Inflammatory Biomarkers of Asymptomatic and Symptomatic Tuberculosis

Abstract

A large proportion of individuals with tuberculosis (TB) are asymptomatic. The biological and inflammatory underpinnings of asymptomatic TB are unknown and may differ from symptomatic TB. We characterised blood transcriptomic and proteomic profiles in South African community screening vs. health facility-based triage cohorts. Asymptomatic TB shared core transcriptomic and proteomic features with symptomatic TB, including upregulation of innate, interferon and inflammatory pathways and downregulation of T and B cell pathways. Integration of transcriptomic and proteomic data from asymptomatic TB individuals identified two distinct sub-clusters characterized by higher or lower bacterial burden, blood IFN-γ responses, BMI, and chest radiographic abnormalities, suggesting different disease severity. We identified a new blood transcriptomic signature of asymptomatic TB. However, diagnostic performance of transcriptomic and proteomic markers was weaker for asymptomatic TB than symptomatic TB, suggesting that policy development for community-based, asymptomatic TB screening should not adopt biomarkers developed for symptomatic TB triage without further optimization.

Keywords: asymptomatic; biomarker; proteomic; transcriptomic; tuberculosis.

Figure 1.. Participant and sample selection for the transcriptomic and proteomic analyses in the three cohorts.

The diagrams illustrate enrolment, classification of TB cases and controls and inclusion and exclusion of samples for transcriptomic (blue) or proteomic (red) analyses for the (A) Household Contact (HHC) Screening Cohort, (B) Clinic Triage Cohort, and (C) Community Screening Cohort. For each cohort, stratification into asymptomatic TB, symptomatic TB, and TB-negative (no TB) is shown. Subsequent exclusions for reasons such as loss to follow-up, poor sample quality (“failed QC”), and case-control selection are indicated, leading the final number of samples analysed for RNA-seq and proteomics for each group.

Figure 2.. Blood transcriptome perturbation and pathways, biological processes, and gene sets modulated in symptomatic and asymptomatic TB.

(A) Box and whisker plots showing molecular distance to health, a composite score that quantifies transcriptome-wide deviation from healthy controls, integrating both the number and magnitude of significantly regulated gene transcripts. On the left are Household Contact (HHC) Screening participants without TB, with asymptomatic TB or symptomatic TB. In the middle are Community Screening participants without TB, with asymptomatic TB or symptomatic TB. On the right are Clinic Triage participants without TB or bacteriologically-confirmed symptomatic TB. The horizontal lines indicate the median, the boxes the interquartile range (IQR) and the whiskers the range of data with 1.5 × IQR from the lower and upper quartiles. The shown p-value was computed with the Mann-Whitney U test. (B) Differential gene set enrichment analysis using QuSAGE on blood RNA-sequencing data, for the gene sets/modules listed on the left, from individuals with asymptomatic TB or symptomatic TB, relative to TB-negative controls, in the HHC Screening Cohort (left), Community Screening Cohort (middle), and Clinic Triage Cohort. Dots denote the enrichment score estimate and error bars the 95% confidence interval. The order of the gene sets is based on ranking from highest to lowest enrichment score estimates in prevalent TB (symptomatic and asymptomatic combined) from the HHC Screening Cohort.

Figure 3.. Blood transcriptome perturbations increase progressively along the TB spectrum.

Analysis of RNA-seq data from individuals categorised and ranked along the TB spectrum, from healthy controls, asymptomatic TB, symptomatic TB in Community Screening participants and symptomatic TB in Clinic Triage participants, by weighted gene co-expression network analysis using CEMiTool. For these analyses, data from the HHC Screening (RePORT SA) and Community Screening (CORTIS) cohorts were combined. (A) M1 through M6 represent six modules of co-expressed gene networks identified by CEMiTool. Dots denote the mean enrichment score estimate and error bars the 95% confidence interval. (B) Gene set overrepresentation analysis for the identified gene modules. Displayed are the top five significantly enriched pathways and processes for each of the six modules, ranked within each module according to Benjamini-Hochberg-adjusted p-value. The colour intensity for each module corresponds to the magnitude of the FDR p -value. Expression levels of representative top regulated genes within each module and pathway along the TB spectrum is shown in Supplementary Figure 3.

Figure 4.. Proteomic biomarkers of asymptomatic TB.

Scatter plots summarizing the diagnostic performance (AUROC) of proteomic biomarkers for distinguishing between asymptomatic TB and TB-negative controls in HHC Screening (RePORT SA, Y-axis) and Community Screening (CORTIS, X-axis) participants or between symptomatic TB and from TB-negative controls in Clinic Triage participants. Each point represents a protein, with its position reflecting AUROC values for the relevant comparison. Points are color-coded according to their lower 95% confidence interval (CI) value for the AUROC; biomarkers with lower 95%CI values below 0.5 were considered to display no significant differentiation between the groups (purple), those with lower 95%CI values above 0.5 in only one cohort are in yellow or blue, and those with lower 95%CI values above 0.5 in both cohorts are in red. (A) AUROC in asymptomatic TB vs control (left) and symptomatic TB vs control (right) in the Community HHC Screening (Y-axis), and Community Screening (X-axis) cohorts for proteins measured in both cohorts. (B) AUROC in symptomatic TB vs controls in Community HHC Screening (Y-axis), and in Clinic Triage participants (X-axis) for protein markers measured in both cohorts. (C) AUROC in asymptomatic TB vs control (Y-axis) and symptomatic TB vs control (X-axis) in the Community HHC Screening (left), and in Community Screening (right) cohorts using all proteins measured in the respective cohorts. This visualization highlights both cohort-specific and phenotype-specific performance as well as biomarkers with reproducible diagnostic potential. Data for each protein are in Supplementary Figure 5.

Figure 5.. Discovery and validation of a transcriptomic signature of asymptomatic TB.

Data from Asymptomatic TB and Control participants from the HHC Screening (RePORT SA) and Community Screening (CORTIS) cohorts were combined and then partitioned into discovery (training, 75%, 21 + 35 asymptomatic TB cases and 135 + 108 healthy controls, respectively) and hold-out validation (25%, 6 + 11 asymptomatic TB cases and 45 + 36 healthy controls, respectively) sets. Feature selection of genes differentiating between asymptomatic TB cases and healthy controls and model fitting into a 3-gene signature of asymptomatic TB, comprising CARD16, LIMK1 and PSTPIP2 was performed in the discovery subset. (A) Expression levels of genes in the novel asymptomatic TB signature in the discovery set and the receiver-operating characteristic (ROC) curve plot for its performance (model fit) in the discovery set. (B) Asymptomatic TB signature scores in asymptomatic TB cases and TB-negative controls in the hold-out validation set. The horizontal lines indicate the median, the boxes the interquartile range (IQR) and the whiskers the range of data with 1.5 × IQR from the lower and upper quartiles. The shown -value was computed with the Mann-Whitney U test. The ROC curve shows the performance of the novel signature in the hold-out validation set for differentiating between asymptomatic TB cases and healthy controls. The right-hand plot shows AUROC values (dots) and 95% CIs (error bars) of the novel asymptomatic TB signature as well as a set of previously published TB signatures in the hold-out validation set. (C) ROC curve showing the performance of the novel asymptomatic TB signature for differentiating between symptomatic TB and symptomatic TB-negative controls in the Clinic Triage cohort (RePORT SA). The right-hand plot shows AUROC values (dots) and 95% CIs (error bars) of the novel signature and a set of previously published TB signatures in Clinic Triage cohort. (D) ROC curves showing the performance of the novel asymptomatic TB signature for differentiating between asymptomatic TB cases (blue) or symptomatic TB cases (red) and controls in the datasets (GSE157657) published by Tabone et al., 2017. The right-hand plot shows AUROC values (dots) and 95% CIs (error bars) of the novel asymptomatic TB signature as well as a set of previously published TB signatures in the GSE157657 dataset.

Figure 6.. Two asymptomatic TB clusters based on transcriptomic and proteomic data.

(A) Similarity network fusion (SNF) analysis on the full transcriptomic and proteomic datasets from the HHC and Community Screening Cohorts combined identified two clusters among asymptomatic TB individuals. Stability of these clusters is addressed in Supplementary Figure 6 (B) Sputum bacteriological outcomes, including days to liquid MGIT culture positivity (only those with culture-positive results included), sputum smear microscopy grade and Xpert semiquantitative category (only those with Xpert-positive or trace results included) among asymptomatic TB clusters 1 and 2 individuals in the HHC and Community Screening Cohorts. IFNγ levels from QuantiFERON Gold-Plus assays for TBAg1 or TBAg2, with levels from the Nil condition subtracted, and BMI values for cluster 1 and 2 individuals in the HHC Screening Cohort (RePORT SA). BMI data were only available for the Community Screening Cohort (CORTIS); analysis of smear grade, and IFNγ levels from QuantiFERON Gold-Plus assays, and CXR scores was based on the Community HHC Screening Cohort (RePORT SA) only, whereas culture and Xpert grading was available for both cohorts. (C) Transcriptomic signature scores for the indicated TB signatures in cluster 1 and 2 individuals. The shown -values were computed with the Mann-Whitney U test. For box-and-whisker plots horizontal lines indicate the median, the boxes the interquartile range (IQR) and the whiskers the range of data with 1.5 × IQR from the lower and upper quartiles.