Structure of Collagen Receptor Integrin α1I Domain Carrying the Activating Mutation E317A*

Background: The integrin αI domain undergoes a conformational change during activation. Results: The crystal structure of an activated αI domain is different from the reported open and closed states. Conclusion: Our structure mimics the state where the Arg287-Glu317 ion pair is just broken during the activation process. Significance: The activation mechanism of the collagen receptor integrins differs from the other integrins. We have analyzed the structure and function of the integrin α1I domain harboring a gain-of-function mutation E317A. To promote protein crystallization, a double variant with an additional C139S mutation was used. In cell adhesion assays, the E317A mutation promoted binding to collagen. Similarly, the double mutation C139S/E317A increased adhesion compared with C139S alone. Furthermore, soluble α1I C139S/E317A was a higher avidity collagen binder than α1I C139S, indicating that the double variant represents an activated form. The crystal structure of the activated variant of α1I was solved at 1.9 Å resolution. The E317A mutation results in the unwinding of the αC helix, but the metal ion has moved toward loop 1, instead of loop 2 in the open α2I. Furthermore, unlike in the closed αI domains, the metal ion is pentacoordinated and, thus, prepared for ligand binding. Helix 7, which has moved downward in the open α2I structure, has not changed its position in the activated α1I variant. During the integrin activation, Glu335 on helix 7 binds to the metal ion at the metal ion-dependent adhesion site (MIDAS) of the β1 subunit. Interestingly, in our cell adhesion assays E317A could activate collagen binding even after mutating Glu335. This indicates that the stabilization of helix 7 into its downward position is not required if the α1 MIDAS is already open. To conclude, the activated α1I domain represents a novel conformation of the αI domain, mimicking the structural state where the Arg287-Glu317 ion pair has just broken during the integrin activation.

In ␣ 1 ␤ 1 , as in all ␣I domain integrins, the ligand binding takes place at the MIDAS of the ␣I domain. The atomic structures of the ␣ L , ␣ M , ␣ X , ␣ 1 (human and rat) and ␣ 2 I domains have been solved (6, 18 -23). The structure of the closed form of ␣ 1 I and ␣ 2 I (PDB code 1AOX) (6) are almost identical, as their structural superimposition gives a root mean square deviation of 0.62 Å for 177 C ␣ atoms. In the active site, the metal ion and the surrounding residues occupy identical positions in both structures, and the ␣C helix is stabilized by the Arg 287 -Glu 317 (␣ 1 I numbering) ion pair interaction as well. The structures of the ␣ L I domain in complex with ICAM-1 or ICAM-5 (24,25), as well as ␣ 2 I in complex with a collagenous GFOGER peptide have also been solved (26). Ligand binding opens the ␣I domain by triggering large conformational changes in the ␣C helix and helix 7 and some spatial adjustments in the MIDAS. The ␣C helix in the ␣ 2 I domain unwinds and swings away from the vicinity of the MIDAS. Helix 7 moves significantly downward, allowing a specific glutamate to bind to the MIDAS of the I domain of the ␤ subunit. Glu 310 in ␣L, Glu 320 in ␣M, and Glu 336 in ␣2 are supposed to stabilize helix 7 in the open conformation and mediate structural changes between integrin ␣ and ␤ subunits (24,(27)(28)(29). These conformational modifications can also lead to the separation of the integrin leg parts and promote cell signaling.
Here, we have characterized the collagen binding properties of an activated variant of the ␣ 1 I domain, harboring a gain-offunction mutation E317A (30). To promote protein yield, solubility, and crystallization, a double variant with an additional C139S mutation was used. In cell adhesion assays E317A mutation promoted binding to collagen. Similarly, double mutation C139S/E317A increased adhesion compared with C139S alone. Furthermore, E317A could increase cell adhesion even after the connection between ␣ 1 I and ␤ 1 subunits had been prevented by E335A mutation. The solved 1.9 Å crystal structure of the activated variant of ␣ 1 I has a novel conformation, which is different from the previously reported "open" and "closed" ␣I domains.

EXPERIMENTAL PROCEDURES
Materials-Rat collagen I was purchased from Sigma-Aldrich, and collagen IV from mouse Engelbreth-Holm-Swarm tumor was from BD Biosciences. Bovine serum albumin (BSA) was from Bovogen Biologicals.
Cell Adhesion Assays-Attachment and spreading of chimeric ␣ 1 ␤ 1 (human ␣ 1 , hamster ␤ 1 ) expressing CHO cells was tested with xCELLigence real-time cell analyzer (RTCA; Roche Applied Science). This technology measures impedance at the bottom of a microtiter plate well and allows estimating the progression of cell attachment and spreading. E-plates 96 (Roche Applied Science) were coated either with collagens I, IV (5 g/cm 2 /well; 16.4 g/ml) or BSA (0.1%) in Dulbecco's phosphate-buffered saline (PBS; Sigma) overnight at 4°C. Before coating collagen I was diluted in acetic acid (0.1 M) to keep it in monomeric, triple helical form. Coated wells were washed with PBS and blocked with BSA for 1 h (0.1% in PBS, 5% CO 2 , 37°C). After blocking, BSA was removed, and ␣-MEM (without FCS) was added to the wells. The background signal was measured, and 40,000 cells/well were added. BSA was used as the negative control, and its signal was measured for each cell line with three parallel wells (data not shown). The Mann-Whitney U test with SPSS software (version 16.0; IBM) was used for statistical analysis of the data collected at the 1-h time point. Cell adhesion was followed for 2 h at 37°C (5% CO 2 ). The cells used in these experiments were collected from almost confluent culture plates. Trypsin-EDTA solution (Sigma) was used to remove the cells from the plate, and trypsin inhibitor (Sigma) was added. The cells were spun down at 500 ϫ g for 5 min at 37°C, and the pellet was resuspended using ␣-MEM medium without FCS.
Protein Expression and Purification-For solid phase binding assays, protein expression and purification was performed as described earlier (30). Briefly, glutathione S-transferase (GST) fusion proteins were expressed in Escherichia coli BL21 Tuner TM (Novagen). For crystallization of ␣ 1 I C139S/E317A, the protein was produced using 100 ml of overnight culture to inoculate 15 liters of LB AMP -medium and cultured in Bioengineering fermentor (Bioengineering AG) at 37°C until the A 600 nm reached 0.5-0.6. To induce expression of the protein, isopropyl ␤-D-1-thiogalactopyranoside was added to a final concentration of 0.4 mM. Protein production was continued at room temperature for 6 h and at 10°C overnight. The cells were harvested by centrifugation, and the GST fusion protein was purified as described earlier (30). GST was removed by thrombin digestion (GE Healthcare). 100 units of thrombin were added every 2nd h for 8 h. GST was separated from the ␣ 1 I domain by GST-affinity chromatography using disposable columns (Bio-Rad). Protein was further purified by gel filtration chromatography (HiLoad TM 26/60 Superdex TM 200 preparation grade; GE Healthcare) using Äkta FPLC. The purified protein was concentrated using Centriprep centrifugal filter device (Millipore) followed by buffer exchange to 40 mM Tris (hydroxymethyl) aminomethane, pH 7.2, 2 mM MgCl 2 , and 20% glycerol. Finally, the protein was concentrated using Centricon centrifugal filter device (Millipore) to 23-28 mg/ml. The purity of the protein was checked by electrophoresis on 8 -25% gradient polyacrylamide gels in the presence of 0.55% SDS using the Phast System (Amersham Biosciences).
Solid Phase Binding Assays-Binding assays were carried out as described previously (30,32). 96-well plates were coated overnight at 4°C with either collagens I, IV (5 g/cm 2 /well) or BSA (3.75%) that was used for blocking the wells. GST fusion ␣ 1 I domains were allowed to bind for 1 h in the presence of 2 mM MgCl 2 in Delfia® assay buffer (PerkinElmer Life Sciences). Wells were washed three times, and the signal was detected with Delfia® europium-labeled anti-GST antibody (PerkinElmer Life Sciences). Label was dissociated with Delfia® enhancement solution (PerkinElmer Life Sciences), and the signal was determined using a time-resolved fluorescence spectrophotometer (Victor3 multilabel counter; PerkinElmer Life Sciences). Estimates for the dissociation constants (K d ) were obtained using the following equation: measured binding ϭ maximal binding/(1 ϩ K d /[␣I]).
Crystallization and Data Collection-Initial screening for crystallization conditions was done with the sparse matrix screen JCSGϩ Suite (Qiagen) using the hanging drop vapor diffusion method. The crystals were grown at 8°C by mixing 1 l of 26 mg/ml protein solution in 40 mM Tris, pH 7.4, 2 mM MgCl 2 , and 20% glycerol with 1 l of the well solution, which contained 1.6 M trisodium citrate. In a week, the crystals grew to the final size of 0.3 ϫ 0.1 ϫ 0.1 mm 3 . The crystal was picked directly from the crystallization plate and flash frozen with liquid nitrogen. The crystal diffracted to 1.9 Å, and the data were collected on ADSC Quantum Q210 detector installed on beamline ID 14-1 at the European Synchrotron Radiation Facility (ESRF, France). The data were integrated and scaled in space group P3 using the XDS and XSCALE programs (33).

Structure Determination, Model Building, and Refinement-
The solvent content of the crystal was 45.5%, with two chains in the asymmetric unit ( Table 1). The structure of ␣ 1 I C139S/E317A was determined by molecular replacement using the program Molrep (34,35). We searched for two ␣ 1 I domains using closed ␣ 1 I (PDB code 1PT6) (23) as a search model. 5% of the reflections (1,428) were randomly selected for R free calculation, and the remaining data (28,552) were used in refinement (Table 1). Using Refmac5 (36), the initial R-factor after rigid body refinement was 27.6% (R free ϭ 26.6). ARP/wARP Solvent (37) was used for the addition of water molecules. The Refmac5 refinement was followed by manual model building in COOT (38). Iterative model building and refinement gave a final model with R-factor 18.6% (R free ϭ 22.9). The electron density quality was good throughout both chains except for the weak electron density of the residues 286 -291 in chain B and the total lack of electron density for the residues 286 -290 in chain A. The final model was validated using Molprobity (39). All the structural figures were made with PyMOL (40).

RESULTS
Activation of Integrin ␣ 1 ␤ 1 by ␣ 1 E317A Mutation Leads to High Avidity Cell Adhesion to Collagens I and IV-In our previous studies, we have identified ␣ 1 I E317A as a gain-of-function mutation, which improves collagen binding considerably (30). To investigate the integrin activation further, we used CHO cells transfected to express wild-type (WT) ␣ 1 and mutant integrins (␣ 1 E317A, ␣ 1 C139S, ␣ 1 C139S/E317A). Flow cytometry experiments confirmed comparable expression levels for the variant and WT ␣ 1 integrins (supplemental Fig. S1). The effect of integrin mutations on cell adhesion was analyzed using the xCELLigence instrument. This technology measures The collagen binding properties of the soluble, recombinant ␣ 1 I C139S and ␣ 1 I C139S/E317A variants were studied in a solid phase binding assay. The double variant containing the activating E317A mutation (␣ 1 I C139S/ E317A) bound to both collagen I and IV significantly better than ␣ 1 I C139S (Fig. 2). Furthermore, based on K d values, ␣ 1 I C139S binds to collagens I and IV with similar avidity, whereas ␣ 1 I C139S/E317A binds to collagen IV more tightly than to collagen I. The results suggest that E317A activates the ␣ 1 I domain even in the presence of the C139S mutation. Because of its higher avidity, we have named ␣ 1 I C139S/E317A as an activated form of ␣ 1 I.
X-ray Structure of Activated ␣ 1 I C139S/E317A-The structure of the double variant C139S/E317A of the ligand-free ␣ 1 I domain was solved by molecular replacement using the closed conformation of ␣ 1 I (PDB code 1PT6) (23) as a search model. We refined the structure using Refmac5 (36) to a final R-factor of 18.6% (R free ϭ 22.9%). Statistics for the data processing and structural refinement are summarized in Table 1. The asymmetric unit in the crystal contains two molecules, which are identical, and their structural superimposition gives a root mean square deviation of 0.06 Å for 177 C ␣ atoms. Each chain comprises residues 142-333, 185 water molecules, and a Mg 2ϩ bound at the MIDAS. The region of residues 138 -141 in the N terminus of both chains is disordered, so the mutation C139S is not visible in the structure. The electron density map was good throughout the structure except for the loop region 286 -290 (Fig. 3). The residues 286 -290 in the loop region in chain A could not be built because of the lack of electron density, and only the main chain atoms are built for chain B (Fig. 3B). In the following, we therefore discuss only chain B. The core structure is similar to the closed form of the WT ␣ 1 I with a central hydrophobic ␤ sheet sandwiched by six amphipathic ␣ helices.
Activating E317A Mutation Caused Significant Conformational Changes in ␣C Helix-The ␣C helix (residues 283-287) stabilized by the Arg 287 -Glu 317 ion pair interaction plays important role in collagen binding (7). The ␣C helix unwinds when collagen binding to the MIDAS disrupts Arg 287 -Glu 317 interaction (26). Compared with the closed form of ␣ 1 I (Fig. 4A,  green), where Tyr 285 restricts the ligand binding to the MIDAS, in our activated C139S/E317A structure (Fig. 4A, salmon) the ␣C helix unwinds and Tyr 285 is moved away from the MIDAS. As a result of this movement, the active site opens. This is to our knowledge the first structure of the ␣ 1 I domain with the unwound ␣C helix in the absence of collagen.
MIDAS of Activated ␣ 1 I-In the MIDAS of the integrin ␣I domain residues from three loops (L1, L2, and L3) surround the metal ion in the active site and coordinate the metal directly or indirectly (Fig. 5). Compared with the hexacoordination of the metal in previously published ␣ 1 I and ␣ 2 I structures, the metal ion in our activated ␣ 1 I structure forms a pentacoordinated complex (see below), where Ser 152 and Ser 154 from L1 and three water molecules coordinate the metal through their hydroxyl oxygen (Fig. 5B).
Activated ␣ 1 I C139S/E317A Shows Differences in MIDAS, ␣C Helix, Helix 6, and Helix 7 Compared with Open ␣ 2 I Structure-The core structure of our activated ␣ 1 I is very similar to the open ␣ 2 I. However, it is different from the open ␣ 2 I with respect to the MIDAS, the ␣C helix, helix 6, and helix 7. The metal coordination in our activated ␣ 1 I structure is slightly different from both closed and open ␣I (Fig. 5, A-C). The direct binding of Ser 152 and Ser 154 (␣ 1 I numbering) to the metal is conserved in all the structures. In the activated ␣ 1 I, the metal moves (2.2 Å) toward L1 with respect to its position in the closed ␣ 1 I (Fig. 5B). This is similar but not identical to the metal ion movement between the open and closed ␣ 2 I, where the metal moves 2.6 Å toward L2 (26). Because of the metal ion movement toward L1 in the activated ␣ 1 I, the conserved threonine (Thr 220 ) on L2 interacts with the metal via another water molecule (Fig. 5B) unlike in the open ␣ 2 I where Thr 221 binds directly to the metal (Fig. 5C). Similar to the open ␣ 2 I structure, Asp 253 in the activated ␣ 1 I makes a water-mediated contact to the metal (Fig. 5E).
As in the open ␣ 2 I structure, the ␣C helix of the activated double variant has unwound to a loop structure. Although the residues near the MIDAS are mostly conserved, the residues Ser 284 and Tyr 285 in the ␣C helix of ␣ 1 I correspond to Tyr 285 and Leu 286 in ␣ 2 I, respectively (Fig. 5D). Despite these differences in the amino acid sequences, the tyrosine residue is turned away from the MIDAS in both open ␣ 2 I and activated Binding of the ␣ 1 I C139S and ␣ 1 I C139S/E317A variants to collagen I (A) and collagen IV (B) was studied in a solid phase binding assay. 96-well plates were coated either with collagen I, IV, or BSA in PBS. BSA served as a negative control and was also used for blocking the wells. GST fusion ␣ 1 I domains were allowed to bind for 1 h in the presence of 2 mM MgCl 2 in Delfia® assay buffer. Wells were washed, and the signal was detected with Delfia® europium-labeled anti-GST antibody. Label was dissociated with Delfia® enhancement solution, and the signal was determined using a time-resolved fluorescence spectrophotometer. Each sample was measured with three parallel wells. Estimates for the dissociation constants were obtained using an equation: measured binding ϭ maximal binding/(1 ϩ K d /[␣I]).     (Fig. 5F). Unlike in the open ligand-bound ␣ 2 I structure, the uncoiling of the ␣C helix does not lead to an extra turnover helix 6 (Fig. 5, B and C). Furthermore, helix 7 in the collagen-bound open ␣ 2 I structure moves downward but does not show such a change in the activated ␣ 1 I structure (Fig. 5F).
Link between ␣ 1 I and ␤ 1 I Domains Is Not Required for E317A-related ␣ 1 I Activation-Previous research has indicated that helix 7 in the ␣I domain is linked to the ␤ subunit by a specific glutamate residue that acts as an "intrinsic ligand" for the MIDAS in the ␤I domain. Glu 310 in ␣ L and Glu 320 in ␣ M are supposed to mediate conformational changes between the two integrin subunits (24,27,28). Similarly, mutation of an equivalent residue Glu 336 in ␣ 2 subunit inactivates ␣ 2 ␤ 1 integrin (29). Because in our novel activated ␣ 1 I domain structure the activation was not associated with a concomitant movement of helix 7, we decided to study whether also at the cellular level ␣ 1 ␤ 1 activation could take place without the relocation of helix 7. In this purpose we introduced mutation E335A alone and in combination with E317A into the ␣ 1 subunit, expressed the variant integrins in CHO cells, and measured their binding to collagen IV using xCELLigence technology (Fig. 6). At the 1-h time point E335A mutation decreased cell adhesion about 10-fold compared with ␣1 WT transfected CHO cells (Fig. 1), whereas the second mutation E317A partially reversed the effect resulting in 2.5-fold increase (Fig. 6). Thus, E317A-related structural changes in the ␣ 1 I domain do not require the relocation or stabilization of helix 7 in a Glu 335 -dependent manner.

DISCUSSION
The mutation E318A in ␣ 2 I activates collagen and laminin binding (30,41). We have previously introduced the equivalent E317A mutation to ␣ 1 I and shown its activating character (30). To get the first three-dimensional structure of an activated collagen receptor ␣ 1 I without a ligand, we had to introduce an additional mutation, namely C139S, to increase the protein yield and solubility. Equivalent cysteine to serine mutation has been used to decrease the aggregation of ␣ M I domain (42). We used both solid phase binding assay and cell adhesion assay to confirm that the mutant ␣ 1 I C139S/E317A still represents an activated ␣I domain. Kinetically, ␣ 1 I C139S/E317A resembles ␣ 1 I WT, binding more tightly to collagen IV than to collagen I. Interestingly, C139S mutation alone decreased both the ␣I domain binding to collagen and cell adhesion. We also know based on our earlier experiments that the short ␣ 1 I and ␣ 2 I domain constructs lacking this particular cysteine show weaker binding to collagens. 4 In the ␣ X ␤ 2 heterodimer (PDB code 3K6S) (43), the corresponding cysteine links the ␣ X I domain to the ␤ propeller domain in the ␣ X subunit by forming a disulfide bond with a cysteine residue in the ␤ propeller. The loss of this stabilizing disulfide would explain the decreased activity of the ␣ 1 ␤ 1 heterodimer harboring the ␣ 1 C139S mutation. However, the stabilization mechanism of C139 in the recombinant ␣ 1 I domain remains unknown.
Collagen binding to the ␣I domain initiates the loss of the Arg 287 -Glu 317 ion pair interaction (Fig. 4C), which leads to the unwinding of the ␣C helix. We originally predicted that the loss of the salt bridge in the ␣ 1 I E317A variant would trigger large conformational changes similar to those seen in the collagenlike peptide bound ␣ 2 I (26), including the downward movement of helix 7 (30). However, our activated mutant structure revealed unexpected and interesting mechanism for the ␣I domain activation without any change in the position of helix 7. Even though the activation of ␣ 1 I domain is due to the gain-offunction mutation and, therefore, not physiological, the unveiled atomic structure can be used to explain the activation mechanism.
Enhanced ligand binding of the ␣ 1 I C139S/E317A variant could be explained by the fact that Tyr 285 , which covers the MIDAS in the closed form (23), has moved aside and unwinds the ␣C helix making the MIDAS more accessible for the ligand. Tyr 285 in ␣ 1 I is not conserved and does not correspond to Tyr 285 in ␣ 2 I (Fig. 5D); however, it seems to take the role of Tyr 285 in ␣ 2 I (Fig. 5F). This could result in differences between the open forms of ␣ 1 I and ␣ 2 I, because the ␣C helix might not be long enough in ␣ 1 I to reorganize an additional turn on helix 6. This would also mean that the open form of ␣ 1 I is different from the homology model of the open ␣ 1 I (23, 30), which was predicted based on the ligand-bound ␣ 2 I structure (26). It seems reasonable to speculate that the change in the metal ion coordination (from penta-to hexacoordinated) upon ligand binding to the activated mutant ␣ 1 I would induce the change of the activated form to a fully open form of ␣ 1 I where helix 7 is in the downward position. Because the opening of the ␣C helix does not have an effect on helix 7, we propose that interaction with collagen or a collagen like peptide is necessary to push helix 7 into the position required for interaction with the MIDAS of the ␤I domain. Recently, the structure of the ␣ L I domain in complex with its ligand ICAM-5 showed the unusual allosteric mobility of helix 7 (25). Previous studies have proposed that the open ␣I domain conformation is also possible without a ligand and represents an activated integrin conformation. Gain-of-function mutations can be used to open the ␣I domain structure and experimentally mimic the activated ␣I domains. The activated ␣I domains, which enhance helix 7 movement, are reported for ␣ X I I314G (44), ␣ M I F302W (45), and ␣ M I I316R and I316G (42). Also, cysteine substitutions that stabilize open conformations via disulfide bonds can be used to activate integrin ␣I domains: ␣ 2 I G172C/L322C (46), ␣ L I K287C/K294C, ␣ L I E284C/E301C (24), ␣ M I C128A/D132C/K315C (47), ␣ M I Q163C/Q309C, ␣ M I D294C/Q311C (48). In another study, stable open conformation of ␣ M I was computationally designed, and numerous mutations were introduced to obtain the desired conformations (49). Based on simulations of molecular dynamics and energy minimization ␣ L I and ␣ M I, unlike ␣ 1 I and ␣ 2 I, are considered to also have an intermediate form (50). The structure of the activated ␣ 1 I variant solved in this study is also clearly different from the intermediate form of ␣ L I (24). In the activated ␣ 1 I domain, all of the structural changes had taken place close to the MIDAS, whereas in the intermediate-affinity I domain structure helix 7 was partially shifted down, and the MIDAS was closed.
The functional stage of all integrins is strictly regulated (51,52). In general, a bent integrin conformation is considered to be unable to bind to large ligands. Integrin preactivation has been considered to take place after the inside-out signals have induced the binding of cytoplasmic proteins, e.g. talin or kindlin, to the cytoplasmic domain of the integrin ␤ subunit (53). The movement of integrin legs apart may then lead to integrin extension and a conformational change in the ␤I domain. A specific glutamate residue (e.g. Glu 310 in ␣ L and Glu 320 in ␣ M ) in ␣I domain acts as a link to the MIDAS in ␤I domain and may transfer the conformational change from the ␤ subunit to the ␣I domain. In this integrin activation model, helix 7 is first pulled down, which then opens and activates the ␣I domain structure (51,52). The activation of integrin ␣ X ␤ 2 requires both extension and an open headpiece (54). However, the collagen receptor ␣I domain integrins may behave in a different manner. We have shown that the nonactivated ␣ 2 ␤ 1 integrin can effectively bind to a large size ligand, namely human echovirus-1 (EV-1) (55). Furthermore, the collagen receptor integrin ␣I domains purified as recombinant proteins can recognize ligands with relatively good avidity even in the closed conformation (7,16,30,41,56). We have used mutagenesis to address the question, whether the stabilization of the downward-moved position of helix 7 by ␤ 1 I is required for the activation of ␣ 1 I domain. In ␣ 1 ␤ 1 integrin, the residue in the ␣ 1 I domain that acts as an intrinsic ligand for the ␤ 1 I domain is Glu 335 . Mutating Glu 335 to alanine in ␣ 1 ␤ 1 -expressing CHO cells leads to weaker cell adhesion on collagen IV. Interestingly, the double variant of ␣ 1 containing E335A and the activating mutation E317A restored adhesion of the CHO cells, which indicates that the Glu 335mediated binding to ␤ 1 I is not necessary for the activation of ␣ 1 ␤ 1 integrin (Fig. 6).
Integrin functions can also be regulated by lateral interactions with other membrane proteins, e.g. syndecan family heparan sulfate proteoglycans (57), tetraspanins, and urokinase-type plasminogen activator receptor (58). There seem to be direct or indirect interactions between integrins and these proteins, but very little is known about the structural basis of the mechanism that they use in integrin regulation. It is possible to speculate that the integrin conformation can be modified by other membrane proteins, which would e.g. induce the breakage of the Arg 287 -Glu 317 salt bridge and consequently unwind the ␣C helix. It remains to be seen, however, whether the novel ␣ 1 I domain conformation represents this kind of activation mechanism.
To conclude, we have solved to our knowledge the first atomic structure of an activated collagen receptor ␣ 1 I domain. The conformational changes explain the enhanced ligand binding, but are clearly different compared with the previously pub-