The mRNA content of plasma extracellular vesicles provides a window into molecular processes in the brain during cerebral malaria

Abstract

The impact of cerebral malaria on the transcriptional profiles of cerebral tissues is difficult to study using noninvasive approaches. We isolated plasma extracellular vesicles (EVs) from patients with cerebral malaria and community controls and sequenced their mRNA content. Deconvolution analysis revealed that EVs from cerebral malaria are enriched in transcripts of brain origin. We ordered the patients with cerebral malaria based on their EV-transcriptional profiles from cross-sectionally collected samples and inferred disease trajectory while using healthy community controls as a starting point. We found that neuronal transcripts in plasma EVs decreased with disease trajectory, whereas transcripts from glial, endothelial, and immune cells increased. Disease trajectory correlated positively with severity indicators like death and was associated with increased VEGFA-VEGFR and glutamatergic signaling, as well as platelet and neutrophil activation. These data suggest that brain tissue responses in cerebral malaria can be studied noninvasively using EVs circulating in peripheral blood.

Fig. 1.. The solid tissue atlas of circulating EV-RNA in cerebral malaria.

(A) The relative comparison of blood and solid tissue RNA fractions in plasma EVs from patients with cerebral malaria. (B) The relative distributions of solid tissue fractions of circulating EV-RNA in cerebral malaria. (C) The estimated absolute proportion of RNA expressed by the brain and nerves is higher in retinopathy-positive (CM-R⁺) and negative (CM-R⁻) cerebral malaria compared to community controls (CC). (D) Brain cell relative fractions estimated from the plasma EV-RNA data. (E) Heatmap clustering of differentially enriched genes between patients with CM-R⁺ and CM-R⁻. (F) Top terms associated with genes significantly increased (red) and decreased (blue) in CM-R⁺. The top terms in the blood and brain domains are shown.

Fig. 2.. Manifold learning infers disease trajectory from plasma-EV transcriptomes.

(A and B) Schematic representation of two models of disease progression in cerebral malaria; (A) CM-R⁻ is either a disease variant distinct from CM-R⁺ or (B) CM-R⁻ precedes CM-R⁺. (C) Scatterplot showing EV-RNA samples colored by disease trajectory. The scatter plot shows cerebral malaria evolves in a single lineage. (D) EV-RNA samples colored by retinopathy status, depicting that late-stage trajectory is enriched for the CM-R⁺ sample set. (E) Boxplots comparing disease trajectory between CC, CM-R⁻, and CM-R⁺. The inferred disease trajectory is significantly more advanced in CM-R⁺ than in CM-R⁻. (F) Receiver operating characteristic (ROC) curve shows that retinopathy is 100% sensitive and 78% specific that cerebral malaria has progressed to late stages of the trajectory. (G and H) Forest plots showing linear regression results (95% bootstrap estimates) comparing retinopathy and clinical parameters, (G) unadjusted and (H) adjusted for age. Red shows positive correlations, black nonsignificant correlations, and blue negative correlations. The estimated molecular trajectory is concordant with known clinical parameters.

Fig. 3.. Brain-derived plasma EV-RNA varies over disease trajectory.

(A) Phaseogram showing the variation of plasma EV-RNA as a function of disease trajectory. The transcripts are clustered along the disease trajectory, revealing four clusters, c1 to c4. (B) Enrichment analysis using PanglaoDB cell markers (31) shows that early trajectory clusters (c1 and c2) are enriched for neuronal markers, while late-trajectory clusters are enriched for glial (oligodendrocyte, astrocyte, and microglia) and immune cells. (C) Enrichment analysis using Darmanis brain cell markers (32) also shows that cerebral malaria is characterized by decreased and increased neuronal and glial gene expression respectively. (D) Representative EV-RNA profiles of neuronal and glial cell markers. (E) KEGG enrichment analysis results showing that cluster 3 is enriched for transcripts belonging to vascular processes and synaptic-related neural functions, while cluster 4 is enriched for transcripts linked to pathways of neurodegeneration.

Fig. 4.. Disease trajectory stratifies patients with CM-R⁻ into three subgroups.

(A) Gaussian mixture modeling (GMM) shows that the patients with CM-R⁻ cluster optimally into three subgroups: early trajectory (yellow), mid-trajectory (blue), and late trajectory (red). (B) Patients in the early CM-R⁻ subgroup tended to be older and less anemic than those in the mid- and late-trajectory subgroups. (C) Supervised heatmap clustering showing differentially altered genes between the CM-R⁻ subgroups. (D) Top terms from the domains of blood and brain that are enriched in at least one CM-R⁻ subgroup.