A New Piece of the Shigella Pathogenicity Puzzle: Spermidine Accumulationby Silencing of the speG Gene

The genome of Shigella, a gram negative bacterium which is the causative agent of bacillary dysentery, shares strong homologies with that of its commensal ancestor, Escherichia coli. The acquisition, by lateral gene transfer, of a large plasmid carrying virulence determinants has been a crucial event in the evolution towards the pathogenic lifestyle and has been paralleled by the occurrence of mutations affecting genes, which negatively interfere with the expression of virulence factors. In this context, we have analysed to what extent the presence of the plasmid-encoded virF gene, the major activator of the Shigella regulon for invasive phenotype, has modified the transcriptional profile of E. coli. Combining results from transcriptome assays and comparative genome analyses we show that in E. coli VirF, besides being able to up-regulate several chromosomal genes, which potentially influence bacterial fitness within the host, also activates genes which have been lost by Shigella. We have focused our attention on the speG gene, which encodes spermidine acetyltransferase, an enzyme catalysing the conversion of spermidine into the physiologically inert acetylspermidine, since recent evidence stresses the involvement of polyamines in microbial pathogenesis. Through identification of diverse mutations, which prevent expression of a functional SpeG protein, we show that the speG gene has been silenced by convergent evolution and that its inactivation causes the marked increase of intracellular spermidine in all Shigella spp. This enhances the survival of Shigella under oxidative stress and allows it to better face the adverse conditions it encounters inside macrophage. This is supported by the outcome of infection assays performed in mouse peritoneal macrophages and of a competitive-infection assay on J774 macrophage cell culture. Our observations fully support the pathoadaptive nature of speG inactivation in Shigella and reveal that the accumulation of spermidine is a key determinant in the pathogenicity strategy adopted by this microrganism.


Introduction
Polyamines are ubiquitous, small polycationic compounds associated with a variety of biological processes: protein translation, gene regulation, stress resistance and differentiation [1,2]. Major representatives of this class of molecule are putrescine, cadaverine, spermidine and spermine. In bacteria, the global level of polyamines is regulated on the one hand by collective effects of catabolism and efflux mechanisms and, on the other, by biosynthetic pathways and uptake mechanisms [2,3]. Figure 1 reports the superpathway of polyamine biosynthesis I in Escherichia coli (from http:\\ecocyc. org database), which is able, like most c-proteobacteria, to synthesize cadaverine, putrescine and spermidine, but not spermine [2,4]. Cadaverine is produced through the combined action of an inducible and a constitutive lysine decarboxylase, encoded respectively by the cadA and ldc genes [5,6]. It is then converted to aminopropylcadaverine by the SpeE protein.
Putrescine results from direct ornithine decarboxylation, mediated by the SpeC decarboxylase, and from arginine decarboxylation followed by agmatine ureohydrolization determined by the SpeA and SpeB proteins, respectively. Spermidine originates from the condensation of putrescine with decarboxylated S-adenosylmethionine, performed by the SpeE [2,7]. High levels of spermidine are toxic for E. coli cells, but spermidine acetylation, catalysed by SpeG, inactivates the polyamine. Acetylspermidine is thought to be either stored by the cells or secreted [8].
During recent years, strong evidence has accumulated on the role of polyamines in microbial pathogenesis. In Pseudomonas aeruginosa, the deletion of genes involved in spermidine uptake significantly decreases the expression of Type III Secretion Systems (TTSS) [9]. In Streptococcus pneumoniae, polyamine biosynthesis and transport mechanisms are intricately linked to the fitness, survival and pathogenesis of this pathogen in host microenvironments [10]. In Yersinia pestis and in Vibrio cholerae, polyamines have been implicated in the control of biofilm formation [11,12]. In Proteus mirabilis, the inactivation of the speAB genes, involved in putrescine biosynthesis, leads to the loss of the swarming phenotype [13] linked to the expression of some virulence genes [14]. The fungal pathogen Pneumocystis jirovecii produces high levels of spermidine, N 1 -acetylspermine and N 1acetylspermidine, thus inducing apoptosis of alveolar macrophages [15].
We have focused our analysis on spermidine metabolism in Shigella, a facultative intracellular pathogen causing a severe enteric syndrome in humans, mainly in the developing world. Shigellosis is extremely contagious and, although usually self-limiting, may be fatal in children [16]. The highly sophisticated infectious strategy of Shigella banks on the capacity of this pathogen to invade, disrupt, and cause inflammatory destruction of the intestinal epithelial barrier. Once ingested, Shigella moves directly down to the colon where it gains access to the intestinal mucosa by invading specialized epithelial cells, the M cells in Peyer's patches, and subsequently infecting adjacent cells in intestinal crypts. Once the bacteria reach the lymphoid follicles, they encounter resident macrophages, where they multiply, induce apoptosis and give rise to an inflammatory response, the hallmark of this enteric disease. This, in turn, induces transmigration of polymorphonucleated leukocytes (PMN) through the tight junctions between epithelial cells. As PMNs begin to migrate, bacteria released from killed macrophages can invade the epithelial monolayer, accessing the basolateral surfaces of the colonic epithelium. Bacterial entry into the host cells is induced by the TTSS-secreted Ipa proteins, which activate host signaling pathways and induce a focused reorganization of the cytoskeletal actin around the bacterial cell. Inside the host cell, Shigella disrupts the vacuole membrane and escapes into the cytoplasm, where it multiplies, and moves by inducing local actin polymerization at one pole of the bacterium. The actin-based motility propels Shigella through the cytoplasm and facilitates intercellular dissemination towards the neighboring cells [17,18].
The cellular pathogenesis and clinical presentation of shigellosis are the sum of the complex action of a large number of bacterial virulence factors mainly located on a large virulence plasmid (pINV) [19]. The availability of complete sequenced genomes of several Shigella strains has given new insight about the molecular evolution of this bacterial pathogen from its ancestor, the commensal E. coli [20]. While the acquisition of pINV is regarded as one of the most critical events in the evolution of Shigella towards a pathogenic lifestyle, a significant complementary step has been the emergence of so-called pathoadaptive mutations [21]. This has led to the inactivation of several chromosomal genes, which negatively interfere with the expression of virulence factors required for the survival within the host [22,23]. In particular, the silencing of the cad genes, involved in the synthesis of a specific polyamine, cadaverine, appears crucial for the optimization of the pathogenicity process in Shigella [22,23]. Cadaverine negatively interferes with Shigella-induced pro-inflammatory events by inhibiting PMN migration to the infection loci [24] and may stabilize the endosomal membrane, hindering the release of Shigella cells into the cytoplasm of infected cells [25].
In this study, we show, by convergent evolution, that Shigella has lost another crucial gene involved in polyamine metabolism, speG. The loss of this gene, which encodes spermidine acetyltransferase, allows for higher concentrations of endogenous spermidine. We also show that restoring SpeG activity confers upon Shigella a higher sensitivity to oxidative stress and reduces bacterial survival inside macrophages. This strongly supports the hypothesis that speG inactivation constitutes a previously unrecognized pathoadaptative mutation common to all member of Shigella genus.

speG expression depends on the VirF regulatory protein
The Shigella pINV plasmid contains, besides genes involved in the invasive process, positive activators necessary for the induction of host cell colonization. Among them, a critical role is played by the transcriptional regulator VirF. It is encoded by a gene activated in response to host temperature [26] and is located outside the large pathogenicity island carrying most virulence determinants [19]. VirF triggers a cascade of events: it activates the transcription of the gene coding for the secondary regulator, VirB, which activates several operons coding for the invasion genes [27].
To understand whether the arrival of VirF by acquisition of pINV might have altered the transcriptional program of the ancestor E. coli and promoted the inactivation of genes potentially detrimental to the full expression of the invasive phenotype, we performed a global transcriptional analysis of E. coli cells expressing or lacking the virF gene. To this end we set up a microarray experiment using the E. coli K12 MG1655 strain [28], carrying the virF-encoding plasmid pMYSH6504 [29] or its virFdepleted derivative pMY6504R (Table S1). This analysis was performed on an E. coli K12-V2 array (MWG) containing 4288 gene-specific oligonucleotide probes representing the complete E. coli (K12) genome. This experiment revealed the presence of several E. coli genes activated at least two fold by VirF, either directly or indirectly (Table S2). Comparative genome analysis with Shigella reveals that these genes can be subdivided into two groups: genes which are common to Shigella and E. coli, and genes that are deleted or inactivated in Shigella. Interestingly in the first group we have identified several, highly induced, genes coding heat shock proteins including ibpA, htpG, GroL/GroS, dnaK and lon (Table S2).
The existence of the second group suggests that some VirFactivated genes might have exerted a perturbing effect on the Shigella invasive process, thus becoming silenced during evolution optimizing bacterial survival in the host. Most of the VirFactivated genes silenced in Shigella are poorly characterized. An exception is represented by speG, which encodes spermidine acetyltransferase (SAT) ( Table S2). The speG gene belongs to the ynfB-speG operon. While no function has been yet attributed to the ynfB gene in E. coli, SAT prevents spermidine accumulation, and the consequent toxic effects, by modifying spermidine to an inert form [8].
To confirm speG activation by VirF, we analysed the activity of the ynfB-speG operon by constructing a translational P ynfB speG -lacZ fusion reporter plasmid (pULS7). The b-galactosidase assay ( Fig. 2A) performed on strain ULS153 pULS7, in the presence or in the absence of a virF-containing plasmid (pMYSH6520 or pMY6520R), confirms that the expression of speG is induced by VirF. The induction is observed only at 37uC, as expected considering the thermodependency of virF expression [26]. A further confirmation of the role played by VirF on speG induction has been obtained in a Shigella background by monitoring speG transcription in a real-time PCR assay. This was performed using S. flexneri strain 2457T (which harbours a frameshift mutation in the speG gene inducing the synthesis of a truncated SAT protein without altering the transcriptional activity of the ynfB-speG operon) and its virF-deleted derivative 2457TFd. As reported in Fig. 2B, the lack of a functional virF gene in strain 2457Fd is paralleled by a two-fold reduction of speG expression, thus confirming the results obtained in the E. coli background.
These results indicate that VirF is able to interfere with the regulation of several genes present on the E. coli chromosome and that this may, in turn, promote the inactivation of genes potentially detrimental to the full expression of the invasive phenotype.

Molecular characterization of the speG locus in Shigella
An in silico analysis, performed on genome sequences currently available on public databases (http://www.mgc.ac.cn/ShiBASE/ and http://www.ncbi.nlm.nih.gov/), highlights that speG is always defective in Shigella and that its inactivation has been obtained by diverse strategies. To verify the widespread nature of speG inactivation in Shigella and analyse the molecular rearrangements that might have led to speG silencing, we sequenced the speG locus of a large collection of Shigella strains (S. flexneri, S. boydii, S. dysenteriae and S. sonnei) isolated over several years in different geographic areas (Table S1). The results are reported in Fig. 3.
The in silico approach indicates that, in S. flexneri strains 2002017, 8401, 2457T and 301, speG inactivation is due to a dinucleotide deletion, which produces a TAA stop codon in the initial part of the coding sequence (see Fig. 3). The presence of the same mutation was confirmed in all S. flexneri strains but two: namely SfZM49 and SfZM43 [30]. Southern analysis revealed that remnants of the speG locus are actually present in SfZM49, while in SfZM43 the speG-containing region is completely lost (data not shown). Interestingly, SfZM43 belongs to serotype 6, previously considered phylogenetically distant from all other S. flexneri serotypes [20]. Moreover, in all Shigella strains analysed we found a non-synonymous mutation in the ynfB gene, responsible for an I14P amino acid substitution (JF737027, JF737028, JF737029, JF737030).
In S. dysenteriae, loss of speG functionality has been attained through two diverse strategies. Two strains, SdZM603 and Sd96.29 (both of serotype 1A), harbour an IS1N insertion (wwwis.biotoul.fr and JF742750, JF742751) within the speG gene, as previously observed for sequenced strain Sd197. The overall genetic organization of strains Sd12, Sd16.81 and Sd4105.65 (all of serotype 2A) is identical to that of E. coli K12, but sequence analysis reveals that these S. dysenteriae strains share several point mutations, out of which only one gives rise to a non-synonymous mutation determining a S56R substitution in the SpeG protein sequence (JF737021, JF737025, JF737026). All S. dysenteriae strains analysed share the same non-synonymous mutation found in the ynfB gene of S. flexneri strains. Finally, we observe that no relevant mutations are located in the ynfB-speG promoter nor in the intergenic region. To check whether the S56R non-synonymous mutation could account for the synthesis of a defective protein, we cloned a fragment containing the entire ynfB-speG operon of S. dysenteriae Sd12 and E. coli MG1655 into the pGEM-T easy vector, thus obtaining plasmids pULS12 and, respectively, pULS11. We then compared the polyamine patterns of an E. coli speG defective strain (ULS117) complemented with pULS12 or with pULS11. As opposed to pULS11, the introduction of pULS12 does not restore production of acetylspermidine (Table 1), confirming that the ynfB-speG operon of S. dysenteriae Sd12 encodes an inactive SpeG protein.
The analysis of the S. boydii Sb227 and CDC 3083-94 sequenced genomes reveals that speG has been inactivated by the insertion of an IS911 element, which carries, within its sequence, another IS element (IS600). A short deletion is present at the IS600-IS911 junction. Among the S. boydii strains analysed only two, Sb481 and Sb483, carry a speG locus with the same genetic organization as strains Sb227 and CDC 3083-94 (JF737022, JF737024). In particular, with the exception of strains Sb51 and Sb485, which harbour a complete deletion of speG locus, in the remaining strains we observe a progressive reduction of the Sb227 speG-IS arrangement: strain Sb411 shows a 365 bp deletion starting from within the hipA sequence and ending within the IS600 sequence (JF737020), while strain Sb484 (JF737023) carries a 1446 bp deletion extending from within the hipA sequence to a position beyond the speG orf terminus (Fig. 3).
As for S. sonnei, the in silico analysis of strain Ss046 indicates that the speG locus has been completely lost. The absence of speG sequences in all strains from our collection (SsIP1-6, SsZM279 and  . Inactivation of the ynfB-speG locus in Shigella spp. has been attained by convergent evolution. The operon on the top is based on the E. coli K12 MG1655 sequenced (http://www.ncbi.nlm.nih.gov/genome). Arrows indicate the orientation of ynfB and speG genes. Point mutation, leading to the Il4P substitution, within the ynfB gene found in most S. flexneri and S. dysenteriae strains analysed is indicated by a white cross. Point mutation, leading to the S56R substitution, detected in some S. dysenteriae strains is indicated by a black cross. All the S. flexneri strains show the interruption of SpeG coding sequence due to a stop codon (7 th ) resulting from a dinucleotide (GT) deletion indicated by two full stops. The comparative analysis of speG sequences from S. boydii strains shows three different structures that may result from at least two deletion steps, which are schematized by dashed lines. Finally, the ynfB-speG locus of S. sonnei is not reported since it has been completely lost. doi:10.1371/journal.pone.0027226.g003 SsZM328) has been confirmed by Southern blot assays (data not shown). These results are consistent with the well-known clonal nature of S. sonnei strains [20].
All together, the observations on the molecular arrangement of the speG gene in Shigella clearly indicate that speG silencing in this microorganism has been attained by convergent evolution. Moreover, our results suggest that this process might have facilitated the adaptation of Shigella to the host environment.

Polyamine pattern in Shigella strains
To ascertain whether in Shigella spp the loss of a functional speG may have altered the polyamine pattern, we analysed the intracellular polyamine level of eight Shigella strains by means of HPLC. To this end, S. flexneri M90T [31] and SfZM49, S. sonnei SsIP3 and SsIP4, S. dysenteriae Sd96.29 and Sd4105.65, and S. boydii Sb483 and Sb485 were selected as representatives of the diverse arrangements of the speG locus ( Fig. 3 and Table S1) and grown in polyamine-free medium. Despite the high homology between Shigella and E. coli, the analysis reveals several relevant differences. In all Shigella strains analysed, putrescine and spermidine predominate, while cadaverine, spermidine and acetylspermidine are absent ( Table 2). The absence of cadaverine in Shigella is well documented [23,32] and is considered as a major pathoadaptive mutation. The absence of endogenous spermine is not surprising since it has been reported also in E. coli [2], which is considered as the commensal ancestor of Shigella [20]. As far as spermidine is concerned, interestingly its concentration in Shigella was found to be approximately 2-to 3-fold higher than in E. coli MG1655. This is almost certainly due to speG inactivation and, consequently, to the lack of conversion of spermidine to acetylspermidine. Finally, the polyamine patterns observed in S. flexneri SfZM49 and S. boydii Sb485 indicate that, despite the presence of specific bands in Southern blot assays (data not shown), speG is inactive in these strains too. Based on these data, the absence of acetylspermidine in Shigella and the consequent spermidine accumulation can be regarded as a new biochemical feature related to the absence of a functional speG gene.

Spermidine accumulation increases resistance to oxidative stress
It is known that polyamines play a role in response to oxidative stress [33,34]. In order to investigate on the effect of spermidine accumulation upon speG inactivation, under oxidative stress in Shigella, we deleted the speE gene, coding for the SpeE protein responsible for spermidine synthetase (Fig. 1), constructing M90TEd, a Shigella M90T derivative unable to synthesize spermidine. We then compared M90T, M90TEd and M90T complemented with a plasmid carrying the entire ynfB-speG operon (pULS37) or the ynfB gene alone (pULS55), for survival on minimal medium agar plates in the presence of H 2 O 2 . Plasmids used to this end are derivatives of the low copy plasmid pACYC184 [35], in order to minimize the copy number effects. Measurement of growth halos after 18 hours at 37uC indicated a higher sensitivity to H 2 O 2 for M90TEd and M90T pULS37 as compared to M90T and M90T pULS55 (Fig. 4). In addition, the oxidative stress resistance was not altered in the M90TEd background by the introduction of pULS37 and pULS55 plasmids (data not shown). This suggests that in M90T spermidine accumulation contributes to increased survival during oxidative stress.
Next, a deeper analysis of the relative survival of M90T, M90TEd and M90T pULS37 under oxidative stress was carried out on strains grown in LB. To confirm that the observed effect was mediated only by speG inactivation, we cloned the speG coding sequence of MG1655 downstream of a tac promoter, obtaining plasmid pULS13, and we also used the M90T pULS13 strain in this analysis. Setting the survival of M90T as 100%, the relative survival of M90T complemented with speG (pULS13) or with the ynfB-speG operon (pULS37) drops to 23% and 31%, respectively ( Fig. 5A top panel). Despite its inability to synthesize spermidine, strain M90TEd exhibited 54% relative survival. Analysis of the polyamine content reveals that the reduced survival of the M90T strains complemented with speG-containing plasmids is paralleled by a low level of intracellular spermidine (Fig. 5A bottom panel). The higher level of spermidine in M90TEd is likely to be dependent on its uptake from the LB medium, which we have measured as containing 2.4 mM spermidine, and may be mediated by the conserved spermidine-preferential uptake system consisting of the PotA-D proteins [36]. Therefore, we repeated the experiments in M9, which we verify to be a polyamine-free medium. Under these conditions, we confirmed that the presence of a functional speG gene reduces survival to oxidative stress and, in addition, we observed that the strain impaired in spermidine synthesis (M90TEd) displays the lowest survival ( Fig. 5B top  panel). Hence, in S. flexneri a direct correlation exists between cellular spermidine levels and oxidative stress resistance. No correlation was observed for the other polyamines involved. Higher putrescine concentration in M90TEd, both in LB and in M9 media, is not related to the outcome of relative survival to oxidative stress (Fig. 5AB). Moreover, the lack of acetylspermidine in M90T and M90TEd strains does not account for the different relative survival to oxidative stress (Fig. 5A, 5B).
To further and definitively confirm the relationship between intracellular spermidine and oxidative stress resistance in Shigella, we analysed the survival of M90TEd grown under oxidative stress in polyamine-free medium after addition of exogenous spermidine. Under this growth condition, the intracellular spermidine level depends exclusively on spermidine uptake. As reported in Fig. 5C, survival decreases with decreasing spermidine concentration, strongly supporting that spermidine and oxidative stress resistance are strictly connected. In E. coli, the addition of exogenous spermidine and putrescine (the spermidine precursor) stimulates the expression of the OxyR and katG protein, both involved in cellular defence against oxidative stress [37]. OxyR is the global regulator of oxidative stress and acts as positive transcriptional activator, among others, of the katG gene encoding hydroperoxidase I, which catalyses the conversion of H 2 O 2 to water and oxygen [33]. Starting from this observation, we can hypothesize that the higher concentration of spermidine in the Shigella wild type strain, compared to that in the speG-complemented strain, could increase the expression of the katG gene (through OxyR induction) and explain the higher oxidative stress resistance exhibited. In order to verify this hypothesis, we decided to check the katG transcriptional activity in M90T and in its speG-complemented derivative (M90T pULS13) by means of a real time PCR assay. Interestingly, katG mRNA transcription results 8 times higher than that observed in the speG-complemented strain. This supports the hypothesis that spermidine accumulation in Shigella strains promotes higher expression of the katG gene, thereby conferring this microorganism an evolutionary advantage in the response to oxidative stress.

The patho-adaptative nature of speG defectiveness in Shigella
It is widely accepted that the oxidative stress response may explain the ability of bacterial cells to survive within macrophages [38,39]. Interestingly, during the first steps of the invasion process Shigella is able to persist within macrophages [40]. We used a wellestablished assay based on infecting BALB/c mice intra-peritoneally, recovering infected peritoneal macrophages and monitoring the survival of intracellular bacteria within in vitro-maintained macrophages over a 72 h period [39,41]. We compared the intracellular survival of S. flexneri M90T with that of isogenic strains containing plasmids carrying the entire ynfB-speG operon (pULS37) or only ynfB (pULS55). No significant difference was observed among strains recovered 8 h after infection, suggesting that all strains are equally able to infect macrophages (Fig. 6). This reinforces previous experimental observations, obtained by plaque assays (data not shown), that indicated no difference in infectivity and spreading among S. flexneri M90T, M90T pULS37 and M90T pULS55 on HeLa cell monolayers. The ability to survive intracellularly 24 h, 48 h, and 72 h after infection, of M90T and M90T pULS55 decreased to a comparable extent, whereas strain M90T pULS37 is significantly more susceptible to macrophage killing from the 24 h time point on (Fig. 6). This indicates that in Shigella restoration of SpeG activity reduces the ability of bacterial cells to withstand hostile conditions within macrophages.
Further evidence supporting this conclusion was obtained by means of an in vitro competitive assay analysing the survival within macrophages of S. flexneri M90T complemented with the entire ynfB-speG operon (pULS37) or only with the ynfB gene (pULS55). Strain M90T and its derivatives carrying pULS37 or pULS55 were grown to OD 600 0.3-0.4, mixed and used to infect a murine macrophage cell line (J774). Bacterial survival was monitored two hours after infection by lysing the macrophages and plating appropriate dilutions on LB plates. To discriminate M90T from its pULS-derivatives, 200 colonies were replicated on LB plates containing tetracycline. As shown in Table 3, when comparing M90T with M90T pULS55 the competitive index (C.I.) corresponds to 1.11 and 1.05 at 1 h and 2 h, respectively, while it drops to 0.74 (1 h) or 0.43 (2 h) when comparing M90T with M90T pULS37 indicating that the M90T strain, in the absence of a functional speG gene, is more competitive for survival in macrophages. All together, these observations indicate that the evolutionary acquired absence of SpeG activity in Shigella confers the bacterium with an increased capability to defy antagonistic host environments. Thus, it can be assumed that the major functional impact of the lack of speG resides in its pathoadaptive significance.

Discussion
The evolution of bacterial pathogens from harmless ancestors mainly depends on the acquisition of virulence gene clusters on plasmids, phages and pathogenicity islands by lateral gene transfer [42,43]. Complementary to this process is the progressive adaptation to a specific niche by pathoadaptive events involving mutations, rearrangements or deletions of genes unnecessary, or even deleterious, for optimal fitness to the new environment [23].  These events usually involve the concomitant arrival or loss of regulatory factors and this might modify the transcriptional profile of the host to a significant extent. In this work, we have analysed the genetic consequences of the uptake of the virF gene by the harmless ancestor of Shigella, the gut commensal E. coli. A crucial step in the evolution of Shigella from E. coli is the acquisition of the large pINV plasmid, which carries the genes required for the invasion of the colonic epithelium [20]. The primary regulator of these genes, the virF gene, encodes an AraC-like transcriptional activator that resides on pINV. The results we report in the present study provide evidence that the activity of VirF is not restricted to the regulation of the virulence system: many other chromosomal E. coli genes are subject to direct or indirect activation by virF.
Comparative sequence analysis of the genomes of Shigella strains, available in current databases, reveals that VirF-activated genes fall into two different groups: one containing genes still conserved in Shigella and the other containing genes which are inactivated or deleted in Shigella. In the first group, among the genes more susceptible to virF induction, we found the genes encoding the heat shock proteins IbpA, GroESL, HtpG, DnaK and Lon. Interestingly, HtpG, which belongs to the HSP90 family, is implicated in the inflammatory response of Shigella in infected mice and is considered a potential subunit vaccine candidate against shigellosis [44]. These observations suggest that VirF, besides operating as a primary virulence regulator, also activates genes whose products contribute to better withstanding of adverse conditions inside the host.
The existence of a group of genes activated by VirF in E. coli, but no longer present or non-functional in Shigella, is an intriguing result. We focused our attention on the speG gene, the best characterized one, whose product is involved in the biosynthesis of polyamines. Putrescine, cadaverine, spermine and spermidine are implicated in several aspects of cellular life. They affect membrane permeability, gene expression, intracellular signalling, oxidative stress resistance, pH stress resistance and apoptosis through noncovalent interactions with nucleic acids or specific interplay with proteins [45]. The SpeG protein is a spermidine-acetyltransferase (SAT), which transfers an acetyl group to either the N-1 or the N-8 position of spermidine. Acetylation converts polyamines, in particular spermidine, into a physiologically inert form. Acetylpolyamines cannot substitute for polyamine in RNA binding, in growth enhancement of E. coli polyamine-deficient mutants nor in the stimulation of in vitro translation [46]. Acetylation has a major impact on spermidine catabolism: the lack of SAT activity produces spermidine accumulation in E. coli [8].
In order to evaluate the extent of speG inactivation in Shigella, we integrated the genomic observations with a specific comparative study of the speG locus of several Shigella strains (Table S1). The speG gene is inactivated in all strains analysed. Inactivating mutations include point mutations and entire gene deletions, suggesting the existence of a strong selective pressure towards the loss of SpeG function in Shigella. The absence of speG hybridization signals in all S. sonnei strains analysed and in some strains of S. flexneri (SfZM53) and S. boydii (Sb51 and Sb485) may represent the final results of several progressive steps. Between the two extremes, i.e. the presence of a complete ynfB-speG operon with a missense mutation and the deletion of the entire speG locus, we observe a series of genetic rearrangements, mainly induced by IS elements. In particular, the results obtained by the analysis of S. boydii nicely exemplify how a progressive erosion of the speG locus has occurred after accumulation of IS sequences next to the speG gene (Fig. 3), leaving a non-functional truncated 39 sequence. The sequence of the speG locus in the S. boydii strains analysed in this work reveals at least three subsequent steps, from an initial IS911 insertion in the speG sequence, strains Sb481 and Sb483, to the complete deletion of the speG locus, strains Sb484 and Sb51.
Since the molecular rearrangements we observed clearly hint at the existence of selective pressure towards the loss of speG functionality, we looked at the consequences of these mutations on the intracellular polyamine balance. The results obtained by considering a pool of eight representative Shigella strains (Table 2) show that a common feature is the absence of spermine, cadaverine and acetylated spermidine, coupled to the presence of putrescine and spermidine. In particular, the endogenous spermidine concentration is 2-to 3-fold higher in Shigella strains as compared to E. coli K12. While the lack of spermine in E. coli is well known [2] and the deficiency of cadaverine in Shigella is documented [32], the absence of the acetylated form of spermidine, determined by speG defectiveness, is a novel finding. The lack of SAT activity, the presence of efficient systems for spermidine and putrescine uptake [36] and the likely absence of known efficient spermidine secretion systems cause accumulation of this polyamine. Moreover, neither a spermidine deacetylating activity nor a spermidine or polyamine oxidase activity can be detected in E. coli or in Shigella. As for the presence of putrescine, spermidine accumulation is known to inhibit ornithine decarboxylase and arginine decarboxylase, both involved in putrescine biosynthesis, maintaining the endogenous putrescine concentration at physiological levels [47].
Besides constituting a new physiological trait of Shigella, how does the abundance of spermidine in this microorganism relate to its virulence specificity? In this context, it is worth stressing that other polyamines are involved in Shigella virulence. In particular, the addition of exogenous putrescine, as well as of methionine and arginine (both implicated in putrescine/spermidine biosynthesis) can restore virulence in S. flexneri mutants that are unable to synthesize modified nucleosides required for tRNA synthesis [48]. More recently, putrescine has been shown to relieve the ornithine repression exerted on Shigella virulence in minimal medium [49]. Overall, during the last years an increasing number of studies related to polyamines in bacteria report new insights about the active role of polyamines during diverse steps of the pathogenic process of different virulent species [50]. Therefore, we asked whether spermidine accumulation, due to speG inactivation, is advantageous for cellular physiology, for the full expression of virulence determinants and for the correct progress of the virulence program. First, we tested the oxidative stress resistance in Shigella since this microrganism experiences a certain degree of oxidative stress within the macrophage cytosol [51] and the role of polyamines in this process, although not yet completely characterized, is well documented [34]. Polyamines are involved in katG expression since they favor the translation of OxyR, a key  regulator of the stress response [33]. By treating wild type speEdefective and speG-complemented S. flexneri strains with H 2 O 2 , we observed a direct correlation between intracellular spermidine concentration and oxidative stress resistance (Fig. 5A, B, C). In this context, we observed that the absence of a functional SpeG in Shigella induces an increase of KatG expression. Moreover, it is worth remembering that cadaverine is reported as the best polyamine providing bacterial resistance to oxidative stress [34], that Shigella is typically a cadaverine-defective microorganism [22,23] and that, while cadaverine is secreted, spermidine is preferentially retained intracellularly [36]. Based on these observations and on the possible functional complementarity among different polyamines in cell [2], the increase of spermidine in Shigella may compensate for the absence of cadaverine in order to maintain an effective response to oxidative stress. As for the impact of spermidine accumulation on the expression of virulence determinants and on the progress of the virulence program, our plaque assay experiments on HeLa cells (data not shown) reveal no difference in infectivity and spreading between S. flexneri and its speG-complemented derivative. On the other hand, the outcome of intracellular macrophage survival assays, performed in mouse peritoneal macrophages (Fig. 6), and of a competitive-infection assay on J774 macrophage cell culture (Table 3), shows a decrease of survival properties in the speGcomplemented Shigella strains. These data suggest that bacterial survival inside macrophages may also depend on the ability of Shigella to exploit the polyamine-mediated neutralization of the oxidative stress to which these bacteria are exposed into the macrophage upon infection.
All together our observations reveal the patho-adaptative nature of speG inactivation in Shigella, suggesting a supportive role of this adaptation in the pathogenicity of Shigella. It is tempting to speculate that, during the evolutionary transition from E. coli to Shigella, the acquisition of virF by means of lateral gene transfer might have caused an increased expression of speG, thus lowering the intracellular spermidine content. This new set up could have hindered the survival strategies of the bacterium within the infected host cells. Hence, speG inactivation would have been favoured in order to increase the intracellular levels of spermidine. This is supported by the presence of an efficient putrescine/ spermidine importer [36] and by the absence of an effective spermidine secretion system in Shigella. It is worth stressing that the recently identified MdtIJ protein complex, belonging to the small multidrug resistance (SMR) family and encoded by the mdtIJ operon, is able to secrete spermidine effectively only when cloned on a multicopy plasmid [52].
Shigella is acquired by oral contamination and is able to cross different parts of the digestive tract. In the intestinal lumen, the major polyamines (putrescine and cadaverine) are produced mainly by bacteria and are mostly free. The existence of a negative putrescine gradient from the jejunum to the ileum has been surmised due to the rapid absorption of luminal polyamines by the intestinal mucosa [53]. On account of these observations, we hypothesize that speG inactivation enables Shigella to compete with the intestinal mucosa for putrescine uptake so, when Shigella reaches the polyamine-poor colon and crosses the epithelial mucosa, it survives inside resident macrophages, even by virtue of the high level of intracellular spermidine synthesized from by the absorbed putrescine. This step is then followed by the induction of macrophage apoptosis and by the invasion of enterocytes, the final target of this pathogen. Other studies are in progress to investigate the involvement of cytoplasmatic spermidine accumulation in different steps of the Shigella's virulence program, taking account of the involvement of this polyamine in gene regulation and of its pro-apoptotic and immuno-modulating properties [54,55].

Bacterial Strains, plasmids and general procedures
The bacterial strains used are listed in Table S1. E. coli strains ULS153 and ULS117 and S. flexneri strains 2457TFd and M90TEd were obtained using the one-step method of gene inactivation [56]. Strain ULS153, carrying a deletion of the lacZ gene, and strain ULS117, carrying a deletion of the entire speG gene, were constructed by transforming MG1655 pKD46 with amplicons obtained using plasmid pKD13 as template and the oligo pairs dlf/dlr or dgf/dgr. The same procedure was used to construct the speE defective S. flexneri strain M90TEd (oligo pair def/der) and the virF defective S. flexneri strain 2457Fd (oligo pair dff/dfr).
b-galactosidase assays were performed as previously described [57] on sodium dodecyl sulfate-chloroform-permeabilized cells grown in LB supplemented with ampicillin. Units of b-galactosidase were calculated by the method of Miller [57]. PCR reactions were routinely performed using Dreamtaq DNA polymerase (Fermentas). Ex taq DNA polymerase (Takara) was adopted to obtain longer transcripts and high fidelity. Oligos used are listed in Table S3. Genomic DNA purifications were performed using the ''mi-Bacterial Genomic DNA isolation kit'' (MetaBion). DNA sequencing was performed by Synergene Biotech.

Plasmid construction
Plasmid pULS7 was constructed by cloning a fragment carrying the ynfB-speG promoter region and the first 35 codons of the MG1655 speG gene into plasmid pRS414, which is a vector suitable for generating protein fusions: the first eight codons of the lacZ gene were removed and a multicloning site has been inserted upstream the lacZ gene [58]. The amplicon obtained with oligo pair pgf/pgr, modified to contain a BamHI site, and MG1655 DNA as template was digested with BamHI and cloned into BamHI linearized pRS414.
Plasmids pULS11 and pULS12, containing the entire ynfB-speG operon, were constructed by cloning into pGEM-T easy a DNA fragment obtained by PCR with the oligo pair pgf/ygt and total DNA of MG1655 or S. dysenteriae Sd12 as template.
In order to construct plasmid pULS37, we subcloned the EcoRI fragment containing the ynfB-speG region from pULS11 to the pACYC184 EcoRI site. pULS55 was obtained by cloning into pGEM-T easy a functional copy of the ynfB gene with its regulatory region, obtained by PCR with the oligo pair pgf/pgr and MG1655 DNA as template. The EcoRI fragment containing the ynfB gene was then subcloned from pGEM-T easy to the pACYC184 EcoRI site. Plasmids pMY6520R and pMY6504R were obtained by re-ligating a HindIII partial digest of pMYSH6520 and pMYSH6504 plasmids in order to delete the virF gene. The loss of virF HindIII fragments was verified by sequencing.
To monitor speG expression independently from ynfB transcription, we cloned the speG gene into pGIP7, a pACYC184 vector carrying a tac promoter and the LacI encoding gene [59]. To this end, a PCR fragment obtained using oligo pair gof/gor and MG1655 DNA as template was digested with BamHI and cloned into the pGIP7 BamHI site, thus obtaining plasmid pULS13.

Polyamine quantification
Polyamines were extracted from suspensions of bacteria with 0.25 M percloric acid containing 5 mM 1,6-diaminehexane as a polyamine internal standard. They were then quantified after derivatization with dansyl cloride and separation by HPLC. The simultaneous fluorimetric determination of intracellular polyamines was performed by reverse-phase high-performance liquid chromatography [60]. Polyamine concentration in the total cellular homogenates was normalized with respect to the corresponding protein content and expressed as nmol/mg of proteins.

Microarray analysis
Bacterial strains MG1655 pMY6504R and MG1655 pMYSH6504 were grown at 37uC in 10 ml LB to OD 600 ,0.6. Cells were immediately chilled and total cellular RNA was isolated by hot phenol extraction [61] and DNase I-treatment for 30 min at 37uC. RNA pellets were suspended in 30 ml diethylpyrocarbonate (DEPC)-treated water. Cy5-dCTP and Cy3-dCTP (GE Healthcare) were used to synthetize labelled cDNA using the direct labelling procedure of the LabelStar TM Array Kit (Qiagen). The E. coli K12-V2 Array (MWG), hybridized to labelled probe pools, were scanned and quantified using a ScanArray lite scanner (Packard Bioscience) and the ScanArray Express software. Two arrays were used and analysed as replicates. Resulting data were processed by Global Lowess normalization and averaged using J-Express software (MolMine AS). We filtered the data to exclude artefacts and low signal spots. Finally, only genes with an error rate lower than 30% and with ratio of 2 and above were considered. Normalized ratio data of microarray experiment is provided in Table S2. All microarray data reported in the manuscript is described in accordance with MIAME guidelines and the data from the experiments are deposited in GEO (accession no. GSE30207).

Real Time PCR
Total RNA was extracted as previously described and cDNA synthesis was performed using the High Capacity cDNA Reverse Transcription Kit from Applied Biosystems in a 20 ml reaction mix containing 20 mg total RNA. Real time quantitative PCR was performed with the aid of a 7300 Real-Time PCR System (Applied Biosystems) in a 30 ml reaction mix containing 2 ml cDNA and Power SYBRHGreen PCR Master Mix (Applied Biosystems). At least three wells were run for each sample. The relative amounts of speG transcript was analysed using the 2 2DDCt method [62] and the results were indicated as a n-fold increase relative to the reference sample. Primers for the mdh transcript, used as endogenous control, and for speG and katG transcripts were designed with the aid of the Primer ExpressH software v2.0 (Applied Biosystems) and experimentally validated for suitability to the 2 2DDCt method. The following oligos pairs were used: mdf/mdr for the mdh gene; rgf/rgr for the speG gene and kgf/kgr for katG gene.

Bacterial susceptibility to oxidative stress
Bacterial susceptibility to oxidative stress was tested as follow: bacterial cultures were grown overnight, diluted in fresh LB or M9 minimal medium and allowed to growth to OD 600 0.6-0.8.
15 ml for each culture were centrifuged and pellets suspended in 1 ml 16 PBS. 1 ml of 16 PBS containing 10 mM H 2 O 2 was added and left to react for 30 minutes at 37uC. The reaction was stopped by adding Catalase to 0.1 mg/ml (Sigma-Aldrich). The number of bacteria surviving the oxidative stress was then quantified by plating aliquots on LB Agar.

Survival assays in mouse peritoneal macrophages
Survival of S flexneri M90T strain in mouse peritoneal macrophages was tested using an in vivo-in vitro infection model as described previously [39]. Briefly, strains M90T and its derivatives M90T pULS37 or pULS55 (Table S1) were grown at 37uC in BHI to OD600,0.4. E. coli strain DH5a grown in LB at 37uC was used as control. The bacteria were harvested by centrifugation and suspended in PBS at 5610 7 cells/ml. Male BALB/c mice (10 weeks old) were infected by intra-peritoneal injection of each strain. After a 6 h infection period, peritoneal macrophages were collected by peritoneal lavage, centrifuged and suspended in Dulbecco's modified Eagle's medium (DMEM) containing 10 mM HEPES, 2 mM glutamine, 10% bovine fetal serum, 16 non-essential amino acids, and 150 mg/ml gentamicin. The cell suspension was dispensed into 24-well tissue-culture plates, incubated at 37uC under 5% CO 2 for 2 h, and bacterial survival was monitored at 24, 48 and 72 h. The animal experiments were performed under a protocol approved by the Institutional Animal Use and Care Committee at Università Cattolica del S. Cuore, Rome, Italy (Permit number: H21, 07/24/ 2008) and authorized by the Italian Ministry of Health, according to Legislative Decree 116/92, which implemented the European Directive 86/609/EEC on laboratory animal protection in Italy. Animal welfare was routinely checked by veterinarians of the Service for Animal Welfare.

Culture of macrophages and bacterial infection
The murine macrophage-like cells J774 (American Type Culture Collection, Manassas, VA) were grown in RPMI 1640 (Gibco) medium containing 10% heat-inactivated fetal bovine serum (Euroclone) and 2 mM L-glutamine at 37uC in a humidified 5% CO 2 atmosphere. For bacterial infection, the cells were seeded in 24-well tissue culture plates (Falcon) at a density of 10 5 cells/ cm 2 and grown over night at 37uC in fresh medium without antibiotics. Bacterial uptake, survival and replication were measured by a gentamicin protection assay [63]. Before infection, cell monolayers were washed twice with phosphate-buffered saline (PBS; pH 7,2), and the medium was replaced by 1 ml of RPMI 1640 supplemented with 10% heat-inactivated foetal bovine serum. In order to produce a competitive infection, M90T and M90T pULS37 (or M90T and M90T pULS55) were used to simultaneously infect J774 monolayers with at a multiplicity of infection of 100 bacteria per macrophage. After 5 min of centrifugation at 900 rpm and a 15 min incubation at 37uC with 5% CO 2 , the infected macrophages were washed twice with PBS. Fresh cell culture medium containing 25 mg/ml of gentamicin was added to kill extracellular bacteria and the cells were incubated further at 37uC for 1 h and 2 h. To determine the number of intracellular bacteria, the cells were washed once with PBS and lysed by adding 0.5 ml of 1% Triton X-100 (Sigma) to each well for 5 min. Samples were mixed, diluted and plated onto LB agar plates to determine the number of CFU recovered from the lysate. The number of intracellular bacteria was determined after 1 and 2 h of gentamicin treatment and compared to bacteria plated at time zero. To calculate the competitive index (C.I.), the ratios of strains M90T pULS55/M90T and of strains M90T pULS37/ M90T recovered from the infected cultures were determined and then normalized by dividing by the corresponding ratio in the initial inoculum.