Identification of the MxiH Needle Protein Residues Responsible for Anchoring Invasion Plasmid Antigen D to the Type III Secretion Needle Tip*

The pathogenesis of Shigella flexneri requires a functional type III secretion apparatus to serve as a conduit for injecting host-altering effector proteins into the membrane and cytoplasm of the targeted cell. The type III secretion apparatus is composed of a basal body and an exposed needle that is an extended polymer of MxiH with a 2.0-nm inner channel. Invasion plasmid antigen D (IpaD) resides at the tip of the needle to control type III secretion. The atomic structures of MxiH and IpaD have been solved. MxiH (8.3 kDa) is a helix-turn-helix, whereas IpaD (36.6 kDa) has a dumbbell shape with two globular domains flanking a central coiled-coil that stabilizes the protein. These structures alone, however, have not been sufficient to produce a workable in silico model by which IpaD docks at the needle tip. Thus, the work presented here provides an initial step in understanding this important protein-protein interaction. We have identified key MxiH residues located in its PSNP loop and the contiguous surface that uniquely contribute to the formation of the IpaD-needle interface as determined by NMR chemical shift mapping. Mutation of Asn-43, Leu-47, and Tyr-50 residues severely affects the stable maintenance of IpaD at the Shigella surface and thus compromises the invasive phenotype of S. flexneri. Other residues could be mutated to give rise to intermediate phenotypes, suggesting they have a role in tip complex stabilization while not being essential for tip complex formation. Initial in vitro fluorescence polarization studies confirmed that specific amino acid changes adversely affect the MxiH-IpaD interaction. Meanwhile, none of the mutations appeared to have a negative effect on the MxiH-MxiH interactions required for efficient needle assembly.

The pathogenesis of Shigella flexneri requires a functional type III secretion apparatus to serve as a conduit for injecting host-altering effector proteins into the membrane and cytoplasm of the targeted cell. The type III secretion apparatus is composed of a basal body and an exposed needle that is an extended polymer of MxiH with a 2.0-nm inner channel. Invasion plasmid antigen D (IpaD) resides at the tip of the needle to control type III secretion. The atomic structures of MxiH and IpaD have been solved. MxiH (8.3 kDa) is a helix-turn-helix, whereas IpaD (36.6 kDa) has a dumbbell shape with two globular domains flanking a central coiled-coil that stabilizes the protein. These structures alone, however, have not been sufficient to produce a workable in silico model by which IpaD docks at the needle tip. Thus, the work presented here provides an initial step in understanding this important protein-protein interaction. We have identified key MxiH residues located in its PSNP loop and the contiguous surface that uniquely contribute to the formation of the IpaD-needle interface as determined by NMR chemical shift mapping. Mutation of Asn-43, Leu-47, and Tyr-50 residues severely affects the stable maintenance of IpaD at the Shigella surface and thus compromises the invasive phenotype of S. flexneri. Other residues could be mutated to give rise to intermediate phenotypes, suggesting they have a role in tip complex stabilization while not being essential for tip complex formation. Initial in vitro fluorescence polarization studies confirmed that specific amino acid changes adversely affect the MxiH-IpaD interaction. Meanwhile, none of the mutations appeared to have a negative effect on the MxiH-MxiH interactions required for efficient needle assembly.
Shigella flexneri is a Gram-negative bacterium and the causative agent of shigellosis, a potentially life-threatening form of bacillary dysentery. Although shigellosis is often considered a disease restricted to the developing world, it is also a problem in industrialized nations as an infectious agent in daycare centers, nursing homes, and situations where public health infrastructure becomes compromised (1). The problems posed by shigellosis are compounded by increasing numbers of antibiotic-resistant isolates and the need to send aid personnel to endemic areas (2).
Once ingested, Shigella passes through the stomach to the colonic mucosa where it crosses M cells into gut-associated lymphoid tissues (3). The bacterium then enters macrophages and induces them to undergo apoptosis (4), which leads to release of the pathogen at the basolateral side of the colonic epithelium. Shigella then promotes its own uptake into epithelial cells by inducing cytoskeletal rearrangements in these cells (5,6). The pathogenesis of S. flexneri requires a functional type III secretion system (6). Type III secretion systems are a shared virulence determinant for many Gram-negative animal and plant pathogens, and they are used to deliver protein effectors to targeted cells to subvert normal cellular functions (7,8). The type III secretion apparatus (TTSA) 2 visually resembles a molecular needle and syringe (9,10). It is composed of a basal body that spans both bacterial membranes and an exposed needle that is a polymer of MxiH. The needle is ϳ50 nm in length, has an outer diameter of 7.0 nm, and possesses an inner channel that is ϳ2.0 -3.0 nm in diameter (10,11). We recently demonstrated that invasion plasmid antigen D (IpaD) localizes to the tip of the TTSA needle (12), where it controls secretion of the translocator proteins IpaB and IpaC and directs their insertion into host cell membranes (13). From its position at the TTSA needle tip, IpaD is responsible for recruiting and stably maintaining IpaB to this site in the presence of bile salts (14). This is an essential step for Shigella invasiveness, and the interactions between IpaD and MxiH that are responsible for tip complex stabilization are thus critical for Shigella pathogenesis (12,14).
The high resolution structures of MxiH and IpaD were recently reported (15,16). MxiH monomers possess a helixturn-helix head group with an elongated C-terminal helix (15). The head group is stabilized by hydrophobic contacts of residues at the interface between the two helices that flank a central PSNP loop (15). Likewise, the TTSA needle appears to pack by polar contacts between the individual MxiH monomers with the PSNP loop facing outward and upward relative to the central channel (15). In contrast to MxiH, IpaD has a more complex structure that gives it a dumbbell shape (16). IpaD has a central coiled-coil that defines the handle of the dumbbell and is flanked at each end by distinct N-and C-terminal domains (16). This shape defines the tip proteins as a unique class (17) and appears to be important for IpaD localization at the tip of the needle (12). Based on in silico modeling of the Yersinia tip complex protein LcrV on the Shigella MxiH needle tip, the LcrV N-terminal globular domain and the C-terminal end of the coiled-coil are both required in docking LcrV on the needle (15). Although the IpaD structure shows that the N-terminal domain can mimic a MxiH molecule and possibly serve to chaperone the C-terminal coil to prevent the binding of MxiH and IpaD in the cytoplasm, how IpaD docks to MxiH could not be resolved even after the structure of the two proteins became available (16). Nevertheless, interactions in addition to those that dictate needle assembly would be expected to be involved in maintaining IpaD at the needle tip. We therefore chose to examine the structural basis for the interaction between MxiH and IpaD. The data presented here show that specific residues within the MxiH head group are located at the MxiH-IpaD interface. These residues are needed for stable maintenance of IpaD at the TTSA needle tip but are distinct from those needed for needle assembly.

EXPERIMENTAL PROCEDURES
Materials-Antibodies to MxiH and IpaD were a kind gift from E. V. Oaks (Walter Reed Army Institute for Research, Silver Springs, MD). The S. flexneri mxiH null strain SH116 was from A. Allaoui (Brussels, Belgium) and grown at 37°C in trypticase soy agar containing 0.025% Congo red to ensure the presence of the virulence plasmid. Sheep red blood cells were from Colorado Serum Company (Denver, CO). Fluorescent probes and Alexa Fluor-labeled secondary antibodies were from Molecular Probes (Eugene, OR). Escherichia coli Tuner(DE3), pET15b, pET22b, and ligation reagents were from Novagen (Madison, WI). Restriction enzymes were from New England Biolabs (Tozer, MA). Oligonucleotide primers were from IDT (Coralville, IA). 15 N-Ammonium chloride and 13 C 6 -glucose were from Cambridge Isotope Laboratories, Inc. (Andover, MA). Iminodiacetic acid affinity resin was from Sigma. Other chemicals were reagent grade.
Protein Expression and Purification-Recombinant IpaD was prepared by standard methods as previously described (18). The mxiH⌬5/pET22b was subcloned from pRKmxiH C⌬5 (19,20). This was the construct used to solve the crystal structure of MxiH C⌬5 (15). To maintain MxiH in monomeric form, five codons have been deleted from the 3Ј-end of the mxiH coding sequence, mxiH C⌬5 , giving rise to MxiH C⌬5 . This mutation also prevents polymerization of needles in vivo, resulting in secretion of MxiH C⌬5 monomers but a phenotype that is otherwise like a mxiH null mutant (19). To obtain uniformly 15 N, 13 Clabeled MxiH C⌬5 , mxiH⌬5/pET22b in E. coli Tuner(DE3) was grown in minimal medium supplemented with 15 N NH 4 Cl and 13 C-glucose as the sole nitrogen and carbon sources, respectively (21). After reaching mid log phase, bacteria were induced to produce protein with 1 mM isopropyl-1-thio-␤-D-galactopyranoside and grown at 15°C overnight. Bacteria were collected, resuspended in IMAC binding buffer (20 mM Tris⅐HCl, pH 7.9, 50 mM NaCl, 5 mM imidazole) and sonicated, and the cellular debris was removed by centrifugation. Recombinant protein was purified via His 6 tag by IMAC chromatography as described (20,22). Fractions were pooled and dialyzed into appropriate buffers (see below), and the protein concentration was estimated by UV absorbance at 280 nm.
NMR Spectroscopy-Uniformly 15 N-or 15 N, 13 C-labeled MxiH C⌬5 and unlabeled IpaD were dissolved in NMR buffer (10 mM sodium phosphate, pH 6.0, 10 mM NaCl, 10% D 2 O, and 90% H 2 O). NMR data were acquired at 25°C on a Bruker Avance 800 MHz spectrometer fitted with a cryogenic triple-resonance probe and a Z-axis pulse field gradient. NMR data were processed using NMRPipe (23) and analyzed using NMRView (24). Backbone resonances of free MxiH C⌬5 were assigned using 1 H-15 N HSQC, HNCA, CBCA(CO)NH, and HNCACB acquired using a 0.8 mM 15 N, 13 C-MxiH C⌬5 sample as described for BsaL C⌬5 (21). Chemical shift mapping was done by acquiring two-dimensional 1 H-15 N HSQC at MxiH C⌬5 :IpaD molar ratios of 0.5, 1, 2, 4, and 8. Both IpaD and MxiH C⌬5 were dialyzed into the same buffer prior to titration to assure that the chemical shift changes were only due to protein-protein interactions and not to changes in solution condition upon mixing.
Preparation of mxiH Mutants for Expression in S. flexneri SH116-pRKmxiH containing the mxiH gene has been described (19). mxiH mutants were made by inverse PCR using pRKmxiH as template, a primer composed of GAGAGA, a restriction site, and 18 nucleotides flanking the nucleotides to be deleted and a primer going the opposite direction encoding GAGAGA, a restriction site, and 18 nucleotides (19). The generation of MxiH Q51A has been described (19). The primers used to introduce mutations into MxiH are given in supplemental  Table S1. The PCR product was digested with the appropriate restriction enzyme, intramolecularly ligated, and used to transform E. coli NovaBlue. Double-stranded sequencing was performed on each new mutant mxiH gene to ensure only the desired mutation was incorporated. Each plasmid was electroporated into S. flexneri SH116. Ampicillin selection ensured presence of the plasmid while kanamycin resistance and/or Congo red binding indicated the presence of the Shigella virulence plasmid.
Phenotypic Characterization of S. flexneri SH116 Expressing Different mxiH Mutants-The phenotype of each newly generated mxiH point mutant strain of Shigella was determined by standard assays. Shigella invasion functions were tested using a gentamycin protection invasion assay with cultured Henle 407 as described (25). Contact-mediated hemolysis with sheep erythrocytes was used to measure the IpaD-dependent insertion of IpaB and IpaC into target cell membranes as described (25). The overnight secretion of Ipa proteins into Shigella culture supernatants was used to ascertain the presence of functional TTSAs and, more specifically, the assembly of functional MxiH needles with IpaD at the tip as recently described (12). Induction of type III secretion was tested using Congo red as an artificial inducer of secretion (19). In S. flexneri, type III secretion occurs upon host cell contact; however, this event can be mimicked by the addition of Congo red (26). The surface local-ization of IpaD was determined using immunofluorescence microscopy while the ability to polymerize MxiH needles was determined by transmission electron microscopy as recently described (12,19).
Preparation of mxiH C⌬5 Mutants for Expression in E. coli Tuner(DE3)-To produce monomeric protein, 5 residues were deleted from the C terminus of MxiH (15). Therefore, the mutant mxiH C⌬5 genes were copied from the corresponding pLZ plasmid by PCR using a 5Ј-primer containing GAGAGA, an NdeI restriction site, and the first 18 bases of mxiH and a 3Ј-primer containing GAGAGA, a XhoI restriction site, and the last 18 bases of mxiH excluding the stop codon and the last 5 amino acids. The PCR product was digested with NdeI and XhoI and ligated into correspondingly digested pET22b, and the ligated product was transformed into E. coli NovaBlue. Double-stranded sequencing was performed on each new mutant mxiH C⌬5 gene to confirm that only the desired mutation was incorporated. The resulting plasmid was used to transform E. coli Tuner(DE3). This strain was used for purification of the mutant MxiH C⌬5 proteins as previously described (22).
Fluorescence Polarization-The interaction of IpaD with MxiH C⌬5 and MxiH C⌬5 mutants was measured by fluorescence polarization (FP) (27). FP provides a measure of the rotational diffusion for a fluorescent molecule in solution. Small molecules rotate quickly, which produces rapid depolarization of the population of fluorescent molecules selected using excitation by polarized light, thus giving rise to low millipolarization (mP) values. In contrast, larger molecules rotate more slowly and display a reduced rate in depolarization in aqueous solution, thereby leading to larger mP values. The association of a small fluorescent species with a larger nonfluorescent species results in a significant increase in overall molecular volume, thus resulting in an increase in the mP of the fluorescent species.
IpaD was labeled with fluorescein at its single Cys residue by combining 1 ml of IpaD (3 mg/ml) in phosphate-buffered saline (pH 7.5) with 50 l of fluorescein maleimide (FM) prepared as a 1 mg/ml stock prepared in N,N-dimethylformamide. After incubating the mixture for 1 h at 4°C, the reaction was stopped by adding dithiothreitol to 1 mM. The FM-IpaD was isolated from the free fluorophore by gel filtration using Sephadex G-50. FP measurements were obtained using a Beacon fluorescence polarimeter (Panvera Corp.) as described (27). Briefly, a constant concentration of FM-IpaD (20 nM) in phosphate-buffered saline was mixed with increasing concentrations of MxiH C⌬5 . The change in FP, measured in millipolarization units (⌬mP), was monitored after adding nonfluorescent MxiH C⌬5 .

RESULTS AND DISCUSSION
MxiH C⌬5 Binding to IpaD-We recently showed that IpaD localizes to the tip of the TTSA needle (12). Although our findings show that the C-terminal coil of IpaD is responsible for its maintenance at the MxiH needle tip, no such data are available to directly demonstrate the structural role of the TTSA needle protein in the IpaD-MxiH interaction. Using FP to detect in vitro protein-protein interactions, a direct interaction between MxiH C⌬5 and IpaD was identified (Fig. 1). IpaD was labeled with FM at Cys-322, which is the only free thiol in IpaD. Cys-322 is 10 residues away from the IpaD C terminus, and it is this portion of IpaD that mediates its stable association with the Shigella TTSA needle tip (12). The initial polarization value for IpaD was 247.9 Ϯ 0.7, a high value that is consistent with the attachment of the probe to a large (36.6-kDa) protein. FM-IpaD association with MxiH C⌬5 resulted in a measured increase of 21.2 mP whereas no change was seen upon the addition of IpgC, the bacterial cytoplasmic type III secretion system translocator chaperone (Fig. 1). Because the maximum ⌬mP achieved by the addition of anti-IpaD IgG is 30.8, the change in polarization upon addition of MxiH C⌬5 represents a significant change based on the fact that its mass is only one fourth that of IpaD. The fact that a detectable change is observed most likely stems from the previous observation that IpaD association with the MxiH needle involves the C terminus of IpaD (12) and this is near where the FM labeling occurs. The data presented thus indicate that IpaD and MxiH C⌬5 interactions are detectable in vitro. Furthermore, the binding data suggest that the two proteins have an apparent dissociation constant (k d ) of ϳ0.48 M (Fig. 1).
Identification of MxiH C⌬5 Residues Interacting with IpaD by NMR Spectroscopy-We used NMR chemical shift mapping to identify the MxiH C⌬5 residues that contact IpaD based on the reasoning that the MxiH C⌬5 -IpaD interaction should result in changes in the NMR resonances of MxiH C⌬5 . Because the smaller MxiH C⌬5 (8.3 kDa) is more amenable to NMR characterization compared with IpaD, we titrated unlabeled IpaD into 15 N-labeled MxiH C⌬5 and monitored the changes in the chemical shifts of MxiH C⌬5 by acquiring several 15 N HSQC spectra with increasing amounts of IpaD. Indeed, there are NMR peaks of MxiH C⌬5 that shift upon adding IpaD (Fig. 2). We assigned the resonances of free MxiH C⌬5 using standard three-dimensional NMR experiments (supplemental Fig. S1). Assigning the NMR peaks of free MxiH C⌬5 was challenging because of the broad peaks between 7.8 -8.5 ppm in the center of the twodimensional 15 N HSQC spectra (Fig. 2A). The narrow proton chemical shift range is consistent with the predominantly ␣-helical structure of MxiH C⌬5 (15), and the disordered N-terminal 19 residues in the x-ray structure of MxiH C⌬5 (15) likely contribute to the broad peaks between 7.8 -8.5 ppm. Nevertheless, using three-dimensional HNCA, CBCA(CO)NH, and  HNCACB (supplemental Fig. S1), we assigned 62 of 78 backbone amides of MxiH C⌬5 .
Upon adding IpaD to labeled MxiH C⌬5 , many MxiH C⌬5 peaks shifted. Further additions of IpaD changed peak positions even more, indicating that binding was in fast exchange NMR time scale. This behavior is consistent with the micromolar binding affinity obtained by fluorescence polarization. The peaks of the PSNP loop residues (Asn-40, Ser-42, Asn-43, and Gln-45) and the head group residues (Leu-26, Leu-30, Leu-32, Ala-33, Leu-34, Tyr-50, Gln-51, Leu-54, and Tyr-57) clearly shifted to a different part of the spectrum (Fig. 2). The latter set of residues form a contiguous surface (Fig. 2, B and C) when mapped on the x-ray crystal structure of MxiH C⌬5 (15). The side chains of Leu-30, Leu-34, Tyr-50, Gln-51, Leu-54, and Tyr-57 form one face of the contiguous surface (Fig. 2B), and the side chains of Leu-26, Leu-32, and Ala-33 are on the "backside" (Fig. 2C). Therefore, we propose that the main site of IpaD binding on MxiH C⌬5 is the surface formed by residues in the contiguous surface and the PSNP loop.
In Vivo Analysis of MxiH Mutant Strains-We have also done NMR chemical shift mapping of the MxiH homolog from Burkholderia pseudomallei, BsaL, and found that the BsaL residues equivalent to MxiH Leu-30, Leu-54, and Tyr-57 are part of the needle monomer-monomer surface contact. 3 Therefore, we propose that there are two overlapping sites on MxiH: one for IpaD binding and another for MxiH binding and that the Leu-30-Leu-54-Tyr-57 surface (see Fig. 2B) is common for both IpaD and MxiH binding. Hence, MxiH mutations were designed to disrupt IpaD binding while maintaining the MxiH-MxiH association. Not all residues that showed chemical shift changes upon IpaD addition were mutated. Gln-51 was previously mutated (see Table 1) and it had no effect on IpaD association with the Shigella surface (12,19). Mutation of Ser-42 was also shown to have no effect on MxiH function (19). Residues that are not conserved among needle proteins from different pathogens were not mutated since the side chain composition at these positions does not appear to be critical. This was the case for Leu-26, Leu-32, Ala-33, and Gln-45 (15,28). Where NMR assignments were lacking, mutations were not introduced because subsequent interpretation of the data would not be possible. Most importantly, those residues that appeared to have a role in the packing of MxiH monomers 3 were not mutated because the mutations would disrupt needle assembly. Therefore, we did not mutate the MxiH residues (Leu-30, Leu-54, Tyr-57) that appeared to be necessary for MxiH-MxiH contacts. In contrast, Leu-47 showed minor chemical shift changes; however, it is next to the PSNP loop and Tyr-50, and it is invariant among the needle proteins of Shigella, Salmonella, and Burkholderia. In the MxiH structure (15), the side chain of Leu-47 (which is on helix ␣ 2 ) is oriented toward the surface of the two-helix bundle (Fig. 2B). This suggests that Leu-47 is not needed in maintaining the hydrophobic core that dictates the overall fold of the MxiH structure and thus may be involved in IpaD binding. Therefore, Leu-47 was mutated.
Asn-40 and Asn-43 of the MxiH PSNP loop and Leu-34, Leu-47, and Tyr-50 of the contiguous surface were selected for sitespecific mutations. Mutations were made to give MxiH L34A , MxiH N40A , MxiH N43A , MxiH N43K , MxiH L47A , MxiH L47D , MxiH Y50F , MxiH Y50L , and MxiH Q51A (Table 1). For residues Asn-43, Leu-47, and Tyr-50, two substitutions per residue were introduced to explore the importance of side chain composition in needle tip association. MxiH N43A and MxiH N43K were designed to test the impact of replacing a polar residue with either a small nonpolar or a somewhat larger, positively charged residue. MxiH Y50F kept the phenyl ring while MxiH Y50L maintained a hydrophobic side chain. In contrast, the MxiH L47A mutation decreased the length of the hydrophobic side chain while MxiH L47D changed the polarity of the region by adding a negative charge.
The ability of the MxiH mutants to complement the S. flexneri mxiH null strain SH116 was assessed using established phenotypic assays with the data summarized in Table 2 (12). First, invasion of Henle 407 cells and contact-mediated hemolysis were used to detect the proper introduction of IpaB/IpaC translocons into targeted cells, which is dependent upon IpaD localization to the TTSA needle tip (10,(12)(13)(14). Second, overnight secretion of Ipa proteins provided a measure of functional needle formation. Excessive overnight secretion can provide a measure of the loss of IpaD at the needle tip because ipaD null mutants and IpaD mutants that fail to localize at the needle tip are known to be hypersecretive (12,13,29). Third, induced (rapid) secretion of Ipa proteins into culture supernatants upon adding Congo red is lost in ipaD null mutants and this correlates with the loss of IpaD, but not MxiH, at the bacterial surface (12). In contrast, loss of needle polymerization results in no Ipa Previously shown to retain function and does not interfere with IpaD localization at needle tip (12,19) a Mutations were introduced into full-length MxiH for complementation analysis using S. flexneri SH116. A MxiH C⌬5 version with a C-terminal His 6 was then made for L47A, L47D, Y50F, Y50L, and Q51A for generation of recombinant protein in E. coli BL21(DE3). The recombinant protein was used for analyzing interactions with FM-labeled IpaD by fluorescence polarization. b Chemical shifts refer to the NMR shifts described in Fig. 2. Conserved is a measure of the sequence conservation seen among MxiH and its needle protein homologues from Salmonella typhimurium (PrgI) and B. pseudomallei (BsaL).
secretion or detection of MxiH on the bacterial surface (12,14,19). Thus, if a MxiH mutant fails to maintain IpaD at the needle tip, there will be no invasion, no contact hemolysis, uncontrolled overnight secretion, a loss of Congo red-induced Ipa protein secretion, and no IpaD detected on the bacterial surface (12). Because the MxiH mutants were designed not to interfere with MxiH polymerization, however, MxiH will be detected on the Shigella surface and it should be possible to overexpress mxiH to form long versions of the TTSA needle that can be sheared from the bacterial surface. The phenotypes of the mxiH mutant strains are presented in Table 2.
Asn-40 and Asn-43, which are part of the PSNP loop, were mutated because they showed some of the largest chemical shift perturbations upon addition of IpaD (Fig. 2). Previous results concluded that the polar residues in the PSNP region can tolerate mutation to Ala and still give rise to functional needle tip complexes (19). Similarly, although contact-mediated hemolysis was reduced, N40A and N43A had minimal effects on the phenotypes that measure the needle tip complex formation ( Table 2). In contrast, mutation of Asn-43 into the somewhat larger and basic residue Lys decreased needle tip complex formation ( Table 2). The decreased ability for this strain to stably maintain IpaD at the needle tip (50% reduction) resulted in lack of secretion control as well as inability to invade Henle 407 cells and lyse erythrocytes (Table 2). Thus, the PSNP loop interacts directly with IpaD and is therefore required for the formation of the MxiH-IpaD tip complex.
Adjacent to the PSNP loop, Leu-34, Leu-47, Tyr-50, and Gln-51 form a contiguous surface on the two-helix bundle and were perturbed upon adding IpaD (Fig. 2B). Although the chemical shift of Leu-34 was small, it lies within the contiguous surface, suggesting that Leu-34 could be involved in IpaD binding. Indeed, the L34A mutation partially reduced invasion and hemolysis, suggesting that it has a minor effect on IpaD binding to the needle. Q51A has been characterized (19) and found to have only a partial defect in secretion control, indicating a minor deficiency in IpaD affinity for MxiH.
In contrast to Leu-34 and Gln-51, mutation of Leu-47 and Tyr-50 had significant consequences with regard to S. flexneri phenotype and the observed interactions with IpaD in vitro. It is also noteworthy that Leu-47 and Tyr-50 are highly conserved among the needle proteins from different species. Both L47A and L47D mutants abolished IpaD surface presentation, resulting in a noninvasive, nonhemolytic strain that also completely lacks secretion control (Table 2). These mutations, however, did not affect MxiH-MxiH interactions since they were capable of needle formation as detected by immunofluorescence microscopy (Table 2) as well as electron microscopy (see Fig. 3A for L47D). When recombinant MxiH C⌬5 proteins possessing mutations at Leu-47 were prepared so that their interaction with FM-labeled IpaD could be monitored in vitro using FP, a measurable decrease in binding to IpaD was observed (Table 3). Thus, Leu-47 is critical for IpaD-MxiH binding and the proper assembly of IpaD at the needle tip.
Tyr-50 is next to Leu-47 on helix ␣ 2 of MxiH and its phenyl ring forms a network of hydrophobic contacts with residues on helix ␣ 1 that contribute to the stability of the two-helix bundle (15). The hydroxyl group of Tyr-50, however, is pointed toward the MxiH surface and is not involved in any intramolecular contacts (15). Therefore, the Tyr-50 hydroxyl group is not a Wild type is pRKMxiH/SH116, the mxiH null mutant strain is SH116, and the ipaD null mutant is SF622. All other strains are SH116 transformed with the plasmid encoding one of the MxiH point mutants as described here. b Surface localization of IpaD and MxiH was determined by immunofluorescence microscopy. At least 100 bacteria in each of three randomly selected fields were counted (n ϭ 3). In most cases, surface localization was either seen on all bacteria or none in a given field. For some mutants, reduced overall fluorescence was visualized and a reduced number of bacteria were seen with IpaD on their surfaces. In these cases, an approximate percentage of cells having surface IpaD is given. c Invasion is relative to pRKMxiH/SH116 Ϯ S.D. (n ϭ 3). d Contact-mediated hemolysis is relative to pRKMxiH/SH116 Ϯ S.D. (n ϭ 3). e Uninduced secretion is the secretion of TTSS proteins (IpaB and IpaC) into the culture medium during overnight growth. Induced secretion is the secretion of TTSS proteins by bacteria after a 30-min incubation with Congo red (12,19 required in stabilizing the overall fold of the two-helix bundle. Structurally, a phenylalanine or leucine in this position would be expected to maintain the hydrophobic core and the overall fold of the two-helix bundle, yet Tyr-50 is invariant among the needle proteins of Shigella, Salmonella, and Burkholderia. Thus, the mutations Y50F and Y50L were made and the phenotypes of each mutant strain determined (Table 2). Although not as severe as the Leu-47 mutations, Y50L and Y50F resulted in a 50 -80% reduction in IpaD surface presentation, which negatively affects invasion, hemolysis, or secretion control ( Table 2). Our data, however, show that both Y50F and Y50L form stable needles in vivo when examined by either immunofluorescence ( Table 2) or electron microscopy (Fig. 3). As with Leu-47, the recombinant MxiH C⌬5 proteins possessing the Tyr-50 mutations were purified to examine their binding interaction with FM-labeled IpaD. Indeed, like the Leu-47 mutants, the Tyr-50 mutants have a notable increase in the apparent dissociation constant, with Y50F being 3-fold higher and Y50L being 5-fold higher (Table 3). Thus, although the hydroxyl group has a role in the stable maintenance of IpaD at the TTSA needle tip, the hydrophobic interactions through the phenyl ring appear to contribute more to the in vitro interaction between IpaD and MxiH C⌬5 . Taken together, the data indicate that key residues of the contiguous face and those of the PNSP loop contribute to the proper binding, maintenance, and function of the MxiH-IpaD tip complex (Fig. 4).
Previously, we have shown that the C-terminal coil of IpaD is required for its association with the MxiH needle and that the N-terminal globular domain is external to the needle (12). Modeling using the recently solved individual crystal structures of MxiH and IpaD, however, was unable to explain how IpaD docks atop the MxiH needle other than that the orientation is likely to be the same as that predicted for LcrV (15,16,30). Based on their crystal structures, Johnson et al. (16) proposed a model for the interaction of MxiH with IpaD based on the fact that the IpaD N-terminal domain is structurally similar to MxiH and therefore could serve as a cytoplasmic intramolecular chaperone that prevents the interaction of IpaD with MxiH prior to recruitment of the IpaD to the bacterial surface. The IpaD N-terminal helices H1 and H2 pack together in such a way that the MxiH structure can be superimposed on the MxiH two-helix bundle (␣ 1 and ␣ 2 ) with a C ␣ root mean square deviation of 3.2 Å (Fig. 4, left). This structural similarity is quite striking considering that the C ␣ root mean square deviation between the two needle homologs BsaL and MxiH is also 3.2 Å (28). This suggests that IpaD has a built-in surface that complements the size and shape of that portion of MxiH that is responsible for maintaining IpaD at the TTSA needle tip. Our NMR chemical shift-mapping results provide the first biological data to support such a model in that the MxiH residues that are perturbed upon adding IpaD appear to form a contiguous face (Fig. 4, right) that may actually be contacting a yet unidentified surface on IpaD and this interaction is blocked in the bacterial cytoplasm by the presence of the IpaD N terminus. MxiH residues within the region capable of interacting with IpaD would include the side chains of Leu-47 and Tyr-50, which upon mutation in this study abolished or severely perturbed IpaD localization. Furthermore, in this model, the MxiH PSNP loop would directly abut helices H3 and H7 of IpaD and such an interaction would be sterically impossible based upon the crystal structure of full-length IpaD unless the N-terminal domain were moved away upon tip complex assembly. Indeed, the chemical  shifts of Asn-40, Ser-42, and Asn-43 demonstrate an interaction between the PSNP loop and IpaD does exist, suggesting that it is possible that the N-terminal domain moves away upon tip complex assembly. Unfortunately, at this time we are unable to predict how the N terminus (helix H1-H2Ј) "swings away" during tip complex assembly to permit docking atop the MxiH needle. Because it has been postulated that the needle channel cannot accommodate a fully folded IpaD (15,31), it is possible that it travels through the needle channel in a partially unfolded manner (31,32) and that refolding at the needle tip forces it to assume a conformation that is compatible with the MxiH-IpaD interactions demonstrated here. High resolution structures are also available for the MxiH/ IpaD counterparts in B. pseudomallei, BsaL/BipD, and a similar BsaL-BipD binding model can be generated in which the twohelix bundle of BsaL can be superimposed to the N-terminal helices of BipD, also with a C ␣ root mean square deviation of 3.2 Å (data not shown). We have also completed the NMR structure of PrgI (28) and shown that it is structurally homologous to MxiH and BsaL. The homology of SipD with BipD and IpaD suggests similarities in the structure of SipD. This suggests that the model of MxiH-IpaD or BsaL-BipD interaction might be a common feature of needle tip interaction for Shigella, Salmonella, and Burkholderia.