Comprehensive plasma proteomic profiling reveals biomarkers for active tuberculosis
Diana J. Garay-Baquero, Cory H. White, Naomi F. Walker, Marc Tebruegge, Hannah F. Schiff, Cesar Ugarte-Gil, Stephen Morris-Jones, Ben G. Marshall, Antigoni Manousopoulou, John Adamson, Andres F. Vallejo, Magdalena K. Bielecka, Robert J. Wilkinson, Liku B. Tezera, Christopher H. Woelk, Spiros D. Garbis, Paul Elkington
Diana J. Garay-Baquero, Cory H. White, Naomi F. Walker, Marc Tebruegge, Hannah F. Schiff, Cesar Ugarte-Gil, Stephen Morris-Jones, Ben G. Marshall, Antigoni Manousopoulou, John Adamson, Andres F. Vallejo, Magdalena K. Bielecka, Robert J. Wilkinson, Liku B. Tezera, Christopher H. Woelk, Spiros D. Garbis, Paul Elkington
View: Text | PDF
Clinical Research and Public Health Infectious disease

Comprehensive plasma proteomic profiling reveals biomarkers for active tuberculosis

Abstract

BACKGROUND Tuberculosis (TB) kills more people than any other infection, and new diagnostic tests to identify active cases are required. We aimed to discover and verify novel markers for TB in nondepleted plasma.METHODS We applied an optimized quantitative proteomics discovery methodology based on multidimensional and orthogonal liquid chromatographic separation combined with high-resolution mass spectrometry to study nondepleted plasma of 11 patients with active TB compared with 10 healthy controls. Prioritized candidates were verified in independent UK (n = 118) and South African cohorts (n = 203).RESULTS We generated the most comprehensive TB plasma proteome to date, profiling 5022 proteins spanning 11 orders-of-magnitude concentration range with diverse biochemical and molecular properties. We analyzed the predominantly low–molecular weight subproteome, identifying 46 proteins with significantly increased and 90 with decreased abundance (peptide FDR ≤ 1%, q ≤ 0.05). Verification was performed for novel candidate biomarkers (CFHR5, ILF2) in 2 independent cohorts. Receiver operating characteristics analyses using a 5-protein panel (CFHR5, LRG1, CRP, LBP, and SAA1) exhibited discriminatory power in distinguishing TB from other respiratory diseases (AUC = 0.81).CONCLUSION We report the most comprehensive TB plasma proteome to date, identifying novel markers with verification in 2 independent cohorts, leading to a 5-protein biosignature with potential to improve TB diagnosis. With further development, these biomarkers have potential as a diagnostic triage test.FUNDING Colciencias, Medical Research Council, Innovate UK, NIHR, Academy of Medical Sciences, Program for Advanced Research Capacities for AIDS, Wellcome Centre for Infectious Diseases Research.

Authors

Diana J. Garay-Baquero, Cory H. White, Naomi F. Walker, Marc Tebruegge, Hannah F. Schiff, Cesar Ugarte-Gil, Stephen Morris-Jones, Ben G. Marshall, Antigoni Manousopoulou, John Adamson, Andres F. Vallejo, Magdalena K. Bielecka, Robert J. Wilkinson, Liku B. Tezera, Christopher H. Woelk, Spiros D. Garbis, Paul Elkington

×

Figure 2

In-depth quantitative plasma proteome profiling in TB.

Options: View larger image (or click on image) Download as PowerPoint
In-depth quantitative plasma proteome profiling in TB. (A) Violin plots ...
(A) Violin plots with median and interquartile range show molecular weight frequency distributions of proteins quantified (peptide confidence ≤ 1% FDR) in each independent HP-SEC segment. The number of proteins with relative quantitative data in all profiled samples is indicated. Four plasma samples from TB patients, 3 healthy controls, and 1 master pool were analyzed. (B) Abundance of quantified proteins from all HP-SEC segments. Only proteins with circulating levels reported in the reference PaxDb4.1 protein abundance database or in the literature were annotated. Proteins considered as classical plasma proteins are indicated in red, tissue leakage proteins in green, proteins with signaling functions in purple, and proteins associated with extracellular vesicles in yellow. Concentrations of detected proteins span 11 orders of magnitude. (C) Principal components analysis (PCA) based on quantified proteins from all HP-SEC segments of 8 profiled samples. iTRAQ tags and groups are indicated. Overall, TB patients were separated from healthy controls by the principal components PC1 and PC2, collectively explaining the 62% of total variance. The TB sample labeled with tag 121 clustered with the healthy control samples. The master pool, a combination of all samples, was located in the center of the samples. (D) Log2-transformed relative protein expression heatmap of all proteins profiled in the 4 HP-SEC segments. Purple indicates TB patients and green healthy controls. Pearson correlation was used for clustering of proteins and Spearman’s for samples. Two clusters were defined based on the relative protein expression, and Gene Ontology (GO) analysis of these was performed using g:Profiler. Cyan, downregulated proteins; magenta, upregulated proteins.