Biomarkers of central and peripheral inflammation mediate the association between HIV and depressive symptoms

Depression is highly prevalent among people living with HIV (PLWH), approximately twice that of the general population. Emerging work suggests depression comprises distinct subtypes, including an inflammation-associated subtype characterized by sickness behaviours and elevated pro-inflammatory cytokines. HIV persists in the CNS even under suppressive antiretroviral therapy and is associated with chronic systemic and neuroinflammation. The study aimed to determine whether biomarkers of peripheral and central inflammation mediate the association between HIV status and depressive symptoms. The primary hypothesis was that inflammatory biomarkers in plasma and CSF, and neurochemical measures from MRS, would at least partly mediate the increased risk of depressive symptoms observed in PLWH.

Prior literature indicates: (1) Depression is associated with elevated peripheral and central inflammatory markers, including cytokines and chemokines; anti-inflammatory treatments can alleviate depressive symptoms in some cases, suggesting a causal role for inflammation in a subset of depression. (2) HIV infection leads to neuroinflammatory responses, including microglial activation, and chronic inflammation persists despite suppressive ART. (3) IL-6 has been repeatedly associated with depression in longitudinal studies and meta-analyses; other chemokines such as MIP-1α and MIG have been less well studied in depression, particularly within HIV. This background supports investigating inflammatory biomarkers as mediators between HIV and depressive symptoms.

Design and cohort: Cross-sectional analysis of baseline data from the prospective COBRA cohort (recruited 2013–2014 in London and Amsterdam). Participants: adults ≥45 years; PLWH were virally suppressed and on ART ≥12 months; demographically similar HIV-negative controls were included. Ethics approvals obtained in the UK and Netherlands. Measures: Depressive symptoms via PHQ-9 (0–27); at cohort screening, individuals with PHQ-9 ≥10 were excluded from enrollment. For the current analysis, depressive symptoms were summarised as a dichotomous variable ('Any Depressive Symptoms' using PHQ-9 >4 as per abstract and table indications) and as a continuous score for sensitivity analyses. Biomarkers: Plasma—high-sensitivity CRP (immunobinding), hFABP, sCD14, sCD163, neopterin, sCD16 (ELISA), NFL (Simoa), tryptophan and kynurenine (HPLC) with kynurenine:tryptophan ratio; in a subset of 78, multiplex assay for IL-6, TNF-α, MIG/CXCL9, IP-10/CXCL10, MCP-1/CCL2, IL-8/CXCL8, RANTES/CCL5. CSF—sCD163, sCD16, neopterin, NFL (ELISA), IL-6 and TNF-α (immunobinding), and IL-8, MCP-1, RANTES, sCD14 (multiplex). Neuroimaging: Single-voxel 1H-MRS at two sites (Siemens 3T Verio; Philips 1.5T Intera and 3T Ingenia); regions: frontal white matter (FWM) and putamen/prefrontal cortex; neurochemicals included myo-inositol (mI) and choline (Cho) as markers of glial activation/cellular metabolism. Statistical analysis: Descriptive statistics with appropriate univariate tests by HIV status. Handling of values below detection involved imputing half the lower limit; certain values set to enable log-transformations per prior COBRA assumptions. Spearman correlations computed among biomarkers in plasma and CSF (full sample and subgroups). Primary analysis: logistic regression estimating odds of Any Depressive Symptoms by HIV status (reference = HIV-negative), adjusted for age, sex, ethnicity, and years of education; models including MRS measures additionally adjusted for MRI scanner. Mediation screening: sequentially added each (log-transformed) biomarker to the adjusted model; a >10% reduction in the HIV odds ratio was considered evidence of potential mediation. Sensitivity analysis: linear regression of PHQ-9 total score as continuous outcome with similar adjustments. Listwise deletion applied; analyses in R 4.1.0; no multiple testing correction.

- Sample: N = 204 (PLWH n = 125; HIV-negative controls n = 79); median age 57 years; 92.6% male; 91.7% White; 82.8% men who have sex with men; all PLWH on cART with HIV-RNA <200 copies/mL. PLWH were less likely to be White; alcohol use differed between groups. - Depressive symptoms: Any Depressive Symptoms prevalence 26.4% in PLWH vs 11.4% in controls (p = 0.02). Logistic regression (adjusted for sociodemographics) showed higher odds of Any Depressive Symptoms in PLWH: OR 3.72 (95% CI 1.76, 8.09). - Biomarker differences: Multiple plasma and CSF inflammatory biomarkers (e.g., sCD14, sCD16, sCD163, TNF-α, neopterin, Kyn:Trp) were higher among PLWH (p < 0.05; details in supplementary files). - Mediation screening (criterion >10% OR reduction): • Plasma: MIG/CXCL9 reduced the HIV–depression OR by 15.0%; TNF-α reduced it by 11.4%. • CSF: MIP-1α/CCL3 reduced the OR by 21.0%; IL-6 reduced it by 18.0%. • No other plasma, CSF, or MRS biomarkers met the mediation criterion. - Neuroimaging: MRS neurometabolites (mI and Cho in FWM/putamen) did not mediate the HIV–depression association. - Correlations: Moderate correlations observed between IL-6 in plasma and CSF and with clusters of chemokines; correlation matrices provided for full sample and subgroups. - Sensitivity analyses using PHQ-9 as continuous outcome were consistent with primary findings.

The study demonstrates that PLWH have higher prevalence and severity of depressive symptoms and elevated inflammatory biomarkers compared to HIV-negative controls. Importantly, specific inflammatory markers—plasma MIG and TNF-α, and CSF MIP-1α and IL-6—attenuated the HIV–depressive symptoms association, supporting a mediating role of both peripheral and central inflammation. These findings align with literature implicating IL-6 in depression risk and extend this evidence to virally suppressed PLWH. The notable mediation by CSF MIP-1α suggests a potentially underrecognized chemokine pathway relevant to HIV-associated depression. Moderate cross-compartment correlations (e.g., IL-6 plasma–CSF) and associations with chemokine clusters point to coordinated inflammatory processes spanning the periphery and CNS. Conversely, MRS-derived neurometabolites (mI, Cho) did not mediate the association, indicating that these neuroimaging proxies may not capture the specific neuroinflammatory mechanisms linking HIV to depressive symptoms in this context or may lack sensitivity compared to soluble biomarkers.

Inflammatory processes in both the periphery and CNS appear to partly mediate the increased risk of depressive symptoms among PLWH. Plasma MIG and TNF-α, and CSF MIP-1α and IL-6 emerged as potential mediators, highlighting biomarker pathways that may inform mechanistic understanding and future intervention targets for HIV-associated depression. The absence of mediation by MRS neurometabolites suggests that soluble inflammatory markers may be more informative in this context. Future work should replicate these findings in larger, more diverse cohorts, include participants with greater depressive symptom severity, evaluate additional candidate biomarkers (e.g., IL-1β, BDNF), explore interactions and combined mediator effects, and leverage more sensitive neuroimaging techniques to refine understanding of neuroinflammatory mechanisms.

- Cross-sectional design precludes causal inference; residual confounding cannot be excluded. - Mediation defined by >10% OR reduction is an exploratory threshold and may be considered low. - Biomarker panel constrained by parent study; key markers such as IL-1β and BDNF were not measured. - Detection-limit issues for some markers (e.g., plasma MIP-1α, CSF RANTES, CSF TNF-α) may reduce sensitivity. - Cohort screening excluded PHQ-9 ≥10, limiting range of depressive severity and potentially underestimating associations. - Small numbers with Any Depressive Symptoms, particularly in subgroups, may yield underpowered analyses. - All PLWH were on ART; treatment side effects (e.g., fatigue) may overlap with depressive symptoms. - Possible selection bias due to willingness to undergo lumbar puncture and attend extra visits. - Demographically narrow sample (predominantly White men who have sex with men) limits generalisability to women and non-Western settings. - No interaction terms or multi-mediator/synergy analyses were conducted.