Lipophilic Prodrugs of FR900098 Are Antimicrobial against Francisella novicida In Vivo and In Vitro and Show GlpT Independent Efficacy

Bacteria, plants, and algae produce isoprenoids through the methylerythritol phosphate (MEP) pathway, an attractive pathway for antimicrobial drug development as it is present in prokaryotes and some lower eukaryotes but absent from human cells. The first committed step of the MEP pathway is catalyzed by 1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR/MEP synthase). MEP pathway genes have been identified in many biothreat agents, including Francisella, Brucella, Bacillus, Burkholderia, and Yersinia. The importance of the MEP pathway to Francisella is demonstrated by the fact that MEP pathway mutations are lethal. We have previously established that fosmidomycin inhibits purified MEP synthase (DXR) from F. tularensis LVS. FR900098, the acetyl derivative of fosmidomycin, was found to inhibit the activity of purified DXR from F. tularensis LVS (IC50 = 230 nM). Fosmidomycin and FR900098 are effective against purified DXR from Mycobacterium tuberculosis as well, but have no effect on whole cells because the compounds are too polar to penetrate the thick cell wall. Fosmidomycin requires the GlpT transporter to enter cells, and this is absent in some pathogens, including M. tuberculosis. In this study, we have identified the GlpT homologs in F. novicida and tested transposon insertion mutants of glpT. We showed that FR900098 also requires GlpT for full activity against F. novicida. Thus, we synthesized several FR900098 prodrugs that have lipophilic groups to facilitate their passage through the bacterial cell wall and bypass the requirement for the GlpT transporter. One compound, that we termed “compound 1,” was found to have GlpT-independent antimicrobial activity. We tested the ability of this best performing prodrug to inhibit F. novicida intracellular infection of eukaryotic cell lines and the caterpillar Galleria mellonella as an in vivo infection model. As a lipophilic GlpT-independent DXR inhibitor, compound 1 has the potential to be a broad-spectrum antibiotic, and should be effective against most MEP-dependent organisms.


Methylerythritol phosphate pathway
Isoprenoids are involved in many critical cellular functions. They participate in electron transport, signal transduction, and maintenance of cell wall and membrane structural integrity. All isoprenoids are formed through either the mevalonic acid (MVA) or the methylerythritol phosphate (MEP) pathways [1]. Plants, algae, and bacteria utilize the MEP pathway to generate isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) from pyruvate and glyceraldehyde-3-phosphate [2]. The MVA pathway is the only pathway used by animals, making enzymes of the MEP pathway attractive targets for novel therapeutics [1].
The first committed step of the MEP pathway is catalyzed by 1deoxy-D-xylulose 5-phosphate reductoisomerase (DXR or MEP synthase) [2]. DXR catalyzes the reaction that generates MEP from 1-deoxy-D-xylulose 5-phosphate (DXP) ( Figure 1A) [1]. MEP pathway genes have been identified in many biothreat agents, including Francisella, Brucella, Bacillus, Burkholderia, and Yersinia [1,3,4]. DXR has been cloned from many different bacteria, including Escherichia coli, Pseudomonas aeruginosa, and Francisella tularensis [1,2]. The importance of the MEP pathway to F. tularensis is demonstrated by the fact that MEP pathway mutations are lethal [5]. It has also been demonstrated that the DXR gene is essential for Mycobacterium tuberculosis [6].
1.1.1 Fosmidomycin and FR900098 are two potent inhibitors of DXR. Fosmidomycin and FR900098 are phosphonate antibiotics naturally produced by Streptomyces lavendulae and S. rubellomurinus, respectively [7,8]. The structures of fosmidomycin and FR900098 are related, with FR900098 being the acetyl derivative of fosmidomycin ( Figure 1B). Fosmidomycin is currently being used in clinical trials in conjunction with clindamycin to treat malaria [9]. FR900098 is twice as active as fosmidomycin against the malaria parasite in vitro and in vivo [10]. Two functional groups of these antibiotics are important for their binding efficacy and inhibition of DXR: the hydroxymate moiety that chelates with a divalent metal ion (Mn 2+ , Mg 2+ , or Co 2+ ) and a negatively charged phosphonate group [11]. Fosmidomycin and FR900098 are both effective antimalarial agents, but they are limited in their effect due to re-emergence of an active infection after completing treatment [9], and low bioavailability, likely due to low lipophilicity [10]. Fosmidomycin requires the glycerol-3phosphate transporter (GlpT) to enter cells [12]. Both compounds are well tolerated in mice up to 300 mg/kg and demonstrate antimalarial efficacy following both intraperitoneal and oral administration [13].
We have previously established that fosmidomycin inhibits purified DXR from F. tularensis LVS with half maximal activity of 247 nM [1]. This is comparable to its effect against DXR from M. tuberculosis (310 nM) [14] and less active against the same enzyme from E. coli (35 nM) [2]. Jawaid et al. suggest that the difference in concentration required for half-maximal activity may be due to structural differences of the DXR homologs [1].

Lipophilic FR900098 prodrugs
Fosmidomycin and FR900098 are effective against purified DXR from M. tuberculosis [6,14,15] but have no effect on whole cells of this bacterium because the compounds are too polar to penetrate the thick cell wall [6]. In addition, fosmidomycin requires the GlpT transporter to enter cells, and this is absent in M. tuberculosis [14]. This data prompted us to question whether lipophilic analogs of fosmidomycin and FR900098 might better penetrate bacteria with thick cell walls and/or no GlpT transporter. Recently, Ortmann et al. generated a series of acyloxyalkyl ester prodrug derivatives of FR900098, including compound 1 ( Figure 1B), which demonstrated improved in vivo antimalarial activity [10]. These compounds are considered prodrugs of FR900098, and are metabolized to FR900098 in bacteria [10]. We have shown that these analogs have antimicro-bial activity against a broad range of bacteria [16] and may also be better at penetrating the cell membranes of eukaryotic cells, which is important for access to intracellular pathogens. For example, both F. tularensis and M. tuberculosis colonize host cells during the course of infection. We tested the ability of some of these compounds to inhibit F. novicida intracellular infection, using both cultured eukaryotic cell lines, and the caterpillar Galleria mellonella as an in vivo infection model.

Francisella tularensis
F. tularensis is a highly infectious Gram-negative facultative intracellular bacterium. Inhalation of as few as ten organisms can cause disease in humans [17]. It is of particular interest due to its historical use as a bioweapon, and is on the Centers for Disease Control's list of Category A select agents [18]. F. tularensis causes the disease tularemia in mammals, including humans, which can be spread via arthropod vectors, such as ticks [19], or by aerosol. F. tularensis can cause a pneumonic disease if it is inhaled, but more commonly causes the ulceroglandular form of the disease that occurs via skin contact. F. tularensis can replicate in many different cell types of mammalian hosts (for example, dendritic cells, neutrophils, hepatocytes, and lung epithelial cells), but macrophages appear to be the main target of this bacterium [19,20]. There are four closely related subspecies of F. tularensis, known as tularensis, holarctica, mediasiatica, and novicida [19]. F. novicida is a model organism of the more virulent F. tularensis species [21]. F. novicida is attenuated for disease in humans, but can still cause disease in small mammals, such as mice [19]. F. tularensis NIH B38 is classified as the type strain for F. tularensis tularensis, but is attenuated for virulence [22][23][24], and thus is a good model for F. tularensis Schu S4, the fully virulent strain. F. tularensis is of concern due to its historical use as a bioweapon in an aerosolized form [18]. Such an event could cause severe pulmonary disease in thousands of individuals and would impose a severe strain and high costs on the health care and public safety systems [20]. Prompt treatment would be important in decreasing the impact of  such an attack. The potential of engineered antibiotic resistant strains suggests that new classes of antibiotics with different modes of action from the standard antibiotics, such as ciprofloxacin, should be developed against F. tularensis.

Identification of the GlpT homolog in Francisella
The uptake of fosmidomycin into many bacteria is an active process dependent on the 12 transmembrane-spanning protein, glycerol-3-phosphate transporter, GlpT [12]. GlpT is a member of the organophosphate:phosphate antiporter family that is part of the major facilitator superfamily (MFS) [12]. E. coli mutants defective in the glpT gene are resistant to fosmidomycin [12]. M. tuberculosis lacks a GlpT homolog, partially accounting for its resistance to fosmidomycin [4,6]. Brucella acquires fosmidomycin sensitivity when it expresses E. coli GlpT [4]. We identified a gene in F. tularensis (FTT0725c) and in F. novicida (FTN_0636) as a potential GlpT homolog (Table 1), and transposon insertion mutants in this locus ( Table 2) were tested for their ability to be inhibited by fosmidomycin and analogs.

Hypothesis
The aim of this research is to test the effectiveness of lipophilic FR900098 prodrugs against F. novicida as potential platforms for novel antibiotic development. Our hypothesis is that these new compounds will be more effective at crossing biological membranes than FR900098 or fosmidomycin, and thus may be more effective antimicrobial compounds against Francisella. Thus, we determined the minimum inhibitory concentration (MIC) and EC 50 for fosmidomycin, FR900098, and compounds 1-3 against F. novicida. In addition, the in vitro inhibition of F. tularensis LVS DXR by FR900098 was determined in comparison to fosmidomycin [1]. The ability of the lipophilic compounds to cross the Francisella membrane was examined using a glpT mutant of F. novicida, the transporter required for fosmidomycin activity in many bacteria. We then assessed the efficacy of fosmidomycin, FR900098, and compound 1 in treating intracellular F. novicida in A549 human alveolar epithelial cells and RAW264.7 mouse macrophages, as well as intracellular glpT mutants in A549 cells. The cytotoxicity of the compounds was determined by measuring the LDH released from the infected cells. Finally, G. mellonella was used as a model system to test the efficacy of fosmidomycin, FR900098, and compound 1 against F. novicida infection in vivo. As a lipophilic GlpT-independent DXR inhibitor, compound 1 has the potential to be a broad-spectrum antibiotic, and should be effective against many MEP-dependent organisms [16].

Results
2.1 Susceptibility of F. novicida to fosmidomycin, FR900098, and analogs 2.1.1 MICs and EC 50 s of fosmidomycin, FR900098, and lipophilic analogs. The susceptibility of F. novicida to the FR900098 analogs was determined in a 96-well plate assay. An initial screening of a panel of compounds demonstrated that compound 1 and compound 2 were the best inhibitors of the analogs tested (data not shown). Compound 3 did not effectively inhibit F. novicida, but its structure, along with the others, is included in Figure 1B for comparison.
The MIC values of fosmidomycin, FR900098, compound 1, and compound 2 were determined against F. novicida ( Table 3). The MIC was best for fosmidomycin at 136 mM. The MIC of FR900098 was 254 mM. The MICs of compound 1 and compound 2 were 202 mM and 1.1 mM, respectively. Thus, compound 1 had approximately 2-fold weaker activity compared with fosmidomycin, but was more active than FR900098.
These compounds were assessed for the number of surviving bacteria at increasing concentration of compound, from which we calculate the EC 50 , a value we use to compare activities of the different compounds. This value was determined for fosmidomycin, FR900098, compound 1, and compound 2. The EC 50 of fosmidomycin against F. novicida was determined to be 3.6 mM ( Figure 2A). The EC 50 of FR900098 was 23.2 mM ( Figure 2B). The EC 50 of compound 1 was 45.2 mM ( Figure 2C) and that of compound 2 was 481 mM ( Figure 2D). While compound 1 had a better MIC than FR900098, it had a slightly poorer EC 50 .

Inhibition of purified F. tularensis LVS DXR by
FR900098. The IC 50 of FR900098 against the DXR of E. coli is reported to be 62 nM [25]. The activity of FR900098 was determined by monitoring the enzyme-catalyzed oxidation of NADPH, and using a plot of enzyme fractional activity as a function of inhibitor concentration, as we previously reported, to determine the IC 50 (Table 4) [1]. The half maximal inhibition of F. tularensis LVS DXR by fosmidomycin was previously determined to be 247 nM [1]. FR900098 was found to have an IC 50 of 230 nM ( Figure S1). Since compound 1 is metabolized to FR900098 in bacteria [10], we did not test the activity of compound 1 against DXR in vitro. DXR from Francisella novicida (FTN_1483), F. tularensis LVS (FTL_0534), and F. tularensis Schu S4 (FTT1574) share .99% homology ( Figure S2). The highlighted differences are not in critical enzymatic residues, thus we conclude that DXR from all Francisella species will have similar sensitivities.
2.1.3 Susceptibility of F. novicida glpT mutants to fosmidomycin, FR900098, and lipophilic analogs. Fosmidomycin requires the glycerol-3-phosphate transporter (GlpT) to enter bacterial cells [6]. Bioinformatic analysis suggests that all Francisella species contain GlpT homologs ( Table 1). One of the goals of this work was to test compounds that were more efficient at crossing biological membranes independent    of the GlpT transporter, presumably by increased penetration of the membrane due to increased lipophilicity. The data for the glpT-1 mutant strain is shown here (Figure 3), and the data for the glpT-2 mutant strain was very similar. Even at a concentration of 1 mM, fosmidomycin was not able to inhibit F. novicida glpT growth at all (Table 3). At this same concentration, FR900098 was only able to inhibit 50% of F. novicida glpT growth. These concentrations inhibited 100% of wild-type F. novicida growth.
Since inhibition of bacterial growth was reduced against the glpT mutant, FR900098 is at least partially dependent on this transporter to enter the bacterial cell. It is possible that FR900098 uses another transporter system, accounting for its 50% inhibition of the glpT mutant. Alternatively, FR900098 may be slightly lipophilic and have some ability to cross the membrane independent of GlpT. Compound 2 demonstrated reduced bacteriostatic activity against the glpT mutant than wild-type F. novicida, suggesting that this analog may also be partially dependent on this transporter to cross the membrane, or a small amount of compound 2 is able to diffuse across the membrane. Both FR900098 and compound 2 are more lipophilic compared with fosmidomycin. In contrast, the MIC of compound 1 was the same for both wild-type F. novicida and the glpT mutant. This indicates that compound 1 entry into the bacteria is independent of the transporter, and is most likely due to the increased lipophilicity of the acyloxyalkyl ester and its ability to cross the membrane.

Cell line infections and treatments
2.2.1 Infection and treatment of human type II alveolar epithelial (A549) cells. When tested at an equal concentration (250 mM), all three antibiotics significantly inhibited the growth of intracellular F. novicida (p,0.05) ( Figure 4A). Untreated cells contained 3.8860.72610 3 CFU/well of F. novicida. At this concentration, fosmidomycin treated cells contained 0.0760.03610 3 CFU/well of F. novicida (p,0.0001), representing a 2-log reduction of intracellular bacteria. Cells treated with FR900098 and compound 1 contained 0.5660.26610 3 CFU/ well and 0.1360.03610 3 CFU/well, respectively (p,0.0001). In this experiment, fosmidomycin was significantly more effective than both FR900098 and compound 1 (p = 0.005 and 0.006, respectively), likely due to the fact that it was dosed at twice its MIC in the extracellular media. Compound 1 was significantly more effective at inhibiting intracellular F. novicida than FR900098 (p = 0.009), likely due to compound 1's extracellular concentration being slightly higher than its MIC (202 mM), while FR900098's extracellular concentration was just at its MIC (254 mM). Although the penetration rate of these compounds across the eukaryotic cell membrane is not precisely known, these results suggest that it was high enough for compound 1 to cross the eukaryotic membrane to achieve intracellular concentrations that approach the MIC.
In a second experiment, each compound was tested at twice the MIC (26 MIC) to encourage maximal diffusion of compound across the eukaryotic cell membrane ( Figure 4A). At 26 MIC, compound 1 was as effective as fosmidomycin at inhibiting intracellular F. novicida, and was more effective than FR900098. Fosmidomycin at 250 mM was already at 26 MIC, thus its performance did not change. Under these 26 MIC conditions, compound 1 (400 mM) inhibited F. novicida intracellular growth to a much greater extent (0.0560.03610 3 CFU/well intracellular bacteria), and was as effective as fosmidomycin. FR900098 (500 mM) did not significantly inhibit more intracellular bacteria (0.5960.40610 3 CFU/well) than at 250 mM, and was the worst performing compound in this assay, suggesting that FR900098 may have some relative inability to penetrate the eukaryotic cell membrane at high efficiency, unlike compound 1. These results demonstrate the antimicrobial activity of compound 1 against F. novicida was better than FR900098 and as effective as fosmidomycin.  was as effective as fosmidomycin (p = 0.94) and significantly more effective than FR900098 (p,0.001) ( Figure 4B).

2.2.3
Infection of A549 cells with the glpT mutant of F. novicida and treatment with fosmidomycin, FR900098, and compound 1. To determine if the glpT mutant could be used as an intracellular model for these compounds, A549 cells were infected with F. novicida and the glpT mutant as previously described. The glpT mutant was able to infect cells as efficiently as wild-type (p = 0.10). Fosmidomycin and FR900098 had no effect on the intracellular bacteria (p = 0.12) ( Figure 4C). Compound 1 demonstrated a significant effect on the intracellular glpT mutant. Untreated cells infected with the glpT mutant contained 514.006206.47610 3 CFU/well, while cells treated with compound 1 contained 22.3262.23610 3 CFU/well (p,0.01) ( Figure 4C).

LDH assay for determination of cytotoxicity of
antibiotics. The release of lactate dehydrogenase (LDH), a stable cytoplasmic protein, can be visualized and quantified in the cellular supernatants when it interacts with a tetrazolium salt (INT) to form a red formazan product. This can be used to measure the percent cytotoxicity of antibiotics against eukaryotic cells. None of the three antibiotics (fosmidomycin, FR900098, compound 1) causes significant cytotoxicity of human lung epithelial A549 cells or mouse macrophage RAW264.7 cells (p.0.05). Treatment did not induce cytotoxicity in either infected or uninfected cells (1.5-4% LDH release, compared with 2.8% LDH release for the control).

G. mellonella infection and treatment
G. mellonella larvae were infected with 3610 4 CFU of F. novicida by injection into the left proleg. After incubation for 2 hours at 37uC, the larvae were treated with the appropriate antibiotic by injection of 10 ml into the right proleg, as previously described [26,27]. Fosmidomycin, FR900098, and compound 1 were tested at 30 mg/kg (9 mg per caterpillar), which is the dose that has been used in mice to treat malaria [13]. Ciprofloxacin was used at 20 mg/kg (6 mg per caterpillar) as a positive control [26]. All antibiotics were effective at prolonging the survival of F. novicida infected larvae ( Figure 5). The mean time to death of untreated larvae was 59 hours. The mean times to death of caterpillars treated with fosmidomycin, FR900098, or compound 1 were 102 hours (p = 0.001), 84 hours (p = 0.009), and 93 hours (p = 0.0006), respectively. The mean time to death for caterpillars treated with ciprofloxacin was 103 hours (p,0.0001). There was no significant difference between the survival of fosmidomycin, compound 1, or ciprofloxacin treated larvae.

Discussion
The two phosphonate antibiotics, fosmidomycin and FR900098, are both able to effectively inhibit F. novicida. Fosmidomycin is able to inhibit F. novicida replication in vitro with an MIC of 136 mM and an EC 50 of 3.5 mM. The action of fosmidomycin was confirmed to be absolutely GlpT-dependent in F. novicida. FR900098 also inhibited F. novicida replication in vitro with an MIC of 254 mM and an EC 50 of 23 mM. However, the action of FR900098 is at least partially GlpT-dependent. Fosfomycin, an antibiotic related to fosmidomycin and FR900098, uses both GlpT and its close homolog, the glucose-6phosphate transporter (UhpT) in E. coli [28]. A BLAST searched revealed that the UhpT transporter is not found in Francisella.
The benzoyloxyethyl ester prodrug of FR900098 (compound 1) inhibited in vitro replication of F. novicida with an MIC of 202 mM and an EC 50 of 45 mM. Compound 1 was found to be entirely GlpT-independent in its activity, illustrating a significantly improved and novel property of this compound for antimicrobial activity against Francisella. This property of compound 1 could be particularly important for bacteria that lack GlpT, such as M. tuberculosis. This compound gets metabolized to FR900098 once inside the bacteria.
Compound 2 is similar to compound 1, with a hexyl ester replacing the benzoyloxyethyl ester. Compound 2 was not as effective against F. novicida in vitro (MIC ,1 mM, EC 50 481 mM), and its action was partially GlpT-dependent. Thus, in vitro compound 2 is less effective than compound 1 and does not demonstrate improved antimicrobial properties against F. novicida compared to compound 1. Compound 3 was ineffective against F. novicida.
Fosmidomycin is an effective inhibitor of DXR, but may not be an optimal drug candidate because it does not penetrate bacterial membranes without a transporter. Fosmidomycin requires the GlpT transporter to carry it across the membrane to the internal cytoplasm of the bacteria, where it can target DXR. Some bacteria that do not have the GlpT transporter, such as Brucella, are relatively unaffected by fosmidomycin [4]. Sangari et al. showed that by introducing the GlpT transporter to this organism, the recombinant B. abortus became sensitive to fosmidomycin [4]. M. tuberculosis (TB) does not have the GlpT transporter, and this may be one of the reasons that fosmidomycin is ineffective against M. tuberculosis in vitro [6], in addition to the impenetrability of the mycolic acid layer. Indeed, for F. novicida, the GlpT mutants were completely resistant to 1 mM fosmidomycin, indicating that the efficacy of this compound is significantly GlpT dependent. The acetyl analog, FR900098, had better performance in this regard against F. novicida, with approximately 50% inhibition against the glpT mutant, compared to 100% inhibition of the wild-type F. novicida. The lipophilic prodrug of FR900098, compound 1, demonstrated complete GlpT independence, inhibiting 100% of both the wild-type and glpT mutant F. novicida. Although its efficacy against Francisella in the mouse model is not yet known, our results in G. mellonella suggest that compound 1 is as effective as FR900098. This finding is similar to those from Ortmann et al., who found that this same compound (compound 1) had Figure 4. Inhibition of intracellular F. novicida replication in two cell lines following treatment with selected compounds. Cell lines were first infected with F. novicida at an MOI of 500. The cells were treated with the following concentrations of antibiotics. The cells were lysed after 20 hours of treatment and the intracellular bacteria were enumerated. A) Inhibition of intracellular F. novicida in A549 cells with fosmidomycin (Fos), FR900098 (FR), or compound 1 (C1) for 20 hours. Each compound was tested at 250 mM and at 26 MIC (26 MIC for fosmidomycin = 250 mM, 26 MIC for FR900098 = 500 mM and 26 MIC for compound 1 = 400 mM). All three compounds significantly inhibited intracellular F. novicida growth. At 250 mM, intracellular growth was inhibited 98.060.7% by fosmidomycin, 8666% by FR900098, and 97.060.8% by compound 1. At 26 MIC, FR900098 (500 mM) inhibited 85610% of intracellular growth, while compound 1 inhibited 99.060.7% of intracellular growth. B) Inhibition of F. novicida in RAW264.7 cells with fosmidomycin, FR900098, and compound 1 for 20 hours. Similar results were seen for the RAW264.7 cells as were seen for the A549 cells. C) Inhibition of the F. novicida glpT mutant intracellular replication in A549 cells. The intracellular-replication inhibition experiment was performed using the glpT mutant as previously described for wild-type F. novicida. Replication of intracellular glpT mutant was not affected by fosmidomycin (250 mM) and FR900098 (500 mM), but was susceptible to compound 1 (400 mM).   antimalarial activity in the mouse model that was similar to, but not better than, FR900098 [10].
A useful result of this study is the development of the glpT mutant of F. novicida as a model for testing the bacterial-cell penetrating ability of lipophilic fosmidomycin and FR900098 analogs. The glpT mutant of F. novicida may serve as a model organism to easily screen compounds to identify those analogs with improved membrane penetrating ability (both eukaryotic and prokaryotic membranes). GlpT has not previously been identified as a significant virulence factor of F. tularensis using various screens for virulence, intracellular replication, or in vivo infection [29]. Other comprehensive studies of genes involved in Francisella pathogenesis or intracellular replication have not identified GlpT homologs in any species of Francisella as playing a role in these processes [29,30]. We have also shown that glpT mutants can infect and replicate within human lung epithelial A549 cells. Thus, there is no overt role of GlpT in Francisella pathogenesis or intracellular replication, supporting the usefulness of this model as a potential screening tool for lipophilic DXR inhibitors.
Our studies with human lung epithelial cells (A549) and mouse macrophages (RAW264.7) demonstrate that fosmidomycin, FR900098 and compound 1 can cross eukaryotic cell membranes efficiently. The LDH assay demonstrates that these compounds are not cytotoxic, suggesting that reduced bacterial counts are due to the antibiotics inhibiting intracellular bacteria, and not due to eukaryotic cell death. Although we did not quantitatively measure the intracellular concentration of these compounds, we can conclude that when concentrations greater than or equal to the MIC are applied extracellularly, the MIC was achieved inside of the eukaryotic cells, as indicated by the complete (fosmidomycin, compound 1) or significant (FR900098) reduction in intracellular bacterial growth. This information suggests that our model using F. novicida glpT mutants to screen membrane-penetrating analogs of fosmidomycin can be expanded to be an intracellular infection model, as illustrated in Figure 6.
Our results with the in vivo G. mellonella model suggest that fosmidomycin, FR900098, and compound 1 could be effective at treating Francisella infections in mammals. The G. mellonella infection system has been used to study many human pathogens, including Francisella [26,27]. G. mellonella demonstrates a humoral immune response similar to that of mammals, and this response is carried out by hemocytes within the hemolymph of the caterpillar [26]. The host defense response includes phagocytosis, nodulation, and melanization [26,31]. In the mouse model, 30 mg/kg of fosmidomycin has been used to treat Plasmodium infections [13]. This translates to approximately 9 mg per caterpillar. At these concentrations, fosmidomycin, FR900098, and compound 1 were able to prolong the survival of F. novicida-infected caterpillars. This suggests that in a mammalian model, these antibiotics could be used to treat Francisella infections.
Although the activity of compound 1 against wild-type F. novicida is relatively similar to that of fosmidomycin and FR900098, its GlpT-independent activity is of particular interest, especially when considering the problem of antibiotic resistance [32]. For many species of bacteria that cause significant infectious disease, there exists strains that are resistant to commonly used antibiotics [33]. Despite an increase in multiple drug resistant (MDR) pathogens, there has been a dramatic decrease in antibiotic research performed by pharmaceutical companies [34]. No successful discovery of a novel antibacterial agent has occurred since 1987 [35]. Furthermore, resistance to every main class of antibiotic arises within one to ten years of their introduction to clinical use [36], indicating that antibiotic resistance must be planned for.
Antibiotic resistance can occur through many different mechanisms, including the inactivation or modification of drugs, the alteration of the drug target, or efflux pumps that remove drugs from the cytoplasm [37]. Studies suggest that for naturally occurring antibiotics, the rate of resistance is higher than that of synthetic compounds, but horizontal gene transfer can spread resistance for both types of compounds [35,38]. Thus there is a need for constant antibiotic discovery and improvement on current antibiotics to combat the resistance problem.
The importance of the GlpT-independent activity of compound 1 becomes apparent when considering the possible resistance mechanisms to fosmidomycin and FR900098 [32]. In organisms that utilize the MEP pathway, DXR is an essential enzyme, and mutations are lethal [5,6] so it is considered a validated drug target. This indicates that antibiotics that target DXR, such as fosmidomycin and FR900098, could potentially be excellent candidates for broad-spectrum use. DXR is an essential enzyme, and while mutations could theoretically occur in the active site that would alter drug activity, loss of this enzyme is unlikely. However, the same cannot be said for the GlpT transporter, as demonstrated by our use of a glpT mutant [5,32]. GlpT is an inorganic phosphate/glycerol-3-phosphate antiporter that depends on the phosphate gradient, so any environmental conditions that affect this gradient could also affect the activity of the transporter [39]. As stated previously, organisms such as B. abortus and M. tuberculosis use DXR, but lack GlpT [4,14], thus these compounds may also have utility against these organisms. Figure 6. Model for screening method to identify lipophilic, fosmidomycin-derived analogs effective against intracellular pathogens. This is the model for a new screening method for a library of fosmidomycin/FR900098 analogs against host cells infected with intracellular bacteria to identify lipophilic derivatives that can cross both eukaryotic and prokaryotic membranes. In this example, the mammalian cell (orange line) is infected with intracellular bacteria, F. novicida (green GlpT) or F. novicida glpT mutant (red GlpT) separately. If intracellular bacterial growth of F. novicida glpT mutant is inhibited by a compound (inhibition of the DXR enzyme), the fosmidomycin analog is likely able to cross both eukaryotic (orange) and prokaryotic (blue) cell membranes. Such analogs would be good candidates for further testing in other models such as host cells infected with TB. If the analog does not inhibit growth, it may be GlpT-dependent for bacterial cell entry, and thus could not reach the intra-bacterial DXR enzyme in the glpT mutant. This would be verified by further testing against wild-type F. novicida infected host cells. Thus, by screening eukaryotic cells infected with F. novicida glpT mutants, we are able to simultaneously screen for the three critical functional properties of the desired compound (eukaryotic & prokaryotic membrane penetration and GlpT independence). doi:10.1371/journal.pone.0038167.g006 A different gene has been identified in E. coli that confers resistance to fosmidomycin. It has been labeled fsr and is homologous to other drug efflux proteins, such as those that export tetracycline and chloramphenicol [40]. Interestingly, this gene did not confer resistance to fosfomycin, another phosphonate antibiotic dependent on GlpT [40]. A BLAST search revealed that this protein is not present in F. tularensis.
Drug targets can be upregulated as a mechanism of resistance. P. falciparum that are resistant to fosmidomycin have been shown to upregulate the target enzyme, DXR [41]. This has also been seen as a resistance mechanism of bacteria to fosfomycin, which is in the same class of antibiotic [35]. However, it has recently been shown that fosmidomycin also targets the enzyme following DXR in the MEP pathway, 2-C-methylerythritol-4-phoshpate cytidyltransferase (or IspD) in E. coli and P. falciparum [42]. This is a promising discovery, as there is a lower frequency of resistance to antibiotics that have multiple targets [35].
Another challenge with both fosmidomycin and FR900098 as therapeutics is their low bioavailability in the serum, probably as a result of their low lipophilicity [10]. The ionization of the phosphonate groups of fosmidomycin and FR900098 at physiological pH is believed to contribute to the low bioavailability of these antibiotics [10,43]. The plasma concentration of FR900098 was more than twice as high in mice after the administration of an FR900098 prodrug, as opposed to FR900098 on its own [10]. Analogs with higher bioavailability in the serum may be important if bacteria become resistant by upregulating the target enzyme. An antibiotic that has fosmidomycin-or FR900098-like activity, but can freely cross cell walls and membranes, with increased bioavailability, would be an excellent contribution to current therapeutics.
In conclusion, we have shown that the FR900098 prodrug, compound 1, is effective at inhibiting the growth of F. novicida, both in vitro and in vivo. Compound 1 can inhibit intracellular F. novicida growth as well as fosmidomycin, and better than FR900098. This prodrug enters the bacterial cell entirely independent of GlpT. Our studies with the in vivo model organism, G. mellonella, indicate that compound 1 may be effective at clearing mammalian Francisella infections. Its effectiveness at treating mice infected with P. falciparum further supports this hypothesis [10]. Our collaborators have recently demonstrated the effectiveness of compound 1 against a variety of pathogens [16]. Thus, compound 1 is an effective lipophilic, GlpT-independent prodrug of FR900098 with the potential to be used as a broad-spectrum antibiotic.

Growth of bacteria and tissue culture cells
F. novicida (# NR13, BEI Resources, Manassas, VA) was grown in Trypticase Soy Broth with 0.1% cysteine HCl (TSB-C), on TSB-C agar plates, or Chocolate II Agar plates (BD Biosciences) as noted for each assay. Francisella on plates was grown in a 37uC incubator with 5% CO 2 . F. novicida was grown from bacterial stocks stored in 20% glycerol in a 280uC freezer.

Identification of the GlpT homolog in
Francisella. The GlpT coding region (GlpT) was identified in the F. tularensis SchuS4 (FTT0725c) and F. novicida (FTN_0636) genome (accession numbers YP_169738.1 and YP_898283.1) via a BLAST search using the E. coli K12 homologous sequence (accession number NP_416743.1) as the query [10]. Transposon insertion mutants in this locus (FTN_0636, Table 1) were obtained through the NIH Biodefense and Emerging Infectious Disease Research Resources Repository, NIAID, NIH: Francisella tularensis subsp. novicida, ''Two-Allele'' Transposon Mutant Library. The F. novicida glpT mutants were grown as above, with the addition of 10 mg/ml of kanamycin to select for the transposon mutant.
3.2.2 Stock solutions of Fosmidomycin, FR900098, and compounds 1-3. All antibiotic stocks were made to have a high starting concentration between 10 and 20 mg/ml. Fosmidomycin (Invitrogen #F-23103) and FR900098 (Sigma-Aldrich # F8307) were obtained as dry powders and were dissolved in water. Compounds 1-3 were dissolved in 100% DMSO to the final concentrations indicated. Gentamicin (Cellgro #61-098-RF), used for cell infection assays, was dissolved in water to a final concentration of 50 mg/ml and was diluted either in Ham's F12 media or RPMI.
3.2.3 Determination of MIC and EC 50 . The minimum inhibitory concentration of an antibiotic compound is the minimum concentration that will completely inhibit all visible growth of bacteria in culture media (bacteriostatic concentration). The EC 50 is the concentration of compound that will inhibit 50% of bacterial growth. Bacteria were grown in a 96-well flat bottom plate to determine both of these values. To begin the assay, the appropriate amount of compound (or antibiotic) was added to 1610 5 CFU/150 ml of F. novicida in TSB-C to obtain a final volume of 1 ml and final concentration of compound to 200 mg/ ml. The DXR inhibitors were first screened for bacterial growth inhibition by adding 150 ml of the bacteria/antibiotic mixture to a 96-well plate in three separate wells. The plate was incubated at 37uC for 48 hours and read for the MIC assay. Those compounds that inhibited growth were further tested. For the EC 50 assay, 300 ml of the bacteria/compound mixture were added to three wells. In all other wells there was 1610 5 CFU/150 ml of F. novicida. 1:2 serial dilutions were performed in the plate. The plate was wrapped in tin foil to protect it from the light and placed at 37uC with 5% CO 2 for 24-48 hours. The OD 600 is read before the plate was placed in the incubator and after 24 or 48 hours.
3.2.4 Data analysis. Data were analyzed using the following equation and GraphPad Prism 4 (GraphPad Software Inc., San Diego, CA).
Y corresponds to bacterial mortality (% OD, where zero drug = 100%) at a given antibiotic concentration (mg/ml), with X being the logarithm of that concentration (log mg/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 a standard sigmoidal dose-response curve (Equation 1) with a Hill slope of 1. Errors were reported based on the standard deviation from the mean of the log EC 50 values.

FR900098 IC 50 determination
The F. tularensis LVS DXR was cloned, expressed in E. coli, purified, and assayed essentially as described previously [1]. All assays were performed in duplicate, with the 1-deoxy-D-xylulose 5-phosphate (DXP; Echelon Biosciences, Salt Lake City, UT) concentration fixed at the K M (104 mM) and a saturating concentration of NADPH (150 mM). A plot of enzyme specific activity as a function of inhibitor concentration was nonlinear regression fit to a sigmoidal dose-response curve to determine the half-maximal inhibition value (IC 50 ) of FR900098. Because fosmidomycin is known to be a slow, tight binding inhibitor of DXR [2], the enzyme was pre-incubated with FR900098 for 10 minutes prior to addition of substrate.

Quantification of intracellular F. novicida
3.4.1 Seeding a 96-well plate. A549 and RAW264.7 cells were grown to 75-90% confluence, and then passaged as previously described. Three T75 flasks of cells were combined and the cells were counted using a hemocytometer [49]. They are then spun down (2506 g) and resuspended in media for a final concentration of 1610 5 cells/100 ml. Cells are added to a 96-well plate so that each well contained 1610 5 cells. The plate was incubated overnight (less than 24 hours) at 37uC with 5% CO 2 , allowing the cells to adhere to the plate.

Infection of cells with F. novicida and treatment
with antibiotics. After overnight growth, A549 or RAW264.7 cells were infected with an overnight culture of F. novicida at a multiplicity of infection (MOI) of 500. The MOI is the number of infectious agents divided by the number of cells to be infected. After F. novicida is added to the cells, the plate was incubated for 2 hours at 37uC with 5% CO 2 . Extracellular bacteria were then removed by washing the plate 2 times with the appropriate cell culture media. Cells were then incubated for 1 hour with 50 mg/ ml of gentamicin to eliminate extracellular bacteria. After this incubation period, cells were again washed twice with media and treated with fosmidomycin, FR900098, or compound 1 at the indicated concentrations. This was the 0 hour time point. The antibiotic solutions also contained 2 mg/ml of gentamicin to prevent extracellular bacterial growth, ensuring that only intracellular bacteria would be counted. The controls (untreated and uninfected) received the 2 mg/ml gentamicin solution with an appropriate amount of DMSO, to account for any effect of the DMSO in the compound 1 solution.

3.4.3
Quantification of intracellular F.
novicida. Intracellular F. novicida was quantified at 0 and 20 hours. For the 0 hour time point the concentration of intracellular F. novicida was determined immediately after the addition of the antibiotics. For 20 hour time point, the cells were incubated for an additional 20 hours at 37uC, 5% CO2. The media was aspirated and the cells were washed twice with 16PBS. Cells were lysed to release the intracellular bacteria by the addition of 100 ml of sterile distilled water and serially diluted in 16 PBS. Dilutions were plated on TSB-C agar plates and incubated at 37uC with 5% CO 2 for 24 hours. Colonies were counted to determine the CFU/well of intracellular bacteria that remained following each treatment.

LDH assay for determination of cytotoxicity of
antibiotics. The cytotoxicity of antibiotics against eukaryotic cells can be determined by measuring the release of cytoplasmic proteins, which is a sign of apoptosis. Lactate dehydrogenase (LDH) is a stable cytoplasmic protein that is released when the cell is lysed. LDH can be visualized and quantified in the cellular supernatants when it interacts with a tetrazolium salt to form a red formazan product, which has absorbance at 490 nm.

Galleria mellonella infection and treatment
G. mellonella (wax moth caterpillars) larvae were used to test the in vivo effect of fosmidomycin, FR900098, and compound 1 against F. novicida infection. G. mellonella larvae were used in their final larval stage and stored at room temperature in the dark. Ten caterpillars of 0.30-0.35 g were used in each group. The following control groups were used for each experiment: caterpillars that received no injection, caterpillars injected with 16 PBS to determine any mortality associated with the trauma of injection, caterpillars injected with PBS and antibiotics to determine any mortality caused by antibiotics alone, and caterpillars injected with bacteria and ''treated'' with 16 PBS. All injected caterpillars received two injections. A 1 cc tuberculin syringe was used for all injections. The syringe was used to inject 10 ml of 3610 6 CFU/ml of F. novicida into the hemocoel via the left proleg of all caterpillars to be infected. Control caterpillars were injected with 10 ml of 16 PBS in the left proleg. Caterpillars were then incubated at 37uC for 2 hours. After the incubation period, the next round of injections was performed using the right proleg. Control caterpillars received either 10 ml of 16PBS or 10 ml of antibiotic. Treated caterpillars received 10 ml of antibiotic. Fosmidomycin, FR900098, and compound 1 were tested at the concentrations indicated in the results section. Live caterpillars were counted daily. The mean time to death was determined for each group.

Statistical analysis
Two tailed T-tests were performed to determine statistical significance of bacterial infection and LDH release data, assuming unequal variance in the different populations. For in vivo testing in G. mellonella, Kaplan-Meier survival analysis was used (http:// www.medcalc.be/). Figure S1 IC 50 of FR900098 against DXR. Nonlinear regression fitting the sigmoidal dose-response curve resulted in an IC 50 of 230 nM, slightly more potent than fosmidomycin (EC 50 = 247 nM) [1]. All assays were performed in duplicate. The R 2 value is indicated. (TIFF) Figure S2 Amino acid alignment of DXR from Francisella species. The DXR from Francisella novicida (FTN_1483), F. tularensis LVS (FTL_0534), and F. tularensis Schu S4 (FTT1574) share .99% homology. The highlighted differences are not in critical enzymatic residues. (TIFF) Author Contributions