Report

Immune Control of Tuberculosis by IFN-γ-Inducible LRG-47

See allHide authors and affiliations

Science  24 Oct 2003:
Vol. 302, Issue 5645, pp. 654-659
DOI: 10.1126/science.1088063

Abstract

Interferon-γ (IFN-γ) provides an essential component of immunity to tuberculosis by activating infected host macrophages to directly inhibit the replication of Mycobacterium tuberculosis (Mtb). IFN-γ–inducible nitric oxide synthase 2 (NOS2) is considered a principal effector mechanism, although other pathways may also exist. Here, we identify one member of a newly emerging 47-kilodalton (p47) guanosine triphosphatase family, LRG-47, that acts independently of NOS2 to protect against disease. Mice lacking LRG-47 failed to control Mtb replication, unlike those missing the related p47 guanosine triphosphatases IRG-47 or IGTP. Defective bacterial killing in IFN-γ–activated LRG-47–/– macrophages was associated with impaired maturation of Mtb-containing phagosomes, vesicles that otherwise recruited LRG-47 in wild-type cells. Thus, LRG-47 may serve as a critical vacuolar trafficking component used to dispose of intracellular pathogens like Mtb.

Mycobacterium tuberculosis (Mtb) currently infects a third of the human population, claiming more lives each year than any other bacterial pathogen and rivaled only by the acquired immunodeficiency syndrome (AIDS) virus as a communicable cause of death (1). In developing countries, as many as 40 to 80% of individuals with AIDS will also develop tuberculosis (TB), indicating a key role for CD4+ T cells in the immune control of Mtb infection (1, 2). CD4+ T cells secrete IFN-γ that in turn activates macrophages, the host cell in which Mtb chiefly resides and replicates, to inhibit bacterial growth (2). IFN-γ–induced expression of the antimicrobial enzyme NOS2 (3) is largely considered responsible for restricting Mtb replication via NO generation (46). Yet whether NOS2 accounts for all of the activity ascribed to IFN-γ has long been subject to debate (2, 3). This question takes on greater importance in light of three recent developments: (i) the choice of IFN-γ as a surrogate marker for vaccine efficacy and latent TB detection in humans (2, 7), (ii) the discovery of IFN-γ–related genetic mutations that predispose people to TB and other intracellular bacterial infections (8), and (iii) a realization that the response to IFN-γ is far more complex in mammals than was first envisaged (6, 9). Each development highlights the possibility of additional TB defense pathways existing within the broad transcriptional programs elicited by this critical cytokine.

In initial experiments, we observed that NOS2-deficient mice (NOS2–/–) infected with virulent Mtb (Erdman strain) survived significantly longer and exhibited some control of lung Mtb growth when directly compared with mice lacking IFN-γ, IFN-γR1, or the IFN-γR signaling protein STAT-1 (10) (Fig. 1A). This contrasted with the resistance seen in wild-type mice (Fig. 1A) and established the existence of IFN-γ–dependent, NOS2-independent immunity against TB.

Fig. 1.

IFN-γ–dependent, NOS2-independent immunity to TB. (A) (Left) Survival of IFN-γ–/–, IFN-γR1–/–, STAT-1–/–, NOS2–/–, and wild-type (WT) mice (n in brackets) after intravenous infection (105 CFU) with Mtb Erdman. Asterisk indicates P < 0.0001 as compared with NOS2–/– mice. One of two similar experiments shown. (Right) Lung Mtb burdens (mean ± SD) of each group (n = 4 to 5 mice per time point). (B) Lrg-47 (Ifi1), Irg-47 (Ifi47), and Igtp loci (mouse chromosome 11) with 5′ GAS elements denoted. (C) qRT-PCR traces for (top) Mtb-infected lung Lrg-47 mRNA transcripts (day 7 p.i.) and (bottom) duplicate cDNA standard (13). (D) (Top) Lung p47 GTPase and Nos2 expression (13) (mean ± SD) in WT and gene-deficient animals (four to five mice per group). (Bottom) p47 GTPase and Tnf expression in AMΦs after 4 hours of stimulation with different agonists. UT, untreated.

To identify IFN-γ–responsive genes conferring NOS2-independent activity, we turned our attention to recent suppression-subtractive polymerase chain reaction (PCR) and microarray analyses (6, 11, 12). As many as 1300 genes, or 12% of the monitored genome, is transcriptionally elicited after stimulation of macrophages with IFN-γ (6). Of these, mRNAs encoding two new families of guanine triphosphatases (GTPases) predominate: a 47-kilodalton (p47) and a 65-kilodalton [guanylate-binding protein (GBP)] class of proteins (11). The p47 family members (LRG-47, IRG-47, IGTP, TGTP, IIGP, and GTPI) respond rapidly to IFN-γ, probably via direct STAT-1–DNA binding (11, 12). In contrast, GBPs appear later as classical secondary response genes that require interferon regulatory factor–1 (IRF-1) to initiate promoter activity (11). Because the observed NOS2-independent activity against TB was reliant on STAT-1 (Fig. 1A), we focused on members of the p47 family.

Bioinformatic and TESS promoter software analysis (13) enabled murine Lrg-47 (Ifi1), Irg-47 (Ifi47), and Igtp loci to be assembled along with 5′ 1.5-kilobase sequences containing γ-activated sites [GAS; 5′-TTC(N)xGAA-3′ (where N is any nucleic acid and x is from 2 to 4)] (10), the consensus elements used to bind STAT-1 homodimers (Fig. 1B). Quantitative real-time reverse transcription PCR (qRT-PCR) hinted that such elements were used in vivo. Lrg-47, Irg-47, and Igtp genes were all robustly expressed in the lungs of Mtb-infected wild-type and NOS2–/– mice, whereas mRNA transcripts were essentially absent from STAT-1– and IFN-γR1–deficient tissues (Fig. 1, C and D).

A similar pattern emerged at the cellular level. In primary explanted alveolar (AMΦ)- and bone marrow–derived (BMMΦ) macrophages, recombinant mouse IFN-γ (rMuIFNγ) substantially increased p47 GTPase expression (Fig. 1D and fig. S1). Induction was further enhanced if cells were also Mtb-infected, whereas alone Mtb or its cell-wall components [mannosylated lipoarabinomannan, (ManLAM) mycolic acids (MAME), and a 19-kilodalton lipoprotein (LP)] acted only as weak agonists (Fig. 1D and fig. S1). Again, STAT-1 deficiency diminished transcriptional responses to IFN-γ. These findings underscore two points. First, TB serves as a powerful natural stimulus for p47 GTPase induction, evident as early as day 7 postinfection (p.i.) and still detectable >5 months later (Fig. 1D). Second, in vivo and ex vivo expression was contingent on IFN-γR signaling (via STAT-1) and independent of NOS2.

To test whether the marked p47 GTPase response to Mtb infection had functional consequences, we exposed mice harboring null mutations in Lrg-47 (LRG-47–/–) (14), Irg-47 (IRG-47–/–) (14), and Igtp (IGTP–/–) loci (15) to TB either as an aerosol or systemic (intravenous) challenge. Regardless of the route of inoculation, LRG-47–/– mice were profoundly susceptible (Fig. 2B), resembling IFN-γR1–/– hosts even when infected with as little as ∼45 colony-forming units (CFU) (Fig. 2A). Massive coalescent tubercles, often occupying >70% of the available air space, formed within the lungs of LRG-47–/– mice by 4 to 5 weeks p.i. (Fig. 2, A and C). Moreover, Mtb replication continued unabated (Fig. 2D) despite the presence of well-organized granulomas replete with infiltrating lymphocytes (Fig. 2C and fig. S2). In contrast, wild-type, IRG-47–/–, and IGTP–/– groups controlled the disease for many months (Fig. 2B). Both vehicle-treated wild-type and LRG-47–/– controls lived at least 600 to 700 days p.i., making it unlikely that the early TB onset in infected LRG-47–/– mice was because of any underlying growth or developmental defects.

Fig. 2.

LRG-47 provides essential NOS2-independent protection against lethal Mtb infection. (A) Gross pathology of Mtb-infected lungs in aerosol-challenged (∼45 CFU) mice. (B) Survival of individual p47 GTPase-deficient mice after aerogenic (∼500 CFU, left) or systemic [105 CFU intravenously (iv); ∼500 CFU seeded the lungs, right] inocula. Vehicle-treated mice received 0.2 ml of phosphate-buffered saline and 0.05% Tween-80 iv. Individual symbols aligned horizontally at the top of each panel denote mice still alive at the end of the observation period, coinciding with 210 days p.i. for aerosol and 700 days p.i. for intravenous experiments. Double asterisks, P < 0.0005 and triple asterisks, P < 0.0001 as compared with WT. (C) TB progression despite granuloma formation (hematoxylin and eosin stain) (13). Scale bar, 2 mm. (D) Mtb organ burdens in mice challenged as in (B). n = 5 mice per group per time point. Single asterisk, P < 0.001; double asterisks, P < 0.0005; and triple asterisks, P < 0.0001 as compared with WT. (E) Mtb-inducible NOS2 activity (as oxidized end products NO 2 plus NO 3) in individual WT and p47 GTPase-deficient mice (2 × 105 CFU iv). UI, uninfected. (F) (Left) Predicted TB outcome if LRG-47 and NOS2 act in the same or different pathway(s). (Right) Mortality of p47 GTPase–deficient, IFN-γR1–/–, gp91phox–/–, and WT mice given AG (2.5% v/v; 2 × 105 CFU iv). Asterisk, P < 0.001 as compared with AG-treated WT or LRG-47–/– (Fig. 2B, right) and NOS2–/– mice (Fig. 1A) at a similar intravenous dose.

Impaired NOS2 activity could constitute another reason for the heightened TB susceptibility of LRG-47–/– mice. Rac-family GTPases help translocate NOS2 monomers or phagocyte oxidase (phox) subunits for assembly into multimeric enzymes (3, 16, 17). As a small GTPase, LRG-47 may fulfill a similar role. NOS2 activity, however, was intact in LRG-47–/– mice (Fig. 2E), and in fact the antimicrobial effects of LRG-47 and NOS2 could be functionally separated. Here, treatment with the NOS2 inhibitor aminoguanidine (AG) (3) had an additive effect in LRG-47–/– but not in IRG-47–/–, IGTP–/–, or gp91phox–/– mice, accelerating TB progression as compared with AG-treated wild-type animals (Fig. 2F) or drug-free LRG-47–/– (Fig. 2B, right panel, intravenous infection) and NOS2–/– controls (Fig. 1A). This result suggests that LRG-47 and NOS2 act via different mechanisms, because sharing the same pathway would have left TB susceptibility unchanged (Fig. 2F). The comparable survival times of AG-treated LRG-47–/– and IFN-γR1–/– hosts also indicated that the combined actions of LRG-47 and NOS2 account for much of the IFN-γ–dependent immunity to TB (Fig. 2F). In this respect, the contributions made by the antimicrobial phox pathway, itself IFN-γ–responsive (6) and implicated as an important defense against TB (17), appeared negligible (Fig. 2F).

We next sought to ascertain where and how LRG-47 exerts its control of TB. Initial experiments revealed that primary macrophages express high LRG-47 levels, even when treated with 10 U ml–1 of IFN-γ (Fig. 1D and fig. S1). We stimulated BMMΦ with varying amounts of this cytokine and assayed for bacterial killing up to 7 days later. A clear defect in the ability of IFN-γ–activated LRG-47–/– macrophages to eliminate Mtb was evident (Fig. 3A), albeit less than that seen in NOS2–/– macrophages. Loss of resistance was noted at a dose of IFN-γ that elicits strong LRG-47 (Fig. 1D and fig. S1) but weak NOS2 expression (fig. S3A), confirming the distinction between the two pathways. In contrast, macrophages from the other p47 GTPase knockouts retained their tuberculocidal profile (Fig. 3A).

Fig. 3.

LRG-47 inhibits Mtb by a previously unknown antimicrobial mechanism. (A) Impaired ability of IFN-γ–activated LRG-47–/– BMMΦs to kill Mtb (13). Viability (mean ± SD) given as a percentage of the starting (4 hours, t = 0) CFU uptake. Asterisk, P < 0.01 as compared with WT. Representative of three to six experiments. (B) (Top) IFN-γ (100 U ml–1, 48 hours) elicits Nos2, Tnf, P27rx, and Ido expression with limited effect on Tlr2 in WT (left) and LRG-47–/– (right) BMMΦs (RT-PCR, 2.0% agarose gel). (Bottom) IFN-γ–induced macrophage antimycobacterial pathways evaluated at day 7 p.i. (13) (except P2 purinergic receptors, evaluated at 24 hours p.i.) with the use of agonists (P2 receptors, 300 μM ATP; IDO, 320 μM final l-tryptophan) or inhibitors [NOS2, 2.5 mM AG; TNFα, 20 μg ml–1 goat antibody (Ab) against mouse TNF; P2 receptors, 300 μM oxidized oATP; IDO, 1 mM α-Me-dl-tryptophan]. Addition of 20 μg ml–1 goat serum controlled for the TNFα neutralizing Ab. TLR-2–mediated effects in the absence of IFN-γ examined with the use of natural (1 μg ml–1 19-kilodalton LP) or synthetic agonists (1 μg ml–1 Pam3-Cys-Ser-Lys4). The ligand control was 1 μg ml–1 Pam3-Cys-OH. Representative of two to three experiments. (C) IFN-γ (48 hours) suppresses BMMΦ Trfr expression while eliciting Defb3 compared with untreated control cells (qRT-PCR). One of two similar experiments shown.

The higher viability of Mtb in LRG-47–/– macrophages was not due to differences in phagocytic uptake, NO synthesis, secretory export of tumor necrosis factor–α (TNFα), or an oxidative burst (fig. S3), all processes that require the assistance of small Rab-, Rac-, or Rho-family GTPases (16, 17). Indeed, functional expression of several IFN-γ–induced macrophage antimycobacterial pathways—NOS2, TNFα, P2 purinergic (P2X7) receptors, indolamine-2,3-dioxygenase (IDO) (17, 18)—was unimpaired in LRG-47–/– cells as shown via agonist or inhibitor treatment (Fig. 3B). Similarly, IFN-γ–dependent transferrin receptor (TrfR) downregulation, which helps limit iron availability for microbial growth (19), was comparable in LRG-47–/– and wild-type macrophages (Fig. 3C). So, too, was IFN-γ–induced expression of β–defensin 3 (DEFβ3), a homolog of the tuberculocidal human DEFβ2 lysin (17) (Fig. 3C). An antitubercular pathway issuing from Toll-like receptor 2 (TLR2) (20) was also excluded with the use of natural (Mtb 19-kilodalton LP) or synthetic TLR2 ligands (Fig. 3B). Thus, LRG-47 appeared to inhibit Mtb by a previously unknown intracellular mechanism.

To identify the mechanism(s) by which LRG-47 interferes with pathogen survival, we focused on the replicative niche of Mtb: the phagosome (PG). In unactivated cells, this specialized vacuole undergoes maturational arrest, resisting fusion with hydrolase-rich lysosomes and enabling Mtb to avoid the degradative fate that awaits most ingested microbes (21). A paucity of adenosine triphosphate (ATP)–dependent vacuolar proton pump [H+ V–adenosine triphosphatase (ATPase)] subunits on the PG surface is thought to account for its limited acidification (pHpg ∼ 6.0 to 6.5) and reduced protease activity, because these enzymes have acidic pH optima (21). Once macrophages become activated with IFN-γ, however, the block in PG maturation is largely overcome, and they sequentially fuse with late endosomes and/or lysosomes, leading to acidification (pHpg ∼ 4.5 to 5.0), processing of immature hydrolases, and death of PG-encircled mycobacteria (21).

The drop in pHpg that normally accompanies IFN-γ activation was partly attenuated in LRG-47–/– macrophages, stabilizing at ∼1.0 pH unit higher than that in wild-type and NOS2–/– cells or in macrophages with recessive mutations in the Nramp1 locus (13) encoding a phagosomal H+-divalent cation antiporter (Fig. 4A). The altered acidification in LRG-47–/– macrophages appeared to be confined to Mtb PGs, because endosomal (pHend) and cytosolic (pHc) pH were unaffected (fig. S4, A and B). An elevated pHpg due to intralumenal anionic charges generating a Donan equilibrium (22) was also discounted, because treatment of purified LRG-47–/– Mtb PGs with the H+ ionophore CCCP (carbonyl cyanide-m-chlorophenyl hydrazone) allowed cation efflux (fig. S5A). Hence, LRG-47–/– Mtb PGs were not at electrochemical equilibrium most likely because of the action of H+ V-ATPase pumps. Dissipation of the proton gradient using inhibitors [bafilomycin A1 and omeprazole] confirmed that V-ATPases were still functional on LRG-47–/– PGs (Fig. 4A). Neither a loss of counterion conductance, needed to provide electrogenic equivalents for H+ pumping (22, 23), nor excess H+ consumption by superoxide dismutation, generating H2O2 (22), could account for the higher pHpg of LRG-47–deficient cells (figs. S3C and S5B).

Fig. 4.

Arrested maturation of LRG-47–/– PGs. (A) Limited acidification (mean ± SD) of Mtb PGs in IFN-γ–treated (48 hours) LRG-47–/– (Bcgr) as compared with WT (Bcgr), NOS2–/– (Bcgr), and NRAMP1–/– [Bcgs, NRAMP1G169D (where G169D indicates Gly169 → Asp169)] BMMΦs. Lysosomotropic agents (Bafilomycin A1 and omeprazole) dissipated the pH gradient (6 hours shown) without affecting bacterial uptake. UT, untreated. Single asterisk, P < 0.01 and double asterisks, P < 0.005 as compared with WT. One of four independent assays shown. (B) Purified Mtb PG composition (20 min p.i.). Brefeldin A(BFA) added 1 hour preinfection. Samples (5 μg per lane) were Western blotted with Abs to LRG-47, TrfR, V-ATPase E subunit, and Rab5. Anti-LAMP1 indicates the amount of the vacuolar material loaded. (C) Fusion (mean ± SD) of biotin-labeled-Mtb PGs with preloaded avidin–horseradish peroxidase endosomes or lysosomes in BMMΦs (13) treated as in (A). The 4°C fusion assays served as controls. Single asterisk, P < 0.008 as compared with WT. One of three experiments shown. (D) Heightened accessibility of LRG-47–/– Mtb PGs to hTrf–fluorescein isothiocyanate in BMMΦs. Single asterisk, P < 0.01 and double asterisks, P < 0.001 as compared with WT. One of two experiments shown. (E) (Left) Electron micrograph of Mtb PGs fused with bovine serum albumin–Au+ (15 nm)–loaded lysosomes in IFN-γ–treated (48 hours) BMMΦs at 45 min p.i. Scale bar, 0.5 μm. Arrows depict 15-nm gold particles. Blue boxes, lysosome-fused PGs; red boxes, unfused PGs. (Right) Quantitation of PG-lysosome fusion (mean ± SD) from sets of 120 Mtb PGs counted for each condition. UT, untreated. Single asterisk, P < 0.01 and double asterisks, P < 0.002 as compared with WT.

A more likely explanation rests with the lower V-ATPase subunit levels found on the LRG-47–/– Mtb PG surface (Fig. 4B). In wild-type macrophages, immunologically inducible V-ATPase E subunits (23) were recruited to PGs via cargo traffic exiting the golgi and/or endoplasmic reticulum (ER), because brefeldin A (BFA) largely blocked their acquisition (Fig. 4B) (24). LRG-47 recruitment to the Mtb PG surface followed a similar route (Fig. 4B). This may resemble other p47 GTPases, IGTP and IIGP, which at least in uninfected cells are also ER- and golgi-associated (25, 26). In the absence of LRG-47, however, Mtb PGs still retained low levels of TrfR and Rab5 (Fig. 4B), endosomal proteins synonymous with immature, nonfused PGs (21, 27).

Thus, a more general model of aberrant or arrested PG development could account for the higher pHpg of LRG-47–/– macrophages, a hypothesis confirmed by direct intact cell trafficking assays (13). Here, LRG-47–/– Mtb PGs were found to be selectively impaired for fusion with lysosomes (Fig. 4, C and E) while accessible to fluoresceinated Trf, a marker of recycling endosomes (27) (Fig. 4D). Moreover, bafilomycin A1 and the weak base ammonium chloride (NH4Cl) (28) had limited effects in knockout macrophages but inhibited PG-lysosome fusion in wild-type cells (Fig. 4C) (29), suggesting that trafficking may already be partly compromised in LRG-47–/– cells.

Is the altered PG profile likely to account for the profound TB susceptibility of LRG-47–/– mice? Mtb is unusual among Mycobacterium sp. in its sensitivity to acidic pH (<5.5 to 6.0), especially when Mg2+ is also limiting (30). Such conditions probably exist inside Mtb PGs, because acid-sensitive genes are turned on (31, 32) and bacterial mutants lacking the mgtC (Mg2+ transporter) gene are severely attenuated within this environment (30). Indeed, blocking lysosome fusion and alkalinizing the PG lumen with NH4Cl or omeprazole helped restore Mtb viability in IFN-γ–activated wild-type macrophages (fig. S6). This evidence supports the idea that fully mature, hydrolytically competent PGs provide a hostile and nutrient-restrictive dwelling for Mtb (21). Faced with these challenges, Mtb appears to have evolved means of inhibiting IFN-γ–dependent host transcription (33) or switching to alternate metabolic pathways within activated macrophages (32).

Our finding of LRG-47 as critical to innate immunity, equal in importance yet functionally distinct from NOS2, TNFα, NRAMP1, phox, IDO, defensins, and TLR2-mediated killing, opens up avenues for understanding the ancient relationship between Mtb and its human host (17). Indeed, the existence of human LRG-47 homologs (29) suggests that this GTPase could fulfill a similar role during natural TB infection.

Supporting Online Material

www.science.org/cgi/content/full/302/5645/654/DC1

Materials and Methods

Figs. S1 to S6

References and Notes

View Abstract

Navigate This Article