CsoR Is Essential for Maintaining Copper Homeostasis in Mycobacterium tuberculosis

Mycobacterium tuberculosis, a pathogen infecting one third of the world population, faces numerous challenges within the host, including high levels of copper. We have previously shown that M. tuberculosis CsoR is a copper inducible transcriptional regulator. Here we examined the hypothesis that csoR is necessary for maintaining copper homeostasis and surviving under various stress conditions. With an unmarked csoR knockout strain, we were able to characterize the role of csoR in M. tuberculosis as it faced copper and host stress. Growth under high levels of copper demonstrated that M. tuberculosis survives copper stress significantly better in the absence of csoR. Yet under minimal levels of copper, differential expression analysis revealed that the loss of csoR results in a cell wide hypoxia-type stress response with the induction of the DosR regulon. Despite the stress placed on M. tuberculosis by the loss of csoR, survival of the knockout strain was increased compared to wild type during the early chronic stages of mouse infection, suggesting that csoR could play an active role in modulating M. tuberculosis fitness within the host. Overall, analysis of CsoR provided an increased understanding of the M. tuberculosis copper response with implications for other intracellular pathogens harboring CsoR.


Introduction
Mycobacterium tuberculosis is one of the world's most successful bacterial pathogens, infecting approximately one third of the human population. Despite widespread vaccination and the existence of antibiotic therapies, this causative agent of tuberculosis leads to over 1 million deaths each year [1]. Part of the success of this pathogen is attributed to its ability to quickly adapt and survive within harsh host microenvironments. The stress conditions M. tuberculosis must face within the host include reactive nitrogen and oxygen species [2], low pH [3], and hypoxia [4]. Recently, copper has been recognized as an additional weapon in the host macrophage's arsenal as it localizes high levels of copper to the mycobacterial phagosome [5]. This finding led our group to investigate the M. tuberculosis response to copper stress revealing a set of 30 genes responsive to copper and the damage it can cause [6]. Among these genes were two that encode for copper-responsive transcriptional repressors-paralogs now identified as ricR and csoR-the latter of which we continue to characterize in this report.
The bactericidal capabilities of copper have been known for some time [7]. The mechanisms of copper damage include production of oxidative stress through Fenton reactions [8], displacing metal cofactors in proteins, and destabilization of Fe-S clusters [9]. Indeed, high physiological levels of copper have been demonstrated to be bactericidal for M. tuberculosis [5,6]. Conversely, copper serves as a cofactor of metalloenzymes and is necessary in trace amounts for various cellular activities [10]. In M. tuberculosis such metalloenzymes include cytochrome c oxidase and superoxide dismutase [11], thus the role of copper in metabolism and the copper stress response must be carefully regulated. For example, under hypoxic conditions, which increase copper toxicity, M. tuberculosis makes use of copper-independent cytochrome bd oxidase [12]. This effectively reduces the requirement for cytochrome c oxidase, and therefore copper, under conditions where copper poses the greatest threat to the cell.
To better understand how the copper stress response is managed, we began to characterize one M. tuberculosis copper-induced regulator, CsoR. The crystal structure of CsoR revealed that the protein forms a homodimer with each monomer binding one molar equivalent of Cu (I) [13]. In its apo form, CsoR binds the promoter of its own copper-sensitive operon (cso), however upon binding copper, repression is released inducing expression of the cso in a graduated manner leading to the export of copper ions. Here we investigated the importance of the negative regulatory role of csoR in M. tuberculosis copper homeostasis. Our initial results, examining the ΔcsoR strain under copper stress and during early chronic murine infection, suggested that under certain conditions M. tuberculosis may have an advantage upon suppressing the expression of csoR. Closer examination of the transcriptome of M. tuberculosis lacking csoR, however, revealed a hypoxic stress response during growth in copper-free media. This response may have better prepared the bacilli for survival during the early chronic stage of murine infection. We expect that M. tuberculosis must strike a balance when it comes to handling copper stress and that CsoR is integral, though perhaps not alone, in this role. Furthermore, in this role CsoR may make an important contribution during entry to the chronic phase of tuberculosis.

Generation of M. tuberculosis csoR constructs
To create a knockout mutant of csoR in the virulent, wild type M. tuberculosis strain H37Rv, homologous recombination was used to replace 5 base pairs (bp) of the 360bp coding region of csoR with a hygromycin resistance cassette (hyg R ), an~2.4kB region, using the cosmid pYUB854 ( Fig 1A) [14]. The mutant was confirmed by sequencing and with Southern blot, which showed a 2.4kB shift between bands from wild type or mutant strains ( Fig 1B). However, reverse transcriptase PCR (RT-PCR) testing for expression of the downstream genes revealed loss of transcription of all three downstream genes in the cso: Rv0968, a conserved hypothetical protein; ctpV, coding for a copper-exporter [15]; and Rv0970, a conserved membrane protein of unknown function (Fig 1C). In order to create a nonpolar ΔcsoR strain, an additional step was taken to remove the hygromycin cassette. The vector pYUB870 [14] was electroporated into ΔcsoR::hyg R removing hyg R and leaving behind a 150 bp insertion sequence at the 14 th codon of csoR. The resulting unmarked mutation led to a frame shift which introduced 6 stop codons either at the insertion site or within the remainder of the csoR gene. RT-PCR results confirmed transcription of the remaining members of the cso in the unmarked mutant strain (Fig 1C). A complementation strain, ΔcsoR::csoR, was constructed by incorporating csoR under its own promoter into the ΔcsoR genome using the integrative vector pMV306.
To ensure that no other possible disruptions of the genome in or outside of the cso were contributing to the examined phenotypes, whole genome sequencing was carried out comparing M. tuberculosis ΔcsoR with its wild type H37Rv parent strain. A total of 95 single nucleotide polymorphisms (SNPs) were found in M. tuberculosis ΔcsoR when compared to the published H37Rv reference [16]. Of these, 38 SNPs corresponded to those identified as sequencing errors in the original reference [17] and 49 of the remaining SNPs were found in the sequenced parent strain. In all 8 unique SNPs were found in M. tuberculosis ΔcsoR differentiating it from the parent strain. Two of these SNPs had been previously identified as they were annotated at the start of the insertion sequence in csoR. Five of the other six SNPs were found to be non-synonymous. One SNP each was found in PPE5 (Rv0304c) and PPE55 (Rv3347c). Three more, including the one synonymous SNP, were found in PE_PGRS9 (Rv0746). The last was found in a probable cyclase, Rv2435c. When M. tuberculosis ΔcsoR was searched for larger genomic variations, only the 150bp scar at the csoR deletion site was identified. Given the distance of these variations from the cso and the function of the genes they target, it seems unlikely that the observed phenotypes could be due to anything other than the deletion of csoR.

Survival of ΔcsoR is enhanced under copper stress
To test the growth kinetics of M. tuberculosis ΔcsoR during copper stress, wild type and mutant strains were grown in copper-free Sauton's liquid media with no added copper or with 50 or 500μM CuCl 2 . As expected, both strains showed similar levels of growth in untreated cultures suggesting no generalized growth defects (Fig 2A), and showed a decline in growth in the presence of 500μM CuCl 2 (Fig 2B). While the gradual decline of the ΔcsoR strain in 500μM CuCl 2 was similar to wild type at early time points, by day 15 the culture of the knockout strain maintained its total CFU/ml while the wild type culture continued to decline (Fig 2B). The difference at this time point was significant (P = 0.002) and observed across two separate experiments. Growth of the two strains in 50μM CuCl 2 did not vary from that of untreated cultures (S1 Fig).
To investigate the possibility that csoR and its regulon may play a greater role during later stages of growth as the culture transfers into stationary phase, the growth curves were repeated To knock out csoR, the first gene of 4 in its operon, a 2.4kB hyg R cassette (black bar) was inserted near the 5' region of csoR. Grey arrowheads indicate primers flanking the genomic region amplified to be used as a probe for Southern blot. The BamHI cut site within the cso used for Southern blot is shown. (B) Southern blot targeting the csoR region after BamHI digestion of wild type (WT) or ΔcsoR:: hyg R genomic DNA (gDNA) demonstrating insertion of the hyg R cassette. (C) The polar nature of the different inoculating from cultures that had reached stationary phase, rather than from actively growing cultures as was done above. Interestingly, the difference in survival between M. tuberculosis wild type and ΔcsoR, while not affected in untreated cultures (Fig 2C), differed dramatically in the presence of 500μM CuCl 2 ( Fig 2D). This difference was noted after day 4 when wild type survival fell two logs lower than the mutant. By day 8 it remained near the limit of detection, 10 CFU/ml, while ΔcsoR survival declined much more gradually, remaining above 10 4 CFU/ml through day 15, indicating the ability of ΔcsoR to adapt to increased Cu levels.
After testing survival under two concentrations of copper, we sought to investigate the impact of copper on the growth of M. tuberculosis ΔcsoR at a higher resolution. This was done by employing a modified microplate Alamar blue assay (MABA) [18,19]. Growth inhibition of M. tuberculosis wild type H37Rv, ΔcsoR, and ΔcsoR::csoR by two-fold dilutions of CuCl 2 ranging from 16μM to 4000μM was tested (S2 Fig). Microplates were inoculated with cultures growing at mid-log phase and read after approximately 10 days of growth. Susceptibility of both wild type and complement strains was equivalent, both showing a minimum inhibitory concentration (MIC) of 250μM CuCl 2 . The mutant, on the other hand, showed an MIC of 500μM CuCl 2 . These data confirmed the idea that loss of csoR improves survival of M. tuberculosis under copper stress.

Absence of csoR improves survival of M. tuberculosis during the early stages of chronic infection
To gain a broader understanding of the role of csoR during M. tuberculosis infection, BALB/c mice were infected with H37Rv wild type or ΔcsoR by aerosol infection. The groups infected with wild type received on average 370 CFU per animal while ΔcsoR infected mice received closer to 1500 CFU per animal; however this difference in inoculum did not reach significance. In fact, by week 2 ΔcsoR showed on average slightly less colonization than wild type. Despite this early trend, by weeks 4 and 8 ΔcsoR showed a significant increase in survival compared to wild type (P = 0.015 and P = 0.005 respectively). This increase, however, was not sustained long term, as by week 25 no significant difference in survival between wild type and knockout was apparent (Fig 3A). While the colonization of the wild type strain was consistent from week 8 to week 25, ΔcsoR showed a significant drop from week 8 to week 25 (P = 0.007), suggesting that the improved growth rate during early chronic stage was not sustainable throughout the duration of chronic infection. Unlike the lungs, no significant difference in survival was seen between wild type and ΔcsoR when spleen and liver were assayed, although mean CFU/g were consistently higher in the mutant strain in both organs (S3 Fig). Body weight of the mice did not vary significantly between groups after week 2 of the experiment, when mice infected with ΔcsoR were on average 1.2g lighter than those infected with wild type (P = 0.005) (S3 Fig). Histopathology of the lungs, spleen, and liver ( Fig 3B) was scored by a trained pathologist. For both strains, lymphocytic infiltration and granulomatous reaction in the lungs worsened as the infection progressed, while minimal pathology was seen in the spleen and liver. No significant difference was noted between strains at any time point for any of the tissues (S4 Fig). In sum, csoR conferred a moderate disadvantage to wild type M. tuberculosis during the early weeks of murine infection, but may be required for long term survival.

The global transcriptional response of M. tuberculosis ΔcsoR
To better understand the breadth of influence csoR has directly as a regulator, and indirectly over copper homeostasis within the cell, the ΔcsoR transcriptome was interrogated and compared to that of wild type M. tuberculosis H37Rv using RNA-Seq analysis. Our previous work [13] and that concerning csoR in other species such as B. subtilis [20] has suggested that under copper stress CsoR derepresses its regulon, and therefore few transcriptional differences between wild type and csoR knockout strains would be noted when compared under copper stress. Thus, we compared the transcriptomes of both wild type M. tuberculosis and ΔcsoR grown in copper-free Sauton's media. Details of the number of reads and how they mapped can be found in S1 Table. Differential expression analysis using strict cutoffs requiring differentially expressed genes to have a fold change ! |2.0| and a false discovery rate (FDR) of 0.05 left us with 223 genes that were significantly, differentially expressed between the mutant and wild type strains (S2 Table). Of those genes 152 were induced, while 71 were repressed in ΔcsoR compared to wild type ( Fig 4A). To confirm these results, 10 genes were selected-6 induced and 4 repressed-for quantitative, real-time PCR (qRT-PCR). All 10 were confirmed to be induced or repressed in the same direction, and with similar magnitude as the RNA-Seq data (S5 Fig).
It has already been established by our group and others that csoR represses its own operon in the absence of copper [13,21]. Therefore, we first looked at the expression levels of the cso. As expected under low levels of copper, the cso was moderately induced in the ΔcsoR strain as compared to wild type ( Fig 4A; Table 1). This included the observed induction of csoR itself as sufficient remnants of the mutated csoR gene remained for successful RNA-Seq analysis. As CsoR is an established copper-responsive regulator, the list of differentially expressed genes in ΔcsoR compared to wild type was interrogated for the 30 genes known to be responsive to copper in M. tuberculosis as determined by previous microarray analysis [6]. Very few genes overlapped; in fact, the only overlapping gene outside of the cso, Rv0848, was repressed in the ΔcsoR strain despite being induced under copper stress (Fig 4A and 4B; Table 1). These results suggest that csoR may indeed only directly control expression of its own operon as previously suggested by Festa, et al [21]. While no significant changes in the majority of copper responsive genes were noted, most outside the cso (n = 18) were down regulated, short of our cutoff values. These include another copper sensing regulator, ricR (-1.5, FDR = 0.18). Among the RicR regulon [21], two genes, Rv0847 and Rv0848, were significantly repressed. Additionally, several of the other genes under the control of RicR (n = 7), including mymT a copper binding metallothionein [22] and mmcO, a multicopper oxidase which is exported across the cytoplasmic membrane [23], were slightly, though not significantly down regulated (Fig 4A), and this trend was also confirmed by qRT-PCR. These results clearly demonstrate that the CsoR and RicR regulons are distinct; however deregulation of the CsoR regulon may contribute to a slight, perhaps compensatory, increase in repression of the RicR regulon.
With evidence that CsoR directly regulates only its own operon, it seemed likely that other forms of regulation were at play in the ΔcsoR strain. While no sigma factors were found to be differentially expressed, numerous transcriptional regulators were among those differentially expressed (n = 16) (Table 1). Therefore, the genes under transcriptional control of these regulators may represent a significant portion of the differentially expressed genes in our dataset.
One striking pattern was the induction of 44 out of 48 members of the dosR regulon, known for its role in response to hypoxia, nitric oxide, and dormancy [24,25] (Fig 4B). These members of the dosR regulon were among the most highly induced genes in our dataset ( Fig 4A). Another notable regulator involved in broad regulation of genes throughout various stages of hypoxia, Rv0081, was also induced [26]. This again suggests that csoR itself may not be directly controlling the expression of these genes, but rather derepression of its own regulon may be creating hypoxia-like stress conditions to which dosR, Rv0081, and other transcriptional regulators respond. Overall, our analyses suggest that CsoR represses its own operon, and absence of csoR may lead to a disruption in copper homeostasis, leading to a hypoxia or NO type stress response.

The impact of CsoR on drug resistance
During analysis of the csoR impacted genes it was noted that genes involved in the response to isoniazid (INH) were enriched, with 7 of the 23 known responders being induced. These 7 genes included several genes involved in fatty acid synthesis as well as membrane efflux protein, efpA (Table 1). Based on this finding and a previous study that showed the absence of M. Highlighted are members of the cso (yellow), members of the RicR regulon (orange), genes responsive to copper, but not part of the cso or RicR regulon (green), members of the DosR regulon (blue), and significantly differentially expressed genes as determined by an FDR ! 0.05 not included in the above groups (red). Two parallel, black lines demarcate the region outside of which differential expression values exceed our cutoff of a 2.0 fold difference between strains. Genes not meeting both cutoff values are shown as semi-transparent grey diamonds. Data points show the mean of two biological replicates. (B) Overlap of the induced csoR regulon (red) with 500 CuCl 2 inducible genes (green) and genes induced under the control of dosR (blue). Of the 152 genes induced in the ΔcsoR strain compared to H37Rv wild type, only 4 overlapped with the 24 genes induced in H37Rv wild type when exposed to copper stress at 500μM CuCl 2 . DosR inducible genes showed substantial overlap with ΔcsoR with 44 out of 48 overlapping. The diagram is area-proportional. Expression dynamics of the copper sensing operon in the absence of csoR during copper stress While we have shown that the cso is induced in the absence of csoR and copper stress, it is still unclear if CsoR deregulation alone accounts for the induction of the cso under copper stress. To better examine the dynamics of CsoR regulation of the cso during copper stress, we utilized qRT-PCR to profile the expression levels of csoR, Rv0968, ctpV, and Rv0970. Samples for analysis were taken from M. tuberculosis H37Rv, ΔcsoR, and ΔcsoR::csoR left untreated or exposed to either 50 or 500μM CuCl 2 for 3 hours. As expected, qRT-PCR analysis of all 4 cso genes showed induction of the operon in the absence of csoR (Fig 5A). This level of induction was on par with the induction of the operon under moderate copper stress (50μM CuCl 2 ) in the wild type strain, but was less than the expression levels seen in wild type under high copper stress (500μM CuCl 2 ). Expression levels of the complement strain were similar to that of the knockout. Interestingly, as the level of copper stress increased for the ΔcsoR strain, no change in cso expression was seen at 50 or 500μM CuCl 2 compared to untreated ΔcsoR, unlike in wild type (Fig 5B). Analysis of ΔcsoR::csoR on the other hand did reveal a modest rescue of the copper responsive phenotype.
To test the hypothesis that the cso may be under the control of multiple copper sensitive regulators and may not be responding to copper stress in ΔcsoR due to increased copper export, we sought to investigate the levels of copper present within each strain. Direct measures of copper such as neutron activation analysis are not reliably precise enough to detect small differences in copper ions within M. tuberculosis [15,21]. Therefore mmcO, part of the copper responsive RicR regulon, was picked as an indirect indicator of copper stress experienced by the cell [18]. Thus, samples of wild type, ΔcsoR, and complement strains under different levels of copper stress were examined by qRT-PCR analysis for mmcO. No significant difference in mmcO induction was detected between strains ( Fig 5C). As with the global transcriptome analysis, mmcO was slightly down-regulated in ΔcsoR as compared to wild type in copper-free media. This could be indicative of reduced levels of copper present in the M. tuberculosis cytoplasm in ΔcsoR in the absence of copper stress. However, once copper was added, mmcO, unlike the cso, increased drastically in the mutant strain. A similar profile was also observed when copper responsive gene mymT was analyzed (S7 Fig). These results indicated that under copper stress, copper levels between the three strains were equivalent, excluding the possibility that differences in intracellular copper were contributing to the differential expression of the cso. We cannot, however, rule out the possibility that other regulators, responsive to copper or related stress, participate in the regulation of the cso.

Discussion
Recently, great interest has been taken in the importance of copper homeostasis and its role in bacterial pathogenesis and host defense [28]. Previously, our group elucidated the regulatory mechanisms behind the copper responsive transcriptional repressor CsoR [13]. We have also shown the shuttling activity of one of its directly regulated genes, ctpV, in response to copper stress [15]. In this study, we examined the importance of this negative regulator, CsoR, to both copper and host-mediated stress in a murine model of tuberculosis. While a multifaceted role of CsoR had been previously suggested-in which CsoR has a buffering effect preventing damage from free copper [13]-our results suggest that the genes within the CsoR regulon, rather than CsoR itself, are primarily responsible for directly mitigating the harmful effects of copper stress. Consistent with this hypothesis, the survival curves of ΔcsoR at 500μm CuCl 2 are reminiscent of those seen for M. tuberculosis ΔctpV::ctpV, which overexpresses ctpV, a copper exporting member of the csoR regulon [15]. Less expected was the dramatic difference in strain survival depending on the use of a fresh, log stage inoculum or one that had been grown to stationary phase before inoculation in copper containing media. One possible cause of this difference could be a slower response to copper stress in the stationary phase cultures, exaggerating the otherwise slight copper resistant phenotype of ΔcsoR. On the other hand, it may be that the Expression of cso genes during copper stress in the absence of csoR. qRT-PCR was used to analyze expression levels of csoR, Rv0968, ctpV, and Rv0970. A schematic of the operon (not drawn to scale) is shown above the graph and the order of the genes in the operon corresponds to the order in which expression levels are graphed for each gene. (A) M. tuberculosis strains H37Rv (black), ΔcsoR (striped), or ΔcsoR::csoR (white) were exposed to CuCl 2 at 50 or 500μM or left untreated (0μM) for 3 hours. Values shown are the mean fold change between each gene and its untreated wild type counterpart after normalization to sigA expression levels. (B) A second comparison of the same samples was done showing the mean fold change between each gene after normalization to sigA expression levels at 50μM (striped) or cso has a more important role to play during dormancy than during logarithmic phase growth, particularly after lengthy exposures to copper that may allow copper to build up within the cell.
The role of the cso in dormancy is supported by the observed induction of csoR throughout the enduring hypoxic response [29]. An increased, hypoxia-specific copper response may be physiologically important as not only is copper more toxic under hypoxic conditions [30], but hypoxia stimulates macrophages to increase copper uptake [31]. In fact hypoxia and high copper levels have both been found within granulomas of guinea pigs during M. tuberculosis infection, emphasizing the need to adapt to combat both stress conditions simultaneously [4,32]. At the same time, the M. tuberculosis requirement for copper under hypoxia is decreased as copper-independent cytochrome bd oxidase is induced, reducing the need for copper-dependent cytochrome c oxidase [12]. The relationship of csoR, copper stress, and hypoxia-induced dormancy is currently being investigated further.
As the host environment [32], particularly the mycobacterial phagosome [5] is known to harbor high levels of copper during M. tuberculosis infection as a potential antimicrobial defense mechanism, we were interested in testing if the increased resistance to copper of the ΔcsoR strain translated to our murine aerosol infection model. Mouse infections revealed that despite similar CFU counts at week 2, the mutant strain multiplied and survived nearly a log fold above wild type through the early stages of chronic infection at weeks 4 and 8. These results might be attributed to the preparedness of the ΔcsoR strain to face host stress conditions even before they were detected; as transcriptional analysis indicated, the mutant strain already had genes induced that typically respond specifically to host stressors such as copper (ctpV), nitric oxide, and hypoxia (dosR regulon). Toxic copper levels and nitric oxide are stressors faced by M. tuberculosis within the phagosome of macrophages, especially those that have been activated. This activation follows the initiation of the adaptive immune response to M. tuberculosis infection taking place around weeks 2 to 4 [33], after which the ΔcsoR strain shows improved survival over wild type. Previously published data examining our Δcso strain demonstrated a phenotype opposite of that observed for ΔcsoR during mouse infection [34]. Unlike our ΔcsoR strain, where the remaining members of the cso are induced, the Δcso strain was attenuated, particularly during the later time points of chronic infection [34]. This earlier study, and our transcriptional study showing the induction of the cso during mouse infection [35], highlight the importance of the whole cso operon to the survival of M. tuberculosis during infection.
To better understand the control CsoR has over its own operon, we studied transcriptional expression of the cso in wild type, knockout, and complement strains in untreated culture and under two levels of copper (50, and 500μM). As expected, we found that without csoR, the cso genes were induced in the absence of copper. Notably, this induction is not to the same level as the induction seen in wild type under high levels of copper stress (500μM). This indicates that at high levels of copper, CsoR derepression may not be solely responsible for cso induction. This hypothesis is consistent with the idea that genes outside of the core copper responsive regulon (n = 15 induced at 50μM CuCl 2 ) are induced to address secondary stress such as the presence of reactive oxygen and nitrogen species at higher copper levels (n = 30 induced at 500μM 500μM (white) and its untreated counterpart (black) from the same strain. Data represent one of two similar biological replicates. Error bars represent the standard error of the mean from three technical replicates. (C) Expression levels of mmcO as determined by qRT-PCR analysis of the same samples-H37Rv (black), ΔcsoR (striped), or ΔcsoR::csoR (white)-left untreated (0μM) or stressed with 50μM or 500μM CuCl 2 . Fold change is shown as expression levels of each gene relative to expression levels in untreated wild type culture after normalization to sigA expression levels. Data represent one of two similar biological replicates. Error bars represent the standard error of the mean from two technical replicates. CuCl 2 ) [6]. It could be that a second transcriptional regulator is required to further induce the cso in response to this toxic level of copper stress. Therefore, disruption of sequence near the binding site of a second regulator may have contributed to a lesser level of cso induction. One such candidate may be transcriptional regulator Rv2324 predicted to bind within csoR just upstream of the deletion site [26]. Therefore, the attenuated cso induction under high levels of copper stress may have attributed to a more modest survival phenotype of the ΔcsoR strain under copper and host stress than might otherwise be expected.
Despite the apparent advantage of losing csoR under both copper and host stress, our global transcriptional analysis reveals the important role csoR plays in the cell under normal growth conditions. While this work has not ventured to define the direct targets of CsoR regulation, our results support previous findings that CsoR may only regulate its own operon [21]. Both studies found the other members of the cso to be induced in the csoR mutants to similar levels. The previous study [21], however, found no differentially expressed genes outside of the cso, in stark contrast to our list of 223 differentially regulated genes. It is possible, as the authors point out, that the previously studied strain is a polar knockout, unlike ours. It may be that translation of the cso in their ΔcsoR:: hyg R strain was not as efficient as in the ΔcsoR strain discussed here, therefore mitigating any stress we posit is placed on the cell when the operon is deregulated in the absence of csoR. We hypothesize that this stress is derived from the consequences of cso overexpression in the absence of copper. Interestingly, outside of the cso, the only copper responsive genes to be differentially expressed were down regulated [6]. It may be that lower levels of copper in the ΔcsoR strain, due to increased Cu export by CtpV, tightened copper mediated repression of these genes such as mmcO and mymT, by regulators such as RicR. This tightening of repression was observed in our RNA-Seq and confirmed by qRT-PCR above. This compensatory response suggests a hierarchy of copper regulation in M. tuberculosis with CsoR having a key role. It should also be noted, however, that these previous studies were conducted using microarray analysis, which can be less sensitive to differential expression than RNA-Seq analysis [36]. Therefore, the absence of more extensive overlap among these studies may be partially due to the difference in technique used.
A major consequence of the loss of csoR and the deregulation of the cso was the induction of the DosR regulon. While culture manipulation can cause a slight but significant induction of the DosR regulon [37], great care was taken to quickly process all samples to be used for RNA analysis in an identical fashion. The observed induction of the DosR regulon was substantial, and likely too high to be caused by handling differences. While the mechanism for this induction remains unresolved, we have modeled two opposing hypotheses (Fig 6). The first is based on data suggesting that disruption in the electron transport system or cytochrome c oxidase induces expression of the DosR regulon [25,38] (Fig 6A). As cytochrome c oxidase, involved in the terminal steps of cellular respiration, is a copper metalloenzyme, cso induction resulting in excessive copper export could negatively impact its function. This block in respiration may contribute to the activation of the DosR regulon.
A second possibility could be that NO is directly inducing the DosR regulon in our ΔcsoR strain (Fig 6B). While we hypothesize that copper may be present at lower levels in the cytoplasmic space, CtpV is thought to export copper across only the cell membrane. We do not know if expression levels of genes involved in the export of copper across the mycobacterial cell wall were affected, as the only gene proposed to fulfill this role in M. tuberculosis, Rv1698, was later excluded as a potential candidate [32,40]. Thus it is possible that copper could be accumulating and causing damage in the mycobacterial periplasmic space of ΔcsoR. This damage could be exacerbated by the slight down-regulation of a recently characterized periplasmic multicopper oxidase, mmcO [23]. Cu(I) can generate NO through interactions with S-nitrosothiols [39], which could subsequently induce the DosR regulon.
These scenarios and the stressed transcriptional profile of the ΔcsoR strain highlight the potential harm that can be done even at low levels of copper and emphasize the need for balance in regulating the level of intracellular copper, as aided by csoR. This increased understanding of the balance of copper M. tuberculosis must maintain during infection may ultimately help guide development of antimycobacterial therapies that wish to utilize copper or target the copper stress response. Overall this study has revealed new insights into the importance of copper regulation and csoR, not only under copper stress, but also in its absence. Similar copper regulators are present in other intracellular pathogens and similar scenarios could be at play in these organisms as well. Potential consequences of ΔcsoR disrupted copper homeostasis. Overexpression of the cso and the proteins it codes for after deletion of csoR (dark red) may result in excessive levels of copper export by CtpV (red). (A) As a result, pools of copper (orange) for metalloenzyme usage may be depleted, reducing the levels of functional proteins such as cytochrome c oxidase (yellow), as seen in the panel on the right. This could potentially slow the electron transport chain leading to the induction of the DosR (blue) regulon through the redox sensor, DosS (blue) [38]. The wild type strain is shown in the panel on the left for comparison. (B) Alternatively, the increased cytoplasmic export of copper may lead to a buildup of copper in the mycobacterial periplasmic space where MmcO (green) usually assists in copper detoxification. The increased copper stress in this region may be magnified, however, by the down regulation of mmcO in the mutant strain. Free copper can lead to the production of reactive nitrogen species, such as NO [39]. NO can trigger expression of the DosR regulon [25]. The wild type strain is shown in the panel on the left for comparison. CM; cytoplasmic membrane. MM; mycomembrane. Cu; copper. NO; nitric oxide. COX; cytochrome c oxidase. P; phosphoryl group.

Strain construction
Strains used in this work include Mycobacterium tuberculosis H37Rv, and its derivatives M. tuberculosis ΔcsoR and M. tuberculosis ΔcsoR::csoR. To create an inactivating mutation in csoR in M. tuberculosis, we first generated a ΔcsoR::hyg R strain using specialized transduction and homologous recombination as outlined previously [14,34] and confirmed by sequencing and Southern blot [34]. To determine the polar or nonpolar nature of the construct, RNA was isolated from the strain using a TRIzol1 Reagent based method before treatment with TURBO™ DNase (Ambion1, Austin, TX) until PCR negative. RNA was then reverse transcribed to cDNA using Superscript1 III (Invitrogen, Carlsbad, CA). cDNA was then screened by PCR for the expression of the genes downstream of csoR: Rv0968, ctpV, and Rv0970. Our analysis indicated that the mutant was polar.
To construct an unmarked mutant strain, a helper plasmid, pYUB870, was used to remove the hyg R gene as described previously [14], leaving behind a stop codon at the 14 th amino acid of the csoR gene. Again these results were confirmed by sequencing and the nonpolar nature of this mutant was demonstrated as described above. This strain was used for the construction of a complement strain. The csoR gene and the 200bp promoter region upstream of the csoR transcriptional start site were cloned into the integrative, E. coli-mycobacterial shuttle vector pMV306, originally derived from pMV361 [41]. The vector, containing a kanamycin resistance marker, was electroporated into M. tuberculosis ΔcsoR. Successful electroporants were identified by screening colonies growing on Middlebrook 7H10 media supplemented with 10% ADC and 30 μg/ml kanamycin by PCR targeting csoR and the upstream portion of the pMV306 vector. The final clone was verified by sequencing.

Growth conditions
Starter cultures for experiments were grown from frozen stocks in liquid media to late log stage (~OD 600 1.0) in Middlebrook 7H9 (Remel™, Lenexa, KS) with 10% ADC and 0.05% TWEEN 80. For all experiments, 30μg/ml kanamycin were included in growth media used for M. tuberculosis ΔcsoR::csoR. Before use in downstream experiments, bacteria taken from 7H9 cultures were washed twice with copper-free Sauton's media. Similarly, all reagents and media used in copper sensitive experiments were treated with 0.2% (w/v) Chelex1 100 (Sigma-Aldrich, St. Louis, MO) to remove traces of copper prior to addition of other metal supplements before use. All glassware was acid washed with 1M nitric acid to remove any traces of metal ions. For transcriptional experiments, washed M. tuberculosis strains H37Rv, ΔcsoR, and ΔcsoR::csoR were inoculated to an OD 600 of 0.10 in 30 or 100 ml Sauton's media containing 0.05% TWEEN 80. At late log phase (~OD 600 1.0) cultures were exposed to untreated media or 50 or 500μM CuCl 2 for 3 hours before collecting cultures for RNA extraction as detailed previously [6].

Microplate Alamar Blue Assay (MABA)
To quantify the level of resistance of M. tuberculosis constructs to copper, MABA assays were performed as described previously [18,19] with a few modifications. Briefly, the outer wells of flat bottom 96-well microplates were filled with sterile ddH 2 O to prevent evaporation. Inner wells were filled with 100ul Sauton's broth containing serial dilutions of CuCl 2 in duplicate or no added CuCl 2 as a control and M. tuberculosis strains H37Rv, ΔcsoR, and ΔcsoR::csoR inoculated to a theoretical OD 600 of 0.05. A series of nine, two-fold dilutions were made with CuCl 2 from 16 to 4000μM. On day 5 after inoculation, 30μl Alamar Blue (Invitrogen; Carlsbad, CA): TWEEN 80 (1:1) were added to a control well and incubated an additional 24hrs. When a shift in color from blue to pink was seen on the following day, 30μl Alamar blue:TWEEN 80 (1:1) was added to each of the remaining wells. Data were collected the following day and the blue wells containing the lowest concentration of CuCl 2 were recorded for each series and each strain as the minimum inhibitory concentration (MIC).

Mouse infections
BALB/c mice (Harlan, Indianapolis, IN) were infected with M. tuberculosis H37Rv wild type or ΔcsoR strains using an aerosol infection model with the Glas-Col1 Inhalation Exposure System (Glas-Col, LLC, Terre Haute, IN) as outlined before [15]. At 2, 4, 8, and 25 weeks groups of 3 to 5 mice were euthanized by cervical dislocation following anesthetization by isoflurane, and lung, spleen, and liver were collected for colony counts and histopathology [15]. Aside from the 25 week time point, the experiment was completed in duplicate.

Ethics statement
Animal experiments were approved by the Institutional Animal Care and Use Committee, University of Wisconsin-Madison (V1095) in compliance with the regulations set in place by the Public Health Service Policy on the Humane Care and Use of Laboratory Animals, overseen by the NIH Office of Laboratory Animal Welfare (OLAW). Mice were monitored daily by caretakers for signs of illness or distress and all efforts were taken to prevent animal suffering. No unexpected deaths occurred during the study and all mice reached predetermined endpoints without showing signs such as behavioral changes, fur ruffling, anorexia, or lethargy that would require the use of a humane endpoint.
Quantitative, real-time PCR RNA was isolated for qRT-PCR immediately after collecting the cultures using a TRIzol based method described previously [15,34]. RNA was treated with TURBO DNase until PCR negative to remove contaminating DNA. For qRT-PCR analysis of the cso, 1-2μg RNA were used as template for cDNA using Superscript III. A SYBR green based qRT-PCR protocol utilizing GoTaq1 qPCR Master Mix (Promega, Madison, WI) and the StepOnePlus™ Real-Time PCR System (Applied Biosystems1, Foster City, CA) were used. Expression of all genes was normalized to sigA expression levels. Analysis was carried out using LinRegPCR [42]. Two biological replicates with no less than two technical replicates each were completed.

Transcriptional profiling through RNA sequencing
For transcriptional profiling with RNA-Seq technology, RNA was isolated from cultures grown in Sauton's broth to late log phase and treated with TURBO DNAse as done with samples for qRT-PCR. RNA integrity was confirmed using the 2100 Bioanalyzer (Agilent, Santa Clara, CA). The 23S, 16S, and 5S rRNAs were depleted from the samples using the Ribo-Zero™ Magnetic Kit for bacteria (Epicentre, Madison, WI). Biological replicates of prepared rRNAreduced RNA were submitted to the University of Wisconsin Biotechnology Center's DNA Sequencing Facility where they were prepared for sequencing using the TruSeq RNA Sample Preparation Kit (Illumina1, San Diego, CA). The samples were run on an Illumina HiSeq™ 2000 to generate 100bp single reads. Base calling was done in CASAVA 1.8.2. Before differential expression analysis the quality of the FASTQ files was checked using the FASTX-Toolkit developed in the lab of Gregory Hannon at Cold Spring Harbor (http://hannonlab.cshl.edu/ fastx_toolkit/index.html). Sequence reads were mapped to the M. tuberculosis H37Rv genome using Bowtie [43]. Reads mapping to more than one location within the genome were excluded from further analysis. BEDTools was used to determine counts for each coding sequence in the M. tuberculosis H37Rv genome [44]. The above analyses were carried out within the web-based Galaxy platform [45,46,47]. Differential expression analysis using the output count data was carried out in R with the Bioconductor software package, edgeR [48]. Genes demonstrating !2 fold differential expression between ΔcsoR and wild type strains, and having an FDR < 0.05 were considered to be differentially expressed. Confirmation of 10 differentially expressed genes was carried out by qRT-PCR as detailed above.

Whole genome sequencing
Genomic DNA was extracted as described previously [49]. Quality was confirmed by gel electrophoresis and measuring absorbance at 260nm and 280nm.

Statistical analysis
For large scale differential expression analysis, false discovery rates were calculated using the Benjamini and Hochberg algorithm [50] with a cutoff for significance at FDR < 0.05. For all other experiments, statistical significance was determined using Student's t-test with cutoff value of P < 0.05. M. tuberculosis H37Rv (red circles), ΔcsoR (green squares), and complement (blue triangles) strains were exposed to two-fold dilutions of CuCl 2 in Sauton's broth from 16μM to 4000μM. Percent reduction of Alamar blue reagent as compared to untreated wells was used to measure growth. Data are representative of two biological replicates. tuberculosis H37Rv (red) or ΔcsoR (green). Once a lawn of growth was detected, the radius of the zone of inhibition around the disc was measured. Error bars indicate the standard deviation. Data are representative of two biological replicates. (B) M. tuberculosis H37Rv (red circles) and ΔcsoR (green squares) were exposed to two-fold dilutions of isoniazid in 7H9 broth with ADC from 3ng/ml to 80ng/ml. Percent reduction of Alamar blue reagent as compared to untreated wells was used to measure growth. Data are representative of two biological replicates. (TIF) S7 Fig. Assessing copper levels in the cell using mymT transcription levels as an indirect indicator. Expression levels of mymT as determined by qRT-PCR analysis of samples H37Rv, ΔcsoR, or ΔcsoR::csoR left untreated (0μM) (black) or stressed with 50μM (striped) or 500μM (white) CuCl 2 . Fold change is shown as expression levels of each gene relative to expression levels in untreated wild type culture after normalization to sigA expression levels. Data represent one of two similar biological replicates. Error bars represent the standard error of the mean from two technical replicates. (TIF) S1