A pivotal role for a conserved bulky residue at the α1-helix of the αI integrin domain in ligand binding

The ligand-binding βI and αI domains of integrin are the best-studied von Willebrand factor A domains undergoing significant conformational changes for affinity regulation. In both βI and αI domains, the α1- and α7-helixes work in concert to shift the metal-ion–dependent adhesion site between the resting and active states. An absolutely conserved Gly in the middle of the α1-helix of βI helps maintain the resting βI conformation, whereas the homologous position in the αI α1-helix contains a conserved Phe. A functional role of this Phe is structurally unpredictable. Using αLβ2 integrin as a model, we found that the residue volume at the Phe position in the α1-helix is critical for αLβ2 activation because trimming the Phe by small amino acid substitutions abolished αLβ2 binding with soluble and immobilized intercellular cell adhesion molecule 1. Similar results were obtained for αMβ2 integrin. Our experimental and molecular dynamics simulation data suggested that the bulky Phe acts as a pawl that stabilizes the downward ratchet-like movement of β6-α7 loop and α7-helix, required for high-affinity ligand binding. This mechanism may apply to other von Willebrand factor A domains undergoing large conformational changes. We further demonstrated that the conformational cross-talk between αL αI and β2 βI could be uncoupled because the β2 extension and headpiece opening could occur independently of the αI activation. Reciprocally, the αI activation does not inevitably lead to the conformational changes of the β2 subunit. Such loose linkage between the αI and βI is attributed to the αI flexibility and could accommodate the αLβ2-mediated rolling adhesion of leukocytes.

Integrins are cell adhesion molecules that transmit both the mechanical and chemical signals into and out of the cells and thus allow the cells to communicate with their surroundings (1)(2)(3)(4). The different combinations of 18 ␣ and 8 ␤ subunits lead to 24 integrin ␣/␤ heterodimers that play important roles in diverse physiological and pathological conditions such as hemostasis, development, immune responses, thrombosis, inflammation, and cancer (1). All the ␤ integrin subunits con-sist of 10 subdomains, ranking from the N terminus to the C terminus are PSI (plexin, semaphorin, and integrin), hybrid, I (inserted), integrin epidermal growth factor (I-EGF) 2 1-4, ␤-tail, transmembrane, and cytoplasmic tail domains (Fig. 1, A-F). All the ␣ integrin subunits are composed of seven subdomains, namely ␤-propeller, thigh, genu, calf-1, calf-2, transmembrane, and cytoplasmic tail (Fig. 1, A-F). In addition, half of the ␣ subunits including the leukocyte-specific integrins ␣ L and ␣ M contain an extra ␣I domain that is inserted into the ␤-propeller domain (Fig. 1, D-F). For integrins without the ␣I domain, the ␤-propeller and the ␤I domain form the ligandbinding site at the top of their interface (Fig. 1, A-C), whereas the ␣I domain is solely responsible for binding ligands when it is present in an ␣ subunit (Fig. 1, D-F). It has been demonstrated that the conserved Glu residue at the ␣I C terminus acts as an internal ligand for the ␤I domain ( Fig. 1F) (5). EM, crystallography, small-angle X-ray scattering, and mutagenesis studies have revealed at least three conformational states of integrin: bent (Fig. 1, A and D), extended with closed headpiece (Fig. 1, B and E), and extended with open headpiece (Fig. 1, C and F), which have been proposed to represent the inactive, intermediate, and active integrin conformations, respectively (5). Such a bent-to-extended global conformational transition can be triggered from the ligand-binding head region and transmit in the outside-in direction or propagated from the cytoplasmic tail and transmit in the inside-out direction. How the wave of conformational change is relayed mechanically across the tandem multiple domains of integrin has been an active area of research in the last decades (6 -10).
Integrin global conformational rearrangements are triggered by the local structural changes (11), which have been very welldefined for the ␣I and ␤I domains by the crystal structures (12)(13)(14)(15)(16)(17). The ␣I and ␤I domains (also known as A domains) are structural homologs that belong to the large von Willebrand factor A (VWA) domain superfamily, which is widespread in the eukaryotes, bacteria, and viruses and perhaps beyond them (18). Both ␣I and ␤I domains contain a metal ion-dependent adhesion site (MIDAS) defined by a signature Asp-Xaa-Ser-Xaa-Ser (DXSXS) motif (Fig. 1G), which is connected to the N terminus of an ␣1-helix and also is seen in many other VWA domains (Fig. 1J). In addition, the ␤I domains have two additional metal ion-binding sites flanking the MIDAS, namely adjacent to MIDAS (ADMIDAS) and synergetic metal ion-binding sites (Fig. 1H). Both ␣I and ␤I domains can shift between two conformations defined as the closed inactive and the open active states. In the active state, the ␤1-␣1 loop moves toward the MIDAS, increasing its potential for binding ligands (Fig. 1, H and  I). For ␤I domain, this is accompanied by a bent-to-straight structural transition, resulting in an inward movement of the ␣1/␣1Јhelix (Fig. 1H), whereas for ␣I domain, it is associated with only a swung-in motion of the N terminus of ␣1-helix (Fig. 1I). For both ␤I and ␣I domains, such conformational changes of ␣1/(␣1Ј)-helix are linked with a downward displacement of ␤6-␣7 loop and ␣7-helix, leading to hybrid domain swinging out ( Fig. 1, C, F, and H) or the engagement of ␣I and ␤I domain via an internal ligand at the C terminus of ␣7-helix (Fig. 1, F and I). In our recent studies, we defined the role of an absolutely conserved Gly residue at the ␣1/␣1Ј-helix (Fig. 1H), which is critical in maintaining the inactive conformation of ␤I domains of ␤ 1 , ␤ 2 , and ␤ 3 integrins (19). Interestingly, a conserved Phe is present at a homologous position of the ␣I ␣1-helix among all the ␣I-containing integrins (Fig. 1, G and I). The equivalent Phe is also seen in the ␣1-helix of many VWA domains other than integrin (Fig. 1J). We speculated that such a conserved Phe residue could be important for the activation of ␣I domains.
In this study, we tested our above-mentioned hypothesis on ␣ L ␤ 2 integrin, given that the activation assay of ␣ L ␤ 2 has been very well-established and the high-resolution crystal structures of the ␣ L ␣I domain are available in both closed and open states (15, 19 -21). The ␣ L ␤ 2 integrin binds ligand intercellular adhesion molecule 1 (ICAM-1) and plays important roles in leukocyte adhesion and migration during the immune responses (5). To investigate a potential structural role of the conserved Phe of the ␣I ␣1-helix in regulating the activity of ␣ L ␤ 2 integrin, we performed the soluble ICAM-1 binding, cell adhesion, and the conformation-specific mAb-binding assays. Our results revealed an unexpected function of the Phe residue in ␣ L ␤ 2 activation, which may be generalized to all the integrin ␣I domains or other related VWA domains that couple large-scale conformational changes to activation and ligand binding.

A bulky residue in the mid position of ␣1-helix of ␣ L ␣I domain is critical for ligand binding
A hallmark structural change of ␣I domain in the transition from the low to the high-affinity state is the inward movement of the N terminus of ␣1-helix and the piston-like movement of the entire ␣7-helix ( Fig. 1, E, F, and I). Structural superimposition of the active conformation onto the resting conformation of the ␣ L ␣I domain shows steric clashes between the N-terminal interfacial residues of ␣1-helix and ␣7-helix ( Fig. 2A), indicating that these helixes should mechanically work in concert to avoid the collisions. Interestingly, the invariant phenylalanine residue ␣ L -Phe-153 at the middle of ␣1-helix in the active position has no van der Waals overlaps with the ␣7-helix in the resting position ( Fig. 2A). To test whether the conserved ␣ L -Phe-153 is important for the activation of ␣ L integrin, we mutated it to Gly, Ala, Val, Leu, Met, Tyr, and Trp, respectively, ranging from the slimmest Gly to the bulkiest Trp (Fig. 2B). The activation of ␣ L ␤ 2 integrin was assessed by the ICAM-1-binding assay in the presence of the universal integrin activator Mn 2ϩ . Remarkably, the ␣ L -F153G and ␣ L -F153A mutations almost completely abrogated Mn 2ϩ -induced ICAM-1 binding to HEK293FT cells expressing ␣ L ␤ 2 integrin (Fig. 2C). Remarkably, the level of ICAM-1 binding correlated well with the side chain volumes of the substituted amino acids for ␣ L -Phe-153, i.e. Gly ϭ Ala Ͻ Val Ͻ Met Ͻ Leu ϭ Phe ϭ Tyr Ͻ Trp (Fig. 2C). As a comparison, the ␣ L -Glu-146 and ␣ L -Lys-149 were mutated to Ala. These residues show steric clashes with the ␣7-helix Phe-292 when overlapping their high-affinity conformation with the lowaffinity conformation of ␣7-helix ( Fig. 2A). As expected, both ␣ L -E146A and ␣ L -K149A mutations greatly dampened Mn 2ϩ -induced ICAM-1 binding, but to a lesser extent compared with the ␣ L -F153A mutation (Fig. 2D). Interestingly, the ␣ L -F147A and ␣ L -L151A mutations, located at the opposite side of the ␣1and ␣7-helix interface ( Fig. 2A), also decreased ICAM-1 binding (Fig. 2D). By contrast, the ␣ L -K160A mutation at the bottom of ␣1-helix had no effect on Mn 2ϩ -induced ICAM-1 binding of ␣ L ␤ 2 integrin (Fig. 2, A and D). Most of the mutations had little effect on the ␣ L ␤ 2 cell surface expression except the Leu substitution decreased the expression by ϳ50% (Fig. 2, C and D). In addition, the ␣ L -F153I mutation completely abolished ␣ L ␤ 2 cell surface expression probably because the position of ␣ L -Phe-153 cannot accommodate an Ile residue because all the Ile rotamers show severe steric clashes with the surrounding residues when mutated in silico (data not shown). These data demonstrated that a bulky amino acid in the mid ␣1-helix is critical for the high-affinity ligand binding of ␣I domain. In addition, the N-terminal residues of ␣1-helix also contribute to the high-affinity conformation of ␣1-helix.

Phe-153 mutation
It has been shown that the downward movement of the ␣7-helix is important for the activation of ␣I domain (15,22,23). We speculated that the inactivating effect of the ␣ L -F153G or F153A mutation could be due to the destabilization of the downward movement of ␣7-helix. In line with this possibility, enforcing the downward movement of ␣7-helix, for example by mutations, may rescue the inactivating effect of the ␣ L -F153 mutations. This hypothesis was tested using four activating mutations of ␣ L ␤ 2 integrin. The ␤ 2 -G128A/G129T mutation (Fig. 3A) renders ␣ L ␤ 2 activation by facilitating the active conformation of ␤I domain ␣1-helix (19). The ␣ L -F265S mutation induces the downward movement of ␣I ␣7-helix by facilitating the movement of the ␤5-␣6 loop (Fig. 3, A and B) (24), whereas the ␣ L -K287C/K294C mutation directly stabilizes the downward movement of the ␤6-␣7 loop and ␣7-helix of ␣I domain by forming a disulfide bond (Fig. 3, A and C) (15). The ␣ L -GFFKR/GAAKR (FFAA) mutation renders ␣ L ␤ 2 constitutively active by disrupting the ␣-␤ cytoplasmic association, which mimics the inside-out activation of integrin (Fig. 3A) (25). As shown in Fig. 3D, the presence of the ␤ 2 -G128A/G129T Function of the ␣1-helix bulky residue of ␣I domain mutation did not rescue the inactivating effect of ␣ L -F153G and ␣ L -F153A. However, compared with the ␤ 2 wild type (Fig. 2C), the requirement of residue size at the position of ␣ L -153 for Mn 2ϩ -induced ICAM-1 binding was weakened by the ␤ 2 -G128A/G129T mutation (Fig. 3D). With the ␤ 2 -G128A/ G129T mutation, even the ␣ L -F153V rendered the same level of ICAM-1 binding as the ␣ L wild type in the presence of Mn 2ϩ but not Ca 2ϩ /Mg 2ϩ (Fig. 3D). Consistently, the bulkiest amino acid Trp further enhanced the ICAM-1 binding at both metal ion conditions (Fig. 3D). There was ϳ20% decrease of the ␣ L ␤ 2 cell surface expression because of these mutations (Fig. 3D). In sharp contrast, both the ␣ L -F265S and ␣ L -K287C/K294C mutations completely rescued the inactivating effect of ␣ L -F153A (Fig. 3E). Interestingly, the ␣ L -F153A mutation even enhanced the activating effect of ␣ L -K287C/K294C (Fig. 3E). Although the ␣ L -FFAA mutation induced the similar level of ␣ L ␤ 2 activation as the ␣ L -F265S and ␣ L -K287C/K294C mutations, it failed to rescue the inactivating effect of ␣ L -F153A (Fig.  3E). All of these mutations clearly decreased the cell surface expression of ␣ L ␤ 2 integrin (Fig. 3E). These data suggest that the bulky ␣ L -Phe-153 of ␣1-helix is critical for stabilizing the downward displacement of ␣7-helix, required for the high affinity ligand binding of ␣ L ␤ 2 integrin.

The ␣ L -F153A mutation has little effect on the overall conformational change of ␤ 2 integrin
Integrin activation is associated with the long-range conformational rearrangements that are relayed among the connecting domains (5). Such conformational changes on the cell surface can be measured by the exposure of epitopes that are masked in the bent inactive conformation (26). Next, we asked whether the inactivating ␣ L -F153A mutation could affect the global conformational change of ␣ L ␤ 2 integrin. Two conformation-specific mAbs were used to report the conformational states of ␣ L ␤ 2 integrin. The mAb m24 binds to the ␤ 2 ␤I domain and is specific to the open headpiece conformation (27) (Fig.  3A). The mAb KIM127 binds to the I-EGF2 domain of ␤ 2 leg and is specific to the extended conformation of ␤ 2 (28, 29) (Fig.  3A). Mn 2ϩ induced the binding of both m24 and KIM127 to the ␣ L ␤ 2 integrin expressed in HEK293FT cells (Fig. 4, A and B), indicating the headpiece opening and leg extension. The inactivating mutations ␣ L -K149A and ␣ L -F153A decreased ICAM-1 binding by more than 90% (Fig. 2B). By contrast, these mutations had no effect on Mn 2ϩ -induced binding of m24 (

Closed-αI
Open-βI Open-αI The activating ␣ L -FFAA mutation, which mimics the integrin inside-out activation, induced spontaneous binding of m24 and KIM127, which was further enhanced by Mn 2ϩ (Fig. 4, A and B). Although the ␣ L -F153A mutation significantly reduced the binding of both mAbs when combined with the ␣ L -FFAA mutation, it only decreased the mAb binding by 14 or 18% (Fig. 4, A and B), which is in sharp contrast with the almost 100% decrease of ICAM-1 binding (Fig. 3E).
In line with these findings, neither ␣ L -K149A nor ␣ L -F153A affected the binding of m24 and KIM127 in the presence of the activating mutant ␤ 2 -G128A/G129T that stabilizes the active conformation of ␤I ␣1-helix (Fig. 4, A and B). As seen in our previous study on ␤ 1 and ␤ 3 integrins (19), the ␤ 2 -G128A/G129T mutation increased the binding of both m24 and KIM127 compared with the wild type (Fig. 4). These data demonstrated that the inactivating mutations at the ␣I ␣1-helix had little effect on the headpiece opening and extension induced by the signals of outside-in or inside-out activation. This is consistent with the previous study showing that locking the ␣I domain in a high-affinity conformation by the ␣ L -K287C/ K294C mutation did not induce the binding of m24 or KIM127 (22). Thus, the conformational changes between the ␣I and the ␤ 2 subunit can be uncoupled during the transmission of conformational changes. To further define the structure requirement for ␤ 2 headpiece opening and extension, we introduced a ␤ 2 -KKGG mutation at the ␣1-helix of ␤I domain (Fig. 3A). This mutation has been shown in our previous study to dramatically reduced ICAM-1 binding to ␣ L ␤ 2 integrin by stabilizing the inactive conformation of ␣1-helix (19). In opposition to the ␤ 2 -G128A/G129T mutation of ␣1-helix, the ␤ 2 -KKGG mutation greatly impaired the binding of m24 and KIM127 when co-expressed with the ␣ L wild type (Fig. 4). Moreover, it reduced the m24 binding by more than 90% (Fig. 4A), and the KIM127 binding by more than 50% (Fig. 4B), even when combined with the activating ␣ L -FFAA mutation. These data clearly demonstrated that the active conformation of ␤I ␣1-helix; i.e. the inward movement toward the MIDAS (Fig. 1, F and H), is essential for ␤ 2 headpiece opening and extension.

The ␣ L -F153A mutation abolishes ␣ L ␤ 2 -mediated cell adhesion on immobilized ICAM-1
Having established the critical role of ␣ L -F153 in the binding of ␣ L ␤ 2 integrin to soluble ICAM-1, we asked whether it is also Function of the ␣1-helix bulky residue of ␣I domain important in ␣ L ␤ 2 -mediated cell adhesion and spreading on immobilized ICAM-1, given that the receptor-ligand binding kinetics could be different in solution and in the solid phase. When seeded onto the ICAM-1-coated dish, the HEK293FT cells transfected with wild-type ␣ L ␤ 2 or ␣ L -F153Y/␤ 2 integrin spontaneously adhered and spread on the surface (Fig. 5A). However, the ␣ L -F153A mutation completely abolished cell adhesion under the tested coating concentration of ICAM-1 (Fig. 5A), despite its comparable cell surface expression with the wild type (Fig. 5B). This is in agreement with the pivotal role of the bulky residue of ␣ L -F153 in the high-affinity ligand binding of ␣ L ␤ 2 .

The ␣ M -F156A mutation reduces ␣ M ␤ 2 -mediated cell adhesion on immobilized fibrinogen
We next tested whether the conserved Phe residue plays the same role in ␣ M (Mac-1) integrin that binds multiple ligands including fibrinogen. The ␣ M -Phe-156 was mutated to the small amino acid Ala and the bulky amino acid Trp. The ␣ M ␤ 2 -mediated cell adhesion on immobilized human fibrinogen was induced by Mn 2ϩ ions. The ␣ M -F156A mutation significantly reduced the ␣ M ␤ 2 -mediated cell adhesion (Fig. 6A). By contrast, the bulky ␣ M -F156W mutation significantly enhanced the cell adhesion (Fig. 6A). This is not due to the differences in the cell surface expression of ␣ M ␤ 2 because both the ␣ M -F156A and ␣ M -F156W mutations decreased the ␣ M ␤ 2 expression to a similar extent compared with the wild type (Fig. 6B). In addition, the cell adhesion could be blocked by the anti-␣ M inhibitory mAb ICRF44 (data not shown). These data demonstrated that a bulky residue is also required for the ␣1-helix of the ␣ M ␣I domain to support ligand binding. Function of the ␣1-helix bulky residue of ␣I domain affinity ligand binding. To gain a structural insight into how the Phe affects the conformation of ␣I domain, we performed the all-atom molecular dynamics (MD) simulation for wild-type and mutant ␣I domains in water solvent. We used the crystal structure of ␣ L ␣I domain in the open conformation for the MD simulation in line with our hypothesis that the bulky Phe may stabilize the active conformation, whereas a small amino acid substitution like Ala would have the opposite effect. A 60-ns MD simulation was performed for both ␣ L wild type and ␣ L -F153A mutant using exactly the same parameters. To analyze the overall structural similarity of the recorded snapshots to the starting structure, the root mean square deviations (RMSDs) of backbone atoms were calculated. The ␣ L wild type showed stable RMSD values over the 60-ns simulation. However, the RMSD values of the ␣ L -F153A increased substantially after the first 30-ns simulation, indicating the structure instability (data not shown). To evaluate the conformational dynamics for each residue, we calculated the root mean square fluctuation (RMSF, i.e. standard deviation) of C␣ positions in the snapshots relative to the starting structure. As shown in Fig. 7A, most of the residues of the wild-type ␣ L ␣I domain are relatively stable over the 60-ns simulation. Remarkably, in the presence of F153A mutation, the regions of ␤5-␣6 and ␤6-␣7 loops showed significant fluctuations (Fig. 7A). These differences became more obvious when the RMSF values were converted to B-factors and represented as cartoon models (Fig. 7B). We next performed the same MD simulations for the open-conformation crystal structure of ␣ M ␣I domain. The simulation time was extended to 100 ns. Similar to ␣ L -F153A, the ␣ M -F156A mutation rendered the ␤5-␣6 and ␤6-␣7 loops more flexible than the wild type (Fig. 7, C and D), although to a lesser extent compared with the ␣ L ␣I domain (Fig. 7, A-D). These data clearly demonstrated that the bulky Phe residue of ␣1-helix contributes to the high affinity state of ␣I domain by stabilizing the active conformation of ␤5-␣6 and ␤6-␣7 loops.

Discussion
Integrin is an excellent example of protein machinery in which local conformational changes in one site are propagated to a distal site, altering the ligand-binding affinity and function. Structure-based mutagenesis studies have determined certain residues that are critical for the ligand-binding activity of ␤I and ␣I domains (19,22,24,30,31). Most of the studies focused on the residues that either directly participate in the coordination with the metal ion and/or ligand or that are involved in the movements of the loops or ␣-helixes surrounding the active site. The importance of these residues is obvious according to the structural changes because the gain-of-function mutations are predicted to shift the conformation to the open state, whereas the loss-of-function mutations are predicted to shift it to the closed state (15,19,24). However, a functional role of the conserved Phe of ␣I ␣1-helix is unpredictable because this residue seems not directly participate in the movement of either ␣1-helix or ␣7-helix based on the structural comparison between the active and inactive conformations. Thus, our findings of the loss-of-function mutation of ␣ L -Phe-153 are unexpected. Our current study further advanced our understanding of the conformational requirement for integrin ligand binding.
How does the ␣ L -Phe-153 contribute to the ligand binding of ␣ L ␣I domain? It has been indicated that the downward movement of ␣7-helix is required for the high affinity ligand binding of ␣I domain because mutations constraining it at the upward position blocked, whereas mutations facilitating the downward movement enhanced ligand binding (22,23). In addition, small molecule inhibitors bound to the cavity under the ␣7-helix allosterically block ligand binding of ␣ L ␤ 2 by restraining the downward movement of ␣7-helix (32-37). Our mutagenesis study showed that the function of a residue at the position of ␣ L -Phe-153 is determined by its size (or volume). A bulky residue is specifically required to maintain the ligand-binding capability of ␣I domain. At the downward position, the ␤6-␣7 loop moves close to the middle of ␣1-helix (Figs. 1I and 3, B and  C), where a hydrophobic and bulky side chain would act as a pawl to anchor the downward position of ␤6-␣7 loop and ␣7-helix during their ratchet-like movement (15). This was indicated by the MD simulation data showing that the ␣ L -F153A or ␣ M -F156A mutation rendered the ␤5-␣6 and

Function of the ␣1-helix bulky residue of ␣I domain
␤6-␣7 loops less stable in the active conformation (Fig. 7). The bulky ␣ L -Phe-153 is solely required for the downward movement of ␤6-␣7 loop and ␣7-helix but dispensable for the inward movement of ␣1-helix because once the ␤6-␣7 loop is enforced to the downward position, the trimmed ␣1-helix by the ␣ L -F153A mutation still supports ligand binding. Crystal structures revealed the flexibility of ␣7-helix C terminus both in the isolated form of ␣I domain and in the context of integrin ectodomain (38 -41), indicating that the ␣7-helix and the connecting ␤6-␣7 loop can sample rapidly between the closed and the open conformational states. We did not observe the ␤6-␣7 loop and ␣7-helix moving upward to the closed conformation in our MD simulation even in the presence of ␣ L -F153A mutation. This is very likely due to the short time scale used for the simulation. Such a large conformational change may require much longer time of simulation to be observed. Indeed, the hydrogendeuterium exchange kinetics measured via NMR revealed a large conformational fluctuation of ␣7but not ␣1-helix in the resting state of the ␣ 1 integrin ␣I domain (42). The conserved bulky residue in the middle of ␣1-helix is critical to regulate the shifting between the two states of ␣7-helix by facilitating the transition of the ␣I domain to the high affinity conformation. It should be noted that although Leu and Met have very close side chain volumes, the ␣ L -F153L mutation exerted a similar level of ICAM-1 binding as the ␣ L WT, whereas the ␣ L -F153M mutation greatly reduced ICAM-1 binding (Fig. 2C). This is consistent with the numbers of rotamers that Leu and Met have. Leu can only sample four rotamer conformations, which is likely to stabilize the downward position of ␤6-␣7 loop and ␣7-helix. In contrast, Met is more flexible by having more than Function of the ␣1-helix bulky residue of ␣I domain ten rotamer conformations, one of which may adapt to the free movements of ␤6-␣7 loop and ␣7-helix and thus lack the activation supporting function. The ␤I and ␣I domains are the best-studied VWA domains that undergo conformational changes for affinity regulation. Although their ␤6-␣7 loops and ␣7-helixes move in a similar fashion to accommodate for the active position of MIDAS loop and to communicate with the neighboring domains, their ␣1-helixes change in a different fashion (Fig. 1, H and I). This is in accordance with a conserved Gly and a conserved Phe at the homologous position of ␣1-helix, which play opposite roles in the ␤I and ␣I domains. As we found for a conserved negative role of the Gly in the ␤I ␣1-helix (19), the positive role of the Phe in the ␣I ␣1-helix could also be generalized to all the ␣I-containing integrins, given that the same conformational changes of ␣1and ␣7-helix have been observed in the crystal structures of the ␣I domains of ␣ L , ␣ M , ␣ X , ␣ 1 , and ␣ 2 integrins (12-15, 40, 43, 44). This was demonstrated in the current study on ␣ L and ␣ M integrins. Interestingly, an equivalent Phe but not Gly is also present in the ␣1-helix of many VWA domains other than integrins (Fig. 1J). Moreover, most of the VWA domains have only one or lack an intact MIDAS, whereas the integrin ␤I domains have three metal ion-binding sites that regulate ligand binding. Given the different structural features and conformational regulations of ␤I domains, it is tempting to speculate that the ␤I domains might be evolved differently from a typical VWA domain and could be classified as VWA variants.
Although a conserved Phe residue at the equivalent position of ␣I ␣1-helix is found in many non-integrin VWA domains, generalizations for the similar function of this residue as seen in the ␣I domains can only be made with caution. For example, the VWF A1, A2, and A3 domains and the collagen VI ␣3N5 domain all have an equivalent Phe at the ␣1-helix, but they lack an intact MIDAS (Fig. 1J), and no conformational changes were observed in their ␣1and ␣7-helixes (45)(46)(47)(48)(49). Similarly, the presence of an intact MIDAS does not seem to correlate with the presence of the equivalent Phe in the ␣1-helix. Several examples are found in the complement components C2a and factor B, in the thrombospondin repeat anonymous proteins (TRAPs) of the Plasmodium vivax and Plasmodium falciparum, and in the pilus-related adhesion RrgA of Streptococcus pneumoniae, in which a non-Phe residue is present at the equivalent position of ␣1-helix (Fig. 1J). We compared the crystal structures of several representative VWA domains that have an intact MIDAS, which has been suggested or confirmed to participate in ligand binding (Fig. 8). Crystal structures of the VWA domain of human anthrax toxin receptor 2 (ANTR2) in complex with anthrax toxin or pseudo-ligands highly resemble the active open conformation of integrin ␣I domain (Fig. 8A) (50). Although a closed conformation has yet to be seen for the ANTR2 VWA domain, mutagenesis studies suggested that it might undergo conformational changes for affinity regulation and signaling (51). Other VWA domains that have similar conformational regulations as integrin ␣I domains are from the adhesins of the parasites Toxoplasma gondii micronemal protein 2 (TgMIC2) and the P. vivax TRAP or P. falciparum TRAP as mentioned above. The TRAP VWA domain has been seen in both the closed and open conformations (Fig. 8B) (52). The MIC2 VWA domain has only been crystalized in the closed conformation so far (Fig. 8C), but the transition between the closed and the open states is highly predictable (53). A recent crystal structure of the Blue Mussels adhesion protein, the PTMP1 (proximal thread matrix protein 1) revealed two tandem VWA domains both in the closed conformation with the A1 domain MIDAS occupied by Zn 2ϩ (Fig. 8D) (54). It was suggested that the MIDAS of PTMP1 might participate in collagen binding (54), which resembles the interaction of the ␣ 2 integrin ␣I domain and collagen. All the above-mentioned VWA domains contain the equivalent Phe at the ␣1-helix except that some TRAP proteins have a Leu or Met (Figs. 1J and 8, A-D). However, our data showed that a Met or Leu residue at the equivalent position of ␣1-helix could still support the ligand binding of ␣I domain (Figs. 2C and 3D). Therefore, it is tempting to speculate that these VWA domains may follow the activating mechanism of integrin ␣I domains.
We presented two examples of VWA domains from the bacteria adhesin RrgA (Fig. 8E) (55) and the complement factor B (Fig. 8F) (56), which have a small amino acid at the equivalent position of ␣1-helix. The crystal structure of RrgA VWA domain shows a closed conformation of MIDAS and ␣7-helix and an Ala at the Phe-equivalent position of ␣1-helix (Fig. 8E). Notably, the loop connecting the MIDAS motif and the ␣1-helix in RrgA is much longer than that of ␣ L ␣I domain (Figs. 1J  and 8E). This longer MIDAS loop may move freely toward the metal ion to accommodate with the high affinity ligand binding even without the inward movement of the ␣1-helix and the downward movement of the ␣7-helix. As such, a bulky residue in the mid ␣1-helix is not essential. The VWA domain of complement factor B was crystallized in both the closed and open  8F). The movement of ␣7-helix seems to be induced by the binding of factor D (56). In addition, the factor B VWA domain has an extra ␣L-helix, which interacts with the ␣7-helix and influences its conformation (Fig. 8F). All these differences may minimize the requirement of a bulky residue at the ␣1-helix. In summary, our study revealed a unique feature of integrin ␣I ␣1-helix in regulating ligand binding, which may be applicable to certain VWA domains of other proteins that undergo large conformational changes for signal transduction.
Another important finding in the current study is that integrin conformational transmission can be uncoupled between the ␣I and the ␤I domains. We found that although the conserved Phe of ␣1-helix is critical for ICAM-1 binding to ␣I domain, it is dispensable for ␤ 2 extension and headpiece opening. The downward displacement of ␣7-helix that requires the Phe of ␣1-helix is not essential for the conformational changes of ␤ 2 induced by Mn 2ϩ or by inside-out activation. In addition, previous studies also showed that locking the downward conformation of ␣7-helix of ␣I domain did not result in ␤ 2 extension or headpiece opening (22). Interestingly, the internal ligand ␣ L -Glu-310 at the C terminus of ␣7-helix is required for the headpiece opening of ␤ 2 subunit (58). Recent crystal structures of ␣ X ␤ 2 and ␣ L ␤ 2 suggested the intrinsic flexibility of ␣I domain on the platform formed by the ␤-propeller and ␤I domains (39 -41). This is consistent with the loose conformational linkage between the ␤I and ␣I domains. The ␣ L ␤ 2 integrin mediates both the rolling adhesion and migration of leukocytes by binding the ICAM-1 expressed on the endothelial cells (59). Under the rolling condition, the loose linkage between the ␣I and ␤I enables the cells to quickly bind and release ICAM-1. When the cell signals for firm adhesion and migration are turned on, the tight engagement of the ␣I and ␤I domains could be enforced by the mechanical forces generated by the actin polymerization and ICAM-1 binding at each end of integrin (60).

DNA constructs
DNA constructs of human ␣ L and ␤ 2 integrins were as described previously (19). The full-length cDNA of human ␣ M integrin was cloned into the pcDNA3.1 vector using the 5-Ј KpnI and 3-Ј XbaI sites. Mutations were introduced by site-directed mutagenesis with the QuikChange kit (Agilent Technologies).

Conformation-specific antibody binding
Binding of the active conformation-specific anti-␤ 2 mAbs KIM127 and m24 to the HEK293FT transfectants was performed as described before (19,21). In brief, the cells were first incubated with 10 g/ml biotin-labeled KIM127 or m24 in HBSGB buffer containing 1 mM Ca 2ϩ /Mg 2ϩ or 0.2 mM Ca 2ϩ plus 2 mM Mn 2ϩ at 25°C for 30 min and then washed and incubated with 10 g/ml FITC-labeled TS2/4 and Alexa Fluor 647-labeled streptavidin on ice for 30 min. TS2/4-positive cells were acquired for calculating the MFI by flow cytometry. The KIM127 or m24 binding was presented as normalized MFI, i.e. streptavidin MFI as a percentage of TS2/4 MFI.

Cell adhesion assay and fluorescence microscopy
For ␣ L ␤ 2 -mediated cell adhesion assay, the Delta T dishes (Bioptechs) or 24-well cell culture plate (BD Falcon) were first Function of the ␣1-helix bulky residue of ␣I domain coated with 5 g/ml mouse anti-human IgG in pH 7.4, PBS, at 4°C for 12 h, and then blocked with 1% BSA at 37°C for 1 h before finally coated with 5 g/ml ICAM-1-Fc at 37°C for another 1 h. Transfected HEK293FT cells suspended in DMEM without FBS were allowed to adhere on the ICAM-1-Fc-coated Delta T dishes at 37°C for 1 h. The unattached cells were washed off by DMEM, and the attached cells were fixed with DMEM containing 3.7% formaldehyde, blocked with 5% nonfat dry milk in PBS, and stained with 10 g/ml mAb TS2/4 and Alexa Fluor 488-labeled goat anti-mouse IgG. The nuclei were stained with 5 g/ml DAPI. The stained cells were fixed with 3.7% formaldehyde in PBS and imaged with an EVOS digital inverted fluorescence microscope with a 20ϫ objective. The level of integrin expression of the transfected cells was accessed by the mAb TS2/4 staining, followed by flow cytometry. For ␣ M ␤ 2 -mediated cell adhesion assay, the 12-well cell culture plate was coated with 3 mg/ml human fibrinogen in PBS at 37°C for 1 h. HEK293FT cells transfected with ␣ M ␤ 2 plus EGFP were seeded onto the fibrinogen-coated plate in HEPES buffer for 10 min before adding 0.2 mM Ca 2ϩ plus 2 mM Mn 2ϩ . After incubation at 37°C for 1 h, the plate was washed with PBS buffer for three times and fixed with 3.7% formaldehyde in PBS. The adhered cells were imaged with an EVOS digital inverted fluorescence microscope with a 10ϫ objective. At least 20 images were randomly taken for each sample, and the EGFPpositive cells were counted for each image. The data were presented as the averaged cell numbers that were normalized to the WT level. The cell surface expression of ␣ M ␤ 2 was measured by flow cytometry after staining with the allophycocyanin (APC)labeled anti-␣ M mAb ICRF44 (BioLegend).

Molecular dynamics simulation
The crystal structures of the ␣ L ␣I (PDB code 1T0P) and the ␣ M ␣I domains (PDB code 1IDO) in the open conformation were used for MD simulation. ␣ L -Phe-153 and ␣ M -Phe-156 were mutated to Ala in silico using PyMOL. For the ␣ L ␣I structure (PDB code 1T0P), the two cysteine mutations, K287C and K294C, were mutated back to their native Lys residue by PyMOL to remove the engineered disulfide-bond that was used to stabilize the open conformation of ␤6-␣7 loop and ␣7-helix. GRO-MACS version 2016.3 was used to prepare the structures and perform the production MD simulation using the OPLS-AA/L allatom force field (64,65). The protein was centered in a cubic box with a minimum distance to any wall of 1.2 nm. The box was filled with TIP3 water model and 100 mM NaCl. Energy minimization and equilibrations of temperature and pressure were performed before conducting the unrestrained 60-or 100-ns production MD simulation. The RMSDs and RMSFs of the backbone or C␣ atoms were calculated for the recorded trajectories to evaluate the stability of the conformations.

Statistical analysis
The two-tailed Student's t test was performed using the GraphPad Prism software to calculate the p values for comparing the data between two experimental groups.
Author contributions-Z. W. and A. M. M. T. performed the experiments and analyzed the data. J. Z. designed the study, analyzed data, and wrote the manuscript. All authors contributed to the manuscript preparation.