Two Synergistic Activation Mechanisms of α2β1 Integrin-mediated Collagen Binding*

Activation of protein kinase C by 12-O-tetradecanoylphorbol-13-acetate (TPA) induces ligand-independent aggregation of a cell surface collagen receptor, α2β1 integrin. Concomitantly, TPA increases the avidity of α2β1 for collagen and the number of conformationally activated α2β1 integrins. The structural change was shown using a monoclonal antibody 12F1 that recognizes the “open” (active) conformation of the inserted domain in the α2 subunit (α2I). Amino acid residue Glu-336 in α2 subunit is proposed to mediate the interaction between α2I domain and β1 subunit. Glu-336 seems to regulate a switch between open and “closed” conformations, since the mutation α2E336A inhibited the TPA-related increase in the number of 12F1 positive integrins. E336A also reduced cell adhesion to collagen. However, E336A did not prevent the TPA-related increase in adhesion to collagen or α2β1 aggregation. Thus, α2β1 integrin avidity is regulated by two synergistic mechanisms, first an α2E336-dependent switch to the open α2I conformation, and second an α2E336-independent mechanism temporally associated with receptor aggregation.

Integrin-type cell adhesion receptors undergo functional regulation via different conformational states (1). In an inactive integrin ␣/␤ heterodimer, the ligand binding headpiece is bent toward the cell surface (2), while an active integrin stands tall but may still have multiple activation states. Nine ␣ subunits out of 24 human integrin heterodimers have an inserted ␣I (or ␣A) ligand binding domain, which can exist in multiple configurations of differing affinity. The conformation of ␣I is regulated by a spring-like mechanism formed by a conserved glutamate residue in the ␣7 helix acting as an intrinsic ligand of the I-like domain of the ␤ subunit (␤I; Refs. 3 and 4).
Avidity, the overall strength of integrin mediated interactions between cells and extracellular ligands, is multideterminant, involving the affinity of individual receptors and varia-tions in the spatial arrangement of the integrins. The total number of receptors participating in the interaction, or valency, is expected to be larger if integrins form clusters. The relationship between conformational regulation and cluster formation has been thoroughly studied in ␣L␤2 integrin, a leukocyte receptor for intercellular adhesion molecules (5)(6)(7). Recent studies suggest that ␣L␤2 clustering takes place only after binding a multivalent ligand and that clustering strengthens adhesion after ligand binding (8). The requirement of a ligand for receptor clustering has also been reported for ␣V␤3 (9) and ␣IIb␤3 integrins (10).
Four collagen receptors, ␣1␤1, ␣2␤1, ␣10␤1, and ␣11␤1, form a subgroup of ␣I domain containing integrins. Integrin ␣2␤1 is expressed on platelets, epithelial cells, and many mesenchymal cell types (11). The ␣2I domain, like other integrin I domains, can exist in both an "open" and a "closed" conformation (12), as well as a less frequent, intermediate form (13). Thus, ␣2␤1 can also be found in more than one conformation (14), likely regulated similarly to other ␣I domain integrins. The unique functions of collagen receptor integrins may also be governed by the regulation of the receptor distribution and local density. For example, ␣2␤1 and ␣11␤1 participate in the formation and rearrangement of collagen fibrils (15,16), and echovirus-1-dependent clustering leads to rapid relocation and internalization of ␣2␤1. In this process, ␣2␤1 behaves differently from other integrin family members, such as ␣V integrins (17).
The phorbol ester TPA 2 can increase the avidity of ␣2␤1 binding to collagen (18). We report here that while TPA induces rapid (5-15 min) clustering of cell surface ␣2␤1 in the absence of ligand and increases ␣2␤1 avidity for collagen, it also activates the integrin via a conformational mechanism involving ␣2E336 and intermolecular cross-talk between the ␣2I and ␤1I domains. Aggregation of ␣2␤1 is independent of E336 and temporally associated with increased avidity of adhesion. We conclude that the avidity of ␣2␤1 for collagen is regulated by two synergistic mechanisms.
Cell Spreading Assay-Costar 96-well plates were coated with collagen I (col I, 5 g/cm 2 ; Vitrogen) or 0.1% BSA. Cells were allowed to spread in serum-free medium with 50 M cycloheximide (Sigma) for 2 h at 37°C. Fixed cells were analyzed by phase-contrast microscopy as described earlier (26).
Receptor Staining-100 nM TPA (Calbiochem) was added to subconfluent cells in serum free Dulbecco's modified Eagle's medium. Released cells were fixed (4% formalin) and suspended in phosphate-buffered saline with 3% BSA and 0.6% rat IgG at time points. Cells were stained with mAb (12F1 or 16B4, Serotec) and anti-mouse Alexa-488-IgG (Molecular Probes) and assayed by flow cytometry.
Integrin Clustering and Internalization-1) Cells cultured on chambered coverslips in CO 2 -independent medium (Sigma) were stained with Alexa-555-conjugated 16B4 (labeling kit, Molecular Probes) and treated with 1 M TPA, anti-mouse IgG (positive control), or left untreated (negative control). Fourdimensional (three-dimensional ϩ time) data sets were obtained at 37°C at 3-min intervals for a total time of 24 min. 2) Cells were starved overnight in medium containing 0.1% fetal bovine serum and treated either with 10 M BIS or 1 M TPA for 40 min at 37°C, or cells were pretreated with 10 M BIS for 15 min followed by addition of 1 M TPA for 30 min. Control cells were untreated. Fixed cells (4% paraformaldehyde) were permeabilized (0.2% Triton X-100) and stained with mAb ␣2 integrin (2AE11, a kind gift from Dr. Fedor Berditchevski, University of Birmingham, Birmingham, UK) and anti-mouse Alexa-488-IgG. In each sample, 200 cells were analyzed. 3) Cells on coverslips were untreated (negative control), treated with 1 M TPA for 15 min or 16B4 and anti-mouse Alexa-488-IgG (positive control). Fixed cells were permeabilized, stained with Alexa-555-conjugated 16B4, and embedded in Mowiol. Data were obtained by laser scanning confocal microscopy (LSM510, Carl Zeiss), and all images and volume and side view renderings were produced with BioImageXD software.
On-line Supplemental Material-Four-dimensional microscopy shows TPA-induced clustering and internalization of ␣2␤1 integrin in Saos-␣2 cells. The movie (see supplemental material) was created using BioImageXD.

RESULTS
TPA Induces a Ligand-independent but PKC␣-dependent Aggregation and Internalization of ␣2␤1 Integrins-Virus and antibody-mediated ␣2␤1 clustering are known to activate PKC␣ (17). Here, the effect of the phorbol ester TPA, a PKC activator, on integrin clustering and internalization was investigated using Saos osteosarcoma cells transfected to express ␣2␤1 integrin (22). Cells were stimulated with TPA, and the experiments were done in the absence of collagen ligands. In four dimensional confocal microscopy studies, both fast macroclustering and internalization of ␣2␤1 integrin induced by TPA were observed ( Fig. 1A and supplemental movie). The effect of TPA was comparable with, but not as extensive as, that obtained by secondary antibody-induced clustering, which was used as a positive control (Fig. 1A). As shown in Fig. 1B, the average time for the appearance of clear clusters after TPA addition was about 10 min, and the average time for their internalization was about 14 min. For secondary antibody-induced clustering, these times were about 3.5 and 7 min, respectively. No clustering or internalization was observed in negative control cells that were not stimulated in any way. Furthermore, BIS, a chemical inhibitor of PKCs, prevented the formation of TPAinduced integrin aggregates and their internalization (Fig. 1C). Thus, we propose bidirectional dialog between PKC␣ activation and integrin aggregation: receptor clustering activates PKC␣ (17) and PKC␣ further promotes the aggregation.
The binding of CHO-␣2WT cells to collagen I coated beads was detected by flow cytometry at successive time points ( Fig.  2A). Cell adhesion to collagen I appeared to increase nearly linearly for up to 50 min before equilibrating. However, when CHO-␣2WT cells were treated with TPA, a significantly faster adhesion, which reached its equilibrium in 40 min, was obvious. At every time point, TPA caused a 50% increase in cell adhesion compared with untreated CHO-␣2WT cells ( Fig. 2A). Increased adhesion was already clear at the 10 min time point, indicating the rapidity of TPA-induced activation (Fig. 2B).
Previously, it has been shown that the cytoplasmic tail of the ␣2 subunit is required for outside-in signaling mediated by ␣2␤1 integrin (22,33). In these studies, a chimeric ␣2 integrin subunit bearing an ␣1 cytoplasmic tail (␣2/␣1) appeared to be unable to mediate heterodimer-specific signals. To investigate the role of the ␣2 cytoplasmic tail in TPA-activated adhesion, CHO-␣2/␣1 cells were tested in time series. Temporal profiles of both untreated and TPA-treated CHO-␣2/␣1 cell adhesion to collagen I resembled the adhesion of CHO-␣2WT cells. However, the TPA mediated increase in cell adhesion was substantially higher in CHO-␣2/␣1 cells than in CHO-␣2WT cells (Fig. 2C). Thus, the swap mutation did not affect the ability of the receptor to become activated. CHO cells bearing ␣10␤1 integrin were analyzed similarly (CHO-␣10WT; Fig. 2D). In this case, the adhesion in both TPA treated and untreated cells reached saturation at earlier time points than in CHO-␣2WT cells (Fig. 2D). Furthermore at the 10-min time point, TPA treatment increased CHO-␣10WT cell adhesion to collagen I more than seen in CHO-␣2WT or CHO-␣2/␣1 cells indicating an earlier overall response. Altogether, the results show that the intracellular domains of integrin ␣ subunits may not directly participate in regulation by TPA and that the effect is not limited to the ␣2␤1 heterodimer.
To verify that the observed increases in adhesion were mediated by ␣2␤1 integrin, a series of control experiments were performed using CHO-␣2WT, CHO-␣2/␣1, and CHO-pAW cells. All cell types showed an invariably low level of interaction with negative control beads, which were either left uncoated or coated with BSA (Fig. 2E). Additionally, vectortransfected control cells, CHO-pAW cells, did not interact with collagen I-coated beads (Fig. 2E). However, CHO-␣2WT and CHO-␣2/␣1 cells adhered strongly to collagen I-coated beads, revealing a specific interaction. This adhesion, both basal and TPA-mediated, was significantly decreased by pretreating cells with the ␣2 integrin-specific function blocking mAb P1H5, thereby verifying that adhesion occurred in all conditions via the ␣2␤1 integrin (Fig. 2E).
To confirm that the TPA-mediated increase in ␣2␤1 integrin avidity to collagen I uses a PKC-related signaling pathway, the PKC inhibitor, BIS, was tested in the presence and absence of TPA. When CHO-␣2WT and CHO-␣2/␣1 cells were pretreated with BIS, the TPA-mediated cell adhesion to collagen decreased about 35%. However, BIS did not significantly alter the basal adhesion of CHO-␣2WT cells to collagen I. In comparison, the PKC inhibitor slightly reduced basal adhesion in CHO-␣2/␣1 cells (Fig. 2F).
Integrin ␤1 Subunit Is Essential for TPA-related Increase in ␣2-mediated Cell Adhesion to Collagen I-Talin is a cytoskeletal protein and a well know substrate for PKC (34). Talin binding to the cytosolic part of the integrin ␤ subunit is typically required for the initiation of integrin activation (35). To analyze the importance of the ␤1 subunit in TPA-mediated integrin avidity regulation, we used GD25 cells expressing similar levels of either wild type ␣2␤1 or ␣2␤1 with Y783/795F mutations in the cytoplasmic tail of the ␤1 subunit (25). Substituting these tyrosine residues in conserved NPXY motifs of the ␤ tail has previously shown to be sufficient to inhibit talin binding (35,

TPA Increases the Number of
Conformationally Activated Integrins on the Cell Surface-Recent papers have identified monoclonal antibodies that recognize epitopes corresponding to specific activation states of the ␣2 integrin subunit (14,37). Therefore, the possibility that treatment with TPA causes ␣2 integrin conformational activation in CHO-␣2WT cells was tested. Cells were stimulated at different time points with TPA, and relative amounts of "total" and open ␣2 integrin subunit were determined by flow cytometry. Monoclonal ␣2 integrin-specific antibody 16B4, whose epitope is located outside the ligand binding region of the ␣2 subunit, was used for determination of total ␣2. Open ␣2 conformation was determined using mAb 12F1, which has recently been shown to favor an activated conformation of ␣2 subunit on platelets (14).
The expression of integrin ␣2 on the cell surface decreased in the presence of TPA between the first two measured time points (0 and 10 min) in all three independent experiments as shown in Fig. 4. This was not surprising given that some internalization of ␣2␤1 integrin was seen by confocal microscopy in TPA treated Saos cells (Fig. 1). Notwithstanding the initial decrease, the amount of ␣2 integrin on the cell surface remained stable after 10 min in all three experiments. In contrast, a significant increase in the number of 12F1 epitopes corresponding to the active conformation of ␣2 integrin was apparent in each experiment beginning at between 20 and 40 min after stimulation with TPA. The increase was transient in each case, lasting from 10 to 20 min before declining to near baseline levels (Fig. 4). Interestingly, the transient conformational activation of the ␣2 integrin seemed to occur later than the increase in avidity (within 10 min).
CHO-␣2E336A Cells Retain Adhesion to Collagen, but the Mutation Inhibits TPA-related Conformational Change-In leukocyte ␤2 integrins, e.g. in ␣M␤2 and ␣L␤2 integrins, ligand binding is shown first to cause a change in the conformation of the ␣I domain, including a downward movement of the ␣7 helix. This is followed by a similar change in the ␤I domain, leading to separation of the leg domains of the ␣ and ␤ subunits (1, 3, 4, 13, 38). A conserved glutamate residue on the ␣7 helix of the ␣I domain was suggested to act as an intrinsic ligand for the ␤1I domain. The same structural link between the ␣I and ␤I domains seems to be important in integrin inside-out signaling and regulation of affinity.
The corresponding glutamate in the ␣2I domain, the conserved Glu-336, may interact in a similar manner with the ␤1I domain and participate in the switch between open and closed ␣2I domain conformations (Fig. 5A). We wanted, however, also to test the hypothesis that the conformational regulation of ␣2␤1 takes place in a different manner than the regulation of ␣L␤2 and ␣M␤2. Thus, we mutated amino acid residues Glu-336, Glu-309, and Tyr-410. Glu-309 is located close to the C terminus of the ␣7 or terminal helix of the ␣2I domain and potentially could interact with the ␤1I domain metal binding sites (Fig. 5A). This residue was mutated to alanine to remove the negative charge of the side group. The mutation Y410A was constructed to probe the importance of the interaction between the ␣2 ␤-propeller and the ␤1I domain. A large surface area is buried between these two subunits, and there appears to be a specific interaction at the center of the ␣2 ␤-propeller (Fig. 5A). From our molecular models of this specific interaction, based on the ␣V␤3 crystal structures (2), Tyr-410 could be one of the three aromatic residues involved in a cation-interaction with ␤1K269. Disruption of this interaction might lead to defects in signaling and/or conformational activation.
CHO cells transfected to express integrin ␣2 subunits carrying either E309A, Y410A, or E336A mutations (CHO-␣2E309A, CHO-␣2Y410A, and CHO-␣2E336A) were tested by flow cytometry for their ability to adhere to collagen I-coated polystyrene beads. It appeared that the base level adhesion of  CHO-␣2E336A to collagen I was clearly affected compared with CHO-␣2WT cells (Fig. 5B). Mutation Y410A also decreased ␣2 integrin binding to collagen I, whereas mutation E309A had no effect (Fig. 5B). Reduced CHO-␣2E336A cell binding to collagen I is in agreement with previous studies indicating that ␣L and ␣M integrins lose their ability to bind their ligands after the mutation of the residue corresponding to E336 in ␣2 (E310A in ␣L and E320A in ␣M; Refs. 3 and 39). Notably, CHO-␣2E336A cells retained a low level of avidity to collagen and were able to spread on collagen I-coated surfaces in 2 h (Fig.  5C). However, CHO-␣2WT cells, as well as CHO-␣2Y410A and CHO-␣2E309A cells, spread significantly more quickly on collagen (Fig. 5C).
The TPA-induced conformational activation of the ␣2 integrin with the E336A mutation was studied in a flow cytometry based assay by using CHO-␣2E336A cells and antibodies recognizing either total or open ␣2 integrins on the cell surface. In CHO-␣2E336A cells, the level of total ␣2 expression was stable over the course of the experiment (Fig. 5D). Interestingly, in these cells, as opposed to cells bearing the wild type ␣2 subunit ( Fig. 2A), the TPA treatment did not induce an increase in the amount of the active ␣2 integrin conformation recognized by mAb 12F1 (Fig. 5D).
The clustering was more difficult to detect in CHO cells than in Saos cells because of the dot-like appearance of the integrin stain. However, clustering is clearly visible in optical sections from the tops of the cells in which it looks similar to clusters induced by antibody cross-linking, while it is absent in negative control cells (Fig. 6B). These results indicate that, in addition to conformational regulation, TPA also induces other mechanisms regulating integrin activity. We propose that ligand-independent ␣2␤1 integrin clustering is one of the mechanisms, which increases avidity.

DISCUSSION
The interaction of ␣2␤1 integrin and collagenous matrix has been implicated in a number of biological and pathological processes such as thrombosis, inflammation, angiogenesis, and wound healing. The precise regulation of integrin ␣2␤1 activation is very important for example for platelet function (40). In general, integrin activation is an essential mechanism for the formation of matrix adhesion sites. Thus, understanding the mechanism of ␣2␤1 activation thoroughly will provide insight into a cellular event of broad biological significance. Here, we report that ␣2␤1 integrin activation is regulated by two synergistic mechanisms: receptor clustering and conformational activation.
The structural basis of outside-in signaling in leukocyte ␤2 integrins has been investigated in depth (1,38). The integrins are considered to have at least three different conformations with distinct abilities to bind their ligands. In inactive integrins, the ligand binding head part is bent toward the plasma membrane in a manner that makes the binding of large ligands difficult. However, inside-out signaling mechanisms such as talin binding to the cytoplasmic domains of the ␤ subunits (35) can induce a switchblade-like movement in the integrin raising the integrin head part up. The avidity of standing integrin for its ligand can still be increased by further conformational modifications. In ␣I domain integrins the high avidity conformation has been recognized as the open conformation of the ␣I domain. Here, we have studied a model in which ␣2␤1 integrins were activated by TPA. It was clear that the TPA effect on integrin avidity could, at least partially, be explained by the ability of TPA to switch ␣2I domains from the closed to the open conformation. This conformational change was studied using 12F1 anti-␣2 mAb binding to the cell surface. In a recent study, the monoclonal antibody 12F1 was shown to favor the activated, open conformation of the ␣2 subunit on platelets (14). Monoclonal antibody 16B4, which recognizes an epitope located outside the ligand binding region of the ␣2 subunit, was used to determinate the total ␣2 on the cell surface. Previous studies on ␣L␤2 and ␣M␤2 integrins have proposed that a glutamate residue in the ␣7 helix of the ␣I domain (E320 in ␣L and E310 in ␣M) can act as an intrinsic ligand for the ␤ subunit and mediate conformational regulation between the ␣I domain and the ␤ subunit in two directions (3,4). Our molecular model of ␣2␤1 integrin supports the idea that Glu-336 in the ␣2 subunit could have a similar role. This hypothesis was tested using CHO cells transfected with the ␣2 subunit harboring the E336A mutation. E336A prevented TPA-dependent increases in the number of 12F1 epitopes on the cell surface indicating the essential role of Glu-336 in conformational regulation.
Despite the fact that integrin conformation was shown to participate in the regulation of avidity, it seemed clear that additional mechanisms are likely to exist. This conclusion was based on two observations: first, the conformational change in ␣2␤1 integrins was observed only after 10 min, while the avidity started to increase earlier. Typically, a transient increase in the number of 12F1 epitopes occurred 20 -40 min after stimulation with TPA. Second, E336A mutation could not block the effect of TPA. In contrast, the E310A mutation in ␣L, which is an equivalent to E336A in ␣2, prevents the effect of TPA on avidity (8). In our experiments, base-line cell adhesion was lower in CHO-␣2E336A than in CHO-␣2WT cells, which is in agreement with previous papers that have described the loss of ligand binding by ␣L and ␣M integrins after mutation in the corresponding residue (3).
The clustering of ␣L␤2 is seen after ligand binding, and it is an important mechanism in the strengthening of adhesion (8). Ligand binding to ␣2␤1 integrin is known to induce clustering and concomitant activation of specific signaling pathways (17). Thus, ligand-dependent receptor clustering is a well known phenomenon. However, the role of ligand-independent integrin clustering in the regulation of avidity has been a controversial issue (41,42). Here, using confocal microscopy, TPA was FIGURE 6. Mutation ␣2E336A prevents neither the TPA-related increase in avidity for collagen nor ␣2␤1 aggregation. A, cells were incubated for 50 min with either collagen I-or BSA-coated beads Ϯ 100 nM TPA and assayed by flow cytometry (mean Ϯ S.D., n ϭ 3). B, cells were treated for 15 min with either TPA or secondary antibodies (positive control), fixed, and imaged by confocal microscopy. Optical sections from the tops of the cells (left) and differential interference contrast images (right) are shown. No clustering was observed in untreated cells. Dotted lines indicate that cells were in separate visual fields.
documented to induce the macroclustering of ␣2␤1 integrins in ␣2␤1 overexpressing Saos-␣2␤1 osteosarcoma cells and CHO-␣2␤1 cells. The role of integrin aggregation was supported by two facts: clustering took place very quickly, in less than 10 min after TPA treatment similarly to the increase in avidity. Additionally, the E336A mutation could not prevent integrin clustering, and it could not prevent the increase in avidity.
The effects of TPA on ␣2␤1 avidity and clustering are likely mediated by activation of PKC, since a chemical inhibitor, bisindolylmaleimide, could block the effect of TPA. Interestingly, we have previously shown that ␣2␤1 clustering can activate the ␣ isoform of PKC, and specific inhibitors of PKC␣ prevent clustering-dependent integrin internalization (17). Thus, our results propose that ␣2␤1 integrin clustering and PKC may regulate each other in both directions. Here, the structural requirements of TPA-mediated ␣2␤1 integrin activations were also studied. Our earlier studies have shown that an integrin chimera having the ␣2 extracellular domain, and the cytoplasmic domain of ␣1 the subunit (␣2/␣1) cannot mediate heterodimer-specific outside-in signals (22). In this study, however, this swap mutation neither destroyed the ability of the receptor to become activated by TPA nor affected the time course of activation. In both ␣2and mutant ␣2/␣1-mediated collagen I binding, the TPA-dependent increase was already clear at the 10-min time point. These results suggest that the intracellular domains of integrin ␣ subunits may not directly participate in regulation by TPA. In fact, collagen binding activity by CHO cells transfected to express ␣10␤1 integrin was also activated by TPA. One previous study suggests that PKC might interact directly with ␤1 subunit (43) or through talin-containing integrin activation complex (44). For example ␣IIb␤3 integrin has previously shown to be activated by TPA only when PKC␣ and talin are present (44). In agreement, cells expressing wild type ␣2 and talinbinding defective ␤1 subunit (GD25-␣2␤1mut) were not stimulated by TPA. The results suggest that instead of the ␣ subunit, the ␤ subunit is required for TPA-mediated activation.
We conclude that ␣2␤1 integrin forms clusters upon TPA induced activation and that receptor aggregates may contribute to the increase in the avidity of adhesion. Similarly, a recent paper has suggested that ligand free clusters of ␣L␤2 integrin may occur (45), while in other experimental models, receptor clustering has only followed ligand binding (8,9). In addition, some earlier studies have suggested that clustering would regulate integrin avidity even without conformational activation of individual receptors (46,47). This report and our previous results suggest that the same signaling proteins that are activated by ligand mediated clustering may also initiate ligandfree clustering. Furthermore, structural activation seems to be required for the high avidity configuration of the ␣2␤1 integrin. Thus, we propose a model in which ␣2␤1 clustering and conformational activation take place in a synergistic manner.