Novel α2β1 Integrin Inhibitors Reveal That Integrin Binding to Collagen under Shear Stress Conditions Does Not Require Receptor Preactivation*

Background: Integrin α2β1 is a platelet collagen receptor. Results: Novel sulfonamide derivatives are conformation-selective inhibitors of α2β1, especially when tested under shear stress conditions. Only inhibitors that block non-activated integrins inhibit platelet binding to collagen. Conclusion: Non-activated α2β1 integrin plays an important role in platelet binding to collagen. Significance: We propose an alternative model for α2β1 activation during thrombosis. The interaction between α2β1 integrin (GPIa/IIa, VLA-2) and vascular collagen is one of the initiating events in thrombus formation. Here, we describe two structurally similar sulfonamide derivatives, BTT-3033 and BTT-3034, and show that, under static conditions, they have an almost identical effect on α2-expressing CHO cell adhesion to collagen I, but only BTT-3033 blocks platelet attachment under flow (90 dynes/cm2). Differential scanning fluorimetry showed that both molecules bind to the α2I domain of the recombinant α2 subunit. To further study integrin binding mechanism(s) of the two sulfonamides, we created an α2 Y285F mutant containing a substitution near the metal ion-dependent adhesion site motif in the α2I domain. The action of BTT-3033, unlike that of BTT-3034, was dependent on Tyr-285. In static conditions BTT-3034, but not BTT-3033, inhibited collagen binding by an α2 variant carrying a conformationally activating E318W mutation. Conversely, in under flow conditions (90 dynes/cm2) BTT-3033, but not BTT-3034, inhibited collagen binding by an α2 variant expressing E336A loss-of-function mutation. Thus, the binding sites for BTT-3033 and BTT-3034 are differentially available in distinct integrin conformations. Therefore, these sulfonamides can be used to study the biological role of different functional stages of α2β1. Furthermore, only the inhibitor that recognized the non-activated conformation of α2β1 integrin under shear stress conditions effectively blocked platelet adhesion, suggesting that the initial interaction between integrin and collagen takes place prior to receptor activation.

domains of integrin ␤ subunits (11). As a result of this insideout regulation, the integrin "stands up" taking on an extended conformation that is capable of binding large ligands. Alternatively, some experimental evidence has led to speculation that ligand binding to a bent integrin may also induce extension of the receptor (4).
In leukocytes, the integrin ␣I conformation is linked to the conformation of the ␤-subunit by a specific glutamate residue that can act as an intrinsic ligand for the MIDAS in the ␤I domain and mediate structural changes related to both insideout and outside-in signaling (12)(13)(14). Structural studies of ␣X␤2 integrin have indicated that the connection between the ␣I domain and the rest of the heterodimer is flexible and may allow the ␣I domain to move relatively freely (15).
The ␤1-integrins are thought to function in a manner similar to ␤2and ␤3-integrins. Still, direct evidence for most details of this process is lacking. We have previously shown that the Glu-336 residue may act as a link between ␣2I and ␤1I domains (16) in a manner similar to that of Glu-320 in ␣M (12,13) and Glu-310 in ␣L (14) and regulate the activity of the receptor. However, the function of ␣2␤1 does not seem to be entirely dependent on conformational activation because ␣2␤1 can bind even a large ligand (human echovirus-1) in a non-activated conformation (17).
To develop novel ␣2␤1 integrin inhibitors, we used a ligandbased drug discovery approach to combine properties of the integrin-binding RKKH peptide (18,19) with sulfonamide inhibitors (20,21) of ␣2␤1 integrin. The urea moiety was found to have properties that appropriately mimicked the arginine of the RKKH peptide. Taking advantage of several urea-substituted integrin ligands that have been previously discovered (22,23), we generated a novel class of urea-substituted sulfonamide derivatives. Here, we describe two molecules that have different ␣2I domain-binding mechanisms and distinct functional properties. On the basis of our experiments utilizing ␣2 loss-offunction and gain-of-function mutations, we propose that the two sulfonamides are selective for different ␣2␤1 integrin conformations, especially under shear stress conditions. Accordingly, they can be used to test the biological role of distinct functional stages of ␣2␤1. Our data support the idea that ␤1-integrins also have several different activation stages, although there are differences compared with those of ␤2-integrins. Collectively, our results suggest a re-evaluation of the role of nonactivated ␣2␤1 under shear stress conditions.
For continuous measurement of cell adhesion, CHO-␣2wt, CHO-␣2E318W, and CHO-␣2E336A cells (5 ϫ 10 4 cells/well) were plated on rat tail collagen I-coated (Sigma) E-plates (Roche Applied Science) and monitored for 3 h with xCelligence (Roche Applied Science), which measures changes in impedance at the bottom of a cell culture well caused by cell attachment and spreading.
Analysis of Integrin Expression Levels-Integrin ␣2 expression levels on the CHO cell surface were determined by flow cytometry. Cells (3 ϫ 10 5 cells/ml) were incubated with mouse anti-human CD49b primary antibody (7.5 g/ml; BD Biosciences) for 1 h at 4°C, followed by incubation with fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mouse IgG secondary antibody (115 g/ml; DAKO) for 30 min at 4°C. Cells were washed and analyzed with FACScan (BD Biosciences). Secondary antibody only or ␣2-negative cells were used as negative controls. Cytometry data were analyzed with Flowing Software 2 (Turku, Finland).
Protein thermal stability was determined using a Bio-Rad C1000 Thermal cycler and CFx96 real-time system. Protein unfolding was monitored over a range of 20°C-95°C at a rate of 0.5°C/30 s by measuring the fluorescence of the environmentsensitive fluorescent dye, SYPRO Orange (5ϫ; Invitrogen). The final protein concentration in the samples was 7 M, and the concentrations of BTT-3033 and BTT-3034 were 42 and 49 M, respectively. The total volume of samples was 25 l.
Statistics-Continuous data are reported as means with S.E. using GraphPad Prism (GraphPad Software, San Diego). To evaluate the differences between groups, Wilcoxon Rank-Sum test or Student's t test was utilized.

RESULTS
Two Novel Sulfonamide Derivatives Selectively Block Collagen Binding by ␣2␤1 Integrin-To understand the role of different substituents in the sulfonamide, we developed novel structural analogs based on previously identified ␣2␤1 integrin modulator molecules (20,21,24,26). First, the keto group in the benzophenone moiety (20,21,26) at the amide site was replaced with urea to test the effect of a slightly bulkier substituent at that site. Second, the bi-phenyl moiety of BTT-3016 (20) was replaced with analogs that have a similar shape. Third, all developed analogs were tested with and without amide methylation. A cell-based assay utilizing CHO-␣2wt cells was used to test two potential ␣2␤1 integrin-binding molecules, BTT-3033 and BTT-3034 ( Fig. 1, A and B). BTT-3033 and BTT-3034 inhibited cell adhesion to rat tail collagen I with EC 50 values of 130 and 160 nM, respectively, and corresponding E max values of 97 and 86% (Fig. 1, C and D). The relative selectivity for ␣2␤1 and ␣1␤1 integrins was determined in an assay using CHO cells stably overexpressing wild-type human ␣1 (CHO-␣1wt) and measuring attachment to collagen IV ( Fig. 1, C and D). A comparison of EC 50 values in CHO-␣1wt/collagen IV assays and CHO-␣2wt/collagen I assays showed that BTT-3033 selectivity for ␣2␤1 integrin (8-fold) was greater than that of BTT-3034 (2-fold). Neither BTT-3033 nor BTT-3034 was cytotoxic at concentrations up to 200 M (data not shown). The chemical characteristics of the synthesized compounds, BTT-3033 and BTT-3034, are summarized in supplemental Table 1. The potential inhibitory effects of BTT-3033 and BTT-3034 on the function of various integrins were tested using MG-63 and PC3 cells in adhesion assays utilizing different matrices. Neither BTT-3033 nor BTT-3034 (at EC 50 concentrations) inhibited the adhesion of MG-63 cells to vitronectin, 120-kDa fibronectin or 40-kDa fibronectin, assays that measured ␣V, ␣5␤1, and ␣4␤1 integrin function, respectively (supplemental Table 2). Furthermore, BTT-3033 did not inhibit adhesion of PC3 cells to laminin-332 (previously termed laminin-5; LN-332), indicating that this sulfonamide does not interact with ␣3␤1 integrin at its EC 50 concentration. BTT-3034, however, did slightly inhibit PC3 cell adhesion to LN-332 (20% inhibition) at its EC 50 concentration (supplemental Table 2).
The Sulfonamide Derivative BTT-3033, but Not BTT-3034, Inhibits Platelet Binding to Collagen I under Flow-The effects of BTT-3033 and BTT-3034 on platelet aggregation in capillaries coated with collagen I was studied using a Cellix platform (Cellix, Ltd.). This technology allows the function of platelets to be tested under near-physiological conditions. BTT-3033 (10 M) inhibited human platelet adhesion to collagen I-coated capillaries under flow ( Fig. 2A). (Note that the free concentration of the sulfonamide derivative was remarkably reduced (to 1-3% of the total concentration) in the presence of plasma proteins.) With mouse whole blood, the EC 50 value for BTT-3033 was determined to be 6 M (Fig. 2C). In sharp contrast to BTT-3033, BTT-3034 (10 M) had no inhibitory effect on platelet adhesion to collagen I-coated capillaries under flow (Fig. 2, B and D).
The two major platelet collagen receptors, ␣2␤1 integrin and GPVI, have been shown to activate each other when platelets adhere to collagen I (27). We have previously shown that the sulfonamide inhibitors have no effect on platelets derived from ␣2 integrin-deficient mice (20), suggesting that they do not affect GPVI-mediated adhesion to collagen I. Here, we also coated capillaries with convulxin, a C-type lectin from rattlesnake venom that is a ligand for GPVI, but not ␣2␤1 (28). BTT-3033 (10 M) did not inhibit human platelet aggregation or binding to convulxin; unexpectedly, it actually slightly increased binding (supplemental Fig. 1). Mouse platelets did not show significant binding to convulxin when tested at a high flow rate (120 dynes/cm 2 ). However, under these conditions, a large concentration (8 g/ml) of a specific antibody against mouse GPVI (JAQ1) inhibited platelet binding to collagen (by 27%), suggesting that GPVI may still participate in the binding process (data not shown). Importantly, at 120 dynes/cm 2 , the maximal inhibition by BTT-3033 was ϳ75% (Fig. 2C). These data indicate that GPVI-related mechanisms cannot explain the differential effects of the two sulfonamide derivatives on platelet binding to collagen I under flow.
The Two Sulfonamide Derivatives Bind ␣2␤1 Integrin through Distinct Mechanisms-To obtain a more detailed view of the ␣2␤1 binding sites and the mechanisms of BTT-3033 and BTT-3034, we first tested the ability of these agents to inhibit collagen binding by CHO-␣2Y285F cells. The Y285F mutation in the ␣2I domain has no effect on collagen binding but prevents MIDAS binding of some ␣2␤1 inhibitors, including the previously described sulfonamide derivative, BTT-3016 (Fig.  3A) (20). The function of BTT-3033, unlike that of BTT-3034, was shown to be fully dependent on Tyr-285 (Fig. 3A).
The results of differential scanning fluorimetry (29) showed that both BTT-3033 and BTT-3034 bound to the recombinant ␣2I domain in a metal-dependent manner and reduced the thermal stability of the ␣2I domain (⌬T m , Ϫ2.0°C and Ϫ1.4°C for BTT-3033 and BTT-3034, respectively; Fig. 3B; supplemental Table 3). Importantly, these differences between BTT-3033 and BTT-3034 suggest that BTT-3033 binds more efficiently to the ␣2I domain. We did not obtain similar results with a recombinant ␣2I domain containing the gain-of-function E318W Integrin Activation DECEMBER 28, 2012 • VOLUME 287 • NUMBER 53 JOURNAL OF BIOLOGICAL CHEMISTRY 44697 mutation ( Fig. 3B; supplemental Table 3), which is thought to promote adoption of an open/active ␣2I domain conformation (17,30). The distinct binding sites for BTT-3033 and BTT-3034 were confirmed in competition assays using CBL027, a compound that fluoresces in a Tyr-285-dependent manner (26). BTT-3033 almost completely eliminated CBL027 fluorescence, indicating that BTT-3033 and CBL027 have the same binding site; in contrast, BTT-3034 only partially competed with CBL027 (supplemental Fig. 2). Taken together, our results suggest that BTT-3033 and BTT-3034 have distinct binding preferences and that whereas BTT-3033 favors a site close to the ␣2 MIDAS motif, BTT-3034 acts at least partially through another as yet uncharacterized allosteric site.
The Sulfonamide Derivatives Recognize Different Conformations of ␣2␤1 Integrin-To gain further insight into the specificity of BTT-3033 and BTT-3034, we compared their effects on CHO cell clones transfected with different ␣2 variants (Fig. 4A). The E318W mutation activates ␣2␤1 integrin at the ␣2I-domain level (17,30), whereas the E336A loss-of-function mutation prevents cross-talk between the ␣2I domain and the ␤1 subunit. Equivalent loss-of-function mutations, namely E310A in ␣L and E320A in ␣M (12)(13)(14), are thought to force the integrin into a bent conformation. Before assaying the effects of sulfonamides, we tested the ability of CHO cells expressing variant integrins to attach and spread on collagen I. For this purpose, CHO-␣2wt, CHO-␣2E336A, and CHO-␣2E318W cell clones were plated on collagen I and continuously monitored for 3 h using xCelligence technology (supplemental Fig. 3A), which measures changes in impedance at the bottom of a cell culture well caused by cell attachment and spreading and allows detailed temporal comparisons between cell clones. In general, CHO-␣2E318W and CHO-␣2wt cells showed similar attachment properties when plated on collagen I (supplemental Fig.  3A), whereas CHO-␣2E336A cells attached less effectively to collagen I (supplemental Fig. 3A). Interestingly, the early adhesion and spreading of CHO-␣2E318W cells to collagen I was consistently slightly less than that of CHO-␣2wt cells (supplemental Fig. 3B) but was always greater at the end of the 3-h period. These results suggest that, under static conditions, preactivation of integrins plays a minor role in the early attachment phase but makes the final adhesion stronger.
The ability of the two sulfonamide derivatives to inhibit collagen binding by CHO-␣2E318W and CHO-␣2E336A cells was tested under static conditions and after TPA (100 nM) treatment (Fig. 4B). TPA is a phorbol ester that is known to induce ligand-independent clustering of ␣2␤1 integrin (16). It also induces a transient shift in the wt ␣2I domain from a closed to an open state (16). BTT-3034 was a potent inhibitor of CHO-␣2E318W cells, whereas BTT-3033 was not (Fig. 4B). In these conditions, BTT-3033 inhibited 74% and BTT-3034 53% of binding by CHO-␣2E336A cells (difference not statistically significant).

Under Shear Stress the Sulfonamide Derivatives Have Distinct Effects on the E336A-induced Conformation of ␣2␤1
Integrin-To directly study the function of different integrin conformations under flow conditions, we measured the attachment of transfected CHO cell clones to collagen I-coated capillary walls under flow using a Cellix microfluidic platform (Fig.   FIGURE 3. Both BTT-3033 and BTT-3034 bind to the ␣2I domain but only the function of BTT-3033 is dependent on Tyr-285. A, the adhesion of CHO-␣2Y285F cells to collagen I in the presence or absence of the integrin inhibitors BTT-3033, BTT-3034, or BTT-3016 (at EC 50 concentrations) was evaluated. Adherent cells were detected using the WST-1 reagent. BTT-3034, unlike BTT-3033 or BTT-3016, inhibited the adhesion of CHO-␣2Y285F cells to collagen I, indicating that BTT-3033 and BTT-3016 bind to the ␣2I MIDAS motif. The data expressed are the mean Ϯ S.E. of three independent experiments. Significant difference between BTT-3033 and BTT-3034 is indicated by an asterisk (Student's t test, *, p ϭ 0.05). B, the thermal stability of wt and E318W recombinant ␣2I domains in the presence or absence of BTT-3033 (42 M) or BTT-3034 (49 M) and 1 mM Mg 2ϩ was studied using differential scanning fluorimetry. Both BTT-3033 and BTT-3034 decreased the thermal stability of the ␣2I wt domain in the presence of Mg 2ϩ . 5A). A comparison of low-flow (0.01 dynes/cm 2 ) and high-flow (90 dynes/cm 2 ) conditions showed that CHO-␣2wt cells adhered more effectively to collagen I-coated capillary wall at low than at high shear rates (not shown). The adhesion of CHO-␣2wt cells to collagen I-coated capillaries was inhibited by an ␣2 integrin neutralizing antibody, indicating that, under the test conditions used, attachment is mediated almost entirely by ␣2 (Fig. 5B). Importantly, when cells were tested under fast flow conditions, CHO ␣2E336A cells could bind to collagen, whereas E318W mutation in ␣2 integrin did not improve the binding (Fig. 5C). Similar with the results obtained with E318W integrins, TPA did not activate the binding of CHO-␣2wt cells when the shear rate was 90 dynes/cm 2 (data not shown). Using the lower shear rate, some increase in cell adhesion was detected (data not shown).
Thus, these results suggest that in under flow conditions, the "closed," loss-of-function ␣2␤1 variant can bind to collagen. Preactivation of the integrins, either by E318W mutation (Fig.  5C) or TPA-induced inside-out signaling (data not shown) may even decrease adhesion. Under shear stress conditions (90 dynes/cm 2 ), BTT-3033 inhibited (maximally 70%) the binding of CHO-␣2wt cells to collagen I, an effect not observed with BTT-3034 (data not shown). Thus, CHO-␣2wt cells behaved in a similar manner as platelets. Importantly, BTT-3033 inhibited adhesion of CHO-␣2E336A cells to collagen I (p ϭ 0.028; Fig.  6A), whereas BTT-3034 had no inhibitory effect (Fig. 6B). Thus, BTT-3033 and BTT-3034 recognize different ␣2␤1 integrin conformations, especially under high shear stress conditions. On the basis of these data, it is possible to propose a model for ␣2␤1 action under flow conditions in which the initial interaction with collagen is mediated by a non-activated integrin conformation.

DISCUSSION
Integrin ␣2␤1 is an ␣I domain-containing collagen receptor that is abundantly expressed in human tissues. It has a potential role in inflammation, cancer, and thrombosis and, as such, is a target of many active drug development projects. The structure-function relationship of leukocyte ␣I domain integrins, especially ␣L␤2 and ␣M␤2, has been extensively studied (4, 5). However, much less is known about the conformational regulation of the ␤1-integrins, including the ␣I domain-containing collagen receptors.
In this study, we describe two new sulfonamides that can be used to probe the structural regulation of ␣2␤1 function. BTT-3033 and BTT-3034 have almost identical effects on CHO-␣2wt cell adhesion to collagen I, but only BTT-3033 blocks platelet attachment under flow. Our experiments suggest that the function of BTT-3033 is dependent on ␣2 Tyr-285, whereas that of BTT-3034 is not, despite the fact that BTT-3034 also binds to the ␣2I domain. Thus, BTT-3033 may bind close to MIDAS, whereas BTT-3034 appears to bind to other allosteric sites. We have shown that BTT-3034 interacts with the ␣2I domain, but we cannot exclude the possibility that BTT-3034 also binds other sites (e.g. on the ␤1I domain-␣2 subunit inter-face). The exact binding mechanism of BTT-3034 remains to be solved, but the existence of a potential allosteric regulatory site in the ␣2I domain has been described previously (23). Importantly, the binding sites for BTT-3033 and BTT-3034 appear to be differentially available in distinct integrin conformations. This was shown using CHO cells (which normally have no collagen receptors) transfected with cDNAs encoding variant ␣2 integrins (17). In the ␣2 subunit, amino acid residue Glu-336 corresponds to Glu-310 in ␣L and Glu-320 in ␣M (12)(13)(14). These glutamate residues may act as intrinsic ligands that mediate conformational regulation between ␣ and ␤ I-domains. Mutation of ␣L Glu-310 changes the balance of integrin conformations on the cell surface toward the bent stage (31). In general, it is not known whether ␤1-integrins can adopt a bent conformation, and there is no direct evidence that the E336A  substitution in ␣2␤1 leads to a shift from an extended to a bent structure. However, the obvious inactivation of ␣2␤1, which we have noted in the E336A mutant (16), is difficult to explain in any other way. Mutation of this residue may also prevent preactivation of the ␣2I domain by inside-out signals (16); however, in collagen receptors, closed ␣I domains also bind to their ligands with relatively high avidity (30,(32)(33)(34)(35). Thus, collagen receptors should not be critically dependent on preactivation at the ␣I-domain level.
Another mutation in the ␣2I domain, namely E318W, breaks an intradomain salt bridge (Arg-288/Glu-318) that regulates the shift between closed (non-activated) and open (activated) ␣I domain conformation (30,35). When the two sulfonamides were tested with variant integrins, it was suggested that BTT-3034 is a more effective inhibitor of the gain-of-function ␣2E318W mutant. This difference was seen in assays with transfected cells, but not with recombinant ␣2I domains. Conversely, under flow the inhibition of E336A variant by BTT-3033 was statistically significant, whereas BTT-3034 had no effect.
These data indicate that sulfonamide derivatives can be used to study the biological roles of preactivated and non-activated integrins, especially under shear stress conditions. Surprisingly, only the inhibitor that was selective for non-activated conformation could block platelet-collagen interactions. This contradicts reports proposing that platelet ␣2␤1 must be preactivated by inside-out signals before binding to collagen (36,37). We evaluated the significance of ␣2␤1 preactivation further, testing CHO cells in a Cellix microfluidic platform at high (90 dynes/ cm 2 ) flow rates. With fast flow, CHO-␣2E336A cells could clearly bind to collagen I. Surprisingly, ␣2E318W mutation did not improve cell attachment. Thus, we conclude that preactivation of ␣2 provides no advantage in collagen binding under flow. The results from the Cellix experiments were in accordance with measurements performed using xCelligence technology. In these latter experiments, integrin activation seemed to be more important for strengthening the final adhesion sites than for the initial attachment.
Our results suggest a model in which, under flow, non-activated integrins are capable of low-affinity interactions with collagen. This may generate signals that further activate integrins and lead to firm adhesion. Importantly, the initial interaction between non-activated integrins and collagen seems to be essential for the entire process of integrin-mediated attachment. Low-avidity interactions, leading to platelet rolling but not firm adhesion, may play an important role in platelet-vessel wall interaction (38). The molecular basis for platelet rolling is not completely understood, but the interaction between GPIb␣ and immobilized von Willebrand factor can at least partially explain this phenomenon (38). Leukocytes exhibit an analogous rolling on an endothelial cell layer before attachment and extravasation (38). This process was originally shown to be dependent on the action of P-or E-selectin, whereas activated ␣L␤2 integrin (LFA-1) was proposed to mediate firm adhesion only (39). More recently, several reports have indicated that leukocyte integrins may also participate in this process and stabilize selectin-mediated rolling (40). In ␣L␤2 integrin, the transition from rolling to firm adhesion is associated with a change in the ␣I domain from a closed to an open or intermediate conformation (41,42). However, the overall mechanism of ␣L␤2 action seems to be different compared with that of ␣2␤1 integrin: The E310A mutation in ␣L results in an inactive receptor that can support neither rolling nor firm adhesion (31), whereas our experiments showed that the corresponding mutation (E366A) in ␣2 only weakens collagen binding but does not completely prevent it. In conclusion, in a continuation of our previous research approach (20), we have developed two novel ␣2␤1 inhibitors that show selectivity for non-activated and activated ␣2␤1 under flow. These inhibitors were used to test the role of different ␣2␤1 conformations during cell and platelet adhesion to collagen I. The results support the idea that, under flow, no preactivation of ␣2␤1 integrin is needed for collagen recognition.