Autoproteolysis and Intramolecular Dissociation of Yersinia YscU Precedes Secretion of Its C-Terminal Polypeptide YscUCC

Type III secretion system mediated secretion and translocation of Yop-effector proteins across the eukaryotic target cell membrane by pathogenic Yersinia is highly organized and is dependent on a switching event from secretion of early structural substrates to late effector substrates (Yops). Substrate switching can be mimicked in vitro by modulating the calcium levels in the growth medium. YscU that is essential for regulation of this switch undergoes autoproteolysis at a conserved N↑PTH motif, resulting in a 10 kDa C-terminal polypeptide fragment denoted YscUCC. Here we show that depletion of calcium induces intramolecular dissociation of YscUCC from YscU followed by secretion of the YscUCC polypeptide. Thus, YscUCC behaved in vivo as a Yop protein with respect to secretion properties. Further, destabilized yscU mutants displayed increased rates of dissociation of YscUCC in vitro resulting in enhanced Yop secretion in vivo at 30°C relative to the wild-type strain.These findings provide strong support to the relevance of YscUCC dissociation for Yop secretion. We propose that YscUCC orchestrates a block in the secretion channel that is eliminated by calcium depletion. Further, the striking homology between different members of the YscU/FlhB family suggests that this protein family possess regulatory functions also in other bacteria using comparable mechanisms.


Introduction
In 1952, Hills and Spurr showed that virulent strains of Yersinia pestis (Pasturella pestis) were unable to grow and divide when incubated at 37uC; instead, they required incubation at 27uC [1]. This phenotype was surprising, because Y. pestis causes lethal infections in rodents and humans, which have a body temperature close to 37uC. Moreover, no typical nutritional requirements could explain this phenotype. Later, Kupferberg and Smith demonstrated that addition of 2.5 mM calcium to the growth medium supported growth of Y. pestis at 37uC [2]. This unusual requirement for calcium was later shown to be correlated to the massive synthesis and secretion of a number of proteins, called Yersinia outer proteins (Yops). This was based on the observation that 2.5 mM calcium in the growth medium blocked Yop secretion, while depletion of calcium induced massive Yop secretion that also results in stop of bacteria proliferation [3,4,5]. Synthesis and secretion of Yops are dependent on a virulence plasmid [6], a common feature of all human pathogenic Yersinia (Y. pestis, Y. enterocolitica, and Y. pseudotuberculosis). Yops are synthesized during infection, which indicates their importance in virulence [3]. Yop secretion involves the type III secretion system (T3SS) of Yersinia, which is encoded by the same virulence plasmid that carries the yop genes. The T3SS is a dedicated secretion system that forms a multi-protein complex of around 25 proteins spanning the inner and outer bacterial membranes [7]. It is built up by a basal body located in the membrane showing high homology with a corresponding structure of the bacterial flagellum. A needle is anchored to the basal body forming hollow tube measuring around 60 to 80 nm in length and 8 nm in external width with an inner diameter of 3 nm [8,9]. It has been postulated that Yops are transferred to the target cell through the needle structure [8]. This model has however been challenged in recent work from our laboratory [10] where we show that bacterial surface localized Yop-effectors can be translocated into the target cell. Hence, translocation can occur via a mechanism that is distinct from the postulated micro-injection model. Yersinia employs the T3SS to secrete Yops into the external environment and to translocate Yops into the cytoplasm of eukaryotic target cells [11]. These processes are highly regulated. It has been shown that Y. pseudotuberculosis up-regulates yop expression after contact with eukaryotic cells, and this requires a functional T3SS [12,13]. Importantly, target cell contact can be mimicked by depleting calcium in the growth medium and simultaneously shifting the temperature from 26uC to 37uC [11]. Modulation of calcium levels in the growth medium has been an invaluable tool for increasing our understanding of T3SSs in Yersinia virulence. Several seminal and general discoveries have been made based on the calcium effect, including T3SS mediated secretion, translocation, and target cell induced expression of effector proteins [12,13,14].
The YscU protein of Yersinia is an integral inner-membrane protein with four membrane spanning segments ( Figure 1A) and is required for T3SS function. It belongs to a family of proteins (YscU/FlhB class) that is characterized by auto-cleavage at a highly conserved NqPTH motif (amino acids 263-266) [15]. Autoproteolysis of YscU is required for proper regulation of Yops synthesis and secretion. Furthermore, the Yop synthesis and secretion is lost when the full yscU gene or the NqPTH coding sequence are deleted, indicating the importance of YscU for T3SS function. Similar phenotypes are observed when point mutations affect cleavage at the NqPTH motif; this illustrates the importance of cleavage for calcium regulation [16,17,18,19]. Full length YscU (denoted YscU) contains two domains, the transmembrane domain (TM) and a soluble cytoplasmic domain, denoted YscU C ( Figure 1A and Figure 1B). Autoproteolysis of YscU occurs between asparagine 263 and proline 264 at the NqPTH motif and results in a 10 kDa C-terminal polypeptide fragment, denoted YscU CC that is attached to the remainder of the protein through protein-protein interactions. In context of the cytoplasmic domain, YscU C (which is used extensively in this article), cleavage generates two fragments; the YscU CC fragment and a 6 kDa N-terminal fragment denoted YscU CN [16,18] ( Figure 1A).
Both YscP (FliK) and YscU (FlhB) have been linked to the ''substrate specificity switch'', first identified by MacNab and coworkers in the flagellum T3SS [15,20]. This switching machinery changes secretion specificity from early hook substrates to late filament substrates as one step in the assembly of the flagellum [21]. It has been suggested that the C-terminal domain of FliK (FliK C ) binds to the C-terminal cytosolic domain of FlhB (FlhB C ), causing a conformational change in FlhB C that is required for the switch [20]. An yscP mutant was impaired in switching from the early secretion of needle subunits (YscF) to the late export of Yops [16,22]. This led to a phenotype with unusually long needles unable to secrete Yop proteins, thus YscP is an essential protein for T3SS mediated secretion [16,22,23]. A similar phenotype was observed for the yscU mutant, N263A, which highlighted the importance of YscU autoproteolysis in the substrate specificity switch [16]. Interestingly, an yscP null mutant was suppressed by single amino acid substitutions in YscU C , and these suppressor mutants partially restored Yop secretion [17]. This suggested that YscU and YscP interact, and that this interaction was essential for proper control of needle formation and Yop secretion [16]. A direct interaction between the YscU and YscP orthologs, Flik and FlhB has been shown with surface plasmon resonance experiments [24]. In analogy, mutations in the corresponding yscP gene in Shigella flexneri (spa32) and Salmonella thypimurium (invJ) [25] also caused defective substrate switching. It has been shown that Spa32 (YscP) and Spa40 (YscU) interact [26,27]. Given the high functional similarity between the T3SSs of different species, it is likely that the substrate specificity switch is regulated by a similar mechanism in different pathogens.
Here, we studied the functional role of YscU in Yop secretion by exploiting the calcium regulation of substrate switching in Y. pseudotuberculosis. We combined in vivo and in vitro methods to examine the steps of YscU autoproteolysis, subsequent dissociation, and secretion of YscU CC and how they affect Yop secretion in Y. pseudotuberculosis during growth in calcium depleted media.

Bacterial Strains, Plasmids, and Growth Conditions
Bacterial strains and plasmids used in this study are listed in the supporting material (''Table S1''). Escherichia coli strains were grown in Luria-Bertani broth (LB) or on Luria agar plates at 37uC. Y. pseudotuberculosis was grown at either 26uC or 37uC in Hepes buffered LB or on Luria agar plates (unless specified in the text). Antibiotics were used for selection according to the resistance markers carried by the strains at the following concentrations: kanamycin, 50 mg/ml; chloramphenicol, 25 mg/ml; and carbenicillin, 100 mg/ml. EGTA was added to the media at a final concentration of 5 mM and 20 mM MgCl 2 to create calcium depleted conditions. The full-length protein contains 354 amino acid residues. The N-terminal 210 residues constitute four transmembrane helices (TM). The cytosolic domain of YscU (YscU C ) undergoes autoproteolytic cleavage at the NqPTH-motif (amino acids 263-266), which leaves an N-terminal cytoplasmic polypeptide, denoted YscU CN , and a C-terminal polypeptide, denoted YscU CC . (B) Ribbon drawing of the cleaved cytosolic domain YscU C (2JLI.PDB) from Y. pestis [39]. YscU CN and YscU CC resulting from cleavage at the NqPTH motif are colored in orange and grey, respectively. doi:10.1371/journal.pone.0049349.g001

Yop Secretion Assay
Cultures were started at an absorbance of OD 600 = 0.1 in Hepes buffered LB with the appropriate antibiotics. Bacteria were grown at 26uC for 2 h and shifted to 37uC for 3 h in calciumsupplemented or calcium-depleted conditions (except where specified in the text). Cultures were harvested and centrifuged for 10 min at 4 0006g. Aliquots (4.5 ml) of filtrated supernatant were combined with 10% (v/v) trichloroacetic acid (TCA) for protein precipitation. Precipitated proteins were solubilized in SDS-PAGE loading buffer. The pelleted cells were resuspended in an equal volume of LB and lysed with SDS-PAGE loading buffer. Cells and supernatants were loaded at equivalent protein concentrations (according to OD 600 ) and separated by SDS-PAGE. Proteins were either stained with Coomassie R250 or, alternatively, transferred onto a PVDF membrane (GE Health-care) for immunoblotting. Anti-Yop antibodies were diluted at 1:5 000 and horseradish peroxidase-conjugated anti-rabbit IgG was diluted at 1:10 000 (GE Healthcare). Proteins were detected with a chemiluminescence detection kit (GE Healthcare).

YscU CC Overexpression and Secretion Assay
We grew YPIII/pIB102 bacterial strains, which contained the pBADmycHis B plasmid (Invitrogen) with the yscU CC expression sequence (see supporting ''Material and methods S1''), and control strains contained an empty vector. The growth conditions were as described above, except 0.2% (v/v) of L-arabinose was added after 1 h at 26uC to induce biosynthesis of YscU CC . After separation by SDS-PAGE, proteins were stained with Coomassie R250 (Yop secretion) or transferred onto a PVDF membrane (GE Healthcare) for immunoblotting. Anti-YscU CC peptide antibodies were diluted at 1:5 000 [16] and horseradish peroxidase-conjugated anti-rabbit IgG was diluted at 1:10 000 (GE Healthcare). Proteins were detected with a chemiluminescence detection kit (GE Healthcare).

GST-pulldown Assay
We purified GST-YscU C -His 6 and GST-A268F-His 6 proteins in 2 steps, by combining GST-and Ni-NTA affinity chromatography. Purified proteins (80 mM) were incubated in 25 mM Tris, pH 7.4, 1 mM EDTA, and 150 mM NaCl at 37uC and 30uC. At different time points, 250 ml samples were taken, centrifuged to remove aggregates (15 min, 16 0006g at 4uC), and loaded on GST-SpinTrap TM columns (GE Healthcare). Columns were washed twice with 25 mM Tris, pH 7.4, 150 mM NaCl buffer, and eluted twice by adding 20 mM GSH solution, pH 8.0. Eluted samples were mixed with SDS-sample buffer and boiled. Proteins were subsequently separated and visualized with 4-12% Bis-Tris Gel SDS-PAGE (Invitrogen).

Protein Purification of YscU C Variants
All YscU C constructs were cloned as GST fusion with a cleavage site for PreScission Protease between the GST domain and YscU C (see supporting ''Material and methods S2''). After transformation into E. coli BL21 (DE3) pLysS the protein synthesis was induced with IPTG and performed overnight at 30uC in LB medium containing carbenicillin and chloramphenicol. The bacterial cells were harvested by centrifugation at 5 000 rpm at 4uC and stored at 280uC until use. The protein purification of YscU C variants was performed with an Ä KTA purifier system (GE Healthcare). The bacterial pellet was resuspended in 50 mM Tris pH 7.4 and 2 mM DTT, and cells were disrupted by sonication. Thermal up-and down-scans of YscU C at pH 7.4 monitored with CD spectroscopy at 220 nm in the presence of 2.5 mM calcium. Up-and down scans of YscU C are shown in red and blue circles, respectively. (B) Titration of calcium to YscU C monitored with CD spectroscopy at 220 nm. The calcium binding isotherm to YscU C was fit to a one-site binding model (red line). The resulting K d was 800 mM. (C) Western Blot analysis of YopB, YopD, and YopE in wild-type Y. pseudotuberculosis. The bacteria were grown for 2 h at 26uC and shifted to 37uC for 3 h (temperature shift for induction of Yop secretion) with varying concentrations of free calcium. ''Pellet'' indicates intracellular proteins; ''supernatant'' denotes secreted proteins. The LB growth medium was initially supplemented with 1 mM EGTA to complex residual calcium content (approximately 500 mM); thereafter, calcium was added to set the indicated concentrations of free calcium. doi:10.1371/journal.pone.0049349.g003 The lysate was clarified by centrifugation at 15 000 rpm and 4uC, and the supernatant passed through a 0.45 mm syringe filter (Corning). The lysate, containing the soluble GST fusion protein, was loaded on a 5 mL GSTrap FF column (GE Healthcare) and eluted with 20 mM GSH solution at pH 8.0. Fractions with the fusion protein were pooled, dialyzed at 4uC against cleavage buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM DTT). The GST protein was cleaved from the target protein by adding PreScission Protease (GE Healthcare). To remove GST and non-cleaved GST-fusion protein the solution was passed through a Glutathione Sepharose 4B column. The flowthrough with YscU C , was subjected to cation exchange chromatography (5 ml SP Sepharose, GE Healthcare). All eluted fractions with YscU C were pooled, concentrated with Amicon Ultra-15 Centrifugal Filter Units (Millipore, Billerica, MA), and polished with sizeexclusion chromatography (HiPrep 26/60 Sephacryl S-100HR, GE Healthcare) in phosphate buffered saline at pH 7.4. Fractions with YscU C were pooled stored as 100 mM stock at 20uC until use.

Analytical Size Exclusion Chromatography
Protein samples (YscU CC ) were applied to a Superose 6 10/ 300 GL column (GE Healthcare) and size-exclusion chromatography (SEC) was performed with a flow rate of 0.5 ml/min in phosphate buffered saline at pH 7.4. Protein elution was followed by monitoring the UV absorption at 260 nm, 280 nm, and 220 nm. All samples with signal peaks at 280 nm were analyzed by SDS-PAGE. Prior to analytical SEC, the column was calibrated with a gel filtration protein standard (Bio-Rad).
The time series of one-dimensional 1 H NMR spectra to probe dissociation were acquired with a pulse program from the Bruker library, which incorporated excitation sculpting for water suppression. For each spectrum, 64 scans were accumulated with a relaxation recovery delay of 2 s between scans. For each protein, time series were acquired at 30uC and 37uC. To quantify dissociation kinetics, we integrated the methyl group resonances in the 0.2 to 0.4 ppm spectral region. Each time course of the NMR signal was fit with a single exponential decay function of the form: I = I 0 exp (2t/tdiss) +A, where t diss was the lifetime of the decay, and A was a baseline offset. NMR data was processed with NMRPipe [30] and visualized in ANSIG for Windows [31].

Circular Dichroism
Circular dichroism (CD) spectra were recorded on a Jasco J-810 spectropolarimeter, equipped with a Peltier element for temperature control and a 0.1 cm quartz cuvette. The proteins were measured at 10 mM in a buffer of 10-fold diluted phosphate buffered saline at pH 7.4. For all experiments with calcium, different buffers were used (phosphate, MOPS, and Pipes) to exclude possible calcium precipitation effects. Thermal dissociation experiments were performed by monitoring the CD signal at 220 nm as a function of temperature. All thermal profiles were acquired in the interval of 20uC to 90uC. The thermal scan rate was varied from 0.5 to 2uC/min, without any significant change in protein behavior; we selected 1uC/min as the standard condition in this study for CD spectroscopy-based temperature perturbation experiments. The inflection point for dissociation, T diss , was quantified by fitting thermal curves with a two-state equation [32]. Calcium binding affinity was quantified by fitting a one-site binding model to the CD data.

Dissociation of YscU CC from YscU CN
We recently published results showing that specific yscU mutants (N263A and P264A), defective in autoproteolysis, were impaired in their ability to secrete Yops into the culture supernatant at wildtype levels [16]. These mutations strongly reduced the autoproteolytic activity of the YscU protein. Especially the yscU mutant P264A was severely suppressed in autoproteolysis leading to an almost complete inhibition of Yop secretion [16]. These results and earlier findings showing that deletion of the autoproteolytic cleavage motif NPTH leads to a complete loss of Yop secretion indicated that the cleavage of YscU is required for Yop secretion [18]. Here, we decided to study YscU in more detail. Because YscU is an integral inner-membrane protein, we produced a polypeptide that comprised the cytosolic segment of YscU including the motif for autoproteolytic cleavage (YscU C ; Figure 1A and Figure 1B) for in vitro studies. For detailed investigation of conformational changes in YscU C we have used the spectroscopic methods nuclear magnetic resonance (NMR) and circular dichroism (CD).
It was previously proposed that the YscU CC polypeptide dissociates from the remaining, membrane-anchored segment of YscU to allow Yop secretion [16]. To test this hypothesis, we developed biophysical NMR and CD based protocols for the quantification of YscU CC dissociation from YscU CN in vitro. The high quality, 1 H-15 N HSQC spectrum of YscU C at 20uC (Figure 2A, blue contours) and the chemical shift dispersion showed that YscU C was a folded protein under the experimental conditions. We assigned 91% of the non-proline backbone resonances in our protein construct that contained residues 211-354. In the YscU C crystal structure (2JLI.PDB) the N-terminal residues 241-255 are in a helical conformation. The helix protrudes into solution and the first residue that makes contact CD spectroscopy at 220 nm was used to monitor thermal up-and down-scans of YscU C in presence of different divalent cations at a scan rate of 1uC/min to determine dissociation temperatures (T diss , compare Figure 3A). To measure the binding isotherms (K d ) of different divalent cations towards YscU C (compare Figure 3B) titrations monitored with CD spectroscopy at 220 nm were performed. doi:10.1371/journal.pone.0049349.t001 Bacteria transformed with empty pBADmycHis B (pBAD), or pBAD with one additional yscU CC copy (pBAD(YscU CC )), were grown for 2 h at 26uC and 3 h at 37uC in calcium depleted (2) or calcium supplemented (+) medium. The expression of yscU CC was induced by addition of arabinose. Yop secretion is coupled to the secretion of YscU CC in all analysed Yersinia strains and required a functional T3SS. Secreted Yops visualized on Coomassie stained PAGE gels; YscU CC visualized on immunoblots with anti-YscU CC peptide antibodies. ''pellet'' indicates intracellular proteins; ''supernatant'' denotes secreted proteins. The YopJ protein (black box) was subjected to densitometric analysis for quantification of secretion levels (see Table 2). doi:10.1371/journal.pone.0049349.g004 with the remainder of the protein is residue number 250. In solution the first helical residue that we could identify based on NOE contacts was residue 251, and the assigned preceding residues are adopting an unstructured and flexible conformation as inferred from high signal intensities and narrow chemical shift dispersion. After incubation at 60uC for 10 min, the NMR spectrum of YscU C at 20uC showed a dramatic perturbation (Figure 2A, red contours); only resonances that corresponded to the YscU CN polypeptide were visible. The narrow chemical shift dispersion and high peak intensities of the YscU CN resonances showed that the polypeptide adopted an unfolded conformation also after dissociation from YscU CC . The absence of signals that corresponded to the YscU CC polypeptide was due to the formation of aggregated particles too large for detection with NMR spectroscopy (see supporting material, ''Results S1'').
The NMR experiments showed that YscU CC dissociation from YscU CN was triggered by subjecting YscU C to thermal perturbation, and that dissociation was an irreversible process. The irreversibility provided a tool for quantifying dissociation kinetics (discussed below). To obtain an accurate value of the dissociation temperature (T diss ), we observed YscU C dissociation by subjecting the protein to a thermal cycle, and we followed this event with CD at 220 nm. The CD signal at 220 nm contains contributions from both alpha helical and b-strand secondary structures [33]. Thus, because YscU C contains, both alpha helices and b-strands, 220 nm was a suitable wavelength for following changes in the YscU C structure ( Figure 1B). The thermogram of YscU C ( Figure 2B) was composed of two distinct transitions (55uC and 77uC), and the overall thermal response was not reversible, as the CD signal did not reach its initial value after a complete thermal cycle (see supporting material, ''Results S2''). The transition at 55uC corresponded to the dissociation of YscU C ; this was in good agreement with the NMR results that showed an upper limit of T diss equal to 60uC. The high temperature transition (77uC) was reversible, but with distinct signs of hysteresis ( Figure S1). Because this transition was not relevant in the context of dissociation, we did not study it further. Of note, because dissociation is an irreversible process, T diss is scan-rate dependent. Therefore, all CD spectroscopy-based temperature perturbation experiments were conducted at a fixed scan rate of 1uC/min.
To address the questions whether the dissociation of YscU C might have biological relevance we subjected the non-cleavable YscU C mutant P264A to a thermal denaturation. The thermal signature of P264A was dramatically perturbed compared to the wild-type, and the thermogram displayed one irreversible transition at 55uC ( Figure 2C). To investigate the biological relevance further, we asked whether calcium might affect the YscU C dissociation in vitro. The thermogram in the presence of calcium was remarkably similar to that of the non-cleavable variant P264A in the absence of calcium ( Figure 3A). Hence, calcium mimicked the effect of a mutation that suppressed the YscU C auto-processing activity. P264A contained one polypeptide chain that could not dissociate; thus, the data suggested that calcium prevented dissociation of YscU CC from the YscU CN polypeptide.
Next we investigated the ability of YscU C to bind calcium in vitro and compared it to the calcium concentration needed to block the Yop secretion in vivo. The CD signal at 220 nm indicated the YscU C binds calcium with a dissociation constant (K d ) of 800 mM, assuming a one-site binding model ( Figure 3B). To benchmark the K d value of calcium against the calcium concentration required for inhibition of the Yersinia T3SS in vivo, we analyzed Yop secretion and expression at different calcium levels in the growth medium. The T3SS was down regulated at calcium concentrations between 0.5 to 1 mM ( Figure 3C). Hence, the in vitro K d value for calcium interaction with YscU C (800 mM) was well in the concentration interval that inhibited Yop secretion in vivo. Comparative analysis with different divalent cations revealed no exclusive specificity of YscU C towards calcium. Different alkaline earth metals, Mg 2+ , Ca 2+ , Sr 2+ and Ba 2+ showed comparable effects in vitro on YscU C interaction and dissociation (Table 1). It was not surprising that Ba 2+ and Sr 2+ showed a similar effect as Ca 2+ since these ions have been shown to effect Yop secretion similarly to Ca 2+ [34]. On the other hand Mg 2+ has no effect on Yop secretion suggesting that the in vivo regulation of calcium controlled secretion is dependent on additional factors. This promiscuous metal binding property of YscU C is consistent with the absence of any known calcium binding motif in YscU C . In accordance we observed with NMR spectroscopy that calcium binding is mediated through a large set of residues confined to the YscU CC polypeptide ( Figure S2).
Nevertheless the in vitro data clearly showed that YscU C was poised for dissociation into YscU CN and YscU CC fragments, and that this event can be inhibited by calcium and other divalent cations. Thus we wondered whether the dissociation of YscU C could be monitored directly in Yersinia, and whether it can be linked to T3SS regulated Yop secretion. To test this idea, we first probed for the presence of YscU CC in the culture supernatants  Figure 4 and Figure 5E) was selected for quantitative analysis. Growth kinetics in media with different pH's ( Figure S4B) showed attenuation of bacterial growth directly linked to the observed Yop secretion efficiency. a secretion efficiency was set to 100%; b secretion efficiency at pH 7.5 was set to 100%. doi:10.1371/journal.pone.0049349.t002 after incubating the wild-type Y. pseudotuberculosis strain at 37uC in the absence or presence of 2.5 mM calcium. Remarkably, YscU CC was found in the culture supernatant from the calcium depleted cultures, and no YscU CC was found in cultures with 2.5 mM calcium (Figure 4). To explore this finding further, the gene for the YscU CC polypeptide (amino acids 264 to 354 of YscU) was cloned into the pBAD vector under the control of an inducible araC promoter. This allowed the in trans overexpression of yscU CC in Y. pseudotuberculosis. After promoter induction, the levels of secreted YscU CC were analyzed in cultures grown with or without calcium. We found that induction of yscU CC expression caused increased secretion of YscU CC into the culture supernatant when compared to the wild-type levels secreted by a strain that contained the control vector. Further, secretion of YscU CC was blocked in the presence of 2.5 mM calcium in the medium. To investigate the requirement of a functional T3SS for YscU CC secretion we analyzed the secretion behavior of a Y. pseudotuberculosis DyscC null mutant. In absence of YscC no secretion of either Yops or YscU CC was observed (Figure 4) showing that secretion of YscU CC is dependent on a functional T3SS. To link our in vitro findings of YscU C and calcium directly to in vivo events we analyzed the secretion behavior of a DyopN mutant that has lost its calcium regulation and secretes Yops in presence and absence of calcium in similar amounts [35]. It was found that the DyopN mutant secreted YscU CC independently of the calcium concentration. Thus, YscU CC was secreted from the DyopN mutant in the presence of calcium showing a similar secretion profile as the Yop substrates ( Figure 4).
In conclusion, YscU CC showed a secretion pattern similar to that of Yops. This suggested that YscU CC constituted a novel substrate of the T3SS in Yersinia. Furthermore, the fact that YscU CC was secreted in the wild-type strain demonstrated that YscU CC was able to dissociate from the remaining membrane bound part of YscU in vivo.

pH-dependencies of YscU CC Dissociation/secretion and Yop Secretion
Unpublished observations from our laboratory indicated that Yop secretion, but not bacterial growth, was influenced by the pH of the growth medium. Further it has been shown that autoproteolysis of YscU C was pH dependent [15]. Here, we addressed the question of whether dissociation of YscU C was also pH dependent. First, we subjected cleaved YscU C to a thermal cycle at pH 6.0 (instead of pH 7.4) by monitoring the thermal signature with CD spectroscopy at 220 nm ( Figure 5A). When the resulting thermogram was super-imposed on the thermograms of the wild-type YscU C at pH 7.4 in presence of 2.5 mM calcium and the non-cleavable mutant P264A in the absence of calcium, they were virtually identical. From this observation, we concluded that YscU C dissociation was prevented by low pH. YscU C contains two histidine residues (positions 266 and 324) in the YscU CC fragment that may explain the observed pH-dependency. By monitoring the CD signal at 220 nm as a function of pH, we identified one ionization event in the pH interval of 6.0 to 7.4, with a pKa value of around 6.3 ( Figure 5B), and one ionization event around pH 8.0. NMR analyses revealed that both histidines were protonated in response to a pH drop from 7.4 to 6.0 ( Figure 5C, 5D). This indicated that the protonation of histidines 266 and 324 was responsible for the observed differences in thermally induced dissociation of YscU C at pH 6.0 and 7.4. This observation was interesting, because all known YscU orthologs in other T3SSs harbor a conserved histidine residue at the position that corresponds to amino acid 266 in the NqPTH motif. To further dissect the relevance of the two histidines we replaced histidine 324 with alanine (H324A) and studied the in vitro response of this mutant to both thermal-and pH perturbations. Since it has been shown that mutation of histidine 266 leads to an yscU mutant affected in autoproteolysis this position is not suitable for an alanine replacement [36]. The H324A variant showed similar dissociation behavior in vitro but with reduced thermal stability at pH 7.4 and pH 6.0 compared to wild-type ( Figure S3A and S3B). . pH-dependencies of YscU C dissociation in vitro and Yop/YscU CC secretion in vivo. (A) Thermal up-and down-scans of YscU C at pH 6.0, monitored with CD spectroscopy at 220 nm in the absence of calcium. The thermal signature of YscU C displayed one large-amplitude transition at 55uC. Up-and down scans of YscU C are shown in red and blue circles, respectively. (B) The pH-dependency of the YscU C monitored with CD spectroscopy at 220 nm. (C) Chemical shift perturbations of YscU C quantified from 1 H-15 N HSQC spectra, in response to a pH-shift from 7.4 to 6.0, displayed against the primary sequence. The blue line indicates the threshold value (0.05 ppm) used in Figure 5D. The chemical shift perturbation of the two histidines at positions 266 and 324 are shown in red. (D) Structural distributions of residues that show significant chemical shift perturbations in response to a pH-shift from 7.4 to 6.0 are shown in red on the YscU C structure (2JLI.PDB). The YscU CN and YscU CC fragments are colored orange and gray, respectively. The two histidine residues (266 and 324) in the folded part of YscU C are indicated. (E) Coomassie stained gels show Yop secretion under different pH conditions. (top panel) ''pellet'' indicates intracellular proteins; (middle panel) ''supernatant'' denotes secreted proteins. The YopJ protein (black box) was subjected to densitometric analysis for quantification of secretion levels (see Table 2). (bottom panel) The pH-dependency of YscU CC secretion was visualized on immunoblots with anti-YscU CC peptide antibodies. doi:10.1371/journal.pone.0049349.g005 The pH-dependency of the CD-signal at 220 nm of the histidine mutant H324A showed a distinct difference compared to the wildtype protein ( Figure S3C). Whereas the wild-type protein displayed two ionization events (at pH 6.3 and 8.0) the mutant only displayed the ionization event at pH 6.0. Hence, the pKa of the NqPTH histidine is around 6, and protonation/deprotonation of this histidine is likely responsible for the difference in thermally induced dissociation at pH 6.0 and 7.4. Next, we wondered whether these biophysical observations reflected biological effects in vivo. To investigate the pH-dependency of Yop secretion, we cultivated wild-type Y. pseudotuberculosis in Hepes buffered media at pH values between 6.0 and 7.5 in calcium depleted media and monitored the secretion of YscU CC and Yops ( Figure 5E and Table 2). Both Yop and YscU CC secretion was maximal at pH levels between 7.0 and 7.5. Secretion gradually decreased when pH was lowered, and at pH 6.0, we observed a pronounced inhibitory effect on the secretion of YscU CC as well as the Yops (albeit not as strong as the inhibition by calcium). Importantly, bacterial growth was not affected by changing the pH of the growth medium; thus, perturbations of external pH values Dissociation temperatures (T diss ) of single suppressor mutants at pH 7.4 were quantified with the CD signal at 220 nm and a scan rate of 1uC/min; YscU C (black), A268F (blue), Y287G (green), and V292T (red). All suppressor mutants are destabilized compared to wild-type YscU C . T diss values are summarized in Table 3. (B), (C) Dissociation kinetics quantified as dissociation life-times (t diss ) of YscU C (black) and V292T (red) at pH 7.4 followed with NMR-spectroscopy at (B) 37uC and (C) 30uC, respectively. Primary NMR data for (B) is shown in Figure S6. Solid lines correspond to fits of the experimental data to single exponential decays. (D) Time dependent GST-pulldown experiments show the dissociation of wild-type YscU C and the suppressor mutant A268F at 30uC and 37uC after varying incubation times. The suppressor mutant A268F displayed pronounced dissociation of YscU CC -His 6 at 37uC and moderate dissociation at 30uC; wild-type YscU C displayed no dissociation of YscU CC -His 6 at 37uC or at 30uC over the observed time period. Note! Dissociation of YscU CC is manifested as disappearance of YscU CC -His 6 over time since the dissociation is irreversible and YscU CC -His 6 cannot bind itself to the used resin. doi:10.1371/journal.pone.0049349.g007 did not cause any general effects on bacterial proliferation ( Figure  S4). These in vivo and in vitro results suggested that the pHdependent secretion of the secretion of Yops and YscU CC was a consequence of the molecular behavior of YscU; i.e., the dissociation and secretion of YscU CC was a prerequisite for maximal secretion of Yop substrates into the culture medium.

Yops are Secreted at Lower Temperatures in Destabilized YscU CC Variants Compared to Wild-type
The results described above suggested that dissociation of YscU CC from the remainder of YscU, followed by secretion of the YscU CC polypeptide, was important for Yop secretion. We and other groups previously showed that single amino acid substitutions in YscU C could suppress the Yop secretion-deficient phenotype of the Y. pseudotuberculosis DyscP mutant [17,21,37]. Identification of such suppressor mutations is generally considered as strong genetic evidence for protein-protein interactions. Thus, second-site suppressor mutations are expected to be localized at protein surfaces that are prime positions for protein-protein interactions. In sharp contrast, all amino acid substitutions on the YscU CC polypeptide were either fully or partially buried in the protein structure ( Figure 6). Consequently, the underlying mechanism of these mutations must be more complex than a direct protein-protein interaction with YscP. Mutations at buried positions generally act to destabilize proteins; therefore, we reasoned that the altered secretion behavior of the suppressor mutants might be attributed to perturbed stabilities (T diss ) and dissociation rates (k diss = 1/t diss ) compared to wild-type YscU C .
To test this notion, we monitored the dissociation kinetics of YscU C suppressor mutants A268F, Y287G, V292T and wild-type YscU C with CD and NMR spectroscopy in vitro. The resulting dissociation kinetics are reported as dissociation lifetimes (t diss ), or the reciprocal of the dissociation rate ( = 1/k diss ). Because Yop secretion is triggered by a temperature shift from 26uC to 37uC [3,4,6], we initially performed the assays at 37uC. We found that all YscU C mutants were folded ( Figure S5A) and displayed decreased thermal stabilities compared to wild-type YscU C , evident from the reduced dissociation temperatures ( Figure 7A and Table 3). We quantified the dissociation kinetics with both CD and NMR spectroscopy for the YscU C variant V292T; with CD, we detected changes in ellipticity at 220 nm that accompanied dissociation ( Figure S5B); with NMR, we detected the loss of resonance intensities for residues in the YscU CC fragment ( Figure  S6). The time-dependent signals were well described by first order processes; accordingly, all kinetic traces could be fit accurately with single exponential decay functions. Both CD and NMR results indicated that wild-type YscU C displayed very slow dissociation kinetics at 37uC in the observed time frame; in comparison, the YscU C variant V292T displayed a significantly enhanced rate of dissociation ( Figure 7B, Figure S6A, S6B). It should be noted that the observed variations in the dissociation lifetimes (t diss ) for the suppressor mutants are dependent on the method used for quantification (Table 3). For instance, t diss for the V292T variant was 61 min and 140 min, based on CD and NMR, respectively. This discrepancy could be attributed to differences in sensitivity; the CD signal was directly sensitive to the dissociation process, but NMR required both dissociation and aggregation for a reduction in signal intensity. Hence, both dissociation and aggregation were slow processes that occurred at similar time scales in vitro. After one hour at 37uC, a significant fraction of the V292T variant displayed dissociated species (65% and 38% from CD and NMR, respectively), but wild-type YscU C remained intact under the same conditions.
V292T displayed a significantly reduced thermal stability compared to wild-type YscU C ; therefore, we also performed Table 3. Dissociation temperatures and kinetics of wild-type YscU C and the suppressor mutants V292T, Y287G, and A268F probed with NMR and CD spectroscopy. CD spectroscopy at 220 nm was performed at a scan rate of 1uC/min to determine the dissociation temperature (T diss ) of YscU C and suppressor mutants ( Figure 7A). The kinetics of the dissociation process (t diss ) was monitored with CD spectroscopy at 220 nm and NMR spectroscopy at 37uC and 30uC. See also Figure 7B and 7C; Figure S5B  NMR-based kinetic experiments at 30uC (Figure 7C). At this temperature, wild-type YscU C was stable over the entire experiment (860 min), but the V292T variant dissociated at a rate equal to the rate for wild-type YscU C at 37uC, within experimental error (Table 3). We confirmed these results with the A268F suppressor mutant. Further we used a GST pull down assay, and the dissociation of GST-YscU C -His 6 respectively GST-A268F-His 6 was monitored at 30uC and 37uC. Indeed, GST-YscU C -His 6 was completely stable at both temperatures, but GST-A268F-His 6 dissociated to a significant extent over time, visible in the decrease of the YscU CC -His 6 content at both 30uC and 37uC ( Figure 7D). Because the dissociation of YscU C is required for Yop secretion in vivo, we expected suppressor mutant strains to secrete Yops at lower temperatures compared to the wild-type strain. To test this hypothesis, we probed for the presence of secreted Yops and YscU CC in the culture supernatant in strains that carried either yscU or yscU suppressor mutants A268F or V292T after a temperature shift from 26uC to 30uC. Both the A268F and V292T mutant strain secreted Yops and YscU CC already at 30uC; in contrast, the wild-type strain showed almost no secretion at this temperature ( Figure 8). Probing for LcrV and YscI as early substrates in T3SS confirmed the direct link between secretion of YscU CC and T3SS substrates ( Figure S7B). Despite the increased secretion of Yops at reduced temperatures, the pH regulation was still active in these Yersinia mutants. A268F and V292T revealed the same secretion pattern like the wild-type when grown in media with different pH ( Figure 5E and Figure S7A). Importantly, all strains secreted Yops in a calcium-regulated manner; this showed that the suppressor mutants retained calcium sensing capability (data not shown).
Our results showed a strong link between in vivo and in vitro results for dissociation of the YscU CC polypeptide from the remainder of the protein. This demonstrated that not only autoproteolytic cleavage but also dissociation and secretion of YscU CC is a key step in the regulation of Yop secretion.

Discussion
The YscU protein of Yersinia pseudotuberculosis and orthologs in other bacteria display autoproteolytic activity, with cleavage at a conserved NqPTH motif. It is reasonable to assume that a strictly conserved autoproteolytic activity in a protein is linked to a specific function in the organism. Auto-processing has been observed in other proteins; e.g., in the SEA domain of the membrane-bound MUC1 protein, the processing occurs at a conserved GDqPH site. It has been suggested that this cleavage introduces a molecular-mechanical fracture that protects epithelial cells from rupture [38]. It was previously postulated that autoproteolysis of YscU exposes a new binding surface to other T3SS proteins by changing the charge distribution at the cleavage site [39]. Here, we propose an alternative model, where dissociation of YscU CC from the membrane anchored segment of YscU, followed by YscU CC secretion via the T3SS, plays a central role in the substrate specificity switch [21].
We showed that T3SS mediated Yop secretion correlated with the secretion of YscU CC and presumed a fully functional T3SS. Secretion of YscU CC required autocatalytic cleavage of the cytosolic domain (YscU C ) of the inner membrane protein YscU that was linked in earlier studies to be one key regulator in the substrate specificity switch of T3SS mediated Yop secretion [17]. In vitro analysis with recombinantly produced YscU C confirmed the dissociation capacity of the protein and revealed potential regulating factors, like divalent ions (i.e. calcium), pH and temperature. We showed that T3SS mediated Yop secretion was strongly affected by 0.5 to 1 mM calcium and pH values below 6.5. The calcium concentration had an ''all or nothing'' effect on Yop secretion. In contrast, the inhibitory effect of low pH values was less pronounced. Surprisingly Ca 2+ -ions bound to YscU C with a K d of 800 mM in vitro; notably, this value was consistent with the threshold value for calcium dependent down-regulation of Yop secretion in vivo. Additional in vitro analysis including different divalent cations Mg 2+ , Sr 2+ and Ba 2+ indicated that the alkaline earth metal ions interacted and stabilized YscU CC binding to about the same extent. These results indicated that the calcium regulation of the Yersinia T3SS in vivo is complex and cannot be explained only on basis of the YscU C calcium interaction. The NqPTH histidine, is unfortunately not suitable for mutations to other amino acids since it affects the autoproteolytic activity [36]. However experiments where histidine 324 was replaced with alanine indicated that, overall, the NqPTH histidine is responsible for the pH-dependency of YscU C dissociation. Thus these findings suggest that pH is also an extracellular queue regulating the Yersinia T3SS.
It is known that the YscP protein plays an important role in the regulation of the substrate specificity switch [17]. We recently proposed that YscP activity may stimulate displacement of YscU CC from the remaining YscU part, and that this activity was triggered by target cell contact or calcium depletion. We further suggested that, in yscU suppressor mutants, the point mutations in YscU CC induced a perturbation in the YscU structure, which destabilized the interaction between YscU CC and the remaining membrane anchored part of YscU [16]. This was suggested because the whole yscP gene was deleted; thus, it was likely that the YscU CC mutations caused a gain of function. In the present study, we confirmed this hypothesis by showing that destabilization of the suppressor mutants led to premature YscU CC dissociation, which then induced Yop secretion. Strains that carried the suppressor mutations secreted Yops at 30uC in calcium-depleted medium. This finding contrasted with findings in the wild-type strain, which showed almost no Yop secretion at 30uC. Nevertheless, all suppressor mutants retained the wild-type calcium regulation of Yop secretion indicating that calcium exhibits a stabilizing effect on the interaction between YscU CC and the remainder of YscU.
Surprisingly, we found that YscU CC was also secreted via the T3SS; this was remarkable, given that YscU is an inner membrane protein [40]. Our results favor a model where YscU orchestrates obstruction of the T3SS secretion channel. This block is thus, relieved through YscU CC secretion in calcium depleted and in pH $6.5 conditions. In support for this model is the observation that yscU mutants impeded for autoproteolysis and subsequent secretion of YscU CC (mutations within the NPTH motif) are unable to secrete the Yop-proteins [18]. Further, these mutants secrete elevated amounts of the early substrate YscF showing that the T3SS is active in these mutants and allows secretion of early but not late substrates [16]. The model explains why these mutants are also defective in Yop secretion. Further support for this model was our finding that the ''Ca 2+ -blind'' yopN mutant, unable to respond to extracellular calcium, secreted YscU CC as well as Yops in Ca 2+ containing conditions indicating that YscU CC secretion is tightly coupled to Yop secretion. Thus, YopN is essential for the calcium response in Yersinia and in addition our results suggest that YopN has a role in the Ca 2+ response in vivo leading to stabilization of YscU. Moreover, Mg 2+ blocks YscU CC displacement in vitro similarly to Ca 2+ but in contrast the pathogen secretes Yops in vivo at high concentrations of Mg 2+ , indicating that Mg 2+ is not interacting with YscU C during in vivo conditions. Thus it is possible that YopN discriminates between the ions, allowing transport/uptake of Ca 2+ but not Mg 2+ during growth at 37uC in vivo. YopN has also been shown to be surface located when Yersinia is grown at 37uC in presence of calcium which agrees with its putative role as calcium scavenger [35].
This study has provided a new handle for investigating the function of YscU/FlhB proteins in other bacteria. Given the conservation within this family of proteins, it is likely that they also exhibit regulatory roles that involve secretion of the processed Cterminus of the protein. However, depending on life style, the actual triggering signal may differ among different pathogens. For example, Salmonella invasion is controlled by environmental pH, low oxygen, and acetate [41]. Although different intracellular regulatory pathways have been linked to these signals, virtually nothing is known about signal reception and interpretation by the pathogen. Based on the results presented here, it would not be surprising to find that some signals stimulated secretion of the Salmonella YscU CC homolog, SpaS CC . Figure S1 Reversibility of the high temperature (776C) transition of YscU C . Thermal up-and down-scans of 10 mM YscU C were monitored with CD spectroscopy at 220 nm. YscU C was subjected to two sequential thermal cycles (one cycle: 20uC to 95uC and then back to 20uC).  Figure S4 Y. pseudotuberculosis growth kinetics were pH independent between pH 6.0 and pH 7.5. Bacterial growth of wild-type Y. pseudotuberculosis was analyzed by monitoring the optical density (OD 600 ) under (A) calcium-supplemented and (B) calcium-depleted conditions. Bacteria were cultivated 2 h at 26uC, then 3 h at 37uC. Samples were taken every hour to monitor the growth based on the measured OD 600 . No significant differences in growth kinetics were observed within the monitored pH-interval of 6.0 to 7.5. (TIF) Figure S5 Secondary structure analysis and dissociation kinetics of YscU C suppressor mutants monitored with CD spectroscopy. (A) Far-UV CD spectra of 10 mM wildtype YscU C (black) and the suppressor mutants A268F (blue), Y287G (green), V292T (red). The similarity in the shape of the CD signals indicates that all YscU C variants have similar secondary structure. (B) Dissociation kinetics of YscU C (black) and suppressor mutants A268F (blue), Y287G (green), and V292T (red) were monitored with CD spectroscopy at 220 nm and 37uC. The solid lines represent the best fit of a single exponential decay function to determine t diss . Dissociation kinetics are summarized in Table 3. Note that the data points at time = zero have been normalized for clarity. (TIF) Figure S6 Primary NMR data used to quantify dissociation kinetics of wild-type YscU C and V292T at 376C. Shown are expansions of methyl resonances from one-dimensional 1 H spectra at various time points for (A) wild-type YscU C and (B) the V292T mutant (see Figure 7B and 7C). (TIF) Figure S7 Secretion analysis of yscU suppressor mutants A268F and V292T. (A) Secretion analysis of A268F and V292T after cultivation in Hepes-buffered LB at different pH values. Yop secretion was induced by calcium depletion and a temperature shift from 26uC to 37uC. A268F and V292T showed elevated Yop secretion at pH$6.5 and strong inhibition at pH 6.0 (B) Secretion analysis of A268F and V292T at pH 7.5 and 30uC compared to wild-type. The T3SS was induced by calcium depletion and a temperature shift to 30uC. A268F and V292T showed a strongly elevation of Yop secretion. Coomassie stained gels demonstrate secreted Yops. ''pellet'' indicates intracellular proteins; ''supernatant'' denotes secreted proteins. The secretion of YscI and LcrV was visualized on immunoblots with anti-YscI and anti-LcrV antibodies. (TIF) Figure S8 Size exclusion chromatography-based estimation of the YscU CC aggregate size. Chromatogram of analytical size exclusion chromatography of purified YscU CC . YscU CC eluted in the void volume of the SEC column. (TIF) Figure S9 CD-based analysis of YscU CC produced by thermal stimulation or by recombinant protein production. Comparison of YscU C CD spectra at 20uC after one completed thermal cycle to 95uC. Filled circles show the YscU CC produced by thermal stimulation and open circles show the purified YscU CC fragment, produced in vitro by recombinant protein production. The similarity of the spectra shows that the residual CD signal of YscU C after one thermal cycle is dominated by the YscU CC fragment. (TIF) Results S1 YscU CC aggregated after dissociation from YscU CN. Analytical size exclusion chromatography was used to show that YscU CC aggregates after dissociation from the YscU CN polypeptide.

(RTF)
Results S2 Persistent secondary structure in aggregated YscU CC from CD spectroscopy. CD spectroscopy on thermally and recombinantly produced YscU CC aggregates was used to show that YscU CC contains elements of secondary structure in the aggregated state. (RTF)