Lack of OxyR and KatG Results in Extreme Susceptibility of Francisella tularensis LVS to Oxidative Stress and Marked Attenuation In vivo

Francisella tularensis is an intracellular bacterium and as such is expected to encounter a continuous attack by reactive oxygen species (ROS) in its intracellular habitat and efficiently coping with oxidative stress is therefore essential for its survival. The oxidative stress response system of F. tularensis is complex and includes multiple antioxidant enzymes and pathways, including the transcriptional regulator OxyR and the H2O2-decomposing enzyme catalase, encoded by katG. The latter is regulated by OxyR. A deletion of either of these genes, however, does not severely compromise the virulence of F. tularensis and we hypothesized that if the bacterium would be deficient of both catalase and OxyR, then the oxidative defense and virulence of F. tularensis would become severely hampered. To test this hypothesis, we generated a double deletion mutant, ΔoxyR/ΔkatG, of F. tularensis LVS and compared its phenotype to the parental LVS strain and the corresponding single deletion mutants. In accordance with the hypothesis, ΔoxyR/ΔkatG was distinctly more susceptible than ΔoxyR and ΔkatG to H2O2, ONOO−, and O2-, moreover, it hardly grew in mouse-derived BMDM or in mice, whereas ΔkatG and ΔoxyR grew as well as F. tularensis LVS in BMDM and exhibited only slight attenuation in mice. Altogether, the results demonstrate the importance of catalase and OxyR for a robust oxidative stress defense system and that they act cooperatively. The lack of both functions render F. tularensis severely crippled to handle oxidative stress and also much attenuated for intracellular growth and virulence.


INTRODUCTION
Francisella tularensis, a Tier 1 select agent and the causative agent of tularemia, is a zoonotic, facultative intracellular bacterium with two clinically relevent subspecies, tularensis and holarctica, the former of which causes an aggressive disease with high mortality if left untreated (Oyston et al., 2004). Although there is no licensed vaccine against this potential bioterrorism agent, the subspecies holarctica live vaccine strain, LVS, is used to vaccinate laboratory workers, and is widely used in Francisella research as it is attenuated in humans, but retains its virulence in mice (Sjöstedt, 2006;Conlan, 2011).
Francisella tularensis is capable of infecting numerous cell types, including professional phagocytes, like macrophages. Upon phagocytosis, it transiently resides within the phagosome before escaping into the cytosol to replicate (Bröms et al., 2010;Chong and Celli, 2010). Phagocytes constitute a hostile environment utilizing a wide array of anti-bacterial mechanisms, such as phagosome acidification, disruption of pathogen membrane integrity, removal or sequestration of nutrients, and the production of reactive oxygen species (ROS) (Flannagan et al., 2009) and since F. tularensis is an intracellular bacterium, it will encounter a continuous exposure to ROS. Vital macromolecules, such as proteins and DNA, will react with ROS, thereby disrupting their functions (Fridovich, 1998;Schaible and Kaufmann, 2004;Flannagan et al., 2009). There are several ROS with potent antibacterial effects, such as superoxide and H 2 O 2 . The former is produced at high levels by the phagocyte oxidase (phox) and it rapidly combines with nitric oxide (NO), which is produced at high levels by inducible nitric oxide synthase (iNOS), to form peroxynitrite, a highly reactive compound. H 2 O 2 is toxic per se, but the damage it exerts can be exacerbated in combination with intracellular ferrous iron, resulting in the formation of hydroxyl radicals (HO • ) and hydroxide anions (OH − ) through the Fenton reaction.
Reactive oxygen species (ROS) are not only formed during host attack, but low levels are also formed as by-products of normal aerobic metabolism. Thus, pathogens, in particular intracellular pathogens, have a pressing need for defense mechanisms to combat the ever present levels of ROS, but even more so to combat the assault of ROS experienced within a host (Betteridge, 2000). The critical roles of ROS and NO for the host defense against tularemia are illustrated by the extreme susceptibility of phox-deficient and iNOS-deficient mice to an F. tularensis infection (Lindgren et al., 2004). Moreover, ex vivo, it has been demonstrated that the requirements for host protection vary depending on the cell type investigated, since killing of F. tularensis by mouse peritoneal cells is NO-dependent, but NOindependent by mouse pulmonary cells (Anthony et al., 1992;Polsinelli et al., 1994;Lindgren et al., 2005).
The oxidative stress defense system of Escherichia coli has been extensively studied and includes numerous detoxifying enzymes, such as catalase, superoxide dismutases (SODs), alkyl hydroperoxide reductase (Ahp), and the H 2 O 2 -activated transcriptional regulator OxyR. The latter combats the effect of H 2 O 2 by dual mechanisms, since it regulates the expression of both catalase and the ferric uptake regulator (Fur) (Farr and Kogoma, 1991;Zheng et al., 1998Zheng et al., , 1999Pomposiello and Demple, 2001). Catalase renders H 2 O 2 harmless by degrading it to oxygen and water, whilst Fur down-regulates the expression of genes involved in iron uptake, thus limiting the amount of iron with which H 2 O 2 can combine in the Fenton reaction (Andrews et al., 2003;Troxell and Hassan, 2013). Catalase, SODs, AhpC and other detoxifying enzymes are employed as oxidative stress defense mechanisms also by F. tularensis (Bakshi et al., 2006;Lindgren et al., 2007;Melillo et al., 2009;Binesse et al., 2015). The F. tularensis catalase, encoded by katG, mediates H 2 O 2 tolerance and is known to be important for the virulence of F. tularensis LVS (Lindgren et al., 2007). SodB, FeSOD, and SodC, CuZnSOD, are both known to be important for the dismutation of O − 2 in F. tularensis, and SodB further acts in the defense against oxidative stress by harnessing iron (Bakshi et al., 2006;Melillo et al., 2009). The F. tularensis AhpC enzyme is important for the detoxification of O − 2 and peroxynitrite (ONOO − ), but not of H 2 O 2 , in the highly virulent SCHU S4 strain (Binesse et al., 2015), but the importance in the LVS strain is yet unknown. F. tularensis also encodes an oxyR homolog, the role of which has been studied recently (Ma et al., 2016). It was found that the absence of OxyR rendered LVS defective for oxidative stress defense, growth in macrophages and epithelial cells, and virulence in mice. Moreover, it was demonstrated that OxyR regulates the expression of the ahpC, katG, and sodB genes, with the most pronounced regulatory effect exerted on ahpC.
A more thorough understanding of the F. tularensis antioxidant system will undoubtedly reveal virulence mechanisms of this bacterium, since ROS constitute such an essential threat to the pathogen. As aforementioned, antioxidant enzymes, such as catalase, AhpC, SodC, and SodB, all contribute to the virulence of F. tularensis in mice, although each appears to render the bacterium only moderately attenuated and this indicates that the antioxidant system of F. tularensis is complex and may in part possess overlapping functions (Lindgren et al., 2007;Ma et al., 2016). Indeed, a double deletion mutant of katG and ahpC has not been possible to generate in F. tularensis (Binesse et al., 2015) and this demonstrates that the cooperative functions of these enzymes are crucial, although either one is not essential. The aim of the present study was to better understand this interconnecting web of antioxidants in F. tularensis. To this end, a double deletion mutant, oxyR/ katG, was generated since this mutant, besides lack of catalase activity, should have a repressed expression of OxyR-regulated antioxidant genes, one of which is AhpC (Ma et al., 2016). We hypothesized that the lack of both KatG and OxyR would lead to a severely impaired phenotype of F. tularensis LVS. We therefore characterized the phenotypes of single deletion mutants, oxyR and katG, and a double deletion mutant, oxyR/ katG, in comparison to the parental LVS strain.

Bacterial Strains
The F. tularensis LVS strain was obtained from the Francisella strain collection (FSC) at FOI, Swedish Defense Research Agency. The katG deletion mutant ( katG) has been described previously (Lindgren et al., 2007).
The oxyR and oxyR/ katG mutants of the LVS strain were generated by allelic replacement as described previously (Golovliov et al., 2003). Briefly, sequences up-and down-stream of oxyR were amplified by PCR. The fragments contained complementary sequences, which were joined together by a second PCR. The resulting fragment was cloned into the pDM4 suicide-vector, which was transformed into Escherichia coli S17λpir and thereafter transferred to LVS by conjugation. Clones with a successful recombination event were selected on plates supplemented with Cm and polymyxin B. Correct integration was confirmed by PCR. Positive clones were subjected to sucrose selection to select for a second recombination event and clones were screened by PCR to identify successful deletion mutants. The double deletion mutant oxyR/ katG was generated using the same procedure, apart from using the pDMK3 plasmid carrying kanamycin resistance. The deletions were verified by sequencing 1500 bp on each side of the deleted region.

Aerobic and Microaerobic Growth
Bacteria were cultivated overnight on plates based on modified GC-agar (MC plates) and then inoculated to an OD 600 of 0.1 in Chamberlain's chemically defined medium (CDM). All cultures were split into triplicates and were incubated at 37 • C and 200 rpm in an aerobic (normal air) or a microaerobic (10% O 2 and 10% CO 2 ) milieu up to 48 h with monitoring of the OD 600 .

H 2 O 2 Susceptibility Assay
Bacteria were cultivated overnight on MC plates, inoculated to an OD 600 of 0.1 in CDM and H 2 O 2 was added to the final concentration of 0.02, 0.1, or 0.5 mM, respectively. Controls were grown without the addition of H 2 O 2 . All cultures were split into triplicates and were incubated at 37 • C and 200 rpm up to 24 h with monitoring of the OD 600 .

Catalase Activity Assay
Catalase degrades H 2 O 2 to O 2 and H 2 O. H 2 O 2 absorbs light at 240 nm and degradation of H 2 O 2 can therefore be measured as a reduction of A 240 nm over time.
Strains were cultivated overnight after being diluted to an OD 600 of 0.1 in CDM. For each strain, one set of tubes were left untreated and another set of tubes were supplemented with H 2 O 2 to a final concentration of 0.02, 0.1, or 0.2 mM. All cultures were split into triplicates and incubated at 37 • C, 200 rpm for 2, 4, and 24 h before sampling for evaluation of catalase activity. Depending on the density and growth phase of the culture, a volume of 10-50 µl were withdrawn and diluted in PBS to reach a final volume of 120 µl in UV-clear 96-well plates (Greiner Bioone, Frickenhausen, Germany). Then, 80 µl 100 mM H 2 O 2 in PBS was added to each sample immediately before placing the plate in a Tecan Infinite 200 pro plate reader and measuring the reduction in absorption at 240 nm for 10 min. A molar extinction coefficient of H 2 O 2 at 240 nm of 43.6 M −1 cm −1 was used to calculated the concentration of H 2 O 2 using the Beer-Lambert law, A = εcl. One unit of catalase is defined as the amount that decomposes 1 µmol of H 2 O 2 per minute per OD 600 at 25 • C. The catalase units were normalized against the OD of the culture.

Paraquat Susceptibility Assay
Susceptibility of F. tularensis strains to O − 2 was determined by use of the O − 2 generating compound paraquat dichloride hydrate (Sigma-Aldrich, St. Louis, USA) in a disc diffusion assay. Paraquat generates O − 2 through reacting with parts of the respiratory chain in bacteria, causing the reduction of O 2 to O − 2 (Hassan and Fridovich, 1979). Bacterial strains were cultivated on MC plates overnight, re-suspended in phosphate-buffered saline (PBS) and approximately 3 × 10 5 CFU were plated onto MC plates. Sterile filter discs (Oxoid Blank Antimicrobial Susceptibility Discs, Thermo Scientific, MA, USA) were placed in the center the plates once they had dried, and 10 µl of MQwater, 1.25 mM, 5 mM or 20 mM paraquat solution was added to each disc. The plates were incubated for 4 days at 37 • C, 5% CO 2 before the size of the growth inhibition zone surrounding each disc was determined.
Strains were cultivated in CDM to logarithmic growth phase and diluted to a density of approximately 2 × 10 6 bacteria/ml in PBS. The bacterial suspensions were incubated with or without the addition of 0.48 mM SIN-1 with equal amounts of SIN-1 added at the start of the experiment and again after 1.5 h to ensure stable levels of ONOO − (Lindgren et al., 2005). After 3 h samples were collected, diluted and plated on MC plates for determination of viable bacteria.

Analysis of Gene Expression by Real Time PCR
Bacteria were cultivated overnight on MC plates, inoculated to an OD 600 of 0.1 in CDM and incubated at 37 • C, 5% CO 2 for 10 h before sampling. RNA extraction, cDNA synthesis and Real Time PCR (RT-PCR) were all performed as described previously (Honn et al., 2012).
Briefly, RNA was extracted using Trizol reagent (Invitrogen, CA, USA) from pelleted bacteria, 3 × 10 9 CFU/sample. Contaminating DNA was removed using the DNA-free kit (Ambion, Inc, Austin, TX, USA) and RNA was quantified by Nanodrop (Thermo Fisher Scientific, Wilmington, DE, USA). cDNA was synthesized from 1 µg RNA/sample using iScript (BioRad, Hemel, Hampstead, UK), RT-PCR was performed using the Power SYBR green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) and the ABI Prism 7900Ht Sequence Detection System (Applied Biosystems) as described (Honn et al., 2012). Trizol, DNA-free, iScript and Power SYBR green were all used in accordance with the instructions provided by the manufacturers. Forward and reverse primers were obtained from Invitrogen and have been published previously for fslA (FTL_1832), fslB (FTL_1833), (Lindgren et al., 2009) (Bröms et al., 2009), mglA (FTL_0260), feoB (FTL_0133), and katG (FTL_1504) (Honn et al., 2012) The Ct values of the selected genes were normalized to the Ct value of the house keeping gene FTT0901 (lpnA) and relative copy numbers (RCN) were calculated according to the following equation: RCN = 2 − Ct × 100, where Ct is Ct(target)−Ct(FTT0901) (Gavrilin et al., 2006). Thus, the copy number of a given gene is related to the copy number of FTT0901. Normalized Ct values were used for statistical evaluation of the data by One way ANOVA followed by Tukey's honest significant difference (HSD).

Preparation and Infection of BMDM
The capacity of LVS and the mutants to proliferate intracellularly were assessed in bone marrow-derived macrophages (BMDMs). BMDMs were generated from C57BL/6 mice essentially as described previously (Bröms et al., 2011).
The day before infection, BMDM cells were seeded at a density of 4 × 10 5 cells/ml in 24-well tissue-culture plates and incubated at 37 • C, 5% CO 2 with or without murine recombinant 1000 U/ml of IFN-γ (Peprotech, Rocky Hill, NJ, USA) The next day, the cells were washed and reconstituted with fresh, prewarmed culture media. Bacteria were grown overnight on MC plates and re-suspended in PBS to a density of approximately 3 × 10 9 bacteria/ml. Bacteria were diluted in DMEM and added to each well at multiplicity of infection of 30 and bacterial uptake was allowed to occur for 90 min at 37 • C, 5% CO 2 . Remaining extracellular bacteria were removed by rinsing the monolayers three times with DMEM and incubating with gentamicin for 45 min followed by rinsing the monolayers three times. This time-point was defined as 0 h. After 0, 4 and 24 h incubation the macrophages were lysed in 0.1% deoxycholate in PBS. The lysate were serially diluted in PBS and plated on MC plates for determination of viable bacteria.

Mouse Experiments
Virulence of the mutant strains was determined by subcutaneous infection of female C57BL/6 mice with 4 × 10 3 CFU/mouse of LVS, oxyR, katG, and oxyR/ katG. Mice were monitored for signs of illness and were euthanized by inhalation of isoflurane followed by CO 2 asphyxiation after 3 or 6 days, whereupon the number of viable bacteria in spleens and livers were determined by homogenizing the organs in PBS and plating dilutions on MC plates. All animal experiments were approved by the Local Ethical Committee on Laboratory Animals, Umeå, Sweden (no. A 1-09, A 99-11, and A 67-14).

Statistical Analysis
One way ANOVA followed by Tukey's HSD test was used to determine statistical significant difference between groups.

Growth under Aerobic vs. Microaerobic Conditions
CDM effectively supports growth of LVS. We therefore compared growth of the bacterial strains, LVS, oxyR, katG, and oxyR/ katG. The former three strains all replicated to the same extent, whereas oxyR/ katG showed intact growth to late log phase, but impaired growth thereafter. Therefore, it did not reach as high densities as LVS and the other strains at 24 h (P < 0.001; Figure 1A). To explore if a reduced oxygen tension could rescue the growth of oxyR/ katG, the strains were cultivated under microaerobic conditions, i.e., 10% O 2 and 10% CO 2 . Indeed, oxyR/ katG grew as well as the other strains and reached an optical density of > 2.0 within 48 h ( Figure 1A). As noted before (Honn et al., 2012), the growth rate of LVS under microaerobic conditions was reduced compared to aerobic conditions ( Figure 1A).

Catalase Activity under Aerobic vs. Microaerobic Conditions
The results so far suggested that LVS experienced oxidative stress during growth in an aerobic environment and to handle this stress, required either the function of catalase, or the expression of OxyR-regulated detoxifying mechanisms. OxyR is known to respond to oxidative stress by inducing antioxidant enzymes, such as catalase. As an indicator of oxidative stress and to investigate if catalase is under the regulation of oxyR in LVS, we measured the activity of the enzyme during growth of the bacteria in CDM. The catalase activity in LVS gradually increased during the two to 24 h period, whereas the catalase activity in oxyR was sustained at a constant, but lower level compared to LVS from two to six h (P < 0.05 at 2 and 4 h and P < 0.001 at 6 h; Figure 1B). However, the catalase activity of the two strains was similar at 24 h ( Figure 1B). In the microaerobic environment, the catalase activity of LVS and oxyR was similar, but for both lower than in the aerobic environment ( Figure 1B). The H 2 O 2 decomposition in samples containing katG or oxyR/ katG was below 1 µmol, regardless of growth condition and time point, indicating the absence of catalase activity ( Figure 1B).
In summary, oxyR demonstrated a basal catalase activity, but did not induce this activity further during the aerobic logarithmic growth phase as LVS did. oxyR/ katG, which lacks this basal catalase activity, failed to grow to high densities under the aerobic condition, but grew as well as LVS in the microaerobic milieu.
H 2 O 2 Tolerance oxyR and katG grew as well as LVS in CDM despite the reduced, or lack of catalase activity (Figure 2A). To investigate their adaptation to stress, H 2 O 2 , the substrate of catalase, was added to the cultures. Growth of LVS or oxyR was not affected by 0.02 mM H 2 O 2 , whereas, initially, the growth rate of katG was reduced (P < 0.01) and growth of oxyR/ katG almost completely inhibited (P < 0.001; Figure 2B). At 0.1 mM of H 2 O 2 , LVS and oxyR still grew rapidly, in contrast to katG and oxyR/ katG that did not grow at all (P < 0.001; Figure 2C). Growth of oxyR was significantly reduced in the presence of 0.5 mM of H 2 O 2 compared to LVS (P < 0.001; Figure 2D). Exposure of the strains to H 2 O 2 did not significantly change their catalase activity (data not shown).
In summary, the mutant strains displayed increased susceptibility to H 2 O 2 as compared to LVS, with the effect being most pronounced for oxyR/ katG, followed by katG, and the least affected strain being oxyR.

Susceptibility to Paraquat-Mediated Killing
O − 2 is continuously generated as a by-product of the respiratory chain during growth of bacteria. To investigate the capacity of the bacteria to defend against such ROS, LVS, oxyR, katG, and oxyR/ katG were exposed to paraquat in a disc diffusion assay (Figure 3). Paraquat dichloride hydrate generates O − 2 through a reaction with parts of the respiratory chain in bacteria, causing the reduction of O 2 to O − 2 (Hassan and Fridovich, 1979). oxyR displayed a significantly larger zone of inhibition than did LVS in the presence of 1.25 and 5 mM paraquat (P < 0.001 and 0.01, respectively), but the zones were similar when exposed to 20 mM (Figure 3). The zone of inhibition for katG was larger compared to LVS at 1.25 mM (P < 0.05), but similar at the two higher concentrations (Figure 3). A significantly larger zone of inhibition was observed for oxyR/ katG vs. LVS and katG at The results shown illustrate one representative experiment of at least three performed. Each value represents the average for triplicate samples and error bars represent the SEM. *P < 0.05, **P < 0.01, ***P < 0.001 vs. LVS.
Frontiers in Cellular and Infection Microbiology | www.frontiersin.org FIGURE 3 | F. tularensis strains were exposed to the O − 2 -generating compound paraquat in a disc diffusion assay. Each bar represents the average from three separate experiments with triplicate samples in each and error bars represent the SEM. *P < 0.05, **P < 0.01, ***P < 0.001 vs. LVS for each concentration.
all three concentrations of paraquat (P < 0.001 for 1.25 and 5 mM and P < 0.01 for 20 mM) and also larger compared to oxyR at 1.25 and 5 mM (P < 0.01; Figure 3).
In summary, the results demonstrated that oxyR and oxyR/ katG were more susceptible to paraquat-mediated killing compared to LVS, with oxyR/ katG being the most susceptible, whereas katG was only slightly more susceptible than LVS.

Susceptibility to SIN-1-Mediated Killing
Peroxynitrite (ONOO − ) is a highly reactive and bactericidal ROS formed through the reaction between (NO) and O − 2 and it is active against F. tularensis in activated macrophages (Lindgren et al., 2005). Experimentally, SIN-1 can be used to mimic a continuous exposure to ONOO − . SIN-1 slowly decomposes, thereby releasing both NO and O − 2 that combine to form ONOO − , which quickly is internalized since it passes through lipid bilayers (Hogg et al., 1992;Murphy et al., 1998).
The exposure to 0.48 mM SIN-1 for 3 h reduced the viability of all strains in comparison to un-treated cultures (P < 0.001 for all strains), but affected the mutant strains to a greater extent compared to LVS (P < 0.001 vs. LVS for all; Figure 4). The viability of LVS decreased approximately 0.8 log 10, of oxyR 2.8 log 10 , of katG 3.0 log 10 , and of oxyR/ katG 4.6 log 10 CFU. The latter was significantly more susceptible than any of the other strains (P < 0.001; Figure 4).
In summary, all mutant strains displayed increased susceptibility to ONOO − as compared to LVS, with the effect being similar for oxyR and katG and most pronounced for oxyR/ katG.

Gene Expression
oxyR/ katG did not grow after the late logarithmic growth phase (Figure 1A), and we therefore found it of interest to explore the gene expression of the strains at 10 h, i.e., during the late logarithmic growth phase. The analysis was focused on genes expressing proteins influencing the oxidative stress response of FIGURE 4 | F. tularensis strains were exposed to the peroxynitrite generating compound SIN-1 for 3 h. After 1.5 h of incubation, additional SIN-1 was added to the tubes to ensure a constant generation of peroxynitrite during the whole incubation period. Each bar represents the average from three separate experiments with triplicate samples in each and error bars represent the SEM. ***P < 0.001 vs. LVS.
the bacterium, such as antioxidant enzymes, chaperones and iron-related proteins. Genes found to be differentially expressed vs. LVS are shown in Figure 5. Of all genes examined, ahpC was the sole gene significantly repressed in oxyR (P < 0.001; Figure 5A). A similar degree of repression, about 3-fold, was observed in oxyR/ katG, which in addition, had a 1.5 to 2-fold increased expression of sodB, sodC and FTT0086 (P < 0.001 for all genes; Figure 5A). ahpC was not repressed in katG and as expected, katG transcripts were not detected in either katG or in oxyR/ katG ( Figure 5A). All chaperone genes examined were upregulated 1.6 to 2.5-fold in oxyR and katG (P < 0.001 for all genes; Figure 5B). In contrast, these genes, except for clpB, were suppressed 2.4 to 3.1-fold in oxyR/ katG (P < 0.001 for all genes).
fslA, the first gene of the siderophore operon, was slightly upregulated in oxyR and katG, although only about 1.2-fold, whereas the other iron-related genes were expressed at similar levels as in LVS. In contrast, fslA was suppressed 1.8-fold in oxyR/ katG and fslE, fslF and feoB were upregulated 2.5 to 2.9-fold (P < 0.001; Figure 5C).
In summary, the absence of OxyR resulted in a suppressed expression of ahpC and an up-regulated expression of genes encoding chaperone proteins. Except for ahpC, the expression profile of katG was similar to oxyR. In contrast, loss of both oxyR and katG changed the expression profile and low expression of chaperone-encoding genes was observed in oxyR/ katG, together with high expression of antioxidant genes, except for ahpC and katG, and an altered expression of genes related to iron-uptake.

Intracellular Replication in BMDM
Based on the increased susceptibility to various ROS displayed by oxyR, katG, and oxyR/ katG, it was of interest to test whether the strains were defective for replication in professional phagocytes. Non-stimulated or IFN-γ-stimulated BMDMs were infected with LVS, oxyR, katG, or oxyR/ katG at an MOI of 30, and the viability of internalized bacteria was determined after 0 h, 4 h, and 24 h. In non-stimulated BMDM, LVS grew from approximately 2.5 log 10 CFU to more than 5.0 log 10 CFU within 24 h and also oxyR and katG grew to similar extent ( Figure 6A). oxyR/ katG grew in non-stimulated cells, but reached approximately 10-fold lower numbers compared to the other strains after 24 h (P < 0.001; Figure 6A).
IFN-γ-stimulation of BMDM prior to infection reduced the numbers of LVS, katG, and oxyR about 10-fold at 24 h vs. the numbers in non-stimulated cultures (P < 0.001; Figures 6A,B). There was no growth of oxyR/ katG in IFN-γ-stimulated cultures and, thus, significantly lower bacterial numbers compared to non-stimulated cultures at 24 h (P < 0.001; Figures 6A,B) and vs. all the other strains exposed to IFN-γ (P < 0.001; Figure 6B).
Thus, the oxyR and katG mutants showed intact capacity of intracellular replication, whereas the oxyR/ katG mutant showed impaired replication in BMDM, both in the presence and absence of IFN-γ.

Virulence in Mice
The virulence of LVS, oxyR, katG, and oxyR/ katG was determined by subcutaneous infection of C57BL/6 mice with 4 × 10 3 CFU/mouse, a non-lethal dose, and enumeration of viable bacteria in spleen and liver on day 3 and 6 of infection. Compared to LVS, there were lower numbers of both oxyR and katG on day 3 in the liver of the mice (P < 0.05; Figure 7A), whereas there were no differences between these strains in either the liver or spleen at the other time points (Figures 7A,B). Numbers of oxyR/ katG in both organs were at least 100fold lower vs. all other strains at both time points (P < 0.001). Thus, both oxyR and katG showed slight attenuation in mice, whereas oxyR/ katG was highly attenuated.

DISCUSSION
Francisella tularensis is a versatile bacterium capable of surviving in many different hosts, vectors and in various cell types, including the normally bactericidal macrophages. Upon phagocytosis, F. tularensis is encased in a phagosome, a membrane-bound compartment designed for the annihilation of phagocytosed microbes, which is rich in antimicrobial molecules, such as reactive oxygen and nitrogen species. Although F. tularensis only transiently resides in this compartment, it must still muster defenses against highly reactive species in order to survive and escape to the cytosol, where it proceeds to replicate. By entering the cytosol, F. tularensis gains access to a nutrient-rich, protected niche in which it multiplies. As survival and replication in the intracellular niche is essential for the life cycle of F. tularensis, a thorough understanding of how the bacterium survives intracellularly is essential to fully grasp its defense mechanisms against oxidative stress. To this end, the study focused on understanding the interplay between catalase and OxyR, the latter being important for the expression of several antioxidant enzymes, in the defense against ROS and their impact on the survival of the bacterium in professional phagocytes.
To investigate if OxyR is involved in the oxidative stress response of LVS, we constructed an in-frame deletion of oxyR. A similar investigation has been performed recently by Ma et al. which studied the role of OxyR in LVS (Ma et al., 2016).
It was found that OxyR controlled transcription of katG and the findings agree with the reduced catalase activity of oxyR observed in the present study. Nevertheless, our study revealed that even in the absence of OxyR, there was still prominent catalase activity. Overall, it appears that OxyR, as expected, regulates katG in the LVS strain, however, the regulation does not completely abolish its expression as is the case observed for various other bacterial species, e.g., E. coli (Michán et al., 1999), Salmonella enterica (Morgan et al., 1986), Haemophilus influenza (Whitby et al., 2012), or Moraxella catarrhalis (Hoopman et al., 2011). In both the present study, and in the previous study, it was observed that the lack of OxyR led to marked suppression of ahpC2 (Ma et al., 2016). In addition Ma et al. demonstrated suppressed expression of both katG and sodB in oxyR by realtime PCR and demonstrated that OxyR binds to the upstream promoter regions of each gene. In contrast, there was no downregulation of katG or sodB observed in the present study. Likely, this is a consequence of the rapid on/off switch of the promoter binding capacity of OxyR in response to the oxidative levels in the bacteria leading to a limited window when elevated mRNA levels can be detected (Wei et al., 2012).
Besides antioxidant genes, our study revealed an aberrant expression of genes encoding chaperone proteins of the mutants. Such proteins are induced in response to various stresses, including oxidative stress (Hartl et al., 2011). Thus, the induced expression of these genes in oxyR and in katG, also observed by Ma et al. (2016), likely is a reflection of oxidative stress encountered by the mutants. The chaperone network likely helps the bacterium to handle this stress through unfolding and/or degradation of mis-folded/damaged proteins. The reason behind the suppressed expression of multiple chaperone genes in oxyR/ katG is obscure, but should lead to an accumulation of damaged or mis-folded proteins and may explain why it was so impaired for growth in broth. The intact growth of oxyR/ katG under microaerobic conditions likely reflects that reduced levels of ROS are formed and therefore that antioxidant defenses are less important. The aberrant expression of genes related to iron-uptake did not result in a skewed iron content of oxyR/ katG (data not shown) and it is therefore not obvious that this would influence the susceptibility of the strain to various ROS.
The F. tularensis ahpC2 gene is divergently transcribed from the oxyR promoter, a feature commonly seen for genes transcriptionally regulated by OxyR (Hahn et al., 2002;Maddocks and Oyston, 2008). AhpC belongs to the peroxiredoxin family, which is ubiquitously found in nature (Rhee et al., 2005) and is known to be involved in defenses against peroxides in E. coli (Storz et al., 1989), and both peroxides and peroxynitrite in, e.g., Salmonella typhimurium (Bryk et al., 2000), and in the defense against superoxide and peroxynitrite in the virulent SCHU S4 strain of F. tularensis subsp. tularensis (Binesse et al., 2015). In agreement with this, and in view of the reduced expression of AhpC in oxyR, this mutant was also highly susceptible to ONOO − . katG was as susceptible as oxyR to ONOO − and in view of the substantial catalase activity remaining in oxyR, this result implies that the function of catalase overlaps with other OxyR-regulated detoxifying mechanisms, presumably AhpC, to protect against ONOO − . Further corroborating the importance of AhpC and catalase was the failure to generate a katG and ahpC double deletion mutant and even an ahpC mutant in LVS. Hence, AhpC seems indispensable to LVS, which is in stark contrast to SCHU S4, where deletion of ahpC resulted in only slight attenuation (Binesse et al., 2015). This indicates that there is a disparity regarding the importance of the enzyme between the SCHU S4 and LVS strains, possibly a factor that to some extent explains the difference in virulence between the strains, since it implies that the detoxifying mechanisms of SCHU S4 are much more elaborate. Nevertheless, as for LVS, it has not been possible to generate a katG and ahpC double deletion mutant of SCHU S4 Binesse et al., 2015). Collectively, this indicates that the mechanisms of protection conferred by these enzymes may be overlapping and the lack of both is detrimental to the survival of both LVS and highly virulent F. tularensis strains.
Based on the failure to generate a katG and ahpC double deletion mutant and the marked suppression of ahpC in the oxyR mutant, we hypothesized that the absence of OxyR together with the absence of catalase would severely disarm the capability of the bacterium to handle ROS. Indeed, we observed that the oxyR/ katG mutant was hyper-susceptible to H 2 O 2 , ONOO − , and O − 2 ; much more so than either oxyR or katG. Collectively, the results demonstrate that the roles of OxyRregulated antioxidant enzymes and catalase overlap to protect LVS against various ROS. We find it likely that the reduced activity of catalase and expression of ahpC observed in oxyR contributed to the increased susceptibility of the mutant to H 2 O 2 , O − 2 , and ONOO − through the increase of both Fenton-mediated toxicity and direct O − 2 -and ONOO − -mediated damage. We further suggest that the reduced levels of AhpC together with the lack of catalase in the oxyR/ katG strain, despite an increased expression of sodB, sodC and FTT0086, resulted in enhanced Fenton-mediated toxicity and ONOO − -mediated damage, which likely account for the extreme susceptibility of the double mutant to O − 2 , H 2 O 2 , and ONOO − . Our findings concur with those of Ma et al. (2016), and, in addition, demonstrate that the combined activity of catalase and OxyR-regulated detoxifying mechanisms are critical for ROS detoxification by F. tularensis.
Despite the enhanced susceptibility of both oxyR and katG to various ROS, the strains replicated as efficiently as LVS in mouse BMDM, but importantly, the capacity to replicate in professional phagocytes required either OxyR or catalase, since oxyR/ katG failed to replicate. IFN-γ-activation of BMDM restricted growth of LVS, katG, and oxyR to a similar degree and completely blocked the growth of oxyR/ katG. The majority of F. tularensis LVS escapes the phagosome of IFN-γ-activated macrophages (Lindgren et al., 2004), but the mechanism of growth inhibition appears to vary depending on the cell model used (Edwards et al., 2010). IFN-γ-mediated inhibition of intracellular growth of F. novicida is dependent on the expression of IRGB10 and various guanylate-binding proteins (Meunier et al., 2015;Man et al., 2016), however, the role of this pathway is unknown for other F. tularensis species.
Our results reveal elaborate interconnecting roles between OxyR-regulated ROS-detoxifying mechanisms and catalase and demonstrate that either needs to be intact for the bacterium to thrive in professional phagocytes. The roles of the antioxidative mechanisms could be to protect the bacterium from direct damage by various ROS, such as ONOO − , which has been demonstrated to be crucial for killing of F. tularensis in peritoneal cells (Lindgren et al., 2005). Alternatively, or additionally, the antioxidants may restrict macrophage activation through their ability to preserve phosphatase activity required for kinase signaling and proinflammatory cytokine production (Melillo et al., 2010).
Our finding that oxyR replicated as efficiently as LVS in BMDM is in contrast to findings in a previous study, which reported that an oxyR mutant of LVS was markedly impaired with regard to escape from the phagosome, replication in professional phagocytes, and virulence in the mouse model (Ma et al., 2016). Notably, the LVS strain used by Ma et al. replicated less than 10-fold during 24 h in C57BL/6 BMDM, whereas the LVS strain used in the present study replicated about 500-fold. Isolates of LVS with different virulence are used in the research community (Griffin et al., 2015) and the distinct differences in the intracellular growth of these two LVS strains are additional examples of such distinct phenotypes. The phenotypic differences between the two LVS strains likely explain the discrepant findings of the two studies. The observation in the present study of the intact growth of the single mutants in BMDM was corroborated by findings in vivo, since the oxyR and katG mutants showed essentially intact growth in organs of mice, whereas oxyR/ katG hardly grew at all. Despite their effective growth in the organs, a previous study demonstrated a more distinct growth defect of the katG mutant, most likely because a 100fold higher dose was given (Lindgren et al., 2007). Moreover, by the intranasal route, oxyR was demonstrated to be moderately attenuated (Ma et al.). Based on these collective findings, it can be concluded that both OxyR and KatG contribute to the virulence of F. tularensis LVS and that the concomitant loss is detrimental to the virulence of the bacterium.
Altogether, the results presented in this study clearly demonstrate the mutual importance of catalase and OxyR for a robust oxidative stress defense system and that either of these systems is vital for the intracellular replication of F. tularensis and for its virulence.