Characterization of a Clp Protease Gene Regulator and the Reaeration Response in Mycobacterium tuberculosis

Mycobacterium tuberculosis (MTB) enters a non-replicating state when exposed to low oxygen tension, a condition the bacillus encounters in granulomas during infection. Determining how mycobacteria enter and maintain this state is a major focus of research. However, from a public health standpoint the importance of latent TB is its ability to reactivate. The mechanism by which mycobacteria return to a replicating state upon re-exposure to favorable conditions is not understood. In this study, we utilized reaeration from a defined hypoxia model to characterize the adaptive response of MTB following a return to favorable growth conditions. Global transcriptional analysis identified the ∼100 gene Reaeration Response, induced relative to both log-phase and hypoxic MTB. This response includes chaperones and proteases, as well as the transcription factor Rv2745c, which we characterize as a Clp protease gene regulator (ClgR) orthologue. During reaeration, genes repressed during hypoxia are also upregulated in a wave of transcription that includes genes crucial to transcription, translation and oxidative phosphorylation and culminates in bacterial replication. In sum, this study defines a new transcriptional response of MTB with potential relevance to disease, and implicates ClgR as a regulator involved in resumption of replication following hypoxia.


Introduction
A hallmark of infection with Mycobacterium tuberculosis (MTB) is the organism's ability to survive for months to decades in an asymptomatic state, before reactivating in a subset of infected individuals to cause frank disease. Roughly 1/3 of all people worldwide are thought to harbor MTB in a clinically latent state, and 2-10% of these individuals will reactivate during their lifetimes [1,2,3]. The threat posed by latency and reactivation is emphasized by the markedly higher 5-10% annual risk of TB disease observed among individuals co-infected with HIV and MTB [4,5,6].
The environmental cues that MTB recognizes during latency and reactivation are poorly characterized. Of these, however, oxygen tension may be the best understood [7]. Oxygenation and mycobacterial growth rate are intimately linked, both in vitro and in vivo [8,9,10]. Hypoxia is relevant during infections since in animal models with a similar course of disease to that seen in humans, such as the macaque, granuloma oxygen tension is quite low [11,12], and human granulomas without airway communication are hypoxic [13]. In vitro studies show that hypoxic bacilli halt replication, shift to the glyoxylate cycle, and increase nitrate reduction [14,15]. In this state MTB requires NAD and ATP synthesis, and maintenance of the proton-motive force [16,17], indicating that the bacteria remain metabolically active despite halted replication. Underscoring these adaptations are two distinct transcriptional responses: the initial hypoxic response controlled by dosR [18,19,20], and the enduring hypoxic response or EHR [21].
In contrast to the considerable efforts devoted to elucidating the MTB response to hypoxia, the return to favorable growth conditions is poorly studied. Here we exploit a simple culture model to investigate mechanisms by which MTB resumes growth following reaeration. Prior to replication, MTB upregulates a selection of genes indicating reversal of the adaptations employed during long term hypoxia. These genes encode proteins involved in transcription, translation, cell wall modification and oxidative phosphorylation, and were earlier repressed by the transition from log phase to hypoxia. In addition, our data also reveal a subset of genes induced during reaeration relative to both hypoxia and log phase. This transcriptional profile, which we call the Reaeration Response, is enriched in genes involved in protein degradation and refolding. We also demonstrate that the Reaeration Response transcription factor Rv2745c directly regulates the Clp proteases.

Bacterial Strains and Growth Conditions
H37Rv (ATCC 27294) was grown at 37uC in Middlebrook 7H9, 0.05% Tween, 0.2% glycerol and ADC (Becton Dickinson). Stocks were expanded from frozen aliquots within two weeks of experiments. Defined hypoxic assays were performed as previously described [20]. For reaeration, we transferred hypoxic cultures into roller bottles (5:1 head space ratio) and incubated with rolling at 37uC. Bacteria were enumerated by CFU and most-probable number (MPN) analysis [22]. Alternatively, bacteria were pelleted at 2000 xg for 5 minutes, frozen on dry ice and stored at 280uC for RNA extraction.

RNA Extraction and Purification
RNA was extracted from cell pellets as previously described [20]. After precipitation, RNA was purified using an RNeasy kit (Qiagen) as recommended by the manufacturer.

Microarray Analysis
Microarray analysis was performed using arrays provided by JCVI/PFGRC under the NIAID contract N01-AI-15447 using published protocols [23]. Arrays were scanned and spots quantified using Genepix 4000B with GenePix 6.0 software. Data were exported to Acuity 4.1, with spots 'Not Found' by the GenePix algorithms excluded from downstream analysis. Identical arrays were analyzed using Statistical Analysis of Microarrays (SAM) [24]. Unless otherwise noted, we defined significance as a mean log 2 fold change of 1 with a false discovery rate (FDR) of ,0.5% in a minimum of three arrays. All data is MIAME compliant and has been deposited in the NCBI Gene Expression Omnibus [25]and are accessible through GEO Series accession number GSE21590 (http://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc = GSE21590).
Quantitative Real-Time PCR cDNA synthesis and real time PCR were performed as described [21], with each sample normalized to sigA expression level.

Western Blot
Lysates were made by washing and resuspending MTB in 50 mM sodium phosphate buffer pH 8.0 300 mM NaCl 10% glycerol with 1X protease inhibitors (Sigma) then shaking with Lysing Matrix B (Qbiogene) in a Fastprep 120 homogenizer (Qbiogene) and filtering (0.2 mm). SDS-PAGE was performed with 5 ug of lysate per lane, transferred to nitrocellulose then blocked, incubated with primary antibody (anti-ClpP and anti-PtpB rabbit antibodies diluted 1:5000 and 1:200 respectively), followed by secondary antibody (Licor Goat anti-rabbit IR680 diluted 1:5000). Membranes were scanned at 700 nm, quantified and normalized using the Licor Odyssey imaging system.

Rv2745-S Expression and Purification
The Rv2745c coding sequence was PCR amplified and cloned into pET50b (Novagen), using primers designed to encode two mutations V111D and A112D at the C terminus. Protein expression was induced in E.coli BL21(DE3) with IPTG for 3 hours, and cells were lysed with Bugbuster (Novagen). Rv2745c was affinity purified using His-Select Ni-NTA agarose (Sigma), eluted with imidazole (Sigma) and dialyzed into HRV3C protease buffer (Novagen). The NusA fusion protein was removed using HRV3C protease (Novagen) overnight at 4uC, then protease and fusion protein were removed by affinity purification.
Binding reactions were performed with 200 ng protein and 2 ug of MTB lysate in 25 mM sodium phosphate pH 7.5, 150 mM NaCl, 1 mM EDTA, 2.5 mM MgCl 2 , 1 mM DTT, 1 mg/ml BSA, 10% glycerol containing 1 mg dI:dC DNA and competitor DNA as indicated for 15 minutes at room temperature. 200pmol Cy5.5 labeled probe was added and incubated at 37uC for 5 minutes then 4uC for 25 minutes. Samples were electrophoresed in 5% native polyacrylamide gels for 1 hour at 10 V/cm and scanned by Licor Odyssey.

Results
The M. tuberculosis Reaeration Response To begin modeling reactivation, we characterized the transcriptional response of reaerated MTB. Starting with the previously reported defined hypoxia model [18,21], we transferred hypoxic cultures into roller bottles (5:1 head space ratio, 15 rotations per minute) to promote rapid re-equilibration with atmospheric air, and growth at 37uC was assessed. Reaeration of MTB cultures was performed after 7 days of hypoxia, because earlier observations indicated that MTB transcription at 7 days closely resembles that at time points up to 3 weeks in this system (Rustad and Sherman, unpublished). Contrary to earlier reports [26], we found only very modest drops in viability following rapid shifts in oxygen tension ( Figure 1).
Hypoxia-Repressed, Reaeration-Induced Genes. We hypothesized that reaerating MTB might express a distinct transcriptional response that could provide insight into the process by which MTB resumes replication following a period of bacteriostasis. Accordingly, we performed transcriptional profiling of reaerating MTB by whole genome microarray analysis. After 7 days hypoxia (time 0 for reaeration) and at subsequent times, samples were processed for microarray analysis as described [20,21]. Reaeration induced a wave of transcription, which peaked at 24 hours with 370 genes induced at least two-fold ( Figure 2A, Table 1 and Table S1). Not surprisingly, many induced genes are among those repressed during the transition from log phase to prolonged bacteriostasis [21] This transcriptional profile reflects the significant modifications MTB makes in order to survive hypoxia. MTB's return to a log phase transcriptional response represents the reversal of these bacteriostasis-associated adaptations. As cultures adapted to reaeration and neared replication, there was a progressive increase in the proportion of reaeration-induced genes that are originally hypoxia-repressed ( Figure 2A). Analysis of these genes indicates that during reaeration MTB induces the transcriptional and translational machinery, including synthesis of ribosomes, RNA polymerase, purines, pyrimidines, and amino acids (Table S1). MTB undergoes significant cell wall remodeling during the adaptation to bacteriostasis [27], and the reversal of this process is likely necessary in order for cell division to proceed. In this context, we note renewed transcription of the dTDP-rhamnose synthesis genes rmlB and rmlC, and peptidoglycan-synthesis genes including murC, murF, murX, and ponA1. Hypoxic repression and reaeration-induced transcription of genes for ATP synthase, cytochrome C oxidase and reductase, and NADH dehydrogenase suggest resumption of oxidative phosphorylation in reaerated MTB. Other genes that fit this transcriptional profile include the serine-threonine kinase pknB, virulenceassociated esxA and esxB, and the cell division associated wag31. Consistent with the idea that expression of these hypoxia-repressed genes represents a return to log phase metabolism, our analysis shows that 28-50% of the hypoxia-repressed, reaeration-induced genes at any time point are predicted to be essential for growth of MTB in vitro [28].
Reaeration-Specific Genes. In addition to this progressive upregulation of log phase-associated genes, we also noted a subset of reaeration-induced genes that had not previously been repressed during hypoxia. These reaeration-specific genes, which we defined as those induced at least two-fold during reaeration relative to both late-stage hypoxia and log phase, were evident at the earliest time point (1 hour) and peaked by 12 hours. At these early times, reaeration-specific genes comprised about half of all induced loci ( Figure 2A, gray bars).
There was substantial overlap in the particular reaerationspecific genes induced at each time point. To facilitate analysis of this transcriptional response, we defined the Reaeration Response as those reaeration-specific genes induced at both six and twelve hours, the two time points where the number of reaeration-specific genes is highest (Table 1). This list of 103 genes includes 8 transcription factors including the mce2 operon repressor mce2R, ferric uptake regulator furB, cell division regulator whiB2, moxR3 and Rv2745c. Reaeration Response genes involved in metabolic adaptation include a sugar transporter (Rv1235 and Rv1236) predicted to be essential for MTB survival in macrophages and mice [29][30], the triacylglycerol lipase lipY (Rv3097c), recently shown to cleave triacylglycerides upon reaeration, and a ribonucleotide reductase (Rv3048c) essential for aerobic growth [31]. The Reaeration Response also includes seven genes encoding protease or chaperone proteins, suggesting a need to degrade or refold proteins. Among these proteases are clpP1, clpP2 and their associated ATPase clpC1. Recent reports by Mehra et al. and Barik et al. have suggested that these Clp proteases may be involved in the response of M. tuberculosis to oxidative and surface stress, under the control of the putative transcription factor Rv2745c [32,33]. These reports showed that transcript levels of clpP1, clpP2 and clpC were modulated by altered expression of Rv2745c. However, these studies did not address whether regulation of clp genes was directly controlled by Rv2745c. We sought to determine whether Rv2745c  Table 1. Genes induced at 6 hours and 12 hours of reaeration, in comparison to 7 days hypoxia (Num .1, q,0.5).

Characterization of Transcriptional Regulator ClgR (Rv2745c)
The Rv2745c protein shows high homology (54-60% identity) with ClgR, a protein that activates transcription of Clp proteases in the related actinomycetes Streptomyces coelicolor and Corynebacterium glutamicum. As part of a self-regulation loop, ClgR both induces and is degraded by the Clp proteases in these organisms [34,35,36,37]. As mentioned above, Rv2745c is part of the MTB Reaeration Response (Table 1). It is up-regulated in late hypoxia, and is further induced upon reaeration ( Figure 2B). The two MTB Clp protease genes responsible for Clp protease substrate specificity, clpC1, clpC2, and clpX, are also up-regulated during reaeration ( Figure 2B). Analysis of the promoter region of the clpP1P2 operon reveals a perfect match with the Corynebacterium ClgR consensus binding site (WNNWMGCYNNNRGCGWWS) ninety-six base pairs upstream of the ClpP1 start codon [36]. The clpC1 promoter contains a copy of this consensus site with a single mismatch. Based upon the strong homology to ClgR, the consensus binding site and the expression pattern, we suspected that Rv2745c directly regulates clp protease gene expression in MTB. Recent reports have suggested that modulating transcription of Rv2745c or the M. smegmatis homologue affect transcript levels of the clp proteases [32,33]; however, these studies did not assess whether or not this is a direct effect.
To test whether Rv2745c regulates MTB clp genes, we employed a tetracycline-inducible overexpression system. However, as noted above ClgR is a Clp substrate in other bacteria. In Streptomyces, cleavage depends upon the ClgR C-terminus, which bears strong similarity to the MTB Rv2745c C-terminus. Since induced degradation of the regulator would confound results, we generated an altered Rv2745c (Rv2745-S, for stabilized) in which the final two amino acids (V111, A112) were mutated to aspartates. This mutation prevents degradation of ClgR by the Clp proteases in Streptomyces [35]. In the presence of anhydrotetracycline (AHT), Rv2745-S mRNA copy number increased 25fold compared to the strain without AHT ( Figure 3A). We used this inducible system to assess whether overexpression of Rv2745-S resulted in increased clpP1P2 transcription. In the presence of excess Rv2745-S mRNA, ,4-fold more clpP1P2 mRNA was detected, consistent with the expression level of these genes during reaeration (Figures 2B and 3B). Furthermore, levels of the ClpP1 and ClpP2 proteins in MTB were also increased by Rv2745-S overexpression ( Figure 3C and D, ClpP antibody recognizes both subunits).
We assessed direct interaction of the Rv2745c protein with the clpC1 and clpP1P2 promoters. Due to insolubility of the wild-type protein, we used the Rv2745-S described above. Purified Rv2745-S protein was incubated with DNA in an electrophoretic mobility shift assay (EMSA). A shift of the clpC1 promoter was observed in the presence of Rv2745-S, and DNA-binding could be outcompeted by 25-fold excess unlabeled probe, but not by a non-specific competitor (Figure 4). Similar results were obtained with the clpP1P2 promoter fragment (data not shown). To assess if overexpression of MTB clpP1P2 is really due to transcriptional activation by Rv2745-S, we sought to interrupt the DNA binding activity of this protein. The Rv2745c sequence predicts a helixturn-helix DNA-binding motif [34,38], and by overlaying the Rv2745c protein sequence on a recently published structure of the Corynebacterium glutamicum ClgR protein [38,39,40] we identified residues with predicted direct DNA interactions. We used sitedirected mutagenesis to create proteins with alterations in the putative helix-turn-helix region, R25D, R43D and L24SR25D, and assessed the ability of these mutant proteins to act as transcription factors. Overexpression of these mutant proteins occurred on addition of anhydro-tetracycline ( Figure 3A, L245SR25D data shown are representative of results for all three mutants), but a concomitant increase in transcription of clpP1P2 was not observed (Figures 3B and C). In addition, EMSA analysis demonstrated that the mutants failed to bind clpC1 and clpP1P2 promoter DNA (Figure 4 and data not shown). From these results, as well as the high degree of homology to clgR in related organisms, we conclude that Rv2745c is the MTB clgR orthologue.

Discussion
While the MTB adaptations to hypoxic bacteriostasis have received considerable attention, little is known about the mechanism by which bacilli reactivate. Early studies suggested that growth-arrested MTB responds to re-oxygenation with immediate RNA synthesis, while DNA synthesis and replication are delayed [10,14]. More recent work revealed that triacylglycerols stored during hypoxia are cleaved during reaeration of Mycobacterium bovis bacillus Calmette-Guerin by lipY (Rv3047c) and other lipases [41]. The resuscitation promoting factors (RPFs) encoded by MTB also appear to have a role in chronic infection and reactivation in a mouse model [42,43,44,45].
To gain more mechanistic insights, we developed a model in which MTB resumes replication following reaeration under defined conditions in vitro. Replication is preceded by two distinct transcriptional phases. The latter of these, steadily building in number as MTB nears replication, includes genes involved in transcription, translation, oxidative phosphorylation and fatty acid synthesis (Table S1). Since genes involved in each of these functions are repressed as MTB adapts to hypoxia [21], the bacillus would be expected to resume their expression while adapting to reaeration and preparing for log phase growth.
Notably, approximately one-third of the genes in this late prereplication transcriptional profile are predicted to be essential for MTB survival in vitro [28]. The increased transcription we observed for genes involved in purine and pyrimidine synthesis, RNA polymerase, amino acids synthesis, and ribosome components agrees with previous estimates that levels of protein, RNA and ribosomes increase in actively growing as compared to slower growing bacilli [46]. In addition, it was shown recently that MTB adapted to a non-replicating state in response to hypoxia or nutrient deprivation has a substantially lower ATP content than replicating MTB [16,47]. These data suggest that metabolically slowed MTB synthesize fewer of these crucial components of transcription and translation, possibly due to energy limitations. Perhaps during bacteriostasis, these adaptations are part of a strategy to conserve energy for survival.
The earlier transcriptional adaptation we observed, the Reaeration Response, is a set of ,100 genes induced when compared to both long-term hypoxia and log-phase ( Figure 2A and Table 1). We hypothesized that it would be necessary for MTB to undergo metabolic changes prior to replication and that an identifiable transcriptional modulation would characterize this shift. Consistent with this possibility, the Reaeration Response includes several genes involved in reordering bacterial metabolism, including a ribonucleotide reductase gene essential for aerobic growth along with genes involved in fatty acid metabolism, sugar transport, and a triacylglycerol lipase (lipY), which may facilitate utilization of stored triacylglycerols during reaeration [31,41]. In addition, enrichment of chaperone proteins and proteases in the Reaeration Response suggests an effort by MTB to repair or replace proteins and cellular components damaged by either hypoxia and/or re-introduction of oxygen. Within the Reaeration Response, we note a two-to ten-fold enrichment of genes upregulated in response to other stresses including diamide, heat shock, acid stress, and macrophage infection [48,49,50,51], suggesting that MTB senses and responds to stress during reaeration.
The Reaeration Response also includes eight putative transcription factors. One of these is Rv2745c, a protein similar to the ClgR regulators of related actinomycetes [36,37,52]. In these species, ClgR binds to a conserved sequence upstream of clpP and activates transcription of this gene. We demonstrate that Rv2745c is a ClgR homologue that directly activates transcription of Clp  proteases in MTB. While this work was ongoing, two independent studies reported modulation of clpP gene transcription upon alteration of Rv2745c mRNA levels [32,33]. The first study demonstrated that inducible overexpression of Rv2745c led to higher transcript levels of clpP1, clpP2 and clpC1 [33]. The second study showed that depletion of the Mycobacterium smegmatis homologue of Rv2745c by antisense prevented upregulation of clpP1,clpP2, and clpC1 transcription upon exposure to vancomycin [32]. Our data confirm and extend these findings, showing that ClpP protein levels as well as transcript levels increase with Rv2745c overexpression (Figure 3C and D), and also that Rv2745c directly interacts with the clpC1 and clpP1P2 promoters (Figure 4 and data not shown). Interestingly, Rv2745c transcription in response to diamide stress is dependent on the sigma factor sigH [33], which we previously have shown is upregulated during extended hypoxia [21]. These findings suggest shared mechanisms in MTB responses to stress conditions likely to be relevant in vivo. It will be interesting to decipher which other components of the MTB Reaeration Response are shared with other stress responses, and which are specific to increased oxygenation.
Clp proteases and their regulators play crucial roles in maintaining and overcoming replication checkpoints in other bacteria. For example, the Caulobacter crescentus regulator CtrA both transcribes a set of genes involved in replication and binds to the chromosomal origin, silencing initiation of replication, and degradation of CtrA by the Clp proteases is needed for replication to begin [53]. In S. lividans, ClgR overexpression delays a morphological transition that the Clp proteases accelerate, and stabilizing ClgR to prevent its degradation intensifies this effect [37,54].
The set of genes regulated by ClgR is not yet fully characterized. A pattern search for the conserved ClgR binding motif within MTB shows matches upstream of cysK1, Rv2224c, Rv0037c, hemD, and PPE55. Allowing a single mismatch yielded a list of 101 genes with a putative ClgR binding site, including the ATPase clpC1 and clgR itself. The transcriptional patterns of these genes during reaeration are varied, suggesting that this process is mediated by more than one regulator. Seven other transcription factors were identified in the Reaeration Response, and their regulons remain to be determined.
These studies introduce a model for probing the mechanisms of reactivation of MTB in vitro. Here, we describe the transcriptional response of MTB to reaeration, and identify the Clp proteases and their transcription factor ClgR as upregulated during the transition of MTB from bacteriostasis to growth. Further exploration of this model may reveal other players in MTB's adaptation to an active, replicating state, and it will be interesting to assess the extent to which induction of the Reaeration Response occurs in other strains, including clinical isolates. In addition, while the adaptations of MTB to hypoxic/reaerated microenvironments are likely central to the pathogen's success, the in vivo relevance of the Reaeration Response remains to be determined.
Determining the means by which the bacterium adapts to these changing conditions will open doors to potential new drug therapies and diagnostics. The roles of protein degradation and refolding in MTB stress response are just beginning to be revealed. Recently, the MTB proteasome was shown to have a role in adaptation to nitrosative stress, and compounds were identified which kill MTB by selectively targeting the MTB proteasome over the human counterpart [55,56]. The Clp proteases might prove a similarly viable new drug target. Additionally, genes of the Reaeration Response could contribute to the development of new and effective diagnostics. Identifying which latently-infected individuals are most likely to suffer from reactivation of TB is an important priority in TB research [57]. Further studies are needed to explore whether immunological recognition of Reaeration Response proteins correlates with the likelihood of reactivation among MTB-infected individuals.