Importance of PdpC, IglC, IglI, and IglG for Modulation of a Host Cell Death Pathway Induced by Francisella tularensis

Abstract

Modulation of host cell death pathways appears to be a prerequisite for the successful lifestyles of many intracellular pathogens. The facultative intracellular bacterium Francisella tularensis is highly pathogenic, and effective proliferation in the macrophage cytosol leading to host cell death is a requirement for its virulence. To better understand the prerequisites of this cell death, macrophages were infected with the F. tularensis live vaccine strain (LVS), and the effects were compared to those resulting from infections with deletion mutants lacking expression of either of the pdpC, iglC, iglG, or iglI genes, which encode components of the Francisella pathogenicity island (FPI), a type VI secretion system. Within 12 h, a majority of the J774 cells infected with the LVS strain showed production of mitochondrial superoxide and, after 24 h, marked signs of mitochondrial damage, caspase-9 and caspase-3 activation, phosphatidylserine expression, nucleosome formation, and membrane leakage. In contrast, neither of these events occurred after infection with the ΔiglI or ΔiglC mutants, although the former strain replicated. The ΔiglG mutant replicated effectively but induced only marginal cytopathogenic effects after 24 h and intermediate effects after 48 h. In contrast, the ΔpdpC mutant showed no replication but induced marked mitochondrial superoxide production and mitochondrial damage, caspase-3 activation, nucleosome formation, and phosphatidylserine expression, although the effects were delayed compared to those obtained with LVS. The unique phenotypes of the mutants provide insights regarding the roles of individual FPI components for the modulation of the cytopathogenic effects resulting from the F. tularensis infection.

INTRODUCTION

Intracellular bacteria have evolved distinct strategies to survive and multiply within host cells, the most challenging of which is to evade the antibacterial immune mechanisms of the phagocytic cells. This can be achieved either by preventing the oxidative burst, by manipulating the phago-lysosome fusion, by having sufficient physical stability to withstand the antimicrobial mechanisms within the phagosome, or by escaping from the phagosome (1). Such evasion strategies can be counteracted by the host by induction of cell death, thereby efficiently removing the intracellular pathogen's habitat and exposing it to extracellular antibacterial mechanisms. However, induction of host cell death can also be advantageous for pathogens since it may avoid phagocytosis or allow escape from a habitat depleted of nutrients. Thus, manipulation of host cell death pathways is a critical step in the ever-continuing host-parasite battle. Host cell death occurs via several mechanisms, either without inflammation, apoptosis, or with accompanying inflammation, pyroptosis or necrosis, and all of the three pathways have been found to result from various types of infections (2).

Francisella tularensis is a Gram-negative, facultative intracellular bacterium that is able to infect many mammalian species and the etiological agent of the zoonotic disease tularemia (3). The pathogenicity of the bacterium is not fully understood, but replication in the macrophage appears to play an essential role. It rapidly escapes from the phagosome before fusion with the lysosome occurs, allowing it to multiply within the cytosol (4, 5). The initial encounter with the host cell leads to a rapid proinflammatory response, but this is repressed after internalization of the bacterium (6–8). The infected cells, regardless of whether they are dendritic or macrophages, are unable to secrete cytokines in response to secondary stimuli (9–11). Francisella novicida is a closely related species, and it has become a prototypic agent for studies of absent in melanoma 2 (AIM2)-mediated inflammasome activation, which leads to pyroptosis, a form of programmed cell death characterized by inflammasome formation resulting in caspase-1-dependent secretion of interleukin-1β (IL-1β) and IL-18 (12, 13). Infection with the live vaccine strain (LVS) also leads to secretion of IL-1β, but it has previously been observed that in addition, the infection leads to several cytopathogenic features characteristic of apoptosis but not pyroptosis, such as release of cytochrome c, caspase-3 activation, and nucleosome formation (12, 14–16). Thus, the cytopathogenic effects do not fully conform to either apoptosis or pyroptosis, and it is clear that F. tularensis is able to modulate many essential host cell pathways to facilitate its intracellular survival.

Much work aimed at understanding the bacterial factors responsible for modulation of the host cell signaling has been focused on the components constituting the Francisella pathogenicity island (FPI), a large genomic region duplicated in all strains of the highly virulent Francisella tularensis subsp. tularensis and subsp. holarctica and in the live vaccine strain (17, 18). The latter is an empirically attenuated strain of F. tularensis subsp. holarctica that has been used for many decades as a human vaccine but also in experimental models of tularemia, since it still shows marked virulence in, e.g., mice (19). The FPI encodes a type VI secretion system (T6SS), containing some 17 proteins, most of which have been found to be essential for the intramacrophage replication and virulence of the bacterium (20–23). When mutants have been generated in the FPI, many of them show a uniform phenotype characterized by lack of phagosomal escape, no intracellular replication, and a loss of virulence in vivo. Examples of such mutants include those of the iglABCD operon and the core components VgrG and DotU (20–25). In contrast, we have observed that the ΔiglI and ΔiglG mutants showed aberrant intracellular replication and very diminished cytopathogenic effects (26). The ΔpdpC mutant of the LVS and the highly virulent F. tularensis subsp. tularensis strain SCHU S4 strain shows no intracellular replication but other phenotypes distinct from those of the T6SS core component mutants (27, 28). To understand the contribution of T6SS components to the cytopathogenic effects of the F. tularensis infection, we analyzed how cell death pathways of infected macrophages were affected by infection with the LVS strain or the ΔpdpC, ΔiglC, ΔiglG, and ΔiglI T6SS mutants.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The live vaccine strain was originally obtained from the United States Army Medical Research Institute for Infectious Diseases (USAMRIID), Frederick, MD, USA. The ΔiglC, ΔpdpC, ΔiglG, and ΔiglI mutants and complemented strains have been described earlier (20, 26, 27). F. tularensis strains were grown on modified gonococci (GC) agar at 37°C, 5% CO₂. When required, kanamycin (10 μg/ml) was added to the medium.

Cultivation and infection of macrophages.

The murine macrophage-like cell line J774A.1 was used in all cell infection assays. Macrophages were cultured and maintained in Dulbecco's modified Eagle medium (DMEM) (Gibco BRL, Grand Island, NY, USA) with 10% heat-inactivated fetal bovine serum (FBS; Gibco). For all experiments, cells were seeded either in tissue culture plates or in 50-ml Falcon tubes, incubated overnight, and reconstituted with fresh culture medium 30 min prior to infection. A multiplicity of infection (MOI) of 200 was used, and bacteria, grown on agar plates, were suspended in PBS and kept on ice prior to infection. Intracellular replication experiments were performed essentially as described earlier (20).

Detection of cleaved caspase-9.

One million J774 cells were seeded in 6-well plates and incubated overnight. Infection with F. tularensis strains was allowed for 2 h before extracellular bacteria were removed by washing and fresh medium containing 5 μg/ml gentamicin was added. At indicated time points, plates were placed on ice and ice-cold PBS supplemented with complete EDTA-free protease inhibitor (Roche, Basel, Switzerland) was used to wash the monolayers once before Laemmli buffer was added. Cells were scraped and transferred to precooled microcentrifuge tubes. After 10 min of boiling, samples were separated on SDS-PAGE and transferred to nitrocellulose membrane. Mouse-specific anti-caspase-9 antibody (Cell Signaling Technologies) was used to detect procaspase-9 and cleaved caspase-9, followed by a secondary horseradish peroxidase (HRP)-conjugated donkey anti-rabbit antibody (GE Healthcare, United Kingdom). As a loading control, antibodies against β-actin were used.

Flow cytometry analysis.

For all flow cytometry experiments, cells were grown overnight and reconstituted with fresh medium the next day. Cells were allowed to recover for 30 min before being infected at an MOI of 200. After 2 h, the cells were once again pelleted and extracellular bacteria were removed by washing. Then, cells were resuspended in fresh medium and incubated for indicated times. A Vi-cell-XR cell viability analyzer (Beckman Coulter, Fullerton, CA, USA) was used to estimate cell viability at the time of sampling. Irrespective of the type of analysis, a minimum of 10,000 events were acquired for each analysis using a LSRII flow cytometer (BD Biosciences, San Jose, CA, USA) with FACSDiva software (BD Biosciences). Results were analyzed using FlowJo (Tree Star) software.

To measure the levels of reactive oxygen species (ROS), the MitoSOX red mitochondrial superoxide indicator kit (Molecular Probes) was used according to the manufacturer's instructions. Briefly, cells were incubated with medium containing 5 μM MitoSOX reagent for 10 min at 37°C and washed two times in PBS before analysis.

The changes in mitochondrial membrane potential (Δψ) were detected using the flow cytometry mitochondrial membrane potential detection kit MitoScreen (BD Biosciences) according to the manufacturer's instructions. In live cells with normal, polarized Δψ, JC-1 is taken up, leading to the formation of JC-1 aggregates, which results in high levels of red fluorescence. In mitochondria with depolarized Δψ, JC-1 remains in the cytoplasm as monomers, resulting in green fluorescence.

For measurement of phosphatidylserine (PS)-positive cells, the fluorescein isothiocyanate (FITC) annexin V apoptosis detection kit II (BD Biosciences) was used according to the manufacturer's instructions. Briefly, cells were stained with annexin V-FITC and propidium iodide (PI). Early apoptosis was defined by annexin V⁺/PI⁻ staining and late apoptosis by annexin V⁺/PI⁺ staining.

For staining for active caspase-3, cells were transferred to 96-well plates and centrifuged for 3 min at 500 × g. Supernatants were removed, and cells were washed once in PBS. After fixation and permeabilization by Cytofix/Cytoperm (BD Biosciences) for 20 min at room temperature (RT), cells were pelleted and washed with Perm/Wash (BD Biosciences) and subsequently stained with phycoerythrin (PE)-conjugated rabbit anti-human active caspase-3 antibody (clone C92-605; BD Biosciences). Cells were then washed and resuspended in Perm/Wash buffer. Staurosporin (Sigma) was used as a positive control for cell death.

Terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) staining was carried out using the In Situ Cell Death Detection kit (Roche) according to the manufacturer's instructions.

Multiplex cytokine analysis.

Supernatants from infected cells were collected and stored at −80°C until analyzed using the Bio-Plex Pro Mouse cytokine 23-plex kit and a Bio-Plex 200 system (Bio-Rad Laboratories) according to the manufacturer's instructions. Samples were analyzed in duplicates. For statistical analysis, samples below the standard range were set to 0.5 times the detection limit for the respective cytokine, and samples above the standard range were set to the highest standard value for the respective cytokines.

Data analysis and statistical methods.

Statistical significances were determined using paired, two-tailed Student's t tests or Pearson's chi-square test. To compare the cytokine and chemokine data, principal component analysis (PCA) was used. All data were log transformed and analyzed by principal component analysis using the SIMCA program, version 13.

RESULTS

ΔiglG and ΔiglI mutants but not ΔpdpC or ΔiglC mutants are able to replicate in J774 cells.

Previously, we and others have shown that the FPI mutants display a spectrum of intracellular phenotypes, since some replicate as well as the wild type whereas others show compromised or no growth (20, 23–26, 28–30). We followed bacterial numbers of the ΔpdpC, ΔiglC, ΔiglG, and ΔiglI mutants and the LVS strain in J774 cells after infection. In agreement with previously published data (20, 23–26, 28–30), the ΔiglC and ΔpdpC mutants showed very marginal net replication, whereas the ΔiglG and the ΔiglI mutants replicated as well as LVS (Table 1). In separate experiments, we observed that the complemented strains showed the same degree of replication as did LVS (not shown).

Table 1.

Intracellular replication of F. tularensis strains in J774 cells^a

F. tularensis strain	Net increase of bacterial population (mean log10 ± SD)/well) from 0 to 24 h
LVS	3.6 ± 0.2
ΔiglC mutant	0.7 ± 0.1b
ΔpdpC mutant	1.0 ± 0.2b
ΔiglG mutant	4.2 ± 0.1
ΔiglI mutant	3.6 ± 0.1

^aThe bacterial strains were added to the J774 cells at an MOI of 200. Data are from one representative experiment (of 3) for 3 cultures.

^bSignificantly lower than the LVS increase by Student's t test.

Mitochondrion-specific superoxide is induced upon infection with LVS or ΔpdpC or ΔiglG mutant strains.

The mitochondrial production of ROS increases in response to various stimuli and is indicative of mitochondrial dysfunction related to many pathophysiological processes, such as inflammation, infection, cancer, and aging (31). Accumulation of mitochondrial ROS occurs concomitantly to redistribution of cytochrome c from the mitochondria into the cytoplasm, where it causes activation of caspases and triggering of the intrinsic pathway of apoptosis (32). We measured the amount of mitochondrial superoxide in J774 cells using the indicator dye MitoSOX, which penetrates cell and mitochondrial membranes and, upon contact with mitochondrial superoxide, emits red fluorescence. We observed that the number of superoxide-positive cells was drastically increased upon infection with LVS (Fig. 1; see Fig. S1 in the supplemental material). Within 12 h, around 50% of the cell population was MitoSOX positive in LVS-infected cells, in comparison to <4% of uninfected cells. At this time point, no cells infected with the LVS mutants differed markedly from uninfected cells, whereas all the complemented strains had a phenotype more or less identical to that of LVS. After 24 h, however, 42% and 25% of the ΔpdpC and ΔiglG mutant-infected cells, respectively, but <15% of the ΔiglI and ΔiglC mutant-infected and uninfected cells were MitoSOX positive (Fig. 1; see Fig. S1 in the supplemental material).

Fig 1.

Mitochondrial superoxide accumulation in response to infection. J774 cells were infected with the indicated strains of F. tularensis for 2 h, after which the cells were washed and incubated for 12 or 24 h. The amount of mitochondrial superoxide was measured by assessing the oxidation of the fluorogenic dye MitoSOX, and the percentages of MitoSOX-positive cells are shown. The data shown are representative of one experiment of two performed. Pearson's chi-square test was used to test if the numbers of gated cells were significantly different from those of LVS-infected cells (***, P < 0.001).

Thus, infection with LVS or ΔpdpC or ΔiglG mutant strains leads to mitochondrial superoxide production, albeit with delayed kinetics for the last two infections. In contrast, very little or no production was seen in cells infected with ΔiglC or ΔiglI mutants.

Mitochondrial membrane instability is induced in cells infected with LVS or the ΔpdpC mutant.

Among the first events of the mitochondrially triggered intrinsic apoptotic pathway is the perturbation of the mitochondrial transmembrane potential (32). We measured it by utilizing the JC-1 dye, the fluorescence of which will be affected by the mitochondrial membrane potential; as the dye accumulates inside the mitochondria with intact membrane potential, it forms aggregates with red fluorescence. In cells with depolarized mitochondrial membranes, this kind of aggregate is not formed and the JC-1 dye exists as green fluorescent monomers. Previously, we have demonstrated that the LVS infection resulted in no significant change of the potential during the first 12 h, whereas a significant decrease occurred at 18 h and onwards (16). Therefore, the membrane potentials were determined at 24 and 48 h. After this time point, very few of the LVS-infected cells are intact and total cell numbers decrease drastically; therefore, we did not follow the effects beyond 48 h. At both time points, only minimal changes were observed in the ΔiglC, ΔiglI, and ΔiglG mutant-infected cells compared to uninfected cells, whereas already at 24 h, 72% and 56%, respectively, of the LVS- and ΔpdpC mutant-infected cells demonstrated significant decreases of the potential (Fig. 2; see Fig. S2 in the supplemental material). The effects were even more marked at 48 h as similar proportions, >83%, of both LVS- and ΔpdpC mutant-infected cells were affected (Fig. 2; see Fig. S2 in the supplemental material). Thus, the ΔpdpC mutant infection resulted in marked changes in the mitochondrial potential, albeit with slightly delayed kinetics compared to the LVS infection. In contrast, very little or no changes in the potential were observed in cells infected with ΔiglC, ΔiglG, or ΔiglI mutant strains. All complemented strains had a phenotype similar to that of LVS.

Fig 2.

Effect of F. tularensis infection on mitochondrial membrane stability. J774 cells were infected with the indicated strains of F. tularensis for 2 h, after which the cells were washed and incubated for 24 or 48 h. They were then stained with MitoScreen and analyzed by flow cytometry. The MitoScreen reagent, JC-1, indicates mitochondrial membrane stability; intact membrane potential leads to red fluorescence and reduced potential to green fluorescence. The data shown are representative of one experiment of two performed. Pearson's chi-square test was used to test if the numbers of gated cells were significantly different from those of LVS-infected cells (***, P < 0.001).

Cells infected with LVS or the ΔpdpC strain become markedly annexin V positive.

Surface exposure of phosphatidylserine (PS) is considered to be an early event of apoptosis, preceding the full sequence of morphological changes at the ultrastructural level (33). At this initial stage of cell death, the barrier function of the plasma membrane remains intact. At a later stage of apoptosis, the plasma membrane becomes leaky for compounds like propidium iodide (PI). Thus, we monitored the presence of surface-exposed PS by staining with annexin V, a phospholipid-binding protein with high affinity for PS, and concomitantly the integrity of the plasma membrane by staining with PI.

At 24 h, 21% of the LVS-infected cells were both annexin V and PI negative, 15% were annexin V positive only, and 53% were both annexin V and PI positive. In contrast, the corresponding percentages for the ΔiglC mutant strain-infected cells were 78%, 11%, and 10%, clearly distinct from the percentages for LVS-infected cells but similar to those of uninfected cells (Fig. 3). Cells infected with either the ΔiglG or ΔiglI mutants were similar to uninfected cells, whereas ΔpdpC mutant-infected cells and those infected with the complemented strains had phenotypes more similar to that of LVS. Of the ΔpdpC mutant-infected cells, 36% were both annexin V and PI negative, 31% were annexin V positive only, and 23% were both annexin V and PI positive (Fig. 3).

Fig 3.

Annexin V and propidium iodide (PI) staining of infected J774 cells. Cells were infected with the indicated F. tularensis strains for 2 h and thereafter washed. Cells were incubated for 24 or 48 h before being subjected to annexin V/PI staining and analyzed using flow cytometry. Unaffected cells (lower left quadrant) are both annexin V and PI negative; apoptotic cells with intact membrane integrity (lower right quadrant) are annexin V positive but PI negative; apoptotic cells that have lost their membrane integrity (upper right quadrant) are both annexin V and PI positive. The scatter plots show the entire cell population for each sample and time point. The data shown are from one representative experiment of three performed.

The differences became more marked at 48 h, as only 3% of the LVS-infected cells were negative for both annexin V and PI, 29% positive for annexin V, and 67% positive for both PI and annexin V. In contrast, the proportions seen in ΔiglC mutant strain-infected cells and uninfected cells did not differ much from those seen at 24 h, whereas cells infected with either the ΔiglG or ΔiglI mutants showed intermediate phenotypes. Of the latter, 45% and 38%, respectively, were annexin V positive only and 24% and 15%, respectively, both annexin V and PI positive (Fig. 3). The ΔpdpC mutant-infected cells were distinct, since a large majority, 88%, were only annexin V positive and 7% were positive for both annexin V and PI. The complemented strains had a phenotype similar to that of LVS-infected cells; at 24 h the proportions of annexin V- and PI-positive cells were 38 to 48% and at 48 h 58 to 66% (data not shown).

These results indicate that upon an LVS infection, a significant increase in the relative numbers of annexin V-positive cells occurs within 24 h and that the proportion of cells that are both annexin V and PI positive, presumably dying cells, increases over time. In contrast, very few of the ΔiglC mutant-infected cells became annexin V and/or PI positive within 48 h, whereas the ΔpdpC mutant infection resulted in higher proportions of annexin V-positive, PI-negative cells at both time points than did the LVS infection but showed very few double-positive cells. Cells infected with the ΔiglG or ΔiglI mutants showed intermediate phenotypes with larger proportions of annexin V-positive cells than for ΔiglC mutant infection but much lower numbers than for LVS- or ΔpdpC mutant-infected cells.

Detection of caspase-9 in infected J774 cells.

As a response to the mitochondrial membrane destabilization, cytochrome c is released into the cytoplasm and interacts with ApaF1 to activate caspase-9, which in turn will activate caspase-3 and other effector caspases (34). The proform and mature form of caspase-9 were distinguished using Western blot analysis. Infection with LVS or the complemented strains led to distinct cleavage already within 24 h, whereas none of the four mutants showed any distinctly cleaved caspase-9 at 24 or 36 h (Fig. 4). Not even at 48 h was marked cleavage of caspase-9 detected, although weak cleavage in cells infected with the ΔiglG mutant was observed (Fig. 4).

Fig 4.

Activated (cleaved) caspase-9 (C-9) in response to F. tularensis infection. J774 cells were infected with the indicated strains of F. tularensis and cell lysates prepared after 24, 36, or 48 h. A caspase-9 antibody was used to detect procaspase-9 and activated caspase-9 by standard Western blot techniques. An antibody against β-actin was used as a loading control. The data shown are from a representative experiment of two performed.

Thus, infection with LVS and the ΔiglG mutant strain, although the latter much less prominently, led to activation of caspase-9.

Caspase-3 cleavage occurs in LVS-, ΔpdpC mutant-, or ΔiglG mutant-infected cells.

Caspase-3 is a crucial effector in many forms of caspase-dependent cell death and is activated through both the extrinsic and intrinsic pathways and required for condensation of the nucleus and DNA fragmentation (35). The infected cell populations were analyzed using flow cytometry and an antibody specific to the activated form of caspase-3 at 24 or 48 h; uninfected cells or cells treated with the apoptosis-inducing substance staurosporin were used as controls. Within 24 h, there were clear differences among the cell populations; infection with LVS and the ΔpdpC strain resulted in at least 35% positive cells, whereas less than 10% of uninfected cells or cells infected with the ΔiglC, ΔiglG, or ΔiglI mutants were caspase-3 positive (Fig. 5; see Fig. S3 in the supplemental material). After 48 h, the proportions of caspase-3-positive cells were approximately 20% for the ΔiglC and ΔiglG mutant-infected cells, whereas all other infected cells showed only small changes compared to the 24 h time point (Fig. 5; see Fig. S3 in the supplemental material). Cells infected with the complemented strains were caspase-3 positive almost to the same extent as LVS-infected cells, with the exception of the complemented ΔiglI mutant, which showed intermediate values. In staurosporin-treated, uninfected cells, 86% were caspase-3 positive.

Fig 5.

Percentage of infected J774 cells containing active caspase-3. Cells were infected with the indicated strains of F. tularensis, washed, and incubated for either 24 or 48 h before the portion of cells containing active caspase-3 was measured using flow cytometry. The data shown are representative of one of two experiments performed. Pearson's chi-square test was used to test if the numbers of gated cells were significantly different from those of LVS-infected cells (***, P < 0.001).

Nucleosome formation occurs within 24 h in cells infected with LVS but is delayed in cells infected with ΔpdpC or ΔiglG strains.

At the terminal stage of apoptosis, DNA fragmentation occurs, leading to nucleosome formation (33). Here, it was assessed by following the appearance of TUNEL-positive cells. At 12 h, there were very low numbers of such cells even after infection with LVS (data not shown). At 24 h, and in repeated experiments, infection with either of the mutants resulted in no significant fragmentation, <4%, whereas the corresponding number for LVS-infected cells was 20% (Fig. 6; see Fig. S4 in the supplemental material). At 48 h, the ΔpdpC infection also resulted in significant fragmentation and 22% of the cells were positive versus 17% of the LVS-infected cells, whereas <5% of the cells infected with either of the other mutants were TUNEL positive (Fig. 6; see Fig. S4 in the supplemental material). It should be noted that staining of bacterial DNA may also occur with this method. However, since the infection with the replicating ΔiglI mutant consistently resulted in very low values, <5%, the data indicate that the fragmentation occurred only in the eukaryotic DNA.

Fig 6.

DNA fragmentation in cells infected with F. tularensis. J774 cells were infected with the indicated strains for 2 h and thereafter washed. They were incubated for 24 or 48 h before lysates were prepared and analyzed using the TUNEL assay and flow cytometry. Shown is the percentage of TUNEL-positive cells for each infection and time point. The data are representative of one experiment of three performed. Pearson's chi-square test was used to test if the numbers of gated cells were significantly different from those of LVS-infected cells (*, P < 0.05; ***, P < 0.001).

Collectively, the data indicate that infection with the LVS and the ΔpdpC strains, but not the other mutants, resulted in significant levels of nucleosome formation.

Cytokine secretion patterns reveal clear distinctions of the effects from infections with different strains of F. tularensis.

In view of the distinct phenotype of several of the mutants in the aforementioned assays, we performed a comprehensive multiplex analysis of the secretion of 22 cytokines and chemokines from J774 cells infected with either of the four mutants or the LVS strain. At 24 h, levels were much higher in ΔiglC mutant-infected than in LVS-infected cells and higher than or identical to those of uninfected cells. The differences between ΔiglC mutant-infected cells and LVS-infected cells were 2- to 4-fold for granulocyte-macrophage colony-stimulating factor (GM-CSF), gamma interferon (IFN-γ), IL-12p70, IL-13, IL-17, IL-1α, IL-2, IL-3, IL-4, IL-5, IL-9, KC (also known as chemokine [C-X-C motif] ligand 1), and tumor necrosis factor alpha (TNF-α) and >4-fold for eotaxin, granulocyte colony-stimulating factor (G-CSF), IL-10, IL-12p40, IL-6, monocyte chemoattractant protein 1 (MCP-1), macrophage inflammatory protein 1α (MIP-1α), MIP-1β, and RANTES (Table 2). The secretion patterns from ΔiglG or ΔiglI mutant-infected cells were intermediate. In contrast, the secretion pattern from ΔpdpC mutant-infected cells was much more similar to that of LVS-infected cells than that of ΔiglC mutant-infected cells; all cytokine levels were <2-fold different compared to the levels seen for LVS.

Table 2.

Cytokines secreted by J774 cells 24 h postinfection with F. tularensis^a

Cytokine	Amt (pg/ml) of cytokine secreted after infection with:
	LVS	ΔiglC mutant	ΔpdpC mutant	ΔiglG mutant	ΔiglI mutant
Eotaxin	105	840	171	437	290
G-CSF	5.0	32.9	7.1	17.5	15.4
GM-CSF	36.6	92.9	42.2	68.8	53.5
IFN-γ	1.2	3.4	1.7	2.3	1.9
IL-1α	4.9	13.5	5.5	9.2	8.6
IL-2	1.6b	4.8	1.7b	1.7b	1.7b
IL-3	0.8b	2.9	0.8	1.8	0.8b
IL-4	3.9b	15.5	3.9b	8.0	3.9b
IL-5	1.5b	3.0	1.5b	1.5b	1.5b
IL-6	1.3	0.2b	0.2b	0.2b	0.2b
IL-9	72.8	268	94.6	183	134
IL-10	11.3	58.5	15.3	32.5	26.2
IL-12(p40)	3.9	28.3	6.5	17.4	10.3
IL-12(p70)	12.4	40.7	15.6	26.4	22.4
IL-13	61.8	194.3	67.6	125	98.4
IL-17	1.4b	4.2	1.4b	1.4b	1.4b
KC	1.2b	4.4	1.2b	3.1	1.2b
MCP-1	1,709	25,520	1,980	8,491	5,125
MIP-1α	486	28,540c	600	3,931	2,031
MIP-1β	327	12,384c	441	4,156	1,457
RANTES	24.9	121	31.9	73.7	49.1
TNF-α	8.5	28.4	9.5	15.4	11.2

^aJ774 cells were infected at an MOI of 200 and incubated for 24 h before supernatants were collected and cytokine amounts were measured.

^bValue below the standard range, set to 0.5 times the lowest standard value.

^cValue above the standard range, set to the highest standard value.

In conclusion, the secreted cytokines indicate that the ΔpdpC mutant-infected cells showed the phenotype most similar to that of the LVS-infected cells, whereas the other types of infected cells demonstrated distinct patterns. This was also confirmed with principal component analysis (PCA), and the analysis resulted in a model that explained 91% of the variation of the data (Fig. 7). The secretion pattern of the ΔiglC mutant-infected cultures was separated from that of the other infections and that of the uninfected cultures was clearly separated from that of all infected cultures. The secretion patterns of LVS- and ΔpdpC, ΔiglI, and ΔiglG mutant-infected cultures were each distinct, and their distances to the ΔiglC mutant-infected cultures were in the corresponding order, with the LVS-infected culture being the most distant.

Fig 7.

Principal component analysis (PCA) of cytokine gene expression or secretion of peritoneal exudate cells infected with indicated F. tularensis strains. The data are presented in Table 2. The two components shown explained 91% of the variation of the data set.

DISCUSSION

The virulence of F. tularensis is intimately linked to its ability to replicate in macrophages since mutants defective for intramacrophage replication are rendered avirulent. A prerequisite for the intracellular replication is a functional type VI secretion system, encoded by the FPI. Many of the FPI components are presumed to be essential core components since they appear critically required and in their absence, a lack of phagosomal escape, no intracellular replication, and lack of virulence are manifested (17). The best-characterized example of such mutants is the ΔiglC mutant. Not all of the investigated mutants so far fit into this uniform pattern, however, since the ΔiglI and ΔiglG mutants of LVS show delayed cytopathogenicity and lack of virulence though more or less intact intracellular replication in J774 cells, whereas only the latter mutant replicated in peritoneal exudate cells and bone marrow-derived macrophages (26). In addition, in a recent study, we observed that the ΔpdpC mutant of LVS showed a unique phenotype, characterized by a lack of intracellular replication, incomplete phagosomal escape, and marked attenuation in the mouse model; however, unlike a phagosomally contained FPI mutant, it triggered secretion of IL-1β and an MOI-dependent release of lactate dehydrogenase (LDH), a marker of membrane damage (27). Another recent publication demonstrated likewise that the ΔpdpC mutant of the highly virulent F. tularensis subsp. tularensis strain SCHU S4 displays incomplete phagosomal escape and lacks intracellular replication (28).

We wanted to understand how the unique FPI mutant phenotypes were related to the marked cytopathogenic effects that are hallmarks of the F. tularensis infection. This will provide important information regarding the roles of individual FPI components for the modulation of the cellular pathways required for the successful intracellular lifestyle of F. tularensis as well as provide a more complete understanding of its T6SS.

A number of publications have demonstrated that F. novicida triggers cell death via release of its DNA into the cytosol, leading to recruitment of absent in melanoma 2 (AIM2) and apoptosis-associated specklike protein (ASC), which result in inflammasome activation, cleavage of caspase-1, and release of IL-1β, a form of cell death designated pyroptosis (12, 13, 36, 37). Although a variant of programmed cell death, it is distinct from the immunologically silent cell death taking place during apoptosis and it occurs independently of proapoptotic caspases (38). On the cellular level, it is characterized by plasma membrane rupture, water influx, cellular swelling, osmotic lysis, release of proinflammatory cellular content, and DNA cleavage (39). The DNA degradation is executed by an unknown nuclease and is distinct from the nucleosome formation characteristic of apoptosis, and conversely, caspase-1 is not involved in apoptosis (38). As demonstrated in many studies, not only the F. novicida infection but also the F. tularensis macrophage infection is characterized by secretion of IL-1β, indicating a pyroptotic mechanism (26, 36, 37, 40–44); however, at the same time, it is also characterized by activation of proapoptotic caspases and other features of the mitochondrially triggered intrinsic apoptotic pathway (14–16). In fact, studies in vivo using virulent strains demonstrated very little activation of caspase-1 and normal pathology in caspase-1-deficient mice, whereas cell death strongly correlated to caspase-3 activation (14). Thus, the evidence indicates that the effects of the F. tularensis infection may not unambiguously fit with the characteristics of either apoptosis or pyroptosis. This conclusion was corroborated by our recent findings that ΔiglG and ΔpdpC mutant-infected cells secrete IL-1β (26, 27) but, as we show in this study, also display many features characteristic of activation of the intrinsic apoptotic pathway, such as mitochondrial damage, caspase-3 activation, and nucleosome formation. The reasons for these dichotomous findings are unclear but may be linked to the presence of simultaneous signals that trigger both types of pathways or, alternatively, that the unusual form of inflammasome activation triggered by F. tularensis also leads to engagement of the intrinsic apoptotic pathway; however, there is no direct evidence for the latter hypothesis. It should be noted that regardless of stimulus, IL-1β secretion from J774 cells is very low (26), indicating that the inflammasome activation may be defective and, therefore, alternative cell death pathways may be more easily discernible in this cell type. An interesting finding on F. novicida was recently published demonstrating that an AIM2/ASC-dependent, caspase-3-mediated apoptosis, which also involves caspase-8 and caspase-9, occurs in caspase-1-deficient macrophages (45). If these findings also are relevant for the LVS strain, then they would support the hypothesis that the inflammasome activation triggered by Francisella may lead to several types of terminal events and the ultimate cause of the cell death will depend on both bacterial and host factors. An example of this is our previous demonstration that the F. novicida U112 strain more potently induces release of IL-1β than does LVS (26). Thus, the lower potency of LVS to induce IL-1β may suggest that cell death pathways other than pyroptosis are also important for the cell death occurring during the LVS infection.

Besides our findings that the LVS infection resulted in activation of features characteristic of the intrinsic apoptotic pathway, we observed that each of the mutants showed very distinct features in this regard. Although the ΔiglI and ΔiglG mutants replicated and the ΔiglC and ΔpdpC mutants did not, this did not correlate to the resulting cytopathogenic effects. In almost all aspects, infection with the ΔiglC mutant showed very marginal or no cytopathogenic effects and also infection with the ΔiglI mutant resulted in minimal effects. In contrast, the ΔiglG mutant showed an intermediate phenotype, and in most assays, it showed minimal effects after 24 h but significant effects after 48 h. The most unexpected effects resulted from infection with the ΔpdpC mutant since despite its lack of replication and unlike the other mutants, it resulted in marked activation of most mechanisms analyzed, although with delayed kinetics compared to the LVS infection. Specifically, the ΔpdpC mutant markedly triggered mitochondrial damage, caspase-3 activation, PS expression, and DNA fragmentation. In contrast to the LVS infection, infection with the mutant, somewhat unexpectedly, did not lead to caspase-9 cleavage or PI-positive cells.

The ΔiglC mutant has served as a prototype for F. tularensis T6SS mutants, and it has been found to lack phagosomal escape, intracellular replication, and virulence in the mouse model (20–22). In addition, it induces increased expression of a subset of TLR2-dependent, proinflammatory genes (11). In the present investigation, we observed that the cytopathogenic effects on the ΔiglC mutant-infected cells were very discrete, in most respects not very different from those of uninfected cells, although the cytokine secretion patterns were clearly distinct. Thus, the lack of phagosomal escape appears to result in minimal cytopathogenic effects but distinct secretion of cytokines. Most likely these features are not modulated directly by IglC but result from the lack of a functional T6SS due to the essential role of the protein for the secretion system. It should be noted, however, that no defined role of IglC for the secretion machinery has been demonstrated, and in fact, a recent publication demonstrated that it is secreted during macrophage infection (46).

The ΔiglI and ΔiglG mutants of LVS demonstrate aberrant intramacrophage replication and very diminished LDH release (26). The former strain replicated only in J774 cells, and the latter showed slightly delayed replication in bone marrow-derived macrophages (BMDM), but their phagosomal escape in J774 cells was only slightly delayed compared to that of LVS. Additionally, a recent publication demonstrated that the SCHU S4 ΔiglI mutant did not replicate in BMDM (28). These features appear to coincide with distinct patterns of activation of the host cell death pathways investigated, since infection with the ΔiglI mutant resulted in effects similar to those of the ΔiglC mutant strain infection, whereas the ΔiglG mutant infection resulted in an intermediate phenotype with marginal cytopathogenic effects after 24 h but significant effects after 48 h, although not as marked as those due to the LVS infection. Still, the infection with the ΔiglG mutant led to a secreted cytokine pattern similar to that of infections with the ΔiglI and ΔiglC mutants. The findings suggest that IglI and IglG are indirectly or directly involved in modulating the cytopathogenic effects of the F. tularensis infection. Although it is possible that their functions are essential for T6SS and their effects are thereby indirect, as for IglC, our previous findings that the corresponding mutants can replicate in certain types of macrophages and escape from the phagosome in J774 cells (26) show that they are distinct from mutants such as the ΔiglC mutant and that it is more likely that they affect the cytopathogenic effects by other means than the T6SS core components do.

The present investigation reveals that the ΔpdpC mutant of LVS is another example of an FPI mutant with a very distinct and paradoxical phenotype, since it in some aspects mimics that of the LVS strain; for example, it shows induction of IL-1β and inhibition of LPS-induced TNF-α release, whereas in other aspects it is very distinct since it does not show normal escape into the cytosol, lacks intramacrophage replication, and is highly attenuated in the mouse model (27). Our present findings further corroborate the unusual phenotype of the ΔpdpC mutant, since it was found to be distinct from all other investigated mutants and infection led to marked mitochondrial damage, caspase-3 cleavage, expression of PS, and DNA fragmentation, although with delayed kinetics compared to LVS. Moreover, the secreted cytokine pattern was very similar to that of LVS-infected cells. We have previously proposed that the aberrant behavior of the mutant is due to its partial degradation of the phagosomal membrane since this could lead to the release of bacterial components into the cytosol that activate pathways resulting in cytopathogenic effects and a cytokine pattern like that induced by LVS (27). This hypothesis assumes that PdpC does not directly modulate the cytopathogenic signaling but rather is involved in the degradation of the phagosomal membrane. The phenotypic differences between the ΔpdpC mutant and the ΔiglG and ΔiglI mutants, both of which also degrade the phagosomal membrane, may be dependent on specific roles of IglG and IglI to modulate the cytosolic signaling that leads to cytopathogenicity.

In contrast to our findings on macrophages, it has recently been demonstrated that F. tularensis infection leads to a delay of the host cell death of neutrophils (47). A fundamental difference between the two cell types is that neutrophils are very short-lived cells that undergo constitutive apoptosis (48). Therefore, it may be that the rapid, spontaneous host cell death may not be advantageous for an intracellular bacterium since the life span may not be sufficient for significant replication to occur, and thus delay of apoptosis is advantageous for the pathogen, whereas this is not a problem in the more long-lived macrophage. In the latter cell type, however, deprivation of nutrients eventually may be an issue, and therefore, induction of host cell death may be beneficial since it allows the spread of the pathogen to nutrient-replete cells.

In summary, we have shown that the F. tularensis LVS infection leads to induction of host cell death that is similar to the intrinsic apoptotic pathway, although it is well established that the infection leads to secretion of IL-1β, a hallmark of pyroptosis, whereas infection with each of four FPI mutants results in cytopathogenic effects that are distinct from the LVS infection and, in some instances, with essentially no cytopathogenic effects. The findings provide novel insights regarding the roles of individual FPI components for the modulation of the cytopathogenic effects resulting from the F. tularensis infection and contribute to a more thorough understanding of its enigmatic T6SS.