The NlpD Lipoprotein Is a Novel Yersinia pestis Virulence Factor Essential for the Development of Plague

Yersinia pestis is the causative agent of plague. Previously we have isolated an attenuated Y. pestis transposon insertion mutant in which the pcm gene was disrupted. In the present study, we investigated the expression and the role of pcm locus genes in Y. pestis pathogenesis using a set of isogenic surE, pcm, nlpD and rpoS mutants of the fully virulent Kimberley53 strain. We show that in Y. pestis, nlpD expression is controlled from elements residing within the upstream genes surE and pcm. The NlpD lipoprotein is the only factor encoded from the pcm locus that is essential for Y. pestis virulence. A chromosomal deletion of the nlpD gene sequence resulted in a drastic reduction in virulence to an LD50 of at least 107 cfu for subcutaneous and airway routes of infection. The mutant was unable to colonize mouse organs following infection. The filamented morphology of the nlpD mutant indicates that NlpD is involved in cell separation; however, deletion of nlpD did not affect in vitro growth rate. Trans-complementation experiments with the Y. pestis nlpD gene restored virulence and all other phenotypic defects. Finally, we demonstrated that subcutaneous administration of the nlpD mutant could protect animals against bubonic and primary pneumonic plague. Taken together, these results demonstrate that Y. pestis NlpD is a novel virulence factor essential for the development of bubonic and pneumonic plague. Further, the nlpD mutant is superior to the EV76 prototype live vaccine strain in immunogenicity and in conferring effective protective immunity. Thus it could serve as a basis for a very potent live vaccine against bubonic and pneumonic plague.


Introduction
Yersinia pestis is the etiological agent of plague, which has caused millions of deaths in three world pandemics and is still a public health issue in some regions of the world. The most prevalent form of the disease is the bubonic plague, which develops following transmission of the pathogen from rodent reservoirs to humans via infected fleas [1]. Primary pneumonic plague is less abundant in nature and results from inhalation of Y. pestis droplets or aerosols. It is a rapidly progressing disease leading to high mortality rates in untreated patients and can spread from person to person [1]. These characteristics led to the recognition of Y. pestis as a potential threat agent [2].
The ability of Y. pestis to respond to the host environment and to overcome immune systems is attributed to the combined activity of multiple virulence mechanisms. Among these mechanisms, only few have been found to be absolutely required for virulence in animal model systems. The type III secretion system (TTSS) is essential for survival of the pathogen within the mammalian host environment. This was demonstrated by the inability of Y. pestis strains devoid of the plasmid carrying the TTSS (pCD1 2 ) genes to colonize host tissues and to produce systemic disease following infection via both subcutaneous (s.c.) and airway routes [3][4][5][6]. The TTSS, shared by the closely related enteropathogens Y. enterocolitica and Y. pseudotuberculosis, comprises a secretion apparatus, chaper-ones and several effectors (Yops) and leads to the modulation of cell signaling networks necessary for an effective immune response [7,8]. Other virulence factors were found to be indispensable for Y. pestis pathogenesis via the s.c. route of infection. These include the plasminogen activator factor, which is encoded on the pPCP1 plasmid [9][10][11]; the Yersiniabactin (Ybt) iron acquisition system, which is encoded within the high pathogenicity island [12,13]; the chromosomally encoded PurH involved in the synthesis of purines [4,14]; adenylate kinase, which is involved in nucleotide metabolism [15]; and the recently characterized YadBC [16].
In a previous study, we isolated a highly attenuated mutant designated Kimberley53pcm (Kim53-K9) in which the pcm gene was disrupted by a mini-Tn5 transposon insertion [4]. When administered subcutaneously to mice, colonization and persistence of Kim53-K9 in the spleen and liver were substantially reduced and the median lethal dose (LD 50 ) of the mutant was more than seven orders of magnitude higher than the LD 50 of the wild type Kimberley53 strain [4]. The homologous E. coli pcm gene codes for a Protein-L-isoaspartate O-methyltransferase (Pcm), which can specifically catalyze the transfer of the methyl groups from Sadenosylmethionine to atypical L-isoaspartyl residues in proteins [17,18]. The protein was found to be involved in the survival of stationary phase cells when exposed to environmental stress conditions [18]. Immediately upstream of the E. coli pcm gene lies the surE gene (Fig. 1), which overlaps with pcm by four nucleotides.
In E. coli, these two genes share a bicistronic operon, but pcm can be transcribed independently from its own promoter [19]. The nlpD gene is located downstream of pcm and codes for an outer membrane lipoprotein that is assumed to be involved in cell wall formation and maintenance [20,21]. The last gene in the locus is rpoS, which encodes an alternative RNA polymerase Sigma factor (RpoS). The rpoS and the nlpD genes constitute an operon; however, the major rpoS promoter is located within the nlpD gene [22][23][24][25]. The expression of rpoS is induced during stress conditions, such as starvation and extreme pH, and during stationary growth phase [26,27]. This factor has been found to be involved in S. typhimurium virulence in a mouse infection model [28].
Genes within the pcm ''stress locus'' were extensively studied in many species of the Enterobacteriaceae family, including Yersinia enterocolitica [29,30]. However, their importance for the pathogenesis of plague has not been evaluated so far. In the present study, we have characterized the pcm genomic locus in Y. pestis. The expression pattern of the genes was evaluated along with their respective contributions to Y. pestis virulence using mouse models of bubonic and pneumonic plague. Our findings indicate that within the pcm locus, nlpD is the only essential gene for Y. pestis virulence. In addition, a highly attenuated Y. pestis nlpD-null mutant has been shown to induce an efficient protective immunity against bubonic and pneumonic plague.

Results
The pcm locus of Y. pestis In a previous screen designed to isolate attenuated Y. pestis Kimberley53 mutants, a transposon insertion was identified at the 39 end of the pcm gene of the Kim53-K9 mutant [4]. When administered subcutaneously to mice, the virulence of Kim53-K9 was severely attenuated [4]. Inspection of the genome sequence of Y. pestis CO92 [31] revealed that the pcm locus shows high conservation of the gene order with respect to the related enteropathogens E. coli and Salmonella (Fig. 1A). However, whereas the Y. pestis SurE, Pcm and RpoS proteins share high levels of identity with their related E. coli/Salmonella proteins (88-97%), the NlpD lipoproteins are more divergent (64-65%, Fig. 1B). The most noticeable disparity between the NlpDs of Y. pestis and E. coli/Salmonella is a protein sequence containing a unique proline and glutamine-rich repeat region at the N-terminus that is present in the proteins of the enteric pathogens but is completely absent from NlpD of Y. pestis and all other Yersinia species. Examination of Figure 1. In silico analysis of the pcm genomic locus of Y. pestis. (A) A similarity plot representing the alignment of the Y. pestis CO92 surE, pcm, nlpD and rpoS genomic locus (gray arrows) with the corresponding orthologous regions of E. coli K12 and S. typhimurium LT2. Peaks represent regions of sequence conservation; regions that are conserved among all three genomes are in light gray, while regions that are conserved only between E. coli and S. typhimurium are in dark gray. (B) Similarity comparison of the amino acid sequences of the Y. pestis CO92 pcm locus-encoded proteins to the corresponding proteins in E. coli and S. typhimurium. Y. pestis protein length is indicated in parenthesis and the similarity and E-values with respect to the E. coli or S. typhimurium proteins are provided. In the bottom row, the similarity between the E. coli and S. typhimurium proteins is presented. doi:10.1371/journal.pone.0007023.g001 the NlpD sequences within the Yersinia genus revealed that the nlpD gene products in Y. pseudotuberculosis and Y. enterocolitica have relatively high levels of sequence similarity to the corresponding gene product of Y. pestis (98% and 94%, respectively).
To characterize the factor/s involved in the attenuated phenotype of Kim53-K9, we used the virulent Y. pestis Kimber-ley53 strain to generate a series of isogenic deletion mutants within the pcm locus genes (Table 1). First, the DNA sequence of the pcm locus of Kimberley53 was determined (GenBank acc. no. FJ666123) and was found to be identical to CO92. In each of the newly constructed mutants, a defined region of a single gene was deleted and replaced with a kanamycin resistance cassette (see details in ''Materials and Methods''), resulting in Y. pestis Kimberley53 derivatives that were designated Kim53DsurE, Kim53Dpcm, Kim53DnlpD and Kim53DrpoS ( Fig. 2A). PCR analysis verified that all of the Kimberley53-derived strains carry the pMT1, pCD1, pPCP1 plasmids and the chromosomal pgm locus. To evaluate the expression of Pcm, NlpD and RpoS in these strains, bacterial cultures were grown to stationary phase (24 hours) at 37uC and 28uC in heart infusion broth (HIB) and then subjected to Western blot analysis with highly specific antibodies. The patterns of expression were independent of the growth temperature (data not shown). In the Kim53DsurE mutant, Pcm was not detected, whereas the levels of NlpD and RpoS were comparable to the wild type strain (Fig. 2B). This result suggests that a control element influencing pcm expression resides within the surE gene, as has been reported for E. coli pcm [19]. In the Kim53-K9 strain, Pcm and NlpD were not detected, whereas the level of RpoS was comparable to the wild type strain (Fig. 2B). This finding suggests that expression of nlpD is regulated by an element within the pcm gene. This assumption was further supported by the observation that the level of NlpD was reduced in Kim53Dpcm, whereas that of RpoS was not altered. Of note, in E. coli and Salmonella, expression of nlpD is driven by promoters located within the intergenic region between pcm and nlpD [21,22,25,27,32]. The genomic sequence in this region differs considerably between Y. pestis and the enteric pathogens (Fig. 1A). The rpoS gene resides downstream of nlpD. In bacterial enteropathogens, the major rpoS promoter has been identified within the nlpD gene [33 and references therein, 34,35]. We attempted to locate these known enterobacterial rpoS promoter sequences in the corresponding Y. pestis Kimberley53 and CO92 genomic regions. Putative rpoS promoter elements within the nlpD gene (547 bp upstream of the ATG start codon of rpoS) were identified according to tandem presence of 235 and 210 consensus sequences (Fig. S1). We therefore constructed two nlpD-null mutants: KimDnlpD L, in which the putative rpoS promoter elements were deleted, and KimDnlpD, in which these sequences were preserved ( Fig. 2A). As expected, the level of RpoS was decreased in KimDnlpD L but not in KimDnlpD (Fig. 2B). Of note is that the Kan r phenotype does not appear to derive expression of downstream ORFs, see for example the expression levels of Pcm in Kim53DsurE, NlpD in Kim53Dpcm, and RpoS in Kim53DnlpD L .
The last gene in the pcm locus that was analyzed was rpoS, which is predicted to encode an alternative sigma factor expressed during stress conditions [36]. The Pcm and NlpD expression levels in Kim53DrpoS were comparable to the levels found in the wild type strain (Fig. 2B). Likewise, in E. coli, expression of the pcm and nlpD genes is RpoS-independent [19,37].
Identification of the sites of transcription initiation of the pcm, nlpD and rpoS genes of Y. pestis and their relation to the observed expression profiles To further characterize the elements regulating the expression of the pcm, nlpD and rpoS genes, total RNA was prepared from cultures of Y. pestis bacteria grown to stationary phase (24 hours) at 37uC and then was used for primer extension analysis and RT-PCR. Using a primer complementary to the 59 region of the pcm gene, a single transcription initiation site designated Txn2 was identified within the surE gene, 176 bp upstream of the ATG start codon of the pcm gene ( Fig. 3A left panel). Similar analysis performed with a primer complementary to the 59 region of nlpD allowed identification of two transcriptional initiation sites within the surE and pcm genes located 988 bp (Txn1) and 300 bp (Txn3) upstream of the ATG start codon of nlpD, respectively ( Fig. 3A middle panel). To corroborate these findings, RT-PCR analysis was performed using primers complementary to regions within the nlpD and surE genes. An 1123-bp fragment was observed extending from the nlpD gene to the surE gene at a site between Txn1 and Txn2 ( Fig. 3B   Integration of the data obtained from primer extension, RT-PCR and Western blot analyses ( Fig. 2 and Fig. 3) helped to unravel the complex regulation controlling the expression of Y. pestis pcm, nlpD and rpoS genes (Fig. S2). Transcription of the pcm gene is driven by control elements within the surE gene, and a major transcription start site was identified at Txn2 (Fig. 3C). Consistent with this result, Pcm was not detected in the Kim53DsurE mutant (see Fig. 2). Transcription of the nlpD gene can start in principle from three different positions: Txn1, Txn2 and Txn3. However, in our analysis we failed to see nlpD transcripts from Txn2. Transcription starting at position Txn1 leads to the synthesis of a pcm-nlpD bicistronic message, whereas transcription starting from position Txn3 drives the synthesis of an nlpD-rpoS bicistronic message (Fig. 3C). The finding that the NlpD level was severely reduced in the Kim53Dpcm mutant in which the genomic region containing Txn3 was deleted (Fig. 2) indicates that this region includes a major nlpD control element/s. In agreement with this result, deletion of the Txn1 genomic region hardly interfered with NlpD expression (see Fig. 2, Kim53DsurE). Putative Y. pestis rpoS promoter sequences were identified within the nlpD gene by in-silico analysis (Fig. S1), and the rpoS transcription start point was indeed identified at that region (Txn4, Fig. 3C). A significant reduction in the RpoS level was observed upon deletion of the Txn4 region, indicating that this region indeed contains major rpoS control elements (see Fig. 2, Kim53DnlpD L ). The low RpoS level detected in Kim53DnlpD L (Fig. 2) probably represents the basal expression level of the nlpD-rpoS bicistronic message transcribed from Txn3 (see Fig. 3C). In accordance with these results, a deletion generated within the nlpD gene that did not affect Txn4 did not cause a significant reduction in the RpoS level (see Fig. 2 Kim53DnlpD).
To characterize the expression patterns of the Y. pestis pcm, nlpD and rpoS genes during logarithmic and stationary growth phases, wild type Kimberley53 and the Kim53Dpcm and Kim53DnlpD mutants were grown in HIB for 48 hours, and equal amounts of total protein were loaded on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Western blot analysis of Kimberley53 cells indicated that Pcm and NlpD levels remained essentially constant during the transition from logarithmic to stationary growth phase, whereas the RpoS expression pattern was growth phase-dependent (Fig. 4). One should note that during the stationary growth phase an NlpD precursor starts to accumulate, appearing as a slightly slower migrating band than NlpD. Similar findings were reported for E. coli [21,36,38]. The Western blot analyses with the various mutants ( Fig. 4), together with the RT-PCR results (Fig. 3B), indicate that the low expression levels of NlpD and RpoS were governed by distal control elements (Txn1 and Txn3, respectively). We therefore evaluated the contribution of each of the distal elements to NlpD (Txn1) and RpoS (Txn3) expression in vitro during growth. In Kim53Dpcm, the NlpD level was constant but was lower than that of the wild type strain (Fig. 4). Consequently, the region in proximity to Txn3 contains the major nlpD control element. In Kim53DnlpD L , in which the Txn4 region was deleted, the level of RpoS was low and growth phasedependent (Fig.4). This finding indicates that the major control elements regulating RpoS expression are located in close proximity to the Txn4 genomic region. NlpD is an essential factor for the development of bubonic and pneumonic plague To test the contribution of SurE, Pcm, NlpD and RpoS to Y. pestis pathogenesis, we evaluated the virulence of Kimberley53deletion mutants in mouse models of bubonic plague (s.c. infection) and pneumonic plague (intranasal infection). In the bubonic plague mouse model, Kim53DsurE, Kim53Dpcm and Kim53DrpoS were found to be highly virulent as reflected by the mortality rate, the mean time to death and the LD 50 value (Fig. 5A, Table 2). In contrast, The Kim53-K9 and Kim53DnlpD strains were avirulent (Fig. 5A, Table 2).
The inability of Kim53-K9 and Kim53DnlpD to overcome host defense systems was further emphasized by the finding that infecting mice with higher doses of these mutants, up to 10 7 cfu of each strain, was non-lethal (Table 2) and that infected mice did not show any visible disease symptoms, such as ruffled hair or a hunched back.
Intranasal (i.n.) infection of mice was demonstrated recently to serve as a model for studying the development of primary pneumonic plague [3]. We therefore used this model system to study the involvement of our genes of interest in development of pneumonic plague. All mice infected intranasally with 2610 4 cfu of wild type Kimberley53 succumbed to the infection by day 4 (data not shown), and the LD 50 was found to be 550 cfu ( Table 2). Similar LD 50 values were reported recently for other virulent Y. pestis strains [39,40]. The Kim53Dpcm and Kim53DrpoS mutants were able to produce disease upon i.n. infection of mice, and the LD 50 values of these mutants were 6 and 2 fold higher (respectively), than the LD 50 of the wild type strain ( Table 2). In contrast, i.n. infection of mice with up to 4610 7 cfu of Kim53DnlpD was non-lethal (Table 2), and infected mice did not show disease symptoms.
To substantiate our observation that the attenuated phenotype of the NlpD-null mutants (Kim53DnlpD and Kim53-K9) resulted from the loss of NlpD expression, we examined the ability of a trans-complemented Y. pestis nlpD gene to restore the mutant's virulence. Due to the toxicity associated with over-expression of NlpD ( [21] and our unpublished data), we expressed the gene using its native control element. The complete coding sequence of the Y. pestis Kimberley53 nlpD gene preceded by the 442 bp upstream of the nlpD ATG, a region that includes the Txn3 transcription initiation region, was cloned into pBR322. The plasmid was introduced into Kim53DnlpD and Kim53-K9 to give the complemented strains Kim53DnlpD(pnlpD) and Kim53-K9(pnlpD). Western blot analysis performed using an overnight culture of the complemented strains grown at 37uC or 28uC (not shown) confirmed that in both strains the NlpD expression level was comparable to expression in the wild type strain (Fig. 5B). All mice infected subcutaneously with 100 cfu of the NlpDcomplemented strains (the bubonic infection model) succumbed to the infection by day 8 (Fig. 5C). Moreover, Kim53DnlpD(pnlpD) also regained virulence following i.n. infection (the pneumonic infection model) ( Table 2). These results confirmed that NlpD is a novel virulence factor of Y. pestis and is essential for development of both bubonic and pneumonic plague. Finally, NlpD expression on the background of the Kim53-K9 mutant also restored virulence, confirming that the attenuation of Kim53-K9 was not due to the absence of the pcm gene product but resulted from the elimination of nlpD expression.

Deletion of nlpD impairs Y. pestis colonization of internal organs in mice
In an attempt to evaluate the level of attenuation of Kim53DnlpD, we used mice to monitor the colonization of internal organs by Kim53DnlpD and the wild type strain Kimberley53. A dramatic disparity between the two strains was observed following infection via both the s.c. and the airway routes. In contrast to the wild type strain, which reached high bacterial loads in all organs within 48 hours, Kim53DnlpD could not be detected in the draining inguinal lymph nodes (I-LN), the spleen, the lungs or the blood of the subcutaneously infected mice at 24 hours post infection (data not shown) and up to 10 days post infection (Fig. 6A). No visible lesions were observed at the site of the injection.
When mice were infected intranasally with Kim53DnlpD, the bacterial load in the lungs decreased rapidly by two orders of magnitude within 48 hours to an average of 110 cfu (Fig. 6B), and bacteria were cleared within 3 days post infection (data not shown). In contrast, the bacterial load increased rapidly when the wild type strain was used. Moreover, Kim53DnlpD was unable to colonize the mediastinal lymph node (M-LN), the spleen or the blood, whereas the number of colony forming units in all organs increased rapidly to an average of 10 4 -10 7 cfu within 48 hours for the wild type strain (Fig. 6B), causing the death of all mice within 4 days.
These observations indicate that NlpD is required for Y. pestis propagation and dissemination to target organs, and they corroborate the non-virulent phenotype of Kim53DnlpD in both pneumonic and bubonic infection models.
In vitro characteristics of the nlpD-null mutant Bacterial lipoproteins are components of the cell envelope of Gram-negative bacteria and are usually localized at the periplasmic space anchored to either the outer or the inner membrane [41]. Y. pestis NlpD has a threonine residue immediately after the fatty-acylated cysteine and is therefore predicted to reside in the outer membrane similarly to the E. coli NlpD [41,42]. Microscope analyses indicated that in vitro culturing of Kim53DnlpD at 28uC leads to formation of unsegmented chains containing an average of 7.265.6 cells/chain as opposed to the more common aggregative morphology of Kimberley53 ( Figures 7B and 7A, respectively). Elevation of the culturing temperature to 37uC (Fig. 7D), reduced significantly the number of Kim53DnlpD cells/chain to 462.5 (Ttest P,1E-14), while the wild type morphology is temperature independent (Fig. 7C). In spite of this change, NlpD is not essential for Y. pestis cell growth, at least under laboratory conditions, as  shown by the normal growth rate of the nlpD mutant (Fig. 7E). Growth rate of Kim53DnlpD was found to be similar to that of the wild type strain under in vitro culture conditions at 37uC (Fig. 7E) and at 28uC (data not shown) as determined by ANOVA test of regression (P.0.15). Of note is that there is a significant difference between the intercept of the nlpD mutant and the wild type strain (P,0.05) due to the filamentous phenotype of the mutant (Fig. 7B  and 7D). Complementation of nlpD expression eliminated completely the abnormal cell morphology and restored the wild type morphology (data not shown).
It can be assumed that deletion of nlpD in Y. pestis may influence membrane-related functions, such as TTSS activity, which is known to be essential for Y. pestis virulence, and thus could lead to the attenuated phenotype. However, preservation of TTSS functionality, at least in vitro, was demonstrated by retention of calcium-dependent growth at 37uC and expression and secretion of Yop effectors such as YopE (Fig. 8A and 8B) and YopJ (data not shown), into the culture medium following exposure to inducing conditions. Moreover, the nlpD-null mutant preserves the ability to translocate Yop effectors into host cells, like the wild type strain, as evidenced by suppression of TNF-a secretion from infected RAW264.7 macrophages (Fig. 8C). The genes flanking nlpD were found in many enteropathogens to be involved in survival during stationary phase and under other environmental stress conditions [36]. In search of a possible explanation for the attenuation of Kim53DnlpD, we evaluated the involvement of Y. pestis NlpD in resistance to prolonged growth in rich (HIB) and minimal (M9) broth, and in acidic and oxidative conditions that simulate the intra-phagosomal milieu. Exposure of Kim53DnlpD and Kimber-ley53 to prolonged in vitro growth (up to 96 h), and to oxidative stress (10-50 mM H 2 O 2 ) indicated that NlpD is not essential for resistance to these conditions (data not shown). In contrast, Kim53DnlpD was slightly more susceptible than wild type  Kimberley53 to acidic pH (Fig. 8D). Moreover, the Kim53DnlpD(pnlpD) strain, in which NlpD expression was restored, regained resistance to these conditions (Fig. 8D). The possible linkage between acidic susceptibility and attenuation of Kim53DnlpD awaits further study.

Evaluation of the immunization potential of the Kim53DnlpD strain
The high level of attenuation of Kim53DnlpD motivated us to evaluate the potential of this defined mutant to serve as a vaccine. In a model of bubonic plague, mice were injected subcutaneously with 1610 5 cfu of either Kim53DnlpD or the Y. pestis EV76 prototype vaccine strain. Fifty days later, mice were challenged subcutaneously with 1610 5 LD 50 of the fully virulent Kimberley53 strain. As shown in Table 3, the geometric mean ELISA titers of aF1 and aV antibodies measured in sera from the mice at 30 days post immunization with Kim53DnlpD were 17,100 and 26, respectively. Similar aF1 and aV titers were measured following immunization with Kim53-K9. In contrast, mice vaccinated with the EV76 strain demonstrated anti-F1 geometric mean titers of less than 30 and anti-V antibody ELISA titers below the limit of detection (Table 3). These results suggest that although Kim53DnlpD was highly impaired in colonization of host internal organs, its persistence within the host lasted long enough for the host to mount an effective immune response. Consistent with these results, all mice vaccinated with Kim53DnlpD survived s.c. challenge with 1610 5 LD 50 of Kimberley53 without showing signs of illness whereas a similar dose of EV76 failed to elicit significant protective immunity ( Table 3). The EV76 strain appears to be effective only when applied at a high vaccination dose (10 7 cfu , Flashner et al 2004).
In the mouse model of pneumonic plague, a single s.c. immunization with 1610 5 -1610 7 cfu of Kim53DnlpD or EV76 was followed 50 days later by i.n. challenge of virulent Y. pestis Kimberley53 ( Table 4). The challenge dose was 5,500 cfu, which is equivalent to 10 LD 50 . While all control mice died within 4 days, a protection level of 33% was obtained following immunization with 10 5 cfu of Kim53DnlpD. Higher protection levels of 66% and 82% were obtained by immunization with increasing doses of Kim53DnlpD (10 6 cfu and 10 7 cfu, respectively). In contrast, EV76 was unable to elicit protection following immunization with 10 5 cfu, and a protection level of 33% was attained only after the immunization dose was increased to 10 7 cfu (Table 4). Comparing total survival in the nlpD-vaccinated mice to EV76-vaccinated mice by Chi-square analysis revealed highly significant difference (P = 0.0013). These findings strongly accentuate the potency of Kim53DnlpD in establishing effective protective immunity and suggest that it may be a suitable platform for a live vaccine.

Discussion
Studies performed in the last decade with several Gram negative bacteria have demonstrated that the genomic region including the surE, pcm, nlpD and rpoS genes is important for survival under environmental stress conditions [18,36,43,44]. In the present study, we have further characterized the Y. pestis pcm locus genes and analyzed their expression and involvement in the pathogenesis of Y. pestis using mouse models of bubonic and pneumonic plague. Systematic deletion mutagenesis of the surE, pcm, nlpD and rpoS genes and complementation studies allowed us to identify the NlpD lipoprotein as the only essential factor for Y. pestis pathogenesis in this locus.
Subcutaneous and intranasal administration of a Y. pestis nlpDnull strain to mice demonstrated that this strain is severely attenuated (LD 50 .10 7 cfu, Table 2), and is impaired in its ability to colonize internal organs (Fig. 6). In addition, the nlpD-null mutant was unable to produce a systemic disease following intravenous inoculation with 10 6 cfu and infected mice lacked any visible disease symptoms (data not shown).
Trans-complementation experiments verified that NlpD is an essential Y. pestis virulence factor (Fig. 5C, Table 2). Using a similar strategy, virulence was also restored to the original transposon insertion mutant (Kim53-K9), confirming that its attenuation resulted from elimination of nlpD expression and not from disruption of pcm expression (Fig. 5C, Table 2; note also that the Kim53Dpcm is as virulent as the wild type strain). To the best of our knowledge, this is the first demonstration of a single chromosomal Y. pestis factor that is essential for development of both bubonic and pneumonic plague.
Yersinia enterocolitica NlpD was also suggested to be involved in pathogenesis [45]. In the latter study, transposon insertion within the Y. enterocolitica nlpD gene led to attenuation of virulence in a mouse infection model. Interestingly, as opposed to the dramatic attenuation of the Y. pestis nlpD mutant shown in the present study, the Y. enterocolitica nlpD mutant displayed only a limited impairment of virulence [45]. This finding might imply that Y. pestis nlpD is required for an activity specific to the development of plague disease.
Recently, an additional Y. pestis lipoprotein, Braun's lipoprotein (Lpp), was found to be important for virulence [40]. It should be noted, however, that in contrast to NlpD, Lpp contributes only to the production of bubonic plague but not to that of the pneumonic plague [40]. Lipoproteins have been implicated in the pathogenesis of other bacterial pathogens as well. For example, disruption of the lsp and lgt genes, which are involved in lipoproteins   [46][47][48].
The dramatic loss of virulence of the Y. pestis nlpD-null strain is reminiscent of the non-virulent phenotype of Y. pestis strains lacking the pCD1 plasmid that carries the genes encoding the type III secretion system. However, our results indicate that the nlpD mutant is able to properly express and secrete Yop effectors during in vitro inducing conditions ( Fig. 8A and 8B) and is able to translocate effectors into host cells, as reflected by its ability to suppress TNFa secretion from infected macrophages (Fig. 8C). Moreover, the V-antigen multifunctional virulence factor important for type III secretion system activity is expressed by the nlpD mutant during infection as reflected by development of aV antibodies following immunization of mice (Table 3 and 4). Taken together, these observations suggest that the type III secretion system is functional in the nlpD mutant.
The finding that the Y. pestis nlpD gene is expressed from control elements shared with either pcm or rpoS raised the possibility that NlpD might also be important for survival under environmental stress conditions. However, the nlpD-null mutant did not differ significantly from the wild type strain in the ability to survive during prolonged growth in rich and minimal broth and during exposure to oxidative conditions. Yet, we found a slight increase in the mutant's sensitivity to in vitro acidic conditions (Fig. 8), similar to the conditions present in the intra-phagosomal milieu. While the latter does not seem to account for the dramatic loss of virulence, further studies are needed to understand the importance of the sensitivity of the mutant to acidic conditions. The aberrant shape of the Y. pestis DnlpD bacilli ( Fig. 7B and 7D) suggests that NlpD is important for cell separation. Consistent with this assumption is the structure of NlpD (Fig. 9), which contains an N-terminal LysM domain found in a variety of enzymes involved in bacterial cell wall degradation [49][50][51]. NlpD also contains a Cterminal M23 metallopeptidase region (Fig. 9). The M23 family of endopeptidases includes proteins that are involved in bacterial cell separation [52], however, no proteolytic activity has been demonstrated for bacterial lipoproteins of this family (see the MEROPS peptidase database, http://merops.sanger.ac.uk [53]). Of note is that in vitro cell growth of the nlpD mutant was not affected (Fig. 7E). The observed phenotype of the nlpD-null mutant ( Fig. 7B and 7D) suggests a linkage between impairment of cell separation and attenuation of virulence. Interestingly, similar links have already been described for other bacterial pathogens [54,55]. Furthermore Kajimura and colleagues [54] hypothesized that cluster formation of the attenuated S. aureus Sle1 mutant, which lacks the lysM-containing N-acetylmuramyl-L-alanine amidase, inhibits the dissemination of daughter cells and thus could affect the spread of the bacteria during infection. Indeed, the Y. pestis nlpD mutant is totally impaired in its ability to disseminate from the site of infection to the internal organs (Fig. 6).
Interestingly, the highly attenuated phenotype of the nlpD mutant and its inability to colonize host organs did not seem to prevent the development of immunity against plague following s.c. infection. Rather, this strain seemed to effectively stimulate a longterm adaptive immune response as demonstrated by the generation of high F1 antibody titers. Accordingly, immunization of mice with low doses of Kim53DnlpD resulted in remarkably higher humoral response and protection levels against bubonic and pneumonic plague than did immunization with the Y. pestis EV76 vaccine strain (Tables 3 and 4). The differential behavior of the two vaccine strains may result from the complete absence of the chromosomal pgm locus (102 kb including the pathogenicity island), from EV76. The contribution of the pgm locus to Y. pestis survival in host cells is well documented [56]. The observed development of protective immunity could have practical implications in the design of future Y. pestis vaccines or therapies against both bubonic and pneumonic plague.

Bacterial strains, plasmids, mutant construction and routine growth conditions
The Y. pestis strains and the plasmids used in this study are listed in Table 1. The primers used for construction of the Y. pestis derivatives are listed in Table S1. Construction of the Kimber-ley53 deletion mutants was performed by replacing part of the gene coding sequence with a linear fragment containing the Kan r GeneBlock TM resistance cassette (pUC4K plasmid, Pharmacia) by homologous recombination, using previously established methodologies [60]. In all constructs the Kan r resistance cassette is oriented in the same direction as the pcm locus genes ( Fig. 2A). To preserve expression of downstream genes, the resistance cassette does not include transcription terminator sequences.
Genotype verification of all obtained phenotypes was done by PCR and Western blot analysis. All of the Kimberley53-derived strains carry the pMT1, pCD1, pPCP1 plasmids and the pgm locus.
Bacteria were isolated from stocks on selective BIN medium [61]. Routine propagation was performed on brain heart infusion agar (BHIA) (Difco) or HIB (Difco). For generation of nlpDcomplemented strains, a 1444 kb fragment extending from the 39 end of the nlpD gene up to 442 nt upstream of the ATG start codon of nlpD was amplified from the DNA of the virulent Y. pestis Kimberley53 strain. The resulting PCR product was cloned into pBR322 between the NheI and PstI sites and its sequence was verified. Plasmids were introduced into the Kim53DnlpD and Kim53-K9 strains by electroporation to give the Kim53DnlpD(pnlpD) and Kim53-K9(pnlpD) strains. These strains were routinely grown in media supplemented with 100 mg/ml ampicillin (Sigma, Israel).
DNA sequence determination of the Y. pestis Kimberley53 pcm locus and in-silico analysis The Y. pestis CO92 sequence corresponding to the pcm locus (nt 3744222 to 3747841) was compared to the Kimberley53 sequence at the equivalent region and was found to be identical) GenBank acc. no. FJ666123). All sequencing reactions were performed with an ABI310 genetic analyzer (Applied Biosystems) using the ABI PRISM BigDye terminator reaction kit. The genome alignment of the orthologous regions comprising the surE, pcm, nlpD and rpoS genes was generated by the multiple genome alignment software Mauve [62]. The sequences were extracted from the following GenBank (NCBI) entries: NC_003143 (Y. pestis CO92), NC_003197 (S. typhimurium LT2), NC_000913 (E. coli K12).

Growth curves
Y. pestis strains were grown by shaking (100 rpm) at 28uC in HIB supplemented with 0.2% (+)xylose (Sigma) and 2.5 mM CaCl 2 (Sigma). The resulting cultures were diluted in fresh broth to an OD 660 of 0.01 and allowed to grow for 48 hours at either 28uC or 37uC while shaking at 180 rpm. At the indicated time points, aliquots were removed for determination of the OD 660

Western blot analysis
Bacteria (OD 660 = 0.1) were lysed with Laemmli Sample buffer (Bio-Rad) and protein concentrations were determined using bicinchoninic acid (BCA Protein Assay Reagent, Pierce) according to the manufacturer's instructions. Equal amounts of protein were loaded and separated by 12% SDS-PAGE. After transfer to nitrocellulose membranes, duplicate membranes were developed with rabbit polyclonal anti-peptide Pcm and NlpD antibodies, with rabbit polyclonal S. typhimurium RpoS antibodies (a generous gift from François Norel, Institute Pasteur), or with goat anti-YopE antibodies (bL-20, Santa Cruz Biotechnology) and with rabbit polyclonal anti-peptide YopJ [60]. Probing with the primary antibody was followed by incubation of the membranes with HRPconjugated second antibody, and then the reactive protein bands were visualized by ECL. The Pcm and NlpD anti-peptide antibodies were raised by immunizing rabbits with maleimideactivated KLH (Pierce) conjugated to the synthetic peptides CERLLQAIEAVPRER (amino acids 21-34 of Pcm) and CVGGDRSGTMLSKANT (amino acids 39-53 of NlpD). Antibodies were affinity purified using a Sulfolink-peptide coupling gel column (Pierce) according to manufacturer's instructions.
Total RNA isolation, primer extension and RT-PCR analyses HIB supplemented with 0.2% (+)xylose and 2.5 mM CaCl 2 was inoculated with Y. pestis bacteria (initial OD 660 = 0.01), and incubated for 24 hours at 28uC or 37uC with shaking at 100 rpm. Total RNA was isolated using a RiboPure TM -bacteria Kit (Ambion) and was then treated with DNase I (Ambion) according to the manufacturer's instructions.
Primer extension (PE) reactions were carried out according to Lloyd and colleagues [63]. Briefly, for each reaction, FAM-labeled primer (Table S1, final concentration 5-10 nM) was added to 20 mg of total RNA, and first strand cDNA synthesis was performed using the AMV RT enzyme (Promega). FAM-labeled cDNAs were cleaned using a PreformaH DTR Gel Filtration Cartridge (Edge BioSystems) according to the manufacturer's instructions and then air-dried with a heated SpeedVac centrifuge and resuspended in 25 ml of UltraPure TM Formamide (Invitrogen) with 1.5 ml MapMarkerH1000 (BioVentures, Inc.). The mixture was heated to 90uC for 2 minutes, chilled on ice for 5 minutes and then used for electrophoresis, using an ABI310 genetic analyzer (Applied Biosystems). The DNA fragments were sized using the GeneScanR Analysis Software version 3.1 (Applied Biosystems). For reverse transcriptase-PCR analyses (RT-PCR), primers were added to 1 mg of total RNA and synthesis was carried out with AMV RT (Promega) at 42uC for 1 hour according to the manufacturer's instructions. After first strand synthesis the cDNA was diluted 1:2 and an aliquot (3 ml) was taken for PCR amplification. Positive and negative control reactions were performed with 200 ng of DNA or with total RNA, respectively.

Infection of mice
All experiments were performed in accordance with the Israeli law and were approved by the Ethics Committee for animal experiments at the Israel Institute for Biological Research. Female OF1 mice, 5-6 weeks old (IFFA CREDO S.A., France), were used in all infection studies. For s.c. infections of mice, the Y. pestis strains were grown on BHIA for 48 hours at 28uC. Bacterial colonies were harvested from the BHIA plates at the end of the incubation period, suspended in 5 ml of saline solution (0.9% NaCl) and diluted in saline solution to the required infection dose. Bacteria were quantified by counting colony forming units after plating and incubating on BHI agar plates. In s.c. infections, samples of 100 ml were administered into the lower back of the mice. Under these conditions, the LD 50 of the Kimberley53 strain is 1-3 cfu [4]. For i.n. infections, the Y. pestis strains were grown on BHIA for 48 hours at 28uC. Bacterial colonies were suspended in HIB supplemented with 0.2% (+)xylose and 2.5 mM CaCl 2 to give an initial OD 660 of 0.01, and the cultures were incubated for additional 22 hours at 28uC or 37uC with shaking at 100 rpm. At the end of the incubation period, the cultures were washed, diluted in saline solution to the required infection dose and quantified by cfu counting as described above. Prior to infection, mice were anaesthetized with a mixture of 0.5% ketamine HCl and 0.1% xylazine injected intraperitoneally and were then infected intranasally with 35 ml of the bacterial suspension. The LD 50 experiments were performed with groups of 8 mice (i.n.) or 5 mice (s.c.). The LD 50 values were calculated according to the method described by Reed and Muench [64], protection analyses were performed by Chi-square test.

Monitoring disease progression
Mortality rates were expressed in accordance with the required information by either the proportion of dead animals at the end of the monitoring period (14 days unless otherwise indicated), by MTTD, or by the 50% lethal dose (LD 50 ). Monitoring of disease progression was performed by tracking bacterial dissemination to the internal organs and blood. In s.c. infection experiments, organs were aseptically removed 48 hours after infection with 1610 4 cfu of the wild type Kimberley53 strain (a terminal stage of disease) or with 1610 5 of Kim53DnlpD. Groups of four mice were anaesthetized, tail vein blood was collected and spleens, lungs and draining inguinal lymph nodes (I-LN) were harvested. Tissue homogenates were prepared in 1 ml PBS/organ. Bacterial enumeration in tissue homogenates or in blood samples was done by plating serial dilutions in PBS on BHI agar and calculating the cfu/organ or cfu/1 ml blood. In i.n. infection experiments, groups of four mice were anesthetized, and tail vein blood was collected one hour and 48 hours post infection with 1610 5 cfu. The lungs, spleens and the mediastinal lymph nodes (M-LN) were harvested and prepared as described above for bacterial enumeration.

Morphology analysis
Y. pestis bacilli were observed by light microscopy using a Nikon Eclipse E200 microscope at a magnification of 61000. Gram staining was performed using a HT90A kit (Sigma-Aldrich) according to the manufacturer's instructions. Y. pestis cells in at least 20 random non-overlapping microscopic fields were counted.

TTSS functionality analyses and stress survival assays.
For secretion of TTSS proteins, bacteria were grown at 28uC in HIB for 18 hours with shaking at 150 rpm. The bacteria were then diluted to an OD 660 of 0.05 in HIB medium supplemented with either 20 mM sodium oxalate (Sigma) and 20 mM MgCl 2 (Sigma) (inducing conditions) or 2.5 mM CaCl 2 (non-inducing conditions) and were grown for additional 4.5 hours at 37uC with shaking at 100 rpm. Cultures were centrifuged at 4500 g for 10 min, the supernatants were collected and the proteins were precipitated with 10% TCA overnight at 4uC and the subjected to SDS-PAGE. Analysis of the suppression of TNF-a secretion from macrophages was preformed as described in Zauberman et al. [60]. Briefly, RAW264.7 macrophages (2610 5 ) were seeded in 24-well plates and infected at the indicated MOI (determined by the cfu count) for 2 hours. The concentration of TNF-a in the culture medium was determined by ELISA using the DuoSet mouse TNFa immunoassay system (R&D systems). The susceptibility of the Y. pestis strains to acidic pH was determined by plating stationaryphase bacteria on BHI agar after incubation for 10-60 minutes in 16PBS adjusted to pH 4.2 with citric acid. The susceptibility of the Y. pestis strains to hydrogen peroxide was determined by plating stationary-phase bacteria on BHI agar after exposure to hydrogen peroxide (10-50 mM in 16PBS, pH = 7) for up to 60 minutes.

Antisera and serological tests
Mouse anti-F1 and anti-V antigen IgG antibody titer determinations were performed by ELISA as previously described by Flashner and colleagues. [4]. Briefly, microtiter plates were coated with purified recombinant F1 antigen [65] or with 500 ng of purified Glutathione-S-Transferase -V-antigen fusion protein prepared according to Leary et al. [66]. The tested sera were serially diluted by 2-fold dilutions in a final volume of 50 ml and were incubated in the wells for 1 hour at 37uC. Alkaline phosphatase-labeled rabbit anti-mouse IgG (1/2500 dilution, Sigma) was used as the secondary antibody. Titers were defined as the reciprocal values of the endpoint serum dilutions, which had OD 405 values two fold higher than the normal serum controls. The expression level of each gene is indicated by the intensity of its color: dark gray -comparable to the wild type strain, light gray -lower than the wild-type strain, colorless -no expression. The mini-Tn5 transposon inserted within Kim53-K9 is indicated by a black arrow (the arrow points to the direction of transcription of the kanamycin resistance cassette). The (*) symbol indicates a putative transcription start site (see also in Results). Found at: doi:10.1371/journal.pone.0007023.s002 (0.60 MB TIF)

Supporting Information
Table S1 Sequences of primers used in this study. a. Numbers in parenthesis indicate the position of the primer relative to the first nucleotide in the ORF. b. Underlined bases indicate restriction sites for cloning of the amplified sequence. c. FAM (6-carboxyfluorescein) labeled primers for primer extension. These primers where also used for first strand synthesis in the RT-PCRs. Found at: doi:10.1371/journal.pone.0007023.s003 (0.05 MB DOC)