Meta-analysis of human gene expression in response to Mycobacterium tuberculosis infection reveals potential therapeutic targets

Abstract

Background: With the global emergence of multi-drug resistant strains of Mycobacterium tuberculosis, new strategies to treat tuberculosis are urgently needed such as therapeutics targeting potential human host factors.

Results: Here we performed a statistical meta-analysis of human gene expression in response to both latent and active pulmonary tuberculosis infections from nine published datasets. We found 1655 genes that were significantly differentially expressed during active tuberculosis infection. In contrast, no gene was significant for latent tuberculosis. Pathway enrichment analysis identified 90 significant canonical human pathways, including several pathways more commonly related to non-infectious diseases such as the LRRK2 pathway in Parkinson's disease, and PD-1/PD-L1 signaling pathway important for new immuno-oncology therapies. The analysis of human genome-wide association studies datasets revealed tuberculosis-associated genetic variants proximal to several genes in major histocompatibility complex for antigen presentation. We propose several new targets and drug-repurposing opportunities including intravenous immunoglobulin, ion-channel blockers and cancer immuno-therapeutics for development as combination therapeutics with anti-mycobacterial agents.

Conclusions: Our meta-analysis provides novel insights into host genes and pathways important for tuberculosis and brings forth potential drug repurposing opportunities for host-directed therapies.

Keywords: Drug repurposing; Gene expression signature; Host-direct therapies; Parkinson’s disease; Tuberculosis.

Fig. 1

Flowchart of the statistical meta-analysis of human gene expression in response to Mtb infection. The process consists of eight major steps which were detailed in the grey boxes. The output of each data analysis step was indicated in the corresponding pink box. Detailed criteria for each major step were described in Methods

Fig. 2

The heatmap of the subset 407 DEGs identified in the meta-analysis. For each DEG, its normalized expression value in each sample of the nine datasets was indicated in the heatmap. Two hundred forty one genes were up-regulated and 166 genes were down-regulated. The genes were clustered using the Ward’s method [56]. The samples were grouped first by comparison group then by individual studies

Fig. 3

Pathway map for “LRRK2 and immune function in Parkinson’s disease”. Significant up-regulation of genes was denoted as up-pointing bars colored in red, and significant down-regulation of genes was denoted as down-pointing bars colored in blue. The length of the colored bar was proportional to the fold change of the gene in the meta-analysis

Fig. 4

Pathway map for “Inhibitory PD-1 signaling in T cells”. Significant up-regulation of genes was denoted as up-pointing bars colored in red, and significant down-regulation of genes was denoted as down-pointing bars colored in blue. The length of the colored bar was proportional to the fold change of the gene in the meta-analysis

Fig. 5

Protein-protein interaction network of the 1655 DEGs in the meta-analysis. a The subnetwork of all DEGs including their functional partners. Each node represents a gene and each edge represents an interaction between two genes supported by experimental evidence. The up-regulated genes were colored in red. The down-regulated genes were colored in green. The non-DEG functional partners were colored in grey. The minimum network mode was chosen for display purposes. b The subnetwork of only DEGs exclusive of their functional partners