Phenotypic analyses of mutants under PQ and H₂O₂ stress. (A) Growth analysis of SOD single and double gene knockout mutants in the presence of 0.125 mM PQ. (B) Survival of mutants after a 2 h treatment with either 4 mM PQ or 25 mM H₂O₂. (C) Survival of mutants after 2 h treatment with 5, 15, or 25 mM H₂O₂. H. salinarum NRC-1 Δura3 is the parent strain that was used as a control in all experiments.

Growth and survival of H. salinarum NRC-1 with H₂O₂ and PQ. Effect of increasing concentrations of H₂O₂ and PQ on growth (A, B) and survival (C, D) of H. salinarum NRC-1 in complete media (CM). Growth was measured by determining increase in cell density (OD₆₀₀) at 30 min intervals using a Bioscreen instrument. Survival was calculated as ratio of number of viable cells (N) after 2 h treatment with varying doses of H₂O₂ or PQ to number of viable cells (N_o) in control. Black arrows represent dosage at which microarray analyses were performed. Data shown is representative of the experiments, which were repeated three times.

Temporal transcriptome changes in H. salinarum NRC-1 treated with sub-lethal doses of H₂O₂ (25 mM) or PQ (4 mM). Relative changes in transcription of all genes were measured immediately before and up to 160 min after adding 25 mM H₂O₂ or 4 mM PQ. The mRNA log₁₀ ratios were normalized to the first condition (control) and data were filtered (log₁₀ ratio >0.05, λ=15 in at least one condition; see Materials and methods for details). I. Differentially regulated genes are grouped into different functional groups (A–U) (see Supplementary Table S2 for details). II. Transcriptional profiles that distinguished OSRs to H₂O₂ and PQ (Groups V and W) were identified using the significance analysis of microarrays (SAM) algorithm (Tusher et al, 2001).

Discovery of conditionally co-regulated genes reveals novel physiological links. Conditionally co-regulated genes were discovered using the cMonkey algorithm (see text for details). (A, B) Shows two example biclusters: bc84 (23 genes) and bc12 (21 genes), respectively. The plots show mean and variance-normalized transcriptional changes (log₁₀ ratios) for all genes (gray) in each bicluster with an overlaid averaged profile (black). Filled triangles on x axis indicate time course analysis of responses during continued exposure to 25 mM H₂O₂ or 4 mM PQ. The open triangles, on the other hand, indicate time course analysis of recovery from 2 h exposure to 25 mM H₂O₂ or 4 mM PQ. Genes and their putative functions are indicated in the table. Sequence logos for the conserved motifs are shown in lower right of each panel, with numbers indicating nucleotide position and size of letters indicating relative conservation of bases. Genes possessing conserved DNA sequence motifs are marked with squares, triangles, or circles. (B) Lower right: operon-like organizations of nine genes within bicluster 12.

Secondary protective mechanisms against OS. (A) Endogenous H₂O₂ levels during aerobic growth of Δura3 and ΔperA strains were quantified using the Amplex Red assay. Although H₂O₂ levels were consistently low throughout growth of the parent strain, as expected the levels were significantly higher in the peroxidase mutant. However, the unexpected growth of this mutant coincided with drop in H₂O₂ level, suggesting the presence of an alternate mechanism(s) to detoxify H₂O₂. (B) Defective gas vesicle biogenesis in the peroxidase mutant. To assay for flotation properties, cells were grown to stationary phase at 37°C and then left undisturbed (i.e. without shaking) at room temperature on the bench top. Although the parent Δura3 strain floated to the surface, the peroxidase/catalase mutant sank to the bottom of the flask. Microscopy analysis confirmed that gas vesicles (observed as bright refractile bodies in the parent strain) are completely missing in the peroxidase/catalase mutant. (C) The function of carotenoids and other pigments was predicted to be important based on the bleached appearance of stationary phase cultures of the double Δsod1/sod2 mutant. This was confirmed on measuring and calculating a fourfold increase in the rate of degradation of pigments relative to single sod mutants (Δsod1, Δsod2) and the parent Δura3 strain. Pigments were assayed using UV-VIS spectroscopy.

Coordination of frontline defense mechanisms under extreme OS. (A) Degree of co-regulation of genes within 67 biclusters (residuals <0.45) in diverse environmental stress conditions was evaluated by calculating correlations (R²) among their transcript level changes in those experiments. The R²-values were hierarchically clustered to identify the generalized and specific components of OSR. (B) Pearson's correlation (R) among mean and variance-normalized transcriptional changes for four representative biclusters are shown. Biclusters 31 and 37 are examples of the generalized stress response component of OSR with co-regulation across most environmental conditions. In contrast, co-regulation of genes within biclusters 12 and 84 is much more significant under severe OS conditions. For instance, genes of bc84 are co-regulated under severe OS resulting from treatment with sub-lethal dose PQ and to a lesser degree H₂O₂, but not on irradiation with a sub-lethal dose of γ-irradiation (Whitehead et al, 2006), subjecting cells to sudden and extreme changes in O₂ tension (Schmid et al, 2007) or treatment with sub-inhibitory dose of transition metals (Kaur et al, 2006). Genes within bc12, on the other hand, were better co-regulated on treatment with H₂O₂. Genes within both biclusters (especially bc84) encode important functions associated with frontline defense mechanisms that provide protection against OS (see Figure 3 and http://baliga.systemsbiology.net/drupal/content/egrin-oxidative-stress for characteristics of biclusters). (C) A specific example of OS-specific co-regulation: two genes of bc84, sod1 and msrA, are co-induced under severe OS—especially with PQ treatment—but not coordinately controlled by any other stressful environmental conditions.

EGRIN_OS: environment and gene regulatory influence network for oxidative stress. (A) EGRIN_OS was constructed by applying the Inferelator algorithm to infer predictive regulatory influences of TFs and EFs (H₂O₂ and PQ) on biclusters of conditionally co-regulated genes discovered using cMonkey. To simplify visualization of this model, the network was filtered for biclusters with residuals <0.4 and regulatory influence weights >0.1 and restricted to physiological functions relevant to OS. Rectangles with numbers indicate specific biclusters; width of each box is proportional to number of conditions in that bicluster and height is proportional to number of genes. Likewise, width of each arrow is proportional to the strength of the regulatory influence, with red arrows indicating activation and green indicating repression. Circles indicate single TFs or groups of TFs. ‘TF groups' are averaged expression profiles of their member TFs (see table). Triangles indicate a logical ‘AND' combination of regulatory influences. A red star indicates that the influence of a TF (red font) within a particular group has been experimentally validated. Corresponding position of the TFgroups in the EGRINos network are indicated with a red border on relevant nodes. (B) Cross-correlation with independently measured protein–DNA interactions by ChIP-chip (Bonneau et al, 2007; Facciotti et al, 2007; Reiss et al, 2008) is consistent with the architecture of the network. TFs in different TFgroups bind promoters of genes within the same bicluster (bc88). Further, these TFs also bind promoters of genes in other biclusters that they are predicted to regulate. The network is drawn as a composite of P–D interactions and statistically inferred influences by Inferelator. Thickness of edges are proportional to significance of promoter binding in ChIP-chip experiments, whereas the activate (red arrows) and repress (green edges) influences are based on Inferelator predictions. (C) TFBd binds the promoters of at least four genes grouped into as many distinct biclusters. Consistent with its predicted influence a knockout in TFBd results in significantly lower transcript levels of these genes at all stages of growth in batch culture. This example validates the EGRIN_OS predicted regulation by TFBd and illustrates how a TF coordinates transcription of genes involved in different processes during OSR. (D) Finally, the deletion of TFBc, which is predicted to coordinate amino-acid metabolism and iron metabolism genes, results in increased sensitivity to 4 mM PQ [transcription factor-binding sites for the six TFs (indicated with red stars in panel A) are listed in Supplementary Table S9].