Compartmentalized biomarker correlations in HIV: plasma-CSF relationships vary by blood-brain/blood-CSF barrier permeability and viral suppression.

Abstract

Neuroinflammation is common in people with HIV (PWH) and may be reflected also in plasma biomarkers; the latter are sometimes used as surrogates for CSF. However, use of plasma biomarkers in this way risks obscuring compartment-specific processes since distribution across the blood-brain barrier (BBB) varies between proteins with some reaching the CNS more readily than others. We tested the hypothesis that BBB and viral suppression status shape cross-compartment biomarker coupling, clarifying when plasma proteins do or do not represent neuroinflammation. Paired CSF and plasma samples from 567 PWH in the CHARTER cohort were analyzed with the Olink Target-96 Inflammation Panel. Using regression, canonical correlation, and machine learning, we evaluated viral suppression and BBB permeability (indexed by CSF total protein levels, which are more readily available clinically and highly correlated with the CSF to serum albumin ratio) as effect modifiers of CSF-plasma biomarker correlations. Intra-CSF and intra-plasma correlations were consistently strong, but cross-compartment correlations were weak and inconsistent. Some proteins had strong correlations when CSF total protein was high, including CD8A, IL-12B, TNFRSF9, and TNFB. Unsuppressed viremia amplified broader cross-compartment signaling (e.g., IL-12B, CXCL9, CXCL10). BBB permeability and viral suppression moderate biomarker compartmentalization in PWH. These findings support a hypothesis-driven framework in which biomarkers can be classified as CNS-restricted, peripherally driven, or BBB-dependent. This mechanistic structure informs biomarker selection for clinical trials and provides testable models of neuroinflammation in PWH.