Disruption of disulfide restriction at integrin knees induces activation and ligand-independent signaling of α₄β₇

Integrins are a family of α/β heterodimeric cell adhesion molecules that mediate cell–cell, cell–matrix, and cell–pathogen interactions, and signal bidirectionally across the plasma membrane (Ekblom et al., 1986; Luo et al., 2007). In vertebrates, 18 α-subunits and 8 β-subunits combine to yield 24 distinct integrins, which play vital roles in a wide range of biological and pathological events including immune responses, homeostasis, embryonic development, and diseases such as cancer and autoimmunity (Hynes, 2002). In contrast to most integrins, which only mediate firm cell adhesion upon activation, α₄β₇ mediates both firm cell adhesion when activated and rolling adhesion before activation through its ligand mucosal addressing cell adhesion molecule-1 (MAdCAM-1), making it indispensable in the tissue-specific homing of lymphocytes to intestine and associated lymphoid tissues (Berlin et al., 1993). Dysregulation of α₄β₇-mediated recruitment of lymphocytes to inflamed intestinal tissues has been implicated in the pathogenesis of intestinal inflammatory disorders (Feagan et al., 2005).

The biological functions of integrins rely on dynamic regulation of integrin affinity and signaling. Integrin affinity can be uniquely regulated through a process called inside-out signaling, in which stimuli received by other cell surface receptors initiate intracellular signals that impinge on integrin cytoplasmic domains and alter integrin conformation and ligand-binding affinity (Carman and Springer, 2003). In addition, binding of ligand to integrins can trigger the transduction of extracellular signals into the cytoplasm and activate downstream signals in the classic outside-in signaling (Litvinov et al., 2004; Luo et al., 2007). Moreover, lateral assembly of integrins (integrin clustering) is known to augment integrin-mediated adhesion by increasing the bond valency and also contributes to signal transduction (Miyamoto et al., 1995).

Recent research advances have shed light on the structural basis for integrin affinity regulation and signaling. The entire ectodomain of integrin α_Vβ₃, α_IIbβ₃ and α_Xβ₂ was crystallized in a bent conformation, which is believed to exist on the cell surface and to represent the inactive, low-affinity state (Xie et al., 2010; Xiong et al., 2001; Zhu et al., 2008). A number of negative staining electron microscopy and mutagenesis studies indicate that integrin activation is accompanied by a switchblade-like opening of the ectodomain during which the bent conformer undergoes a large angle rotation at the knee fulcrum and becomes extended to expose the ligand binding site on top of the integrin headpiece for high-affinity ligand binding (Takagi et al., 2002). However, other studies have indicated that such large-scale conformational change was not always required to render integrins competent to bind physiological ligands (Butta et al., 2003; Calzada et al., 2002; Chigaev et al., 2001; Xiong et al., 2003). Compelling evidence for this is the observation that the bent conformer of α_IIbβ₃ ectodomain in solution is able to form a stable complex with its physiological ligand (Adair et al., 2005). These studies suggest a level of complexity in the relationship between the conformation and affinity of integrins.

All integrins contain a large number of cysteine residues in their ectodomains. Previous studies have shown that breakage of some disulfide bonds in integrins by reducing agents or mutagenesis can induce the activation of integrins (Mor-Cohen et al., 2012; Yan and Smith, 2001). However, the underlying mechanism remains elusive. Crystal structures of the ectodomain of integrins α_IIbβ₃, α_Vβ₃ and α_Xβ₂ revealed two disulfide bonds located at the apex of the knees, with one in the knee of the α-subunit and the other in the β-subunit I-EGF2 domain (Xie et al., 2010; Xiong et al., 2009; Zhu et al., 2008) (Fig. 1A–C). Sequence alignment demonstrated that these two disulfide bonds (α₄C589–C594 and β₇C494–C526) are highly conserved between all integrins (Fig. 1D,E). Considering that they are exposed to solution and are easily accessed by reducing agents, it is tempting to speculate that these two disulfide bonds could be disrupted under certain reducing conditions and that loss of the disulfide restriction might regulate integrin function.

Here, we demonstrate that breaking either or both α₄C589–C594 and β₇C494–C526 induces the activation of α₄β₇ by increasing integrin ligand-binding affinity. Surprisingly, the highly activated α₄β₇ that lacks the restriction of these disulfide bonds exhibits a unique conformation that is different from the characteristic extended conformation of integrin that is activated by inside-out signaling or Mn²⁺. In addition, breaking these disulfide bonds in α₄β₇ increases phosphorylation of focal adhesion kinase (FAK) and paxillin, and promotes cell spreading on immobilized poly-L-lysine, indicating the activation of integrin outside-in signaling independent of ligand binding, possibly as a result of enhanced integrin clustering before ligand binding. Moreover, these disulfide bonds can be specifically reduced by 0.1 mM dithiothreitol (DTT), although they were not found to be reduced during the activation of α₄β₇ by inside-out signaling or Mn²⁺. Thus, our findings suggest a novel mechanism of integrin affinity and signaling regulation through reduction of two conserved disulfide bonds at the knees of integrin. The resulting activated α₄β₇, with its unique conformation, can spontaneously cluster on the cell membrane and trigger integrin outside-in signaling and cell spreading independent of ligand binding.

To date, there is no published experimental structure of the full-length integrin α₄β₇ ectodomain. Thus, we constructed the full-length ectodomain of human α₄β₇ in bent conformation by using homology modeling technique (Fig. 1A). The head domain was modeled after the crystal structure of human α₄β₇ headpiece (pdb3V4P), which consists of β-propeller and thigh domains of the α₄ subunit and the βI and hybrid domains of the β₇ subunit (Yu et al., 2012). The other domains were built based on their counterparts from the crystal structure of human α_Vβ₃ (pdb3IJE). Sequence alignments reveal that two disulfide bonds at the knee region of integrin α₄β₇ (α₄C589–C594 and β₇C494–C526) are highly conserved among all integrins (Fig. 1D,E). The disulfide bond α₄C589–C594 occludes the α-subunit knee domain (Fig. 1B) and β₇C494–C526 is the first disulfide bond in the βI-EGF2 domain (Fig. 1C). To investigate the role of these disulfide bonds in the regulation of integrin α₄β₇ function, we substituted α₄C589, α₄C594, β₇C494 and β₇C526 with Ser and generated α₄ mutations [α₄C589S, α₄C594S, α₄C589SC594S (α₄CC)], β₇ mutations [β₇C494S, β₇C526S, β₇C494C526S (β₇CC)] and a α₄β₇ double mutation (α₄CC/β₇CC) to break each or both of the two disulfide bonds. CHO-K1 cell lines stably expressing either wide-type (WT) human integrin α₄β₇ or each of the above mutants were established (supplementary material Table S1). Adhesive behavior of these transfectants in shear flow was characterized in a parallel wall flow chamber by allowing the cells to adhere to human MAdCAM-1 adsorbed to the lower wall. The velocity of the cells remaining bound at a wall shear stress of 1 dyn/cm² was determined (Fig. 2A). Rolling and firm adhesion represents the low- and high-affinity interaction of MAdCAM-1 with inactive and activated α₄β₇, respectively. WT α₄β₇ CHO-K1 transfectants behaved as previously described for lymphoid cells expressing α₄β₇ (Chen et al., 2003). In 1 mM Ca²⁺ and 1 mM Mg²⁺ (1 mM Ca²⁺/Mg²⁺), ∼80% of the bound α₄β₇ transfectants rolled on MAdCAM-1 substrates (Fig. 2A). By contrast, the majority of cells were firmly adherent after activation of integrin by 0.5 mM Mn²⁺ (Fig. 2A). α₄β₇ transfectants treated with the α₄β₇ blocking antibody Act-1 or 5 mM EDTA did not accumulate on MAdCAM-1 substrates (Fig. 2A), indicating the specificity of the interaction between α₄β₇ and MAdCAM-1. In contrast to the robust rolling cell adhesion mediated by WT α₄β₇ in 1 mM Ca²⁺/Mg²⁺, all mutations that break the two disulfide bonds significantly increased the number of firmly adherent cells, demonstrating that α₄β₇ is activated by breaking either or both of the two disulfide bonds (Fig. 2A). Therefore, the two disulfide bonds at the knees of integrin are essential for keeping integrin α₄β₇ inactive.

To further study the binding characteristics of WT α₄β₇ and α₄CC/β₇CC mutant to MAdCAM-1, we purified soluble α₄β₇ proteins that contain the full-length ectodomain as described previously (Qi et al., 2012; Zhu et al., 2008) and measured the binding of WT α₄β₇ and α₄CC/β₇CC proteins to MAdCAM-1 using surface plasmon resonance. In the presence of 1 mM Ca²⁺/Mg²⁺, soluble integrins bound to immobilized MAdCAM-1 in a concentration-dependent manner (Fig. 2B). Compared with WT α₄β₇, α₄CC/β₇CC showed increased binding to MAdCAM-1 (Fig. 2B), suggesting that breaking these disulfide bonds increases integrin ligand-binding affinity. As a positive control, addition of 0.5 mM Mn²⁺ maximally increased the binding of WT α₄β₇ (Fig. 2B). The binding curves fitted well to a two-step binding model (supplementary material Table S2) as described before (Takagi et al., 2002). For simplicity, we reported here apparent association and disassociation rate constants derived from the initial phase of the association and dissociation curves, respectively. The k_on values for WT α₄β₇ and α₄CC/β₇CC were 0.871×10⁵ M⁻¹ second⁻¹ and 3.13×10⁵ M⁻¹ second⁻¹, respectively, and the k_off values were 5.31×10⁻³ second⁻¹ and 1.43×10⁻³ second⁻¹. The calculated binding affinity of integrin to MAdCAM-1 is 2.71 nM for WT α₄β₇ and 0.628 nM for the α₄CC/β₇CC mutant. Thus, disruption of the two disulfide bonds affects both association and dissociation rate, resulting in an overall 4.32-fold increase in the binding affinity of α₄β₇ to MAdCAM-1.

Metal-ion binding sites have been shown to play important roles in the regulation of integrin affinity (Chen et al., 2003; Xie et al., 2004; Zhang and Chen, 2012). There is a conserved Ca²⁺ binding site in the loop occluded by α₄C589–C594 in the α₄ knee (Xiong et al., 2001); therefore, breaking this disulfide bond could change the orientation of the loop that contains the Ca²⁺-coordinating residues (C589, E592 and D635) and affect Ca²⁺ binding. To exclude the possibility that integrin activation induced by breaking the disulfide bond is a result of the loss of Ca²⁺ binding, we disrupted the Ca²⁺ binding site by mutating E592 and D635 to Gly and Ala, respectively. CHO-K1 cell lines stably expressing α₄E592G, α₄D635A or α₄E592GD635A (α₄ED) mutants were established (supplementary material Table S1). Cells expressing WT and mutant α₄β₇ integrin showed similar adhesion behavior in shear flow on MAdCAM-1 substrates (Fig. 2C), demonstrating that abolishment of Ca²⁺ binding does not activate α₄β₇.

The α₄C589–C594 and β₇C494–C526 disulfide bonds occlude 4- and 13-amino-acid loops, respectively. It is tempting to speculate that disruption of the two disulfide bonds increases the flexibility of the integrin knee region, which might facilitate the extension of the integrin ectodomain and allosterically activate α₄β₇. To test our hypothesis, we individually deleted the two loops to reduce flexibility at the knees of α₄β₇, which was expected to keep integrin inactive. Unexpectedly, deletion of either the α₄ 4-amino-acid loop (α₄Δ4) or the β₇ 13-amino-acid loop (β₇ I-EGF2 C1–C2 loop; β₇Δ13) markedly increased the number of firmly adherent cells on MAdCAM-1 substrates (Fig. 2C), indicating the activation of α₄β₇. Thus, integrin activation induced by breaking the two disulfide bonds at the knees of α₄β₇ is not simply due to the change of flexibility at the knee region.

Upon activation, integrin converts from the low-affinity bent conformation to the high-affinity extended conformation. A previous negative staining EM study indicated that the complete α₄β₇ ectodomain in 1 mM Ca²⁺/Mg²⁺ had a compact rather than extended conformation (Yu et al., 2012). Extension of the integrin ectodomain has been shown to decrease the retention volume of purified integrin proteins in gel filtration because of an increase in the hydrodynamic radius of integrin molecules (Takagi et al., 2002; Xiong et al., 2009). To investigate the conformational change induced by disruption of these disulfide bonds, we compared the isocratic elution profiles of purified WT α₄β₇ and α₄CC/β₇CC mutant with full-length ectodomain on a molecular sieve chromatography column. WT α₄β₇ was eluted at 10.84 ml in 1 mM Ca²⁺/Mg²⁺, whereas its elution volume in 0.5 mM Mn²⁺ decreased to 10.65 ml, which is consistent with the more extended conformation of Mn²⁺-activated α₄β₇ (Fig. 3A). Surprisingly, the α₄CC/β₇CC active mutant showed a larger retention volume of 10.98 ml in Ca²⁺/Mg²⁺. Thus, our results suggest that breaking these disulfide bonds induces a relatively compact instead of the typical extended active ectodomain of α₄β₇ (Fig. 3A).

To demonstrate that this unique conformation exists in α₄CC/β₇CC on the cell surface, we used fluorescence resonance energy transfer (FRET) to examine the conformational changes of α₄β₇. To assess the orientation of α₄β₇ ectodomain relative to the plasma membrane, β₇ I domain was labeled with Alexa-Fluor-488-Act-1 Fab as donor, and the outer leaflet of cell membrane was labeled with FM4-64-FX as acceptor (Pan et al., 2010). The FRET efficiency of α₄CC/β₇CC transfectants was significantly higher than that of WT α₄β₇ transfectants when cells adhered to poly-L-lysine (Fig. 3B), indicating that the β₇ I domain in α₄CC/β₇CC was closer to the cell membrane before ligand binding than that in WT α₄β₇. Binding of MAdCAM-1 to α₄β₇ significantly decreased the FRET efficiency of both WT α₄β₇ and α₄CC/β₇CC transfectants, suggesting that the integrin head domain stands away from the cell membrane (Fig. 3B). However, the FRET efficiency of α₄CC/β₇CC transfectants was still substantially higher than that of WT α₄β₇ transfectants, even after ligand occupancy (Fig. 3B). Activation of integrin by 0.5 mM Mn²⁺ maximally decreased the FRET efficiency of both WT α₄β₇ and α₄CC/β₇CC transfectants to a similar level (Fig. 3B), suggesting full extension of α₄β₇ ectodomain. To further confirm that breaking the two disulfide bonds does not induce the typical conformational change in integrin α₄β₇ on the cell surface, we examined the epitope expression of J19, a monoclonal antibody (mAb) that recognizes an epitope located in the α₄β₇ head domain and only expressed in the extended α₄β₇ (Qi et al., 2012) (Fig. 3C). The conformational change in α₄β₇ upon activation by chemokine and Mn²⁺ could be well reported by J19 binding (Qi et al., 2012). Consistent with the FRET results, both WT α₄β₇ and α₄CC/β₇CC mutant transfectants were stained with J19 only after stimulation with 0.5 mM Mn²⁺ but not in 1 mM Ca²⁺/Mg²⁺ (Fig. 3C), suggesting that the α₄CC/β₇CC mutant does not have a typical active integrin conformation in Ca²⁺/Mg²⁺. Collectively, these findings show that breaking the two disulfide bonds at the knees of α₄β₇ induces a unique active conformation that is different from the typical extended conformation of active integrin.

Integrin activation is reported to be coupled with the separation of the α- and β-cytoplasmic domains (Kim et al., 2003). We next used FRET to investigate whether the activation of integrin induced by breaking the disulfide bonds was accompanied by separation of integrin cytoplasmic tails. Monomeric mutants of cyan fluorescent protein (mCFP) and yellow fluorescent protein (mYFP) were fused to the C-termini of integrin α₄ and β₇ subunits to act as donor and acceptor, respectively (Kim et al., 2003; Pan et al., 2010). Addition of C-terminal mCFP and mYFP did not affect the binding of WT and mutant α₄β₇ to MAdCAM-1. Interestingly, cells expressing α₄CC/β₇CC showed the same FRET efficiency as WT α₄β₇-expressing cells in all conditions (Fig. 3D). Therefore, activation of α₄β₇ induced by breaking the two disulfide bonds at the knees of integrin is unique because it does not induce the separation of α₄ and β₇ cytoplasmic domains, a typical conformational change coupled with integrin activation.

To examine the influence of the two disulfide bonds on integrin outside-in signaling, we studied α₄β₇-mediated cell spreading. α₄β₇ CHO-K1 stable transfectants were allowed to adhere to immobilized poly-L-lysine or MAdCAM-1 in serum-free F12 medium for 2 hours, followed by fixation and microscopic analysis (Fig. 4A). WT α₄β₇ transfectants spread substantially on MAdCAM-1 substrates. Interference reflection microscopy showed an irregular shape and extensive areas of cell-substrate contact (Fig. 4A,B). By contrast, the same cells did not spread on immobilized poly-L-lysine, and exhibited the same area of projection as cells in suspension (Fig. 4A,B). Interestingly, α₄CC/β₇CC mutant transfectants showed obvious cell spreading and irregular shape of cell-substrate contact even on poly-L-lysine (Fig. 4A,B), suggesting activation of integrin outside-in signaling. We further investigated the activation of FAK and paxillin, two downstream molecules of integrin outside-in signaling (Turner, 2000), by measuring their phosphorylation (Fig. 4C–E). Compared with WT cells, the expression and phosphorylation of FAK and paxillin were significantly elevated in α₄CC/β₇CC mutant cells adhered to immobilized poly-L-lysine, suggesting the activation of integrin downstream signaling independent of ligand binding (Fig. 4C–E). These data strongly suggest that breaking the two disulfide bonds can trigger integrin outside-in signaling in a manner that is independent of ligand binding. In addition, cells that adhered to MAdCAM-1 substrates showed further increased expression and phosphorylation of FAK and paxillin as a result of enhanced integrin outside-in signaling triggered by ligand binding (Fig. 4C–E).

Although it is believed that integrin outside-in signaling is dependent on its global conformational change and ligand binding (Legate et al., 2009; Takagi et al., 2002), the above results showed that breaking the two disulfide bonds at the knees of α₄β₇ could trigger outside-in signaling in the absence of integrin global conformational changes and ligand binding. To address the underlying molecular mechanisms, we investigated the overall capacity of WT α₄β₇ and α₄CC/β₇CC integrins to undergo cell surface redistribution and clustering (Fig. 5), because lateral diffusion and clustering of integrins can contribute to rearrangement of the cytoskeleton and transduction of outside-in signaling (Cluzel et al., 2005). In α₄CC/β₇CC CHO-K1 transfectants adhered to immobilized poly-L-lysine, numerous α₄β₇ microclusters were observed at the cell periphery (Fig. 5A), whereas no discernible clusters of α₄β₇ were observed in cells expressing WT α₄β₇ on immobilized poly-L-lysine (Fig. 5A). In cells adhered to MAdCAM-1, both WT and mutant integrins formed clusters at the cell periphery (Fig. 5B). Moreover, α₄CC/β₇CC formed many more integrin clusters underneath the main cell body than WT (Fig. 5B), which might account for the higher phosphorylation levels of FAK and paxillin in α₄CC/β₇CC transfectants than in WT α₄β₇-expressing cells adhered to MAdCAM-1 (Fig. 4C–E).

Next, we addressed whether α₄CC/β₇CC clustering is induced by linkage of integrins to the cytoskeleton. Integrins have no intrinsic actin-binding sites, and their linkage to the cytoskeleton relies on the binding of scaffold proteins (Legate et al., 2009). Therefore, we generated α₄β₇ constructs that could not link to the cytoskeleton by deleting regions of the cytoplasmic domains that contain binding sites for scaffold proteins in both α₄ and β₇ subunits of WT α₄β₇ (α₄Δ976/β₇Δ736) and the α₄CC/β₇CC mutant (α₄CCΔ976/β₇CCΔ736) (Fig. 5C) (Legate et al., 2009; O'Toole et al., 1994). α₄Δ976/β₇Δ736 and α₄CCΔ976/β₇CCΔ736 were transiently expressed in CHO-K1 cells and their distribution in cells adhered to immobilized poly-L-lysine was examined (Fig. 5C). Truncation of the cytoplasmic domains in WT α₄β₇ did not change the even distribution of α₄β₇ on the cell surface. By contrast, although the capability to support cell spreading on poly-L-lysine was lost due to the disrupted linkage of integrin to the cytoskeleton, α₄CC/β₇CC with truncated cytoplasmic domains (α₄CCΔ976/β₇CCΔ736) still formed microclusters on the cell surface, suggesting that α₄CC/β₇CC integrins can spontaneously cluster on the cell surface independent of linkage to the cytoskeleton. To further confirm that breaking the two disulfide bonds can induce integrin clustering, and to exclude any effect of poly-L-lysine, we investigated the distribution of α₄β₇ on the cell surface when cells were in suspension (Fig. 5D). Consistently, integrin clusters were detected in α₄CC/β₇CC transfectants, but not in WT α₄β₇ transfectants. Notably, CHO-K1 cells expressing α₄CC/β₇CC mutant are basically round in suspension but the shape is less regular than WT cells (Fig. 5D). We speculate that the clustered α₄CC/β₇CC might induce cytoskeleton rearrangement to some extent, resulting in slight cell shape change and triggering cell spreading that is independent of ligand binding. Collectively, these results clearly demonstrate that breaking these disulfide bonds can spontaneously induce α₄β₇ clustering independent of ligand binding, resulting in increased integrin avidity and contributing to rearrangement of the cytoskeleton and transduction of outside-in signaling.

To study whether reduction of the two disulfide bonds at the knees of α₄β₇ is involved in α₄β₇ activation, we examined free cysteine residues in α₄β₇ before and after activation by different stimuli, including talin overexpression, phorbol-12-myristate-13-acetate (PMA) stimulation and Mn²⁺ treatment. Talin has been reported to activate integrins by binding to integrin cytoplasmic domains (Critchley and Gingras, 2008). PMA is a phorbol ester that can activate integrins through the protein kinase C (PKC) pathway (Banno and Ginsberg, 2008). Both agents activate integrins by inside-out signaling. By contrast, Mn²⁺ activates integrin independent of cytoplasmic signaling. α₄β₇ mutant (α₄/β₇C526S) that contains one free cysteine residue in the β₇ subunit was used as a sensitivity control for the detection of single free sulfhydryl. In addition, 5 mM DTT was used to maximally reduce the disulfide bonds in integrins.

To detect free cysteine residues in α₄β₇, we stained free sulfhydryls and α₄β₇ on the surface of α₄β₇ and α₄/β₇C526S 293T transient transfectants with Alexa-Fluor-488-maleimide and anti-β₇ mAb FIB27, respectively. DTT treatment resulted in robust Alexa Fluor 488 signals on the cell surface, some of which colocalized with integrin β₇, suggesting the reduction of disulfide bonds in integrins after DTT treatment (Fig. 6A,B). α₄/β₇C526S transfectants showed clear Alexa Fluor 488 signals with strong colocalization with β₇, indicating that this system is sufficiently sensitive to detect a single free sulfhydryl (Fig. 6A,B). However, there were no detectable free sulfhydryls on WT α₄β₇ transfectants on immobilized poly-L-lysine or MAdCAM-1 before and after activation of integrin by PMA, talin or Mn²⁺ (Fig. 6A,B). These data suggest that no disulfide bond is reduced in integrins activated by either inside-out signaling or Mn²⁺. To confirm the above results, we used biotin-BMCC to conjugate biotin to the exposed free sulfhydryls in proteins on the surface of WT and α₄/β₇C526S mutant α₄β₇ 293T transfectants. Integrin α₄β₇ was then immunoprecipitated and the biotinylated sulfhydryls were visualized by probing with avidin-HRP (Fig. 6C). Consistently, only the β₇ subunit from α₄/β₇C526S and α₄ and β₇ subunits from DTT-treated α₄β₇ were biotinylated and detected by avidin-HRP, suggesting that free cysteine residues only exist in these proteins, but not in WT α₄β₇ before and after activation by inside-out signaling or Mn²⁺ (Fig. 6).

To confirm the above results in a more physiological system, three kinds of chemokines were used to activate α₄β₇ in freshly isolated human peripheral blood lymphocytes (PBLs) (Laudanna et al., 2002). Consistently, free cysteine residues were detected in PBLs only after treatment with DTT, but not before or after activation by SDF-1α, CCL21, CCL25 or Mn²⁺ (supplementary material Fig. S1). These results demonstrate that no integrin disulfide bonds are reduced before or after activation of α₄β₇ by either inside-out signaling or Mn²⁺, suggesting that the two disulfide bonds at the knees of α₄β₇ are not reduced under normal activating conditions.

Although the two disulfide bonds at the knees of α₄β₇ remain intact during integrin activation by inside-out signaling or Mn²⁺, they might be disrupted under certain reducing conditions that are produced in vivo during particular physiological and pathological processes (Alderete and Provenzano, 1997). To determine the minimal reducing strength required for the reduction of these two disulfide bonds in α₄β₇, we examined the free cysteine residues in WT α₄β₇ and α₄CC/β₇CC mutant after treatment with serial concentrations of DTT (Fig. 7A). After treatment with 0.1 and 0.2 mM DTT, we detected biotinylated sulfhydryls only in α₄ and β₇ subunits from WT α₄β₇, but not in the α₄CC/β₇CC mutant that lacks the two disulfide bonds at the knees of α₄β₇ (Fig. 7A). These data suggest that only these two disulfide bonds in α₄β₇ are selectively reduced by 0.1 and 0.2 mM DTT.

Next, we examined whether treatment of WT α₄β₇ with DTT could recapitulate the activation and outside-in signaling mediated by the α₄CC/β₇CC mutant. As expected, treatment of WT α₄β₇ 293T transfectants with 0.1 and 0.2 mM DTT significantly increased the number of cells firmly adhered to immobilized MAdCAM-1, indicating activation of integrin (Fig. 7B). In contrast to the decreased FRET between β₇ I domain and cell membrane (Fig. 7C) and between α₄ and β₇ cytoplasmic tails (Fig. 7D) in α₄β₇ activated by Mn²⁺, α₄β₇ activated by 0.1 and 0.2 mM DTT showed increased FRET between the β₇ I domain and cell membrane and unchanged FRET between α₄ and β₇ cytoplasmic tails, suggesting that α₄β₇ activated by 0.1 and 0.2 mM DTT has a compact conformation with clasped α₄ and β₇ cytoplasmic tails (Fig. 7C,D). Consistent with the results obtained using α₄CC/β₇CC transfectants, treatment of WT α₄β₇ CHO-K1 transfectants with 0.1 and 0.2 mM DTT resulted in cell spreading on immobilized poly-L-lysine with formation of remarkable α₄β₇ clusters (Fig. 7E). Moreover, phosphorylation levels of FAK and paxillin were elevated (Fig. 7F,G).

To further confirm that α₄C589–C594 and β₇C494–C526 were selectively reduced by 0.1 mM DTT and their biological roles are unique, we disrupted each of the other eight disulfide bonds that could be exposed to solution and examined the effects of these mutations on integrin α₄β₇-mediated cell adhesion under flow conditions (Fig. 8A). Among the eight disulfide bonds, only disruption of C459–C478 or C469–C481 in β₇ subunit significantly increased the number of firmly adherent cells, indicating the activation of integrin α₄β₇ (Fig. 8A). However, neither of these two mutations could induce integrin clustering and cell spreading on immobilized poly-L-lysine as did the 0.1 mM DTT treatment or disruption of α₄C589–C594 or β₇C494–C526 (Fig. 8B). Thus, only the two disulfide bonds at integrin knees, α₄C589–C594 or β₇C494–C526, could be selectively reduced by 0.1 mM DTT. Taken together, these data suggest a novel mechanism for integrin activation and signaling by selectively reducing the two disulfide bonds at the knees of α₄β₇ under certain reducing environments.

A major finding of our present study is that integrin α₄β₇ can be activated by selectively breaking two conserved disulfide bonds, α₄C589–C594 and β₇C494–C526, located at the knees of integrin. Interestingly, integrin activated by this mechanism has a unique active conformation that is different from the global conformation induced by Mn²⁺ stimulation. In addition, activated α₄β₇ integrin can spontaneously cluster on the cell membrane and trigger integrin outside-in signaling independent of ligand binding.

All integrins contain a large number of disulfide bonds that are generally believed to facilitate protein folding and stabilize three-dimensional structures (Calvete et al., 1991). The two disulfide bonds at the knees of integrin α₄β₇ are exposed to solution, making them easily accessible to reducing agents. In this study, these two disulfide bonds were selectively reduced by 0.1 mM DTT, which induced the activation of integrin α₄β₇ and triggered outside-in signaling. Accumulating evidence indicates that reducing environments are produced and play essential roles in some physiological and pathological processes. Dendritic cells, monocytes, macrophages and B lymphocytes have been shown to release free thiols and oxidoreductase to generate and maintain reducing environments, which have been demonstrated to be essential for immune cell activation and successful antigen presentation (Angelini et al., 2002; Castellani et al., 2008). In addition, a reducing condition equivalent to 40 µM to 1 mM DTT has been reported in pathogen invasion processes, such as vaginal infection by Trichomonas vaginalis (Alderete and Provenzano, 1997). Such conditions should be capable of inducing α₄β₇ activation by reducing the two disulfide bonds at the knees of integrin. Thus, integrin α₄β₇ and its downstream signaling could be activated by this mechanism in vivo to promote ligand-independent cell spreading under particular reducing environments.

It is important to note that breaking the two conserved disulfide bonds at knees of α₄β₇ triggered the activation of integrin downstream signaling (Fig. 4; Fig. 7E–G) without the extension of α₄β₇ ectodomain and separation of α₄ and β₇ cytoplasmic tails (Fig. 3B–D and Fig. 7C,D), suggesting that ectodomain extension and tail separation are not absolutely required for the activation of integrin downstream signaling. Also, breakage of the two disulfide bonds induced spontaneous α₄β₇ clustering independent of ligand binding and cytoskeleton linkage (Fig. 5). Thus, it is tempting to speculate that some particular interfaces might be formed in this unique activated α₄β₇ with compact structure, which can mediate integrin clustering through interactions between integrin ectodomains and subsequently trigger integrin downstream signaling.

The importance of disordered regions in the regulation of protein allostery has been increasingly recognized (Hilser and Thompson, 2007). Our study showed that deletion of either the α₄ knee (α₄Δ4) or the β₇ I-EGF2 C1–C2 loop (β₇Δ13) strongly activated integrin α₄β₇ (Fig. 2C). The head domains of α₄Δ4β₇ and α₄β₇Δ13 mutants were much closer to the cell membrane than that of WT α₄β₇ both before and after ligand occupancy, suggested by the increased FRET efficiencies between mutant α₄β₇ head and plasma membrane compared with those of the WT (supplementary material Fig. S2A). Furthermore, the cytoplasmic tails of α₄Δ4β₇ and α₄β₇Δ13 mutants were more difficult to dissociate upon MAdCAM-1 binding, as reflected by the slightly higher intracellular FRET efficiencies relative to those of the WT (supplementary material Fig. S2B). Previous studies on α_Vβ₃ crystals revealed that the β₃ I-EGF2 C1–C2 loop forms an important interface with the α_V thigh domain to stabilize integrin in an inactive bent state, and that deletion of this loop induced constitutive activation of α_Vβ₃ (Xiong et al., 2009); Benoit and colleagues detected similar activation of both α_IIbβ₃ and α_Vβ₃ by shortening the β₃ I-EGF2 C1–C2 loop (Smagghe et al., 2010) and advanced a new scenario whereby the β₃ knee functions as an entropic spring to adjust integrin conformation equilibrium between bent and extended conformations at different set points. Partially consistent with these studies, we observed robust activation of integrin α₄β₇ upon deletion of the β₇ I-EGF2 C1–C2 loop (α₄β₇Δ13). However, β₇ C1–C2 loop deletion resulted in a compact conformation instead of the extended conformations of α_IIbβ₃ and α_Vβ₃ induced by shortening the β₃ C1–C2 loop. Considering that α₄β₇ is unique among most integrins for its ability to mediate both rolling and firm cell adhesion (Berlin et al., 1993; Hamann et al., 1994), our results suggest that α₄β₇ has a different regulatory mechanism for integrin conformation to support the unique two-phase cell adhesion.

In conclusion, our findings reveal a novel mechanism of integrin activation under specific reducing conditions by which integrin can signal and promote cell spreading in the absence of ligand in some particular physiological and pathological processes. Moreover, this study suggests the possibility of integrin activation and signaling in the absence of global conformational changes.

The full-length human integrin α₄β₇ ectodomain in a bent conformation was constructed by modeling the crystal structure of α₄β₇ headpiece (pdb3V4P) and building the legs homologically using the counterparts in the structure of α_Vβ₃ (pdb3IJE) with Schrödinger software. This initial model was processed using the following steps. Molecular dynamics simulation was performed with Gromacs4.5.3. To simulate the solvent environment, the system was placed in the center of a rectangular water box. The closest distance from the protein to the water box was 10 Å. Na⁺ and Cl⁻ ions were added to neutralize the system. The resulting system consisted of ∼240,000 atoms. A Gromos 0653A6 force field was used for α₄β₇ (Oostenbrink et al., 2004). The simple point charge model was used for water. First, the system was energy-minimized for 5000 conjugate-gradient steps with all heavy atoms of the protein constrained with a force constant of 1000 kJ/mol.nm². Then the system was equilibrated for 2 nseconds with all heavy atoms of the protein constrained followed for another 2 nsecond simulation, with Ca²⁺ constrained under 1 atm at 300 K. Finally a 1.0 µsecond equilibration MD was carried out. For the simulation, periodic boundary condition was imposed. Bonds involving hydrogen were set to be rigid with LINCS (Hess, 2008) algorithm for protein and SHAKE (Kollman, 1992) algorithm for water, thus permitting use of an integration step of 2 fseconds. Non-bonded forces were cut off at 14 Å and electrostatics were treated by utilizing the Particle Mesh Ewald method (Kollman, 1992). For the NPT simulation, the Berendsen weak coupling method (Berendsen, 1991) was applied with coupling constants of 0.1 pseconds and 0.5 pseconds respectively, to maintain the system at 300K and 1 atm. Images of trajectory were recorded every 200 picoseconds. The last 2000 images were used to obtain the time-averaged model used in this paper.

mAb goat anti-rat IgG-Alexa-Fluor-488 was from Invitrogen. mAb against paxillin and β₇ integrin (FIB27) were from BD Biosciences. pY118-paxillin antibody was from Cell Signaling. mAb against FAK and pY397-FAK were from Upstate Biotechnology. mAb J19 human IgG was prepared as described (Qi et al., 2012). mAb FIB504 against human β₇ was prepared from hybridoma (Developmental Studies Hybridoma Bank). Human MAdCAM-1/Fc fusion protein (MAdCAM-1), and Act-1 mAb specific for α₄β₇ were as previously described (Tidswell et al., 1997). Complete protease inhibitor cocktail tablets and PhosSTOP phosphatase inhibitor cocktail tablets were from Roche. Chemokines were from R&D.

The full-length α₄β₇ ectodomain, consisting of β-propeller, thigh, knee, calf-1, calf-2 domains in α₄ subunit and βI domain, hybrid, PSI, I-EGF1-4, β-Tail domains in β₇ subunit, was purified as previously described (Qi et al., 2012; Yu et al., 2012; Zhu et al., 2008).

The flow chamber assay was performed as described (Chen et al., 2003). A velocity of 1 µm/second was the minimum velocity required to define a cell as rolling instead of firmly adherent.

Immunofluorescence flow cytometry was carried out as described (Chen et al., 2006). Expression levels of integrin α₄β₇ were determined by staining with FIB504, and followed by staining with goat anti-rat IgG-Alexa-Fluor-488. Before staining with J19 human IgG, cells were washed with HBS buffer with 5 mM EDTA, and then resuspended in HBS buffer containing either 1 mM Ca²⁺/Mg²⁺ or 0.5 mM Mn²⁺. Stained cells were then measured using FACS Calibur (BD Biosciences) and analyzed using WinMDI 2.9 software.

All molecular sieve chromatography analyses were performed as previously described (Adair et al., 2005) using pre-calibrated Superdex 200 (10/300 GL) columns on an ÄKTA purifier system running Unicorn 5.01 software (GE Healthcare) at a flow rate of 0.5 ml/minute at room temperature. The elution profiles were monitored in-line by UV adsorption at 280 nm. HBS buffer containing either 1 mM Ca²⁺/Mg²⁺ or 0.5 mM Mn²⁺ was used throughout. 6 µg of purified WT or mutant integrin α₄β₇ was injected.

Experiments were performed using the BIAcore T100. Dilutions of WT or mutant α₄β₇ in HBS containing either 1 mM Ca²⁺/Mg²⁺ or 0.5 mM Mn²⁺ were injected over 3 minutes at 30 µl/minute into the flow cell containing 720 RU of MAdCAM-1 coupled to a sensor chip CM5. After each cycle of association and dissociation, the surface was regenerated by injecting 30 µl of regeneration buffer containing 20 mM EDTA and 25 mM NaOH. This treatment did not affect the subsequent binding reaction after up to 60 repetitions. All measurements were baseline corrected by subtracting the sensorgram obtained with control surface and kinetic parameters were determined by fitting the data to two-step binding model using BIAevaluation software ver2.0.2.

FRET was measured as described (Kim et al., 2003; Pan et al., 2010; Xiong et al., 2009). For extracellular FRET, cells were seeded on poly-L-lysine (100 µg/ml) or MAdCAM-1 (10 µg/ml) substrates in serum-free medium, and incubated for 10 minutes at 37°C. 0.5 mM Mn²⁺ was added to activate integrin where indicated. For DTT treatment, cells were pretreated with 0.1 or 0.2 mM DTT at 37°C for 10 minutes. Adherent cells were fixed with 3.7% formaldehyde (PFA)/PBS for 15 minutes at room temperature and nonspecific sites were blocked by incubation with 10% serum-rich medium at room temperature for 10 minutes. Then cells were stained with 20 µg/ml Act-1 Fab conjugated with Alexa Fluor 488 for 20 minutes at 37°C. After two washes, cells were labeled with 10 µM FM-4-64 FX for 4 minutes on ice, washed once, fixed and mounted with GVA mount (Invitrogen) under a coverslip.

For detecting the association of integrin cytoplasmic tails, α₄-mCFP/β₇-mYFP 293T transient transfectants were treated as above. Then cells were fixed with 3.7% paraformaldehyde in PBS for 15 minutes at room temperature and subjected to photobleach FRET imaging.

FRET image acquisition, image registration, background subtraction and data analyses were performed with Leica TCS SP5 under a 63× oil objective. FRET efficiency (E) was calculated as E = 1–[Fdonor(d)_Pre/Fdonor(d)_Post], where Fdonor(d)_Pre and Fdonor(d)_Post are the mean donor emission intensity of pre- and post-photobleaching.

Glass coverslips were coated with 100 µg/ml poly-L-lysine or 10 µg/ml MAdCAM-1 overnight at 4°C and blocked by 2% BSA at 37°C for 1 hour. Cells were seeded on coated coverslips in serum-free F12 medium at 37°C for 2 hours, with 0.5 mM Mn²⁺ added to the medium during spreading where indicated.

Differential interference contrast and interference reflection microscopy were conducted on the OLYMPUS IX71 microscope with a 63× oil objective coupled to an Orca CCD camera (Q-IMAGING). For quantification of cell spreading, outlines of 50 randomly selected adherent cells from three independent experiments were generated, and the number of pixels contained within each of these regions was measured by using Image-Pro plus v6.0.

For western blotting, after cell spreading, cells were washed and lysed with lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.05% Tween-20, Complete protease inhibitor cocktail tablets and PhosSTOP phosphatase inhibitor cocktail tablets) on ice for 30 minutes. Cell lysates were then analyzed by blotting for FAK, pY397-FAK, paxillin, pY118-paxillin and β-actin. Intensity analyses were conducted using Image J software.

After cell spreading, cells were fixed with 3.7% paraformaldehyde in PBS at room temperature for 10 minutes. Anti-β₇ mAb FIB27 (5 µg/ml) was used to stain β₇ at room temperature for 2 hours, followed by staining with goat anti-rat IgG-Cy3 (5 µg/ml) at room temperature for 1 hour. Finally, coverslips were mounted with GVA mount and images were obtained with Leica TCS SP5 under a 63× oil objective.

All the buffers used in these assays were degassed for at least 30 minutes. Cells were pretreated with 5 mM DTT at 37°C for 30 minutes where indicated. PMA (10 ng/ml) stimulation was applied at 37°C for 3 minutes. To visualize free cysteine residues on cell surfaces using immunofluorescence staining, 293T transfectants with or without stimulation were allowed to adhere to poly-L-lysine or MAdCAM-1 surfaces for 5 minutes before staining. For human peripheral blood lymphocytes (PBLs), the indicated chemokines were coated together with poly-L-lysine or MAdCAM-1 at 2 µg/ml and cells were allowed to adhere for 3 minutes before staining. Alexa-Fluor-488-maleimide (Invitrogen) was applied for 45 minutes to label free cysteine residues, and then cells were washed and fixed. β₇ detection was the same as for the integrin clustering test. DAPI (1 µg/ml) was applied for 15 minutes at room temperature to detect cell nuclei. Finally, coverslips were mounted with GVA mount and images were obtained with Leica TCS SP5 under a 63× oil objective.

In the immunoprecipitation assay, chemokines used to activate integrin α₄β₇ on PBLs were applied at 500 ng/ml for 3 minutes at room temperature. For DTT titration, 293T transfectants were treated with different concentrations of DTT at 37°C for 10 minutes. Thereafter cells were labeled with 400 µM 1-biotinamido-4 (4′-[maleimodoethyl-cyclohexane]-carboxamido) butane (biotin-BMCC) (Pierce) at room temperature for 30 minutes and subsequently washed and lysed. α₄β₇ immunoprecipitates were analyzed by blotting biotinylated sulfhydryls with horseradish peroxidase-conjugated avidin (avidin-HRP).

The paired Student's t-test was used for statistical analyses of experiments with two conditions. In the cases of three or more conditions, analysis of variance (ANOVA) was used with Bonferroni post-tests (two factors) or Dunnett post-tests (one factor).

We thank Dr DianQing Wu and Dr JunLin Guan for advice on experiments and data.