A gene network-driven approach to infer novel pathogenicity-associated genes: application to Pseudomonas aeruginosa PAO1

Abstract

We present here a new systems-level approach to decipher genetic factors and biological pathways associated with virulence and/or antibiotic treatment of bacterial pathogens. The power of this approach was demonstrated by application to a well-studied pathogen Pseudomonas aeruginosa PAO1. Our gene co-expression network-based approach unraveled known and unknown genes and their networks associated with pathogenicity in P. aeruginosa PAO1. The systems-level investigation of P. aeruginosa PAO1 helped identify putative pathogenicity and resistance-associated genetic factors that could not otherwise be detected by conventional approaches of differential gene expression analysis. The network-based analysis uncovered modules that harbor genes not previously reported by several original studies on P. aeruginosa virulence and resistance. These could potentially act as molecular determinants of P. aeruginosa PAO1 pathogenicity and responses to antibiotics.

Keywords: Pseudomonas aeruginosa; antibiotic resistance; gene co-expression network; pathogenicity.

Fig 1

Schematic diagram of the gene co-expression network-based pipeline for the identification and functional characterization of likely differentially expressing genes (LDEGs) that are missed by standard protocols for differential gene expression analysis. This pipeline involves retrieving RNA-Seq data sets from NCBI-SRA (Step 1) and alignment of reads onto the PAO1 transcriptome using the tool Salmon (Step 2). Next, a PAO1 gene co-expression network is constructed, and differential gene expression analysis is performed to identify DEGs (Step 3). LDEGs were identified in modules enriched in DEGs and having over 65% perturbed genes (Step 4). Finally, a functional analysis of LDEGs is performed (Step 5). DEGs refer to differentially expressing genes.

Fig 2

(a) Pseudomonas aeruginosa PAO1 gene co-expression network. The network comprises 48 gene modules, which are shown in different colors. The network was constructed using WGCNA, and the gene modules were retrieved at a soft-thresholding power of 5, minimum module size of 10, merge cut height of 0.1, and deep split of 2 (b) The degree distribution of the PAO1 co-expression network is shown as a histogram of degree (number of connections for a node, k), which reveals that many genes have few connections and few genes have many connections in the network. (c) The log-log plot depicting the characteristic of network connectivity, highlighting the approximate linear relationship of P(k) with k on the log-log scale and a high R² value of 0.87, which indicates an approximate scale-free topology of the network. P(k) refers to the fraction of nodes in the network having k connections to other nodes.

Fig 3

LDEGs from a chronic wound infection RNA-Seq data set, which were found following mapping of genes with expression fold change > X onto the P. aeruginosa PAO1 gene co-expression network and identification of DEG-enriched modules that have over 65% genes with expression fold change > X (see text). The LDEGs are shown here in DEG-enriched module (a) #31, (b) #42, and (c) #8. LDEGs in each module are highlighted in pink.

Fig 4

(a) Module 36 and (b) module 41 are not significantly enriched in DEGs but comprise over 65% of genes that are either DEGs or LDEGs from a chronic wound infection PAO1 expression data set. LDEGs in each module are highlighted in pink.

Fig 5

(a) Module 44 and (b) module 7 are not significantly enriched in DEGs but comprise over 65% of genes that are either DEGs or LDEGs in PAO1 during its treatment with azithromycin. LDEGs in each module are highlighted in pink.