Cutting Edge: Mutation of Francisella tularensis mviN Leads to Increased Macrophage Absent in Melanoma 2 Inflammasome Activation and a Loss of Virulence

The ability of Francisella tularensis to disrupt phagocyte function plays a key role in its virulence. Following phagocytosis, F. tularensis escapes the phagosome prior to phagosome/lysosome fusion and subsequently replicates within the cytosol (1, 2). Intracellular survival of F. tularensis is also associated with its ability to inhibit NADPH oxidase activity as well as subsequent NF-κB activation (3, 4). In activated macrophages (Mϕs), the escape of F. tularensis from the phagosome into the cytosol triggers activation of the absent in melanoma 2 (AIM2) inflammasome (5–7), a multiprotein complex containing AIM2 (an IFN-inducible HIN-200 family member), apoptosis-associated speck-like protein containing a caspase recruitment domain (ASC), and the cysteine protease caspase 1 (8–11). Activation of the AIM2 inflammasome results in autocatalytic cleavage of caspase 1, resulting in the processing and secretion of IL-1β and IL-18. Inflammasome activation plays a crucial role in innate immune responses against F. tularensis, as AIM2-, caspase 1-, and ASC-deficient mice have markedly increased bacterial burdens and mortality following infection with F. tularensis (5, 6).

In this study, we demonstrate that F. tularensis live vaccine strain (LVS) avoids efficient AIM2 inflammasome activation. We show that mviN is required to inhibit F. tularensis-induced caspase 1-dependent processing and secretion of IL-1β and IL-18 as well as to limit F. tularensis-mediated cytotoxicity. Mutation of mviN resulted in a marked change in the morphology of the bacterium and was required for full virulence of F. tularensis LVS in an in vivo model of infection.

F. tularensis ssp. holarctica LVS was obtained from American Type Culture Collection (ATCC 29684; Manassas, VA). The LVS strain containing a chromosomal mutation in mviN was transformed with the Tn5 delivery plasmid pBB109 as previously described (12). Complementation of the mviN::Tn5 strain with wild-type (WT) mviN containing the ribosomal binding site in the plasmid pBB103 was accomplished by cryotransformation as previously described (13). For in vitro studies, LVS strains were grown on Difco cysteine heart agar (Becton Dickinson, Sparks, MD) supplemented with 9% SRBC for 48 h at 37°C. A total of 25 μg/ml spectinomycin was added to plates for growth of the mviN::Tn5 + mviN strain. For in vivo studies, bacteria were grown overnight in modified Mueller-Hinton broth (BD Biosciences, San Jose, CA) supplemented with 1% (w/v) glucose, 0.025% ferric pyrophosphate, and 2% IsoVitaleX.

Bacterial strains were grown in modified Mueller-Hinton broth for 6 h to an OD₆₀₀ between 0.2 and 0.4. Samples were placed on silicon wafers, fixed in 2.5% glutaraldehyde, and dehydrated using a standard graded ethanol series, with a final clearance in hexa-methyl-disilazane. Samples were coated with gold/palladium and viewed on a Hitachi S-4800 scanning electron microscope (Hitachi, Tokyo, Japan).

The generation of ASC-, caspase 1-, NLRP3-, NLRC4-, and AIM2-deficient mice have been described previously (6, 14–16). The University of Iowa Institutional Animal Care and Use Committee (Iowa City, IA) approved all protocols used in this study. Mice (6–8 wk old) were injected i.p. with the indicated dose of F. tularensis LVS or the mviN mutant. Mice were monitored every 12 h for lethality; mice found to be in a moribund state for >4 h were considered terminal and euthanized.

Bone marrow-derived Mϕs were generated as previously described (14). Unless indicated, Mϕs were activated by stimulating with 50 ng/ml LPS from Escherichia coli serotype 0111:B4 (Invivogen, San Diego, CA) for 3–6 h prior to infection. Mϕs were infected with F. tularensis LVS or the mviN mutant at a multiplicity of infection (MOI) of 50:1 unless otherwise indicated. Nine hours later, or at the indicated time, supernatants were collected and assayed for IL-1β, IL-18, TNF-α, and IL-6 by ELISA (14). Mϕ cell death was determined by measuring lactate dehydrogenase (LDH) release using a cytotoxicity detection kit (Promega, Madison, WI). Western blotting was performed as previously described (14).

Two-tailed Mann-Whitney U tests were performed using Prism software (GraphPad, La Jolla, CA). The p values <0.05 were considered statistically significant. Unless stated otherwise, determinations are expressed as the mean ± SEM.

Peptidoglycan, a major component of bacterial cell walls, is composed of glycan chains that are cross-linked by peptide bridges. A lipid-linked peptidoglycan precursor (lipid II) is generated in the cytoplasm by a series of highly conserved enzymatic steps. MviN was recently identified as being responsible for flipping the lipid II across the cytoplasmic membrane (17, 18); once on the periplasmic side, penicillin-binding proteins use lipid II to generate mature peptidoglycan. F. tularensis LVS possesses an mviN gene (FTL_1305) that shares homology to that of E. coli (30% identity, 54% similarity) as well as Salmonella typhimurium (28% identity, 52% similarity) and Legionella pneumophila (31% identity, 54% similarity), two pathogens that, upon infecting Mϕs, can also activate caspase 1 (19–21). mviN is also highly conserved among other F. tularensis subspecies with homology to the mviN genes of F. tularensis ssp. novicida (98% identity, 99% similarity) and F. tularensis ssp. tularensis (98% identity, 99% similarity).

To determine if mviN is important for F. tularensis LVS virulence in vivo, we used an mviN mutant, mviN::Tn5, isolated from a F. tularensis LVS transposon library. WT mice were challenged i.p. with either LVS or the mviN mutant and survival monitored. Infection with 3 × 10⁴ CFU LVS resulted in 100% mortality by day 6 (Fig. 1A). In contrast, all mice infected with 3 × 10⁴ CFU mviN mutant survived until day 14 (Fig. 1A). Infection with 3 × 10⁵ or 3 × 10⁶ CFU mviN mutant resulted in a 90% and 70% survival rate, respectively (Fig. 1A). The increased survival of mice infected with the mviN mutant relative to WT LVS was also reflected in the 10–100,000-fold lower bacterial burdens in infected organs 3 d postinfection (Fig. 1B). Histology of liver 3 d postinfection revealed fewer hepatic lesions in mice infected with the mviN mutant compared with LVS (Fig. 1C, 1D). Although fewer in number, the necropurulent foci seen in mviN mutant-infected mice trended toward being larger than those of LVS-infected mice (Fig. 1D). Similarly, histology of spleen 3 d postinfection demonstrated occasional necropurulent foci in mviN mutant-infected mice, whereas LVS-infected mice had extensive coalescing necropurulent foci in the splenic red pulp that extended into the white pulp (Supplemental Fig. 1A, 1B). Collectively, these data suggest that the F. tularensis LVS mviN mutant was highly attenuated in vivo.

Mutant E. coli strains lacking functional mviN fail to grow and undergo lysis (17, 18); similarly, mviN was essential for the viability of Sinorhizobium meliloti and Burkholderia pseudomallei (22, 23). However, the F. tularensis LVS mviN mutant grew normally in broth (Supplemental Fig. 2A), consistent with the findings that mviN mutations in S. typhimurium and mutations of mviN homologs in Bacillus subtilis do not result in defective growth (24, 25). These results suggest that mviN is dispensable for the growth of F. tularensis LVS, and it is possible that LVS may possess redundant pathways that compensate for the lack of mviN.

Because Mϕs support Francisella replication in the host during natural infection, we examined the intracellular replication of the mviN mutant in vitro. Both LVS and the mviN mutant replicated with similar kinetics within unprimed Mϕs (Supplemental Fig. 2B), suggesting that the attenuation of infection by the mviN mutant in mice was not simply due to an inability of the intracellular organism to replicate in this cell type.

Although we did not detect a defect in bacterial replication of the mviN mutant in vitro, scanning electron microscopy revealed a striking change in cellular morphology of the mviN mutant; mviN mutants appeared round and lacked the pleomorphic structure typical of LVS (Supplemental Fig. 2C). These morphological results suggest that mviN serves an important role in the synthesis of Francisella cell wall structures, although additional studies are required to assess the full impact of mviN on F. tularensis peptidoglycan synthesis.

Because Francisella induces cell death in primed Mϕs in a caspase 1-dependent manner (5), we tested whether mviN was necessary for Francisella-mediated cytotoxicity. LPS-primed WT Mϕs infected with LVS underwent cell lysis with 25% cytotoxicity 12 h postinfection. However, infection of LPS-primed WT Mϕs infected with the mviN mutant resulted in more extensive cell death compared with LVS, with all cells lysed at 12 h (Fig. 2A). Complementation of the mviN mutant induced Mϕ death at levels similar to WT LVS (Fig. 2B), confirming the role of mviN in this process. The Mϕ death induced by the mviN mutant was dependent on caspase 1, as caspase 1-deficient Mϕs infected with the mviN mutant had a delayed cell death compared with WT Mϕs (Supplemental Fig. 3A). We also examined the intracellular replication of the mviN mutant within LPS-primed Mϕs and found decreased growth of the mviN mutant compared with LVS at 24 h (Supplemental Fig. 3B), consistent with the increased Mϕ death induced by the mviN mutant.

In addition to its role in mediating F. tularensis-induced Mϕ death, caspase 1 is required for the processing and secretion of IL-1β and IL-18 by infected cells (5). The mviN mutant induced greater secretion of IL-1β and IL-18 from LPS-primed Mϕs compared with LVS (Fig. 2C–E), at all MOIs tested (Supplemental Fig. 3C). Mutant bacteria complemented with mviN failed to induce increased IL-1β and IL-18 secretion (Fig. 2D, 2E). In contrast to the augmented IL-1β and IL-18 release, the secretion of IL-6 and TNF-α by unprimed Mϕs infected with F. tularensis LVS was unaffected by the absence of mviN (Fig. 2F, 2G).

Caspase 1 activation is a two-step process culminating in the autocatalytic processing of the 45-kDa procaspase 1 to generate two subunits, p20 and p10. In the first step, treatment with an agent, such as LPS, not only results in the generation of pro–IL-1β, but also primes the inflammasome for subsequent activation (26). In the case of AIM2 inflammasome activation, phagosomal escape of F. tularensis into the Mϕ cytoplasm, likely with concurrent release of bacterial DNA into the cytosol, serves as the second signal (6). The mviN mutant did not enhance IL-1β secretion by bypassing the need for a priming step, as neither WT LVS nor the mviN mutant was capable of inducing IL-1β secretion from unprimed Mϕs (Supplemental Fig. 3D). Whereas primed Mϕs infected with LVS activated caspase 1 by 9 h postinfection, as judged by Western blot detection of the p10 cleavage product (Fig. 2H), those infected with the mviN mutant induced caspase 1 activation more rapidly (by 6 h postinfection) and to a greater extent compared with WT LVS (Fig. 2H). Taken together, these data suggest that F. tularensis LVS expression of mviN was required to limit caspase 1 activation and subsequent cytotoxicity and secretion of IL-1β and IL-18; although the mviN mutant was able to replicate within unprimed Mϕs in vitro, attenuation of infection by the mviN mutant in vivo was most probably due to increased death of F. tularensis-infected Mϕs and increased proinflammatory cytokine production.

In addition to AIM2, the nucleotide-binding domain leucine-rich repeat containing (NLR) family members NLRP3 and NLRC4 can activate caspase 1 in response to bacterial pathogens (27). To determine if NLRP3 or NLRC4 was responsible for the increased IL-1β secretion induced by infection with the mviN mutant, Mϕs from WT, caspase 1^−/−, ASC^−/−, NLRP3^−/−, and NLRC4^−/− mice were infected with either LVS or the mviN mutant (Fig. 3A). Secretion of IL-1β in response both to WT LVS and to the mviN mutant was dependent on caspase 1 and ASC but independent of NLRP3 and NLRC4 (Fig. 3A). In contrast, caspase 1 activation and IL-1β secretion in response to infection both with WT LVS and with the mviN mutant were completely dependent on AIM2 (Fig. 3B, 3C). As expected, AIM2-deficient Mϕs were fully capable of secreting IL-1β in response to the NLRP3 agonist alum (Fig. 3B). Taken together, these data demonstrate that the augmented caspase 1 activation induced by the mviN mutant did not reflect the mviN mutant engaging a different inflammasome pathway, such as NLRP3 or NLRC4, but rather increasing activation of the AIM2-dependent pathway. Consistent with our in vitro findings, ASC^−/− and caspase 1^−/− mice infected i.p. with the mviN mutant succumbed to infection at similar rates, as did mice infected with WT LVS (Fig. 3D, 3E). In contrast to WT mice, no difference in bacterial burdens was observed in the spleen and liver of caspase 1^−/− mice infected with either LVS or the mviN mutant (Supplemental Fig. 4). Similarly, no difference in bacterial burdens was observed in the liver of ASC^−/− mice (Supplemental Fig. 4). Although ASC^−/− spleens did have statistically lower burdens of the mviN mutant compared with LVS (Supplemental Fig. 4), this difference was dramatically smaller than that observed in WT mice (3-fold versus 150,000-fold, respectively). These observations suggest that the attenuation of survival of the mviN mutant in vivo was due to its increased activation of the AIM2 inflammasome.

Given the importance of the AIM2 inflammasome in host defense against F. tularensis (6), it is not unexpected that a successful pathogen, such as F. tularensis, would evolve mechanisms to limit activation of this important innate immune pathway. Our results indicate that mviN was a critical for virulence of F. tularensis by virtue of its ability to restrict inflammasome activation. In fact, two genes have been identified in the related organism F. novicida, FTT_0748 and FTT_0584, that also inhibit Mϕ cytotoxicity and IL-1β secretion (28). Although the role of F. tularensis LVS genes FTL_1364 and FTL_1327, homologs of FTT_0748 and FTT_0584, respectively, remain to be elucidated, they do not appear to compensate for a deficiency in mviN. It appears that Francisella possesses multiple genes that contribute to evading inflammasome activation, raising the possibility that strains that elicit more severe systemic disease, such as the type A strain F. tularensis ssp. tularensis, may also more efficiently evade AIM2 inflammasome activation.

It is postulated that acidification of the Mϕ phagosome results in lysis of some ingested F. tularensis with the subsequent release of bacterial DNA (6). This DNA enters the cytoplasm, possibly through F. tularensis-induced phagosomal membrane damage, and activates the AIM2 inflammasome, resulting in the pyroptotic death of the Mϕs along with processing and secretion of IL-1β and IL-18. The aberrant morphology of the mviN mutant may leave the bacteria more susceptible to lysis, with increased release of bacterial DNA. Alternatively, bacteriolysis may be occurring within the cytosol once the bacteria have escaped the phagosome. Similar to our findings in this study, a recent study by Sauer et al. (29) identified an lmo2473 mutant of Listeria monocytogenes that induces hyperactivation of the AIM2 inflammasome following bacteriolysis in the macrophage cytosol. In conclusion, our data demonstrate that mviN played a critical role in limiting F. tularensis LVS-induced AIM2 inflammasome activation and in doing so subverted an important innate immune defense pathway.

We thank Richard Flavell, Anthony Coyle, Ethan Grant, and John Bertin for providing knockout mice, Leobaldo Solorzano and Vickie Knepper-Adrian for technical assistance, and Jian Shao for assistance with electron microscopy. We also thank Lee-Ann Allen for critical review of this manuscript.

Disclosures The authors have no financial conflicts of interest.