Mass Spectrometry-Based Proteomic and Metabolomic Profiling of Serum Samples for Discovery and Validation of Tuberculosis Diagnostic Biomarker Signature

Abstract

Tuberculosis (TB) is a transmissible disease listed as one of the 10 leading causes of death worldwide (10 million infected in 2019). A swift and precise diagnosis is essential to forestall its transmission, for which the discovery of effective diagnostic biomarkers is crucial. In this study, we aimed to discover molecular biomarkers for the early diagnosis of tuberculosis. Two independent cohorts comprising 29 and 34 subjects were assayed by proteomics, and 49 were included for metabolomic analysis. All subjects were arranged into three experimental groups—healthy controls (controls), latent TB infection (LTBI), and TB patients. LC-MS/MS blood serum protein and metabolite levels were submitted to univariate, multivariate, and ROC analysis. From the 149 proteins quantified in the discovery set, 25 were found to be differentially abundant between controls and TB patients. The AUC, specificity, and sensitivity, determined by ROC statistical analysis of the model composed of four of these proteins considering both proteomic sets, were 0.96, 93%, and 91%, respectively. The five metabolites (9-methyluric acid, indole-3-lactic acid, trans-3-indoleacrylic acid, hexanoylglycine, and N-acetyl-L-leucine) that better discriminate the control and TB patient groups (VIP > 1.75) from a total of 92 metabolites quantified in both ionization modes were submitted to ROC analysis. An AUC = 1 was determined, with all samples being correctly assigned to the respective experimental group. An integrated ROC analysis enrolling one protein and four metabolites was also performed for the common control and TB patients in the proteomic and metabolomic groups. This combined signature correctly assigned the 12 controls and 12 patients used only for prediction (AUC = 1, specificity = 100%, and sensitivity = 100%). This multiomics approach revealed a biomarker signature for tuberculosis diagnosis that could be potentially used for developing a point-of-care diagnosis clinical test.

From simple to complex: Protein‐based biomarker discovery in tuberculosis

“…Jiang et al., [22] identified a three‐marker signature consisting of SAA, vitamin K‐dependent protein Z, and C4b‐binding protein beta chain with 97.06% sensitivity and 95.45% specificity to discriminate between patients with TB and healthy controls and with 96.32% sensitivity and 87.50% specificity to discriminate patients with TB from those with pneumonia. Fernandes et al., [25] described a four‐protein signature consisting of alpha‐1‐antichymotrypsin (SERPINA3), ceruloplasmin, hemopexin, and complement factor H that discriminated between TB and other respiratory diseases with an AUC = 0.963 and a specificity of 93% and a sensitivity of 91%. In a study by Sandhu et al.…”

“…In a study using both proteomic and metabolomic data, Fernandes et al. [25] identified a signature comprising four metabolites (indole‐3‐lactic acid, trans‐3‐indoleacrylic acid, hexanoylglycine, and N‐acetyl‐L‐leucine) and one protein (hemopexin), with a specificity of 100%, and sensitivity of 100%.…”

“…1B and Table 1). Five studies [22][23][24][25][26] using MS discovered biosignatures fulfilling the TPP criteria for a triage test (Table 2). Xu et al, [24] found that the combination of serum proteins including protein S100-A9 (S100A9), extracellular superoxide dismutase [Cu-Zn] (SOD3), and MMP9 levels could achieve 92.5% sensitivity and 95% specificity to discriminate between pulmonary TB and healthy controls, and 90% sensitivity and 87.5% specificity to discriminate between pulmonary TB and pneumonia.…”

“…Jiang et al, [22] identified a three-marker signature consisting of SAA, vitamin K-dependent protein Z, and C4b-binding protein beta chain with 97.06% sensitivity and 95.45% specificity to discriminate between patients with TB and healthy controls and with 96.32% sensitivity and 87.50% specificity to discriminate patients with TB from those with pneumonia. Fernandes et al, [25] described a four-protein signature consisting of alpha-1-antichymotrypsin (SERPINA3), ceruloplasmin, hemopexin, and complement factor H that discriminated between TB and other respiratory diseases with an AUC = 0.963 and a specificity of 93% and a sensitivity of 91%. In a study by Sandhu et al [23] patients with TB disease were distinguishable from undifferentiated symptomatic controls with an accuracy of 87% (sensitivity 84%, specificity 90%), from symptomatic controls with latent TB infection (accuracy of 87%, sensitivity 89%, specificity 82%) and from symptomatic controls without latent TB infection (accuracy 90%, sensitivity 90%, specificity 92%).…”

“…We further discuss different sample types, experimental platforms, and briefly the signaling pathways associated with the identified host-derived proteins. [12,13], [17], [29], [63] # , [66], [103][104][105][106][107][108][109][110][111][112][113][114] Mass spectrometry [22][23][24], [25] # , [115][116][117][118][119][120][121][122][123][124][125][126][127]] SOMAscan [19,20] Other [21], [70], [78], [128,129] Sputum/Saliva ELISA [7] Multiplex bead assay [14], [29] Mass spectrometry [26], [30], [31] # , [32]<...…”