Modulation of yersinia type three secretion system by the S1 domain of polynucleotide phosphorylase.

Both low temperatures and encounters with host phagocytes are two stresses that have been relatively well studied in many species of bacteria. Previous work has shown that the exoribonuclease polynucleotide phosphorylase (PNPase) is required for Yersiniae to grow at low temperatures. Here, we show that PNPase also enhances the ability of Yersinia pseudotuberculosis and Yersinia pestis to withstand the killing activities of murine macrophages. PNPase is required for the optimal functioning of the Yersinia type three secretion system (TTSS), an organelle that injects effector proteins directly into host cells. Unexpectedly, the effect of PNPase on the TTSS is independent of its ribonuclease activity and instead requires its S1 RNA binding domain. In contrast, catalytically inactive enzyme does not enhance the low temperature growth effect of PNPase. Surprisingly, wild-type-like TTSS functioning was restored to the pnp mutant strain by expressing just the approximately 70 amino acid S1 domains from either PNPase, RNase R, RNase II, or RpsA. Our findings suggest that PNPase plays multifaceted roles in enhancing Yersinia survival in response to stressful conditions.

Bacteria can rapidly modulate their metabolism to enhance their survival during periods of environmental change. For example, the cold shock response of Escherichia coli involves the induction of several genes including cold shock proteins (CSP) 1 as well as exoribonucleases PNPase and RNaseR involved in RNA metabolism (1,2). In addition to E. coli, Bacillus subtilis and Yersinia enterocolitica have been shown to require PNPase for growth in the cold (3)(4)(5). It is believed that PNPase is required for the restart of growth following cold shock by degrading the remaining CSP mRNA molecules (5,6).
Similar to cold-induced responses, interactions with immune cells also triggers bacterial responses that serve to enhance bacterial survival. Viewed in this light, at least some virulence gene products can be thought of as host cell-induced stress response proteins. Several bacterial pathogens secrete virulence factors that mollify anti-microbial killing systems of host cells. Many plant-and animal-interacting Gram-negative bacteria, including the pathogenic Yersiniae, utilize a type three secretion system (TTSS) to inject virulence factors directly from the bacterium into the host cell (7). Yersinia outer membrane protein virulence factors (Yops) are either effector or facilitator proteins, the latter being necessary for the direct injection of the effector Yops into the host cell. YopE, an effector Yop, has GTPase-activating protein (GAP) activity, and disrupts the host cell cytoskeleton (8,9). YopB and YopD have been shown to be required for the directed translocation of effector Yops into the host cell (10,11). Therefore, it appears as though TTSS has evolved to directly address host cell-induced stresses.
Through microarray analysis, a Salmonella enterica pnp mutant strain was recently shown to have altered patterns of gene expression in 3.44% of all genes analyzed including TTSS and fimbriae genes (12). Differential expression of TTSS and fimbriae-encoding genes may account for the differences observed between the wild-type and pnp strains in cell culture and animal infection assays (12). Here, we examined whether Yersinia ribonucleases are involved in host cell-induced stress responses and specifically tested whether PNPase affects the functioning of Yersinia TTSS.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-Oligonucleotide primer sequences used for constructing strains and plasmids are provided under Supplemental Materials. The Y. pseudotuberculosis (YPT) pnp strain was derived from the parental YPIII/pIB100 yopE::gfp (13) using a recombinase-based method (14). In brief, a kanamycin or a chloramphenicol resistant cassette was PCR-amplified by using forward and reverse oligonucleotide primers flanked by 30 nucleotides of complementary target gene sequence. These PCR products were transformed into YPT strains previously transformed with a plasmid encoding the lambda red recombinase. Following transformation of the PCR product and subsequent recombination, the inserted resistance cassette was then excised by virtue of a flanking FRT site. The recombinase-encoding plasmids were then cured, and deletion of the entire pnp coding sequence was confirmed by PCR and immunoblotting. The YPIII/pIB604 strain was used as the TTSS-deficient (yopB) control (10). All Y. pestis (YP) strains originate from the KIM5-3001 strain (15). A KIM5-3001 yopB mutant strain was generated by an allelic replacement strategy as described by Day et al. (16). The YP ribonuclease mutant strains were generated using the same method as described above, and the YP pnp strain was deleted from residues 79 -578. For the Elk-tagged translocation assay (Fig. 7), the KIM5-3001.P39 and P.41 strains (PNPaseϩ/YopB ϩ and PNPaseϩ/YopB, respectively) were used (16). The PNPaseϪ/YopBϩ strain was generated by transferring the pCD1 virulence plasmid from KIM-3001.P39 into the pnp strain described above.
Constructs encoding either the full-length E. coli PNPase or PNPase variants were supplied by Claude Portier (17). Various S1 domains from YP proteins (Fig. 10) were cloned into Bluescript (pSK, Stratagene), which was previously converted into a Gateway-compatible destination vector (Invitrogen).
Cell Infection Assays-Viability assays ( Fig. 1) were performed essentially as described by Bartra et al. (13). YPT wild-type and yopB strains were transformed with the empty control vector pCL-1920, and the YPT pnp strain was transformed with either the control vector pCL-1920, or a pCL-1920 construct encoding the E. coli full-length PNPase. In brief, a murine macrophage-like cell line (RAW 264) was seeded in 24-well plates at densities of 2.5-4.0 ϫ 10 5 per well. Cultures were grown to saturation in either defined TMH (19) medium supplemented with 2.5 mM CaCl 2 at 37°C (Figs. 1A, 5, 6, and 11) or tissue culture medium at 26°C (Figs. 1B, 2, 8, and 10). Similar results were obtained when cultures were grown under either condition. Strains were added to each well at low multiplicity of infections (MOIs) to minimize background. Following removal of unattached bacteria after 30 min, fresh RPMI medium was added to each well. At the indicated times, the number of viable bacteria in each well was determined by first removing the medium followed by lysing the infected cells (but not bacteria) with 500 l of distilled H 2 O, a portion of which was spread on a Luria-Bertani plate, and colony forming units (cfu) were counted 2-3 days later.
For cytotoxicity assays (Figs. 2, 8, 10, and Supplemental Materials Fig.  S5), HeLa cells were seeded in 6-well plates at densities of 1.5 ϫ 10 5 -2.0 ϫ 10 5 in RPMI ϩ10% bovine serum. Strains were added to wells at MOIs as indicated in the figure legends, and cells were monitored for Yersinia-induced cytotoxicity. The degree of cytotoxicity was determined using a defined scoring system described under Supplemental Materials (Fig. S2).
TTSS Assays-In the Elk-tagged translocation assay (Fig. 7), YP strains were transformed with the expression plasmid pYopE-Elk1 that contained yopE sequences (codons 1-129) translationally fused to an elk1 fragment (codons 375-392) derived from pFA2-Elk-1 (16). These transformants were grown to saturation at 26°C in Heart Infusion Broth (HIB) with appropriate antibiotics, and saturated cultures were diluted to an A 620 of 0.15. These diluted cultures were then grown in HIB with appropriate antibiotics at 26°C for 3 h after which the number of bacteria were again normalized and added to HeLa cells at a MOI of ϳ30. After a 3-h infection period, HeLa cells were lysed with sample buffer containing phosphatase and protease inhibitors, and lysates were fractionated by SDS-PAGE. The resulting blots were probed with rabbit anti-phospho-Elk Ser-383 and Elk antibodies, respectively (Cell Signaling). The secondary antibody used was Alexa Fluor 680 (Molecular Probes) that possesses a near-infrared fluorophore, which is excited at 680 nm and emits at 702 nm when scanned on the Odyssey Imaging System (Licor). The digital image obtained by the Odyssey Infrared Imaging System was then quantified using the Odyssey v1.1 software (Licor).
The method of Plano and Straley (18) was employed for the growth restriction assay (Fig. 4). In brief, cultures were grown to saturation in defined TMH medium (19), and strains were diluted in 3 ml of TMH (either without or with 2.5 mM of Ca 2ϩ ) to optical densities (620 nm) of 0.2. Diluted cultures were then grown for 1 h at 26°C and then shifted to 37°C where optical densities at 620 nm were read every hour for a 5-h period.
For flow cytometric analysis of yopE promoter activity (Fig. 5), cultures were grown to saturation in 2 ml of TMH containing 2.5 mM Ca 2ϩ and 5.0 mM MgCl 2 at 37°C. Cultures were diluted to an A 620 of 0.2 in 5 ml of TMH that was prewarmed to 37°C and contained 2.5 mM Ca 2ϩ and 5.0 mM MgCl 2 . After 1 h of shaking at 37°C, the TTSS was induced by adding 5.0 mM EGTA, a calcium chelator. At various times, 300-l samples were removed and placed in prechilled tubes and kept on ice. Bacterial cells were then washed once in Hank's Buffered Saline Solution (HBSS, Invitrogen Life Technologies, Inc.) and resuspended in 300 l of HBSS. Flow analysis was then carried out using a Facscan 488 with an argon laser (BD Biosciences), and mean fluorescent intensities (MFIs) were analyzed using the Cellquest Pro software (BD Biosciences).
In the secretion assay shown in Figs. 6 and 11, cultures were grown exactly the same way as for the above flow cytometric analysis except that some cultures did not receive EGTA. At the indicated time points, 1-ml samples were removed and placed in prechilled tubes and were kept on ice. Cells and supernatants were fractionated by centrifugation at 12,000 ϫ g for 5 min. 500 l of the resulting supernatant were removed and precipitated with 55 l of trichloroacetic acid overnight on ice after which samples were centrifuged 12,000 ϫ g for 10 min at 4°C. The resulting pellets were resolved by SDS-PAGE and analyzed by immunoblotting with a rabbit polyclonal mix of anti-YopE and anti-YopD antibodies (Fig. 6) or anti-YopE alone (Fig. 11). Cell pellets were resuspended directly in sample buffer and were analyzed together with the secreted proteins.
Cold Growth Assays-For the experiment shown in Fig. 9, strains were grown to saturation at 26°C in HIB with the appropriate antibiotics, diluted 10 Ϫ1 , and grown for 1 h at 26°C. Cultures were then diluted 10 Ϫ5 , and 50 l were plated on LB plates containing antibiotics, if necessary, and were incubated at 4 -6°C for 10 days.

Viability of PNPase and Various Ribonuclease-deficient
Strains in an Infection Assay-The viability of a YPT PNPasedeficient mutant strain was tested in a cell infection assay A, YPT wild-type, yopB, and pnp strains were transformed with either the control vector pCL-1920 or pCL-1920 encoding E. coli PNPase (pnp/PNPϩ). Transformants were added to tissue culture wells containing RAW 267 mouse macrophages at MOIs of 0.2-0.5. Following a 30-min attachment period, excess bacteria were removed, and the number of viable cell-associated bacteria was determined by plating either immediately (0 h) or 6 h later. Three independent wells per strain were analyzed, and the average fold-increases over the 6-h infection period for each strain are shown graphically. Using the Student's t test and calculating for unequal variance, p Ͻ 0.01 for the differences between the wild type and yopB, wild type and pnp, and pnp and pnp/PNPϩ. For an independent YPT viability assay showing actual numbers of cfu per well see Supplemental Materials ( Fig. S1A). B, YP wild-type, yopB, and pnp strains were assayed in the same manner as the YPT strains shown in A with the exception that a 8-h infection period was used. p values Ͻ0.01 were calculated when comparing differences between wild type and yopB, and wild type and pnp. For an independent YP viability assay showing actual numbers of cfu per well see Supplemental Materials ( Fig. S1B). using a mouse macrophage-like cell line. Previously, it had been demonstrated that this assay is sensitive to well characterized virulence determinants of YPT (13). There was a 14.8fold increase in cfu of the wild-type YPT strain between the start of the infection and the 6-h end point (Fig. 1A). In contrast, there was a decrease in the number of cfu of the YPT TTSS-deficient yopB strain during the 6-h infection period. The cfu fold-increase of the YPT pnp strain was significantly reduced when compared with the wild-type strain in this assay (p Ͻ 0.01). Wild-type-like levels of viability were restored to the YPT pnp strain by supplying E. coli PNPase in trans ( Fig. 1A; p Ͻ 0.05 between the pnp and pnp/PNPϩ). For an independent YPT viability assay showing actual numbers of cfu recovered per well see Supplemental Materials ( Fig. S1A).
We also evaluated whether PNPase contributed to YP survival vis-à -vis cultured macrophages. There was a 15.5-fold increase in cfu of the wild-type YP strain between the start of the infection and the 8-h end point (Fig. 1B). There was a small increase (1.3-fold) in the number of cfu of the YP yopB strain during the 8-h infection period. Similar to what was observed in YPT, the cfu fold-increase of the YP pnp strain was significantly less than the wild-type strain (p Ͻ 0.01). For an independent YP viabilty assay represented by actual numbers of cfu recovered per well see Supplemental Materials ( Fig. S1B). Differences observed in viability between the various strains shown in Figs. 1A, 1B, S1A, S1B were not a result of differential binding to host cells since approximately the same number of viable bacteria were recovered after the initial 30-min attachment period. Collectively, these data show that both the YPT and YP pnp mutant strains behaved similarly relative to their respective wild-type and TTSS-deficient strains in a viability-based infection assay.
To test whether the loss of viability of the PNPase strain in the infection assay is caused by a general loss of ribonuclease activity, we tested YP strains mutated in RNaseII, RNaseIII, and RNaseR in our infection assay. We found that the performance of these mutant strains were indistinguishable from that of the isogenic wild-type strain (Supplemental Materials, Fig.  S1B, and data not shown).
PNPase and Cytotoxicity of Yersinia-infected HeLa Cells-Since PNPase appeared to play a specific role in Yersinia viability in the infection assay, we examined whether PNPase was necessary for Yersinia to affect the morphology of HeLa cells. Yersinia enhances its viability vis-à -vis the host cell by employing its TTSS to disrupt the cytoskeleton of the infected cell (13,20). HeLa cells infected by the wild-type YPT and YP strains retract following a 2-3-h infection period (Fig. 2, A and B respectively). Yersinia-mediated HeLa cell cytotoxicity was quantified using a defined scoring system ( Fig. 3 and Supplemental Materials, Fig. S2). In contrast, the morphology of cells infected with the TTSS-deficient yopB strains appeared unaltered and resembled uninfected cells. Cells infected with the pnp strains did display a cytotoxic effect indicating that these strains were delivering effector Yops into host cells. However, the kinetics of the HeLa cell cytotoxicity induced by the pnp strain appeared delayed relative to the wild-type-infected cells. Transforming the YP and YPT pnp strains with a plasmid encoding E. coli PNPase restored the HeLa cell cytotoxicity to levels observed in wild-type-infected cells (Fig. 2). These results indicate that the pnp strain possesses a translocationcompetent TTSS, which may function suboptimally. A compromised TTSS likely accounts for the decreased viability of the pnp strains in the infection assays shown in Fig. 1.
PNPase and TTSS Functioning-Based upon the resemblance between the pnp strains and the TTSS-deficient yopB strains in the above assays, we further analyzed TTSS func-tioning in the YP and YPT pnp strains. Normally, Yersinia possessing an intact TTSS arrest their growth when shifted from their optimal growing temperature of 26 -37°C in medium lacking calcium; this is referred to as "growth restriction" (21). Shown in Fig. 4, similar to the wild-type strain, the YP pnp strain ceased growing when shifted to 37°C in medium lacking calcium in contrast to YP strains with disregulated TTSSs that either constitutively secrete Yops in both the absence and presence of calcium termed calcium blind (22) or constitutively grow in either the presence or absence of calcium, calcium-independent. 2 These data indicate that PNPase is not required for TTSS-dependent growth restriction.
We then tested whether PNPase plays a role in Yop virulence factor expression and export by the TTSS. We measured Yop secretion in YPT since YP possesses a unique extracellular protease, Pla that rapidly degrades secreted Yops (23). An additional feature of the YPT strains used in Fig. 1A is that they contain a yopE::gfp transcriptional fusion gene that allows for yopE promoter activity measurements without affecting the expression or function of YopE (13). Strains were grown at 37°C in the presence of calcium, and the expression and secretion of Yops were induced by adding a calcium chelator to the medium. For evaluating yopE promoter activity in the wildtype and pnp YPT strains, samples were removed from the 2 G. V. Plano and F. Ferracci, unpublished data. medium at various times shortly after induction and analyzed by flow cytometry. As previously reported (13), the wild-type strain had a detectable increase in yopE promoter-driven GFP signal within 5 min following induction that steadily increased to 5.5-fold compared with initial levels by 60 min (Fig. 5). In the pnp strain, initial GFP levels were comparable to those observed in the wild-type strain, and following TTSS induction, there was a similar increase in GFP signal (Fig. 5). These data indicate that PNPase does not grossly affect either steady-state or inductive yopE promoter activity.
For the direct analysis of cellular Yop protein levels, YPT strains were grown and induced in the same fashion as shown in Fig. 5. As measured by immunoblotting, YopE protein levels increased in both wild-type and pnp cells 20 min following induction whereas cellular YopD protein levels changed very little in either strain during the course of the experiment (Fig.  6). These data closely matched the results obtained in the flow cytometric analysis (Fig. 5) in that, essentially, no differences in steady-state or inductive Yop protein levels were observed between the wild-type and pnp YPT strains.
To test whether PNPase affected Yop secretion, YopE and YopD protein levels were determined in the culture supernatants from the above samples. Prior to induction, low levels of YopE and YopD were detected in the culture supernatant from the wild-type strain (Fig. 6, lanes 1 and 2). By either 5 or 20 min following induction, there were several fold-increases in YopE and YopD protein levels, respectively, in the wild-type culture supernatants (lanes 5 and 9). In stark contrast, very little YopE protein, and no detectable YopD protein were observed in the cultural supernatant from the pnp strain prior to induction (lanes 3 and 4). In fact, YopE and YopD protein were not detectable in the pnp culture supernatants until either 5 or 20 min, respectively (lanes 7 and 11). Although YopE protein was detected 5 min after induction, the levels were severalfold lower than that present in the culture supernatant from the wild-type strain (compare lanes 5 with 7 and 9 with 11). However, by 60 and 120 min following induction, there were little appreciable differences in YopE and YopD protein levels between the wild-type and pnp culture supernatants (data not shown). These data suggest that PNPase is involved in configuring Yersinia TTSS in such a way that maximal Yop effector proteins are exported from the bacterium upon induction.
A translocation assay was employed to test whether the reduced initial Yop secretion rates in the pnp strain translated into reduced Yop injection into host cells. A plasmid encoding a hybrid protein consisting of YopE (residues 1-130) and a 40residue Elk tag was transformed into the YP wild-type, yopB, and pnp strains. The relative level of YopE-Elk translocation into HeLa cells can be determined by probing for Elk phosphorylation that only occurs within the host cell (16). Phosphorylated YopE-Elk was readily detected in lysates prepared from cells infected with the wild-type strain after a 3-h infection period (Fig. 7A, lanes 1 and 2). As described previously by Day et al. (16), phosphorylated YopE-Elk was not detected in lysates prepared from cells infected with the TTSS-deficient yopB mutant strain (lanes 3 and 4). In lysates prepared from cells infected with the pnp strain, phosphorylated YopE-Elk was detected after a three hour infection period but at levels greatly reduced compared with levels in lysates from wild-type-infected cells (lanes 5 and 6; signal quantification shown in Fig.  7B). These data are consistent with what was observed in the secretion and infection assays and further supports a model in which PNPase is required for the optimal functioning of Yersinia TTSS following host cell contact.
PNPase Determinants Required for Affecting TTSS Functioning-The exoribonuclease PNPase is comprised of several recognizable domains including 2 distinct RNA binding domains as well as 2 catalytic centers (17,24). To determine whether any of these specific domains are required for PNPase effect on TTSS functioning, a panel of plasmids encoding E. coli PNPase variants (17) were transformed into the Yersinia pnp strains. PNPase variants tested included a PNPase R100D that is deficient in all of the enzymatic activities associated with PNPase, as well as a PNPase deltaKH, and PNPase deltaS1 variants that have internal deletions in the KH and S1 RNA binding domains, respectively. By immunoblotting, it was found that all of the PNPase variants were stably expressed in YP (Fig. 8, panel A, inset). In fact, PNPase R100D , PNPase delta KH, and PNPase delta S1 were expressed at notably higher levels than the wild-type PNPase; this observation was consistent with the pre- Fig. 2A were quantified using a scoring system in which all cells from two fields were evaluated (200 cell minimum) based upon the severity of the morphological alteration of the infected HeLa cell. Definition of the scoring system is provided in Supplemental Materials (Fig.  S2). B, cytotoxicity of the YP-infected HeLa cells shown in Fig. 2B were quantified as described above and in Supplemental Materials. viously reported finding that the PNPase catalytic and RNA binding domains are involved in autoregulation of PNPase expression (17,25).

FIG. 3. Quantification of Yersiniainduced HeLa cell cytotoxicity. A, cytotoxicity of the YPT-infected HeLa cells shown in
As documented above, wild-type E. coli PNPase fully complemented the YP and YPT pnp strains in the viability-and cytotoxicity-based infection assays. Surprisingly, HeLa cells infected with the YP and YPT pnp strains expressing the catalytically inactive PNPase R100D displayed levels of cytotoxicity comparable to cells infected with pnp strains expressing wildtype PNPase (Fig. 8 and data for YPT not shown). In fact, we consistently observed slightly higher levels of cytotoxicity in cells infected with the PNPase R100D -expressing strain. Additionally, in the viability assay, the PNPase R100D -expressing strain was as viable as the pnp strain expressing wild-type PNPase (data not shown). Like the PNPase R100D -expressing strain, the YP pnp strain expressing the PNPase deltaKH variant caused similar levels of cytotoxicity in HeLa cells as the levels observed in cells infected with YP-expressing wild-type PNPase (Fig. 8).
In contrast to the catalytically inactive and KH-deleted PNPases being fully active in our infection assay, reduced levels of cytotoxicity were observed in cells infected with the pnp strain expressing the PNPase delta S1 variant. In fact, the level of cytotoxicity observed in cells infected with PNPase del -taS1-expressing YP was essentially equivalent to the levels observed in cells infected with the YP pnp strain (Fig. 8). Collectively, these data indicate that the enzymatic activity and the KH RNA binding domain are not required for PNPase affect on the TTSS. However, our results suggest that PNPase S1 domain is required for TTSS functioning.
PNPase Determinants Required for Low Temperature Growth-We were surprised by our finding that the PNPase effect on Yersinia TTSS was independent of PNPase ribonuclease activity. This observation prompted us to test whether the phenomenon of PNPase enhancing growth at low temperatures was also independent of its ribonuclease activity, an issue that to the best of our knowledge has not been addressed. Similar to what has been previously reported for E. coli, B. subtilis, and Y. enterocolitica (3)(4)(5), PNPase is clearly required for YPT and YP to grow at low temperatures ( Fig. 9A; data for YP not shown). The YPT strains expressing the various PNPases described above were also tested for their ability to grow at low temperature. Plasmid-encoded PNPase of E. coli was able to enhance the ability of the YPT pnp strain to grow at low temperatures (Fig. 9B). In stark contrast, the pnp strain expressing the catalytically inactive PNPase R100D remained clearly defective for growth at low temperature. The PNPase delta S1-and PNPase delta KH-expressing strains were also compromised in their ability to grow at low temperature although not to the same degree as the PNPase R100D -expressing strain (Fig. 9B). These data indicate that a catalytically active PNPase is required for Yersinia to grow at low temperature.
The S1 Domain and TTSS Functioning-To address whether the S1 domain alone was sufficient to complement the pnp strain in regard to TTSS functioning, the sequence encoding the YP 69-amino acid S1 domain from PNPase was cloned into an expression vector and transformed into the YP pnp strain. Strikingly, HeLa cells infected with the S1-expressing strain displayed levels of cytotoxicity that were similar to those observed in cells infected with strains expressing full-length PNPase (Fig. 10). The pnp strain expressing PNPase's S1 domain alone induced as severe a degree of HeLa cell cytotoxicity as did the pnp strain expressing the catalytically inactive FIG. 5. yopE promoter activity in the wild-type and pnp YPT strains. A, the wild-type and pnp strains (solid and dashed lines, respectively) were grown to mid-log phase at 37°C in defined medium containing calcium, at which time EGTA was added. In these strains, gfp expression is under the control of the yopE promoter, and GFP levels were determined by flow cytometry at the indicated time points following the addition of EGTA. B, mean fluorescent intensities of data shown in A.  Fig. S5). To similarly test S1 domains from other proteins, we cloned S1 domains from YP proteins ribosomal S1, RNase II, and RNase R. These various S1 domains had similar effects on YP TTSS as the S1 domain from PNPase (Fig. 10). As might be expected, expression of S1 domains did not restore low temperature growth to the YPT pnp strain in an assay similar to the one shown in Fig. 8 (data not shown). The data suggest that these various S1 domains share some structural attribute that is sufficient to enhance the activity of Yersinia TTSS.
To determine whether S1 domains affected the yersinial TTSS, the YPT pnp strain was transformed with Bluescript vectors containing the cloned S1 domains from either PNPase or RNaseR (see Fig. 10). We used the YPT strains here for the same reason as in the experiment shown in Fig. 6. Samples were collected in the same manner as in Fig. 6 in that they were removed from the cultures as quickly as possible after the addition of the TTSS inducer EGTA. YopE levels in the whole cell pellets of all strains were similar (Fig. 11). As shown previously in Fig. 6, YopE levels in the supernatant fraction of FIG. 7. Translocation assay using a YopE-Elk hybrid protein. A, YP strains either expressing or not the PNPase and/or YopB proteins (see "Experimental Procedures") were transformed with a YopE-Elk-encoding plasmid. The resulting transformants were used to infect HeLa cells at a MOI of 50 for 3 h. Infected cells were then lysed, lysates were resolved by SDS-PAGE, and the resulting blots were probed with either antibodies specific for phosphorylated Elk or total Elk. B, signals were quantified by the Odyssey Infrared Imaging System, and normalized values were determined by dividing the samples phospho-Elk signal by its total Elk signal (absolute values provided in Supplemental Materials).
FIG. 8. PNPase determinants required for YP-induced HeLa cell cytotoxicity. Cells in A were infected with the YP pnp strain at a MOI of 100. Shown in B-E are cells similarly infected with the pnp strain transformed with plasmids encoding the following E. coli PNPase variants: B, wild-type PNPase, C, catalytically inactive PNPase R100D, or PNPase deleted in either the KH or S1 RNA binding domains, D and E, respectively. The above photographs were taken after 2 h of infection at 37°C. A (inset) endogenous PNPase protein levels in the wild-type and pnp strain (lane A) as well as levels of plasmid-encoded PNPases from the strains shown in the corresponding B-E were determined by immunoblotting. Similar results were obtained for YPT strains expressing PNPase variants.
FIG. 9. PNPase and cold growth of YPT. A, the YPT wild-type and pnp strains were grown for 1 h to mid-log phase, plated on Luria-Bertani medium and incubated at 5°C for 10 days. B, the YPT pnp strain shown in A was transformed with plasmids encoding either the wild-type, PNPase R100D , PNPase delta S1, or the PNPase delta KH variants. The resulting transformants were analyzed as in A except that transformants were plated on antibiotic-containing medium. the wild-type strain was severalfold greater than the YopE levels observed in the supernatant fraction from the pnp strain (compare lanes 1 and 3). Strikingly, the expression of just PNPase S1 domain restored wild-type-like YopE secretion to the pnp strain (lanes 5 and 6). An even more pronounced effect on YopE secretion was observed in the pnp strain expressing the RNaseR S1 domain (lanes 7 and 8). These data, together with the data shown in Fig. 8, indicate that overexpressed S1 domains are sufficient to enhance Yop secretion and translocation by the pathogenic Yersiniae.

DISCUSSION
The yersinial TTSS is charged with rapidly delivering Yop effector proteins into the host cell where they modulate cellular processes that normally serve to limit bacterial viability. In order to outpace the microbial killing responses of the host cell, the Yersiniae possess both a preformed pool of Yop effectors for immediate injection, as well as the ability to quickly synthesize Yops following host cell attachment (26). Yersinia survival vis-à -vis the host cell is dependent on the proper functioning of both of these levels of regulation.
We present evidence here that the exoribonuclease PNPase is required for the proper functioning of Yersinia TTSS. Functional studies revealed that pnp mutant strains had reduced viability in infection assays that was likely due to a malfunctioning Yop delivering system. Despite the pnp strain having similar Yop steady-state expression levels as the wild-type strain, we found that the pnp strain was defective in rapidly exporting Yop effector proteins upon both TTSS induction in culture as well as injecting Yops into host cells. Thus, it appears that PNPase is somehow required for the Yersiniae to rapidly deliver preformed Yops into the host cell. This defect could conceivably be caused by either fewer TTSS complexes present in the bacterial membrane or to qualitative alterations resulting in a suboptimally functioning TTSS. Whatever the defect, the TTSS in the pnp strain appears to eventually function normally following its induction. For an individual bacterium, however, by that time the host cell has had ample opportunities to initiate and complete its microbial killing program.
Ribonucleases have been shown to be involved in bacterial virulence. In a transposon mutagenesis screen, a Shigella flexneri was isolated that displayed reduced host cell invasion. The disruption was mapped to a locus that was later identified as the gene (vacB/rnr) encoding RNaseR (27,28). More pertinent to this study and as mentioned in the introduction, a pnp mutant of Salmonella enterica was found to have a disregulated TTSS (12). In contrast to the findings presented here though, Clements et al. (12) found that a lack of a fully functional PNPase in Salmonella correlated to increased TTSS activity. In the Salmonellae, TTSS activity mediates host cell invasion and intracellular proliferation whereas in the Yersiniae, TTSS activity results in the diametrically opposite effect of blocking the host cell from internalizing surface-bound bacteria (7). It is unclear why or how PNPase exerts opposite effects on the Yersiniae and Salmonellae TTSSs. It is interesting to note though, that for both genera, disrupting the pnp locus results in an apparent increase in the levels of bacterial internalization. In the case of the Yersiniae, however, this does not bode well for its viability (29).
How does PNPase affect the functioning of Yersinia TTSS? One possible mechanism could be that the pnp cells suffer from a general metabolic sickness. Previously, it has been shown that an E. coli pnp strain had slightly slower doubling times compared with the isogenic wild-type strain (30). Similarly, when grown in some media, we observed a slightly slower growth rate of the YP pnp strain when the strains were tested FIG. 10. S1-mediated cytotoxicity of YP-infected HeLa cells. A, cells were infected with either the wild-type or pnp YP strains at a MOI of 100 (panels A and B, respectively). Shown in C-F are cells similarly infected with the pnp strain transformed with plasmids encoding the ϳ70-amino acid S1 domains from the following proteins: C, PNPase; D, ribosomal protein S1; E, RNase II; and F, RNase R. The above photographs were taken after 2 h of infection at 37°C. Cytotoxicity of the corresponding panels shown in A were quantified and described in Supplemental Materials (Fig. S4).
FIG. 11. S1-mediated effects on YPT YopE expression and secretion. The wild-type and pnp strains used in Fig. 5 were transformed with either an empty Bluescript vector (lanes 1-4) or a Bluescript vector expressing either the PNPase S1 domain or RNaseR S1 domain (S1 PNPase or S1 RNaseR , respectively). All strains were grown and induced as described in the legend to Fig. 4 or left uninduced. Samples were removed as quickly as possible after the addition of the TTSS inducer EGTA (ϳ20-s time point), and YopE levels from both whole cell and supernatant fractions were determined by immunoblotting.
at Yersinia optimal growth temperature (data not shown). However, when the YP pnp strain was grown in defined TMH medium, no differences were observed between its growth rate and that of the wild-type strain (Fig. 4). A slightly slower growth rate could, in theory, contribute to the reduced viability of the pnp strains shown in Fig. 1. This explanation, though, we find unlikely since the PNPase R100D -expressing YP cells, like PNPase-deficient cells, displayed a slightly reduced growth rate when grown in the same conditions. Despite this defect, however, the PNPase R100D -expressing strain performed as well as, if not better than, the wild-type strain both in the viability assay (data not shown) as well as the cytotoxicity assay ( Fig. 8 and Supplemental Materials, Fig. S5). Additionally, Clements et al. (12) found that the growth rates of the wild-type and pnp strains of Salmonella were identical when tested at 37°C. Therefore, although full-length, enzymatically active PNPase is clearly required for optimal growth of Yersinia at low temperature (see below), at higher temperatures it appears to be dispensable for maximal growth in most laboratory conditions. At low temperatures, or in cells recovering from a cold shock, PNPase plays a central role in reprogramming cellular metabolism (4,5,31). Following a cold shock, restart of growth in both E. coli and Y. enterocolitica is preceded by PNPase-mediated degradation of transcripts encoding socalled CSPs (5,31). It is believed that PNPase might act to free ribosomes that are inactivated by bound CSP-encoding transcripts. In fact, our data support this model in that PNPase catalytic activity was found to be essential for YPT cold growth ability (Fig. 9B). CSPs, which consist of a 5-stranded anti-parallel ␤-barrel (32), share structural similarity to the S1 domains found in PNPase and the ribosomal protein S1 and have been shown to melt out mRNA secondary structures that form at lower temperatures (33). Xia and Inouye (34) reported that a quadruple deletion of 4 of the 9 CSPs rendered E. coli cold-sensitive, and that overexpressing any 8 of the 9 CSPs or the S1 domain from E. coli PNPase alone enabled the quadruple-deleted strain to grow in the cold. Thus, several proteins with bona fide S1 domains as well as proteins that resemble S1 domains (e.g. CSPs) are involved in low temperature stress responses in Gram-negative bacteria.
In this report, we show that PNPase S1 domain is required for optimal growth of YPT at low temperature (Fig. 9). PNPase KH RNA-binding domain, and to a greater extent, its catalytic activity, were also found to be important determinants for its enhancement of low temperature growth. Unexpectedly, the catalytic and KH domains of PNPase were not required for its effect on Yersinia TTSS ( Fig. 8 and Supplemental Materials). Even more surprisingly, just the 69 amino acid S1 domain of PNPase complemented the TTSS deficiency of the pnp mutant Yersinia strains ( Fig. 10 and Supplemental Materials). In contrast, expression of the S1 domain by itself did not restore low temperature growth to the YPT pnp strain (data not shown) indicating further the necessity of PNPase catalytic activity during this condition. The fact that S1 domains from other proteins could likewise complement the pnp strain (Fig. 10) suggests that there is a general feature of these peptides, at least when expressed at relatively high levels, that positively affects Yersinia TTSS. This latter finding is puzzling given the fact that although the S1 domains from various proteins are structurally similar, they differ substantially in their surface residue composition that is thought be the basis for their specificity for their respective cognate RNAs (32).
Another issue that is unclear is why other S1 domains, as they naturally occur in their native contexts in a variety of other proteins (e.g. RNase R, RNase II, etc.), do not functionally complement the pnp strain in our infection assays. Is there a unique feature of PNPase S1 domain in the context of the entire PNPase, or are these other S1 domains not present at sufficient levels? Our S1-related data may offer an as of yet undiscovered aspect of the process by which the Yersiniae enhance their viability by quickly injecting virulence factors into the host cell.