The Salmonella effector protein SifA plays a dual role in virulence

The virulence of Salmonella relies on the expression of effector proteins that the bacterium injects inside infected cells. Salmonella enters eukaryotic cells and resides in a vacuolar compartment on which a number of effector proteins such as SifA are found. SifA plays an essential role in Salmonella virulence. It is made of two distinct domains. The N-terminal domain of SifA interacts with the host protein SKIP. This interaction regulates vacuolar membrane dynamics. The C-terminal has a fold similar to other bacterial effector domains having a guanine nucleotide exchange factor activity. Although SifA interacts with RhoA, it does not stimulate the dissociation of GDP and the activation of this GTPase. Hence it remains unknown whether the C-terminal domain contributes to the function of SifA in virulence. We used a model of SKIP knockout mice to show that this protein mediates the host susceptibility to salmonellosis and to establish that SifA also contributes to Salmonella virulence independently of its interaction with SKIP. We establish that the C-terminal domain of SifA mediates this SKIP-independent contribution. Moreover, we show that the two domains of SifA are functionally linked and participate to the same signalling cascade that supports Salmonella virulence.

The Gram-negative bacterium Salmonella Typhimurium is an intracellular pathogen whose virulence relies on the capacity to survive and replicate inside cells of the infected host. The intracellular phase requires the expression of the type 3 secretion system-2 (T3SS-2), which is expressed by the intracellular bacterium in response to the vacuolar environment 1 . T3SS-2 mediates the translocation across the vacuolar membrane of a set of bacterial effector proteins that support collectively the intra-vacuolar replication (for review see 2 ).
The T3SS-2 effector protein SifA 3 plays a significant role in Salmonella virulence and several cellular phenotypes are linked to its translocation. SifA is required to maintain the integrity of the Salmonella-containing vacuole (SCV) 4 . It promotes the formation of tubular membranous structures connected to SCVs that are named Salmonella-induced tubules [5][6][7][8] .
Several functional peptide stretches and domains of SifA have been identified. The N-terminal residues direct the T3SS-2-mediated secretion/translocation of this effector 9 . The last C-terminal residues form a CAAX motif, which is found in eukaryotic Rab GTPase but also in many bacterial effectors 10 . Following translocation, the CAAX motif of SifA is isoprenylated and S-acylated by the eukaryotic enzymatic machinery 11 and these modifications enable the membrane anchoring of the effector 12 . The resolution of the crystal structure of SifA has shown that the protein is divided into two distinct major domains 13,14 separated by a potential caspase-3 cleavage site 15 . Thus, the two domains of SifA might act independently of each other upon cleavage. However, the caspase-3-mediated cleavage of SifA has not been demonstrated.

Results
Characterization of SKIP −/− mice. SKIP −/− mice were obtained from the Sanger Institute that performs systematic phenotyping of knock-out mice 22 and makes data available online 23 . SKIP −/− mice are declared as presenting slight increases for leukocyte cell number and for level of circulating alkaline phosphatase in females and males, respectively. We analysed blood samples but did not find significant differences in the hematology profiles of 8-weeks-old females C57BL/6 versus SKIP −/− mice (Supplementary Table S1). As an increased circulating alkaline phosphatase is a possible sign of liver dysfunction, we assayed two transaminases (AST and ALT) that are commonly use to monitor liver damage. Mean AST and ALT levels (18 unsexed mice in each group) were slightly higher in SKIP −/− as compared to C57BL/6 mice (Supplementary Table S1), but these differences are not significant (p = 0.5 and 0.19, respectively).
We used an anti-SKIP antibody 16 to confirm the absence of the protein in SKIP −/− mice. By Western blotting, the antibody recognizes in lysates of HeLa cells and of peritoneal macrophages derived from C57BL/6 mice a protein with an apparent molecular mass of ≈150 kDa that corresponds to SKIP. This protein was not detected in SKIP −/− peritoneal macrophages (Fig. 1A).
The consequences of the lack of SKIP have been previously described in cultured cells using a siRNA-mediated knockdown 16,17 . In Salmonella infected HeLa cells, one observes an accumulation of kinesin-1 and of T3SS-2 effectors on SCVs in the absence of SifA or of SKIP. We checked these phenotypes using mouse embryonic fibroblasts and bone marrow-derived macrophages (BMM) prepared from C57BL/6 and SKIP −/− mice and infected with wild type or ∆sifA strains of S. Typhimurium (Salmonella enterica subsp. enterica, strain NCTC 12023). A microscopic analysis of immunostained SKIP −/− mouse embryonic fibroblasts showed that kinesin-1 accumulates both on wild type and ∆sifA SCVs (Fig. 1B). We found more than 30% of kinesin-1-positive SCVs in SKIP −/− cells infected by one or the other strain (Fig. 1C). By contrast, in C57BL/6-derived cells, we detected 10 ± 2% of wild type SCVs decorated by the anti-kinesin-1 antibody as compared to 32 ± 6% for ∆sifA SCVs (Fig. 1C). Likewise, we observed in both cell types an accumulation of the T3SS-2 effectors SseJ and PipB2 (not shown) on SCVs in the absence of SifA or of SKIP (see Fig. 1D,E for SseJ in BMMs). We concluded that, as far as the accumulations of kinesin-1 and effectors are concerned, the consequences of the absence of SKIP are similar in SKIP −/− -derived cells and in HeLa cells. SKIP −/− mice are less sensitive to Salmonella infection than congenic C57BL/6. To investigate whether SKIP mediates susceptibility to salmonellosis, we inoculated perorally (P.O.) C57BL/6 or SKIP −/− mice with 10 5 CFU of S. Typhimurium and monitored their survival. SKIP −/− mice succumbed to wild type Salmonella infection significantly later than C57BL/6 mice. C57BL/6 and SKIP −/− mice had a median survival time of 7.5 and 9 days, respectively ( Fig. 2A). We obtained very similar results for wild type and ∆sifA Salmonella strains in C57BL/6 mice (Fig. 2B), with a median survival time of 7 and 9 days, respectively. The absence of either SifA or SKIP results in a longer survival of mice exposed to a Salmonella challenge.
We examined how the lack of SKIP could decrease the susceptibility of mice to a Salmonella challenge. At day five post-inoculation, we found lower bacterial counts in organs of SKIP −/− as compared to C57BL/6 mice (Fig. 2C). The ratios of bacterial burdens between SKIP −/− and C57BL/6 mice were 1:23, 1:32, 1:25 and 1:51 in the spleen, the liver, the mesenteric lymph nodes and the small intestine, respectively. Although statistical analysis revealed significance only for bacterial numbers in the small intestine (see P values in Fig. 2C), it suggests that the absence of SKIP limits Salmonella replication. We supported this point using BMMs prepared from the two mouse lineages. Cells were infected with wild type or ∆sifA strains and the fold increase of intracellular bacteria between 2 and 16 h after infection was determined. As expected, in C57BL/6 macrophages we observed a dramatic replication defect for the ∆sifA mutant with respect to the wild type strain (Fig. 2D). In contrast, we detected in SKIP −/− macrophages a lower Salmonella replication and a no significant difference between the two bacterial strains. We concluded that SKIP is an important mediator of the role played by SifA in Salmonella intracellular replication. All together our data indicate that SKIP, by interacting with SifA, mediates susceptibility to salmonellosis. Characterization of a SKIP-independent function of SifA. Next, we examined if the role of SifA in virulence was exclusively mediated by its interaction with SKIP. For this purpose, we compared the virulence of diverse Salmonella strains in C57BL/6 and SKIP −/− mice. Groups of mice were inoculated intraperitoneally (I.P.) or P.O. with different two strains combinations (1:1 mix) (Fig. 3) and bacteria were recovered from mouse spleens after two (I.P.) or five (P.O.) days to determine the competitive index (CI) 24 . We found that a ∆sifA mutant was still significantly attenuated as compared to wild type Salmonella in SKIP −/− mice inoculated I.P. or P.O. (CI of 0.83 ± 0.13 and 0.23 ± 0.11, respectively) ( Fig. 3A,D,I). As a control, we tested the virulence attenuation of a ∆sseG mutant 25 , which did not differ significantly between mice expressing or not SKIP (Fig. 3J). These data indicate that SifA mediates a SKIP-independent function in virulence.
SKIP binds the distal part of the SifA N-term. This interaction, which involves a network of hydrogen bonds and van der Waals contacts is abolished by the substitution of the interacting leucine 130 with an aspartic acid (L130D) 13,14 . This point mutated form of SifA is secreted and translocated by the T3SS-2 14 . To confirm that SifA acts also independently of its interaction with SKIP, we performed a mixed inoculation of wild type Salmonella and a strain expressing chromosomally SifA L130D (Fig. 3B,G). The CIs in C57BL/6 mice inoculated I.P. or P.O. were both of ≈0.5 (Fig. 3K). These results reveal the contribution of SKIP to the virulence mediated by SifA and indicate that this point mutant form of SifA is active despite its lack of interaction with SKIP and its propensity to be more rapidly degraded than the wild type protein in an eukaryotic environment (half-lives of 8 h and 5.3 h for SifA and SifA L130D , respectively, see Supplementary Fig. S1). In SKIP −/− mice we obtained CI values of ≈1 (Fig. 3K), which ascertain that the CI values obtained in C57BL/6 mice are strictly reflecting the contribution of SKIP to the virulence mediated by SifA. We also infected C57BL/6 mice with a mix of sifA L130D and ∆sifA strains (Fig. 3E) and observed a CI of 0.62 ± 0.22 (Fig. 4L) that substantiates a SKIP-independent function in virulence for SifA.
The SKIP-independent function of SifA is associated with its C-term domain. To determine which of the N-or C-term domain of SifA is responsible for the SKIP-independent function in virulence, we constructed a Salmonella strain expressing chromosomally SifA deleted of its C-term domain [sifA ]. As small deletions throughout SifA are sufficient to block its secretion by T3SS-2 26 , we first verified that SifA(1-136) was secreted. For this, we engineered strains expressing SifA with an internal double haemagglutinin tag at the boundary between the N-and C-term domains and deleted [sifA(1-136)-2HA] or not (sifA-2HA) of its C-term domain. SifA(1-136) was efficiently secreted by the T3SS-2 ( Supplementary Fig. S2) and as stable as SifA (Supplementary Fig. S1). These strains were tested in mixed Values are means ± SD of three independent experiments. Survival curves were compared using the logrank test and two-tailed P values are reported. Unpaired t-tests were used to compare two values. P values: *P < 0.05; **P < 0.01; ***P < 0.001. A onesample t-test was used to determine whether a CI was significantly different of one, and unpaired t-tests to compare two values. P values: ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. P.O. infections. We found that as compared to wild type Salmonella, the sifA(1-136) mutant presented in C57BL/6 mice a strong attenuation of virulence similar to that of a ∆sifA strain (CI = 0.05 ± 0.02, Fig. 3C,M). Therefore, we compared the virulence of sifA  and ∆sifA mutant strains (Fig. 3F,M) and obtained in C57BL/6 mice a CI of 0.45 ± 0.29. This value is similar to that found for the SKIP-dependent role of SifA to virulence (Fig. 3B,K), suggesting that the contribution to virulence of the SifA N-term is solely mediated by SKIP. In SKIP −/− mice (Fig. 3H) this mixed infection gave a CI of ≈1 (1.03 ± 0.42, Fig. 3M) that validated the hypothesis. We concluded that the SKIP-independent function observed in other mixed infections (Fig. 3D,E) is borne by the SifA C-term domain.
SifA and SseJ contribute independently to Salmonella Virulence. The T3SS-2 effector SseJ exerts a lipase activity that increases the esterification of cholesterol in host cell membranes 27 . SseJ is activated by binding the GTP-bound form of the eukaryotic RhoA GTPase 28 while SifA binds preferentially GDP-bound RhoA. Though a GEF activity of SifA toward RhoA could not been demonstrated 21 , it has been proposed that SifA participates to the activation of SseJ by favouring the recruitment of RhoA to the SCV and possibly its activation 29 . Thus, we tested whether the role of the SifA C-term in virulence could be linked to this proposed activity. In I.P inoculated mice we compared the virulence attenuation of strains deleted of sseJ, sifA or both genes. We found that a ∆sifA∆sseJ strain is more attenuated than each individual mutant as compared to wild type Salmonella (Fig. 3N) and that the CIs of ∆sifA versus ∆sifA∆sseJ is not different from the CI of wild type versus ∆sseJ (0.54 ± 0.01 and 0.47 ± 0.16, respectively). These findings suggest that the two genes are involved in distinct signalling pathways. Therefore the function of the SifA C-term in the mouse model of infection is probably not linked to function of RhoA-SseJ in virulence.
The SifA C-term is important for the recruitment of LAMP1 to the SCV and for intracellular replication. We finally investigated how the C-term domain contributes to the SifA function.
Compared to the wild type, the ∆sifA SCV is characterized by very low levels of lysosomal glycoproteins 4 . Thus, we explored whether the SifA C-term could be important for the recruitment of these membrane proteins to the SCV. Firstly, we explored this phenotype in BMMs prepared from C57BL/6 and SKIP −/− mice and infected with wild type or ∆sifA Salmonella strains. The presence of LAMP1 on SCVs at 16 hours post-infection was examined and illustrated by confocal microscopy (Fig. 1D,F). Most wild type SCVs were LAMP1-positive in macrophages expressing or not SKIP (91 ± 3% and 85 ± 6%, respectively) while the fraction of ∆sifA SCVs decorated by the anti-LAMP1 was below 50% in both types of macrophages (48 ± 3% versus 35 ± 12%). This indicates that the recruitment of LAMP1 requires SifA but not its interaction with SKIP. To confirm this result, we infected HeLa cells with the same Salmonella strains and also a strain expressing chromosomally SifA L130D . We observed that expression of SifA L130D increased noticeably the fraction of LAMP1-positive SCVs (Fig. 4A,B) thus confirming the prominent role of the SifA C-term in this phenotype. This result prompted us to compare the capacity of these various strains to replicate intracellularly. These experiments were performed in mouse macrophages as strains deleted of sifA 4 or expressing SifA L130D14 partly escape the vacuole and replicate well in the epithelial cell cytosol 30 . We found that the expression of SifA L130D increases the intracellular bacterial growth by a factor ~1.5 (Fig. 4C). Although modest this increase was significant. Collectively, these results establish that the SifA C-term supports the recruitment of LGPs to SCVs. This is associated with a fairly small but robust positive effect on Salmonella intracellular replication.

Discussion
For this study, we developed a SKIP −/− mouse model of acute Salmonella infection. It helped us to answer the question of whether the virulence mediated by SifA is only supported by its interaction with SKIP and subsequently whether the C-terminal domain of SifA is functional on its own. Our results show that both the SifA N-term via SKIP and the SifA C-term play a role during Salmonella infection. Remarkably, the two domains are probably involved in the same signalling cascade.
SKIP −/− mice were maintained as a colony of homozygote animals. These mice are viable, have a good prolificity and do not present signs of disease or debility. Thus, SKIP is dispensable for the life of C57BL/6 mice indicating that in laboratory stabling conditions the function of this protein is not required and/ or that another protein is capable of complementing the functions of SKIP. At the cellular level we did not observed the Golgi scattering 12 or the clustering of the LAMP1 compartment 31 that were previously described upon siRNA-mediated knock-down of SKIP expression. Thus, the function of SKIP regarding the positioning and the organization of these organelles is complemented and these results rather support the hypothesis of a protein compensating for the lack of SKIP.
We found that SKIP −/− mice are more resistant to a Salmonella challenge than congenic C57BL/6 mice and we observed very similar profiles for the survival curves of wild type Salmonella in SKIP −/− mice and for a ∆sifA strain in C57BL/6. This SKIP-mediated susceptibility to salmonellosis likely reflects the role played by the SifA-SKIP interaction during the infection. The SifA-SKIP complex interacts and activates kinesin-1 and thereby favours the formation of vesicles and tubules 17 . By sequestering rab9, it also inhibits the retrograde transport of mannose 6-phosphate receptors to the trans-Golgi network and the delivery of lysosomal enzymes to lysosomes 32 . Both functions are probably affected by the absence of SKIP and their deficiencies contribute to the decreased susceptibility of SKIP −/− mice to salmonellosis.
These results are seemingly contradictory with the higher susceptibility of SKIP −/− mice to a Salmonella challenge reported online by the Sanger Institute. This laboratory infects mice I.V. with a sub-lethal dose of an attenuated strain of S. Typhimurium. In these conditions, infected C57BL/6 mice survive and resolve the infection within approximately one month while they succumb within one week when challenged P.O. or I.P. with the dose of S. Typhimurium 12023 used in this study ( Fig. 2A). A rough estimation based on the Sanger Institute's data indicates that bacterial loads in the liver and the spleen fourteen days post I.P. inoculation are several logs (> 4) lower than those we observed after two days. Therefore, SKIP is required for acute infection of susceptible mice by a virulent Salmonella strain while it is necessary for the host resistance during later stage of infection by an attenuated Salmonella strain.
By performing in vivo competitions of Salmonella strains we found that SifA exerts a SKIP-independent function that is borne by its C-term domain. Hence, SifA has two domains and two functions. A comprehensive analysis of the CI results emphasizes other important information regarding the functional and physical interactions between the two domains.
A putative caspase 3 cleavage site separates the N-and C-term domains of SifA 15 but it is not known whether the two domains are really split apart after SifA translocation. We assessed the role of the SifA N-term domain in virulence in the context of a membrane bound (Fig. 3B,K, WT versus sifA L130D ) or cytosolic (Fig. 3F,M, sifA  versus ∆sifA) protein. In both cases we obtained CI values of ≈0.5 indicating that the N-term domain of SifA is equally functional in either context. Therefore, these data do not exclude the possibility that translocated SifA is cleaved.
The consequences of the absence of one or the other domain of SifA on virulence are markedly different. While the lack of a functional N-term domain impacts moderately the virulence of Salmonella (Fig. 3B,K, CI ≈ 0.5), the absence of the C-term domain (Fig. 3C,M, CI ≈ 0.05) diminishes dramatically the virulence of the strain. Thus, despite its lack of characterization, the C-term is clearly crucial for SifA functions. The role in virulence of this domain was analysed in the context of a fully functional molecule (Fig. 3C,M, WT versus sifA ) or in the absence of a functional N-term domain (Fig. 3E,L, sifA L130D versus ∆sifA, and Fig. 3D,I, WT versus ∆sifA in SKIP −/− mice). The CI values were considerably divergent and this observation reveals that the contribution of the SifA C-term to virulence is far more important in the presence of a functional N-term domain. The SifA C-term is poorly functional in the absence of SKIP or a N-term domain capable of interacting with SKIP and this entails that the SifA C-term acts upstream of the SifA-SKIP signalling cascade. Therefore, it appears that the two domains of SifA are functionally but not necessarily physically linked.
Very recent studies have highlighted the role of the host protein Plekhm1 in Salmonella infection. This protein interacts with the SifA N-term domain and supports Salmonella intracellular replication 18 . These results are apparently inconsistent with our present data indicating that SKIP only mediates the functions of the SifA N-term domain. One can hypothesise that Plekhm1 is unnecessary for the virulence in the mouse model of infection or that Plekhm1 needs SKIP to play its role in the infectious process. This last hypothesis is however unlikely as the two proteins are competing for binding SifA 18 .
We have previously shown that the plasmidic expression sifA L130D in a ∆sifA strain increases the frequency of mouse macrophages enclosing more than 10 bacteria 14 and this suggested that SifA L130D supports Salmonella replication. The present study confirms this observation. In addition, we observed that SifA L130D increases the proportion of LAMP1-positive SCVs. Owing to the increased recruitment of LGPs, the sifA L130D SCV is probably more stable and it might explain the slightly better replication of this Salmonella strain in macrophages. SifA L130D is more rapidly degraded than wild type SifA in a eukaryotic context, may be because of its lack of interactions with SKIP. Thus, our results may underestimate the functional impact of the SifA C-term. This possibility is however balanced by the consistency of our observations regarding the composition of the SCV and the virulence for the sifA L130D mutant and the wild type strain in the absence of SKIP.
Considering the GEF-like conformation of the SifA C-term 13,14 , we questioned the function of this domain and paid a special attention to its possible interaction with host GTPases. SifA binds the GDP-bound RhoA and may support the SseJ function by recruiting the GTPase to the SCV membrane 29 . Our results do not support this hypothesis as we found that the products of sifA and sseJ contribute independently to Salmonella virulence in the mouse model of infection. However, the sifA and sseJ interaction might be host-specific.
The present study reveals that the two domains of SifA are functional and participate to the same pathway supporting the important role of SifA in virulence. Considering the GEF-like structure of the SifA C-term, this domain likely interacts and modulates the activity of a GTPase. Studies to identify this protein may provide important insights into the mechanisms by which SifA regulates the membrane dynamics of the bacterial vacuole.

Material.
A hematology analyser (HORIBA ABX Pentra 60 C, calibrated for mouse cells) was used for analysis of whole-blood specimens. Plasma AST and ALT activities were assayed using a Cobas C 501/502 analyser (Roche Diagnostics). For quantification of western blotting, chemiluminescence signal was read with an Azure biosystems C300.
Antibodies. HA-tagged proteins were detected using a mouse monoclonal anti HA (Covance, clone 16B12). A rabbit polyclonal serum was used to detect SKIP 16 . The rabbit anti-kinesin HC (PCP42) (a kind gift from R. Vale) was absorbed with Salmonella acetone powder 33 . Secondary antibodies for Western blotting were goat anti-mouse or anti-rabbit IgG HRP conjugate (Sigma-Aldrich). Fluorescent Alexa secondary antibodies were obtained from Jackson ImmunoResearch Mouse Strains. C57BL/6 were obtained from Charles River Laboratory. SKIP −/− mice (B6N;B6J-Tyr c-Brd Plekhm2 tm1a(EUCOMM)Wtsi/Wtsi ) were obtained from the Wellcome Trust Sanger Institute (EUCOMM Consortium). A mouse colony was maintained by incrossing homozygotes, which had been genotyped as described by the Sanger Institute.
Ethic statement. Animal  strain, respectively. The amplicon of pKD4 with the oligo pairs O-745/O-71 was recombined in the sifA-2HAsc4 strain (WZ012sc4) to obtain the sifA(1-136)-2HA::Km R (WZ019). For WZ039, the amplicon of pKD4-SifA L130D -2HA with oligos O-686 and O-687 was used to perform a chromosomal recombination in a ∆sifA::FRT strain. For WZ041 (sifA-2HA, ∆ssaV), a P22 lysate of the strain AAG057 was used to transduce ∆ssaV::Km R into the WZ012sc4 strain. WZ042 and WZ043 strains were obtained by P22-mediated transduction of ∆sifA::Km R or pipB2-2HA::Km R into a ∆sseJ strain, respectively. Eukaryotic cells and culture conditions. RAW 264.7, HeLa, primary bone marrow-derived macrophages and embryonic fibroblasts were grown in DMEM (GibcoBRL) supplemented with 10% foetal calf serum (FCS; GibcoBRL), 2 mM nonessential amino acids, and glutamine (GibcoBRL) at 37 °C in 5% CO 2 . For peritoneal macrophages, mice (C57BL/6 or SKIP −/− ) were injected intraperitoneally with a thioglycollate solution for a volume of 1 ml per mouse. Four days post injection, the thioglycollate-pretreated mice were sacrificed and macrophages were harvested form the peritoneal cavity by washing with 5 ml of cold PBS. Cells were collected from the washing solution by centrifugation, washed again twice with cold PBS, resuspended in DMEM based growing medium and seeded at a density of 10 5 cells per cm 2 in 6-or 24-well plates. Cells were used for infection after 1 day culture. For preparing mouse embryonic fibroblasts C57BL/6 or SKIP −/− mice were sacrificed at 13-14 days gestation. The uterine horns were collected and washed 3 times with 10 ml PBS. Then, visceral tissues were separated from embryos. Embryos were washed again for 3 times with PBS and then finely minced with a curved dissecting scissors. A volume of 2 ml of trypsin was added and incubated for 5 min, during which the tissue was minced. 5 ml trypsin were added and the cells were pipetted vigorously up and down. The cells were placed into incubator for 20-30 min and again pipetted vigorously up and down. Cells were diluted in DMEM-derived growing medium and seeded in 75 cm 2 flasks and incubated at 37 °C in a tissue culture incubator until the flasks are at least 90% confluent. Then, the cells were split using trypsin and seeded in 6-or 24-well plates for infection.

Bacterial infection and replication assays.
Bone marrow-derived macrophages, HeLa and RAW 264.7 macrophages were grown, infected and treated as previously described 7 .
Competitive index. C57BL/6 or congenic SKIP −/− mice (eight to ten weeks old) were inoculated intraperitoneally or perorally with equal amounts of two bacterial strains for a total of 10 5 bacteria per mouse. The spleens were harvested two (I.P.) or five (P.O.) days after inoculation and homogenized. Bacteria were recovered and enumerated after plating a dilution series onto LB agar with the appropriate antibiotics. Competitive indexes (CI) were determined for each mouse 38,24 . The CI is defined as the ratio between the mutant and wild type strains within the output (bacteria recovered from the mouse after infection) divided by their ratios within the input (initial inoculum).