Identification of serum tRNA-derived small RNAs biosignature for diagnosis of tuberculosis

Abstract

The tRNA-derived small RNAs (tsRNAs) are a new class of non coding RNAs, which are stable in body fluids and can be used as potential biomarkers for disease diagnosis. However, the exact value of tsRNAs in the diagnosis of tuberculosis (TB) is still unclear. The objective of the present study was to evaluate the performance of the serum tsRNAs biosignature to distinguish between active TB, healthy controls, latent TB infection, and other respiratory diseases. The differential expression profiles of tsRNAs in serum from active TB patients and healthy controls were analyzed by high-throughput sequencing. A total of 905 subjects were prospectively recruited for our study from three different cohorts. Levels of tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1, and tsRNA-Lys-CTT-2-M2 were significantly elevated in the serum of TB patients compared to non-TB individuals, showing a correlation with lung injury severity and acid-fast bacilli grades in TB patients. The accuracy of the three-tsRNA biosignature for TB diagnosis was evaluated in the training (n = 289), test (n = 124), and prediction (n = 292) groups. By utilizing cross-validation with a random forest algorithm approach, the training cohort achieved a sensitivity of 100% and specificity of 100%. The test cohort exhibited a sensitivity of 75.8% and a specificity of 91.2%. Within the prediction group, the sensitivity and specificity were 73.1% and 92.5%, respectively. The three-tsRNA biosignature generally decreased within 3 months of treatment and then remained stable. In conclusion, the three-tsRNA biosignature might serve as biomarker to diagnose TB and to monitor the effectiveness of treatment in a high-burden TB clinical setting.

Keywords: Tuberculosis; biomarker; biosignature; diagnosis; transfer RNA-derived small RNAs.

Figure 1.

Study design and classification of study participants. TB, tuberculosis; LTBI, latent TB infection; ORD, other respiratory diseases.

Figure 2.

Expression profiles of tsRNAs in the serum of tuberculosis patients and healthy individuals. (A) A heat map displays the clustering of differentially expressed tsRNAs in three TB patients and three healthy individuals. (B) The ratio of each category of differentially expressed tsRNAs. (C) Venn diagram illustrating the distribution of expression profiles. (D) Visual representation of tsRNA differences between two groups in a scatter plot. (E) Volcano plot displaying the outcomes of differential examination of tsRNAs.

Figure 3.

Verification of the elevated expression levels of tsRNAs through qRT-PCR. (A-F) Using qRT-PCR, we analyzed the levels of tsRNAs in the serum of 50 active TB patients, 50 ORD individuals, 50 LTBI individuals, and 50 HC. Non-parametric data was assessed for statistical differences using the Kruskal-Wallis and Mann-Whitney tests. *p < 0.05, **p < 0.01, ***p < 0.001. (G-I) Receiver operating characteristic (ROC) curve analysis to evaluate the diagnostic value of validated tsRNAs with significant differences (tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1 and tsRNA-Lys-CTT-2-M2) in the validation set. TB, tuberculosis; LTBI, latent TB infection; ORD, other respiratory diseases; HC, healthy control.

Figure 4.

Correlations between confirmed tsRNA levels and lung injury severity and AFB grades in tuberculosis patients. (A-C) In TB patients, lung damage was categorized into three classes using a double-blind process. The minimal (1) (n = 20), moderate (2) (n = 19), and advanced (3) (n = 11) illness phases are represented by the images. A correlation was established between the severity of lung damage in active TB patients and the levels of verified tsRNAs (tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1, and tsRNA-Lys-CTT-2-M2) using the Spearman's rank correlation test. Each chart specifies the values of p and r. (D-F) The expression of tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1 and tsRNA-Lys-CTT-2-M2 in serum from active TB patients with different AFB grades. *p < 0.05. AFB, acid fast bacilli; TB, tuberculosis.

Figure 5.

Receiver operator characteristic (ROC) curve analyses of individual cohorts and the value of three-tsRNA biosignature in monitoring tuberculosis treatment response. Utilizing the random forest (RF) algorithm model, ROC curves were generated for the (A) training, (B) test, and (C) prediction groups, with AUCs indicated for each group. (D-E) Monitoring treatment response in tuberculosis using tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1, tsRNA-Lys-CTT-2-M2 and three-tsRNA biosignature. The levels of tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1, tsRNA-Lys-CTT-2-M2 and three-tsRNA biosignature were measured in microbiologically cured patients before starting TB treatment and at 3 and 6 months into treatment. The values of the individual patients are represented by the coloured dots. AUC, area under curve; ROC, receiver operator characteristic. ***p < 0.001; ns, no significance.

Figure 6.

Prediction of the downstream regulation mechanism of three-tsRNA biosignature. (A-C) Vienna RNA Web Services provides secondary structure prediction diagrams for tsRNA-Gly-CCC-2, tsRNA-Gly-GCC-1 and tsRNA-Lys-CTT-2-M2. (D) Analysis of GO functional enrichment for potential target genes of three-tsRNA biosignature. (E) Analysis of the KEGG biological pathway for potential target genes of three-tsRNA biosignature. GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.