α₄β₁-dependent adhesion strengthening under mechanical strain is regulated by paxillin association with the α₄-cytoplasmic domain

Circulating leukocytes rapidly develop firm adhesion to vessel wall ligands through their various integrin receptors α₄β₇, α₄β₁ (VLA-4), α_Lβ₂ (LFA-1), and α_Mβ₂ (Mac-1; Alon and Feigelson, 2002). Integrins bind their respective endothelial ligands under shear flow at lower efficiency than selectins (Springer, 1994). Adhesive tethers form over a fraction of a second and depend on the ability of the nascent adhesive bond to withstand disruptive shear force. In contrast to selectins, all leukocyte integrins can undergo instantaneous up-regulation of their affinity or avidity to endothelial ligands upon exposure to endothelial chemokines (Kinashi, 2005). In addition, integrins can undergo conformational changes upon ligand binding (Hynes, 2002). Cytoskeletal constraints of integrins may also control integrin adhesiveness (van Kooyk and Figdor, 2000). Previous studies on leukocyte (L)-selectin function regulation have shown that preformed cytoskeletal associations of L-selectin with the actin cytoskeleton control the ability of ligand-occupied selectin to stabilize nascent tethers under shear flow and capture leukocytes under physiological shear stresses (Kansas et al., 1993; Dwir et al., 2001). This raised the possibility that specialized subsets capable of interacting with their respective endothelial ligands under physiological shear flow may also need to properly anchor to the cytoskeleton. Although selectins and integrins are structurally distinct, we hypothesized that α₄ integrin bonds forming under disruptive shear stresses may share a common regulatory mechanism with L-selectin bonds. However, as alterations in cytoskeletal constraints of integrins can modify affinity, clustering, and ligand-induced conformational rearrangements (Carman and Springer, 2003), the direct contribution of integrin anchorage to adhesive outcome has been difficult to dissect.

In this study, we unraveled novel adhesive properties of an α₄-tail mutant with disrupted association with the cytoskeletal adaptor paxillin (Liu et al., 1999). We found that blocking the α₄–paxillin interaction markedly impaired the integrin's ability to anchor to the cytoskeleton in Jurkat T cells. Although not essential for α₄β₁ affinity, ligand-induced conformational changes, surface clustering and topography, or redistribution at short static contacts, paxillin association with α₄β₁ was crucial for α₄β₁–VCAM-1 bonds to resist mechanical stress. These results suggest that subsecond stabilization of α₄ tethers depends on the ability of ligand-occupied α₄ integrins to properly anchor to the cytoskeleton. This work also highlights the key role of the α subunit of α₄β₁ in postligand binding adhesion strengthening of the integrin under mechanical strain.

Paxillin binding to the α₄-cytoplasmic domain is important for integrin α₄β₁ signaling but not for adhesion developed in shear-free conditions (Rose et al., 2003). To examine the role of paxillin binding in α₄β₁-mediated adhesion under shear stress, we analyzed the resistance to shear-induced detachment from the α₄β₁ ligand VCAM-1 of α₄-deficient JB4 Jurkat T cells transfected with either wild-type (wt) α₄ (JB4-wt) or the paxillin binding–defective α₄(Y991A) mutant JB4-α₄(Y991A) (Rose et al., 2003). JB4-α₄(Y991A) cells were less resistant to shear-induced detachment than their JB4-wt counterparts (Fig. 1 A). Notably, bivalent VCAM-1 (VCAM-1–Fc) was much more potent than monovalent soluble VCAM-1 (sVCAM-1) in supporting α₄β₁-specific adhesion (Fig. 1 A), but it still could not rescue the adhesive defect of the α₄(Y991A) mutant. These results were confirmed with multiple clones expressing similar levels of α₄ and β₁ subunits as well as the β₁ activation epitope 15/7 (Fig. 1 B and not depicted). Nevertheless, resistance to detachment from different densities of either ICAM-1–Fc or ICAM-1 was comparable between wt- and mutant α₄β₁–expressing cells (Fig. 1 C). In agreement with these results, VLA-4–dependent adhesion to TNFα-stimulated human umbilical vein endothelial cells (HUVECs) was reduced in Jurkat cells expressing the α₄(Y991A) mutant (Fig. 1 D), in particular at shear stresses ≥5 dyn/cm², within the upper range of shear stresses prevailing in postcapillary venules where the majority of lymphocyte extravasation takes place (Firrell and Lipowsky, 1989). Whereas most cells expressing the wt α₄ firmly arrested on the stimulated HUVEC via their VLA-4, a significant fraction of α₄(Y991A) mutant–expressing cells failed to arrest and established endothelial (E)-selectin–dependent rolling on the HUVEC (Fig. 1 D). In the absence of functional E-selectin, the shear resistance of cells expressing the α₄(Y991A) mutant was much lower than the shear resistance of cells expressing wt α₄ (Fig. 1 D). Because the contribution of LFA-1 to Jurkat arrest was minimal, these data suggest that the Y991A α₄ mutant is deficient in establishing α₄β₁-mediated shear resistance on endothelial cells expressing VCAM-1 as well as on substrates coated with isolated VCAM-1.

Notably, preformed clustering of wt and mutant α₄ subunits on JB4 cells was essentially identical (Fig. 2 A). Real time imaging of JB4 cells that adhered on VCAM-1 also showed identical cell spreading as well as the distribution of both mutant and wt α₄ during 1-min cellular contacts before shear application (WT: n = 44, 16% round, 54% polarized with uniform α₄, 30% polarized with patched α₄; Y911A: n = 27, 18% round, 52% polarized with uniform α₄, 30% polarized with patched α₄; Fig. 2 B). Notably, the strength of resistance to detachment developed by wt α₄ did not correlate with the degree of patching (Fig. 2 B) in contrast to reports on LFA-1–dependent systems (Constantin et al., 2000; Kim et al., 2004). Thus, a mutation of the α₄ tail defective in paxillin binding prevents α₄β₁-mediated resistance to shear-induced cell detachment independent of cell spreading and α₄ patching on VCAM-1.

The α₄(Y991A) mutation blocks paxillin association with the α₄ tail selectively (Liu et al., 1999). As an alternative test of the role of the α₄–paxillin interaction, we exploited a recently identified small molecule inhibitor of this interaction. The compound, designated A7B7C7, blocks the α₄–paxillin interaction and interferes with α₄β₁-dependent cell migration (Ambroise et al., 2002). This inhibitor, but not a control compound (A6B6C6), attenuated the shear resistance of wt α₄β₁–mediated Jurkat cell adhesion to VCAM-1 (Fig. 3 A, left) but had no effect on the residual shear resistance developed by the JB4-α₄(Y991A) cells (Fig. 3 A, right). Adhesion mediated by the α_Lβ₂–ICAM-1 interaction was also insensitive to the inhibitor (not depicted). Knocking down paxillin expression by up to 75% using transient short inhibitory RNA (siRNA) silencing (Fig. 3 B) resulted in reduced adhesiveness of wt α₄β₁–mediated Jurkat cell adhesion to VCAM-1 (Fig. 3 C), with no inhibition of adhesiveness mediated by the α₄(Y991A) mutant (Fig. 3 C). Notably, LFA-1–dependent adhesion to ICAM-1 was also insensitive to identical paxillin silencing (not depicted). Thus, both genetic and pharmacological approaches indicate that the α₄–paxillin interaction increases the resistance of α₄β₁–VCAM-1 contacts to detachment by disruptive shear stresses.

Paxillin binds a number of actin-binding proteins such as talin and vinculin (Brown and Turner, 2004) and does so at sites distinct from the α₄-binding site (Liu and Ginsberg, 2000). We next quantified the fraction of detergent-resistant wt α₄ or α₄(Y991A) retained on NP-40–solubilized cells using fluorescence-tagged integrin-bound α₄ mAb (Fig. 4 A). Retention of intact wt and α₄(Y991A) was similar and low (20% of the total surface α₄; Fig. 4 A). However, the addition of anti–mouse Ig to cluster the mAb-bound wt α₄ markedly increased the association of α₄β₁ surface integrin with the detergent-insoluble cytoskeleton (Fig. 4 B). In contrast, the same treatment produced a negligible increase in the cytoskeletal association of α₄(Y991A)β₁ (Fig. 4 B). Thus, the α₄(Y991A) mutant fails to anchor properly to the actin cytoskeleton in Jurkat T cells.

In light of this poor cytoskeletal anchorage of α₄(Y991A)β₁, we next compared the level of talin associated with the wt or mutant α₄β₁ complex in nonadherent Jurkat cells. Notably, constitutive talin binding to the α₄(Y991A)β₁ complex was significantly reduced compared with the wt integrin, as was evident from coprecipitation analysis (Fig. 5 A). Knocking down up to 65% of the total talin content in wt α₄β₁–expressing Jurkat cells (Fig. 5 B) retained integrin expression (not depicted) but resulted in significant reduction in their α₄β₁-mediated shear resistance on both sVCAM-1 and VCAM-1–Fc (Fig. 5 C, left and right insets, respectively). Notably, identical suppression of talin expression in the α₄(Y991A)β₁-expressing Jurkat cells (Fig. 5 B) had no effect on their low shear resistant adhesion to identical VCAM-1 substrates (Fig. 5 C, right). These results collectively suggest that paxillin association with α₄β₁ also recruits talin to the α₄–paxillin complex and may enhance talin association with the β₁ subunit tail. Thus, both paxillin and talin associations promote α₄β₁-dependent cell resistance to detachment from VCAM-1 under shear stress.

Although the affinity of integrin α₄(Y991A)β₁ to soluble VCAM-1–Fc is retained (Rose et al., 2003), we considered that Jurkat cells expressing the α₄(Y991A)β₁ mutant might fail to develop shear resistant adhesion as a result of reduced avidity for surface-bound VCAM-1. Comparing wt α₄β₁ with α₄(Y991A)β₁ adhesiveness to VCAM-1–coated beads in the absence of applied shear stress, we found that JB4-wt and JB4-α₄(Y991A) cells bound identically to magnetic beads coated with increasing site densities of VCAM-1–Fc (Fig. 6 A), which is supportive of the normal adhesion of α₄(Y991A)β₁-expressing cells under static conditions. Nevertheless, when VCAM-1–coated beads that bound to JB4-wt cells were exposed to abrupt mechanical stress, these beads were displaced significantly less than beads prebound to JB4-α₄(Y991A) cells (Fig. 6 B and Fig. S1, available). α₄ resistance to displacement required an intact actin cytoskeleton, as JB4-wt cells pretreated with the F-actin–severing drug cytochalasin D exhibited even greater displacement in response to abrupt magnetic stress (Fig. 6 C). These findings, together with the shear-based detachment assays (Fig. 1), collectively suggest that paxillin association with the α₄β1 heterodimer and an intact actin cytoskeleton are both required for ligand-occupied α₄β₁ to develop stress-resistant adhesive bonds.

The α₄β₁ and α₄β₇ integrins mediate leukocyte capture under physiological shear flow (Alon et al., 1995; Berlin et al., 1995). Therefore, we compared the ability of mutant α₄(Y991A) versus wt α₄ to support α₄β₁-dependent T cell capture by monovalent and bivalent VCAM-1 under continuous shear flow. Consistent with its defective resistance to shear force, α₄(Y991A)β₁ mediated reduced cell capture and arrest on a large range of densities of either monovalent (sVCAM) or bivalent (VCAM-Fc) VCAM-1 (Fig. 7, A and B). This differential behavior was also manifested at different levels of shear stress that were tested (Fig. 7 B, first two panels). Notably, whereas disruption of the actin cytoskeleton by cytochalasin D resulted in marked inhibition of both cell capture and arrest mediated by wt α₄β₁, cytochalasin D had no effect on the residual adhesions mediated by the α₄(Y991A)β₁ mutant (Fig. 7 A). Interestingly, the duration of individual α₄β₁ tethers, which is a measure of integrin affinity to the ligand (Feigelson et al., 2001), was not altered upon the loss of paxillin binding (Fig. 7 A). Thus, for optimal cell capture under shear flow, the α₄ tail of α₄β₁ requires associations with the intact actin cytoskeleton.

Jurkat cells express low levels of the β₇ integrin subunit; thus, >95% of their α₄-integrin subunits are found in α₄β₁ heterodimers. Nevertheless, JB4-wt cell capture on high density of the bivalent α₄β₇-integrin ligand MadCAM-1–Fc (Berlin et al., 1993) was inhibited by the anti-α₄β₇ antibody Act-1 (unpublished data). The JB4-α₄(Y991A) cells formed fivefold fewer tethers on MadCAM-1 than JB4-wt cells, with a diminished fraction of tethers followed by immediate arrests (Fig. 7 B, right). Thus, paxillin association with the α₄ subunit enhances the ability of α₄ to promote adhesive tethers in the context of both β₁ and β₇ integrins under continuously applied disruptive shear stress.

Preferential localization of receptors to microvilli increases their availability for interactions with counter ligands under shear flow (von Andrian et al., 1995). Electron microscopic analysis of wt α₄ and the α₄(Y991A)-tail mutant revealed identical distribution of these variants to microvillar compartments (82 ± 4% for wt α₄, n = 222; 80 ± 6% for the α₄(Y991A) mutant, n = 196; Fig. 5 C). Furthermore, a higher ratio of the α₄(Y991A)-tail mutant localized on microvillar tips than wt α₄ (a 4.5 tip/base ratio for the mutant vs. only 1.8 tip/base ratio for wt α₄). The number and size of microvillar projections in JB4-wt and JB4-α₄(Y991A) cells were also comparable (Fig. 7 C). Thus, the enhanced ability of wt α₄β₁ to promote adhesive tethers under shear flow was not the result of its preferential distribution to cellular microvilli or to increased localization on microvillar tips.

To examine the effects of disrupting the α₄–paxillin interaction at a single molecule level and in the presence of an external force other than shear stress, we next measured the force of unitary adhesive interactions between wt α₄β₁ or the α₄(Y991A)β₁ mutant and immobilized VCAM-1 by atomic force microscopy (AFM). JB4-wt or JB4-α(Y991A) cells were coupled to the end of an AFM cantilever (Fig. 8 A) and lowered onto a VCAM-1–Fc-coated surface. After a 0.5-s contact, the frequency of productive adhesive events and their strength were analyzed by the degree of deflection experienced by the cantilever during its retraction from the adhesive substrate. Both cell types detached from the VCAM-1 substrate through single jumps, suggesting the breakage of individual bonds during cantilever retraction (Fig. 8 B). The adhesion frequencies of all experiments were maintained below 30%, a level assumed to reflect single α₄β₁–VCAM-1 interactions (Zhang et al., 2004). The specificity of the adhesive events detected in this system were confirmed by a similar 70% reduction in total binding events by the blockade of wt α₄β₁ with Bio1211 (Lin et al., 1999) or by omission of VCAM-1 from the substrate (Fig. 8 C and not depicted). As indicated by the force histograms derived for JB4-wt or JB4-α₄(Y991A) cells (Fig. 8 C), the frequency of productive adhesive events developed by α₄(Y991A)β₁ was up to 10-fold lower than those developed by α₄β₁ after background subtraction. The distribution of unbinding (rupture) forces measured for the two integrin variants was, however, similar (Fig. 8 C). Thus, paxillin association with the α₄ subunit dramatically augments the ability of α₄β₁ to form adhesive tethers that resist disruptive forces irrespective to whether these forces are applied during a vertical force loading (AFM) or during cell rotation (shear stress).

The aforementioned data suggest that paxillin binding to α₄ is required for mechanical stabilization of cell attachments rather than for cytoplasmic induction of high affinity integrin conformations. Ligand binding to integrins can induce conformational changes in the integrin, resulting in high affinity conformations (Du et al., 1991; Shimaoka et al., 2002). Therefore, we considered the possibility that the reduced tether formation by the α₄(Y991A) mutant could reflect defective, instantaneous ligand-induced conformational changes in the integrin under shear stress. We first verified that the α₄β₁ ligand Bio1211 provoked similar conformational changes in α₄β₁ and α₄(Y991A)β₁ under shear-free conditions, as indicated by the identical induction of the β₁ ligand–induced binding site reporter 15/7 epitope by increasing doses of the monovalent α₄β₁-specific ligand Bio1211 (Fig. 9 A; Lin et al., 1999). We next tested the intrinsic attachment efficacy of either wt α₄ or the α₄Y991A mutant to surface-immobilized α₄ mAb in the absence of ligand occupancy of the integrin. Notably, the HP1/2 mAb binding to α₄ integrins is not sensitive to their affinity to or rearrangement by native ligands (Feigelson et al., 2001) and, thus, should be insensitive to intrinsic or ligand-induced affinity changes under shear stress. Notably, in the presence of shear flow, the α₄(Y991A) mutant formed adhesive tethers to immobilized HP1/2 mAb much less efficiently than wt α₄ (Fig. 9 B), as was observed for VCAM-1 (Fig. 1). In addition, adhesive contacts generated by the α₄(Y991A) mutant after 1 min of static contact also exhibited poor resistance to detachment by increasing shear forces relative to wt α₄–mediated contacts (Fig. 9 C). Thus, paxillin association with the α₄-integrin tail enhances the ability of the integrin subunit to generate resistance to detachment forces independently of ligand-induced conformational rearrangements under shear stress conditions.

Cellular stimulation by various agonists increases integrin adhesiveness in various contexts (Hynes, 2002). We next examined the effects of two prototypic agonists, PMA, a direct agonist of diacyl glycerol–dependent PKCs, and the chemokine SDF-1 (CXCL12) on mutant and wt α₄. Exposure of JB4-α₄(Y991A) cells to soluble PMA or to immobilized SDF-1 resulted in enhanced resistance to detachment from VCAM-1–bearing surfaces (Fig. 10 A), although the overall adhesion strength developed by the α₄(Y991A) mutant was lower than that developed by the intact integrin. Further analysis of adhesive tethers formed on VCAM-1 at subsecond contacts also indicated that the ability of initial α₄(Y991A)β₁-dependent Jurkat tethers to convert to immediate firm arrests was enhanced by PMA (Fig. 10 B). Likewise, JB4-α₄(Y991A) cells efficiently responded to in situ subsecond signals from SDF-1 with a twofold elevated frequency of α₄β₁-dependent tethers on VCAM-1 (Fig. 10 B, right). However, overall SDF-1–stimulated tethers mediated by α₄(Y991A) cells remained lower than wt α₄–mediated tethers. Thus, although the mutant α₄ underwent robust activation in response to both chemokine and PMA inside-out signals, its impaired cytoskeletal associations resulted in overall reduced adhesion to VCAM-1 under shear stress.

This study shows that the disruption of paxillin binding to the integrin α₄ tail abrogates its anchorage to the actin cytoskeleton and impairs the ability of integrin ligand bonds to withstand immediate rupture by shear stress, an AFM-pulling device, or abruptly applied magnetic force. Despite normal distribution on the cell surface and retained avidity to immobilized VCAM-1, in the presence of applied forces, this anchorage-deficient mutant poorly mediates tether formation and rapid adhesion strengthening on its ligand. Paxillin-dependent cytoskeletal anchoring of ligand-occupied α₄ integrins may thus underlie their unique capacity to resist disruptive forces and support leukocyte adhesion under shear flow. Thus, although cytoskeletal constraints of integrins were predicted to restrict mobility and clustering on the cell surface and reduce cell adhesiveness (Kucik et al., 1996; Yauch et al., 1997; Kim et al., 2004), we propose that α₄ integrins must retain correct cytoskeletal associations to resist immediate rupture by shear stresses exerted at leukocyte contacts with target blood vessels. Our findings indicate that α₄-integrin anchorage to the cell cytoskeleton is critical for nascent adhesive contacts to resist immediate rupture by shear stress, but it is not required for integrin binding to the ligand nor for ligand-induced conformational rearrangements in the absence of external force. The anchorage deficiency of the α₄-tail mutant resulted in an inability to develop adhesion to a high affinity mAb, which binds the integrin independently of affinity to native ligands (Feigelson et al., 2001; Kinashi et al., 2004). Altogether, these data suggest that paxillin associations with the α₄ tail control (a postligand occupancy anchorage step that is critical for tether stabilization under stress), which is a mechanical property underlying the ability of lymphocytes to capture and arrest on endothelial α₄-integrin ligands under shear flow. Our experiments on cytokine-activated endothelial cells also predict an increased contribution of this α₄–paxillin association to T cells interacting with endothelial beds expressing α₄-integrin ligands in the absence of endothelial selectins.

Integrin affinity is controlled by β-subunit associations with the talin head domain (Tadokoro et al., 2003) and by Rap1 (Kinashi, 2005) via effectors such as RAPL (regulator of adhesion and cell polarization enriched in lymphoid tissues; Katagiri et al., 2004). The effect of RAPL requires α-tail sequences (Katagiri et al., 2004) that are distant from the paxillin-binding site on α₄ (Liu and Ginsberg, 2000). The retained affinity of the α₄(Y991A) mutant and its capacity to mediate static adhesion suggest that lack of paxillin association with the α₄-integrin tail does not interfere with Rap1-dependent signals whether mediated through RAPL or other Rap1 effectors. The reduction in talin association with α₄(Y991A)β₁ did not alter basal α₄β₁ affinity for VCAM-1 (Rose et al., 2003), suggesting that the paxillin-mediated association of talin with α₄β₁ does not contribute to affinity modulation. On the other hand, the extent of these cytoskeletal associations and an intact actin cytoskeleton critically determine the mechanical strength of α₄β₁ VCAM-1 bonds (i.e., tether formation, adhesion strengthening, and resistance to mechanical stress). Thus, we propose that the ability of α₄ integrins to translate ligand occupancy into immediate mechanical stability of subsecond adhesive contacts requires paxillin and talin-mediated linkages of α₄β₁ to the actin cytoskeleton. The regulation of α₄β₁ adhesiveness by talin has never been addressed, especially not under shear stress conditions. The finding that talin suppression impairs the strengthening of wt α₄β₁ bonds under strain is reminiscent of results reporting the involvement of talin1 in ligand-driven α₅β₁-cytoskeletal bonds in fibroblasts (Jiang et al., 2003). Although different integrins may anchor differently to the cytoskeleton in distinct cell types, this involvement of talin in both α₄- and α₅-integrin associations with the actin cytoskeleton is consistent with the notion that talin, apart from its role in integrin affinity regulation (Tadokoro et al., 2003), is a key postligand occupancy adaptor that promotes integrin bond stabilization in distinct mechanical contexts and cellular environments.

Our results highlight the role of the α-integrin subunit rather than the β subunit in postligand binding adhesion strengthening of the α₄β₁–VCAM-1 bond under mechanical strain. Previous findings suggested that a nearly complete truncation of the α₄-cytoplasmic tail impairs α₄β₁ adhesion strengthening without altering initial cell capture to VCAM-1 under shear flow (Alon et al., 1995; Kassner et al., 1995). This truncation of the α₄-integrin tail also reduced integrin mobility (Yauch et al., 1997) and may have increased integrin cytoskeletal anchorage via the intact β1 subunit, although this was not experimentally demonstrated. Therefore, in these earlier studies, it was impossible to distinguish between the contributions of α₄ anchorage versus mobility to rapid mechanical stabilization of α₄ integrin–mediated tethers. Our present results provide the first direct evidence for a positive role of α₄ anchorage for the earliest stabilization events of α₄β₁–VCAM-1 bonds subjected to mechanical strain.

In addition to the α₄ Y991 residue, the phosphorylation level of the α₄ serine 988 has been shown to control the degree of α₄ association with paxillin (Han et al., 2001). A dephosphorylated serine variant mimicked by the phosphodeficient mutant α₄ S988A was reported to bind paxillin at enhanced levels (Nishiya et al., 2005). Interestingly, this phosphodeficient mutant did not properly anchor to the cytoskeleton and supported reduced adhesiveness to VCAM-1 (unpublished data). These findings, together with the paxillin-silencing data of this study, suggest that paxillin binding to the α₄ subunit is required but is insufficient to anchor α₄ to the cytoskeleton. Thus, α₄ phosphorylation, which is postulated to attenuate paxillin binding to the α₄ subunit, is in fact required, at least at a basal level, for proper cytoskeletal α₄ association and productive adhesiveness under shear stress. Overphosphorylation of α₄, which reduces paxillin association, may, on the other hand, attenuate both anchorage and adhesiveness. Studies are ongoing to address both the positive and negative effects of serine phosphorylation on α₄ anchorage and function under strain.

α₄ association with paxillin enhances the activation of the focal adhesion kinases FAK and PYK-2 after α₄β₁ ligation (Liu et al., 1999; Rose et al., 2003). This association also restricts Rac activation at the leading edge via recruitment of the Arf GTPase-activating protein (Nishiya et al., 2005). Nevertheless, at 1-min contacts with VCAM-1, cells expressing the α₄-tail mutant spread normally on VCAM-1. Suppression of tyrosine phosphorylation or inhibition of PYK-2 activity in Jurkat T cells had no effect on α₄β₁-mediated adhesion strengthening developed under shear stress (unpublished data). Thus, the ability of paxillin association with α₄β₁ to enhance integrin anchorage to the cytoskeleton and promote mechanical stability of adhesive tethers at short-lived contacts is distinct from its roles in focal adhesion turnover and Rac deactivation during cell spreading on VCAM-1–containing substrates (Nishiya et al., 2005). Altogether, our findings suggest that modulating mechanical properties of α₄ integrins by the inhibition of specific associations between α₄-cytoplasmic tails and the cytoskeleton may be a selective strategy to fine tune integrin-mediated adhesion under shear stress without altering integrin affinity.

Recombinant seven-domain human VCAM-1 (sVCAM-1) was provided by B. Pepinsky (Biogen, Cambridge, MA). VCAM-1–Fc fusion protein containing seven-domain VCAM-1 fused to IgG was generated as described previously (Rose et al., 2000). VCAM-1–Fc constructed from domains 1 and 2 of VCAM-1 fused to Fc, termed 2D VCAM-1–Fc, was provided by B. Pepinsky. Affinity-purified human full-length spleen-derived ICAM-1 was a gift from T. Springer (Harvard University, Boston, MA). ICAM-Fc and SDF-1α were purchased from R&D Systems. BSA (fraction V), poly-l-lysine, and Ca²⁺/Mg²⁺-free HBSS were obtained from Sigma-Aldrich. Human serum albumin (fraction V) and PMA were purchased from Calbiochem. The α₄-integrin function–blocking HP1/2 mAb, the E-selectin–blocking mAb BB11, the β₁-specific TS2/16 mAb, the α₄-specific nonblocking 5BG10 mAb (all provided by B. Pepinsky), the β₁-integrin subunit mAb 15/7 (provided by T. Yednock, Elan Pharmaceuticals, San Francisco, CA; Yednock et al., 1995), and the anti-α₄β₇ Act-1 (a gift from M.J. Briskin, Millennium Pharmaceuticals, Cambridge, MA) were all used as purified Ig. Antitalin mAb (clone 8d4) was purchased from Sigma-Aldrich. Antipaxillin mAb (clone 349) was purchased from BD Transduction Laboratories. Goat polyclonal anti-α₄ Ab (clone C-20) was purchased from Santa Cruz Biotechnology, Inc.

Jurkat cells deficient in α₄ (JB4) were stably transfected with either wt α₄ or α₄(Y991A) cDNA as described previously (Liu et al., 1999). Cells were subcloned, and multiple clones expressing identical levels of α₄ and β₁ subunits were taken for functional analysis. Clones were maintained in RPMI 1640 medium supplemented with 10% heat-inactivated FCS, 2 mM l-glutamine, 1 mM sodium pyruvate, 100 mM nonessential amino acids, and antibiotics (Biological Industries). Primary HUVECs were established as previously described (Shamri et al., 2005). HUVECs were left intact or stimulated for 4 h with 0.1 ng/ml TNF-α (R&D Systems) before experiments. Staining and FACS analysis were performed as previously described (Feigelson et al., 2003).

For α₄-integrin immunostaining, Jurkat cells were washed in PBS and incubated with 10 μg/ml anti-α₄ B5G10 mAb for 30 min at 4°C. Cells were washed once with PBS + 5 mM EDTA and twice with PBS/0.1% BSA, and α₄ integrins were stained with AlexaFluor546-conjugated anti–mouse Ab (Invitrogen) and fixed in 3% PFA in PBS (30 min at RT). Control cells were fixed before mAb incubation steps. Cells were attached to poly-l-lysine–coated glass slides, and coverslips were mounted with elvanol overnight and analyzed with a confocal microscope (TE300; Nikon) and a laser-scanning system (model 2000; Bio-Rad Laboratories).

α₄ localization was assessed by immunoelectron microscopy as previously described (Chen et al., 1999). In brief, cultured Jurkat cells were washed and prefixed in 0.1 M phosphate buffer, pH 7.4, containing 2% PFA and 0.05% glutaraldehyde. Washed cells in H/H medium (HBSS containing 2 mg/ml BSA and 10 mM Hepes, pH 7.4, supplemented with 1 mM CaCl₂ and 1 mM MgCl₂) were incubated with 10 μg/ml anti-α₄ B5G10 mAb for 40 min at 22°C. Washed cells were stained with 10 μg/ml rabbit anti–mouse Ig, washed, and incubated for 45 min with 5 nm gold particle–conjugated goat anti–rabbit (Aurion). Ultrathin sections (70–90 nm) of 40–60 cells for each experimental group were examined with an electron microscope (Tecnai 12; FEI) under 120 kV, and images were taken using a CCD camera (Megaview 3; Soft Imaging System). In each experimental group analyzed, the number of α₄-specific gold particles on microvillar projections was compared with that on adjacent cell body compartments of identical dimensions.

Silencing of talin expression in Jurkat cells was achieved by a talin1-specific 21-nucleotide siRNA (Dharmacon) corresponding to positions 6,043–6,063 relative to the talin1 mRNA start codon (Shamri et al., 2005). Silencing of paxillin was conducted as described previously (Nishiya et al., 2005). Control transfections were performed with a fluorescein-labeled 21-nucleotide duplex directed to Luciferase GL2. Transfection of T cells was performed by electroporation using the Nucleofection system (Amaxa). Transfected cells were maintained in culture medium. Talin and paxillin expression monitored by immunoblotting was maximally suppressed 72 h posttransfection, and time points were chosen for subsequent functional assays. Immunoprecipitation of α₄ was performed as previously described (Feigelson et al., 2003).

Cells were stained with 10 μg/ml of the FITC-conjugated anti-α₄ mAb HP1/2 at 4°C for 30 min, washed twice with H/H medium supplemented with 1 mM CaCl₂ and 1 mM MgCl₂, and left either untreated or cross-linked with secondary antibodies at 4°C for 30 min followed by two washes as described previously (Geppert and Lipsky, 1991; Evans et al., 1999). All cells were then incubated at RT for 30 min with the cytoskeletal stabilizing buffer (CSB; 50 mM NaCl, 2 mM MgCl₂, 0.22 mM EGTA, 13 mM Tris, pH 8.0, 1 mM PMSF, 10 mM iodacetamide, and 2% FCS) alone or supplemented with 0.1% NP-40. The intact cells or their recovered detergent-insoluble cytoskeletal fractions were washed in detergent-free CSB, fixed in 1% PFA/PBS, and analyzed by flow cytometry. Under these conditions, the majority of detergent-treated cells retain their shape and size. The ratio of mean fluorescence intensity recovered in detergent-treated cells divided by that of cells not exposed to the detergent yields the fraction of mAb-bound α₄ integrin that is resistant to detergent extraction; i.e., anchored to the (detergent resistant) cytoskeletal fraction of the cell.

Protein A–coated magnetic M-280 Dynabeads (Dynal) were coated at RT with various concentrations (0.004–1 μg/ml) of 2D VCAM-Fc in H/H binding medium, washed according to the manufacturer's instructions, and stored on ice. Cells and VCAM-1–coated beads were mixed at RT for 1 min in binding medium at a concentration of 10⁷ cells/ml at a cell/bead ratio of 1:8 followed by a threefold dilution in binding medium. The cellular side scatter, distinguishing between bead-bound and bead-free cells, was analyzed immediately in a FACScan flow cytometer (Becton Dickinson). Background binding determined with protein A–coated beads was <10% of the maximal binding observed at VCAM-1 saturation and was subtracted from the total binding results.

The mechanical stiffness of α₄β₁/VCAM-1 adhesions was measured by electromagnetic pulling cytometry using VCAM-1–coated beads (Matthews et al., 2004). The detailed method is described in supplemental material (available).

All force measurements were conducted at 35 ± 2°C using a previously described AFM apparatus (Benoit et al., 2000). In brief, a microfabricated Si₃N₄ cantilever tip (Park Scientific Instruments) was coated with 0.1 mg/ml of the anti-CD43 mAb (R&D Systems). The spring constants of the cantilevers used were determined at ∼4.7 ± 0.6 mN/m. A single cell was immobilized on the cantilever tip shortly before experimentation. The device was mounted with a piezo-actuator (Piezosystem Jena) on an inverted optical microscope (Carl Zeiss MicroImaging, Inc.) containing a heating stage. A diode laser beam focused on the sensor was used to measure the displacement of the cantilever by the laser beam deflection on a two-segment photodetector. The cell adhering to the cantilever was positioned above an adhesive substrate coated with 2D VCAM-1–Fc captured via human IgG Fc mAb (Jackson ImmunoResearch Laboratories). The cantilever was lowered until the sensor detected a contact force equal to a preselected value (typically 50 pN). After the contact was established for a dwelling time of 500 ms, the cell-bearing cantilever was lifted up by the piezo-actuator, and the de-adhesion force was monitored by a force–distance plot (Fig. 5 B). From this plot, the last detectable de-adhesion force was calculated. For each cell, ∼50–200 force–distance plots were collected within ∼30 min. All de-adhesion events collected in at least 10 independent experiments were presented in histograms (Fig. 5 C).

Purified ligands or mAbs were coated on polystyrene plates as previously described (Grabovsky et al., 2000). Site densities of coated sVCAM-1 and VCAM-1–Fc were determined as previously described (Grabovsky et al., 2000; Sigal et al., 2000). The polystyrene plates were each assembled on the lower wall of the flow chamber (260-um gap) as previously described (Dwir et al., 2000; Feigelson et al., 2001). Cells were washed with cation-free H/H medium, resuspended in binding medium (H/H medium supplemented with 1 mM CaCl₂ and 1 mM MgCl₂), and perfused through the flow chamber at the desired shear stress. To disrupt actin cytoskeleton, cells were pretreated for 15 min with 20 μM cytochalasin D (Calbiochem) or carrier solution (0.1% DMSO). All flow experiments were conducted at 37°C. Tethers were defined as transient if cells attached briefly (<2 s) to the substrate and as arrests if they immediately arrested and remained stationary for at least 5 s of continuous flow. Frequencies of adhesive categories within differently pretreated cells or rates of cell accumulation on adhesive substrates were determined as a percentage of cells flowing immediately over the substrates, as previously described (Grabovsky et al., 2000). To assess rapid development of integrin avidity to the ligand at 1-min stationary contacts, cells were allowed to settle onto the substrate for 1 min at stasis. Flow was then initiated and increased step-wise every 5 s by a programmed set of rates. At the indicated shear stresses, the number of cells that remained bound was expressed relative to the number of cells originally settled on the substrate. Over 95% of tethers to VCAM-1 were blocked by pretreating cells with 10 μg/ml of the α₄-blocking mAb HP1/2. Live imaging of α₄ on Jurkat cells prelabeled with AlexaFluor488-conjugated B5G10 mAb that settled on VCAM-1 was conducted with Delta Vision Spectris RT (Applied Precision). α₄ patching was quantified using Image J software (National Institutes of Health).

Fig. S1 shows the analysis of VCAM-1–coated bead displacement during a magnetic force pulse applied on wt Jurkat cells. The supplemental Materials and methods section describes the experimental setup.

We thank S. Schwarzbaum for editorial assistance.

R. Alon is the Incumbent of The Tauro Career Development Chair in Biomedical Research. This research was supported by the Fogarty International Research Collaboration Awards and was partially supported by the Israel Science Foundation and MAIN, the EU6 Program for Migration and Inflammation. This work was also supported by National Institutes of Health (NIH) grants AR27214 and HL57009 to M.H. Ginsberg, NIH grant CA45548 to D.E. Ingber, and NIH grant P30AR47360 to D.M. Rose. The work of D.M. Rose was additionally supported by the Department of Veterans Affairs Merit Review Entry Program Award and by the Arthritis Foundation.