Plasma metabolomics identifies lipid abnormalities linked to markers of inflammation, microbial translocation, and hepatic function in HIV patients receiving protease inhibitors

Abstract

Background: Metabolic abnormalities are common in HIV-infected individuals on antiretroviral therapy (ART), but the biochemical details and underlying mechanisms of these disorders have not been defined.

Methods: Untargeted metabolomic profiling of plasma was performed for 32 HIV patients with low nadir CD4 counts (<300 cells/ul) on protease inhibitor (PI)-based ART and 20 healthy controls using liquid or gas chromatography and mass spectrometry. Effects of Hepatitis C (HCV) co-infection and relationships between altered lipid metabolites and markers of inflammation, microbial translocation, and hepatic function were examined. Unsupervised hierarchical clustering, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), Random forest, pathway mapping, and metabolite set enrichment analysis (MSEA) were performed using dChip, Metaboanalyst, and MSEA software.

Results: A 35-metabolite signature mapping to lipid, amino acid, and nucleotide metabolism distinguished HIV patients with advanced disease on PI-based ART from controls regardless of HCV serostatus (p<0.05, false discovery rate (FDR)<0.1). Many altered lipids, including bile acids, sulfated steroids, polyunsaturated fatty acids, and eicosanoids, were ligands of nuclear receptors that regulate metabolism and inflammation. Distinct clusters of altered lipids correlated with markers of inflammation (interferon-α and interleukin-6), microbial translocation (lipopolysaccharide (LPS) and LPS-binding protein), and hepatic function (bilirubin) (p<0.05). Lipid alterations showed substantial overlap with those reported in non-alcoholic fatty liver disease (NALFD). Increased bile acids were associated with noninvasive markers of hepatic fibrosis (FIB-4, APRI, and YKL-40) and correlated with acylcarnitines, a marker of mitochondrial dysfunction.

Conclusions: Lipid alterations in HIV patients receiving PI-based ART are linked to markers of inflammation, microbial translocation, and hepatic function, suggesting that therapeutic strategies attenuating dysregulated innate immune activation and hepatic dysfunction may be beneficial for prevention and treatment of metabolic disorders in HIV patients.

Figure 1

Metabolomic profiling of plasma from two cohorts identifies metabolites that distinguish HIV subjects from controls. (A) Schematic of strategy used to identify plasma metabolites altered in initial and validation cohorts. LC, liquid chromatography; GC, gas chromatography; MS, mass spectrometry; ESI, electrospray ionization; EI, electron ionization. (B) Unsupervised hierarchical clustering of signature metabolites (n=35) that distinguish HIV vs. control subjects in initial and validation cohorts (left and right heatmaps, respectively) regardless of HCV serostatus (FC>1.3, p<0.05, FDR <10%). Red and blue indicate increased and decreased levels, respectively. (C) Top VIP scores with expression heatmap from PLS-DA models. PLS-DA models were constructed with signature metabolites (n=35) from the initial and validation cohorts. Red and green indicate increased and decreased levels, respectively.

Figure 2

Metabolic pathways altered in HIV⁺HCV^- and HIV⁺HCV⁺ subjects on PI-based ART. (A) PCA analysis shows separation of control (blue), HIV⁺HCV^- (green), and HIV⁺HCV⁺ (red) metabolomes (n=227 metabolites) in initial and validation cohorts. (B) Venn diagram showing the distribution of metabolites altered in HIV⁺HCV⁺ subjects in initial and validation cohorts (n=46; FC>1.3, p<0.05, FDR<10%). (C) Box plots of metabolite classes altered in HIV⁺HCV^- (n=7) or HIV⁺HCV⁺ (n=10) subjects compared to controls (n=8) matched for age, gender, and race/ethnicity. Medians are represented by horizontal bars, boxes span the interquartile range (IQR) and whiskers extend to extreme data points within 1.5 times IQR. Outliers plotted as open circles lie outside 1.5 times the IQR. Grey, orange, and red box plots represent controls, HIV⁺HCV^- and HIV⁺HCV⁺ subjects, respectively. P-values were calculated using Welch’s t-tests. K:T ratio; kynurenine to tryptophan ratio.

Figure 3

Metabolites altered in HIV subjects compared to healthy controls map to multiple biosynthetic pathways. Altered metabolites with KEGG ID from the merged dataset (n=32 HIV and n=20 control subjects; FC>1.3, p<0.05, FDR<0.1) were mapped to KEGG and SMPDB reference pathways and interaction networks were generated in Cytoscape. Green and red nodes represent metabolites with increased and decreased levels, respectively. White nodes represent pathways. Asterisks indicate related metabolites detected in only one cohort.

Figure 4

Lipid pathways altered in HIV subjects on PI-based ART. (A) Quantitative Enrichment Analysis (QEA) performed using MSEA. QEA is based on the globaltest algorithm which uses a generalized linear model to estimate a Q-statistic for each metabolite set. Lipid metabolites (n=113) from the merged dataset consisting of all HIV subjects (n=32) and controls (n=20) were inputted into MSEA and enrichment was assessed using the MSEA Metabolic Pathway library (n=88) and custom metabolite sets derived from Lipid Maps (n=5). Pathways were considered enriched when p<0.05 and FDR<5%. (B) Box plots of major lipid classes altered in HIV subjects (n=32) compared to controls (n=20). Medians are represented by horizontal bars, boxes span the interquartile range (IQR) and whiskers extend to extreme data points within 1.5 times IQR. Outliers plotted as open circles lie outside 1.5 times the IQR. Blue and red represent controls and HIV subjects, respectively. P-values were calculated using Welch’s t-tests.

Figure 5

Correlation matrix reveals relationships between clusters of lipids and markers of inflammation and hepatic function. The Pearson correlation matrix was constructed in R using the heatmap.2 function. Significant correlations had a −0.35> r >0.35, p<0.05, and FDR<10% (Additional file 5). Red and blue indicate positive and negative correlations, respectively (see Color Key).

Figure 6

The HIV plasma lipidome has overlapping yet distinctive features compared to alterations in NAFLD/NASH. Venn diagram depicting overlap between metabolites altered in the same direction in HIV subjects on PI-based ART and HIV-negative subjects with NAFLD or NASH (n=45, Additional file 6). NAFLD and NASH data sets were based on 3 published studies [27-29].

Figure 7

Increased plasma bile acids are associated with intermediate/high FIB-4 and APRI scores and high YKL-40. Bar graphs showing bile acid levels (scaled intensity) in healthy controls (white), and HIV subjects with low (pink) vs. intermediate/high FIB-4 or APRI scores or high YKL-40 levels (red). P-values were calculated using Welch’s t-tests.