|Home | About | Journals | Submit | Contact Us | Français|
Francisella tularensis is a fastidious Gram-negative bacterium responsible for the zoonotic disease tularemia. Investigation of the biology and molecular pathogenesis of F. tularensis has been limited by the difficulties in manipulating such a highly pathogenic organism and by a lack of genetic tools. However, recent advances have substantially improved the ability of researchers to genetically manipulate this organism. To expand the molecular toolbox we have developed two systems to stably integrate genetic elements in single-copy into the F. tularensis genome. The first system is based upon the ability of transposon Tn7 to insert in both a site- and orientation-specific manner at high frequency into the attTn7 site located downstream of the highly conserved glmS gene. The second system consists of a sacB-based suicide plasmid used for allelic exchange of unmarked elements with the blaB gene, encoding a β-lactamase, resulting in the replacement of blaB with the element and the loss of ampicillin resistance. To test these new tools we used them to complement a novel d-glutamate auxotroph of F. tularensis LVS, created using an improved sacB-based allelic exchange plasmid. These new systems will be helpful for the genetic manipulation of F. tularensis in studies of tularemia biology, especially where the use of multi-copy plasmids or antibiotic markers may not be suitable.
Tularemia has a variety of clinical manifestations depending on route of entry, subspecies of bacteria and inoculum size, while the disease state can range from mild to fatal (Ellis et al., 2002). There are three official subspecies of the causative agent Francisella tularensis: the highly pathogenic F. tularensis subsp. tularensis, and the less pathogenic F. tularensis subsp. holarctica and F. tularensis subsp. mediaasiatica. The high mortality of F. tularensis subsp. tularensis pneumonic tularemia and its high infectivity has raised concerns about its potential use as a biological weapon (Dennis et al., 2001). These concerns have prompted the US Centers for Disease Control (CDC) to classify F. tularensis as a Select Agent.
A wide variety of genetic tools have recently been developed for the manipulation of Francisella. These include Escherichia coli–Francisella shuttle vectors (Bina et al., 2006; LoVullo et al., 2006; Ludu et al., 2008; Maier et al., 2004; Rasko et al., 2007), random transposon mutagenesis systems based on EZ-Tn5, Himar1 and Tn5 (Buchan et al., 2008; Kawula et al., 2004; LoVullo et al., 2006; Maier et al., 2006; Qin & Mann, 2006), as well as methods for allelic exchange (Golovliov et al., 2003; LoVullo et al., 2006; Ludu et al., 2008; Rodriguez et al., 2008; Twine et al., 2005). One methodology that has not been fully explored is the integration of genetic elements into the F. tularensis chromosome. In other bacteria this has been accomplished using non-replicative vectors containing an attachment site and integrase gene from a lysogenic bacteriophage (Hoang et al., 2000; Stover et al., 1991). This approach is not possible for F. tularensis, since phages that are infective for this organism have yet to be discovered.
In this paper, we report the development of two single-copy integration systems for incorporating genetic elements into the Francisella genome. The first system is based on the transposon Tn7 and takes advantage of its ability to insert in both a site- and orientation-specific manner at high frequency into the attTn7 site, located downstream of the highly conserved glmS gene, which encodes the essential glucosamine-6-phosphate synthetase (Peters & Craig, 2001). This system has been used in a number of pathogens, including Pseudomonas aeruginosa, E. coli, Salmonella typhimurium and the Select Agents Burkholderia mallei, Burkholderia pseudomallei and Yersinia pestis (Choi et al., 2005, 2006, 2008; McKenzie & Craig, 2006). As the insertion occurs in an intergenic region the fitness of the modified organisms appears to be unchanged (Choi et al., 2005; Peters & Craig, 2001).
We constructed a mini-Tn7 vector, which has a kanamycin-resistance marker flanked by γδ-res sites, and a helper plasmid, which encodes the site-specific Tn7 transposase complex TnsABCD, expressed from an F. tularensis promoter. We confirmed the ability of the mini-Tn7 to stably insert at the attTn7 site in the F. tularensis chromosome. We also showed that the kanamycin marker can be efficiently excised by the γδ-resolvase. The ability to remove the kanamycin marker is important, because F. tularensis genetics has a limited repertoire of Select Agent approved markers (Titball et al., 2007).
The second system uses a sacB-based suicide plasmid expressing kanamycin resistance that is used for allelic exchange of unmarked elements with the blaB gene, which encodes the only functional β-lactamase in F. tularensis (Bina et al., 2006; LoVullo et al., 2006). The deletion of the blaB gene allows for convenient screening of desired recombinants based on their sensitivity to ampicillin.
E. coli DH10B (Table 1) was used for routine cloning procedures and was grown in Luria–Bertani (LB) broth (BD Biosciences) or on LB agar. E. coli HB101 was used to maintain all plasmids containing the γδ-res cassettes and was grown as described above. F. tularensis strains (Table 1) were grown as previously reported (LoVullo et al., 2006). Specifically, strains were grown at 37 °C in liquid modified Mueller–Hinton medium (MMH), which is Mueller–Hinton broth (BD Biosciences) supplemented with 1.0% (w/v) glucose, 0.025% (w/v) ferric pyrophosphate (Sigma-Aldrich) and 0.05% (w/v) l-cysteine free base (Calbiochem), or on MMH agar, which is the MMH medium described above supplemented with 1.0% (w/v) proteose peptone (BD Biosciences), 2.5% (v/v) defibrinated sheep blood (Remel) and 1.5% (w/v) bacto-agar (BD Biosciences). When necessary, ampicillin (Ap; Sigma-Aldrich) was added at 100 or 50 μg ml−1, respectively, for E. coli or F. tularensis, while kanamycin (Km; Sigma-Aldrich) was used at 50 μg ml−1 for E. coli and 5 μgml−1 for F. tularensis strains LVS and Schu. Kanamycin stock solutions were made by accounting for the concentration of active kanamycin in each lot. Hygromycin B (Hyg; Roche Applied Science) was used at 200 μg ml−1 for all species and strains. Sucrose was used at a final concentration of 8 or 5% (w/v) depending on the sacB vector. The β-galactosidase substrate X-Gal (Invitrogen) was used at 50 μg ml−1 in MMH agar lacking sheep blood. d-Glutamic acid (Sigma-Aldrich) was used at a final concentration of 200 μg ml−1 in MMH broth and in MMH agar lacking proteose peptone and sheep blood.
Electroporations and allelic exchange experiments were done as described previously (LoVullo et al., 2006).
DNA methods were performed essentially as described by Ausubel et al. (1987). DNA fragments were isolated using agarose gel electrophoresis and QIAquick spin columns (Qiagen). Oligonucleotides were synthesized by Invitrogen Life Technologies. Oligonucleotides flanking the Tn7 attachment site were attF, 5′-ATGCAGGACATGATTTTAGTG (forward 5′ to attTn7), and attR, 5′-TTATGTTGAGTCCATATTTCAG (reverse 3′ to attTn7). All restriction endonucleases and DNA modifying or polymerase enzymes were from New England Biolabs or Fermentas. PCRs were performed with Iproof High-Fidelity DNA Polymerase (Bio-Rad) according to the manufacturer's recommendations. All plasmids used in this study (Table 1) were from the authors' collections. Preparation of plasmid and genomic DNA from E. coli and F. tularensis was done as previously reported (LoVullo et al., 2006).
The assays were performed on whole cell suspensions according to a standard protocol (Miller, 1972).
Plasmids used in this study are described in Table 1. Detailed descriptions of the construction of the plasmids used in this study can be obtained from the corresponding author. Information about plasmid construction that is pertinent to the understanding of this work is described below.
The R6K origin from plasmid pTNS2 (accession no. AY884833) was replaced with the pUC ori from pBluescript II KS(+) to produce pMP650. The tnsABCD operon from pMP650 was cloned into the multiple cloning site of pMP658 to produce pMP685. Inverse PCR was performed on pMP685, eliminating 380 bp between the tnsABCD operon and the blaB promoter, producing pMP720.
The R6K origin from plasmid pUC18R6KT mini-Tn7T (GenBank accession no. AY712953) was replaced with the pUC ori from pBluescript II KS(+) to produce the empty mini-Tn7 pMP651. The PgroESL–aphA-1 from pMP527 was flanked with γδ-res sites and ligated into pMP651 to produce pMP749.
The R6K origin from plasmid pUC18R6KT mini-Tn7T gnELp-Kan-GFP was replaced with the pUC ori from pBluescript II KS(+) to produce pMP661. The F. tularensis rpsL promoter region was amplified from Schu genomic DNA by PCR and placed upstream of gfp, encoding GFPmut3 (Cormack et al., 1996), which produced pMP761. This PrpsL–gfp fragment was then inserted into the multiple cloning site of pMP749, generating pMP793.
The γδ-resolvase gene, tnpR, was obtained from pGH542 and inserted into the multiple cloning site of pMP658 to generate pMP672.
Inverse PCR was performed on the suicide vector pMP590 to eliminate the pFNL10 ori yielding pMP671. The flanking regions of blaB were amplified from Schu genomic DNA and cloned upstream (5′ flanking) and downstream (3′ flanking) of the multiple cloning site of suicide vector pMP671 to produce the blaB integration vector pMP719. The lacZ gene was cloned into the multiple cloning site of pMP719 to form pMP741, and then the rpsL promoter region was amplified from Schu genomic DNA and placed upstream of the lacZ gene to form pMP790. This construct was created to test for heterologous gene expression in the blaB locus.
To improve the counterselectable marker, the repA promoter region upstream of sacB of pMP719 was removed by restriction digestion. This region was replaced with the dnaK promoter region amplified from F. tularensis Schu genomic DNA using PCR and cloned upstream of sacB to form pMP799. Inverse PCR was then performed with pMP799 to eliminate 192 bp of unnecessary DNA upstream of PdnaK to produce pMP815.
The Francisella ori region and repA promoter region upstream of sacB were removed from suicide vector pMP590 by restriction digestion. This region was replaced with the dnaK promoter and cloned upstream of sacB to form pMP780. Inverse PCR was performed as above to eliminate 192 bp of unnecessary DNA upstream of PdnaK to produce pMP812.
A DNA fragment containing murI was obtained from strain LVS genomic DNA using PCR and cloned into pMP812 to yield pMP880. An in-frame deletion of 768 bp within murI was made using PCR to yield pMP884. There are 962 bp upstream and 982 bp downstream of the ΔmurI1 allele in this plasmid.
The F. tularensis murI gene was amplified from LVS genomic DNA by PCR and placed downstream of PrpsL in pMP767, forming pMP889. This PrpsL–murI fragment was then inserted into the multiple cloning site of pMP749, generating pMP890.
The PrpsL–murI fragment was amplified from pMP889 by PCR and inserted into the multiple cloning site of pMP815, generating pMP895.
The most common implementation of the Tn7 system is to simultaneously deliver both transposon and transposase to the cells on separate suicide plasmids (Choi et al., 2005). This allows for transient expression of the transposase and subsequent integration of the transposon in the chromosome without replication of the delivery plasmids. This approach did not work with F. tularensis using the suicide transposase plasmid pTNS2 (Table 1) and an early generation Tn7 suicide plasmid. We first hypothesized that the lac promoter driving the 6 kb tnsABCD operon in the helper plasmid was not functional in F. tularensis. We created another suicide helper plasmid with the operon cloned downstream of the Francisella groESL promoter, which is the same promoter that we have used for the expression of selectable markers. However, this approach was also unsuccessful. We found that the tnsABCD genes are not optimal for the codon preference of F. tularensis and this, coupled with the size of the operon, probably prevented the cells from producing enough transposase proteins to catalyse transposition during the transient expression period. We then hypothesized that expressing the operon from a replicating plasmid prior to introduction of the Tn7 suicide plasmid would allow time for the cell to produce sufficient amounts of transposase. A similar method using a temperature-sensitive helper plasmid has been reported to express the transposase in E. coli and S. typhimurium (McKenzie & Craig, 2006). Toward this end, we created the helper plasmid pMP720 (Fig. 1a) from the unstable hygromycin-resistant shuttle plasmid pMP658 (LoVullo et al., 2008), which contains the transposase operon expressed from the Francisella blaB promoter. This revised strategy proved successful, as described below.
In addition to the helper plasmid, we created a mini-Tn7 element on a suicide vector. The plasmid pMP749 (Fig. 1b) contains the kanamycin-resistance marker aphA-1 driven by the Francisella groESL promoter, flanked by γδ-res DNA binding sites for the site-specific γδ-resolvase of E. coli transposon Tn1000 (Bardarov et al., 2002). It also contains two terminators (T0 and T1) to prevent read-through from the glmS promoter after chromosomal insertion (Choi et al., 2005) and a multiple cloning site for cloning DNA elements. An additional mini-Tn7 construct, pMP793 (Fig. 1c), was made to express GFP, and has a Francisella rpsL promoter driving gfp cloned into the multiple cloning site of pMP749.
The methodology we developed is shown in Fig. 2. First, the helper plasmid pMP720 is electroporated into F. tularensis and transformants selected by hygromycin resistance. One clone is prepared for electroporation while maintaining hygromycin selection. We then transform the strain with the mini-Tn7 plasmid pMP749, and select for kanamycin-resistant clones. We routinely obtain ~104 kanamycin-resistant LVS transformants per electroporation with ~1 μg pMP749 DNA. We obtained similar results with the mini-Tn7 plasmid expressing GFP, pMP793, with both LVS and Schu, resulting in ~104 kanamycin-resistant transformants per electroporation. We grew kanamycin-resistant transformants overnight in liquid media lacking selection and these were subcultured 1:10 and grown for an additional 24 h, after which they were plated on medium lacking antibiotics. The antibiotic-resistance phenotypes of the resulting clones were then screened, and we found that the hygromycin-resistance helper plasmid pMP720 was lost from the population at a frequency of 50–80%, while kanamycin resistance, encoded in Tn7, was maintained at 100% in the population. This confirmed our expectations that the helper plasmid would be readily lost from the population but that the Tn7 would be stably maintained. We confirmed the presence of the kanamycin-resistant transposon insertion at the attTn7 site using PCR with primers attF and attR (see Methods), which lie outside the attTn7 site (Fig. 3). Sequence analysis of five LVS and five Schu clones determined that the insertion site occurs at either 25 bp (eight insertions) or 26 bp (two insertions) downstream of the glmS stop codon (data not shown), regardless of strain. This is similar to the behaviour of Tn7 in P. aeruginosa, in which the transposon inserts at two sites, either 24 or 25 bp downstream of glmS (Choi et al., 2005). In contrast, Tn7 inserts into a single site 25 bp downstream of glmS in Y. pestis and E. coli (Choi et al., 2005; DeBoy & Craig, 1996). Southern blot analysis using the aphA-1 gene as a probe confirmed that there were no additional insertions in the LVS chromosome (data not shown). This is in agreement with the observations that transposition mediated by TnsABCD yields insertions only at attTn7 (Peters & Craig, 2001), and that Tn7 also confers immunity, whereby it blocks transposition into a site already occupied by a Tn7 element (DeBoy & Craig, 1996). We used confocal microscopy to visualize the GFP in the LVS Tn7 strain, but only ~10% of cells in each field expressed GFP at one time (data not shown). We believe that a multitude of factors could have been responsible for the poor visualization, including promoter strength, improper folding of GFP, degradation and photobleaching.
After confirming the loss of the helper plasmid we tested the γδ-resolvase system. We transformed LVS and Schu containing Tn7 insertions with plasmid pMP672 (Fig. 1d), an unstable hygromycin shuttle vector expressing the γδ-resolvase from the F. tularensis blaB promoter. Select clones were then grown in liquid media containing hygromycin overnight and plated for single colonies on hygromycin medium. These were then screened for loss of kanamycin resistance, which occurred at a frequency of ~80% in both LVS and Schu. Kanamycin-sensitive clones were cured of the γδ-resolvase plasmid in the same manner as the helper plasmid. We then confirmed the loss of the kanamycin marker with PCR, utilizing the Tn7 attF and attR primers (Fig. 3). Sequence analysis of a resolved clone confirmed that the two γδ-res sites recombined into one γδ-res site with loss of the aphA-1 marker (data not shown).
We have previously shown that LVS and Schu contain only one functional β-lactamase, blaB (LoVullo et al., 2006). The blaB gene lies in the Schu S4 chromosome with a hypothetical gene transcribed in the opposite direction 435 bp upstream of its start site and a potential DNA/RNA endonuclease family protein transcribed in the opposite direction overlapping the blaB stop codon by 10 bp (Larsson et al., 2005). Based on our sacB-based suicide plasmid, we created a blaB integration vector that contains 1002 bp upstream of the blaB gene, a multiple cloning site, and 593 bp downstream of the blaB gene that includes the overlapping DNA/RNA endonuclease sequence.
The advantage of the blaB integration system is that we can quickly screen for the clones with the unmarked insertion in the secondary recombinant pool as they will be ampicillin-sensitive and kanamycin-sensitive. This is in contrast to the integration system developed for Francisella novicida, which integrates into a gene that is present only in the F. novicida chromosome and retains the kanamycin selectable marker (Ludu et al., 2008).
Our blaB integration vector, pMP719 (Fig. 4a), is based on the suicide vector pMP671, a pFNL10 Δori derivative of pMP590 (LoVullo et al., 2006). To test the ability of the system to integrate elements by allelic exchange, we cloned lacZ under the Francisella rpsL promoter into the multiple cloning site to form pMP790 (Fig. 4b). We selected the rpsL promoter because we knew from our studies with the PrpsL–gfp cassette that the promoter is active in E. coli (data not shown), and therefore allowed us to confirm β-galactosidase production in E. coli before moving the system into F. tularensis. We performed an allelic exchange experiment similar to that shown in Fig. 5 for both LVS and Schu. We confirmed the expression of lacZ from the chromosomes of both strains by patching ampicillin-sensitive colonies onto medium with X-Gal and observing blue colonies (data not shown). We also confirmed the activity of the protein produced in LVS by performing β-galactosidase assays. We compared two ΔblaB::PrpsL–lacZ strains with wild-type LVS. Three assays were performed on each strain: wild-type LVS averaged 4 Miller units and the two integrated strains averaged 300 Miller units, indicating expression of lacZ in the novel location within the chromosome.
Our previously described sacB suicide vector, pMP590 (LoVullo et al., 2006), has proven itself useful for the sucrose counterselection-mediated construction of unmarked, in-frame deletions in both LVS and Schu, but needed improvement (LoVullo et al., 2006). This allelic exchange vector was developed from a shuttle vector that was made non-replicative in F. tularensis by replacement of the repA and ORF2 genes by the sacB gene, driven by the repA promoter. However, the pFNL10 ori sequence is still present in this plasmid, which could be problematic in experiments that test gene essentiality by deleting a gene in the presence of a plasmid carrying a wild-type copy of the gene. In such an experiment, trans-acting replication proteins from the plasmid could recognize the suicide vector-borne ori sequence in the chromosome and initiate replication that would likely be lethal due to incompatibility with the natural chromosomal origin of replication. To solve this problem, we removed the ori sequences and repA promoter and inserted the F. tularensis dnaK promoter region upstream of the sacB gene, which allowed for strong expression of sacB such that the concentration of sucrose in the selection medium could be reduced from 8 to 5%, while maintaining a very clean selection (data not shown). This new plasmid was then subjected to inverse PCR to remove 192 bp of DNA upstream of the dnaK promoter region. This was done to prevent the integration of the suicide vector into the chromosomal dnaK region, which occurred often in test experiments (data not shown). The final suicide vector, pMP812, is shown in Fig. 6(a). Note that the dnaK promoter does not seem to be recognized by E. coli, as pMP812 transformants are not sensitive to sucrose. These improvements to our basic sacB vector led us to modify our original blaB integration vector, pMP719, in the same manner, resulting in pMP815 (Fig. 4c).
To test these new tools, we sought to construct a strain with a novel mutation, characterize it, and then complement it with a wild-type copy of the gene inserted into the chromosome using each integration system. We chose to interrupt the production of d-glutamate by deleting the murI gene, which encodes a glutamate racemase that is essential to E. coli (Doublet et al., 1993). d-Glutamate is indispensable for the biosynthesis of peptidoglycan in most eubacteria, and is produced through two known routes: by d-amino acid transferase (d-AAT), which converts α-ketoglutarate to d-glutamate by transamination with d-alanine provided by the alanine racemase reaction; and by glutamate racemase, which produces d-glutamate through the racemization of l-glutamate (Liu et al., 1998). In a number of bacteria, most notably in certain Bacillus species, d-glutamate is also shuttled into the production of poly-γ-d-glutamic acid (PGA). In Bacillus anthracis, there are two glutamate racemases (RacE1 and RacE2) that produce d-glutamate for peptidoglycan as well as the biosynthesis of PGA, which constitutes an antiphagocytic capsule, one of the two major virulence factors of B. anthracis (Dodd et al., 2007; Mock & Fouet, 2001). F. tularensis has capB and capC genes, which share sequence homology at the amino acid level of 38 and 29% with the CapB and CapC proteins of B. anthracis that synthesize PGA (Su et al., 2007), and a number of groups have shown that the capBC genes of F. tularensis are important for tularemia pathogenesis (Maier et al., 2007; Su et al., 2007; Weiss et al., 2007). However, there is as yet no evidence that F. tularensis produces PGA.
Our inspection of the genomes of F. tularensis strains and of F. novicida indicates that the gene FTT1197c, annotated as murI, is the only glutamate racemase present in these organisms, and we could find no genes encoding a d-AAT, suggesting that there is only one source of d-glutamate in these bacteria. However, a comprehensive transposon library of the F. novicida genome has one mutant with an insertion in the middle of the murI gene, and d-glutamic acid was not included in the selection medium (Gallagher et al., 2007). This might indicate that murI is dispensable.
To clarify this matter, we sought to determine whether a murI deletion mutant would be auxotrophic for d-glutamate, indicating the lack of any additional amino acid racemase or d-AAT capable of compensating for the loss of murI. We constructed an in-frame deletion of murI in our improved sacB-based suicide vector, pMP812. Since it has been shown that it is possible to rescue d-glutamate auxotrophs with exogenous d-glutamate in Bacillus subtilis and E. coli B/r and K-12 strains (Ashiuchi et al., 2007; Doublet et al., 1993; Hoffmann et al., 1972), we performed a standard two-step allelic exchange (LoVullo et al., 2006) using pMP884 (Table 1) with d-glutamic acid present in the medium. We then picked and patched 24 sucrose-resistant secondary recombinants onto media with or without d-glutamic acid. This yielded two recombinants that could not grow on media lacking d-glutamic acid. PCR was used to confirm that the secondary recombinants auxotrophic for d-glutamic acid were murI deletions and that the d-glutamic acid prototrophs were wild-type recombinants (Fig. 6b). The murI deletion mutants formed smaller colonies than the wild-type, suggesting a growth defect. This is probably the reason why the phenotypes of the secondary recombinants were skewed towards the wild-type. One mutant, PM2181, was selected for further study.
We performed complementation studies on PM2181 utilizing the Francisella rpsL promoter driving murI in the Tn7 system and the blaB system. The mini-Tn7 vector pMP749 bearing PrpsL–murI+ was introduced into PM2181 as shown in Fig. 2. The resulting attTn7::PrpsL–murI+ strain, PM2194, was prototrophic for d-glutamic acid. We then resolved the kanamycin marker to ensure that it had no effect on complementation of the murI lesion. As expected, the resolved strain, PM2209, was able to grow in the absence of d-glutamic acid. For the blaB system, a two-step allelic exchange as shown in Fig. 5 was performed with pMP815 bearing PrpsL–murI+. The resulting strain, PM2210, was also able to grow in the absence of d-glutamic acid. These results confirm that there is only one pathway for d-glutamate biosynthesis in wild-type Francisella. We do not know whether MurI also supplies d-glutamate for PGA biosynthesis, but this mutant could be used to identify such a polymer if it exists, since it should be possible to label PGA with radioactive d-glutamic acid supplied to PM2181 in culture.
The existence of a murI transposon mutant of F. novicida (Gallagher et al., 2007) obtained without d-glutamate supplementation suggests that there may be an extragenic suppressor mutation in this mutant. To test this possibility we performed suppressor analysis of our LVS d-glutamate auxotroph. The strain was grown to saturation, washed, and serially diluted on plates with and without d-glutamic acid. We found that the strain produced suppressor mutants at a frequency of ~1×10−6 per viable d-glutamic acid-requiring colony-forming unit. This high frequency of suppression in PM2181 supports the idea that the F. novicida transposon mutant likely contains an extragenic suppressor.
We anticipate that these integration systems will be useful for studies requiring single-copy gene expression, such as the complementation of mutant genes when expression of the wild-type gene from multi-copy plasmids is toxic. Furthermore, these systems will be helpful where the use of multi-copy plasmids may not be suitable for cell culture or animal experiments. They may also be helpful for developing live vaccine strains containing additional antigens, where the use of antibiotic-resistance markers is undesirable.
This work was supported by the Molecular Pathogenesis of Bacteria and Viruses NIH grant T32 AI007362-18 to E.D.L., NIH grant AI068013 to M.S.P., and NIH grants AI058141 and AI065357 to H.P.S. We wish to thank the members of the Pavelka lab for reviewing this manuscript.