Dysbiosis of the gut microbiota is associated with HIV disease progression and tryptophan catabolism

Abstract

Progressive HIV infection is characterized by dysregulation of the intestinal immune barrier, translocation of immunostimulatory microbial products, and chronic systemic inflammation that is thought to drive progression of disease to AIDS. Elements of this pathologic process persist despite viral suppression during highly active antiretroviral therapy (HAART), and drivers of these phenomena remain poorly understood. Disrupted intestinal immunity can precipitate dysbiosis that induces chronic inflammation in the mucosa and periphery of mice. However, putative microbial drivers of HIV-associated immunopathology versus recovery have not been identified in humans. Using high-resolution bacterial community profiling, we identified a dysbiotic mucosal-adherent community enriched in Proteobacteria and depleted of Bacteroidia members that was associated with markers of mucosal immune disruption, T cell activation, and chronic inflammation in HIV-infected subjects. Furthermore, this dysbiosis was evident among HIV-infected subjects undergoing HAART, and the extent of dysbiosis correlated with activity of the kynurenine pathway of tryptophan catabolism and plasma concentrations of the inflammatory cytokine interleukin-6 (IL-6), two established markers of disease progression. Gut-resident bacteria with capacity to catabolize tryptophan through the kynurenine pathway were found to be enriched in HIV-infected subjects, strongly correlated with kynurenine levels in HIV-infected subjects, and capable of kynurenine production in vitro. These observations demonstrate a link between mucosal-adherent colonic bacteria and immunopathogenesis during progressive HIV infection that is apparent even in the setting of viral suppression during HAART. This link suggests that gut-resident microbial populations may influence intestinal homeostasis during HIV disease.

Figure 1. Gut bacterial microbiota composition in HIV-infected viremic untreated (VU) subjects differs from that of HIV-uninfected risk-matched controls

Community composition dissimilarity was analyzed using microbiota profiles generated using PhyloChip. Non-metric dimensional scaling (NMDS) was used to plot each sample community based on a Canberra community dissimilarity matrix. Each dot represents the microbiome of a single subject, and color indicates subject group as denoted in legend. Ellipses represent 95% confidence intervals for standard error of weighted NMDS score means of respective groups, and are colored based on subject group (red denotes HIV+VU subjects, while blue denotes HIV-). Community differences were verified using Adonis, P = 0.002.

Figure 2. Phylogenetic distribution of the disease-associated microbial community (DMC), defined as significantly depleted or enriched in VU subjects compared with HIV- subjects

Phylogenetic tree representing 16S rRNA gene sequence relatedness of taxa identified as significantly enriched or depleted in relative abundance between VU and HIV- subjects (see Materials & Methods for statistical analyses). The 625 taxa identified belonged to 172 unique genera, and representative taxa for each unique genus are depicted. Tree branches are colored-coded by phyla as indicated in legend. Red and blue bars depict bacterial genera in greater (red bars) and lesser (blue bars) abundance in VU subjects. Dashed lines indicate magnitude of relative abundance change (inner dashed circle denotes 500 MFI, outer denotes 2000 MFI difference between comparative groups; 1,000 MFI change represents a log₂ difference in relative abundance).

Figure 3. Bacterial community enriched in untreated infection associates with immunopathologic markers of HIV disease progression within HIV-infected subjects

(A) Spearman correlations between individual taxa of the DMC and immunologic parameters were calculated among all HIV-infected subjects (treated and untreated), collated, and represented by color. Correlation coefficients for taxa with representative species belonging to the same genus and displaying similar FI trends were averaged for visualization purposes. Intersections of each genus within the DMC (rows) and immunologic parameters (columns) represent the strength of the positive (red) or negative (blue) correlation as determined by Spearman correlation coefficient (R_S). Bacteria depleted in VU subjects correlate positively with protective immune markers (low T cell activation in the blood and gut, higher gut IL-17 secretion), while bacteria enriched in VU subjects correlate with markers of disease progression (converse of above, as well as lower T cell-mediated IL-22 secretion in the gut, higher Kyn:Trp, and higher concentrations of plasma markers of inflammation [IP-10, IL-6, and soluble TNF receptor II]). Families and genera of taxa that exhibited numerous strong correlations with immunologic parameters are indicated. Intersections enclosed by a black box represent associations with unadjusted P < 0.02. (B) First principal component (PC1) representing trends in DMC bacterial community composition accounts for 52.8% of bacterial community variance, and separates VU subjects from HIV-, while subjects undergoing HAART have PC1 values distributed amongst those for VU and HIV- subjects (PC1, Principal Component 1; LTNP, Long Term Non-Progressor). (C) First principal component correlates with markers of disease progression: plasma concentration of inflammatory marker IP-10, IDO activity (Kyn:Trp), gut T cell activation, and peripheral blood CD8 T cell activation (R_S, Spearman rho correlation coefficient; dotted lines represent 95% confidence interval bands for linear regression coefficients).

Figure 4. Relative abundance of DMC members is diminished in HAART subjects compared to VU and falls along a spectrum of VU-like or uninfected-like bacterial communities

(A) Shown are relative abundances of genera within the DMC averaged within subject groups. Each row represents a unique genus. A higher relative abundance is denoted by yellow squares while blue denotes lower relative abundance. After assessment of data distribution, a two-tailed, unpaired T-test assuming unequal variance was used to test significance of differences (***P < 5×10⁻¹⁰). (B) Heatmap of DMC member relative abundances, ordered first by subject grouping, then by subject PC1 values (relative values denoted by color below each subject column). Rows represent genera of the DMC while columns represent individual subject samples. In the treated subject grouping, subjects were ranked for each variable (i.e. Kyn:Trp and IL-6 concentration), and these rankings are represented by colored squares below each subject column (red indicates high Kyn:Trp and high plasma IL-6 in the first and second rows, respectively). (C) Correlations between PC1 and Kyn:Trp or plasma IL-6 concentration. P values displayed represent the contribution of PC1 to the best-fit linear regression model that best predicts each immunologic variable, adjusted to account for multiple comparisons (see Materials & Methods for details).

Figure 5. Bacterial tryptophan catabolism machinery is genetically and functionally homologous to IDO1 enzymatic activity and is enriched in the DMC

(A) Homologs to tryptophan catabolism pathway enzymes are genetically encoded in numerous bacterial genera as denoted by ‘Totals.’ Phylogenetic distribution of annotated enzymes in the kynurenine pathway was collated and represented using pie charts as defined in legend. (B) A Fisher’s exact test for enrichment found that a greater proportion of genera overrepresented in VU encoded 3–4 genes involved in the tryptophan catabolism pathway as compared to genera not differing abundance between VU and HIV- (P = 0.037). (C) All genera that were both detected in our cohort and contained members for which genome sequence data was available were ranked based on their strength of non-parametric Spearman correlation to Kyn:Trp ratios in HIV-infected subject plasma. Genera with 3 or 4 genetically encoded homologs to kynurenine pathway enzymes exhibited significantly stronger correlations to plasma Kyn:Trp ratios than those with only 1 or 2 homologs, and even more so than those with no annotated homologous enzymes. (D) Pseudomonas fluorescens strain A506 and a lab strain of E. coli (DH5α) were grown in King’s B Medium for 20 hours in the presence of varying concentrations of tryptophan. Supernatants from these cultures were assayed for the presence of kynurenine. Error bars indicate standard deviations calculated from three independent experiments. Non-parametric Mann-Whitney tests were used for group comparisons (*P < 0.05, **P < 0.005).