatergic signaling contributes to executive control and reward processing relevant to food-cue reactivity. Together, these lines of evidence support investigating discrimination-related alterations across brain food-cue processing and gut glutamate metabolites within the BGM framework.

Design and participants: Cross-sectional study of 107 community-recruited adults (87 women) from Los Angeles. Exclusions included major medical/neurological or psychiatric conditions, vascular disease/diabetes, weight-loss/abdominal surgeries, substance use disorders, tobacco dependence (≥1/2 pack daily), CNS-interfering medications, regular analgesics, pregnancy/breastfeeding, extreme strenuous exercise (>8 h/week), MRI contraindications and weight >181 kg. Women were scanned during the follicular phase; peri-/post-menopausal women were excluded. IRB-approved; written informed consent obtained. Measures: Demographics (age, sex), BMI, race, subjective socioeconomic status (MacArthur Scale), and diet (standard vs nonstandard American diet). Discrimination was assessed by the Everyday Discrimination Scale (EDS). Based on the sample median, participants were dichotomized into high (EDS > 10; n=50) and low (EDS ≤ 10; n=57) discrimination exposure groups; EDS=0 were excluded. Food-cue fMRI task: Participants fasted ~6 h prior to scanning (confirmed). Visual blocks of images: unhealthy (high-calorie) savory, unhealthy (high-calorie) sweet, healthy (low-calorie) savory, healthy (low-calorie) sweet, and nonfood (pixelated controls). Images (blocks of six; 3 s each) alternated with 12 s fixation; two slideshows with different orders were presented. Post-scan, willingness to eat the viewed foods was rated from 0 (not at all) to 10 (very much). MRI acquisition and preprocessing: 3.0 T Siemens Prisma. Preprocessing in FSL FEAT v6.0: motion correction, brain extraction, high-pass filtering (100 s), 5-mm FWHM spatial smoothing; registration to MNI space via FLIRT. First-level analyses produced 10 contrast maps per participant. Whole-brain analysis: Group-level mixed-effects (FLAME) unpaired t-tests contrasted high vs low discrimination across: (1) unhealthy sweet vs nonfood, (2) unhealthy savory vs nonfood, (3) healthy food vs nonfood, (4) unhealthy sweet vs healthy sweet, (5) unhealthy savory vs healthy savory, plus reverse contrasts. Covariates: BMI, age, sex, race, diet, SES. Family-wise error cluster correction: Z > 2.3, cluster P < 0.05. ROI analysis: Significant clusters from whole-brain contrasts (high vs low) were combined into a discrimination-related food-cue ROI mask. For each participant, β values (first-level) were extracted and tested via multiple linear regression for association with EDS scores, adjusting for BMI, age, sex, race, diet, SES; FDR correction by Benjamini-Hochberg. Fecal metabolomics: Subsample (n=62; high discrimination n=30; low discrimination n=32) provided fecal samples stored at −80 °C and processed by Metabolon using UHPLC-MS/MS in a single batch. Data curation included median imputation for <3% missingness, interquartile range denoising, internal standard normalization, and compilation for analysis in MetaboAnalyst. A priori focus on 12 glutamate-pathway metabolites. Metabolite analysis: Generalized linear models compared metabolite levels between discrimination groups controlling for BMI, age, sex, race, diet, SES. Multiple testing controlled using FDR. Pearson correlations assessed associations with SES. Willingness-to-eat analysis: Generalized linear models compared ratings (unhealthy and healthy foods separately) between discrimination groups with the same covariates. Structural equation modeling (SEM): Using lavaan in R, a latent gut metabolite variable comprised glutamate-pathway metabolites differing by discrimination (N-acetylglutamate, N-acetylglutamine). Three SEMs related EDS (discrimination), brain reactivity (ROI β for the relevant contrast), and gut metabolites for: unhealthy sweet vs nonfood, unhealthy savory vs nonfood, and healthy vs nonfood, controlling for BMI, race, diet, SES. Fit indices: CFI > 0.9, RMSEA < 0.08, GFI > 0.9, SRMR < 0.08. Significance set at P < 0.05. Baseline stats: Student’s t-tests and χ2 tests compared demographics/clinical characteristics; two-way ANOVA tested discrimination group by dietary style interaction on BMI.

• Participant characteristics: No significant differences by discrimination group in sex, age, BMI, education, marital status, income, or diet; SES was lower in the high discrimination group (P = 0.01). No interaction effects of discrimination and diet on BMI.
• Whole-brain fMRI: Compared to low discrimination, the high discrimination group showed greater food-cue reactivity: • Unhealthy sweet vs nonfood: increased activation in insula, inferior frontal gyrus (pars opercularis), lateral orbitofrontal cortex, frontal operculum; additional right orbitofrontal cortex cluster (Table 2; cluster-corrected Z > 2.3, P < 0.05). • Unhealthy savory vs nonfood: greater activation in caudate, putamen, insula, frontal pole, lateral orbitofrontal cortex. • Healthy food vs nonfood: greater activation in superior frontal gyrus and middle frontal gyrus (dorsolateral prefrontal regions). • Unhealthy sweet vs healthy sweet: high discrimination exhibited lower reactivity in vmPFC relative to low discrimination. Unhealthy savory vs healthy savory: no significant differences. • Reverse contrasts (low > high) were not significant for these comparisons.
• ROI analysis: EDS scores correlated positively with greater reactivity in the composite food-cue ROI to: • Unhealthy sweet foods: β = 0.29, q = 0.03 • Unhealthy savory foods: β = 0.32, q = 0.03 • Healthy foods: β = 0.72, q < 0.001
• Gut metabolites (glutamate pathway; n=62): High discrimination associated with higher levels of N-acetylglutamine (P = 0.0021, FDR q = 0.0247) and N-acetylglutamate (P = 0.0490, FDR q = 0.2938). Neither metabolite correlated significantly with SES.
• Willingness to eat: High discrimination group reported higher willingness to eat unhealthy foods (P = 0.048); no significant difference for healthy foods (P = 0.174).
• SEM results (controlling for BMI, race, diet, SES): • Unhealthy sweet model: discrimination → brain reactivity (standardized coeff = 0.31, P = 0.009); discrimination → glutamate metabolism (0.42, P = 0.004); significant brain ↔ gut association (0.34, P = 0.048). Fit: RMSEA 0.0, CFI 1.0, GFI 0.955, SRMR 0.071. • Unhealthy savory model: discrimination → brain (0.249, P = 0.043); discrimination → glutamate metabolism (0.462, P = 0.006); brain ↔ gut not significant (−0.227, P = 0.216). Fit: RMSEA 0.0, CFI 1.0, GFI 0.964, SRMR 0.058. • Healthy model: discrimination → brain (0.454, P < 0.001); discrimination → glutamate metabolism (0.445, P = 0.004); brain ↔ gut not significant (0.155, P = 0.373). Fit: RMSEA 0.0, CFI 1.0, GFI 0.969, SRMR 0.058.

Findings support that discrimination exposure, as a chronic stressor, is associated with heightened neural reactivity to food cues in frontal–striatal networks mediating reward, motivation and executive control, especially for unhealthy foods, aligning with increased willingness to eat unhealthy items. Elevated gut glutamate-pathway metabolites (N-acetylglutamine and, nominally, N-acetylglutamate) suggest stress-related oxidative/inflammatory processes within the BGM system. SEM indicated that discrimination relates to both brain and gut signatures, with significant bidirectional brain–gut coupling specifically for unhealthy sweet foods, consistent with the unique rewarding and analgesic properties of sweetness and potential modulation by stress and the opioidergic system. The results imply that discrimination-related stress may bias food valuation and weaken self-regulatory control, while concurrently altering gut metabolites linked to glutamatergic signaling and inflammation, thereby fostering unhealthy eating patterns that increase obesity risk. These BGM disruptions may be further exacerbated by stress-induced dietary changes (high-fat/high-sugar), inflammation, and altered glutamatergic neurotransmission impacting fronto-limbic circuits and decision-making.

Using a systems-biology approach integrating fMRI and fecal metabolomics, the study shows that higher discrimination exposure is associated with increased food-cue reactivity in reward and executive-control regions, greater preference for unhealthy foods, and altered gut glutamate-pathway metabolites linked to stress and inflammation. Brain–gut coupling was most evident for unhealthy sweet foods. These alterations may heighten vulnerability to obesity and related comorbidities among individuals experiencing discrimination. Potential interventions include brain-targeted neuromodulation to dampen overactive food-reward circuits or enhance frontal control, and targeting glutamatergic/inflammatory pathways via probiotics or adherence to an anti-inflammatory Mediterranean diet. Future work should employ longitudinal designs to test causality and explore moderators such as sex, race/ethnicity, and sources of discrimination.

The study is correlational; causal inferences cannot be made, underscoring the need for longitudinal research. Men were under-represented, and prior work suggests possible sex-specific gut–CNS effects. The sample did not include sufficient numbers within racial/ethnic subgroups to conduct stratified analyses; thus, potential moderating effects of sex, race/ethnicity, and discrimination source remain to be tested. The fecal metabolomics analyses were conducted in a subsample (n=62). Although key confounders (BMI, age, sex, race, diet, SES) were controlled, unmeasured confounding cannot be excluded.