|Home | About | Journals | Submit | Contact Us | Français|
Nitric oxide (NO), which is an important component of immunity to Mycobacterium tuberculosis, has both cytotoxic and immune regulatory functions. We examined the way that this molecule interacts with M. tuberculosis in vivo by screening for bacterial mutations that alter growth in mice that are unable to produce inducible NO synthase (iNOS), the dominant source of NO during infection. We found that very few bacterial genes appeared to be specifically required for resistance to NO in vivo. Instead, mutations in several virulence factors caused greater attenuation in the absence of iNOS. Among these were mutants incapable of transporting the lipid phthiocerol dimycocerosate (PDIM). Although PDIM has been implicated in NO defense, this result indicates that PDIM has other roles during infection. We additionally found that PDIM transport is required for virulence in mice lacking interferon-γ. Thus, PDIM is important for resisting an interferon-γ–independent mechanism of immunity.
Mycobacterium tuberculosis is a deadly pathogen that infects >8 million people each year . Macrophage-mediated killing of mycobacteria requires activation by interferon-γ (IFN-γ), which stimulates maturation of the arrested mycobacterial phagolysosome and targets high concentrations of antimicrobial compounds to this compartment . Despite this activity, a subset of bacilli can resist killing and persist for decades, posing a constant threat to host survival .
Mycobacterial growth in macrophages can be limited by the production of nitric oxide (NO), which is stimulated by IFN-γ . NO is potently toxic in vitro and necessary for murine control of M. tuberculosis [5, 6]. During human infection, both inducible NO synthase (iNOS) expression and the products of NO activity, nitrotyrosines, have been detected in lung tissue at sites of M. tuberculosis infection [7, 8]. Mice lacking iNOS activity quickly succumb to disseminated M. tuberculosis infection, and latently infected mice use iNOS to prevent reactivation [6, 9]. To resist iNOS activity, M. tuberculosis has developed defenses that protect it from reactive nitrogen intermediates, as shown by bacterial mutations that cause increased sensitivity to NO in vitro [10–12].
In addition to its role as a cytotoxic component of the macrophage response to M. tuberculosis, NO is a pleiotropic signaling molecule that affects several aspects of immunity . NO can negatively regulate immune signaling through direct nitrosylation of signal transduction machinery or by activation of guanylate cyclase and guanosine 3′,5′-cyclic monophosphate–dependent protein kinase [14–16]. This can lead to the inhibition of T cell development and the impairment of cytokine production and signaling. iNOS expression dramatically influences the transcriptional response to IFN-γ and M. tuberculosis, altering over a third of the genes that respond to these stimuli .
Because of the importance of NO during M. tuberculosis infection, we studied the effect that this molecule has on the bacterial requirements for infection. We did this by identifying bacterial survival requirements that change in the absence of iNOS. We found that mutations that interrupted transport of a complex cell wall lipid, phthiocerol dimycocerosate (PDIM), produce a competitive disadvantage in the absence of iNOS. We further showed that the transport of this lipid is required for virulence in the absence of IFN-γ. Our results show that PDIM transport is required for resisting a host immune mechanism that is independent of 2 key aspects of murine immunity to M. tuberculosis, NO and IFN-γ.
Wild-type M. tuberculosis H37Rv and derivative strains were propagated at 37°C in Middlebrook 7H9 broth containing 10% oleic acid-albumin-dextrose-catalase (OADC; Difco), 0.2% glycerol, and 0.05% Tween 80 or Middlebrook 7H10 agar containing 10% OADC and 0.5% glycerol. Hygromycin (50 μg/mL) and kanamycin (25 μg/mL) were added when appropriate. Escherichia coli strains were grown on lysogeny broth agar containing kanamycin (50 μg/mL), chloramphenicol (20 μg/mL), or hygromycin (100 μg/mL).
All genetic deletions created for this study were generated using homologous recombination and counterselection according to the following strategy. Primers L1 and R1 (table 1) were used to amplify a lox site-flanked hygromycin-chloramphenicol resistance cassette with primers containing homology to the target gene. Primers L2 and R2 (table 1) were used to amplify the products of this reaction to extend homology to the target gene. Lambda red recombination was used to integrate the L2R2 products onto a cosmid containing a region of the H37Rv chromosome that included the target gene, such that the target gene was replaced in the cosmid [18, 19]. Primers HCL1 and HCR1 were used to amplify the resistance cassette with ~1000 base pairs upstream and downstream of the integrated cassette. Primers HCL2 and HCR2 were used to amplify the products of this reaction with extended homology to pJM2 (Genbank accession number FJ009256). The HCL2/HCR2 polymerase chain reaction (PCR) product was recombined onto the pJM2 plasmid using lambda red recombination. Suicide vectors were each treated with 80,000 μJ cm−2 ultraviolet (UV) irradiation using a HL-2000 UV crosslinker (UVP Laboratory Products) to increase homologous recombination efficiency. Five micrograms of each vector were transformed into H37Rv, and double crossover recombinants were selected on agar containing hygromycin and 5% sucrose. PCR analysis and sequencing were used to confirm that the appropriate regions were deleted.
The DrrA complementing plasmid was generated by ligating the drrA open reading frame into pJEB403 (Genbank accession number FJ009257), an integrating plasmid downstream of a constitutive “mop” promoter . Transformation of M. tuberculosis with this plasmid resulted in single-copy integration at the phage L5 attB site.
Female C57BL/6J, IFN-γ−/−, and iNOS−/− mice, 9–13-weeks old (Jackson Laboratories), were maintained in specific pathogen-free conditions in accordance with Harvard University Institutional Animal Care and Use Committee protocols. IFN-γ−/− and iNOS−/− strains were congenic with the C57BL/6J strain. Mice were infected by intravenous tail vein injection, and spleens and lungs were harvested at the indicated time points, as described elsewhere . Bacteria were quantified by plating diluted organ homogenates on 7H10 agar containing the appropriate antibiotics and counting colonies after 18–21 days at 37°C.
RAW 246.7 cells were seeded onto 24-well, tissue culture–treated plates at 2 × 105 cells/well and infected as described elsewhere . Where indicated, cells were treated 24 h prior to infection with or without 100 U/mL IFN-γ and/or 1 mM aminoguanidine. IFN-γ and aminoguanidine were maintained whenever the medium was replaced. After 0 (4 h after infection), 3, and 5 days after infection, monolayers were lysed, and serial dilutions were plated on agar containing the appropriate antibiotic.
An M. tuberculosis transposon library containing ~100,000 independent clones was generated using the MycoMarT7 phage, as previously described . C57BL/6J (wild-type) mice and iNOS−/− mice were infected with 106 colony-forming units (CFU) of the mutant library. At 3 or 4 weeks after infection, ~200,000 CFU of surviving bacteria were recovered from spleen homogenates. Previous work shows that iNOS is expressed and required for M. tuberculosis resistance in splenocytes [6, 24]. Five and 8 mice from each strain were sacrificed at 3 and 4 weeks, respectively. Genomic DNA was purified for each pool, and Cy3- or Cy5-labeled TraSH probes were generated, hybridized to microarrays, and analyzed as described elsewhere [21, 25]. Two technical replicates were generated from each mouse. Data for each time point were averaged and filtered to include only those features that had wild-type or iNOS−/− signal >200 in 13 of the 26 microarrays and were significantly different from 1 (P < .05 by Student’s t test). Full raw data from each microarray is available in the National Institutes of Health Gene Expression Omnibus database. On the basis of previous TraSH experiments, those mutants with iNOS−/−/wild-type ratios that were 2.5-fold above or below 1 were defined as having an altered growth rate in iNOS−/− mice [21, 22].
Fourteen strains with defined mutations were pooled at roughly equal ratios. Six strains were predicted by TraSH to replicate better in iNOS−/− mice than in wild-type mice: ΔRv1680, Δmez, ΔRv0598c, ΔkefB, Rv3425::Tn, and ΔRv1991c. Six strains were predicted to replicate worse in iNOS−/− mice than in wild-type mice: drrA::Tn.1, drrA::Tn.2, fadD10::Tn, ΔRD1 , Δmce1 , and ΔyrbE4A . Two strains were predicted to replicate equally well in iNOS−/− and wild-type mice: Rv1720c::Tn and wild-type H37Rv with pJEB402 . Wild-type and iNOS−/− mice were infected with the pool, and surviving bacteria (~60,000 CFU) were recovered from spleen homogenates (5 mice per strain per time point). Genomic DNA was purified, and the level of individual mutants was quantified relative to the total pool, as determined by sigA levels, using quantitative real-time PCR (qRT-PCR) with SYBR green, an ABI7900HT sequence detection system, and the indicated oligonucleotides sequences (table 1). Serial dilutions of purified PCR product were amplified to generate standard curves of DNA product versus threshold cycle. The relative amounts of DNA produced in the qRT-PCRs were interpolated from these standard curves and expressed as the ratio of target strain to sigA.
Labeling, extraction, and analysis were performed essentially as described elsewhere . In brief, log-phase cultures were grown for 16 h in minimal medium containing [3–14C] propionate. Cell pellets were washed, and the apolar fraction was extracted with petroleum ether. The extracts were then dried under nitrogen and resuspended in the same solvent. PDIM was resolved by thin-layer chromatography with single development of hexane:ether (9:1) and detected using a Phosphoimager.
To identify M. tuberculosis transposon mutations that altered replication in the absence of iNOS, we used a complex transposon pool to intravenously infect iNOS−/− and wild-type C57BL/6J mice. Transposon mutants that survived infection in either strain were recovered from spleens after 3 or 4 weeks and were quantified using a modified form of the TraSH method. These time points were chosen because they allow the development of adaptive immunity and the production of IFN-γ but are also within the expected survival time for iNOS-deficient mice . Mutations that altered competitive fitness in iNOS−/− mice relative to wild-type mice were defined as those producing significantly increased or decreased ratios of iNOS−/−/wild-type probes (defined in Materials and Methods). To ensure reproducibility, we focused our analysis on those genes that gave consistent results at both 3 and 4 weeks after infection (table 2).
These results measure growth in one mouse strain relative to another, but they do not show how each mutation affects growth during infection relative to growth in vitro. We were able to determine this by comparing our results with those of a previous TraSH experiment (figure 1). This allowed us to search for genes that were specifically required to resist NO production during infection. Mutations in such genes would be expected to cause attenuation in wild-type mice, where iNOS is functional, while causing less attenuation in iNOS−/− mice that lack the dominant source of NO during infection. Mutations in 4 genes that produced significant attenuation in wild-type mice at 4 weeks also increased growth in iNOS−/− mice relative to wild-type mice at 4 weeks (Rv1224, Rv0326, Rv1422, and Rv3864; figure 1). None of these mutants was significantly overrepresented in iNOS−/− mice relative to wild-type mice at 3 weeks. The differences observed at these time points could be attributable to variations in the expression of iNOS in wild-type mice or stochastic variations inherent to the analysis.
Surprisingly, several mutations that were attenuating in wild-type mice were further attenuated in mice lacking iNOS (mmpL7, fadE30, embA, drrA, Rv3869, Rv2707, Rv3868, Rv3168, and drrC; figure 1). Mutations in 4 of these genes caused decreased fitness at both tested time points (table 2). Five of these 9 genes are involved in 2 functional groups: PDIM transport (mmpL7, drrA, and drrC) and esx-1 secretion (Rv3868 and Rv3869). Several other genes that are required for these functions showed a similar trend (figure 1).
To assess the predictive value of the TraSH data, we performed a secondary screen with a small group of mutant strains. We pooled roughly equal amounts of each mutant strain and measured the behavior of 9 strains at 1 day and 28 days after infection using qRT-PCR (figure 2). Among these strains were 2 mutants that were predicted to be overrepresented at both time points (ΔRv1680 and Δmez) and 2 mutants that appeared to be overrepresented in the TraSH results but did not meet the stringent criteria used in table 2 (ΔkefB and ΔRv1991c). None of these 4 strains had an apparent advantage in the secondary screen (figure 2A). The enrichment of overrepresented strains in this secondary pool may have obscured subtle differences detected in the TraSH screen, which uses a more complex pool that is unlikely to be influenced by the behavior of individual mutants. However, among the strains that were predicted to be underrepresented in iNOS−/− mice, we found that 2 independent transposon insertions in drrA and a transposon insertion in fadD10 caused relatively decreased fitness in iNOS−/− mice (figure 2B). A strain lacking the RD1 region (which includes much of the esx1 locus) was equally attenuated in wild-type and iNOS−/− mice (figure 2B). This could indicate that strains with transposon insertions in individual genes in the locus behave differently from this deletion strain, as suggested by the TraSH experiment, in which data from individual genes in the esx1 locus did not consistently produce the same phenotype (figure 1).
Previous work characterizing a drrC::Tn insertion mutant indicated that the drrABC operon is involved in transporting PDIM . We compared the lipid content in the wild-type strain, 2 strains carrying independent insertions in drrA, and those same strains with the wild-type drrA gene expressed by an exogenous promoter. We found that lipid 1 and lipid 2 matched the migration patterns for PDIM and phthiodiolone dimycocerosate, respectively (figure 3). Mutants that cannot export PDIM from the cytosol typically produce an increased amount of phthiocerol dimycocerosate relative to phthiodiolone dimycocerosate [29, 30]. The 2 independent drrA::Tn strains produced a similar pattern, and this phenotype was reversed when drrA expression was restored (figure 3). These results support the hypothesis that DrrA transports PDIM across the lipid membrane.
NO plays a major role in macrophage signaling during M. tuberculosis infection . Among the effector responses altered by the absence of iNOS is the respiratory burst, which is elevated in iNOS−/− macrophages . We reasoned that the dramatic effect that NO production has on macrophage signaling might be responsible for the differential growth of drrA mutants seen in iNOS−/− mice. To assess this, we infected RAW 264.7 murine macrophages with a mixture of wild-type M. tuberculosis and drrA::Tn.1. We measured replication rates in unactivated macrophages, IFN-γ–activated macrophages, and IFN-γ–activated macrophages that were pretreated with the iNOS inhibitor aminoguanidine (figure 4). drrA::Tn.1 mutants have a similar competitive disadvantage in wild-type mice fed aminoguanidine and in iNOS−/− mice (data not shown). After growth in macrophages, the drrA mutant strain was equally attenuated under each condition (figure 4). This suggests that altered NO signaling in macrophages is not responsible for the competitive disadvantage observed with the drrA mutants in the absence of iNOS.
Nitric oxide has been shown to inhibit the production of IFN-γ . As a result, the deletion of iNOS can elevate IFN-γ-mediated immunity and increase resistance to specific pathogens [31, 32]. We hypothesized that PDIM transport mutants have a competitive disadvantage in iNOS−/− mice because they are hypersensitive to an IFN-γ–dependent mechanism of immunity that is elevated in iNOS−/− mice. If this is the case, then PDIM transport mutants should gain a greater advantage than wild-type bacteria when IFN-γ is removed.
We tested this prediction by infecting IFN-γ−/− mice with wild-type, drrA mutant and complemented strains (figure 5). Each of these strains replicated to similarly elevated levels in the spleens and lungs of IFN-γ−/− mice relative to wild-type mice (figure 5A). In both wild-type and IFN-γ−/− backgrounds, the mutant strain was significantly attenuated relative to the complemented strain (figure 5B), indicating that the drrA mutant did not gain a greater advantage than complemented bacteria when IFN-γ was removed. On the contrary, there was a trend in both the spleens and lungs towards increased attenuation in IFN-γ−/− mice, although this difference was not sig-nificant. Thus, PDIM transport mutants are sensitive to IFN-γ-independent immune mechanisms.
Although PDIM is critical for virulence, its molecular role during infection remains elusive [30, 33–35]. Previous studies have proposed roles for PDIM in permeability, phagocytic uptake and trafficking, and defense against reactive nitrogen intermediates [29, 35, 36]. We, therefore, expected that PDIM mutants would have a relative advantage in mice lacking iNOS. However, we found the opposite—PDIM transport mutants have a competitive disadvantage when iNOS is removed. Furthermore, we found that a mutant in the PDIM transport gene, drrA, is still attenuated in mice lacking IFN-γ. This shows that M. tuberculosis uses PDIM to resist an immune mechanism that can function independently of reactive nitrogen intermediates and IFN-γ.
Our data show that drrA is similar to drrC and mmpL7 in that it is not required for PDIM production but does affect the total cellular levels of PDIM relative to phthiodiolone dimycocerosate [29, 30]. This suggests a model in which the export of phthiodiolone dimycocerosate across the inner mycobacterial membrane prevents phthiodiolone from being modified to phthiocerol, presumably by removing it from the site of enzymatic activity. Several recent biochemical studies have demonstrated that Rv2951c, Rv2952, and Rv2953 complete the final steps in PDIM biosynthesis [34, 37–39]. We find that mutations in these enzymes do not reduce fitness during infection .
Our screen for differential growth in iNOS−/− and wild-type mice did not clearly identify any virulence factors that are exclusively required for resisting the antimicrobial activity of NO in vivo. This might be attributable, in part, to the stringent requirements used to identify reproducible attenuation and rescue. For example, mutations in Rv1422 consistently caused attenuation in mice and increased replication in iNOS−/− mice at 4 weeks but were not statistically significant at 3 weeks (P=.055). Similarly, mutations in pstP and mez consistently increased replication in iNOS−/− mice at both time points and caused mild attenuation in wild-type mice after 2 weeks, but they did not appear attenuated at the 4-week time point that we used for comparison . Even with more-relaxed criteria, however, we still find that very few mutants seem to be specifically rescued in iNOS−/− mice.
The scarcity of these mutants indicates that defense mechanisms used to resist NO are either redundant or multifunctional. Consistent with this, many virulence factors that have been shown to be involved in NO resistance through in vitro screens are also required for resisting other stresses. For example, mutants lacking sigH are hypersensitive not only to NO but also to heat shock and oxidative stress [10, 40]. Attenuating mutations in proteasome components cause less pathology in iNOS−/− mice, but they do not replicate to higher levels in iNOS−/− mice . Similarly, attenuated mutants lacking uvrB do not exhibit increased replication in mice lacking iNOS alone [11, 12]. In each of these cases, mutations that impart sensitivity to NO also appear to impart sensitivity to other stresses in vivo.
Although we did not find mutants that were rescued in iNOS−/− mice, we found several mutations that cause attenuation in wild-type mice and increased attenuation in iNOS−/− mice. Among these were mutations that disrupted the ability to transport PDIM. We found no evidence for altered drrA mutant susceptibility to NO in vitro (data not shown), in macrophages, or in mice. These observations differ from a previous report, which demonstrated that mutations in fadD26 increase NO susceptibility during macrophage infection . Taken together these observations indicate that individual mutations that affect PDIM have differential effects on NO susceptibilities and that there are also other mechanisms of immunity that restrict the growth of mutants lacking functional PDIM.
Inhibition of caspase signaling by NO has previously been shown to inhibit IFN-γ production, such that the deletion of iNOS increases IFN-γ levels . This elevated IFN-γ makes iNOS−/− mice resistant to infection by Mycobacterium avium and influenza virus [31, 32]. This suggested that PDIM transport mutants might be further attenuated in iNOS−/− mice because they are sensitive to other aspects of IFN-γ-mediated immunity that are up-regulated in this background. This does not appear to be the case, because we found that both wild-type and PDIM transport mutants have similar increases in replication in IFN-γ−/− mice. Consistent with this, we have also found that PDIM transport mutants are not hypersensitive to restriction mediated by the small GTPase Irgm1 or autophagy (data not shown), both of which are IFN-γ–regulated means of controlling mycobacterial growth [41, 42].
What, then, is the immune mechanism responsible for restricting the growth of PDIM mutants? Recent evidence indicates that PDIM can alter receptor-mediated uptake of bacilli by macrophages . However, we show that some mutants lacking PDIM also have reduced growth rates after entering phagocytes, as has been previously reported [22, 33, 43]. Alternatively, the increased permeability observed with PDIM deficient mutants may make them more susceptible to host antimicrobial mechanisms [44–46]. Finally, PDIM mutants may be sensitive to an IFN-γ–independent restriction mechanism that has yet to be identified.
Much of the work on effector mechanisms that restrict M. tuberculosis has focused on IFN-γ–dependent immunity, and correspondingly, much of the work on vaccine development has focused on optimizing IFN-γ production. Although IFN-γ is clearly important, recent work indicates that vaccination can also elicit IFN-γ–independent immunity [47, 48]. Bacteria that lack PDIM could be useful tools in the elucidation of these mechanisms. Bacteria unable to transport PDIM have also been shown to elicit higher levels of protective immunity than Mycobacterium bovis bacillus Calmette-Guérin . Our results show that these mutants are attenuated in severely immuno-compromised mice and suggest that these mutants may be safer than was previously expected. Thus, these mutants may be useful in the development of novel preventative strategies.
Financial support: National Institute of Health (grants AI48704 and AI51929).
We thank Noman Siddiqi, Larry Pipkin, and Ilona Breiterene, for expert technical assistance; David Sherman, for the ΔRD1 mutant strain; and Sarah Fortune, Jeffery Cox, and members of the Rubin lab, for helpful conversations.
Potential conflicts of interest: none.