Mutations of Francisella novicida that Alter the Mechanism of Its Phagocytosis by Murine Macrophages

Infection with the bacterial pathogen Francisella tularensis tularensis (F. tularensis) causes tularemia, a serious and debilitating disease. Francisella tularensis novicida strain U112 (abbreviated F. novicida), which is closely related to F. tularensis, is pathogenic for mice but not for man, making it an ideal model system for tularemia. Intracellular pathogens like Francisella inhibit the innate immune response, thereby avoiding immune recognition and death of the infected cell. Because activation of inflammatory pathways may lead to cell death, we reasoned that we could identify bacterial genes involved in inhibiting inflammation by isolating mutants that killed infected cells faster than the wild-type parent. We screened a comprehensive transposon library of F. novicida for mutant strains that increased the rate of cell death following infection in J774 macrophage-like cells, as compared to wild-type F. novicida. Mutations in 28 genes were identified as being hypercytotoxic to both J774 and primary macrophages of which 12 were less virulent in a mouse infection model. Surprisingly, we found that F. novicida with mutations in four genes (lpcC, manB, manC and kdtA) were taken up by and killed macrophages at a much higher rate than the parent strain, even upon treatment with cytochalasin D (cytD), a classic inhibitor of macrophage phagocytosis. At least 10-fold more mutant bacteria were internalized by macrophages as compared to the parent strain if the bacteria were first fixed with formaldehyde, suggesting a surface structure is required for the high phagocytosis rate. However, bacteria were required to be viable for macrophage toxicity. The four mutant strains do not make a complete LPS but instead have an exposed lipid A. Interestingly, other mutations that result in an exposed LPS core were not taken up at increased frequency nor did they kill host cells more than the parent. These results suggest an alternative, more efficient macrophage uptake mechanism for Francisella that requires exposure of a specific bacterial surface structure(s) but results in increased cell death following internalization of live bacteria.


Introduction
Francisella tularensis is a Gram-negative facultative intracellular pathogen and the causative agent of tularemia. Four subspecies of F. tularensis are recognized according to Bergey's Manual of Systematic Bacteriology [1]: F. tularensis subsp. tularensis (type A), F. tularensis subsp. holarctica (type B), F. tularensis subsp. mediasiatica, and F. tularensis subsp. novicida. F. novicida strain U112 is pathogenic for mice, but not for man and the genes are on average .95% identical to F. tularensis strain SchuS4, making it an acceptable model system. Another advantage of working with F. novicida U112 is the availability of a comprehensive transposon library containing two insertion alleles for the majority of non-essential genes [2].
Lipopolysaccharide (LPS) is the major structural component of the outer membranes of all Gram-negative bacteria and recognition of this unique structure by the host is a key factor in activating a robust immune response. LPS is a tripartite macromolecule comprised of lipid A, the core, and an O-antigen and is assembled sequentially beginning with lipid A. Lipid A anchors the LPS in the outer membrane and is a disaccharide of glucosamine that is both acylated and phosphorylated [3]. The core and O-antigen carbohydrate domains are linked to the lipid A moiety through the eight-carbon sugar 3-deoxy-D-mannooctulosonic acid (Kdo) [3]. Francisella (and other bacteria) begin with synthesis of di-phosphorylated tetra-acylated lipid IVA in the inner leaflet of the inner membrane. In Francisella, this structure is initially penta-acylated and transferred to the outer leaflet of the inner membrane or, alternatively, the core sugars can also be added before the transfer [4]. Two Kdo (a disaccharide) are transferred to lipid A by KdtA while the molecule is still on the inner leaflet of the inner membrane. Once transferred to the periplasm, the nascent chain is dephosphorylated at the 49 position by lpxF and one of the two Kdo saccharides is removed by Kdo hydrolase followed by the addition of mannose di-saccharide to form the core [3,5]. The structure is transferred to the outer membrane by the Lpt ABCFG transport system. In the outer membrane, one of the acyl groups is removed via the 39 Oacylase, leaving a tetra-acylated form. Two of the genes described in this report, wbtA and wzx, are both involved in the synthesis of the Oantigen [3]. WbtA functions as a dehydratase and is likely required for the synthesis of the first sugar attached to the core and thus all subsequent sugars as well [6,7]. Based on homology with E. coli, the Oantigen flippase, encoded by wzx, translocates individual carbohydrate residues from the cytosol to the periplasm where Wzy polymerizes them into the O-antigen, which is then transferred to the core [8]. Strains that lack wbtA or wzx synthesize a LPS that lacks an O-antigen [8,9].
The lipid A moiety without the attached Kdo unit represents the minimal LPS substructure required for bacterial viability in Francisella subspecies [10]. Compared to the E. coli or Salmonella LPS, the Francisella LPS is a much less potent endotoxin and does not stimulate inflammatory pathways via TLR4 signaling. The lack of proinflammatory response is related to hypoacylation and the addition of very long chain fatty acids attached to the diglucosamine backbone [11]. The absence of a phosphate at the 49 position is also important as compared with LPS from enteric organisms. In E. coli, removal of the 49 phosphate reduces lipid A toxicity [12].
While the LPS biosynthesis genes have been only partially characterized in Francisella, the orthologous genes in E. coli and Salmonella are well characterized and likely have the same function. The addition of mannose to the core of the LPS requires the action of at least three genes, lpcC, manB and manC ( Figure 1). Group 1 glycosyl transferase (lpcC) adds a mannose unit to the inner Kdo moiety of the LPS precursor in E. coli. Among the genes required for the mannose pathway in E. coli, manB (pmm or rfbk) encodes phosphomannomutase, which converts mannose-6-phosphate to mannose-1-phosphate while manC encodes a mannose-1-phosphate guanylyltransferase, which catalyzes mannose-1-phosphate to GDP-D-mannose (the substrate for the mannose glycosyltransferase, LpcC). Without lpcC, manB, and manC, mannose would not be added to the LPS structure thereby preventing the addition of the O side chains and leaving Kdo as the outer most polysaccharide [13,this report].
Francisella is readily phagocytosed by macrophages by an actindependent mechanism [14][15][16]. Following phagocytosis, F. tularensis strains replicate intracellularly and eventually kill both cultured macrophage cell lines and primary macrophages [15,[17][18][19]]. In the current study, mutations in 28 genes identified from the comprehensive F. novicida transposon library were found to increase cell toxicity upon infection of J774 macrophage-like cells. To determine whether the mutant strains must be internalized by the macrophages to promote cell death, we inhibited actin polymerization by adding cytochalasin D (cytD), a classic inhibitor of macrophage phagocytosis, prior to cell infection. Surprisingly, three of the mutant strains (lpcC, manB, manC) were cytotoxic to macrophages treated with cytD. In-frame deletions that removed more than 90% of the coding sequence and were designed to not interfere with expression of downstream genes in the same operon were constructed in these and other genes involved in LPS biosynthesis. Mutations in only four genes were cytotoxic to macrophages treated with cytD: kdtA (Kdo transferase) and the three identified during the screen for hypercytotoxic mutants (lpcC, manB and manC). All four of these genes are required for LPS core biosynthesis.

Results
One class of F. novicida mutants kills J774 cells faster than the parental strain U112 Like other intracellular pathogens, intracellular replication of Francisella eventually kills the host cell. Bacteria are released following host cell death, allowing them to infect other cells. F. novicida type strain U112 has been known to cause damage to J774 monolayers at 18 hours post infection [20]. By directly visualizing morphological changes of infected J774 macrophages, we screened an arrayed F. novicida library of more than 3000 mutants containing two allelic mutations of most nonessential genes [2]. Twenty-eight transposon mutants killed J774 cells within 12 hours of infection (Table 1) while the parental strain U112 required 24 hours to result in substantial cell death. The degree of cytotoxicity among the 28 mutants was determined by measuring the release of lactate dehydrogenase (LDH), a cytoplasmic protein commonly used to identify cell lysis. Twelve hours after infecting J774 macrophages with approximately the same number of bacteria, cell culture supernatants were removed and assayed for LDH. The 28 hypercytotoxic mutants killed J774 macrophage-like cells more quickly (within 12 hours post infection) in comparison to the wild-type strain U112 (Column 3 in Table 1). In vitro growth did not differ among the hypercytotoxic mutants (Column 5 in Table 1), nor did they display any other obvious phenotype. Among the genes identified, 12 were pil genes involved in pilus production and protein secretion and four were involved in LPS synthesis (htrB, lpcC, manB and manC).
Twelve F. novicida hypercytotoxic mutants are attenuated in mice The virulence of Francisella is connected to its ability to evade the immune response of a host. By increasing the rate of macrophage cell death, an infection with the hypercytotoxic mutants may induce a proinflammatory response and as a result, diminish the virulence of the strain. In order to determine whether any of the hypercytotoxic mutants are attenuated, BALB/c mice were infected with each of the 28 transposon mutant strains. Mice were inoculated intraperitoneally (i.p.) with 1 LD 50 (60 CFUs) and monitored for survival. After 28 days, the survival rate for 12 of the 28 hypercytotoxic mutants was 100% (Column 4 in Table 1), suggesting that they are attenuated for virulence. The remaining 16 had lower survival rates that were more similar to the survival rate of mice infected with U112. The transposon strain with a mutation in pilP (FTN1138) appeared hypervirulent, as it killed all three mice within two days. Typically, mice infected with the parental strain began dying three to four days after infection.
The LD 50 of the 12 mutant strains showing attenuation in the initial screen was determined by infecting mice with increasing doses of each mutant strain. All 12 of the mutants had a LD 50 that was at least one order of magnitude higher than the parental strain U112 ( Table 2). Mutants identified to be hypercytotoxic in vitro were more likely to be avirulent (12 of 28 strains, 43%) as compared to transposon insertion mutations screened at random for virulence in which 4-6% of insertion mutations were avirulent [21].

Role of cell contact in cytotoxicity
Francisella is less inflammatory than many pathogens in part because its LPS is not recognized by TLR4. An alternative, although less likely scenario, is that Francisella secretes proteins that function to inhibit cell death. For example, as hypothesized in Hager et al. [22], secreted proteases could remove cell surface proteins (e.g. Fas and TNFaR) that are involved in signaling the cell to undergo apoptosis. Therefore, to determine if cell contact between the bacteria and macrophages is necessary for cell death, we divided a transwell (Nunc 137044) with a membrane containing 0.22 mm pores (too small for bacterial passage); on one side we grew macrophages, and on the other side we inoculated bacteria. We did not observe any increase in cell death when macrophages were separated from the wild-type U112 or mutant bacteria, indicating that contact with bacteria was necessary for cell death.
Next, in order to determine whether internalization of the mutant Francisella strain is required for macrophage cell death, we inhibited actin polymerization by adding cytochalasin D (cytD) before and during infection with the 28 transposon mutants and U112 in J774 macrophages. CytD has been previously shown to be an inhibitor of phagocytosis of Francisella [14][15][16]23]. Three of the 28 transposon mutant strains (lpcC, manB and manC) were highly cytotoxic when infection took place in the presence of cytD ( Figure 2A). For the remaining mutants, the cell toxicity was greatly reduced following treatment with cytD ( Figure 2A). In order to observe differences in U112-infected macrophages, LDH release was quantified after 24 hours. As shown in Figure 2B, cytD significantly reduces the cell toxicity of the parent strain 24 hours p.i. Internalization of three mutant strains occurs independently of actin polymerization The three transposon mutant strains that remained hypercytotoxic in the presence of cytD contained transposon insertions in genes necessary for biosynthesis of LPS core (lpcC, manB, and manC). Transposon insertions can have polar effects on downstream genes, and manB and manC are within an operon; therefore, we constructed deletions of lpcC, manB, and manC in the U112based restriction-deficient strain MFN245 (see Materials & Methods). In this background, we confirmed that individual deletions of these genes still result in a hypercytotoxic phenotype that is independent of actin polymerization. LDH release from J774 macrophages infected with the deletions of FTN1253 (DlpcC), FTN1417 (DmanB), and FTN1418 (DmanC) with a multiplicity of infection (MOI) of 100 or 1000 was significantly (p,0.01) higher than macrophages infected with the same MOI of the parent strain and similar to the LDH release observed after infection with the transposon mutants (see below). LDH release from macrophages infected with MFN245 was comparable to LDH release from macrophages infected with wild-type U112. Furthermore, the presence of cytD did not affect LDH release in J774 macrophages infected with DlpcC, DmanB, and DmanC (data not shown). To confirm that the hypercytotoxicity was not specific to a macrophage cell line, we repeated the experiment in bone marrow derived macrophages prepared from BALB/c mice. As shown in Figure 3, DlpcC, DmanB, and DmanC were more toxic to primary macrophages than parent strain MFN245.
Because the hypercytotoxicity of the lpcC, manB, and manC mutant strains is not dependent upon actin polymerization, we wanted to determine if the bacteria were being internalized by the macrophages. Bacteria were visualized in macrophages infected with either one of the mutant strains or MFN245 using a fluorescently conjugated anti-Francisella antibody. To our surprise, we observed approximately one hundred times more DlpcC, DmanB, and DmanC bacteria in macrophages two hours after infection as compared to the parent strain ( Figure 4A). In the presence of cytD, very few MFN245 bacteria were visualized internally. In contrast, a similar number of mutant bacteria were visualized inside macrophages in the presence of cytD as were visualized in the absence of cytD. ( Figure 4B). Z sections prepared using the API Deltavision deconvolution microscope confirmed that the bacteria were inside the cell and not simply associated with the cell externally ( Figure 4). Increasing the concentration of cytD to 50 mm did not inhibit the entry or cytotoxicity of the DlpcC, DmanB, and DmanC bacteria (data not shown). The localization of wild-type U112 bacteria in infected macrophages was indistinguishable from MFN245 under similar conditions (data not shown).
As a way to quantitate the above results, we determined the number of intracellular bacteria in J774 macrophages after infection with DlpcC, DmanB, DmanC and MFN245 for two hours in the presence or absence of cytD. After two hours, the macrophages were washed and treated with gentamicin to remove extracellular bacteria and then lysed with saponin to determine the number of intracellular bacteria. As shown in Figure 5, the addition of cytD significantly lowered the uptake of the parent (p,0.01) but not the number of internal mutant bacteria ( Figure 5). These data combined with the microscopy above suggest that some bacteria may be killed upon phagocytosis. Furthermore, at 8 hours p.i., the number of intracellular MFN245 bacteria was greater than the number of DmanB and DmanC (data not shown) suggesting that either the mutant strains grow more slowly than the wild-type strain or they are being killed inside macrophages.

Increased cell uptake alone does not result in hypercytotoxicity
We reasoned that the hypercytotoxic phenotype of DlpcC, DmanB, and DmanC might be solely due to the increased number of bacteria internalized by the macrophages. To test this hypothesis we varied the multiplicity of infection (MOI) for the wild-type parent and mutant strains and assayed for LDH release. Increasing the MOI of MFN245 did not increase the LDH released from infected J774 macrophages 10 hours p.i. ( Figure 6A). LDH released from macrophages infected with the mutants increased with increasing MOI ( Figure 6A) and was always higher than the LDH released following MFN245 infection at the same MOI. The increased LDH release in the mutants was abolished when the wild-type gene was expressed in trans indicating that the phenotype is specific to the deleted gene. The number of  . High numbers of mutant bacteria were visualized intracellularly in infected J774 macrophages even in the absence of actin polymerization. J774 macrophages were infected with the three deletion mutants or parental strain MFN245 in four-well microscope chambers for two hours at an MOI of 100 either in the absence (A) or presence (B) of cytD. The cells were fixed in 4% paraformaldehyde, permeabilized, and probed with a rabbit polyclonal antibody against Francisella followed by a secondary goat anti-rabbit antibody conjugated with Alexa 488 (green). J774 nuclei were identified by staining DNA with DAPI (blue). Cells were imaged with an Applied Precision DeltaVision deconvolution microscope system. Experiment was repeated six times with similar results and representative images are shown. Eukaryotic cell boundary can be observed in the phase-contrast images of the same fields. Scale bar 10 mm (lower left corner). X-Z stack images show that bacteria were within cells. doi:10.1371/journal.pone.0011857.g004 parent bacteria internalized at an input MOI of 10,000 was confirmed visually to be comparable to the mutant strains infected at a MOI of 100 ( Figure 6B vs. Figure 4) and internalization of MFN245 at 10,000 MOI remained dependent upon actin polymerization ( Figure 6B). These data indicate that the parental strain is remarkably non-toxic to cells even if there are very high numbers of intracellular bacteria.

A bacterial surface structure is sufficient for increased macrophage invasions
In the case of Salmonella and many other intracellular pathogens, cell invasion requires synthesis of specific proteins. To determine if the exposed LPS core polysaccharide or some other exposed structure on the surface of the bacteria was sufficient to promote uptake, we determined whether bacteria treated with 4% formaldehyde, a cross-linking reagent, were internalized by J774 macrophages. Internalization of the mutant strains was compared to the parent strain MFN245 at the same input MOI by direct microscopic observations using polyclonal anti-Francisella antibodies to visualize the bacteria. As shown in Figure 7A, the number of internalized lpcC mutant bacteria is at least 10-fold higher than the parent at all concentrations of fixed bacteria. Similar results were observed for DmanB and DmanC strains (data not shown). These results indicate that the mutant bacteria possess a structure that promotes their uptake by macrophages, which is not exposed on the parent strain. Even though the dead parent and mutant strains were internalized by the macrophages, LDH release was low for all macrophages ( Figure 7B). These results demonstrate that viable Francisella bacteria are necessary to promote macrophage cell death.

Internalization of live Francisella bacteria is essential for macrophage killing
Because fixed mutant strains failed to induce macrophage cell death, we wanted to determine how long the intracellular bacteria had to be viable for cell death to occur. Ciprofloxacin, a cell permeable bactericidal antibiotic, was added at different times following infection of J774 cells with DlpcC, DmanB, or DmanC strains. The macrophages were infected with the bacterial strains for two hours and then washed and treated with gentamicin to remove extracellular bacteria. Ciprofloxacin was added to infected macrophages at six separate time points after the initial treatment with gentamicin (hr 0) and then LDH release was measured 10 hours later (12 hours after initial infection). Without addition of ciprofloxacin, the DlpcC, DmanB, and DmanC bacteria killed .95% of J774 within 12 hours of infection at an input MOI of 100 ( Figure 8). LDH released from cells infected with the parent at the same input MOI showed ,10% LDH release at 12 hours p.i. (data not shown). The addition of ciprofloxacin simultaneously with gentamicin reduced LDH release by 90%. Ciprofloxacin added at one or two hours postgentamicin treatment reduced LDH release by 80% or 70%, respectively ( Figure 8). These results show that four to five hours after infection with DlpcC, DmanB, or DmanC, the majority of J774 macrophages are committed to cell death.
lpcC, manB, and manC mutants have a shortened LPS structure Because lpcC, manB, and manC may be involved in LPS synthesis, we decided to analyze whether deleting these genes alters the LPS. The structure of the LPS can be partially deduced by its size following electrophoresis because it is assembled sequentially and as a consequence, mutants missing the core or O-antigens will migrate faster. The size of LPS synthesized in the deletion mutants was compared to wild-type strain U112, a strain that synthesizes a LPS without the core and O-antigen (DkdtA), two strains that synthesize a LPS lacking the O-antigen (DwbtA, wzx), and hypercytoxic transposon mutant htrB, which is predicted to alter the acylation of lipid A [26]. As shown in Figure 9, lpcC, manB, and manC mutant strains expressed a shortened LPS that resembled the LPS from the DkdtA mutant strain (Figure 9). These strains lack the O-antigen and likely contain a defect in the core. The O-antigen is also absent in DwbtA and wzx but these two strains seem to synthesize a complete core. The core and O-antigen were present in both wild-type U112 and transposon mutated htrB. These data suggest that the LPS of the lpcC, manB, and manC deletion strains lack an O-antigen and have an absent or altered core. LPS is synthesized starting with lipid A, followed by the core, followed by the O-antigen; therefore, these three genes are involved in the synthesis of the core or of the O-antigens. As lpcC, manB, and manC are all involved in mannose biosynthesis, it is very likely that these enzymes are required for complete synthesis of the core (Figure 1). The exact changes in LPS structure after deleting lpcC, manB, and manC remain to be determined because altering LPS at one position may influence other modifications [24,25].

Other LPS biosynthesis mutants
In our initial screening of the two-allele library, the other transposon mutants of LPS biosynthesis genes behaved like U112 or were hypocytotoxic, as judged by direct microscopic observation. However, as shown above, the LPS structure of DlpcC, DmanB, and DmanC is similar to DkdtA and both DwbtA and wzx synthesize a LPS lacking an O-antigen; therefore, we determined the cytotoxicity of these mutants by assaying LDH release from infected macrophages 10 hours after infection. As shown in Figure 10A, DkdtA killed infected macrophages as quickly as the three LPS mutants, DlpcC, DmanB, and DmanC. In contrast, DwbtA and wzx had a similar release of LDH as the parental strain MFN245. When a full-length copy of the kdtA gene was expressed in trans in DkdtA, LDH released from infected macrophages was not significantly different from the LDH released from MFN245infected macrophages ( Figure 10A). KdtA was probably not identified in the initial library screen because of polar effects on expression of downstream genes.
To determine whether internalization of DwbtA, DkdtA, and wzx was independent of actin polymerization, LDH release, and the number of intracellular bacteria were determined in infected J774 macrophages in the presence of cytD. High LDH release was observed for DkdtA at 10 hours, which was unchanged in the presence of cytD ( Figure 10A). As shown by microscopy in Figure 10B, the number of intracellular mutant wzx and wbtA bacteria was similar to what was observed for MFN245 ( Figure 5A) and reduced by the presence of cytD. In contrast, the number of intracellular DkdtA bacteria was greater than wzx   Figure 10B). Furthermore, the number of internal DkdtA bacteria was not affected by the presence of cytD ( Figure 10B). These results show that modifications to the O-antigen, as observed with wzx and Dwbt, are not sufficient to result in a hypercytotoxic phenotype and that alterations to the core were required.

Discussion
Our results have identified four F. novicida genes involved in LPS core synthesis/assembly that have two surprising phenotypes when mutated. First, they are phagocytosed by macrophages at a much higher rate than the parent via a novel mechanism that does not require actin polymerization. Second, they are more toxic to macrophages than the parent. Phagocytosis of wild-type Francisella strains by primary murine macrophages and macrophage cell lines has previously been shown to require actin polymerization [14][15][16]. Our results represent one of the first examples of bacterial mutants that can enter macrophages without actin polymerization. However, as we discuss in more detail below, there is indirect evidence in other Gram-negative bacteria that mutations that alter the LPS may have phenotypes similar to the four mutants we identified in this manuscript. The increased phagocytosis observed in the Francisella mutants occurs even for bacteria that have been fixed with formaldehyde, suggesting that a Francisella surface structure(s) mediates uptake. In addition, formaldehyde-fixed Francisella treated with trypsin are also phagocytosed at a higher rate, suggesting that the bacterial ligand is not proteinacious (X.H. Lai and F. Heffron, unpub. data). While dead bacteria continue to be internalized at a high rate, the increased toxicity requires viable Francisella. Infected macrophages are not killed when they are treated with a bactericidal antibiotic immediately following phagocytosis of the mutant bacteria. In contrast, approximately half of the infected macrophages die if mutant bacteria remain alive four hours after infection. We assume infection with the mutant activates a macrophage cell death pathway, although which pathway is not yet known.
Several LPS mutants in addition to lpcC, manB, manC and kdtA were identified and studied in this work. A transposon insertion in htrB, which is 99% identical to lpxL and as such, is a late acylase that modifies lipid A [26], was found to be hypercytotoxic in vitro and attenuated in mice ( Table 1). We found that this mutant was phagocytosed at a rate similar to that of the parent and synthesizes a LPS containing an O-antigen. Mutations in two other genes that are required for the addition of the O-antigen to the LPS including wzx and DwbtA were attenuated in mice [9,X.H. Lai and F. Heffron unpub data] but were not more cytotoxic to macrophages nor were they internalized at a higher rate ( Figure 10B). In fact, none of the hypercytotoxic transposon mutants we identified through our screen were in genes required for O-antigen synthesis. The four mutants we identified as being more cytotoxic to macrophages and phagocytosed via an actin-independent pathway are all normally involved in synthesis of the LPS core. We speculate that the LPS structure remaining in these mutants alters the mechanism of infection; however, it is also possible that the shortened LPS exposes other surface structures that signal import. The uptake by macrophages may be mediated by a lectin receptor. One such lectin is dectin-1, which is present on macrophages and mediates resistance to fungal infections [27]. It binds 1,3 linked glucose oligomers, but it requires an 11-13mer oligosaccharide. A single 1,6 linked glucose is observed on the surface of lipid A in F. novicida but not other Francisella species.

LPS mutants in other bacteria are phagocytosed at a high rate
Although this is the first example of a LPS mutant in Francisella being internalized at a higher rate, enhanced host cell uptake of LPS mutants in other microbes has been shown. Wick et al. found that peritoneal macrophages phagocytose rough mutants of Salmonella typhimurium four to ten times better than the smooth wild-type strain [28]. A recent paper described increased invasion of and adhesion to HeLa and MDCK cells for Salmonella mutants entirely lacking O-antigen [29]. Rough mutants of different Brucella species, including those due to mutation of the manB ortholog, are about 10 times more invasive to Vero cells [30], monocytes [31] and J774 macrophages [32,33]. In the cases described above, we do not know whether the increased uptake was associated with increased cytotoxicity or whether the internalization was dependent upon actin polymerization. There are examples of microbes entering host cells in the absence of actin polymerization. Oelschlaeger et al. described entry into T24 bladder cells by some strains of Campylobacter jejuni and Citrobacter freundii that is independent of actin polymerization but requires microtubule assembly [34]. Schramm and Wyrick reported that disruption of microfilaments with cytochalasin D markedly reduces infection of host epithelial (HEC-1B) and fibroblast (McCoy) cells by Chlamydia trachomatis serovar E, but not serovar L2 [35]. Furthermore, in the presence of cytochalasin D, Actinobacillus actinomycetemcomitans strains SUNY 523 and 4065 exhibit enhanced ability to enter epithelial cells [36].

The Francisella hypercytotoxic mutants are heterogeneous
We identified 28 F. novicida transposon mutants that increased the rate of cell death of infected macrophages. Even though all the mutants were hypercytotoxic, they differed in their attenuation in mice, their LPS patterns, and the rate and mechanism of their uptake in macrophages. Given the differences in phenotype, the exact mechanism of increased cell killing likely differs between the mutant strains. Weiss et al. identified two genes in F. novicida strain U112, FTN0757 (FTT0584) and FTN0720 (FTT0748), that are hypercytotoxic when mutated not because of a change in bacterial replication but rather because the mutants are more proinflammatory than the parental strain [37]. Similarly, a very recent report described a mutation in tolC of F. tularensis subsp. holarctica that is hypercytotoxic to macrophages and enhances a proin-flammatory response from infected host cells [38]. In contrast, mutations in both FTN1592 (FTT0123, oppB oligopeptide permease) and FTN1186 (FTT1209c, pepO, endothelian converting enzyme) of F. novicida have been found to induce higher levels of cytotoxicity in macrophages because these mutants replicate to higher levels in macrophages than wild-type Francisella [18]. Among the mutants listed above only FTN0757 was identified in our screen because of its exceptionally strong hypercytotoxic phenotype. The cytotoxicity was not as robust in strains containing transposon mutations in FTN0720, FTN1209, FTN1592 and FTN1703; therefore, these strains were not pursued further.
The Francisella lpcC, manB, manC and kdtA mutants described in this study require viable bacteria for cell toxicity. Although these mutants were internalized at a high rate, this increase in initial internalization is not sufficient to explain the hypercytotoxic phenotype. The mutant strains are more cytotoxic as compared to the parental strain at all MOIs used for infection, and we did not observe an increase in macrophage cell death even when the number of intracellular parental bacteria was equivalent to the mutant strains. In addition, once internalized the number of mutant bacteria did not significantly increase eight hours p.i. (see Supporting Figure S1), which indicates that either the mutant bacteria replicate more slowly than the parental strain or are killed by the host cell at a rate that is approximately equal to their rate of replication. Therefore, the cytotoxicity of these strains is not the result of an increase in bacterial uptake or by an accelerated rate of replication. Based on previous studies, a possible explanation is that these mutants increase the inflammatory response by signaling through TLR4. However, primary macrophages prepared from TLR4 2 mice (C3H/HeJ mice) were killed at about the same or greater rate than macrophages from congenic TLR + mice (X.H. Lai and F. Heffron, unpubl. obs.). Other possible explanations that are being explored include measuring expression of different virulence factors after infection, comparing the phagocytic pathway of the mutants to the parent, and directly measuring the inflammatory pathways that are being stimulated.
The largest class of mutants identified in this study was that which affects type IV pili synthesis. Eleven of the thirteen F. novicida pil genes were found to be more cytotoxic to J774 cells than the wild-type U112 strain ( Table 1). The finding that some of the pil genes are involved in both pilus production and protein secretion may help explain these results. U112 and LVS have been shown to secret proteins into culture supernatant via the type IV pilus [39,40]. Among the pil genes, it has been shown that four mutants (pilA, pilF, pilG, and pilQ) reduced their protein secretion; mutations in pilF and pilQ resulted in fewer pili [39,40]; and pilA and pilF mutants were less virulent in mice [39,40]. It is possible that type IV pili are normally anti-inflammatory by themselves or alternatively, that Francisella secretes a factor via these structures that inhibits cell death. These possibilities are being tested by directly adding purified pilus system proteins to cells to determine if we can complement the cytotoxicity of a pil mutant, as well as by performing studies to determine if other secreted proteins could be anti-inflammatory.
The current study reports the finding of a class of hypercytotoxic mutants in F. novicida that are heterogeneous in their entry into macrophages, LPS structures and virulence in vivo. More extensive analyses of four of these mutants (lpcC, manB, manC and kdtA) showed that they were in genes necessary for synthesis of the LPS core. These F. novicida mutants utilize a highly efficient mechanism for entry into macrophages that does not involve eukaryotic microfilaments. The results suggest that Francisella has an alternative mechanism to enter macrophages, e.g. a vestigal mechanism left from the evolution of Gram-negative bacteria. The signal transduction pathways stimulated by the mutants vis-à-vis the parent, as well as the cognate cytoskeletal changes that take place, remain to be investigated. It is interesting that infection of lung alveolar type II (ATII) epithelial cells by Francisella is highly efficient and likely accounts for much of the pulmonary pathology [41,42]. These same authors observed that the invasion of a lung ATII type epithelial cell line (TC-1) took place even with formaldehyde fixed F. tularensis LVS bacteria [43]. During LPS synthesis, some incomplete molecules are always expressed (R. Ernst, pers. comm.), perhaps resulting in bacteria that are taken up by ATII cells by the same mechanism that we describe.

Culture and infection of J774 murine macrophage-like cells and murine bone marrow-derived macrophages (BMDM)
The J774 murine macrophage-like cells (American Type Culture Collection, Manassas, VA) were cultured in Ham's F10 medium (Gibco-BRL, Rockville, MD) supplemented with 10% fetal bovine serum (FBS) (Gibco-BRL), 1 mM nonessential amino acids (Gibco-BRL), and 0.2 mM sodium pyruvate (Gibco-BRL) at 37uC in the presence of 5% CO 2 . For infection, bacteria were added to 70% confluent cells in 6-, 24-, or 96-well culture dishes (Corning, Corning, NY) at the multiplicities of infection (MOI) indicated in the results section, and incubated at 37uC in the presence of 5% CO 2 . Two hours after infection, the cells were washed twice with phosphate-buffered saline (PBS), and cultured in Ham's F10 medium containing 10 mg/ml of gentamicin to prevent the growth of extracellular bacteria. For some experiments J774 cells were treated with 1 mg/ml cytochalasin D (Sigma, St. Louis, MO) 30 min prior to infection to inhibit actin filament polymerization.
BMDM were collected by flushing the femurs of BALB/c (TLR4+) mice with Dulbecco's modified Eagle's medium (DMEM, Invitrogen) and cultured in DMEM with 10% heatinactivated FBS, 30% sterile filtered L-cell conditioned media (made in house), and penicillin/streptomycin (10,000 U/ml each) for six to seven days. The cells were split and infected as above for J774 macrophages.

Lactate Dehydrogenase (LDH) release assay for cytopathogenicity
The LDH release assay was conducted as described [20]. Briefly, J774 macrophages (about 4-5610 4 /well) seeded in 96-well culture plates were infected in triplicate with the transposon mutants, the non-polar deletion mutants, the restriction-deficient strain MFN245, or wild-type F. novicida U112 at the designated input MOI and washed at two hours post-infection (p.i.). At 12 and 24 hours p.i., the supernatants were removed and assayed for release of LDH using the CytoTox 96 nonradioactive cytotoxicity assay (Promega, Madison, WI). Cytotoxicity was determined by calculating LDH release as a percentage of the maximal amount released from macrophages lysed with detergent.
Phase-contrast and fluorescent microscopy J774 cells were infected as previously described [20] at the indicated input MOI, in four-well chamber plates (Nalgene Nunc/ Thermo Scientific, Rochester, NY). After two hours, the cells were washed twice with PBS, fixed for one hour with 4% paraformaldehyde at 4uC. After three washes for 10 min in phosphate-buffered saline (PBS), the cells were permeabilized with 0.5% Triton X-100 (Sigma) in PBS for 20 min at room temperature, blocked with 5% FBS in PBS for 30 min, and incubated for one hour at 4uC with a polyclonal antibody against F. tularensis at a 1:2,000 dilution (Becton, Dickinson and Company). After three washes for 10 min in PBS, the cells were again blocked with 5% FBS. A goat anti-rabbit antibody conjugated to Alexa 488 (Molecular Probes, Eugene, OR) was applied to the cells at a 1:500 dilution for one hour at 4uC. The cells were again washed three times for 10 min in PBS and incubated with a 1:1,000 dilution of FM 4-64 membrane stain (Molecular Probes) and 1:1,000 dilution of DAPI DNA stain in PBS (Alexis Biochemicals, San Diego, CA) for 10 min at room temperature. The cells were washed twice with PBS and mounted in Fluormount-G antifade solution (Southern Biotechnology, Birmingham, AL), and images were obtained with an Applied Precision DeltaVision deconvolution microscope system (Advanced Precision Instruments, Issaquah, WA). All images were taken with a 606 objective. Stacks of 10 z-plane images that were 1 mm apart were captured at 102461024 pixels and deconvolved for seven iterations. Selected images were saved in TIFF format and imported into Adobe Photoshop to be formatted.

Mouse studies
Six-to eight-week old female BALB/c mice were purchased from the Jackson Laboratory (Bar Harbor, ME) and acclimatized for one week. The animals were fed autoclaved food and water ad libitum. All experiments were performed in accordance with Animal Care and Use Committee guidelines (Institutional Approval B11220). Mutant strains were cultured in TSBC, OD normalized, and diluted with PBS. Mice were inoculated intraperitoneally (i.p.) with bacteria in 150 ml (total volume) of PBS and checked for signs of illness or death daily following infection for a total of 28 days. The 50% lethal doses (LD 50 ) were calculated by the method of Reed and Muench [45].

in vitro growth index assay
Overnight cultures of the transposon mutants and wild-type U112 strains were diluted in fresh TSBC to a starting OD 600 of 0.05. At 18 hours post inoculation OD 600 was taken and an in vitro growth index was calculated as (mutant OD 18h 2 mutant OD 0h )/ (wt OD 18h 2 wt OD 0h ) to determine if the mutant shows an obvious difference in vitro growth as compared to wild-type [21]. A value of 630% (i.e., 60.3) for the mutant strain is considered to be significantly different than the wild-type.

Construction and complementation of deletions in F. novicida
All deletions were generated in F. novicida mutant strain MFN245, a quadruple mutant that substantially reduces the restriction barrier and thereby increases the efficiency of transformation [44]. The cytotoxicity and virulence of this modified host strain in mice is comparable to U112 (X.-H. Lai and F. Heffron, unpub. data). Plasmid pKD13 [46] was modified such that expression of kanamycin was more efficient in Francisella and segments of Francisella DNA would be easier to clone into the vector. First, the groE promoter was amplified from F. novicida genomic DNA by PCR using oligonucleotides 59 -CGCGGATCCGTATG-GATTAGTCGAGC and 59-CGCGGATCCTGCACGACGA-ACTAATACTC. The oligonucleotides also contained a recognition site for BamHI at the 59 end. The DNA fragment containing the groE promoter was digested with BamHI and ligated into the BglII site directly upstream of the kan gene in pKD13. A pKD13 plasmid with the groE promoter in the correct orientation and free of PCR errors was identified by sequence analysis. Second, complementary pairs of oligonucleotides were annealed to generate DNA fragments containing ApaI and SbfI restriction sites (59-TCGAGGGCCCG-CACCTGCAGGGC and 59-TCGACCTGCAGGTGCGGGCC-C) and AscI and SmaI restriction sites (59-CTAGAGGCGCG-CCGCCCCGGG and 59-CTAGCCCGGGGCGGCGCGCCT). After annealing, the complementary oligonucleotides had singlestrand 59 overhangs that allowed the fragments to be cloned into pKD13 digested with SalI and AvrII, respectively. In the final modified pKD13 plasmid, both the ApaI and SmaI restriction sites were most distal to the FRT sites. For each gene to be deleted, two sets of oligonucleotide pairs were designed ( Table 3). The first pair (labeled Up F and Up R) amplifies the first 50-70 bp of the open reading frame (ORF) plus ,500 bp upstream and includes recognition sites for ApaI and SbfI. The second pair (labeled Down F and Down R) amplifies the last 50-70 bp of the ORF plus ,500 bp downstream and includes recognition sites for AscI and SmaI.
For one-step allelic replacement, pKD13 containing the Francisella fragments was digested with ApaI and SmaI and transformed into MFN245 as described previously [47]. Kanamycin resistant transformants were streaked for single colonies and correct integration of only the linear fragment was verified by PCR. Although the gene deletions marked with kanamycin are inframe, we proceeded to remove the kanamycin resistance gene by transforming the temperature-sensitive plasmid pFFLP [44], which expresses the flippase recombination enzyme, into the kanamycin resistant colonies by electroporation as described in [48]. Briefly, a 10 ml culture in Mueller-Hinton (MH) broth containing 0.1% glucose, 0.025% ferric pyrophosphate, and 2% IsoVitaleX (Becton Dickinson) was inoculated to an OD of ,0.15 using overnight cultures of each of the deletion mutants. After approximately four hours of growth the cultures (OD between 0.3 to 0.5) were washed twice with 10-15 ml of 0.5 M sucrose (Fisher Scientific) and suspended in 200 ml of 0.5 M sucrose. One microliter pFFLP (500 ng/ml) was mixed with 200 ml of cells, incubated at room temperature for 10 min, transferred to a 0.2-mm cuvette and electroporated using a GenePulser (BioRad) at 2.5 kV, 25 mF, and 600V. Following electroporation, cells were suspended in 1 ml of MH broth and incubated at 30uC for two hours before plating on CHA plates containing 8 mg/ml tetracycline. As detailed in Gallagher LA et al. (2008), tetracyline resistant transformants were streaked for single colonies on tetracycline and grown at 30uC. Single colonies were transferred to CHA, grown at 37uC and tested for kanamycin sensitivity. Kanamycin sensitive colonies were streaked again for single colonies on CHA, grown at 37uC and tested for tetracycline sensitivity. The tetracycline sensitive colonies, which indicate loss of the pFFLP plasmid, were used in the experimental analyses. The resolved deletion was confirmed by PCR and sequencing. The gene deletions in DlpcC, DmanB, DmanC and DkdtA strains were complemented in trans by transforming into the mutant strains plasmids that express the wild-type gene. The promoter and ORF of lpcC (FTN1253), manC (FTN1418) and kdtA (FTN1469) were amplified by PCR from F. novicida genomic DNA using oligonucleotides complementary to ,500 bp upstream of the start codon and 100-300 bp downstream of the stop codon. manB (FTN1417) lies downstream of manC in an operon; in order to express manB from its endogenous promoter, we designed oligonucleotides that amplified 503 bp upstream of manC and 403 bp downstream of manB. The resulting FTN1417/18 PCR fragment contains the entire operon. The oligonucleotides used to amplify the wild-type copies of lpcC, manB, manC and kdtA are described in Table 3 and contain recognition sites for NotI and ApaI. DNA fragments digested with NotI and ApaI were cloned into similar sites into a modified pKK202 that contains unique restriction sites for NotI, XhoI and SfiI [20].

LPS gel analysis
The LPS from F. novicida wild-type and mutant strains was isolated from whole cells after growth in TSBC for 24 hours. Bacteria (1 ml) were pelleted, resuspended in 100 ml lysis buffer (187 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 4% 2mercaptoethanol, 0.03% bromophenol blue), heated to 100uC for 10 minutes, and cooled to room temperature. Subsequently, 25 mg proteinase K was added to each sample and incubated at 60uC for one hour. Finally, samples were incubated at 100uC for 10 minutes, cooled briefly on ice and subjected to sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) using Bio-Rad Ready Gel precast 10-20% gradient Tris-Tricine/ Peptide polyacrylamide gels (Hercules, CA). After electrophoresis, the LPS was stained with Pro-Q Emerald 300 LPS stain kits (Invitrogen, Carlsbad, CA), according to the manufacturer's recommendation, visualized, and photographed using the AlphaImager digital imaging system (Alpha Innotech Corp., San Leandro, CA).

Statistics
Statistical significance of data was determined by using an unpaired analysis of variance and the Tukey-Kramer multiplecomparisons test (GraphPad Prism 4, San Diego, CA). Figure S1 Mutant bacteria are internalized at higher levels than parent bacteria but do not replicate robustly inside host cells. RAW264.7 macrophages were infected with parent strain MFN245 or the DlpcC mutant strain at an input MOI of 100. One hour after infection, the cells were washed three times with PBS and incubated in DMEM containing 100 mg/ml gentimicin. One hour later, the cells were washed again and incubated in DMEM containing 10 mg/ml gentimicin. At two, four, or eight hours p.i., the macrophages were lysed in TSBC with 0.5% saponin, and intracellular bacteria were quantified by plating. Each infection was performed in triplicate and with a mockinfected control. MFN245 was less abundant intracellularly than DlpcC at two hours (p,0.001), four hours (p,0.05), and eight hours (p,0.05) p.i. The number of internalized DlpcC at two hours was not statistically significantly different from the level of internalized DlpcC observed at either four or eight hours p.i. Found at: doi:10.1371/journal.pone.0011857.s001 (0.68 MB TIF)