Outer membrane adhesion factor multivalent adhesion molecule 7 initiates host cell binding during infection by Gram-negative pathogens

Anne Marie Krachler, Hyeilin Ham, and Kim Orth1

Department of Molecular Biology, University of Texas Southwestern Medical Center, Dallas, TX 75390 Edited by Harry L. T. Mobley, University of Michigan Medical School, Ann Arbor, Michigan, and accepted by the Editorial Board June 1, 2011 (received for review February 10, 2011)

The initial binding of bacteria to host cells is crucial to the delivery of virulence factors and thus is a key determinant of the pathogens success. We report a multivalent adhesion molecule (MAM) that enables a wide range of Gram-negative pathogens to establish high-afnity binding to host cells during the early stages of infection. MAM7 binds to the host by engaging in both proteinprotein (with bronectin) and proteinlipid (with phosphatidic acid) interactions with the host cell membrane. We nd that MAM7 expression on the outer membrane of a Gram-negative pathogen is necessary for virulence in a nematode infection model and for efcient killing of cultured mammalian host cells. Expression of MAM7 on nonpathogenic strains produced a tool that can be used to impede infection by Gram-negative bacterial pathogens. Targeting or exploiting MAM7 might prove to be important in combating Gram-negative bacterial infections.
adhesin

whether MAMs, which constitute a distinct class of predicted outer membrane proteins from Gram-negative bacteria, are involved in cellular attachment. We initially used V. parahaemolyticus as the representative Gram-negative bacterium for analysis of MAM7s, and followed those studies with analyses on the role of MAMs in attachment for other Gram-negative pathogens (6). Results
MAM7 Is an Outer Membrane Protein Mediating Host Cell Attachment. We analyzed the intracellular localization of V. para-

| microbiology | bacterial attachment

acterial pathogens have a large repertoire of virulence factors that target and manipulate the host cellular machinery to enable infection. Delivery of effector proteins to the host cytosol by type III, type IV, and type VI secretion systems, as well as delivery of extracellular toxins, is a common strategy used by bacterial pathogens to abrogate the host immune response and alter cellular pathways to the pathogens advantage (1, 2). Because the secretion of effector and toxin proteins is contact-dependent, the bacteria need to establish tight binding to the host to successfully start an infection. We hypothesize that a common strategy exists across species enabling the pathogen to establish strong initial host binding that is complemented by other species-specic adhesion factors for efcient activation and secretion of virulence factors and toxins. During infection, a variety of adhesion factors are expressed by pathogens to facilitate hostpathogen interactions (35). Many of these adhesins are induced during infection and thus likely would not be involved in the initial adhesion of the bacterial pathogen with the host cell. Using bioinformatics, we searched the genome of Vibrio parahaemolyticus, a Gram-negative bacterium that occurs in marine and estuarine environments and can cause shellsh-borne food poisoning, for a constitutively expressed protein that might be involved in the initial binding of bacteria to a host cell (6). We discovered a predicted outer membrane molecule, multivalent adhesion molecule (MAM), which includes a putative transmembrane motif followed by six (MAM6) or seven (MAM7) mammalian cell entry (mce) domains (Fig. 1A and Fig. S1A Unexpectedly, we found that MAM6 or MAM7 is encoded in a wide range of Gram-negative animal pathogens, but not in Grampositive or plant pathogenic bacteria (Fig. 1A and Figs. S1B and S2). In contrast, proteins containing a single mce domain are widespread (Fig. 1A). In Mycobacterium spp. and some Grampositive bacteria, including Rhodococcus spp. and Streptomyces spp., the mce domain occurs in conjunction with a second domain of unknown function (DUF3407) (7, 8). Proteins containing one mce domain and a C-terminal low-complexity region are thought to represent an accessory component of ABC transporters occurring in algae, higher plants, and bacteria (9). Here we tested
1161411619 | PNAS | July 12, 2011 | vol. 108 | no. 28

haemolyticus MAM7 (VP1611) by replacing the endogenous gene with a plasmid-borne, C-terminally myc-tagged version of MAM7 under the control of its endogenous promoter using a noncytotoxic strain of V. parahaemolyticus, POR2 (10) (Fig. S3). MAM7-myc was constitutively transcribed when the strain was grown in marine LB (Fig. S3) and, based on subcellular fractionation, localized exclusively to the outer membrane (Fig. 1B). Outer membrane localization was also observed after arabinoseinduced heterologous expression of V. parahaemolyticus MAM7myc in Escherichia coli strain BL21 (Fig. 1C), which does not contain a MAM7 ortholog. The rst 44 N-terminal amino acids of MAM7 contain a stretch of hydrophobic residues (aa 2140) predicted to form a transmembrane helix. Deletion of the 44 Nterminal amino acids (MAM7N144-myc) led to cytoplasmic retention of the protein (Fig. 1D). To assess whether MAM7 was present on the surface of the bacteria, BL21-MAM7 and BL21MAM7N144-myc were tested for protease sensitivity. Treatment of these strains with increasing concentrations of papain led to a gradual loss of the epitope tag on cells expressing MAM7myc but not on those expressing MAM7N144-myc, further supporting localization of the C-terminal epitope to the extracellular space (Fig. 1E). To explore whether the N-terminal sequence is embedded in the outer membrane, we introduced a tobacco etch virus (TEV) protease-cleavable peptide between the hydrophobic N-terminal sequence (residues 144) and the mce domains (N144-TEV-MAM7-myc). The protein expressed from this construct was correctly localized and successfully cleaved by TEV protease incubated with intact bacteria (Fig. 1F), demonstrating that the N-terminal peptide contains the information necessary for outer membrane targeting and membrane anchoring of MAM7, whereas the rest of the protein is exposed extracellularly. Having established that MAM7 is an outer membrane protein, we next tested whether MAM7 is important for early attachment of V. parahaemolyticus to host cells. Using V. parahaemolyticus POR2

Author contributions: A.M.K., H.H., and K.O. designed research; A.M.K. and H.H. performed research; A.M.K. and H.H. contributed new reagents/analytic tools; A.M.K., H.H., and K.O. analyzed data; and A.M.K., H.H., and K.O. wrote the paper. The authors declare no conict of interest. This article is a PNAS Direct Submission. H.L.T.M. is a guest editor invited by the Editorial Board.
1

To whom correspondence should be addressed. E-mail: Kim.Orth@utsouthwestern.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1102360108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1102360108

Fig. 1. MAM7 is an outer membrane protein involved in host cell adhesion. (A) Classication of mce-containing proteins using PFAM (26). (BD) Western blot analyses of subcellular fractions of V. parahaemolyticus POR2 (B), E. coli BL21 expressing MAM7-myc (C ), and E. coli BL21 expressing MAM7N144myc (D). (), total lysate before induction; (+), total lysate after induction; sup, culture supernatant; OM, outer membrane; peri, periplasm; IM, inner membrane; cyto, cytoplasm. (E ) Detection prole of MAM7 and MAM7N144 after papain degradation. (F ) Detection prole of N144-TEV-MAM7-myc after TEV treatment. P, cell pellet; S, supernatant. (G) Attachment of V. parahaemolyticus POR2, POR2MAM7, and complemented POR2MAM7 to HeLa, RAW264.7, Caco-2, and 3T3 cells. (H) Attachment of BL21-MAM7 or MAM7N144.

(10) and the POR2MAM7 derivative, we observed that in the absence of MAM7, attachment of V. parahaemolyticus was decreased from 80% to 3540% for all tested host cell lines, including HeLa and Caco-2 epithelial cells, RAW264.7 macrophages, and 3T3 broblasts. The attachment of POR2MAM7 was recovered by a plasmid expressing MAM7 with or without a TEV cleavage site or myc tag (Fig. 1G and Fig. S4). The nonadherent BL21 strain was converted to an adherent strain by inducing expression of V. parahaemolyticus MAM7, but not of the mutant MAM7N144 (Fig. 1H). We conclude that MAM7 contributes to the attachment of V. parahaemolyticus to a broad range of mammalian cells and is sufcient to mediate efcient cellular attachment of a Gram-negative strain in the absence of other adhesion proteins.
MAM7-Mediated Attachment Augments Type IIIMediated Cell Death.

Gentamycin protection assays were used to assess how long bacteria must remain attached to induce 100% lysis. V. parahaemolyticus strains were used to infect 3T3 cells, and gentamycin was added at various time points during the infection. After 4 h of infection, the cells were tested for cell lysis by a lactate dehydrogenase (LDH) release assay. When gentamycin was added at the start of the infection (0 min), minimal lysis was observed after 4 h, whereas without the addition of gentamycin, almost 100% lysis was observed after the same period. When infected cells were treated at 10, 30, 60, or 90 min after infection with gentamycin, minimal lysis was observed. These results support the hypothesis that V. parahaemolyticus must remain associated with 3T3 broblasts for at least 90 min to efciently mediate T3SS-induced cell lysis (Fig. 2A). To test whether MAM7-mediated attachment contributes to host cell binding through this initial phase of infection, we performed infections using the POR1MAM7+N144TEV-MAM7 strain and treated cells with TEV protease at various time points during the infection. Cell lysis was measured at 4 h after infection. As predicted, cells infected with either the POR1 strain or the POR1MAM7+N144-TEV-MAM7 strain, but not with the POR1MAM7 strain, displayed 100% cell lysis. However, POR1MAM7+N144-TEV-MAM7infected cells treated with TEV protease immediately or at 10 or 30 min after the start of infection demonstrated a 40% decrease in cytotoxicity (Fig. 2B). This toxicity level is comparable to that seen with the POR1MAM7 strain, supporting the hypothesis that MAM7 plays an important role in attachment during the early stages of infection. To further investigate the role of MAM7 in adhesion and cytotoxicity, we performed time course infection studies on a range of mammalian cell lines using POR1, POR1MAM7, and POR1MAM7 complemented strains. The POR1MAM7 strain lysed 3T3 broblasts and RAW264.7 macrophages less efciently than either POR1 or POR1MAM7+ MAM7 (Fig. 2 C and D). In addition, the onset of lysis induced by the POR1MAM7 strain was delayed by 3040 min (Fig. 2 C and D). In contrast, no signicant difference between POR1 and POR1MAM7 was seen in either Caco-2 or HeLa epithelial cells (Fig. 2 E and F). Taken together, these results support the hypothesis that, with some cell lines, MAM7-mediated attachment plays an important role in mediating the initial attachment of bacteria to host cells during the early stages of infection. It also supports the hypothesis that other molecules, some of which might be host cell type dependent, play

The translocation of effector proteins to manipulate host signaling pathways is a key step in the pathogenesis of many Gram-negative organisms. V. parahaemolyticus features two type III secretion systems (T3SS) that translocate at least eight different effector proteins into the host cytosol with the aim of altering the cellular response to infection to the pathogens advantage (11, 12). We hypothesized that intimate association that can be mediated through a variety of adhesion mechanisms between pathogen and host is a prerequisite for successful T3SS effector translocation during infection. POR1, a V. parahaemolyticus strain containing both T3SSs but lacking the thermostable direct hemolysins tdhA and tdhS, causes T3SSdependent cell lysis within 23 h after infection (10, 11). To test the contribution of MAM7 in POR1-mediated cell lysis, we created a POR1 strain deleted for MAM7 (POR1MAM7) (Fig. S3) and a POR1MAM7 strain complemented with MAM7 containing a TEV-cleavable sequence inserted between residues 44 and 45 and a C-terminal myc tag (N144-TEV-MAM7-myc). The N144-TEVMAM7-myc protein was conrmed to localize to the outer membrane (Fig. S4B). In addition, treatment of the POR2MAM7+ N144-TEV-MAM7-myc strain with TEV-protease for 5 min resulted in a strain displaying a decrease in attachment comparable to that observed for POR2MAM7 (Fig. S4A). We next used the POR1 strains to assess the contribution of MAM7 attachment to host cell cytotoxicity during V. parahaemolyticus infections.
Krachler et al.

Fig. 2. Impact of MAM7-mediated adhesion on V. parahaemolyticus infection. (A and B) Cytotoxicity of 3T3 broblasts after infection with POR1 (black bar), POR1MAM7 (white bar), or POR1MAM7+pMAM7 (gray bar) and treatment with gentamycin (A) or TEV protease (B). (CF ) Lysis of 3T3 broblasts (C ), RAW264.7 macrophages (D), Caco-2 epithelial cells (E ), or HeLa epithelial cells (F ) with POR1 (), POR1MAM7 (), or POR1MAM7 +pMAM7 (6).

PNAS | July 12, 2011 | vol. 108 | no. 28 | 11615

MICROBIOLOGY

a role in attachment during later stages of infection with V. parahaemolyticus.

MAM7 Is Required for V. parahaemolyticusInduced Pathogenicity in Caenorhabditis elegans. The nematode C. elegans has been used as

a model host for a variety of bacterial pathogens, including Vibrio cholerae and Vibrio vulnicus (13, 14). In the absence of a relevant animal model for V. parahaemolyticus, we tested whether MAM7mediated adhesion plays a role during infection of nematodes. Synchronized germline-decient L4 stage worms were fed with RIMD 2210633, POR1, POR1MAM7, or POR1MAM7+MAM7 strains. Worms fed either nonpathogenic E. coli HB101 or POR1MAM7 exhibited a normal life expectancy prole (Fig. 3A). RIMD 2210633 and POR1-infected worms displayed severe phenotypic changes by day 2 of feeding, including growth retardation and increased frequency of intestinal tract distention leading to abdominal rupture, and died at a much faster rate than worms fed either HB101 or POR1MAM7 (Fig. 3 AC). Both RIMD 2210633 and POR1 killed the worms within 13 d, supporting the hypothesis that most of the lethality is mediated by T3SS effectors rather than by the thermostable hemolysins (10). Full virulence was reconstituted in the POR1MAM7 by reconstituting the strain with plasmid-encoded MAM7 (Fig. 3A). Analysis of worms for abdominal rupture on day 7 revealed a signicant increase in injury of the RIMD 2210633- and POR1fed worms (Fig. 3D). Overall, the results support the hypothesis that MAM7 adhesion plays an important role in pathogenicity in the nemotode infection model.
Multiple mce Domains Are Required to Mediate Stable Attachment of MAM to Host Cells. The number of tandem mce domains in path-

seven domains). We thus hypothesized that six or seven mce domains are the minimum number of domains required for stable host cell attachment and thus explored the relationship between domain number and host cell afnity. We produced recombinant proteins containing one, two, six, or seven mce domains in tandem with a maltose-binding protein (MBP) tag and a single cysteine residue between mce domains and the MBP tag to allow labeling of the proteins with a single uorophore. The amount of protein bound to host cells was measured using uorescence spectroscopy, and the afnities of individual constructs were determined with saturation-binding experiments. The afnity increased nonlinearly with the number of mce domains, with equilibrium dissociation constants ranging from 15 3 M for one mce domain to 0.2 0.1 M for seven mce domains in tandem (Fig. 4A and Table S1). The afnities of MAM proteins for host cells also were determined indirectly, using unlabeled MAM proteins to block the host cell surface before measuring residual binding of E. coli BL21 expressing MAM7. Afnities determined by this method were consistent with those determined by uorescence assays (Fig. 4B and Table S1). Having established the binding afnities for proteins containing varying numbers of mce domains, we analyzed whether strains expressing one mce domain were able to compete for host ad-

ogenic Gram-negative bacteria is strikingly constant (always six or

Fig. 3. MAM7 adhesin is an important factor for V. parahaemolyticus pathogenicity in the nematode C. elegans. (A) Lethality assays with C. elegans strain SS104 glp-4(bn2) fed on RIMD 2210633, POR1, POR1MAM7, or a POR1MAM7 complemented strain. (B and C ) Morphology of worms fed on HB101 (B) or POR1 (C ) for 48 h and pictured using Nomarski optics. (D) Quantitation of intestinal rupture on day 7.

Fig. 4. Relationship between mce domain number, afnity, and competitiveness. (A) Saturation binding with HeLa cells and puried, Alexa Fluor labeled MBP-MAM1, -MAM2, -MAM6 and -MAM7. KD values are listed in Table S1. (B) Indirect determination of binding afnities using MAM proteins to block the cell surface before attachment of BL21-MAM7. KD values are listed in Table S1. (C ) HeLa cell attachment of BL21-MAM1, -MAM6, and -MAM7N144 compared with BL21-MAM7 determined as CI. (D) Ability of BL21-MAM1, -MAM6, -MAM7, or -MAM7N144 to block POR1 attachment to host cells and POR1-mediated cytotoxicity.

11616 | www.pnas.org/cgi/doi/10.1073/pnas.1102360108

Krachler et al.

hesion with strains expressing six or seven mce domains in tandem. All MAM constructs used were correctly localized to the outer membrane, as demonstrated by subcellular fractionation and attachment experiments (Fig. S5). E. coli BL21 strains expressing MAM1, MAM6, or MAM7N144 were mixed with a strain expressing MAM7, and the ratio of attachment to host cells was determined as the competitive index (CI) between the strains. Whereas MAM6 and MAM7 were similarly effective in conferring adhesive properties on E. coli BL21 (CI of 0.8), MAM1 and MAM7N1-44 were far less competitive (CI of 0.27 and 0.19, respectively) (Fig. 4C). We next investigated the ability of E. coli BL21 expressing MAM1, MAM6, or MAM7 to inhibit attachment of and cytolysis by V. parahaemolyticus POR1. When POR1 infections were performed in the presence of E. coli BL21 expressing MAM7 or MAM6 at an identical multiplicity of infection, the attachment of POR1, and thus pathogen-mediated cytotoxicity, were decreased signicantly. The addition of BL21 expressing MAM1 or MAM7N144, however, had little or no effect on the outcome of POR1 infection (Fig. 4D). These data demonstrate how the requirement for high-afnity attachment to host cells necessitates the presence of a large number of mce domains. Although we observed low-afnity binding with one or two mce domains, the expression of six or seven mce domains in tandem results in a steep increase in afnity, allowing the bacterial strain expressing these constructs to successfully compete for host binding. Because of problems with misfolding and insolubility, we could not study constructs containing between three and ve mce repeats, and we hypothesize that such proteins might not occur in nature for similar reasons.
MAM7 Establishes Both ProteinProtein and ProteinLipid Interactions with Host Cells. The secondary structure of MAM7 is predicted to be

rich in -strands connected by exible loop regions, a composition similar to that of bronectin-binding proteins from Gram-positive bacteria (15). We thus tested whether MAM7 also could bind bronectin. Immobilized GST-MAM7, but not GST alone, was able to pull down puried bronectin from human plasma (Fig. 5 A and B). Titrations of uorophore-labeled MBP-MAM7 against immobilized Fn showed that the interaction between hFn and MAM7 was of moderate afnity (KD of 15 4 M), whereas no measurable interaction between bronectin and MAM1 was detected (Fig. 5C). To further explore the possibility that MAM7 binds bronectin on cells, we treated the cells with trypsin to degrade extracellular proteins and then assessed whether MAM7 can bind to cells. Cells treated with trypsin reduced the number of MAM7 molecules on cells by >100-fold (Fig. 5D). Furthermore, the specicity of binding by MAM7 to bronectin on cells was demonstrated by either blocking MAM7 binding to cells with an anti-bronectin antibody or by competing with soluble bronectin (Fig. 5E). Arabidopsis Tgd2 is the substrate-binding component of a chloroplast lipid transporter that is involved in phosphatidic acid (PA) trafcking and is required for the biogenesis of thylakoid membrane lipids (9). Tgd2 contains a single mce domain that has been shown to display weak binding to PA (16). Using lipid overlay assays, we found that both MAM1 and MAM7 bound to PA, but detected no binding with the MBP tag alone (Fig. 5 FH and Fig. S6). We compared the binding afnities of MAM1 and MAM7 to PA using liposome association assays (Fig. 5I). Whereas MAM7 demonstrated stoichiometric binding to PA when present in the liposomes at concentrations as low as 1 mol%, MAM1 bound only to liposomes containing at least 3 mol% PA. To assess whether MAM7 binding to PA occurs in vivo, we analyzed the binding of MAM7 to 3T3 cells incubated without and with previous phospholipase C (PLC) treatment. In the presence of PLC, PA is converted to diacylglycerol, and we noted compromised binding of BL21 expressing MAM7 to cells (Fig. 5J). To further assess whether MAM7 binds to PA on cells, we preincubated BL21-MAM7 with phospholipids and assessed the attachment of the bacteria to 3T3 cells. BL21-MAM7 could bind to cells after incubation with liposomes containing 1,2-Dioleoyl-sn-glycero-3-phosphocholine (PC), but not after incubation with liposomes containing 20 mol%
Krachler et al.

Fig. 5. MAM7 attaches to host cells via bronectin and phospholipid interactions. (A and B) Pull-down of hFn with GST-tagged MAM7 (A) or GSTtag (B). (C ) Saturation binding of uorescent- labeled MBP-MAM1 and MBPMAM7 on immobilized Fn. MBP-MAM7 binds to Fn with an afnity of 15 4 M, whereas no binding was detected with MAM1. (D) Attachment of puried, labeled MBP-MAM7 or MBP to untreated (black) and trypsin-treated (white) 3T3 cells. (E ) Attachment of BL21-MAM7 to 3T3 cells after preincubation with anti-bronectin, anti-mouse IgG antibodies, or hFn. (F and G) Lipid overlay assays with MBP (F ) and MBP-MAM7 (G). (H) Key for lipid strip. (I) Liposome association assays with MBP-tag, MBP-MAM1, and MBPMAM7. Proteins were incubated with liposomes containing only PC (lane 1) or containing mixtures of PC and PA, with PA content in mol% as indicated (lanes 29). Supernatant (S) and pellet (P) fractions were analyzed by SDS/ PAGE. (J) Effect of PLC treatment on BL21-MAM7 attachment to 3T3 cells. (K ) Attachment to 3T3 cells after preincubation of bacteria with liposomes containing either PC or a mixture of 20 mol% PC and 80 mol% PA.

PC and 80 mol% PA (Fig. 5K). These ndings further support the model that MAM7 binds PA on the surface of cells.
Nonpathogenic E. coli Heterologously Expressing MAM7 Ameliorates Effects of Infection by Gram-Negative Pathogens. Because we suc-

cessfully used nonpathogenic E. coli BL21 expressing MAM7 to inhibit infection by V. parahaemolyticus POR1 (Figs. 4D and 6B; Fig. S7 GI), we hypothesized that preincubation of host cells with BL21 expressing MAM7 would ameliorate infections caused by a broad range of pathogenic Gram-negative strains that are predicted to encode MAM7. We investigated the protective effect of E. coli BL21 + MAM7 on infection with the V. parahaemolyticus RIMD 2210633 strain that is equivalent to POR1 but features two thermostable direct hemolysins that are thought to contribute to cell lysis (17) (Fig. 6A and Fig. S7 DF), V. cholerae El Tor N16961 (Fig. 6C and Fig. S7 JL), Y. pseudotuberculosis YP126 (Fig. 6D and Fig. S7 MO), and the enteropathogenic E. coli (EPEC) strain O127:H6 E2348/69 (Fig. 6E and
PNAS | July 12, 2011 | vol. 108 | no. 28 | 11617

MICROBIOLOGY

Fig. 6. Importance of MAM7 in V. parahaemolyticus, Y. pseudotuberculosis, V. cholerae and EPEC infection. (AE ) HeLa cells were infected with V. parahaemolyticus RIMD (A), V. parahaemolyticus POR1 (B), V. cholerae (C ), Y. pseudotuberculosis (D), or EPEC (E ). P, addition of pathogen; C, competition of pathogen with BL21-MAM7; Cm, competition of pathogen with BL21-MAM7N144. (F ) Attachment of BL21-MAM7 (black), V. parahaemolyticus MAM7TM (gray), and MAM7 homologs from Y. pseudotuberculosis (Yp, red), V. cholerae (Vc, blue), or EPEC (green) to 3T3 broblasts. (G) Attachment of WT, MAM7 (), and complemented MAM7 (C) strains of Yp, Vc, and EPEC to 3T3 broblasts. (H) Cytotoxicity (LDH release) of Yp WT (red), MAM7 (pink), and complemented MAM7 (purple) strains toward 3T3 broblasts over time. (I) Cytotoxicity of Vc WT (blue), MAM7 (cyan), and complemented MAM7 (dark-blue triangles) strains toward 3T3 cells. (J) Pedestal formation of EPEC WT (green), MAM7 (green checkered), and complemented MAM7 (green striped) strains on 3T3 cells over time.

Fig. S7 PR). For all infections we performed, we observed a drastic decrease in pathogenicity as manifested either by decreased cytotoxicity (V. parahaemolyticus, V. cholerae, and Y. pseudotuberculosis) or decreased actin pedestal formation (EPEC). Preincubation with E. coli BL21-MAM7 alone did not induce any phenotypic changes in host cells (Fig. S7 AC), and thus residual cytotoxicity was presumably due to the secretion of soluble toxins into the extracellular medium. Based on these results, we propose that MAM7 expressed on the nonpathogenic E. coli strain is masking sites needed by the pathogenic bacteria to initiate binding with the host cells. We hypothesized that MAM7 from the various species are functionally redundant in that all mediate binding to host cells. To test this hypothesis, we cloned and expressed the MAM7 from Y. pseudotuberculosis, V. cholerae, or EPEC into nonadhesive E. coli BL21. Each of these MAM7 homologs enabled BL21 cells to attach to 3T3 cells at a level similar to that observed for BL21 expressing V. parahaemolyticus MAM7 (Fig. 6F). To assess whether MAM7 from Y. pseudotuberculosis, V. cholerae, or EPEC plays a role in host cell adhesion during infection, we created MAM7 deletion strains for each of these pathogens (YpMAM7, VcMAM7, and EPECMAM7) and then reconstituted the MAM7 deletion strains with a WT copy of MAM7 (YpMAM7+ MAM7, VcMAM7+MAM7, and EPECMAM7+MAM7). After incubating these various strains with 3T3 cells, we found compromised attachment for the MAM7 deletion strains, but not for the WT or reconstituted deletion strains (Fig. 6G). We next tested whether the absence of MAM7 might attenuate cell culture cytotoxicity induced by Y. pseudotuberculosis and V. cholerae or reduce pedestal formation by EPEC. 3T3 cells infected with either YpMAM7 or VcMAM7 displayed reduced cytotoxicity over time compared with WT or reconstituted Y. pseudotuberculosis and
11618 | www.pnas.org/cgi/doi/10.1073/pnas.1102360108

V. cholerae strains (Fig. 6 H and I). When 3T3 cells were infected with EPECMAM7, fewer pedestals were observed over time compared with WT or reconstituted EPEC. Interestingly, the phenotype induced by the pathogen appeared to be attenuated only during the early time points in the infection (Fig. 6J). In agreement with previous ndings from other groups, the attachment at the later times of infection was mediated by adherence molecules induced during infection, including Yersinia invasin, E. coli Tir, and type IV pili (4, 5, 18). Discussion Although many bacterial adhesins are known, most of these are species-specic, and many are induced during infection (19, 20). Here we have described the adherence factor, MAM7, used by a wide range of Gram-negative bacterial pathogens to mediate an initial, high-afnity interaction with host cells. Whereas the Nterminal hydrophobic sequence of MAM7 is necessary for outer membrane localization and anchoring of the protein, a dedicated system is not required for protein transport or membrane insertion, because MAM7 can be expressed heterologously and correctly localized by E. coli BL21. MAM7 is used for bacterial adhesion to a variety of mammalian cells and is a crucial factor contributing to host cell infection, as was observed in a tissue culture model for Gram-negative pathogens, including V. parahaemolyticus, Y. pseudotuberculosis, V. cholerae, and EPEC. Through multivalent interactions with the host cell surface, MAM7 contributes to the adherence of bacterial pathogens during the early stages of infection, thereby facilitating injection of T3SS effectors into the host cytoplasm. However, specicity of a pathogen for a certain cell type must be mediated by other, strain-specic adhesion molecules, which most likely are strengthened by the initial MAM7 interaction with host cells. The other adhesion facKrachler et al.

tors are likely to dominate attachment during later phases of infection, when MAM7 binding becomes dispensable (Fig. 2). The binding of the outer membrane adhesion factor MAM7 to host cells is mediated by a multivalent proteinprotein interaction between the adhesin and the extracellular matrix component bronectin. In contrast, the proteinlipid interaction between MAM7 and membrane-bound PA can be mediated by a single mce domain, albeit at a lower apparent afnity. Thus, even though mce domains have been integrated into a diverse range of proteins, the number of mce domains plays an important role in determining ligand-binding afnities and thus the functionality of mce containing proteins. Mce proteins involved in lipid transport require a low-afnity transient interaction with their ligand and contain only one mce domain (e.g., Tgd2). In contrast, constitutively expressed proteins with multiple mce domains appear to mediate an early, highafnity interaction with the host cell surface. We show that this not only increases lipid-binding afnity, but also gives rise to a new functionality (bronectin binding), which further strengthens the interaction between MAM7 and host cells. Although interactions with extracellular matrix components, such as bronectin, are a common strategy used by both Gram-positive (FnBPs of Staphylococcus spp.) and Gram-negative (e.g., Yersinia invasin YadA, Salmonella MisL, or Campylobacter CadF) pathogens to achieve host cell adhesion, the direct binding of an adhesin to membrane phosphatidic acid is, to our knowledge, a previously unobserved mechanism for hostpathogen interaction (2125). Although at 23 mol%, phosphatidic acid is only a minor component of eukaryotic membranes, we have shown that these concentrations are sufcient to mediate stable attachment of MAM7. The strategies for protein and lipid interactions are combined in MAM7mediated adhesion to achieve efcient binding of Gram-negative pathogens to their host. We have demonstrated that nonpathogenic BL21 expressing MAM7 can be used to prevent binding of a range of Gram-negative pathogens to host cells, thus offering protection against pathogenmediated cytotoxicity. Based on bioinformatic analysis, it appears that a large number of Gram-negative pathogens contain a MAM7 (or MAM6) and are predicted to use this protein to mediate the initial, high-afnity attachment to host cells. Our studies indicate that MAM7 molecules from several Gram-negative pathogens, including V. parahaemolyticus, Y. pseudotuberculosis, V. cholerae,
1. Alouf JE (2000) Bacterial protein toxins: An overview. Methods Mol Biol 145:126. 2. Galn JE (2009) Common themes in the design and function of bacterial effectors. Cell Host Microbe 5:571579. 3. Lebreton F, et al. (2009) ace, which encodes an adhesin in Enterococcus faecalis, is regulated by Ers and is involved in virulence. Infect Immun 77:28322839. 4. Heroven AK, Dersch P (2006) RovM, a novel LysR-type regulator of the virulence activator gene rovA, controls cell invasion, virulence and motility of Yersinia pseudotuberculosis. Mol Microbiol 62:14691483. 5. Boekema BK, Van Putten JP, Stockhofe-Zurwieden N, Smith HE (2004) Host cell contact-induced transcription of the type IV mbria gene cluster of Actinobacillus pleuropneumoniae. Infect Immun 72:691700. 6. Daniels NA, et al. (2000) Vibrio parahaemolyticus infections in the United States, 19731998. J Infect Dis 181:16611666. 7. Arruda S, Bomm G, Knights R, Huima-Byron T, Riley LW (1993) Cloning of an M. tuberculosis DNA fragment associated with entry and survival inside cells. Science 261: 14541457. 8. Chitale S, et al. (2001) Recombinant Mycobacterium tuberculosis protein associated with mammalian cell entry. Cell Microbiol 3:247254. 9. Awai K, Xu C, Tamot B, Benning C (2006) A phosphatidic acidbinding protein of the chloroplast inner envelope membrane involved in lipid trafcking. Proc Natl Acad Sci USA 103:1081710822. 10. Park KS, et al. (2004) Functional characterization of two type III secretion systems of Vibrio parahaemolyticus. Infect Immun 72:66596665. 11. Burdette DL, Yarbrough ML, Orvedahl A, Gilpin CJ, Orth K (2008) Vibrio parahaemolyticus orchestrates a multifaceted host cell infection by induction of autophagy, cell rounding, and then cell lysis. Proc Natl Acad Sci USA 105:1249712502. 12. Broberg CA, Zhang L, Gonzalez H, Laskowski-Arce MA, Orth K (2010) A Vibrio effector protein is an inositol phosphatase and disrupts host cell membrane integrity. Science 329:16601662. 13. Vaitkevicius K, et al. (2006) A Vibrio cholerae protease needed for killing of Caenorhabditis elegans has a role in protection from natural predator grazing. Proc Natl Acad Sci USA 103:92809285.

and EPEC, mediate early attachment of the pathogen to host cells. Based on these studies with the aforementioned pathogens, we propose that MAM7 could play an important role during the initial phase of infection for many Gram-negative pathogens that express this adhesion molecule. The identication and initial characterization of this broadly expressed adhesion factor is important for understanding molecular interactions between Gram-negative pathogens and their target host cells. The adhesion factor MAM7 appears to be constitutively expressed, allowing the bacteria to be primed for immediate attachment when encountering a host cell. The MAM7-mediated attachment appears to be important during the initial phases of infection to allow for the production or presentation of other factors that might be involved in later stages of infection (Figs. 2 and 6). These initial studies demonstrate how bacteria expressing MAM7 have an advantage over other bacteria that have no MAM7 on their cell surface (Fig. 4). In addition, the multivalent interaction of the mce repeats in MAM7 allows for binding of substrates not recognized by a single repeat. Finally, future microbial and biochemical studies will address the possibility of countering Gram-negative pathogens with their own outer membrane adhesion factor, MAM7, to attenuate infection. Materials and Methods
A detailed discussion of the materials and methods used in this study is provided in SI Materials and Methods. Protease Protection Experiments. Cells expressing myc-tagged MAM constructs were resuspended in PBS to give an OD600 of 1. Papain was added to nal concentrations as indicated, and cells were incubated at 22 C for 10 min. Reactions were stopped by adding Complete Protease Inhibitor Mixture (Roche) and 5 SDS loading buffer and then boiled for 10 min. Samples were separated by SDS/PAGE, and myc-tagged proteins were detected by Western blot analysis. ACKNOWLEDGMENTS. We thank A. Torres, the K.O. laboratory, E. L. Reddick, and L. Avery for their advice and discussion; N. Alto and V. Sperandio for providing EPEC and V. cholerae strains; and T. Silhavy and I. Collinson for providing antibodies against OmpA and SecE. K.O., H.H., and A.M.K. are supported by grants from the National Institute of Allergy and Infectious Diseases (R01-AI056404 and R01-AI087808) and the Welch Foundation (I1561). K.O. is a Burroughs Wellcome Investigator in Pathogenesis of Infectious Disease and a W.W. Caruth Jr. Biomedical Scholar.

Krachler et al.

PNAS | July 12, 2011 | vol. 108 | no. 28 | 11619

MICROBIOLOGY

14. Dhakal BK, et al. (2006) Caenorhabditis elegans as a simple model host for Vibrio vulnicus infection. Biochem Biophys Res Commun 346:751757. 15. Schwarz-Linek U, et al. (2003) Pathogenic bacteria attach to human bronectin through a tandem beta-zipper. Nature 423:177181. 16. Lu B, Benning C (2009) A 25-amino acid sequence of the Arabidopsis TGD2 protein is sufcient for specic binding of phosphatidic acid. J Biol Chem 284:1742017427. 17. Nishibuchi M, Fasano A, Russell RG, Kaper JB (1992) Enterotoxigenicity of Vibrio parahaemolyticus with and without genes encoding thermostable direct hemolysin. Infect Immun 60:35393545. 18. Roe AJ, Hoey DE, Gally DL (2003) Regulation, secretion and activity of type IIIsecreted proteins of enterohaemorrhagic Escherichia coli O157. Biochem Soc Trans 31:98103. 19. Boland A, Cornelis GR (2000) Interaction of Yersinia with host cells. Subcell Biochem 33:343382. 20. Kline KA, Flker S, Dahlberg S, Normark S, Henriques-Normark B (2009) Bacterial adhesins in host-microbe interactions. Cell Host Microbe 5:580592. 21. Frman G, Switalski LM, Speziale P, Hk M (1987) Isolation and characterization of a bronectin receptor from Staphylococcus aureus. J Biol Chem 262:65646571. 22. Tertti R, Skurnik M, Vartio T, Kuusela P (1992) Adhesion protein YadA of Yersinia species mediates binding of bacteria to bronectin. Infect Immun 60:30213024. 23. Dorsey CW, Laarakker MC, Humphries AD, Weening EH, Bumler AJ (2005) Salmonella enterica serotype Typhimurium MisL is an intestinal colonization factor that binds bronectin. Mol Microbiol 57:196211. 24. Konkel ME, Garvis SG, Tipton SL, Anderson DE, Jr, Cieplak W, Jr. (1997) Identication and molecular cloning of a gene encoding a bronectin-binding protein (CadF) from Campylobacter jejuni. Mol Microbiol 24:953963. 25. Henderson B, Nair S, Pallas J, Williams MA (2011) Fibronectin: A multidomain host adhesin targeted by bacterial bronectin-binding proteins. FEMS Microbiol Rev 35:147200. 26. Finn RD, et al. (2010) The Pfam protein families database. Nucleic Acids Res 38(Database issue):D211D222.