Host transcriptomic signatures of tuberculosis can predict immune reconstitution inflammatory syndrome in HIV patients

Abstract

Immune reconstitution inflammatory syndrome (IRIS) can be a complication of antiretroviral therapy (ART) in patients with advanced HIV, but its pathogenesis is uncertain. In tuberculosis (TB) endemic countries, IRIS is often associated with mycobacterial infections or Bacille-Calmette-Guerin (BCG) vaccination in children. With no predictive or confirmatory tests at present, IRIS remains a diagnosis of exclusion. We tested whether RISK6 and Sweeney3, validated immune-based blood transcriptomic signatures for TB, could predict or diagnose IRIS in HIV+ children and adults. Transcripts were measured by RT-qPCR in BCG-vaccinated children and by microarray in HIV+ adults with TB including TB meningitis (TBM). Signature scores before ART initiation and up to IRIS diagnosis were compared between participants who did or did not develop IRIS. In children, RISK6 and Sweeney3 discriminated IRIS cases from non-IRIS controls before ART, and at diagnosis. In adults with TB, RISK6 discriminated IRIS cases from controls after half-week on ART and at TB-IRIS onset. In adults with TBM, only Sweeney3 discriminated IRIS cases from controls before ART, while both signatures distinguished cases from controls at TB-IRIS onset. Parsimonious whole blood transcriptomic signatures for TB showed potential to predict and diagnose IRIS in HIV+ children and adults.

Keywords: Bacille-Calmette-Guerin; host biomarker; immune reconstitution inflammatory syndrome; transcriptomic signature; tuberculosis.

Figure 1

Evaluation of RISK6 and Sweeney3 performance for discrimination between IRIS cases and controls in an HIV‐positive paediatric cohort. Transcript signatures were quantified in whole blood by RT‐qPCR in HIV+ children at ART initiation and IRIS diagnosis. Distribution of (A) IRIS cases (total n = 37) by type and relative proportions, (B) time to IRIS diagnosis in weeks, (C) log10(VL), and (D) %CD4 at ART initiation in cases (n = 37) and controls (n = 53). (E) ROC curves depicting the performance of HIV viral load (VL, grey), Sweeney3 score (green), and RISK6 score (red) to discriminate between IRIS cases (n = 37) and controls (n = 53) at ART initiation. (F) ROC curves depicting performance of Sweeney3 (green) and RISK6 (red) to discriminate between IRIS cases (n = 36) and controls (n = 36) at IRIS diagnosis. Comparison of (G) Sweeney3 and (H) RISK6 scores in the subgroups of IRIS (BCG n = 21, TB n = 6, other n = 10) and all controls (n = 53) at ART initiation. Comparison of (I) Sweeney3 and (J) RISK6 scores in all controls (n = 36) and the subgroups of IRIS (BCG n = 21, TB n = 6, other n = 9) at IRIS diagnosis. “Other” cases of IRIS were oral candida, eczema, CMV colitis, molluscum contagiosum, cryptococcal meningitis, papular pruritic eruption, seborrheic dermatitis, tinea capitis, and zoster. Shaded areas in the ROC curves depict the 95% CI. The horizontal lines in the boxplots depict the median; boxes represent the IQR and whiskers the range; dots represent individual sample scores. p‐values were calculated using the Wilcoxon rank‐sum test with Bonferroni correction for multiple comparisons. Adjusted p‐values <0.05 were considered significant (bold text).

Figure 2

Distribution of TB signature scores in TB‐IRIS patients and non‐IRIS controls. Gene expression was quantified by microarray in whole blood from HIV+ adults with TB. Sweeney3 and RISK6 scores were computed on the normalised log2‐transformed transcript level intensities in samples collected at 0, half‐, 1‐, and 2‐weeks (median time to IRIS onset) on ART. (A) Schematic of the sampling timeline. Comparison of (B) RISK6 and (C) Sweeney3 scores in TB‐IRIS patients (n = 16) and non‐IRIS controls (n = 15). The AUC is depicted for each timepoint and is considered significant where the 95% confidence interval > 50 (bold text). The whiskers above and below represent the interquartile range of values and the dot indicates the median.

Figure 3

Distribution of TB signature scores in patients with TBM with and without IRIS. Transcript levels were quantified by microarray in whole blood from HIV+ adults with TBM. Sweeney3 and RISK6 scores were computed on the normalised log2‐transformed transcript level intensities in samples collected at TBM diagnosis, two weeks after tuberculosis treatment (2wpTBRx) when ART started, and two weeks after ART initiation (2wpART, typical time of IRIS onset). (A) Schematic of the sampling timeline. Comparison of (B) RISK6 and (C) Sweeney3 scores between controls (n = 14) and TB‐IRIS cases (n = 18). The AUC is depicted for each timepoint and is considered significant where the 95% confidence interval > 50 (bold text). The whiskers above and below represent the interquartile range of values and the dot indicates the median.