A Family of IFN-γ–Inducible 65-kD GTPases Protects Against Bacterial Infection

See allHide authors and affiliations

Science  06 May 2011:
Vol. 332, Issue 6030, pp. 717-721
DOI: 10.1126/science.1201711


Immune interferon gamma (IFN-γ) is essential for mammalian host defense against intracellular pathogens. IFN-γ induces nearly 2000 host genes, yet few have any assigned function. Here, we examined a complete mouse 65-kilodalton (kD) guanylate-binding protein (Gbp) gene family as part of a 43-member IFN-γ–inducible guanosine triphosphatase (GTPase) superfamily in mouse and human genomes. Family-wide loss-of-function analysis found that at least four Gbps—Gbp1, Gbp6, Gbp7, and Gbp10—conferred cell-autonomous immunity to listerial or mycobacterial infection within macrophages and gene-deficient animals. These Gbps solicited host defense proteins, including the phagocyte oxidase, antimicrobial peptides, and autophagy effectors, to kill intracellular bacteria. Thus, specific 65-kD Gbps coordinate a potent oxidative and vesicular trafficking program to protect the host from infection.

Immune interferon gamma (IFN-γ) is critical for resistance to infection, exerting its effects through broad transcriptional programs involving ~2000 genes, many of which remain uncharacterized (1, 2). Prominent within this transcriptional signature are several families of guanosine triphosphatases (GTPases). These include the 47-kD immunity-related GTPases (p47 IRGs), 65- to 73-kD guanylate-binding proteins (p65 Gbps), and 285-kD very large inducible GTPases (Vligs/Gvins) (2).

Recent mapping efforts have uncovered 43 members of this IFN-γ–inducible GTPase superfamily within mouse and human genomes (37) (fig. S1, A and B). p47 IRGs represent the largest subgroup (~18 to 21 genes) important for host defense (812). These GTPases bind phosphoinositides, cardiolipin, soluble NSF attachment protein receptor adaptor proteins, and other p47 IRGs to direct their membrane regulatory activities against compartmentalized bacteria and protozoa (1012). In contrast, little is known about the p65 Gbp and Gvin subfamilies, despite accounting for ~20% of the relative abundance of all proteins induced by IFN-γ (2). Weak antiviral or antibacterial properties have been ascribed individually to Gbp1, Gbp2, and Gbp5 (1315); however, integrated family contributions remain untested (2, 5).

We thus conducted loss-of-function screens across the complete 11-member mouse Gbp family in macrophages where it is strongly induced by IFN-γ compared with other IFNs (IFN-αβ, IFN-λ, and IFN-Ζ) and Toll-like receptor (TLR) ligands (fig. S1, C and D) (6). These immune cells were infected with two intracellular bacteria particularly sensitive to IFN-γ–mediated killing: Listeria monocytogenes (Lm), a gram-positive bacterium responsible for food-borne infection in humans, or Mycobacterium bovis BCG (Mb BCG), which causes lethal mycobacteriosis in IFNγR-deficient patients (1). Using short 21-bp (base pair) Gbp small interfering RNA (siRNA) duplexes that gave robust gene-specific silencing (fig. S2, A and B), we found that Gbp1, Gbp6, Gbp7, and Gbp10 were critical for control of virulent Lm (EGD strain) or Mb BCG (Phipps strain). In resting RAW264.7 macrophages, Lm multiplied by a factor of ~85 (log101.9 growth) over 6 hours after uptake. IFN-γ activation, however, curtailed replication (to a factor of ~18, or log101.2 growth), a restriction that was reversed with siRNAs for Gbp1, Gbp6, Gbp7, Gbp10, and to a lesser extent Gbp5 (to a factor of 39 to 58 or ~log101.5–1.75 growth; P < 0.0062) (Fig. 1A and fig. S3A). Protective Gbps functioned cooperatively, with siRNA combinations exacerbating the loss of IFN-γ–induced killing (Fig. 1A and fig. S3A). siRNA phenotypes were not attributable to single- and double-stranded RNA sensing. Primary bone-marrow–derived macrophages (BMMs) and RAW264.7 cells defective in either TLR (MyD88/Trif/) or Rig-1/Mda5/Ips-1–dependent RNA recognition (expressing the viral Rig-1/Mda5/Ips-1 inhibitor, NS34A) yielded comparable results (Fig. 1A and fig. S3A).

Fig. 1

IFN-γ–inducible Gbps protect against bacterial infection. (A) Gbp siRNA screen for loss of Lm or Mb BCG killing in IFN-γ–activated macrophages. Mean (triplicate wells) for each siRNA shown (SD removed for clarity; see fig. S3). Gbp10 siRNAs also silence Gbp6 due to 99.1% nucleotide identity. Dual silencing: Gbp1+Gbp7 (red); Gbp7+Gbp10 (blue); Gbp10+Gbp1 (green). Control, scRNA (2x). *, P < 0.0112, analysis of variance (ANOVA). 1 of 14 similar experiments. (B) Antibacterial activity in RAW264.7 cell lines expressing GbpDN mutants or BMMs lacking Gbp1 (fig. S6). Mean ± SD, some values falling within symbols. *, P < 0.046, ANOVA. N = 3 experiments each. (C) Lm burdens in orally infected (109 CFU) Gbp1+/+ and Gbp1–/– mice. Day 1 colonization included local mesenteric lymph nodes (MLNs). N = 8 to 12 mice per group per time point. Error bars, mean ± SD. *, P < 0.040, ANOVA. N = 3 experiments. (D) Listeriosis-induced weight loss (% of starting weight) over 8 days (8 to 12 per group). Error bars, mean ± SD. *, P < 0.039, ANOVA. One of three experiments shown. (E) Diseased lungs of Gbp1+/+ and Gbp1–/– mice infected with Mb BCG Phipps (105 CFU intravenously) at day 70 p.i. Scale bar, 1 cm. (F) Organ mycobacterial burdens at 1, 35, and 70 days p.i. N = 6 mice per group. Error bars, mean ± SD. *, P < 0.025, ANOVA. One of two similar experiments shown.

Mb BCG challenge showed similar Gbp-dependent resistance. In short 48-hour killing assays that were necessitated by waning siRNA effectiveness at 96 hours, IFN-γ reduced Mb BCG by ~log100.8 [72 to 77% reduction in colony-forming units (CFU)] in untreated and scrambled RNA (scRNA)–treated macrophages. Inhibition was partly reversed by Gbp1, Gbp5, Gbp7, or Gbp6/Gbp10 siRNAs (56 to 64% CFU reduction) but not siRNAs for the remaining Gbp genes (77 to 82%; P < 0.0112) (Fig. 1A and fig. S3B).

To amplify the smaller phenotypes for slow-growing Mb BCG, we devised a system of long-term Gbp inhibition. Dominant-negative (DN) mutants were identified and stably expressed under tetracycline-repressible control [tTA-TRE2-Gbp (DN)] to avoid polyketide antibiotics during infection. Two conserved P-loop residues (GxxH/RxKS) required for nucleotide-dependent self-assembly or a C-terminal CaaX box (CVIL) used for C20 isoprenyl membrane tethering were mutated (16, 17). All Gbp1 (Gbp1H48P, Gbp1S52N, and Gbp1SVIL) and Gbp7 (Gbp7R48P and Gbp7S52N) mutants failed to target vesicle membranes; however, those of Gbp10 (Gbp10R46P and Gbp10S50N) showed a less robust phenotype and were not pursued further as potential DN candidates (fig. S4A). Stable expression of Gbp1 and Gbp7 P-loop (S52N) mutants disrupted endogenous Gbp vesicle localization in IFN-γ–activated macrophages (fig. S4B), underscoring their DN action by binding wild-type partners for incorporation into “dead-end” multimeric complexes (fig. S5, A to C). This resembles dynamin-1 P-loop (S52N) mutants that inhibit self-assembly and endocytic trafficking (18). Cell-free assays corroborated these results: α-32P-GTP hydrolyzing activity and nucleotide-dependent homo-tetramerization (16) were abolished in recombinant Gbp1S52N and Gbp7S52N proteins, like DynS45N (18) (fig. S5, D and E).

Subsequent infection of four tTA-TRE2-Gbp1S52N and tTA-TRE2-Gbp7S52N macrophage cell lines showed impaired mycobactericidal activity in each case. Mb BCG was reduced 43 to 57% versus 83 to 88% for parental tTA cells over 7 days of IFN-γ treatment (P < 0.0013); the DN phenotype exceeded that of 48-hour siRNA assays (Fig. 1B and fig. S6) and was independent of other antimycobacterial pathways [inducible nitric oxide synthase (Nos2)] (8) or loss of host cell viability. Defective listerial killing was also seen in all mutant cell lines (57 to 78% versus 83 to 91% reduction in Lm CFU of tTA controls; P < 0.0024) (Fig. 1B and fig. S6).

Next, we validated bactericidal deficiencies in vivo. We generated Gbp1/ mice (fig. S7) that exhibited normal T, B, and phagocytic (CD11b+) cell profiles in target lymphoid organs like the spleen and LyG6+ granulocytes in peripheral blood (fig. S8). Macrophages from these mice, however, had impaired killing activities despite intact responses to IFN-γ as shown by NO release (8) (Fig. 1B and fig. S9A). Such killing defects also manifested in vivo. Gbp1/ mice allowed Lm to replicate by a factor of 100 to 1000 in livers and spleens after natural orogastric challenge, whereas Gbp1+/+ mice curtailed replication despite initial colonization of these organs being similar for both groups (0.8- to 9-fold growth during days 1 to 3) (Fig. 1C). Genotypic differences persisted through day 8 postinfection (p.i.), by which stage Gbp1/ mice exhibited discernible weight loss and sporadic diarrhea (Fig. 1, C and D, and fig. S9B).

Mb BCG infection yielded similar results. Gbp1/ mice had increased mycobacterial burdens as early as 5 weeks p.i. before becoming moribund at ~10 to 14 weeks, with up to 350 times more Mb BCG in their lungs than Gbp1+/+ controls (Fig. 1, E and F, and fig. S9, C to E). Susceptibility to either Lm or Mb BCG was not attributable to impaired granuloma formation (fig. S9, D and E) or to global IFN-γ defects, because production of this cytokine (fig. S9F) and responses to it (macrophage NO secretion) (fig. S9A) were intact in Gbp1/ mice. Thus, our three independent loss-of-function approaches—siRNA silencing, DN inhibition, and chromosomal deletion—implicate the 65-kD Gbps as a class of bacterial host defense proteins that operate in vitro and in vivo.

To understand how the protective Gbps conferred host resistance, we focused on trafficking to the bacterial compartment because pathogen-susceptible S52N mutations interfered with membrane targeting (figs. S4, A and B, and S10A). All the protective Gbps translocated to mycobacteria-containing vacuoles (MCVs) and Lm-containing vacuoles (LCVs) within 0.5 to 2 hours of uptake (Fig. 2, A and B), which correlates with the known ability of IFN-γ–activated macrophages to sequester Lm or Mb BCG inside phagosomes for lysosomal delivery and killing (810, 19). Live imaging showed that Gbps arrived on ~50- to 100-nm vesicles over a 20- to 30-min period before fusing with MCVs (Fig. 2C, fig. S10B, and movie S1). This parallels Toxoplasma gondii infection, where different Gbps target the parasitophorous vacuole (6), potentially helping deliver antimicrobial cargo to the pathogen compartment.

Fig. 2

Protective Gbps target bacteria-containing vacuoles. (A) Vesicular Gbp1, Gbp7, and Gbp10, but not Gbp3, recruited to 30-min LCVs (antibody to Lm, Alexa 594) or (B) 2 hours MCVs (GFPS65T-Mb BCG) in IFN-γ–activated RAW264.7 macrophages. Scale bar, 5 μm. Gbp1 detected with antibody to M18 (Alexa 488); remaining Gbps as (A) enhanced yellow fluorescent protein (eYFP)– or (B) monomeric red fluorescent protein (mRFP)–tagged constructs. Colocalization (mean ± SD) from 120 to 270 confocal images. N = 8 experiments. (C) Live imaging (frames 7.41 to 38.21 min) of IFN-γ–activated macrophages stably expressing eGFP-Gbp7 given Cy5-labeled Mb BCG. Scale bar, 5 μm. Three-dimensional surface intensity plots (boxed area) shown below. One of four experiments shown.

To identify the type of cargo Gbps transport, we combined three unbiased partner interactive screens with microscopic analysis of Gbp-containing vesicles. Macrophages stably expressing doxycycline-repressible Gbp1, Gbp7, or Gbp10 with different N-terminal tags for single-step or two-step capture via tandem affinity purification (TAP) were generated. In addition, glutathione S-transferase (GST)–Gbps were immobilized to ensure sufficient bait for isolating partners from stringent detergent-liberated fractions enriched for membrane complexes (10). Each method identified interacting proteins involved in the antimicrobial effects of the Gbps.

Most abundant among these partners was the membrane-bound gp91phox-p22phox (cytochrome b558) component of NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase that generates superoxide (O2) for killing listeria and mycobacteria (19, 20) (Fig. 3A and table S6). Gbp7 coprecipitated endogenous p22phox and both shared a subpopulation of vesicles that targeted MCVs (Fig. 3B and fig. S11A). Gbp7 also bound other NADPH oxidase subunits—principally p67phox, which is recruited as a heterodimer with p40phox to the cytochrome b558 complex for assembling the holoenzyme (Fig. 3B). An N-terminal G domain of Gbp7 bound p67phox, whereas its C-terminal helix engaged gp91phox-p22phox (Fig. 3A and fig. S11B). This suggested that Gbp7 could act as a bridging protein to help deliver cytosolic p67phox-p40phox to gp91phox-p22phox for NADPH oxidase assembly on phagosomal membranes.

Fig. 3

Gbp7-dependent oxidative host defense. (A) Silver-stained gel of macrophage membrane-enriched fractions incubated with the GST-Gbp7 C-terminal helical domain (amino acids 311 to 636). Controls, GST-Gbp1 (amino acids 311 to 589) and GST. TEV protease-released Gbps, p22phox, and gp91phox plus liquid chromatography–tandem mass spectrometry peptides (red font, overlapping peptides in green) are shown. (B) Coprecipitation of endogenous p22phox and p67phox by stable eGFP-Gbp7 in IFN-γ–activated macrophages. One of three experiments shown. (C) (Left) Gbp7 siRNAs disrupted p67phox-HA interaction with endogenous gp91phox (asterisk) but not p40phox; (Right) NADPH oxidase subunit and Gbp1 expression. One of two experiments shown. (D) Impaired p40phox-HA targeting to MCVs (blue arrows) with Gbp7 siRNAs or Gbp7S52N DN mutant at peak 2 hours recruitment time using GFPS65T-Mb BCG. 280 to 430 MCVs (mean ± SD) per condition. P < 0.012, ANOVA. siRNA treatment: scRNA (black bars); Gbp7 siRNAs (white bars). DN expression: tTA control (black bars); Gbp7S52N DN (gray bars). (E) Diminished O2 production in Gbp7 siRNA-treated or DN RAW264.7 cells given IFN-γ (100 U/ml, 16 hours) ± Lm (multiplicity of infection, 1:1; 30 min) or Mb BCG (MOI, 2:1; 3 hours). ACP, adherent cell protein. P < 0.017, ANOVA. N = 3 experiments.

We tested this proposal in IFN-γ–treated macrophages stably expressing hemagglutinin (HA)–tagged p67phox that allows robust coprecipitation of endogenous gp91phox; Gbp7 siRNAs largely abolished this interaction, without altering p67phox-HA binding to cytosolic p40phox or the expression of NADPH oxidase subunits (Fig. 3C). Gbp7 siRNAs and Gbp7252N DN mutants both diminished p67phox-p40phox targeting to MCVs as well as IFN-γ–induced O2 production (Fig. 3, D and E). Such defects were complemented by provision of an O2 donor that helped restore Lm killing in Gbp7-deficient cells (fig. S11C). Thus, Gbp7 exerts its action in part by downstream O2 production. Enforced Gbp7 expression also increased O2 release by a factor of ~10, like p67phox-HA but not Gbp7S52N, which served as positive and negative controls, respectively (fig. S11D). p67phox-HA–mediated O2 secretion was inhibited by Gbp7 but not Gbp1 or Gbp10 siRNAs, further underlining the specific role of Gbp7 in p67phox translocation. Thus, Gbp7 is important for IFN-γ–induced oxidant protection against intracellular bacteria.

Our Gbp-interaction screens also identified a second antimicrobial pathway that generates bacteriolytic peptides (2123). Here, Gbp1 bound p62/Sqstm1 and Gbp7 captured Atg4b, respectively (Fig. 4, A to C). p62 delivers ubiquitinated cargo to autolysosomes, generating ubiquitin-derived peptides that kill mycobacteria once MCVs fuse with this compartment (21, 22). Gbp1 vesicles harbored p62 and monoubiquitinated proteins (fig. S12) that were delivered to larger LC3b+ vacuoles (Fig. 4D) for liberating mycobactericidal peptides (detected in Gbp1 TAP screens) (fig. S13A and table S6). Lysosomal delivery was blocked in Gbp1-deficient cells or with siRNAs against Tsg101 (also required for p62 targeting and mycobacterial killing) (23) because large amounts of p62 went undigested (Fig. 4E). Gbp1 deficiency did not affect proteasomal degradation of inhibitor of nuclear factor κB (IκB) for p62-stimulated nuclear factor κB (NF-κB) signaling (fig. S13B), nor was Gbp1 itself ubiquinated (fig. S13C), degraded (fig. S13D), or targeted to lysosomes (fig. S12). Thus, Gbp1 probably recycles off this compartment before enclosure, consistent with a lack of LC3b coprecipitation along with p62, the latter of which Gbp1 bound outside its ubiquitin-associated (UBA) domain (Fig. 4A). Such binding could promote p62 oligomerization through Gbp1 self-assembly to capture monoubiquitinated proteins (24).

Fig. 4

Gbp1 and Gbp7 traffic antimicrobial peptides. (A) eYFP-Gbp1 coprecipitated full-length or UBA-deleted DsRed-p62, and (B) eYFP-Gbp7 captured Myc-Atg4b or Myc-Atg4BS74A in mammalian cell (HeLa) extracts. One of four experiments shown. (C) His6-Gbp7 retrieved endogenous Atg4b from RAW264.7 lysates. One of three pulldown (Pd) experiments shown. (D) Targeting endogenous Gbp1 (top) or eGFP-Gbp7 vesicles (bottom) to LC3b autophagic membranes in IFN-γ–activated macrophages. 245 to 260 images counted. Scale bar, 10μm. (E) Impaired p62 degradation in 1% triton-X 100 extracted fractions of Gbp-deficient macrophages activated with IFN-γ (100 U/m; 48 hours) ± Mb BCG (MOI, 2:1, 3 hours) at peak p62 turnover (fig. S13D). One of three experiments shown. (F) Impaired MCV targeting by Atg4b (K25 antibody, Alexa 594) at peak 4 hours recruitment time. Targeted MCVs (white arrows); non-targeted MCVs (blue arrows). Scale bar, 15μm. 220 to 240 MCVs (mean ± SD) for each condition. P < 0.009, ANOVA. One of two experiments shown.

Gbp7 appeared to regulate the next step—cargo engulfment by autophagic membranes—because it bound native Atg4b (Fig. 4C), localized to sites of LC3b membrane elongation (Fig. 4D), and promoted p62 degradation (Fig. 4E). Gbp7 siRNAs disrupted membrane closure; here, incomplete ubiquinated ring structures were evident (fig. S14A) along with Atg5-Atg12, which normally disassembles after membrane closure, like Atg4bC74A mutants that prevent LC3b processing (25) (fig. S14, B to D). Gbp7, but not Gbp7S52N, rescued these defects and accelerated LC3b lipidation plus p62 turnover, probably by recruiting native Atg4b to unoccupied sites of LC3b attachment (fig. S14, D to G). Indeed, Atg4b recruitment to early MCVs was also blocked by Gbp7 siRNAs, reminiscent of p67phox-p40phox translocation (Fig. 4F). Hence, we favor Gbp7 operating as a membrane trafficking protein rather than protease cofactor (it bound both active and inactive Atg4b equally well) (Fig. 4B) that transports different substrates, like NADPH oxidase subunits or Atg4b, to the site of infection.

In sum, the IFN-γ–inducible Gbps promote oxidative killing and deliver antimicrobial peptides to autophagolysosomes (fig. S15). Such cooperative effects could provide broad host protection against different pathogen classes (2, 5).

Supporting Online Material

Materials and Methods

Figs. S1 to S15

Tables S1 to S6

Movie S1


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We thank Y. Matsuura (NS34A DN cells); S. Ivanov (Ub-K63-GFP cells); R. Medzhitov (MyD88–/– Trif–/– mice); M. Pypaert (electron microscopy); and P. Cresswell, W. Mothes, C. Roy, G. Superti-Furga, T. Yoshimori, and H. Zhu (plasmids and antibodies). Supported by NIH National Institute of Allergy and Infectious Diseases (R01 AI068041-01A1), Burroughs Wellcome Fund Investigator in Pathogenesis of Infectious Disease Award, Searle Scholars Program, Cancer Research Institute Investigator Award, W.W. Winchester Foundation (to J.D.M.) and Browne-Cox Fellowship (to A.R.S.).
View Abstract

Stay Connected to Science

Navigate This Article