Deletion of Two-Component Regulatory Systems Increases the Virulence of Mycobacterium tuberculosis

Abstract

Two-component regulatory signal transduction systems are widely distributed among bacteria and enable the organisms to make coordinated changes in gene expression in response to a variety of environmental stimuli. The genome sequence of Mycobacterium tuberculosis contains 11 complete two-component systems, four isolated homologous regulators, and three isolated homologous sensors. We have constructed defined mutations in six of these genes and measured virulence in a SCID mouse model. Mice infected with four of the mutants (deletions of devR, tcrXY, trcS, and kdpDE) died more rapidly than those infected with wild-type bacteria. The other two mutants (narL and Rv3220c) showed no change compared to the wild-type H37Rv strain. The most hypervirulent mutant (devRΔ) also grew more rapidly in the acute stage of infection in immunocompetent mice and in gamma interferon-activated macrophages. These results define a novel class of genes in this pathogen whose presence slows down its multiplication in vivo or increases its susceptibility to host killing mechanisms. Thus, M. tuberculosis actively maintains a balance between its own survival and that of the host.

Mycobacterium tuberculosis is an extraordinarily successful pathogen that currently infects approximately one-third of the global population and causes 8 million new cases of tuberculosis annually (41). The tubercle bacillus functions within several hostile environments in order to survive within the human host and cause disease. For instance, the bacteria must be able to gain entry into macrophages, multiply intracellularly, survive within the lung granuloma for years, and disperse to a new host via aerosols (9). It is therefore likely that the expression of different sets of genes by M. tuberculosis at various stages of infection is crucial to its survival.

Two-component regulatory systems (2CRs) are widely distributed among bacteria and plants and enable the organisms to respond to many different external stimuli (20, 37, 38). These systems form a large family of related proteins that consist of a membrane-bound sensor protein that activates an effector protein, generally a transcriptional regulator, by phosphorylation. 2CRs have been shown to play a crucial role in the controlled expression of virulence genes in other bacteria (11). For example, in Salmonella enterica serovar Typhimurium, the membrane-bound sensor protein PhoQ is activated by changing levels of magnesium, and this activates the PhoP response regulator. PhoP is a DNA-binding protein that binds in a sequence-specific manner to selected promoters, thereby inducing or repressing their transcription, and S. enterica serovar Typhimurium phoP mutants are highly attenuated (14, 17).

The genome of M. tuberculosis contains 11 paired 2CRs, four isolated regulators, and three isolated sensors (6). Mutants in two of these, phoP and mprA are highly attenuated in a murine model (31, 43), while a prrA mutant grows more slowly in macrophages (12), indicating that these regulators control genes which are important for successful infection by the pathogen. In this work, we have investigated the role of other M. tuberculosis 2CRs in infection by systematically constructing defined mutations in these systems and phenotypically characterizing the resultant mutants in vivo and in vitro.

MATERIALS AND METHODS

Growth of bacteria and extraction of DNA.

M. tuberculosis was grown at 37°C in Middlebrook 7H9 medium (Difco) supplemented with 10% (vol/vol) oleic acid-albumin-dextrose-catalase (Becton Dickinson) and 0.05% (wt/vol) Tween 80 or on Middlebrook 7H10 agar (Difco) supplemented with 10% (vol/vol) oleic acid-albumin-dextrose-catalase. Hygromycin at 100 μg/ml, kanamycin at 20 μg/ml, and X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) at 50 μg/ml were used where appropriate. DNA for Southern analysis was extracted as described previously (2).

Mutant construction.

Delivery vectors were constructed using the pNIL and pGOAL series vectors (30). Mutants were constructed using a two-step strategy as previously described (30). Briefly, 1 to 5 μg of vector DNA was pretreated with UV to stimulate homologous recombination and used to electroporate M. tuberculosis (19, 30). Single-crossover strains were selected on agar containing hygromycin, kanamycin, and X-Gal. An individual colony was streaked out onto agar (without antibiotics) to allow the second crossover to occur. The cells were resuspended in medium, and serial dilutions were plated onto X-Gal with 10% (wt/vol) sucrose. Sucrose-resistant white colonies were tested for kanamycin sensitivity and analyzed by PCR and Southern hybridization. Genomic DNA was prepared according to the method of Belisle and Sonnenberg (2). Southern hybrizidation was carried out using the AlkPhos Direct kit (Amersham) according to the manufacturer's instructions. In this way, individual strains carrying six unmarked and one marked mutation were isolated (Fig. 1 and Table 1).

FIG. 1.

Genomic contexts of 2CRs analyzed in this study. The genomic contexts of the two-component system genes are shown. Response regulators are shown as open arrows, sensors are shown as hatched arrows, and other genes are shown as solid arrows. The open boxes show the regions deleted in the mutants made, and the names of the mutants are given alongside. All were unmarked deletions, except for tcrXY, which contained a deletion and an additional insertion of a hygromycin resistance gene. In the case of mtrB, the open box shows the region that we attempted to delete.

TABLE 1.

M. tuberculosis H37Rv strains used

Name	Genotype	Description
H37Rv		Wild-type laboratory strain (ATCC 25618)
Tame12	trcSΔ	882-bp BalI deletion in sensor
Tame13	tcrXY::hyg	786-bp EcoRV-StuI deletion in sensor and regulator; adjacent hyg insertion
Tame14	kdpDEΔ	1,691-bp SphI deletion in sensor and regulator
Tame16	devRΔ	447-bp BalI deletion in regulator
Tame19	Rv3220Δ	1,170-bp SphI deletion
Tame37	narLΔ	317-bp PmlI-SmaI deletion in regulator
Devcomp5	Tame16::pSOUP86	Contains integrating plasmid carrying devR

Complementation of devRΔ.

In order to complement the devR mutation in Tame16, the integrating vector pSOUP86 was made by cloning a 1.7-kb region containing 150 bp upstream, Rv3134c, and devR, but not devS, into the unique NarI site of pUC-Gm-Int (24).

In vitro tests.

Mutants were tested in a variety of in vitro situations, none of which showed any differences from the wild type.

(i) Extreme pH.

Liquid medium was adjusted to pH 2 with HCl or to pH 12 with NaOH. Strains were inoculated into 10 ml of liquid medium to a theoretical optical density at 600 nm (OD₆₀₀) of 0.05. The cultures were incubated standing at 37°C. Serial dilutions were plated to determine colony counts at 2, 4, and 24 h.

(ii) Hydrogen peroxide.

Cultures were diluted to a theoretical OD₆₀₀ of 0.01 in phosphate-buffered saline and treated with 5 mM H₂O₂ (final concentration) for 4 h at 37°C. Serial dilutions were plated to determine colony counts.

(iii) Long-term stationary-phase culture.

Strains were inoculated into 10 ml of medium and incubated standing at 37°C. Serial dilutions were plated to determine colony counts over several months.

(iv) Total starvation.

Strains were inoculated into 10 ml of sterile water to a theoretical OD₆₀₀ of 0.05. The cultures were incubated standing at 37°C. Serial dilutions were plated to determine colony counts over several weeks.

In vivo studies.

Experimental infections of SCID and DBA/2 mice and tissue analyses were carried out as described previously (36). For survival analysis, groups of six mice were infected with 10⁶ viable mycobacteria in 200 μl of normal saline via a lateral tail vein. The level of inoculum was routinely checked by CFU analysis. Where appropriate, infected mice were killed by cervical dislocation in accordance with humane endpoint protocols under the Animals Scientific Procedures Act, 1986 (United Kingdom). Median survival times were calculated for each group, and statistical analysis was performed using the log rank tests of survival.

For tissue analysis, lungs, livers, and spleens were collected aseptically from three mice per group into 10 ml of medium and passed through a 100-μm-pore-size sieve (Falcon) in 7H9 medium containing 0.05% Tween 80. Serial 10-fold dilutions were plated in 100-μl volumes, and CFU were counted after 4 weeks and checked at 6 weeks to allow for any change in the growth rate for the mutants ex vivo.

Macrophage infections.

Bone marrow-derived macrophages from BALB/c mice were isolated and infected in the absence of penicillin and streptomycin as described previously (36). Macrophage monolayers were prestimulated with gamma interferon (IFN-γ) (Gibco) at a concentration of 200 U/ml for 4 h prior to infection. The cells were infected for 4 h, then washed six times in warm tissue culture medium. The infection dose was assayed independently by plating the inoculum. The number of viable mycobacteria was assessed by lysis of the macrophage monolayer with 1 ml of sterile distilled water containing 0.1% Triton X-100 per well, followed by plating the bacteria on Middlebrook 7H10 plates.

RESULTS

We initially selected six paired 2CRs as targets for gene disruption: devRS (Rv3132c/Rv3133c), kdpDE (Rv1027c/Rv1028c), mtrAB (Rv3245c/3246c), narLS (Rv0844c/Rv0845), tcrXY (Rv3764c/Rv3765c), and trcRS (Rv1032c/Rv1033c) (Fig. 1). We also selected one isolated sensor (Rv3220c). These gene names have all been used in other publications, except narS and tcrXY, for which we use our own nomenclature. All response regulators belong to the class 2 DNA-binding regulator family (29) typified by Escherichia coli OmpR, except for DevR and NarL, which belong to the class 3 (LuxR) family typified by E. coli NarL. Little is known about the functions of these genes, and only the kdpDE genes show convincing homology to a well-characterized system; in E. coli, this is an emergency high-affinity potassium transport system synthesized only at low potassium concentrations (40).

Six mutants were successfully constructed (trcSΔ, kdpDEΔ, narLΔ, devRΔ, tcrXYΔ::hyg, and Rv3220cΔ) (Fig. 2), but we were unable to isolate an mtrAB mutant, although we tried to generate both marked and unmarked deletions. No differences were observed between the growth rates in axenic culture of the mutants and the wild type. In addition, bacterial survival under a variety of culture conditions, including extreme pH, long-term stationary-phase culture, total starvation, and hydrogen peroxide, was not affected by any of the mutations (see Materials and Methods) (data not shown).

FIG. 2.

Southern blot analysis of mutants. DNA was extracted from wild-type (WT) and mutant (mut) strains and analyzed. The gels show diagnostic digests that confirm their genotypes. devR, EcoRI/XhoI digest probed with 0.8-kb NotI fragment; trcS, XhoI digest probed with 2.3-kb XhoI fragment; trcXY, EcoRI digest probed with 2.5-kb SmaI fragment; kdpDE, XhoI digest probed with 3.7-kb XhoI fragment; narL, BamHI digest probed with 2.4-kb mutagenesis fragment; Rv3220c, XhoI digest probed with 1-kb mutagenesis fragment.

In vivo assays.

We proceeded to test the ability of each mutant to cause disease in the SCID mouse model (Fig. 3A and B). We had previously shown that this is a rapid and sensitive model for assaying changes in mycobacterial virulence, with mutants that are attenuated in these mice also being attenuated in immunocompetent mice (16, 36). Infection with two of the mutants (narLΔ and Rv3220Δ) resulted in the same time to death as with the wild-type strain. In contrast, the other four mutants (devRΔ, tcrXYΔ, trcSΔ, and kdpDEΔ) all showed an increase in virulence, with significantly shorter survival times (P < 0.001). This contrasts with all previous studies of this pathogen, in which gene deletion has resulted in attenuation or no change in virulence (3, 5, 7, 12, 15, 21, 27, 31, 36).

FIG. 3.

Virulence of mutant M. tuberculosis strains in SCID mice. The mice (six per group) were infected i.v. with 10⁶ wild-type or mutant bacteria. (A) Solid squares, wild-type H37Rv (median survival time [MST], 40.5 day); solid triangles, devRΔ (MST, 30.5 days; P < 0.0001); open inverted triangles, tcrXYΔ (MST, 33.5 days; P < 0.0001); crosses, trcSΔ (MST, 35 days; P < 0.001); diamonds, kdpEΔ (MST, 35 days; P < 0.001); circles, Rv3220Δ (MST, 40 days; no significant difference [NS]). The initial inocula were 4 × 10⁶ (H37Rv), 5 × 10⁶ (devRΔ), 2.5 × 10⁶ (tcrXYΔ), 2 × 10⁶ (trcSΔ), 2.5 × 10⁶ (kdpEΔ), and 2 × 10⁶ (Rv3220Δ) bacteria. (B) Solid squares, wild-type H37Rv (MST, 34 days); solid circles narLΔ (MST, 35.5 days; NS). The initial inocula were 8 × 10⁵ and 8.9 × 10⁵, respectively. (C) Complementation of the M. tuberculosis devR mutant. Solid squares, wild-type H37Rv (MST, 34 days); solid triangles, devRΔ (MST, 27 days); open triangles, complemented devRΔ (MST, 35.5 days). The initial inocula were 5.4 × 10⁶, 7.2 × 10⁶, and 8.1 × 10⁶, respectively. The differences between the mutants and the other strains were significant at a level of <0.001. The results for each mutant are representative of two or three individual experiments.

The most hypervirulent mutant in the SCID model, devRΔ (median survival time, 30.5 days compared to 40.5 days for the wild-type bacteria) was studied further. First, the role of devR in causing the increase in virulence was confirmed by complementation. An integrating plasmid carrying the Rv3134 and devR genes expressed from the promoter upstream of Rv3134 was introduced into devRΔ, producing strain Devcomp5. The SCID experiment was repeated and clearly showed that the increased virulence shown by the devRΔ mutant was reduced to the level of the wild type in the complemented strain (Fig. 3C).

The phenotype of the devRΔ mutant was then examined in immunocompetent DBA mice. Bacterial loads in different organs were counted 15, 30, and 60 days following infection (Fig. 4). The numbers of bacteria were elevated in the mice infected with the devRΔ mutant in the lung, liver, and spleen at all time points. This was statistically significant in the lung at 15 days and in all three tissues after 30 days. By day 60, the numbers of devRΔ bacilli recovered from all three organs were still higher than in the controls, but significantly so only in the liver. The increase in bacterial load was also observed by histological analysis and was coincident with increased inflammatory responses (data not shown).

FIG. 4.

Virulence of M. tuberculosis devRΔ in DBA mice. Mice (three per group) were infected i.v. with 10⁶ wild-type or mutant bacteria. The P values of bacterial loads that differ significantly from those of the wild type are given above the relevant bars. The results represent means plus standard errors for three mice per group, with significance measured using Student's t test. Solid bars, wild-type H37Rv; open bars, devRΔ. *, P < 0.05; ***, P < 0.0001. The results shown are representative of three individual experiments.

Macrophage studies.

One of the most important early effector cells in controlling infection by M. tuberculosis is the macrophage. Although M. tuberculosis is able to survive in resting macrophages by inhibiting the maturation of phagosomes (32), macrophages that are activated by IFN-γ are able to kill the bacteria using a variety of methods, including tumor necrosis factor alpha and reactive oxygen and nitrogen intermediates (1, 34). We therefore compared the abilities of activated macrophages to control the growth of wild-type and devRΔ bacilli. Whereas wild-type bacteria were killed rapidly, there was a reproducible rise in the numbers of devRΔ bacteria in the first 24 h after infection (Fig. 5); we saw a 13% increase in the growth of devRΔ bacteria in 24 h compared to a reduction of 50% in wild-type bacteria. This effect was abrogated by complementation of the mutation. As an additional control, we showed that activated macrophages were able to kill narLΔ bacilli with kinetics similar to those of wild-type bacteria, mirroring the in vivo results with this strain.

FIG. 5.

Virulence of M. tuberculosis devRΔ in activated murine macrophages. Squares, wild-type H37Rv; Solid triangles, devRΔ; open triangles, complemented devRΔ; circles, narLΔ. The bacterial counts after 24 h were significantly different in the devRΔ and wild-type strains using an unpaired t test (P < 0.001); this is an experiment representative of three, all of which showed devRΔ to be hypervirulent at 24 h.

DISCUSSION

We chose 2CRs as targets for gene inactivation, as they have been implicated in the virulence of many other pathogens. M. tuberculosis has 11 2CRs, and we reasoned that they may play an important role in the survival of the pathogen. We selected seven systems (six 2CRs and one isolated sensor) as targets, of which six were amenable to gene disruption, indicating they were not essential for growth in culture. We were unable to isolate a mutant lacking the mtrB gene (8, 39), suggesting that the gene is essential. This is consistent with the studies of Zahrt and Deretic (42), who were unable to isolate a mutant lacking the mtrAB system. The reason for the essentiality of this system is unknown, as are the genes which the system controls.

SCID mice infected with four of the six mutants (devRΔ, trcSΔ, tcrXYΔ, and kdpDE) succumbed to infection more rapidly than those infected with wild-type M. tuberculosis. SCID mice act as a model for assaying the innate immune response; we have previously found that mutants that show altered responses in SCID mice show similar changes in immunocompetent mice. However, this preliminary screen will miss effects that are due to the acquired immune response. Thus, the increase in virulence in SCID mice was seen in the absence of any T-cell-mediated response, although these mice retain the capacity to generate organized inflammatory cell responses surrounding foci of bacteria.

We carried out further experiments on the devRΔ mutant, which showed the greatest hypervirulence in the SCID model. This gene was originally identified by its differential expression in the M. tuberculosis strain H37Rv and its avirulent derivative H37Ra (23). We saw increased bacillary load in the devRΔ mutant-infected immunocompetent mice. Significantly greater numbers of devRΔ bacteria (∼8 to 10-fold) than wild-type bacteria were seen on day 15 in the lungs and on day 30 in the lungs, liver, and spleen. By day 60, the wild-type bacteria in the lungs had almost reached the levels seen with the mutant, though the numbers of CFU in the liver were still significantly higher. These early differences in CFU in immunocompetent mice are consistent with the SCID mouse results, as T-cell-specific controlling immunity only develops 21 to 28 days after infection. As with the SCID data, the CFU differences observed were not due to an intrinsic difference in the axenic growth rate, as we described earlier. Sherman et al. (35) also reported no difference in the in vitro exponential-phase growth of a Mycobacterium bovis BCG strain lacking devR activity, although stationary-phase survival was reduced.

In order to investigate the hypervirulent phenotype in a more defined model, we investigated the growth of the devRΔ mutant in activated macrophages. Again, the growth of the bacteria was greater than that of the wild-type strain. This is the first time to our knowledge that mycobacteria have been found to increase in number in IFN-γ-activated macrophages. The phenotype was consistently seen at 24 h (Fig. 5) but seemed to be lost by 72 h (data not shown). Our results did not show whether the increase in bacteria compared to the wild type was due to increased growth, reduced killing, or a combination of the two. The transient loss of susceptibility to killing was not due to a deficiency in the production of nitric oxide or tumor necrosis factor alpha, previously described as potential key mechanisms of mycobacterial killing (1, 4). Both of these were in fact elevated in supernatants collected at 24 h from devRΔ mutant-infected macrophages compared to those infected with the wild-type strain (data not shown).

It has been reported that virulence in a trcS mutant is unaltered (12), which is inconsistent with our results. There are a number of experimental differences which may explain this discrepancy. We tested our trcS mutant using intravenous (i.v.) inoculation of SCID mice and assayed survival, while Ewann et al. used aerosol infection of C57Bl/6 mice and assayed CFU in the lungs. In addition, the strains of M. tuberculosis were different, as we used H37Rv whereas they used strain Mt103. Further work will be needed to clarify this; in particular, it would be valuable to test our mutants using an aerosol model.

Our results, together with the previous reports on mprA, phoP, and prrA, show that a large proportion of 2CR mutants are attenuated or hypervirulent, demonstrating the critical roles 2CRs play in determining the outcome of mycobacterial infection. Virtually nothing is known about the mechanisms by which this is achieved, either at the level of the signals they respond to or the genes they modulate. It is difficult to predict the signals to which 2CRs respond; infection and multiplication within a host requires the ability to adjust to multiple stimuli, thereby exposing the bacteria to many different microenvironments. Of the systems studied here, we know that transcription of devR is induced directly or indirectly by hypoxia, a signal that is associated with persistence (35). Our results suggest either that hypoxia is relevant at more than one stage of the infection or that M. tuberculosis embarks down the road to a persistent state much earlier than has been previously suggested. The trcR system has been reported to be induced in early exponential phase (18), while the kdp system appears to be orthologous to the E. coli system, which is a high-affinity potassium transport system (40). This is induced during potassium-limited growth and is also affected by oxidative stress (33). The further use of a macrophage killing model with combinations of bacterial and murine mutants may help us to elucidate both signals and killing mechanisms.

It has been shown that there is a spectrum of virulence in clinical isolates (22, 25, 28), which presumably is related to loss or alteration of genes. This work represents the first report of an increase in mycobacterial virulence following targeted gene deletion. We propose that the four regulatory systems identified here, inactivation of which leads to hypervirulence, are involved in suppressing the growth rate of M. tuberculosis at early stages of infection. It would be expected that most gene deletions will lead to pathogens that are less virulent. The few precedents for gene knockout causing increased virulence are found in regulatory genes, where mutations lead to derepression of virulence genes (10, 13). We could be observing the same phenomenon, as we too have studied regulatory mutations. However, it is remarkable that this phenotype was seen in several of the mutants, while we have recently reported that insertion of a transposon into a nonregulatory gene can also lead to hypervirulence (26). This suggests that we are observing an important and previously undescribed aspect of the biology of M. tuberculosis.