Phylogenetic analyses of PrrA and PrrB in mycobacteria
Since prrAB orthologues are present in all mycobacterial species and prrAB is essential for viability in M. tuberculosis , it is reasonable to believe that PrrAB fulfills important regulatory properties in mycobacteria. We therefore questioned the evolutionary relatedness or distance between PrrA and PrrB proteins in mycobacteria. The M. tuberculosis H37Rv and M. smegmatis mc2155 PrrA and PrrB amino acid sequences share 93% and 81% identity, respectively. Maximum-likelihood phylogenetic trees, based on PrrA (Fig. 1a) and PrrB (Fig. 1b) multiple sequence alignments, were generated. Using the Gupta et al.  recent reclassification of mycobacterial species, the results suggested that, with a few exceptions, PrrA and PrrB evolved with specific mycobacterial clades (Fig. 1). While subtle differences in the PrrA or PrrB sequences may represent evolutionary changes as mycobacterial species of the same clade adapted to similar environmental niches, additional experiments are needed to determine if prrAB is essential in other pathogenic mycobacteria.
We next questioned if the distinct phylogenetic separations between clades could be mapped to specific PrrA or PrrB amino acid residues. We separately aligned mycobacterial PrrA and PrrB sequences in JalView using the default MUSCLE algorithm . Within species of the Abscessus-Chelonae clade, two unique PrrA signatures were found: asparagine and cysteine substitutions relative to serine 38 (S38) and serine 49 (S49), respectively, of the M. smegmatis PrrA sequence (See Fig. S1, Additional file 1). These Abscessus-Chelonae clade PrrA residues were not found at similar aligned sites in other mycobacteria (See Fig. S1, Additional file 1). Similarly, members of the Abscessus-Chelonae clade (except Mycobacteriodes abscessus) harbored unique amino acid substitutions in PrrB, including glutamate, valine, lysine, aspartate, lysine, and valine corresponding to threonine 42 (T42), glycine 67 (G67), valine 90 (V90), methionine 318 (M318), alanine 352 (A352), and arginine (R371), respectively, of the M. smegmatis PrrB sequence (See Fig. S2, Additional file 1).
Transcriptomics analysis of the M. smegmatis WT, DprrAB mutant, and complementation strains
We previously generated an M. smegmatis mc2155 prrAB deletion mutant (mc2155::ΔprrAB; FDL10) and its complementation strain (mc2155::ΔprrAB::prrAB; FDL15) . Since the prrAB regulon and the environmental cue which stimulates PrrAB activity are unknown, a global transcriptomics approach was used to analyze differential gene expression in standard laboratory growth conditions. RNA-seq was used to determine transcriptional differences between the DprrAB mutant, mc2155, and the complementation strains during mid exponential growth, corresponding to an OD600 of ~0.6 (See Fig. S3, Additional file 1), in supplemented Middlebrook 7H9 (M7H9) broth. Total RNA was isolated from three independent, biological replicates of each M. smegmatis strain. Based on multidimensional scaling (MDS) plot, one mc2155 biological replicate which was deemed an outlier and excluded from subsequent analyses (details in Methods, see Fig. S4, Additional file 1). Principal component analysis of the global expression patterns of the samples demonstrated that samples from the mc2155 and FDL15 complementation strains clustered together, apart from those of the FDL10 ΔprrAB strain with the majority of variance occurring along PC1 (See Fig. S5, Additional file 1), indicating complementation with ectopically-expressed prrAB in the ΔprrAB background.
Identifying the PrrAB regulon
To identify differentially-expressed genes (DEGs), pair-wise comparisons of normalized read counts between the DprrAB mutant and WT (FDL10 vs. mc2155) as well as the DprrAB mutant and prrAB complementation (FDL10 vs. FDL15) datasets were performed using EdgeR. Deletion of prrAB resulted in induction of 95 genes and repression of 72 genes (q < 0.05), representing 167 transcriptional targets (Fig. 2a) that are repressed and induced, respectively, by PrrAB in the WT background (Fig. 2c). Less conservative comparisons revealed 683 DEGs (p < 0.05) between the WT and DprrAB mutant strains (See Fig. S6a, Additional file 1). Between the prrAB complementation and DprrAB mutant strains, 67 DEGs (q < 0.05) were identified (Fig. 2b), representing 35 repressed and 32 induced genetic targets by the complementation of PrrAB (Fig. 2c), while less conservative comparisons (p < 0.05) revealed 578 DEGs (See Fig. S6a, Additional file 1). Overall, pair-wise DEG analyses revealed that during mid-logarithmic M. smegmatis growth, PrrAB regulates transcription through a relatively balanced combination of gene induction and repression. In addition, comparison between the two DEG sets (i.e., for mc2155 vs. FDL10 and FDL15 vs. FDL10) datasets revealed 40 (Fig. 2e) and 226 (See Fig. S6b, Additional file 1) overlapping DEGs at the significance levels of q < 0.05 and p < 0.05, respectively. Hierarchical clustering with the overlapping DEGs further illustrated that gene expression changes induced by the prrAB deletion were partially recovered by prrAB complementation (Fig. 2d). We randomly selected six DEGs for qRT-PCR analyses and verified the RNA-seq results for five genes in both the FDL10 vs. mc2155 and FDL10 vs. FDL15 comparisons (See Fig. S7, Additional file 1). [See Additional file 2 for a complete list of DEGs between all pair-wise comparisons.]
Gene ontology and clustering analyses
To infer function of the genes regulated by PrrAB, enrichment of gene ontology (GO) terms (biological processes and molecular functions) in the DEGs of the mc2155 vs. FDL10 comparison was assessed by the DAVID functional annotation tool (See Additional file 3 for a complete list of functional annotations returned from the DAVID results). The two sets of DEGs from the mc2155 vs. FDL10 comparison (See Fig. S6, Additional file 1), were examined. In general, genes repressed by PrrAB were associated with numerous metabolic processes (Fig. 3a) and nucleotide binding (Fig. 3b), while PrrAB-induced genes were associated with ion or chemical homeostasis (Fig. 3c) and oxidoreductase, catalase, and iron-sulfur cluster binding activities (Fig. 3d). Similar GO enrichment terms in the two group comparisons (mc2155 vs. FDL10 and FDL15 vs. FDL10) suggested evidence of genetic complementation (Fig. 3; Fig. S8, Additional file 1). GO term enrichment was also found for metabolism, nucleotide binding, oxidoreductase, and catalase activity, based on conservative (q < 0.05) DEG comparisons (See Figs. S9 and S10, Additional file 1). The GO enrichment analyses suggested that during M. smegmatis exponential growth in M7H9 medium, PrrAB negatively regulates genes associated with diverse components of metabolic and biosynthetic processes and positively regulates expression of genes participating in respiration (qcrA, cydA, and cydB), ion transport (via the F1F0 ATPase), redox mechanisms, and recognition of environmental signals (dosR2) (Fig. 3; Figs. S8, S9, and S10, Additional file 1).
Classification of genes (q < 0.05) based on clusters of orthologous groups (COGs) analyses were then performed using the online eggNOG mapper program. Of all COG categories in each gene list, 32% (n=22) and 24% (n=20) of genes repressed or induced by PrrAB, respectively, participate in diverse aspects of metabolism (Fig. 4), thus corroborating the GO results. Of the COG categories induced by PrrAB, 17% (n=14) were associated with energy production and conversion (COG Category C). The relatively even proportions of COG categories associated with PrrAB-induced and repressed genes (Fig. 4) suggest that this TCS, as both transcriptional activator and repressor, fine-tunes diverse cellular functions to maximize and/or optimize growth potential during exponential replication.
PrrAB regulates dosR expression in M. smegmatis
Differential expression analysis revealed significant repression of MSMEG 5244 and MSMEG 3944, two orthologues of the dosR (devR) response regulator gene, in the ΔprrAB mutant strain (Fig. 2a). In M. tuberculosis, the hypoxia-responsive DosRS (DevRS) TCS (along with the DosT histidine kinase) induces transcription of ~50 genes that promote dormancy and chronic infection . Here, we designate MSMEG 5244 as dosR1 (due to its genomic proximity to dosS) and MSMEG 3944 as dosR2. Among the 25 M. smegmatis homologues of the M. tuberculosis DosRS regulon genes, 23 genes were differentially expressed (+/- 2-fold changes) in pair-wise comparisons among the three strains (Fig. 5 and Additional file 4). Importantly, 21 of the 25 M. smegmatis DosRS regulon homologues were induced by PrrAB in the WT and complementation backgrounds, corroborating the activity of the DosR as a positive transcriptional regulator .
PrrAB is necessary for M. smegmatis adaptation to hypoxia
The cytochrome bd oxidase respiratory system is a high-affinity terminal oxidase that is important for M. smegmatis survival under microaerophilic conditions . Because the cydA, cydB, and cydD genes were repressed in the ΔprrAB mutant during aerobic growth (Fig. 2a; Additional file 2), we questioned if the ΔprrAB mutant was more sensitive to hypoxia than the WT strain. Compared to WT and the prrAB complementation strains, the ΔprrAB mutant exhibited reduced viability (See Fig. S11a, Additional file 1) and produced smaller colonies (See Fig. S11b, Additional file 1) after 24 h hypoxia exposure. In contrast, cell viability and colony sizes were similar for all strains cultured under aerobic growth conditions (See Fig. S11, Additional file 1).
Next, we questioned if differential expression of cydA, cydB, and cydD correlated with growth deficiencies in the ΔprrAB mutant during hypoxia. We compared transcriptional profiles of cydA, cydB, and cydD by qRT-PCR from each strain incubated in M7H9 broth under hypoxic and aerobic conditions for 24 h. After 24 h hypoxia, cydA, cydB, cydD, dosR1, and dosR2 were universally downregulated in the ΔprrAB mutant relative to the WT strain (Fig. 6a-e), demonstrating that PrrAB is inducing gene expression under hypoxic conditions. Expression levels of cydA and cydB were significantly reduced in the ΔprrAB mutant relative to the WT strain during aerobic growth (Fig. 6a, b). Furthermore, both dosR1 and dosR2 were significantly downregulated in the ΔprrAB mutant under aerobic conditions (Fig. 6d, e). Overall, these qRT-PCR results verify the RNA-seq data (Additional file 2) and confirm that PrrAB induces cydA, cydB, cydD, dosR1, and dosR2 in both oxygen-rich and oxygen-poor environmental conditions.
The ΔprrAB mutant is hypersensitive to cyanide exposure
Cyanide is a potent inhibitor of the aa3 cytochrome c oxidase in bacteria. Conversely, cytochrome bd oxidases in Escherichia coli , Pseudomonas aeruginosa , some staphylococci , and M. smegmatis  are relatively insensitive to cyanide inhibition. In the absence of alternative electron acceptors (e.g., nitrate and fumarate), aerobic respiratory capacity after cyanide-mediated inhibition of the M. smegmatis aa3 terminal oxidase would be provided by the cytochrome bd terminal oxidase (CydAB). Because cydA, cydB, and cydD were significantly repressed in the ΔprrAB mutant (Fig. 2a), as were most subunits of the cytochrome c bc1 – aa3 respiratory oxidase complex (See Additional file 2), we hypothesized that the ΔprrAB mutant would be hypersensitive to cyanide relative to the WT and complementation strains. Cyanide inhibited all three strains during the first 24 h (Fig. 6f). While the WT and complementation strains entered exponential growth after 24 h of cyanide exposure, the ΔprrAB mutant exhibited significantly delayed and slowed growth between 48-72 h (Fig. 6f). These data demonstrated that the ΔprrAB mutant strain had defects in alternative cytochrome bd terminal oxidase pathways, further supporting that genes controlling cytochrome c bc1 and aa3 respiratory oxidases are induced by PrrAB.
PrrAB positively regulates ATP levels
KEGG pathway analysis of DEGs (p < 0.05) induced by PrrAB revealed oxidative phosphorylation as a significantly enriched metabolic pathway (Additional file 3; enrichment = 3.78; p = 0.017). Further examination of the RNA-seq data generally revealed that genes of the terminal respiratory complexes (cytochrome c bc1-aa3 and cytochrome bd oxidases) were induced by PrrAB, whereas F1F0 ATP synthase genes were repressed by PrrAB (Fig. 7a). Therefore, we hypothesized that ATP levels would be greater in the ΔprrAB mutant relative to the WT and complementation strains despite the apparent downregulation of terminal respiratory complex genes (except ctaB) in the ΔprrAB mutant (Fig. 7a). While viability was similar between strains at the time of sampling (Fig. 7b), ATP levels ([ATP] pM/CFU) were 36% and 76% in the ΔprrAB mutant and complementation strains, respectively, relative to the WT strain (Fig. 7c). Ruling out experimental artifacts, we confirmed sufficient cell lysis with the BacTiter-Glo reagent (See Methods) and that normalized extracellular ATP in cell-free supernatants were similar to intracellular ATP levels (See Fig. S12, Additional file 1). These data suggested that PrrAB positively regulates ATP levels during aerobic logarithmic growth, although prrAB complementation did not fully restore ATP to WT levels (Fig. 7c). Additionally, ATP levels correlated with PrrAB induction of respiratory complex genes rather than PrrAB-mediated repression than F1F0 ATP synthase genes (Fig. 7a). To verify the RNA-seq data which indicates PrrAB repression of nearly all F1F0 ATP synthase genes (Fig. 7a), we directly measured transcription of three genes in the atp operon: atpC (MSMEG 4935), atpH (MSMEG 4939), and atpI (MSMEG 4943).
The qRT-PCR results revealed that PrrAB represses atpC, atpH, and atpI in the WT and prrAB complementation strains (See Fig. S13, Additional file 1).