The Role of relA and spoT in Yersinia pestis KIM5+ Pathogenicity

The ppGpp molecule is part of a highly conserved regulatory system for mediating the growth response to various environmental conditions. This mechanism may represent a common strategy whereby pathogens such as Yersinia pestis, the causative agent of plague, regulate the virulence gene programs required for invasion, survival and persistence within host cells to match the capacity for growth. The products of the relA and spoT genes carry out ppGpp synthesis. To investigate the role of ppGpp on growth, protein synthesis, gene expression and virulence, we constructed a ΔrelA ΔspoT Y. pestis mutant. The mutant was no longer able to synthesize ppGpp in response to amino acid or carbon starvation, as expected. We also found that it exhibited several novel phenotypes, including a reduced growth rate and autoaggregation at 26°C. In addition, there was a reduction in the level of secretion of key virulence proteins and the mutant was>1,000-fold less virulent than its wild-type parent strain. Mice vaccinated subcutaneously (s.c.) with 2.5×104 CFU of the ΔrelA ΔspoT mutant developed high anti-Y. pestis serum IgG titers, were completely protected against s.c. challenge with 1.5×105 CFU of virulent Y. pestis and partially protected (60% survival) against pulmonary challenge with 2.0×104 CFU of virulent Y. pestis. Our results indicate that ppGpp represents an important virulence determinant in Y. pestis and the ΔrelA ΔspoT mutant strain is a promising vaccine candidate to provide protection against plague.


Introduction
Plague remains one of the most feared infectious diseases in humans. The etiological agent of the disease, Yersinia pestis, is disseminated by fleas and infects both humans and rodents. Y. pestis rapidly invades from the infection site into the lymphatic system and circulation, to produce the systemic and often fatal disease [1]. Globally about 2000 cases of plague are reported to the World Health Organization each year [2]. Most of these cases are the bubonic form of the disease, usually a consequence of the transmission of bacteria to humans via bites from fleas that have previously fed on infected rodents although contact with domestic cats that have been exposed to Y. pestis is another important transmission mode because of the higher than average incidence of pneumonic plague that occurs in these cases [1]. More rarely, cases of pneumonic plague are reported that are characterized by a short incubation period of 2 to 3 days and a high rate of mortality, even if treated. Pneumonic plague can be transmitted person to person or animal to person via the inhalation of contaminated air droplets [1]. Pneumonic plague is the most likely form to be encountered if Y. pestis is used as a biological weapon [3].
Y. pestis overwhelms its mammalian host during systemic growth by evading phagocytosis and by inhibiting the inflammatory response [4]. These properties are associated with a 70-kb plasmid, termed pCD1, which codes for a virulence-associated type III secretion system (T3SS) [5]. The analogous 70-kb pYV (Yersinia virulence) plasmid is found in Yersinia pseudotuberculosis and Yersinia enterocolitica which primarily cause gastrointestinal disease. In addition, Y. pestis has two plasmids that are absent in other yersiniae: the 100-kb plasmid pMT1 proposed to contribute to the survival of the bacteria in the flea [6], and the 9.5-kb plasmid pPCP1, which is responsible for the invasive character of plague in the mammalian host [7]. After subcutaneous (s.c.) administration of Y. pestis in mice, the pPCP1 plasmid potentiates the spread of bacteria into the circulation [7]. It was recently demonstrated that pPCP1 also enhances invasion of Y. pestis into human epithelial cells [8].
The stringent response is a generalized adaptive response to nutritional deprivation and environmental stress. The production of a specific nucleotide, guanosine-59,39-(bis)pyrophosphate (ppGpp), is the primary signaling and initiating event in the stringent response. In Escherichia coli and Salmonella enterica serovar Typhimurium, the relA gene encodes an enzyme with guanosine 39,59 bis(diphosphate) synthetase activity, which carries out ppGpp synthesis in response to a nutritional imbalance caused by amino acid starvation leading to increased synthesis of ppGpp and a reduction in stable RNA synthesis. This response to amino acid limitation is absent in DrelA strains [9]. Basal levels of ppGpp synthesized in the absence of relA activity during balanced growth is regulated by spoT, which encodes an enzyme with both guanosine 39, 59-bis(diphosphate) 39-pyrophohydrolase and guanosine 39,59 bis(diphosphate) synthetase activity [10]. The spoT gene modulates ppGpp levels in response to a number of nutritional factors including carbon starvation [10]. In Grampositive bacteria, a single RelA/SpoT protein is responsible for both functions [11]. The relA and spoT genes play an important role in the virulence of a number of pathogenic bacteria, including Mycobacterium tuberculosis [12], Listeria monocytogenes [13], Legionella pneumophila [14,15], Vibrio cholerae [16], and Pseudomonas aeruginosa [17]. A recent study indicated that a DrelA DspoT S. Typhimurium is effectively noninvasive for epithelial cells in vitro and is attenuated in BALB/c mice [18,19]. These genes were shown to play a crucial role in the regulation of genes in Salmonella pathogenicity islands 1 and 2 (SPI1, SPI2) and the spv virulence plasmid genes. Taken together, these results suggest that ppGpp may play a universal role in bacterial virulence gene expression.
The ppGpp molecule is part of a highly conserved regulatory system for mediating the growth response to various environmental conditions. This mechanism may represent a common strategy whereby facultative intracellular pathogens regulate the virulence gene programs required for invasion, survival and persistence within host cells to match the capacity for growth. However, the role of the ppGpp in Y. pestis physiology and virulence has not been investigated.
The goal of our study was to determine what role relA and spoT play in Y. pestis physiology and virulence by constructing DrelA and DrelA DspoT mutants and characterizing them for both in vitro and in vivo characteristics. We examined the effect of these mutations on transcription and protein levels at 26uC (flea temperature) and at 37uC (human temperature) and the effect on host colonization, immune responses and virulence. We also evaluated the double mutant for its capacity to induce protective immunity. Our results showed that the relA spoT mutant was attenuated for virulence and induced protective immunity by s.c. vaccination against bubonic and pneumonic plague.

Materials and Methods
Bacterial strains, culture conditions and plasmids All bacterial strains and plasmids used in this study are listed in Table 1. All strains were stored at 270uC in phosphate-buffered glycerol. Y. pestis cells were grown routinely at 28uC on Congo red agar from glycerol stocks and then grown in heart infusion broth (HIB) or on tryptose-blood agar base (TBA) [20]. The chemically defined medium PMH2 was used routinely [21]. All E. coli strains were grown routinely at 37uC in LB broth [22] or LB solidified with 1.2% Bacto Agar (Difco).

Plasmid construction
All primers used in this paper are listed in Table S1. The original source for the tightly regulated araC P BAD in pYA3700 was E. coli K-12 strain x289 [23]. For construction of the P BAD spoT insertion/deletion into lacZ, primer sets of LacZ1/LacZ2 and LacZ3/LacZ4 were used for amplifying lacZ-U (upstream gene sequence of lacZ), and lacZ-D (downstream gene sequence of lacZ) fragment, respectively. The lacZ-U and lacZ-D fragments were cloned into the SphI/PstI sites and SacI/EcoRI sites of pYA3700 to form pYA4573. The spoT gene fragment was amplified using SpoT-1 and SpoT-2 primers. The primer SpoT-1 containing the new SD sequence is shown Table S1. The spoT fragment was cloned into pYA4573 to construct pYA4574. Plasmid pYA4574 was digested with PstI, blunt ended with T4 DNA polymerase and dephosphorylated with Shrimp Alkaline phosphatase (Promega). The cat-sacB fragment was cut from pYA4373 using PstI and SacI restriction endonucleases and blunted by T4 DNA polymerase. Then, the cat-sacB fragment was ligated into PstI site of pYA4574 to form plasmid pYA4575.
To construct a spoT-3x-flag-kan fusion, a C-terminal spoT gene fragment (510 bp) was amplified using SpoTC-1 and SpoTC-2 primers and cloned into HindIII and BamHI sites of pUC18 to construct pYA4642. The spoU' gene fragment (sequence downstream of spoT) and lacZ-D gene fragment (sequence downstream of lacZ) were amplified from genomic DNA using SpoTD-1/ SpoTD-2 and LacZ3/LacZ4 primers, respectively. The spoU' and lacZ-D fragment were cloned into SacI and EcoRI sites of pYA4642 to form pYA4643 and pYA4644, respectively. Then the 36flag-kan gene fragment amplified from pYA4045 was cloned into SacI and BamHI sites of pYA4643 and pYA4644 to construct pYA4645 and pYA4646.
To delete the pla gene from plasmid pPCP1, plasmids pYA4647, pYA4648, and pYA4649 were constructed. The pla-U fragment was amplified from total DNA of Y. pestis KIM6 + using Pla1 and Pla2 primers and cloned into the EcoRI and PstI sites of pUC18 to form pYA4647. The pla-D fragment was amplified using Pla3 and Pla4 primers. The pla-D fragment was cloned into pYA4647 to construct pYA4648. The cat cassette (including Flp recombination site) amplified using Cm1 and Cm2 primers was cloned into the PstI site of pYA4648 to form pYA4649.

Construction of Y. pestis mutant strains
The construction of strains x10003 and x10004 using a two-step recombination method was previously described [24]. Strain x10019 was constructed from strain x10004 using similar methods. Briefly, plasmid pKD46 was introduced into x10004 by electroporation. A linear lacZ-U-cat-sacB-TT araC P BAD spoT-lacZ-D fragment was purified from plasmid pYA4575 by digestion with EcoRI and SphI and transformed into x10004 (pKD46) competent cells. Electroporants were isolated on TBA+Cm (10 mg/ml) plates. Integration of the lacZ-U-cat-sacB-TT araC P BAD spoT-lacZ-D fragment into the correct site of the chromosome was verified by PCR. Colonies with the correct PCR profile were streaked onto TBA+ Cm (10 mg/ml)+5% Sucrose plates to verify sucrose sensitivity and onto HIB Congo Red+Cm (10 mg/ ml) plates to confirm the presence of the pgm locus. To remove the cat-sac cassette from the chromosome, electrocompetent cells were prepared from a sucrose-sensitive isolate and electroporated with approximately 1 mg of a linear DNA (lacZ-U-TT araC) cut from pYA4574 using SphI and BamHI. Electroporants were selected on TBA+5% sucrose plates incubated at 30uC. Colonies were tested using PCR to validate that the cat-sacB cassette was eliminated. Plasmid pKD46 was cured from a single colony isolate of a sucrose-resistant, chloramphenicol-sensitive strain to yield x10019.
To construct strains expressing spoT tagged with the Flag epitope [25], plasmid pKD46 was introduced into Y. pestis KIM6 + and x10019. The resulting strains were electroporated with ,0.5 mg of spoTC-36flag-kan-spoU' and spoTC-36flag-kan-lacZ-D cut from pYA4645 and pYA4646, respectively. Electroporants were selected on TBA+Kan (20 mg/ml) plates at 37uC. The resulting colonies were verified using PCR to confirm that the 36flag-kan fragment was correctly inserted into the chromosome. Plasmid pKD46 was cured from single colony isolates of Y. pestis KIM5 + or x10019 derivatives to yield x10021 and x10022, respectively.
ppGpp assay ppGpp was detected using a slight modification of previously described procedures [27,28]. To starve cells for amino acids, strains were grown overnight in HIB medium at 26uC. The cells were then harvested and washed three times with PBS and resuspended to an OD 620 of 0.15 in 1 ml of modified PMH2 medium lacking L-phenylalanine [28]. The culture was shaken at 250 rpm at 26uC for approximately 5 h until the OD 620 reached 0.25, whereupon, [ 32 P] H 3 PO 4 was added to 100 mCi/ml. Cells were incubated for an additional 1 h at 26uC. Following incubation, an equal amount of chilled 90% formic acid was added to the cell suspension. The ice-cold suspensions were then  [21] pCP20 Ap r Cm r , FLP recombinase expression [26] pKD3 Ap r Cm r , cat cassette template [26] pKD46 Ap r , l Red recombinase expression [26] pYA3700 TT araC P BAD cassette plasmid, Ap r [85] pSUB11 Kn r , 3xFlag-tagged [25] pYA4373 The cat-sacB cassette in the PstI and SacI sites of pUC18. pUC18 pYA4573 The lacZ-U (upstream gene sequence of lacZ), and lacZ-D (downstream gene sequence of lacZ) fragment were cloned into the SphI/PstI sites and SacI/EcoRI sites of pYA3700 respectively. pYA3700 pYA4574 The spoT gene with new SD sequence was cloned into the XhoI and SacI sites of pYA4573. pYA4573 pYA4575 The cat-sacB cassette from pYA4373 was ligated into PstI site of pYA4574. pYA4574 pYA4642 The C-terminal spoT gene fragment (510 bp) was cloned into HindIII and BamHI sites of pUC18. pUC18 pYA4643 The spoU' gene fragment (downstream sequence of spoT) was cloned into SacI and EcoRI sites of pYA4642. pYA4642 pYA4644 The lacZ-D gene fragment (downstream sequence of lacZ) was cloned into SacI and EcoRI sites of pYA4642. pYA4642 pYA4645 The 36Flag::kan gene fragment was cloned into SacI and BamHI sites of pYA4643. pYA4643 pYA4646 The 36Flag::kan gene fragment was cloned into SacI and BamHI sites of pYA4644. pYA4644 pYA4647 The pla-U fragment (upstream sequence of pla) was cloned into the EcoRI and PstI sites of pUC18. pUC18 pYA4648 The pla-D fragment (downstream sequence of pla) was cloned into the SphI and PstI sites of pYA4647. pYA4647 pYA4649 The cat cassette (including Flp recombination site) was cloned into the PstI site of pYA4648. pYA4648 doi:10.1371/journal.pone.0006720.t001 rigorously vortexed followed by three freeze-thaw cycles. The acid extracts were centrifuged in a minifuge set at the highest speed for 5 min, and 5 ml of supernatant was then applied to a polyethyleneimine-cellulose thin-layer chromatography plate (TLC). The TLC plates were developed at room temperature with 1.5 M KH 2 PO 4 (pH 3.4). The developed plates were then air-dried and visualized by autoradiography using X-ray film at 270uC. To starve cells for carbon, strains were grown overnight in HIB medium. For strain x10019, two cultures were grown, one with and one without the addition of 0.05% arabinose. The cells were harvested, washed three times using PBS and resuspended to an OD 620 of 0.15 in 1 ml of modified PMH2 medium without glucose or arabinose. Cultures were grown, labeled and evaluated by TLC as described above.

Analysis of virulence factor transcription by RT-PCR
Total RNA was extracted from bacterial cells using TRIzol Reagent (Invitrogen) according to the manufacturer's recommendations. RNA samples were treated with DNase I for 10 min at 37uC to degrade contaminating DNA followed by inactivation of DNase I with 2 mM EDTA and heating to 65uC for 10 min. RNA was then precipitated with sodium acetate and ethanol and washed with 70% ethanol prior to performing RT-PCR. RNA samples of 200 ng were used for reverse transcription, using random hexamer primers and Superscript II reverse transcriptase as described by the manufacturer (Invitrogen). PCR amplification was performed using the lcrV yopB, yopD, yopE, yopH, yopJ, yopK, yopM, yopT or 16S rRNA primer pairs listed in Supplementary Table S1. RNA samples were used as templates in PCR reactions for RT minus controls. Twenty cycles of amplification were performed using an annealing temperature of 58uC. Products were then separated on a 1% agarose gel, stained with ethidium bromide and imaged for visualization of appropriately sized PCR products. In all cases, reactions were performed in triplicate.

Protein analysis
Secreted virulence factors were prepared by using a modification of previously described methods [29]. Y. pestis was grown in HIB medium overnight at 26uC. The cells were then harvested and washed three times using PMH2, inoculated to 40 ml of fresh PMH2 medium to an OD 600 of 0.05 and shaken overnight at 26uC. Cultures were shifted to 37uC for 6 h with shaking to provide mild aeration. Bacterial cells were removed by centrifugation for RNA extraction. Secreted virulence factors from the culture supernatants were concentrated by precipitation with 10% (w/v) trichloroacetic acid overnight at 4uC. Precipitated proteins were collected by centrifugation, washed with ice-cold acetone, and dissolved in 0.05 M Tris-HCl buffer (pH 9.5). Insoluble materials were removed by centrifugation at 12 500 g for 15 min and the protein concentration in the supernatant was determined using the DC protein assay kit (Bio-Rad Laboratories, Hercules, CA). Samples containing 200 mg proteins were heated at 95uC for 5 min in protein sample buffer containing 2-mercaptoethanol and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) with 10% polyacrylamide. Proteins were transferred to nitrocellulose membranes. The membranes were blocked with 5% skim milk in PBS, incubated with rabbit polyclonal antibodies specific for the indicated Yop proteins or LcrV, and washed with PBS-Tween 20. Then alkaline phosphatase-conjugated goat anti-rabbit immunoglobulin G (IgG) (Sigma, St. Louis, MO) was added in PBS-Milk. Immunoreactive bands were detected by the addition of NBT/BCIP (Sigma, St. Louis, MO). The reaction was stopped after 5 min by washing with several large volumes of deionized water.

Two-dimensional gel electrophoresis
Comparison of two dimensional protein profiles was carried out as previously described [30]. Y. pestis KIM5 + and x10004(pCD1Ap) were grown at 26uC or 37uC in 5 ml of bestcase-scenario (BCS) medium without Ca 2+ . The cultures were harvested by centrifugation and washed once with low salt PBS (0.16). Cells were resuspended in 1 ml lysis buffer containing 8M Urea, 0.05M DTT, 2% (w/v) CHAPS and 0.2% (w/v) ampholytes. Proteins were extracted by vortexing 1 ml cell samples in lysis buffer with 0.2 mm glass beads ten times for 30 s with cooling between vortexing. The samples were centrifuged at 2500 g for 5 min to remove the beads. The bead-free supernatant was centrifuged at 15000 g for 15 min at 4uC to remove cellular debris. The cell-free lysates were immediately placed on ice and protease inhibitor was added. The lysates were retreated with a 2D protein cleanup kit (Bio-Rad, Hercules, CA) and protein concentration was determined using the Bio-Rad Protein Assay kit.
Protein lysates (300 mg) were mixed with rehydration buffer (Bio-Rad) in a total volume of 300 mL. Equal amounts (300 mg) of protein were isoelectrically focused using 17 cm pH 4-7 strips followed by 18.3619.3 cm 8-16% SDS-PAGE using Midi-Protean II 2D cell (Bio-Rad). Gels were stained with Coomassie Brilliant Blue R-250 (Bio-Rad) and visualized using Gel Doc XR system (Bio-Rad). Protein expression levels from protein spots on gels were compared between the different samples. Gel analysis was performed using the PDQuest3 2-D Analysis Software (Bio-Rad) to determine differential expression. Differentially expressed protein spots were excised and were digested with In-Gel Tryptic Digestion Kit (Pierce, Rockford, IL). Peptide digests were analyzed using a Voyager DE STR MALDI-TOF mass spectrometer (Applied Biosystems, Framingham, MA). Data were searched in bacterial proteomics database using Aldente in ExPASy Proteomics Server. This experiment was performed four times with similar results.

Virulence studies in mice
Single colonies of each strain were used to inoculate HIB cultures and grown overnight at 26uC. To select for plasmid pCD1Ap, ampicillin was added into the medium at a concentration of 25 mg/ml. Bacteria were diluted into 10 ml of fresh HIB enriched with 0.2% xylose and 2.5 mM CaCl 2 to obtain an OD 620 of 0.1 and incubated at 26uC for s.c. infections (bubonic plague) or at 37uC for intranasal (i.n.) infections (pneumonic plague). Both cultures were grown to an OD 620 of 0.6. The cells were then harvested and the pellet resuspended in 1 ml of isotonic PBS.
All animal procedures were approved by the Arizona State University Animal Care and Use Committee. Female 7-week-old Swiss Webster mice from Charles River Laboratories were inoculated by s.c. injection with 100 ml of bacterial suspension. Actual numbers of colony-forming units (CFU) inoculated were determined by plating serial dilutions onto TBA agar. To determine 50% lethal dose (LD 50 ), five groups of six mice were infected with serial dilutions of the bacterial suspension. For in vivo complementation of strain of x10019(pCD1Ap), 120 mg of Larabinose dissolved in PBS was intraperitoneally administered to mice on the day of inoculation and once a day thereafter [31]. Mice were monitored twice daily for 21 days, and the LD 50 was calculated as described [32].
For colonization/dissemination analysis, 3 mice per time point were infected by s.c. injection in the front of the neck. At the indicated times after infection, mice were euthanized, and samples of blood, lungs, spleen and liver were removed. The bacterial load for each organ was determined by plating dilutions of the homogenized tissues onto TBA with ampicillin plates and reported as CFU per gram of tissue or CFU per ml blood. Infections were repeated in at least two independent experiments.

Preparation of bacterial antigens
Bacterial antigens used for ELISA were prepared from fresh cells. Briefly, single colonies of Y. pestis KIM5 + were inoculated into HIB media and cultured overnight at 26uC. Cells were switched to 37uC for 6 h. Bacterial cultures were centrifuged at 5,0006g for 10 min, the pellet was washed once with sterile PBS and resuspended in sterile PBS. Bacterial cells were broken using 0.2 mm glass beads 10 times for 60 s with cooling between vortexing (with 2 min incubation on ice between cycles). The whole bacterial lysate was sterilized by UV light and sterility was confirmed by TBA agar culture. The lysate was frozen at 280uC until use. Protein content was determined by BCA analysis per manufacturer's instructions (Sigma).

Enzyme-linked Immunosorbent Assay (ELISA)
Mice were lightly anesthetized using ketamine and xylazine mixture administered intramuscularly. Blood was collected by retro-orbital sinus puncture for the determination of antibody titers at different time points. ELISA was used to assay serum antibodies against the whole cell lysate of Y. pestis KIM5 + . Sera were tested for IgG at a starting dilution of 1:1000, and for IgG1 and IgG2a at 1:100, respectively.
Polystyrene 96-well flat-bottom microtiter plates (Dynatech Laboratories Inc., Chantilly, VA) were coated with 200 ng/well of Y. pestis whole cell lysates. Antigens suspended in sodium carbonate-bicarbonate coating buffer (pH 9.6) were applied in 100 ml volumes to each well. The coated plates were incubated overnight at 4uC. Free binding sites were blocked with a blocking buffer (phosphate-buffered saline [PBS; pH 7.4], 0.1% Tween 20, and 1% bovine serum albumin). A 100 ml volume of serially diluted sample was added to individual wells in triplicate and incubated for 1 h at 37uC. Plates were treated with biotinylated goat anti-mouse IgG, IgG1, or IgG2a (Southern Biotechnology Inc., Birmingham, AL). Wells were developed with streptavidinhorseradish peroxidase conjugate (Invitrogen, Carlsbad, CA), followed by 2,2-azino-bis-(3-ethylbenzthiazoline-6-sulfonic acid) (ABTS) (Sigma) in sodium citrate buffer containing 0.03% hydrogen peroxide (H 2 O 2 ). After a 10 min incubation at 37uC in the dark, color development (absorbance) was recorded at 405 nm using an automated ELISA plate reader (model EL311SX; Biotek, Winooski, VT). Absorbance readings that were 0.1 higher than PBS control values were considered positive.

In vivo cytokine analysis
Cytokines were quantitated by a double-sandwich enzymelinked immunosorbent assay (ELISA) as described previously [33]. Mice in groups of three were euthanized at intervals by terminal bleeding under anesthesia. Pooled blood was allowed to clot overnight at 4uC, and serum was separated by centrifugation at 10,000 g for 10 min. Sera were filtered once through a 0.22 mm syringe filter, cultured on TBA to confirm that bacteria had been removed and stored at 270uC prior to assay for cytokines.
Commercial solid-phase enzyme immunoassays utilizing the multiple-antibody sandwich principal were used to determine cytokines in biological samples. In these experiments, IL-10, TNFa and IFN-c were determined with Mouse IL-10, IFN-c and TNF-a Ready-SET-Go kits (ebioscience), respectively. Concentrations of cytokines were measured by reading optical density at 450 nm and then calculated in reference to values obtained in standard curves generated for each assay. Assays of pooled sera were repeated three times.

Protective efficacy
Two groups of Swiss Webster mice (10 mice/group) were immunized by s.c. injection with 2.5610 4 CFU of x10004(pCD1Ap) cells in 100 ml of isotopic PBS on day 0. Two groups of mice (4 mice/group) were injected with 100 ml of PBS as controls. On day 35, animals were challenged by s.c. injection with 100 ml of virulent Y. pestis KIM5 + or lightly anesthetized with a 1:5 xylazine/ketamine mixture and challenged by the intranasal route with 20 ml of bacterial suspension. The challenge dose for s.c. injection was 1610 5 CFU and for i.n. challenge was 2.0610 4 CFU. Protective efficacy was determined by the number of surviving animals. All infected animals were observed over a 15day period for the development of signs of plague infection.

Statistical analysis
Data are expressed as means6SE. One-way analysis of variance with Student t-test were used for statistical analysis. A P-value of,0.05 was considered significant.

Sequence analysis of the relA and spoT genes
Analysis of the Y. pestis KIM5 + database revealed the presence of relA and spoT genes homologous to E. coli K-12 and S. Typhimurium LT-2 [34,35,36]. The Y. pestis RelA protein shares 84.7% identity with E. coli K-12 and 83.9% identity with S. Typhimurium LT-2 RelA proteins. The Y. pestis SpoT protein has 91.3% identity with E. coli K-12 and 91.8% identity with S. Typhimurium LT-2 SpoT proteins.
Our analysis indicated that Y. pestis SpoT, but not RelA, possesses the HD domain that is conserved in a superfamily of metal-dependent phosphohydrolases [37]. Histidine (H) and aspartate (D) residues in the HD domain are thought to be involved in (p)ppGpp degradation [37]. Both Y. pestis RelA and SpoT proteins possess the conserved ATP/GTP-binding and GTP binding domains, TGS [38] and ACT [39,40], respectively, that are present in the E. coli RelA and SpoT proteins [41]. The presence of these conserved motifs in the Y. pestis proteins is in agreement with their biochemical functions because ATP and GTP are substrates of the reaction catalyzed by (p)ppGpp synthetase.
The relA and spoT genes are involved in synthesis of ppGpp and physiological differentiation To evaluate the linkage between relA and spoT and the production of ppGpp, we constructed DrelA, DrelA DspoT [24] and DrelA DspoT DlacZ::TT araC P BAD spoT mutants of Y. pestis KIM6 + strain (Fig. 1). To construct a strain with arabinoseregulated spoT expression, a TT araC P BAD promoter cassette was inserted in front of the spoT gene. The spoT gene is located in the middle of an operon. To avoid affecting the transcription of nearby genes, the TT araC P BAD spoT construct was inserted at another location, lacZ (Fig. 1).
Because of the high degree of similarity between Y. pestis RelA and SpoT proteins and their E. coli and Salmonella counterparts, it is likely that the function of RelA and SpoT in Y. pestis will be the same. To evaluate the effect of relA and spoT on ppGpp synthesis during amino acid starvation, Y. pestis was grown in PHM2 media [21] without L-phenylalanine. ppGpp accumulation was observed in wild-type Y. pestis, but not in the relA null strains (Figure 2A), illustrating that Y. pestis is indeed capable of ppGpp biosynthesis in response to amino acid starvation. We also evaluated the effect of carbon starvation. When glucose was exhausted in the medium, ppGpp accumulated in the wild type and DrelA spoT + strains, but not in DrelA DspoT strains ( Figure 2B). These results indicate that Y. pestis has a RelA-dependent response to amino acid starvation and a SpoT-dependent response to glucose starvation, comparable to what is observed in E. coli [42]. The SpoT deficiency could be complemented in strain x10019 (DrelA233 DspoT85 DlacZ516::TT araC P BAD spoT) by the addition of arabinose. Synthesis of SpoT in strain x10019 in the presence of 0.05% arabinose was nearly identical to wild-type SpoT synthesis ( Fig. S1 and S2). The addition of arabinose to strain x10019 also restored ppGpp synthesis when cells were starved for carbon (Fig. 2B).
A cursory examination of the Y. pestis DrelA DspoT double mutant after growth on solid rich medium indicated that the DrelA DspoT double mutants grew more slowly than wild-type or DrelA mutants. When growth was assessed in liquid medium, the DrelA DspoT mutants exhibited a longer lag phase and did not reach as high a final OD 600 than the wild-type and DrelA mutant strains at both 26uC and 37uC ( Fig. 3A and B). The DrelA DspoT strains were prone to autoaggregate and precipitate to the bottom of the culture tube at 26uC, but not at 37uC (data not shown). The addition of 0.05% arabinose restored wild-type growth characteristics to strain x10019 (DrelA233 DspoT85 DlacZ516::TT araC P BAD spoT) (Fig. 3), but it continued to autoaggregate and precipitate at 26uC. However, the addition of higher concentrations of arabinose reduced autoaggregation in a concentration-dependent manner. The addition of 0.4% arabinose resulted in the complete absence of detectable autoaggregation at 26uC.

The effect of ppGpp on production of virulence factors of Y. pestis
The virulence of the pathogenic Yersinia species depends on a plasmid-encoded type III secretion system (T3SS) that transfers effector proteins called Yops (Yersinia outer proteins) into host cells, interfering with mammalian cell signaling pathways, inhibiting phagocytosis, modulating cytokine production, and inducing apoptosis [43]. In S. Typhimurium, pathogenicity islands 1 and 2 (SPI1 and SPI2) encode T3SSs required for invasion and replication within host cells, respectively [44]. SPI1 and SPI2 gene transcription and expression are severely reduced in the absence of ppGpp [45]. To determine if ppGpp had a similar effect on Y. pestis, transcription of the genes encoding T3SS substrates LcrV and Yop proteins was analyzed using RT-PCR. Our results indicated that relA or relA spoT status did not have a significant effect on the transcription of lcrV and or the yop genes (Fig. 4A).
To examine the effect of ppGpp on protein synthesis, the proteome of wild-type and DrelA DspoT mutant Y. pestis strains was compared at different temperatures using two-dimensional electrophoresis (Fig. S3). Our results indicate that deletion of relA and spoT led to reduced synthesis of some metabolic enzymes at flea (26uC) and human (37uC) temperatures, and also reduced synthesis of virulence factors such as Pla, LcrH and LcrV at 37uC ( Table 2 and Table 3).
The results indicate that LcrV and YopM secretion was reduced slightly in absence of ppGpp (DrelA DspoT), but secretion of YopH, YopD and YopE were significantly decreased (Fig. 4B).

A DrelA DspoT mutant is attenuated in mice
To investigate the contribution of ppGpp to the virulence of Y. pestis, we infected groups of three Swiss Webster mice subcutaneously with wild-type, x10003(pCD1Ap) (DrelA233), x10004(pCD1Ap) (DrelA233 DspoT85) and x10019(pCD1Ap) (DrelA233 DspoT85 DlacZ516::TT araC P BAD spoT), in which spoT expression is regulated by arabinose availability. Strain x10019(pCD1Ap) was grown in the presence of arabinose prior to inoculation of mice. Once this strain colonizes host tissues where there is no free arabinose [23], it will become phenotypically SpoT 2 . In preliminary experiments we determined that the LD 50 of the wild-type strain in mice is,10 CFU, consistent with previous results [47,48]. Mice given wild-type Y. pestis KIM5 + and x10003(pCD1Ap) (DrelA) succumbed to the infection in a highly synchronous manner (Fig. 5). Only 50% of the mice infected with 5.8610 5 CFU of DrelA DspoT strain x10004 developed plague after 6 days, and the rate at which the mice died was slower than the rate of those infected with the wild-type strain. The LD 50 of x10004(pCD1Ap) was 5.8610 5 CFU. Thus, the lack of ppGpp resulted in a ,100,000-fold increase in the LD 50 obtained by subcutaneous (s.c.) infection. The LD 50 of x10019(pCD1Ap) strain, administered after growth in arabinose was intermediate, at 3.3610 2 CFU (,100-fold increase). The LD 50 of x10019 (pCD1Ap) was the same as KIM5 + (LD 50 ,10) when inoculated mice were injected with arabinose, indicating full complementation of the attenuation phenotype.
To further evaluate the ability of Y. pestis to disseminate to the bloodstream and internal organs, we monitored the growth of both Y. pestis KIM5 + and x10004(pCD1Ap) in the lungs, spleens, livers and blood of infected mice over a 7-day period after s.c. injection. Because of the difference in LD 50 between the two strains, we inoculated mice with different doses of each, 1.5610 3 CFU of Y. pestis KIM5 + or 1.6610 6 CFU of x10004(pCD1Ap). The kinetics of colonization was similar for both strains (Fig. 6). Despite the difference in dose, the levels of bacteria in blood, spleen and liver were similar for both strains on days 3 and 5. There was an approximate 1.5 log difference in bacteria isolated from lung tissue, indicating that the DrelA DspoT mutant had was less efficient than KIM5 + at reaching the lungs. By day 7, the number of the DrelA DspoT mutant began to decline in all tissues, indicating clearance by the host, while all of the mice inoculated with wildtype Y. pestis had succumbed to the infection.
The immune responses to DrelA DspoT Y. pestis strain x10004(pCD1Ap) Because x10004 was attenuated, we explored its potential as a vaccine. To evaluate the immune responses to DrelA DspoT Y. pestis strain x10004(pCD1Ap), two groups of 10 mice each were immunized s.c. with 2.5610 4 CFU on day 0. Two groups of 4 mice each were injected with PBS as controls. Mice were challenged on day 35 with either 1.5610 5 (s.c.) or 2.0610 4 (i.n.) CFU of Y. pestis KIM5 + . Blood was taken at 2 and 4 weeks post immunization and 2 weeks after challenge. Serum IgG responses to Y. pestis whole cell lysates (YpL) from immunized mice were measured by ELISA (Fig. 7A). At two weeks after immunization, the reciprocal anti-Y. pestis serum IgG titers were greater than 1,000 and increased at 4 weeks and after challenge.
The serum immune responses to YpL were further examined by measuring the levels of IgG isotype subclasses IgG1 and IgG2a. Th1 cells direct cell-mediated immunity and promote class switching to IgG2a, and Th2 cells provide potent help for B-cell antibody production and promote class switching to IgG1 [49]. The level of anti-YpL IgG1 and IgG2a isotype antibodies rapidly increased after vaccination and gradually increased at 2 weeks, 4 weeks and post-challenge (Fig. 7B). At 2 and 4 weeks postimmunization, the ratio of IgG1 to IgG2a was 1.06:1 and 1.2:1 respectively, indicating an initial mixed Th1/Th2 response, which developed into a slight Th2 bias by week 4. This Th2 bias continued after challenge as well.
Immunization with a DrelA DspoT Y. pestis strain x10004(pCD1Ap) can protect against plague challenge To evaluate the protective efficacy of DrelA DspoT Y. pestis strain x10004(pCD1Ap) against the bubonic and pneumonic forms of plague, immunized mice were challenged on day 35 with either 1.5610 5 (s.c.) or 2.0610 4 (i.n.) CFU of Y. pestis KIM5 + . Postchallenge survival was monitored for 14 days. A single s.c. vaccination could provide complete protection against s.c. challenge without any symptoms (Fig. 8A) and 60% protection against pulmonary challenge (Fig. 8B). None of the mice immunized with PBS survived either challenge (Fig. 8).
Induction of cytokines by Y. pestis KIM5 + and DrelA DspoT strain x10004(pCD1Ap) Cytokines are critical to the development and functioning of both the innate and adaptive immune responses. They are often secreted by immune cells that have encountered pathogens, thereby activating and recruiting additional immune cells to increase the system's response to the pathogen. Previously, LcrV has been demonstrated to be an immunomodulator (TNF-a and IFN-c down-regulation and IL-10 induction) both in vivo and in vitro [50,51,52]. Since the synthesis and secretion of LcrV is reduced in the DrelA DspoT mutant, we compared production of three cytokines (IL-10, INF-c and TNF-a) in mice infected with Y. pestis KIM5 + and x10004(pCD1Ap). For this experiment, groups of three Swiss-Webster mice were inoculated via the s.c. route with 1.5610 3 CFU of Y. pestis KIM5 + or 1.6610 6 CFU of x10004(pCD1Ap). A group of uninfected mice served as controls. Blood was collected via cardiac puncture 3 and 5 days later for cytokine analysis. Measurements indicated that levels of IL-10 were higher in the sera of animals infected with Y. pestis KIM5 + than that of x10004(pCD1Ap) (Fig. 9). The pro-inflammatory cytokines IFN-c and TNF-a were not detected in sera from mice inoculated with either strain (data not shown).

Discussion
The bacterial global signal molecule, ppGpp, plays a major role in translating the nutritional state of bacteria into appropriate expression of virulence genes. Our studies indicated that relA and spoT deletion mutants of Y. pestis and S. Typhimurium have similar, but not identical, effects on growth and virulence. A DrelA deletion does not affect the growth of either Y. pestis (Fig. 3) or S. Typhimurium (data not shown). In addition, as is the case for S. Typhimurium, the Y. pestis DrelA DspoT mutants fail to reach the same final cell density as their wild-type parents (Fig. 3) [18]. However, the DrelA DspoT Y. pestis strain showed a slight growth defect in vitro, a phenomenon not reported for S. Typhimurium [18].
The DrelA DspoT mutant was prone to autoaggregate and precipitate at 26uC. This phenotype was not apparent at 37uC. This phenomenon was never reported in DrelA DspoT mutants of Salmonella and E. coli. However, we could find no published experiments where the mutants had been grown at 26uC or a similar low temperature. To investigate whether this phenotype could be observed in S. Typhimurium, we cultured wild type, DrelA, and DrelA DspoT S. Typhimurium UK-1 at 26uC and 37uC. We did not observe autoaggregation or precipitation at 26uC (data not shown), indicating that this phenotype is specific for Y. pestis. In addition, autoaggregation was not alleviated in strain x10019 (DrelA DspoT araC P BAD spoT) by addition of 0.05% arabinose, a concentration that allowed complementation of other phenotypes relating to ppGpp formation (Fig. 2), growth (Fig. 3) and Yop and LcrV secretion (Fig. 4B). Autoaggregation was, however, eliminated by the addition of 0.4% arabinose, indicating that a higher level of spoT expression was required to alleviate this phenotype.
DlacZ::TT araC P BAD spoT) without arabinose; ¤ , x10019(pCD1Ap) (DrelA233 DspoT85 DlacZ::TT araC P BAD spoT) with 0.05% arabinose.   ppGpp regulates expression of many genes involved in the virulence and pathogenesis of S. Typhimurium including SPI-1 and SPI-2 encoded genes essential for invasion and the spv operon, involved in systemic infections [44,45]. In Y. pestis, the 70-kb virulence plasmid, pCD1, encodes a complex virulence property called the low-Ca 2+ response (LCR) [53,54]. The LCR was first observed in vitro, where the bacteria respond to the absence of Ca 2+ at 37uC by the strong expression and secretion of Yops and LcrV [4]. This is accompanied by growth restriction, in which the yersiniae undergo an orderly metabolic shutdown and cease growth [55,56]. The phenomenon is phenotypically similar to the stringent response which is induced by ppGpp synthesized when cells encounter amino acid or carbon starvation [10]. However, early research indicated that Y. pestis, although capable of ppGpp biosynthesis, did not utilize this pathway to mediate its temperature-dependent response to Ca 2+ -deficient environments. Early lesions in RNA synthesis associated with the growth restriction of Ca 2+ -deprived Y. pestis reflected a block in stable RNA synthesis and this effect is not mediated by ppGpp [28]. Yops accumulate in the membrane fraction after 3 h of induction, whereas transcription of the yop genes during Ca 2+ -deficiency is dramatically reduced [57]. Our results showed that the DrelA DspoT mutations did not affect yop transcription, but resulted in reduced synthesis and secretion of LcrV and a number of Yops, including YopD. YopD is essential for several discrete steps during efficient Yop effector translocation [58], therefore, the reduction of YopD synthesis might directly impair the secretion of other Yops.
A number of studies report the involvement of ppGpp in processes related to growth, stress, starvation, and survival that affect pathogenicity. A frequent scenario is that in relA spoT mutants, pathogenicity is compromised for reasons that vary with the organism studied and include pathogen/host interactions, invasiveness and persistence [59]. Our observations indicate that the relA-dependent accumulation of ppGpp does not play a major role in Y. pestis pathogenicity. The DrelA Y. pestis mutant was as virulent as wild type Y. pestis KIM5 + (Fig. 6), while the DrelA DspoT Y. pestis mutant was dramatically attenuated. The growth defect observed in vitro for the DrelA DspoT strain may have contributed to this reduction in virulence. Zusman et al showed that an L. pneumophila relA mutant is capable of wild-type intracellular proliferation in both human macrophage and the protozoan host Acanthamoeba castellanii [15]. A S. Typhimurium DrelA mutant also showed no significant differences in invasion, intracellular growth, virulence, or expression of a number of SPI-1 genes [18]. Unlike S. Typhimurium, wild-type Y. pestis and L. pneumophila are not capable of growth on defined medium unless supplemented with amino acids that also serve as a carbon source for these pathogens [60,61,62]. The pattern of amino acid requirements of Y. pestis is similar to that of the Legionella [60] and S. Typhimurium DrelA DspoT strains [63], including two branched chain amino acids, phenylalanine and threonine. But amino acid auxotrophs designed to duplicate the amino acid requirements of DrelA DspoT mutants do not affect intracellular growth and virulence of Legionella [64,65]    and Salmonella [18]. Therefore, we conclude that it is the basal (SpoT-dependent) ppGpp levels that affect pathogenicity of Y. pestis and not the stringent response induced by amino acid starvation.
The increased LD 50 and lung colonization deficiency of the ppGpp null mutant indicates that ppGpp may regulate genes important in establishing a lethal infection during bubonic plague. Only a modest degree of attenuation was observed in the pneumonic model (unpublished data). The pneumonic model bypasses the need to reach the lungs, indicating that it is the ability to reach the lungs, and not the ability to colonize the lungs, per se, that is at least partially responsible for the attenuation phenotype. This suggests that relA and spoT are required for expression of genes that facilitate lung invasion after subcutaneous entry into the host. Furthermore, this observation highlights the point that mutant strains are not universally attenuated for virulence, and route of infection is an important consideration when investigating the phenotype of a mutant.
Results of proteome analysis indicated that the absence of ppGpp in Y. pestis reduced Pla and LcrH synthesis at 37uC (Table 3). This may be another important factor leading to attenuation. Pla is a member of the omptin surface protease family and cleaves host plasminogen and components of the complement pathway [66]. Independent of this protease activity, Pla binds to the extracellular matrix component laminin and promotes invasion of endothelial cells [67]. Inactivation of pla severely attenuates Y. pestis during bubonic infection [68,69]; however, a pla mutant is still lethal during intranasal or intravascular infection [70]. In addition, the YopE and YopH type III effector proteins disrupt the host cell cytoskeleton allowing the bacterium to resist phagocytosis [71,72] and are required for Y. pestis virulence [20,73]. Therefore, the reduction in YopE and YopH synthesis observed in the DrelA DspoT mutant may also contribute to attenuation.
High titers of anti-Y. pestis serum IgG were produced by vaccination with the DrelA DspoT Y. pestis mutant and antibodybased humoral immunity provides good protection against bubonic plague [74,75]. Although a mixed Th1/Th2-type immune response was initially induced by this attenuated strain, the response became slightly biased to a Th2-type humoral immune response. Y. pestis secretes LcrV and Yops during infection. LcrV triggers the release of IL-10 by host immune cells and suppresses proinflammatory cytokines such as TNF-a and INF-c as well as innate defense mechanisms required to combat the pathogenesis of plague [76,77]. Our results were similar to previous findings. We were unable to detect TNF-a and IFN-c in sera of mice inoculated with the wild-type or the ppGpp null mutant, but the IL-10 levels induced by the ppGpp null mutant were significantly lower than that of wild type strain (Fig. 7). The reduction in LcrV expression in the ppGpp null mutant could be responsible for the decrease in IL-10 production. However, the reduced expression of LcrV and Yops was not sufficient to permit induction of the proinflammatory cytokines TNF-a and IFN-c.
Subcutaneous immunization with DrelA DspoT strain x10004(pCD1Ap) protected mice against s.c. challenge (simulating bubonic plague) and i.n. challenge (simulating pneumonic plague), but protection against i.n. challenge was not complete (Fig. 8) in spite of high serum titers of anti-Y. pestis IgG (Fig. 7). This result is consistent with what is known about the immunogenicity requirements for protection against Y. pestis. Protection against were challenged via the i.n. route with 2610 4 CFU of Y. pestis KIM5 + . Immunization provided significant protection against both challenge routes (P,0.001). For each experiment, there were 10 mice in the vaccinated group and 4 mice in the control group. doi:10.1371/journal.pone.0006720.g008 Figure 9. IL-10 production in sera of mice inoculated with Y. pestis KIM5 + or x10004(pCD1Ap). *, the P value was less than 0.01; **, the P value was less than 0.05. doi:10.1371/journal.pone.0006720.g009 bubonic plague is primarily antibody-mediated [74,75], while protection against pneumonic plague may require cell mediated immunity induced by INF-c and TNF-a [78,79]. This view is supported by vaccine trials with nonhuman primates indicating that humoral immunity may not protect humans against pulmonary Y. pestis infection. Specifically, studies by researchers at the USAMRIID found that a significant number of nonhuman primates immunized with the F1-V fusion protein vaccine succumbed to aerosol Y. pestis challenge, despite high-titer serum antibodies against F1-V at the time of challenge [80,81,82]. Therefore, antibodies alone are not sufficient to protect against pneumonic plague. Recent studies demonstrated that cytokinemediated immunity could augment protection against lethal pulmonary Y. pestis challenge in the presence or absence of specific antibody [79,83]. Therefore, absence of TNF-a and IFN-c in sera may be the major reason we observed poor protection efficacy against pneumonic plague.
Based on these results, an effective plague vaccine needs to prime not only humoral immunity but also strong Th1-type cellular immunity [79,84]. To induce effective cellular immunity we will further modify our DrelA DspoT strain to permit induction of cellular immune responses.