High-throughput proteomic analysis reveals systemic dysregulation in virally suppressed people living with HIV

Abstract

BACKGROUNDPeople living with HIV (PLHIV) receiving antiretroviral therapy (ART) exhibit persistent immune dysregulation and microbial dysbiosis, leading to development of cardiovascular diseases (CVDs). We initially compared plasma proteomic profiles between 205 PLHIV and 120 healthy control participants (HCs) and validated the results in an independent cohort of 639 PLHIV and 99 HCs. Differentially expressed proteins (DEPs) were then associated to microbiome data. Finally, we assessed which proteins were linked with CVD development in PLHIV.METHODSProximity extension assay technology was used to measure 1,472 plasma proteins. Markers of systemic inflammation (C-reactive protein, D-dimer, IL-6, soluble CD14, and soluble CD163) and microbial translocation (IFABP) were measured by ELISA, and gut bacterial species were identified using shotgun metagenomic sequencing. Baseline CVD data were available for all PLHIV, and 205 PLHIV were recorded for development of CVD during a 5-year follow-up.RESULTSPLHIV receiving ART had systemic dysregulation of protein concentrations, compared with HCs. Most of the DEPs originated from the intestine and lymphoid tissues and were enriched in immune- and lipid metabolism-related pathways. DEPs originating from the intestine were associated with specific gut bacterial species. Finally, we identified upregulated proteins in PLHIV (GDF15, PLAUR, RELT, NEFL, COL6A3, and EDA2R), unlike most markers of systemic inflammation, associated with the presence and risk of developing CVD during 5-year follow-up.CONCLUSIONOur findings suggest a systemic dysregulation of protein concentrations in PLHIV; some proteins were associated with CVD development. Most DEPs originated from the gut and were related to specific gut bacterial species.TRIAL REGISTRATIONClinicalTrials.gov NCT03994835.FUNDINGAIDS-fonds (P-29001), ViiV healthcare grant (A18-1052), Spinoza Prize (NWO SPI94-212), European Research Council (ERC) Advanced grant (grant 833247), and Indonesia Endowment Fund for Education.

Keywords: AIDS/HIV; Bioinformatics; Cardiovascular disease; Innate immunity.

Figure 1. Dysregulation of protein concentrations in PLHIV compared with HCs.

(A) Study overview. (B) PCA of protein levels from the discovery (left: n = 205 PLHIV vs. 120 HCs) and replication cohorts (right: n = 639 PLHIV vs. 99 HCs) using the first 2 principal components. The ellipses were centered on the basis of the median of PC1 and PC2 for each group (PLHIV and HCs) and the median differences between groups were assessed by the Mann-Whitney U test. ***P < 0.0001. (C) Four-quadrant plot of the fold-change of sDEPs (n = 276) in the discovery (x axis; n = 205 PLHIV vs. 102 HCs) and replication cohorts (y axis; n = 639 PLHIV vs. 84 HCs). DE analysis was performed using a linear regression model with age, sex, and smoking status as covariates. Fold-change in the x axis label refers to the difference in the mean of log₂ NPX values between PLHIV and HCs. Only proteins with FDR < 0.05 and log₂ fold-change ≥ 1.5 are annotated. See also Supplemental Table 4. Var, variation.

Figure 2. Origin and functional characterization of DEPs between PLHIV and HCs.

(A) Bar chart showing significantly enriched tissue- and cell-specific proteins using the upregulated sDEPs from the discovery and replication cohorts. Enrichment analysis was performed using a tissue- and a cell-specific gene list according to the consensus transcriptomic data from the HPA and GTEx. See also Supplemental Tables 5 and 6. (B) Network of functional pathways based on enrichment analysis of sDEPs. Circular nodes represent pathways, with the colors showing a gradient of the enrichment P values; node size represents the number of genes in the pathway; weighted edges represent the degree of gene overlap score between pathways, calculated by the average between the jacquard and overlap coefficients. Clusters of nodes were identified by Markov cluster algorithm (see Methods). See also Supplemental Table 8.

Figure 3. Gut bacterial species were related to sDEPs originated from the intestine and lymphoid tissues in PLHIV.

(A) Bar chart showing the proportion of significant associations (P < 0.05) among 4 different groups of proteins and bacterial species using linear regression analysis corrected for age, sex, smoking status, and read counts. Differences in the proportion of significant associations between bacterial species and protein groups were tested using post hoc pairwise χ² tests. ***P < 0.0001. (B) Heatmap depicting the associations between sDEPs specific to the intestine and lymphoid tissues with the abundance of enriched and depleted bacterial species in PLHIV of the discovery cohort (n = 143). The analysis was performed using linear regression analysis corrected for age, sex, smoking status, and read counts. Only significant associations (P < 0.05) are shown. See also Supplemental Table 9.

Figure 4. Plasma proteomic signatures of CVD in PLHIV.

(A) Circular heatmap of associations of plasma proteins and CVD in PLHIV from the discovery (n = 205) and replication (n = 639) cohorts. sDEPs are highlighted in red. The associations were explored using linear regression corrected for age, sex, BMI, and smoking status. Only significant associations in the discovery (P < 0.05) and/or replication cohort (FDR < 0.05) are shown. (B) Forest plot showing positive associations between proteins and CVD events in PLHIV in the discovery cohort (P < 0.05; n = 205). P values were calculated using a binomial logistic regression model adjusted for age, sex, BMI, smoking status, and the presence of dyslipidemia, hypertension, and type 2 diabetes at baseline. The relative effects of protein concentration are presented as ORs with 95% CIs. See also Supplemental Table 10. (C) Venn diagram showing the number of shared proteins identified in 3 groups: (i) sDEPs, (ii) proteins associated with the presence of CVD at baseline, and (iii) risk of developing CVD events during 5-year follow-up.