A region of the Yersinia pseudotuberculosis invasin protein that contributes to high affinity binding to integrin receptors.

The entry of Yersinia pseudotuberculosis into cultured mammalian cells is mediated by the bacterial protein invasin. The mammalian receptors for invasin are five beta1 chain integrins. Site-directed mutagenesis of the aspartate and lysine residues in the 192-amino acid integrin binding domain of invasin was performed to identify regions, in addition to the previously characterized 903-913 region, that are important for integrin binding. One mutation, D811A, resulted in depressed ability of invasin to bind purified alpha5beta1 and to promote bacterial entry. Further mutational analysis of Asp-811 indicated that an oxygen-containing side chain is required at this position. A second nearby residue, Phe-808, was also shown to be important for integrin binding, as an alanine substitution at this site had properties similar to the Asp-811 mutation. This mutational analysis has therefore identified a second region that, in conjunction with residues 903-913, is required for wild type levels of integrin binding. The contribution to binding by two noncontiguous sites in the primary sequence parallels results that indicate two domains of fibronectin are involved in integrin binding.

The interaction of bacterial pathogens with host cells is an important step in the establishment of a productive infection. Bacterial adhesion to the mammalian cell surface allows either extracellular colonization (1)(2)(3) or penetration within the host cell (4 -7). Once internalized, the pathogen may replicate within the protective niche of the cell or translocate into deeper tissues where multiplication occurs (8,9). The latter strategy is used by the enteropathogenic bacterium Yersinia pseudotuberculosis (10 -12). 1 Y. pseudotuberculosis is internalized by normally nonphagocytic cultured mammalian cells via two pathways (13). The best studied and most efficient of these mechanisms is mediated by the bacterial protein invasin (14). Members of the invasin family of proteins are found in a variety of enteropathogens, each of which contributes to the interaction with host cells (15,16). Invasin is a 986-amino acid outer membrane protein encoded by the inv gene which, when expressed by laboratory strains of Escherichia coli, confers the ability to penetrate cultured cells (14,17). The carboxyl-terminal 192-amino acid region of invasin binds mammalian cells and is necessary and sufficient to promote internalization (18,19).
The receptors on the mammalian cell surface to which invasin binds belong to a subset of the integrin superfamily of receptors (20). These receptors are ␣␤ heterodimeric proteins found on the surface of most mammalian cells and are involved in cell-cell interactions as well as adhesion to the extracellular matrix (21). Each of the invasin receptors thus far identified has the ␤ 1 chain (20). 2 The site on the integrin receptor that binds invasin appears to be close, or identical, to the site bound by other integrin substrates. Invasin and fibronectin bind to mutually exclusive sites on this receptor, although there is no sequence homology between the two ligands and invasin lacks the Arg-Gly-Asp (RGD) 3 sequence shown to be important for binding by fibronectin (17,22,23). One outstanding difference between these two ligands is that invasin binds to the ␣ 5 ␤ 1 integrin with a much higher affinity than does fibronectin (22). This tight binding is required for intracellular entry (24).
Several features of the cell binding domain of invasin are important for receptor binding. Mutations that disrupt a 76amino acid disulfide loop between residues 907 and 982 destroy integrin binding (25). In addition, residues within an 11-amino acid region encompassing residues 903-913 are also critical for cell binding (26). One of these residues, aspartate 911 (Asp-911), appears to be absolutely essential for binding, as changes at this position result in complete loss of integrin binding. The extreme sensitivity of this residue to amino acid changes is reminiscent of the aspartate residue of the GRGDS sequence found in the fibronectin repeat III-10 required for cell adhesion (27)(28)(29)(30).
In fibronectin, the repeat III-10 domain containing the RGD sequence is not sufficient to promote levels of adhesivity observed with wild type protein (31)(32)(33)(34). Sequences amino-terminal to this domain, in repeat III-9, are also required (35,36). This upstream region, called the synergy region, appears to contain critical residues approximately 100 amino acids aminoterminal to the RGD sequence. In this study, we report the analysis of a region of invasin that, like the synergy region of fibronectin, enhances the binding of invasin to its receptor.
Plasmid Constructions-pSelect2-8, which contains the complete Y. pseudotuberculosis inv gene, was constructed by inserting a HindIII to BamHI fragment containing the invasin gene (nucleotides 1-3933 (17)) into pSelect (Promega, Madison, WI). Plasmid pLHS22 was derived from pSelect2-8 by site-directed mutagenesis and contains an aspartate to alanine change at amino acid 811 of invasin (D811A). pLHS11 contains the E. coli malE gene (encoding MBP) fused at its 3Ј end to the region of inv encoding the carboxyl-terminal 192 amino acids of invasin, under the control of Ptac. The plasmid was constructed by inserting a KpnI to BamHI fragment from pSelect2-8 containing the 3Ј end of inv into pMAL-c1 (New England BioLabs, Beverly, MA). pLHS6 was similarly constructed using the KpnI to BamHI fragment from pLHS22 and introducing it into pMAL-c1. pRI284 that expresses the carboxyl-terminal 197 amino acids of invasin fused to the carboxyl-terminal amino acid of MBP was constructed by inserting the 3Ј end of the wild type inv gene downstream of the sequence encoding the factor Xa site in plasmid pMAL-c1. pLHS8, which expresses MBP fused to the carboxyl-terminal 165 amino acids of invasin, was constructed using polymerase chain reaction amplification of the region of inv encompassing nucleotides 2934 to 3933 from plasmid pJL277, generating a KpnI restriction site at the 5Ј end and a BamHI site at the 3Ј end. A KpnI to BamHI fragment from this amplified DNA was then cloned into pMAL-c1. pJL277 is a derivative of pJL274 (25) containing a silent mutation at nucleotide 2882 in inv to delete a HpaI restriction site. 4 All site-directed mutations in the full-length invasin gene were made in plasmid pSelect2-8. Each of these mutations was then moved into an MBP-inv fusion by ligating the KpnI to BamHI fragment encompassing nucleotides 2859 -3933 of inv into pMAL-c1 and verified by nucleotide sequencing using standard protocols (U. S. Biochemical Corp.) (26). A schematic representation of the MBP-Inv constructs is shown in Fig. 1B.
Site-directed Mutagenesis-Site-directed mutagenesis of inv was performed on the phagemid pSelect2-8 using the Altered Sites Mutagenesis System (Promega) following the directions of the manufacturer. Briefly, single-stranded pSelect2-8 DNA, which contains a point mutation in its bla gene, was annealed simultaneously to an oligonucleotide containing a predetermined nucleotide change in inv and a 27-base pair oligonucleotide containing the bla sequence that corrects the point mutation. The oligonucleotides were used to prime DNA synthesis, and the resulting double strand product was ligated and transformed into E. coli BMH71-18 mutS selecting for ampicillin resistance. The plasmids were then transformed into DH5␣ and potential inv mutants were analyzed by nucleotide sequencing of DNA mini-preps (Quiagen Inc., Chatsworth, CA). Oligonucleotides for mutagenesis and DNA sequencing were synthesized by the Howard Hughes Medical Institute Microchemical Facility, Harvard Medical School.
Immunoblots of Invasin Proteins-E. coli strains that expressed MBP-Inv protein fusions were grown to early logarithmic phase and induced for 1 h with 1 mM isopropylthio-D-galactoside before lysis. Extracts were fractionated on SDS-polyacrylamide gels, transferred to Immobilon filters (Millipore Corp., Bedford, MA), and immunoprobed to determine the relative amount of invasin expressed (41). Cultures were resuspended in one-tenth volume SDS-polyacrylamide gel electrophoresis sample buffer (42) and boiled 2 min, and then 10-l aliquots were subjected to electrophoresis on a 7.5% SDS-polyacrylamide gel under reducing conditions. Fractionated proteins were transferred to Immobilon filters and probed with anti-invasin monoclonal antibodies 3A2-1 or 1E8 (43) as described previously (19).
Internalization of Bacteria into Cultured Mammalian Cells-Penetration of bacteria into cultured mammalian cells was assayed by a gentamicin protection assay (19). Approximately 5 ϫ 10 5 bacteria were added to subconfluent monolayers of HEp-2 cells in 24-well tissue culture dishes in RPMI 1640 containing 20 mM HEPES, pH 7.0, 0.2% BSA and allowed to penetrate cells for 90 min at 37°C. Extracellular bacteria were then killed by a 90-min treatment in the identical media containing 50 g/ml gentamicin. The surviving intracellular bacteria were released with 0.1% Triton X-100, and viable counts were titered on bacteriological media.
The ability of MBP-Inv fusion proteins to promote entry into HEp-2 cells was determined using a S. aureus coating assay (18,44). S. aureus Cowan I strain was coated with dilutions of anti-MBP antiserum (gift from Dr. Carol Kumamoto) followed by purified MBP-Inv hybrid proteins labeled with carboxytetramethyl succinimidyl rhodamine (Molec-ular Probes, Inc., Eugene, OR), as described (45). The coated bacteria were tested in a gentamicin protection assay as described above. The number of molecules of invasin on the surface of each coated bacteria was determined by quantitating the specific activity of the rhodaminelabeled protein and then determining the amount of fluorescence per bacterium using a microtiter fluorometer (Flow Laboratories Inc., McLean, VA).
Protein Purification-MBP-Inv hybrid proteins were purified by affinity chromatography on cross-linked amylose after isopropylthio-Dgalactoside induction of either MC4100 or DH5␣ harboring plasmids that express MBP-Inv fusions, as described previously (19,44,46). Protein concentration was determined by scanning densitometry (Molecular Dynamics, Sunnyvale, CA) of polyacrylamide gels, using protein standards of known concentration, or by the Bradford microassay (Bio-Rad). MBP-Inv192 contains the carboxyl-terminal 192 amino acids of invasin fused to MBP (Fig. 1B). Point mutations in MBP-Inv192 are indicated by single-letter amino acid designation followed by the invasin residue altered and the resulting amino acid. For example, MBP-InvD811A carries an Asp to Ala mutation at invasin residue 811.
HEp-2 Cell Binding to Purified MBP-Inv Derivatives Coated on Microtiter Wells-The ability of HEp-2 cells to adhere to MBP-Inv derivatives coated on 96-well electroimmunoassay microtitration plates (ICN Biochemicals, Costa Mesa, CA) was determined as described previously (19,25). Wells were coated in triplicate with dilutions of each protein in 50 mM sodium bicarbonate, pH 9.6, for 2 h at 25°C and then blocked overnight in PBS containing 0.2% BSA. 1 ϫ 10 5 HEp-2 cells in RPMI 1640, 0.4% BSA, 10 mM HEPES, pH 7.0, were added per well, allowed to attach to proteins for 90 min at 37°C, in 5% CO 2 , and fixed with methanol for 20 min. The number of cells attached was quantitated by crystal violet staining followed by solubilization in sodium deoxycholate (47).
The effect of divalent cation concentration on mammalian cell adhesion was tested in a modification of the above procedure. K562 cells were stripped of extracellular divalent cations by incubating for 10 min at 37°C in PBS, 10 mM EDTA, pelleting, resuspending in PBS, 2 mM EDTA, and incubating a further 10 min, before washing three times with PBS. The stripped cells were resuspended in HEPES-buffered saline, pH 7.05 (48), containing 0.4% BSA and increasing concentrations of MgCl 2 and then tested for cell adhesion by incubating in wells of microtiter dishes coated with a saturating concentration (2.1 ϫ 10 Ϫ7 M) of MBP-Inv derivatives.
The efficiency of coating of microtiter wells by MBP-Inv hybrids was tested by ELISA. Microtiter wells were coated with MBP-Inv hybrids exactly as described above, blocked overnight in PBS containing 0.2% BSA, and then proteins revealed with anti-MBP antiserum and alkaline phosphatase-conjugated goat anti-rabbit IgG (Zymed Laboratories, South San Francisco, CA) as described previously (43).
Binding of Invasin to Purified Receptor-Competition of mutant invasin derivatives with wild type invasin for binding to ␣ 5 ␤ 1 integrin was performed as described (26). Briefly, 96-well microtiter plates were coated with 2.1 ϫ 10 Ϫ8 M MBP-Inv192 in HBS (125 mM NaCl, 25 mM HEPES, 7.0) for 4 h at 25°C and then blocked overnight with HBS containing 1% BSA. This concentration of MBP-Inv192 was shown to give half-maximal binding to placental ␣ 5 ␤ 1 . The wells were then washed with HBS containing 1 mM MnCl 2 , and dilutions of MBP-Inv hybrid proteins in HBS containing 1 mM MnCl 2 , 20 mM octyl-␤-glucopyranoside were added to triplicate wells. Placental ␣ 5 ␤ 1 integrin (a generous gift from Dr. John Leong), in the same buffer, was immediately added to each well at a final concentration of 7 ϫ 10 Ϫ9 M and incubated for 3 h at 25°C. Wells were washed with HBS containing 1 mM MnCl 2 , 0.1% Nonidet P-40 and the bound receptors fixed in 3% paraformaldehyde, 2 mM MgCl 2 , PBS for 30 min at 25°C. Wells were washed three times within PBS and blocked overnight in PBS containing 1% BSA. Bound receptor was revealed by standard ELISA using anti-fibronectin receptor antiserum (Telios Pharmaceuticals, San Diego, CA) and alkaline phosphatase-conjugated goat anti-rabbit IgG.
Isolation of Monomeric Forms of MBP-Inv192 Hybrid Proteins-To separate monomeric MBP-Inv192 from larger complexes, protein preparations isolated by affinity chromatography were subjected to gel filtration chromatography on a commercially prepared 10 mm ϫ 30 cm Superose 12 column (Pharmacia Biotech, Inc.). The column was equilibrated with 20 mM HEPES, pH 8.0, 150 mM NaCl, and calibrated using the following size standards: blue dextran (void volume), thyroglobulin (M r ϭ 660,000), aldolase (M r ϭ 158,000), bovine serum albumin (M r ϭ 67,00), and ovalbumin (M r ϭ 43,000). 200-l samples of several MBP-Inv192 hybrids were then subjected to chromatography on this column, collected in 200-l fractions, and the fractions corresponding to either the protein peak found in the column void volume or to a 60-kDa species 4 J. Leong, unpublished data.
were each pooled and analyzed by SDS-polyacrylamide gel electrophoresis. The volume of each pool was adjusted, and protein concentrations were determined by the Bradford microassay (Bio-Rad). The pools corresponding to the 60-kDa monomeric MBP-Inv192 species were concentrated using a Centricon 30 filter (Amicon, Beverly, MA) and refractionated on an identical gel filtration column to ensure that no aggregation of the purified sample occurred on storage or concentration. Fractions from the single peak of this column corresponding to monomeric MBP-Inv192 were pooled, protein concentrations were determined, and microtiter wells were coated with dilutions of these preparations to analyze the ability of each preparation to support adhesion of HEp-2 cells.

Isolation of Invasin Mutations Defective for Cell Adhesion-
Mutants previously isolated that resulted in defective binding of invasin to mammalian cells were selected based on the requirement that the carboxyl terminus of the protein was presented properly on the bacterial cell surface but unable to promote efficient bacterial uptake (26). To isolate mutants that would not survive this screen, but which identified regions of invasin that contribute to integrin binding, systematic replacement of the 10 aspartate and 8 lysine residues by alanine was performed by site-directed mutagenesis (Fig. 1A), and each mutant was tested for the ability to promote entry. The aspartates were chosen because the previously isolated mutations with the strongest phenotypes were alterations of the Asp-911 residue, which apparently performs a functionally similar role to the aspartate in RGD found in many integrin ligands (26). We wanted to determine if the previously characterized properties of mutations at this site were unique to this residue or were a property of simply reducing the amount of negative charge associated with the cell binding domain. The lysines were chosen as there are several clusters of lysines within the cell binding domain that may serve important functional roles, as well as to test the model that altering the charge of this protein affects cell adhesion.
Six aspartate to alanine changes, at amino acid positions 829, 831, 852, 892, 929, and 959, had no affect on invasinpromoted entry of E. coli into HEp-2 cells in tissue culture. The aspartate to alanine change at residue 945 decreased the efficiency of bacterial penetration 50 -100-fold. Alanine changes at positions Asp-811, Asp-911, and Asp-966 abolished all detectable entry (Ͼ500-fold decrease in bacterial penetration). The result of the D911A change was expected based on previous work (26), but the defective entry of the other derivatives was surprising, as alterations in these residues were not isolated in the previous mutant selection.
To determine whether these mutant proteins were expressed efficiently, immunoblots were performed on lysates of E. coli expressing the invasin derivatives. As expected, the Asp to Ala changes that had no affect on bacterial uptake also had no detectable affect on steady state protein levels (data not shown). The Ala substitution at position 945, on the other hand, had decreased steady state invasin levels that were comparable with its decreased efficiency of promoting HEp-2 penetration. Mutations at residues 811 and 966 resulted in extremely unstable proteins that were almost undetectable by Western blot analysis using two different anti-invasin monoclonal antibodies. As the previous selection for invasin mutants defective for promoting bacterial uptake required that the protein be stable, these mutants would not have been isolated by the prior protocol (26).
Single alanine substitutions replacing the lysine residues at positions 812, 816, 820, 871, 873, and 874 also eliminated bacterial entry. Western blot analysis of these proteins expressed by E. coli showed that all six of the derivatives generated extremely low steady state levels of invasin (data not shown). A double mutant, however, replacing lysines at residues 954 and 955 with alanines, was stable and promoted entry into mammalian cells as efficiently as wild type invasin.
Cell Binding Assays Reveal a Second Aspartate Residue Critical for HEp-2 Cell Binding-The results from the above analysis indicate that simple change of an aspartate to an uncharged residue is not sufficient to cause defective integrin binding by invasin. Two Asp to Ala changes and five Lys to Ala changes, however, showed no ability to promote entry and therefore identified residues potentially involved in receptor binding. The inability to detect any steady state levels of these derivatives, however, made further analysis of these constructions impossible. In the hope that fusing the carboxyl-terminal 192 amino acids of each of these derivatives to MBP would result in stable proteins, gene fusions to the E. coli malE gene were constructed and analyzed further (Fig. 1B). All the resulting hybrid proteins were stable and expressed at high levels, allowing large amounts of each protein derivative to be isolated. Hybrid proteins were purified by affinity chromatography, and the ability of HEp-2 cells to adhere to microtiter wells coated with each protein was tested. The results showed that with the exception of MBP-InvD811A, all Asp to Ala changes promoted mammalian cell adhesion at levels comparable with the wild type MBP-Inv derivative (Fig. 2, A and B). Ten-to twenty-fold more MBP-InvD811A than the corresponding wild type derivative was required to promote half-maximal adhesion of HEp-2 cells ( Fig. 2A and Table I). In contrast, MBP-InvD966A bound mammalian cells as efficiently as the wild type protein (Fig. 2B). Each of the Lys to Ala changes showed only small effects on adhesion. As compared with wild-type MBP-Inv, less than a 4-fold increase in the protein coating concentration was required to give half-maximal cell binding for each of the Lys to Ala mutants (data not shown). Therefore, with the exception of D811A, most of the mutations analyzed in this fashion had little effect on mammalian cell adhesion.
InvD811A Is Defective in Its Ability to Promote Cell Entry-Bacterial uptake promoted by the MBP fusion protein, MBP-InvD811A, was tested. Wild type hybrid protein MBP-Inv197 and mutant MBP-InvD811A were used to coat the surface of S. aureus, and the coated bacteria were tested for the ability to penetrate HEp-2 cells ("Materials and Methods"). When S. aureus was coated with approximately 10,000 monomers of each MBP-Inv derivative per viable count, the amount necessary to allow the wild type derivative to promote efficient entry, uptake promoted by the D811A derivative was negligible. Even if the bacteria were coated with saturating amounts of protein (approximately 30,000 monomers per viable count), bacterial uptake was significantly depressed if S. aureus was coated with the D811A derivative (Fig. 3). Therefore, the binding-defective D811A mutation was also impaired in its ability to promote cell entry.
Analysis of Amino Acid Changes at Position 811-To determine what features of the aspartate at position 811 were required for full invasin activity, the aspartate residue was replaced by amino acids with various side chains. MBP-Inv hybrid proteins were purified, and the effect of each mutation on mammalian cell adherence to invasin was tested ( Table I). Substitution of the aspartate residue for the oxygen-containing residues serine (MBP-Inv811S), threonine (MBP-Inv811T), or asparagine (MBP-Inv811N) had negligible affects on HEp-2 cell adhesion. The derivative harboring a glutamate substitution (D811E) required 4-fold more protein than wild type invasin to promote equivalent cell adhesion (Table I). Two other mutants bound HEp-2 cells as poorly as the original MBP-InvD811A construct. These proteins, which had half-maximal binding concentrations 15-20 times higher than the wild type invasin (MBP-Inv192), were a change to a lysine (MBP-InvD811K) or a leucine (MBP-InvD811L). The above analysis indicates that an oxygen-containing residue at position 811 could allow wild type levels of mammalian cell adhesion to the invasin derivatives. The analysis of a double mutant (MBP-InvD811A, K812D) indicated that the proper positioning of such a side chain in the protein structure is critical for cell adhesion. To show this, we took advantage of the fact that the neighboring residue Lys-812 can be replaced without causing a binding defect (MBP-InvK812A, Table I, Fig. 6B). The double mutant was thus constructed, in which the D811A mutation was harbored with a change of residue 812 to aspartate, to determine if the presence of the neighboring aspartate could result in pseudoreversion of the D811A lesion. MBP-InvD811A, K812D, however, was indistinguishable from the derivative harboring the single D811A change (Table I). Therefore, the positioning of this oxygen-containing residue is critical for full protein function.
The Region Surrounding Asp-811 Is Required for Integrin Binding-To further elucidate the importance of the 811 region of invasin in receptor binding, a hybrid protein containing only the carboxyl-terminal 165 amino acids of invasin fused to MBP (MBP-Inv165, Fig. 1B) was constructed. If the region of invasin identified by the Asp-811 lesions is essential for function, then   MBP-Inv hybrid proteins lacking this region should be defective for binding ␣ 5 ␤ 1 . The ability of both MBP-Inv165 and MBP-InvD811A to compete with MBP-Inv192 binding to ␣ 5 ␤ 1 was then analyzed (Fig. 4). Microtiter wells were coated with MBP-Inv192 and probed with purified ␣ 5 ␤ 1 integrin in the presence of competing invasin derivatives. Bound receptor was then detected by ELISA. Wild type MBP-invasin (MBP-Inv192) competed with itself for binding ␣ 5 ␤ 1 at an IC 50 of 80 nM, whereas the IC 50 for MBP-InvD811A was 200 nM. MBP-Inv165 was unable to compete with wild type invasin at any concentration tested. The results of this assay indicate that the region of invasin between residues 794 and 821 is required for cell and receptor binding and that lesions in Asp-811 disrupt cell adhesion.
The D811A Mutation Causes a Requirement for High Concentrations of Divalent Cations to Promote Cell Adhesion-The requirement for an oxygen-containing residue at amino acid position 811 in invasin is reminiscent of other proteins that require divalent cations for activity (49,50). Mutagenesis of residues involved in binding a divalent cation can cause a reduction in activity that is overcome by increasing the divalent cation concentration (51). To determine if the D811A lesion could be similarly overcome by increasing the divalent cation concentration, the effect of Mg 2ϩ ion concentration on the binding of K562 cells to MBP-Inv192 and its mutant derivative MBP-InvD811A was examined. K562 cells were stripped of divalent cations by incubation in buffer containing EDTA (see "Materials and Methods") and tested for the ability to bind saturating amounts of the wild type and D811A derivatives of MBP-Inv in the presence of limiting concentrations of MgCl 2 . HEp-2 cells were not used in this assay, because in the presence of saturating concentrations of invasin, the cells were unable to be sufficiently stripped of divalent cations to abolish all cell binding.
The concentration of MgCl 2 necessary to allow half-maximal adhesion of K562 cells to microtiter wells coated with MBP-Inv192 was 8 M, whereas the mutant protein MBP-InvD811A required 60 M MgCl 2 for the identical level of cell adhesion (Fig. 5). These results indicate that an Asp to Ala change at residue 811 results in a protein that requires higher levels of divalent cations than wild type to promote the identical amount of cell adhesion.
Targeted Mutagenesis Identifies an Additional Residue Critical for Integrin Binding-Targeted mutagenesis has identified Asp-811 as important for protein function. To identify neighboring residues that may be important for integrin binding, site-directed mutagenesis of the residues surrounding Asp-811 was performed. Six additional mutants were constructed, each replacing the flanking residues in the sequence GQNFATD 811 K with an Ala residue. When proteins were analyzed from each of the mutant derivatives, all were unstable compared with the wild type (data not shown). Following the previous results that showed instability could be overcome if the mutations were expressed in hybrid proteins, each mutation was moved onto an MBP-Inv hybrid. The resulting stable proteins were purified by affinity chromatography and tested for the ability to support adhesion of HEp-2 cells to microtiter wells. Hybrid proteins MBP-InvG805A, MBP-InvQ806A, MBP-InvN807A (Fig. 6A) and MBP-InvK812A (Fig. 6B) all promoted cell adhesion as efficiently as MBP-Inv192, with a coating concentration of about 2 nM required to give half-maximal cell adhesion. MBP-InvT810A resulted in only a 2-fold increase in the amount required to support mammalian cell adhesion relative to wild type (Fig. 6B). However, a marked increase in the coating concentration required to give half-maximal cell adhesion (21 nM) was seen for the protein MBP-InvF808A (Fig. 6B). The magnitude of the defect seen in this derivative is nearly identical to that seen as a result of the D811A change and is unique in that only these two residues in this contiguous region appear sensitive to changes to alanine.
Size-fractionated Monomeric MBP-InvD811A Is Defective for Supporting Mammalian Cell Adhesion-One possible explanation for the defective cell adhesion of the F808A and D811A derivatives is that the hybrids containing these mutations may aggregate into large complexes that have low binding affinities for integrin receptors. To investigate this possibility and determine the multimerization states of the proteins, five of the protein preparations were subjected to gel filtration chromatography on a size-calibrated Superose 12 column. For each preparation, the majority of the protein fractionated in the void volume (corresponding to apparent molecular mass Ͼ700,000 daltons) with the remaining present as a species of apparent mass ϭ 60,000 daltons, the predicted size of monomeric MBP-Inv192. The percentage of total hybrid found in the void volume varied from approximately 63% for the wild type hybrid MBP-Inv192 to between 81 and 85% for each of the remaining derivatives ( Fig. 7A; compare lanes V and M). Among the hybrids containing variant sequences, there was no correlation between the binding properties and the amount of protein found in the void volume. For instance, the binding-defective MBP-InvF808A hybrid fractionated identically to the binding-proficient derivatives MBP-InvG805A and MBP-InvD811S (Fig.  7A), with each protein preparation having an identical percentage of the high molecular weight form (Fig. 7A, compare lanes  V and M).
As most of the variant derivatives had less than the wild type of the apparent monomer, the aggregation state of the hybrids could have affected their relative proficiencies at promoting cell adhesion. To address this possibility, the peak fractions corresponding to monomeric MBP-Inv192 and MBP-InvD811A were each pooled, concentrated, and refractionated on Superose 12 to obtain identical homogeneous preparations (Fig. 7B). The peak fractions from each preparation corresponding to this second purification, displayed in Fig. 7B, were then pooled and subjected to cell adhesion studies. Once again, cell adhesion by the monomeric D811A derivative required a coating concentration that was 10 times greater than the monomeric hybrid having the wild type sequence (Fig. 7C). The defect in cell adhesion exhibited by this mutant, therefore, appeared independent of its multimerization state relative to identically fractionated protein having the wild type sequence.

DISCUSSION
Mutational analysis of invasin indicates that only 2 of the 10 aspartate residues in the cell binding domain are critical for integrin binding. One of these, Asp-911, was previously determined to be absolutely required for binding (26) and is unique in that it is the only aspartate absolutely required for adhesion. The second aspartate residue, Asp-811, is identified in this study.
Data suggest that Asp-911 directly interacts with its receptor (26). Mutagenesis of this residue does not affect the stability of the protein, and even conservative amino acid changes showed drastic effects on the ability of invasin to bind ␣ 5 ␤ 1 . A direct role for Asp-811 interaction with the receptor is more difficult to argue. In contrast to the mutation at Asp-911, mutagenesis of Asp-811 results in a protein with residual binding activity. In addition, unlike the Asp-911 lesion, the mutant protein retained some ability to promote entry into mammalian cells in tissue culture. Although further work needs to be done to determine how these two regions of invasin interact to form an integrin binding site(s), it is interesting to note that the invasin derivative with the 811 region deleted retained some residual binding to purified integrin receptor immobilized on plastic, although it was unable to promote cellular adhesion (data not shown). The aspartate at residue 911 may provide the major contact with the receptor, and the residues in the 811 region may support this primary interaction.
Changes of Asp-811 to residues varying in size and charge indicated that an oxygen-containing side chain is required at this site. It is known that divalent cations are required for integrin-ligand interactions (52)(53)(54)(55), and mutations in this residue may alter the binding of divalent cations. Many integrin ligands contain aspartate residues that are critical for binding (21), and it has been proposed that these residues either complex a divalent cation in conjunction with the integrin chains (56 -59) or allow displacement of a divalent cation that is involved in maintaining the receptor in a binding-competent conformation. The finding that the diminished activity of MBP-InvD811A can be partially overcome by increased Mg 2ϩ concentration supports the idea that Asp-811 is involved in one of these activities.
The identification of Asp-811 as an important residue in integrin binding suggested that additional residues in this region might also be involved. Mutagenesis of six of the flanking amino acids, GQNFATD 811 K, showed that only Phe-808 is required for receptor binding. A mutation at this site results in a protein with reduced ability to promote cell adhesion. Interestingly, the sequence FATDK that contains both of these critical residues is also found as part of a larger sequence in the Listeria monocytogenes protein internalin that is repeated three times as FAT(D/S)K (60). Of note, also, is that replacement of Asp-811 in invasin by serine had no detrimental affect on the protein's ability to promote adhesion to mammalian cells, indicating that aspartate and serine could be interchangeable. Like invasin, internalin is a surface protein that enables L. monocytogenes to invade cultured mammalian cells (60). Although the receptor for internalin has not been identi-  7. HEp-2 cell binding to fractionated monomer MBP-D811A is defective. A, the five hybrid protein derivatives noted in A were fractionated on a Superose 12 column ("Materials and Methods"), and the protein peaks corresponding to the void volume of the column (V) or that of apparent monomeric molecular mass ϭ 60,000 daltons (M) were pooled, quantitated for protein ("Materials and Methods"), and subjected to SDS-gel electrophoresis. B, the protein preparations corresponding to the monomer (M) forms of D911A and MBP-Inv192 were each pooled, concentrated, and refractionated on a Superose 12 column. Each preparation gave a single peak predicted for a protein of apparent molecular mass ϭ 60,000 daltons. Void, elution volume of blue dextran. 158K, elution volume of aldolase. 67K, elution volume of bovine serum albumin. C, microtiter wells were coated with decreasing concentrations of MBP-invasin hybrids, which were isolated as the peak monomeric fractions displayed in B and then challenged with 10 5 HEp-2 cells. Binding was quantitated as in Fig. 2. Each point represents the average of three trials from a single experiment. Points are displayed as amount binding relative to maximum. Ⅺ-Ⅺ, MBP-Inv192; q-q, MBP-InvD811A. fied, the FATDK sequence may play analogous roles in both proteins.
The finding that the region surrounding Asp-811 is critical for cell binding parallels results regarding the binding of fibronectin to its integrin receptors. Although no sequence homology exists between the two ligands, it has been proposed that Asp-911 performs a function equivalent to the aspartate in the RGD sequence of fibronectin (26). In addition, our finding that a second region in invasin 100 residues upstream of Asp-911, is involved in ␣ 5 ␤ 1 binding, mimics the involvement of a synergistic site in fibronectin that is also approximately 100 amino acids N-terminal to RGD (35,36,61,62). These two proteins may share similar structural recognition elements in spite of their lack of primary sequence similarity. A functional difference does seem to exist between the identified binding regions of invasin and fibronectin. Studies have shown that each of the binding elements in fibronectin individually retains binding activity and the two together act synergistically (36). Invasin D911A mutants, however, retain no detectable binding (26), indicating that at least one of these two sites is absolutely required for binding. The residual activity seen with the Asp-811 and Phe-808 mutants, however, suggests that this second region may not be absolutely required for binding. This idea is supported by the finding that MBP-Inv165, which lacks the 811 region entirely, has residual ability to bind purified receptor although this low activity renders it unable to promote cell adhesion (data not shown).
Although the results of our alanine scanning mutagenesis do not directly prove that these residues make contact with their receptor, it does implicate their importance in binding. Further high resolution structural analysis of invasin binding should provide more clues to understanding how Phe-808, Asp-811, and Asp-911 interact with ␣ 5 ␤ 1 and elucidate how integrin receptors bind their ligands.