YopD Self-assembly and Binding to LcrV Facilitate Type III Secretion Activity by Yersinia pseudotuberculosis*

YopD-like translocator proteins encoded by several Gram-negative bacteria are important for type III secretion-dependent delivery of anti-host effectors into eukaryotic cells. This probably depends on their ability to form pores in the infected cell plasma membrane, through which effectors may gain access to the cell interior. In addition, Yersinia YopD is a negative regulator essential for the control of effector synthesis and secretion. As a prerequisite for this functional duality, YopD may need to establish molecular interactions with other key T3S components. A putative coiled-coil domain and an α-helical amphipathic domain, both situated in the YopD C terminus, may represent key protein-protein interaction domains. Therefore, residues within the YopD C terminus were systematically mutagenized. All 68 mutant bacteria were first screened in a variety of assays designed to identify individual residues essential for YopD function, possibly by providing the interaction interface for the docking of other T3S proteins. Mirroring the effect of a full-length yopD gene deletion, five mutant bacteria were defective for both yop regulatory control and effector delivery. Interestingly, all mutations clustered to hydrophobic amino acids of the amphipathic domain. Also situated within this domain, two additional mutants rendered YopD primarily defective in the control of Yop synthesis and secretion. Significantly, protein-protein interaction studies revealed that functionally compromised YopD variants were also defective in self-oligomerization and in the ability to engage another translocator protein, LcrV. Thus, the YopD amphipathic domain facilitates the formation of YopD/YopD and YopD/LcrV interactions, two critical events in the type III secretion process.

The ability to evade the innate immune responses of an animal, plant, fish, or insect host is a feature common to many bacteria and is often mediated by homologous type III secretion (T3S) 5 systems, used to deliver immunosuppressive toxins into target host cells (1). Pathogenic Yersinia sp. all harbor a common T3S system (T3SS) consisting of a needle-like complex of numerous Ysc (Yersinia secretion) components. Upon target cell contact, this system secretes two protein classes of Yop (Yersinia outer proteins): the immunosuppressive effector toxins and those required for their efficient delivery into target cells (2). These latter molecules, termed the translocator proteins, integrate into the plasma membrane of infected cells forming a pore through which the secreted toxins may gain entry into the cell cytosol (3)(4)(5). How this pore forms and how it functions to deliver toxins into target cells essentially remain a mystery.
Pathogenic Yersinia possesses at least three translocators, YopB, YopD, and LcrV, that collectively are essential for pore formation and subsequent effector delivery (6 -17). Although complexes among YopD, YopB, and LcrV are known to occur, direct evidence linking these to pore formation and effector delivery has often been lacking (16, 18 -20). These proteins are encoded on the lcrGVHyopBD translocon operon (21), which is similar in gene organization and sequence to operons in other pathogens, including the pcrGVHpopBD operon from the opportunist Pseudomonas aeruginosa. Experiments to compare and contrast Yersinia and P. aeruginosa have proven useful in probing the functions of each operon component (6,(22)(23)(24)(25)(26).
The 306-residue YopD protein is of interest because it is involved in multiple steps in T3S by Yersinia. It is a key negative regulator, controlling the levels of Yop synthesis and secretion (27), which is achieved through formation of a complex with its cognate chaperone, LcrH (28). This complex may bind to the untranslated region of yop mRNA inhibiting translation, perhaps until YopD has been secreted (29,30). It is also possible that the role of YopD in effector delivery extends beyond just forming pores in infected eukaryotic cell plasma membranes, because a portion of the YopD protein pool is also localized to the cytosol of these cells (9). Furthermore, pore formation and effector delivery are separable events, dependent on discrete functional domains (11). Therefore, demystifying how YopD integrates these multiple functions will bridge the gaps in our understanding of the molecular mechanisms underlying protein delivery by T3SS into eukaryotic cells.
So far, only limited structural data are available for YopD or related proteins. A soluble internal 150 -287-residue YopD polypeptide has been found to exist stably in an unfolded state (31). This is comparable with the behavior of the PopD N-terminal region, which exists in a loosely folded, highly flexible, molten, globule state (32). Evidence also indicates that EspB from enteropathogenic Escherichia coli and IpaC from Shigella flexneri both assume a partially folded state with a significant degree of disordered structure (33,34). All of these structural characteristics may confer several benefits to this protein family; limiting the need for unfolding may promote their efficient secretion and/or pore formation into biological membranes or increase the potential surface area to facilitate physical contacts with multiple protein interaction partners (31)(32)(33)(34).
In silico analysis revealed that the YopD family of proteins often possesses C-terminal amphipathic ␣-helices (35) and/or coiled-coil domains (36). Both secondary structures are believed to be important for the full activity of YopD (9,11,26,37), quite possibly through the formation of inter-or intramolecular protein-nucleic acid, protein-protein, or protein-lipid interaction interfaces (38,39). However, the formation of molecular complexes involving YopD has rarely been investigated in a biological context. For this reason, we first performed a widespread site-directed mutagenesis of the YopD C terminus to localize residues critical for YopD function. Although a functional requirement for residues encompassing the putative YopD coiled-coil domain could not be demonstrated, individual substitutions of certain hydrophobic residues in the amphipathic domain altered the ability of YopD to maintain yop regulatory control, form pores, or deliver Yop effectors into infected cells. Significantly, these defects often correlated to an impaired interaction with LcrV or the formation of homo-oligomers. Hence, this study reports on the need for both YopD-YopD and YopD-LcrV complexes at critical stages of T3S by Y. pseudotuberculosis.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Growth Conditions, and Plasmids-Bacterial strains used in this study are listed in Table 1. We used Y. pseudotuberculosis YPIII/pIB102 (serotype III) as the parental strain. The Ysc-Yop T3SS is encoded on pIB102, which also contains a kanamycin resistance cassette inserted into the yadA gene (40,41). This strain also harbors a duplication within phoP, inactivating the PhoP response regulator (42). Unless otherwise indicated, bacteria were routinely cultivated in Luria-Bertani (LB) agar or broth at either 26°C (Y. pseudotuberculosis) or 37°C (E. coli) with aeration. Where required, antibiotics were added at the final concentrations of carbenicillin (100 g/ml), kanamycin (50 g/ml), gentamicin (10 g/ml), tri-methoprim (10 g/ml), and chloramphenicol (25 g/ml). Individual plasmids can be viewed online in supplemental Table S1.
Mutant Construction-Amplified DNA fragments used for constructing the deletion and point mutations were generated by overlap PCR (43). The primer combinations used to create each mutation are listed in supplemental Table S2 and were synthesized by DNA Technology A/S (Aarhus, Denmark), TAG Copenhagen A/S (Copenhagen, Denmark), or Sigma-Aldrich. Amplified fragments were first cloned into pCR4-TOPO TA (Invitrogen) and then sequenced commercially by Eurofins MWG Operon (Ebersberg, Germany). Confirmed fragments were subsequently purified and subcloned into a suitably digested suicide mutagenesis vector, pDM4 (44). Conjugal mating experiments with Y. pseudotuberculosis and selection for the appropriate allelic exchange events used conventional sucrose sensitivity methodology (44).
Protein Stability-Intrabacterial protein stability was assessed by the method of Feldman et al. (45). Protein fractions were analyzed by SDS-PAGE and Western blot. YopD was detected by sequential treatment of nitrocellulose membranes with rabbit ␣-YopD polyclonal antiserum, horseradish peroxidase-conjugated anti-rabbit antibodies (Amersham Biosciences), and a homemade luminol-based detection kit.
Growth Phenotypes and the Magnesium Oxalate Test-Yersinia plating frequencies and subsequent low calcium response (LCR) growth phenotypes (grown under high and low Ca 2ϩ conditions at 37°C) were determined using the magnesium oxalate test (MOX (11,21,46)).ParentalYersinia(YPIII/pIB102)isdefinedascalciumdependent (CD), because it is unable to grow in the absence of Ca 2ϩ at 37°C, whereas the ⌬yopD null mutant (YPIII/pIB621) is termed temperature-sensitive (TS), reflecting its inability to grow at 37°C. In parallel, some mutants were further scrutinized by measuring absorbance at 600 nm during growth in liquid Thoroughly Modified Higuchi's (TMH) medium (ϪCa 2ϩ ) or medium supplemented with 2.5 mM CaCl 2 (ϩCa 2ϩ ) (28,47). Growth phenotypes were scored similarly, with CD and TS defining the opposing extremes.
Analysis of Protein Synthesis and Secretion-Routine analysis of T3S substrate synthesis and secretion was performed using established methods (9,11,28,37). Protein fractions were separated by SDS-PAGE and subjected to immunoblotting, enabling individual proteins to be identified with rabbit polyclonal antisera raised against various Yop substrates or ExoS. The Yop antibodies were a generous gift from Hans Wolf-Watz (Umeå University, Sweden), and the ExoS antibodies were a kind gift from Bengt Hallberg (Umeå University, Sweden). Anti-rabbit secondary antibodies conjugated to horseradish peroxidase were used according to the manufacturer's direction (GE Healthcare). A homemade luminol kit was used for detection.
Cultivation and Infection of HeLa Cells-The cultivation and infection of HeLa cells for cytotoxicity assays was performed using our standard methods (9). At short time intervals following infection, the extent of morphological change was visualized by light microscopy. The cytotoxicity induced by infections with parental Y. pseudotuberculosis (YPIII/pIB102) defined the upper limit. The lower limit was defined by infections with isogenic Yersinia producing a truncated YopD lacking the 278 -292-residue amphipathic domain (YPIII/pIB622). A detailed description of the Ras modification assay, involving the ExoS-mediated ADP-ribosyl modification of eukaryotic Ras in Yersinia-infected HeLa cells, is provided in detail elsewhere (11). Modified Ras was visualized by immunoblotting with a purified mouse anti-Ras antibody (BD Biosciences). Cytosolic levels of Erk, used as a loading control, were detected using a purified mouse anti-Erk1 antibody (BD Biosciences). Contact-dependent Lysis and Pore Formation by YopD-Contact hemolysis of sheep erythrocytes and osmoprotection analysis using the carbohydrates raffinose, dextrin 15, and dextran 4 were performed as described previously (6,11,17,48).
YopD Chemical Cross-linking-In preparation for chemical cross-linking, wild type and mutant yopD alleles were amplified by PCR using the respective Y. pseudotuberculosis parental and mutant strains as a source of template DNA. The primers used are listed in supplemental Table S2. Amplified DNA was cloned into the pET22b(ϩ) (Merck KGaA/Novagen, Darmstadt, Germany) expression vector following NdeI/BamHI restriction. For YopD cross-linking, the expression constructs were transformed into E. coli BL21(DE3). Overnight cultures were back-diluted by a factor of 10 into fresh LB broth supplemented with 0.4 mM IPTG. Bacteria were grown at 30°C to exponential phase (A 600 of ϳ1.0) and then washed twice with phosphate-buffered saline (PBS: 140 mM NaCl, 8.1 mM Na 2 HPO 4 , 2.7 mM KCl, and 1.5 mM KH 2 PO 4 ). Bacterial pellets were resuspended in 0.1 volume of PBS, and 100-l aliquots were generated for cross-linking. A freshly prepared 25 mM solution of cross-linker (dithiobis(succinimidyl propionate) (DSP) and ethylene glycol bis(succinimidylsuccinate) (EGS) (Thermo Fisher Scientific and Pierce Biotechnology)) was added to the bacterial suspension on ice to final concentrations of 0.1, 0.5, 1.0, and 2 mM. For a placebo control, only dimethyl sulfoxide was added to one aliquot. Following a 2-h incubation on ice, the cross-linking reactions were terminated with the addition of Tris-HCl, pH 7.5, to a final concentration of 50 mM. After 15 min, bacteria were harvested by centrifugation and the pellets resuspended in sample buffer (2% SDS, 6.25 mM Tris base, 5% ␤-mercaptoethanol, and 10% glycerol). Because DSP is thiol-cleavable, duplicate samples were also resuspended in sample buffer from which ␤-mercaptoethanol was omitted. Samples were separated on a 12% SDS-polyacrylamide gel followed by immunoblotting with rabbit anti-YopD antiserum.
Construction of pGEX-derived Expression Plasmids-Two sets of plasmid combinations were created, the first based on pGST-LcrV (pMF777) and the second on pGST-YopD (pMF778). All primer combinations are available in supplemental Table S2. The original pGST-LcrV vector (a gift from Hans Wolf-Watz) is a derivative of pGEX-KG. To generate pMF777, a (HindIII-)SalI-SpeI-KpnI-SacI-NotI-HindIII polylinker was inserted into a HindIII opening downstream of lcrV. The cloning was such that only the 3-prime HindIII site was retained. Expression vectors allowing for the co-production of both GST-LcrV with YopD (pMF781) and GST-LcrV with LcrH and YopD (pMF782) were established as follows. A SalI-NotI PCR-amplified fragment of yopD with the Shine-Dalgarno sequence was cloned downstream of lcrV to complete pMF781. To generate a contiguous lcrHyopD coding sequence, overlap PCR was employed, deleting the yopB sequence lying between lcrH and yopD. This SalI-NotI fragment of lcrHyopD with the Shine-Dalgarno sequence was cloned downstream of lcrV to complete pMF782. Control vectors lacking gst::lcrV were generated by cloning the yopD (pMF779) and lcrHyopD (pMF780) fragments into pGEX-5X-3.
To establish the gst::yopD-expressing vector pMF778, a BamHI-SalI PCR-amplified product was cloned in-frame into pGEX-5X-3. Three derivatives of pMF778 were then constructed to permit co-production with YopB and/or LcrH. yopB was amplified on a SalI-NotI fragment, either alone or together with an upstream-located lcrH, and with the Shine-Dalgarno sequence was cloned downstream of yopD, giving rise to pMF794 and pMF795, respectively. Similarly, lcrH was also cloned on a SalI-NotI fragment to give pMF793. Once again, control vectors lacking gst::yopD were generated by cloning the lcrH (pMF789), yopB (pMF790), and lcrHyopB (pMF791) SalI-NotI PCR-amplified fragments into pGEX-5X-3. The creation of equivalent expression vectors with various derivatives of the yopD allele encoding for mutants within the amphipathic ␣-helical domain followed the same protocols used to clone wild type yopD.
Production and Affinity Chromatography of Hybrid GST Fusion Proteins-For analysis of GST-LcrV/YopD and GST-YopD/LcrH interactions in Y. pseudotuberculosis, cultures incubated overnight at 26°C in LB broth supplemented with 2.5 mM CaCl 2 were diluted to absorbance at 600 nm (A 600 ) of 0.2 in 5 ml of fresh medium containing the appropriate antibiotics. The cultures were then incubated with aeration at 26°C until an A 600 of 0.3-0.4 was reached. IPTG was then added to a final concentration of 0.4 mM, and the incubation continued until the cultures had attained an A 600 of 0.6 -0.7. All cultures were then standardized according to this specific absorbance, harvested by centrifugation, and sonicated for 2 min in 1 ml of ice-cold PBS additionally supplemented with 0.5 M NaCl and Complete EDTA-free protease inhibition mixture (as directed by the manufacturer (Roche Applied Science). Cell debris and insoluble material were pelleted in a microcentrifuge at maximum speed for 20 min at 4°C. The total protein concentration of the clear soluble extract was standardized by absorbance at 280 nm before 1 ml was incubated with 50 l of glutathione-Sepharose 4B (GE Healthcare) for 2 h at 4°C with gentle agitation. The glutathione beads were then recovered by centrifugation for 4 min at 1500 rpm and washed at least four times with 10 ml of ice-cold washing buffer (PBS with 0.5 M NaCl). Prior to SDS-PAGE and immunoblot analysis, bound protein was recovered from the glutathione beads by incubation in sample buffer for 10 min at 95°C. Recovered GST fusions were detected with purified mouse anti-glutathione antibody (BD Biosciences), whereas non-tagged protein was visualized by rabbit polyclonal antibodies to LcrV, LcrH, YopD, and YopB (Agrisera AB, Vännäs, Sweden). Where appropriate, antimouse or anti-rabbit secondary antibodies conjugated to horseradish peroxidase were used according to the manufacturer's direction (GE Healthcare).
Because of the bactericidal effect of YopB overexpression, analysis of GST-YopD/YopB interactions necessitated some minor modifications to the above procedure. In particular, overnight cultures were diluted into 40 ml of fresh medium, and after IPTG induction, cells were collected in 4 ml of ice-cold PBS with 0.5 M NaCl and sonicated for 6 min. Finally, the cleared soluble extracts (standardized at 4 ml) were incubated with 100 l of glutathione-Sepharose 4B.

RESULTS
Site-directed Mutagenesis of the YopD C Terminus-Phenotypic analysis of YopD variants lacking a putative coiled-coil domain between residues 248 and 277 (36) and the amphipathic domain encompassing residues 278-292 (35,49) suggested that both features were central to YopD-mediated effector delivery into eukaryotic cells (9,11). Furthermore, the hydrophobic residues on one side of the amphipathic ␣-helix primarily contributed to an interaction with the cognate chaperone LcrH (37) and subsequent yop regulatory control (28,29). These two domains may therefore provide an interface to which YopD mediates functionally relevant interactions with other T3S components. Hence, we sought to identify which residues within this C-terminal region were important for YopD function and T3S activity in Y. pseudotuberculosis.
Coiled-coil domains are usually defined by the classic sevenresidue periodicity ((aXXdXXX) n ) where hydrophobic residues predominate at the "a" and "d" positions. In YopD, however, only on two occasions are these positions filled by highly hydrophobic amino acids (valine and methionine), with the remainder occupied by alanine or even the hydrophilic amino acid lysine (Fig. 1B). Additionally, the predicted coiled-coil length is modest, probably consisting of only three to four repeating units. Nevertheless, we performed site-directed mutagenesis over this region, generating in cis mutants that produced YopD variants with an exchange of either two or three residues with alanine (supplemental Fig. S1A). This meant that whenever an alanine appeared in the wild type sequence, it was not manipulated. These 14 bacterial strains were subjected a series of phenotypic tests to analyze the functional status of YopD. Although altered electrophoretic mobility was observed for several of these YopD variants (supplemental Fig. S1A), all were indistinguishable from parental bacteria with respect to yop regulatory control and effector delivery into HeLa cells (Table 2). Hence, using this mutagenic approach we were unable to identify a role for the predicted coiled-coil domain of YopD, at least according to the routine functional assays utilized in this study. Indeed, on the basis of this evidence and in the absence of any tertiary structural information, it is difficult to assert that this is actually a coiled-coil domain.
Next, we focused on the "sidedness" of the 15-residue amphipathic ␣-helix to generate single in cis point mutants in hydrophobic (18 mutants total) and hydrophilic (13 total) amino acids. Each was substituted with alanine or a residue with opposing chemical properties (Fig. 1A). In some cases, the individual mutants were combined forming seven double mutants. These 38 bacterial strains were then subjected to the same series of phenotypic tests to analyze the functional status of YopD. Interestingly, substitution of the hydrophilic residues did not result in the loss of any facet of YopD function (Table 2). In contrast however, substitutions of some hydrophobic residues generated variants that had a profound effect on YopD function and T3S activity. These are described below and, for the first time, enabled identification of biologically relevant YopD-dependent protein-protein interactions.
Hydrophobic Residues within the Amphipathic Domain Are Essential for YopD Function and T3S Activity-Although the mechanisms are poorly defined, YopD is a vital player in the ability of Yersinia to maintain yop regulatory control and to deliver anti-host Yop effectors into target eukaryotic cells (9,11,27,29). Remarkably, single (YopD V284R and YopD I288K ) and double (YopD F280R,V284R , YopD M281K,L285R , and YopD L287R,Y291K ) substitutions of hydrophobic residues within the amphipathic domain caused the most pronounced defects in YopD function ( Table 2). In fact, these five Yersinia mutants behaved somewhat similarly to a complete deletion of yopD or bacteria producing a YopD ⌬278 -292 truncate devoid of the amphipathic domain (9,11). Moreover, the double YopD M281K,L285R and YopD L287R,Y291K substitutions appeared to represent an additive effect, because the respective single mutants behaved more like parental bacteria ( Table 2).
We first examined Yop synthesis and secretion by these mutant bacteria during growth in BHI medium at 37°C either in the presence (ϩCa 2ϩ ) or absence (ϪCa 2ϩ ) of inducing signals. Clearly, Yop synthesized by these mutants occurred in both Ca 2ϩ replete and deplete media ( Fig. 2A). These Yop levels were comparable to the constitutive synthesis observed by bacteria producing YopD ⌬278 -292 . In contrast, parental bacteria produced abundant amounts of Yop only when Ca 2ϩ was depleted from the growth medium. Additionally, although secretion was typically observed for all strains in the absence of Ca 2ϩ , the mutants also secreted abundant LcrV in the presence of Ca 2ϩ , when the T3SS is normally closed (Fig. 2B). This contrasted with parental bacteria, which secreted only minor  (73), the amphipathic ␣-helical (AH) domain is revealed by a helical wheel projection of residues 278 -292, shown in singleletter code. Hydrophobic residues are boxed, and charged residues are circled. Superimposed on the projection is a summary of the site-directed substitution mutations made in this region. B, by means of the COILS web server (74), the sequence of the predicted YopD coiled-coil (CC) domain is also shown. The letters a and d highlight the characteristic aXXdXXX hydrophobic periodicity of a coiled-coil structure. Moderately hydrophobic amino acids are underlined, and strongly hydrophobic amino acids are double overlined. An earlier study revealed that deletion mutants disrupting this putative coiledcoil domain (YopD ⌬234 -254 and YopD ⌬256 -275 ) prevented effector delivery (11); this region (bounded by a gray box) defines the sequence targeted for specific alanine-scanning mutagenesis.

TABLE 2 Phenotypic characterization of C-terminal mutations in YopD
amounts of LcrV during growth under these same conditions. The altered synthesis and secretion profiles of these bacteria could be corroborated with their growth response to low calcium levels. Specifically, all five mutants were essentially unable to grow at 37°C (TS) regardless of the calcium concentration (Fig. 3, D and E). Temperature sensitivity corresponds to constitutive Yop synthesis. Bacteria producing the YopD ⌬278 -292 truncate were also TS, being unable to grow regardless of the Ca 2ϩ concentration (Fig. 3F). As anticipated, parental bacteria were able to grow at 37°C as long as Ca 2ϩ was present (CD) (Fig. 3A). Calcium dependence for growth mirrors normal yop regulatory control, with Yop synthesis occurring only in the absence of Ca 2ϩ .
To investigate the delivery of Yop effectors into eukaryotic cells, our first approach was to employ a HeLa cell cytotoxicity assay that scores for the extent of cell rounding caused by activity of the internalized YopE cytotoxin. Bacteria producing YopD V284R , YopD F280R,V284R , or YopD L287R,Y291K did not induce any morphological change on infected HeLa cell monolayers, whereas cytotoxicity induced by YopD I288K -and YopD M281K,L285Rproducing bacteria was visibly impaired (Table 2 and data not shown). These results suggest that the five mutant bacteria were defective in their ability to deliver YopE to the cell interior. In a parallel assay, we looked at the ability of mutant bacteria to deliver the ADP-ribosylating ExoS toxin from P. aeruginosa. Internalized ExoS ribosylates intracellular targets such as the Ras family of proteins, but this activity occurs only in the presence of a family of cytosolic host proteins, 14-3-3 (50). Intracellularly localized ExoS can therefore be measured by Western blot as a shift in mobility of modified Ras after SDS-PAGE. For this assay, we introduced the respective yopD mutations into a ⌬yopE,yerA background to avoid any complications that might occur when YopE functions to feedbackinhibit the delivery process (51). Consistent with the previous YopE-dependent cytotoxicity assay, bacteria producing YopD V284R , YopD F280R,V284R , or YopD L287R,Y291K failed to cause any Ras modification, but this effect was notably delayed in YopD I288K -or YopD M281K,L285R -producing bacteria (Fig. 4). These results confirmed that these five mutants possessed general defects in their ability to deliver anti-host effectors inside target eukaryotic cells.
During the delivery process, YopD has been attributed a role in translocon pore formation in the target cell plasma membrane through which the effectors are believed to traverse en route to the cell interior. Contact-dependent lysis of sheep erythrocytes is used as a measure of translocon pore formation, which is quantified by determining the release of hemoglobin into the cleared supernatant of infected erythrocytes. In Yersinia infections, assay sensitivity is improved by using bacteria devoid of YopK, a regulator of pore formation (48). Therefore, our yopD mutations were first introduced into the  Bacteria were grown at 37°C in nonsupplemented Thoroughly Modified Higuchi's medium (without Ca 2ϩ , Ⅺ) or supplemented with 2.5 mM CaCl 2 (with Ca 2ϩ , f). Four different growth phenotypes were detected: CD (A and B); CD-like, moderate calcium-dependent growth (C); TS, bacteria that are sensitive to elevated temperature regardless of the presence of calcium (E and F); TS-like, minimal growth observed in the presence of calcium (D). The parental strain is YPIII/pIB102, and YopD ⌬278 -292 is YPIII/pIB622. Other strains listed contain derivatives of YopD, each containing an amino acid substitution in the hydrophobic side of the C-terminal amphipathic ␣-helix. These are described in more detail under "Results" in the text and/or Table 2.
⌬yopK background before we assessed contact-dependent lysis of red blood cells. Bacteria producing YopD V284R , YopD F280R,V284R , or YopD L287R,Y291K did not cause contact-dependent lysis of infected erythrocytes (Fig. 5A). The mutants producing YopD I288K or YopD M281K,L285R still induced the release of a little hemoglobin (Fig.  5A), and those formed by the latter were significantly smaller (Student's t test, p Ͻ 0.05) than pores created by parental bacteria as assessed by an osmoprotection assay (Fig. 5B). Thus, the inability to form a fully functional translocon pore offers one reason why these bacteria are defective in the delivery process.
These data indicate that a few hydrophobic residues are particularly critical for all-around YopD function. To address whether exchange of these residues destabilizes YopD, we compared the sensitivity of wild type YopD and these variants with endogenous protease digestion. Significantly, all five variants (YopD V284R , YopD I288K , YopD F280R,V284R , YopD M281K,L285R , and YopD L287R,Y291K ) were more sensitive to proteolytic digestion (Fig. 6). Our interpretation of this reduced stability is that these mutants are deficient in either intraor intermolecular interactions that impact their folding. Such folding defects could result in the functionally null phenotypes observed for these mutants.

Specific Loss of yop Regulatory Control in Some Amphipathic
Domain Variants-Up until now it has proven difficult to study one YopD function in isolation from the other. However, a few additional mutants in the YopD amphipathic domain possess a more pronounced defect in the control of Yop synthesis, at least when grown in vitro in defined laboratory media. We examined the effect of generating progressively less conservative substitutions of the phenylalanine residue at position 280. Bacteria producing YopD F280Y , with a conservative exchange of tyrosine for phenylalanine, did not impact on  YopD function in any detectable way. However, substitution with alanine caused these YopD F280A -producing bacteria to produce low amounts of Yop inappropriately during growth under non-inducing conditions ( Fig. 2A). This regulatory defect became even more obvious in bacteria producing YopD F280R , in which a substitution with arginine was used ( Fig.  2A). This incremental regulatory defect was supported by the low calcium growth response of the individual mutants. Expectantly, the YopD F280Y mutant displayed a typical CD phenotype (Fig. 3B). However, the YopD F280A mutant was less able to cope with growth at 37°C, although it still displayed a CDlike pattern of growth (Fig. 3C). In contrast, YopD F280R was much more sensitive to elevated temperature, displaying a TSlike phenotype (Fig. 3D). This scenario was also observed when substituting methionine at position 281; a gross exchange with lysine (YopD M281K ) induced a severe yop regulatory defect, whereas a subtle exchange with alanine (YopD M281A ) did not induce any phenotypically measurable alteration (Figs. 2 and 3). Significantly, these yop regulatory defects (as assessed by in vitro methods) did not perturb effector delivery into eukaryotic cells (Table 2 and Fig. 4). Consequently, these data identify YopD mutants solely defective in yop regulatory control, which should be advantageous for studying the molecular mechanisms of YopD-dependent control of T3S.
Spatial Distribution of Functionally Important YopD Residues within the Amphipathic Domain-Previously, we used circular dichroism and nuclear magnetic resonance spectroscopy to define the solution structure of a peptide encompassing the amphipathic domain of YopD (Protein Data Bank accession code: 1KDL) (49). We used this information to gain a preliminary spatial perspective on those residues critical for YopD function. Shown by red highlighting on a molecular surface diagram of the YopD amphipathic ␣-helix is the co-localization of the Val 284 and Ile 288 residues identified by mutagenesis to be important for the dual function of YopD in yop regulation and effector delivery (Fig. 7). Additionally, the blue highlighting (Fig. 7) show the co-localization of Phe 280 and Met 281 , which are involved predominately in ensuring appropriate YopD-dependent yop regulatory control. The amino acids in this region presumably represent hot spots for the initiation of intra-or intermolecular contacts that are critical for preserving YopD function during T3S by Y. pseudotuberculosis.
Interactions with LcrV Require the YopD Amphipathic Domain-We have identified a number of YopD variants defective in one or more of the following: yop regulatory control, contact hemolysis, and effector delivery. It is likely that these defects are borne out through an inability of YopD to engage in one or more molecular interactions. To initiate an investigation into the protein contacts of YopD, we set up a protein interaction assay based on GST pulldowns. We specifically investigated the previously reported YopD interactions with the cognate chaperone LcrH (18, 37) and the two translocators YopB  When substituted for a residue of opposing properties, these variants are essentially indistinguishable from the full-length yopD null mutant (9,11,27), being defective for both yop regulatory control and Yop effector translocation. Highlighted in blue are the side groups of Phe 280 and Met 281 . When these are substituted for a residue of opposing characteristics, mutant YopD is unable to maintain yop regulatory control, although effector translocation remains essentially unaffected. The images display the molecular surface generated by SwissPdb Viewer (76). Images are related by a rotation out of the page about the horizontal axis by ϳ90°to reveal two sides of the structure. (18,20,52) and LcrV (19). We demonstrated that GST-YopD wt (wild type) pulled down LcrH and/or YopB from Y. pseudotuberculosis-soluble lysates lacking endogenously produced YopD, YopB, LcrH, and LcrV (YPIII/pIB191) (supplemental Fig. S2, A and B, respectively). Interestingly, the GST-YopD wt /YopB interaction protected bacteria from YopB toxicity that occurred when it was produced in isolation (supplemental Fig. S3A) (18). Moreover, GST-LcrV readily bound to YopD wt , in agreement with an earlier study (Fig. 8) (19).
We wondered whether the phenotypic defects of certain YopD variants were due to an inability to interact with one or more of LcrH, YopB, and LcrV. We selected the YopD F280Y and YopD F280A variants that essentially behaved like the wild type, YopD F280R that was predominantly defective in yop regulatory control, and the inactive YopD V284R , YopD I288K , and YopD M281K,L285R variants that possessed prominent defects in both yop regulatory control and Yop delivery. All GST-YopD variants maintained the ability to interact with LcrH (supplemental Fig. S2A). Moreover, all GST-YopD variants also maintained the ability to interact with YopB (supplemental Fig. S2B) and to suppress YopB-mediated toxicity thus permitting continued growth of the host strain (supplemental Fig. S3B). As YopB is a known pore-former (7,17), these data serve to reinforce the notion that a YopD/YopB interaction cannot be the sole requirement for translocon pore formation and Yop effector delivery. Furthermore, given that mutations disrupting the YopD amphipathic ␣-helix still permit YopB binding, this may indicate that residues within this domain are not necessary for mediating YopD/YopB interaction. Rather, the YopB interaction domain within YopD might primarily be located peripheral to the amphipathic ␣-helix. Significantly however, this contrasts with the YopD/LcrV interaction. We observed that GST-LcrV was unable to pull down YopD I288K or YopD M281K,L285R , whereas the recovered levels of YopD F280R and YopD V284R were clearly diminished (Fig. 8). Hence, by analyzing defective YopD variants in this binding assay, the accumulated data define a biologically significant role for the LcrV/YopD interaction in pore formation and effector delivery, which can now definitively be attributed to an LcrV interaction interface within the YopD amphipathic domain. Reduced GST-LcrV/YopD F280R binding also evidently indicates a previously underappreciated role for an LcrV/YopD interaction in yop regulatory control.
YopD Multimerization-Formation of translocon pores in the target cell plasma membrane suggests that the translocators form multimers (53). To determine whether YopD was capable of multimerization, we overexpressed YopD from pET22b(ϩ) under the control of an IPTG-inducible promoter in E. coli. This background was chosen to avoid any contaminating cross-linking events that would otherwise occur between YopD and other interacting components of the Ysc-Yop T3SS produced by Yersinia. Intact bacterial suspensions were incubated in the presence and absence of the membrane-permeable chemical cross-linkers DSP and EGS. Following SDS-PAGE fractionation, boiled lysates were subjected to immunoblotting, and the anti-YopD antiserum was used to visualize YopD. These antibodies recognized a protein band of molecular weight comparable with monomeric YopD when overexpressed in E. coli. Significantly, however, in the presence of increasing concentration of EGS cross-linker (Fig. 9A), and to a lesser extent DSP cross-linker (Fig. 9B), several higher order oligomers were readily detected. Thus, these data indeed indicate that YopD can form multimers.
YopD F280R ) (Fig. 9C), which is in accordance with their level of functional attenuation. Taken together, these data show for the first time that YopD is able to multimerize and that this occurs via the amphipathic ␣-helical domain. Significantly, this multimerization propensity is biologically relevant because it correlates directly with YopD function and T3S activity.
Mutagenesis of the Extreme C-terminal Residues Fails to Impinge on YopD Function-To complete a systematic mutagenesis of the YopD C terminus, we generated alanine substitution mutants in the extreme end of YopD encompassing the region of residues 293-305. Past studies have indicated that this region is sensitive to manipulation. Bacteria producing a YopD ⌬293-305 variant fail to deliver Yop effectors into eukaryotic cells (11), whereas appending epitopes to the YopD C ter-minus also compromise YopD function (unpublished data) (37). We also observed that the helical structure of a peptide encompassing the amphipathic domain was dependent on downstream residues (49). Because helix-capping motifs are found at or near the ends of helices in proteins and peptides (55), the YopD C-terminal residues may function like a clamp to maintain the stability of the amphipathic ␣-helix structure. C-terminal manipulations could therefore result in disruption of this functionally significant domain. To investigate this hypothesis, a total of 16 mutants were generated: 11 single alanine substitutions in positions 293-305 and five double or triple alanine substitutions within the same region. These were assayed for YopD function in T3S. All YopD variants maintained standard levels of YopD production and secretion, yop regulatory control, and effector delivery ( Table 2 and data not shown). Thus, a study of these mutants failed to identify any one extreme C-terminal residue as absolutely required for YopD function. This raises a question as to whether this region acts as a clamp to maintain the ␣-helical structure of the amphipathic domain.
YopD and the T3S Translocon Pore-Of the 68 YopD variants scrutinize in this study, only seven were defective in effector delivery into cells and/or yop regulatory control. However, a contact-dependent hemolysis assay revealed several other mutant bacteria that were significantly impaired (Student's t test, p Ͻ 0.05) in formation of the YopD-dependent T3S translocon pore. Within the amphipathic domain, mutants specifically producing YopD L285A , YopD L287A , YopD L287R , YopD Y291K , YopD V292D , YopD F280A,Q290A , YopD V284A,Y291A , YopD L285A,V292A , or YopD L287A,Y291A were defective in contactdependent hemolysis ( Table 2 and Fig. 5A). Reduced hemolysis could not easily be explained by a reduction in translocon pore size, because an osmoprotection assay with carbohydrates of varying size suggested that the pore diameter was essentially equivalent to that formed by parental bacteria producing native YopD ( Fig. 5B and  . In fact, their only defect was a reduction in the ability to cause contact-dependent lysis of erythrocyte membranes (supplemental Fig. S1B). Osmoprotection analysis even revealed that pores formed by bacteria producing YopD IQV(242-244)AAA were smaller in size compared with those formed by parental bacteria (supplemental Fig. S1C). In addition, of the 16 mutants producing YopD with alterations in the very extreme C terminus, seven of these (YopD H295A , YopD T296A , YopD M299A , YopD HTH295-297AAA , YopD MK299 -300AA , YopD FG303-304AA , and YopD VV305-306AA ) also exhibited considerably less contact-dependent lysis of erythrocytes, although all pores formed were similar in size to those generated by native YopD (Table 2 and data not shown). Despite these multiple examples, in no case did the hemolysis defect impact on the efficiency of any of these mutant bacteria to deliver effector toxins into target eukaryotic cells as measured by two independent assays ( Table 2, Fig. 4, and data not shown). This is quite an extraordinary finding, because these data could hint at an alternative to the notion that effectors are delivered into eukaryotic cells through membrane-inserted T3S translocon pores. At the very least they suggest that T3S systems can tolerate a considerable degree of disruption to the translocon pore assembly process while still maintaining efficient effector delivery into eukaryotic cells.

DISCUSSION
The YopD C terminus is predicted to possess coiled-coil and amphipathic ␣-helical structural domains, which are believed to be functionally relevant. Presumably, these domains function to establish important molecular interactions involving YopD. To begin to identify these critical interactions, we first embarked on a comprehensive mutagenesis screen of the C terminus to discover functionally important YopD residues.
The C-terminal residues most critical for YopD function are the hydrophobic amino acids in the amphipathic ␣-helix. Bacteria producing YopD derivatives solely defective in their ability to mediate yop regulatory control were isolated; others were identified that ranged in their abilities to lyse erythrocyte membranes and to deliver Yop effectors into the cell interior. Significantly, residues that were essential for the full functional repertoire of YopD were grouped together in one functional hot spot within the YopD amphipathic ␣-helical structure. These residues were vital in facilitating two molecular interactions: YopD/YopD and YopD/LcrV. This was the first time that these interactions had been validated in the context of YopD function and T3S activity in Yersinia. Given our limited molecular understanding of how YopD functions as both a regulator and a translocator, this study has revealed YopD mutants that should be useful in future experiments designed to improve the way we view controlled effector delivery via T3SS.
T3SS translocator proteins are indispensable for effector delivery into the interior of eukaryotic cells, presumably functioning as a translocon pore that inserts into the target cell plasma membrane (3,5,56). However, experimental proof that effectors actually passage through this translocon pore en route to the cell interior is not available. Using the contact hemolysis assay, we essentially identified two classes of YopD variants seemingly defective for pore formation in red blood cells: those that could still efficiently deliver effectors into eukaryotic cells and those that could not. Although the latter phenotype aligns with the tenet that effectors pass through the pore on their way to the target cell cytosol, the former does not, thereby opposing current dogma that the so-called translocon pore is necessary for effector delivery. Although YopD and related proteins may well exhibit pore-like effects as measured by standard in vitro assays, this could be an artifact. However, it is also true that these in vitro assays are insufficient for studying the vagaries of T3S pore formation and its configuration. Indeed, translocon pores formed in red blood cell membranes may not be the equivalent of those formed in epithelial cells (57). Moreover, pore formation is a highly dynamic process responding to multiple molecular interactions between individual pore constituents as well as with the target cell plasma membrane. For instance, the YopD and YopB translocators of Yersinia probably do not function alone but require at least transient input from several additional components including LcrV, YopE, YopK, and YopN (23,24,48,51,58,59). However, the consequences of this molecular network essentially remain an enigma because it is difficult to reach meaningful interpretations of data collected from inadequate studies of target cell lysis induced by bacteria lacking individual translocon components. Clearly, therefore, there is a pressing need for improved assays that would subsequently enable the temporal and spatial constraints governing T3S translocon assembly to be investigated.
To address T3S translocon assembly in Yersinia, we identified inactive YopD variants and then determined their ability to associate with fellow pore constituents LcrV and YopB. Although it would have been ideal to analyze the interactions of YopD with LcrV/YopB at the bacterial surface, we were not successful in cross-linking extracellular YopD with any other T3S component, a situation also experienced by others (15). Therefore, we resorted to a GST pulldown assay to examine the ability of intracellular YopD to bind either LcrV or YopB. Although all tested YopD variants still engaged YopB, reduced binding to LcrV correlated with the extent of YopD inactivity. This demonstrates that the YopD amphipathic domain is functionally important because it is needed to establish a critical interaction with LcrV although apparently not with YopB. LcrV is therefore likely to augment YopD function in effector delivery, befitting its strategic location at the "needle" tip (15,60). However, it would be prudent also to examine whether the aforementioned YopE, YopK, and YopN components, which also influence aspects of Yop effector delivery, do so by promoting the extracellular function of YopD via an interaction with its amphipathic domain.
Seeing as oligomeric complex formation may contribute to translocator pore development (53), using in vitro cross-linking studies we also explored whether YopD had a tendency to selfoligomerize. Although native YopD efficiently formed higher order oligomers, this ability was impaired in YopD lacking the amphipathic ␣-helical domain. Substitution of residues within this amphipathic ␣-helix, which caused YopD inactivity, also reduced the propensity for self-oligomerization. The severity of this reduction was directly proportional to the degree of YopD inactivity. Significantly, these data are the first to demonstrate that this highly structured amphipathic ␣-helix (49) is involved in the creation of higher order structures of YopD that are essential for its functionality. Oligomerization of the needle tip of T3SS may act as a host cell sensor and an initiator of the pore formation process (61,62). In Yersinia, therefore, YopD may oligomerize upon secretion as a result of interacting with LcrV at the needle tip or as a consequence of interacting with or inserting into the host cell plasma membrane (in combination with YopB).
YopD oligomerization at the interface between bacteria and the target cell might even constitute a signal perceived by Yersinia to coordinate effector delivery. This signal would probably act through the YscF needle that extends from the bacterial surface. It has been suggested that the YscF needle undergoes conformational change as a means of sensing and transducing signals to the bacterial cytoplasm upon close target cell contact to promote synthesis and secretion of Ysc-Yop T3SS components (63). This must also involve translocated YopE and YopK effectors, which by a poorly understood feedback-inhibitory mechanism restricts the levels of intracellularly delivered Yop effectors following prolonged target cell contact (51). Thus, it could be rewarding to visualize needle-like complexes purified from the surface of our YopD mutants and to examine their molecular composition. After all, a recent study of the Sh. flexneri IpaB translocator gives precedent to the idea that translocator proteins can affect T3S needle composition (57).
At least in vitro, pathogenic yersiniae display a fascinating nutritional dependence for Ca 2ϩ at elevated temperature. At 37°C in the absence of Ca 2ϩ , bacteria cease to divide and replicate DNA. Remarkably, synthesis of the Ysc-Yop T3SS continues under these same conditions (64,65). The molecular basis for this so-called low calcium response (66) remains shrouded in mystery. It might be concerned with defective proton motive force generation, reduced ATP pools, and the inability to export toxic levels of Na ϩ from the cytoplasm (64,67,68). At 37°C Yersinia appears to be sensitive to Na ϩ but only during active T3S; the presence of Ca 2ϩ in the growth medium inhibits Ysc-Yop T3S, concomitantly relieving Na ϩ sensitivity and reinitiating growth (69). It follows that the Na ϩ and temperature sensitivity displayed by a yopD null mutant constitutively producing an active T3SS cannot be suppressed by the addition of Ca 2ϩ (a so-called Ca 2ϩ blind phenotype) (69). This led to a proposal that another component of the Ysc-Yop T3SS may facilitate the elevation of toxic Na ϩ levels inside the bacteria, thus inhibiting a fundamental cellular process needed for vegetative growth (64,69). Indeed, one possible candidate could be YscV (also termed LcrD), a core inner membrane component of the Ysc-Yop T3SS, because its depletion from a yopD null mutant subsequently restored Ca 2ϩ -dependent growth to these bacteria at elevated temperature (69). It is possible that our YopD mutants with intermediary temperature sensitivity, unlike either parental Yersinia or the full-length yopD null mutant, may provide opportunities to better understand this intriguing YopD-YscV regulatory connection.
Finally, another important regulatory control mechanism involves the YopD-LcrH complex (28). This complex seems to cooperate with the LcrQ regulatory element to specifically bind yop mRNA and impose post-transcriptional silencing on Yop synthesis (29,70). We were therefore surprised that GST fusions of YopD variants unable to maintain yop regulatory control could still pull down the cognate LcrH chaperone, especially because earlier results from a yeast two-hybrid analysis indicated that these mutants fail to engage LcrH (37). Most likely, these contrasting readouts are due to inherent differences between the two assays, coupled with the fact that YopD possesses two chaperone-binding domains (CBDs), CBD1 located at the N terminus and CBD2 encompassing the C-terminal amphipathic domain. The pulldown assay was performed in the natural environment of the Yersinia cytoplasm, albeit with higher levels of exogenously expressed protein. In all likelihood, these higher protein levels still enabled LcrH to engage CBD1 of YopD, even though CBD2 was defective in all of the mutants tested. In contrast, the level of protein produced in yeast was comparatively miniscule, and these pools were further diluted by transport to the nucleus, a necessity of the yeast two-hybrid assay. These two constraints probably mean that a YopD/LcrH interaction is detectable in the yeast two-hybrid assay only when both CBD1 and CBD2 of YopD are functionally intact and available to engage LcrH. In view of these discrepancies, it is imperative to scrutinize the YopD-LcrH complex involvement in the recently proposed post-transcriptional regulatory model. For example, neither YopD nor LcrH display any obvious nucleic acid binding motif, so how they specifically engage the yop mRNA template, individually or together as a complex, essentially remains unexplored. Furthermore, it is not yet known whether this binding is influenced by the presence of LcrQ. This is a critical point considering that SycD (synonymous with LRH) can interact directly with YscM2, one of two LcrQ homologues produced by Yersinia enterocolitica (71,72). We therefore propose to utilize infuture studies, the YopD vari-ants isolated from this present study, which are solely exhibiting yop regulatory defects of varying severity, to better define the contributions of YopD (and LcrH) to post-transcriptional silencing of Yop synthesis.