Francisella tularensis IglG Belongs to a Novel Family of PAAR-Like T6SS Proteins and Harbors a Unique N-terminal Extension Required for Virulence

The virulence of Francisella tularensis, the etiological agent of tularemia, relies on an atypical type VI secretion system (T6SS) encoded by a genomic island termed the Francisella Pathogenicity Island (FPI). While the importance of the FPI in F. tularensis virulence is clearly established, the precise role of most of the FPI-encoded proteins remains to be deciphered. In this study, using highly virulent F. tularensis strains and the closely related species F. novicida, IglG was characterized as a protein featuring a unique α-helical N-terminal extension and a domain of unknown function (DUF4280), present in more than 250 bacterial species. Three dimensional modeling of IglG and of the DUF4280 consensus protein sequence indicates that these proteins adopt a PAAR-like fold, suggesting they could cap the T6SS in a similar way as the recently described PAAR proteins. The newly identified PAAR-like motif is characterized by four conserved cysteine residues, also present in IglG, which may bind a metal atom. We demonstrate that IglG binds metal ions and that each individual cysteine is required for T6SS-dependent secretion of IglG and of the Hcp homologue, IglC and for the F. novicida intracellular life cycle. In contrast, the Francisella-specific N-terminal α-helical extension is not required for IglG secretion, but is critical for F. novicida virulence and for the interaction of IglG with another FPI-encoded protein, IglF. Altogether, our data suggest that IglG is a PAAR-like protein acting as a bi-modal protein that may connect the tip of the Francisella T6SS with a putative T6SS effector, IglF.


Introduction
Francisella tularensis is a Gram-negative bacterium that causes tularemia [1]. The severity of tularemia is highly variable depending on the route of inoculation of the bacterium and the infecting strain. F. tularensis subspecies tularensis is the most virulent subspecies with a 50% lethal dose estimated to be at most 10 bacteria by the intranasal route for humans [2]. A Live Vaccine Strain (LVS), derived from a F. tularensis subspecies holarctica strain, is widely used to study the pathogenesis of tularemia. F. novicida is another closely related species, which is avirulent for immunocompetent humans but highly virulent in mice. Due to its ability to reproduce the intracellular life cycle of the more virulent subspecies, F. novicida is widely used as a model system to study tularemia [3].
The ability of Francisella strains to cause disease is linked to their ability to replicate within host cells such as macrophages. Upon phagocytosis, Francisella escapes from the vacuole to reach the cytosol, where it replicates rapidly. Escape from the vacuole into the host cytosol is dependent on a genomic island termed the Francisella Pathogenicity Island (FPI) [4,5]. In addition, the FPI is implicated in the inhibition of macrophage pro-inflammatory response (e. g. TNF-α secretion) [6,7]. The island is highly conserved between F. tularensis and F. novicida with greater than 97% identity at the nucleotide level [8,9], and contains 17 genes, 8 of which encode for proteins that share homology with proteins from type VI secretion systems (T6SS) ( [10] and S1 Fig). Two copies of the FPI are present in F. tularensis strains while a single copy is present in F. novicida [8]. In the latter species, another genomic island termed the "Francisella novicida Island (FNI)" demonstrates some similarities with the FPI suggesting it might encode another atypical T6SS ( [10] and S1 Fig).
T6SS are specialized machineries involved in the delivery of toxins and effector proteins to prokaryotic and eukaryotic cells [11][12][13]. They are functionally related to the bacteriophage contractile tail [14,15]. Contraction of an external sheath consisting of TssB and TssC results in the secretion of an inner tube made of stacks of Hcp protein hexamers. The T6SS is tethered to the bacterial envelope by a membrane complex composed of the inner membrane proteins TssL and TssM/IcmF and the outer membrane lipoprotein TssJ [14]. The Hcp tube is surmounted by a complex made of a VgrG trimer capped with the recently identified PAAR protein [16]. PAAR proteins, originally named TagD [17], are characterized by three prolinealanine-alanine-arginine (PAAR) motifs. A zinc atom bound to three histidines and one cysteine is believed to stabilize their three-dimensional structure [16]. This distal VgrG 3 -PAAR protein complex is thought to act as a membrane-puncturing device allowing the delivery of toxins and effector proteins into the target cell. The mode of secretion may involve physical interactions between the secreted effector and the Hcp, VgrG or PAAR proteins [16,18,19].
Phylogenetic analysis of T6SSs led to the classification of the F. tularensis T6SS as a unique evolutionary outlier [20]. Indeed, out of the 13 proteins that define the core of prototypical T6SSs, the FPI lacks obvious homologues for at least 5 of them. Furthermore, while the FPI encodes a TssM/IcmF family protein (PdpB), this protein lacks the conserved Walker A box required to bind ATP and to provide energy to the secretion process [9,21]. Although Francisella possesses a VgrG protein, it is much smaller than prototypical VgrG proteins [9]. IglC has structural homology with Hcp, the stacking unit of the T6SS inner tube. Yet, it is still unclear whether IglC can also form hexameric rings [22]. Importantly, Clemens et al. recently strengthened the T6SS nature of the FPI proteins by demonstrating that IglA and IglB (TssB and TssC homologues, respectively) interact to form the T6SS sheath [23]. While the presence of a functional T6SS encoded by the FPI is now established [9,23,24], it is still largely unknown which FPI genes encode for structural component of the T6SS machinery per se and if the FPI proteins demonstrated to be secreted are indeed effector proteins.
IglG is an 18-kDa protein of unknown function encoded by the FPI. Intriguingly, a LVS ΔiglG mutant presents delayed kinetics of phagosomal escape compared to parental LVS [7]. Thus, in contrast to vgrG, iglA (tssB), iglB (tssC) or iglC (hcp) mutants [25,26], the mutant is not fully confined to the vacuole and replicates efficiently in macrophages [7]. Despite efficient replication, the LVS ΔiglG mutant shows delayed cytopathogenicity and impaired inhibition of TNF-α secretion [7,27]. Moreover, IglG is required for virulence of LVS in mice, indicating that this protein is central to the pathogenicity of subsp. holarctica [7]. However, it is still unclear whether IglG is a structural component of the T6SS or a T6SS effector. We therefore decided to perform a comparative study using F. novicida and F. tularensis subspecies holarctica and tularensis to identify the nature of IglG. Our results demonstrate that IglG is absolutely required for F. novicida escape into the host cytosol, triggering of the cytosolic innate immune responses and replication within macrophages. In addition, we demonstrate that IglG is a member of the DUF4280 family, which comprises features of the recently described PAAR protein family located at the tip of the T6SS. While the PAAR motifs are poorly conserved within this family, 4 cysteine residues are highly conserved. These residues were shown to be essential for IglG function in the T6SS and in virulence and to contribute to the coordination of a metal ion. This study thus defines the members of the DUF4280 as novel PAAR-like proteins. In addition, we found that IglG is unique among PAAR-like proteins in bearing an N-terminal extension required for virulence. We demonstrate that IglG, in an N-terminal domain-dependent manner, interacts with the FPI-encoded IglF protein, suggesting that IglG may act as a cargo to connect the tip of the Francisella T6SS with other T6SS proteins/effectors. Finally, the importance of IglG for the virulence of F. novicida as well as the highly virulent F. tularensis SCHU S4 strain was demonstrated in vivo.

Results
IglG and proteins from the DUF4280 family feature a predicted PAARlike structure To gain insight into the function of IglG, we performed a bioinformatic analysis. A BLAST search using the IglG sequence identified a conserved domain of unknown function (DUF4280; IPR025460; PF14107) in the C-terminus of the protein (residues 59-173) (Fig 1A). DUF4280 is a domain present in numerous proteins of unknown function and is found in more than 250 different bacterial species. One of the DUF4280 proteins, Fjoh_3275/Fte1 from Flavobacterium johnsoniae was recently predicted to adopt a fold closely related to the PAAR domain [28] suggesting that other proteins containing the DUF4280 domain might also share these sequence signatures. We thus performed homology modeling using the Phyre [29] and I-Tasser [30] servers with IglG or the DUF4280 consensus sequence as templates. Both servers clearly identified the structures of Escherichia coli or Vibrio cholerae PAAR proteins as suitable templates for modeling and produced high confidence models (e.g. 85% confidence according to Phyre) for the DUF4280 consensus sequence and for IglG residues 59-173 ( Fig 1B). The   Fig 1. IglG, a DUF4280-containing protein, is a structural homologue of PAAR proteins with four conserved cysteines potentially implicated in metal ion binding. (A) Protein sequence alignment of IglG, DUF4280 (consensus sequence), FTN_0054, VcPAAR (pdb code 4JIV) and EcPAAR (4JIW). Predicted secondary structure elements of IglG are indicated above the sequence (orange rectangles for α-helices and magenta arrows for β-strands). Secondary structures of EcPAAR are indicated below the alignment in blue. Green boxes indicate the PAAR motifs, residues binding Zn in VcPAAR and EcPAAR are shaded in yellow. Conserved cysteines in the sequences of IglG, FTN_0054 and DUF4280 are shaded in red. (B) A comparison of the EcPAAR structure (blue) with homology models of DUF4280 (wheat) and IglG (magenta). Segments colored in green in the three structures correspond to the PAAR motifs. Spheres indicate the position of Zn ion (colored in yellow) in EcPAAR and the putative metal binding site in IglG. The side chains of metal binding residues are shown as ball and stick and colored in yellow (EcPAAR) or red (DUF4280 and IglG). Structures are otherwise colored as in A. (C) Profile hidden Markov model (HMM) generated by iterative searches with IglG sequence using the Jackhmmer software. The positions of the four conserved cysteine residues within IglG are indicated. models generated by I-Tasser for residues 59-173 of IglG and for the DUF4280 consensus sequence were all very similar to that of a conical β-barrel fold (Fig 1B). The three PAAR motifs were, however, not strictly conserved neither in IglG, nor in the DUF4280 proteins ( Fig 1A). Yet, one to one threading [29] with a typical PAAR protein from V. cholerae (VCA0105, here denoted VcPAAR) identified partial sequence conservation for two out of three of these motifs (Fig 1A). Our confidence in the model was strengthened by a molecular dynamics simulation with a 400 ns trajectory on the IglG C-terminus structure presented above. Indeed, the root mean square fluctuation obtained (RMSF = 0.27 + / -0.11) was very similar to the RMSF of a well-characterized structural domain (RMSF SH3 domain = 0.24 + / -0.06) (S2 Fig and S1 Movie). This simulation thus suggests that the predicted protein fold would be stable over time.
We were unable to purify IglG or any of the other DUF4280 proteins tested (FTN_0054 from F. novicida, PA_2375 from P. aeruginosa and ROD_34101 from Citrobacter rodentium) in soluble form and in sufficient amount to perform crystallization assays. This observation is in line with the described low solubility of PAAR proteins [16] and highlights the challenge of working on individual T6SS proteins out of the context of the whole machinery. However, His 6 -tagged IglG could be purified from in E. coli under denaturing conditions and refolded using step-wise dialysis. Importantly, the analysis of the refolded IglG protein by circular dichroism (CD) spectroscopy supported our in silico analysis since the proportion of α-helices and β-strands determined with this method (25% and 22%, respectively) was highly similar to the one predicted in our model (23% and 22%, respectively). Altogether, these data demonstrated that IglG and the members of the DUF4280 family define a novel family of PAAR-like proteins formed by a β-barrel.
IglG is essential for F. novicida phagosomal escape and replication within macrophages To investigate the role of IglG for the T6SS, we made a full-length deletion of iglG (FTN_1314) in F. novicida strain U112. The F. novicida ΔiglG mutant was unable to replicate in J774 macrophages (Fig 2A) as well as in primary bone marrow-derived macrophages. This phenotype resembled that of a mutant lacking the whole pathogenicity island (ΔFPI) [31] and could be complemented by expression of the IglG protein in trans (Fig 2A). The main function of the FPI during the Francisella intracellular life cycle is to promote the escape of Francisella into the host cytosol [32]. We thus investigated the occurrence of phagosomal rupture by using the βlactamase/CCF4 assay [33,34]. IglG was required for vacuolar escape since we did not observe any phagosomal rupture in ΔiglG mutant-infected macrophages at 2 h post-infection (Fig 2B). At this time point, phagosomal rupture could be detected in more than 25% of the cells infected with the WT strain or the complemented ΔiglG mutant strain. As expected, no phagosomal rupture was detected in cells infected by the ΔFPI mutant or by a Δbla mutant lacking a functional F. novicida β-lactamase [35]. In addition, we did not observe any phagosomal rupture in macrophages infected with the ΔiglG mutant strain even at 18 h (S3 Fig) suggesting that in F. novicida, IglG is essential for the escape from the vacuole. Escape of F. novicida into the host cytosol triggers cytosolic innate immune responses, namely secretion of type I IFN [36] and activation of the AIM2 inflammasome [37]. In agreement with the lack of phagosomal escape, the F. novicida ΔiglG mutant failed to induce type I IFN secretion ( Fig 2C) and inflammasomemediated death as measured by real-time propidium iodide incorporation ( Fig 2D). Similarly, the ΔiglG mutant was not cytopathogenic towards J774 macrophages even at 48 h as determined by measuring the release of LDH into the culture supernatant (S4 Fig). Altogether, these data demonstrate that in F. novicida, deletion of iglG leads to the same phenotype as the deletion of iglC (the hcp homologue) or of vgrG, in agreement with a PAAR-like role of IglG in the T6S machinery. IglG is secreted in a T6SS-dependent manner and is required for secretion of the Hcp homologue, IglC Based on its PAAR-like nature, we expected IglG to be secreted as a key component of the Hcp/IglC tube. We first attempted to monitor IglG secretion into host cells by flow cytometry using the β-lactamase reporter [24]. Using this method and in agreement with our previous microscopy-based assay [24], we could not detect translocation of IglG, VgrG, IglI or IglF, although PdpE and IglC were readily secreted (S5 Fig, left panel). Since all fusion proteins were expressed (S5 Fig, right panel), this suggests that the large tag may have an adverse effect on the secretion of some FPI proteins in F. novicida. We thus switched to a recently described in vitro secretion assay in which addition of KCl triggers T6SS-dependent secretion [23]. Upon KCl addition, IglG was secreted into the culture supernatant (Fig 3). As expected VgrG was also specifically secreted during this condition [23], while PdpB, the inner membrane-associated TssM homologue, was not detected in the supernatant (Fig 3). Importantly, IglG secretion was abolished in a ΔvgrG mutant indicating that a functional T6SS is required for IglG export. Similarly, we could not detect secretion of IglG or VgrG in the absence of KCl indicating that environmental cues associated with T6SS activation are required for their secretion.
Deletion of PAAR proteins highly reduces Hcp secretion by V. cholerae and Acinetobacter baylyi [16]. We thus tested whether a deletion of iglG in F. novicida would impair secretion of the Hcp homologue, IglC. Indeed, secretion of IglC was absent in the ΔiglG mutant, although in a few experiments low levels of IglC were still detected in the mutant (Fig 3 and S6 Fig). Similarly, IglC was not detected in the supernatant of the ΔvgrG mutant or in the supernatant of the WT strain if KCl was absent from the growth medium (Fig 3). Importantly, expression of IglG and VgrG in trans restored IglC secretion by ΔiglG and ΔvgrG mutants, respectively ( Fig  3). Altogether, these data, together with the results from the bioinformatic analysis, are consistent with a PAAR-like role of IglG in the F. novicida T6SS. The four cysteine residues conserved in DUF4280 proteins are required for the function of IglG in F. novicida and contribute to metal binding PAAR proteins bind Zn 2+ via 3 histidine and 1 cysteine residues (Fig 1A and 1B). The central Zn 2+ ion is thought to be important to stabilize the protein structure [16]. While the cysteine residue is conserved in IglG (Cys 152) and in the PAAR-like proteins, no histidines are present neither in IglG nor in the DUF4280 consensus sequence (Fig 1A). However, both the sequence alignment and the 3D models suggest that this residue, together with 3 other cysteines (Cys 64, 105, 122 in IglG) could potentially form the metal-binding site of the PAAR-like proteins ( Fig  1A and 1B). The strong conservation of these residues in IglG homologues was highlighted by iterative blast searches using the Jackhmmer software (Fig 1C), suggesting that their presence is essential to the function of these PAAR-like proteins. To determine if IglG is able to bind metal ions, we expressed IglG fused to GST and tested whether the purified fusion protein could bind metal by using inductively coupled plasma mass spectrometry (ICP-MS). Although we did observe some variability in metal binding between the different batches of purified proteins (five independent experiments are shown), iron and zinc were specifically associated with GST-IglG. In contrast, we detected only trace amounts of nickel and copper while the other metals tested (e.g. Mn, Co) were undetectable ( Fig 4A). The observed ratio between iron and zinc binding to IglG ([Fe]/[Zn] = 1.6) suggests that, under our conditions (see discussion), iron binding to GST-IglG is slightly favored over zinc binding. Of note, the same experiments performed on a GST-fusion of a typical PAAR protein from Pseudomonas aeruginosa (PA_0824) demonstrated a similar proportion of molecules binding metal (about 30% of the total purified molecules) but with a strong binding preference for zinc ( The contribution of each IglG cysteine in Fe/Zn binding was evaluated using purified mutant proteins and ICP-MS. Although we could not directly compare the metal-binding capacity of mutant proteins purified on different days, the direct comparison of each mutant IglG protein to the WT IglG protein purified the same day demonstrated a 20 to 40% reduction in metal binding ( Fig 4B). These results suggest that the four cysteine residues contribute to the ability of IglG to bind metal ion but that mutating a single cysteine residue is not sufficient to abolish metal binding in vitro.
To strengthen our findings, we performed a metal binding assay using circular dichroism (CD) spectroscopy [38]. CD spectra of IglG protein purified under denaturing conditions and refolded as described above ( Fig 4C) were recorded upon addition of different concentrations of Fe 2+ . We observed a modification of the CD spectrum indicative of a change of conformation upon addition of 0.2 molar equivalent of iron (i.e. 0.2 molecule of iron per molecule of IglG) (Fig 4D). The CD spectra recorded at increasing iron concentrations intersected at a single point (isodichroic point) indicating that addition of increasing concentrations of iron led to a change in the proportion of two unique conformations. Importantly, no further modifications of the CD spectra were observed after addition of one equivalent of iron strongly suggesting that the modifications observed resulted from the binding of one single iron ion per protein. Similar observations were obtained with Zn 2+ (Fig 4D), Mg 2+ and Co 2+ (S8 Fig) suggesting, as observed in the ICP-MS experiments presented above, a promiscuous ability of IglG to bind divalent cations. No change of conformation was observed upon addition of up to 50 molar equivalent of TCEP (Tris (2-carboxyethyl) phosphine), a reducing agent suitable for CD application (Fig 4E), which suggests that the cysteine residues are not involved in forming disulfide bonds within the purified IglG protein.
Altogether, these experiments demonstrate that IglG binds divalent cations. While this remains to be confirmed with other PAAR-like proteins, it suggests that PAAR and PAAR-like proteins have a conserved ability to bind a metal ion to stabilize this specific protein structure. The nature of the bound metal ion remains to be determined in vivo.
To next investigate the role of the cysteines for PAAR-like protein function in the T6SS, we mutated each of these residues in IglG. We first assessed whether the conserved cysteine residues were required for IglG and IglC secretion using the KCl secretion assay. While the wildtype IglG was easily detected in the supernatant in the presence of KCl, we did not detect secretion of any of the four cysteine mutant proteins when expressed in U112 (Fig 5A).
Similarly, when each of the mutant proteins was expressed in the ΔiglG mutant, the secretion of the Hcp homologue IglC was strongly decreased and at the same level as in the noncomplemented ΔiglG mutant (S6 Fig). Moreover, mutating each individual cysteine completely abolished the ability of the ΔiglG mutant to replicate in macrophages (Fig 5B), to escape from the phagosome (Fig 5C), and to trigger cytosolic innate immune responses (Fig 5D and 5E). To further demonstrate the key roles of the conserved cysteines, we selected the first cysteine mutant (Cys 64) to assess its virulence in mice. Whereas intradermal injection of U112 or of  Francisella IglG Is a T6SS PAAR-Like Protein of the experiment (Fig 5F). Altogether, these findings highlight the importance of the four conserved cysteine residues to maintain the T6SS-dependent secretion of IglG and of the Hcp homologue IglC as well as the functional role of IglG in F. novicida virulence. Our data suggest that the role of the conserved cysteines in the PAAR-like proteins may be linked to their metalbinding ability, which is likely to stabilize the PAAR-like fold.
IglG is a distinct member of the PAAR-like family by featuring a unique N-terminal alpha helix extension associated with F. novicida virulence Mirroring PAAR proteins [16], a subset of PAAR-like proteins harbors N-or C-terminal extensions including rearrangement hotspot (RHS) repeats [39], toxin and hydrolase domains ( Fig 6); suggesting they could act as T6SS effectors.
Interestingly, the primary sequence alignment (Fig 1A) and the three-dimensional structure modeling ( Fig 1B) both clearly identified an extension of the PAAR-like domain at the N-terminus of IglG. This extension, encompassing residues 1 to 57, is not present in the other PAAR-like protein (FTN_0054) encoded in the FNI (Fig 1A and S1 Fig). The specificity of the N-terminal extension is strengthened by the lack of homology to other PAAR-like proteins ( Fig 1A) or to any polypeptides outside of the genus Francisella (except for an IglG-like protein in Piscirickettsia salmonis, a species closely related to Francisella [41], S9 Fig). The N-terminal extension of IglG is predicted to contain predominantly α-helices (Fig 1A and 1B). The tertiary structure of this domain was modeled with poor reliability, resulting in one or two helices clearly protruding outside the core PAAR domain albeit with various orientations in the different models (S10 Fig). To experimentally assess the role of this extension in T6SS function and virulence, we engineered a series of N-terminal deletion mutants deleting one or both of the two predicted α-helices (IglG Δ2-17 and IglG Δ2-39 , respectively), the whole N-terminal extension (IglG Δ2-58 ) or the whole extension together with the first 8 residues of the PAAR-like domain (IglG Δ2-66 ). We first analyzed secretion of the IglG proteins in vitro in the presence of 5% KCl, using the strain U112 to overcome any problem with lack of complementation ( Fig 7A). Interestingly, the two α-helices of the N-terminal extension were not required for in vitro secretion, since IglG Δ2-17 and IglG Δ2-39 were both efficiently secreted. In contrast, longer deletions (Δ2-58 or Δ2-66) abolished IglG secretion. When tested in the macrophage assays, even the smallest deletion (Δ2-17) failed to complement the ΔiglG mutant for intracellular replication ( Fig  7B), escape from the vacuole (Fig 7C), or triggering of the cytosolic innate immune responses (Fig 7D and 7E). Finally, the importance of the N-terminal extension was validated in vivo. Indeed, the ΔiglG mutant expressing IglG Δ2-39 -2HA was unable to kill mice, while, as previously mentioned, the ΔiglG mutant expressing IglG-2HA killed 100% of the infected mice within 4 days ( Fig 7F). Importantly, these results dissociate for the first time secretion of a FPI protein from its requirement for F. novicida virulence. Furthermore, these results highlight two distinct domains of IglG critical for protein function; a unique α-helical N-terminal extension dispensable for secretion but required for virulence and a conserved PAAR-like C-terminal domain, which is a key component of the T6SS.
The four conserved cysteine residues and the N-terminal α-helices are required for IglG function in the F. tularensis LVS and F. tularensis SCHU S4 strains The IglG proteins are highly conserved between Francisella species with greater than 98% identity (S11 Fig). Therefore, it was puzzling that the phenotypes of F. novicida and LVS ΔiglG mutants (the latter being deleted of the two iglG copies [7]) differed so strongly with regard to phagosomal escape, replication and cytopathogenicity. Indeed, the initial defect in phagosomal escape observed for the LVS ΔiglG mutant is only transient, eventually leading to efficient intracellular replication but a delayed cytopathogenic response [9,25]. We thus decided to check whether the critical features of the F. novicida IglG protein were also required for IglG function in LVS and in the highly virulent F. tularensis SCHU S4 strain. We first analyzed whether F. novicida-derived IglG could restore the defects of LVS ΔiglG with regard to cytopathogenicity and inhibition of LPS-induced TNF-α secretion [9]. While the wild-type F. novicida protein was equally competent as LVS-derived IglG at restoring the defects, F. novicida derived mutant proteins IglG Δ2-39 , IglG C64G , IglG C105G , IglG C122G and IglG C152G were all unable to restore the defects of LVS ΔiglG. We also confirmed these findings by constructing an IglG Δ2-39 mutation within the LVS-derived IglG protein, and by individually exchanging the cysteines of LVS-derived IglG to either serine or alanine. Again, only the wild-type IglG construct was able to restore the defects of the LVS ΔiglG mutant back to LVS levels, while the A subset of PAAR-like proteins is extended by T6SS effector domains. The 10 most prevalent domain architectures of DUF4280-containing proteins are shown. Data were retrieved from CDART using the IglG protein sequence as input. The name of the DUF4280-associated domain (and when required a short description) is given below together with the superfamily accession number. Lysozyme-like superfamily (cl00222); NLpC_P60 (cl21534), a family of cell-wall peptidases [40]; Esterase-lipase superfamily (cl21494); Hint superfamily (cl22434): Hedgehog/Intein domain; Colicin-DNase superfamily (cl15861); PLN02733 superfamily (cl22774): phosphatidylcholine-sterol O-acyltransferase; LysM superfamily (cl00107). RHS repeats (cl11982 & cl14012) [39].
Francisella IglG Is a T6SS PAAR-Like Protein phenotype of ΔiglG expressing either of the mutant variants was very similar to that of the non-complemented ΔiglG mutant (Fig 8A and 8B). This was not a consequence of a loss of expression, since all of the mutant forms were efficiently expressed (S12 Fig). We were unable to induce type VI secretion in vitro for LVS or SCHU S4 using high KCl concentrations since the two strains were severely impaired for growth in such a medium. As previously described Francisella IglG Is a T6SS PAAR-Like Protein by a microscopy-based assay [24], in cellulo secretion of IglC-TEM was clearly detected by flow cytometry when expressed in a LVS ΔiglG mutant (S13 Fig). Similarly, VgrG-TEM was secreted by the LVSΔiglG mutant at a significant level. Still, the number of cells displaying detectable TEM activity was strongly reduced upon infection with the LVSΔiglG mutant compared to WT LVS (S12 Fig) suggesting that IglG is required for optimal Type VI secretion in macrophages infected with LVS.
The role of IglG in highly virulent F. tularensis had until now not been investigated. We therefore generated a SCHU S4 ΔiglG mutant by deletion of the two iglG gene copies and assessed its ability to grow within macrophages as well as to cause cytotoxicity. In J774 cells, the phenotype was very similar to that of LVS ΔiglG in that the mutant exhibited no detectable growth defect (Fig 9A) [25], and it showed an intermediate cytotoxic effect, although with a slightly faster kinetics compared to that of the LVS ΔiglG mutant (compare Figs 8A and 9B). Thus, the SCHU S4 ΔiglG mutant exhibited an intermediate cytotoxic effect at 24 h, but not at 48 h, which could be restored to wild-type levels by expressing LVS-derived IglG (Fig 9B). The extensive cell death observed in SCHU S4-infected cells was likely responsible for the drop in bacterial counts observed between 24h and 48h. Accordingly, the delayed cytotoxicity kinetics observed upon infection with the ΔiglG mutant correlated with a lower decrease in bacterial counts between 24h and 48 h vs. SCHU S4 (Fig 9A and 9B). As expected, the ΔiglC mutant was severely defective for both growth and cytotoxicity (Fig 9A and 9B). We also confirmed that F. novicida-derived IglG was equally efficient as that of LVS in restoring the defective cytopathogenic response of the SCHU S4 ΔiglG mutant in J774 cells, while the F. novicida mutant variants that either lacked the two α-helices of the N-terminal extension (IglG Δ2-39 ) or carried a C64G substitution were unable to complement the mutant. We then analyzed the virulence of the SCHU S4 ΔiglG mutant in a mouse model of tularemia. The mutant was avirulent in mice, since it did not cause any mortality even when 10 7 cfu of the mutant were inoculated intradermally ( Fig 9C). As expected, both the WT strain and the complemented ΔiglG mutant killed 100% of the mice at much lower inoculum (10 2 and 10 3 , respectively). Taken together, this comparison validates the role of the two identified IglG domains for F. tularensis T6SS function and virulence.
IglG interacts with IglF in an N-terminal α-helix-dependent manner N-and C-terminal extensions of PAAR proteins possess either enzymatic activities or are cargo domains enabling effector secretion [16,42]. To test whether the IglG N-terminal extension could act as such a cargo domain, we performed a bacterial two-hybrid (B2H) interaction screen for IglG against all other FPI proteins. We identified a specific interaction between IglG and IglF, the latter encoded by the gene immediately upstream of iglG (Fig 10A upper panel  and S1 Fig). This interaction was validated by co-immunoprecipitation in F. novicida (Fig 10B  and 10C). To identify the IglF-interacting domain within IglG, we generated specific mutations within the latter protein and tested them in the B2H system. The amount of β-galactosidase activity, indicative of a stronger interaction, increased gradually when smaller truncations were introduced in the IglG protein (Fig 10A, upper panel). Strikingly, the interaction was completely abolished when the first 39 residues of IglG were removed (IglG Δ2-39 ) (Fig 10A,  upper panel). This result was validated by co-immunoprecipitation (Fig 10B). In contrast, deletion of the last 39 of the protein (IglG Δ134-173 ) had only a minor effect on the interaction ( Fig  10A, upper panel). In agreement with these results, we did not observe any interaction between IglF and the other F. novicida PAAR-like protein FTN_0054, which lacks an N-terminal extension. While the N-terminal extension was clearly essential for the interaction, it was not sufficient in this system, since the N-terminal extension alone (IglG Δ58-173 ) did not interact with  (Fig 10A, upper panel). This lack of interaction may be partly due to improper folding of the truncated protein. In support of this hypothesis, a protein consisting of the first 60 residues of IglG fused to the PAAR-like protein FTN_0054 (starting at the shared alanine at position 10, see Fig 1A) established an interaction with IglF, albeit partially reduced (Fig 10A, upper  panel). To identify key residues that may participate in the binding, a helical wheel analysis of the IglG N-terminus was performed using the first 56 residues in the analysis. Charged residues, both positively and negatively, at the different sides of the predicted helix were targeted for alanine scanning resulting in single substitutions K10A, R11A, D19A, E20A, D28A and D32A. Strikingly, when tested in the B2H assay, two of these substitions either abolished (K10A) or severely diminished (D32A) binding to IglF (Fig 10A, lower panel). Interestingly, these residues were predicted to lie close to each other in the same side of the helix. Importantly, individual substitutions of each of the single cysteines to either serine or alanine had no effect on the interaction (Fig 10A) further suggesting that the PAAR-like domain per se and the metal-binding ability are not essential for the interaction with IglF. Altogether our results indicate that IglG harbors a PAAR-like domain and an N-terminal extension, which interacts with IglF. The N-terminal extension is required for virulence, possibly leading to the translocation of this putative effector protein into the host cell.

Discussion
Despite the well-recognized role of the FPI in Francisella virulence in vitro and in vivo, the function of most of the FPI-encoded proteins is still elusive. The homology between FPI proteins and components of a T6SS [44,45] was described almost 10 years ago [8]. Yet, in spite of tremendous progress in the understanding of T6SS assembly and function in numerous bacterial species [11,14], the translation of this knowledge to the Francisella FPI has been challenging. This difficulty lies in part in the limited similarity of the Francisella FPI with other T6SS [46].
In this work, we focused on IglG, a protein required for the virulence in mice of F. novicida, F. tularensis LVS and of the highly virulent F. tularensis SCHU S4 strain ( [7], this work). We identified IglG as a protein containing a DUF4280 domain. This work and another recent study [28] identified IglG and two other DUF4280-containing proteins (FTN_0054 and Fjoh_3275/FteI) as three proteins likely to adopt a tertiary structure homologous to that of PAAR proteins. In addition, we found that the DUF4280 consensus sequence itself was also predicted to adopt a PAAR-like fold suggesting that the results we obtained on IglG will be valid for most of the DUF4280 proteins present in a large number of bacterial species. We thus propose to assign the term PAAR-like domain to this domain of unknown function, characterized by four highly conserved cysteines. Although we cannot completely rule out that the cysteines might form disulfide bridges, our biochemical analyses and the predicted homology between the zinc-binding PAAR protein and the PAAR-like proteins strongly suggests that the structural fold of PAAR-like proteins is stabilized by the cysteine-mediated metal coordination. Accordingly, we identified IglG as a metal-binding protein. In vitro, IglG had a small preference for iron over zinc in contrast to the typical PAAR protein PA_0824 from P. aeruginosa, which in our experimental conditions, preferentially binds zinc (S7 Fig). Many metalloproteins a 2-sided t-test with equal variance (*, P 0.05; ***, P 0.001). (B) J774 cells left uninfected or infected with F. tularensis at an MOI of 500 for 2 h, were washed and subsequently incubated in the presence of E. coli-derived LPS (50 ng/ml) for an additional 2 h. The average TNF-α secretion (%) compared to LVS, which was set as 100%, and the SD of quadruple samples (n = 4) from one representative experiment out of two is shown. The asterisks indicate that the cytokine levels were significantly different than those of LVS-infected cells as determined by a 2-sided t-test with equal variance (**, P 0.001; ***, P 0.001).
doi:10.1371/journal.ppat.1005821.g008 have the potential to bind different metal ions. The binding of a specific divalent cation to a protein is controlled by the natural order of stability of complexes of bivalent transition metals (also known as the Irving-Williams series: Mg 2+ and Ca 2+ (weakest binding)< Mn 2+ < Fe 2+ < Co 2+ < Ni 2+ < Cu 2+ > Zn 2+ ), the relative concentrations of the different cations as well as specific mechanisms, such as the presence of metallochaperones and chelators [47,48]. Indeed, we have observed that purified IglG can bind different cations in vitro (Fig 4C). The nature of the metal contained within IglG protein when expressed at physiological level in Francisella within its mammalian host may be influenced by the localization of the protein and the relative concentrations of the different divalent cations in each compartment as well as by specific mechanisms such as the presence of metallochaperones. Furthermore, innate immune mechanisms such as Zn 2+ chelation by calprotectin, one of the most abundant protein of the neutrophil cytosol [49] may influence the bioavailability of divalent cations and the nature of the IglGbound metal. Mutation of single cysteine residues only moderately decreased metal binding, suggesting that three functional side chains are sufficient to bind metal, a feature that has been observed in other proteins [50][51][52]. However, in vivo, each individual cysteine is required to sustain the role of IglG in F. novicida virulence suggesting that tetravalent coordination of the metal is required for IglG stability and function. Interestingly, the tip of the membrane-attacking complex of bacteriophages P2 and φ92, which is homologous to the VgrG 3 -PAAR complex of the T6SS binds iron [53,54]. These results highlight the evolutionary conservation of spike proteins and identify PAAR-like proteins as potential evolutionary intermediates between phage spike proteins and bacterial T6SS PAAR proteins.
The predicted PAAR-like fold suggests that PAAR-like proteins are located at the tip of the T6SS. Despite its very small size, Francisella VgrG displays a β-structural repeat consistent with an ability to bind PAAR-like proteins [16] although we failed to experimentally demonstrate an interaction between IglG and VgrG by both bacterial two-hybrid and pull-down experiments. Strikingly, while the deletion of the PAAR-like gene iglG confers to F. novicida a phenotype identical to the one observed after the deletion of the hcp homologue iglC, this is not true in F. tularensis LVS and F. tularensis SCHU S4 strain. These results suggest that in F. tularensis, the activity of the T6SS is less stringently dependent on the capping of the T6SS by a PAARlike protein. Accordingly, using TEM fusion proteins and a β-lactamase/CCF4 flow cytometry assay, we observed a significant (although strongly reduced compared to WT LVS) secretion of both IglC and VgrG in the LVS ΔiglG mutant. IglG may thus play an auxiliary role in F. tularensis T6SS function. Yet, it is clearly required for F. tularensis virulence in vivo ( [7], this work) highlighting the importance of a fully functional T6SS to trigger disease. IglG is required for SCHU S4 virulence in vivo but is dispensable for in vitro replication and cytopathogenicity. Intracellular growth and cytopathogenicity in J744 cells (A and B, respectively) of F. tularensis subsp. tularensis strains. (A) J774 cells were infected at an MOI of 200 for 2 h. Upon gentamicin treatment, cells were allowed to recover for 30 min after which they were lysed immediately (corresponds to 0 h) or after 24 or 48 h with PBS-buffered 0.1% sodium deoxycholate solution and plated to determine the number of viable bacteria (log 10 ). Infections were repeated twice and a representative experiment is shown. Each bar represents the mean values and the error bar indicates the SD from triplicate data sets. The asterisks indicate that the log 10 number of CFU was significantly different from the parental SCHU S4 strain as determined by a 2-sided t-test with equal variance (*, P 0.05; **, P 0.01; ***, P 0.001). (B) Culture supernatants of infected J774 cells were assayed for LDH activity at 0, 24  Francisella IglG Is a T6SS PAAR-Like Protein As described above, another PAAR-like protein (FTN_0054) is found in the Francisella novicida island (FNI). We could not find evidence of this genomic island playing any major role in the virulence of F. novicida in vitro nor in vivo in a mouse model of tularemia (S14 Fig). It remains to be demonstrated whether this island encodes a functional T6SS involved in bacterial competition and/or in the targeting of specific eukaryotic cells. Interestingly, one FNI protein (FTN_0052) exhibits similarity to proteins from the phosphoesterase family (pfam04185), which includes both bacterial phospholipase C enzymes and eukaryotic phosphatases making this protein a likely T6SS effector [55]. FTN_0052 has no homologue in the FPI, suggesting that the two genomic islands may encode different arrays of effectors. On the other hand, unique features of the FPI include the N-terminal extension of the PAAR-like protein IglG (compared to FTN_0054) as well as IglF. The comparison of the FPI and the FNI might thus help us to discriminate the proteins involved in the T6SS machinery sensus stricto from specific the ω subunit of E. coli RNAP respectively, induces transcription from the lacZ promoter of the E. coli reporter strain KDZif1ΔZ, resulting in β-galactosidase activity. The ability of mutant forms of IglG to interact with IglF was analyzed and the levels of interaction expressed as percentage of the wild-type interaction, which was set to 100%. The negative controls correspond to vectors missing inserts, while the positive control is IglA-IglB in the upper panel and IglF-IglG in the lower panel [43]. Shown is the percentage of mean β-galactosidase activity ± SEM from four independent experiments, where two to three independent transformants were tested on each occasion. Unpaired t-tests were performed to compare the β-galactosidase activity obtained with the indicated mutant IglG protein to that of wild-type IglG (*, P 0.05; ***, P 0.001). (B, C) Lysates from F. novicida expressing 2HA-tagged VgrG, IglG, IglG Δ2-39 (B) or IglF (C) proteins were incubated with anti-HA agarose beads. The loaded beads were then incubated with lysates from F. novicida expressing or not TEM-tagged IglF (B) IglG or VgrG (C) and used for co-immunoprecipitation. Proteins in bacterial lysates and immunoprecipitated proteins were revealed by immunoblotting using specific antibodies against the HA-or TEM tags. Francisella IglG Is a T6SS PAAR-Like Protein proteins associated with virulence towards the mammalian host. Interestingly, the iglF gene lies in between the vgrG and the iglG genes. This specific genetic linkage supports our experimental data demonstrating the key role of the IglG N-terminal extension in establishing an interaction with IglF.
PAAR-like domains are frequently extended in an N-or C-terminal manner by domains with known T6SS effector functions (Fig 6), strengthening the functional homology between PAAR [16] and PAAR-like proteins. The N-terminal extension that we identified in IglG is predicted to protrude outside of the PAAR-like core domain (Fig 1B and S10 Fig) suggesting it might act as an adaptor domain to connect the tip of the T6SS with other T6SS proteins/effectors. The adaptor nature of the N-terminal extension of IglG is supported by the fact that IglG interacts with IglF in a manner that requires the first 39 residues of IglG. The same region within IglG is dispensable for in vitro secretion of IglG, while the PAAR-like domain is required for IglG secretion. In line with identified PAAR-protein functions [16], these data suggest that IglG could act as a cargo molecule aiding in the T6SS-mediated delivery of IglF. While deciphering the role of IglF in the T6S machinery or in infected cells will be the goal of future studies, IglF is a secreted protein [24] required for in vitro replication [21], suggesting that IglF may be a Francisella effector secreted into the host cell in an IglG-dependent manner. Altogether, this work in addition to increasing our understanding of the structure of F. tularensis T6SS (see model in Fig 11) identifies a novel family of PAAR-like proteins with conserved cysteines and metal binding ability, involved in connecting the Type VI secretion machinery to effector activities.

Ethics statement
All experiments involving animals were reviewed and approved by the animal ethics committees of the University of Lyon, France under the protocol numbers #ENS_2014_017 and #ENS_2012_061, by the IRB of the National Research Council, Ottawa, Canada and by the Local Ethical Committee on Laboratory Animals, Umeå, Sweden (no. A67-14).

Mouse infections
For testing of SCHU S4 and derivatives, BALB/c mice were purchased from Charles River Laboratories (St. Constant, Quebec, Canada). The mice were maintained and used in accordance with the recommendations of the Canadian Council on Animal Care Guide to the Care and Use of Experimental Animals in a federally licensed, Select Agent-approved, small animal containment level 3 facility, National Research Council, Ottawa, Canada. F. tularensis strains were injected in a volume of 50 μl intradermally in groups of five (n = 5). The mice were examined daily for signs of infection and were euthanized by CO 2 asphyxiation as soon as they displayed signs of irreversible morbidity. For testing of U112 and derivatives, C57BL/6J mice were injected in a volume of 100 μl intradermally in groups of five (n = 5). Aliquots of the diluted cultures were also plated on GC-agar to determine the number of CFU injected. Actual doses were the following: 830 (U112), 915 (ΔiglG), 1,635 (ΔiglG/IglG-2HA), 1,065 (ΔiglG/IglG Δ2-39 -2HA) and 600 (ΔiglG/IglG C64G -2HA). Previous studies have demonstrated that the LD 50 of the U112 strain is approximately 500 CFU [56]. Mice were examined twice daily for signs of severe infection and euthanized by CO 2 asphyxiation as soon as they displayed signs of irreversible morbidity. In our experience, such mice were at most 24 h from death, and time to death of these animals was estimated on this premise.
F. novicida chromosomal deletion mutants of iglG, vgrG or FTN_0037-FTN_0054 (FNI) locus were constructed as previously described [57] by allelic exchange using PCR products followed by Flp-mediated excision of the antibiotic-resistance marker [58] or by using the suicide vector pJEB753 [59]. Primers are presented in S2 Table. To construct the ΔiglG deletion mutant in SCHU S4, the suicide vector pJEB866 was used. pJEB866 was constructed by lifting the deletion fragment from pJEB753 into pDMK2 [60] using XhoI/SacI digestion. Conjugal mating experiments using S17-1λ pir as the donor strain and sucrose-selection allowed for the allelic exchange of the suicide plasmids within regions of complementary sequence on the chromosome of U112 or SCHU S4 [3]. To remove both copies of the iglG gene in the latter strain, the procedure was repeated. PCR screening and genomic sequencing were used to verify that the anticipated genetic event had occurred.

Construction of expression vectors
Plasmids used in this study are listed in S1 Table. Primer combinations and restriction sites used to generate the plasmids are listed in S2 Table. All amplified fragments were first cloned into pCR4-TOPO TA cloning vector to facilitate sequencing. PCR or overlap PCR was used to introduce substitution or deletion mutations within iglG. A hybrid gene consisting of IglG and FTN_0054 was constructed using LVS and U112, respectively as template in the first PCR reaction step. Upon overlap PCR, the resulting hybrid consisted of the first 60 residues of IglG, and where the remaining part was derived from FTN_0054, starting at alanine at position 10. 2xHA-tagged IglG was generated by cloning iglG in frame with a single HA tag sequence in the popHA plasmid (kindly provided by P. Mangeot) followed by addition of a second HA tag sequence by PCR amplification. Plasmids used for complementation of ΔiglG in trans were constructed by introducing C-terminally 6xHis-tagged or 2xHA-tagged wildtype or mutated versions of iglG into the NdeI/EcoRI sites of pKK289Km [61] or into a cyaAcontaining pFNLTP6 derivative [62,63] respectively. The latter plasmid was also used to create TEM fusion after removal of the plasmid encoded β-lactamase gene. E. coli TEM-1 amplified from pUC19 (ThermoFisher Scientific) was then cloned under the Gro promoter using NheI/BamHI sites. FPI and controls genes were cloned in frame with TEM-1 using EcoRI and NheI sites. Plasmids used for B2H analysis were constructed by introducing NdeI/NotI fragments of mutated iglG into pBRGPω [64]. For GST-IglG expression, the F. novicida iglG gene was PCR-amplified and cloned into the pGEX-6-P3 (Novagen) in frame with the GSTencoding gene. Plasmids were transferred into bacteria by chemical transformation or electroporation.  [13]. The Francisella T6SS is presented in the contraction phase. While the interaction of IglF and IglG was demonstrated in this study, the function and the localization of IglF within the T6SS remain to be determined. While the current knowledge on VgrG and PAAR proteins suggests that the two Francisella homologues may be located at the tip of the T6SS, this remains to be demonstrated. Bacterial 2-hybrid analysis E. coli strain KDZif1ΔZ was used as the reporter strain for the bacterial-2-hybrid experiments. It harbors an F9 episome containing the lac promoter-derivative placZif1-61 driving expression of a linked lacZ reporter gene [64]. Cells were grown with aeration at 37°C in LB supplemented with 1 mM IPTG (Isopropyl β-D-1-thiogalactopyranoside). Cells were permeabilized with SDS-CHCl3 and assayed for β-galactosidase (β-gal) activity as described previously [65].
Infections and replication assay BMDM were infected as described before [68] at the indicated multiplicity of infection (MOI). For F. novicida intracellular replication assay, macrophages were lysed with 1% (w/v) saponin (Sigma) in water for 5 min. Dilution, plating on TSA supplemented with 0.1% (w/v) cysteine and counting was performed using the easySpirale Dilute (Interscience). For LVS and SCHU S4, cells were infected for 2 h at an MOI of 200, washed three times, and incubated in the presence of gentamicin (5 μg/ml; LVS and 2 μg/ml; SCHU S4) for 30 min (corresponds to time zero). At 0, 24 and 48 h, the macrophage monolayers were lysed in PBS with 0.1% deoxycholate, serially diluted in PBS and plated on modified GC-agar base plates for determination of viable counts.

Interferon and TNF-α secretion
Type I interferon secretion was determined by an ISRE-luciferase bioassay [67]. L929 ISREluciferase cells were plated the day before at 10 5 cells per well in a 96 wells-plate. Supernatants from infected BMDMs were added for 4 h onto the ISRE-luciferase cells. Luciferase luminescence was detected using Bright Glo Assay (Promega) following the manufacturer's instructions. TNF-α secretion of J774 cells upon 2 h of LPS stimulation was performed according to our previously established protocols [7], using the BD OptEIA Mouse TNF-α ELISA Set (BD Biosciences) according to the manufacturer's instructions.

Cell death assays
Cell death was monitored in BMDMs by monitoring in real time incorporation of propidium iodide (used at 5 μg/ml) through measurement of fluorescence emission at 635 nm every 15 min on a microplate reader (Tecan). Quantification of cytopathogenicity in J774 cells was performed by analysis of LDH release in the cell supernatant, using the CytoTox96 LDH kit (Promega, France), following manufacturer's instructions.

Phagosomal rupture assay
Quantification of vacuolar escape using the β-lactamase/CCF4 assay (Life technologies) was performed as previously described [33]. Briefly, bone marrow-derived macrophages seeded onto non-treated plates were infected as described above for 2 h, washed and incubated in CCF4 for 1 h at room temperature in the presence of 2.5 mM probenicid (Sigma). Propidium iodide negative cells were considered for the quantification of cells containing cytosolic F. novicida using excitation at 405 nm and detection at 450 nm (cleaved CCF4) or 510 nm (intact CCF4).

Bioinformatic analysis
Identification of conserved domains, of the DUF4280 consensus sequence and of the species presenting DUF4280 proteins were performed using CD database, the Conserved Domain Architecture Retrieval Tool (CDART) [69] and the Interpro database. Three dimensional modeling was performed using the default parameters in the I-Tasser and Phyre-2 webservers [30]. Five models were obtained for each template (DUF4280 and IglG) with very high scores. Figures of structure were generated with Pymol (Delano Scientific, https://www.pymol.org). Conserved residues were identified using the iterative search algorithm Jackhmmer [70]. The N-terminal α-helix was analysed using EMBOSS::pepwheel software (http://www.tcdb.org/ progs/?tool=pepwheel). For the molecular dynamics protocol, IglG was modeled with the AMBER99SB force field [71]. The model was ionized and solvated with TIP3P water molecules, setting unit cell dimensions to 61 Å × 57 Å × 73 Å (residues 53-173). The resulting~24000 atoms system was minimized and equilibrated locally with ACEMD [72] for 2 ns under NPT conditions, of 1 atm at 300 K, nonbonded cutoff of 9 Å, rigid bonds, and PME electrostatics. A time step of 4 fs was used, in conjunction with a hydrogen mass-repartitioning scheme. During minimization and the first 2 ns of equilibration, the protein's heavy atoms were restrained by a harmonic potential with k = 1 kcal mol −1 Å −2 . The complete molecular dynamics resulted in a 400ns trajectory where the rmsf was computed by means of the Gromacs subroutines [73].

Recombinant IglG production and purification
For protein expression, GST, GST-IglG, His 6 -IglG encoding vectors were introduced into the E. coli Rosetta strain, which was cultured at 37°C in LB medium supplemented with 100 μg/ml ampicillin to an optical density at 600 nm of 0.8-1.0. Protein expression was induced by adding 1 mM IPTG at 20°C for 16 h. Cells were harvested by centrifugation and re-suspended in lysis buffer (20 mM Tris pH 8.0, 0.5 M NaCl, 10% Glycerol (V/V), 1% triton (V/V) supplemented with 1 mg/ml of lysozyme (Sigma), DNAse (Sigma) and Ethylenediamine tetra-acetic acid (EDTA)-Free protease inhibitors cocktail (Roche)). Cells were disrupted by sonication and the resulting lysate was cleared by centrifugation (20 min at 16,000 x g at 4°C). For GST and GST-IglG, the supernatant was loaded onto a GST-HiTrap column (GE Healthcare), equilibrated with 15 ml of buffer (20 mM Tris pH 8.0, 300 mM NaCl, 5% glycerol). The column was washed with 20 ml of buffer containing 20 mM Tris pH 8.0, 300 mM NaCl, 5% glycerol (V/V). The proteins were eluted with 10mM glutathione and concentrated using Amicon 0.5 mL 10K (Millipore) centrifugal devices. The proteins were dialysed against 20 mM Tris pH 8.0, 300 mM NaCl, 5% Glycerol (V/V) at 4°C. To reach more than 95% purity (as assessed by sodium dodecyl sulphate-polyacrylamide-gel electrophoresis), the proteins were further purified by gel filtration using the Superdex 75 column or by the anion exchange monoQ hiTrap Q HP column (GE Healthcare). His 6 -IglG protein was extracted from the insoluble fraction. Following lysate centrifugation, the pellet was re-suspended in 8M Urea buffer, 20mM Tris pH8 under strong agitation during 20 minutes at room temperature. The resulting solubilized proteins were cleared by centrifugation (15 min at 10,000 x g at 4°C). Supernatant was collected and loaded onto a 5 ml His-trap column (GE Healthcare). The column was washed once with 10mL buffer containing 8M Urea buffer, 20mM Tris pH8, 1M NaCl, once with buffer containing 25 mM imidazole. The protein was eluted with a linear gradient of imidazole (25-500 mM) in Urea 8M, 20mM Tris pH8. His 6 -IglG was refolded by dialysis overnight in buffer containing 20mM Tris pH8, 250mM arginine, 100mM NaCl, 5% glycerol, 1mM EDTA. Arginine was eliminated by three sequential 2 hours dialysis at 4°C, the first one in buffer containing 20mM arginine, 20mM Tris pH8, 300mM NaCl, 5% glycerol, 1mM EDTA, the second and the third ones in 20mM Tris pH8, 300mM NaCl, 5% glycerol, 1mM EDTA. Protein concentrations were determined by measuring the UV light absorbance at 280nm.

Metal concentration determination by inductively coupled plasma mass spectrometry (ICP-MS)
100 μl of each sample was diluted to 10 ml in 2% HNO 3 with Indium as an internal standard. Zn, Fe, Ni, Cu, Pb and In were measured by ICP-MS (iCAP Q ThermoFisher Scientific) with running conditions presented in S3 Table. Circular dichroism (CD) spectroscopy and IglG-metal binding Far-UV CD spectra (190nm-260nm) of His 6 -IglG were recorded on a Chirascan Circular Dichroism Spectrometer (Applied photophysics) calibrated with (1S)-(+)-10-camphorsulfonic acid. Measurements were carried out at room temperature in a 0.1cm path length quartz cuvette. Parameters were set as followed: wavelength range 180-260nm, 0.2nm increment, bandwith 0.5nm; scan speed, 50 nm.min -1 ; response time 1s. Spectra were corrected by subtracting buffer contributions and protein dilution factors before calculating the mean residue molar ellipticity. Spectra deconvolution was performed using Dichroweb server [74] and the Selcon3 algorithm [75] on His 6 -IglG refolded in the presence of FeCl 2 at 1mM and dialyzed four times in a buffer consisting of NaH 2 PO4 20mM, NaF 300mM, pH 7.5. To follow metal binding by CD spectrometry, His 6 -IglG in Tris 20mM pH8, NaCl 200mM, glycerol 5% was diluted to 10 μM in NaH 2 PO4 20mM pH 7.5, NaF 300mM, glycerol 5% buffer. Spectra were measured immediately after dilution or following addition of increasing amount of FeSO 4 , ZnSO 4 , MgSO 4 , CoSO 4 from 0.2 to 2 molar equivalent of protein. TCEP was incubated with His 6 -IglG at 0.5, 5 and 50 molar equivalent of protein for up to 4 hours on ice before spectra acquisition. Spectra were smoothed using Prism software.

Immunoblotting analysis
Protein lysates for immunoblotting were prepared by using Laemmli sample buffer. Protein lysates corresponding to equal OD 600 were loaded on 4-12% Bis/Tris gels (Invitrogen), and run in TGS buffer. Protein transfer was performed with iBlot gel transfer stacks (Invitrogen). Membranes were probed with the following monoclonal antibodies: mouse anti-PdpB, mouse anti-IglC, and mouse anti-IglB (all provided from BEI Resources, Manassas, VA, USA) or via commercially available anti-HA (clone HA-7, Sigma), anti-Penta-His (Qiagen, MD, USA). A secondary horseradish peroxidase (HRP)-conjugated goat anti-mouse antibody (Santa Cruz Biotechnology, CA, USA) and the Enhanced Chemiluminescence system (ECL) (Amersham Biosciences, Uppsala, Sweden) were used.

Secretion assays
Protein secretion during infection was estimated using TEM-β-lactamase fusion proteins and CCF4 substrate as previously described [24] by flow cytometry analysis using a Canto II analyser (BD). The in vitro KCl assay was performed as previously described [23]. Briefly, F. novicida was grown overnight in TSB containing 0.1% cysteine. The overnight culture was diluted to OD 600 = 0.3 in medium with or without 5% KCl. Upon reaching an OD 600 = 1.5, the culture was centrifuged at 4,700 x g for 20 min at 4°C. The supernatant was filtered-sterilized using a 0.22 μm filter and proteins were extracted by methanol-chloroform precipitation. Protein extracts corresponding to 0.045 and 1 OD 600 of the pellet and the concentrated supernatant fraction, respectively were loaded onto the SDS-PAGE gel and immunoblotted with the indicated antibodies.

Statistical analysis
Statistical data analysis was performed using unpaired t-tests and Prism 5.0a software (Graph-Pad Software, Inc.). Two-tailed P-values are shown. For survival experiments, Log-rank (Mantel-Cox) test were performed. The following convention is used: Ã , P 0.05; ÃÃ , P 0.01; ÃÃÃ , P 0.001.

Accession numbers
For each strain, when applicable, the discontinued NCBI gene record and the current NCBI locus tag are indicated. For F. tularensis strains, the two FPI loci are indicated. NA: Not applicable.
VgrG  . At 48h post-injection, spleens were collected and the bacterial counts were determined on TSA cysteine plates with and without kanamycin. CI was calculated as followed: CI = (Strain A/StrainB) in output/ (Strain A/Strain B) in input. To avoid any interference of the inherent kanamycin resistance with the assay, the CI displayed combines two independent CI: ΔFNI::Kan vs U112 and ΔFNI vs Δbla::Kan. One sample t-tests were performed to analyze whether the experimental means were statistically different from 1 (ns: not significant, ÃÃÃ : P 0.001). (PDF) S1