Mutations in the Cytoplasmic Domain of the Integrin (cid:98) 1 Chain Indicate a Role for Endocytosis Factors in Bacterial Internalization*

Mutations that result in defective (cid:98) 1 -integrin focal ad- hesion formation were analyzed for effects on bacterial internalization. Mutations in the cytoplasmic domain of the (cid:98) 1 chain that disrupt the sequence NPIY resulted in integrins deficient in bacterial uptake. Other mutations in the (cid:98) 1 chain that reduced cytoskeletal association showed enhanced bacterial uptake. Replacement of the NPIY sequence of the (cid:98) 1 subunit by the endocytosis in- ternalization sequence PPGY resulted in integrin receptors highly proficient in bacterial internalization, yet severely defective in focal contact localization. Electron microscopy indicated that coated structures associated specifically with bacteria-binding (cid:98) 1 -integrins, with an apparent recruitment of coated pits from ventral cell surfaces to apical surfaces corresponding to nascent bacterial phagosomes. Clathrin inhibition studies indi- cated a role for the adaptor molecule AP2 as well as clathrin in integrin-mediated bacterial internalization. These results indicate that association of (cid:98) 1 -integrins with the cytoskeleton at focal contacts interferes with integrin-mediated bacterial internalization. Also, although actin polymerization is required for bacterial uptake, clathrin is probably involved in bacterial uptake promoted by (cid:98) 1 -integrins. linked to linked to substrate 2,2 -azinobis(3-ethylbenzthiazoline-6-sulfonic acid) Laboratories, Inc. Anti-mouse beads NHS-LC-biotin Cell Transfection and Integrin Expression Analysis— HEp-2 cell transfection was performed as described (23) by electroporation at 450 V and 500 microfarads in a Gene Pulser (Bio-Rad). For each transfection, between 24 and 48 individual G418-resistant transfectants were cloned and tested for chicken integrin receptor expression by an ELISA-based procedure (24). Briefly, individual clones were plated in 96-well tissue culture plates to obtain confluent monolayers, and the cells were fixed with 3.7% paraformaldehyde for 15 min at 22 °C. The wells were washed three times with PBS containing 1 m M MgCl 2 and 0.5 m M CaCl 2 (PBS/Ca/Mg) and blocked in RPMI 1640 medium containing 20 m M HEPES, pH 7.0, and 1% BSA for 60 min at 22 °C. The monolayers were then incubated with CSAT hybridoma supernatant for 2 h and washed three times with PBS/Ca/Mg, and the secondary antibody (rabbit anti-mouse IgG linked to horseradish peroxidase) was added for 60 min. Bound antibody was detected with the peroxidase substrate 2,2 (cid:57) -azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (0.6 mg/ml final concentra- tion), reading the absorbance at 405 nm using an EIA microtiter plate reader (Bio-Rad

Integrins are heterodimeric receptors involved in numerous cellular processes including migration, differentiation, and adhesion to extracellular matrix and cell-surface proteins (1)(2)(3). Members of this family are ␣␤ heterodimeric transmembrane proteins that are involved in both inside-out and outside-in signaling (2,4,5). Ligand specificity of each receptor is determined by the particular ␣ and ␤ chains found in the heterodimer. The unique cytoplasmic domains of the different subunits allow diversity and regulation of receptor function (6). For example, the cytoplasmic domain of the ␤ 1 subunit interacts with the cytoskeleton by binding to actin-binding proteins such as talin (7) and ␣-actinin (8). Mutational analysis as well as peptide inhibition studies have identified regions of the cytoplasmic domain of the ␤ 1 chain important for this cytoskel-etal association (9 -11). The ␤ chain cytoplasmic domain also plays a prominent role in transduction of signals originating either from outside or inside the cell (12)(13)(14). The cytoplasmic domain of the ␣ subunit appears to regulate ligand affinity (15) and participates in processes such as gel contraction (16).
Integrins also promote the internalization of various microorganisms. In the macrophage, the integrin receptor CR3 (or ␣ Mac ␤ 2 ) can mediate internalization of complement-coated particles (17). In normally non-phagocytic cells, viruses and bacteria have been shown to be internalized by integrins containing ␤ 1 , ␤ 3 , and ␤ 5 chains (18,19). The uptake of the enteroinvasive bacterium Yersinia pseudotuberculosis by ␤ 1 -integrins has been studied in detail. Uptake of this microorganism occurs via the bacterial surface protein invasin (20), which binds multiple ␤ 1 -integrins (18). Invasin-mediated bacterial uptake is accompanied by a local rearrangement of the actin network (21) and is inhibited by drugs that antagonize actin polymerization (22). This has led to speculation that direct association of integrins with the cytoskeleton is required during internalization.
By using various anti-integrin antibodies to coat the Grampositive bacterium Staphylococcus aureus, we have been able to mimic the phenotype of invasin-expressing bacteria with respect to uptake. We have shown that integrin-mediated internalization following bacterial attachment to the cell is primarily dependent on the affinity of the bacterial ligand for the integrin (23,24). In this study, we use this technique to study the role of the cytoplasmic domain of the integrin ␤ 1 subunit. We identify a region of the ␤ 1 subunit cytoplasmic domain that is critical for internalization and show a potentially antagonistic relationship between cytoskeletal association of the cytoplasmic domain and bacterial uptake.

MATERIALS AND METHODS
Bacterial Strains, Cell Lines, and Media-To assay bacterial uptake mediated by chicken integrin ␤ 1 chains, S. aureus strain 377 McCowan was grown in Penassay broth and coated with the anti-chicken mAb 1 CSAT as described (24). MC4100/pRI203 is Escherichia coli harboring the inv gene (25). It was grown at 37°C in L broth containing ampicillin at a 100 g/ml final concentration. Salmonella typhimurium strains 1344 and EL451, obtained from Dr. Catherine Lee (Harvard Medical School, Boston), were grown as described (61) in L broth at 37°C without agitation. Human cultured HEp-2 cells were grown in RPMI 1640 medium containing 5% newborn calf serum, and mouse cultured 3T3 fibroblastic cells were grown in Dulbecco's modified Eagle's medium containing 10% calf serum at 37°C in an incubator containing 5% CO 2 .
Plasmid Constructions-Plasmid pRSVneo␤ 1 c containing the fulllength cDNA encoding the chicken integrin ␤ 1 subunit under the control of the Rous sarcoma virus promoter as well as the selectable marker for neomycin was described previously (10). Plasmids containing single amino acid substitutions (F768A, F771L, E784L, N785I, P786A, Y788A,  S790D, N797I, and Y800A) or deletions between residues 759 and 771 and between residues 771 and 790 in the cytoplasmic domain of the chicken integrin ␤ 1 subunit were described previously (10). WAFB was obtained by replacing the 1.1-kilobase pair ClaI-XhoI fragment containing the 3Ј-end of the chicken integrin ␤ 1 chain cDNA of pRSVneo␤ 1 c with the corresponding fragment of pRSVneo␤ 1 c-F768A, and clone FAWB was obtained by the reverse manipulation. To replace Asn 785 -Tyr 788 of the cytoplasmic tail of the integrin ␤ 1 chain with PPGY or to produce the Y788F substitution, the 1.1-kilobase ClaI-SalI fragment of plasmid pRSVneo␤ 1 c-N785I was introduced into plasmid pSELECT (Promega). Mutagenesis was performed following the manufacturer's instructions using the Amp repair primer (Promega) and the mutagenic primer 5-AGTTGTCACTGCACTCTTGTAACCAGGAGGTTCACCCG-TATCCCACTT-3 for the PPGY substitution or the mutagenic primer 5-AGTTGTCACTGCACTCTTGAAAATAGGATTTTCACCCGTATCC-CACTT-3 for the Y788F substitution. The mutagenized fragments were introduced back into plasmid pRSVneo␤ 1 c cut by ClaI and XhoI. Confirmation of the mutations was obtained by DNA sequence analysis (26).
Cell Transfection and Integrin Expression Analysis-HEp-2 cell transfection was performed as described (23) by electroporation at 450 V and 500 microfarads in a Gene Pulser (Bio-Rad). For each transfection, between 24 and 48 individual G418-resistant transfectants were cloned and tested for chicken integrin receptor expression by an ELISAbased procedure (24). Briefly, individual clones were plated in 96-well tissue culture plates to obtain confluent monolayers, and the cells were fixed with 3.7% paraformaldehyde for 15 min at 22°C. The wells were washed three times with PBS containing 1 mM MgCl 2 and 0.5 mM CaCl 2 (PBS/Ca/Mg) and blocked in RPMI 1640 medium containing 20 mM HEPES, pH 7.0, and 1% BSA for 60 min at 22°C. The monolayers were then incubated with CSAT hybridoma supernatant for 2 h and washed three times with PBS/Ca/Mg, and the secondary antibody (rabbit antimouse IgG linked to horseradish peroxidase) was added for 60 min. Bound antibody was detected with the peroxidase substrate 2,2Ј-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) (0.6 mg/ml final concentration), reading the absorbance at 405 nm using an EIA microtiter plate reader (Bio-Rad Model 2250). The cell density in each well was quantitated by the crystal violet method (27). For each individual clone, the relative expression of chicken integrin receptor was determined by calculation of the value (Y x Ϫ Y h )/Y h , where Y x represents the ELISA absorbance value obtained for the clone, and Y h represents the corresponding value obtained for nontransfected HEp-2 cells at the same cell density (24). After transfection, G418-resistant clones showing significant expression of the chicken ␤ 1 chain were expanded and analyzed further. On the average, for the different constructs, 3-6 clones out of 24 -48 tested showed significant expression levels of the chicken ␤ 1 chain. The reason for the low frequency of G418-resistant clones expressing the chicken ␤ 1 chain is unclear; perhaps the chicken ␤ 1 chain competes somewhat inefficiently with the human ␤ 1 chain for association with endogenous ␣ subunits in HEp-2 cells. Transfection of 3T3 cells with chicken ␤ 1 chain derivatives was performed as described previously (10), and ϳ30 -50% of the G418-resistant clones showed significant expression of the chicken ␤ 1 chain.
Bacterial Internalization Analysis-Cellular clones that showed a positive reaction by ELISA were tested for their ability to internalize S. aureus coated with anti-chicken mAb CSAT as described (24). The day preceding the assay, clones were seeded in 24-well plates at a density of ϳ10 4 cells/well in medium free of any antibiotic, and the number of internalized bacteria was determined by the gentamicin protection assay (25). For each individual clone, the relative level of bacterial internalization was determined by calculating the value (U x Ϫ U h )/U h , where U x represents the number of gentamicin surviving colony-forming units for the cloned transfectant, and U h represents the number of gentamicin surviving colony-forming units for the same density of nontransfected HEp-2 cells (24).
Electron Microscopy-HEp-2 cells were seeded in a 6-well tissue culture plate at a density of 5 ϫ 10 5 cells/well. After overnight incubation, the cell monolayers were washed three times with PBS, and 2 ϫ 10 7 E. coli MC4100/pRI203 cells were added in 2 ml of ice-cold RPMI 1640 medium containing 20 mM HEPES, pH 7.0, and 1% BSA (binding medium) to synchronize the bacterial infection. The bacteria were allowed to adhere for 3 h at 22°C before removing unbound bacteria with five washes of PBS. Prewarmed binding medium was then added, and the cultures were shifted to 37°C for 15 min to allow bacterial internalization.
The tannic acid post-fixation, which has been described as an efficient technique to enhance clathrin-coated membrane visualization, also leads to intense staining of bacterial membrane components, 2 rendering difficult the interpretation of results. To visualize coated membrane structures, the samples were processed for transmission electron microscopy as described previously (28). The extent of membrane was quantitated from micrographs enlarged at least 40,000 times using the Bioquant system IV (R & M Biometrics, Inc., Nashville, TN). The incidence of coated pits was determined by scoring coated pits from micrographs over 50 randomly selected cell sections. The total number of coated pits on the apical or ventral membrane was related to the total length of the apical or ventral membrane, respectively. The relative percentage of coated membranes over the ventral plasma membrane could not be unambiguously determined because of numerous cell areas of dense cytoplasmic background.
Potassium Depletion and Analysis of Bacterial Internalization and Transferrin Uptake-The potassium depletion treatment, previously shown to inhibit clathrin-mediated endocytosis, was performed as described (29) with modifications that allowed quantitation of bacterial internalization. Monolayers of HEp-2 cells seeded in 24-well plates at 50% confluency were incubated overnight; rinsed twice in RPMI 1640 medium containing 20 mM HEPES, pH 7.0, and 1% BSA; and incubated in this buffer for 30 min at 37°C. Triplicate monolayers were then washed twice in different buffers and incubated in the appropriate buffers for 30 min at 37°C to test their effects on bacterial uptake. The following buffers were used: buffer K ϩ (140 mM NaCl, 0.5 mM CaCl 2 , 1 mM MgCl 2 , 3 mM KCl, and 50 mM HEPES, pH 7.0) and buffer K ϩ ⌬ (140 mM NaCl, 0.5 mM CaCl 2 , 1 mM MgCl 2 , and 50 mM HEPES, pH 7.0). To test the effect of K ϩ depletion, after 30 min of incubation in buffer K ϩ , the cells were subjected to a hypotonic shock for 5 min in buffer K ϩ ⌬ diluted 1:1 with H 2 O prior to incubation in buffer K ϩ ⌬. The cell monolayers were then analyzed for bacterial internalization and transferrin uptake. Bacterial uptake was quantitated by challenging the treated cells either with MC4100/pRI203 (inv ϩ ) or with S. aureus coated with the anti-␤ 1 -integrin mAb AIIBII (gift of Dr. Caroline Damsky) or with the anti-␣ 5 ␤ 1 mAb VD1 as described previously (23) at a multiplicity of infection of 10 bacteria/cell for 30 min at 37°C. The percentage of internalized bacteria was determined by the gentamicin protection assay (25). Transferrin uptake was determined as described previously (30). Briefly, cells were incubated with 125 I-ferrotransferrin (20 g/ml, 2.4 ϫ 10 3 cpm/ng) in the corresponding buffer for 5 min at 37°C, and unbound transferrin was removed with three washes of ice-cold PBS. the monolayers were treated with 0.3 ml of serum-free medium containing Pronase at a 0.3% final concentration for 1 h on ice. The cells were then transferred to Eppendorf tubes and centrifuged for 2 min, and the pellet and the supernatant were counted for ␥-radiations. Nonspecific internalization was determined by incubation with 125 Iferrotransferrin for 5 min on ice. Specific internalization of 125 I-ferrotransferrin, corresponding to the Pronase-resistant cell-associated counts, was obtained by substracting the nonspecific counts and is expressed as a percentage of total cell-associated counts.
Syringe Loading of HEp-2 Cells with Anti-clathrin mAb-HEp-2 cells were loaded with mAb as described previously (31). Briefly, monolayers of HEp-2 cells were trypsinized, washed once in RPMI 1640 medium containing 5% newborn calf serum, and resuspended in RPMI 1640 medium containing 5% newborn calf serum at a density of 5 ϫ 10 6 cells/ml. The cells were allowed to recover at 37°C for 30 min, and for each sample, 70 l of the cell suspension was aliquoted in a microcentrifuge tube containing 20 g of mAb. The suspension was subjected to 20 strokes using a 27-gauge needle and transferred to a 24-well tissue culture plate. 200 l of RPMI 1640 medium containing 5% newborn calf serum prewarmed to 37°C was added; the plates were centrifuged for 5 min at 800 ϫ g; and the cells were allowed to adhere to the plastic surfaces for 60 min at 37°C. Preliminary experiments using a fluorescently labeled mAb indicated that under these conditions, 20 strokes resulted in optimal mAb-cell association with an average efficiency of ϳ400 ng of mAb/10 6 cells as determined using a fluorometer (Fluoros-kan II, Perkin-Elmer) and ϳ50% of the initial input of cells adhering to plastic. After loading, the cells were washed twice with RPMI 1640 medium containing 20 mM HEPES, pH 7.0, and 1% BSA and infected with E. coli MC4100/pRI203 at a multiplicity of infection of 1 bacterium/ cell or with S. typhimurium SL1344 at a multiplicity of infection of 10 bacteria/cell. The plates were centrifuged for 10 min at 1000 ϫ g and incubated for 30 min at 37°C. The percentage of internalized bacteria was determined using the gentamicin protection assay (25). In each experiment, the results are expressed as a percentage of the number of gentamicin-resistant counts obtained when loading the cells with the irrelevant mAb CSAT. In three experiments performed, the samples were diluted before plating to obtain a number of gentamicin-resistant counts ranging from 100 to 600 viable counts for the CSAT mAb control.

RESULTS
The Chicken Integrin ␤ 1 Subunit Forms Chimeric Receptors in HEp-2 Cells and Mediates Bacterial Uptake-With the exception of B cells, most cultured cell lines express endogenous ␤ 1 -integrins (3) that act as receptors for the bacterial protein invasin (18). To overcome this problem and to study the effect of integrin ␤ 1 chain mutations on bacterial internalization, we took advantage of the fact that species-specific high affinity anti-integrin mAbs can promote bacterial internalization when used to coat the bacterium S. aureus (24). Human HEp-2 cells transfected with mutant derivatives of the chicken ␤ 1 chain were challenged with S. aureus coated with the chicken-specific anti-integrin ␤ 1 chain mAb CSAT. This allowed bacterial internalization by hybrid integrin receptors containing the transfected mutant chicken ␤ 1 subunit to be analyzed in the presence of wild-type endogenous human ␤ 1 -integrins (24).
The chicken integrin ␤ 1 subunit expressed in mammalian cells is reported to associate with the endogenous ␣ chains, forming heterodimers functional in adhesion and cytoskeletal association (3,9). After transfection with the chicken ␤ 1 clone, the expression levels of the chicken/human hybrid integrins on the surface of various transfected clones of HEp-2 cells were determined by an ELISA-based assay (see "Materials and Methods"). From immunoprecipitation experiments, the levels of chicken ␤ 1 /human hybrid integrins in the various transfectants were generally Ͻ25% of the endogenous levels of human integrin ␤ 1 chains (data not shown). Transfectants expressing various derivatives of the chicken ␤ 1 chain bearing the F768A, P786A, or N785I substitution in the ␤ 1 cytoplasmic domain (Table I) were also analyzed by immunoprecipitation using the CSAT mAb. No obvious qualitative or quantitative differences in association with the endogenous ␣ subunits could be observed among the different mutants when compared with the wild-type chicken ␤ 1 chain (data not shown). The values obtained by scanning the intensity of the band corresponding to the chicken ␤ 1 chain were in good agreement with the values obtained in the ELISA-based assay, and these ELISA-based determinations were used to quantitate the expression levels of chicken/human hybrid receptors at the surface of the various transfectants (Table I).
Transfectants expressing the wild-type chicken ␤ 1 chain internalized significant levels of S. aureus coated with the antichicken ␤ 1 -integrin mAb CSAT (Table I, ␤ 1 c-wild type, uptake/ expression equal to 4.1) when compared with nontransfected HEp-2 cells. Therefore, the chicken ␤ 1 /human hybrids could be used to analyze the effect on bacterial uptake of mutations located in the cytoplasmic domain of the integrin ␤ 1 chain.
The NPIY Sequence Located at Residues 785-788 of the Cytoplasmic Tail of the Integrin ␤ 1 Subunit Is Critical for Bacterial Internalization-Mutational analysis of the cytoplasmic domain of the integrin ␤ 1 chain has identified three clusters of residues important for localization of ␤ 1 -integrins to focal contacts (cyto-1, -2, and -3) (Fig. 1) (10). Both cyto-2 and cyto-3 contain NPXY motifs that have been implicated in rapid endocytosis of the low density receptor receptor (33), although their role in integrin function is unclear. To determine whether these residues of the ␤ 1 -integrin cytoplasmic domain are critical for bacterial internalization, transfectants of HEp-2 cells expressing derivatives of the chicken/human hybrid integrins were tested for their ability to internalize S. aureus coated with the anti-chicken integrin mAb CSAT as described (24).
Transfectants were cloned, and the efficiency of bacterial internalization as well as the levels of expression of chicken/ human hybrid integrins were normalized relative to nontransfected HEp-2 cells. Several substitutions in the cytoplasmic region cyto-2 (NPIY) (Fig. 1) (10) resulted in no detectable levels of bacterial internalization when compared with non- a The expression levels of chicken ␤ 1 -integrin were determined by an ELISA-based procedure (see "Materials and Methods"). Each value is the mean of three to six determinations from independently derived clones.
b The percentage of bacterial uptake was determined by the gentamicin protection assay (25) after normalization to the cell density (24). Units are defined under "Materials and Methods." Each value is the mean of three to six determinations from independently derived clones.
c Uptake levels of 0.00 represent uptake efficiency at or lower than levels of internalization by nontransfected controls.
d The relative percentage of bacterial uptake was obtained by arbitrarily dividing the values of "% bacterial uptake" by the values of "relative chicken ␤ 1 -integrin expression." Ϯ refers to standard deviation. transfected HEp-2 cells (Table I, N785I, P786A, and Y788E). In addition, the deletion (residues 759 -771) removing the cyto-1 region as well as the deletion (residues 771-790) removing the cyto-2 region also abolished the ability of integrin-mediated bacterial internalization (Table I). In contrast, the single amino acid substitutions F768A, F771L, and E784L as well as other mutations in the region called cyto-3 ( Fig. 1) (10) resulted in increased efficiency of bacterial uptake relative to wild-type transfectants ( Table I). Several of the mutants that were more efficient at promoting bacterial uptake than the parental integrin ␤ 1 chain exhibited a reduction in focal contact association (Table I) (10).
These results indicate that residues in the membrane-proximal NPIY (cyto-2) region as well as the integrity of the cyto-1 region are critical for integrin-mediated bacterial internalization. The fact that single amino acid substitutions in the cyto-1 and cyto-3 regions depressing cytoskeletal association resulted in an increased uptake efficiency suggests that cytoskeletal association of integrins at focal contacts interferes with bacterial internalization.
The NPIF and PPGY Internalization Sequences Can Substitute for the Integrin ␤ 1 Chain NPIY Sequence in Integrinmediated Bacterial Internalization-NPXY internalization sequences are presumed to allow localization of receptors mediating endocytosis to clathrin-coated pits, perhaps by interacting with adaptors or associated molecules (34). In the case of the low density receptor receptor, replacement of NPVY by the NPVF sequence allows rapid endocytosis (35) because the Tyr 3 Phe change does not disrupt the tight-turn conformation of the wild type (36). As shown in Table I, substitution of Tyr 788 with Phe resulted in integrins that were more efficient at promoting bacterial uptake than the parental integrin ␤ 1 chains (Table I). This is consistent with the proposition that the presence of an aromatic amino acid at residue 788 allows internalization.
To determine if the membrane-proximal NPIY sequence could be substituted with another internalization sequence, the cyto-2 NPIY sequence of the integrin ␤ 1 chain was replaced by PPGY, which determines rapid internalization of the lysosomal acid phosphatase via coated pits (37,38). As shown in Table I, the integrin receptor containing the PPGY sequence was competent to mediate bacterial internalization (Table I, NPIY 3 PPGY) with an efficiency comparable to that of other cytoplasmic mutants of the ␤ 1 chain that were more proficient than the wild type. Binding studies performed on cells expressing mu-tant integrins bearing the PPGY or N785I substitution indicated that these receptors bound the CSAT mAb at a similar efficiency (Fig. 2), ruling out the possibility that the variations in bacterial uptake were due to a change in ligand-receptor affinity (23).
To analyze the ability of this integrin derivative to localize to focal contacts, immunofluorescence staining of 3T3 transfectants using the anti-chicken ␤ 1 -integrin mAb W1B10 was performed on cells plated onto fibronectin-coated surfaces as described (10). As opposed to parental chicken ␤ 1 -integrins (Fig.  3C), the PPGY substitution demonstrated a strong defect in localization to focal contacts with a diffuse punctate staining (Fig. 3A). The defect in focal contact localization appeared considerably more severe for the PPGY substitution (Fig. 3A) than for the N785I substitution (Fig. 3B), which had been previously reported to show a defect in focal contact formation (10). The phenotype of the PPGY mutation was therefore a defect in cytoskeletal association coupled with an enhancement of bacterial internalization.
Clathrin Associates with Forming Phagosomes during Integrin-mediated Bacterial Internalization-Our data from mutational analysis of the cytoplasmic domain of the integrin ␤ 1 chain suggest a role for clathrin during integrin-mediated bacterial internalization. This is consistent with earlier studies that demonstrate in macrophages an association of clathrin FIG. 2. Saturation curves of the CSAT mAb binding to 3T3 transfectants expressing chicken/mouse integrins containing the N785I or PPGY mutation. Two-fold serial dilutions of the 3T3 transfectants were plated in 24-well plates, and the cells were allowed to adhere to the plastic surfaces. Monolayers were incubated with RPMI 1640 medium containing the CSAT mAb (2 g/ml) for 3 h at 22°C, followed by incubation with rabbit anti-mouse antibody conjugated to horseradish peroxidase for 2 h at 22°C. Bound antibody was quantitated using the 2,2Ј-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) substrate, reading the absorbance at 405 nm (see "Materials and Methods"). q, 3T3 transfectant expressing the chicken/mouse integrins with the N785I substitution; Ⅺ, 3T3 transfectant expressing the chicken/ mouse integrins with the PPGY substitution; Ç, parental 3T3 cells. with phagosomes in the process of internalizing latex beads (39). Indirect immunofluorescence staining of HEp-2 cells indicated that clathrin associates with E. coli (inv ϩ ) during the initial events of internalization (data not shown). To characterize the potential role of clathrin during integrin-mediated bacterial internalization, the distribution of clathrin was analyzed by thin-section transmission electron microscopy.
HEp-2 cell monolayers were incubated with E. coli MC4100/ pRI203 (inv ϩ ) for 2 h at 22°C to allow bacterial binding. After washing to remove unbound bacteria, the cells were shifted at 37°C for 10 min to allow internalization (see "Materials and Methods"), and the samples were prepared using standard protocols (28).
In uninfected HEp-2 cells, coated pits appeared to be associated primarily with the basolateral surface of the cell with a density of ϳ11.2 pits/mm compared with 3.4 pits/mm for the apical surface (Table II, column A). In HEp-2 cells challenged with bacteria, the distribution of coated pits was dramatically different. In these cells, most of the coated pits appeared to be located on the apical surface with an incidence of 11.4 pits/mm of membrane, whereas the density of basolaterally located pits decreased to 5.8 pits/mm of membrane (Table II, column B). The vast majority of the pits (83%) that were on the apical surface appeared to associate with nascent bacterial phagosomes (Table II and Fig. 4C, curved arrows). Also, large coated membrane lattices (Ͼ500 nm) were often visible at the cell membrane region juxtaposing the engulfed bacteria (Fig. 4, A  and B, arrows). In uninfected cells, such structures were detected primarily at the basal surface (Table II). Consistent with these results, quantitation of coated membranes indicated that in uninfected cells, clathrin structures associated with ϳ0.3% of the apical surface (Table II), whereas 4.2% of the apical surface of phagocytizing cells was covered by clathrin-coated membranes, with 90% of these coated membranes associated with bacterial phagosomes (Table II).
In these experiments, the thin sections showed an average of 3.85 cell-associated bacteria/cell section, with ϳ25% of the cellbound bacteria associated with clathrin-coated membranes (Table II). This number increased to 50% for cell sections showing only 1 or 2 bound bacteria (data not shown). Some bacteria were internalized in phagosomes showing several coated pitlike structures (Fig. 4D, arrows), whereas other phagosomes were devoid of such structures (Fig. 4D, arrowhead).
Immunostaining with the anti-adaptor molecule AP2 antibody AP-6 also showed the formation of large adaptin lattices specifically associated with the plasma membrane juxtaposing the bacteria that were not detected in uninfected control cells (data not shown). These data indicate that clathrin tends to redistribute to the apical surface of HEp-2 cells that are internalizing invasin-expressing bacteria and that the majority of FIG. 4. Invasin-promoted uptake generates the formation of large coated membrane lattices containing coated pit-like structures. Monolayers of HEp-2 cells were infected with MC4100/pRI203 (inv ϩ ) on ice for 2 h. Unbound bacteria were washed away, and the cells were shifted to 37°C for 15 min and immediately fixed (see "Materials and Methods"). The cells were post-fixed and processed for transmission electron microscopy as described (28). In A and B, the arrows point to large coated membrane lattices underlying the bacteria and containing coated pit-like structures. The arrowhead in B shows a similar structure at the ventral surface of the cell. In C, the curved arrows point to coated pits underlying bound bacteria, and the arrow indicates a coated vesicle. In D, the cells were immunostained with the anti-clathrin mAb X-22 followed by anti-mouse IgG linked to horseradish peroxidase. Bound mAbs were revealed by staining with 3,3Ј-diaminobenzidine (see "Materials and Methods"). The arrows indicate phagosomes associated with several coated pit-like structures, and the arrowhead points to a bacterial phagosome located closer to the cell nucleus devoid of such structures. Bars ϭ 1.0 m (A, B, and D) and 0.5 m (C).

TABLE II Distribution of clathrin-coated structures on HEp-2 cell plasma membrane Thin-section micrographs of uninfected HEp2 cells (column A) or
HEp-2 cells infected with MC4100/pRI203 (column B) (see "Materials and Methods") were scored for coated pits. For each section, the total number of coated pits was determined as was the percentage of coated membrane relative to the length of the apical or ventral surface (see "Materials and Methods").

HEp-2 cells (A) a
HEp-2 cells ϩ MC4100/pRI203 (B) No. coated pits/mm apical membrane clathrin structures detected on the apical surface are associated with phagosomes surrounding bacteria being internalized. Cell Loading with Anti-clathrin Antibody Results in Inhibition of Integrin-mediated Bacterial Uptake-Potassium depletion has been shown to efficiently inhibit clathrin-mediated endocytosis (29,40). To determine the effects of potassium depletion on the ability of HEp-2 cells to internalize bacteria expressing invasin, HEp-2 cell monolayers were incubated in K ϩ -free medium prior to infection with MC4100/pRI203 (inv ϩ ) or with S. aureus coated with anti-␤ 1 -integrin mAbs (see "Materials and Methods"). As shown in Fig. 5, potassium depletion strongly inhibited bacterial internalization mediated by the anti-␤ 1 mAb AIIBII (Fig. 5A, ␤ 1 ), the anti-␣ 5 ␤ 1 mAb VD1 (Fig.  5A, ␣ 5 ), or invasin (Fig. 5A, INV), with an inhibition level ranging from 90 to 95% when compared with cells treated in K ϩ -containing buffer (Fig. 5A, compare solid and hatched  bars). In control experiments, this treatment was found to inhibit 125 I-ferrotransferrin uptake, a ligand that is endocytosed via clathrin-coated pits (Fig. 5B, compare hatched and  solid bars), although in this case, the inhibition appeared less pronounced (ϳ70%) than the one observed for bacterial uptake. Cell treatment with 0.45 M sucrose or cytosol acidification, both of which have also been described as inhibiting coated pitmediated endocytosis (41)(42)(43), also strongly inhibited bacterial internalization (data not shown).
Cell loading of antibodies against clathrin and AP2 adaptor molecules has been shown to result in inhibition of endocytosis of transferrin (44). To demonstrate a role for clathrin in integrin-mediated bacterial uptake, HEp-2 cells were loaded with the anti-clathrin mAb X-22 (45) or with the anti-adaptor AP2 mAb AP-6 (46) using the syringe loading technique (31). After loading with mAb, HEp-2 cells were allowed to adhere to plas-tic surfaces and infected with MC4100/pRI203 (inv ϩ ) (Fig. 6,  solid bars). As a control, HEp-2 cell samples were also infected with S. typhimurium strain SL1344, which is internalized via an integrin-independent pathway (Fig. 6, hatched bars). Cell loading with the anti-clathrin mAb resulted in a 60% inhibition of invasin-mediated bacterial internalization (Fig. 6, solid bar, X-22) when compared with loading with the irrelevant mAb CSAT (Fig. 6, solid bar, CSAT), whereas loading with the anti-adaptor AP2 mAb resulted in ϳ50% inhibition (Fig. 6, solid bar, AP-6). In contrast, internalization of S. typhimurium was not significantly affected by cell loading with anti-clathrin or anti-adaptor mAb (Fig. 6, hatched bars, X-22 and AP-6), indicating that the inhibition seen was not due to a general defect in bacterial internalization, but was specific for the integrin-mediated uptake pathway. These results are consistent with a functional role for clathrin and AP2 adaptor molecules during integrin-mediated bacterial uptake.
Cell loading with higher concentrations of the anti-clathrin and anti-adaptin mAbs resulted in similar levels of partial inhibition of integrin-mediated bacterial internalization (data not shown). These partial levels of inhibition could reflect the limitations of the technique used (31,44), but they could also indicate the existence of multiple bacterial internalization mechanisms, some of which may be independent of clathrin and the AP2 adaptor. DISCUSSION The results presented here clearly indicate that integrinmediated bacterial internalization and focal contact localization have different requirements. Previous studies demonstrated that the cytoplasmic domain of the ␤ 1 subunit allows association of integrins with the cytoskeleton as well as localization of integrins to focal contacts (9,32). Three clusters of amino acids distributed along the cytoplasmic domain of the ␤ 1 subunit, denoted cyto-1, -2 and -3, contribute to this process ( Fig. 4) (10). Substitution mutations in each of these clusters have weak but distinct effects on focal adhesion formation. In contrast, substitution mutations depressing bacterial internalization are centered in cyto-2. Such mutations have very strong effects on bacterial internalization (Table I) and weak effects on focal contact formation (10). FIG. 5. A, effect of potassium depletion on integrin-mediated bacterial uptake. For potassium depletion, HEp-2 cell monolayers were subjected to a hypotonic shock and incubated for 30 min in K ϩ -free buffer (see "Materials and Methods"). After this period, MC4100/pRI203 (inv ϩ ) or anti-␤ 1 -integrin mAb-coated S. aureus resuspended in the same buffer was added to the monolayers at a multiplicity of infection of 10 bacteria/ cell, and the incubation was allow to proceed at 37°C for 30 min. The percentage of bacterial internalization was determined as described previously (25). ␤ 1 , monolayers challenged with S. aureus coated with the anti-␤ 1 mAb AIIBII; ␣ 5 , monolayers challenged with S. aureus coated with the anti-␣ 5 ␤ 1 mAb VD1; INV, monolayers challenged with MC4100/pRI203 (inv ϩ ). Hatched bars, buffer containing potassium; solid bars, buffer without potassium as described under "Materials and Methods." Each bar represents the mean Ϯ S.D. of three determinations. B, effect of potassium depletion on 125 I-labeled transferrin (TFN) uptake. HEp-2 cells were potassium-depleted as described above, and the uptake of transferrin was determined as described previously (30). Cells were incubated with 125 I-labeled transferrin for 5 min at 37°C. The cells were washed three times with ice-cold PBS, and the monolayers were treated with 0.3% Pronase for 1 h on ice. The cells were then transferred to Eppendorf tubes and centrifuged for 2 min, and the pellet and the supernatant were counted for ␥-radiations. Nonspecific internalization was determined by incubation with 125 I-labeled transferrin for 5 min on ice. The results are expressed as percentage of total cell-associated counts. Each bar represents the mean Ϯ S.D. of three determinations.
FIG. 6. Effect of syringe loading of anti-clathrin and anti-AP2 mAbs on integrin-mediated bacterial uptake. HEp-2 cells were resuspended in RPMI 1640 medium containing 5% newborn calf serum and loaded with mAbs using a 27-gauge needle. Cells were allowed to adhere to plastic surfaces for 60 min at 37°C, washed with prewarmed medium, and infected with bacteria for 30 min at 37°C. The percentage of internalized bacteria was determined by the gentamicin protection assay (25) after normalization to the value obtained for the CSAT mAb control (see "Materials and Methods"). Each bar represents the mean Ϯ S.D. of three independent determinations. Hatched bars, cells infected with S. typhimurium SL1344; solid bars, cells infected with MC4100/ pRI203 (inv ϩ ). Cells were loaded with anti-adaptor AP2 mAb AP-6, anti-clathrin mAb X-22, bovine serum albumin (BSA), or the irrelevant anti-chicken integrin ␤ 1 chain mAb CSAT.
These results were unexpected because it had been thought that association of integrins with the cytoskeleton was required for bacterial uptake. This hypothesis was based on the fact that invasin-mediated uptake by cultured epithelial cells is accompanied by a reorganization of F-actin around the internalized bacteria (21) and is inhibited by cytochalasins (22). Also, the role of actin in phagocytosis has been well established as actin filaments as well as actin-binding proteins associate with nascent phagosomes in macrophages (47). Electron micrographs of HEp-2 cells, however, show that during the early phases of bacterial uptake, actin filaments, visualized by decoration with myosin S-1 pieces, associate mainly with cell protrusions at the tip of the nascent phagosome, with only a few nucleation sites associated with the host cell membrane. 3 Clearly, actin is involved in bacterial uptake, but whether it directly binds to integrin receptors during uptake is questionable based on the results of this work.
In this study, we found that deletions removing the cyto-1 and cyto-2 regions as well as amino acid substitutions of residues 785-788 (NPIY) in cyto-2 abolished the ability of the integrin receptor to mediate bacterial uptake. Physical studies using short peptides similar to NPIY indicate that tyrosinecontaining motifs have a high propensity to promote tight-turn configurations (36,48). When these sequences are proximal to an ␣-helical segment, they have been shown to allow receptor localization to clathrin-coated pits (33), perhaps by interacting with adaptor molecules (34). Although the NPIY sequence corresponding to cyto-2 and the predicted ␣-helical cyto-1 appear to follow this paradigm closely, there has been little previous evidence that these determinants allow integrin association with clathrin-coated pits. For example, it has been reported that mutations in the NPXY sequences do not prevent internalization of the ␣ 5 ␤ 1 -integrin in Chinese hamster ovary cells (49). On the other hand, immunofluorescence and immunocolloidal gold experiments have shown that a portion of the fibronectin receptor population localizes within clathrin-coated pits (50 -52). It is possible that ␤ 1 -integrins are internalized via multiple pathways, one of which requires the ␤ 1 chain cytoplasmic sequence NPIY and clathrin-coated pits.
In this study, large lattices of clathrin and AP2 adaptor complexes are formed beneath bound bacteria in the early stages of the internalization process. These structures were not detected in uninfected cells and appeared to be induced during the bacterial internalization process. Consistent with this observation, a redistribution of the clathrin coat from the basolateral to the apical cell surface occurred in cells that were in the process of internalizing bacteria. The recruitment of coated pits from the basolateral to the apical surface of the cell, in specific association with bacterial phagosomes, suggests that the accumulation of clathrin is a specific event triggered by the bacterial internalization process.
These observations are similar to previous reports that demonstrated that particles internalized by phagocytic cells were encompassed by clathrin-coated phagosomes (39), but the functional significance of these previous findings was unclear. In the results reported here, inhibition of uptake following cell loading with anti-clathrin or anti-adaptin mAb indicated that coat-associated proteins are required for integrin-mediated bacterial uptake. In addition, potassium depletion (Fig. 5), high concentrations of sucrose, and cytosol acidification (data not shown), treatments that have been shown to prevent coated pit-mediated endocytosis, resulted in a strong inhibition of integrin-mediated bacterial uptake. These results suggest that clathrin plays a role in integrin-mediated bacterial internaliza-tion, although it is not clear if processes different from those seen in clathrin-mediated endocytosis are also involved.
Whether the sole role of the proximal NPIY sequence of the ␤ 1 chain cytoplasmic domain during bacterial internalization is to allow association of integrins with AP2 complexes remains to be determined. Mutations of these residues lead to total elimination of bacterial uptake, whereas introduction of anti-clathrin antibody in the cell cytosol only inhibits internalization by ϳ60%. Perhaps these residues are critical for multiple bacterial internalization pathways, some of which do not require clathrin. Alternatively, clathrin association with integrins may be required for bacterial internalization, and the partial inhibition observed in the antibody loading experiment may be due to inherent technical limitations.
Several mutations in the cytoplasmic domain of the ␤ 1 chain caused impaired integrin localization to focal contacts, yet resulted in a bacterial internalization efficiency that was higher than that promoted by the parental ␤ 1 -integrin. Similarly, substitution of the proximal NPIY sequence by a structurally related sequence, PPGY (cyto-2) (Fig. 2), resulted in an integrin receptor highly efficient in promoting bacterial uptake, yet defective in focal plaque localization. One explanation for this observation is that the proximal NPIY sequence is involved in integrin recycling (53), but this process is inefficient unless the receptor breaks its association with the cytoskeleton. Presumably, the PPGY mutation shifts the balance from integrin association with focal contacts to a recycling receptor readily available for bacterial internalization. The role of the more distal NPIY region is less clear. Although the mutations in the more distal NPXY sequence (NPKY) stimulate uptake, we cannot eliminate a possible role for this region in the uptake process. It is possible that the presence of the amino acid changes used in this work may not disrupt signaling functions in cyto-3 that are important for uptake.
Previous studies on Fc receptors in the macrophage have suggested that phagocytosis and endocytosis of the receptor involve different processes (54). Mutational analysis of the cytoplasmic domain of the Fc receptor for IgE in macrophages indicates that residues involved in endocytosis of the receptor are different from residues involved in phagocytosis of IgEcoated erythrocytes (54). On the other hand, as pointed out above, coated pits have been shown to associate with early phagosomes during Fc receptor-mediated phagocytosis in macrophages (55). Perhaps with some receptors, clathrin associates with phagosomes mainly to ensure receptor recycling, rather than playing an active role during phagocytosis, as indicated by our results with integrin-mediated bacterial internalization. To illustrate that phagocytic internalization of different receptors may occur via different pathways, phagocytosis via the receptors for C3 derivatives is not accompanied by a strong release of H 2 O 2 , whereas Fc receptor-mediated uptake results in such a burst (56). It is possible that phagocytosis via Fc receptors in macrophages involves different processes than integrin-mediated bacterial internalization in normally nonphagocytic HEp-2 cells, and the results described here may not apply to all classes of receptors.
Phagocytosis via either Fc receptors or integrins seems to require tyrosine phosphorylation (57,58) and formation of Factin (21,22). Although the integrin ␤ 1 chain presents a potential tyrosine phosphorylation site at Tyr 800 (59), its substitution with an alanine residue did not influence bacterial uptake. It is likely, however, that tyrosine phosphokinases play an active role in transmitting signals necessary for bacterial internalization steps, and a number of studies indicate that a variety of cytoplasmic components are phosphorylated in response to adhesion of integrins to substrates (4, 13, 60). Fur-ther studies aimed at identifying these signals should allow a better understanding of the internalization process.