A library of celecoxib and its derivatives was kindly provided by Arno Therapeutics, Inc. (Flemington, NJ, United States). Chemical structures of celecoxib, AR-13 and AR-16 are presented in Figure 1. Gentamicin, kanamycin, ethidium bromine, and carbonyl cyanide m-chlorophenyl hydrazone (CCCP) were purchased from Sigma (St. Louis, MO, United States). F. tularensis SchuS4 strain (Type A), and F. tularensis LVS used in this study were described previously (Clay et al., 2008; Dai et al., 2012; Hoang et al., 2016). When needed, F. novicida mutant strains were obtained from BEI Resources transposon library¹ (Gallagher et al., 2007). The bacteria were cultured on chocolate II agar (CHA) plates (Becton Dickinson, Sparks, MD, United States) or modified Tryptic Soy Broth (mTSB) or agar (Hoang et al., 2012) for 48 h (F. tularensis SchuS4 and LVS) or for 24 h (F. novicida) at 37°C prior to use in all experiments. Experiments involving the LVS and F. novicida strains were performed in a BLS2 environment. Experiments with F. tularensis SchuS4 were performed in The Ohio State University BSL3 select agent facility in accordance with CDC and locally approved BSL3 facility and safety standards.

The susceptibilities of Francisella to AR-13, AR-16 and other antimicrobials were determined by minimum inhibitory concentrations (MICs) and bacterial killing assays. MIC assays were performed using the microtiter broth dilution method with an initial inoculum of approximately 10⁶ bacteria/ml as described previously (Hoang et al., 2011, 2012). MICs were determined by the lowest concentration of specific antimicrobial showing complete inhibition of bacterial growth after 24 h of incubation at 37°C. For the bacterial killing assays, bacterial strains were grown at 37°C in mTSB or agar by supplementing with 135 μg/ml ferric pyrophosphate and 0.1% cysteine hydrochloride at 37°C for 48 h. Bacteria were suspended in phosphate buffered saline (PBS) to an optical density (OD) of 0.4 at 600 nm, equivalent to 3 × 10⁹ CFU/ml. Approximately 10⁹ or 10⁶ bacteria in 1 ml of PBS or mTSB, respectively, were incubated with 10 μg AR-13 or AR-16 or control dimethyl sulfoxide (DMSO) at 37°C. Viable bacteria at different time-points were evaluated by serial dilution and plating on mTSB agar plates or CHA plates.

To examine the mechanism of AR-13 action, AR-13 was used to select for spontaneous AR-13-resistant (AR-13^r) mutants in vitro by stepwise selection in broth culture. Briefly, 50 μl from an overnight culture of LVS was exposed to increasing concentrations of AR-13 in 5 ml mTSB broth with 1.25 μg AR-13/ml as a starting concentration at 37°C while shaking. 50 μl of this LVS grown culture was then passed into twofold increasing concentrations of AR-13 in 5 ml mTSB. The process was repeated until LVS was able to stably grow in 20 μg AR-13/ml (approximately 20 passes). Approximately 70 AR-13^r clones were selected to determine the MIC to AR-13. Two representative AR-13^r mutants were then passed in non-selective AR-13 free mTSB for 10 passes of an overnight culture. The stable AR-13^r mutants were chosen for the subsequent studies including genomic sequencing, RNA sequencing, and assays regarding the mechanism of AR-13 resistance.

Genomic DNA from wild-type LVS and its stable isogenic AR-13 resistant mutants were purified using the GenElute Bacterial Genomic kit (Sigma–Aldrich, Cat #PLN70). Samples were sequenced on the Illumina MiSeq instrument using paired-end 2 × 300 bp chemistry, generating more than 3 million read pairs per sample. Churchill² (Kelly et al., 2015) was run to identify any statistically supported variants, utilizing BWA_MEM v0.7.12³ and the GCF_000009245.1 assembly of the F. tularensis LVS reference from NCBI⁴ for alignment and GATK HaplotypeCaller v3.5⁵ for variant calling. Variants were annotated with SnpEff v4.1⁶ against the GCA_000009245.1.26 annotation database.

Similarly, the RNA from the wild-type LVS and stable isogenic AR-13 resistant mutants was purified from mid-log cultures in mTSB using the RNAeasy plus mini kit (Qiagen, Cat#74134). Using methods described previously (Jones et al., 2014), after determining the quality of total RNA using Agilent 2100 bioanalyzer and RNA Nano Chip kit (Agilent Technologies, CA), rRNA was removed from 2 μg of RNA with Ribo-Zero^TM rRNA removal kit for Gram-Negative bacteria (Epicentre Biotechnologies, WI). To generate directional signal in RNA seq data, libraries were constructed from first strand cDNA using ScriptSeq^TM v2 RNA-Seq library preparation kit (Epicentre Biotechnologies, WI). rRNA-depleted RNA (50 ng) was fragmented and reverse transcribed using random primers containing a 5′ tagging sequence, followed by 3′ end tagging with a terminal-tagging oligo to yield di-tagged, single-stranded cDNA. After magnetic-bead based purification, the di-tagged cDNA was amplified by limit-cycle PCR using primer pairs that anneal to tagging sequences have adaptor sequences required for sequencing cluster generation. The AMPure XP System (Beckman Coulter) was used to purify RNA-seq libraries. The Agilent 2200 TapeStation using High Sensitivity D1000 ScreenTape was used to determine the quality of libraries, which were then quantified using the Kappa SYBR^®Fast qPCR kit (KAPA Biosystems, Inc., MA, United States). On average, 21 million paired-end 150 bp RNA-Seq reads were generated using the Illumina HiSeq 4000 platform for each sample (the range was 19 to 24 million). Each sample was aligned to the GCF_000009245.1 assembly of the F. tularensis LVS reference from NCBI⁷ using version 0.7.5a of BWAMEM⁸. The GFF file provided with the GCF_000009245.1 assembly from NCBI was used to identify transcript features and raw coverage counts were calculated using HTSeq⁹. After normalization of the RNA-Seq gene expression data a post-alignment statistical analyses was performed using custom analysis scripts written in R and DESeq2¹⁰. Normalized read counts were used for comparisons of gene expression and associated statistical analysis of the different bacteria/conditions. Fold change values were displayed as test condition/control condition, where values less than one were expressed as the negative of its inverse (note that there will be no fold change values between -1 and 1, and that the fold changes of “1” and “-1” represent the same value). A false discovery rate of 10% (DESeq2 adjusted p-value <= 0.1) was used in determining transcripts that were significantly differentially expressed.

Human monocyte-derived macrophages (hMDMs) were derived from human blood acquired via venipuncture from healthy donors following a Ohio State University Institutional Review Board approved protocol. Written informed consent was provided by study participants. The protocol followed published methods (Schlesinger et al., 1990; Hoang et al., 2016). Briefly, peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood over a Ficoll cushion (GE Healthcare Bio-Science, Piscataway, NJ, United States). PBMCs were then cultured in sterile screw-cap Teflon wells in RPMI 1640 plus L-glutamine (Gibco-Life Technologies, Grand Island, NY, United States) with 20% autologous human serum at 37°C in a humidified incubator containing 5% CO₂ for 5 days. Teflon wells were chilled on ice to recover the PBMCs, which were then re-suspended in RPMI 1640 with 10% autologous serum. Cells were then allowed to attach in 24-well or 6-well tissue culture plates for 2–3 h at 37°C in a humidified incubator with 5% CO₂. After washing to remove the lymphocytes, hMDM monolayers were seeded at a density of approximately 2.0 × 10⁵ cells/well for 24-well plates for infection studies.

Human monocyte-derived macrophages viability was assessed in the presence of AR-13 and AR-16 using a lactate dehydrogenase (LDH) assay (Roche Applied Science, Indianapolis, IN, United States) described previously (Hoang et al., 2014). hMDM cells were seeded into 24-well plates at 2 × 10⁵ cells/well with RPMI 1640 supplemented with 20% autologous serum. This medium was replaced with 2% autologous serum in RPMI 1640 containing different concentrations of AR-13 or AR-16 (in 0.1% DMSO) or DMSO vehicle control. Triton X-100 (0.1%) was used as a positive control. After 24 h or 48 h of treatment, supernatants were collected, centrifuged to remove cells, and subjected to the LDH assay (as per manufacturer’s instructions). LDH release was measured (570 nm) and the cytotoxicity was calculated as a percentage of the Triton X-100 treated cells (positive control).

Live vaccine strain growth in hMDMs was performed as described previously (Chiu et al., 2009b) with a minor modification. Briefly, LVS were grown at 37°C in mTSB or on agar plates supplemented with 135 μg/ml ferric pyrophosphate and 0.1% cysteine hydrochloride at 37°C for 48 h. Bacteria were equilibrated in PBS (OD₆₀₀ of 0.4 which is approximately 3 × 10⁹ CFU/ml). LVS was added to hMDMs at a multiplicity of infection (MOI) of 50 in the presence of 2% autologous serum in RPMI 1640 plus L-glutamine (Gibco-Life Technologies). Extracellular bacteria were removed 2 h post-infection by addition of 50 μg/ml of gentamicin (Gibco-Life Technologies) for 1 h followed by three washes of the monolayer with pre-warmed RPMI 1640 to remove additional extracellular bacteria. Various concentrations of AR-13 were then added to culture medium (with 2% autologous serum and 10 μg/ml gentamicin). As a control, the parental compound of AR-13, AR-12, which was previously shown to inhibit Francisella growth in macrophages via induction of autophagy, was included. Soluble AR-12 and AR-13 were dissolved in 10 mg/mL in DMSO and diluted in RPMI 1640 containing 2% autologous serum to the appropriate concentrations. At 22 h post-treatment, the infected hMDMs were lysed with 0.1% Triton X-100 (Calbiochem, San Diego, CA, United States) in PBS for 15 min. The cell lysates were then serially diluted with PBS, plated onto CHA agar and enumerated after 72 h incubation at 37°C.

Six–eight weeks old pathogen free female BALB/c mice were purchased from Harlan Sprague (Indianapolis, IN, United States). Food and water were provided ad libitum to the mice (5 mice/group unless otherwise indicated) in sterile micro isolator cages and allowed to acclimate for 2–3 days prior to challenge. All experimental procedures were carried out in strict accordance with guidelines established by The Ohio State University Institutional Animal Care and Use Committee (IACUC), and all efforts were made to minimize animal suffering.

Intranasal (I.N.) infection of F. tularensis SchuS4 and LVS was performed as previously described (Hoang et al., 2016). Briefly, F. tularensis SchuS4 and LVS were grown on CHA plates for 48 h at 37°C. The bacteria were removed from the agar surface and suspended in PBS to an OD₆₀₀ of 0.4, which is approximately 3 × 10⁹ CFUs/ml. After culture dilution, mice were infected with 10 CFU of F. tularensis SchuS4 in 50 μl PBS. For LVS infection, mice were infected by the I.N. route with 3 × 10³ CFU in 50 μl PBS. Both doses are ∼10x the lethal number of CFU. Prior to the infection, mice were anesthetized with isoflurane as approved by The Ohio State University Institutional Animal Care and Use Committee (IACUC).

To evaluate AR-13 as a treatment for infection, survival studies were performed with F. tularensis SchuS4 and LVS strains. Mice were infected with F. tularensis SchuS4 or LVS and treated, starting about 1 h post-infection, with different doses (2.5, 5.0, or 10 mg/kg/day for LVS and 2.5 or 5.0 mg/kg/day for F. tularensis SchuS4) of AR-13 in 200 μl of polyethylene glycol (PEG) PEG400:0.9% saline:ethanol (50:35:15) given by the intraperitoneal (I.P.) route (Hoang et al., 2016) once daily from day 0 until day 10 post-infection. The infected mice were monitored for survival up to 2 weeks post-infection. To determine the effects of AR-13 on bacterial growth in the infected mice, we infected mice (5 mice/group) with the LVS strain via the I.N. route and treated once daily with 5 mg AR-13 in 200 μl PEG/kg/day from day 0 (the day of infection). The bacterial burdens in the lungs of surviving mice were determined by homogenization of lung tissue followed by plating and subsequent CFU enumeration.

Data are presented as mean ± standard deviation (SD). P-values were calculated using one-way ANOVA for multiple comparisons and adjusted with Bonferroni’s correction; ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001; NS, not significant. A Chi-square test was used for survival analysis. GraphPad Prism 6 was used for statistical analysis.

RNAseq data were deposited into the GEO Repository¹¹ under the record number GSE100069. Genomic sequences and alignment data have been deposited to the Sequence Read Archive¹² under identifier SRP109660.

In our continuing effort to find novel antimicrobial agents to control Francisella infection, we screened 64 compounds derived from celecoxib (Salunke et al., 2015) to identify potential anti-Francisella agents using a standard serial dilution method. We found that two compounds, AR-16 and AR-13, have the ability to inhibit the growth of several Francisella subspecies with MICs at 24 h post-inoculation of 2.5 μg/ml for F. tularensis LVS (Figure 2A) and F. tularensis SchuS4 (data not shown), and 5 μg/ml for F. novicida (data not shown). To explore the modes of action of these two anti-Francisella agents, we performed bacterial killing assays the presence of 10 μg/ml for each compound in PBS (Figure 2) or mTSB (Supplementary Figure 1). Viable bacteria were evaluated at different time points post-treatment by serial dilution, plating and enumeration. To our surprise, the two compounds had distinct modes of action: AR-13 had bactericidal activities and AR-16 had bacteriostatic effects on LVS and F. tularensis SchuS4 at stationary phase (Figures 2B–D and Supplementary Figure 1) and at log phase (data not shown). AR-13 treatment (10 μg/ml) lead to an approximate 2–3.5 log decrease in CFUs of F. tularensis SchuS4 and LVS at 8 h post-treatment (Figures 2B,D and Supplementary Figure 1A), while AR-16 showed no significant decrease over this time period (Figures 2C,D and Supplementary Figure 1B). Since AR-16 was not bactericidal, we focused on our further studies on AR-13. We examined the antimicrobial activities of AR-13 on two Gram-positive bacteria, Listeria monocytogenes strain 10403S and methicillin-resistant S. aureus strain JE2 (MRSA, USA300LAC), as well as two Gram-negative bacteria, Pseudomonas aeruginosa strain PAO1 and Salmonella Typhi. The results showed that AR-13 exerted strong antimicrobial activities on the two Gram-positive bacteria, but not on the two Gram-negative bacteria (Supplementary Figure 2). These data provide strong evidence that AR-13 may serve as a potential antimicrobial with bactericidal properties to control Francisella and possibly other bacterial infectious agents.

Since AR-13 exerts bactericidal effects on Francisella in vitro, we sought to examine the effects of AR-13 on the growth of LVS in its primary cellular target, macrophages. The cytotoxicity of AR-13 on hMDMs was tested by measuring lactate dehydrogenase (LDH) release in the culture supernatants from hMDMs treated with various concentrations of AR-13 at 24 and 48 h post-treatment (Hoang et al., 2014). AR-13 was not toxic toward hMDMs at AR-13 concentrations as high as 20 μg/ml (Supplementary Figure 3). hMDMs were then infected with LVS and treated with different concentrations of AR-13. The intracellular bacterial load was evaluated at 22 h post-treatment. As shown in the Figure 3, AR-13 (at 2.5 and 5 μg/ml) significantly inhibited the growth of LVS in hMDMs, reducing the CFU recovered by ∼0.5 logs. As a control, AR-12, known to inhibit intracellular Francisella growth by the induction of autophagy (Chiu et al., 2009b), reduced LVS growth approximately 1 log at 2.5 μg/ml and was toxic to macrophages as expected at 5 μg/ml (Hoang et al., 2014).

Examination of the resistance of microbes to new antimicrobials provides insight into the mechanism(s) of action of the drug as well as drug targets. We sought to examine whether Francisella was able to develop resistance to AR-13 in vitro. Our first attempt to select AR-13 resistant mutants by plating LVS on mTSB agar plates containing AR-13 at 4x MIC (10 μg/ml), 8x MIC (20 μg/ml), and 16x MIC (40 μg/ml) was unsuccessful. Stepwise selection in mTSB broth with increasing concentrations of AR-13 was performed to attempt to select for AR-13 resistant mutants of LVS. After approximately 20 overnight passages in increasing AR-13 selective pressure, AR-13 resistant mutants were obtained and sensitivity to AR-13 was examined by culturing in mTSB with serial concentrations of AR-13 (0.016–10 μg/ml). As shown in Figure 4A, a representative AR-13 resistant mutant was able to grow in the presence of up to 10 μg/ml AR-13 but was inhibited in 20 μg/ml (eightfold increase in MIC) (data not shown). Examination of the stability of an AR-13 resistant mutant was performed by passaging the mutant in AR-13-free mTSB for 10 overnight passages, with several clones then chosen for bacterial killing assays. As shown in Figure 4B, after 10 overnight passages in AR-13-free mTSB, there was no difference in viable bacteria recovered between the initial AR-13 resistant mutant (MT) and that of 10 overnight passages in AR-13-free mTSB (MT-10 passages). These data suggest that Francisella is able to develop resistance to AR-13 and the resistance is relatively stable.

Comparative transcriptomic analysis revealed that the AR-13 resistance of this mutant is not due to changes in the RNA levels since no gene transcript was differentially expressed greater than 1.7-fold relative to the parental strain (data not shown). Thus, the observed stable AR-13 resistance in Francisella indicated an alteration in the inherent genomic information of the resistant mutant. Comparative genomic analysis was performed between the AR-13 resistant mutant and its parent wild type strain. This analysis identified three non-synonymous mutations in the AR-13 resistant mutant. Interestingly, two of them were found in genes encoding for outer membrane efflux proteins FTL_1107 (FtlC) and FTL_1865 (TolC) (Gil et al., 2006) with amino acid substitutions Leu236Pro and Glu441Lys, respectively (Table 1). TolC is the outer membrane channel component for multidrug efflux and type I secretion systems. The other mutation was in gene locus FTL_0600 (wbtH homolog in F. novicida) (Thomas et al., 2007) that is involved in O-antigen synthesis (amino acid substitution of Pro353Ser). Efflux pumps confer resistance to a variety of antibiotics and detergents in Francisella (Bina et al., 2008; Ahmad et al., 2010; Gestin et al., 2010). We hypothesized that the mutations in the efflux systems resulted in a gain of function mutation that increased efflux activity. As such, treatment with an efflux inhibitor would then sensitize the AR-13 resistant mutant to AR-13. To test our hypothesis, we utilized experimental approaches with AR-13 and the proton-mediated efflux pump inhibitor carbonyl cyanide m-chlorophenyl hydrazone (CCCP). Various concentrations of CCCP were tested against LVS to which dramatic reductions in growth were observed at >∼8 nM (Supplementary Figure 4). Thus, a 4 nM sub-inhibitory concentration of CCCP was chosen and was examined in combination with various concentrations of AR-13. We found that the AR-13 resistant mutant was more susceptible to AR-13 in the presence of sub-inhibitory concentrations of CCCP. The growth of the AR-13 resistant mutant was significantly decreased at 2.5 μg/ml AR-13 and nearly completely inhibited at 5 μg/ml AR-13 in the presence of CCCP while the strain grew normally at all concentrations of AR-13 (up to 10 μg/ml) in the absence of a sub-inhibitory concentration of the inhibitor (Figure 5A). In addition, the AR-13 resistant isolate also conferred decreased sensitivity to ethidium bromide (EtBr) which is mediated by TolC as previously observed (Gil et al., 2006) (Figure 5B), but does not affect sensitivity to kanamycin (Supplementary Figure 5). Similar to what was observed with AR-13, a sub-inhibitory concentration of CCCP sensitized the AR-13 resistant strain to EtBr (Figure 5C). In parallel, we examined independent mutants of two TolC orthologs, (FtlC; FTN_0779 in F. novicida) and FTL_1865 (TolC; FTN_1703 in F. novicida) as well as the O-antigen synthesis locus (FTL_0600; FTN_1421 in F. novicida), which were obtained from a BEI Resources transposon library (Gallagher et al., 2007), for susceptibility to AR-13. This showed that both tolC efflux genes, but not the O-antigen synthesis gene, confer intrinsic resistance to AR-13 in F. novicida (Supplementary Figure 6). These data provide strong evidence that efflux pumps mediate AR-13 resistance in Francisella and suggest that AR-13 has an intracellular target(s).

AR-13 exerts bactericidal effects on Francisella (Figures 2B,D) and causes minimal cytotoxicity to hMDMs (Supplementary Figure 3). Therefore, we tested the ability of AR-13 to protect mice from a lethal intranasal dose of F. tularensis SchuS4 or LVS. Toxicity of AR-13 in mice was first evaluated in non-infected mice I.P. treated with 10 mg AR-13/kg/day once daily for 10 consecutive days. These mice did not show any abnormal clinical signs (sickness, hair ruffling) which indicated that they can tolerate at least total 100 mg AR-13/kg. We next infected mice by the I.N. route with a lethal dose of LVS (3 × 10³ CFUs/mouse) and then treated the infected mice, starting about 1 h post-infection, once daily by I.P. injection with 2.5, 5, or 10 mg AR-13/kg/day in 200 μl PEG (Hoang et al., 2016) for 10 days. A PEG-only treated group was included as a control. As shown in Figure 6A, treatment of 2.5 mg AR-13/kg/day for 10 consecutive days provided the best protective effects (50% survival) from LVS infection (total drug of 25 mg AR-13/kg). There were no culturable bacteria in the lungs of these surviving mice at 15 days post-infection (data not shown). Because the 5 and 10 mg AR-13/kg/day treatments demonstrated less protective effects than mice given 2.5 mg AR-13/kg/day, but alone showed no toxicity at these doses, it is possible that the combination of LVS infection and these higher doses of AR-13 may be toxic to mice.

We then sought to evaluate the ability of AR-13 to control virulent Type A F. tularensis SchuS4 (a human clinical isolate) in the mouse infection model. Our previous publication showed that without treatment, mice died from days 4 to 6 following intranasal infection with 10 CFUs of F. tularensis SchuS4 (Hoang et al., 2016). Since AR-13 treatment (25 mg/kg total drug) only provided 50% protective effects (Figure 6A) the compound was also tested in combination with a sub-optimal dose of gentamicin, one of the primary antibiotics commonly used to treat Francisella infection (Hoang et al., 2016). Mice were infected with 10 CFUs of F. tularensis SchuS4/mouse via the I.N. route and then treated, starting about 1 h post-infection, with 2.5 mg (total dose 12.5 mg/kg) or 5 mg AR-13/kg/day (total dose 25 mg/kg), or with 5 mg AR-13/kg/day plus 0.25 mg gentamicin/kg/day (I.P.) once daily for 5 consecutive days. PEG and 0.25 mg gentamicin treated groups were included as controls. As shown in Figures 6B,C, AR-13 treatment prolonged survival of F. tularensis SchuS4 infected mice but did not protect mice from death. All PEG and gentamicin only treated mice died at day 5 or day 6 post-infection, respectively (Figures 6B,C). However, the combination of a sub-optimal dose of gentamicin with 5 mg AR-13/kg/day protected 60% of the infected mice from death (Figure 6C), with no CFU recovered from the lungs of surviving mice (data not shown). These data suggest that AR-13 could potentially be used as a combinational therapeutic to control Francisella infection.