Systematic Analysis of Bacterial Effector-Postsynaptic Density 95/Disc Large/Zonula Occludens-1 (PDZ) Domain Interactions Demonstrates Shigella OspE Protein Promotes Protein Kinase C Activation via PDLIM Proteins*

Background: PDZ domains provide specificity in protein-protein interactions in eukaryotic systems. Results: Shigella OspE, which is delivered into human cells during infection, binds PDLIM7 via a PDZ interaction and contributes to activation of PKC. Conclusion: During infection, OspE contributes to activation of PKC through PDZ-mediated binding to PDLIM7. Significance: PDZ interactions likely contribute to pathogenesis of several bacterial pathogens. Diseases caused by many Gram-negative bacterial pathogens depend on the activities of bacterial effector proteins that are delivered into eukaryotic cells via specialized secretion systems. Effector protein function largely depends on specific subcellular targeting and specific interactions with cellular ligands. PDZ domains are common domains that serve to provide specificity in protein-protein interactions in eukaryotic systems. We show that putative PDZ-binding motifs are significantly enriched among effector proteins delivered into mammalian cells by certain bacterial pathogens. We use PDZ domain microarrays to identify candidate interaction partners of the Shigella flexneri effector proteins OspE1 and OspE2, which contain putative PDZ-binding motifs. We demonstrate in vitro and in cells that OspE proteins interact with PDLIM7, a member of the PDLIM family of proteins, which contain a PDZ domain and one or more LIM domains, protein interaction domains that participate in a wide variety of functions, including activation of isoforms of protein kinase C (PKC). We demonstrate that activation of PKC during S. flexneri infection is attenuated in the absence of PDLIM7 or OspE proteins and that the OspE PDZ-binding motif is required for wild-type levels of PKC activation. These results are consistent with a model in which binding of OspE to PDLIM7 during infection regulates the activity of PKC isoforms that bind to the PDLIM7 LIM domain.

Diseases caused by Gram-negative bacterial pathogens are mediated in large part by proteins delivered by bacteria into host cells via specialized bacterial secretion systems. Interaction of these effector proteins with host proteins elicits a host response or alters signaling pathways to enhance the survival and spread of the infecting bacteria. Quantification of effector protein delivery suggests that 3000 -10,000 molecules of each effector are delivered per bacterium that contacts the host cell (1,2). Under biologically relevant circumstances, individual cells are likely infected with no more than one bacterium. In contrast, the copy number of eukaryotic proteins has been estimated to range from fewer than 50 to more than 1,000,000 per cell (3,4). Consequently, for most pathogenic interactions, the number of bacterial effector molecules per host target protein is likely to be low. For these effector proteins to efficiently interact with their molecular targets, high affinity specific interactions likely occur.
PDZ domains are commonly conserved domains that mediate specific protein-protein interactions in mammalian cells. Approximately 250 individual PDZ 3 domains have been identified in more than 150 proteins in the human proteome (5). PDZ domains interact with short peptides, designated PDZbinding motifs, typically located in the extreme C terminus of the interacting protein (6). PDZ domain interactions with target PDZ-binding motifs display substantial specificity (7). PDZbinding motifs are categorized into three major classes based on their peptide sequences: -Xaa-(Ser/Thr)-Xaa-(class I); -Xaa--Xaa-(class II); and -Xaa-(Asp/Glu)-Xaa-(class III), where Xaa is any amino acid, and is a hydrophobic residue.
PDZ domains within proteins mediate a wide variety of specific interactions that have diverse cellular functions, particularly relating to cell polarity, cell-cell junctions, recognition of immune cells, control of proliferation, and control of cellular * This work was supported, in whole or in part, by National Institutes of Health Grants F32 AI092967 (to L. A. B.) and R01 AI081724 (to M. B. G.). G. M. and J. E. A. are employees and stockholders of Merrimack Pharmaceuticals, Inc. 1  migration. In bacteria, only a handful of proteins, primarily type III secreted effector proteins of pathogenic Escherichia coli (8,9), have been shown to participate in interactions with mammalian PDZ domains. We systematically assessed the extent of PDZ domain-mediated interactions in infection by Gram-negative bacterial pathogens generally and Shigella flexneri specifically. We show that putative PDZ-binding motifs are significantly enriched among effector proteins delivered into mammalian cells by certain bacterial pathogens. We identify an interaction of S. flexneri effectors OspE1 and OspE2 with the mammalian PDLIM family of PDZ domain-containing proteins and show that the interaction involves up to 11 C-terminal residues of OspE, suggesting that binding energy is extended over a relatively large surface area in the PDZ domain-peptide interface. We found that during infection, OspE-PDZ interactions enhance bacterial spread, and OspE promotes PDLIM7-dependent activation of PKC.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The wild-type S. flexneri strain used in this study is serotype 2a strain 2457T (10). An isogenic ospE1 ospE2 mutant was generated by sequentially deleting the entire coding sequences of ospE1 and ospE2 via phage Red recombinase-mediated homologous recombination and FLP recombinase-mediated removal of the inserted kanamycin cassette (11). Bacteria were grown in tryptic soy broth from individual colonies that were red on agar containing Congo red.
Genetic Methods-A plasmid that carries ospE1 under the control of its endogenous promoter was generated by cloning a PCR fragment containing the coding region of ospE1 plus 148 nucleotides upstream into pBR322. For bacterial production of GST-OspE1, a PCR fragment containing the ospE1 coding sequence was cloned into pET-GST (Novagen/EMD Biosciences). For mammalian expression of EGFP-OspE1, a PCR fragment containing the ospE1 coding sequence was cloned into pEGFP-C1 (Clontech) or pBABE-puro vector (Addgene). Generation of plasmid-encoded ospE1⌬3, ospE1⌬6, and ospE1⌬11 was by the QuikChange site-directed mutagenesis method (Stratagene), according to the manufacturer's protocol. Sequence analysis was performed to verify that each construct was correct. Sequences of the primers used in this study are available from the authors upon request.
cDNAs for human PDLIM1, PDLIM2, PDLIM3, PDLIM4, and PDLIM7 were obtained from OpenBiosystems (pOTB7-PDLIM1, pOTB7-PDLIM2, pSPORT6-PDLIM3, pDNR-LIB-PDLIM4, and pOTB7-PDLIM7, respectively). Derivatives for mammalian expression tagged at the C terminus with Myc were generated by cloning into pcDNA3 (Invitrogen) and pBABE-puro fragments containing the coding sequence, amplified by PCR with primers that include the sequence encoding Myc. For expression of genes in yeast for analysis by the protein interaction platform assay, appropriate sequences were PCR-amplified and cloned into pDNR221, pAG416, and pBYO11 plasmids, as described previously (12).
Peptide Synthesis-Peptide synthesis, expression, and purification of PDZ domains, fabrication and processing of microarrays, and fluorescence polarization were performed as described (7,13). Peptides were synthesized on the solid phase using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry as described previously (7,13). All peptides were labeled on their N terminus with 5(6)-carboxytetramethylrhodamine (TAMRA), purified by reversed-phase high performance liquid chromatography, and verified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.
Expression and Purification of PDZ Domains-Recombinant domains were purified from E. coli as described previously (7,13). Proteins were produced with N-terminal thioredoxin and His 6 tags and were purified in a single step by immobilized metal affinity chromatography. All proteins used in this study were found to be predominantly monomer as judged by analytical gel filtration.
Fabrication and Processing of Protein Microarrays-Protein microarrays were printed on aldehyde displaying glass substrate and assembled into 96-well plates as described (7,13). Proteins attach to the surface through the ␣-amine at their N termini or through the ⑀-amine of accessible lysine residues and hence are shown in many different orientations within each spot. Proteins were arrayed at a concentration of ϳ40 M in a buffer that contains 20% glycerol (v/v) to prevent evaporation. A small amount (typically 100 nM) of cyanine-5-labeled bovine serum albumin (BSA) was combined with each protein to facilitate image analysis. After a 12-h incubation at room temperature, unreacted aldehydes on the glass were quenched, and the surface was blocked by addition of a buffer containing BSA.
Fluorescently labeled peptides were incubated with the array for 1 h. at room temperature. The arrays were then washed, dried, and scanned for both cyanine-5 and 5(6)-TAMRA fluorescence. The cyanine-5 image was used to define the location of each spot, and the mean fluorescence of replicate spots in the 5(6)-TAMRA image was determined for each PDZ domain. Spots were compared with the mean intensity of control spots (thioredoxin), and spots with intensities at least 2-fold over background were further analyzed by fluorescence polarization.
Fluorescence Polarization-Fluorescent peptides were incubated with PDZ domains for 1 h at 25°C in assay buffer (20 mM NaH 2 PO 4 /Na 2 HPO 4 , 100 mM KCl, pH 7.4, supplemented with 0.02% bovine serum albumin (w/v), 0.04% NaN 3 , and 1 mM DTT). Peptides were kept at a fixed concentration (20 nM), and the concentration of the PDZ domains was varied from 20 M to 10 nM (2-fold serial dilutions). Fluorescence polarization was measured in 384-well microtiter plates using an Analyst AD fluorescence plate reader (Molecular Devices), with excitation at 525 nm and emission at 590 nm. Equilibrium dissociation constants (K D ) were calculated from these data as described previously (7,13).
Purification of GST Fusion Proteins-GST-OspE1 and GST alone were synthesized in E. coli BL21 (DE3). Expression was induced at 16°C using 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside. Bacteria were lysed in cold lysis buffer (50 mM Tris, pH 8, 150 mM NaCl, 40 g/ml lysozyme, 1 tablet of protease inhibitor mixture (Roche Applied Science)) by sonication. Following removal of cell debris and unbroken cells by centrifugation, the supernatant was incubated in the presence of glutathione-Sepharose beads overnight at 4°C to bind the GST-tagged protein. Beads with bound protein were collected by brief centrifugation and washed once with low salt wash buffer (20 mM Tris, pH 8, 100 mM NaCl, protease inhibitor mixture, 1 mM DTT), once with high salt wash buffer (20 mM Tris, pH 8, 500 mM NaCl, protease inhibitor mixture, 1 mM DTT), and again with low salt wash buffer, before resuspending in a volume of low salt wash buffer to make a 50% slurry.
GST Pulldown Procedures-PDZ domain containing proteins that were specifically precipitated by GST-OspE1 were identified by mass spectrometry. HeLa cells were lysed with ice-cold lysis buffer (50 mM Tris pH 8, 150 mM NaCl, 1% Triton X-100, 0.05% sodium deoxycholate, 10 mM EDTA, protease inhibitor mixture), and cell debris was removed by centrifugation. Purified GST or GST-OspE1 on beads was incubated with HeLa cell lysates at 4°C for at least 1 h with constant rotation. Beads were recovered by centrifugation and washed once each with wash A (20 mM Tris, pH 8, 500 mM NaCl, 0.5% Triton X-100, 1 mM DTT, protease inhibitor mixture), wash B (20 mM Tris, pH 8, 100 mM NaCl, 0.5% Triton X-100, 1 mM DTT, protease inhibitor mixture), and wash C (20 mM Tris, pH 8, 500 mM NaCl, 1 mM DTT, protease inhibitor mixture). Proteins bound to beads were analyzed by SDS-PAGE stained with Coomassie. Bands specific to precipitation by GST-OspE1 were identified and cut out of the gel. Proteins within these bands were identified by LC/MS/MS of tryptic peptides at the Proteomics Core of the Harvard Partner's Center for Genetics and Genomics, using standard procedures.
HEK293T cells transfected with plasmids encoding Myctagged PDLIM proteins were lysed as described above, and cell debris was removed by centrifugation. Purified GST or GST-OspE1 on beads was incubated with lysate at 4°C for at least 1 h with constant rotation. Beads were recovered by centrifugation and washed as above. Proteins bound to beads were analyzed by SDS-PAGE and Western blot analysis.
Type III Secretion Assay-S. flexneri strains containing plasmids that encode FLAG-tagged wild-type OspE1 or C-terminal truncations of OspE1 were grown to post-exponential phase. Bacterial pellets were collected by centrifugation and resuspended in phosphate-buffered saline containing 10 M Congo red to induce secretion of type III effector proteins, with incubation at 37°C for 30 min with constant rotation. Following separation from the bacteria by centrifugation, proteins that had been secreted into the culture supernatant were precipitated by addition of trichloroacetic acid and were analyzed by Western blot analysis.
Protein Interaction Platform Assay and Analysis-Protein interaction assays in yeast assays were performed as described previously (12). Briefly, yeast cells were co-transformed with plasmids that encode NS or NS-PDLIM7 and GFP or GFP-OspE1. Following induction of gene expression with galactose for 4 -6 h, yeast cells were imaged by fluorescence microscopy, and the formation of foci was quantified.
Lentivirus-mediated Stable Expression of Genes-For stable expression of genes in mammalian cells, lentiviral expression systems were used. Lentiviral stocks were prepared by filtering the supernatants of HEK293T cells that had been transfected with pAdvantage (Promega), pCMV-VSVG, pGag-Pol, or pBABE-puro constructs carrying the genes of interest. Target cells, at 70% confluency in 6-well plates, were infected with filtered lentivirus and supplemented with protamine sulfate (8  (41), and Tobe et al. (42) with the understanding that in some cases the analysis is not comprehensive. The different classes of PDZ-binding motifs are defined by the sequence of C-terminal residues of the protein. The sequences X(S/T)X, XX, and X(D/E)X identify classes I, II, and III, respectively. indicates a hydrophobic amino acid, including valine, leucine, isoleucine, phenylalanine, methionine, cysteine, tryptophan, and tyrosine; X indicates any amino acid.  CT358  MQAEARSLEEHC  III  CT157  IWAFLLKNSSPV  I  CT105  WAQKAGVQSSSI  I  SycE  LPDLHALGMYHL  II  PepF  IERKLEELASLL  I  CT047  PMLSLEMLVSRL  I  CT300  AHCCGSSHQIEI  II  PmpA  SNSLSCGGYVGF  II  PmpC  AHMMNCGARMTF  II  MdhC  DEILQEKASVSL  II  AroC  VDLLLQHRCTQL  I  CT223  DDDDGFNYGSRV  I  CT309  ELINLMEKGIRW  II  FolP  ALAAAAWAGMFV  II  CT469  TFSCKSPFPELF  III  CT387  DYPPAPDGLVIV  II  CT618  SALGNVILSFAF  II  CT138  SAQKKVVNHIAF  II  MutL  MIRRLVLDDDFM  III  CT686  FPKLQKQDLYHV  II  Ppa  RLAHEDYCNLFM  II  BrnQ  VFALTILYKLSV  II  DnaE  TNIPARVLATTV  I   L. pneumophila Philadelphia-1  RalF  EMQKEKGRQLKF  II  Ceg10  LKPDLPKNLFKF  II  Ceg7  QLQEAELRRILY  II  CegC1  QEEEEVRLNLGV  II  CegC3  EEENEESSRFTM  II  CegC4 PEREEPKMGMQL II g/ml final) by spinfection at 2000 rpm for 30 min at 20°C. Infected cells were incubated at 37°C overnight and were allowed to recover in complete medium the next day. Beginning 2 days after transduction, cells were maintained in selection medium (DMEM containing 10% FBS and 5 g/ml puromycin). Cell Culture and Transfection-HeLa cells were maintained in minimal essential medium and supplemented with 10% fetal bovine serum and 1ϫ nonessential amino acids under humidified air containing 5% CO 2 at 37°C. HEK293T and MDCK cells were maintained in Dulbecco's modified Eagle's essential medium (DMEM) and supplemented with 10% fetal bovine serum (FBS) under the same conditions.

C. trachomatis D/UW-3/CX
For transfection of HeLa cells, DNA and FuGENE 6 reagent (Roche Applied Science or Promega) were added according to the manufacturer's protocol to cells that had been seeded in a 6-well dish the day before on acetone-rinsed glass coverslips when appropriate. Transfection was allowed to continue for 36 -48 h. For transfection of HEK293T cells, DNA, combined with 0.25 M CaCl 2 and HBS (140 mM NaCl, 25 mM Hepes, pH 7.1, 0.75 mM Na 2 HPO 4 ), was added to 3 ϫ 10 6 cells that had been seeded in a 10-cm dish the day before. Fresh medium was added to cells the next day, and transfection was allowed to continue for 48 h. For transfection of MDCK cells, GenJet TM in vitro DNA transfection reagent for MDCK cells (Version II, SignaGen Laboratories) was used according to the manufacturer's protocol. For reverse transfection of siRNA (Dharmacon) into HeLa cells, HiPerfect reagent was used according to the manufacturer's protocol (Qiagen).
Stable cell lines expressing either an empty vector or PDLIM7 were reverse-transfected with siRNA targeted to PDLIM7 (Thermo Scientific D-013081-04). Cells were rinsed after 24 h with fresh complete media (DMEM with 10% FBS, penicillin/streptomycin, and 5 g/ml puromycin) and used for infections after 48 h.
Analysis of Phosphorylation of MARCKS-Cells were infected with Shigella strains expressing the AFA-I adhesion from uropathogenic E. coli (14) on a P15A backbone at a multiplicity of infection (bacteria to cells) of 25 or with wild-type S. flexneri not expressing the AFA-I adhesin at a multiplicity of infection of 100. At the indicated infection times, plates were placed on ice; media were aspirated, and cells were rinsed once with ice-cold TBS (20 mM Tris, pH 8.0, 156 mM NaCl) containing PhosStop phosphatase inhibitor tablets (Roche Applied Science 04906837001). Sample buffer was added to each well, and cells were collected by scraping. MARCKS phosphorylation was analyzed by Western blotting with PMARCKS antibody (Cell Signaling rabbit polyclonal 27415) (1:1000) and quantified by densitometry using ImageJ. For infections of untransfected HeLa cells, cell lysates were subjected to centrifugation at 100,000 ϫ g at 4°C for 1 h, and the soluble fraction was assessed for phosphoMARCKS by Western blot analysis.
Shigella Plaque Assay-Plaque assays and measurement of area of individual bacterial plaques within the monolayers were performed as described previously (15). For cells transfected with siRNA, plaque assays were initiated 48 h post-transfection.
Inhibition of PKC-For chemical inhibition of PCK during plaque assays, Gö6976 or Gö6983 (Tocris Bioscience 2253 and 2285) was added at the time the agarose overlay to 10 M, which is above the IC 50 for the PKC isoforms they inhibit. At this concentration, the inhibitors did not appear to cause significant cytotoxicity. For analysis of the effect of inhibition of PKC on MARCKS phosphorylation, Gö6983 was added to 10 M immediately prior to infection of HEK293T cells.
Immunofluorescent Staining-Protein localization in cells was performed essentially as described previously (15). To detect localization of Myc-tagged PDLIM proteins and EGFP-OspE1, HeLa cells seeded on acetone-rinsed glass coverslips were transfected with plasmids encoding genes of interest and were fixed and permeabilized 48 h post-transfection. To detect localization of MAGI-1 in MDCK cells, transfected MDCK cells were seeded in Nunc Lab-Tek II chambered cover glass (Thermo Scientific) and 3 days post-transfection were fixed and permeabilized. Monoclonal antibodies against Myc (Clontech) were used at 1:100 dilution; rabbit antibodies against MAGI-1 (Sigma) were used at 1:100 dilution, and Alexa Fluor 568-conjugated secondary antibodies (Invitrogen) were used at 1:100 dilution. Polymerized actin was labeled with Alexa Fluor 350 phalloidin (Invitrogen).
Microscopy-Epifluorescence and phase-contrast microscopy were performed using a Nikon Eclipse TE300 microscope equipped with Chroma Technology filters and a Photometrics CoolSNAP HQ charge-coupled device camera (Roper Scientific). Images were acquired using iVision software. Confocal microscopy was performed on a Nikon Ti-E inverted microscope, and a Coherent, 4-watt, continuous-wave laser served as a fluorescence excitation light source. Images were acquired using Metamorph (Molecular Devices). Color figures were assembled by separately capturing images with each of the appropriate filter sets and digitally pseudocoloring the images using Adobe Photoshop software.
Statistical Analyses-All experiments were performed in triplicate, and data were analyzed in a blinded fashion. The statistical significance of differences between experimental results was determined using the Student's t or Fisher exact test.

Secreted Bacterial Effectors of Selected Gram-negative Pathogens Contain Conserved PDZ-binding Motif Sequences-
The sequences of individual type III and type IV secretion system effector proteins of S. flexneri, Salmonella enterica serovar typhimurium, Yersinia enterocolitica, enterohemorrhagic E. coli, Chlamydia trachomatis, and Legionella pneumophila were analyzed for the presence of a consensus sequence for PDZbinding motifs at the protein C terminus (Table 1). Except for Salmonella typhimurium, a third or more of the secreted effectors from each of these pathogens contain putative PDZ-binding motifs. Compared with a set of proteins randomly selected from each proteome, the secreted effectors from S. flexneri, Y. enterocolitica, enterohemorrhagic E. coli, and C. trachomatis were significantly enriched 2-4-fold for the presence of PDZbinding motifs (Table 2), raising the possibility that PDZ domains contribute to the function of these bacterial proteins in the host. For S. typhimurium, PDZ-binding motifs were not enriched among secreted effectors, and for L. pneumophila, the observed 1.5-fold enrichment did not reach statistical significance.
Shigella Effectors Interact with PDZ Domains-To test the possibility that PDZ domain interactions play a role in effector function, the nine effectors from S. flexneri that contain putative PDZ-binding motifs were tested for interactions with mammalian PDZ domains. Because the PDZ-binding motif consensus sequences are located within the C terminus of the polypeptides, and some PDZ domains exhibit binding selectivity out to the Ϫ8 position of their target protein (16), we synthesized C-terminal peptides corresponding to the last 10 residues of the parent bacterial effector protein. The tripeptide sequence NNG was added to the N terminus of each peptide to increase its water solubility, and each peptide was capped with a fluorophore (5(6)-TAMRA) at its N terminus to enable visualization. Fluorescently labeled effector peptides were incubated with protein microarrays that contained 158 purified PDZ domains from the mouse proteome ( Table 3). The ability of the peptides to bind was assessed by the intensity of the signal from the fluorescent tag. Using the cutoff of a signal greater than 2-fold over background, we identified 54 potential peptide-PDZ domain interactions ( Table 4). The C-terminal peptide of the enteropathogenic E. coli effector Map, which binds NHERF1 (Ebp50) (8,9), served as a positive control; this interaction gave a signal 2.48-fold over background. Of note, the Map peptide interacted with the first of the two PDZ domains present in NHERF1 but not with the second (Table 4 and data not shown).
We retested a subset of hits from this initial screen, and we quantified the solution phase binding affinity of effector peptides with PDZ domains using quantitative fluorescence polarization (7), which enabled us to assess the strength of each interaction. The results identified the PDZ domain of PDLIM3 and the fourth of six PDZ domains in MAGI-1 as potential host binding partners of OspE, with K d 18.9 and 12.2 M, respectively, for these interactions. S. flexneri encodes two OspE proteins that differ only at position 54; their C-terminal sequences are identical. MAGI-1 is a component of tight junctions, and tight junctions contribute to Shigella spread (17), yet in polarized MDCK cells, OspE and MAGI-1 co-localized minimally (Fig. 1), suggesting that this interaction is unlikely to be relevant during infection; we did not analyze this possibility further.
OspE Proteins Interact with Proteins of the PDLIM Family-In a parallel approach designed to find PDZ domain proteins that might interact with OspE, we identified proteins that were specifically precipitated by OspE from HeLa cell lysates. Mass spectrometry was performed on tryptic peptides of protein bands identified on SDS-PAGE analysis as present in precipitates of GST-OspE1 or GST-OspE2 and absent from precipitates of GST alone ( Fig. 2A and data not shown). Among the top proteins identified by this approach, only PDLIM7 (Enigma) contains a PDZ domain. The multiple peptides that were identified map uniquely to PDLIM7 (Fig. 2B) and not to any other PDLIM family member. The interaction of OspE1 and OspE2 with PDLIM7 was confirmed in an independent GST precipitation experiment by Western blotting with antibodies to PDLIM7 (Fig. 2C). A similar precipitation-based approach used previously identified integrin-linked kinase as an interacting partner of OspE (18), indicating that particular conditions and analyses can identify distinct yet potentially relevant interactions.

Frequency of PDZbinding motifs (%)
Proteins of the PDLIM family contain one PDZ domain and one to three LIM domains, each of which participates in protein-protein interactions. Because PDLIM7 is expressed in the intestine (21) and was the only family member precipitated from HeLa cells, we chose to focus on the interaction of OspE with PDLIM7, realizing that PDLIM7 may not be the only PDLIM family protein that binds OspE. GST-OspE1 specifically precipitated Myc-tagged PDLIM7 from lysates of 293T cells (Fig. 2D). OspE1 also precipitated PDLIM2, again providing evidence that conserved sequences are critical to these interactions. OspE expressed exogenously in mammalian cells was insoluble in nonionic detergent, likely due to its interactions with the cytoskeleton. Thus, to test whether OspE and PDLIM7 interact when co-expressed in vivo, a protein interaction assay was performed in yeast (12). PDLIM7 was fused to the reovirus scaffolding protein NS, which forms large focal inclusions in living cells. Upon co-expression, GFP-OspE1 was recruited to these inclusions, visualized as fluorescent foci (Fig.  5). The targeting of GFP-OspE1 depended on PDLIM7, as no foci were observed upon expression of GFP-OspE1 with NS alone.
PDZ-binding Motif Is Required for Interaction of OspE1 with PDLIM7-To determine whether interaction of OspE with PDLIM7 is mediated by its PDZ domain, precipitation experiments were performed using GST-OspE1 lacking its three C-terminal residues (OspE1⌬3). Absence of the last three residues led to substantially reduced precipitation of PDLIM7 (Fig.  2D), indicating that these residues contribute to the interaction. As some PDLIM7 was still pulled down, we considered whether residues upstream of these three C-terminal residues contribute binding energy to the interaction, as observed for other PDZ interactions (22). Alignment of homologs of OspE1 showed that although OspE proteins from all Shigella species contain a PDZ-binding motif, homologs from other pathogens lack PDZbinding motifs entirely (Fig. 6A, solid line box) and lack conserved charged residues found among the C-terminal 11 residues of Shigella OspE proteins (Fig. 6A, underline). Deletion of the last 6 or the last 11 residues in OspE1 markedly diminished the ability of OspE1 to precipitate PDLIM7 (OspE1⌬6 and OspE1⌬11, Fig. 6B). Similar requirements were observed for the interaction of OspE1 with PDLIM2 (Figs. 2D and 6B), suggesting that this is a conserved feature of the PDLIM2 PDZ domain. Although there was some degradation of the OspE⌬6 and OspE1⌬11 constructs, the decrease in PDLIM7 and   OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 PDLIM2 precipitated was substantially more than the relative decrease in the amount of the full-length constructs, strongly suggesting that the interaction between OspE1 and PDLIM7 or PDLIM2 in vitro involves more than the three C-terminal residues.

Bacterial Effector-PDZ Interactions and Shigella OspE
Truncations of OspE1 Localize to Focal Adhesions-Targeting of OspE to focal adhesions is dependent on binding to integrin-linked kinase; mutation of the Trp residue at position 68 of OspE abolishes this interaction and alters OspE localization (18). We tested whether the C-terminal PDZ-binding motif and adjacent residues are also required for localization to focal adhesions. The distribution in HeLa cells of EGFP-tagged derivatives of each truncation mutant was identical to that of full-length EGFP-OspE1 (Fig. 7A), indicating that deletion of up to 11 amino acids at the C terminus, including the PDZbinding motif and residues required for interaction with PDLIM proteins, does not disrupt the localization of OspE1 to focal adhesions.
C-terminal 11 Residues of OspE1 Are Important for Intercellular Spread of S. flexneri-Whereas S. flexneri encodes two intact copies of ospE, ospE1 and ospE2, Shigella sonnei encodes only a single intact ospE, which is required for efficient intercellular spread through mammalian cell monolayers (18,20). We found that in S. flexneri, OspE1 and OspE2 are redundant with respect to intercellular spread, as strains carrying a deletion of either one or the other ospE gene alone displayed no significant defect, whereas a strain carrying a deletion of both genes displayed a 35-45% reduction in spread (Fig. 7B). The spread defect of the ospE1 ospE2 strain was complemented by expression of ospE1 from a vector in trans (Fig. 7C).
Complementation of the ospE1 ospE2 mutant with OspE1⌬11 was unable to rescue the spread defect of the ospE1 ospE2 mutant, whereas complementation with OspE1⌬3 or OspE1⌬6 rescued the spread defect to levels similar to that of full-length OspE1 (Fig. 7C). The inability of the OspE1⌬11 construct to rescue the spread defect of the double mutant was not due to a defect in protein synthesis, stability, or secretion from S. flexneri, as the levels and efficiency of secretion for each of the truncation mutants was similar to that of the full-length protein (Fig. 7D). These results suggest that the C-terminal 11 residues of OspE1 are required for its role in S. flexneri spread, and because these C-terminal residues are dispensable for the tar-  geting of OspE to focal adhesions, the results suggest that OspE function in S. flexneri spread is not completely dependent on its localization to focal adhesions per se.
It was notable that OspE⌬6 and OspE⌬11 were both defective in precipitation of PDLIM7 from HeLa cell lysates (Fig. 6B), yet only OspE⌬11 was unable to rescue the spreading defect of the ospE1 ospE2 mutant (Fig. 7C). Of note, although defective in precipitating PDLIM7, both OspE⌬6 and OspE⌬11 precipitated small amounts of it (Fig. 6B). These findings suggest that factors that associate with OspE during S. flexneri infection of cells may contribute to the function of OspE in a manner that depends on residues present in OspE⌬6 and absent in OspE⌬11. These observations also raise the possibility that in the context of cellular infection, the interaction of PDLIM7 might be stabilized with OspE⌬6 but not OspE⌬11.
OspE Participates in PKC Signaling Dependent on Its PDZbinding Motif and PDLIM7-In addition to a PDZ domain, each PDLIM family member contains one or more LIM domains (Fig. 4A), protein interaction domains that participate in a wide range of cellular functions (23). As PDLIM7 interacts with PKC (24), and LIM domains regulate PKC activity (25), we tested whether PKC activity might contribute to Shigella pathogenesis. Treatment of HEK293T cells with Gö6976 or Gö6983, which inhibit PKC isoforms ␣, ␤I, ␥, and , and ␣, ␤, ␥, ␦, and , respectively, led to a significant reduction in bacterial spread compared with DMSO carrier alone (Fig. 8A). PKC ␤II/EGF receptor inhibitor, which inhibits PKC isoforms ␣, ␤I, and ␤II, had no effect on bacterial spread (data not shown), suggesting that a PKC isoform other than ␣ or ␤ might be primarily involved.
To test whether OspE modulates PKC signaling through PDLIM7, we compared activation of PKC in the presence or absence of OspE and/or PDLIM7. PKC activation was assessed by measuring phosphorylation of the PKC substrate MARCKS (25). Infection of HeLa cells with wild-type S. flexneri induced an increase in the amount of phosphorylated MARCKS beginning at 30 -40 min of infection (Fig. 8B). We then specifically depleted PDLIM7 with RNAi in HeLa cells that had been stably transfected with either a control vector or a vector expressing an RNAi-resistant PDLIM7 mRNA; thus, RNAi treatment with a PDLIM7-specific siRNA led to depletion of PDLIM7 in the former but not the latter cell line (Fig. 8C). We chose to use this approach because it permitted us to use the same siRNA construct under all experimental conditions, thereby avoiding any RNAi construct-specific effects. At 30 min of S. flexneri infection, an increase in MARCKS phosphorylation was observed, and this was enhanced by ϳ50% in the presence of OspE (Fig. 8, C-F). MARCKS phosphorylation was substantially attenuated in the presence of OspE⌬11 (Fig. 8, E and F), indicating that the PDZ-binding motif is required for efficient OspE-dependent phospho-MARCKS signaling. Inhibition of PCK with Gö6983 during S. flexneri infection of HEK293T cells completely blocked phosphorylation of MARCKS (Fig. 8, G and H), indicating that induction of phosphorylation of MARCKS was due to activation of PKC. Of note, in HEK293T cells, some MARCKS phosphorylation occurred at baseline, and S. flexneri-induced increases in phospho-MARCKS above baseline were entirely dependent on OspE (Fig. 8, G and H). The model that emerges is one in which OspE interacts with PDLIM7 through a PDZ domain-mediated interaction, and this interaction promotes the spread of bacteria through the activation of PKC.

DISCUSSION
In eukaryotic organisms, PDZ domains are common protein interaction domains. Type III-and type IV-secreted effector proteins, delivered into mammalian cells by bacterial pathogens during infection, specifically alter cellular signaling pathways and the functions of a wide variety of host proteins. Our observation that PDZ-binding motifs are enriched among pro-teins delivered into cells by bacterial type III and type IV secretion systems raised the possibility that these motifs serve functional roles in pathogenesis. The finding that C-terminal peptides derived from several different effector proteins were recognized in vitro by a variety of mammalian PDZ domains supported a potential functional role for these interactions in disease. For OspE, these biochemical interactions were validated using complementary in vitro and in vivo protein-protein interaction assays. Together, these findings support the hypothesis that, in the context of specialized secretion systems, bacterial protein sequences have evolved to target the protein interaction platform presented by PDZ domains as an important mechanism to promote virulence.
Although PDZ domains and PDZ-binding motifs each display conserved features, the interactions of PDZ domains with target PDZ-binding motifs display substantial specificity (7). It is notable that although PDZ-binding motifs are common among bacterial effector proteins, the larger PDZ domains themselves are not. The short length of the PDZ-binding motif, combined with the specificity of PDZ domain-mediated interactions, provides a mechanism by which bacteria can use a short motif to direct specific interactions in cells. In principle, the interface can serve to target the effector protein to a particular subcellular localization and/or to define a particular functional interaction.
In the case of OspE proteins, we found that the interaction with host cell PDZ domains serves a functional role, rather than a role in subcellular localization. Targeting of OspE to focal adhesions is lost when integrin-linked kinase is depleted or by mutation of Trp-68 (18), but it is unaffected by deletion of the 11 C-terminal residues that include the PDZ-binding motif  ( Fig. 7A). In contrast, the PDZ-binding motif of Map is critical for its localization to the cortical cytoskeleton (9).
Our findings with respect to the frequency of putative PDZbinding motifs among secreted bacterial effector proteins and the functional importance of the PDZ-binding motif in OspE proteins contribute to a growing body of data implicating specific PDZ interactions in bacterial pathogenesis. In pathogenic E. coli, in addition to Map binding NHERF1, NleA (EspI) interacts with the PDZ domain of Sec24 (22,27), a component of the COPII complex, as well as to NHERF2. The NleA interactions functionally disrupt COPII-dependent protein export (27) and likely contribute to disruption of tight junctions (28). In addition, many viruses encode proteins that interact via PDZ-binding motifs with mammalian PDZ domains in ways that promote pathogenesis (29). Our results expand on these findings both by documenting a role for PDZ interactions in OspE function and by surveying the prevalence of PDZ-binding motifs among effectors of multiple pathogens.
Our results provide insight into the nature of the PDZ domain-mediated interaction with OspE proteins. Our findings that only upon deletion of six or more C-terminal residues was binding markedly diminished in vitro (Fig. 6B) and only upon deletion of 11 C-terminal residues were effects on function observed in vivo (Fig. 7C) suggest that binding energy is extended over a relatively large surface area in the PDZ domain-peptide interface, consistent with structural studies of PDLIM proteins with peptide ligands (30).
PKC activation by Shigella spp. has been observed both in tissue culture and in rabbit intestine, occurring within minutes of bacterial entry (31,32). Activated PKC redistributes to the plasma membrane (33), where it phosphorylates a variety of cellular substrates. PDLIM7 binds several isoforms of PKC, PKC␣, PKC␤1, and PKC and has been shown to regulate the activity of PKC␤1 via its LIM domains (24,25). Our analysis of PKC inhibitors suggests that a PKC isoform other than ␣ or ␤ might be primarily responsible ( Fig. 8A and data not shown), but additional experiments are needed to test this and to identify which isoform(s) are involved. Our data showing that OspE contributes to activation of PKC in a PDLIM7-dependent manner implicate an OspE-PDLIM7-PKC signaling pathway during S. flexneri infection. Of note, in HeLa cells, the absence of OspE reduced but did not eliminate PKC activation (Fig. 8), indicating that at least in selected cell types, OspE-independent factors, perhaps including vacuole escape (34), also contribute to activation of PKC by Shigella. Isoforms of PKC control intracellular pathways that regulate cell morphology and barrier function (35), and PKC activity is correlated with increased cell spreading and adherence (36), suggesting functional roles for PKC activation in Shigella infection.

FIGURE 7. Phenotype of OspE deletion strains and OspE truncations in cells.
A, localization of EGFP-OspE1 wild type (WT) and C-terminal truncation mutants to focal adhesions in HeLa cells. Scale bar, 50 m. B and C, intercellular spread of S. flexneri ospE mutants (B) and rescue of spread defect of ospE1 ospE2 mutant with plasmid-borne ospE1⌬3 and ospE1⌬6, but not ospE1⌬11 (C), by plaque assay in HeLa cells. Data shown are the mean Ϯ S.D. of results from three independent experiments. *, p Ͻ 0.05; N.S., not significant; a.u., arbitrary units. D, C-terminal truncation mutants of OspE1, carried on plasmids, were secreted from an S. flexneri ospE1 ospE2 mutant with similar efficiencies. OspE constructs were internally tagged with FLAG (diagram at top), with the coding sequence for the FLAG tag inserted between the coding sequence for the N-terminal 50 amino acids of OspE1, which encodes the secretion signal, and the coding sequence for full-length ospE1 or C-terminally truncated ospE1. Culture supernatants and bacterial pellets were harvested following induction of type III secretion by incubation in the presence of 10 M Congo red. Western blot of wild type OspE1 (WT) and OspE1 truncations using antibody to the internal FLAG tag. Data represent three or more independent experiments.
Here, we focused on the interaction of OspE proteins with PDLIM7, but our data suggest that OspE may also interact with other members of the PDLIM family of proteins during infection. PDLIM7 and PDLIM1 are expressed in the intestines (21,37), the site of Shigella infection; for other PDLIM proteins, expression is predominantly in other tissues or of unknown distribution. Although we only detected PDLIM7 in OspE-mediated precipitations from HeLa cell lysates (Fig. 2C), OspE also efficiently precipitated PDLIM2-Myc from lysates of cells that expressed it (Fig. 2D). In addition, the C-terminal peptide of OspE interacted on the protein microarray with the PDZ domain of PDLIM3. Given the high degree of conservation of PDZ domains among the PDLIM family members (Fig. 3), it is not surprising that OspE would interact with any of them. Indeed, the lack of an observed interaction of the OspE peptide with the PDZ domain of other PDLIM proteins present on the protein microarray may arise from technical issues, such as improper folding of these PDLIM PDZ domains on the array. The fact that an observed interaction with the PDZ domain of PDLIM3 led us to discover a functional interaction with PDLIM7 demonstrates the potential utility of screens such as the PDZ domain microarray for identifying candidate biologically relevant interactions, as identification of families of proteins can lead to discovery of family members that are biologically relevant. Whether OspE interaction with PDLIM proteins other than PDLIM7 occurs or is relevant during Shigella infec-tion is presently uncertain, but of interest, as different PDLIM proteins have distinct functions.
Overall, our data are consistent with a model in which OspE proteins bind via their C termini to the PDZ domains of the subset of cellular PDLIM proteins that are localized at the junction of focal adhesions and stress fibers. Our data suggest that binding of OspE to PDLIM7 regulates the activity of PKC isoforms that bind to the PDLIM7 LIM domain. The role of PKC in barrier function and cell adherence suggests that OspE-mediated PDLIM-dependent regulation of PKC activity at focal adhesions may be important in a variety of cellular processes during infection. Our findings also suggest that PDZ interactions may be widely relevant to infection by certain bacterial pathogens.