|Home | About | Journals | Submit | Contact Us | Français|
Following phagocytosis by macrophages, Mycobacterium tuberculosis (Mtb) senses the intracellular environment and remodels its gene expression for growth in the phagosome. We have identified an Acid and Phagosome Regulated (aprABC) locus that is unique to the Mtb complex and whose gene expression is induced during growth in acidic environments in vitro and in macrophages. Using the aprA promoter, we generated a strain that exhibits high levels of inducible fluorescence in response to growth in acidic medium in vitro and in macrophages. aprABC expression is dependent on the two-component regulator phoPR, linking phoPR signaling to pH sensing. Deletion of the aprABC locus causes defects in gene expression that impact aggregation, intracellular growth, and the relative levels of storage and cell wall lipids. We propose a model where phoPR senses the acidic pH of the phagosome and induces aprABC expression to fine-tune processes unique for intracellular adaptation of Mtb complex bacteria.
Mycobacterium tuberculosis (Mtb) is an intracellular pathogen that is capable of survival and growth inside the macrophage (MØ) phagosome. For many bacteria, the phagosomal environment is hostile due to a variety of toxic stresses including reactive oxygen and nitrogen species, lysosomal hydrolases and antimicrobial peptides, and an increasingly acidic environment (Rohde et al., 2007a). Mtb, however, is adapted to an intracellular lifestyle; it arrests phagosome maturation and multiplies inside a hospitable phagosome that resembles the early endosome (Sturgill-Koszycki et al., 1996, Russell, 2001). The mechanisms by which Mtb senses the phagosomal environment and adapts its gene expression for intracellular growth are of significant interest given the importance of MØ colonization to the establishment and maintenance of infection.
One of the defining characteristics of the endosomal-lysosomal sorting pathway is its pH gradient. Pathogenic mycobacteria that successfully arrest phagosome maturation in resting mouse bone marrow-derived MØ reside in a phagosome with a pH of ~6.4 (Yates et al., 2005, Sturgill-Koszycki et al., 1994). However, in activated MØ mycobacteria are trafficked to a more acidified vacuole with a pH of ~5.2 (Schaible et al., 1998). Bacteria that are sent to the lysosome encounter a highly acidic, degradative environment with a pH below 5.0. Therefore, the ability to sense phagosomal pH is a potential mechanism by which Mtb defines its intracellular location and modulates its physiology accordingly.
Mtb undergoes a rapid transcriptional reprogramming immediately following phagocytosis by MØ. Gene expression is not altered upon contact with the MØ surface, however, within 20 minutes of internalization over 100 genes exhibit altered expression profiles (Rohde et al., 2007b). This finding demonstrates that Mtb senses and responds to cues within the MØ. Using transcriptional profiling we previously identified 23 genes that are induced by acidic pH in the MØ phagosome, 16 of which are also induced when grown in medium buffered to pH 5.5. These induced genes represent excellent candidates for genes that are required for sensing and adapting to growth inside the MØ. Additionally, the promoters of these genes may be fused with fluorescent proteins to generate reporter strains of Mtb that can be used for genetic screens, high-throughput chemical screens or in vivo imaging. The goal of this study was to define the function(s) of phagosomal pH responsive genes. Here we report functional characterization of a pH-regulated locus encompassing the MT2466, MT2467 and Rv2396 genes. These genes constitute the Acid and Phagosome Regulated (apr) locus and have been designated aprA, aprB and aprC respectively. The locus operates downstream of phoPR and introduces an additional tier of regulation to the physiological response induced by the two component sensor/effector kinase.
The aprABC locus consists of three genes spanning 1.66 kb that are predicted to encode proteins with uncharacterized functions: aprA is predicted to encode a small, highly basic, 71 amino acid protein with no similarity to other proteins; aprB is predicted to encode a small hypothetical protein; aprC is predicted to encode a Pro-Glu polymorphic GC-rich sequences (PE-PGRS) protein (Figure 1A). Given the small size of the aprA and aprB genes they were not annotated in the H37Rv genome (Cole et al., 1998). Analysis of publicly available genomes indicates this locus is only present in species of the Mtb complex, including M. tuberculosis, M. bovis, M. africanum, M. microti and M. canettii. The aprABC locus is absent from the related species M. marinum and M. ulcerans. In more distantly related organisms including M. smegmatis and the M. avium complex, two additional, proximal genes are absent, Rv2394 and Rv2395. The 5 gene region encompassing Rv2394-aprC in the Mtb complex is inserted between two conserved operons that are involved in sulfate transport and reduction (Pinto et al., 2007).
In previous microarray studies we observed that the genes aprABC were induced in MØ two hours after phagocytosis making this locus a part of the “first responder” transcriptome for host adaptation (Rohde et al., 2007b). Treatment of infected MØ with the ATPase inhibitor Concanamycin A (Cca), which increases phagosomal pH from 6.4 to 7.0 (Yates et al., 2005), inhibits phagosomal induction of the aprABC locus, revealing that Mtb uses the acidic pH of the MØ phagosome as a cue to induce aprABC expression (Rohde et al., 2007b). To extend these observations, we examined aprABC locus gene expression using semi-quantitative real-time PCR (qRT-PCR, Figures 1B–1D). Over a 14 d time course (Figure S1A), RNA from Mtb cultures at pH 7.0, pH 6.0 and pH 5.5 were collected and analyzed for the relative expression of genes in the aprABC locus. At pH 6.0 and 5.5, as compared to pH 7.0, we observed a ~2 fold induction of aprA and aprB transcripts after 2 d of acid stress (Figure 1B and 1C). By 6 d we observed a spike in aprA and aprB expression with a sustained ~8–11 fold induction at pH 6.0 and a peak ~8 fold induction at pH 5.5. aprC was also induced at pH 6.0 and 5.5, however, to a lesser degree. Next, we examined the expression of the aprABC locus in MØ over 14 d. Within 2 h of infection aprA and aprB were induced 10 fold in MØ as compared to the MØ medium control (Figure 1D). The gene induction persisted throughout the course of infection with a peak level of induction between days 2 and 6. Together these data demonstrate that the aprABC locus exhibits a sustained induction during growth in acidic medium or within the MØ phagosome. It is notable that near-maximal aprABC induction was observed following 2 d of growth in MØ, as compared to 10 d in vitro. This difference in timing suggests other factors may contribute to induction of the locus in MØ, including growth rate, nutrient availability or signals from other phagosome stimuli.
To probe the activity of the aprA promoter as a dynamic function across the bacterial population we fused a 1 kb region upstream of aprA to the GFP gene to generate the strain CDC1551(aprA′::GFP). When grown in medium buffered at pH 7.0 or pH 6.5, the reporter strain maintained a detectable and stable level of GFP fluorescence at pH 7.0 and pH 6.5 over 6 d (Figure 2A). At the increasingly acidic pHs of 6.0 and 5.5, we observed a ~3 fold or 5 fold induction of GFP fluorescence, respectively. Further analysis of the reporter strain revealed GFP fluorescence was induced at a threshold of pH 6.3 with a peak level of fluorescence between pH 5.75 and pH 5.25 (Figure 2B). Interestingly, this pH is consistent with the pH range of mycobacterium-containing vacuoles in resting and activated MØ, which exhibit pH values of 6.4 and 5.2, respectively (Sturgill-Koszycki et al., 1994, Schaible et al., 1998). These findings are supportive of aprABC locus expression being responsive across the phagosomal pH range experienced by intracellular Mtb in both resting and activated host cells. Because we observed strong induction of GFP fluorescence and reasonable growth at pH 5.7, we performed further induction experiments using pH 5.7.
We and others (Piddington et al., 2000, Vandal et al., 2008) have observed that acidic pH causes reduced growth of Mtb in vitro (Figures S1A). This growth inhibition is partially caused by the presence of the detergent Tween-80 in the growth medium, where acidic pH is proposed to cause the hydrolysis of Tween-80 into lipid species that are inhibitory to Mtb growth (Vandal et al., 2008). To address the potential role of Tween-mediated toxicity in the induction of CDC1551(aprA′::GFP) fluorescence, we examined reporter induction in acidic medium supplemented with the detergent Tyloxapol, which is not hydrolysed at acidic pH and exhibits less toxicity (Vandal et al., 2008). We observed strong, acid inducible reporter fluorescence in medium supplemented with Tyloxapol (Figure 2C), demonstrating that reporter induction is driven by acidic pH, independently of Tween. We tested other environmental stimuli in vitro including treatment of the reporter strain with 5 mM hydrogen peroxide added every other day over the course of 6 d. These conditions induce expression of multiple stress pathways, however, we did not observe induction of the CDC1551(aprA′::GFP) reporter (data not shown), suggesting that aprABC is not generally induced by stress.
To enable ratiometric quantification of reporter fluorescence, we cloned the aprA′::GFP construct into a replicating plasmid carrying a codon-optimized mCherry gene under regulation of the smyc constitutive promoter (Carroll et al., 2010), thus generating the strain CDC1551(aprA′::GFP, smyc′::mCherry). We grew this strain in buffered media at pH 7.0 and pH 5.7 over 12 d and examined GFP fluorescence by flow cytometry. This strain maintained a detectable and stable level of GFP fluorescence at pH 7.0 over the course of the experiment (Figures 3A and S1B). At pH 5.7, however, we observed a steady increase in GFP fluorescence, with a ~5 fold induction of signal by 12 d (Figures 3A, 3B, and S1B). Ratiometric analysis of both constitutive and induced fluorescent signals from CDC1551(aprA′::GFP, smyc′::mCherry) was also examined and the relative fluorescence profile closely matched that of the GFP fluorescence alone (Figure 3A).
CDC1551(aprA′::GFP, smyc′::mCherry) fluorescence was easily detectable by confocal microscopy at pH 7.0 and 5.7 (Figure 3B) suggesting the strain can be used for cell biological studies of individual bacterial gene expression. We infected resting and activated murine MØ with the reporter strain and quantified fluorescence by confocal microscopy. Two hours post infection the fluorescent profiles were homogeneous with low levels of aprA′-driven GFP fluorescence (Figure 3C and 3D). In both resting and activated MØ, we observed increased GFP fluorescence 2 d post infection. Peak GFP fluorescence was detected 6–9 d post-infection. Interestingly, there was a substantial increase in heterogeneity of fluorescence among individual bacteria over time (Figures 3C, 3D, S1C). This population heterogeneity suggests that during the early phase of MØ infection, the infecting Mtb population is either experiencing a variety of phagosomal compartments or responding differentially to similar environments.
We noted that a significant number of the phagosomal pH-regulated genes identified in our transcriptional profiling experiments (Rohde et al., 2007b) matched a list of genes regulated by the two component regulator phoPR (Walters et al., 2006). Additionally, Walters and colleagues reported that aprC/Rv2396 expression was strongly dependent on the presence of phoPR. Therefore, we hypothesized that the aprABC locus may be under control of phoPR. We examined expression of the aprABC locus in WT and phoP transposon (Tn, phoP::Tn) mutant strains by qRT-PCR. We observed a striking reduction in aprABC expression in the phoP::Tn mutant. At pH 7.0 and 5.5, both aprA and aprB were detected at levels >50 fold lower than the WT at pH 7.0 (Figure 4A). aprC also exhibited lower levels of expression in the phoP::Tn mutant. To confirm the role of phoP in aprABC locus expression we transformed the dual fluorescence protein reporter plasmid into the phoP::Tn mutant. The phoP::Tn reporter strain exhibited very low GFP fluorescence at both pH 7.0 and 5.7 (Figure 4B, 4C and 4D), supporting the conclusion that aprABC expression is under the regulation of the phoPR two component sensor kinase. Examination of the aprABC promoter region revealed the presence of a putative PhoP binding consensus sequence (Gupta et al., 2006), suggesting aprABC locus may be under the direct control of PhoP (Figure S2A).
Given the sustained induction of the aprABC locus in MØ we reasoned that the locus may play a role in adaptation for intracellular growth. We deleted the entire aprABC locus by homologous recombination to generate the strain ΔaprABC (Figure S3A). The deletion was confirmed by i) sequencing the junctions between the inserted hygromycin resistance cassette and the chromosome, ii) qRT-PCR analysis of the ΔaprABC mutant (Figure S3B) and iii) microarray analyses (see below). We constructed two complementation strains reintroducing the whole aprABC locus (c-aprABC) or the aprA (c-aprA) gene alone, both driven by the native promoter, inserted elsewhere in the genome using an integrative plasmid. We also generated a deletion of the aprA gene alone to generate the strain ΔaprA and the complemented strain ΔaprA (c-aprA). However, this strain exhibited polar defects on downstream genes, including reduced expression of the downstream aprB gene (data not shown). Therefore, we have focused our phenotypic analyses on the ΔaprABC whole locus deletion mutant.
We first examined the growth of the ΔaprABC locus mutant in vitro at pH 7.0 and 5.7 and did not observe any growth differences between the strains (Figure S4A). This observation revealed that although the aprABC locus is induced by acidic pH, it is not required for growth in acidic environments in vitro. We next examined growth of the ΔaprABC mutant in resting and activated murine MØ. In resting MØ, WT CDC1551 grew steadily over the course of 9 d, whereas the ΔaprABC mutant did not increase in numbers (Figure 5A). The complemented ΔaprABC mutants with either the whole aprABC locus or aprA alone regained the ability to grow in MØ over the first 6 d of infection, demonstrating that the aprABC locus is required for optimal intracellular growth in MØ early in infection. Complementation was only partially effective, with reduced populations in the complemented strains compared to the WT at 9 d. The ΔaprA mutant exhibited a similar defect in intracellular growth to the ΔaprABC mutant and this defect was also partially complemented by aprA (Figures S4B). In IFN-γ and LPS activated MØ, it is known that the growth of WT Mtb is restricted due to the enhanced antimicrobial defenses. Over the course of an experimental infection, WT Mtb grew modestly with a 5 fold increase in bacterial numbers (Figure 5B). In contrast, the ΔaprABC mutant did not grow in activated MØ and exhibited a 9 fold reduction in bacterial numbers over the infecting inoculum. Complementation of the ΔaprABC mutant with aprA alone restored survival to levels comparable to the WT. These intracellular growth assays suggest that: i) the aprABC locus is required for normal intracellular growth early in infection of MØ, and ii) aprA alone is capable of restoring a significant proportion of the intra-MØ growth defect.
Transcription of the aprABC locus is under the genetic control of phoP indicating the existence of a common genetic pathway. Additionally, like the ΔaprABC mutant, phoP mutants are reported to grow poorly in MØ (Perez et al., 2001). We hypothesized that if phoP and the aprABC locus are in the same genetic pathway, then a double mutant of phoP and ΔaprABC should have a similar growth defect to the single mutants alone. We generated a phoP::Tn/ΔaprABC double mutant and infected resting MØ with the mutant. The phoP::Tn/ΔaprABC double mutant exhibited a growth pattern similar to that of the phoP::Tn mutant alone (Figure 5C), confirming that the growth defects observed for phoP and ΔaprABC are in a common genetic pathway.
During routine culture in 7H9 medium supplemented with Tween 80 detergent there were no apparent growth differences between the WT and ΔaprABC mutant. However, when grown in medium in the absence of Tween 80, we detected differences in culture appearance. WT CDC1551 exhibits substantial clumping and aggregation of cells (Figure 5D). This phenomenon has been credited to a variety of factors associated with components of the cell envelope, including cell wall lipids such as trehalose dimycolates (TDM) as well as specific proteins that are present in the envelope including HbhA and PE-PGRS proteins (Menozzi et al., 1996, Brennan et al., 2001). The ΔaprABC mutant, however, did not aggregate and grew as an opaque, almost clump-free suspension (Figure 5D). Complementation of the mutant with aprA or the whole locus partially restored the culture aggregation phenotype. Loss of the aprABC locus also revealed a culture phenotype on agar medium plates where the ΔaprABC mutant grew as a smaller colony than the WT (Figure 5E and S5). The small colony phenotype was complemented by aprA alone (Figure S5), suggesting that aprA is required for normal colony formation.
An aggregation phenotype has previously been reported with phoP mutants (Perez et al., 2001, Asensio et al., 2006). We also observed decreased cellular aggregation in the phoP::Tn mutant (Figure 5D), although this did not present as a complete loss of aggregation. Notably, we detected low levels of aprA transcripts in the phoP::Tn mutant (Figure 4A), suggesting that relative aprABC transcript levels may account for the intermediate aggregation phenotype in the phoP::Tn mutant. The phoP::Tn/ΔaprABC double mutant exhibited an almost complete loss of aggregated cells verifying the role of aprABC in mediating cellular aggregation.
Complementation of the aprABC locus mutant suggests that aprA plays a major role in the defects observed in MØ growth, culture aggregation and colony size. However, in these experiments the complementation was incomplete. To dissect the functions of individual genes in their native context, we generated a series of isogenic deletion strains (Figure S3A) progressing from the 3′ region of the locus, including ΔaprC alone and ΔaprBC. By comparing these strains with the ΔaprABC mutant we can deduce the genetic dependence of individual genes within the locus (Figure S6). For example, defects that are present in the ΔaprABC locus mutant and absent in ΔaprBC and ΔaprC mutants are likely dependent specifically on aprA.
When examined for ΔaprABC-like aggregation and growth defects, we observed that the ΔaprBC and ΔaprC mutants were indistinguishable from the WT in cellular aggregation and colony size (Figure 5D and 5E). These mutants also grew better in MØ than the ΔaprABC mutant (Figure S4C). Because the ΔaprBC and ΔaprABC (c-aprA) mutants both exhibit restored or partially restored growth and aggregation phenotypes as compared to the ΔaprABC mutant, we conclude that these defects are primarily the result of the absence of aprA. These findings support our model where differences between ΔaprABC and ΔaprBC can be attributed to the presence or absence of aprA.
The observed aggregation and colony morphology phenotypes implicate differences in cell wall constituents and lipid metabolism. To examine the role of the aprABC locus on these processes, we metabolically-labeled Mtb cultures with 14C-acetate and determined lipid profiles. The cultures were grown at pH 7.0 or pH 5.7 for 6 weeks and lipids associated with the mycobacterial pellet were extracted and examined by thin layer chromatography (TLC). Equal counts of WT and ΔaprABC total lipids were examined by 2D TLC using a solvent system that resolves apolar lipids. We observed multiple differences in the relative abundance of lipid species. At both pH 7.0 and pH 5.7 we observed higher accumulation of a lipid species (band 1) in the mutant as compared to the WT (Figure 6A, 6C, 6D, S7A). Band 1 migrated at a position consistent with triacylglycerol (TAG). We also observed an altered pattern of lipids in a region associated with phthiocerol dimycocerates (PDIM, bands 2–5). The WT had four radiolabelled spots, whereas the ΔaprABC locus had only two predominant spots. We also observed a highly apolar band that only accumulated in the ΔaprABC mutant (band 6).
To confirm that the altered species were indeed TAG and PDIM, we analyzed bands 1, 2 and 3 by mass spectrometry (MS). For band 1 we examined lipids from both WT and ΔaprABC mutant and in both cases the samples were identified as TAG (Figure 6B). There were no obvious differences between the WT and ΔaprABC mutant in TAG species detected. Specific peaks were further analyzed to identify structural components of the Mtb TAG and all tested peaks were TAG with two common fatty acyl chains, such as 18:1/16:0 or 19:0/16:0, and one long fatty acyl chain, 26:0 or 24:0. Bands 2 and 3 were identified as phthiocerol A dimycocerate and phthiodiolone dimycocerate, respectively (Figure 6B).
Given the observed role of aprA and phoP in culture aggregation, we examined radiolabelled lipids in additional strains including the ΔaprBC mutant and the phoP::Tn and phoP::Tn/ΔaprABC double mutants. Similar to the WT, ΔaprBC did not accumulate the highly apolar band or excessive TAG suggesting that loss of aprA is responsible for the accumulation these lipids (Figure 6C, 6D, S7A). The ΔaprABC mutant accumulated ~2–3 fold more TAG relative to the WT or ΔaprBC mutants in unbuffered medium and pH 7.0 medium. Mtb accumulated lower relative levels of TAG at pH 5.7, where TAG was more abundant in the ΔaprABC mutant. Similarly, complementation of ΔaprABC with aprA returned TAG accumulation to WT levels (Figure S7A) further establishing a role for aprA in the normal accumulation of TAG.
The altered balance of PDIM species indicates a divergence between the aprABC- and phoP-dependent pathways. The ΔaprABC and ΔaprBC mutants exhibited alterations in the relative levels of PDIM-related lipids at both pH 7.0 and 5.7 (Figure 6C), suggesting aprB or aprC plays a role in regulating the balance of PDIM species, independently of pH. Attempts to complement this phenotype with the aprABC locus were unsuccessful (data not shown), possibly owing to partial complementation of the locus and the relatively low levels of phthiocerol A and phthiodiolone dimycocerates detectable in the WT. However, we observed the same altered PDIM profile in two independently generated mutants (ΔaprABC and ΔaprBC), suggesting the phenotype is not a spontaneous mutation in PDIM synthesis. The phoP::Tn and phoP::Tn/ΔaprABC mutants exhibited substantially altered lipid profiles, with the balance of PDIM lipids shifted to the more polar varieties, phthiocerol A and phthiodiolone dimycocerates (Figure 6A and 6C), in a pH-dependent manner. This suggests that the phoP::Tn phenotype is dominant over the ΔaprABC phenotype, given the comparable lipid profiles in the phoP::Tn and phoP::Tn/ΔaprABC mutants. Organizationally, this observation places phoP and aprABC in a pathway associated with PDIM regulation, with phoP in a position upstream of aprABC.
We examined the role of the aprABC locus in regulation of gene expression by comparing transcriptional profiles of WT Mtb with the mutant strains: ΔaprABC, ΔaprBC, ΔaprC and phoP::Tn. Comparison of the WT with the ΔaprABC revealed widespread changes in gene expression with 224 genes significantly higher and 206 genes significantly lower (1.4X, p<0.05) in the ΔaprABC mutant (Figure 7, Tables S3 and S4). Comparisons of the ΔaprC, ΔaprBC and ΔaprABC profiles, revealed that 32 genes that were lower in the ΔaprABC mutant have WT levels of expression in the ΔaprBC and ΔaprC mutants (Figure 7A, Table 1). This finding suggests these genes are normally induced in the presence of aprA. Conversely, 23 genes that were higher in the ΔaprABC mutant have WT levels of expression in ΔaprBC and ΔaprC, suggesting these genes are normally repressed in the presence of aprA (Table 1). For the purposes of discussion we will refer to these aprA-dependent genes as aprA-induced (red genes on Figure 6A) and -repressed (blue genes on Figure 7A) respectively, although, this designation refers to genetic control and is not intended to imply direct control.
The aprA-induced gene list contains several notable cistrons of co-regulated genes. Most striking are the operons Rv2484c-Rv2481c and Rv1806–1809 whose expression were strongly downregulated in the ΔaprABC mutant, but were otherwise normally expressed in all other strains tested (Figure S8A). The Rv2484c operon is predicted to encode genes involved in TAG synthesis, via the Kennedy pathway, or the formation of wax-esters, depending on the activity of the predicted wax-ester/diacylglycerol synthase (WE-DAGS) protein Rv2484c (Kalscheuer et al., 2003). The Rv2484c gene expression was down 3.9-fold (p=0.02) in the ΔaprABC mutant compared to the WT levels of expression in ΔaprBC and ΔaprC. This finding is particularly interesting given the abnormal accumulation of TAG and highly apolar lipids in the ΔaprABC mutant, which was shown to be an aprA-dependent phenotype. The Rv1806–1809 operon is predicted to encode a cluster of PE and PPE proteins. Our previous experiments had identified Rv1806 and Rv1807 as pH- and phagosome-regulated (Rohde et al., 2007b), and these data support a model where aprA controls the expression of these genes in the MØ in response to pH.
Perhaps the most striking group of aprA-induced genes is a cluster of 21 predicted cistrons spanning a 40.4 kb region between Rv3301c and Rv3338. In this region, 20 genes were significantly lower in ΔaprABC compared to ΔaprBC or ΔaprC (Figure S8B). These 20 genes include genes in 13 of the 21 cistrons in the region, including the alternative sigma factor, sigJ, that was significantly downregulated (1.9X, p=0.022) in the ΔaprABC mutant and slightly upregulated in ΔaprBC mutant (1.2X, p=0.046). Predicted functions of some genes in this region include molybdopterin biosynthesis (Rv3322c-MT3427), central carbon metabolism (sdhABCD) and nucleotide metabolism (Rv3307, Rv3309c, Rv3313c–15c).
We compared the genes that were induced and repressed by aprA to genes that were controlled by phoP (Figure 7B and 7C). We found that 19 of 32 (59%) aprA induced and 12 of 23 (52%) aprA repressed genes were also differentially-regulated in the phoP::Tn mutant (Figure 7B). Notably, the phoP::Tn and ΔaprABC mutants showed comparable regulation of genes in the sigJ region, including lower expression of sigJ (Figure S8B). Intriguingly, both the Rv1806 and Rv2484c operons are expressed at WT levels in the phoP::Tn mutant (Figure S8A). These findings suggests either that residual levels of aprA are sufficient to maintain expression of these genes, or alternatively, that the misregulation may be driven by a phoP-dependent process (as presented in the discussion).
Only 3 genes show expression patterns with clear aprB dependence: Rv2934 (ppsD), Rv2939c (papA5) and Rv2341 (lppQ). Although they missed strict cut-offs, the genes near the ppsD and papA5, including ppsE and the drrABC operon also exhibit aprB dependence (Figure S8D). Genes in this region are involved in the synthesis and transport of PDIMs (Gokhale et al., 2007).
Genes that are misregulated in all three mutants ΔaprABC, ΔaprBC and ΔaprC compared to the WT constitute a class of genes that are dependent on the presence of aprC. We found 157 aprC-dependent genes with 63 aprC induced and 94 aprC repressed (Figure 7D, Tables S1 and S2). Notably, 100 of 157 (65%) of these genes were also differentially expressed in the phoP::Tn mutant (Figure 7E). We observed a significant decrease in the expression of phoP in the ΔaprABC (1.6X, p=0.002), ΔaprBC (1.6X, p=0.0006) and ΔaprC (1.8X, p=0.056) mutants, revealing that aprC (or a dependent gene) acts to positively regulate the expression of phoP. The aprC-dependent expression of phoP confounds the analysis of aprC regulated genes given the broad pathways whose regulation intersects with phoP control, including the dosRS dormancy/hypoxia regulon (Gonzalo-Asensio et al., 2008). Indeed, we observed 17 of 50 (34%) dosR induced genes (Park et al., 2003) were expressed lower in the ΔaprABC mutant (Figure S8E). Given the strong influence of phoP feedback regulation on aprC-regulated genes, it is notable that 57 aprC regulated genes have expression patterns that differ significantly from phoP (Figure 7F). Indeed in several cases, genes show opposite trends in misregulation between the ΔaprABC and phoP::Tn mutants, further supporting a genetic interaction between the two loci (as elaborated in the discussion). Overall, the transcriptional profiling data demonstrate that the aprABC locus plays an important role in the biology of Mtb, as evidenced by the hundreds of genes that are differentially regulated in its absence.
The genes of the aprABC locus exhibit high levels of sustained induction during growth in acidic environments and MØ, suggesting they play an important role in MØ adaptation. The biochemical function of AprABC proteins remains to be defined, however, analysis of the amino-acid sequences suggests a potential function of AprA. aprA is predicted to encode a 71 amino acid protein, with three α-helices and a very basic isoelectric point of ~11 (Figure S2B). We have expressed and purified the AprA protein in E. coli indicating aprA encodes a stable protein (data not shown). The strong positive charge of AprA is driven by the presence of two Arginine Rich Motifs (ARM). These features are suggestive of a nucleic acid binding protein and possibly an RNA binding protein given a proven role for ARMs in this process (Weiss et al., 1998). This function is consistent with the broad control of aprA on gene expression observed in the current study. Analysis of aprB provides no clues regarding its function because it is predicted to express as a 54 amino acid protein with little secondary structure except a short β-sheet in the C-terminus. aprC belongs to a common class of Mtb genes that encode the PE-PGRS class of proteins. The PE-PGRS proteins are unique to mycobacteria and their abundance has expanded in slow-growing pathogenic mycobacteria, suggesting they may be involved in adaptation for growth in a host (Gey van Pittius et al., 2006). Indeed, in M. marinum, PE-PGRS proteins are required for survival in macrophages (Ramakrishnan et al., 2000) and in M. tuberculosis PE-PGRS proteins can function to alter macrophage cytokine signaling (Basu et al., 2007).
The generation of pH-responsive, GFP-expressing Mtb strains opens up several avenues of future experimentation. The inducible fluorescence represents a synthetic phenotype on which we can perform novel forward genetic or high-throughput screens to identify genes or small compounds that target pH sensing and adaptation. For example, in forward genetic screens, we can mutagenize this reporter strain to identify mutants that have aberrant reporter fluorescence in response to acid stress. Our observation that the phoP::Tn mutant has a ~40 fold decrease in fluorescence compared to WT proves in principle that such a screen is feasible.
During growth in MØ, the CDC1551(aprA′::GFP, smyc′::mCherry) reporter exhibited a tremendous degree of heterogeneity in inducible GFP fluorescence. Given the homogeneous population of GFP fluorescence in liquid medium at the start of the macrophage infection, this finding suggests that Mtb experiences diverse environments in the macrophage early in infection. Based on the identified threshold of pH 6.3 for activation of GFP fluorescence, it is possible that bacteria with low GFP fluorescence are in more hospitable, higher pH vacuoles, as compared to those with high fluorescence. It is notable that we did not observe an enhanced induction of reporter fluorescence in the more highly acidified activated MØ phagosomes as compared to resting MØ (Figure 3C). This observation suggests that although the aprABC locus is induced by acidic pH, other stimuli associated with the macrophage environment may also influence reporter induction. For example, the activated MØ phagosome may be sufficiently stressful to the bacterium, due to acidified pH below 5.5 and the presence of hydrolytic enzymes, that bacterial growth or survival is limited and gene expression patterns are altered. The use of reporter strains to study gene expression in individual bacteria reveals that the dynamics of MØ infection are complex and not completely described by examining the average of a population.
The aprABC locus shares a genetic pathway with the phoPR two-component regulator system. Because, aprABC is induced by acidic pH in a phoP dependent manner, our data support a model where phoPR is acting to sense environmental pH and modulates expression of a subset of genes through the aprABC locus. Indeed, 53% of genes differentially regulated in the ΔaprABC mutant, were also differentially regulated in the phoP::Tn mutant. Additionally, aprC expression promotes phoP expression, suggesting an explanation for the sustained high levels of aprABC induction. In Salmonella spp., phoP has also been shown to be under control of a positive feedback loop, where this system is proposed to enable rapid adaptation for the host environment (Shin et al., 2006).
It is equally interesting that many aprABC regulated genes exhibit profiles that differ significantly from the phoP mutant (Figure 7C and 7F). This is surprising because the aprABC locus is downregulated strongly in the phoP::Tn mutant. To explain this observation, we propose a speculative model where an aprABC-dependent pathway may be acting on an unknown factor that is generated in a phoP-dependent manner (Figure S9). In the absence of aprABC, the unknown intermediate may accumulate and lead to altered gene expression. However, in the absence of phoP, the intermediate is not generated and therefore relieves the differential expression of aprABC dependent genes. In this model, the observation of genes that are differentially regulated in ΔaprABC, but not phoP::Tn, further supports a functional link between the loci. The highly apolar “Band 6” (Figure 6) accumulates in the mutants with a pattern consistent with this model. This is not a completely novel scenario as similar inhibitory or toxic intermediates have been demonstrated in Mtb mutants defective in enzymes involved in the processing of cholesterol(Chang et al., 2009).
The connections between phoP and aprABC suggest that a specific subset of the phoP regulon is controlled by aprABC. However, many previously characterized targets of the phoP pathway do not appear to be impaired in the ΔaprABC mutant. For example, pks2 and pks3, genes involved in production of acylated trehaloses are strongly downregulated in the phoP::Tn mutant, but are normally expressed in ΔaprABC. These acylated trehaloses are produced at lower levels in phoP mutants (Walters et al., 2006, Asensio et al., 2006), however, we did not observe any major differences in acylated trehaloses between the WT and ΔaprABC mutant (Figure S7B).
The ΔaprABC mutant has altered relative accumulation of apolar lipids including TAG and PDIM. Interestingly, other mutants have been recently shown to accumulate higher levels of TAG and altered PDIM profiles, including the whiB3 mutant and the opp transporter mutant (Singh et al., 2009, Flores-Valdez et al., 2009). In transcriptional profiles of the opp mutant, it was observed that aprA and aprB are downregulated (Flores-Valdez et al., 2009), which may explain the similar phenotype between opp and ΔaprABC mutants. It was hypothesized that whiB3 may alter the flux of lipids to TAG and PDIM to deal with a reductive stress associated with propionate metabolism during growth on host lipids as a carbon source (Singh et al., 2009). Consistent with this model, we observed that several methylcitrate cycle genes are expressed at lower levels in the ΔaprABC mutant, including Rv1130, Rv1131, icl2, sdhABCD, and mdh (Figure S10A) suggesting that the aprABC locus is required for the normal induction of the pathway. Furthermore, the aprABC locus may play a role in controlling the accumulation of TAG as a storage lipid. Mtb has been hypothesized to rely on storage TAG during growth in humans given that human sputum samples contain bacteria with abundant lipid droplets and enhanced expression of a TAG synthase gene (Garton et al., 2008).
In our array studies, none of the putative TAG synthases exhibit induced expression in the aprABC mutants (Figure S8C). However, Rv2484c and another characterized TAG synthase Rv3130c (Sirakova et al., 2006) are downregulated in the aprABC mutant in an aprA dependent manner (Figure S8C). This finding suggests that i) Rv2484c and Rv3130c may be regulated by intracellular TAG accumulation, and ii) the accumulation is not due to increased de novo synthesis resulting from increased TAG synthase expression. It is possible that disruption of an alternate pathway is resulting in an accumulation of long chain fatty acids, that are being sequestered on TAG (which we detected as C24:0 and C26:0 groups). The C24–26 carbons on TAG may be produced by the FAS-I system in Mtb. Pathways downstream of FAS-I, including FAS-II system and genes associated with trehalose and TDM biosynthesis exhibit differential expression in these mutants (Figure S10B) supporting altered flux through this arm of metabolism.
The aprABC locus likely plays a role in the relative accumulation of PDIM species (Figure 6) and expression of PDIM biosynthesis and transport genes (Figure S8D). PDIM biosynthesis is also coupled to propionate metabolism (Jain et al., 2007) suggesting these observed differences may be driven by altered flux through propionate metabolic pathways observed in our transcriptional profiles. pH-driven manipulation of carbon flux to detoxify propionyl-CoA, and to enhance synthesis of lipids implicated in bacterial virulence may represent a host-specific adaptation that increases the success of the Mtb complex as a pathogen (Russell et al., 2010).
Our characterization of the aprABC locus supports a model where phoP uses phagosomal pH to sense its environment and reprogram the transcriptome for growth in the MØ. A key component in this intracellular adaptation is the rapid and sustained induction of the aprABC locus. The aprABC locus functions to genetically modulate the expression of hundreds of genes. Given the broad genetic control of the aprABC locus on Mtb gene expression and lipid metabolism, we have yet to determine the contributions of individual physiological changes to enhanced bacterial survival. However, the unique presence of the aprABC locus within the Mtb complex leads one to speculate that this locus has been selected and maintained to perform a function that is important for the success of Mtb complex bacteria but is dispensable for growth of related Mycobacterium species, or even for Mtb outside the host. The data indicate that it is within the MØ that the activities of the aprABC locus are key to optimal survival and growth. Therefore, elucidating the functional mechanisms of the aprABC locus should provide insights into the molecular strategies essential for Mtb pathogenesis.
All bacterial strains studies are in the Mtb CDC1551 strain background. The phoP::Tn mutant (BEI Cat # NR-14776) contains a Tn insertion at bp 629 of MT0782. Bacteria were grown in standing, vented tissue culture flasks in 7H9 Middlebrook medium, supplemented with OADC and 0.05% Tween 80, unless stated otherwise. For acid stress, the 7H9 medium was buffered with 100 mM MOPS or 100 mM MES for pH 7.0 and pH 6.5-5.0, respectively (Piddington et al., 2000). Tyloxapol was used at a final concentration of 0.025% to substitute for Tween-80 in the experiments presented in Figure 2C. MØ infections and survival assays were performed as described by Pethe et al. with modifications presented in the supplemental methods (Pethe et al., 2004).
RNA isolation, amplification and labeling, microarray hybridizations and data analyses were performed as described by Rohde et al. (2007b, Supplemental methods). For transcriptional profiling experiments, RNA was isolated from strains grown to early-log phase (OD 0.25) in standing, vented T-75 flasks in 7H9 medium buffered at pH 7.0 as described above. The presented transcriptional profiles were submitted to the NCBI GEO database (accession number GSE22854) and are also available at the TB database website (Reddy et al., 2009). For semi-quantitative real-time PCR experiments cDNA was generated from 100 ng of amplified RNA using iScript cDNA synthesis kit (Biorad) and real time PCR was performed using the iTaq SYBR Green kit (Biorad). The sigma factor gene SigA (Rv2703) was used to normalize each sample and approximate fold induction was calculated using the delta-delta CT method (Livak et al., 2001). Average fold induction and standard deviation were calculated based on 3–4 replicates at each time point.
A 989 bp region upstream of aprA was PCR amplified and cloned upstream of the GFP-mut2 (Cormack et al., 1996). This construct was cloned into a modified version of the replicating plasmid pSE100 and transformed into CDC1551 to generate CDC1551(aprA′::GFP). For the dual fluorescent protein reporter the aprA′::GFP-mut2 construct was cloned into the replicating plasmid pCherry3 (Carroll et al., 2010) and transformed into Mtb strain CDC1551 to generate the strain CDC1551(aprA′::GFP, smyc′::mCherry). For acid induction experiments, cultures were grown in 10 mL of buffered 7H9 in standing vented T-25 tissue culture flasks. At each time point samples were fixed in 4% paraformaldehyle (PFA) and GFP fluorescence was measured by counting 25000–50000 cells by flow cytometry on a BD FACScalibur instrument. GFP and mCherry fluorescence ratios were measured on a Perkin-Elmer EnVision plate reader where each sample was split into three replicates and examined in a 96 well plate.
For MØ infections, C57BL/6 mouse bone marrow derived MØ on glass coverslips were infected (MOI ~1:1) with log phase reporter bacteria grown in pH 7.0 buffered medium. Cells were fixed in 4% PFA at the specified timepoints. In experiments with labeled lysosomes, we treated resting MØ overnight with 1 ug of Alexafluor647 10000 MW Dextran (Invitrogen). Relative fluorescence was quantified using a Leica SP5 confocal microscope by measuring the GFP and mCherry signal for ~100 individual bacteria for each sample. To avoid bias, quantitated fluorescent bacteria were chosen by examining only the mCherry channel. The presented GFP induction and heterogeneity were observed in two independent experiments.
Gene knockouts were performed by gene replacement with the hygromycin resistance (HygR) cassette via homologous recombination and counterselection using an RpsL CDC1551 mutant (Sander et al., 2001). For each knockout a flanking region of approximately 1000 bp was PCR amplified and cloned into the plasmid pYUB854 in the corresponding regions flanking the HygR cassette. Gene replacement plasmid DNA was UV irradiated (Hinds et al., 1999) and electroporated into CDC1551 rpsL mutants that are streptomycin resistant. Double crossover events were selected for by recovering colonies that were resistant to both hygromycin and streptomycin. Genomic DNA from putative transformants was isolated and screened for the correct gene replacement by PCR, using primers internal to the HygR cassette and external to the cloned flanking sequence. Positive PCR products were sequenced to confirm the junctions. Successful knockouts were further confirmed by determining the absence of transcript via qRT-PCR (Figure S3B) and appropriate patterns of gene expression in our microarray analyses (Figure 7B and 7E). Complemented strains were generated by cloning fragments containing the entire aprABC locus including the native promoter region, or the aprA gene and promoter region alone, into the integrative vector pMV306. The complementation constructs were transformed into the ΔaprA and ΔaprABC locus mutants and complementation achieved by integration of these fragments at the attB site.
For lipid analysis bacterial cultures were grown in 20 mL of pH 7.0 or pH 5.7 buffered 7H9 medium in vented standing T-75 tissue culture flasks. For radiolabeling experiments we added 4 μCi of [1,2 14C] sodium acetate. Cultures were started at an OD of 0.05 and grown for 6 weeks. Bacteria were pelleted, washed in PBS and the lipids were extracted twice in 2:1 chloroform:methanol and Folch washed. 14C incorporation was determined by scintillation counting of the total extractable lipids.
To analyze apolar lipid species we spotted 10000 counts per minute (CPM) of the lipid sample at the origin of a 100 cm2 and 400 cm2 silica gel 60 aluminum sheets for 2D and 1D TLC, respectively. The 2D TLC was developed in the first dimension with petroleum ether:ethyl acetate (98:2, v/v) solvent system and in the second dimension with petroleum ether:acetone (98:2, v/v). To separate TAG for quantification (Figure 6D) and mass spectrometry, lipids were analyzed by TLC with the hexane:diethyl ether:acetic acid (80:20:1, v/v/v) solvent system. Radiolabelled lipids were detected and quantified using a phosphor screen and a Storm imager.
To prepare samples for mass spectrometry, Band 1 was isolated from WT and ΔaprABC mutants grown at pH 7.0, while Bands 2 and 3 were isolated from the phoP::Tn mutant grown at pH 5.7. The lipids were visualized on TLC plates using iodine vapors. Bands in the silica gel were scraped from the plates and lipids were extracted twice in chloroform:methanol (2:1). MALDI-TOFMS profiling of lipids was performed on a Voyager DE STR mass spectrometer (PerSeptive Biosystems, Framingham, MA, USA) equipped with a 337 nm nitrogen laser and delayed extraction. Analyses were carried out in the reflector mode at a mass range of m/z 500–3000 with an accelerating voltage of 20 kV and a delay time of 300 ns. The instrument was externally calibrated. A low-mass gate value of m/z 500 was selected to avoid saturation of the detector. α-cyano-4-hydroxycinnamic acid at 10 μg μl−1 in 70% ACN-0.1% TFA was used as a matrix. The final mass spectra were from an average of 5–10 spectra, in which each spectrum is a collection from 200 laser shot. The lipid structures were determined by multiple-stage mass spectrometry using a Finnigan LTQ linear ion-trap (Thermo Electron Corporation, CA) equipped with a Finnigan ESI source and Xcalibur software (Hsu et al., 2007).
We are grateful to Tanya Parish for sharing the pCherry3 plasmid and the TARGET group at Colorado State University for providing the the phoP::Tn mutant. We thank Brian VanderVen for insightful discussions and critical reading of the manuscript and TBK Reddy for submission of the microarray data to the GEO and TB databases. This research was supported by fellowship support to RBA from the New York Community Trust and the NIH NIAID (AI081482) and a grant to DGR NIAID (AI067027). This work was also supported as part of collaborative project awarded to JoAnne Flynn, University of Pittsburgh by the Bill and Melinda Gates Foundation.