Francisella tularensis subsp. holarctica LVS (ATCC 29684) was obtained from Karen Elkins (Center for Biologics Research and Evaluation, U.S. Food and Drug Administration, Bethesda, MD, USA). F. novicida (U112) was obtained from ATCC, LVSG (a spontaneous gray phase variant) was provided by F. Nano (University of Victoria, Victoria, BC, Canada). Ft LVSGD (another spontaneous gray variant we found to lack O-antigen) was obtained from lot number 703-0102-080 produced at Cambrex BioSciences in Baltimore, MD, USA. The cell bank was produced from a lyophilized vial of Ft LVS lot NDBR-101, lot 4 (Salk produced). F. tularensis subsp. tularensis strain SchuS4, a Centers for Disease Control and Prevention clinical isolate, was provided by Rick Lyons (University of New Mexico, Albuquerque, NM, USA). For most experiments, bacteria were grown overnight (∼24 h) on Choc II agar (BD Biosciences, San Jose, CA, USA) at 37°C as the frequency of phase variation was minimal in these conditions. Liquid cultures were grown overnight (∼16 h) in tryptic soy broth (TSB; Difco Laboratories, Detroit, MI, USA) containing 0.1% cysteine HCl (Sigma-Aldrich, St. Louis, MO, USA) for specified times as described in the results or Figure legends.

Constructs for the deletion of FTL1611 (flmF2) and FTL1609 (flmK) were made in pJC84 (Wehrly et al., 2009). The ∼1 kb upstream region of flmF2 was amplified using forward primer JG1823 (5′-aaacgagctcgGGGTTATGGTGACTTCTGCATC-3′) with SacI restriction site at the 5′ end and reverse primer JG1824 (5′-cgcggatccCACAAATACAAAATATATTAACCTTAATTAATGCTATTATAACC-3′) with a 5′ BamHI restriction site. The ∼1 kb downstream region was amplified using forward primer JG1825 (5′-cgcggatccAATATTGTTTTAAGCTAATGAAT CAATACTTATTAAATTCTTAG-3′) and reverse primer JG1826 (5′-acgcgtcgacGTATTATATTTTTAGTAGCAGCTGTTGCTGTTAT-3′) with BamHI and SalI 5′ flanking restriction sites, respectively. Similarly, the flmK upstream region was amplified by JG2290 (5′-aaacgagctcgGATCTAATAC TGGATACCACTCATTATC-3′) forward primer with a 5′ SacI site and JG2291 (5′-cgcggatccCTTCTTTACCCTCAAATAGAAACTTATAC-3′) reverse primer with a 5′ BamHI site, and the downstream region using the JG2292 (5′-cgcggatccGATTTATCAGCATTAATTACTTTGATAAGCTAAG-3′) forward primer with a 5′ BamHI site and JG2293 (5′-acgcgtcgacCATAGATAA GCGTACAGTTGTTTCATG-3′) reverse primer with a 5′ SalI site. Fragments were cloned in pJC84 sequentially and the construct was transformed into Ft LVS followed by chromosomal recombination using the procedure described by Wehrly et al. (2009). Mutants were confirmed by PCR amplification and sequencing of the deleted region.

Choc II plates containing bacteria were visualized under oblique light settings as suggested by Robert Miller at Dynport Vaccine Company LLC, Frederick, MD, USA and as described by Eigelsbach (Eigelsbach et al., 1951). Briefly, a focused light source, concave mirror and dissecting microscope with 10× objective magnification and a transparent stage were used to visualize blue and gray variants. The concave mirror was placed horizontally tilted upward so that the light beam would hit the upper concave region and the distance between mirror and the microscope is adjusted so that the light beam would reflect on the plate sitting on the stage of the microscope. Blue and gray colonies were observed and counted using these conditions. Samples were prepared for electron microscopy from overnight (∼16 h) grown cultures of Ft LVS or Ft LVSG in TSB containing 0.1% cysteine HCl using methods as described previously (Mohapatra et al., 2008). In brief, cells were pelleted by centrifugation, washed in PBS, and fixed with 2.5% warm glutaraldehyde for 15 min. followed by fixing with a combination of 2.5% glutaraldehyde and 1% osmium tetroxide in 0.1 M sodium cacodylate (pH 7.3) for 15 min at 4°C. Staining of the cells was accomplished by using 0.25% uranyl acetate in 0.1 M sodium acetate buffer (pH 6.3) for 45 min, and viewed after further processing by transmission electron microscopy using an FEI Technai G2 Spirit microscope at 60 kV. Multiple fields (>50) were examined to determine the average size (diameter and length) and shape of bacteria.

Overnight (∼24 h) grown bacteria from Choc II agar plates were suspended in PBS at a concentration of 3 × 10¹⁰ CFU/ml as determined previously by the optical density (OD₆₀₀) of diluted cultures and subsequent colony counts on solid agar. Bacteria equalized by optical density (OD₆₀₀) were then pelleted, frozen, and lyophilized overnight to obtain ∼20 mg of dry cells. LPS was purified using hot phenol/water method using the standard protocol as described by Apicella et al. (1994).

Lipopolysaccharide was separated by 15% SDS-PAGE and silver stained as described (Clay et al., 2008). Briefly, after fixing overnight in 40% ethanol and 5% acetic acid, gels were incubated in 0.7% periodic acid in fixing solution for 7 min and subsequently washed with multiple exchanges of water. The staining solution (0.013% concentrated ammonium hydroxide, 0.02 N sodium hydroxide, and 0.67% silver nitrate (w/v) was applied with vigorous agitation for 10 min, followed by three washes (each 10 min) in water. Gels were developed using a solution containing 0.275% monohydrous citric acid (w/v) and 0.0025% formaldehyde. Upon completion, 5% acetic acid was used to stop the development.

Purified LPS samples (10 μg/well) were electrophoresed on a 15% SDS-PAGE gel and transferred on nitrocellulose membrane using the Bio-Rad semi-dry transfer system. Immunoblotting was performed using either anti-F. tularensis LVS or F. novicida polyclonal sera (from infected mice) or monoclonal sera specific to the LPS of Ft LVS or F. novicida. Polyclonal sera to Ft LVS or F. novicida (1:1000 dilution), commercial F. tularensis FB-11 (1:1000, Abcam, Cambridge, MA, USA), F. tularensis LPS specific monoclonal (1:10), or F. novicida LPS specific monoclonal (1:10) were used as primary antibodies with alkaline phosphatase conjugated goat anti-mouse IgG (1:4000) as the secondary antibody. The F. tularensis-specific and F. novicida-specific anti-LPS monoclonal antibodies were obtained from monoclonal hybridoma cell lines (ImmunoPrecise, Victoria, BC, Canada). Blots were developed using 5-bromo-4-chloro-3-indolyl phosphate/NBT (Sigma-Aldrich) as the substrate.

Lipopolysaccharide was isolated using the hot phenol/water method (Apicella et al., 1994). Crude LPS was enzymatically treated to remove contaminating nucleic acids and proteins and ultracentrifuged for 18 h. The LPS pellet was collected and the carbohydrate portion of LPS was released from lipid A via 2 h mild hydrolysis with 1% acetic acid at 100°C followed by centrifugation of the lipid A at 3500×g. The carbohydrate fraction in the supernatant was extracted threefold with chloroform to remove any contaminating lipid A, lyophilized, re-suspended in water, filtered through nylon filter 0.2 μm prior the HPLC separation, and lyophilized again. Carbohydrates from Ft LVS and Ft LVSG of were resolved on a Superdex Peptide HPLC column with ammonium acetate used as an eluent. The eluting fractions were pooled and salts removed by repeated evaporations from de-ionized water on a rotary evaporator. The elution profiles for the Ft LVS and Ft LVSG carbohydrates were examined and Fraction 1 contained the O-polysaccharide (OPS), Fraction 2 contained slightly lower molecular weight OPS, and Fraction 3 contained the core oligosaccharides (OSs) with some possibly low molecular weight OPS repeat units. Fractions 1 and 3 were analyzed by NMR spectroscopy. Fraction 1 from both LVS and LVSG were compared to each other using 2D NMR. OSs found in Fraction 3 from LVS and LVSG were analyzed by 1D proton NMR spectroscopy.

RNA from log phase (0.4–0.5 optical density at 600 nm) cultures of Ft LVS and Ft LVSG was extracted using the RNeasy kit (Qiagen, Valencia, CA, USA). The quality and quantity of RNA was determined using the Experion automated electrophoresis system (Bio-Rad, Hercules, CA, USA). One microgram of total RNA was reverse transcribed to cDNA using Superscript II RNase H⁻ reverse transcriptase (Invitrogen, Carlsbad, CA, USA). cDNA was then normalized according to the concentration and 2 ng of the converted cDNA was used for quantitative PCR with the SYBR green PCR master mixture in the Bio-Rad iCycler apparatus (Bio-Rad, Hercules, CA, USA). All primers were designed to give 200- to 220-nucleotide amplicons with melting temperatures of 48–52°C. Relative copy numbers and expression ratios of selected genes were normalized to the expression of the housekeeping gene (dnaK) and calculated as described by Mohapatra et al. (2007).

Human monocyte-derived macrophages (MDMs) were isolated using standard procedure as described elsewhere (Mohapatra et al., 2010) and obtained with informed consent from healthy donors by an OSU IRB approved protocol. Intramacrophage survival assays in human MDMs were performed using following procedure; 2 × 10⁶ PBMCs/well (MDMs plus lymphocytes) were plated in a 24-well plate resulting in 2 × 10⁵ MDMs/monolayer after adherence of MDMs and washing. Francisella spp. were opsonized with 0.1% serum for 30 min at 37°C. Macrophages were infected with Ft LVS, Ft LVSG and F. novicida at an MOI of 50 and incubated at 37°C in a CO₂ (5%) incubator for 2 h. Cells were washed and 50 μg/ml gentamicin was added to each well and incubated for 30 min. Cells were washed and replenished with fresh media containing 10 μg/ml gentamicin. At various time points cells were washed and lysed with 0.1% SDS and plated on Choc II plates to enumerate the colony forming units.

Bacteria grown overnight (∼24 h) on Choc II plates were scraped and suspended, washed twice and diluted in PBS. Four- to six-week-old BALB/c mice were anesthetized and infected with ∼1000 bacteria in a 20-μl volume by the intranasal route and dilutions were plated on Choc II plates to enumerate the inoculum. Mice were anesthetized and challenged with 1000 CFU of overnight (∼24 h) grown F. tularensis subsp. tularensis SchuS4 intranasally 4 weeks post vaccination and observed daily for survival. These procedures were performed as described in an OSU IACUC approved protocol in an inspected and approved biosafety level 3 laboratory.

Francisella tularensis LVS phase has been observed to vary from a blue (wild-type) colony to a gray colony variant (Eigelsbach et al., 1951; Cowley et al., 1996). Such gray variants have been both characterized with an extended LPS O-antigen (Cowley et al., 1996) as well as a truncated O-antigen (Hartley et al., 2006). Our work described here is with the Ft LVSG isolate (a variant with an extended O-antigen), but at times comparisons are made to Ft LVSGD (a variant with no O-antigen). We examined various media conditions and growth phases to determine the conditions that affected the degree of phase variation. We found that Ft LVSG grows slower than Ft LVS, forms smaller colonies on agar surfaces, and appeared gray by eye on Choc II agar plates under oblique lighting (Figures 1A,B). The frequency of blue to gray phase variation was higher (27–31%) in liquid cultures (TSB + 0.1% cysteine HCL) grown to stationary phase (typically 30–48 h) and plated on solid agar. The frequency of phase variation was minimal (2–5%) for bacteria grown on plates 1–2 days and in log phase liquid cultures. We also observed that the frequency of blue to gray phase variation dramatically increased when Ft LVS was passed through macrophages (23–27%) or recovered from organs of infected animals (31–36%). We also observed that the frequency of forward phase variation in broth grown bacteria (blue to gray) was always higher (∼30%) than frequency of reverse (gray to blue) phase variation (5–7%).

To more clearly compare Ft LVS to Ft LVSG bacteria, log phase cultures were examined by scanning electron microscopy. Comparisons of average cell size were not significantly different, but more membrane vesicles were observed in Ft LVSG cultures (Figure 1C). It is not clear what impact this increased vesiculation has on the subsequent phenotypes described for Ft LVSG.

It was shown previously that the LPS of Ft LVS and Ft LVSG had differential reactivity to monoclonal antibodies stated to be O-antigen specific (Cowley et al., 1996), suggesting an O-antigen antigenic switch. To further examine the LPS O-antigen and its antigenic properties, we purified LPS from Ft LVS, Ft LVSG, F. novicida, Ft LVSGD, F. tularensis SchuS4 and a F. tularensis SchuS4 small colony gray variant and performed silver staining on SDS-PAGE separated samples. Consistent with previously published results, the gray variant (Ft LVSG) possessed an O-antigen but Ft LVSGD was rough (lacked O-antigen) (Figure 2A). The F. tularensis SchuS4 small colony gray variant also appeared to produce an LPS with a repeating O-antigen.

Glycosyl composition analysis of the OSs released from the purified LPS preparations show that the Ft LVS and Ft LVSG OSs contain the same glycosyl residues, but there is a large quantitative difference, in that the Ft LVS OS contains much larger amounts of QuiN and Gal than what is found in the OS from the Ft LVSG strain (Table 1). The QuiN could be due to QuiNFo as NMR analysis (data not shown) shows a significant resonance at around 8 ppm, which is consistent with a formyl proton. During the preparation of trimethylsilyl (TMS) methyl glycoside, which is accompanied by N-acetylation, this formyl group would have been replaced by an acetyl group. The large difference in these components between the LPS of Ft LVS and Ft LVSG would indicate that the Ft LVS LPS contains much more of the QuiN/Gal-containing O-antigen chain polysaccharide than Ft LVSG. Thus, these data suggest that the Ft LVSG LPS has an O-antigen but the O-antigen contains fewer repeating units than seen in Ft LVS LPS or that LVSG lipid A-core is capped less frequently with O-antigen. Interestingly, even though the two variants (Ft LVSG and Ft LVSGD) have distinct LPS regarding the amount of O-antigen present (albeit both with amounts less than that of wildtype), they give rise to morphologically similar gray variants.

Western blot analysis was performed on LPS samples using commercially available anti-F. tularensis FB-11 (Figure 2B) antibodies stated to be specific to the O-antigen, as well as anti-F. tularensis (Figure 2C) and anti-F. novicida LPS monoclonal antibodies (Figure 2D). The results showed that Ft LVS LPS reacted with both the F. tularensis monoclonal and FB-11 antibodies showing the typical LPS laddering, and while the laddering was not observed on Ft LVS LPS Western blots with the F. novicida specific antibodies, reactivity was observed to a low molecular weight species that is typically lipid A plus core (Figures 2B–D). F. novicida LPS did not react at all to the anti-F. tularensis LPS or FB-11 antibody, but strongly to a low molecular weight species (plus typical laddering) with the F. novicida specific monoclonal antibody. Ft LVSG reacted with both F. tularensis and the F. novicida anti-LPS monoclonal antibodies. These results are consistent with those of Cowley et al. (1996) with regard to O-antigen ladder reactivity, but the lipid A core region is not clearly visible on their gels. As expected, LVSGD did not demonstrate the typical O-antigen ladder due to its the lack of O-antigen (Figures 2B–D). The F. tularensis SchuS4 gray variant reacted with both anti-F. tularensis monoclonal antibodies and clearly possesses an O-antigen based on the observed laddering. However, the modal chain length or capping frequency of this O-antigen, like that of Ft LVSG, appears reduced versus F. tularensis SchuS4 (Figures 2B–D).

These LPS samples were reacted in a Western blot with anti-F. tularensis or anti-F. novicida polyclonal sera generated from infected mice (Figure 2E). Ft LVS LPS and F. novicida LPS only reacted with their respective antisera while Ft LVSG now reacted only with F. tularensis polyclonal sera. These results suggest that changes in the Ft LVSG LPS are specifically recognized by monoclonal but not polyclonal antibodies.

We next isolated and performed extensive NMR analyses on the Ft LVS and Ft LVSG OPSs to determine if any structural differences could be detected between these molecules. Crude LPS was enzymatically treated to remove contaminating nucleic acids and proteins and ultracentrifuged. The LPS pellet was collected and the carbohydrate portion of LPS was released from lipid A via mild hydrolysis. The carbohydrate fractions from Ft LVS and Ft LVSG of were resolved by HPLC. The carbohydrates eluted in three primary fractions. Fraction 1 contained the OPS, Fraction 2 contained slightly lower molecular weight OPS, and Fraction 3 contained the core OSs with some possibly low molecular weight OPS repeat units. From these results, the Fraction 1/3 ratio for Ft LVS LPS is 17, while it is 4.6 for LVSG (Table 2). These results are consistent with the above data showing that the Ft LVSG strain contains less OPS as reflected by the lower QuiN level during composition analysis. Fractions 1 and 3 were analyzed by NMR spectrometry. Fraction 1 from both Ft LVS and Ft LVSG were compared to each other using 2D NMR experiments – COSY, TOCSY, NOESY, and HSQC (data not shown). These results indicate that the OPS from Ft LVS and Ft LVSG have the same structures. In addition, the data are completely consistent with the structure reported for F. tularensis strain 15, strain SchuS4, and OSU10 (Vinogradov et al., 2002; Prior et al., 2003; Thirumalapura et al., 2005). The results clearly support the conclusion that Ft LVS and Ft LVSG have the following OPS structure as previously reported for the above F. tularensis strains: -4)-α-D-GalpNAcAN-(1 → 4)-α-D-GalpNAcAN-(1 → 3)-β-D-QuipNAc-(1 → 2)-β-D-Quip4Fo-(1 →

The differential staining with the F. tularensis and F. novicida monoclonal antibodies coupled with the fact that the F. novicida monoclonal antibody binds to the LMW LPS and is, therefore, likely binding to the core OS, suggests that the true monoclonal antibody-tracked alteration between Ft LVS and Ft LVSG is related to the core region. Therefore, we analyzed OSs found in Fraction 3 from Ft LVS and Ft LVSG by 1D proton NMR spectroscopy. The results are shown in Figure 3. The spectrum of Ft LVS Fraction 3 indicates that a small amount of truncated OPS is still present. We also see resonances that are consistent with the published core structure with the exception that we do not observe evidence for the core GalNAc residue. The Ft LVS and Ft LVSG proton spectra clearly differ from one another indicating that the Ft LVSG Fraction 3 contains different structures than found in the Ft LVS Fraction 3. Therefore, since the OSs in Fraction 3 would be those that would comprise the core region, as well as some possible truncated OPS, these data support the conclusion that the Ft LVSG has an altered core region compared to that of Ft LVS.

Cowley et al. (1996) demonstrated that Ft LVSG lipid A versus that of Ft LVS elicited increased NO induction in rat macrophages. We confirmed this finding (NO production by rat macrophages measured as nitrate by the Griess reagent system) by both Ft LVSG LPS and purified lipid A (data not shown). This suggested that the lipid A of Ft LVSG was different than that of Ft LVS. To explore these differences, we performed structural analyses on purified lipid A of Ft LVS and Ft LVSG. MALDI-TOF analysis in negative and positive ion mode was performed on replicate lipid A preparations. Both showed the absence of a peak at m/z 1666 in Ft LVSG that was present in Ft LVS. It is known that this peak represents the addition of galactosamine (161 Da) to the basic structure at m/z 1504 (2× GlcN, 3× C18:0 (3-OH), C16:0, P), though previous data suggested that this modification was not observed in the Ft LVS strain (Kanistanon et al., 2008; Figures 4A,B). This was the only structural alteration noted. Since the MALDI-TOF analysis is only semi-quantitative, we performed galactosamine quantitation assays to further demonstrate the reduced galactosamine modification of lipid A in Ft LVSG. The lipid A was derivatized and analyzed by GC–MS. F. novicida showed the highest degree of modification at 25% while Ft LVS was at 14%. Both Ft LVSG and Ft LVSGD showed reduced galactosamine modification, with 6 and 7%, respectively (Figure 5A).

Three genes have been identified that are responsible for galactosamine or mannose lipid A modification (Gunn and Ernst, 2007; Kanistanon et al., 2008). The transferases FlmF1 and FlmF2 are required for adding mannose or glucosamine residues, respectively, to the lipid A. The glycosyltransferase FlmK can add both mannose and galactosamine to lipid A (Gunn and Ernst, 2007; Kanistanon et al., 2008). A real-time PCR assay was performed on the genes flmF2 and flmK from Ft LVS and Ft LVSG to determine if their expression was altered and might be responsible for the observed lipid A galactosamine modification alteration. Expression of both genes was found to be significantly less in Ft LVSG (Figure 5B), correlating with the reduction in galactosamine modification. Mutation of the flmF2 gene in Ft LVS, which eliminates the galactosamine modification, reduced the frequency of phase variation from 30% to 5-7% in stationary phase liquid cultures. Though F. novicida strains carrying this mutation have been shown to affect mouse virulence and cytokine/chemokine induction in macrophages, the Ft LVS flmF2 mutant demonstrated no defect in survival in macrophages of mouse (Raw, J774.1 and MH-S), rat (bone marrow derived, alveolar), or human (THP-1 and monocyte derived macrophages) origin (data not shown, see below section).

It has been shown previously by Cowley et al. (1996) that Ft LVS and Ft LVSG intracellular growth/survival was similar in mouse macrophages, but that differences in growth/survival could be visualized in rat bone marrow-derived macrophages. We examined the survival of Ft LVS and Ft LSVG in various macrophages including J774.1 (a mouse macrophage cell line), MH-S (a mouse alveolar macrophage cell line) and mouse bone marrow-derived macrophages and did not find any significant differences in survival of Ft LVS and Ft LSVG (data not shown). However, we observed that Ft LVSG survived less well in rat bone marrow derived macrophages and a rat alveolar macrophage cell line (ATCC# CRL-2192) and that this inhibition of growth of Ft LVSG can be reversed by using the NO inhibitor NMMA (data not shown). These findings were consistent with the previous findings of Cowley et al. (1996). We then examined intramacrophage survival in human MDMs and the THP-1 macrophage-like cell line. We observed that the Ft LVSG strain survived less well over the first 12 h post-infection in both cell types with macrophage cell death at later time points. The defect was most prominent for the MDMs, where the Ft LVSG strain demonstrated nearly a log defect in survival at 12 h post-infection versus the Ft LVS strain (Figure 6). Thus, the Ft LVSG strain has an intramacrophage survival defect in rat and human but not mouse macrophages.

Gray variants have been shown to be less virulent (Eigelsbach and Downs, 1961) and/or less protective as vaccines against F. tularensis SchuS4 challenge (Eigelsbach and Downs, 1961; Cowley et al., 1996; Hartley et al., 2006; Conlan and Oyston, 2007) However, the specific virulence of the Ft LVSG strain in the mouse model of tularemia has not been determined. To compare the virulence of Ft LVS and Ft LVSG, BALB/c mice were infected with 100 CFU of Ft LVS and Ft LVSG intranasally and observed for survival. Both Ft LVS and Ft LVSG infected mice demonstrated 80% survival (N = 10 mice; Figure 7A). The surviving mice were challenged with 1000 CFU of F. tularensis SchuS4 (∼100-fold above the LD₅₀) intranasally 4 weeks post vaccination (Figure 7B). All Ft LVS vaccinated mice (N = 8) survived the challenge whereas Ft LVSG vaccinated mice (N = 8) could not survive the challenge and succumbed to infection within 5 days. These results suggest that Ft LVSG is as virulent in mice as Ft LVS but it does not protect against Type A challenge.