Bayesian network expansion identifies new ROS and biofilm regulators

Abstract

Signaling and regulatory pathways that guide gene expression have only been partially defined for most organisms. However, given the increasing number of microarray measurements, it may be possible to reconstruct such pathways and uncover missing connections directly from experimental data. Using a compendium of microarray gene expression data obtained from Escherichia coli, we constructed a series of Bayesian network models for the reactive oxygen species (ROS) pathway as defined by EcoCyc. A consensus Bayesian network model was generated using those networks sharing the top recovered score. This microarray-based network only partially agreed with the known ROS pathway curated from the literature and databases. A top network was then expanded to predict genes that could enhance the Bayesian network model using an algorithm we termed 'BN+1'. This expansion procedure predicted many stress-related genes (e.g., dusB and uspE), and their possible interactions with other ROS pathway genes. A term enrichment method discovered that biofilm-associated microarray data usually contained high expression levels of both uspE and gadX. The predicted involvement of gene uspE in the ROS pathway and interactions between uspE and gadX were confirmed experimentally using E. coli reporter strains. Genes gadX and uspE showed a feedback relationship in regulating each other's expression. Both genes were verified to regulate biofilm formation through gene knockout experiments. These data suggest that the BN+1 expansion method can faithfully uncover hidden or unknown genes for a selected pathway with significant biological roles. The presently reported BN+1 expansion method is a generalized approach applicable to the characterization and expansion of other biological pathways and living systems.

Figure 1. Schema for the BN+1 expansion algorithm.

Bayesian networks are generated from discretized microarray data and ranked according to log posterior score. One of the top-scoring networks was selected as a core network for subsequent expansion. Each gene not included in the core network yet appearing in the microarray dataset was independently tested for its ability to acquire the best log posterior score versus the other tested expansion genes.

Figure 2. Consensus network for the ROS detoxification pathway based on gene expression data.

Bayesian networks were generated using twenty-seven genes from the reactive oxygen species (ROS) detoxification pathway as variables or nodes and 305 gene expression microarray observations per variable. Edges which appear in the consensus and are supported by external data (e.g. EcoCyc, RegulonDB, and/or literature) are indicated (see Table S1).

Figure 3. The genes dusB(A) and uspE (B) were the top results for the large network expansion.

(C) Scatter plot for uspE versus gadX highlighting experiments with the word “biofilm” in the experiment title and/or description. High levels of uspE and gadX were observed for all conditions mapped to ‘biofilm’. The dotted lines indicate boundaries for binning used in network learning. A similar profile was shown for gadE (not shown).

Figure 4. Expression profiles of E. coli gadX and uspE upon exposure to hydrogen peroxide.

Change of GFP fluorescence of two reporter strains E. coli BW25113/pgadX-gfp and BW25113/puspE-gfp upon exposure to 0 mM, 1 mM and 10 mM hydrogen peroxide for 20 min. Cells were cultured in LB broth at 30°C overnight and re-suspended in 1×PBS. Different concentration of hydrogen peroxide was added into three aliquots for 20 min before cell density (OD) and fluorescence intensity were measured. Presented GFP fluorescence for each sample was normalized to OD. Error bar indicated standard deviation from two replicated cell cultures.

Figure 5. Analyses of the gadX-uspE interaction through knockout studies.

GFP fluorescence of wild type E. coli BW25113 and single gene knockout mutant ΔgadX carrying the reporter plasmid puspE-gfp, and wild type E. coli and single gene knockout mutant ΔuspE carrying the other reporter plasmid pgadX-gfp. Cells of each reporter strain were cultured in LB broth at 30°C overnight and re-suspended in 1×PBS before cell density (OD) and fluorescence intensity were measured. GFP fluorescence for each strain was normalized to the OD value. Error bars indicated standard deviations from two replicated cultures each with four replicate readings.

Figure 6. Summary of gadX and uspE gene expression under various experimental conditions.

Plot of the expressions of gadX (x-axis) and uspE (y-axis) against each other in different strain backgrounds and tested experimental conditions. The expression of gadX or uspE was represented by the GFP fluorescence of the reporter strains carrying the respective reporter plasmids pgadX-gfp or puspE-gfp. The strain background or experimental conditions were noted by the data. Expression of gene uspE or gadX was assumed as zero in its single gene mutant ΔuspE or ΔgadX, respectively. Wild type strain was used in the ROS exposure experiments using 1 mM and 10 mM hydrogen peroxide. Error bars indicate standard deviation from replicates.

Figure 7. The effect of gadX and uspE on E. coli biofilm formation.

Fluorescent micrograph of biofilms formed by (A) wild type E. coli BW25113, (B) single gene knockout mutant ΔgadX, and (C) single gene knockout mutant ΔuspE. Biomass of biofilms formed by each strain was calculated (D) using the software COMSTAT. Biofilms were formed on glass bottom of 24-well plates for 3 h after inoculation. Suspended cells were gently removed. Biofilms were gently washed with PBS twice and stained with Syto 60 for 10 min before microscopic examination. Images were taken from randomly chosen spots near the center of the well. Error bar in the calculated biomass was standard deviation from three stacks of images. Scale bar = 10 µm.