Machine learning from Pseudomonas aeruginosa transcriptomes identifies independently modulated sets of genes associated with known transcriptional regulators

Abstract

The transcriptional regulatory network (TRN) of Pseudomonas aeruginosa coordinates cellular processes in response to stimuli. We used 364 transcriptomes (281 publicly available + 83 in-house generated) to reconstruct the TRN of P. aeruginosa using independent component analysis. We identified 104 independently modulated sets of genes (iModulons) among which 81 reflect the effects of known transcriptional regulators. We identified iModulons that (i) play an important role in defining the genomic boundaries of biosynthetic gene clusters (BGCs), (ii) show increased expression of the BGCs and associated secretion systems in nutrient conditions that are important in cystic fibrosis, (iii) show the presence of a novel ribosomally synthesized and post-translationally modified peptide (RiPP) BGC which might have a role in P. aeruginosa virulence, (iv) exhibit interplay of amino acid metabolism regulation and central metabolism across different carbon sources and (v) clustered according to their activity changes to define iron and sulfur stimulons. Finally, we compared the identified iModulons of P. aeruginosa with those previously described in Escherichia coli to observe conserved regulons across two Gram-negative species. This comprehensive TRN framework encompasses the majority of the transcriptional regulatory machinery in P. aeruginosa, and thus should prove foundational for future research into its physiological functions.

Figure 1.

Data analysis procedure. (A) Overview of the methodology used in the study. It includes gathering high-quality data from the NCBI-SRA as well as generated in the lab. The RNAseq reads were processed and quality control was done. Further, the independent component analysis (ICA) was applied to generate the iModulons that were characterized to get the regulatory networks of P. aeruginosa (Adapted from Sastry et al. (7)). (B) ICA calculates the independently modulated sets of genes (iModulons). A compendium of expression profiles (X) is decomposed into two matrices: the independent components composed of a set of genes, represented as columns in the matrix M, and their condition-specific activities (A).

Figure 2.

iModulons computed from the Pseudomonas aeruginosa transcriptomic data compendium. (A) Plot showing the amount of passed samples per year which is used in the study. (B) Bar plot showing the explained variance in all the iModulons with overall explained variance of 0.66. Total Explained Variance is the sum of the fraction of explained variance across all iModulons. (C) Scatter plot showing the regulon recall versus iModulon recall for all 104 iModulons found in the P. aeruginosa dataset. The scatter plot is divided into four quadrants: Upper right represents the well-matched iModulons; upper left shows iModulons representing a regulon-subset; lower right depicts the regulon-discovery; lower left contains the poorly-matched iModulons. The size of the circle represents the size of the iModulons (number of genes) and the color represents the functional categories as shown in the color key. (D) Treemap of the 104 P. aeruginosa iModulons. The size of each box represents the size of the iModulons (number of genes) and the color shades of each functional category represented by the explained variance of each iModulon. iModulons are grouped into 12 different categories: AA/Nucleotide Metabolism, Biosynthetic Gene Clusters, Carbon Metabolism, Defense Mechanism, Energy Metabolism, Metal Homeostasis, Miscellaneous Metabolism, Prophages, Quorum sensing, Secretion systems, Stress Responses and Structural Components. Abbreviations: AA, amino acids.

Figure 3.

iModulons can aid in the definition of genomic boundaries of biosynthetic gene clusters (BGCs). (A) Genomic locations of the 14 predicted BGCs in the P. aeruginosa PAO1 by using the anti-SMASH software. (B) Scatter Plot showing the gene weights of the ErbR-2 iModulon with the color depicting the COG categories of the genes that it contains. (C) Venn diagram depicting the status of the genes in the ErbR-2 iModulons, ErbR regulon, and the predicted redox-cofactor BGCs by using the anti-SMASH software. (D) Genomic overview of the redox-cofactor BGCs predicted by the anti-SMASH software, alongside the iModulons whose boundaries are defined by genes between the PA1975-PA1990.

Figure 4.

iModulon responses to GlcNAc culture. (A) Scatter Plot showing the gene weights of the NagQ iModulon; the color depicts the COG categories. The NagQ iModulons have two regulons; one is GlcNAc catabolism and other is related to structural components. (B) Heat map depicting the activity of selected iModulons in different concentrations of GlcNAc (1g/l, 2g/l, 4g/l, and 8g/l). It describes the change in differential activities in NagQ, biosynthetic gene clusters, secretion systems, carbon metabolism, amino acid metabolism, and nucleotide metabolism. (C) Activity plot of the conditions expressed in NagQ iModulon in the Paeru_Precise. (D) Plot showing iModulon activities in the presence of N-acetyl glucosamine (GlcNAc), ZnCl, CuSO₄ and FeSO₄ micronutrients. The iModulons include the micronutrient metabolism (NagQ, CueR, Zur-1, Zur-2, FpvR) and the biosynthetic gene clusters (PvdS, PchR, RiPP, NRPS, QscR-2 and k-opioid).

Figure 5.

iModulons related to Carbon metabolism and Amino acid/Nucleotide metabolism. (A) Heat map depicting the differential activity of glucose, sucrose, fructose, N-acetylglucosamine, pyruvate, glycerol, Ca-MHB (bacteriological media), and acetate with respect to HexR-1, NagQ, EutR, FruR, HexR-2, GlpR, and PtxR iModulons. (B) Correlation plot among the Branched chain amino acid [BCAA (LiuR and MmsR)] and the Aromatic amino acid [AAA (DhcR and PhhR)]. The outer layer is divided into the four arcs which depict the four different iModulons. Thin lines represent the common genes among the iModulons, and the thick line connecting different iModulons depicts the Pearson correlation coefficients (PCC). (C) Bar plot representing the iModulon activities of MmsR, LiuR and PhhR under different conditions. The x-label shows some conditions used in the study. The ‘△yhjH vs. wt’ is the knockout of the yhjH, ‘Biofilm vs. Dispersed’ is the biofilm mode of growth, ‘pAMBL vs. metabolite_wt’ is the pAMBL plasmid showing overexpression of metabolites, ‘△crc vs. wt’ is the deletion of the global regulator of crc, ‘PrePSA_gentamycin vs. wt’ is the pre-PatH-Cap library of P. aeruginosa (‘PSA’ PAO1-GFP) treated with gentamycin,’NaNO2_EDTA vs. wt’ is the presence of sodium nitrite and EDTA in the media, and ‘Cisplatin vs. wt’ is the presence of cisplatin and bile in the media. (D) Scatter plot showing the correlation between the BCAA pathways iModulons, i.e. LiuR and MmsR, with the PCC of 0.69.

Figure 6.

Activity clustering of the iModulons among P. aeruginosa defines stimulons. (A) Sulfur acquisition cluster includes the grouping of AtsR, CysB-1 and CysB-2 iModulons with silhouette score of 58. (B) The scatter plot shows correlation between the CysB-1 and CysB-2 iModulons with PCC of 0.86. Both the iModulons show high activity in the planktonic condition and bile salt medium of P. aeruginosa.

Figure 7.

Fear versus Greed trade-off relationship between iModulons. (A) The RpoS-2 iModulon activities were anti-correlated with the Translational-1 iModulon activities. All the stress conditions (hypoxia, iron starvation, osmotic stress, oxidative stress, and low pH) were highlighted with different colors. (B) Scatter plot showing correlation between the RpoS-2 iModulon activity and the rpoS gene expression with the Pearson's correlation coefficient of 0.61.