Heme Oxygenase-1-derived Carbon Monoxide Induces the Mycobacterium tuberculosis Dormancy Regulon*

The mechanisms that allow Mycobacterium tuberculosis (Mtb) to persist in human tissue for decades and to then abruptly cause disease are not clearly understood. Regulatory elements thought to assist Mtb to enter such a state include the heme two-component sensor kinases DosS and DosT and the cognate response regulator DosR. We have demonstrated previously that O2, nitric oxide (NO), and carbon monoxide (CO) are regulatory ligands of DosS and DosT. Here, we show that in addition to O2 and NO, CO induces the complete Mtb dormancy (Dos) regulon. Notably, we demonstrate that CO is primarily sensed through DosS to induce the Dos regulon, whereas DosT plays a less prominent role. We also show that Mtb infection of macrophage cells significantly increases the expression, protein levels, and enzymatic activity of heme oxygenase-1 (HO-1, the enzyme that produces CO), in an NO-independent manner. Furthermore, exploiting HO-1+/+ and HO-1-/- bone marrow-derived macrophages, we demonstrate that physiologically relevant levels of CO induce the Dos regulon. Finally, we demonstrate that increased HO-1 mRNA and protein levels are produced in the lungs of Mtb-infected mice. Our data suggest that during infection, O2, NO, and CO are being sensed concurrently rather than independently via DosS and DosT. We conclude that CO, a previously unrecognized host factor, is a physiologically relevant Mtb signal capable of inducing the Dos regulon, which introduces a new paradigm for understanding the molecular basis of Mtb persistence.

lon. To test this hypothesis, we examined the induction of the Mtb Dos regulon in response to CO exposure and assessed the independent contribution of DosS and DosT in relaying this signal to the Mtb Dos regulon. We also studied the expression of the Mtb Dos regulon in response to physiological levels of CO generated within macrophages. Lastly, we examined HO-1 levels in macrophages and in the lungs of Mtb-infected and uninfected mice.

EXPERIMENTAL PROCEDURES
Mice-C3HeB/FeJ mice were purchased from the Jackson Laboratory. HO-1 Ϫ/Ϫ mice and age-matched HO-1 ϩ/ϩ littermates (C57BL/6 ϫ FVB background) and iNOS Ϫ/Ϫ (C57BL6 background) were used. The animal protocols were approved by the Institutional Care and Use Committees at the University of Alabama at Birmingham and Harvard School of Public Health. The mice were regularly tested and found to be free of common mouse pathogens and were used for experiments at 6 -12 weeks of age.
Mtb Mutant Strains-Mtb dos mutant strains were kind gifts from Dr. David Sherman (Seattle Biomedical Research Institute, WA). Construction of Mtb⌬dosS and Mtb⌬dosR is described by Sherman et al. (6), and Mtb⌬dosT and Mtb⌬⌬dosS⌬dosT are described by Roberts et al. (9). Mtb⌬dosR was found to be attenuated for growth in the Wayne model for in vitro dormancy (8,13) and is not capable of inducing the Mtb Dos regulon under hypoxic conditions or in response to NO (8). The expression of hspX was reduced 40 -45% in Mtb⌬dosS or Mtb⌬dosT, whereas Mtb⌬dosS⌬dosT abolishes expression of hspX in response to hypoxia (9).
Bacterial Infection-Virulent Mtb (strain Erdman, TMC 107) was obtained from the Trudeau Mycobacterial Culture Collection (Trudeau Institute). Mice were intravenously infected with 1 ϫ 10 5 colony-forming units of Mtb in 100 l of phosphatebuffered saline via tail vein. Mtb H37Rv was used in all other experiments.
Microarray Hybridization and Data Analysis-Microarrays used in this study were produced and processed at the Center for Applied Genomics at the Public Health Research Institute as described earlier (22). See supplemental information for details.
Expression Analysis of the Mtb Dos Regulon in Bone Marrowderived Macrophages (BMM)-Four independent experiments were performed using BMM collected from 8 -12-week-old HO-1 Ϫ/Ϫ mice and HO-1 ϩ/ϩ littermates (C57BL/6 ϫ FVB background). 1-2 ϫ 10 6 macrophages were infected with a multiplicity of infection of 5-10. RNA was isolated and utilized in real-time PCR analysis according to previously described protocols (14,23). The -fold expression of a specified gene transcript of bacilli residing in HO-1 ϩ/ϩ BMM was measured relative to the gene transcript of bacilli residing in HO-1 Ϫ/Ϫ BMM. Expression was normalized using Mtb 16 S rRNA as internal control.
Macrophages were lysed with guanidine thiocyanate, and bacteria were harvested by centrifugation. Approximately 1-2 g of bacterial RNA was isolated from four independent biological samples and analyzed via quantitative PCR (Q-PCR).
Western Blot Analysis-Approximately 2 ϫ 10 6 cells (murine macrophage cell line or BMM) were grown in a six-well plate and infected with Mtb (multiplicity of infection 5-10). For Western blot analysis, cells were harvested and lysed in radioimmune precipitation buffer, samples were sonicated and centrifuged, and the protein concentration of the lysate was determined using the BCA protein assay. Western blot analysis was performed as described earlier (24).
HO Enzymatic Activity Assay-RAW cells grown in 175-cm 2 tissue culture flasks (5 ϫ 10 7 cells) were infected with Mtb (multiplicity of infection 5-10). Uninfected cells were used to serve as controls. HO activity was determined as described earlier (25).
Immunohistochemistry-Tissue sections from infected animals (n ϭ 4) of each group were treated with 10% formalin in neutral buffer and cut into sections to 5 m. These sections were deparaffinated and rehydrated in graded ethanol. Sections were washed in phosphate-buffered saline followed by endogenous peroxidase inactivation in 0.3% H 2 O 2 , labeled with primary antibody against HO-1 (Stressgen, SPA-896), stained with anti-rabbit peroxidase (Jackson Laboratories), and developed with the 3,3Ј-diaminobenzidine (DAB) substrate kit (Vector Laboratories) to produce a characteristic brown color. Control reactions were performed using rabbit IgG instead of the primary antibody, and the secondary antibody alone. Sections were also stained with hematoxylin and counterstained with eosin according to previously described protocols (26).
Statistical Analysis-Results are expressed as mean Ϯ S.E. and are derived from at least three independent experiments. Student's t-test or analysis of variance with Student-Newman-Keuls post-test was used for comparisons.

CO Elicits the Induction of the Complete Mtb Dos Regulon-
We have recently reported that the redox state of the DosS and DosT heme irons, as well as the ligands O 2 , NO, and CO, modulates DosS and DosT autokinase activity (14). Because the effect of CO on the complete Dos regulon is not yet established, we sought to determine whether CO could selectively induce the Mtb Dos "fingerprint," consisting of ϳ48 genes. We used microarray expression profiling and captured the transcriptional response of Mtb cells exposed to 50 M CO. The results demonstrate that CO rapidly induced the complete Dos regulon ( Fig. 1). We repeated the microarray experiment using the Mtb⌬dosR mutant strain and found that induction of the Dos regulon by CO was mediated via DosR (Fig. 1). Thus, the induction of the Dos regulon by hypoxia (6), NO (7,8), and now CO is mediated by the DosR/S/T system. We next tested a series of CO concentrations (5 nM-100 M) and examined fdxA expression via Q-PCR as an indicator (9) for Dos regulon expression. We observed that full induction of fdxA occurred at Ն5 M CO ( Fig. 2A). Similarly, we analyzed the time course for the induction of fdxA expression after exposure to 50 M CO and found that maximal expression occurred within 30 min (results not shown). By 24 h post treatment, expression of fdxA declined to basal levels but rapidly increased upon treatment with a second dose of CO (results not shown).
In sum, these data demonstrate that, similar to hypoxia and NO, exposure to low, non-toxic concentrations of CO specifically induces the complete Mtb Dos regulon in a rapid and coordinated fashion. Thus, we have identified CO as a third signal that induces the Mtb Dos regulon.
Mtb DosS Is the Primary Sensor of CO-Because dosT, as opposed to dosS, is not up-regulated in response to any condition tested to date, including hypoxia (6), NO (7,8), or CO ( Fig.  1), we sought to examine the independent contributions of DosS and DosT in modulating the Dos regulon. First, we exposed Mtb⌬dosS⌬dosT cells to CO and found that similar to the Mtb⌬dosR strain, this strain was impaired in its ability to induce the Dos regulon (Fig. 2B). We also utilized the single knock-out strains, Mtb⌬dosS and Mtb⌬dosT, and found that Mtb⌬dosS was severely attenuated in its ability to induce the Dos regulon compared with wild-type Mtb, whereas Mtb⌬dosT was moderately affected (Fig. 2B). We conclude that CO is sensed through the Mtb DosR/S/T system and that DosS is the preferred, but not the sole, sensor of CO.

Induction of the Mtb Dos Regulon in Macrophages Is
Modulated by HO-1-We next sought to determine whether HO-1generated CO could be a physiologically relevant ligand of Mtb DosS and DosT to ultimately modulate the Dos regulon. We utilized BMM isolated from HO-1 ϩ/ϩ and HO-1 Ϫ/Ϫ mouse strains rather than chemical inhibitors or inducers of HO, which have nonspecific effects, including inhibition of guanylate cyclase activity (27), inhibition (28) or stimulation of NO synthase activity (29), and inactivation of caspase 3 and 8 (30). To analyze the effect of Mtb infection on HO-1 expression, we first infected BMM from 8 -12-week-old HO-1 ϩ/ϩ mice with Mtb and subsequently used Western blot analysis to examine HO-1 levels. Mtb-infected BMM showed significantly increased HO-1 levels compared with uninfected BMM (Fig.  3A), whereas the constitutive isoform of HO-2 was unchanged. Similar to previous iNOS studies (31), we infected BMM from HO-1 ϩ/ϩ and HO-1 Ϫ/Ϫ mice with Mtb and performed Q-PCR using primers specific for fdxA, hspX, and dosR. Prior studies have shown that these genes are highly responsive to hypoxia and NO (6 -8). We observed a 3-to 68-fold up-regulation of these genes in bacilli recovered from HO-1 ϩ/ϩ cells relative to HO-1 Ϫ/Ϫ cells (Fig. 3B), supporting a role for HO-1 in the induction of the Dos regulon. In sum, these data suggest that Mtb-infected macrophages specifically up-regulate HO-1 to increase CO, which is sensed by the Mtb sensor kinases DosS and DosT and then relayed to DosR to induce the Dos regulon.
Up-regulation of HO-1 Is Independent of the NO Pathway-It is well known that during Mtb infection iNOS is induced via IFN-␥ to generate NO (32). Because NO is an inducer of HO-1 (see supplemental note 1 and supplemental Fig. S1) and because the above results demonstrate up-regulation of HO-1 in   Note that HO-1 can exist in a 32-kDa wild type and a cleaved 28-kDa form in immunoblot blots (33). Furthermore, time course experiments demonstrated that HO-1 expression was induced within 2 h of Mtb infection and was maintained for at least 24 h (data not shown). To specifically examine the potential link between HO-1 and the NO signaling pathway, we also infected the RAW 264.7 ␥-NO(Ϫ) cell line, which is defective in the IFN-␥-mediated induction of iNOS (34). Our Q-PCR analysis showed that HO-1 expression following Mtb infection (Fig. 4A) is significantly increased in the RAW 264.7 ␥-NO(Ϫ) cell line at levels comparable with that of the parental cell line (Fig. 4A). Corroborating these data is the Western blot analysis showing increased HO-1 protein levels in Mtb-infected RAW 264.7 ␥-NO(Ϫ) cells (Fig. 4B). We also utilized the highly specific NOS inhibitor N-(3-{amino-methyl} benzyl) acetamidine (1400W) and found that HO-1 mRNA and protein levels were significantly increased (Fig. 4, A  and B) in the presence of the inhibitor during Mtb infection. Finally, we infected iNOS2 Ϫ/Ϫ BMM with Mtb and again observed a significant increase in HO-1 expression upon Mtb exposure (Fig. 4A).
HO Enzymatic Activity Is Increased upon Infection of BMM with Mtb-Having observed that HO-1 mRNA and protein levels increase in macrophages upon infection with Mtb ( Figs. 3A and 4, A and B), we now sought to examine (i) whether increased HO-1 expression reflects an increase in HO enzymatic activity leading to increased CO levels, and (ii) the role of the NO pathway in modulating HO enzymatic activity upon infection. RAW 264.7 cells were infected with Mtb and used in the HO enzymatic activity assay. In this widely used assay, HO stoichiometrically releases CO, molecular iron, and biliverdin. The latter is then converted to bilirubin by biliverdin reductase and is therefore an exact indicator of the amount of released CO. Subsequently, it was observed that macrophages infected with Mtb displayed ϳ8-fold higher HO enzymatic activity compared with uninfected control cells (Fig. 5A). Importantly, identical experiments were conducted with RAW 264.7 ␥-NO(Ϫ) and RAW 264.7 cells treated with 1400W. Again, the results demonstrate that HO activity was not affected by a defective IFN-␥ signaling pathway or a specific iNOS inhibitor (Fig. 5, B and  C). Collectively, these results suggest that the induction of HO-1, and therefore CO, upon Mtb infection is independent of IFN-␥-mediated NO production and point toward an alternative pathway by which Mtb induces HO-1.

HO-1 Is Induced in Lesions of Mtb-infected
Mice-Using a systematic series of in vitro culture and macrophage infection experiments described above, we made the observation that Mtb specifically senses and responds to physiological levels of CO to induce the Dos regulon. To determine whether CO was produced at the site of infection, as is the case for NO (3), we assessed whether HO-1 mRNA is up-regulated in mouse lungs following Mtb challenge. Upon infection with Mtb, HO-1 expression was significantly increased (19.5-fold) (Fig. 6A). Staining of lung lesions with HO-1specific antibodies demonstrated that HO-1-expressing cells were associated with TB lesions. Also, a high proportion of HO-1-positive cells were found within extensive inflammatory lesions containing necrotic areas in the lungs of the mice  ( Fig. 6B, I and II). The staining was localized predominantly in alveolar epithelial cells and infiltrating macrophages (Fig.  6B, I, inset).
In sum, we have shown that increased levels of HO-1 mRNA, protein, and therefore CO are produced in the lungs of Mtbinfected mice. The data suggest both diverging and complementary roles for NO and CO in Mtb pathogenesis and identify CO as a signaling ligand for DosS and DosT to modulate the Mtb Dos regulon.

DISCUSSION
The Mtb DosR/S/T two-component system responds to at least two physiologically relevant dormancy signals, O 2 (6) and NO (7,8). In this study, we discovered for the first time that CO is a third physiologically relevant signal capable of inducing the complete 48-member Mtb Dos regulon. We have also shown that CO is primarily sensed through DosS. Importantly, we have demonstrated that the levels of CO produced within macrophages via HO-1 were sensed by Mtb and resulted in the induction of key members of the Dos regulon. We further demonstrated that HO-1 expression and protein levels, as well as HO enzymatic activity in macrophages, are significantly increased upon Mtb infection and are independent of the NO signaling pathway. Lastly, we have demonstrated that HO-1 is significantly increased in the lungs of Mtb-infected mice. These findings demonstrate an important and previously unknown function for HO-1 and, thus, CO. Notably, the ability of three diatomic gases, CO, NO, and O 2 , to induce an identical set of Mtb genes is an unparalleled finding and presents a useful paradigm for redox signal transduction in prokaryotic cells.
Recently, our laboratory (14) as well as others (15)(16)(17) has shown that DosS and DosT are GAF-containing heme proteins. We proposed a sense-and-lock model that suggests that the ligation state of Mtb DosS and DosT can be "locked" by gradients of NO and CO throughout progression of disease (14). In a recent study demonstrating a protective role for CO in experimental cerebral malaria (21), it was proposed that CO "locks" cell-free hemoglobin in the Fe(II) state, thereby preventing oxidation to the unstable met Fe(III) state, which can react with reactive oxygen species to disrupt the blood brain barrier (21). The overlap between this model and the sense-and-lock model for the Mtb Dos regulon (14) illustrates one of the most important properties of CO: the ability to react only with Fe(II) and not Fe(III). NO, on the other hand, can react with both Fe(II) and Fe(III) species.
In this study, we have utilized microarray technology and Q-PCR and conclusively demonstrated that similar to hypoxia and NO (6 -8), low concentrations of exogenously provided CO induce the complete Dos regulon within minutes of CO exposure. Earlier studies have suggested that M. bovis BCG, M. gordonae, M. smegmatis, and M. tuberculosis H37Ra oxidize high concentrations of CO (35) via CO dehydrogenase (Rv0373c/Rv0375c). However, earlier microarray studies as well as this study have shown that neither hypoxia (6), NO (7,8), nor CO (Fig. 1) differentially regulates expression of these genes.
Importantly, our findings demonstrating that CO is sensed primarily through DosS, with DosT playing a less prominent role, agree with recent kinetic data reporting that DosS (K d ϭ 36 nM) has a higher affinity for CO than DosT (K d ϭ 940 nM) (17). Nonetheless, this should be viewed with caution, because an earlier study has shown that DosS and DosT contributed essentially equally toward regulating fdxA (9) in response to hypoxia despite differences in binding of O 2 to DosS (K d ϭ 3 M) or DosT (K d ϭ 26 M) (9). Also, because dosT expression is not responsive to either O 2 , NO, or CO (or any other condition according to our knowledge) and appears to be constitutively expressed, DosS or DosT protein levels may also influence regulation of the Dos regulon. Thus, the biological significance of the highly regulated nature of dosS, as opposed to constitutively expressed dosT, is an important, albeit unexplained, finding.
Because HO generates equimolar amounts of CO, Fe(II), and biliverdin using heme as a substrate in vivo, it was important to demonstrate that not only exogenously provided CO but also physiological levels of CO are capable of inducing the Mtb Dos regulon upon infection. Using mouse HO-1 ϩ/ϩ and HO-1 Ϫ/Ϫ BMM cells, we demonstrated that CO produced within macrophages significantly induces expression of the Mtb Dos regulon. Because HO generates not only CO, but also Fe(II) and biliverdin, it could be argued that these two components (or even heme) rather than CO induce the Dos regulon. However, neither heme (see supplemental text note 2) nor iron (36) was shown to induce the Dos regulon, and there is no evidence to suggest that biliverdin is capable either. Rather, data from several independent studies (14,15,17) as well as our in vitro cellbased studies (Fig. 1) provide convincing evidence that DosS and DosT are exclusively sensing CO and not the other heme breakdown products.
HO-1 is the most important endogenous source of CO and is highly induced upon oxidative stress (37)(38)(39), whereas HO-2 is the low level, constitutive HO that contributes minimally to overall CO levels (40). Heme-independent CO arises primarily from photo-oxidation, peroxidation of lipids, and xenobiotic compounds and represents a fraction of total endogenous CO production (41).
Because continuous production of IFN-␥ and tumor necrosis factor ␣, which synergistically regulate iNOS expression, were previously shown to be essential to prevent reactivation of persistent Mtb in infected mice (5), we examined the contribution of the NO signaling pathway in modulating HO-1 pathway. We independently infected three different cell types, two of which had defects in the IFN-␥/NO response pathway, and demonstrated that the induction of HO-1 in response to Mtb infection occurs via an unknown mechanism that is independent of NO, and possibly IFN-␥, signaling. Notably, we documented a significant increase of HO-1 in the lungs of Mtb-infected mice. This is not unusual, as HO-1 is rapidly induced in response to oxidative stress and inflammatory reactions, which are the major underlying causes of many diseases. However, a surprising finding was that this response appears to be independent of the NO pathway.
The combined production of NO and CO at the site of infection now provides insight into how gradients of NO and CO (and O 2 ) may shape disease progression. For example, "optimal" levels of O 2 , NO, and CO could maintain Mtb in a metabolically quiescent state. On the other hand, because of the strong anti-apoptotic and anti-inflammatory properties of CO (18,19,42), excessive amounts of CO and also NO (5) could cause destructive immunopathology. Regardless, these findings provide mechanistic insight into how the presence of multiple gases ( Fig. 7 and supplemental note 3) generates a "doubleedged sword" that can either be beneficial or detrimental. Furthermore, differences in the on-and off-rates of the gases for DosS and DosT (17) and differential host and bacillary responses are evidence of a dynamic mechanism that Mtb has evolved to modulate disease during discrete disease states. Mtb may also exploit other strategies to sense O 2 or NO. For example, Mtb WhiB3 was shown to bind NO and react with O 2 via its 4Fe-4S cluster, which might initiate a metabolic switchover to a preferred in vivo carbon source, fatty acids (43). Lastly, the recent link between TB and CO air pollution (44), as well as the role of HO-1 in suppressing human immunodeficiency virus replication (45), suggests that the findings in this study may have important socioeconomic and environmental health implications.
In sum, although the exact role of HO-1 and CO in Mtb pathogenesis remains to be established, the findings in this study significantly advance our understanding of the Mtb Dos dormancy response, conventionally viewed as only being induced by two in vivo relevant signals, hypoxia and NO, and therefore represent a hitherto unexplored area of TB research. We have shown that CO generates expression profiles identical to that of Mtb cells exposed to hypoxia or NO and that DosS is the preferred sensor of CO. We have also shown that physiological levels of CO specifically induce the Mtb Dos regulon. Furthermore, we have demonstrated that HO-1 expression, protein, and enzymatic activity significantly increase in response to Mtb infection and that these events are independent of the NO signaling pathway. Lastly, we have demonstrated that HO-1 is produced at the site of infection, the lungs of Mtbinfected mice. These findings have broad implications for understanding the mechanistic basis for Mtb persistence. Because numerous studies reported striking differences in local O 2 , NO, and CO concentrations in organs, tissues, and single cells (see supplemental note 3), the model predicts that gradients of O 2 , NO, or CO are sensed in conjunction by DosS and DosT, rather than independently. DosS is the preferred sensor of CO (thick blue arrow), whereas DosT is less capable (thin blue arrow) of inducing the Dos regulon. This contrasts with O 2 , which inhibits expression of the Dos regulon (blocked arrow), albeit not during hypoxia. The model predicts that a combination of microenvironmental O 2 , NO, or CO (overlapping circles) is crucial in modulating the state of the disease. Although modeling the above events in the context of chronic TB is technically difficult, the well established role of HO-1 in modulating apoptosis/necrosis, its antiinflammatory properties (specifically the effect on tumor necrosis factor ␣), and the release of Fe(II) have significant implications for understanding the mechanism of Mtb persistence.