GBP5 Promotes NLRP3 Inflammasome Assembly and Immunity in Mammals

See allHide authors and affiliations

Science  27 Apr 2012:
Vol. 336, Issue 6080, pp. 481-485
DOI: 10.1126/science.1217141


Inflammasomes are sensory complexes that alert the immune system to the presence of infection or tissue damage. These complexes assemble NLR (nucleotide binding and oligomerization, leucine-rich repeat) or ALR (absent in melanoma 2–like receptor) proteins to activate caspase-1 cleavage and interleukin (IL)–1β/IL-18 secretion. Here, we identified a non-NLR/ALR human protein that stimulates inflammasome assembly: guanylate binding protein 5 (GBP5). GBP5 promoted selective NLRP3 inflammasome responses to pathogenic bacteria and soluble but not crystalline inflammasome priming agents. Generation of Gbp5–/– mice revealed pronounced caspase-1 and IL-1β/IL-18 cleavage defects in vitro and impaired host defense and Nlrp3-dependent inflammatory responses in vivo. Thus, GBP5 serves as a unique rheostat for NLRP3 inflammasome activation and extends our understanding of the inflammasome complex beyond its core machinery.

Inflammasomes integrate microbial- or danger-associated signals for mobilizing immunity to infection, as well as in sterile settings such as those present in gout and diabetes (1, 2). Canonical inflammasome activation involves assembly and autoproteolysis of the zymogen, procaspase-1, within multiprotein platforms containing different members of the nucleotide binding domain, leucine-rich repeat (LRR) protein family (NLRs) or absent in melanoma 2 (AIM2)–like receptors (ALRs) (1, 2). Once activated, caspase-1 cleaves its cytokine substrates, pro-interleukin (IL)–1β and pro–IL-18, to elicit host effector functions, prime adaptive immunity, and stimulate vicinal inflammation (1, 2).

To generate an inflammasome complex, NLRs or ALRs rely on a set of common structural motifs. These include a ligand-sensing LRR or HIN-200 region; a nucleotide binding domain (NBD) for adenosine triphosphatase–driven oligomerization; and either a caspase activation and recruitment domain (CARD) or a pyrin domain (PYD) for engaging procaspase-1, the latter via the adapter protein ASC (apoptosis-associated speck-like protein containing a CARD). This modular design delimits the structural boundaries of inflammasome-related core proteins (1, 2).

We discovered that some NLR-like CARDs, however, are fused to guanosine 5′-triphosphatase (GTPase) domains (GDs) rather than NBD-like domains in lower organisms using genome-wide in silico screens across 91 taxa (see the supplementary text) (Fig. 1A). These GDs belonged to ancestral 65- to 73-kD guanylate binding proteins (GBPs), recently shown to be critical for mammalian host defense (3) (table S1 and fig. S1, A and B). This raised the possibility of functional interactions between the GBPs and NLRs/ALRs in higher species to regulate the inflammasome.

Fig. 1

Human GBP5 and its murine ortholog promote selective NLRP3 inflammasome activation in macrophages. (A) Phylogenetic tree of GBPs harboring GTPase (G) and helical domains (H) fused to either CARDs (C) or death effector domains (DEDs) (D). GenInfo identifier accession numbers are indicated. Other related subfamilies shown: GVIN, GTPase, very large interferon inducible; IRG; and RHD, root hair defective. Scale bar, 0.5 substitutions per site. (B) IL-1β enzyme-linked immunosorbent assay (ELISA) (supernatants) and GBP5 immunoblot (cell lysates) of differentiated THP-1 cells untreated (UT) or treated with LPS alone or LPS/IFN-γ as indicated. Mean ± SD, one of three similar experiments. (C and D) IL-1β ELISA from IFN-γ/LPS–primed THP-1 (C) or J774A.1 cells (D) given siRNAs or YVAD. Murine cells pulsed with ATP (5 mM) in (D). siRNAs to mouse Gbp10 also silence Gbp6 due to 99.1% nucleotide identity. Human immmunity-related GTPase, M (IRGM) siRNAs served as an IRG control. Mean ± SEM, two to three experiments. *, P < 0.05, Student’s t test. (E and F) IL-1β and active caspase-1 (p20) from supernatants of IFN-γ–primed THP-1 cells stably expressing control or GBP5 shRNA given inflammasome priming agents. GBP5 and β-actin expression shown below. Mean ± SEM, two to three experiments.

We tested this idea in mammalian macrophages exposed to microbial or danger signals for each well-characterized inflammasome pathway: NLRP1, NLRP3, NLRC4, and AIM2 (1, 2). The expression of NLRP3, AIM2, and ASC is readily induced by bacterial lipopolysaccharide (LPS) or type I interferon (IFN-α/β) (47). Likewise, GBP expression is induced by LPS or IFN-α/β; however, the highest levels are achievable by IFN-γ (3, 8), similar to the 47-kD immunity-related GTPases (IRGs) (9). Thus, we examined inflammasome activation in differentiated human THP-1 macrophages or C57BL/6 mouse bone marrow–derived macrophages (BMMs) primed with IFN-γ plus bacterial products to ensure that the GBPs were present at sufficient levels to assess their physiological relevance. In the case of NLRP3 inflammasomes, such priming acts together with a second signal, such as adenosine triphosphate (ATP)–induced potassium ion flux or crystal-induced phagosomal damage, to activate caspase-1. This regimen elicited dose-dependent increases in GBP expression and, as recently reported for human macrophages and keratinocytes (10, 11), heightened IL-1β secretion and caspase-1 cleavage (Fig. 1B and figs. S2 and S3). Addition of IFN-γ also more closely mimics the in vivo setting, where it is released from natural killer (NK) and T cells at sites of inflammation or infection (3, 8, 10, 11).

Small interfering RNAs (siRNAs) against the complete human and mouse GBP families were initially screened in IFN-γ–treated macrophages given LPS/ATP, a potent combination that stimulates the NLRP3 inflammasome complex (13) (fig. S4). IL-1β secretion was inhibited with siRNAs against human GBP5 and mouse Gbp5; this cross-species inhibition resembled siRNAs against human ASC and mouse Asc or the caspase-1 inhibitor Ac-YVAD-cmk (YVAD) (~50 to 80% reduction) (Fig. 1, C and D). Importantly, IL-1β defects were not due to altered caspase-1 substrate (pro–IL-1β) levels, IFN-γ/LPS unresponsiveness (shown by nitric oxide release), or reduced macrophage viability (fig. S5, A to C). Thus, of 17 family members tested, human GBP5 and its mouse ortholog were required for LPS/ATP-dependent IL-1β secretion.

Next, we employed additional priming agents besides LPS to delineate the inflammasome specificity of GBP5. Isolated human THP-1 and mouse J774A.1 macrophage cell lines stably expressing small hairpin RNAs (shRNAs) were used to ensure better GBP5/Gbp5 silencing; the latter was confirmed with GBP5/Gbp5-specific antibodies (figs. S4B and S5D). Notably, caspase-1/IL-1β activation was defective for all bacterial NLRP3 priming agents tested: LPS, muramyl dipeptide (MDP), isoglutamate diaminopimelic acid (iE-DAP), and Salmonella typhimurium [grown under low Salmonella Pathogenicity Island (SPI)-1–inducing conditions to engage NLRP3] (fig. S2D) (12). In contrast, inflammasomes containing AIM2 [directly stimulated with deoxyA;deoxyT (dA:dT)] or NLRC4 (activated with Salmonella flagellin in mouse macrophages) were unaffected, as were inflammasomes triggered by lysosomal disruption using the crystalline NLRP3 activator alum (13) (Fig. 1, E and F, and fig. S5, D and E). Thus, GBP5 appeared to promote NLRP3 inflammasome responses specifically to live bacteria and soluble but not crystalline priming agents in IFN-γ–treated human and mouse macrophages.

To obtain unequivocal genetic evidence for this Nlrp3 selectivity and assess the contribution of Gbp5 in vivo, we generated Gbp5-deficient (Gbp5–/–) mice (fig. S6). Pronounced loss of IL-1β and IL-18 release plus caspase-1 cleavage was evident in primary Gbp5–/– BMMs given LPS, MDP, or iE-DAP followed by ATP (Fig. 2, A and B, and fig. S7A). Similar outcomes were observed when ion flux was triggered with nigericin, ruling out ATP-P2X7R defects; P2X7R channel activity itself was comparable between genotypes (Fig. 2, A and B, and fig. S8A) (1, 2). Examination of NF-κB (nuclear factor κB) signaling, TNF-α (tumor necrosis factor–α), O2, or NO2 synthesis along with Nlrp3, Asc, procaspase-1, and pro–IL-1β expression revealed that upstream IFN-γ/LPS signaling and induction of the core inflammasome components were still intact in Gbp5–/– BMMs (fig. S7, B to F). Responses to alum and monosodium urate (MSU), another crystalline Nlrp3 priming agent, were likewise unchanged, as was Aim2 activation by dA:dT (Fig. 2C and fig. S7G). These experiments verified earlier siRNA/shRNA results and provided definitive genetic evidence for Gbp5 involvement in specific Nlrp3 inflammasome responses.

Fig. 2

Defective Nlrp3-Asc inflammasome activation and assembly in Gbp5–/– BMMs. (A to D) Cytokines [(A), (C), and (D)] or caspase-1 (B) in culture supernatants of IFN-γ/LPS–primed BMMs activated with the indicated stimuli [(A) to (D)] or infected with Listeria or Salmonella (D). Mean ± SD; one of three to five similar experiments. (E) Percentage of BMMs with active caspase-1 (Biotin-YVAD+) (see fig. S8A). IFN-γ–primed BMMs were exposed to LPS, MDP, or iE-DAP followed by nigericin or LPS/MSU, LPS/dA:dT, or Lm infection. *, P < 0.0001, Student’s t test. (F) Immunostaining of endogenous Asc foci (arrows) and caspase-1 in BMMs as in (E). Scale bar, 10 μm. *, P < 0.0001, Student’s t test for >1000 cells counted. Mean ± SEM from two to four experiments. (G and H) Asc multimerization (G) and caspase-1 activity (H) in inflammasomes isolated from IFN-γ/LPS–treated BMMs activated with nigericin (L/Nig) or left untreated (UT). BS3-crosslinked or native Asc (-BS3) detected by immunoblot (IB). *, Nonspecific band in (G). Mean ± SEM, two to three similar experiments. **, P < 0.01; ***, P < 0.001, analysis of variance (ANOVA).

Natural infection by bacteria further underscored this specificity. IL-1β secretion was impaired in Gbp5–/– BMMs exposed to Salmonella or a second bacterium, Listeria monocytogenes (Lm), that can also engage Nlrp3 (Fig. 2D) (1416). In contrast, transfection of Salmonella flagellin itself revealed that Nlrc4 inflammasome responses were normal (17, 18) (Fig. 2D and fig. S7G). Notably, loss of inflammasome activity in Gbp5–/– BMMs was most marked for IFN-γ plus LPS/ATP stimulation versus LPS/ATP alone, the latter of which elicited low levels of autocrine Gbp5 expression in wild-type cells (fig. S7, C and G). Thus, chromosomal deletion of Gbp5 verified its importance for selective Nlrp3 inflammasome responses under native IFN-γ– and LPS-induced conditions.

Next, we determined where Gbp5 operated in this process. Staining for active caspase-1 in situ found that this signal was essentially lost in Gbp5–/– BMMs after bacterial infection or priming with soluble bacterial products (LPS, MDP, or iE-DAP) but not crystalline agents (MSU) or Aim2 (dA:dT) ligands (fig. S8B and Fig. 2E). Hence, caspase-1 activation itself, rather than its export, was defective. Notably, cathepsin B activation via MSU or alum was intact in Gbp5–/– BMMs, again showing that the nonlysosomal Nlrp3 pathway was compromised (fig. S8C). Both mitochondrial reactive oxygen species and membrane potential have been proposed to contribute to this nonlysosomal pathway (7, 19, 20). No differences, however, were observed for either parameter between Gbp5+/+ and Gbp5–/– BMMs (fig. S8D). Thus, Gbp5 likely operates more proximal to the caspase-1 complex itself.

We assayed inflammasome complex formation by tracking the Nlrp3 adaptor, Asc, which oligomerizes as part of this complex to form a single large (1 to 2 μm) perinuclear focus per cell (12, 21). Such foci assemble in a stimulus- and NLR-dependent but caspase-1–independent manner (12, 21). IFN-γ–primed Gbp5–/– BMMs revealed almost complete loss (~85%) of Asc foci after LPS/nigericin treatment for engaging the Nlrp3 inflammasome, whereas Aim2 inflammasome activation with dA:dT was unaffected (Fig. 2F). Asc oligomerization and caspase-1 activity were essentially absent within enriched inflammasome fractions isolated from Gbp5–/– macrophages given LPS/nigericin (Fig. 2, G and H). Thus, Gbp5 promotes Nlrp3-Asc inflammasome assembly to activate caspase-1 under native conditions.

This assembly may require interactions of Gbp5 with Nlrp3. Mouse Nlrp3 captured Gbp5 in a PYD-dependent manner (other PYD/CARD-containing inflammasome proteins Nlrp1, Nlrc4, Aim2, Asc, or procaspase-1 failed to do so) and human GBP5 retrieved endogenous NLRP3 from IFN-γ–treated THP-1 cells (Fig. 3, A and B, and fig. S9, A to C). Such interactions were specific; a recombinant glutathione S-transferase (GST)–fused PYD of NLRP3 (NLRP3PYD) bound GBP5 via its GTPase domain (GBP51-306) but did not bind its closest paralog, GBP1 (Fig. 3C). NLRP3PYD also retrieved endogenous human GBP5 and ASC via distinct subdomains, as shown using an L22P, K23A mutant (GST-NLRP3PYDLK→PA) that captured native GBP5 but not ASC (Fig. 3D). Interaction of GBP5 with NLRP3PYD led to assembly of scaffolds around an inner Asc core as seen by fluorescent microscopy (fig. S9, D to F).

Fig. 3

Tetrameric GBP5 binds Nlrp3 to promote Asc assembly. (A) Coimmunoprecipitation and immunoblot of Myc-GBP5 and Flag-tagged Nlrp3 [intact or pyrin-deleted (Δpyrin)] from HEK293 cells. Yellow fluorescent protein (YFP) served as a nonspecific control. One of three similar experiments. (B) Immunoblot of endogenous NLRP3 immunoprecipitated by hemagglutinin-Flag-GBP5 stably expressed in THP-1 cells. One of two similar experiments. (C and D) GST pulldown of recombinant GBP5 and GBP1 (C) or native GBP5 and ASC (D) from IFN-γ–treated THP-1 cells. GST, GST-PYD (GST-NLRP31-114), or GST-PYDLK→PA used as bait. One of three experiments shown. (E) GBP5 homology model depicting N-terminal GTPase (nucleotide-binding P-loop) and C-terminal α-helical domains based on the GBP1 crystal structure (Protein DataBank code 1F5N). (F) Size-exclusion chromatographs of recombinant GBP5 proteins. Arrows, molecular weight (kD) calibration. (G) Asc oligomerization assay in HEK293 cells transfected with Myc-GBP5 variants plus Flag-tagged NLRP3, NLRP3PYD, or AIM2. Lysate IB of expressed proteins below. One of three similar experiments.

Once bound, how does GBP5 promote NLRP3 assembly? The forerunner of the GBP family, GBP1, displays GTPase-dependent tetramerization (22), a process that could help drive complex formation. We built an energetically favored GBP5 model using the GBP1 crystal structure to assign truncation/mutation sites for examining self-assembly (Fig. 3E). Recombinant GBP5 and a GTPase-deficient mutant (GBP5KS→AA) both formed tetramers (Fig. 3F and fig. S10). This GTPase-independent tetramerization required the last 62 amino acids, which by modeling fold back on its G domain (Fig. 3, E and F). Removal of this region (GBP51-528) yielded GBP5 dimers, while a coiled-coil domain within it also dimerized (GBP5476-586) (Fig. 3, E and F). Hence a “dimer of dimers” model could account for GBP5 tetramer assembly to stimulate Nlrp3-Asc oligomerization seen in wild-type but not Gbp5–/– cells (Fig. 2, G and H).

We tested this idea by reconstituting Nlrp3-Asc complexes in human embryonic kidney (HEK) 293 cells lacking each of these components. Asc multimerization was modest in the presence of either NLRP3 or NLRP3PYD but greatly enhanced if tetrameric GBP5 or GBP5KS→AA was added (Fig. 3G). Asc assembly was abolished when the tetramerization mutants, GBP5307-586 and GBP5476-586, were used. Replacement of NLRP3 by the Asc-binding ALR, AIM2, also made this process GBP5-independent (Fig. 3G). Thus, tetrameric GBP5 promotes Asc oligomerization specifically through its interaction with NLRP3.

Finally, we examined the impact of Gbp5 on Nlrp3-Asc assembly for caspase-1 activation and cytokine secretion in vivo. Three inflammatory and infection models were used. First, LPS-induced sepsis (23, 24) revealed that Gbp5–/– animals had 60 to 75% reduction in serum IL-1β and IL-18 but similar TNF-α levels to wild-type mice, confirming inflammasome specificity (Fig. 4A). This coincided with loss of active caspase-1–expressing (Biotin-YVAD+) Gbp5–/– CD11b+ macrophages in target organs such as the spleen (Fig. 4B and fig. S11, A to C). Second, in a caspase-1–dependent model of peritonitis (13, 25), Gbp5–/– mice recruited significantly fewer Biotin-YVAD+ neutrophils after injection with MDP but not crystalline MSU or alum (Fig. 4C). Hence the selective inflammasome defects seen for soluble Nlrp3 priming agents (LPS and MDP) in vitro also operated in vivo. Third, orogastric challenge with L. monocytogenes, a natural food-borne pathogen detected in part by Nlrp3-Asc inflammasomes and requiring caspase-1 for resistance (15, 16, 26), led to higher bacterial burdens, discernible weight loss, and 50 to 80% fewer Biotin-YVAD+ CD11b+ cells in mesenteric lymph nodes of Gbp5–/– mice, confirming defects in caspase-1 activation in vivo (15, 16, 26) (Fig. 4, D and E, and fig. S11D). Inhibiting this caspase-1 activation in wild-type mice with daily YVAD injections led to susceptibility resembling Gbp5–/– mice, whereas the same treatment failed to render the latter group more susceptible given that they are already compromised for this pathway (Fig. 4, D and E). Thus, Gbp5 promotes caspase-1–mediated protection against oral Listeria infection and Nlrp3-dependent inflammatory responses in vivo.

Fig. 4

Gbp5 promotes Nlrp3-dependent responses in vivo. (A) Serum cytokines 3 hours after LPS injection [30 mg per kg of weight intraperitoneally (ip); n = 6 to 7 mice per group]. Mean ± SEM from two to three experiments. P values, Student’s t test. (B) Flow cytometric staining of active caspase-1 [Biotin-YVAD-Streptavidin-Allophycocyanin (APC)] in CD11b+ splenocytes 3 hours after LPS or phosphate-buffered saline as control. Gbp5–/– cells were restimulated with LPS to ensure that they were not tolerized to LPS in vivo. (C) Caspase-1 activation in peritoneal exudate neutrophils (Ly6G+) of mice (n = 3 to 8 per group) injected intraperitoneally with different Nlrp3 agonists. Mean ± SEM, two to five experiments. P values, Student’s t test. (D and E) Mice orally infected with Listeria and in two groups treated with YVAD (1 mg/day ip) before assaying organ bacterial burdens [colony-forming units (CFU)] (D); and body weight (E) 3 days later (mean ± SEM; n = 4 to 9 per group). P values: two-tailed Mann-Whitney test in (D); ANOVA (P < 0.001) at days 2 and 3 in (E). Similar Listeria colonization (~1000 CFU in spleen and liver; day 1) for both groups in two experiments.

Collectively, our findings identify GBP5 as a unique activator of NLRP3-ASC assembly to live bacteria and their cell wall components but not crystalline agents or double-stranded DNA (fig. S12). How GBP5 elicits this selective response may involve spatial segregation of soluble versus crystalline cytosolic signals, events that could affect the recently discovered noncanonical inflammasome pathway as well (24). Moreover, results from Gbp5–/– mice suggest that a search for human GBP5 mutations—like NLRP3 variants in Crohn’s colitis or autoinflammatory syndromes (1, 2)—could have important implications for human health.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S12

Table S1

References (2738)

References and Notes

  1. Acknowledgments: We thank J. Tschopp, M. Karin, W. Mothes, and C. Roy for plasmids. All data are tabulated in the main paper and the supplementary materials. A.R.S. was supported by Yale Brown-Coxe and Anna Fuller postdoctoral fellowships, P.C. by Howard Hughes Medical Institute, and J.D.M. by NIH National Institute of Allergy and Infectious Diseases (R01 AI068041-06), Burroughs Wellcome Fund Investigator in Pathogenesis of Infectious Disease Award (1007845), Searle Foundation Scholars Program (05-F-114), Cancer Research Institute Investigator Award Program (CRI06-10), Chrohn’s and Colitis Foundation of America Senior Investigator Award (R09928), and W. W. Winchester Foundation.
View Abstract

Stay Connected to Science

Navigate This Article