Paralogous Outer Membrane Proteins Mediate Uptake of Different Forms of Iron and Synergistically Govern Virulence in Francisella tularensis tularensis*

Background: FslE and FupA are Francisella-specific paralogous proteins involved in iron acquisition. Results: fslE mutation disrupts siderophore-mediated ferric iron uptake, fupA mutation impairs high affinity ferrous iron uptake, and both mutations impact virulence. Conclusion: Optimal iron acquisition and virulence require both paralogs. Significance: Iron acquisition mechanisms are potential targets for preventive or therapeutic intervention in F. tularensis infections. Francisella tularensis subsp. tularensis is a highly infectious bacterium causing acute disease in mammalian hosts. Mechanisms for the acquisition of iron within the iron-limiting host environment are likely to be critical for survival of this intracellular pathogen. FslE (FTT0025) and FupA (FTT0918) are paralogous proteins that are predicted to form β-barrels in the outer membrane of virulent strain Schu S4 and are unique to Francisella species. Previous studies have implicated both FupA, initially identified as a virulence factor and FslE, encoded by the siderophore biosynthetic operon, in iron acquisition. Using single and double mutants, we demonstrated that these paralogs function in concert to promote growth under iron limitation. We used a 55Fe transport assay to demonstrate that FslE is involved in siderophore-mediated ferric iron uptake, whereas FupA facilitates high affinity ferrous iron uptake. Optimal replication within J774A.1 macrophage-like cells required at least one of these uptake systems to be functional. In a mouse model of tularemia, the ΔfupA mutant was attenuated, but the ΔfslE ΔfupA mutant was significantly more attenuated, implying that the two systems of iron acquisition function synergistically to promote virulence. These studies highlight the importance of specific iron acquisition functions, particularly that of ferrous iron, for virulence of F. tularensis in the mammalian host.

ExbB-ExbD complex to facilitate receptor-mediated transport of siderophore-iron complexes across the outer membrane (19). F. tularensis is among a small group of bacteria with genomes that do not encode an obvious TonB complex and the first for which a receptor (FslE) was identified (16). Characterization of siderophore uptake facilitated by FslE therefore is of particular interest because it represents a new paradigm.
Studies to date suggest that the siderophore-mediated iron uptake impacts virulence to different extents depending on the Francisella strain and/or infection model used. The siderophore biosynthetic gene fslA gene was found to be important for virulence of LVS in a mouse model of respiratory tularemia, but not for Schu S4 virulence in an intradermal model of infection (20,21). Genes in the fsl operon of F. novicida are required for virulence in mice and arthropod vectors, as well as for intracellular replication (22)(23)(24)(25).
FslE is one of a family of five sequence-related proteins encoded in the Schu S4 genome (26). The FslE paralogs are unique to Francisella species, although they show some variation among the different strains. The paralog with the greatest similarity, FupA was initially characterized as a virulence factor in Schu S4 based on attenuation of a fupA mutant in a mouse model of tularemia (27). fupA was found to be important for intracellular replication of Schu S4 as well as F. novicida (23,27), and the related fupA/B hybrid protein in LVS has also been linked to virulence in mice (28).
Growth assays and transcriptional analysis of a fupA mutant linked this FslE paralog with iron metabolism and iron acquisition in Schu S4, although its exact function was not established (21). Here we have sought to define the primary roles played by FslE and FupA in iron acquisition in Schu S4 and to determine their relative importance for virulence. We used single and double deletion mutants of fslE and fupA to show that both paralogs contribute to optimal growth in vitro. We used 55 Fe transport assays to show that FslE functions in the siderophoremediated uptake process and that FupA mediates high affinity ferrous iron uptake. Our results indicate that although FupA functions as the principal virulence determinant, both paralogs are required for full virulence in a mouse model of tularemia.
Construction of ⌬fupA and ⌬fupB Mutants and Complementing Plasmids-Deletion mutants of fupA and fupB were generated using a sacB plasmid as previously detailed (16). A 1.977-kb region upstream of and including the first 10 codons was amplified with primers 243 and 244 (supplemental Table  S1) using FastStart High Fidelity DNA polymerase (Roche Applied Science). The 3Ј-flanking sequences included the last 13 codons and 1.958 additional kb downstream and were amplified with primers 245 and 246. For the fupB deletion construct, the 1.713 kb of DNA starting 14 bp upstream of the start codon was amplified with primers 256 and 257 to generate the 5Ј-flanking sequence. An insertion element ISFTU1 is located 65 bp downstream of the fupB stop codon. As in our previous work, we avoided use of the transposon sequences in the 3Ј-flanking sequences (28), and instead, amplified the 1.532 kb sequence located downstream of the insertion element using primers 360 and 361. The 5Ј-and 3Ј-flanking sequences were cloned in suicide vector pGIR463 to generate plasmids pGIR476 (for fupA deletion) and pBAS4 (for fupB deletion). The 5Ј-flanking sequences used in pGIR476 were combined with the 3Ј-flanking sequences from pBAS4 to generate the ⌬fupA ⌬fupB deletion construct pBAS5. pGIR476 was introduced into Schu S4 and the Schu ⌬fslE strain GR211 by electroporation, whereas pBAS4 and pBAS5 were electroporated into Schu S4. PCR analysis of genomic DNA was used to identify deletion mutants isolated from sucrose plates. PCR products from deletion mutants GR218 (Schu ⌬fupA), GR219 (Schu ⌬fslE ⌬fupA), GR242 (Schu ⌬fupA ⌬fupB), and GR243 (Schu ⌬fupB) were sequenced to verify the deletions. The mutation in fupA resulted in an in-frame deletion of the central 524-amino acid coding region of fupA. The fupB mutant had a complete deletion of the gene including 14 nucleotides upstream.
Complementation in cis was achieved by integration of a suicide plasmid targeted to the 3Ј region of feoB as previously detailed (28). The integrative plasmids had the promoter of fslA driving expression of the different genes. Plasmid pGIR474 expressing fslE has been described previously (28). For construction of the fupA complementing plasmid pGIR477, the fupA sequences were amplified with primers 256 and 257 using FastStart High Fidelity polymerase (Roche Applied Science). The plasmids were introduced into the different bacterial strains by electroporation or by conjugation from E. coli S17-1. Integrants were confirmed by PCR analysis of genomic DNA.
Growth in Iron-replete and Iron-limiting Media-The bacteria were grown in che-CDM containing 2.5 g/ml (3.36 M, iron-replete) or 0.125 g/ml (0.168 M, iron-limiting) FePPi for analysis of growth under iron limitation as previously described (28). In specific experiments as noted, FeSO 4 was added at 2 or 0.2 g/ml to generate iron-replete and iron-limiting media, respectively. Bacteria from overnight cultures in CDM were washed and inoculated into iron-replete or iron-limiting media to an optical density (A 595 ) of 0.01 and grown at 37°C with shaking. Growth was monitored as increase in optical density. For depletion of intracellular iron pools, the CDM-grown bacteria were washed and allowed to grow for 7 h in che-CDM or for 11 h in C-CDM containing no iron. For testing growth on iron-limiting agar, washed bacteria from overnight cultures in CDM were normalized to an optical density of 1.0 and diluted 10-fold serially in che-CDM, and 5 l of the different dilutions were spotted on the CDM-Fe agar plates. Specific plates contained the standard supplement FeSO 4 (7.19 M) or FePPi (6.8 M). The plates were incubated at 37°C, and growth was recorded over a period of 4 days. Growth assays were repeated at least twice.
Generation of FupA Antibody-The DNA sequences corresponding to amino acids 25-176 of FupA were amplified from Schu S4 genomic DNA using primers 332 and 333 (supplemental Table S1) by PCR with FastStart High Fidelity polymerase (Roche Applied Science) and cloned in the expression vector pGST-parallel (30). Expression of protein in BL21(DE3) cells was induced overnight at room temperature with 5 mM isopropyl ␤-D-thiogalactopyranoside, and the GST-FupAЈ protein was purified by affinity purification on glutathione-agarose (Thermo-Fisher). Recombinant protein was used to raise antibody in guinea pigs (Covance, Denver, PA).
Western Blotting-Cultures of F. tularensis in iron-replete or iron-limiting media were normalized to cell density, and the lysates were separated by SDS-PAGE on 10% gels. The proteins were transferred to PVDF and incubated with different primary and HRP-conjugated secondary antibodies for detection by chemiluminescence. The primary antibodies were rabbit antiserum raised to a FslE peptide (28) used at a 1:2500 dilution, rat antiserum raised to FupB (FTT0919, a kind gift from Michael Norgard) used at 1:1000 dilution, guinea pig antibody to FupA used at 1:100,000 dilution, and rabbit antibody to GroEL (Sigma) used at 1:10,000 dilution.

55
Fe Uptake Assays-55 FeCl 3 (PerkinElmer Life Sciences; 25.04 mCi/mg, 36.88 mCi/ml) was diluted as required for use in transport assays carried out in 96-well filter plates (Millipore). Bacteria were cultured for 18 -24 h in iron-limiting medium, and cell pellets were washed once in che-CDM to remove any iron and siderophore remaining from the culture supernatants. Bacteria were then resuspended in cold uptake medium to an optical density (A 595 ) of 0.2 and stored on ice for no longer than 1 h. The wells of the filter plate were rinsed once with Milli-Q water, and 100 l of bacteria were aliquoted into wells along with 90 l of uptake medium. The plate was incubated on a heat block at 37°C for 10 min to prewarm the cells. Uptake reactions were initiated by the addition of 10 l of the 55 Fe labeling mix at different times, and the bacteria were collected and washed by vacuum filtration at 304.8 mm Hg pressure. All of the experiments used samples in triplicate or quadruplicate. Ten l of 0.1% SDS was added into each well, and the plates were warmed to 56°C for 1 h to kill bacteria and to dry the plates. Individual wells were punched out into scintillation vials and counted after 3-4 h at room temperature in 4 ml of scintillation fluid (Ecoscint A; National Diagnostics). A medium control with no bacteria was used to obtain a value for background correction in some experiments. Bacteria in wells were used in a BCA assay (Pierce) after extraction of proteins with 0.1% SDS to normalize results to protein in place of cell density for some of the experiments.
Siderophore-mediated uptake was carried out in che-CDM containing 10 mM sodium citrate. Siderophore was purified from LVS culture supernatants by anion exchange chromatography on AG1X-8 columns as previously described (14). Siderophore concentrations were assessed in relation to deferroxamine standards using the Cu-CAS reagent (31). A labeling mix was prepared by incubating 30 M 55 Fe with an excess of siderophore (100 M) in uptake medium at room temperature for 30 -60 min before use in the assay. 10 l of mix were added to 190 l of bacteria in wells so as to have a final concentration of 1.5 M 55 Fe and 5 M siderophore in the uptake reaction.
For ferrous iron uptake, 55 FeCl 3 was diluted in 0.1 M ascorbate to keep the iron in the reduced form. 10 l of 55 Fe in ascorbate at varying concentrations was added to 190 l of bacteria in che-CDM to attain the desired final concentration of iron with 5 mM ascorbate in the uptake reaction.
For assessment of uptake rates, incorporation of label by bacteria at time points 4 or 5 min apart early in the uptake process was determined. The rate was calculated by normalizing the incorporation to time and to optical density (A 595 ) or to protein content. Kinetic parameters of transport were determined by nonlinear regression analysis and were plotted using Prism 4.0 software (GraphPad Software, Inc., San Diego, CA).
Intracellular Invasion and Replication-The ability of bacteria to replicate intracellularly was assessed as previously detailed (28). Mouse macrophage-like J774.1 cells (ATCC TIB-67) were cultured in DMEM containing 10% FBS (Invitrogen) at 37°C with 5% CO 2 . The cells were seeded in 24-well plates at a density of 2 ϫ 10 5 cells/well 1 day prior to infection . F. tularensis bacteria grown to midexponential phase in CDM were used to infect the J774A.1 cells at a multiplicity of infection of 10. The numbers of bacteria present in the inocula were verified by plating on MHA. After 1 h at 37°C, the medium was replaced with fresh medium containing gentamycin (100 g/ml), and the cells were incubated for 1 h more. The wells were washed with DMEM lacking FBS and replenished with fresh DMEM ϩ FBS. The cells in sets of four wells were lysed with 1% saponin at 2, 9.5, and 23 h post-infection, and the lysates were serially diluted and plated on MHA plates to determine the titer in colonyforming units (CFU). The results were analyzed using the t test function of Excel 2003 (Microsoft). The experiments were performed three separate times with similar results.
Analysis of Virulence in Mice-6 -8-week-old C57BL/6 mice (Jackson Labs, Bar Harbor, ME) were infected with Schu S4 and mutants using aliquots of frozen stocks grown in CDM and titered prior to use in infection. The bacteria were diluted in PBS and delivered by subcutaneous injection behind the neck in a volume of 100 l. The actual numbers of bacteria delivered were verified by plating on MHA. The animals were monitored daily, and those demonstrating signs of irreversible morbidity were humanely euthanized. The mice surviving the initial challenge were rechallenged with Schu S4 at the doses indicated. All of the mouse infection experiments were conducted with the approval of the animal care and use committee of the University of Virginia.

Generation of Schu ⌬fupA and Schu ⌬fslE ⌬fupA Mutants-
Our studies using a Schu ⌬fslE mutant in cross-feeding assays on iron-limiting CDM agar suggested that FslE functions as a receptor in siderophore-mediated iron uptake (16). Using a similar growth assay on agar, Lindgren et al. (21) concluded that FupA is also involved in siderophore-dependent growth in Schu S4, but phenotypic differences with a siderophore-deficient mutant suggested an additional siderophore-independent role for FupA in iron acquisition. To help identify the primary contributions of the two paralogs to iron uptake processes in Schu S4, we generated an in-frame deletion mutant in fupA and a strain with both fslE and fupA deleted.
Whole cell lysates from these strains grown in iron-limiting or iron-replete media were used in Western blotting to test for expression of the proteins. The predicted 56-kDa band corresponding to FupA was present in Schu S4 lysates, whereas the ⌬fupA mutant was missing this band ( Fig. 1, A and B). FupA expression was not regulated in Schu S4 by iron levels in the growth medium, consistent with transcriptional data reported previously (21). Deletion of fslE or of the fur gene had no effect on FupA levels (Fig. 1A).
Lindgren et al. (21) reported that the fupA mutant had increased siderophore production correlating with transcriptional deregulation of the fsl operon under iron-replete conditions. Our ⌬fupA and ⌬fslE ⌬fupA mutant strains also had significant siderophore production following growth in ironreplete liquid medium (data not shown). We examined FslE levels in cell lysates by Western blotting (Fig. 1B). As expected, a 50-kDa band corresponding to FslE was present in lysates of the parent Schu S4 strain grown under iron limitation but not under iron-replete conditions. The ⌬fupA mutant had elevated levels of FslE, with detectable levels even after growth in ironreplete medium. The chaperone GroEL was used as a loading control for the lysates. The protein levels of FslE correlate with the reported transcriptional deregulation of the fsl operon. Our results confirmed that unlike FslE, protein levels of FupA are not influenced by iron availability. This suggested that FupA function was constitutively required and that in its absence, FslE overexpression was a compensatory mechanism.
FslE and FupA Independently Contribute to Growth under Iron Limitation-Lindgren et al. (21) reported that deletion of fupA affects growth in defined liquid medium with limiting ferrous iron, particularly after iron starvation (21). We therefore tested our ⌬fupA strain for a growth deficiency in che-CDM similarly supplemented with FeSO 4 . The cells were first cultured in iron-rich CDM and then allowed to grow in che-CDM with no added iron to deplete internal iron pools. When these starved cells were inoculated into media with FeSO 4 supplementation, growth in iron-replete and in iron-limiting medium was significantly reduced relative to Schu S4 ( Fig. 2A). We employ ferric salts such as FePPi to study siderophoremediated iron acquisition because the siderophore is specific for ferric iron. We therefore tested growth of iron-starved strains in che-CDM supplemented with limiting or sufficient levels of FePPi and found that the ⌬fupA strain was not deficient in these media but even consistently exceeded growth of Schu S4 under iron limitation ( Fig. 2A). The results presented in Fig. 2A suggested that the fupA mutant had a defect associated specifically with ferrous iron utilization.
In the Lindgren et al. (21) study, deletion of fupA was found to affect growth even in iron-replete medium (2 g/ml FeSO 4 ) if the bacteria had previously been iron-starved (21). The C-CDM medium used in the above study contained additional micronutrient supplements of zinc, copper and manganese. Although we used these supplements when we first started work on this organism (14), we subsequently determined that they are not required for growth by Schu S4 and therefore no longer routinely include them in our media. We tested the fupA mutant in C-CDM and, consistent with the published results, found that growth was affected in ferrous iron-supplemented medium regardless of iron levels (Fig. 2B). However, as shown in Fig. 2B, we found that growth in ferric iron was also affected in this medium. Thus the ⌬fupA strain appeared to lose viability when iron-starved in medium containing these additional metal ions.
Schu S4 and the single mutants showed similar growth kinetics when inoculated without prior iron starvation into che-CDM made iron-replete with FePPi (Fig. 2C). Under ferric iron- limiting conditions, the ⌬fslE mutant showed a significant growth defect as expected, but the ⌬fupA mutant grew as well or better than the parent Schu S4 strain. The double mutant had a slower growth rate and grew to a lower density than the single mutants even in iron-replete media. When iron levels were limiting, the attenuation in growth of the double mutant was greater than that of the ⌬fslE mutant. These results indicated that FslE played a primary role in iron acquisition under ferric iron limitation and that FupA influenced the process in some manner.
We next assessed growth of the mutants on agar. Serial dilutions of washed bacteria were spotted onto iron-limiting CDM (CDM-Fe) agar and compared with growth on the rich medium MHA. The parent Schu S4 strain grew out to several dilutions on the CDM-Fe plate (Fig. 2D). The ⌬fslE mutant showed barely any growth, as previously reported (16); we observed growth on the outline of the spot at the highest density and a faint shadowy growth with increasing dilution that was qualitatively very different from the dense growth seen with Schu S4. The ⌬fupA mutant had a different attenuated phenotype than the ⌬fslE mutant, with heavy growth similar to Schu S4 at the highest density spots, and no growth at all with increasing dilution. The ⌬fslE ⌬fupA mutant showed the most severe growth phenotype, with no hint of growth at even the highest density  JULY 20, 2012 • VOLUME 287 • NUMBER 30 spot. The different growth phenotypes of the ⌬fslE and ⌬fupA mutants suggested that the paralogs function in different capacities in iron acquisition.

Iron Transport by Francisella Outer Membrane Proteins
Complementation of the fslE and fupA Mutants-We sought to better clarify the roles of the two paralogs by reintroducing the fslE gene or the fupA gene individually back into the chromosome of the different mutants. In these strains, the introduced gene was placed under control of the fslA promoter on a suicide plasmid, rendering expression iron-responsive. The expression of the paralogs in the singly complemented ⌬fslE ⌬fupA mutant is shown in Fig. 1C. Although iron-regulated, these constructs expressed detectable levels of protein even in iron-replete medium, and the proteins accumulated to significant levels after prolonged growth.
We found that each of the single mutants could be fully rescued for growth on iron-limiting agar by the corresponding wild type gene ( Table 1), indicating that the defective growth phenotype was a result of the specific mutation. Furthermore, we found that fslE expression in cis partially rescued growth of the ⌬fupA as well as the ⌬fupA ⌬fslE mutants, but restoration of fupA expression in cis could not rescue growth of the ⌬fslE or the ⌬fslE ⌬fupA double mutant on these plates. These results were consistent with a role for both fslE and fupA in growth on iron-limiting agar, with FslE appearing to be more important.
Siderophore-mediated 55 Fe Uptake-We developed a filter plate assay that could permit us to follow bacterial uptake of ferric 55 Fe using siderophore purified from LVS. Schu S4 bacteria grown either in iron-limiting or iron-replete CDM were washed in che-CDM to remove secreted siderophore. They were then incubated for varying periods of time with 55 Fe in complex with purified siderophore in the presence of citrate, and the 55 Fe taken up by cells was quantified by scintillation counting. As shown in Fig. 3A, a steady increase in accumulation of 55 Fe by Schu S4 occurred in bacteria grown under iron limitation, whereas cells grown in iron-replete media showed little uptake. This is consistent with the derepression of iron uptake genes in response to iron limitation. The accumulation of 55 Fe was dependent on the added siderophore; in the absence of purified siderophore, a far lower level of 55 Fe was taken up by cells (Fig. 3B). This basal uptake was likely mediated by residual endogenous siderophore produced by the bacteria because a siderophore-deficient ⌬fslA mutant showed even lower basal uptake of 55 Fe in this assay (data not shown). Siderophore-mediated 55 Fe uptake was inhibited by 10 M carbonyl cyanide m-chlorophenyl hydrazone (CCCP), demonstrating that this was an active process dependent on the proton motive force. Reactions incubated at 4°C also showed poor 55 Fe transport. These assays demonstrated that even though F. tularensis lacks a TonB homolog, the siderophore uptake process in the organism is energy-dependent as in other Gram-negative bacteria (32).
We tested the fslE and fupA mutants for siderophore-mediated 55 Fe uptake. The kinetics of uptake are shown in Fig. 3C, and the deduced rates of uptake are presented in Fig. 3D. The ⌬fslE mutant showed no appreciable accumulation of 55 Fe, consistent with a key role for FslE as receptor mediating siderophore-mediated iron uptake (Fig. 3C). Unexpectedly, the ⌬fupA strain was extremely efficient at siderophore-mediated uptake of 55 Fe, with a significantly higher rate of uptake than the parent strain Schu S4 (Fig. 3C). In the ⌬fupA ⌬fslE mutant, however, the siderophore uptake was negligible, similar to the single fslE mutant. These results indicated that FslE is essential for siderophore-mediated iron uptake, whereas FupA is dispensable for this process. The higher level of uptake observed in the ⌬fupA mutant might be ascribed to the high expression of fsl genes in this strain.
We tested the singly complemented ⌬fslE ⌬fupA strains in the siderophore-mediated 55 Fe uptake assay. Fig. 3E shows the rates of 55 Fe uptake for the different strains. The ⌬fslE ⌬fupA strain with control vector integrated in the chromosome had a negligible rate of uptake. This was significantly enhanced when the fslE gene was reintroduced into the strain. The fupA complement had very low levels of uptake, similar to the parental strain. These experiments definitively demonstrated that (i) fslE can function independently of fupA as a siderophore receptor and (ii) that FupA is not itself a receptor for the siderophore.
Transport of Ferrous Iron in Schu S4 and Mutants-Siderophores are specific for ferric iron, and siderophore uptake mechanisms typically are distinct from mechanisms for acquisition of the reduced (ferrous) form of iron. We sought to characterize ferrous iron uptake in Schu S4 and determine whether the FslE and FupA proteins play a role in this process.
We grew Schu S4 in iron-limiting medium and followed transport of 55 Fe by these cells in the presence of ascorbate to keep the iron in the reduced form. A plot of the uptake with 100 nM iron over a period of 20 min is shown in Fig. 4A. The uptake was initially linear, and accumulation reached a plateau at later times (data not shown). Uptake at the early time points could be used to plot the transport rate. The transport rates for iron concentrations ranging from 12.5 nM to 6.2 M plotted as a function of iron concentration followed a curve consistent with Michaelis-Menten kinetics, as shown in Fig. 4B. Maximal uptake (V s ) of 21.94 pmol min Ϫ1 OD 595 g Ϫ1 was reached at ϳ3 M iron, whereas half-maximal uptake (corresponding to K S ) was seen at 357 nM iron. Concentrations of 100 nM and 3 M, taken to represent high and low affinity uptake, respectively, were used in subsequent experiments to characterize ferrous iron uptake further.
We compared the kinetics of 55 Fe (ferrous) uptake of Schu S4 grown under iron limitation and in iron-replete media. Bacteria grown under iron-limitation had a 17-fold higher rate of uptake at 100 nM ferrous iron and a 4-fold higher rate of uptake at 3 M

TABLE 1 Complementation for growth on iron-limiting agar
Ten-fold serial dilutions of the in cis-complemented bacteria were spotted on ironlimiting CDM agar, and the results after 4 days incubation at 37°C are presented. A minus sign represents absence of growth, and a plus sign represents growth. The numbers in parentheses indicate the number of dilutions for which growth was observed on the plate.

Parent strain
a Control vector integrated in chromosome. b Represents growth only on the outline of the first spot. iron (Fig. 5A). All of the subsequent experiments were carried out with bacteria grown under iron limitation. The protonophore CCCP had a drastic effect on ferrous iron uptake at both 100 nM and 3 M concentrations (Fig. 4C). Incubation at 4°C did not support transport at either concentration. This indicated that ferrous iron uptake is also dependent on proton motive force and on temperature.
We assessed the ferrous uptake capability of different mutants. The fupA mutant was significantly defective for uptake at 100 nM iron concentration (Fig. 4A, Fig. 5). The difference in uptake was much less pronounced but remained significant at 3 M iron concentration (Fig. 5, A and C). To get a better understanding of the uptake deficiency of the fupA mutant, we compared transport rates of Schu S4 and the fupA mutant with 55 Fe concentrations ranging from 100 nM to 3.2 M. The relative rates changed with concentration with the very large difference in rates at low concentrations tapering with increasing iron concentration (Fig. 5B). These results clearly demonstrate that the fupA mutant is defective in ferrous iron uptake and specifically high affinity ferrous iron uptake.
In contrast to the fupA mutant, the ⌬fslE mutant had rates of ferrous uptake that were comparable with Schu S4 both at 100 nM and at 3 M (Fig. 5C). The ⌬fslE ⌬fupA mutant was, however, even more affected than ⌬fupA alone, displaying a much lower transport rate than the other strains even at the higher iron concentration. Thus the double mutant is defective even for the low affinity uptake process. We examined the singly complemented double mutant strain to see whether restoring FupA expression would abrogate the uptake defect. As shown in Fig. 5D, reintroduction of fupA on a plasmid restored both high affinity and low affinity ferrous uptake to the mutant, dem-onstrating that the ferrous iron transport defect was entirely due to the loss of fupA.
We also examined the transport rate for the siderophoredeficient fslA mutant GR206 at 100 nM 55 Fe 2ϩ and found that it was not statistically different from the Schu S4 rate (107 Ϯ 3%). This ruled out the formal possibility that the siderophore was in some way involved in the high affinity ferrous iron uptake process.
These experiments clearly demonstrated that FupA was necessary for optimal ferrous iron uptake in Schu S4. At higher iron concentrations, other transporters could partially compensate for FupA function, but the high affinity uptake important for iron acquisition at low iron concentrations was lacking in the mutant. Our results also demonstrated that FslE played no significant role in high affinity ferrous iron uptake.

Rescue of Growth on CDM Agar by Different Forms of Iron-Iron in an aerobic environment is largely in the ferric form.
Because the ⌬fupA mutant is capable of siderophore-mediated ferric iron uptake, it was not clear as to why it had a growth defect on CDM-Fe agar, which presumably contains trace amounts of iron in the ferric form. We therefore investigated the ability of the two different forms of iron to rescue growth of the mutants on CDM-Fe agar. Supplementation of the agar with 7 M FeSO 4 (the normal iron supplement in CDM) completely rescued growth of the ⌬fslE mutant, whereas growth of the ⌬fupA mutants was only partially rescued by the iron supplement after 4 days of growth (Fig. 2B). FePPi at a similar concentration (6.8 M) in the CDM agar also rescued the ⌬fslE mutant completely. Although it did not completely rescue growth of the fupA mutant, the bacteria grew to a higher dilution than with FeSO 4 . Increasing the ferric supplement concentration 10-fold to 68 M allowed a more complete, FIGURE 3. FslE enables siderophore-mediated 55 Fe transport. A, growth under iron-limitation promotes siderophore-mediated uptake ability. Schu S4 bacteria grown in iron-replete or iron-limiting medium were incubated with 55 Fe siderophore mix, and the incorporation of 55 Fe over time was determined by scintillation counting. B, siderophore-mediated ferric iron uptake is energy-dependent. Schu S4 bacteria grown in iron-limiting media were incubated with 55 Fe in the presence or absence of siderophore, and incorporation was monitored over time. Parallel reactions with siderophore were carried out at 4°C or following pretreatment of bacteria with 10 M CCCP. C, FslE is required for siderophore-mediated 55 Fe uptake. The incorporation of 55 Fe siderophore was followed in Schu S4 and mutant strains after growth under iron limitation. D and E, rates of siderophore-mediated 55 Fe transport in strains. D, Schu S4 and mutant derivatives. E, Schu ⌬fslE ⌬fupA mutant complemented in cis with control vector sequences or with fslE or with fupA. All of the values are expressed as the means Ϯ S.E. *, p Ͻ 0.001; **, p Ͻ 0.005 JULY 20, 2012 • VOLUME 287 • NUMBER 30 albeit slow recovery of growth in contrast to supplementation with an equivalent concentration of FeSO 4 (data not shown).

Iron Transport by Francisella Outer Membrane Proteins
The difference in growth on ferric and ferrous iron supplemented CDM agar was observed only when freshly poured plates were used for the assay; with plates that were more than a few days old, the distinction between ferrous and ferric supplemented plates was hard to discern. One explanation is that there is some level of interconversion of the iron between ferrous and ferric forms in the agar that is influenced by a complex set of factors including oxidation following exposure to air on the surface and reduction promoted by cysteine in the medium (CDM contains 0.02% cysteine as supplement because the bacteria are defective in sulfate assimilation and require cysteine for growth (26)). Cysteine is known to reduce ferric salts in bacterial media (33). We found that increasing cysteine levels in the CDM agar reduced the extent of growth promotion of the ⌬fupA mutant by ferric iron, presumably because more of the iron was converted to the ferrous form (data not shown). We propose that the iron in agar plates used for growth of F. tularensis is in a dynamic equilibrium between ferrous and ferric forms and that optimal growth requires functional acquisition systems for both. This may also explain the different growth phenotypes of the ⌬fslE and ⌬fupA mutants on agar. Furthermore, in a shaking iron-limiting liquid culture, the oxidized form is probably more prevalent; this could explain why the fslE mutant had a strong growth phenotype in medium containing FePPi, but the growth defect of the fupA mutant in medium supplemented with FeSO 4 was more modest.
The fslE Paralog fupB Does Not Play a Key Role in Iron Acquisition-FslE and FupA are the only two of five paralogs encoded in the Schu S4 genome for which any function is known. A third paralog, fupB (FTT0919), is encoded adjacent to fupA and likely forms part of the same transcription unit. We considered the possibility that it may also in some way play a part in iron acquisition. We therefore created deletion mutants Schu ⌬fupB (GR243) and Schu ⌬fupA ⌬fupB (GR242). On Western blots with rat antiserum, the Schu S4 and ⌬fupA strains showed a reactive ϳ45-kDa band corresponding to FupB; this band was missing in lysates of the ⌬fupB mutant (Fig.  1B). The fupB mutant expressed FslE and FupA at levels similar to Schu S4.
The ⌬fupB mutant grew similarly to Schu S4 on iron-limiting agar, whereas the ⌬fupA ⌬fupB mutant behaved similarly to the ⌬fupA mutant (data not shown). Siderophore-mediated 55 Fe uptake by the ⌬fupB mutant was not significantly different from Schu S4 (Fig. 3D), whereas the rate of high affinity ferrous iron uptake with 100 nM 55 Fe was 78% of Schu S4 (Fig. 5A). Overall, the fupB deletion did not have large effects on the iron acquisition capability of the bacteria in our assays. We conclude that not all fslE paralogs play a major role in these iron acquisition processes.
Optimal Intracellular Growth Requires at Least One Paralog Functional in Iron Acquisition-We tested the ability of the fslE and fupA mutants to replicate intracellularly in the mouse macrophage-like cell line J774A.1. We assessed bacterial numbers at different times following infection of the monolayers. As shown in Fig. 6, all the strains had similar intracellular numbers at 2 h, indicating a similar level of invasion. The intracellular numbers of the parental Schu S4 strain titer increased by ϳ2.5 logs over a period of ϳ24 h. The ⌬fslE mutant did not show differences relative to Schu S4 at 9.5 and 23 h. The ⌬fupA mutant showed a relative defect at 9.5 h but at 23 h had caught up to Schu S4. The double mutant, however, was significantly reduced in titer at every time point after invasion. This may reflect the general slow growth of the strain that we observed in vitro, likely caused by the poor ability to assimilate iron.
The mammalian intracellular milieu is considered to be an iron-limiting environment. We found that that there was a growth defect in the fupA mutant in J774A.1 cells at 9.5 h, but this did not persist to later times. A fupA mutant was previously found to be defective for intracellular replication in mouse peritoneal macrophages at 15 h after infection, although intracellular bacterial titers at later time points in this assay were not reported (27). Furthermore, biological differences between J774 tissue culture cells and peritoneal murine macrophages may contribute to differences in experimental outcome. Our results indicated that although neither of the iron uptake mechanisms mediated by the two paralogs was essential, at least one needed to be functional for intracellular replication in the J774A.1 cell line. FslE and FupA govern alternate pathways for iron acquisition in this system.
Both Paralogs Synergistically Contribute to Virulence-Schu S4 is highly virulent in mice, requiring less than 10 CFU for a lethal infection. We evaluated the ability of the different mutants to kill C57BL/6J mice following infection by the subcutaneous route (Fig. 7). We initially challenged mice with ϳ25 CFU (Fig. 7A); at this dose, Schu S4-infected mice all succumbed within 6 days of infection. The fslE mutant was equally virulent in this model. When infected with the ⌬fupA mutant, two of five mice succumbed within 6 days, whereas the remaining three survived. Mice infected with the ⌬fslE ⌬fupA mutant all survived. The survivors from both groups were immune to subsequent challenge with 12 CFU Schu S4.
In a second experiment, we tested higher doses of the ⌬fupA mutants (Fig. 7B). At doses of 528 and 2112 CFU of the ⌬fupA strain, three of four mice succumbed, whereas all of the mice   JULY 20, 2012 • VOLUME 287 • NUMBER 30 receiving the highest dose succumbed, albeit with some delay in time to death. The mice survived infection with the ⌬fslE ⌬fupA mutant at all doses tested. Surviving mice from both groups were largely resistant to subsequent challenge with 500 CFU Schu S4, with just one of four mice in two of the groups succumbing to the challenge.

Iron Transport by Francisella Outer Membrane Proteins
We also tested a ⌬fupB mutant and found it to be as virulent as wild type (data not shown). Upon challenge with a ⌬fupA ⌬fupB mutant, three of four infected mice survived and were immune to further Schu S4 challenge. These data suggest that fupB does not significantly contribute to virulence, in agreement with earlier studies (27).
Previous reports with Schu fupA mutants found the strains to be highly attenuated following intradermal infection of Balb/c mice (21,27). Our ⌬fupA mutant was far less attenuated in our model. This may reflect differences in the route of infection, and additionally, the mouse genotype is known to influence the outcome (4,34).
In our experiments the ⌬fslE ⌬fupA mutant demonstrated extreme attenuation relative to the single fupA mutant. Additionally, the survivors demonstrated significant resistance to subsequent Schu S4 challenge. This indicates potential for the combined mutations to be included in development of a live vaccine strain. Our results suggest that the two paralogs involved in iron acquisition synergistically govern virulence in F. tularensis.

DISCUSSION
Several lines of evidence suggest that mechanisms for acquisition of iron in the limiting host environment are critical for establishment of infection by pathogenic Francisella strains. Early studies with LVS showed that intracellular replication is dependent on the availability of iron (35). In addition, the intracellular pathogen was shown to modulate host iron metabolism to increase the labile iron pool (36). Interfering with iron uptake by use of the competitor gallium inhibited bacterial replication in vitro and in vivo (37). Genetic screens and studies with defined mutants have linked the fsl and fupA genes to virulence in different strains (20 -25, 27, 38).
FslE and FupA are paralogous proteins that share 63% identity and 79% similarity and localize to the outer membrane in Schu S4 (39). The PRED-TMBB program (40,41) predicts that these proteins form ␤-barrels with large periplasmic plugs of 152 residues and 201 residues for FslE and FupA, respectively. Our studies with 55 Fe transport in conjunction with the predicted structure of the proteins have confirmed the role of FslE as the siderophore receptor and identified FupA as a high affinity ferrous iron transporter in Schu S4. Although each protein is able to function independently of the other, the two paralogs function in concert to promote growth under iron limitation.
We have demonstrated here that uptake of ferric 55 Fe by FslE is dependent on the siderophore. Siderophore-dependent iron transport by TonB-dependent outer membrane transporters has been characterized in a number of Gram-negative bacteria (19). How the transport is facilitated by FslE across the outer membrane in F. tularensis in the absence of a TonB complex remains to be determined. An analogous TonB-independent siderophore uptake system has been recently identified in Legionella pneumophila (42), suggesting that this paradigm may extend to other bacteria lacking TonB.
We have established that the proton motive force is important in facilitating the siderophore uptake process in F. tularensis. The fsl operon in F. tularensis encodes two members of the major facilitator superfamily of transporters, fslB and fslD. In a recent study with siderophore uptake in L. pneumophila, the authors noted that the L. pneumophila inner membrane transporter LbtC is a close homolog of FslD (43). Based on this similarity, FslD may be the inner membrane transporter involved in the siderophore uptake process in F. tularensis. In the TonBdependent siderophore uptake systems, transport across the outer membrane relies on the proton motive force (32). Whether the proton-motive force is important for transport across only one or across both membranes in F. tularensis is not yet clear.
The involvement of FupA in ferrous iron uptake was suggested by growth assays where the nature of the iron supplement influenced the growth phenotype, but the function was most clearly defined on the basis of 55 Fe transport. Growth assays can be confounding because of the inherent tendency of ferrous iron to oxidize in an aerobic environment. The presence of cysteine in the medium may also contribute to alterations in the relative levels of ferrous and ferric iron. As such, the forms of iron may be in flux over the period of growth. Although siderophore-mediated uptake plays a prominent role, our results suggest that both ferrous uptake and ferric uptake contribute to optimal growth on iron-limiting agar. fupA expression is not regulated by iron levels, and ferrous iron transport by FupA may well be the default iron uptake mechanism in Schu S4. Under ferric iron-replete conditions, there may be sufficient ferrous iron for uptake by FupA because of the presence of cysteine in the medium. In a fupA mutant, siderophore-mediated ferric iron uptake or low affinity ferrous iron uptake by FslE may compensate for the loss of the ferrous uptake pathway. This could explain the reported lower intracellular free iron pool and the deregulation of fsl gene expression in the ⌬fupA mutant (21).
To our knowledge, this is the first report of an outer membrane ferrous iron transport protein in Gram-negative bacteria, and our studies suggest that specificity for ferrous iron is also a feature of outer membrane transport. A manganese-specific outer membrane channel protein has been recently characterized in Bradyrhizobium japonicum (44). Selective channels, rather than nonspecific porins, may be more widely involved in the transport of different divalent cations across the outer membrane.
We have shown that siderophore-mediated iron uptake capability is enhanced following growth under iron limitation, consistent with ferric uptake regulation of fsl genes in response to iron levels. That FupA mediates high affinity ferrous iron uptake but is not regulated by iron or ferric uptake regulator is surprising; FupA maintains a relatively steady rate of ferrous iron uptake over a wide range of iron levels and may play a housekeeping role in the transport of iron across the outer membrane. Subsequent transport of ferrous iron across the inner membrane is likely mediated by the ortholog of the ferrous iron transporter FeoB (45). FeoB in F. tularensis is Fur-regulated 3 and would provide the selective step in ferrous iron uptake. Because fupA mutation leads to attenuation of virulence of Schu S4, ferrous iron is probably the predominant or preferred source for the pathogen within the host. Siderophore-mediated iron uptake plays a less significant role, and thus the fslE mutant shows no loss of virulence. The fupA mutant retains some level of virulence, possibly because compensatory FslE overexpression permits salvage of ferric iron. However, the loss of both uptake pathways is debilitating to the pathogen.
A recent study in F. novicida identified a role for FupA (referred to as a lipoprotein, FopC) in down-regulation of the interferon-␥ signaling pathway as a mechanism for enhancing virulence (46). The authors also proposed that FupA helps to maintain outer membrane integrity based on their findings of cellular proteins in the culture supernatant of a fupA mutant. Our studies identified a ferrous uptake role for FupA, but additional functions related to stabilization of outer membrane proteins, including FslE, are also possible.
FupA was initially identified as a virulence factor based on the fact that fupA mutation results in attenuation in a mouse model of tularemia (27). In a subsequent study, a siderophoredeficient fslA mutant was reported to be virulent, whereas the fupA mutant was not (21). In our experiments, we found that the ⌬fslE mutant was also as virulent as wild type, as was a ⌬fslA mutant (data not shown). Our ⌬fupA mutant showed some attenuation in virulence, but not as much as that observed in published studies (21), and this may be related to the model of infection employed. The ⌬fslE ⌬fupA mutant had defective growth both in culture and in intracellular growth and was extremely attenuated in virulence in our mouse model. Our data suggest that the two iron acquisition systems work synergistically to support pathogen survival in the host.
In most strains of Francisella, fupA and fupB are independent and adjacent genes on the chromosome. These genes have undergone a recombination to form the hybrid fupA/B gene in LVS (27), which nevertheless contributes both to iron acquisition and virulence to some extent (47). Full virulence could be restored by returning the full-length fupA to the genome (48). We have shown here that fupB does not significantly contribute to iron acquisition. The fupB mutation also did not alter virulence, as has been previously reported (27).
F. tularensis Schu S4 has a relatively small genome of 1.89 Mb and 2097 CDS as most recently catalogued by the Pathosystems Resource Integration center (PATRIC) database (49). Sixteen F. tularensis and eight F. novicida genomes have been analyzed at PATRIC. Although the pathogenic tularensis strains have accumulated a number of gene inactivation events that are considered to be unimportant to the evolutionary niche that they occupy, they have retained an intact set of iron acquisition related genes found in all naturally occurring Francisella strains this far (50,51). These include the fur repressor and fsl operon, the fupA locus, and the feoB gene encoding the ferrous iron transporter. Conservation of these genes may reflect the importance of iron acquisition mechanisms to the pathogenic lifestyle of F. tularensis.