A bacterial secreted translocator hijacks riboregulators to control type III secretion in response to host cell contact

Numerous Gram-negative pathogens use a Type III Secretion System (T3SS) to promote virulence by injecting effector proteins into targeted host cells, which subvert host cell processes. Expression of T3SS and the effectors is triggered upon host cell contact, but the underlying mechanism is poorly understood. Here, we report a novel strategy of Yersinia pseudotuberculosis in which this pathogen uses a secreted T3SS translocator protein (YopD) to control global RNA regulators. Secretion of the YopD translocator upon host cell contact increases the ratio of post-transcriptional regulator CsrA to its antagonistic small RNAs CsrB and CsrC and reduces the degradosome components PNPase and RNase E levels. This substantially elevates the amount of the common transcriptional activator (LcrF) of T3SS/Yop effector genes and triggers the synthesis of associated virulence-relevant traits. The observed hijacking of global riboregulators allows the pathogen to coordinate virulence factor expression and also readjusts its physiological response upon host cell contact.

Introduction Protein secretion plays a pivotal role in the interaction between pathogenic bacteria and their hosts. Pathogenic bacteria utilize different highly sophisticated secretion systems (Type I-VII) to translocate proteins across membranes into eukaryotic target cells in order to manipulate host cell functions and disrupt host homeostasis and immune defenses [1][2][3]. Of these, the type III secretion system (T3SS) is a key virulence factor in many plant, animal and human pathogenic bacteria that contributes to bacterial survival and colonization [1,4]. T3SSs are contact-dependent secretion systems. An intimate contact between the pathogen and the eukaryotic target cell is the signal for the bacterium to induce expression of the secretion machinery and its secreted substrates, the so-called effector proteins, as well as the secretion procedure itself. T3SS gene expression and secretion is activated by growth conditions, certain chemicals, and environmental signals (temperature, pH, oxygen availability, host-associated signals) that mimic host cell contact, and by complex feedback control mechanisms [1,2,4]. One of the best-studied systems is the Yersinia T3SS [5,6].
All human pathogenic Yersinia species, Y. pestis, Y. pseudotuberculosis and Y. enterocolitica, share the ability to survive and proliferate under adverse conditions in lymphatic tissues of their mammalian hosts. The Ysc T3SS and the secreted and translocated effector proteins, the Yops, protect yersiniae from phagocytosis by professional phagocytes. All T3SS-associated genes are encoded on the common 70 kb virulence plasmid (pYV/pCD1) [7,8]. The T3SS forms a complex apparatus of about 20-25 proteins-the injectisome-that functions as a molecular syringe to inject bacterial effector proteins into targeted host cells [6,8,9]. Pathogenic yersiniae use the T3SS to secrete and translocate seven Yop effector proteins mainly into neutrophils and macrophages during infection to facilitate phagocytic avoidance, promote systemic spread, manipulate the inflammatory responses and control cell death programs [5,[10][11][12]. Translocation of Yops and expression of the T3SS/yop genes is strictly regulated and substantially increased after the bacteria established contact with their target cells [13,14]. Transcription of the Yersinia T3SS/yop genes is intimately coupled to the activity of the T3 secretory machinery, i.e. transcription is strongly activated when the machinery is active. Growth under Ca 2+ -limiting conditions at 37˚C mimics the host cell contact, a phenomenon known as the low calcium response (LCR), and is commonly used as a surrogate for the contact signal to convert the injectisome to a secretion competent state [15]. However, the underlying mechanisms of host cell contact-induced signal transduction and T3SS/yop induction remains enigmatic.
Some studies reported that YopD, a pivotal structural component of the translocation pore for direct delivery of Yop effectors into the target cells [16][17][18][19], acts as a sensing device for cell contact and is a negative regulator of T3SS/Yop in the absence of an inductive signal [17,[20][21][22][23][24]. It is further known that (i) loss of YopD causes an increase of the virulence plasmid copy number of pYV [25], and/or (ii) that YopD interacts with untranslated leader segments of T3SS/yop transcripts and blocks T3SS/Yop synthesis [22,23]. These data suggested to us that YopD might be crucial for host cell contact-dependent T3SS/yop induction. To gain a better understanding of the underlying molecular mechanism, we first identified regulatory components implicated in triggering effector secretion upon cell contact and characterized how YopD interferes with this process. This revealed a novel virulence strategy in which a pathogen uses a translocator protein (YopD) to coopt important global riboregulators to control the master regulator (LcrF) of T3SS/effector genes (ysc/yops) in response to host cell contact.

Host cell contact-mediated induction of the Ysc/Yop T3SS occurs through LcrF
Host cell contact has previously been reported to be a potent signal that triggers synthesis of Yop effector proteins of Y. pseudotuberculosis [14,26,27]. To address whether host cell induction is promoted through upregulation of the main T3SS/Yop regulator LcrF, we monitored expression of the lcrF and the LcrF-dependent adhesin yadA fused to gfp or luxCDABE in bacteria infecting human epithelial cells. Cell-associated bacteria were characterized by a strong induction of the reporter fusions upon cell contact at 37˚C, but also at 25˚C when ysc/yop and yadA are usually fully repressed (Fig 1A and 1B). In agreement, a strong increase of LcrF and YadA levels were detectable in bacteria bound to host cells (S1A Fig). In contrast, no or only a very weak signal was detectable from bacteria attached to the wells (Fig 1A and 1B), or when other control fusions were used (S1B Fig).
Immediate contact-triggered induction of lcrF expression and delayed upregulation of yadA ( Fig 1B) suggested that cell contact induction occurs through induction of lcrF. In fact, induction of yadA upon host cell contact failed in a ΔlcrF mutant and in a virulence plasmidcured strain (pYV -) (Figs 1C and S1C). Host cell-contact induction could be reconstituted by integration of the yscW-lcrF locus into the Y. pseudotuberculosis chromosome in the absence of pYV. This indicated that one or more chromosomally-encoded regulator(s) are implicated in cell contact-triggered lcrF induction.

The carbon storage regulator CsrA is required for cell contact-mediated lcrF expression
Our previous studies revealed that the pYV-encoded yadA and T3SS/yop genes are differentially regulated in the absence of the carbon storage regulator CsrA [28,29]. This implied that their common regulator LcrF is likely the target for CsrA regulation. CsrA is a regulatory RNA-binding protein, which together with one or more non-coding RNAs (in Yersinia named CsrB and CsrC) constitute the Csr/Rsm system. This post-transcriptional control system regulates a large set of target transcripts involved in diverse physiological functions, ranging from virulence, metabolism and growth to motility, chemotaxis and stress resistance [30][31][32][33].
To address whether CsrA is implicated in host cell contact-mediated induction of pYVencoded virulence factors, we monitored expression of lcrF and the LcrF-dependent yadA gene in a csrA mutant in response to host cell contact. As seen in Figs 1D, 1E and S2A host cell contact induction of yadA was abolished in the absence of the csrA gene but could be complemented by introduction of a csrA + plasmid. A similar expression pattern was found for the lcrF gene (Fig 1F), indicating that CsrA is essential for cell contact-mediated LcrF induction.

CsrA activates translation and increases stability of the lcrF transcript under secretion condition
CsrA is able to post-transcriptionally activate expression of its target genes by binding to the leader transcript to (i) disrupt an RNA structure that would otherwise block the RBS or (ii) prevent transcript degradation by RNases [31,32]. To further investigate the influence of CsrA on LcrF synthesis, we first tested whether CsrA binds directly and specifically to the 5'-UTR of the lcrF transcript. This region forms a thermo-sensitive secondary structure of two stem- HEp-2 cells infected with Y. pseudotuberculosis wildtype strain YPIII harboring a plasmid-encoded yadA-gfp (pJE9) or a lcrF-gfp fusion (pWO3) were incubated for at least 2 h at 25˚C or 37˚C. Arrows indicate bacteria in contact and without contact to host cells. White bars indicate 5 μm. (B) Strain YPIII harboring a plasmid-encoded yadA-luxCDABE (pWO41) or a lcrF-luxCDABE fusion (pWO42) were incubated with or without HEp-2 cells for 2.5 h at 25˚C. Bioluminescence of the samples was monitored every 10 min to determine the kinetics of host cell contactdependent expression of the reporter fusions. (C) Strains YPIII (WT), YP12 (pYV -), YP66 (ΔlcrF) and YP155 (pYV -, lcrF + ) harboring the yadA-luxCDABE fusion plasmid pTS31 were used to infect HEp-2 cells or were incubated in PBS incubated without cells at 25˚C. Bioluminescence of the samples was monitored after 2 h. The data loops, which sequesters the lcrF ribosomal binding site by a stretch of four uracils (fourU RNA thermometer). Opening of this structure is favored at 37˚C and permits ribosome binding at host body temperature [34]. As shown in Fig 2A (left panel), incubation of a lcrF transcript (extending from -123 to +75 with respect to the lcrF start codon) and an hns control transcript with increasing concentrations of purified CsrA-His 6 resulted in the formation of a higher molecular weight complex of the lcrF transcript, but not of the hns transcript. The shifted lcrF leader transcript included one primary CsrA binding site within the RBS of the lcrF gene (AGGA) and a secondary binding site (AGAGA) shortly downstream of the start codon. The secondary binding site is required for CsrA interaction, as a fragment exhibiting a 60 nt deletion, eliminating the potential secondary binding site caused a drastic reduction of CsrA-lcrF complex formation (Fig 2A, right panel). A very small amount of CsrA-lcrF complexes was only observed at high CsrA concentrations (�100 nM), consistent with a much lower binding affinity to the incomplete binding site. To confirm the postulated primary CsrA binding site, we also exchanged the GGA motif within the RBS to TTC. We found that introduced mutations abolished or strongly reduced CsrA interaction with the lcrF transcript (Fig 2B). Only a very weak band representing the CsrA-lcrF complex was found at high CsrA concentrations (�100 nM) most likely due to CsrA binding to the secondary site.
To further prove whether CsrA interaction with the lcrF leader improves translation initiation, we first tested the expression of a translational P BAD ::lcrF'-'lacZ fusion harboring only the lcrF 5'-UTR to exclude CsrA influence on lcrF transcription. As shown in Fig 2C (left panel), loss of CsrA reduced expression of the P BAD ::lcrF'-'lacZ fusion, and ectopic expression of the csrA gene on a plasmid restored expression to wildtype levels. As the RBS of the lcrF gene is part of an RNA thermometer [34], it is possible that CsrA is required to promote and/or support thermo-induced opening of the stem-loop structure at moderate temperature to allow ribosome access. To investigate the influence of CsrA on the accessibility of the RBS, we also tested the effect of CsrA on a derivative of the P BAD ::lcrF'-'lacZ fusion harboring nucleotide exchanges (GUU-30/-28AAA) that lead to a more 'open' conformation of the thermosensing RNA element (mimicking host temperature) in which the RBS is more accessible [34]. As shown in Fig 2C (right panel), the overall synthesis of the LcrF-LacZ fusion protein was increased, but CsrA was still required for maximal induction of LcrF-LacZ synthesis. This indicated that CsrA binding in the vicinity of the start codon enhances initiation of LcrF synthesis. In fact, in vitro translation assays using a lcrF RNA template showed that addition of CsrA enhanced LcrF synthesis (Fig 2D).
In parallel, we tested whether CsrA-mediated activation of lcrF mRNA translation has an influence on lcrF transcript and protein levels under T3SS/Yop-inducing and non-inducing conditions (+/-Ca 2+ ). As shown in Fig 3A, we found that the amount of the lcrF mRNA, as well as the LcrF protein, is significantly reduced in the csrA mutant under secretion conditions. To test whether the lcrF transcript is less stable in the absence of the riboregulator, we determined the stability of the lcrF mRNA. To exclude transcriptional control and changes of the virulence plasmid copy number [25], we expressed the lcrF gene under control of the P BAD represent the mean ± SEM of the fold change (end/start) from three independent biological replicates and were analyzed with Student's t-test. �� : P<0,01, ��� : P<0,001. (D) HEp-2 cells infected with YPIII (WT), YP53 (ΔcsrA) and YP53 pAKH56 (csrA + ) harboring a yadA-gfp fusion (pJE9) were incubated for 4 h at 25˚C. Arrows indicate bacteria with and without contact to host cells. White bars indicate 5 μm. (E) Strains YPIII pV (empty cloning vector pHSG576), YP53 (ΔcsrA) pV and YP53 pKB60 (csrA + ) harboring a plasmid-encoded yadA-luxCDABE (pWO41) were incubated with or without HEp-2 cells at 25˚C. Bioluminescence of the samples was monitored every 10 min to determine the kinetics of host cell contact-dependent expression of the reporter fusions. (F) Strains YPIII, YP53 (ΔcsrA) pV (empty cloning vector pHSG576) and YP53 pKB60 (csrA + ) harboring a plasmid-encoded lcrF-luxCDABE (pWO42) were applied to infected HEp-2 cells or incubated in PBS without cells at 25˚C. Bioluminescence of the samples was monitored after 2 h. The data represent the mean ± SEM of the fold change (end/start) from three independent biological replicates and were analyzed with Student's t-test. ��  promoter. Absence of CsrA resulted in an approximately two-fold reduction of the full-length lcrF mRNA within 2.5 min after inhibition of transcription by rifampicin, whereas the lcrF transcript was stable in the presence of CsrA (Fig 3B). In summary, our analysis revealed that CsrA interacts with the 5'-UTR of the lcrF mRNA to promote translation initiation and that CsrA is required to stabilize the lcrF transcript. The lcrF transcript (-123 to +75) and a mutant transcript in which the GGA motif is exchanged against TTC were incubated with increasing concentration of purified CsrA-His 6 . The CsrA-lcrF mRNA complex is indicated by a star. (C) To study CsrA influence on translation initiation of lcrF, expression of an arabinose inducible translational lcrF-lacZ fusion encoded on pKB14 and a mutant variant with an open lcrF RNA thermometer element encoded on pKB99 was analyzed in YPIII (WT) pV (empty cloning vector pHSG576) and YP53 (ΔcsrA) pV and YP53 pKB60 (csrA + ). The bacteria harboring the different lcrF-lacZ fusions were grown at 25˚C to exponential phase and were then shifted for 4 h to 37˚C in the presence of 0.1% arabinose and absence of Ca 2+ . β-galactosidase activities represent the mean ± SD from three independent biological replicates which were analyzed with Student's t-test. � : P<0,05: �� : P<0,01, ��� : P<0,001. (D) An amplified lcrF DNA fragment of pMP1 was used for in vitro transcription. The resulting lcrF mRNA fragment was used for in vitro translation in the absence or presence of CsrA. Samples were analyzed by western blotting with a polyclonal antibody directed against LcrF. An assay without the lcrF template RNA was used as negative control. c: loading control (lower panel).

T3SS/Yop-inducing conditions manipulate expression of the Csr system
Since CsrA induces LcrF upon host cell contact in a post-transcriptional manner, we investigated whether Yop secretion-inducing conditions influence the amount and/or ratio of the Csr system components, i.e. the RNA-binding regulator CsrA and the counteracting regulatory RNAs CsrB and CsrC. To test this, we used Ca 2+ limiting conditions as a substitute for host cell contact to induce Yop secretion. As shown in Fig 4A, significantly lower transcript levels of the CsrB and CsrC RNAs components were detectable in the bacterial cell upon Ca 2+ -depletion, whereas the overall amount of CsrA was only slightly reduced, indicating that more non CsrB/ C-bound and hence more active CsrA is available under secretion conditions. These data suggest that a regulatory circuit is operative in which the release of the Yops influences Csr function to control LcrF and thus T3SS/Yop synthesis.

The translocon protein YopD controls the Csr system
Several studies demonstrated that expression of the T3SS/yop genes is under negative feedback control implicating the translocator protein YopD [17,20,22,35]. Therefore, we investigated whether export and removal of YopD from the bacterial cytoplasm is responsible for downregulation of the Csr components under Yop secretion-inducing conditions (Figs 4B and S3). We found that a yopD deficient strain is characterized by a strong reduction of the CsrB and CsrC sRNAs levels and caused a small decrease of the intracellular CsrA protein levels ( Fig  4B), similar to what has been observed under Ca 2+ depletion ( Fig 4A, lower panel).
CsrA was shown to indirectly inhibit the degradation of the CsrB and CsrC sRNAs in E. coli, [36,37], and also in Yersinia CsrA is required to maintain CsrB and CsrC (Figs 4A, middle panels, S3). This suggested that YopD influence on all the Csr components occurs through regulation of CsrA (Fig 4B, lower panel).
To further test this hypothesis, we analyzed expression of two translational csrA'-'lacZ fusions of which one harbored a deletion of the csrA upstream region which eliminated one of the two active promoters determined in previous studies [28,38] (Fig 4C). The deletion of upstream promoter P2 led only to a slight reduction of csrA'-'lacZ expression, demonstrating that P1 is the most active promoter. Absence of YopD reduced the overall activity of both reporters to about 50%, indicating that YopD-mediated activation involves sequences located downstream of position -121 (Fig 4C and 4D). Next, we tested whether purified YopD is able to recognize and bind to the 5'-UTR of csrA using transcripts covering different portions of the 5'-UTR downstream of promoter P2 (Fig 5A). We detected specific binding of YopD only to csrA RNA fragments harboring sequences from position -60 to -40 with respect to the start codon of csrA ( Fig 5B).
The inspection of the csrA 5'-UTR further revealed two potential primary CsrA binding motifs (i.e. ANGGA sequences), one of which overlaps the RBS of the csrA transcript ( Fig 5A). MgCl 2 . After growth for additional 4 h total RNA of the non-induced and induced cultures was prepared and subjected to northern blotting using a lcrF specific probe (upper panel). In parallel, LcrF protein levels were assessed by western blotting. Whole cell extracts were prepared and separated by 15% Tricine-SDS polyacrylamide electrophoresis. The LcrF and the H-NS protein were detected by immunoblotting with polyclonal anti-LcrF or anti-H-NS antibodies. (B) YPIII and YP53 (ΔcsrA) harboring a P BAD ::lcrF (pMP1) expression plasmid were grown to exponential phase (OD 600 of 0.5) at 25˚C. The cultures were shifted to 37˚C, depleted of Ca 2+ and grown for additional 4 h before rifampicin was added to the cultures to stop transcription. Samples of the cultures were taken directly (0 min) and 1, 2, 3, 5 and 7.5 min after rifampicin treatment. Total RNA of the samples was prepared and subjected to northern blotting using a lcrF specific probe. The 16S and 23S rRNAs served as loading controls (upper panel). The lcrF transcript levels were quantified and plotted to determine the mRNA degradation rate (lower panel). The intensity for the 0 min sample was set to 100%, and other samples were normalized to time point 0 sample. The data represent the mean ± SD from three independent biological replicates analyzed with Student's t-test. � : P<0,05.  This suggested that csrA translation is negatively autoregulated, by directly competing with 30S ribosomal binding, similar to what has been shown for E. coli csrA [39]. RNA-binding shift assay with an in vitro csrA transcript harboring the 5'-UTR sequences from position -70 to +10 revealed the formation of csrA-CsrA complexes, but a 5'-deletion to position -50, which eliminates one primary CsrA binding site, abolished CsrA-csrA complex formation ( Fig 5C). This strongly suggests that CsrA binding to ANGGA motifs in the vicinity of promoter P1 and the RBS builds a translational repressing complex. As one CsrA binding site overlaps with the identified YopD binding region, we postulate that YopD-dependent regulation of the Csr system occurs through interference with csrA autoregulation. YopD binding to the csrA transcript might prevent formation of the higher-order CsrA-RNA complex and allows free access of the 30S ribosomal subunit, which stimulates CsrA synthesis from the P2 initiated transcript.

Presence of YopD enhances lcrF transcript degradation under nonsecretion conditions
To further characterize the YopD-CsrA interplay, we compared the synthesis of LcrF and the LcrF-dependent Yops and YadA in the absence of CsrA and/or YopD. LcrF, YadA and Yop synthesis, as well as Yop secretion, were abolished in the ΔcsrA strain under Ca 2+ -limiting conditions and host cell contact, as expected from our previous results, demonstrating that CsrA activates LcrF synthesis (Figs 1D-1F and 6A and 6B,). In contrast, a strong induction of LcrF, YadA and Yop expression was detected in the absence of YopD under non-secretion conditions. This is consistent with former studies, demonstrating that YopD acts as a negative regulator of T3SS/yop gene expression in the absence of an inductive signal [17,20]. However, although lower amounts of the CsrA are produced in the yopD mutant, LcrF, Yop and YadA synthesis is not reduced compared to wildtype but enhanced. This indicated that YopD has an additional influence on lcrF expression, which is independent of CsrA.
To gain information about the individual contribution, we also tested how YopD influences LcrF, YadA and Yop synthesis as well as Yop secretion in the absence of CsrA. A significant reduction in the production levels and a failure of Yop secretion was also observed with the ΔcsrA/ΔyopD double mutant, but the overall influence is not as severe as in the ΔcsrA strain ( Fig 6A and 6B). This indicated that YopD acts upstream of CsrA, but additional control mechanisms appear to exist by which YopD impacts LcrF synthesis independently of CsrA. We next tested how YopD affects lcrF transcript levels in the presence and absence of CsrA. RNA degradation assays revealed that the half-life of the lcrF transcript is considerably enhanced in the absence of yopD, but this stabilization of the lcrF mRNA is not apparent in the ΔcsrA/ΔyopD mutant (Fig 6C). Rapid degradation of the lcrF transcript is observed, demonstrating that stabilization of the lcrF transcript is largely dependent on CsrA. However, slightly higher amounts of YadA and the Yop proteins were detectable in the ΔcsrA/ΔyopD mutant compared to the ΔcsrA strain, indicating that YopD does not solely affect these LcrF-activated virulence factors through regulation of CsrA. pseudotuberculosis strains YPIII (WT) pRS1 (empty vector), YP91 (ΔyopD) pRS1, YP91 (ΔyopD) pRS2 (yopD + ) were grown under non-secretion conditions (+Ca 2+ ). Total RNA of the cultures was prepared and subjected to northern blotting using csrB-and csrC-specific probes. The 16S and 23S rRNAs served as loading controls. Whole cell extracts of the identical non-induced and induced cultures were separated by SDS polyacrylamide electrophoresis and subjected to western blotting using a polyclonal anti-CsrA antibody. H-NS detected by a polyclonal antibody was used as loading control. Lower panel: scheme of the regulatory influence of YopD on the Csr systems components. (C) Schematic overview of the transcriptional csrA-lacZ reporter fusions encoded by pJH4 and pJH6 harboring different fragments of the csrA promoter region including the two identified promoters P1 and P2. The numbers indicate the 5'-end of the csrA upstream region fused to lacZ with respect to the translational start site of the csrA gene. (D) Strains YPIII (WT), YP91 (ΔyopD) pRS15 (empty vector), and YP91 (ΔyopD) pRS16 (yopD + ) harboring the different indicated csrA-lacZ reporter plasmids were grown to late exponential phase at 37˚C. β-galactosidase activity of the different cultures was monitored. The data represent the mean ± SD from at least three independent biological replicates performed in triplicates and were analyzed with Student's t-test; � : P<0,05, ��� : P<0,001, n.s.: P>0.05. https://doi.org/10.1371/journal.ppat.1007813.g004 Host cell contact-dependent type III secretion by riboregulators

YopD interferes with components of the degradosome
Since stabilization of the lcrF transcript in the absence of YopD was very prominent (Fig 6C), we hypothesized that this riboregulator could interfere with the RNA degradation machinery. To test this assumption, we first analyzed whether loss of YopD has an influence on the synthesis of the two RNases, PNPase and RNase E, which are components of the bacterial degradosome. As demonstrated by northern blotting and qRT-PCR analysis (Fig 7A and 7B), both pnp and rne transcript levels were significantly reduced in the yopD deletion strain under nonsecretion conditions. Importantly, secretion alone (which eliminates YopD from the cytoplasm) is sufficient to reduce pnp and rne mRNA levels ( Fig 7B). Moreover, we found that purified YopD is able to specifically interact with the 5'-UTR of the pnp and rne transcripts ( Fig 7C). However, the binding affinity of YopD to the 5'-UTR regions, in particular to the rne, is rather low compared to the 5'-UTR of the csrA transcript, implying that the YopDmediated influence occurs predominantly when the intracellular YopD concentration is increased. To further characterize the influence of YopD on rne and pnp expression we constructed rne-lacZ and pnp-lacZ transcriptional and translational fusion. As shown in S4 Fig, expression of both the rne-lacZ and pnp-lacZ translational fusion was significantly reduced in the absence of YopD. Expression of the rne-lacZ transcriptional fusion is also reduced to 69% in the absence of YopD. This confirmed that YopD activates the synthesis of RNase E, but it also indicated that this influence occurs mainly on the transcriptional level. In contrast, no influence of YopD is detectable with the pnp-lacZ transcriptional fusion. This would be expected when YopD predominantly influences PNPase synthesis on the post-transcriptional level. However, the overall expression of the pnp-lacZ fusion is very low, which makes the interpretation of the result difficult.
Based on the fact that YopD influenced rne and pnp transcript levels, we assumed that in the absence of host cell contact intracellular accumulating YopD could destabilize lcrF transcripts by upregulating PNPase and RNase E levels. To verify this assumption, we constructed a pnp mutant and a dominant-negative rne � (1-465) derivative (a rne mutant is lethal, [40]) and tested whether a decrease of functional degradosomes affects lcrF transcript and LcrF protein levels. As shown in Fig 7D, loss of PNPase or functional RNase E has a positive influence on LcrF synthesis as considerably higher amounts of the virulence regulator are detectable in the pnp mutant and the rne � expression strain. Taken together this demonstrated that YopD manipulates not only LcrF production by modulating CsrA levels, it also influences lcrF mRNA degradation by manipulating RNase E and PNPase synthesis.

Discussion
Triggering of T3SS/effector gene expression upon host cell contact is a well-known hallmark of T3SS regulation in many important Gram-negative plant, animal and human pathogens; however, the molecular mechanism initiating this process is still unclear. Up to this point, a role of the translocon protein YopD in this process in Yersinia was assumed [17,20], but how cell contact is transmitted and translated to trigger this response remained unclear. The Host cell contact-dependent type III secretion by riboregulators translocon protein could have a passive role, e.g. by inducing conformational changes in components of the secretion needle upon translocon formation which then initiates a set of signaling pathways to modulate T3SS/Yop synthesis. Alternatively, YopD could actively control T3SS/yop gene expression. Our data demonstrate that YopD influences synthesis of the T3SS/ Yop machinery and that this involves hijacking of multiple RNA regulators, the carbon storage regulator system and the degradosome components RNase E and PNPase.
The underlying regulatory processes of the Yersinia plasmid-borne T3SS are innately complex. Physical contact between the pathogen and its target cell first triggers YopD secretion, which forms the translocon channel together with exported YopB. This event then induces effector translocation. Consequently, the intracellular concentration of YopD is lowered and, as shown in this study, this changes the levels of the components of the global carbon storage regulatory system (CsrABC) and the degradosome-associated RNases RNase E and PNPase. In this manner, T3S of YopD serves as a regulatory switch signaling host cell contact at body temperature. We found that YopD controls the synthesis of the global riboregulator CsrA directly and specifically in a post-transcriptional manner. Under non-secretion conditions, YopD interacts with sequences in the 5'-UTR of the csrA mRNA that overlap with a CsrA binding site located between promoter P2 and P1. Thereby, YopD interferes with the autoregulatory feedback circuit inhibiting CsrA translation from this transcript. The resulting modest reduction in the amount of CsrA upon YopD secretion is sufficient to destabilize and strongly reduce the overall amount of the antagonistic RNAs CsrB and CsrC, leading to a significant increase of the CsrA:CsrB/CsrC ratio and thus free functional CsrA (Fig 8). This is possible because CsrA, of which multiple molecules (�8) interact with each sRNA, is absolutely essential for the maintenance of CsrB and CsrC, since it hinders CsrD to render them susceptible to RNase cleavage [36,37]. As free CsrA boosts/intensifies translation initiation and enhances the stability of the lcrF transcript, we postulate that liberated CsrA from CsrB/C downregulation upon YopD secretion is responsible for the upregulation of LcrF levels in response to host cell contact (Fig 8). As many metabolic, physiological and stress adaptation/resistance functions, as well as important pathogenicity factors, are under control of the Csr system [30][31][32][33], we assume that an increase of CsrA activity upon host cell contact is used by the bacteria to rapidly remodel their overall fitness and virulence program to adjust to host cell responses.
Often proteins that repress translation work in concert with ribonucleases to accelerate the degradation of untranslatable mRNA [41]. Here we show, that this is also the case for the YopD-CsrA-T3SS/yop mRNA interplay: Presence of YopD reduces CsrA-mediated protection of lcrF mRNA decay and enhances the synthesis of lcrF mRNA degrading RNases RNase E and PNPase. PNPase is a phosphorolytic 3' to 5' exoribonuclease. It associates with RNase E, the RNA helicase RhlB, and enolase, forming the RNA degradosome and plays a major role in RNA turnover [42,43]. RNase E-mediated cleavage often takes place within the 5' mRNA regions; this is followed by the processive exonuclease activity of PNPase to degrade target transcripts [41]. Notably, both ribonucleases were found to influence Yop production and Whole cell extract of the bacterial cultures were prepared and analyzed by western blotting with polyclonal anti-Yop, anti-YadA, anti-LcrF, anti-YopD or anti-CsrA antibodies. H-NS detected with polyclonal anti-H-NS served as loading control. (B) In parallel, proteins in the supernatant were precipitated with TCA, separated by SDS polyacrylamide electrophoresis and stained with Coomassie blue. A lcrF mutant was used as control. (C) To determine the stability of the lcrF mRNA rifampicin was added to cultures of YPIII, YP53 (ΔcsrA), YP91 (ΔyopD), and YP145 (ΔcsrA, ΔyopD) grown at 37˚C to stop transcription. Samples of the cultures were taken directly (0 min) and 1, 2, 3, and 5 min after rifampicin treatment. Total RNA of the samples was prepared and subjected to northern blotting using an lcrF specific probe. The 16S and 23S rRNAs served as loading controls. A representative of at least three independent experiments is presented (left panel). The lcrF transcript levels were quantified using the ImageJ software and plotted to determine the mRNA degradation rate (right panel). The band intensity for the 0-min sample was set to 100%, and other samples were normalized to the 0-min sample. The data represent the mean ± SD from three independent biological replicates. The end point was analyzed using ANOVA with Dunnetts post-test for multiple comparisons. � : P<0,05, �� : P<0,01; ��� : P<0.001; ns: not statistically significant. https://doi.org/10.1371/journal.ppat.1007813.g006 secretion, but the precise molecular mechanism remained unclear [40,[43][44][45]. In this study, we found that rne and pnp transcript levels are reduced in the absence of YopD (yopD mutant and YP138 (Δpnp) were grown at 37˚C. Total RNA of the cultures was prepared and subjected to northern blotting using a pnp (upper panel) or rne (lower panel) specific probe. Total RNA of YP138 (Δpnp) was used as negative control, and the 16S and 23S rRNAs served as loading controls. A representative of at least three independent experiments is presented. (B) Y. pseudotuberculosis strain YPIII (WT) and YP91 (ΔyopD) were grown at 37˚C in the presence (+Ca 2+ , non-secretion conditions) or absence (-Ca 2+ , secretion conditions) of 20 mM MgCl 2 and 20 mM sodium oxalate. Total RNA was prepared and a real-time qRT-PCR analysis was performed with specific primers for the pnp and rne transcripts. Transcript levels were normalized to the sopB reference transcript and are shown as relative expression to the wildtype (YPIII) under non-secretion (+Ca 2+ ) conditions. Relative expression data of both mRNAs show the mean ± SD from at least three independent replicates and were analyzed with a one sample t-test with P-value: � : P<0,05. (C) The in vitro generated pnp (upper panel), rne (lower panel) and hns transcripts were incubated with increasing amounts of purified YopD. The pnp-and rne-YopD complexes are indicated. (D) Y. pseudotuberculosis strain YPIII (WT), YPIII prne 1-465 expressing a dominant negative rne variant (pRS40) and YP138 (Δpnp) were grown at 37˚C. Total RNA of the cultures was prepared and subjected to northern blotting using an lcrF specific probe. The 16S and 23S rRNAs served as loading controls. In parallel, whole cell extracts of the cultures were prepared and western blot analysis using a polyclonal antibody directed against LcrF and H-NS (control) was performed.
https://doi.org/10.1371/journal.ppat.1007813.g007 or upon secretion). Since YopD interacts with the 5'-UTR of the rne and pnp transcripts, we hypothesize that YopD-binding increases the stability of the RNase transcripts. Although the molecular mechanism is still unclear, it is possible that YopD indirectly controls the expression of the RNase genes, e.g. in case of the rne gene. In addition, YopD could interfere with the negative autoregulation of RNase E and PNPase synthesis. For instance, YopD binding to the 5'-UTRs of the pnp transcript could hinder binding of both RNases, which cleave and process their own transcripts [46][47][48]. Alternatively, YopD could interfere with CsrA binding, which was also found to participate in the PNPase autoregulation in E. coli by repressing translation and reducing the stability of the pnp messenger [47,49].
Although YopD binds specifically to 5' UTR sequences of the csrA, pnp and rne transcripts, the molecular mechanism how it recognizes and interacts with RNA is less clear. Work with Y. enterocolitica showed that YopD also directly interacts with multiple yop transcripts, prevents the recognition of translational initiation signals by ribosomes, and accelerates their degradation [22]. It was speculated that the specificity might be conferred by the presence of short linear AU-rich elements (AUAAA) in the YopD target mRNA near or overlapping the RBS. However, transfer of the AU-rich sequences to YopD-independent genes did not confer YopD control [22]. Moreover, sequences that are specifically recognized by YopD in the csrA 5'-UTR are not AU-rich. This observation and relative promiscuity of RNA binding in vitro at higher concentration, indicate that additional factors or sequence features of YopD target genes must exist. Furthermore, YopD interaction with the chaperone LcrH is necessary for repressing ysc/ yop transcript translation [21,50]. LcrH could be implicated in YopD-RNA binding or may only be required to hold YopD in an RNA-binding competent fashion.
Overall, YopD is one of the most abundant Yop proteins. This might ensure that even under non-secretion conditions still low intracellular levels of YopD are present for negative regulation. In parallel, the lcrF transcript is highly susceptible to degradation and particularly responsive to CsrA. CsrA directly activates LcrF synthesis on the post-transcriptional level by specific binding to sites overlapping with the RBS and nucleotides of codons 4-6 of the lcrF mRNA. Interaction of CsrA further stabilizes the lcrF transcript, most likely through the inhibition of RNase E/PNPase function and/or synthesis. Likewise, a deletion of the csrA homolog rsmA was recently shown to impair activation of the master regulator hrpG of the T3SS hrp/hrc genes in plant pathogenic Xanthomonads [51]. This indicates that this mechanism may be conserved among certain T3SS-encoding pathogens.
The model that emerges is that Yersinia T3SS/yop gene expression is multilayered and occurs in a sequential manner, beginning with a low-level induction of lcrF expression upon host entry, e.g. due to the rise of temperature. Formation of a translational block by YopD in the absence of host cell contact arrests T3SS/Yop synthesis on a low level and keeps expression control in a 'ready-to-go/standby' position. Flipping the switch, i.e. eliminating the blockage This requires the activity of CsrA to overcome translational repression by the LcrF RNA thermometer element. At 37˚C, translational repression by the RNA thermometer is relieved and allows activation of lcrF translation and Ysc/Yop protein synthesis. In the absence of host cell contact, the Yop effectors bound to their chaperones accumulate in the bacterial cytoplasm. YopD represses LcrF expression by positive regulation of the degradosome components RNase E and PNPase, which promote lcrF transcript degradation. Moreover, presence of YopD influences the components of the Csr system in a way, that high levels of the antagonistic ncRNAs CsrB and CsrC are present that sequester and inhibit CsrA function. Reduced CsrA function prevents activation of lcrF translation and limits Ysc/Yop protein expression to a low level. Upon host cell contact, YopD is secreted and forms the translocation pore together with YopB, which leads to the depletion of YopD from the cytosol. This negatively affects the expression of RNase E and PNPase. Furthermore, loss of YopD triggers a change in the ratio of the Csr components in which significantly more CsrA is present in relation to CsrB and CsrC. The increased number of free and active CsrA molecules promotes a strong increase of LcrF and LcrF-mediated Ysc/Yop protein synthesis, which lead to the formation of active injectisomes and efficient secretion of the Yops into the host cell. by secretion of YopD upon host cell contact triggers T3SS/Yop synthesis via modulation of the CsrABC-RNases cascade (Fig 7).
Intimate coupling of T3SS/effector gene transcription with cell contact-induced activation of the T3 secretory activity is common among T3SSs and was shown to involve complex regulatory networks and feedback control systems. However, the regulatory mechanisms involved are very diverse and differ significantly from Yersinia. In Bordetella pertussis, differential regulation of T3S genes is mediated by a secreted anti-sigma factor, which acts as an antagonist of an alternative sigma factor that controls the synthesis of T3SS apparatus components and secretion substrates [52]. Several other pathogens, including Pseudomonas aeruginosa and Shigella flexneri couple secretion and T3SS gene activation by a partner-switch mechanism [53][54][55][56][57][58][59]. This more widely used coupling theme implicates the transcriptional regulator of T3SS genes, a secreted protein, and additional interacting proteins. The latter change their binding partner when the secreted factor is injected into host cells and eliminated from the bacterial cytoplasm. This results in the liberation of the blocked transcriptional activator to activate T3SS gene transcription. Notably, the CsrA homolog RsmA of Pseudomonas was found to positively control T3SS gene expression in this microorganism, but the mechanism and whether this is linked to the partner-switching cascade is unclear [60, 61]. In enteropathogenic E. coli, the effector chaperone CesT, which is released and remains in the cytoplasm upon translocation of the effectors was shown to bind and antagonize CsrA. This modulates T3SS gene expression and known CsrA-repressed target genes [59].
The newly emerging regulatory scheme from our study now includes coupling cell contactinduced secretion of an RNA-regulating translocator protein with T3SS/effector gene expression through major global RNA regulators. This not only promotes a fast adaptation of T3SS/ effector gene expression, it also coordinates T3SS with numerous CsrA-and RNase E/PNPasedependent virulence-relevant traits, stress adaptation and metabolic functions in response to host cell contact. This reprogramming is crucial to ensure adaptation, survival, proliferation, and pathogenesis during acute infections. In particular, it not only prevents sudden immune cell attacks, but also adjusts the overall fitness of the pathogen towards the imposed energetic burden entailed by fully active T3SSs. Moreover, the RNA-based regulatory circuit can also trigger down-regulation of T3SS/Yops and other virulence factors during later stages of the infection, when immune cell attack has been largely defeated, and thereby contribute to the initiation of persistent infections [62,63]. We postulate that T3SS activity of Yersinia provokes a global remodeling of the virulence and fitness program through control of CsrA and the RNA degradosome by YopD, which is crucial for the pathogens to adapt to rapidly changing conditions and flourish in different hostile host niches. As the CsrA/RsmA regulator was found to control the expression of many T3SSs genes in Gram-negative bacteria [32], we propose that implementing CsrA/RsmA as an interaction partner of T3SS-secreted proteins herein delineates a general strategy for responding to host cell contact. The challenge ahead lies in unifying all identified interactions and mutant phenotypes in the different pathogens into a regulatory network explaining previous and future observations.

Material and methods
Bacterial strains, cell culture, media and growth conditions E. coli and Yersinia strains were routinely grown under aerobic conditions at 25˚C or 37˚C in LB (Luria Bertani) broth on solid or in liquid media if not indicated otherwise. The antibiotics used for bacterial selection were as follows: carbenicillin 100 μg ml -1 , chloramphenicol 30 μg ml -1 , tetracycline 10 μg ml -1 , and kanamycin 50 μg ml -1 . For northern blot experiments bacteria were diluted 1/50 in fresh medium from overnight cultures, grown to exponential phase (OD 600 0.5) at 25˚C and then shifted to 37˚C for additional 4 h in the absence (+Ca 2+ , T3SS/ yop gene non-inducing conditions) or presence of 20 mM MgCl 2 and 20 mM sodium oxalate to deplete Ca 2+ ions (-Ca 2+ , T3SS/yop-inducing conditions). HEp-2 cells (ATCC, USA) used for the analysis of host cell contact expression assays were grown in a humidified atmosphere with 5% CO 2 at 37˚C (HERA cell 150 incubator, Thermo Scientific) in RPMI1640 supplemented with Glutamax medium containing 7.5% NCS as growth factor.

DNA manipulation, construction of plasmids and mutant strains
All DNA manipulations, restriction digestions, ligations and transformations were performed using standard genetic and molecular techniques [64]. The plasmids used in this work are listed in S1 Table. Oligonucleotides used for PCR, qRT-PCR and sequencing were purchased from Metabion and are listed in S2 Table. Plasmid DNA was isolated using Qiagen plasmid preparation kits. DNA-modifying enzymes and restriction enzymes were purchased from New England Biolabs, Promega and Roche. PCRs were performed in a 100 μl mix for 29 cycles using Phusion High-Fidelity DNA polymerase (New England Biolabs). Purification of PCR products was routinely performed using the QIAquick PCR purification kit (Qiagen). The constructed plasmids were sequenced by the in-house facility.
Plasmid pAKH172 is derived from pET28a, carrying a His 6 -tag. The csrA coding region was amplified by PCR with primers IV783/I68. The resulting insert was digested with NcoI and XhoI and ligated into the NcoI/XhoI site of pET28a. To construct pJE9, the yadA insert was amplified via PCR with primers 90/92. The resulting fragment was digested with BamHI and KpnI and ligated into the BamHI/KpnI site of pGFPmut3.1 to produce pJE2. Next, pJE2 was digested with PstI and SpeI, and the resulting yadA-gfp fragment was ligated into the PstI/ SpeI site of pIV2mob. The kanamycin cassette from pJE9 was exchanged against the ampicillin cassette. To do so, the bla gene was amplified by PCR with primers I972 and I998 using pBAD-mycC as template. The resulting fragment was digested with BglII and ligated into the BglII site of pJE9. The csrA-lacZ reporter gene fusion plasmids pJH4 and pJH6 were constructed by amplification of the csrA upstream region using primer pairs IV436/II275 and IV438/II275. After digest with EcoRI and PstI, the inserts were ligated into the EcoRI/PstI site of pKB63. To construct pJH12 the yopD gene was amplified with primers IV706/III645 and the lcrH gene with primer pair IV708/III647. Next, the yopD, as well as the lcrH inserts, were digested with NdeI/SpeI and SpeI/XhoI, respectively. Subsequently, the fragments were ligated into the NdeI/ XhoI sites of linearized pET28a. For construction of the low copy plasmid pKB60, the csrA coding region was amplified with the primer pair 558/559. The resulting fragment was digested with SalI and BamHI and ligated into the SalI/BamHI sites of pHSG576. Plasmid pKB99 was derived from pED7 carrying the lcrF-lacZ fusion with a mutation which destabilizes the thermoloop structure in the 5'-UTR of lcrF (GUU -30/-28 AAA). The kanamycin cassette was synthesized with primers III902 and III905, digested with XbaI and SphI and ligated into the XbaI/SphI sites of pED07. Plasmid pMP1 was derived from pBAD33. The lcrF gene fragment (-123 to +771 relative to the translational start of lcrF) was amplified by PCR using primer pairs V659/I214. The produced fragment was digested with XbaI and SacI and ligated into the XbaI/SacI sites of pBAD33. For construction of the plasmid pFU100, the Ap R cassette of pFU72 was exchanged against the Cm R cassette of pZA31-luc. Therefore, pZA31-luc was digested with XhoI and SacI, and the isolated fragment was ligated into the XhoI/SacI sites of pFU72. The plasmid pRS1 was constructed from pFU50. The kanamycin cassette of pFU50 was excised with SacI and XhoI and replaced by the chloramphenicol cassette of pZA31luc. The translational rne-lacZ and pnp-lacZ fusions of pIVO22 and pIVO23 were constructed by PCR amplification of the rne and pnp promoter fragments using primer pairs VIII675/VIII677 and VIII672/VIII674, which were ligated into the BamHI/SalI site of pTS02. The corresponding transcriptional rne-lacZ and pnp-lacZ fusions of pIVO25 and pIVO26 were constructed by amplification of the rne and pnp promoter fragments using primer pairs VIII675/VIII676 and VIII672/VIII673, which were ligated into the BamHI/SalI site of pTS03. To construct pRS2, the yopD gene was amplified by PCR with the primer pair II348/II349. The resulting fragment was inserted into the PstI and NotI sites of pRS1. The lcrQ fragment was produced with primers II360 and II361 and ligated into the PstI/NotI sites of pRS1 generating pRS4. To construct pRS18, the midi copy origin p15A was amplified by PCR with primers II496/II538 using pACYC184 as template. The insert was digested with SacI and XbaI and introduced into the SacI and XbaI sites of pRS4. For construction of pRS15, the lcrQ coding region of pRS18 was removed by digestion of the vector with MluI and SphI. The sticky ends of both restriction sites were blunted and the vector was religated using the CloneJET PCR Cloning Kit (Fermentas, USA). Plasmid pRS16 was generated by exchanging the ori29807 of pRS2 with the p15A origin as described for pRS18. To construct pRS40, an rne fragment (-15 to +1395 nt with respect to the rne start codon) was amplified by PCR with the primer pair III233/III234 and cloned into the XbaI and SphI sites of pBAD33. For the construction of pRS50, the kanamycin cassette was amplified with the primer pair I661/I662 using pKD4 as template. Next, about 500 bp of the upstream region of pnp was amplified with the primer pair III243/III244. Primer III244 possesses additional 20 nt at the 5´-end that are homologous to the start of the kanamycin resistance gene. The pnp downstream fragment was amplified with primer pair III245/ III246. The forward primer III245 contained additional 20 nt at the 5´-end that were homologous to the end of the kanamycin resistance gene. Next, a PCR reaction was performed with primer pair III243/III246 using the upstream and downstream PCR products of the pnp and the kanamycin cassette as templates. Finally, the PCR product was digested with SacI and integrated into the SacI site of the suicide vector pAKH3. The mutagenesis plasmid pRS34 (ΔyopD) was constructed by amplification of 300 bp regions upstream (II363/II366) and downstream (II364/II365) of the yopD coding sequence. The reverse (II366) and the forward (II365) primers contain overlapping sequences complementary to each other at their ends. Another PCR, using primers II365/II364 and the two fragments as template, resulted in one fragment with an in-frame deletion of the yopD gene. The fragment was digested with SpeI and SphI, and integrated into the SpeI/SphI site of the suicide plasmid pDM4 [65]. For the construction of pSR1, the gfp mut3.1 gene was excised of pGFPmut3.1 with PstI and SpeI and ligated into the PstI/SpeI sites of pIV2mob. The yscW-lcrF locus was excised from pWO8 with BamHI and SalI and ligated into the BamHI/SalI sites of pFU98 generating pTS34. To construct pWO3, an insert carrying P yscW , the yscW deletion (+113 to +298 nt with respect to the yscW transcriptional start site) as well as the first 72 nucleotides of the lcrF coding sequence was synthesized via a three-step PCR with primers I222/I224 using two fragments amplified from Yersinia genomic DNA with the primer pair I224/I747 and I222/I746, respectively, as template. This insert was digested with PstI and ligated into the PstI site of pGP20, resulting in pKB12. Next, the P yscW ::ΔyscW-lcrF' insert was amplified with the primer pair I224/I964 using pKB12 as template, digested with PstI and BamHI, and finally ligated into the PstI/BamHI sites of pSR1. To generate pWO8, the yscW-lcrF locus was amplified with primers I214/I844. The insert was digested with XbaI and SphI and ligated into pACYC184 resulting in pKB28. Next, the yscW-lcrF fragment was extracted from pKB28 and cloned into pFU51 via the SalI/XbaI restriction sites. Plasmids pWO41 (yadA-luxCDABE) and pWO42 (lcrF-luxCDABE) were constructed by an exchange of the chloramphenicol resistance gene of pTS31 and pTS34 against the ampicillin resistance gene of pFU31 using the SacI/BamHI sites. For the construction of pWO14, the pSC101 � ori of pWO8 was exchanged against the suicide ori R6KmobRP4 of pFU100 via the SacI/XbaI restriction sites generating pWO13. Next, the non-coding region between YPTB1128 and YPTB1129 was amplified by PCR with primers II341/II342 and integrated into the SacI restriction site of pWO13.

Construction of Y. pseudotuberculosis deletion mutants
To generate YP155, plasmid pWO14 was mated from E. coli S17-1 λpir (tra + ) into Y. pseudotuberculosis YP12 and transconjugants were selected on Yersinia selective agar (Oxoid) supplemented with carbenicillin. All other mutant strains with integrations, deletions, resistance cassette insertions or nucleotide substitutions were constructed by homologous recombination using suicide plasmids pRS34 (YP91, YP145) and pRS50 (YP138). Plasmids were mated from E. coli S17-1 λpir (tra + ) into Y. pseudotuberculosis YPIII and transconjugants were selected on Yersinia selective agar (Oxoid) supplemented with chloramphenicol. The recombination of the plasmid into the Yersinia virulence plasmid pYV yielded a merodiploid strain, including a wildtype and the mutant copy. Subsequently, the resulting strain was plated on 10% sucrose to induce expression of the toxic sacB gene on the suicide plasmids. 50 selected fast-growing strains were screened for chloramphenicol sensitivity to prove the loss of the integrated plasmid. All mutant strains were proven by PCR and DNA sequencing.

RNA isolation, northern blotting and RNA degradation assays
Bacteria grown under the required growth conditions were pelleted and RNA was isolated using the SV total RNA purification kit (Promega) as described [34]. Total RNA (20 μg) was separated on MOPS agarose gels (1.2%), transferred by vacuum blotting for 1.5 h onto positively charged membranes (Whatman) in 10 x SSC buffer (1.5 M NaCl, 0.15 M sodium citrate, pH7) using a semi-dry blotting system and UV cross-linked. Prehybridization, hybridization to DIG-labelled probes and membrane washing were conducted using the DIG Luminescent Detection Kit (Roche, Germany) according to the manufacturer's instructions. The DIGlabelled PCR fragments used as probes were produced by PCR using the DIG-PCR nucleotide mix (Roche, Germany) as described [34] with the following primer pairs: for the lcrF transcript-I214/I303, for the csrB and csrC transcripts-555/556 and 583/I82, for the rne transcript-IV529/IV530 and the pnp transcript-IV527/IV528 (see S2 Table).
To determine stability of the lcrF transcript, RNA stability assays were performed. In order to stop the de novo mRNA synthesis 0.5 mg/ml rifampicin (Serva) was added. 0, 1, 2, 3, 5 and 7.5 min after rifampicin treatment, 10% v/v phenol was added and the samples were snap frozen in liquid nitrogen. RNA isolation and northern blot analysis were performed as described above.

Quantitative real-time RT-PCR (qRT-PCR)
qRT-PCR was performed using the SensiFastNoRox Kit (Bioline) with 25 ng/μl of the RNA samples according to the manufacturer's instructions. qRT-PCR was performed in a Rotor-Gene Q lightcycler (Qiagen). Primers used for analyzing relative gene expression purchased from Metabion and are listed in S2 Table. The gene sopB was used for normalization. Data analysis was performed with the Rotor-Gene Q Series Software. Relative gene expression was calculated as described earlier [66]. Primer efficiencies were determined experimentally using serial dilutions of genomic Y. pseudotuberculosis YPIII DNA. Primer efficiencies are: csrB: 2; csrC: 2; pnp: 2; rne: 2; sopB: 2.

Purification of the CsrA protein and the YopD-LcrH complex
E. coli strain BL21 transformed with pAKH172 or pJH12 was grown at 37˚C in LB broth to an A 600 of 0.6. 0.5 mM IPTG was used to induce CsrA-His 6 and YopD-His 6 production. CsrA-His 6 purification was performed as described previously [67]. The His 6 -YopD protein was purified by Ni-NTA affinity chromatography in lysis buffer (50 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 20 mM imidazole). The column was washed two times with three column bed volumes of lysis buffer supplemented with 40 mM imidazole. Bound YopD was eluted by adding six column bed volumes of lysis buffer supplemented with 250 mM imidazole. For RNA electrophoretic mobility shift assays, the proteins were dialyzed over night against the respective RNA-binding buffer (see RNA-EMSA) at 4˚C. The purity of CsrA-His 6 and the YopD-His 6 -LcrH complex was estimated to be >95%.

RNA electrophoretic mobility shift assay (RNA-EMSA)
For the RNA-EMSAs, some RNAs were 3'-end labeled with pCp Biotin (Jena, Bioscience, Germany) by ligation with T4 RNA Ligase I (lcrF transcripts , Fig 2A; csrA (a) transcript, Fig 5B  and 5C; csrA transcript (e) and (f), Fig 5C). The reaction was performed with 1 μg linearized RNA in 1 x T4 ligase buffer, 1 mM ATP, 10% (w/v) DMSO, 1 μM pCp-Biotin (Jena, Bioscience, Germany), 15% (w/v) PEG8000 and 1 μl T4 ssRNA ligase (20 u, NEB, USA) at 18˚C for 2 h or overnight. The labeled RNA fragments were purified by phenol-chloroform extraction [64], and RNA labeling was checked by northern blotting. To do so, 2 nM RNA were separated on a native 4-8% TBE polyacrylamide gel, transferred onto a nylon membrane with a Trans-Blot SD Semi-Dry Transfer Cell (Biorad, Germany) and crosslinked to the membrane with an UV-crosslinker (Stratagene, USA). The biotinylated RNAs were detected with the Chemiluminescent Nucleic Acid Detection Module (Thermo Scientific), and light emission was detected by a ChemiDoc XRS+ (Biorad, USA). All other transcripts used for the RNA-EMSAs were 5'end labelled with radioactive P 32 γ-ATP (10 μCi/μl) using the polynucleotide kinase (Thermo Fisher Scientific). The reaction was performed with 100 ng linearized RNA in 1 x polynucleotide kinase buffer, and 1 mM P 32 γ-ATP at 37˚C for 1 h. The reaction was stopped with STE buffer. The labeled RNA fragments were purified by phenol-chloroform-isoamyl alcohol extraction [64], and RNA labeling was checked by northern blotting.
For RNA-binding studies the purified CsrA and YopD proteins were dialyzed against the RNA-binding buffer (20 mM Na 2 HPO 4 /NaH 2 PO 4 (pH 8,0), 100 mM KCl, 5% (w/v) glycerol, 2 mM DTT). The different biotinylated lcrF, csrA, rne and pnp transcripts used for RNA band shift analysis were diluted in the equivalent RNA-binding buffer, denatured at 70˚C for 10 min and chilled on ice. Subsequently, the transcripts (20 fmol/2 nM RNA) were incubated with increasing concentrations of purified CsrA or YopD for 30 min on ice in the RNA-binding buffer, and immediately loaded on 4-8% polyacrylamide gels. After electrophoresis the RNA and RNA-protein complexes were transferred on a nylon membrane and visualized as described above.

In vitro translation of csrA
To test the influence of CsrA on lcrF mRNA translation, an in vitro translation assay was performed using PURExpress In vitro protein Synthesis Kit (NEB, USA) according to the manufacturer's instructions. In vitro transcribed lcrF RNA from plasmid pMP1 was used as template.

Gel electrophoresis and western blotting
For immunological detection of CsrA, YopD, YadA, LcrF and the Yop proteins, cell extracts of equal amounts of bacteria were prepared and separated on a 15% polyacrylamide SDS gel [64]. Proteins were transferred onto an Immobilon-P membrane (Millipore) and probed with polyclonal antibodies directed against CsrA, YopD, LcrF, YadA or secreted Yop proteins (Davids Biotechnologies, Germany) as described [67].

Host cell contact assays
Host cell contact-dependent expression of Yersinia virulence genes was assessed with Y. pseudotuberculosis strain YPIII and isogenic mutant strains (YP53 ΔcsrA) with or without contact to HEp-2 cells (ATCC, USA) by a) in situ monitoring of gfp reporter fusions by fluorescence microscopy, b) in situ quantification of light emission of luxCDABE reporter fusions and, and c) detection and quantification of LcrF, YadA or YopE.
For the microscopical monitoring Y. pseudotuberculosis strains expressing the respective gfp reporter fusions were grown at 25˚C to stationary growth phase. About 10 6 HEp-2 cells seeded in μ-slide 8-well microscope slides (ibidi, Germany) were infected with about 5 x 10 7 bacteria. The bacteria were centrifuged onto the epithelial cells and incubated up to four hours at 25˚C or 37˚C. Fluorescence emitted by GFP was analyzed by fluorescence microscopy (Axiovert II with Axiocam HR, Zeiss, Germany) using the AxioVision program (Zeiss, Germany).
For the monitoring of bioluminescent luxCDABE reporter fusions Y. pseudotuberculosis strains harboring promoter fusions to the bacterial luciferase operon luxCDABE were grown at 25˚C to stationary growth phase. 5 x 10 4 HEp-2 cells seeded in 96 well assay plates (Corning Incorporated) were infected with 10 6 bacteria. The bacteria were centrifugated onto the cells and incubated for 2.5 h at 25˚C in a Varioscan plate reader (ThermoScientific). Light emission was measured every 10 min and documented with the Varioscan Flash and SkanIt RE software (ThermoScientific). Curves of relative luminescent units (RLU) per time were calculated out of the triplicates, normalized to the starting conditions and visualized by GraphPad Prism.
To visualize a host cell contact-mediated increase of the YadA and the LcrF transcript and protein levels in the different tested Y. pseudotuberculosis strains, 2 x 10 8 bacteria were used to infect 4 x 10 6 HEp-2 cells. The bacteria were centrifuged onto the cells and incubated for 150 min at 25˚C or 37˚C. The infected cell culture was washed with PBS, lysed (0.1% Triton X-100, 0.9% NaCl), and diluted in stop solution for RNA isolation and northern blotting, or in SDS sample buffer for SDS-PAGE and western blotting.

Yop secretion assay
For Yop secretion assays, about 50 ml LB medium were inoculated 1:50 with an overnight culture of the desired strain and grown at 25˚C for 2 h. Subsequently, secretion was induced by addition of 20 mM MgCl 2 and 20 mM sodium oxalate. The cultures were shifted to 37˚C and cultivated for 4 h. Cell quantities of the different bacterial cultures were adjusted according to their OD 600 . The bacteria were harvested in falcon tubes by centrifugation. The supernatant was filtered and proteins were precipitated with 1/10 volume of 100% trichloroacetic acid (TCA). After incubation for 20 min on ice, the proteins were pelleted. The pellets were resuspended in 2 ml acetone-SDS solution (1.75 ml 100% acetone, 0.25 ml 2% SDS), incubated on ice for 20 min, and pelleted again. To wash the proteins, the supernatant was discarded and 500 μl 100% acetone was applied. After the last centrifugation step, the precipitated proteins were resuspended in equal amounts of sample buffer and applied on 15% SDS polyacrylamide gels for electrophoretic separation. Coomassie Brilliant Blue was used to visualize the proteins.

Luciferase and β-galactosidase assays
Bacteria harboring luxCDABE and lacZ reporter fusion plasmids were grown under different growth conditions as described. β-galactosidase was measured in cell-free extracts as described previously [68]. The activities were calculated as follows: β-galactosidase activity OD 420 �6,75� OD 600 -1 �Δt (min) -1 �Vol (ml) -1 β-galactosidase assays were performed in triplicate of cultures grown under indicated conditions. Reporter fusions emitting bioluminescence were measured in non-permeabilized cells with a Varioscan Flash (Thermo Scientific) using the SkanIt software (Thermo Scientific) for 1 s per time point and every 10-15 min for kinetic analyses. The data are given as relative light units (RLU/OD 600 ) from three independent cultures performed in duplicate.
Supporting information S1  Furthermore, whole cell extracts of the virulence plasmid-cured strain YP12 (pYV -) and the wildtype strain YPIII (WT) grown under secretion conditions at 37˚C in the absence of Ca 2+ (right panel) were used as negative and positive controls. (B) YPIII (WT) strains harboring lcrF-luxCDABE (pTS34), yadA-luxCDABE (pTS31), invA-luxCDABE (pTS32), or gapA-lux-CDABE (pFU166) fusion plasmids, were used to infect HEp-2 cells in PBS or incubated in PBS without cells at 25˚C. Bioluminescence of the samples was monitored after 2 h. The data represent the mean ± SD of the fold change (end/start) from three independent biological replicates and were analyzed with Student's t-test: ��� : P<0,001, n.s. (P>0.05). (C) HEp-2 cells infected with Y. pseudotuberculosis wildtype strain YPIII harboring pJE2 (empty vector) or a plasmidencoded yadA-gfp (pJE9) and ΔlcrF strain carrying plasmid-encoded yadA-gfp (pJE9) incubated for 4 h at 25˚C. White bars indicate 5 μm. The csrB-deficient strain YP52 was used as negative control. Total RNA of the cultures was prepared and subjected to northern blotting using a CsrC-specific probe. The 16S and 23S rRNAs served as loading controls. A representative of at least three independent experiments is presented. (TIF)