Azithromycin effectiveness against intracellular infections of Francisella

Background Macrolide antibiotics are commonly administered for bacterial respiratory illnesses. Azithromycin (Az) is especially noted for extremely high intracellular concentrations achieved within macrophages which is far greater than the serum concentration. Clinical strains of Type B Francisella (F.) tularensis have been reported to be resistant to Az, however our laboratory Francisella strains were found to be sensitive. We hypothesized that different strains/species of Francisella (including Type A) may have different susceptibilities to Az, a widely used and well-tolerated antibiotic. Results In vitro susceptibility testing of Az confirmed that F. tularensis subsp. holarctica Live Vaccine Strain (LVS) (Type B) was not sensitive while F. philomiragia, F. novicida, and Type A F. tularensis (NIH B38 and Schu S4 strain) were susceptible. In J774A.1 mouse macrophage cells infected with F. philomiragia, F. novicida, and F. tularensis LVS, 5 μg/ml Az applied extracellularly eliminated intracellular Francisella infections. A concentration of 25 μg/ml Az was required for Francisella-infected A549 human lung epithelial cells, suggesting that macrophages are more effective at concentrating Az than epithelial cells. Mutants of RND efflux components (tolC and ftlC) in F. novicida demonstrated less sensitivity to Az by MIC than the parental strain, but the tolC disc-inhibition assay demonstrated increased sensitivity, indicating a complex role for the outer-membrane transporter. Mutants of acrA and acrB mutants were less sensitive to Az than the parental strain, suggesting that AcrAB is not critical for the efflux of Az in F. novicida. In contrast, F. tularensis Schu S4 mutants ΔacrB and ΔacrA were more sensitive than the parental strain, indicating that the AcrAB may be important for Az efflux in F. tularensis Schu S4. F. novicida LPS O-antigen mutants (wbtN, wbtE, wbtQ and wbtA) were found to be less sensitive in vitro to Az compared to the wild-type. Az treatment prolonged the survival of Galleria (G.) mellonella infected with Francisella. Conclusion These studies demonstrate that Type A Francisella strains, as well as F. novicida and F. philomiragia, are sensitive to Az in vitro. Francisella LPS and the RND efflux pump may play a role in Az sensitivity. Az also has antimicrobial activity against intracellular Francisella, suggesting that the intracellular concentration of Az is high enough to be effective against multiple strains/species of Francisella, especially in macrophages. Az treatment prolonged survival an in vivo model of Francisella-infection.


Background
Bacteria in the Francisella genus are nonmotile, nonsporulating, gram-negative coccobacilli. Francisella causes a zoonotic disease; humans can become infected via a variety of mechanisms including inhalation of an extremely low infectious dose [1]. F. tularensis primarily targets macrophages where bacterial survival and replication occurs [1]. The genus Francisella is divided into two species: tularensis and philomiragia. Francisella tularensis has four subspecies: F. tularensis subspecies tularensis (formerly F. tularensis,) F. tularensis subspecies holarctica (which includes the live vaccine strain, LVS), F. tularensis subspecies mediasiatica, and F. tularensis subspecies novicida (F. novicida) [2]. Subspecies of Francisella tularensis are further separated into two types depending on their virulence. Type A strains include Francisella tularensis subspecies tularensis Schu S4 (F. tularensis Schu S4) and are more virulent [3], except for the ATCC type strain F. tularensis subsp. tularensis NIH B38 which is avirulent [4][5][6]. Francisella Type A strains are normally associated with ticks and rabbits and are restricted to North America. Type B strains (Francisella tularensis subspecies holarctica and mediasiatica) are less virulent and cause tularemia throughout Eurasia [3].
Standard recommended antibiotic treatment for tularemia includes oral tetracycline antibiotics (e.g. doxycycline) and fluoroquinolones (e.g. ciprofloxacin) which have adverse side-effects on pediatric and the elderly patients, and individuals with liver disease. Aminoglycosides such as streptomycin and gentamicin can be injected intravenously or intramuscularly [7], but are not commonly used. Macrolides are oral antibiotics commonly used to treat bacterial respiratory illnesses. Azithromycin (Az), a member of the azalide subclass of macrolides, binds to the 50 s subunit of gram-negative bacterial ribosomes, and inhibits translation of mRNA resulting in inhibition of bacterial growth or death [8]. It has been suggested that the two basic amine sites of Az interact with the negatively charged heptose-phosphate region of lipopolysaccharide (LPS) in order to enter gram-negative bacteria [9]. F. novicida transposon insertion mutants in the genes involved in lipopolysaccharide (LPS) production (wbtN, wbtE, wbtQ and wbtA) were tested to determine if there might be a role of LPS in Az binding and penetration. Mutations in genes responsible for the synthesis of the O-antigen in F. novicida have been previously shown to decrease virulence and resistance to serum killing while macrophage uptake and replication remained unaffected [10].
A primary mode of bacterial resistance to antibacterial drugs is the expression of drug efflux pumps such as ATP-binding cassette (ABC), the Major Facilitator Superfamily (MFS) transporters, and Resistance-Nodulation-Division (RND) efflux system. These inner membrane transport systems are often coupled to the outer membrane TolC system [11]. Francisella novicida has two tolC-like proteins, tolC and the highly related fltC [12]. The ABC Superfamily is thought to be responsible for the export of many different antibiotics. For example, in E. coli, macrolides are thought to be transported by the ABC transporter MacAB [13]. Although a potential macA gene was identified in F. novicida (FTN_1692), no gene corresponding to macB could be identified in the F. novicida genome. The RND efflux system consists of a tripartite transporter with an RND pump protein located in the cytoplasmic membrane (AcrB) and a periplasmic membrane fusion protein (AcrA) coupled to the TolC protein in the outer membrane ( Figure 1). The RND system can pump many compounds, including macrolides [14]. The AcrAB RND efflux pump was recently demonstrated to be required for F. tularensis LVS virulence in mice [15], but not in F. tularensis Schu S4 [16]. The function of the RND efflux system is the removal of harmful substances from inside the cytosol of the bacteria directly to the external medium bypassing the periplasm [15]. Thus we hypothesized that mutants in the RND efflux system would have altered sensitivity to Az. Transposon insertion mutants of components of the RND efflux system in F. novicida, including tolC, fltC, acrA, and acrB, were tested for their sensitivity to Az. The dsbB gene encodes the cytoplasmic membrane protein that is involved in disulfide bond formation in the periplasm. A dsbB mutant in F. novicida was tested because it is transcriptionally linked in an operon with acrA and acrB in Francisella. Mutants ΔacrA and ΔacrB were also tested in the fully virulent strain, F. tularensis Schu S4 [16].
Az, which is commonly prescribed to pediatric patients for treatment of common upper respiratory track and ear infections [17], has low toxicity and few side-effects [18]. When administered, the antibiotic becomes ion-trapped in the acidic lysosomes of white blood cells including macrophages resulting in a high intracellular concentration compared to the plasma during the dose period. Intracellular concentrations remain high after the dose period ends with a half-life of 68 hours [18].
Murine macrophages J774A.1 are a well-studied in vitro model system for tularemia [19,20] and were chosen as a model cell system to study Francisella infection and treatment by Az. The murine macrophage cell line J774A.1 supports the intracellular replication of F. tularensis LVS [19], F. novicida [21], and F. tularensis Schu S4 [16]. For a model of the human system, human lung epithelial cells A549 were chosen. F. tularensis LVS has been previously shown to infect and replicate within A549 cells [22][23][24]. We hypothesized that the ability of Az to concentrate at high levels within the macrophages may result in effectiveness against intracellular infections by Francisella species, even at extracellular Az levels lower than the MIC.
The larval stage of Galleria (G.)mellonella, wax moth caterpillar, has been used as a model to study infections caused by some bacteria including F. tularensis LVS [25]. The larvae do not have an adaptive immune system, but have resistance to microbial infections via cellular and humoral defenses [26]. The analysis of insect responses to  [59], to illustrate the relationship between TolC, AcrA and AcrB.

Inner Membrane
Outer Membrane Periplasm AcrB TolC pathogens can provide an accurate indication of the mammalian response to that pathogen. Physical effects such as color change can be observed when the bacteria replicates and increases in the larvae [25]. We used G. mellonella as an alternative to the mouse model of Francisella infection to test our hypothesis that Az treatment could prolong the survival of Francisella infected caterpillars.

Francisella's sensitivity to Az
It has been reported that European clinical strains of Type B F. tularensis are resistant to Az [27]. However, we observed that commonly used laboratory strains of Francisella are sensitive to Az. In vitro susceptibility testing of Az confirmed that F. tularensis LVS strain was not highly sensitive in vitro to this antibiotic, confirming that the Type B strains are relatively resistant to this antibiotic. Our study demonstrated that F. philomiragia, F. novicida and Type A F. tularensis tularensis, including both F. tularensis tularensis NIH B38 and F. tularensis Schu S4 strains, were susceptible to this drug in vitro and in vivo.
Francisella strains were tested in a Kirby-Bauer disc inhibition assay for sensitivity to Az. F. novicida, F. philomiragia, and F. tularensis tularensis B38 were sensitive to 15 μg Az discs, whereas F. tularensis LVS was not sensitive to this concentration. F. novicida had a zone of inhibition of 28.7 ± 0.7 mm in diameter around the 6 mm Az disc, and F. philomiragia's zone of inhibition was 21.7 ± 0.8 mm in diameter. F. tularensis tularensis NIH B38 had the largest zone of inhibition, 45.9 ± 6.2 mm in diameter around the Az disc (Table 1). These results were all significantly different than F. tularensis LVS (p-value < 0.001). Although F. tularensis tularensis NIH B38 is not virulent, this result suggested the potential sensitivity of the Type A strains to Az. In order to corroborate this with the fully virulent strain, F. tularensis Schu S4 was tested and deter-mined to have a zone of inhibition of 25.5 ± 1.9 mm (pvalue < 0.001 compared to F. tularensis LVS).
The Minimal Inhibitory Concentrations (MIC) for Az and gentamicin were measured in liquid broth assays to determine Francisella sensitivity to Az compared to control antibiotic gentamicin. F. novicida and F. philomiragia were more susceptible to Az than F. tularensis LVS, which was only susceptible to Az at higher concentrations. The MIC of Az for F. novicida is 0.78 μg/ml (EC 50 of 0.16 μg/ ml), and 1.56 μg/ml (EC 50 of 0.13 μg/ml) for F. philomiragia. These results were all significantly different than F. tularensis LVS (MIC of 25.0 μg/ml; EC 50 of 17.3 μg/ml; pvalue ≤ 0.004) ( Figure 2, Table 2). The MIC result for F. tularensis LVS explains why there was no inhibition of growth in the disc-diffusion assay, as there was only 15 μg of Az in the disc, which is below the MIC and the EC 50 (Table 2). This is consistent with the disc inhibition assay results. These results are also similar to results with related macrolide antibiotic, erythromycin, which has a reported MIC of 0.5-4, and EC 50 of 2 μg/ml against Type A and B Francisella strains, though not LVS (MIC > 256 μg/ml) [28]. As a control, we determined the MIC for the antibiotic gentamicin to which all strains of Francisella are susceptible [29]. The MIC of gentamicin for F. novicida was determined to be 0.2 μg/ml (EC 50 of 0.12 μg/ml); for F. philomiragia the MIC was 0.39 μg/ml (EC 50 of 0.22 μg/ml); and for F. tularensis LVS the MIC was 0.39 μg/ml  [30]. Cells were lysed and bacteria were recovered and counted as colony forming units (CFU). Francisella-infected J774A.1 and A549 cells were found to have more than 105 CFU/ml of Francisella after 22 hours after infection. J774A.1 cells infected with Francisella and treated with Az had decreasing CFUs as the antibiotic concentration increased. In J774A.1 cells infected with F. philomiragia, no CFUs were recovered when treated with 0.1 μg/ml Az (less than the MIC). In J774A.1 cells infected with either F. novicida or F. tularensis LVS, bacterial concentrations decreased with the addition of Az. At 5 μg/ml Az, no CFUs were recovered (pvalue < 0.005 compared to 0 μg/ml Az) ( Figure 3A). In this case, the Az concentration was less than the MIC for F. tularensis LVS. Francisella-infected A549 cells required higher concentrations of Az than J774A.1 cells, suggesting that epithelial cells are not able to concentrate Az in the same manner as macrophages. As before, intracellular F. novicida, F. philomiragia, and F. tularensis LVS CFU counts decreased when A549 cells were treated with Az. Recovered intracellular CFU counts for F. philomiragia and F. novicida remained approximately equal when treated with 0.1 and 5 μg/ml Az (p-value > 0.05), but strongly decreased at 25 μg/ml Az (p-value < 0.005 compared to 0 μg/ml Az). For these two organisms, the required external antibiotic concentration was higher than the in vitro MIC. F. tularensis LVS infected A549 cells had a steady decline of intracellular CFU counts as the Az concentration increased and had essentially no colonies recovered at 25 μg/ml extracellular Az (p-value < 0.005 compared to 0 μg/ml Az), which is equivalent to the MIC for that strain ( Figure 3B). The difference between the cell types may reflect the fact that J774A.1 cells are phagocytic macrophages, and the A549 cells are nonphagocytic epithelial cells.
To determine if Francisella bacteria counts were decreased due to Az concentrations or due to cell death, cellular lysis and apoptosis were measured by LDH released [19]. At 22 hours, cell cytotoxicity in noninfected A549 cells and A549 cells infected with F. novicida, F. philomiragia, and F. tularensis LVS remained below 20%. Non-infected A549 cells along with F. philomiragia, F. novicida, and F. tularensis LVS-infected cells had a slightly increased cytotoxicity as Az concentrations increased (Table 3). Cellular apoptosis remained low with all Az doses. These results suggest the decreased Franci-  Azithromycin Concentration (log Pg/ml) % Survival sella counts were due to Az treatment and not due to bacterial release during the experiment from apoptosis or cell lysis.

Francisella LPS mutants
Due to the potential for interaction of Az with LPS [9], four F. novicida transposon LPS O-antigen mutants were tested for their Az susceptibility: O-antigen of LPS (wbtA) biosynthesis of GdNAcAN, an O-antigen unit (wbtE), glycosylatransferase that elongates to form GalNAcAN trisaccharides (wbtQ), and aminotransferase (wbtN) [10]. F. novicida LPS O-antigen mutants including wbtA, wbtE, wbtQ, and wbtN were shown to be less susceptible to Az by decreased zones of inhibition in comparison to the wild-type (p-value < 0.001) ( Table 4). The MICs for Az against the F. novicida LPS-related transposon mutants wbtA, wbtE, wbtQ, and wbtN (MIC's > 3.0 μg/ml Az, EC 50 > 0.50 μg/ml Az) were greater than the wild-type MIC (0.78 μg/ml) (p-value < 0.005) ( Figure 4A, Table 5), suggesting increased resistance to Az. These data are consistent with the disc inhibition studies, suggesting that Francisella LPS plays some role in the sensitivity of the strains for Az.

Francisella RND mutants
Five F. novicida transposon insertion mutants in the multidrug efflux protein genes (acrA and acrB), the transcriptionally linked protein gene (dsbB), as well as the related outer membrane channel genes (tolC and ftlC) were tested to determine if Az susceptibility increases or decreases [12]. Results for the RND efflux mutants varied among the different subspecies (Table 6). In the disc inhibition assay, the tolC mutant was slightly more sensitive to Az compared to the wild-type F. novicida (p-value = 0.007), while ftlC, acrA, and acrB were less susceptible to Az compared to the wild-type (p-value < 0.01) ( Table 6). The MICs for ftlC, tolC, acrA, and acrB (MIC = 25 μg/ml Az) were greater than the wild-type (MIC of 0.78 μg/ml Az) and had a higher EC 50 (EC 50 > 12 μg/ml Az) compared to the wild-type of 0.16 μg/ml Az (p-value < 0.002), indicating decreased sensitivity to the antibiotic. These results are consistent between the MIC and disc inhibition assay for acrA, acrB, and ftlC ( Figure 4B, Table 5).
The tolC sensitivity to Az results in the solid agar and liquid broth assay were inconsistent. The disc-inhibition assay suggests increased sensitivity, while the MIC assay demonstrated increased resistance. We are currently investigating the basis of this difference.
In the disc inhibition assay of the disulfide bond protein mutant dsbB, there was no significant difference compared to the wild-type (p-value = 0.162) ( Table 6). Similarly, the MIC for dsbB was not significantly different than the wild-type value (p-value = 0.400) ( Table 5). Thus, mutation of dsbB does not seem to have a significant impact on the ability of the organism to resist Az, whereas transposon insertion mutants in the tolC, ftlC, acrA and acrB components of the RND efflux system appear to decrease the sensitivity of F. novicida to Az. This result for tolC and ftlC may be in contrast to Gil et al. [12], who found that F. tularensis LVS deletion of tolC or ftlC did not alter the sensitivity to erythromycin (15 μg disc). The MIC of F. tularensis LVS is higher than can be achieved using a 15 μg disc, reported at >256 μg/ml erythromycin [28]. Therefore, any alteration in sensitivity due to tolC deletion would not be observed at this low concentration of antibiotic.
In contrast to the F. novicida results, the F. tularensis Schu S4 ΔacrA mutant and ΔacrB mutants had greater sensitivity to Az compared to the wild-type F. tularensis hours, recovered bacterial counts were measured for F. philomiragia, F. novicida, and F. tularensis LVS infected cells (MOI 500). A) J774A.1 cells infected with F. philomiragia, F. novicida, or F. tularensis LVS had more than 10 5 CFU/ml. Bacterial counts decreased for all strains as the Az concentrations increased and were near 0 CFU/ml at 5 μg/ml Az. B) A549 cells infected with F. philomiragia, F. novicida, or F. tularensis LVS had more than 10 5 CFU/ml at 0 μg/ml Az. Bacterial counts decreased at 0.1 and 5 μg/ml Az and were near 0 CFU/ml at 25 μg/ml Az. CFU counts from no Az treatment compared 0.1, 5, and 25 μg/ml Az treatment for all Francisella strains were significantly different (p-value < 0.005).
Schu S4 (p-value < 0.001) ( Table 6). This is consistent with the findings of Qin et al. [16] who found an increased sensitivity of ΔacrB to 50 μg disc erythromycin. The MICs for Az against F. tularensis Schu S4 RND efflux mutants were also determined. The MICs for ΔacrA and ΔacrB (MIC > 1.5 μg/ml Az) are higher than the wildtype MIC of 0.78 μg/ml Az (p-value < 0.02) ( Figure 4C, Table 5). However, the F. tularensis Schu S4 mutants for ΔacrA (EC 50 of 0.085 μg/ml) and ΔacrB (EC 50 0f 0.049 μg/ ml) have EC 50 s less than the wild-type F. tularensis Schu S4 (EC 50 of 0.145 μg/ml), reflecting the altered shape of the MIC curve and indicating increased sensitivity. Only ΔacrB was statistically significantly different for EC 50 when compared to the wild-type F. tularensis Schu S4 (pvalue < 0.05). Thus, F. tularensis Schu S4 ΔacrA and ΔacrB mutants had greater sensitivity to Az compared to F. novicida mutants, or the parental F. tularensis Schu S4 strain by disc inhibition assay and MIC.

Az inhibition of intracellular Francisella mutant strains
J774A.1 and A549 cells infected with F. novicida transposon LPS mutant wbtA and multidrug efflux mutants ftlC, tolC, acrA, and acrB had more than 10 4 CFU/ml 22 hours post-infection ( Figure 5). ftlC generally had lower CFU counts, whereas the acrA and acrB had higher CFU counts in both cell lines. The CFU of F. novicida transposon mutants decreased as the Az concentration increased for each cell line (p-value < 0.005 for each Az treatment compared to 0 μg/ml Az). At 35 μg/ml Az treatment, the bacterial CFU count was near 0 CFU/ml in J774A.1 and A549 cells ( Figure 5). Thus, wbtA and the RND mutants are capable of replication within J774A.1 and A549 cells, although the overall number of bacteria per cell was lower than for the parental F. novicida infection (1.76 × 10 5 ± 6.36 × 10 3 CFU/ml in J774A.1 and 1.80 × 10 5 ± 1.41 × 10 4 CFU/ml in A549 cells at 0 μg/ml). Mutant trends after Az treatments were significantly different from the wild-type F. novicida with a p-value < 0.05 (wild-type decreased to 0 CFU/ml at 5 μg/ml Az in J774A.1 cells and decreased to 0 CFU/ml at 25 μg/ml Az in A549 cells). Corresponding to the higher MICs identified in vitro, LPS mutants require more Az to eliminate the bacteria from infected cells. Using a LDH release assay, the cell cytotoxicity as a result of antibiotic and/or Francisella infection was determined and is indicated as a percentage (%) of total LDH released.

G. mellonella infection by Francisella and antibiotic treatment
Francisella-infected G. mellonella was used as a model system [25] to study Az treatment. G. mellonella were infected with either 3 × 10 6 CFU bacteria/larva of F. novicida or F. tularensis LVS and then treated with a single dose of 10 μl injections PBS (no antibiotic), 20 μg/ml ciprofloxacin, or 25 μg/ml Az. Control groups (no infection) consisted of no injections, injections of either PBS (to measure trauma related to injections), 20 μg/ml ciprofloxacin, or 25 μg/ml Az (to assess antimicrobial agent effects on the host). All controls had similar survival rates (data not shown for antibiotic injection only controls).
Francisella-infected G. mellonella did not survive past 100 hours post-infection. Control groups survived for more than 300 hours. Infected groups treated with a single dose 20 μg/ml ciprofloxacin (mean time to death > 74 hours) or 25 μg/ml Az (mean time to death > 160 hours) had a statistically significant prolonged survival times when compared to infected groups (p-value < 0.005) (Figure 6A &6B). These results are consistent with previously published results of G. mellonella infected with F. tularensis LVS and treated with 20 μg/ml ciprofloxacin [25]. Although we could not achieve complete recovery, Francisella-infected G. mellonella groups treated with Az had an increased mean survival time compared to ciprofloxacin-treated caterpillars (p-value < 0.02).

Discussion
The macrolide erythromycin has limited efficacy against many gram-negative bacteria due to its hydrophobic nature and lack of permeability of the gram-negative outer membrane [31]. The sensitivity of erythromycin varies between Francisella strains. In the North American Type A Francisella strains, erythromycin MICs range from 0.5 to 4 μg/ml, while F. tularensis LVS has an MIC > 256 μg/ml [32]. The macrolide azithromycin is more effective against gram-negative bacteria than erythromycin [33]. Despite reports that European clinical strains of Type B F. tularensis are resistant to Az (MIC > 256 mg/L) [27], we observed that commonly used laboratory strains were sensitive to Az. In this study, we have demonstrated Az is pumped out of gram-negative bacteria by several drug-efflux systems, including the RND efflux pumps. Az sensitivity differed between F. novicida and F. tularensis Schu S4 RND efflux mutants. Wild-type F. tularensis Schu S4 has similar sensitivity to Az as wild-type F. novicida, but the RND efflux mutants ΔacrA and ΔacrB in F. tularensis Schu S4 are more sensitive to Az, whereas the F. novicida acrA and acrB mutants are more resistant. These F. tularensis Schu S4 ΔacrA and ΔacrB mutants were also reported to be more sensitive to the related antibiotic erythromycin [16]. The difference between the F. tularensis Schu S4 and the F. novicida mutants might be due to the fact that F. tularensis Schu S4 has 254 pseudogenes; many of these genes are intact in F. novicida [34]. For example, in F. tularensis Schu S4, at least 14 genes of the MFS transporter superfamily contain stop codons or frameshifts [34,35] and are thus predicted to be non-functional. Additional types of transporter proteins, including a drug-resistance transporter (FTT1618), are also reported to be non-functional pseudogenes [34] in F. tularensis Schu S4. It could be that the remaining TolC-AcrAB pump is the major means by which F. tularensis Schu S4 pumps out Az. If this pump is compromised, the organism would be more susceptible to the antibiotic, because it may not have an operational alternative pump, such as the MFS or ABC transporters to pump out the drug. This is supported by the finding that ΔacrA and ΔacrB mutants in F. tularensis Schu S4 also displayed increased sensitivity to nalidixic acid (a substrate for the MFS transporter), as well as detergents, streptomycin, tetracycline, and other molecules [16]. In the case of F. novicida, there may be alternate systems that can pump out the drug in the absence of the RND system. Alternatively, the mutation in acrA or acrB may cause an up-regulation of expression of another drug-efflux pump, rendering the bacteria more resistant to the antibiotic [36,37]. Previous studies have shown that dsbB mutant in F. tularensis Schu S4 does not have any effect on antibiotic sensitivity (including the macrolide erythromycin) [16]. Consistent with the F. tularensis Schu S4 dsbB mutant, the F. novicida dsbB mutant showed no difference from the wild-type F. novicida. Another common mechanism of resistance to macrolides is modification of the 23S rRNA. It has been reported that F. tularensis LVS has a point mutation in Domain V of the 23S rRNA, rendering it more resistant to erythromycin than F. novicida or F. tularensis Schu S4 [38]. This modification could also explain the increased resistance to Az in F. tularensis LVS. In addition, there are methylases that can confer increased resistance by targeted modification (methylation) of a specific adenine residue of the 23S rRNA. There are some methylases that have been identified as critical virulence factors for Francisella that might carry out this modification [39]. Some methylases that are present in the genome of F. novicida are either absent or are pseudogenes/nonfunctional genes (such as FTT0010, FTT0770, FTT1430, FTT1719, and FTT1735c) in F. tularensis Schu S4, potentially contributing to the different sensitivities to Az between the strains [34]. Any potential role of these molecules in Az sensitivity or resistance in Francisella has not yet been determined.
It has been suggested that Az attaches to the acidic LPS on the outer membrane of gram-negative bacteria, allowing the drug to penetrate through the outer membrane decreased as Az concentrations increased and was near 0 CFU/ml at 35 μg/ml Az (p-value < 0.005 for all Az treatments compared to 0 μg/ml Az for each mutant). The recovery of mutant strains after Az treatments were significantly different from the wild-type F. novicida with a p-value < 0.05 (1.76 × 10 5 ± 6.36 × 10 3 CFU/ml in J774A.1 at 0 μg/ml Az which decreased to 0 CFU/ml at 5 μg/ml Az and 1.80 × 10 5 ± 1.41 × 10 4 CFU/ml in A549 cells at 0 μg/ml Az which decreased to 0 CFU/ml at 25 μg/ml Az). J774A.1 cells had higher bacterial counts than A549 cells.

Figure 6
Antibiotic treatment of Francisella-infected G. mellonella. High concentrations of antibiotics prolonged the survival of G. mellonella infected with 3 × 10 6 CFU Francisella. Non-infected control groups consisted of no injection, PBS injection, 25 μg/ml Az injection, or 20 μg/ml ciprofloxacin injection. All non-infected controls had similar high survival rates (data not shown for non-injected, 25 μg/ml Az injection, or 20 μg/ml ciprofloxacin injection). A) The infected control group received F. novicida injection, then PBS. A single dose of 25 μg/ ml Az, given 2 hours after bacterial inoculation, was effective when compared to the infected control (p-value = 0.004). Treatment with 20 μg/ml ciprofloxacin prolonged the survival of the caterpillars compared to the control (p-value < 0.01). B) The infected control group received F. tularensis LVS injection, then PBS. A single dose of 25 μg/ml Az, given 2 hours after bacterial inoculation, was effective compared to the infected control (p-value < 0.001). Treatment with 20 μg/ml ciprofloxacin prolonged the survival of the caterpillars compared to control (p-value < 0.01). For F. tularensis LVS and F. novicida infections, survival time was longer in Az treated groups compared to ciprofloxacin treated groups (p-value < 0.02). and enter the bacteria [40]. The wbt locus in Francisella, which is responsible for the production of LPS O-antigen, has been shown to be required for virulence [41]. In published reports, the wbtA mutant in F. tularensis LVS demonstrated a loss of the O-antigen and an inability to replicate in mouse macrophages. F. novicida wbtA mutants replicate normally and have only moderate sensitivity to serum [42,43]. We tested F. novicida transposon-insertion mutants wbtN, wbtE, wbtQ and wbtA, which are involved in the production of LPS, and found that these mutants were less susceptible to Az. Mutations of the LPS in the F. novicida transposon LPS O-antigen mutants may alter the LPS region presumed to bind to Az, resulting in a decreased amount of Az penetration and increased resistance to Az. Our results support the proposed role of LPS O-antigen in Az penetration into gram-negative bacteria such as Francisella.
Az is a weak base that can remain inside host cells for a longer time at a higher concentration than in the serum. This occurs because the basic amine groups of Az neutralize the lysosomal pH and prevent acidification of the lysosome. This process causes the drug to become trapped in the cell due to the positive charge. The drug is slowly released from polymorphonuclear neutrophils, allowing for a long half-life [8]. Az also concentrates in macrophages, which suggested to us that it might be useful as a potential treatment of intracellular pathogens such as F. tularensis. J774A.1 mouse macrophage were infected with F. philomiragia, F. novicida, and F. tularensis LVS and treated with Az. It was determined that 5 μg/ ml Az was effective in eliminating intracellular F. philomiragia, F. novicida, and even F. tularensis LVS infections in J774A.1 cells. Although Type B strains are intrinsically more resistant to macrolides, F. tularensis LVS CFUs were eliminated below the Az MIC values for this strain. We suggest that J774A.1 cells can sufficiently concentrate Az so that the intramacrophage concentration of Az exceeds the MIC. Thus, it may be that Az is effective against LVS in vivo due to the concentration effect in macrophages. A concentration of 25 μg/ml Az was found to be effective against Francisella infections in A549 cells, suggesting that these non-phagocytic cells may be less able to concentrate the antibiotic intracellularly [22].
Az treatment has not been tested sufficiently in the clinic to know if it can be used to treat tularemia infection. In one reported case, the patient's illness was fatal after treatment by Az, trimethoprim-sulfamethoxazole, streptomycin, and ceftriaxone of F. tularensis [44], suggesting that the patient was extremely ill when treatment was initiated. In another case, the patient's symptoms decreased with a one day ceftriaxone treatment followed by a 5 day Az treatment, but symptoms recurred after the treatment was completed [45]. There have been several reports of successful treatment with erythromycin, giving credence to the sensitivity of Type A strains to the macrolide class of antibiotics [46,47]. To test the in vivo effectiveness of Az against Francisella infections, we employed the wax-moth caterpillar model [25]. The timecourse of infection of the caterpillars closely matched the published report. We extended the published report by demonstrating that wax-moth caterpillars can also be infected by F. novicida. We demonstrated that a single injection of Az increased the mean survival time of Francisella infected G. mellonella and is more effective than a similar dose of ciprofloxacin. Within a host, macrolides, including Az, inhibit the production of cytokines that cause inflammation and prevent the accumulation of neutrophils, which suggests immunomodulatory effects separate from their antibacterial effects [48]. It has been shown that after Francisella infection in mice, there is a delayed response in the induction of host proinflammatory cytokines and recruitment of inflammatory cells to the site of infection, resulting in uncontrolled bacterial replication [49]. G. mellonella, however, does not have a similar immune response following Francisella infection. Since the therapeutic efficacy of Az cannot be observed in G. mellonella, future experiments will be conducted using a mouse model. Our results demonstrate efficacy of Az against multiple different Francisella strains and species. In future work, we will extend the Az studies to murine infections with the fully virulent strain, F. tularensis Schu S4.

Conclusion
Az and other macrolide antibiotics may have a secondary benefit to patients with pneumonic tularemia infection since they also have immunomodulatory functions. Az has been used to treat non-infectious respiratory diseases such as diffuse panbronchiolitis (an inflammatory lung disease) and has been shown to reduce cytokine responses in the lungs thereby lessening the acute inflammatory response [48,50], even at sub-antimicrobial doses. Az is also used in the long-term management of lung transplant patients, including those with bronchiolitis obliterans syndrome, a disease occasionally resulting from the chronic immunological and inflammatory status in some post-transplant lungs [51]. Pulmonary tularemia often exhibits a robust pro-inflammatory response. If Az proves to be effective against F. tularensis in vivo, it may provide a dual therapeutic effect by also mitigating the pro-inflammatory response. Thus, there may be additional non-antimicrobial benefits to the lung as a result of using Az to treat pulmonary tularemia, which is often complicated by robust pro-inflammatory responses.
The current established treatment protocol for tularemia in children is ciprofloxacin [52]. However, ciprofloxacin has the potential for significant side effects, including liver toxicity, tendonitis and renal failure [40,53,54]. Az (trade name: Zithromax) is commonly prescribed to pediatric patients for ear infections and other common gram-negative infections, with very safe outcomes [55]. With the finding that Az concentrates in macrophages and is effective against Francisella species (including LVS) in vitro and in an in vivo infection model, we propose that further studies be done to establish the clinical utility of Az against tularemia, as an alternative treatment. In case of a deliberate tularemia infection of the population, such as in a biological weapons attack, there may be patients who can not tolerate the standard treatment. Az could be tested either as a stand-alone therapy or in combination with other chemotherapeutic agents. Developing an alternate effective therapy to treat tularemia in patients that do not tolerate ciprofloxacin well, such as pediatric and elderly patients, will lead to safer therapeutic options for physicians.

Antibiotics
The antibiotics investigated in this study were azithromycin (Az) (Biochemika), gentamicin (ATCC), and ciprofloxacin (Biochemika). Az was obtained as 15 μg discs (Fluka # 68601 or Remel # R33105), and dry powder (Fluka). Az was dissolved in distilled water and ciprofloxacin was dissolved in 0.5 M HCl to appropriate concentration. Gentamicin was obtained in solution at high concentration (50 mg/ml, ATCC) and diluted in distilled water.

Disc inhibition assay
Kirby-Bauer disc inhibition assay protocol was followed [57]. 100 μl of overnight bacterial cultures were spread on Chocolate II agar and Schu S4 strains were spread on Mueller-Hinton agar plate with three discs each containing 15 μg Az placed in a triangle and incubated based on length of time for bacterial growth to be seen on the plate: 24 (for F. novicida, F. philomiragia, and F. tularensis Schu S4), 48 (for F. tularensis LVS), and 72 hours (for F. tularensis NIH B38) at 37°C in 5% CO 2 . The diameter of the zone of inhibition including the 6 mm disc was measured (in mm) with three independent measurements for each zone (n = 9). Inhibition was defined as the area of no bacterial growth around the discs. A reading of 6 mm indicates no inhibition [57].

Minimal inhibitory concentration (MIC)
Assays were performed with small modification following published protocols [58]. The MIC for F. novicida, F. philomiragia, F. tularensis LVS, related F. novicida mutants, F. tularensis Schu S4, and related F. tularensis Schu S4 mutants were determined in TSB-C media by antibiotic dilution in triplicates. The broth was then inoculated with 10 5 CFU/ml per strain. Concentration of the antibiotics ranged from 1 mg/ml to 0.0001 μg/ml. The MIC was read at optical density 600 nm after 24 hours (for F. philomiragia, F. novicida, and F. tularensis Schu S4) and after 48 hours (for F. tularensis LVS) and was defined as the lowest concentration of antibiotic with no visible growth.

Data analysis and statistics
Data were analyzed using the following equation and GraphPad Prism 4 (GraphPad Software Inc., San Diego, CA) [23].
Y corresponds to bacterial mortality (% OD, where zero drug = 100%) at a given antibiotic concentration (μg/ml), with X being the logarithm of that concentration (log μg/ ml). In the equation, "Top" and "Bottom" refer to the upper and lower boundaries, and were constrained to values <100% and >0%, respectively. EC 50 values were determined by fitting the data from the antimicrobial assays to  1) with a Hill slope of 1. Control samples with no antibiotic are plotted as 10^-4 μg/ml for graphing purposes. Errors were reported based on the standard deviation from the mean of the Log EC 50 values. Student's T-test was used to determine whether points were statistically different, using a two tailed test assuming normal distribution.

Cell infection with Francisella strains
J774A.1 cells and A549 cells were plated (10 5 /well) in a 96-well plate and infected with either F. novicida, F. philomiragia, F. tularensis LVS, or F. novicida transposon mutants at MOI 500 for 2 hour incubation. Extracellular bacteria were removed by washing cell wells twice with DMEM for J774A.1 cells or Ham's F-12 for A549 cells. After Francisella infection and removal of extracellular bacterium, cells were incubated with 50 μg/ml gentamicin for 1 hour to eliminate extracellular bacterium but which does not affect intracellular bacteria. Cells were washed with media twice and incubated with Az in the media at final concentrations of 0, 0.1, 5, 15, 25, and 35 μg/ml for 0 or 22 hours at 37°C.

Quantification of intracellular Francisella bacteria
After exposure of cells to Francisella and antibiotics, the numbers of intracellular bacteria were determined. At 0 and 22 hours, the samples were washed twice with PBS. Sterile deionized water was used to lyse cells. Aliquots of cells and cell-associated bacteria were serially diluted onto chocolate agar plates, incubated at 37°C and 5% CO 2 for 1 or 2 days and the CFU were counted.

Quantification of cellular apoptosis
After exposure of cells to Francisella and antibiotics, the numbers of cell-associated bacteria were determined, the CytoTox-96 ® Non-radioactive Cytotoxicity Assay (Promega) was used to quantitatively measure lactate dehy-