Peripheral inflammation is associated with micro-structural and functional connectivity changes in depression-related brain networks

The study addresses how peripheral inflammation relates to alterations in brain micro-structure and functional connectivity within depression-related networks. Prior work shows depression co-associates with elevated CRP and pro-inflammatory cytokines; systemic inflammatory diseases show increased depressive symptoms; and experimental, clinical, and epidemiological studies suggest inflammation can precede and potentially cause depression. Neuroimaging studies have demonstrated inflammation-related changes in task-evoked activity and resting connectivity, particularly in mPFC/ACC, insula, hippocampus, amygdala, and striatum. However, fMRI lacks cellular specificity, and the relationship between inflammation-induced micro-structural tissue changes and distributed functional connectivity remains unclear. This study uses CRP to index low-grade systemic inflammation and tests three hypotheses in a combined micro-structural and resting-state fMRI framework: (i) peripheral inflammation relates to differences in quantitative micro-structural MRI parameters (especially proton density, PD); (ii) peripheral inflammation associates with differences in cortico-subcortical functional connectivity; and (iii) these inflammation-related micro-structural and connectivity differences co-localize anatomically and overlap with depression-related functional network alterations.

Experimental inflammatory challenges (e.g., typhoid vaccination, LPS, interferon-alpha) and observational studies link increased peripheral inflammatory markers (CRP, IL-6) with changes in task-related brain activity in dACC, sgACC/mPFC, insula, hippocampus, amygdala, and striatum. Resting-state studies report inflammation-related alterations in functional connectivity, including sgACC/mPFC connectivity changes, increases or decreases in cortico-subcortical connectivity under endotoxemia or interferon, and reduced connectivity in emotion regulation networks associated with higher inflammatory signaling. A subset of the present cohort previously showed widespread connectivity reductions with higher CRP using a network-based approach. Micro-structural MRI, particularly quantitative magnetization transfer and PD, has shown sensitivity to inflammation in humans and animal models, with increased PD interpreted as increased extracellular free water (edema) in conditions with neuroinflammation (multiple sclerosis, stroke). These findings motivate integrating micro-structure with connectomics to elucidate mechanisms linking peripheral inflammation to depression-related brain network dysfunction.

Design: Case-control study including depressed cases stratified by CRP and healthy controls. Inclusion: Depressed cases screened positive on SCID-5 for current depression, HAM-D > 13 at enrollment and pre-scan, and negative for bipolar/psychosis. Healthy controls screened negative for past/current depression. Major medical disorders (e.g., diabetes, severe cardiovascular disease), and BMI ≥ 36 kg/m² were exclusionary to avoid confounds. Ethics: Approved by NRES East of England (15/EE/0092); written informed consent; reimbursement up to £325. Sample: 143 enrolled into three a priori groups: HC CRP < 3 mg/L (N=53), depressed CRP < 3 mg/L (loCRP, N=55), depressed CRP > 3 mg/L (hiCRP, N=35). After QC, fMRI-analysable sample: controls N=46; cases N=83 (loCRP N=50; hiCRP N=33). MT scans from one site were excluded due to technical incompatibility, reducing the micro-structural analysis sample. Clinical measures: HAM-D, BDI-II, STAI state/trait, Chalder Fatigue Scale, SHAPS anhedonia, CTQ childhood trauma. CRP measured from venous blood. MRI acquisition and processing: - Micro-structural qMT: Magnetization transfer–weighted spoiled gradient echo (voxel 2.4×2.4×2.5 mm). Estimated per voxel: proton density (PD), bound proton fraction (f), MT exchange rate (kbf), and T2 of bound/free water. To account for scanner sensitivity, regional PD values were divided by subject mean PD (global normalization to unity). Parcellation into 360 cortical areas (Glasser et al. atlas) and 8 bilateral subcortical structures (FreeSurfer: thalamus, caudate, putamen, pallidum, hippocampus, amygdala, accumbens, ventral diencephalon) yielding 376 regional values per parameter. - Resting-state fMRI: Multi-echo EPI, TR=2.57 s, 10m42.5s (250 volumes), voxel 3.75×3.75×3.99 mm (slight site variation). First 6 volumes discarded. Preprocessing with ME-ICA to retain BOLD-like components and remove non-BOLD (e.g., motion). Bandpass filtering using MODWT to 0.01–0.1 Hz. Motion QC: exclusion if mean RMS FD > 0.3 mm or max FD > 1.3 mm; one scan excluded for high global correlation r>0.7. Regional mean time series extracted for the same 376 regions. Connectivity estimation: Pearson correlations between all regional pairs, producing a 376×376 symmetric functional connectivity matrix (70,500 unique edges). Weighted degree (nodal hubness) computed as the row/column mean correlation for each node. Statistical analysis: Hierarchical approach. (1) Global comparisons using Kolmogorov–Smirnov tests for distributions of regional micro-structural measures and edge-wise connectivity. (2) Regional (nodal) analyses: linear regressions of CRP with micro-structural measures or weighted degree per region. (3) Edge-wise analyses: linear regressions of CRP with each of the 70,500 connections, and specifically with each subcortical-to-cortical edge (375 per subcortical seed). Multiple comparisons controlled using false discovery rate (PFDR < 0.05). Sensitivity analyses included covariates (age, site in primary; additionally sex, BMI, CTQ, antidepressant variables), alternative preprocessing with global signal regression, and case-only analyses.

Sample characteristics: Cases and controls were matched on age and sex. Cases had higher depression, anxiety, fatigue, anhedonia, and childhood adversity scores. HiCRP cases had higher BMI and were predominantly female versus loCRP cases; symptom severity did not differ between hiCRP and loCRP cases. Micro-structure (PD): Global PD distribution was significantly right-shifted in hiCRP versus loCRP cases (KS, P < 6.4×10^−7) and differed from controls. No global group differences were observed for other qMT parameters (kbf, fb, T1, T2) after correction. No significant case-control regional PD differences were detected. Across all participants, PD correlated with CRP in 22 regions positively and 7 regions negatively (PFDR < 0.05). Positive PD–CRP associations were concentrated in DMN-related regions, notably posterior cingulate/precuneus (RSC, PCV, 7m, POS1, v23ab, d23ab, 31pv, 31pd, 31a, ProS), medial/orbital prefrontal cortex (preMOr, area 25, IFJa, IFSa, a10p, p10p, p47r, OFC), and additional DLPFC/premotor/paracentral/ventral visual regions (9a, 6r, 5m, VMV1). Negative PD–CRP associations occurred in several prefrontal/premotor regions. Functional connectivity (FC): Edge-wise FC distributions were left-shifted in cases versus controls (KS, P < 2.2×10^−16), indicating more negative or less positive correlations; the shift was most evident in hiCRP cases. Weighted degree distributions were also significantly left-shifted in cases (KS, P < 2.2×10^−16) and in hiCRP versus loCRP cases (KS, P < 5×10^−5). At the nodal level, 39 regions showed significantly reduced weighted degree in depression (PFDR < 0.05), including pC/pCC, inferior parietal cortex, mPFC, and hippocampus; 27 (69%) of these were DMN-affiliated. Mean degree within the DMN was significantly reduced in cases; no significant differences were found for other canonical networks. CRP–FC associations: In depressed cases, three connections scaled positively with CRP (PFDR < 0.05): (1) PCC/precuneus v23ab–RSC, (2) POS1–hippocampus, and (3) hippocampus–mPFC (area 10r). One connection scaled negatively with CRP: area 7Pm–frontal operculum FOP1. At a more lenient threshold (PFDR < 0.1), two additional negative PCC–dACC edges (d23ab–a32pr, 31pv–a32pr) emerged. Seed-based analyses confirmed higher CRP associated with increased hippocampal connectivity to pC/pCC and mPFC. Several cortical regions with significant FC–CRP associations also showed significant PD–CRP associations (e.g., POS1, v23ab), indicating anatomical co-localization. Mediation: For POS1–hippocampus, CRP had significant direct effects on POS1 PD and on connectivity, with no significant indirect effect via PD. For d23ab/31pv–mPFC (a32pr), CRP increased PD in PCC regions and connectivity scaled negatively with CRP; mediation analysis indicated that CRP’s association with medial prefrontal connectivity was mediated by its effect on PCC PD.

Findings support that peripheral inflammation relates to local micro-structural alterations and distributed functional connectivity changes in brain networks implicated in depression. PD—interpreted as reflecting tissue free water—scaled with CRP predominantly in posterior DMN (pC/pCC) and mPFC, consistent with localized extracellular edema in low-grade systemic inflammation. Functional connectivity showed depression-related reductions in nodal hubness concentrated in the DMN, consistent with a broader literature on DMN dysconnectivity in major depression. CRP-related alterations in connectivity were observed among DMN nodes and hippocampus/mPFC, and in some cases co-localized with PD–CRP effects. Mediation analyses provided preliminary evidence that inflammation-related micro-structural changes in PCC could mediate effects on specific functional connections to mPFC. Mechanistically, peripheral immune signaling (indexed by CRP as a proxy) may influence posterior DMN tissue properties, perturbing integration within the DMN and between DMN and task-positive networks, contributing to affective and cognitive symptoms characteristic of depression.

This study integrates quantitative micro-structural MRI and resting-state functional connectomics to show that peripheral inflammation (indexed by CRP) is associated with increased proton density in posterior cingulate/precuneus and mPFC and with altered functional connectivity among DMN nodes and hippocampus/mPFC. Depression is characterized by reduced hubness of DMN regions, overlapping anatomically with CRP-associated micro-structural and connectivity changes. These findings suggest that inflammation-induced alterations in DMN micro-structure and connectivity may mediate depressive pathophysiology. Future research should validate the biophysical basis of PD changes as edema, employ longitudinal and interventional designs to establish causality, include larger and more diverse samples (including non-depressed individuals across CRP ranges and patients with comorbidities/obesity), and integrate refined phenotyping of immune profiles and psychopathology to link connectivity changes with specific depressive dimensions.

Cross-sectional design precludes causal inference. Sample size, particularly for subgroup and micro-structural analyses (one site’s MT data excluded), may limit power to detect small effects, especially in controls. Enrichment of high-CRP individuals among cases could confound depression versus inflammation effects; however, case-only sensitivity analyses supported key findings. Generalizability may be limited by exclusion of serious medical comorbidities and high BMI (≥36 kg/m²). Primary analyses controlled for age and site; although key results were robust to additional covariates (sex, BMI, childhood adversity, antidepressant exposure) and alternative preprocessing (global signal regression), residual confounding cannot be excluded. Motion was rigorously addressed via ME-ICA and QC, yet motion-related biases in connectivity are a known concern. No significant associations were observed between CRP and symptom dimension scores (depression, anxiety, anhedonia, fatigue) within cases, and edges associated with CRP did not correlate with these measures after correction.