Glutathione Provides a Source of Cysteine Essential for Intracellular Multiplication of Francisella tularensis

Francisella tularensis is a highly infectious bacterium causing the zoonotic disease tularemia. Its ability to multiply and survive in macrophages is critical for its virulence. By screening a bank of HimarFT transposon mutants of the F. tularensis live vaccine strain (LVS) to isolate intracellular growth-deficient mutants, we selected one mutant in a gene encoding a putative γ-glutamyl transpeptidase (GGT). This gene (FTL_0766) was hence designated ggt. The mutant strain showed impaired intracellular multiplication and was strongly attenuated for virulence in mice. Here we present evidence that the GGT activity of F. tularensis allows utilization of glutathione (GSH, γ-glutamyl-cysteinyl-glycine) and γ-glutamyl-cysteine dipeptide as cysteine sources to ensure intracellular growth. This is the first demonstration of the essential role of a nutrient acquisition system in the intracellular multiplication of F. tularensis. GSH is the most abundant source of cysteine in the host cytosol. Thus, the capacity this intracellular bacterial pathogen has evolved to utilize the available GSH, as a source of cysteine in the host cytosol, constitutes a paradigm of bacteria–host adaptation.


Introduction
Francisella tularensis is a Gram-negative bacterium capable of causing the disease tularemia in a large number of mammalian species. It is a highly infectious bacterium that can be transmitted to humans in numerous ways. Four different subspecies (subsp.) of F. tularensis that differ in virulence and geographic distribution exist. These are designated subsp. tularensis (type A), holarctica (type B), novicida and mediasiatica. F. tularensis subsp. tularensis is the most virulent subspecies causing a severe disease in humans, whereas F. tularensis subsp. holarctica causes a similar disease but of less severity [1]. Because if its high infectivity and lethality, F. tularensis is considered a potential bioterrorism agent [2].
F. tularensis is a facultative intracellular bacterium that infects and replicates mainly inside macrophages, but which can also infect hepatocytes, endothelial cells, epithelial cells, fibroblasts, chicken embryos, and amoebae [3]. Although the molecular mechanisms by which Francisella adapts to life inside host cells are not well understood, a number of genes that are necessary for Francisella to survive inside its host and cause disease are known [4]. These include genes located in the Francisella pathogenicity island (FPI) [5,6,7,8,9,10,11], the genes encoding the regulatory proteins MglA, SspA, and PmrA, which regulate expression of the FPI [5,12,13] and the genes responsible for lipopolysaccharide (LPS) production [14,15,16,17]. Furthermore, several genomescale screening approaches have been formerly utilized in F. tularensis subsp. novicida, tularensis, and holarctica, to identify additional genes that are important for replication inside macrophages or survival in mice [7,18,19,20,21].
In the present work, we designed an in vitro negative selection method, based on the use of a bacteriostatic antibiotic, to recover intracellular growth mutants directly from a pool of mutants. A similar approach has been reported earlier to select intracellular growth-deficient mutants of Listeria monocytogenes [22].
One of the mutants isolated by this procedure (see Table S1) showed a drastic intracellular growth defect. This mutant had an insertion into gene FTL_0766, encoding a putative gammaglutamyl transpeptidase, GGT.
We focus here on the role of this gene in the virulence of F. tularensis. GGT's are involved in hydrolysis of c-glutamyl compounds such as glutathione (GSH). GSH is a non-ribosomal tripeptide (L-c-L-glutamyl-L-cysteinyl-glycine) present in almost all eukaryotic cells and in some prokaryotes. In eukaryotes, GSH is essential for cell homeostasis and GSH-deficiency has been associated with various severe diseases [23]. In mammalian tissues, GSH is generally found in the cytosol at mM levels (1-10 mM) [24]; and GSH biosynthesis and metabolism proceeds through the c-glutamyl cycle [25]. In all living organisms, GGT catalyzes the first step in the degradation of GSH, which involves the cleavage and transfer of the c-glutamyl moiety from GSH to an amino acid acceptor (or hydrolysis to release glutamate) and release of cysteinyl-glycine.
The four subspecies of F. tularensis require cysteine for growth; and defined media, containing cysteine and other amino acids essential to support growth of F. tularensis strains, have been developed [26,27]. In silico analysis of the F. tularensis SCHU S4 genome suggests that the specific requirement for cysteine is due to a nonfunctional pathway for sulfate assimilation, resulting from a pseudogene encoding adenylylsulfate kinase [27].
We demonstrate here that FTL_0766 of F. tularensis encodes a genuine GGT involved in the metabolism of c-glutamylcontaining peptides. GGT allows the utilization of c-glutamyl peptides as a source of cysteine during intracellular multiplication of the pathogen, and is thus critical for virulence.

Selection of F. tularensis mutants unable to multiply in J774 macrophages
In the present work, we adapted an in vitro negative screening procedure, based on a penicillin selection, for the isolation of F. tularensis mutants defective in their ability to replicate intracellularly. F. tularensis has been shown to be sensitive to a series of antibiotics [28]. Although penicillins are generally not very active on F. tularensis, cephalosporins are more active. In particular, cefotaxime, a third generation cephalosporin with bacteriostatic activity, was shown to be active on F. tularensis [28]. This antibiotic was furthermore demonstrated to enter eukaryotic cells [29]. Therefore, since b-lactam antibiotics such as cefotaxime, kill only growing bacteria, a protocol was designed to isolate intracellular growth mutants from banks of transposon insertion mutants.
Banks of mutants were generated in the F. tularensis Live Vaccine Strain (LVS), using the in vivo Himar strategy (HimarFT) [30]. The thermosensitive plasmid pFNLTP16 H3 (which carries a kanamycine resistance cassette between the inverted repeats) was introduced into LVS by electroporation. Transformants were first selected for inheritance and maintenance of the plasmid at permissive temperature (30uC). Then, cultures were shifted at nonpermissive temperature (39uC) to inhibit subsequent plasmid replication, and in the presence of kanamycin to select chromosomal integration of the resistance cassette (see Materials and Methods).
Wild-type LVS multiplies in murine macrophage-like cells J774 with an intracellular doubling time of ca. 6 h. However, the presence of 1.5 mg ml 21 cefotaxime in the culture medium efficiently inhibited multiplication of intracellular LVS ( Figure S1). We infected J774 cells with pools of mutant bacteria at an MOI of 100 bacteria/cell in the presence of 1.5 mg ml 21 cefotaxime. The surviving bacteria were recovered at selected intervals after infection by plating onto chocolate agar plates (Text S1; Figure  S1). Clones isolated by this procedure were tested by Southern blot to select single insertion mutants ( Figure S2). Nucleotide sequence analysis of 30 single transposon insertions led to the identification of 10 distinct chromosomal regions (Table S1, Figure S2).
The fact that the largest numbers of mutant hits (10/30) were found in the ggt gene (FTL_0766), and that the mutation led to the most severe intracellular growth defects (see below), prompted us to focus on this mutant. This gene encodes a putative c-glutamyl transpeptidase (GGT), a protein of 601 amino acid residues whose sequence is highly conserved among the Francisella subspecies (99.2% identity with the subsp. novicida ortholog FTN_1159; and 98.8% with the subsp. tularensis SCHU S4 ortholog FTT1181c); and significant degree of homology with GGT of Helicobacter pylori (47% identity).

Intracellular growth and virulence of the ggt mutant
Growth of the ggt mutant was undistinguishable from that of wild-type LVS in broth (Mueller-Hinton, not shown or Chemical Defined medium, see below). The kinetics of intracellular multiplication of the ggt mutant was first followed in mouse macrophages over a 48 h-period. The mutant showed a drastic growth defect, with more than a 100-fold reduction of intracellular bacteria in macrophage-like J774 cells after 24 h ( Figure 1A). A severe intracellular growth defect of the ggt mutant was also observed in the RAW macrophagic cell line as well as in murine bone marrow-derived macrophages (BMM) ( Figure 1B and 1C). We then followed intracellular survival of the ggt mutant in human macrophages. The mutant showed a severe growth defect in THP1 macrophages, with more than a 10-fold reduction of intracellular bacteria cells after 24 h, and a down to a 20-fold reduction after 48 h ( Figure 1D).
Infection of RAW cells was followed by thin section electron microscopy ( Figure 2). As expected, a significant bacterial replication was observed in the cytosol of infected cells 24 h post-infection with wild-type LVS. Of 363 cells counted, 33 were infected by bacteria (9.1%) and 64% of the infected cells contained more than 10 bacteria. Bacterial multiplication had also occurred at 24 h in cells infected with the ggt mutant. However, only 5% of the cells were infected by bacteria (34/651 cells) and only 20% of the infected cells contained more than 10 bacteria. These data strongly suggest that the intracellular growth defect of the ggt mutant is due to an impaired cytosolic growth.
The impact of ggt inactivation on bacterial virulence was then evaluated by infecting 6-8 weeks old BALB/c mice with the ggt mutant and LVS, by the intraperitoneal route ( Figure 3 and Table  S1). The median lethal dose was determined to be ,10 1.04 for LVS and ,10 4.8 for ggt using the Probit method [31], that is a .1,000-fold attenuation of virulence.

Functional characterization of the ggt mutant
We confirmed that the intracellular growth defect observed in the ggt mutant strain was due to the specific inactivation of FTL_0766 by performing a functional complementation of the gene. For this, the recombinant plasmid pKK-ggt, carrying the wild-type ggt allele was introduced into the ggt mutant strain. As shown in Figure 1A, 1B, and 1C, the kinetics of intracellular multiplication of the complemented strain, monitored in J774, RAW, bone marrow macrophages, and THP1 cells over a 48 h period, was indistinguishable from that of wild-type LVS harboring the plasmid without insert. In mice, complementation was also observed but appeared to be only partial, possibly to due in vivo instability of the complementing plasmid ( Figure 3).
GGT activity was determined by measuring cleavage and transfer of a c-glutamyl moiety, with a standard assay used to monitor blood GGT (performed at the Hospital Necker facility). We monitored c-glutamyl transpeptidase activity in whole cell lysates of LVS and of the ggt mutant strain. Inactivation of ggt gene resulted in very low GGT activity (1.960.6 U/l) whereas the

Author Summary
The role of nutrient acquisition systems in survival and multiplication of intracellular bacterial pathogens within infected cells is yet poorly understood. The data presented here suggest that Francisella tularensis, a highly infectious facultative intracellular bacterium, is capable of utilizing glutathione (GSH) and c-glutamyl-cysteine peptides present in the cytosol of infected host cells. An in vitro negative selection method, based on the use of a bacteriostatic antibiotic, to recover intracellular growth mutants directly from a pool of mutants, allowed us to select one mutant in a gene encoding a c-glutamyl transpeptidase (GGT). The mutant strain showed impaired intracellular multiplication and was strongly attenuated for virulence in mice. The cleavage of these cysteine-containing peptides by GGT activity provides thus the essential source of cysteine required for intracellular multiplication. The capacity F. tularensis has evolved to utilize GSH, the most abundant source of cysteine in the host cytosol, constitutes a model of bacterial adaptation to intracellular lifestyle.
activity detected in wild-type LVS was of 40.7560.25 U/l. This shows that FTL_0766 encodes a c-glutamyl transpeptidase and that no other protein with c-glutamyl transpeptidase activity exists in F. tularensis LVS. No GGT activity was detected in culture supernatants of LVS and mutant strains, indicating that, under the conditions we tested, the enzyme is not secreted in F. tularensis.
The existence of a capsule-like structure surrounding F. tularensis has been evoked in several early descriptive papers [32] but its nature was never characterized neither biochemically nor genetically. Interestingly, a study which evaluated the serum resistance of designated ''Capsule-positive'' and ''Capsule-negative'' mutants of F. tularensis, concluded that LPS is the major part of the capsule and that serum resistance of the Capsule-positive strains is due to the presence of O-side chains of LPS [33]. Moreover, a series of recent studies [14,16,34] demonstrated a direct link between LPS O-antigen biogenesis and serumresistance. In further support of this notion, a very recent paper studied complement-mediated lysis and complement C3 deposition on the F. tularensis surface using a series of LVS strains with altered in surface carbohydrate structures, including previously designated Cap-mutants. This study showed that F. tularensis resistance to complement-mediated lysis was essentially mediated by the O Ag [35].
Presence of homologs of the genes capB and capC, required for capsule biosynthesis in Bacillus anthracis, in the genome of F. tularensis SCHU S4 [27] have led to the hypothesis that F. tularensis might contain a poly-D-glutamic acid (PGA) capsule. Notably, F. tularensis mutants in capB, and capC homologs have been recently isolated in two in vivo screening approaches [19,20]. While these two studies demonstrated the role of these genes in bacterial pathogenesis, they did not provide any information on the actual function of these genes. In B. anthracis, capD allows the covalent linkage of PGA to the multilayered peptidoglycan and prevents antibodies from gaining access to the bacterium [36]. We have recently reported that F. tularensis does not produce poly-Dglutamate in broth [16]. However, since GGT of LVS shares 35% similarity with B. anthracis CapD, we examined the impact of ggt inactivation on sensitivity to non-decomplemented human serum, using a mutant of LVS devoid of LPS O-antigen and highly sensitive to serum killing (wbtA) as a control [16]. The serum had no effect on the viability of the ggt mutant and the wild-type strain at all tested concentrations (0-20%), while wbtA mutant was efficiently killed at serum concentrations higher than 2% (Figure 4 and Text S1). Hence, we can say that ggt is not involved in F. tularensis serum-resistance.

Utilization of extracellular gluthathione and c-glutamylcysteine as cysteine sources
We next examined if the activity of GGT provides a mean for Francisella to utilize c-glutamyl peptides as a source of amino acids and specifically cysteine, which is an essential nutrient for Francisella [37]. For this, LVS, ggt mutant, and the ggt mutant complemented with the plasmid-encoded ggt gene were cultivated in defined minimal medium (Chamberlain medium, CDM+) [26] or in CDM lacking cysteine (CDM-). The medium devoid of cysteine was supplemented with cysteine, c-glutamyl-cysteine (c-Glu-Cys), or GSH at 100 mM final concentrations. Cultures were incubated at 37uC and growth was monitored over a 12 h-period ( Figure 5A). The three strains grew normally in CDM+ medium while no growth was observed in CDM devoid of cysteine (CDM2), confirming the absolute cysteine requirement for growth. Notably, in the absence of cysteine, the addition of glutathione (GSH) or of c-Glu-Cys partially restored growth of LVS ( Figure 5A), indicating that wild-type LVS is able to use these exogenous compounds as a source of cysteine. In contrast, the ggt mutant failed to use these compounds ( Figure 5A), confirming that the GGT activity encoded by FTL_0766 is the only enzyme capable of hydrolyzing the c-glutamyl-cysteine bond of these molecules. As expected, the ggt mutant containing pKK-ggt was able to utilize the c-glutamyl-substates as the LVS strain. The c-Glu-Cys dipeptide promoted a slightly better growth of LVS than GSH. This observation is compatible with the fact that with GSH as a substrate, the action of an additional peptidase is required to provide cysteine (to cleave the cysteinyl-glycine bond of the dipeptide released after the digestion with GGT).
Notably, E. coli, which can utilize exogenous glutathione as a cysteine source in a GGT-dependent manner [38], possesses three cysteineglycinases (the aminopeptidases PepA, B and N), responsible for the subsequent cleavage of the dipeptide cysteinylglycine [39]. We identified the three orthologs of these aminopeptidases in the LVS genome (FTL_1479, FTL_1956, and FTL_1108, sharing 48%, 48% and 38% amino acid identity with PepA, B and N, respectively). It is thus reasonable to assume that LVS has the capacity to efficiently cleave cysteinylglycine peptides to produce cysteine and glycine.

Functional complementation of E. coli ggt by F. tularensis ggt
To evaluate the capacity of F. tularensis GGT to functionally complement an E. coli GGT-deficient strain, the LVS ggt gene (carried on plasmid pKK-ggt) was introduced into E. coli auxotrophic strains SH795 (thr leu) and SH794 (thr leu ggt) [38] (kindly provided by Dr. H. Suzuki, Kyoto Institute of Technology,   Japan). Colonies of the E. coli strains containing the recombinant plasmid were picked from LB plates and streaked onto M9 minimal agar plates supplemented with threonine and either leucine (M9+Thr+Leu) or c-glutamyl-leucine (M9+Thr+c-Glu-Leu). As shown in Figure 6, SH794(pKK-ggt), but not SH794(pKK214) could grow on (M9+Thr+Leu) as well as on (M9+Thr+c-Glu-Leu). This assay strongly suggests that F. tularensis GGT was able to cleave the c-Glu-Leu bond and, thus, to provide leucine to supplement the leucine auxothrophy of the strain.

Cysteine is required for intracellular multiplication of the ggt mutant
As the ggt mutant is unable to utilize c-glutamyl substrates in vitro, we reasoned that the inability of the mutant to replicate inside macrophages results from this deficiency. We speculated that cellular GSH (or other c-glutamyl peptides) that cannot be used by the ggt mutant strain, serve as a source of intracellular cysteine for F. tularensis. To test this hypothesis, we cultivated J774 macrophages in DMEM supplemented with 5 mM cysteine and infected them with LVS or the ggt strain (non-supplemented DMEM does not contain cysteine but cystine at a 0.22 mM concentration). Addition of cysteine to the cell culture medium promotes increased intracellular growth of the LVS strain compared to growth in J774 cells cultivated in traditional medium ( Figure 5B), indicating that cysteine is limiting in the host cytoplasm. When the ggt strain was used for infection, bacteria did not multiply in cells cultivated in normal medium (as seen previously, Figure 1A) but multiplied intracellularly in J774 cells cultivated in cysteine-enriched medium ( Figure 5B, left panel). In contrast, infection of J774 cells cultivated in the presence of c-glutamyl-cysteine did not promote growth in the mutant strain ( Figure 5B, right panel). The results show that the growth defect of the ggt mutant can be circumvented by addition of cysteine to the medium, strongly suggesting that the natural intracellular source(s) of cysteine is a c-glutamyl-cysteine dipeptide and the abundant tripeptide GSH.
The fact that glutathione is involved in many cellular processes, including oxidative stress defense [40], prompted us to monitor bacterial growth in the presence or absence of 1 mM H 2 O 2 . Growth of LVS and of the ggt mutant were unaffected by the presence of 1 mM H 2 O 2 (see Figure S3), indicating that this concentration of H 2 O 2 is non-toxic to LVS and that the ggt mutation does not increase H 2 O 2 sensitivity. Of note, this value is still 100fold above the physiological concentration of H 2 O 2 that bacteria encounter in phagocytes during the oxidative burst generated by NADPH oxidase, which is in the 1-10 mM range [41,42,43,44]. Thus, it reasonable to assume that that the phenotype of the mutant not related to increased susceptibility to oxidative stress.

GGT-dependent incorporation of cysteine from GSH into proteins
To further demonstrate that cysteine from GSH was indeed incorporated into proteins, we then performed a metabolic labeling assay (see Materials and Methods for details). Briefly, LVS and ggt mutant bacteria were grown in Chamberlain medium devoid of cysteine (CMD2) and supplemented with either 100 mM GSH (LVS+GSH, ggt+GSH) or 100 mM cysteine (LVS+cys, Dggt+cys). 35 S Radiolabeled-GSH (Perkin Elmers) was added to each culture and the suspensions were incubated for 8 h at 37uC with agitation (in the radiolabled c-Glu-Cys-Gly molecule, the isotope is on the sulfur of the central cysteine residue CH 2 SH side chain). Bacteria collected by centrifugation were resuspended in SDS-PAGE loading buffer, boiled and loaded onto 10%-SDS-polyacrylamide gels. After electrophoresis, gels were vacuum-dried and scanned with a Molecular Dynamics Phosphorimager. The autogradiograph shown in Figure 7A corresponds to 12 h exposure. As a control, a second gel loaded with the same extracts was Coomassie-blue-stained ( Figure 7B). As expected, in CMD2 supplemented with GSH, only residual growth was observed with the ggt mutant. In contrast, growth of LVS and the ggt mutant were identical in CMD2 supplemented with cysteine. Radiolabeld protein bands were detected ( Figure 7A) only in the two LVS protein extracts (+GSH or +Cys). This assay demonstrates that GSH utilization specifically requires the action of the ggt gene product and that cysteine from GSH is directly incorporated into proteins during bacterial growth.

Discussion
The role of nutrient acquisition systems in survival and multiplication of intracellular bacterial pathogens within infected cells is poorly understood. Using a cefotaxime-based negative selection, we identified a gene, encoding a c-glutamyl transpeptidase (GGT, FTL_0766) that is absolutely required for intracel-lular multiplication and virulence of F. tularensis LVS. The data presented here suggest that F. tularensis is capable of utilizing glutathione (GSH) and c-glutamyl-cysteine peptides present in the cytosol of infected host cells. The cleavage of these cysteinecontaining peptides by GGT activity thus provides the essential source of cysteine required for intracellular multiplication.

Critical role of GGT in Francisella pathogenesis
Screenings of banks of mutants have been previously reported in F. tularensis [7,18,19,20,21,45]. A number of genes were repeatedly found by these different approaches, in particular genes involved in purine/pyrimidine synthesis. Notably, three of these studies identified the ggt gene [19,21,45] but its properties were not further characterized.
GGT shares some sequence similarity with the CapD enzyme responsible for covalent anchoring of the B. anthracis capsule. The existence of a capsule surrounding F. tularensis has been evoked repeatedly but never persuasively demonstrated experimentally. Therefore, we evaluated the impact of ggt inactivation on serum resistance of LVS as the putative Francisella capsule confers serum resistance [46]. The ggt mutant showed a serum resistance comparable to that of wild-type LVS, demonstrating that GGT is not involved in serum resistance.
Inactivation of ggt had a drastic effect on the ability of LVS to multiply in eukaryotic cells. Strikingly, we showed that the intracellular growth defect of the ggt mutant could be eliminated by providing external cysteine to the culture medium, establishing a direct link between GGT activity and cysteine availability. Also, the capacity of the ggt mutant to cause disease in mice was severely decreased, further demonstrating the critical role of GGT in pathogenesis. GGT allows utilization of c-glutamyl peptides as a source of cysteine The involvement of GGT in the utilization of glutathione or other c-L-glutamyl peptides has been demonstrated in both prokaryotes and eukaryotes [38,47,48,49,50]. While some bacterial species have the capacity to directly import GSH from the growth medium and degrade it into glutamate and cysteine in their cytoplasm [51], species such as H. pylori cannot transport this substrate and must secrete or export GGT to degrade GSH extracellularly. GGT of H. pylori has been described as a virulence factor essential for the establishment of the infection in the mouse model [52,53,54] and as an apoptosis-inducing protein [55]. Furthermore, a role for GGT in the metabolism of glutamine and glutathione has been recently established, providing information on the role of the enzyme in the physiopathology of this extracellular bacterial pathogen [56]. We did not detect any GGT activity in culture supernatants of LVS under the conditions we tested, suggesting that GGT-mediated degradation of cglutamyl compounds occurs in the bacterial cytoplasm. Future work will be required to address the molecular mechanisms of uptake of these compounds in F. tularensis.
Our results show that both GSH and c-glutamyl-cysteine can functionally replace cysteine to support growth of Francisella in broth, and that GGT activity is required for this, by cleaving the cglutamyl-cysteine linkage. We also showed that F. tularensis GGT could functionally complement an E. coli GGT-deficient strain. The F. tularensis GGT was able to cleave the c-Glu bond of the dipeptide c-Glu-Leu, allowing, like the genuine E. coli enzyme, the utilization of the dipeptide as a source of amino acid (providing leucine to supplement the leucine auxothrophy of the strain).
Finally, a metabolic labeling assay confirmed that GSH utilization, as a source of cysteine, specifically requires the action of the ggt gene product. Furthermore, the use of radiolabeled-GSH, carrying the 35 S isotope on the cysteine side chain, demonstrated that cysteine from GSH is indeed incorporated into bacterial proteins during growth.

A model of GSH utilization by intracellular F. tularensis
The infectious cycle of F. tularensis is essentially intracellular. Since the severe intracellular growth defect of the ggt mutant could be completely relieved by the addition of free cysteine to the cell culture medium, it seems most likely that GGT provides a source of cysteine to cytosolic bacteria by degrading internalized cglutamyl-cysteine peptides.
GSH is the most abundant non-protein thiol in mammalian cells and the prevalent low-molecular weight peptide in eukaryotic cells, attaining concentrations of 1 to 10 mM in many different cell types, most of which (85-90%) is present in the cytosol. Biosynthesis of GSH is dependent on the availability of the amino acid precursors glutamate, glycine, and cysteine. The intracellular pool of cysteine is relatively small (0.10-0.25 mM) and cysteine is generally the limiting amino acid for GSH synthesis (the other two precursors, glycine and glutamate, are found in considerable higher intracellular concentration). It is estimated that at the low levels of c-glutamyl-cysteine normally present, 95% is converted to GSH (see for reviews [23,57,58] and references therein). We propose a model for intracellular cysteine acquisition by F. tularensis (Figure 8), in which intracellular F. tularensis uses cglutamyl-cysteine dipeptide and GSH as the primary sources of cysteine. The ggt mutant, unable to cleave the c-glutamyl-cysteine bond, does not obtain sufficient free cysteine (or cystine) from its surroundings to support growth. Therefore, only the addition of mM amounts of free cysteine compensates its enzymatic deficiency. Utilization of cytosolic GSH and/or cysteine-containing dipeptides as a source of cysteine might be used by other intracellular bacterial pathogens.
More generally, acquisition of nutrients should be considered as an integral part of the pathogenicity of intracellular bacteria. This concept is comforted by recent transcriptomic analyses, which revealed the upregulation of many bacterial metabolic and transport functions in infected cells (see for examples [59,60,61,62]). A remarkable example of the interplay between intracellular pathogens and their host has been provided with L. monocytogenes. Indeed, it has been shown that a listerial membrane transporter, Hpt, was essential for acquisition of phosphorylated hexoses within the cystosol of infected cells and was, thus, of major importance for bacterial replication [63,64]. Clearly, intracellular bacterial pathogens have adapted a fine balance between their metabolic needs and the nutrients available (nature and concen-tration), to multiply efficiently and to avoid a premature death of the host cell that would lead to an aborted infection.

Materials and Methods
Strains, media, and chemicals F. tularensis LVS strain was obtained from Dr Anders Sjöstedt (Umeå University, Sweden), and grown at 37uC in Schaedler K3 (Biomerieux SA Marcy l'Etoile, France) broth, Chamberlain defined synthetic medium [26] or Chocolate agar enriched with PolyViteX (chocolate agar) (Biomerieux SA Marcy l'Etoile, France). E. coli was grown in Luria-Bertani (LB) medium at 37uC. When required, medium was supplemented with kanamycin (10 mg ml 21 ). Kanamycin was purchased from Sigma-Aldrich (St Louis, MO). Oligonucleotide primers were synthesized by Eurogenetec. Primers and bacterial strains used in this study are listed in Table S2.

Generation of banks of LVS mutants and in vitro screening procedure
Banks of transposon insertion mutants were generated in LVS, by using the in vivo Himar strategy (HimarFT) [30]. The thermosensitive plasmid pFNLTP16 H3 used (kindly provided by Dr. Dara W. Frank, Medical College of Wisconsin), carries, between the inverted repeats, the npt gene (determining kanamycin resistance) under the control of the F. tularensis LVS groEL promoter. Classically, 10 ml of Schaedler K3 medium was inoculated with LVS transformed with Himar 3, and grown overnight at 30uC with shaking to the until late exponential phase (optical density at 600 nm, 0.7-0.9). Cultures were then grown twice overnight at 39uC and once at 40uC in the presence of kanamycin, to eliminate non-integrated plasmid in these bacteria. Cultures were finally stored in 1.5 ml fractions at 280uC.
The murine macrophage-like cells J774 were infected with pools of ca. 10 7 mutant bacteria (at an MOI of 100 bacteria/cell). After 1 h infection, cells were washed three times with fresh culture medium and incubated for 24 h in DMEM supplemented with 5% (v/v) fetal calf serum, 10 mg m1 21 gentamycin (to kill extracellular bacteria) and 1.5 mg ml 21 cefotaxime (to kill growing intra-cellular bacteria). At given time-points, cells were washed and lysed with 1 ml of sterile water and plated on chocolate agar enriched with PolyViteX. Plates were incubated at 37uC for 3 days and clones obtained were isolated and stocked. Classically, after 24 h of infection, each selection (i.e. one 12-wells plate seeded with 10 5 cells per well) led to the subsequent isolation of approximately 20 colonies on chocolate agar plates containing kanamycin.
Screening procedure. Ninety-five clones were isolated after cefotaxime selection and were tested by Southern blot. In 65 clones, multiple bands were detected (from 2 to 5), indicating that several transposition events had occurred (most likely due to noneven loss of the plasmid after passages at non-permissive temperature, as confirmed by a recent study) [21]. These clones were not further studied.
Single insertions were observed in 30 clones ( Figure S2). Nucleotide sequence analysis of these 30 single transposon insertions led to the identification of 10 distinct chromosomal regions (Table S1). One double insertion mutant was also analyzed. The genes identified are involved in various processes including metabolism and cell division. Notably, in many instances (Table S1), the same insertion was isolated several times (up to 10 times), possibly due to a bias in the selection procedure after repeated passages at non-permissive temperature and/or during intracellular survival in the presence of cefotaxime. Figure 8. A model of utilization of GSH and c-Glu-Cys by cytosolic F. tularensis. F. tularensis requires cysteine for growth. It can utilize GSH (and its oxidized form GSSG), c-Glu-Cys, cysteine, and cystine as sources of cysteine. In eukaryotic host cells, the intracellular concentration of GSH depends on the availability and transport of cystine (oxidized form of cysteine) and cysteine. Cytosolic GSH, the most abundant thiol-containing compound (present in the mM range), and c-Glu-Cys serve as sources of cysteine for intracellular growth of F. tularensis. In wild-type bacteria, GSH and c-Glu-Cys are processed by GGT to produce c-Glu and Cys-Gly or c-Glu and Cys, respectively. The Cys-Gly dipeptide is further processed by other amino-acid peptidases to produce free cysteine and glycine. GGT-negative bacteria are unable to process these compounds, and the available concentration of free cysteine is too low to promote bacterial growth. doi:10.1371/journal.ppat.1000284.g008

Cell cultures and macrophage infection
Macrophage infections were performed as described previously [65] (see Text S1). For determination of numbers of intracellular bacteria, cells were washed three times with fresh culture medium and lysed with 1 ml of sterile water, serially diluted, and plated on chocolate agar.

Functional complementation
The plasmid used for complementation of the ggt mutant, pKKggt, was constructed by amplifying a 2,206 bp fragment (corresponding to the sequence 300 bp upstream of the ggt gene to 100 bp downstream the gene) using primers D1 and D2 (see Table S2), followed by digestion with XbaI and cloning into the XbaI site of plasmid pKK214 [66]. The plasmids pKK214 and pKK-ggt were introduced into LVS and the ggt mutant by electroporation.
Functional complementation of E. coli ggt by F. tularensis ggt We used the same E. coli -F. tularensis shuttle plasmids to evaluate the capacity of F. tularensis GGT to functionally complement an E. coli GGT-deficient strain. The plasmids pKK214 and pKK-ggt were introduced into E. coli auxotrophic strains SH795 (thr leu) and SH794 (thr leu ggt) [38] (kindly provided by Dr. H. Suzuki, Kyoto Institute of Technology, Japan). Colonies of the E. coli strains containing the recombinant plasmid were picked from LB plates and streaked onto M9 minimal agar plates supplemented with thiamine (2 mg/ml), threonine (100 mg/ml), and when indicated either leucine (100 mg/ml) or c-glutamylleucine (100 mg/ml), and incubated for 72 hours at 37uC.

Metabolic labeling
LVS and ggt mutant bacteria were collected from Chocolate agar plates and resuspended in Chamberlain medium devoid of cysteine to a final OD 600 of 0.11. Each 10 ml culture was supplemented with either GSH at 100 mM final (LVS+GSH, ggt+GSH) or cysteine at 100 mM final (LVS+cys, ggt+cys). Radiolabeled glutathione ( 35 S-GSH, Perkin Elmer) was then added to each culture (10 mCi of 35 S-GSH at a specific activity of 30 Ci/mmol per 10 ml culture) and the suspensions were incubated for 8 h at 37uC with agitation. Bacteria were finally collected by centrifugation and the pellets were resuspended in SDS-PAGE loading buffer. Five 5 ml of each fraction (corresponding to 0.15 ml of bacterial culture) were loaded onto 10%-SDSpolyacrilamide gels.
The culture of LVS+GSH and the two cultures in the presence of cysteine (LVS+Cys, ggt+Cys) grew to a final OD of 0.6, while that of ggt+GSH reached only an OD 600 of 0.2. Thus, for LVS+GSH, LVS+Cys and ggt+Cys, each well corresponds to ca. 1.8610 8 bacteria; while for Dggt+GSH, it corresponds only to 0.6610 8 bacteria.
After electrophoresis, gels were vacuum-dried and scanned with a Molecular Dynamics Phosphorimager. The autogradiograph shown corresponds to 12 h exposure.
As control, a second gel loaded with the same fractions was Coomassie-blue-stained.

Gammaglutamyl transpeptidase enzyme assays
Ten ml overnight culture of wild type and the ggt mutant were centrifuged and pellets were suspended in 1 ml of Schaedler K3. 0.5 ml of Zorkonium bead was added and mixture was agitated in 3 series of 30 sec at speed of 40 in FastPrep apparatus (GMI, Inc., Minnesota, USA). Samples were centrifuged (3,5006g, 5 min) and supernatants collected (whole cell lysates). GGT activity assay was realized on 50 ml of each sample (corresponding to 10 9 bacteria), using a standard dosage procedure (Hospital Necker-Enfants Malades, on a Roche/Hitachi analyzer). Values are indicated in U/l, as calculated by the analyser. Four independent measurements were performed.

Animal studies
Groups of five 6-8 week old female BALB/c mice (Janvier, Le Genest St Isle, France), were i.p. injected with 0.2 ml of Francisella in 0.15 M NaCl. The number of viable bacteria used for infection was determined by plating serial dilutions on chocolate agar. The survival of mice was followed for 10 days. Animal experiments were performed according to the INSERM guidelines for laboratory animals husbandry.