The NITY motif of the beta-chain cytoplasmic domain is involved in stimulated internalization of the β₃ integrin A isoform

Integrins are a family of adhesion receptors that mediate cell-extracellular matrix or cell-cell interactions (Hynes, 1992; Schwartz et al., 1995). The α/β heterodimeric surface proteins provide dynamic links between cells and extracellular matrix, involving transfer of information between the two key domains of the integrin, the extracellular ligand binding domains and the intracellular cytoplasmic tails (Clark and Brugge, 1995). Signaling through integrins is bi-directional (Schwartz et al., 1995; Dedhar and Hannigan, 1996). Inside-out signals regulate the affinity of integrin for ligand and control cell adhesion (Ginsberg et al., 1992). Outside-in signals emanate from integrin binding to the extracellular matrix, and regulate many fundamental processes (Rosales et al., 1995). These include cell survival and proliferation, cellular differentiation, morphogenesis, cell migration, phagocytosis, gene transcription, and integrin localization (Schwartz et al., 1995).

β₃ integrins play an important role in hemostasis, thrombosis, wound repair, angiogenesis and atherosclerosis (Shattil, 1995; Byzova et al., 1998) and are expressed at high levels by many vascular cells including platelets, neutrophils, endothelial and smooth muscle cells (Shattil, 1995; Byzova et al., 1998). β₃ integrins have emerged as modulators of a variety of cellular functions including growth, intracellular signal transduction mechanisms, or gene regulation (Shattil, 1995). The participation of the β₃ cytoplasmic domain in modulating integrin as well as cellular functions is well-established (Ginsberg et al., 1992). Therefore, structural differences in the primary sequences of the intracellular domains are predicted to carry significant changes in a variety of integrin-mediated events.

The cytoplasmic domain of β₃ interacts with various intracellular factors, including cytoskeletal, regulatory and signal transducing proteins (Hemler, 1998). β₃-endonexin, a recently identified cytoplasmic polypeptide of 111 amino acids, has been shown to associate exclusively with the cytoplasmic domain of β₃ (Shattil et al., 1995). Binding specificity is conferred by the unique, membrane-distal NITY motif within the cytoplasmic domain of β₃ chain (Eigenthaler et al., 1997). When cotransfected with α_IIbβ₃ into chinese hamster ovary (CHO) cells, β₃-endonexin increased the integrin affinity for ligand (Kashiwagi et al., 1997).

Integrins are known to be internalized via endocytosis and the cytoplasmic domains play an important role in this process (Bretscher, 1989; Bretscher, 1992; Arnaout, 1990). Receptor cycling and internalization play an important role in regulation of integrin function. Moreover, the analysis of the regulatory role of β₃ integrin cytoplasmic domains in multiple cellular functions has recently acquired a new level of complexity with the identification and characterization of a number of variants of β₃ integrin (Kuppevelt et al., 1989; Kumar et al., 1997). Such isoforms results from alternative splicing events that generate unique cytoplasmic sequences (β₃-A, -B and -C). Cytoplasmic variants of β₃ integrin affect ligand binding affinity, specificity, subcellular distribution (LaFlamme et al., 1994) and signal transduction (Akiyama et al., 1994). For instance, coexpression of β₃-B with α_IIb in CHO cells does not reveal PAC-1 binding (ligand specific for activated α_IIbβ₃) in contrast to β₃-A (O’Toole et al., 1995). Moreover, in contrast to HEK cells transfected with β₃-A, β₃-C transfected cells failed to adhere to osteopontin, an α_vβ₃ matrix protein (Kumar et al., 1997). Whereas β₃-A is localized in focal adhesions (Ginsberg et al., 1992) β₃-B shows a diffuse cellular distribution (LaFlamme et al., 1994). Furthermore, β₃-A but not β₃-B signal phosphorylation of focal adhesion kinase (FAK) (Akiyama et al., 1994).

The cytoplasmic domains of integrin β subunits, expressed in the context of single-subunit chimeric receptors, have earlier been shown to be sufficient to both mimic and modulate integrin functions (Akiyama et al., 1994; Tahiliani et al., 1997). In an attempt to study the functional significance of β3 variants, we constructed tripartite chimeric receptors, bearing the CH2 and CH3 domains of human IgG1 as extracellular portions, the transmembrane domain of CD7 and various β₃ or β₁ cytoplasmic tails as intracellular elements, respectively (Fig. 1).

We found that β₃-chimeras are differentially cell surface expressed, depending on the cytoplasmic amino acid sequence of the variant isoform. We have further defined an important role of the terminal NITY motif, present exclusively in the cytoplasmic tail of β₃-A, in the stimulated internalization of the β₃-A integrin that involves it’s cytoplasmic binding partner β₃-endonexin.

mAb Ab-15 is directed against the extracellular domain of β3 (kindly provided by Dr Mark Ginsberg, La Jolla, USA) and was conjugated to biotin according to standard protocols. Anti-IgG1 and rabbit polyclonal anti-GFP antibodies were purchased as fluorochrome-conjugates (FITC or Texas Red as indicated) from Dianova. Phycoerythrin (PE)-conjugated streptavidin was from Immunotech. Oligonucleotides were synthesized on a model 391 DNA synthesizer. Vitronectin and echistatin was purchased from Sigma. Echistatin was labeled with fluorescein according to standard protocols.

P5C7, a derivative of pRK5 (Lev, 1995) containing a modified polylinker region, was used as the mammalian expression vector. Cytoplasmic Ig fusion proteins and the transmembrane CD16/CD7 chimeras have been described previously (Kolanus et al., 1993). The transmembrane Ig fusion proteins comprise the leader sequence from CD5, extracellular CH2 and CH3 domains from human IgG1, the CD7 transmembrane domain, and the full length intracellular tail of the integrin as indicated (Fig. 1). An Ig construct that lacks the cytoplasmic domain served as control (Ig-control). The cytoplasmic domains of β₁-A and β₃-A were amplified by the polymerase chain reaction (PCR) from wild-type β₁-A and β₃-A cDNA sequences (generously provided by Dr David Phillips, San Francisco), using oligonucleotides bearing the respective coding sequences and restriction sites. Following digestion with MluI and NotI, the respective PCR products were cloned into P5C7. The β₃-B and β₃-C constructs were generated in an analogous fashion. Cytoplasmic sequences encoding the β₃-A variants β₃-A-NPKY, β₃-A-Y759A, β₃-A-I757P, and the β₁-A variant β₁-A-NITY were also generated by PCR.

Both isoforms of β₃-endonexin were cloned from a natural killer cell cDNA library through amplification by PCR using the oligonucleotides: 5′-GGG GCG ACG CGT ATG ATG CCT GTT AAA AGA TCA CTG AAG TTG GAT GGT CTG-3′ (fw)/5′-GGG GCG GCG GCC GCT TCA CAG AGG TTG TGA CAT CTG AGG CTG ACC TTT GTG-3′ (rev) (β₃-endonexin long) or 5′-GGG GCG GCG GCC GCT TCA CTG TAT ACT ACT TAA ATT TTG CAT TAT CTC CAT-3′ (rev) (β₃-endonexin short). The resulting PCR fragments were fused to the respective 3′ termini of an eGFP (Clontech) cloning cassette. β₃-endonexin mutants with deletion of the putative K62RKK nuclear import sequence were generated in a two-step PCR strategy. All constructs were identified by restriction digestion, purified by CsCl centrifugation, and verified by DNA sequencing before transfection.

Chinese hamster ovary (CHO) cells were obtained from American Type Culture Collection (ATCC, Rockville, MD) and maintained in Dulbecco’s modified Eagle medium (DMEM, Sigma) supplemented with 10% fetal calf serum (FCS) (Sigma), 1% L-glutamine, 1% penicillin and streptomycin, and 1% nonessential amino acids. Stable CHO cell lines expressing α_IIbβ₃ and α_IIbβ₃ (Y759A) have been characterized earlier (Ylänne et al., 1995). The cell lines were cultivated in Dulbecco’s modified Eagle’s medium (Sigma) supplemented with 0.75 mg/ml geneticin (G418)-disulfate.

CDNA constructs were expressed in CHO cells by liposome-mediated transfection (Superfect, Qiagen). Twenty-four hours before transfection, cells were plated on 6-well culture plates. A total of 2 μg of each construct and 10 μl of Superfect (Qiagen) reagent were incubated at room temperature for 10 minutes in 110 μl of unsupplemented DMEM or M-199, respectively. 600 μl of supplemented medium was then added and the DNA-Superfect complexes overlaid onto the cells. The cells were incubated for 2 hours at 37°C, washed with phosphate buffered saline, and then incubated at 37°C with complete medium. Medium was changed after 24 hours and the cells were analyzed at 48 hours.

Transient transfectants were harvested in Tyrode’s buffer containing 0.1 mg/ml L-1-tosylamido-2-phenylathylchloromethyl ketone-treated trypsin (Sigma) and 3.5 mM EDTA. After a 5-minute incubation at room temperature, the cells were diluted with Tyrode’s buffer containing 0.1% soybean trypsin inhibitor (Sigma) and 10% bovine serum albumin and fixed by addition of equal volume of 2% formaldehyde in PBS for 30 minutes at room temperature. Thereafter, cells were collected by centrifugation at 1200 rpm for 5 minutes, and washed once in Tyrode’s buffer. For surface Ig staining, fixed cells were resuspended in 100 μl Tyrode’s buffer containing excess of phycoerythrin (PE)-conjugated mouse anti-IgG₁ mAb (50 μg/ml). After 1 hour incubation cells were washed twice with 2% glycine in PBS and resuspended again for intracellular Ig staining in 100 μl Tyrode’s buffer containing fluorescein (FITC)-conjugated mouse-anti-IgG₁ mAb (50 μg/ml) and 0.2% Triton X-100 to permeabilize the fixed cells. After a further 1-hour incubation, cells were washed twice again and resuspended in 0.5 ml with Tyrode’s buffer and analyzed by flow cytometry on a FACScan (Becton Dickinson). 10,000 of green-fluorescent transfectants (positive for intracellular FITC-anti-IgG staining) were analyzed for red (PE) immunofluorescence (surface PE-anti-IgG staining). In experiments with GFP-tagged proteins, fixed cells were surface stained with PE-anti-IgG₁ without permeabilization. For confocal laser immunofluorescence microscopy transfected cells were cultivated on coverslips coated with vitronectin (5 μg/ml) at 4°C overnight and blocked for 1 hour with 5% BSA in PBS at room temperature. Cell monolayers were then fixed and stained as described above for flow cytometric analysis. Immunofluorescence analysis was performed using a Leica confocal laser microscope equipped with a TCS software program.

For immunoprecipitation, subconfluent monolayers of CHO cells were transiently transfected with Ig or GFP fusion constructs. Cells were lysed by adding lysing buffer containing 100 mM Tris, pH 8.0, 150 mM NaCl, 2 mM EDTA and 1% Triton X-100. After 20 minutes at 4°C, lysates were centrifuged at 20,000 g to remove insoluble material. Thereafter, Ig fusion proteins were directly collected on Protein A-6MB-Sepharose beads (Amersham Pharmacia Biotech.). Immunoprecipitates were washed three times in lysis buffer prior to dissociation in SDS sample buffer. Proteins were separated on SDS-polyacrylamide gels and transferred onto nitrocellulose membranes. Immunodetections were performed using horseradish peroxidase-conjugated secondary antibodies (Dianova) and chemiluminescence (Amersham Pharmacia Biotech). In co-immunoprecipitation studies cells were co-transfected with GFP-En-S and Ig-β₃-A or Ig-β₃-A (Y759A). After incubation of the cell lysate Ig fusion proteins were precipitated with Protein A-6MB-Sepharose beads and associated GFP-En-S was detected with westernblotting using anti-GFP-polyclonal antibodies (Dianova).

Internalization of integrins was measured with the help of a method modified from Bretscher (Bretscher, 1992). Stable cell lines expressing similar levels of α_IIbβ₃ and α_IIbβ₃ (Y759A) (Ylänne et al., 1995) were cultured overnight on 35-mm diameter tissue culture dishes, washed twice with Dulbecco’s modified PBS (Life Technologies, Inc.), pre-incubated with biotin-labeled anti-β₃ monoclonal antibody (mAb) (5 μg/ml) for 15 minutes on ice, and incubated for further 0, 2, 5, 15, 30, or 60 minutes at 37°C in Dulbecco’s modified PBS. Cells were then washed twice followed by three successive incubations of 5 minutes each with a reducing solution containing 50 mM 2-mercaptoethanesulfonic acid (Sigma), 10 mM NaCl, 1 mM EDTA, 50 mM Tris, 0.2% bovine serum albumin, pH 8.6. Cells were detached and an aliquot was fixed with 2% formaldehyde for 30 minutes on ice, washed twice with 2% glycine and incubated with PE-streptavidin in the presence of 0.2% Triton X-100 for further 30 minutes. Thereafter, cells were analyzed by flow cytometry and the mean intensity of PE-streptavidin of 10,000 events was used as a measure of integrin endocytosis. A further aliquot of CHO cells incubated with biotin-conjugated anti-β₃ mAb for the indicated time and after reduction with 2-mercaptoethanesulfonic acid was lysed with 200 μl of 200 mM n-octyl-β-_D-glucopyranoside (Sigma), 1 mM phenylmethylsulfonyl fluoride (Sigma) in Dulbecco’s modified PBS at 4°C and centrifuged at 12,000 g, for 5 minutes. After this, an equal volume of 100 mM Tris, 150 mM NaCl, 1 mM CaCl₂, 1% Triton X-100, 0.1% SDS, 0.1% Nonidet P-40, pH 7.4, was added. Thereafter, 20 μg of proteins per lane was loaded on a 8% acrylamide slab gel and after separation under non-reducing conditions the proteins were transferred to Immobilon membrane (Millipore) using 25 mM Tris, 192 mM glycine, 20% methanol, pH 8.3 (Ylänne et al., 1995). Biotin-labeled anti-β₃ mAb was visualized using 2 μg/ml peroxidase-coupled streptavidin (Dianova) and a chemiluminescence immunoblotting kit (Roche). In experiments with echistatin, cells were incubated for 15 minutes on ice with 1 μg/ml FITC-echistatin in the presence of 1 mM CaCl₂ followed by time-dependent incubation as described above. After 3 rounds of washing to remove surface bound echistatin, internalization of the β₃-integrin ligand was evaluated by flow cytometry.

β₃ integrins are implicated in important biological processes, in which the cytoplasmic domains of the β subunits are thought to play roles, including regulation of extracellular ligand binding (inside-out-signaling), signal transduction, and internalization/recycling of the receptors (Shattil, 1995). In an attempt to study the functional significance of the three known β₃ cytoplasmic domain isoforms, we created single-chain chimeric receptors which bear CH2 and CH3 portions of human IgG₁ at the cell surface, the transmembrane domain of CD7 and the full length cytoplasmic domains of β₃-A, -B, and -C, respectively (Fig. 1). An analogous construct, bearing the cytoplasmic domain of and β₁-A or a tail-less Ig construct were employed as further controls. All constructs were based on previously described chimeric receptors (Kolanus et al., 1993), and similar single chain fusion proteins have successfully been used to investigate inside-out signaling events of integrins (Chen et al., 1994; O’Toole et al., 1994; Pfaff et al., 1997). When we expressed these chimeras in various cell systems, it was discovered that the β₃-A chimera had a drastically reduced cell surface expression as compared to other β₃ isoform fusion proteins or control chimeras. CHO cells (Fig. 2A) were transiently transfected with Ig chimeras, and the cells were sequentially stained with fluorescein-or Texas-red coupled anti-Ig antibodies, to detect both the cell surface fraction and the intracellular expression of the fusion proteins. As shown in Fig. 2A, Ig-β₃-B, Ig-β₃-C, and Ig-β₁-A exhibited a diffuse intracellular and steady-state plasma membrane surface staining pattern similar to the staining pattern of the tail-less Ig chimera. In contrast to this, the Ig-β₃-A was localized primarily intracellularly (Fig. 2A). To obtain a more quantitative estimate of the surface expression of the fusion proteins, CHO cells were transiently transfected with the respective constructs, fixed, double-stained for intracellular and surface expression and evaluated by flow cytometry. As shown in Fig. 2B, surface expression of Ig-β₃-A was significantly reduced compared with the alternative splice mutants (Ig-β₃-B, Ig-β₃-C), the Ig-β₁-A chimera, or the tail-less Ig-construct. The differences in surface expression of the tested chimeras were not due to differences in the overall protein levels, since the mean intensity of total cellular immunofluorescence was equal for all transfected constructs (Table 1), and immunoblot analysis confirmed that the steady state expression levels of all tested chimeras were very similar (Fig. 2C). These results specifically implicate the β₃-A cytoplasmic domain in the regulation of cell surface expression of this particular integrin.

Previous studies have shown that the cytoplasmic domain of integrins interact with a variety of cytoskeletal and signal proteins (Hemler, 1998). Conserved regions such as NPXY and NXXY, that are present within most β cytoplasmic domains mediate localization of integrins within the plasma membrane during cell spreading and formation of focal adhesions (Ylänne et al., 1995). We wondered whether mutations or deletion of the carboxy-terminal NITY motif, which is exclusively expressed in the cytoplasmic β₃-A sequence (Fig. 1), would affect surface expression of the respective chimeras. Indeed, we found that a substitution of the carboxy terminus of β₁ with the carboxy-terminal seven residues of the β₃ cytoplasmic domain (756-762) (NITYRGT, Fig. 1), or, alternatively, substituting the NITYRGT motif in β₃-A for the NPKYEGK motif (772-778) of β₁-A (Fig. 1), resulted in a substantially enhanced surface expression of the Ig-β₃-A-NPKY mutant and a concomitant decrease in surface detection of the Ig-β₁-A-NITY chimera (Fig. 3A and B). Moreover, introduction of Pro⁷⁷³ of β₁-A into the β₃-A cytoplasmic domain, resulting in β₃-A-I757P, increased surface expression of the chimeras significantly (Fig. 3A and B). Similar results with enhanced cell surface expression were obtained with an alanine substitution at Tyr⁷⁵⁹ (β₃-A-Y759A) in β₃-A (Fig. 3A and B). These data indicate that a linear sequence motif of β₃-A, N⁷⁵⁶ITY, is critical for the reduced cell surface exposure of the β₃-A integrin fusion protein. Previous mutational analyses of the β₃ tail had shown that the N⁷⁵⁶ITY motif is required for inside-out-signaling in platelets and CHO cells (Wang and Newman, 1998) and mediates interaction with the cytoplasmic protein β₃-endonexin (Eigenthaler et al., 1997).

Previous studies have shown that β₃-endonexin binds to the cytoplasmic domain of the β₃A subunit (Eigenthaler et al., 1997). This interaction is specific, because it was not observed with the cytoplasmic domains of other integrins. Binding of β₃-endonexin to β₃-A was furthermore shown to be dependent on the N⁷⁵⁶ITY motif, present in β₃-A only (Eigenthaler et al., 1997) (Fig. 1). Two isoforms of β₃-endonexin have been described, which differ in size, and more significantly, in their capacity to bind the cytoplasmic domain of β₃-A. (Eigenthaler et al., 1997) The ‘short’ β₃-endonexin splice variant, β₃-endonexin-S (En-S), consists of 111 amino acids and binds to β₃-A, whereas the longer form of β₃-endonexin (170 amino acids, β₃-endonexin-L (En-L)) fails to interact with the β₃-A tail (Shattil et al., 1995; Eigenthaler et al., 1997). Since the amino termini of both variant polypeptides are identical, we made use of the existence of these two isoforms to determine whether β₃-endonexin is involved in the regulation of cell surface expression of β₃-A. The reasoning behind this was as follows: if β₃-endonexin were involved in internalization or cytoplasmic retention of β₃-A, overexpression of the long form was expected to enhance surface expression of the β₃-A chimera. Owing to their conserved amino termini, En-L and En-S might interact with the same intracellular machinery; however, only En-S is capable of binding to β₃-A. Therefore, En-L was likely to uncouple En-S from this hypothetical apparatus, and was thus expected to behave as a dominant negative mutant. To evaluate the effects of the two isoforms of β₃-endonexin on cell surface expression of β₃-A, cDNAs for both isoforms were isolated from a natural killer cell cDNA library, and fused to the carboxy terminus of the green fluorescent protein (GFP), for convenient subcellular detection. Since GFP/β₃-endonexin is expressed both in the cytoplasm and the nucleus (Kashiwagi et al., 1997) we constructed mutants of both β₃-endonexins isoforms that are constitutively localized in the cytoplasm, i.e. in which the putative nuclear localization signal at position K⁶²RKK had been deleted. As shown in Fig. 4B, nuclear import mutants of both of β₃-endonexin isoforms (En-SΔK⁶²RKK and En-LΔK⁶²RKK) showed substantially stronger cytoplasmic GFP-fluorescence compared to the wild-type β₃-endonexin chimeras. When the wild-type and mutated β₃-endonexin isoforms (Fig. 4A) were cotransfected with Ig-integrin chimeras in CHO cells or HUVEC, we found that surface expression of Ig-β₃-A in cells cotransfected with the long form of β₃-endonexin was significantly enhanced, as compared to cells cotransfected with the short form of β₃-endonexin, or with control constructs (Fig. 4B and C). Surface expression of β₃-A was somewhat more enhanced when cells were cotransfected with the nuclear import mutant of the long form of β₃-endonexin (En-L-ΔK⁶²RKK), whereas the short form mutant (En-S-ΔK⁶²RKK) did not have a significant effect on Ig-β₃-A cell surface expression (Fig. 4B and C). Similarly, employing cotransfection of GFP/β₃-endonexin and the Ig-β₁-A-NITY chimera, we observed that En-L enhances surface expression of this fusion protein significantly (Fig. 4B and C), whereas no substantial effect of both β₃-endonexin isoforms on surface expression of the other Ig-integrin chimeras were observed (Fig. 4B and C). We then evaluated whether En-S directly interacts with β₃-A (Shattil et al., 1995; Eigenthaler et al., 1997) in our experimental set-up. It was found that En-S co-immunoprecipitates with β₃-A but not with β₃-A-Y759A (Fig. 4D). This finding implicates the β₃-endonexin system in the specific regulation of cell surface expression of β₃-A. The fact that both GFP/β₃-endonexins are primarily localized in the nucleus does not appear to play a significant role in this context: although constitutive cytoplasmic localization of GFP-β₃-endonexin-L enhances plasma membrane expression of β₃-

A somewhat better than the wild-type version, both constructs have qualitatively similar effects.

Integrins are known to be internalized via endocytosis, and the cytoplasmic domains play an important role in this process (Bretscher, 1989; Bretscher, 1992; Arnaout, 1990). To evaluate the role of the β₃ cytoplasmic domains in endocytic mechanisms, CHO cells were transiently transfected with Ig-integrin chimeras. After 24 hours, a fluorochrome-labeled anti-Ig monoclonal antibody was added to the living cells for 15 minutes. Thereafter, the antibody was removed, and the cells were fixed by 2% formaldehyde at indicated times. We found that the anti-Ig antibody was rapidly internalized in β₃-A-transfected cells, and became detectable within minutes in a vesicular compartment beneath the plasma membrane (Fig. 5A). Longer incubation time resulted in centralization of anti-Ig-FITC-immunofluorescence (Fig. 5A). Similar results were found with Ig-β₁ constructs containing the N⁷⁵⁶ITY motif ofβ₃-A (data not shown). In contrast, predominant or exclusive cell surface staining was observed with virtual no or minimal intracellular uptake of anti-Ig mAb in cells transiently transfected with the other β₃-isoforms β₃-B or β₃-C, the β₁-A fusion protein or the tail-less Ig chimera (Fig. 5A). These results indicate that although steady-state cell surface expression of β₃-A is strongly reduced (Fig. 2A,B), a portion of the cellular pool of the Ig-β₃-A fusion protein apparently cycles between the plasma membrane and an intracellular, vesicular compartment. In contrast, cells transfected with point mutants of β₃-A, β₃-A-Y759A and β₃-A-I757P, or with other β₃ isoforms that reveal substantial steady-state cell surface expression (Figs 2 and 3), showed substantial less internalization of the anti-Ig antibody (Fig. 5A). Thus, we conclude that the N⁷⁵⁶ITY motif is involved in internalization or endocytosis of the β₃-A integrin.

Next we tested whether β₃-endonexin modulates internalization of β₃-A. To test this, CHO cells were cotransfected with the β₃-endonexin isoform GFP-fusion proteins, or mutants thereof, and an Ig-β₃-A chimera. Uptake of an FITC-conjugated anti-Ig mAb was monitored by immunofluorescence microscopy at indicated time intervals. We found that in cells which were cotransfected with the long-isoform chimeras and Ig-β₃-A, uptake of FITC-anti-Ig mAb was substantially less pronounced as compared to cells cotransfected with En-S constructs (Fig. 5B). These results indicate that internalization of anti-Ig mAb-ligated β₃-A integrin is regulated by the cytoplasmic β₃-A binding protein, β₃-endonexin.

To study whether β₃-endonexin is involved in stimulated internalization of the native β₃ integrin, cell lines expressing wild-type α_IIbβ₃ and α_IIbβ₃ (Y759A) were incubated with FITC-conjugated echistatin or a biotin-labeled anti-β₃ mAb or for the indicated time intervals. In experiments with biotin-labeled anti-β₃ mAb cells were exposed to a membrane-impermeable reducing agent after incubation. This chemical abrogated streptavidin-FITC mediated detection of the antibody, and thus provided a means of measuring internalization of the receptors.

Fig. 6A documents that wild-type and mutant integrins were expressed at similar densities at the surface of CHO cells. Subsequent immunoblot experiments showed that approximately 40% of the biotin-labeled anti-β₃ mAb, which was bound to the cell surface at 4°C, became protected from reduction, when wild-type α_IIbβ₃ expressing cells were employed (Fig. 6B and C). However, only 5-6% of the cold-bound material became protected from reduction in cells expressing α_IIbβ₃ (Y759A) (Fig. 6B and C), indicating that substantial internalization occurred in wild-type, but not in mutant β₃ expressing cells. Similar results were obtained when FITC-echistatin was used to monitor integrin internalization (Fig. 6D). Thus, as described above for the integrin chimera, the cytoplasmic domain of native β₃ integrin and the NITY⁷⁵⁹ sequence are needed for substantial ligand-induced internalization of α_IIbβ₃ in CHO cells.

Analogous experiments were then performed using immunofluorescence techniques. Microscopical analyses showed that substantial amounts of a surface-bound, biotin-labeled anti-β₃ monoclonal antibody were transferred to intracellular sites in wild-type α_IIbβ₃ expressing cells (Fig. 7A, left panel). In contrast, internalization of the biotin-labeled anti-β₃ was significantly reduced in cells expressing α_IIbβ₃ (Y759A) (Fig. 7A, right panel). In the mutant (Fig. 7A, right panel), the tracing antibody never was detectable in large vesicular structures, as was observed in the wild-type β₃ expressing cells. Any surface-bound antibody that became protected at all remained associated with the cell cortex.

To obtain a more quantitative estimate of β₃ integrin-dependent internalization measured by immunofluorescence, cells were evaluated by flow cytometry. As shown in Fig. 7B, uptake of PE-streptavidin fluorescence was significantly reduced in cells expressing α_IIbβ₃ (Y759A), circles), as compared to wild-type α_IIbβ₃ expressing cells (squares). Similar results were obtained with the RGD-containing polypeptide FITC-echistatin (Fig. 7C) with substantially reduced intracellular uptake of FITC-echistatin in cells expressing α_IIbβ₃ (Y759A). Fig. 7D documents that equal amounts of FITC-echistatin binds to α_IIbβ₃ and α_IIbβ₃ (Y759A) and that the washing procedure removes all surface bound FITC-echistatin.

To evaluate whether overexpression of β₃-endonexin also modulates antibody-induced internalization of the native β₃ integrin, CHO cells expressing constitutively α_IIbβ₃ or α_IIbβ₃ (Y759A) were transiently transfected with different GFP/β₃-endonexin fusion constructs and incubated with biotin-labeled anti-β₃ for 15 minutes at 37°C. After reduction with 2-mercaptoethanesulfonic acid, cells were stained with PE-streptavidin and evaluated by flow cytometry. In concordance with the experiments employing β₃-A chimeras, we found that expression of the long form of β₃-endonexin, or of the deletion mutant En-L-ΔK⁶²RKK resulted in a dose-dependent, significant reduction of PE-streptavidin fluorescence-uptake, whereas En-S had no effect (Fig. 8). These results indicate that the cytoplasmic β₃-endonexin is involved in antibody-induced internalization of the native β₃-integrin.

The major findings of this study are as follows: (1) By using single-chain chimeras that bear the cytoplasmic domains of the known β₃ integrins (β₃-A, -B, -C) we found that antibody-induced internalization of β₃ integrin is dependent on the isoform transiently transfected into cells. (2) We show that a strongly reduced steady-state cell surface expression of the β₃-A fusion protein is dependent on the cytoplasmic domain, and identified the importance of the terminal cytoplasmic NITY 756-759 motif in this process. (3) The cytoplasmic, integrin-binding protein β₃-endonexin might be involved in regulation of the β₃-A receptor. Measuring internalization kinetics of FITC-labeled anti-Ig antibodies revealed that the β₃-A chimera cycles constitutively between the cell surface and an intracellular, vesicular compartment, and this specific cycling of β₃-A is substantially reduced by overexpression of β₃-endonexin-L. (4) In concordance with the results obtained with chimeric receptors, ligand-induced internalization of the native β₃ integrin in CHO cells is dependent on the NITY motif, and might be regulated by a β₃-endonexin-dependent mechanism. Three isoforms of β₃ integrins have been described so far, and they differ from each other by their cytoplasmic domain sequences (Kuppevelt et al., 1989; Kumar et al., 1997; Fornaro and Languino, 1997) (Fig. 1). Previous studies have shown that functional differences exist between β₃ integrin isoforms. Coexpression of β₃-B, but not of β₃-A, with an α_IIbα₆ chimera in the CHO line result in PAC-1 binding (a specific antibody ligand for activated α_IIbβ₃) of the respective cells (O’Toole et al., 1995). HEK cells transfected with β₃-C but not with β₃-A failed to adhere to osteopontin, an α_vβ₃ matrix protein (Kumar et al., 1997). Furthermore, β₃-A specifically signals phosphorylation of focal adhesion kinase (FAK) (Akiyama et al., 1994) and β₃-A is localized in focal adhesions (Ginsberg et al., 1992) whereas β₃-B shows a diffuse cellular distribution (LaFlamme et al., 1994).

We have observed that steady state surface expression of a single chain chimera bearing the β₃-A tail is substantially reduced or virtually absent in transiently transfected cells, as compared with the two other isoforms β₃-B and -C. The c-terminal NITY 756-759 motif is not conserved amongst β integrins, including the β₃ splice variants, and terminal deletion mutants of β₃ at amino acid position 757 results in defects in the recruitment of α_IIbβ₃ to preestablished adhesion plaques and in α_IIbβ₃ mediated internalization of fibrinogen-coated particles (Ylänne et al., 1995). Thus, we asked whether the NITY motif regulates cell surface expression of β₃-A. Introduction of point mutations into the NITY motif (I757P, Y759A) and substitution of the terminal NITYRGT sequence for the NPKYEGK motif present in the β₁ integrin, results in substantial enhanced cell steady-sate surface expression levels of β₃-A. This indicates that the NITY motif is crucial for regulation of cell surface expression of the β₃-A isoform. Moreover, substitution of the COOH-terminal seven amino acids of β₁ for the NITYRGT sequence imposes the β₃-A phenotype on the β₁-fusion protein: this suggests that the NITY motif is not only responsible, but sufficient for predominant surface expression of the chimeras bearing this cytoplasmic sequence.

Integrins are known to participate in phagocytosis or in internalization of particles coated with ligands. Distinct motifs within the cytoplasmic domain of integrins have been identified to mediate internalization. A conserved motif (NPXY) is found in many β subunits cytoplasmic domains and is essential for internalization of β₂ integrins. Expression of a β₂ truncation that lacks two NPXF motifs, or an F-A substitution variant of the membrane-proximal NPXF, both result in a loss of internalization of integrins (Raab et al., 1993). On the other hand, NPXY motifs do not mediate internalization of α₅β₁ (Vignoud et al., 1994) or α_IIbβ₃ (LaFlamme et al., 1994). We observed that cells transfected with β₃-A or with the β₁-NITYRGT chimera internalize an FITC-anti Ig mAb rapidly. Although very low amounts of the β₃-A chimera are expressed on the plasma membrane at steady state, our findings indicate that the fusion protein is constitutively cycling between the cell surface and an intracellular compartment, and thus mediate uptake and internalization of a fluorochrome-conjugated anti-Ig antibody.

The cytoplasmic domain of β₃-A had been shown to bind to the cytoplasmic protein β₃-endonexin, and the NITY motif was identified as the core binding site (Eigenthaler et al., 1997). The short isoform of β₃-endonexin was reported to induce the high-affinity state of α_IIbβ₃ (PAC-1 binding) and fibrinogen-dependent aggregation without affecting the levels of surface expression of α_IIbβ₃ (Kawashagi et al., 1997). We found that the long but not the short isoform of β₃-endonexin enhances steady-state surface expression of β₃-A and attenuates β₃-A-mediated internalization of anti-Ig antibodies. Because β₃-endonexin is known to be present both in the cytoplasm and in the nucleus (Kawashagi et al., 1997), we tested whether deletion of the putative nucleus import sequence K⁶²RKK present in β₃-endonexin isoforms would results in an altered cellular distribution, favoring cytoplasmic localization. We found that deletion of the nuclear import sequence K⁶²RKK enhanced the observed functional effects of the long isoform somewhat, but the introduction of this mutation was not necessary to observe β₃-endonexin dependent changes in cycling and surface expression of the β₃-A chimera. It is possible that nuclear expression of both β₃-endonexins, which appears so predominant by immunofluorescence detection, points to as yet unknown functions of these proteins, but this has to be elaborated in the future. Our findings and the results of others (Kashiwagi et al., 1997) are compatible with the notion that cytoplasmic levels of both β₃-endonexins are sufficient for the induction of β₃-A-dependent functions in a direct manner, at least when these constructs are introduced into cells by overexpression.

Therefore, a possible explanation for our findings involves a binding competition of the long form of β₃-endonexin with the short variant for a common, as yet unknown factor inside the cell. Overexpression of the long isoform may thus overrun the β₃-endonexin-S dependent internalization of β₃-A.

These results point to a novel function of the β₃-A/β₃-endonexin interaction, namely internalization and cycling of the respective integrin. Furthermore, our experiments employing native α_IIbβ₃ receptors support this concept: the ligation-induced internalization rate of α_IIbβ₃ Y759A was substantially reduced as compared to α_IIbβ₃, and β₃-endonexin-L interfered only with the ligand-induced internalization of the wild-type receptors. Our results do not support earlier reports that the NITY motif is not involved in internalization of β₃-integrin (Ylänne et al., 1995) and the discrepancies between the two studies are not easily explained. In contrast to the previous study of Ylänne et al. we studied internalization of β₃-integrins induced by ligands (antibody, echistatin) which might substantially be different to the mechanism of internalization of constitutive endocytosis in the absence of spcecific ligands.

It is noted that we did not observe functional effects by overexpression of β₃-endonexin-S in our system. However, this is not unexpected since Kashiwagi et al. had likewise not observed changes in surface expression levels by the introduction of β₃-endonexin-S into CHO cells (Kashiwagi et al., 1997). Overexpression might only affect cellular function, if the respective pathway is not saturated. We therefore suggest that the cytoplasmic expression levels of endogenous β₃-endonexin-S are sufficient to promote maximal internalization rates.

Others had implicated β₃-endonexin in inside-out signal transduction, and specifically in the induction of a high affinity epitope (recognized by antibody PAC-1) of α_IIbβ₃ (Kashiwagi et al., 1997). We do at present not know whether the affinity modulation and internalization functions of β₃-endonexin are connected. However, it has recently become evident that vesicle transport mechanisms and actin cytoskeleton remodeling events at the plasma membrane, including formation of focal adhesions, may be coupled. Norman et al. have shown that ARF1, a small GTPase involved in intracellular vesicle budding, regulates deposition of the actin linker protein paxillin to focal adhesions (Norman et al., 1998). Furthermore, ARF6, another family member, appears to regulate both endocytosis and membrane ruffling (D’Souza-Schorey et al., 1997). It is therefore possible, but remains to be proven, that β₃-endonexin-dependent internalization of β₃-A might contribute to the activation of this receptor and/or to its accumulation into focal adhesions.

The regulation of ligand-induced internalization of β₃ integrin via the NITY motif and by a β₃-endonexin-mediated mechanism might have important pathophysiological consequences. The β₃ integrin, α_IIbβ₃ or glycoprotein IIb-IIIa, is known to mediate endocytosis of fibrinogen, or of antibodies that develop in autoimmune diseases, or of GPIIb-IIIa antagonists into the α-granules of platelets (Santoso et al., 1986; Isenberg et al., 1990, Nurden et al., 1999). Moreover, α_vβ₃ mediates internalization of bacterial and viral pathogens into endothelial cells (Wickham et al., 1993; Ozeri et al., 1998). Furthermore, several integrins are internalized via an endocytic pathway and are rapidly recycled back to the cell surface (Bretscher, 1989; Bretscher, 1992). Different integrins seem to have different abilities to take part in the cycling process and 10-20% of cell surface labeled α_IIbβ₃ is rapidly internalized within minutes (Ylänne et al., 1995). Integrin endocytosis and recycling of receptors to the leading edge is involved in the forward-directed movement of migrating cells (Kucik et al., 1989). Thus, interfering with mechanisms involved in β₃ integrin mediated endocytosis might therefore be of therapeutic interest in treatment of vascular or metastatic diseases.

The study was supported in part by grants from the Deutsche Forschungsgemeinschaft (grants Ga 381/4-1, Le554/2-1 to M.G. and Ko-1034, SFBs 455 and 464 to W.K.) and from the Wilhelm Sander-Stiftung to M.G. and to W.K., and Klinische Fördermittel der TU München (M.G.). The authors appreciate the excellent technical assistance of Kirsten Langenbrink and Antje Wallmuth.