Background:
People with HIV (PWH) can suffer from neuropsychiatric conditions that have complex etiologies. We investigated the plasma proteome in 206 virally suppressed PWH on antiretroviral therapy (ART) and compared the results to multiple dimensions of a comprehensive neuropsychiatric assessment.
Methods:
Neuropsychiatric dimensions included global and domain-specific neurocognitive impairment (NCI), Beck Depression Inventory (BDI)-II and subscales, and instrumental activities of daily living (IADL) dependence. Based on these and other data, 4 biopsychosocial phenotypes (BPSPs) were previously derived by machine learning analysis: healthy (BPSP1), mild NCI with moderate depressive symptoms (BPSP2), mild-to-moderate NCI with severe depressive symptoms and IADL dependence (BPSP3), and mild-to-moderate NCI without depressive symptoms or IADL dependence (BPSP4). Proteins were quantified by the Olink Target-96 Inflammation panel. Linear regression of normalized protein expression values analyzed group differences. Benjamini-Hochberg adjusted alpha was 0.003. Proteins with p<0.10 were next analyzed by multivariable regression adjusting for demographic, disease, and treatment characteristics. Classification accuracy was assessed by discriminant analysis.
Results:
Cohort characteristics: mean age 45.0 years, 22.8% women, 51.5% Black, median CD4+ T-cells 480/µL, median duration: HIV 10.6 years, ART 5.7 years. An example volcano plot is shown in the Figure A. Across all dimensions, the most frequently associated proteins were CD8 antigen, IL-12B, IL-17C, FGF-21, FGF-23, and TGF-α. Adjusted multivariable models with R2>0.15 were found for global NCI, motor impairment, BDI-II somatic subscale, and all BPSPs (Table B). Among all dimensions, discriminant analysis correctly classified >75% of participants in only BPSP3 and BPSP4. The BPSP3 model included CCL19 and DNER and the BPSP4 model included CD8 antigen, IL-33, CCL3, TNFSF11, and stem cell factor.
Conclusions:
Data-driven, cross-sectional analyses identified plasma proteins associated with multiple neuropsychiatric dimensions. While the results reinforce the complexity of neuropsychiatric disorders, novel, biologically plausible targets were identified for validation and future investigation. The BPSPs generally performed better in analyses, suggesting that they may be more inflammation-driven than other dimensions. The focus on inflammation is a limitation; future analyses that include other proteins (e.g., neuronal) should provide additional insights.
![]()