Francisella tularensis subsp. novicida Utah 112 (F. novicida; BEI Resources # NR13) was grown in trypticase soy broth with 0.1% cysteine HCl (TSB-C), on TSB-C agar plates, or Chocolate II Agar plates (CHAB – BD Biosciences) as noted for each assay. F. tularensis subsp. tularensis NIH B38 (ATCC 6223; BEI Resources # NR50, deposited as the type strain for F. tularensis tularensis) was grown on TSB-C agar plates at 37°C with 5% CO₂ for 48 h, harvested by scraping into 20% glycerol, and stored at −80°C. Francisella on plates was grown at 37°C in an atmosphere of 5% CO₂. Minimum inhibitory concentration (MIC) assays were performed in cation-adjusted Mueller–Hinton broth (MHB) according to the guidelines published by the Clinical and Laboratory Standards Institute; M11-A7.

Transposon insertion mutants in the glpT locus [FTN_0636; F. novicida tnfn1_pw060323p08q150 BEI Catalog #NR-5683 and F. novicida tnfn1_pw060418p01q161 BEI Catalog #NR-6558, both Type T20 (ISFn2/FRT)] (Gallagher et al., 2007) were obtained through the NIH Biodefense and Emerging Infectious Disease Research Repository, NIAID, NIH: F. novicida, “Two-Allele” Transposon Mutant Library. F. novicida U112 glpT-1 and glpT-2 (data not shown) transposon insertion mutants were grown as above, but in the presence of kanamycin (10 μg/ml) to select for the mutants. Sequences for the transposon mutants are shown in Figure A1 in Appendix (Gallagher et al., 2007).

All inhibitor stocks were made at high starting concentration between 10 and 20 mg/ml. Fos sodium salt (Invitrogen #F-23103) and FR900098 monosodium salt (Sigma–Aldrich # F8307) were obtained as dry powders and were dissolved in water. Compound 1 (kind gift from C. Dowd) was dissolved in 100% DMSO to the final concentrations indicated.

Following the Kirby–Bauer disk diffusion assay protocols (Bauer et al., 1966), inhibitor-saturated disks were placed on a lawn of bacteria to display a zone of growth inhibition around the disk indicating susceptibility. A concentration of an overnight culture of bacteria equal to a 0.5 McFarland standard was spread on Chocolate II Agar plate (BD Biosciences) using a sterile cotton swab. Inhibitor disks were made by absorbing 20 μl of each inhibitor stock (10 mg/ml) to a sterile, thick Whatman filter paper 6-mm disk, so that each disk contained 200 μg. Three disks of the same inhibitor were added to each plate, except for NIHB38, for which only one disk was used per plate. For NIHB38 experiments, both 200 and 100 μg per disk were tested. The plates were wrapped in tin foil to protect the inhibitors from light and placed in the 37°C incubator for 48 h. The zone of inhibition was determined by measuring the diameter of the zone of no growth, including the 6-mm disk. Each zone was measured three times at different points of the zone (n = 9). A 6-mm zone of inhibition indicates no inhibition.

In accordance with CLSI standards, 150 μl cation-adjusted MHB was inoculated with 1.5 × 10⁵ CFU of F. novicida. For MIC assays against wild-type F. novicida, each well of column 1 was amended with 150 μl of Fos or compound 1 to achieve a final concentration of 300 and 200 μg/ml, respectively. Subsequent MIC assays had a starting concentration of 200 μg/ml regardless of the compound tested. Twofold serial dilutions of column 1 were performed to fill columns 2–11 of a 96-well plate. For compound 1, DMSO vehicle control wells were included in side-by-side assays. Column 12 received no drug and acts as the positive growth control for the assay. Absorbances (A₆₀₀) were read for each well before incubation to record background. Plates were incubated at 37°C plus 5% CO₂, A₆₀₀ were determined at 24 and 48 h, and IC₅₀ and MIC were determined.

Overnight cultures were centrifuged to form bacterial pellets and supernatants were removed. Genomic DNA was prepared using the Wizard Genomic DNA kit (Promega, Madison, WI, USA). Briefly, pellets were re-suspended in 400 μl of nuclei lysis solution and incubated at 80°C for 15 min to lyse bacteria. Total DNA concentrations were analyzed using the Nanodrop spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and adjusted to 40 ng/μl. Primers used for PCR confirmation that colonies were indeed F. novicida were designed to amplify 16S rRNAs (16SrRNAF: CTGTCGTCAGCTCGTGTTGT; 16SrRNA R: CGTAAGGGCCATGATGACTT; Forsman et al., 1994). Primers used to amplify the entire ORF of the glpT gene (glpTF: AACAGCGGTTTAGCTATTTTCAA; glpTR: TGCAATCTGAGCTGACTGAAG) were designed using Primer 3 through the National Center for Biotechnology Information (NCBI). Real-time PCR was performed on the Biorad MyIQ instrument with iQ SYBR Green Supermix in 50 μl reactions (Biorad, Hercules, CA, USA). Thermal cycler profile was set to 95°C for 3 min for initial denaturation and enzyme activation followed by 40 cycles of 95°C for 10 s and 55°C for 30 s. Sequencing primers were the same as the ORF amplification primers, but were synthesized and used by Functional Biosciences (Madison, WI, USA).

A stock culture of F. novicida U112 was grown in TSB-C broth overnight at 37°C with continuous shaking at 165 rpm. Viable bacteria from overnight cultures were enumerated by serial dilution in sterile phosphate-buffered saline (PBS) and plated on TSB-C non-selective agar plates. TSB-C plates were incubated for >24 h, and colony forming units were counted to determine bacterial density of the culture. A diluted subculture was created to achieve a final concentration of 1 × 10³ CFU/ml. Parallel cultures prepared as 1 ml aliquots from the subculture (1 × 10³ CFU) were incubated overnight at 37°C with continuous shaking at 165 rpm (Rosche and Foster, 2000). The number of mutations that resulted in Fos^R was determined by plating 0.1 ml aliquots of the parallel cultures (n = 5) on TSB-C plates containing 122 μM (25 μg/ml) Fos. The total number of bacteria in the cultures was determined by serial dilution and plating on non-selective media (Rosche and Foster, 2000).

The Kirby–Bauer disk-inhibition assay was used to determine the susceptibility of wild-type F. novicida U112 to Fos and FR900098. The zone of inhibition around the disk containing 200 μg of Fos was 37 ± 1 mm (Figure 2A). The zone of inhibition around the disk containing 200 μg of FR900098 was 29 ± 2 mm (Figure 2C). For F. tularensis NIHB38, the type strain for Type A tularemia, both of these compounds lead to a large zone of inhibition. For a disk with 100 μg Fos, the zone of inhibition for F. tularensis NIHB38 was 72 ± 2 mm (data not shown), suggesting that F. tularensis Type A strains are sensitive to Fos. A zone of inhibition of 6 mm (the size of the disk) indicated no inhibition, which was observed for a water-only disk (data not shown).

While performing the Kirby–Bauer disk-inhibition assay with F. novicida against Fos and FR900098, breakthrough colonies were observed growing inside the zone of inhibition (Figure 2A – Fos, Figure 2C – FR900098). However, the zone of inhibition is still clearly defined. The colonies within the zone of inhibition are slightly different in their appearance than the wild-type F. novicida that is growing outside the zone of inhibition. The colonies are slightly larger and more mucoid, but are the same color as the wild-type colonies. These colonies (from Figures 2A,C) were confirmed to be F. novicida by real-time PCR using F. novicida specific 16S rRNA primers (Forsman et al., 1994; data not shown). Numerous breakthrough colonies were also observed for F. tularensis NIHB38 with both of these compounds, indicating that this “type strain” for F. tularensis Type A strains may also develop spontaneous resistance to Fos.

Based on our results in the disk-inhibition assay as well as our previous studies (McKenney et al., in press), we hypothesized that the two most likely bacterial targets for the development of mutants would be mutations of the DXR enzyme itself or mutations of GlpT that allow entry of Fos into the bacteria.

We found that Fos^R mutants are resistant to Fos but are sensitive to compound 1, the lipophilic prodrug of FR900098. The MIC for Fos against wild-type F. novicida is 121 μM (IC₅₀ is 0.101 μg/ml with a 95% confidence interval (CI) = 0.079–0.130 mg/l; Figure 3A). The MIC for the lipophilic compound 1 against F. novicida is 100 μg/ml or 203 μM (IC₅₀ is 11.99 μg/ml, 95% CI = 9.84–14.62 μg/ml; Figure 3B). A Fos-resistant colony (Fos^R#2) isolated from our disk-inhibition assays was grown overnight as described, and the MICs and IC₅₀s of Fos and compound 1 were determined. Concentrations of inhibitor used in this assay ranged from 0.4 to 200 μg/ml. Fos^R#2-resistant bacteria grew well at all concentrations of Fos tested, with significant growth in all wells at or near the levels of no-drug control (Figure 3C). Conversely, the MIC of compound 1 for Fos^R#2 was 100 μg/ml or 203 μM (IC₅₀ of compound 1 for Fos^R#2 is 24.38 μg/ml, 95% CI = 18.75–31.70 μg/ml; Figure 3D). The MIC was identical between wild-type and Fos^R#2 and the IC₅₀s were very close. The IC₅₀ curves can almost be superimposed.

Fosmidomycin-resistant bacteria retained sensitivity to compound 1. Since compound 1 is a lipophilic prodrug it will get metabolized to FR900098 inside the bacteria, which will then target bacterial DXR. This suggests that spontaneous resistance is not due to mutation of DXR, since Fos^R bacteria are inhibited by compound 1. No transposon insertion mutants are available for DXR as this enzyme is thought to be essential to Francisella growth, so the role of DXR could not be directly tested in this study. We thus hypothesized that the mutations that lead to the Fos-resistant phenotype are not mutations in DXR. We then predicted that the GlpT transporter may be mutated, leading to this resistant phenotype. Supporting this, transposon insertion mutants at the glpT locus were also found to be resistant to Fos.

The susceptibility of the glpT transposon insertion mutants to Fos and FR900098 was also tested in this manner. There was no inhibition of the glpT transposon mutant by Fos (Figure 2B), supporting the critical role of glpT in Fos transport across the bacterial membranes in order for it to reach the DXR enzyme target inside the bacteria. glpT mutants with FR900098 disks demonstrated intermediate inhibition (15 ± 1 mm zone; Figure 2D). This effect of FR900098 on the glpT mutant is consistent with our previous studies and published reports suggesting improved bacterial penetration of this compound (McKenney et al., in press). Interestingly, as can be seen in Figure 2D, there were no breakthrough colonies observed in the glpT transposon mutant experiments in multiple repeats, perhaps suggesting that the locus was the origination of the instability, leading to the breakthrough colonies observed with the wild-type F. novicida, and that other loci are perhaps not as susceptible to mutations leading to breakthrough colonies.

glpT transposon mutants are resistant to Fos and sensitive to compound 1. The glpT transposon insertion mutants were tested for their susceptibility to Fos and compound 1 in a liquid broth growth inhibition assay. As expected, the transporter-independent lipophilic prodrug compound 1 was effective at inhibiting growth of the glpT transposon insertion mutants (glpT-1 shown) as well as wild-type F. novicida, while Fos at twice the MIC was not (Figure 4). This result supports the hypothesis that mutation in the glpT transporter sequence can have a significant detrimental effect on transport of Fos across the bacterial membrane, preventing its interaction with the intended target enzyme, DXR. Intermediate inhibition of the glpT mutant by FR900098 was observed in this assay (gray bars) consistent with the disk-inhibition assay results (see Figure 2D), perhaps due to the slightly more lipophilic nature of FR900098 compared to Fos (see Figure 1 for structures), which may allow some bacterial penetration.

The glpT coding region (GlpT) was identified in the F. tularensis tularensis SchuS4 (FTT0725c) and F. novicida (FTN_0636) genome (accession numbers YP_169738.1 and YP_898283.1) via BLAST search using the E. coli K12 homologous sequence (accession number NP_416743.1) as the query. All other Francisella sequences also have a highly conserved glpT homolog (e.g., FTM_1358, FTW_1513, FTF0725c, FTA_1594, FTH_1462, FTL_1510, Fphi_0200). In addition, bioinformatics examination of Francisella genomes revealed no homologous gene for fsr, an efflux pump responsible for removal of the Fos from E. coli (Messiaen et al., 2011). Also, by our bioinformatic analysis, the published genomes for F. novicida, F. tularensis Schu S4, and F. tularensis LVS also do not contain the uhpT transporter (an alternate transporter that could substitute for glpT), although the related species F. philomiragia, which has an environmental habitat and is not a known pathogen for humans, has a gene that may be uhpT (Fphi_0883).

In order to examine the sequence of glpT in the Fos^R colonies, we isolated the glpT region by PCR and sequenced glpT from four different spontaneous Fos^R colonies. Colonies were picked and grown in liquid culture overnight in TSB-C to create stocks and small aliquots were removed for genomic DNA preparation and sequencing of glpT. Comparison of sequences of the four selected Fos^R mutants with wild-type sequence data reveals an array of nucleotide deletions and additions from various portions of the glpT gene (Figure 5; Figure A1 in Appendix). The 14 nucleotide deletion from mutant FnFos^R#2 results in a deletion of amino acids in the translated protein as well as the formation of additional stop codons. Similarly, the addition of two nucleotides and one nucleotide in FnFos^R#1 and FnFos^R#3, respectively, results in missense mutation by amino acid substitutions and the appearance of premature stop codons resulting in truncated proteins. These abbreviated proteins are likely non-functional. FnFos^R#4 has a single deletion of six nucleotides at position 1126 of the open reading frame, corresponding to the deletion of a Leu³⁴⁷ and Ser³⁴⁸ amino acid residues in the protein (Figure 5; Figure A2 in Appendix). When the amino acid sequence of GlpT from the FnFos^R#4 is compared with E. coli, the deleted amino acids occur in a region that is analogous to a peripheral alpha-helix number H9 that spans the lipid membrane (Lemieux et al., 2005) but is not associated with the formation of the pore through which Fos is transported into the cell.

We determined the rate and frequency of Fos^R of F. novicida. Using the Drake (1969) formula, the mutation frequency (f) and the mutation rate (μ) resulting in F. novicida Fos^R are approximately 6.3 × 10⁻⁸ and 0.173, respectively. The f is defined as the number of observed colonies in the presence of the inhibitor divided by the total number of cells in culture (N_t). The μ is determined by the number of observed colonies (m), the frequency of mutation (f), and the N_t (Table 1). Similarly, the rate determined by the Maximum Likelihood Method is 0.211 (95% CI = 0.159–0.268; Rosche and Foster, 2000).