NK Cell-Derived IFN-γ Differentially Regulates Innate Resistance and Neutrophil Response in T Cell-Deficient Hosts Infected with Mycobacterium tuberculosis

Interferon-γ is required for host control of Mycobacterium tuberculosis by inducing NO synthase type 2 (NOS2)² and other effector molecules in infected macrophages. Both CD4⁺ and CD8⁺ T cells have been shown to be sources of this protective cytokine in M. tuberculosis infection (1). In the case of CD4⁺ T lymphocytes, IL-12 is thought to be a key regulator of IFN-γ production and has been shown to be required for both the induction (2, 3, 4) and maintenance (5) of Th1 effector functions in M. tuberculosis-infected mice. Although NK cells produce IFN-γ rapidly in a cell cycle progression-independent manner (6, 7) and contribute to early resistance to many pathogens, their role in host resistance to M. tuberculosis has not been formally established. Indeed, while IFN-γ-producing NK1.1⁺ cells have been shown to expand in the lungs of M. tuberculosis-infected mice, in vivo depletion of NK cells with mAb does not lead to significant impairment in mycobacterial control (8). This finding suggests that if NK cells play a role in host resistance to M. tuberculosis infection, their function must be redundant with respect to other IFN-γ-producing lymphocytes.

Although IFN-γ produced by NK cells does not appear to be required for normal control of mycobacterial infection, this response may become important in situations in which T cell function is compromised, such as in patients with HIV/AIDS. The latter individuals are highly susceptible to tuberculosis, developing a rapidly progressing combined disease (tuberculosis/AIDS). It is possible that in such HIV-positive populations, NK cells may emerge as major IFN-γ-dependent effectors, and, if so, any functional impairment in the NK compartment would lead to increased risk of primary and reactivated tuberculosis. Conversely, if this hypothesis is correct, immunotherapeutic enhancement of NK function might serve as an approach for improving the clinical outcome of coinfected individuals.

The existence of NK-mediated, T cell-independent mechanisms of host resistance to tuberculosis was suggested by experiments in which T lymphocyte-deficient mice were shown to display prolonged survival relative to IFN-γ^−/− mice following M. tuberculosis infection (9). In the current study, we have formally investigated the role of NK cells in T-independent resistance to the pathogen in a similar T-deficient RAG^−/− mouse model. We found that, in these mice, NK cells are the primary mediator of IFN-γ-dependent, T and B cell-independent resistance to M. tuberculosis infection and play a major role in inhibiting bacterial replication as well as pulmonary pathology. Moreover, by studying infection in p40- vs p35-deficient RAG^−/− mice, we establish a central function for IL-12 in this innate defense mechanism. Finally, our data reveal that in the absence of both T cells and IFN-γ, neutrophils can play a compensatory role in the host response to the pathogen. Together, our findings suggest that NK-mediated IFN-γ production can provide a temporary barrier limiting both M. tuberculosis infection and lung pathology in immunodeficient hosts, and therefore might be a factor influencing disease outcome in patients with tuberculosis/AIDS.

Wild-type (WT) C57BL/6 mice were obtained from the Division of Cancer Treatment, National Cancer Institute. IL-12p35-deficient animals were purchased from The Jackson Laboratory, and IL-12p40, common cytokine receptor γ-chain (γ_c), or RAG2-deficient mice were provided by Taconic Farms from the National Institute of Allergy and Infectious Diseases Animal Supply Contract. IL-12p40^−/−RAG^−/−, IL-12p35^−/−RAG^−/−, and γ_c^−/−RAG^−/− knockout mice were generated by crossing single gene-deficient animals. All mice were maintained in specific pathogen-free conditions at an American Association of Laboratory Animal Care-accredited, BioSafety Level 3 physical containment animal facility at the National Institute of Allergy and Infectious Diseases, National Institutes of Health. Both male and female mice between 8 and 12 wk old were used.

M. tuberculosis H37Rv harvested from infected mouse lungs were expanded in Middlebrook 7H9 broth supplemented with oleic acid-albumin-dextrose-catalase (Difco) for 14 days at 37°C and stored in aliquots at −80°C. Mice were exposed to bacteria via aerosol using a nose-only exposure unit (CH Technologies). Each mouse received ∼100 CFU measured in lung at 24 h after exposure. Bacterial loads in infected organs were quantitated by culture on 7H11 agar (Difco).

Mice were injected i.p. with 0.5 mg of mAb specific to IFN-γ (clone XMG-6) or Gr-1 (RB6-8C5) twice weekly for 28 days starting at the time of M. tuberculosis infection. When spleen and lung cells were examined by flow cytometry, mice injected for 28 days with RB6-8C5 mAb displayed a >95% reduction in CD45⁺CD11b^high cells (neutrophils) with no detectable decline in CD45⁺CD11b^low cells (macrophages/monocytes) (data not shown).

Lungs were perfused with PBS through the heart before removal from mice. Following digestion with Liberase and DNase (Roche Biochemicals), single-cell suspensions were prepared by passing lung tissue through 40-μm nylon cell strainers. To determine the percentage of NK, neutrophil, and macrophage populations, the lung suspensions were stained with mAb to CD49b (DX5; BD Pharmingen), CD11b (M1/70; BD Pharmingen), Gr-1 (RB6-8C5; BD Pharmingen), and CD45 (30-F11), and analyzed by flow cytometry. For enrichment of pulmonary leukocytes, lung cells were centrifuged in 35% Percoll (Pharmacia Biotech) for 15 min at 700 × g. The cell pellets were then collected, and erythrocytes were lysed.

To enrich NK cells, pulmonary leukocytes were first incubated with FITC-labeled anti-CD49b and then MACS beads conjugated with anti-FITC mAb (Miltenyi Biotec). CD49b⁺ NK cells were then positively selected using magnetic cell sorting. The resulting NK cell populations contained >90% NK1.1⁺CD49b⁺ cells as determined by flow cytometry (data not shown).

Mouse spleens were injected with Liberase CI solution (400 μg/ml; Roche) and incubated for 30 min at 37°C. Single-cell suspensions were prepared by forcing the digested tissues through a cell strainer (Falcon), followed by washing in PBS containing 0.5 mM EDTA. Total splenocytes (5 × 10⁶/ml) were then stimulated with M. tuberculosis at various multiplicities of infection (MOI) in the presence or absence of murine rIL-12 (10 ng/ml) (provided by Wyeth Research) for 48 h in RPMI 1640 supplemented with 10% heat-inactivated FCS, 2 μM glutamine, 10 mM HEPES, and 50 μM 2-ME. IFN-γ levels in culture supernatants were then determined by ELISA.

To detect intracellular IFN-γ in RAG^−/− splenocytes, cells were cultured for 24 h, as indicated above, and brefeldin A was added during the last 2 h of incubation. For intracellular staining of IFN-γ in freshly isolated lung cells, pulmonary single-cell suspensions were incubated with brefeldin A for 5 h at 37°C without any further in vitro stimulation. Cells were then surface stained with mAb to CD49b, fixed, and permeabilized. Intracellular IFN-γ was detected with anti-IFN-γ mAb (clone XMG1.2; BD Pharmingen). Data were collected using a FACSCalibur (BD Immunocytometry Systems) with CellQuest (BD Biosciences) and analyzed with FlowJo (Tree Star) software. As a specificity control, it was shown that coincubation with an excess of unlabeled anti-IFN-γ mAb reduced staining to background levels.

Purified pulmonary leukocytes were lysed in 650 μl of RLT buffer (Qiagen) containing 2-ME. Lysates were passed through QIA Shredder columns and RNA isolated on Qiagen mini columns (Qiagen), according to the manufacturer’s protocol. Individual RNA samples (1 μg each) were reverse transcribed using Superscript II (Invitrogen Life Technologies) and a mixture of oligo(dT) and random primers, as previously described (10). Real-time PCR was performed on an ABI Prism 7900 sequence detection system (Applied Biosystems). Relative quantities of mRNA for several genes were determined using SYBR Green PCR Master Mix (Applied Biosystems) and by the comparative threshold cycle method, as described by Applied Biosystems for the ABI Prism 7700/7900 sequence detection systems. In this method, mRNA levels for each sample were normalized to hypoxanthine guanine phosphoribosyl transferase mRNA levels and then expressed as a relative increase or decrease compared with levels in uninfected controls. Primers were designed using PrimerExpress software (Applied Biosystems). The sequences of the specific primers are as follows: KC, 5′-TGTCAGTGCCTGCAGACCAT-3′(forward; F) and 5′-CCTCGCGACCATTCTTGAGT-3′(reverse; R); MIP-2, 5′-GTGAACTGCGCTGTCAATGC-3′(F) and 5′-CGCCCTTGAGAGTGGCTATG-3′(R); and IFN-γ, 5′-AGAGCCAGATTATCTCTTTCTACCTCAG-3′(F) and 5′-CCTTTTTCGCCTTGCTGTTG-3′(R). Primers for Ym1 (11), NOS2, Arg-1, and hypoxanthine guanine phosphoribosyl transferase (12) were published previously.

Tissues were fixed with formalin, sectioned, and H&E stained. The extent of the pulmonary inflammatory response and the percentage of granulocytes within lesions were determined under light microscopy at magnifications of ×12.5 and ×400, respectively. The percentage of granulocytes/lesion was obtained by dividing the estimated number of granulocytes by the total numbers of all nucleated cells in each lesion based on cellular morphology. The extent of necrosis in lungs was assessed microscopically using a scale of 1–5, with 5 representing the most severe necrotic response. All histological analyses were performed on blinded specimens by a trained pathologist who scored all lesions on one lung section for each animal in the experiment.

ANOVA was used to analyze the significance of differences in means between multiple experimental groups. The multicomparison significance level for the one-way ANOVA was 0.05. If significance was detected by one-way analysis, pairwise differences were evaluated using Fisher’s protected least significant difference ANOVA post hoc test. Statistical significance was defined as p < 0.05.

As a first step in evaluating the role of NK function in host resistance to M. tuberculosis infection, we determined whether M. tuberculosis can activate NK cells to produce IFN-γ in vitro and in vivo. Naive splenocytes from T cell-deficient RAG^−/− mice were stimulated for 24 h with live M. tuberculosis or rIL-12 as a control, and IFN-γ production was evaluated by intracellular cytokine staining. Both stimuli induced IFN-γ production only in CD49b⁺ NK cells (Fig. 1,A). In agreement with the above finding, incubation with mycobacteria induced high levels of IFN-γ in supernatants of RAG^−/− splenic cultures (Fig. 1,B). Consistent with previously published experiments on Mycobacterium avium-induced IFN-γ production by SCID splenocytes (13), this response was reduced at higher MOI. In contrast, no IFN-γ was detected when splenocytes from γ_c^−/−RAG^−/− mice, which are deficient in T as well as NK cells, were exposed to M. tuberculosis in the same assay (Fig. 1 C).

To address whether M. tuberculosis can activate NK-dependent IFN-γ production in vivo, RAG^−/− mice were aerogenically exposed to the pathogen and pulmonary immune responses analyzed 4 wk later. M. tuberculosis infection resulted in a major expansion of both CD49b⁺ and CD49b⁻ cells in mouse lung (Fig. 1,D). When IFN-γ mRNA expression was measured by RT-PCR in purified CD49b⁺ and CD49b⁻ cell populations at day 28 postinfection (p.i.), only the CD49b⁺ fraction displayed significant cytokine induction (Fig. 1 E). To confirm the latter finding, freshly isolated lung cells were incubated with brefeldin A for 5 h without in vitro restimulation by mycobacteria and stained for surface CD49b and intracellular IFN-γ. Although only 2% of the total cells from infected lungs were found to be IFN-γ⁺ using this assay, >90% of these were also CD49⁺ (data not shown). Together, these findings indicate that NK cells are the major cellular source of IFN-γ in M. tuberculosis-infected T cell-deficient animals.

To investigate the role of NK cells and IFN-γ in innate resistance to M. tuberculosis infection, RAG^−/−, γ_c^−/−RAG^−/−, and IFN-γ^−/− mice were infected with M. tuberculosis, and their survival was monitored. As reported previously (9), IFN-γ^−/− mice rapidly succumbed to the infection, whereas all RAG^−/− animals survived during the same period of time (Fig. 2 A, upper panel). Importantly, the NK-deficient γ_c^−/−RAG^−/− animals were more susceptible to M. tuberculosis than RAG^−/− mice, succumbing by 30 days p.i. with a median survival time indistinguishable from that of IFN-γ^−/− mice.

To directly demonstrate the importance of T cell-independent IFN-γ in host resistance to M. tuberculosis infection in RAG^−/− mice, we next compared the susceptibility of anti-IFN-γ mAb-treated and untreated RAG^−/− mice to M. tuberculosis infection (Fig. 2 A, lower panel). We found that neutralization of IFN-γ in infected RAG^−/− mice completely abolished the resistance of untreated RAG^−/− animals to the pathogen. To examine the role of IL-12p40 in the NK-dependent early control of M. tuberculosis infection in RAG^−/− mice, we compared the resistance of IL-12p40^−/− and IL-12p40^+/+RAG^−/− animals. The p40^−/−RAG^−/− mice were also found to be highly susceptible to M. tuberculosis and displayed median survival time indistinguishable from that of anti-IFN-γ-treated RAG^−/− or γ_c^−/−RAG^−/− animals.

To confirm that the observed increases in mortality reflect decreased control of bacterial growth, mycobacterial loads were examined at 28 days p.i. in the lungs and spleens of the different groups of RAG^−/− animals. As shown in Fig. 2B, RAG^−/− mice showed higher bacterial burdens than WT animals in both organs, and these differences were increased by 0.5–1.0 log in the anti-IFN-γ mAb-treated and IL-12p40-deficient RAG^−/− animals. Increased susceptibility was also observed in the γ_c^−/−RAG^−/− mice, particularly in the lungs, which displayed a >2 log elevation in CFU relative to lungs from WT animals. Together, these findings argued that T-independent host resistance to M. tuberculosis requires IL-12p40-dependent IFN-γ production by NK cells.

We next asked whether NK cells and IFN-γ regulate the pulmonary inflammatory response to M. tuberculosis in the absence of T lymphocytes. To do so, we examined histological sections of lungs from the different RAG^−/− strains at day 28 p.i. Lungs of RAG^−/− mice displayed a few small foci with limited mononuclear cell infiltration (Fig. 3, A and E). In contrast, the same tissue from NK-deficient γ_c^−/−RAG^−/− mice showed massive lesions occupying ∼60% of the lung space (Fig. 3, B and E). A marked, although milder, exacerbation of pulmonary pathology was also observed in the infected anti-IFN-γ-treated RAG^−/− (Fig. 3, C and E) and IL-12p40^−/−RAG^−/− (Fig. 3, D and E) mice. In the latter two sets of animals, the lesions contained significant numbers of granulocytes in direct contrast to those in the γ_c^−/−RAG^−/− mice (Fig. 3 E). Together, these data suggest that IL-12p40-dependent IFN-γ production by NK cells, in addition to controlling bacterial growth, also dramatically down-regulates the early T-independent pulmonary inflammatory response to M. tuberculosis.

In addition to its association with p35 in the IL-12 heterodimer, IL-12p40 can also associate with p19 to form IL-23. Both IL-12 and IL-23 have been implicated in the generation of protective Th1 responses to M. tuberculosis infection (4, 14). To examine the relative contribution of IL-12 vs non-p35-containing IL-12 family members in NK-dependent innate resistance to M. tuberculosis infection, we compared the responses of p40^−/−RAG^−/− and p35^−/−RAG^−/− mice to the pathogen, first examining in vitro IFN-γ production by splenocytes from uninfected animals. We found that splenocytes from IL-12p35-, IL-12p40-, or γ_c-deficient RAG^−/− mice produced only minimal levels of IFN-γ in response to mycobacterial stimulation when compared with the equivalent cell populations from RAG^−/− animals (Fig. 4,A). Addition of exogenous IL-12 to the M. tuberculosis-stimulated p40^−/−RAG^−/− or p35^−/−RAG^−/−, but not the γ_c^−/−RAG^−/− cultures restored the IFN-γ response, confirming that the impairment in cytokine production occurring in the splenocytes from these animals stems from the absence of IL-12 production rather than a defect in NK development or IL-12 responsiveness (Fig. 4 B).

To formally test whether IL-12 itself or some other p40-containing IL-12 family member is required for T-independent resistance to M. tuberculosis infection, RAG^−/−, p40^−/−RAG^−/−, or p35^−/− RAG^−/− mice were aerogenically infected with M. tuberculosis, and their survival and host responses were compared. As described above (Fig. 2,A), p40^−/−RAG^−/− mice rapidly succumbed to the infection, whereas RAG^−/− survived until day 50 (Fig. 4,C). In contrast, p35^−/− RAG^−/− animals displayed a transiently improved resistance, succumbing ∼3 days following the infected p40^−/−RAG^−/− mice. IL-12p40^−/−RAG^−/− animals displayed small, but significantly increased pulmonary bacterial loads vs p35^−/−RAG^−/− animals when measured at day 28 (Fig. 4,D). Analysis of tissue responses revealed that although the numbers of lesions were higher in the lungs of p35^−/−RAG^−/− vs p40^−/−RAG^−/− mice, the size of the inflammatory foci as well as the percentage of the lung space affected was increased in the latter animals relative to the former (Fig. 4 E). These differences were not reflected, however, in the percentage of neutrophils in the lesions. Together, the above findings argued that T-independent IFN-γ-mediated resistance to M. tuberculosis depends primarily on IL-12.

IFN-γ activates macrophages to produce NOS2 and other molecules required for restriction of intracellular pathogen growth, whereas Th2 cytokines stimulate macrophages to alternatively express YM-1 and l-arginase 1, a modulator of NOS2 synthesis (15). To assess IFN-γ-dependent macrophage effector functions in vivo, we analyzed the expression of mRNA for these macrophage activation markers in pulmonary leukocytes isolated from M. tuberculosis-infected RAG^−/− and anti-IFN-γ mAb-treated RAG^−/− animals. Although M. tuberculosis infection led to significantly elevated IFN-γ-dependent NOS2 expression in the lungs of RAG^−/− mice, this up-regulation was nearly abolished in anti-IFN-γ mAb-treated RAG^−/− animals (Fig. 5). Importantly, the expression of YM-1 and l-arginase 1, which were induced minimally in the lungs of infected RAG^−/− mice, were elevated >60- and 3000-fold, respectively, in anti-IFN-γ-treated RAG^−/− animals. These findings suggested that in the absence of NK IFN-γ production, macrophages in M. tuberculosis-infected mice are alternatively, rather than classically activated, and therefore incapable of controlling bacterial growth.

To investigate whether the development of granulocyte-enriched lesions in p40^−/−RAG^−/− or anti-IFN-γ-treated RAG^−/− mice results from dysregulated cytokine/chemokine responses, we examined mRNA expression of the neutrophil chemotactic chemokines KC and MIP-2 in NK-enriched and -depleted pulmonary leukocyte populations isolated from day 28 infected mice (Fig. 6 A). This analysis revealed that expression of KC and MIP-2 by CD49b⁺ NK cells was markedly increased in anti-IFN-γ-treated RAG^−/− as well as IL-12p40^−/−RAG^−/− animals. In addition to the NK cells themselves, elevated expression of KC and MIP-2 was also detected in the NK-depleted leukocyte populations in the lungs of infected anti-IFN-γ-treated RAG^−/− and IL-12p40^−/−RAG^−/− animals. In contrast to MIP and KC, expression of IL-17, another potent neutrophil-attracting mediator, was equally induced in the CD49b⁺ fraction regardless of IFN-γ neutralization or the presence of IL-12p40 (data not shown).

To assess whether the influx of neutrophils seen in the absence of NK cell-derived IFN-γ plays a role in either host resistance or tissue pathology to M. tuberculosis, we depleted this population in anti-IFN-γ-treated RAG^−/− animals by simultaneous administration of mAb to Gr-1. As shown in Fig. 6,B, the neutrophil-depleted mice displayed significantly elevated pulmonary and splenic M. tuberculosis loads compared with nondepleted, anti-IFN-γ-treated RAG^−/− animals at day 28 p.i. Interestingly, neutrophil-depleted, anti-IFN-γ-treated RAG^−/− animals also displayed severe tissue necrosis compared with anti-IFN-γ-treated RAG^−/− mice (Fig. 6 C). Together, these findings suggest that in the absence of innate IFN-γ, granulocytes play a compensatory role not only in bacterial control, but also in the regulation of pulmonary pathology.

NK cells have been shown to provide an innate barrier to infection to a variety of viral, bacterial, and parasitic pathogens (16, 17). Although in viral infections cytotoxicity is a major mechanism of NK function, NK control of some viruses and many intracellular bacteria and protozoa is thought to be mediated indirectly through cytokine production, a pathway first elegantly demonstrated in the innate response to Listeria monocytogenes (18, 19). Although this cytokine-dependent pathway is required for host resistance to certain viral infections (20), it appears to play a nonessential, but contributory role in the response to other infectious agents, perhaps by controlling early pathogen growth before the generation of adaptive immunity. This protective activity is revealed in hosts in which T lymphocyte function is impaired.

In mycobacterial infection, previous studies on the role of NK cells in host resistance have involved use of Abs that deplete NK populations. Although mice depleted of NK cells by this procedure were initially reported to be more susceptible to M. avium infection (21), this finding could not be reproduced in a later, more comprehensive study (22). Similarly, in a related report on M. tuberculosis infection in which NK activation and recruitment were documented in vivo, no effect of NK depletion on bacterial loads or pulmonary pathology was observed (8). In contrast to these studies involving mice with an intact T cell compartment, experiments in SCID mice have strongly implicated a protective role for NK cells in the host response to mycobacteria. Thus, in M. avium (13) and bacillus Calmette-Guerin (23) infection, SCID mice have been shown to be capable of forming hepatic granulomas, and in the case of the M. avium model this response was demonstrated to be dependent on both IFN-γ and TNF. More importantly, TCRαβ-deficient mice infected with M. tuberculosis were shown to survive longer than IFN-γ-deficient mice (9), a finding strongly suggestive of NK involvement.

The data presented in this study support the above observations and directly document a role for NK-produced IFN-γ in T-independent host resistance to aerogenic M. tuberculosis infection. The key findings supporting this mechanism are that: 1) NK cells are the primary source of IFN-γ in M. tuberculosis-infected RAG^−/− mice and these animals are more resistant to the pathogen than either IFN-γ^−/−- or anti-IFN-γ-treated RAG^−/− hosts; and 2) NK-deficient γ_c^−/−RAG^−/− mice are more susceptible to infection than RAG^−/− animals, displaying a time to death equivalent to IFN-γ^−/− mice. A critical role for NK cells in the T-independent IFN-γ response to M. tuberculosis was further indicated by the failure of γ_c^−/−RAG^−/− splenocytes to produce the cytokine when stimulated with mycobacteria in vitro.

Our findings suggest that NK cell-produced IFN-γ mediates early resistance to M. tuberculosis by multiple mechanisms. First, this innate cytokine response appears to be required for T-independent induction of NOS2, a critical mediator of macrophage microbicidal activity against mycobacteria in mice. Indeed, IFN-γ neutralization in infected RAG^−/− animals resulted not only in a loss in pulmonary NOS2 mRNA expression, but also in a dramatic increase in message levels for YM-1 and l-arginase 1, two markers of alternative macrophage activation, further suggesting a switch in macrophage function to a nonmicrobicidal phenotype (24). Secondly, our results indicate that T-independent IFN-γ production influences innate resistance by regulating pulmonary inflammation during early infection. Thus, M. tuberculosis-infected anti-IFN-γ-treated RAG^−/−, IL-12-deficient RAG^−/−, as well as γ_c^−/−RAG^−/− mice developed severe lung pathology not observed in the same tissues of RAG^−/− animals.

Previous studies in the M. avium model have demonstrated a requirement for NK-produced IFN-γ in granuloma formation in livers of infected SCID mice (13). The findings reported in this work reveal that this NK response can also play a down-modulatory role in regulating neutrophil infiltration in the lungs of M. tuberculosis-infected T-deficient hosts. The neutrophilic inflammation observed correlated with increased expression of mRNAs for the neutrophil chemotactic chemokines, KC and MIP-2, in both the NK-enriched and -depleted pulmonary leukocyte populations. These findings are consistent with previously published observations describing increased neutrophilic infiltration in the absence of IFN-γ signaling in mice infected with M. tuberculosis (25) as well as other intracellular pathogens (26, 27). Although the possible effect of elevated pathogen loads on the induction of neutrophil chemotactic factors cannot be excluded in these situations, our observation that IL-17 mRNA is equally induced in the presence or absence of IL-12-dependent IFN-γ argues that in the case of the model studied in this work, bacterial burden is not in itself a necessary regulator of neutrophil chemotactant expression. Instead, we speculate that IFN-γ may regulate the transcription of KC and MIP-2 in NK cells, a hypothesis consistent with previously published data demonstrating an effect of the cytokine on IL-13 and IL-5 expression by this cell population (28).

In the case of mycobacterial infection, it has been assumed that granulocyte infiltration contributes to pathogen containment because depletion of neutrophils leads to a moderate increase in bacterial loads (29, 30, 31, 32). Our data suggest that the neutrophil response also serves to protect the host from severe pathology, as evidenced by the exaggerated tissue necrosis observed in anti-Gr-1 mAb-treated IFN-γ-depleted RAG^−/− or IL-12p40^−/−RAG^−/− (data not shown) animals. One interpretation of the latter observation is that neutrophils, in addition to their antimicrobial function, also produce factors that play a role in tissue remodeling (33, 34). However, because of the prolonged period of anti-Gr-1 mAb treatment required to maintain neutrophil depletion during the 4-wk course of these experiments, we cannot exclude the possibility that the observed dysregulation in tissue pathology results from the depletion of other cell populations in addition to neutrophils.

Interestingly, although displaying the same time to death, γ_c^−/− RAG^−/− animals showed even greater bacterial burdens and pathology than either anti-IFN-γ-treated RAG^−/− or p40^−/−RAG^−/− mice. Because mice deficient in Fas/Fas ligand, granzyme, or perforin (35, 36) do not display increased susceptibility to M. tuberculosis infection, it is unlikely that the decreased resistance of γ_c^−/−RAG^−/− mice is due to the absence of NK cytolytic function (in addition to NK IFN-γ production) in these animals. Alternatively, their phenotype may reflect the role of additional NK-produced cytokines, such as TNF, in innate control of mycobacterial growth. Finally, it is possible that the increased susceptibility of γ_c^−/−RAG^−/− animals is due to the participation of other γ_c-dependent mechanisms not involving NK cell function. In this regard, it is interesting to note that neutrophil recruitment was impaired in M. tuberculosis-infected γ_c^−/−RAG^−/− animals, a finding consistent with previous observations indicating that certain γ_c-dependent cytokines such as IL-15 (37, 38) and IL-21 (39) can regulate neutrophil responses.

Although IL-12 is clearly important in the generation and maintenance of host control of M. tuberculosis, it has been shown recently that IL-23, which shares the p40 subunit of IL-12, also regulates IFN-γ production by CD4⁺ T cells (14). This finding raises the question of whether other p40-containing IL-12 family members also play a role in initiating NK cell-dependent IFN-γ responses and innate resistance to M. tuberculosis infection. In the present study, we demonstrated that in the absence of either p35 or p40, IFN-γ production by splenic NK cells is significantly impaired in vitro. Moreover, although infected p35^−/−RAG^−/− mice displayed a transiently improved bacterial control and pulmonary pathology compared with p40^−/−RAG^−/− animals, this p35-independent innate resistance was not sustained, and both p40^−/− and p35^−/−RAG^−/− were significantly more susceptible to M. tuberculosis infection than their IL-12- and IL-23-sufficient RAG^−/− counterparts. Thus, NK-mediated innate resistance to M. tuberculosis appears to depend primarily on IL-12, rather than IL-23.

In humans, NK cells have been shown to contribute to IFN-γ production in cultures of mycobacterial stimulated human PBMC (40, 41) and to mediate killing of M. tuberculosis-infected monocytes (42, 43). It is unclear whether human NK cells exert the same effector functions in vivo. However, it was found in a recent study that IFN-γ-producing CD56^bright NK cells are selectively enriched in pleural fluid of tuberculosis patients due to increased susceptibility of cytolytic CD56^dim NK cells to M. tuberculosis-induced apoptotic death (44). The finding suggests that production of IFN-γ at the site of active infection may also be a major effector function of human NK cells.

Previous studies in HIV⁺ individuals have indicated that NK cells from viremic patients show decreased cytolytic activity and IFN-γ production compared with aviremic individuals (45, 46), suggesting that NK cell function is partially impaired as a consequence of AIDS. Therefore, it is possible that defective NK activity could be a contributing factor in the increased susceptibility of HIV⁺ individuals to mycobacterial infection. An additional implication of these and other studies (47) is that interventions that augment NK activity such as IL-12 (48, 49, 50) or IL-15 (51, 52, 53) administration may be clinically useful in the prevention or treatment of tuberculosis in patients coinfected with HIV, particularly when combined with chemotherapy (54). Clearly, further research on NK function in mycobacterial infection in both HIV⁻ and HIV⁺ individuals is necessary to assess the potential utility of such an approach.

We thank Pat Caspar and Sandy White for their excellent technical assistance, and are grateful to Drs. Dragana Jankovic and Cristina Tato for their helpful discussions during the preparation of this publication.

The authors have no financial conflict of interest.