Actin polymerization stabilizes α4β1 integrin anchors that mediate monocyte adhesion

The actin cytoskeleton is dynamic and supports a multitude of essential functions in adherent and migrating cells from participating in the formation of protrusions to the generation of tensile forces and cell motility (Vicente-Manzanares and Sánchez-Madrid, 2004). In contrast, the role of the actin cytoskeleton remains poorly understood during recruitment of blood leukocytes into tissues.

Leukocyte recruitment is fundamental to inflammatory and homeostatic processes: host response to infection, pathogenesis of inflammatory disorders, and immune surveillance. Blood leukocytes are recruited through a multistep cascade of highly organized adhesive and signaling steps, including tethering and rolling along vascular endothelium, activation of integrins by chemokine signaling, arrest, stabilization of adhesion, intravascular crawling, and diapedesis (Butcher, 1991; Ley et al., 2007).

Integrins are adhesion and signaling molecules expressed by most nucleated cells (Hynes, 2002). Signaling via chemokine G protein–coupled receptors (GPCRs) triggers rapid up-regulation of leukocyte α₄β₁ (very late antigen-4 [VLA-4]) and α_Lβ₂ (LFA-1) integrin affinity, which stabilizes bonds with VCAM-1 and ICAM-1, respectively (Constantin et al., 2000; Chan et al., 2001). Arrest is mediated by a relatively small number of integrins and must be rapidly stabilized. One mechanism for stabilization is integrin clustering upon ligand binding and intracellular signaling (outside-in signaling) through adapter proteins (Alon and Dustin, 2007). Linkage of VLA-4 to the actin cytoskeleton through the adapter protein paxillin is critical for stabilization of leukocyte adhesion when exposed to fluid flow and transmission of mechanical force (Alon et al., 2005). Despite this, the actin cytoskeleton in leukocytes has generally been viewed as a relatively static structure that tethers integrins in the cell membrane and restricts their lateral diffusion (Kucik et al., 1996; van Kooyk et al., 1999). Pharmacological inhibition of actin polymerization has been shown to increase LFA-1–mediated adhesion and also resulted in increased accumulation of integrins at the adhesion surface (van Kooyk et al., 1999; Kim et al., 2004). These experiments modeled the prolonged adhesion of contact-dependent cells. On the other hand, leukocytes undergoing recruitment adhere rapidly to the endothelium and resist detachment. There is a paucity of information on whether the actin cytoskeleton participates in adhesion stabilization and whether it adapts to external forces.

Mechanotransduction of fluid forces dramatically influences endothelial cell gene expression in vascular endothelium (Tzima et al., 2005) and neutrophil morphology (Coughlin and Schmid-Schönbein, 2004; Makino et al., 2006; Coughlin et al., 2008). All of these responses involve an intact actin cytoskeleton. Integrins are both force-bearing and force-sensing receptors and, thus, are critical to a variety of cell functions (Brakebusch and Fässler, 2003; Critchley, 2004). Integrin mechanotransduction is a key feature in strengthening integrin-mediated adhesion. For example, when force is applied to magnetic beads attached to cell surface integrins, cells are capable of producing a “stiffening” response thought to be a “strengthening” of the linkage to the actin cytoskeleton (Wang et al., 1993; Choquet et al., 1997).

Several signaling molecules have been implicated in mechanotransduction, including Rap1 and Rho family GTPases and phosphoinositide 3-kinase (PI3K). Rap1 is a small GTPase critical for integrin adhesive function (Bos et al., 2001; Shimonaka et al., 2003) and can be activated by tensile forces (Sawada et al., 2001). Rho small GTPases are critical players in actin polymerization. Rac, a member of the Rho family of small GTPases, has been shown to be critical for F-actin rearrangement in endothelial cells exposed to shear stress (Tzima et al., 2002). PI3K plays a key role in chemokine-induced cell adhesion under fluid flow and cell polarization during chemotactic migration (Sasaki et al., 2000; Smith et al., 2006; Heit et al., 2008). In addition, PI3K has been shown to be important for force-mediated activation of integrins (Katsumi et al., 2005).

Collectively, our data reveal that leukocytes adhering via VLA-4 adapt to external hydrodynamic force by forming upstream structures—“anchors” that stabilize adhesion. This process is dependent on actin polymerization and activation of Rap1 and Rac as well as PI3K-γ. We propose that leukocytes sense and respond to external forces through signaling and structural adaptations, which are indispensible for stabilization of rapid adhesion in the blood stream.

We set out to investigate the structural and functional role of the actin cytoskeleton in VLA-4–mediated leukocyte adhesion under conditions of fluid flow. Cytochalasin D (CytoD) and Latrunculin B (LatB), two different inhibitors of actin polymerization, were used at concentrations (2 µM) that dramatically reduced F-actin polymerization in lamellipodia of monocyte-like U937 cells but minimally affected total F-actin content and increased clustering of cortical F-actin (Fig. S1, a–d). VLA-4 on the cell surface remained colocalized with F-actin, as measured by immunofluorescence microscopy and flow cytometry of cytoskeleton ghosts, suggesting that the CytoD or LatB did not disrupt the constitutive linkage between integrins and the cortical actin cytoskeleton (Fig. S1, d–f).

In the context of leukocyte recruitment from the blood into tissues, integrins with intermediate affinity can mediate slow rolling, and upon activation of leukocytes by chemokines, the rapid conversion of integrins to high affinity mediates arrest (Campbell et al., 1998; Chan et al., 2001). Therefore, we investigated the role of actin polymerization in VLA-4 affinity modulation by GPCR inside-out signaling. We found that neither CytoD nor LatB pretreatment blocked rapid fMLF (f-Met Leu Phe)-induced VLA-4 affinity up-regulation (Fig. S2 a), consistent with previous studies for VLA-4 and LFA-1 (Chan et al., 2001; Yu et al., 2010).

CytoD or LatB pretreatment inhibited arrest and accumulation of U937 cells induced by fMLF (Fig. S2 b) or chemokine stromal-derived factor 1 α(SDF-1α; Fig. 1 a). In assays in which maximal VLA-4 affinity was induced by manganese (Mn²⁺), CytoD or LatB also inhibited leukocyte accumulation on VCAM-1–coated flow chambers (Fig. 1 b). Integrin affinity up-regulation by Mn²⁺ does not require intracellular signaling through chemokine receptors, which can modulate leukocyte actin dynamics (Howard and Meyer, 1984). In detachment assays, both CytoD and LatB dramatically reduced the adhesion of Mn²⁺-treated U937 cells (Fig. 1 c), human peripheral blood monocytes (Fig. 1 c), and murine lymphocytes (not depicted) to VCAM-1–coated surfaces. Collectively, these data suggest that actin polymerization is required for blood leukocyte resistance to detachment under conditions of flow and implicate a role for actin polymerization in the stabilization of VLA-4–mediated leukocyte arrest but not integrin affinity modulation.

Adherent CytoD- or LatB-pretreated leukocytes were deformed after exposure to flow, in contrast to buffer-treated cells that maintained a spherical shape. Leukocytes were displaced from the initial adhesion site, and elongated structures were pulled from the cell body (Fig. 1 d). These structures remained attached to the VCAM-1–coated surface and sporadically would fracture and release tethered leukocytes. Scanning EM (SEM) of leukocytes fixed under flow revealed deformation of CytoD- or LatB-treated cells and abundant elongated structures without associated cell bodies (Fig. 1, e and f).

Visualization of F-actin revealed the effects of CytoD and LatB on U937 cells exposed to fluid flow. Immunostaining showed that elongated upstream structures contained F-actin and VLA-4, which were colocalized (Fig. 2 a). Because actin polymerization was blocked, these data suggest that F-actin was extruded from the cell body, as it was displaced from the attachment site. This possibility was investigated in real time using Lifeact-GFP–transfected U937 cells (Riedl et al., 2008). Upon introduction of flow, F-actin was extruded from the upstream cortical actin cytoskeleton of CytoD- or LatB-treated cells (Fig. 2, b and c; and Videos 1 and 2). In some experiments, CytoD or LatB was added to the perfusion buffer, which recapitulated cell deformation and the extrusion of F-actin and suggested that continuous actin polymerization is required for stable leukocyte adhesion under flow (Fig. 2 d and Video 3).

The aforementioned experiments suggest that mechanical force exerted on adherent leukocytes as a consequence of fluid flow induces actin polymerization to withstand detachment and deformation. To directly evaluate this possibility, live-cell imaging was performed using Lifeact-GFP–transfected U937 cells. Each cell was imaged under static conditions and subsequently after flow. Under static conditions, constant random movement of GFP was observed in submembrane regions and microvillous-like projections over the entire cell surface (Videos 4 and 5). These features were maintained in cells exposed to flow, but the most significant difference was the accumulation of F-actin in structures that contacted the adhesion surface upstream of the cell body and in the cortical actin cytoskeleton adjacent to the insertion site of these structures (Fig. 3 a, Fig. S3 d, and Video 6). Upstream accumulation of F-actin did not occur when cells adhered to poly-l-lysine were exposed to flow (unpublished data). We refer to the upstream F-actin–rich structures as anchors because they anchor leukocytes to VCAM-1. Upstream anchors were also observed in human peripheral blood monocytes (Fig. 3 b). Anchors extended from the adhesion surface at ∼1/5 of the cell height, and their mean length was 1.3 µm for monocytes and 1.9 µm for U937 cells (Fig. 3 c and Fig; S3 a). Anchors also formed when monocytes were infused and contacted the VCAM-1–coated surface under constant fluid flow (Fig. 3 d).

Increased F-actin appeared in upstream anchors and at sites of anchor insertion within the cortical actin cytoskeleton within seconds of introducing flow and gradually increased over time (Fig. 3, e and f; and Video 6). Exposure to fluid flow induced a twofold increase in F-actin accumulation at anchor insertion into the cortical F-actin relative to the downstream cortical cytoskeleton (Fig. 3, g and h; and Fig. S3 e). Collectively, our data suggest that leukocytes rapidly undergo a structural adaptation involving F-actin reorganization, which enables them to resist detachment forces imparted by fluid flow.

A biomechanical study has established that adhesion contact sites between microvilli at the rear of a spherical leukocyte and a flat surface experience the highest tension upon exposure to laminar flow (Chen and Springer, 1999); thus, it is possible that tension in leukocyte anchors provides a stimulus for actin polymerization. To evaluate this possibility, detachment assays were performed in which Mn²⁺-treated U937 cells were exposed to a defined hydrodynamic force either directly or in stepwise increments (Fig. 4 a). In the latter group, a significantly higher percentage of leukocytes remained adherent. This effect was not observed in leukocytes pretreated with CytoD or LatB (Fig. 4 a). Live-cell imaging of Lifeact-GFP–transfected cells demonstrated that exposure to stepwise fluid flow correlated with increased F-actin accumulation at upstream anchor insertion points (Fig. 4 b). Collectively, these data suggest that upstream F-actin–rich anchor formation in response to lower hydrodynamic forces preconditions cells to resist a higher force.

Rap1 is a known regulator of cell adhesion and cell spreading (Bos et al., 2001). Rap1 can be activated in response to cell stretching (Sawada et al., 2001, 2006). Therefore, we investigated a role for Rap1 in anchor formation. U937 cells were transfected with Rap1-GFP and Lifeact-RFP, adhered to VCAM-1, and exposed to fluid flow. A significant increase in Rap1 accumulation at the anchor insertion into the cortical cytoskeleton was observed (Fig. 5, a and b; and Fig. S4 a). Experiments with constitutively active Rap1-V12–GFP confirmed that accumulated Rap1 was GTP bound (unpublished data). Leukocytes transfected with a dominant negative (dn) Rap1-N17–GFP failed to form F-actin–rich anchors. Furthermore, these cells were displaced from the initial adhesion site (Fig. 5, c–e). F-actin was extruded in a subset of dnRap1-transfected cells, analogous to cells pretreated with CytoD or LatB (unpublished data). These data confirmed a role for Rap1 in the adaptation to fluid flow by leukocytes.

Small GTPases of the Rac and Cdc42 families are well known inducers of actin polymerization (Ridley and Hall, 1992; Ridley et al., 1992; Hall, 1994). Rap1 effectors include guanine exchange factors (GEFs) that activate Rac (Arthur et al., 2004). Therefore, to determine whether fluid flow activated Rac and/or Cdc42, we transfected leukocytes with the p21-binding domain (PBD) of p21-activated kinase (PAK) fused to YFP (PAK-PBD-YFP), which can bind either GTP-bound Rac or Cdc42 (Manser et al., 1994). Leukocytes exposed to fluid flow had a significant increase in the accumulation of PAK-PBD-YFP (Fig. 6, a and b; and Fig. S4 b) at sites of anchor insertion. To determine the role of Rac versus Cdc42, U937 cells were transfected with dnRac-N17–GFP or dnCdc42-N17–GFP. Upon exposure to a fluid flow, cells transfected with dnRac-GFP were deformed and displaced from the initial adhesion site (Fig. 6, c–f; and Fig. S4 c). Stable anchors did not form, and F-actin was extruded in many cells (Fig. S4 c). Cells transfected with dnCdc42-GFP exhibited normal F-actin dynamics and anchor formation in response to flow (Fig. S4 c). These data support a role for Rac-dependent actin polymerization in the formation of anchors and stabilization of leukocyte adhesion.

The PI3K family of lipid kinases can be activated by mechanical stretch (Katsumi et al., 2005) and can regulate Rac-mediated actin polymerization and polarization of cells during chemotaxis and phagocytosis (Sadhu et al., 2003; Srinivasan et al., 2003; Ferrandi et al., 2007). Previously, we observed that blockade of PI3K function did not affect VLA-4 affinity up-regulation by GPCR signaling (Hyduk et al., 2007), but integrin-mediated leukocyte adhesion was dramatically inhibited (Hyduk and Cybulsky, 2002). These observations are analogous to those made after inhibition of actin polymerization.

Several approaches were adopted to elucidate the function of fluid flow-induced PI3K activation. These included utilization of broad spectrum inhibitors, such as Wortmannin (Wort) and LY294002, a γ isoform inhibitor (AS605240), siRNA inhibition of PI3K-γ expression, and isolation of leukocytes from PI3K-γ knockout mice. All of these experiments yielded consistent data showing that PI3K is required for chemokine-induced leukocyte accumulation (Fig. 7 a), adhesion stabilization, and resistance to fluid flow, independent of chemokine signaling (Fig. 7, a and b). Inhibition of PI3K also recapitulated the deformation defects (Fig. 7 c), which we observed in leukocytes treated with inhibitors of actin polymerization (Fig. 2) and dnRac (Fig. 6). Pretreatment or perfusion of Wort onto adherent U937 cells led to a gradual displacement of cells from their initial adhesion site and distortion (Fig. 7, d–f). Wort perfusion experiments confirm that ongoing activation of PI3K by fluid flow is required for maintenance of stable adhesion.

Because PI3K is a regulator of Rac activity (Han et al., 1998), we investigated whether it is required for Rac and F-actin accumulation within anchors. U937 cells transfected with PAK-PBD-YFP were pretreated with Wort and exposed to fluid flow. Blockade of PI3K reduced the accumulation of PAK-PBD-YFP at sites of anchor insertion (Fig. 8, a and b) in contrast to untreated cells (Fig. 6, a and b). Perfusion of Wort onto adherent Lifeact-GFP–transfected U937 cells led to a gradual degradation of F-actin within anchors (Fig. 8 c and Video 9). These cells were displaced from the VLA-4–mediated attachment site. Collectively, these data support a role for force-induced PI3K in regulating F-actin polymerization and adhesion stabilization.

We investigated spatial and temporal evidence for PI3K activation during formation of anchors. U937 cells were transfected with a GFP probe consisting of the pleckstrin homology (PH) domain of protein kinase B (AKT). The AKT-PH probe specifically binds phosphatidylinositol 3,4,5-triphosphate (PIP₃), the main product catalyzed by PI3Ks (Yu et al., 2003). The basal level of PIP₃ in cells is low, and thus, AKT-PH is evenly distributed. Local activation of PI3Ks converts PIP₂ to PIP₃ in the cell membrane and leads to an increased local concentration of AKT-PH (Vanhaesebroeck and Waterfield, 1999; Botelho et al., 2000). U937 cells adhered under static conditions to VCAM-1 showed no preferential distribution of AKT-PH or a PLC-δ PH domain GFP fusion probe that preferentially binds PIP₂ (Fig. 9, a and b). In contrast, upon exposure of cells to flow, the AKT-PH, but not the PLC-δ–PH, preferentially accumulated within upstream anchors and at their insertion sites (Fig. 9, c–g). These findings are consistent with the rapid activation of PI3K in upstream anchors by fluid flow.

Contact-dependent cells, such as fibroblasts and endothelial cells, continually generate internal traction forces via integrin-mediated adhesion to extracellular matrix (Ridley et al., 2003). In contrast, leukocytes in the vascular compartment must arrest on inflamed endothelium. This suddenly exposes them to high external forces from flowing blood, which requires leukocytes to resist this detaching force. We investigated the cellular and molecular responses of adherent leukocytes subjected to hydrodynamic forces. Our data reveal a mechanism that allows leukocytes to sense and resist these forces (Fig. 10). Exposure of blood leukocytes adherent via high affinity VLA-4 to an external force stretches microvilli and triggers localized accumulation of active Rap1 and PI3K, which regulate Rac-dependent actin polymerization. As a result, F-actin–enriched anchors form and provide the tensile strength necessary to reinforce and resist a detaching force.

PSGL-1–mediated rolling results in a rapid elongation of microvilli and microridges and the formation of tethers (Schmidtke and Diamond, 2000; Ley et al., 2007). Tethers form as a result of separation of the lipid bilayer from the cortical cytoskeleton (Ramachandran et al., 2004), and tether formation has been associated with lower rolling velocities. In part, this may be because each tether acts as a fulcrum that promotes cell rotation, which changes the moment of force, and decreases the strain experienced by PSGL-1–selectin bonds (Schmidtke and Diamond, 2000). Recently, total internal reflectance fluorescence microscopy was used to visualize the formation of membrane tethers that extended as much as 16 µm upstream of rolling neutrophils and counteracted hydrodynamic forces (Sundd et al., 2010). However, unlike anchors, tethers are transient structures consistent with the transient nature of PSGL-1–selectin bonds. In contrast, high affinity integrin bonds are more stable and maintain prolonged contact with activated endothelium. This leads to anchor formation. When we inhibited Rap1, PI3K, Rac, or actin polymerization, the cell body was displaced from the initial site of adhesion. Moreover, anchors were destabilized, and F-actin was extruded from the cortical cytoskeleton, yet integrin-mediated adhesive contacts persisted (Figs. 2, 5, 6, and 7).

Rapid accumulation of F-actin and anchor formation was observed in real time when fluid flow was applied to adherent cells (Fig. 3). Anchors were formed upon stretching of an upstream microvillus or microridge. Anchors were maintained for as long as cells were exposed to a hydrodynamic force (Fig. 4). The length and height of anchors can act as lever arms dissipating forces transmitted by integrin bonds, as previously described for selectin-mediated rolling (Alon et al., 1995). Anchor dimensions were proportionally similar in human monocytes and larger U937 cells, suggesting a potentially universal structural response in leukocytes to defined hydrodynamic forces (Fig. 3 c and Fig. S3, b and c).

The contribution of tensile forces to outside-in integrin signaling initiated by ligand binding and integrin clustering is poorly understood. We did not observe significant changes in cell morphology or in the distribution of F-actin when leukocytes with high affinity integrins were adhered to VCAM-1–coated surfaces under static conditions (Fig. 3 and Videos 4 and 5). Only after the application of fluid flow did F-actin accumulation and the formation of anchors occur. VLA-4 located at the tips of anchors bound to VCAM-1 on the adhesion surface, yet we observed F-actin accumulation throughout anchors and at their insertion in to the cortical cytoskeleton. This suggests that tension transmitted throughout anchors initiates signaling ultimately leading to actin polymerization. Potential mechanosensory molecules include adaptor proteins, such as p130cas (Sawada et al., 2006) stretch-activated receptors and/or channels (Matthews et al., 2006).

Consistent with a previous study (Sawada et al., 2001), we found that force induced active Rap1 accumulation at sites of anchor insertion, which was critical for adhesion stabilization (Fig. 5). The role of Rap1 in leukocyte arrest is controversial. Mn²⁺-induced T lymphocyte adhesion to ICAM-1 or VCAM-1 has been reported to be both dependent (de Bruyn et al., 2002) and independent (Ghandour et al., 2007) of Rap1. In the context of chemokine-induced arrest mediated by VLA-4, Rap1 was shown to be dispensable (Ghandour et al., 2007). The reasons for this discrepancy remain unclear, but it is possible that differences in expression or activation of Rap1 GEFs and GTPase-activating proteins are responsible.

Previous studies indicate that Rap1 signaling functions to activate PI3K (Fukuyama et al., 2006; Kortholt et al., 2010). It is possible that amplification loops might exist that regulate their activity. In Dictyostelium discoideum, activated Rap1 binds and activates PI3K, which in turn is critical for Rac1 activation, cell adhesion, and subsequent pseudopod formation (Kortholt et al., 2010). In E-cadherin–mediated cell–cell adhesion, PI3K was required for Rap1-dependent activation of Rac via Vav2, a Rac-GEF; in addition, overexpression of constitutively active Rap1 could not overcome the effects of PI3K inhibition, suggesting that PI3K is downstream of Rap1 (Fukuyama et al., 2006). Whether PI3K regulates Rap1-GEFs is not currently well understood.

Previous studies have demonstrated the activation of PI3K in cells exposed to tensile forces (Suzuma et al., 2002; Katsumi et al., 2005). In our study, leukocyte structural adaptations in response to fluid flow are dependent on force-induced activation of PI3K and localization of its product, PIP₃, within anchors (Fig. 9). PI3K activation was also required for Rac-dependent F-actin accumulation in anchors (Fig. 8). This finding is consistent with the established role of PI3K as a regulator of actin polymerization (Vanhaesebroeck and Waterfield, 1999; Pollard et al., 2000). The reinforcement of tension-bearing structures by actin polymerization is critical for adaptation of cells to external forces.

Anti–human IgG F(ab′)₂ (mouse adsorbed) for coating of adhesion surfaces was purchased from Invitrogen. Human SDF-1α was purchased from PeproTech, and human VCAM-1/fragment crystallizable (Fc) and ICAM-1/Fc were purchased from R&D Systems. HBSS (containing Ca²⁺ and Mg²⁺), manganese chloride, fMLF (the bacterial chemoattractant f-Met-Leu-Phe), and PKH-26 were purchased from Sigma-Aldrich. CytoD, LatB, LY294002, Wort, and PI3K-γ inhibitor II AS605240 were purchased from EMD. 488 phalloidin was purchased from Invitrogen. Primary antibodies included mouse anti–human CD49d clone HP2/1 (AbD Serotec) and clone 44H6 (gift from M. Letarte, Hospital for Sick Children, Toronto, Ontario, Canada), FITC- or allophycocyanin (APC)-conjugated CD49d clone 9F10 (BioLegend), APC-conjugated anti–human CD14 clone 61D3 (eBioscience), and FITC-conjugated anti–mouse CD8 (BioLegend). Goat anti–mouse-Cy3 secondary antibodies were purchased from Jackson ImmunoResearch Laboratories, Inc.

The myelomonocytic U937 cell line was obtained from American Type Culture Collection. A line of U937 cells transfected with human formyl peptide receptor (FPR; U937-FPR) was obtained from G.P. Downey (University of Toronto, Toronto, Ontario, Canada). U937 cells were cultured in RPMI 1640 supplemented with 10% heat-inactivated FBS. 500 g/ml geneticin was added to the culture medium when culturing U937-FPR cells. Cell culture reagents were purchased from Invitrogen.

Human peripheral blood mononuclear cells depleted of platelets were isolated by an adapted density gradient centrifugation protocol (OptiPrep; Axis-Shield). Blood sample collection from healthy human volunteers was approved by the Research Ethics Board at University Health Network. In brief, venous blood anticoagulated with 2 mM EDTA was mixed with 1.85 g OptiPrep and 2 ml of solution B for every 5 ml of blood (100 ml of solution B = 3.42 ml NaCl, 1 ml Hepes, 2 ml EDTA, 1 ml plasma, and 92.5 ml distilled H₂O) to create a 1.091-g/ml density layer. Layered over the top was a 1.078-g/ml density layer made up of 2 ml of solution B (for every 5 ml of blood) and 3.02 g OptiPrep. Enriched mononuclear leukocytes were separated by centrifugation at 1,600 rpm for 30 min at 4°C. Monocytes were purified from peripheral blood mononuclear cells using the human monocyte enrichment kit (EasySep; STEMCELL Technologies) according to the manufacturer’s protocol and used immediately. The purity of monocytes was determined by flow cytometry with the anti–human CD14 antibody and was routinely >90%.

Axillary, cervical, and inguinal lymph nodes were harvested and gently dispersed, and cells were filtered through a 70-µm mesh. After a PBS wash, a negative selection process was performed using a CD8⁺ T cell isolation kit (MACS; Miltenyi Biotec) according to the manufacturer’s instructions. The purity of CD8⁺ T cells was determined by flow cytometry and routinely was >90%.

U937 cells were suspended in 100 µl Nucleofector solution supplemented with 1 or 2 µg of each construct (Lifeact-GFP, Lifeact-RFP, PH-AKT-GFP, PH-PLC-δ–GFP, dnRac-GFP, dnCdc42-GFP, PAK-PBD-YFP, Rap1-GFP, Rap1-N17–GFP, and Rap1-V12–GFP). For double transfections, 1 µg of each construct was used. Cells were transfected using the Cell line Nucleofector Kit V (Lonza) and the Nucleofector II program V-001. Cells were recovered in 500 µl RPMI 1640 with 10% FBS and transferred into 1-ml preincubated FBS-supplemented RPMI 1640 6-well plates. Cells were allowed to recover in a humidified 5% CO₂, 37°C chamber for 24 or 48 h. The GFP- and/or RFP-expressing cells were analyzed by fluorescence microscopy (FluoView A; Olympus) and flow cytometry (Cytomics FC500; Beckman Coulter) to monitor transfection efficiency. At 24 or 48 h, a trypan blue viability assay was performed before infusion into a parallel plate flow chamber confocal microscopy setup. Predesigned siRNA targeting PI3K-γ was purchased from Invitrogen (Silencer; catalog no. 16708; ID no. 143809). siRNA was transfected using Nucleofector as described in this paragraph. After transfection with siRNA, cells were cultured in RPMI 1640 and 10% FBS for 72 h. Efficiency of knockdown was determined by Western blotting with an anti–p110-γ antibody (Cell Signaling Technology).

The first 17 amino acids of the actin-binding protein 140 probe, referred to as Lifeact, fused to GFP or RFP were gifts from M. Sixt (Institute of Science and Technology, Klosterneuburg, Austria). Plasmids encoding the PH domain of AKT and PLC-δ fused to GFP (PH-AKT-GFP and PH-PLC-δ–GFP, respectively), PAK-PBD-YFP, dnRac-N17–GFP, and dnCdc42-N17–GFP were gifts from S. Grinstein (Hospital for Sick Children, Toronto, Ontario, Canada). Rap1-GFP, dnRap1-N17–GFP, and constitutively active Rap1-V12–GFP were gifts from M. Philips (New York University School of Medicine, New York, NY).

VCAM-1/Fc was immobilized on 35-mm polystyrene tissue-culture dishes as previously described (Hyduk et al., 2007). In brief, goat anti–human IgG F(ab)₂ was incubated at a low (10 µg/ml) or high (100 µg/ml) density for 60 min at 22°C. In some experiments with low density coating, 20 µg/ml human SDF-1α was co-coated (for 120 min at 22°C). Nonspecific binding sites were blocked with 5% FBS for 60 min followed by VCAM-1/Fc for 60 min (5 µg/ml for low density or 20 µg/ml for high density experiments). Leukocyte accumulation and detachment assays were performed using parallel plate flow chambers (Glycotech) as previously described (Hyduk et al., 2007). Adhesion molecule–coated dishes served as the bottom surface, and a silicone gasket formed the flow path (0.254-mm height and 2.5-mm width). Flow chambers were maintained at 37°C with an infrared heat lamp. Cells were observed with an inverted phase-contrast microscope (Diaphot 300; Nikon), recorded with a video camera (DXC-151A; Sony) and time-lapse video cassette recorder (SVT-S3100; Sony), and analyzed offline using Image 4.0.2 (Scion). Where indicated, leukocytes were pretreated (for 30 min at 37°C) with either 2 µM CytoD, 2 µM LatB, 50 µM LY294002, 1 µM Wort, 100 µM PI3K-γ inhibitor II AS605240, or DMSO carrier. Alternatively, inhibitors were included within the perfusion assay buffer.

Leukocytes in assay buffer (HBSS with Ca²⁺ and Mg²⁺, 0.5% FBS, and 10 mM Hepes) were infused (at a concentration of 1.0 × 10⁶ cells/ml) with a programmable syringe pump (Kent Scientific) to achieve a wall shear stress of 1 or 2 dynes/cm². After 2 min of infusion, the number of leukocytes stationary for >5 s was determined in three to six nonoverlapping high power fields. In some experiments, leukocytes were infused at 0.5 dynes/cm² for 30 s followed by an increased flow (2 dynes/cm²). After 60 s, cells were perfusion fixed by infusing 2.5% glutaraldehyde and 2.0% formaldehyde in 0.1 M phosphate buffer, pH 7.4, for 5 min.

Leukocytes were infused at a concentration of 1.0 × 10⁶ cells/ml via the outflow port into a flow chamber and allowed to settle under static conditions onto the adhesion surface for 2 min. After static adhesion, flow was introduced by pulling assay buffer through the flow chamber with a programmable syringe pump. Incremental increases in shear stress occurred at 30-s intervals. The number of cells remaining adherent after each interval was determined in three to six nonoverlapping high power fields. Data were expressed as a percentage of input cells remaining adherent.

U937 cells or monocytes (1.0 × 10⁶ cells/ml) adhered under flow to a VCAM-1–coated 33-mm glass coverslip mounted in parallel plate flow chamber were also fixed under flow. For SEM, 2.5% glutaraldehyde and 2.0% formaldehyde in 0.1 M phosphate buffer, pH 7.4, was infused for 5 min. For immunoconfocal experiments, leukocytes were fixed with 2.0% PFA.

Glass coverslips perfused with 2.5% glutaraldehyde and 2.0% formaldehyde were placed in the same fixative for 1 h (22°C). After washing, coverslips were postfixed with 1% osmium tetroxide in 0.1 M phosphate buffer, pH 7.4, for 1 h. Samples were dehydrated through an ascending graded ethanol series, infiltrated with a graded hexamethyldisilizane (Sigma-Aldrich) in absolute ethanol, and dried overnight in a fume hood. Coverslips were mounted on aluminum stubs, sputter coated with gold-palladium (SCD 050; Bal-Tec), and examined using a scanning electron microscope (S-2500; Hitachi) operated at 15 kV. Digital images were acquired using a frame grabber and PCI (Quartz Imaging Corp.). To visualize the side profile of leukocytes and anchors, glass coverslips from flow chamber experiments processed for EM were scored with a diamond scriber, and half the coverslip was removed to rotate the specimen and image it at a 90° angle. Side-view micrographs were characterized in ImageJ (National Institutes of Health) for anchor thickness, anchor insertion height, and anchor length. Veil-type upstream structures were characterized as an individual anchor.

U937 cells were plated on a 5-mm coverslip coated with 20 µg/ml VCAM-1/Fc for a period of 30 min in assay buffer. Upon cell spreading, leukocyte cell membranes were extracted as previously described (Hartwig and Shevlin, 1986). In brief, before detergent extraction, cells were washed in PHEM buffer (60 mM Pipes, 25 mM Hepes, 20 mM EGTA, and 2 mM MgCl₂, pH 6.9). The PHEM buffer was aspirated, and 200 µl PHEM with 0.75% Triton X-100 and 1 µM phalloidin was added for 2 min. The solution was removed and washed three times with PHEM–1 µM phalloidin. Thereafter, cells were fixed with PHEM–1% glutaraldehyde for 10 min, removed, and replaced with 0.01% sodium azide. Coverslips were processed for EM as previously described (Hartwig and DeSisto, 1991). In brief, coverslips were washed extensively in distilled water, rapidly frozen on a helium-cooled copper block, freeze dried at −80°C, and rotary coated with 1.5 nm tantalum/tungsten at a 25° angle and 5 nm carbon at 90° without rotation. The metal replicas were floated off the coverslip using 25% hydrofluoric acid, washed in water, and collected on carbon-coated Formvar 200-mesh copper grids. Grids were photographed in an electron microscope (1200-EX; JEOL) at 80 kV.

Leukocytes were adhered to coverslips coated with poly-l-lysine or VCAM-1. In static assays, cells were washed with assay buffer and incubated with 5 µg/ml mouse anti–human α₄ integrin antibody (HP2/1 or 44H6; for 1 h at 22°C). Coverslips were blocked with 5% FBS followed by a series of washes. Cells were simultaneously fixed in 2% PFA and permeabilized in 0.1% Triton X-100 (5 min at 22°C), washed three times, and incubated with 488 phalloidin (1:1,000 dilution; Invitrogen) and goat anti–mouse secondary antibody (1:500 dilution). In fluid flow experiments, leukocytes were preincubated for 5 min on ice with the 44H6 antibody before the parallel plate flow chamber assay. At the end of the assay, cells were perfusion fixed with 2% PFA, washed, and stained with a secondary antibody and 488 phalloidin. Coverslips were mounted in antifade mounting medium (Vectashield; Vector Laboratories) and imaged with a laser-scanning confocal microscope (FluoView 1000; Olympus) with a 60× oil immersion objective (Plan Apochromat; 1.42 NA).

Real-time imaging was performed on a confocal microscope (FluoView 1000) equipped with a 60× oil immersion objective, FV10-ASW acquisition software (Olympus), and an environmental chamber that maintained parallel plate flow chambers at 37°C throughout the experiment. The frame acquisition rate ranged from 0.4 to 1.0 s per frame (indicated in the video legend). When required, a digital zoom ranging from 5.0 to 9.0 was used (indicated in the video legend; refer to scale bars). Z stacks were acquired by obtaining 0.2–0.5-µM-thick serial sections of an entire cell. 3D reconstructions from z stacks were performed using FV10-ASW acquisition software.

U937 cells were transfected with GFP-, RFP-, or YFP-tagged constructs using the Nucleofector 24 or 48 h before imaging. Transfected cells were infused into a flow chamber and allowed to settle for 1–2 min. At this time, a transfected U937 cell that was brightly fluorescent was identified, and images (including a z stack) were acquired under static conditions for 1 min. Fluid flow was introduced, and continuous real-time frame acquisition was continued for 2–5 min followed by a z stack. If the cell detached upon exposure to fluid flow or shortly thereafter, the experiment was terminated. Therefore, postshear serial z stacks and 3D reconstructions were obtained only when a selected cell did not detach. Occasionally, real-time imaging continued for 2–5 min once fluid flow was stopped. In some experiments, transfected U937 cells were pretreated with a pharmacological inhibitor (30 min at 37°C) before infusion into the flow chamber, whereas in other experiments, inhibitors were introduced into the assay buffer after stabilizing anchors formed (in response to fluid flow), and real-time imaging revealed the effect of the inhibitor on preformed anchors.

To quantify total leukocyte F-actin and α4β1 expression, U937 cells at a concentration of 1.0 × 10⁶ cells/ml were placed in assay buffer and stained with 5 µg/ml anti–human CD49d (clone 44H6) or APC-conjugated anti–human CD49d (clone 9F10) for 30 min on ice. Cells were washed with PBS and fixed in 2% PFA. Cells were then blocked in 5% FBS for 1 h, permeabilized for 5 min with 0.1% Triton X-100, and incubated with Cy3-conjugated anti–mouse IgG secondary antibody (1:500 dilution) and 488 phalloidin (1:1,000 dilution; for 1 h at 4°C in the dark). Cells were washed, centrifuged, and reconstituted in 500 µl PBS. A flow cytometer (LSR II; BD) was used.

Soluble LDV-FITC binding assay FITC-conjugated peptide 4-((N′-2-methylphenyl)ureido)-phenylacetyl-l-leucyl-l-α-aspartyl-l-valyl-l-prolyl-l-alanyl-l-alanyl-l-lysine (LDV-FITC) was synthesized by Commonwealth Biotechnologies, Inc. and used as a ligand for VLA-4 integrins. The LDV-FITC binding assay was performed as previously described (Hyduk et al., 2011). In brief, U937-FPR cells (0.5 × 10⁶ cells in 50 µl) pretreated with vehicle, CytoD, LatB, or LY294002 and siRNA-transfected cells were suspended in HBSS (with 1 mM Ca²⁺/Mg²⁺) supplemented with 20 mM Hepes and 0.5% FBS (assay buffer) and were incubated with 3 nM LDV-FITC (30 min at 37°C). Cells were stimulated at 37°C with 100 nM fMLF (2 min), 50 nM PMA (30 min), or 1 mM MnCl₂ (5 min), diluted with 1 ml HBSS, and immediately fixed by adding 1 ml of 2% PFA at 22°C. Cells were washed with HBSS, and bound LDV-FITC was detected using biotin-conjugated anti-FITC followed by streptavidin-fluorochrome (Jackson ImmunoResearch Laboratories, Inc.). Cells were analyzed by flow cytometry using a flow cytometer (Cytomics FC500).

The association of VLA-4 with the actin cytoskeleton was assessed by extracting U937 cell membranes with 1% Triton X-100 (5 min on ice) and preparing cytoskeletal ghosts. This approach was adapted from Nebe et al. (1997). Initially, cells were stained with PKH-26, a dye that labels membrane lipids, because the loss of PKH-26 fluorescence indicated the extent of membrane extraction. Once membrane extraction was optimized, experiments were then performed to assess VLA-4 association with the actin cytoskeleton. Cells were treated with inhibitors (CytoD or LatB for 30 min), stained with mouse anti–human α4 integrin antibody (9F10), and washed, and the membrane was extracted. Cells were washed, stained with a phycoerythrin-conjugated secondary antibody (goat anti–mouse IgG), washed, fixed with PFA (20 min), washed, and analyzed with a flow cytometer (LSR II).

The displacement of cells from the initial adhesion site after exposure to fluid flow in a parallel plate chamber was quantified using ImageJ software. The mean displacement of cells that resisted detachment for >30 s was determined. ImageJ software was also used to quantify fluorescence intensity in images. The appropriate image slice was selected, and a rectangular box or line was drawn along the x axis that intersected a leukocyte. The box encompassed the upstream anchors and the downstream cortical actin cytoskeleton as well as regions of upstream and downstream background fluorescence. A fluorescence intensity profile within the box was generated across the entire cell. Cells transfected with different fluorescent probes were analyzed, and multiple cells were analyzed in each experiment. In some experiments, analyses were performed on sequential image slices using the same line or rectangle to gain insight into the cell’s height as a measure of the anchor fluorescence profile. A plot profile analysis was applied in ImageJ to determine the fluorescence distribution. Pixel intensity values were obtained and imported into an Excel microscopy worksheet (Microsoft). A mean of the most intense pixels coinciding with the upstream location of anchors was compared with the downstream cortical actin cytoskeleton, and a ratio of upstream to downstream mean fluorescence intensity was obtained. This analysis consistently showed higher Lifeact-GFP fluorescence signals within upstream anchors relative to the downstream cortical actin cytoskeleton, a region where anchors do not form. 3D deconvolution was performed using AutoQuant deconvolution software (Media Cybernetics) with the following settings: deconvolution method with adaptive point spread function (PSF; blind); PSF setting with theoretical PSF; and deconvolution setting with 12 iterations, noise level high, and a noise value of 200.

All data reported are expressed as mean values ± SD. Multiple group analysis was performed using one-way analysis of variance with Tukey-Kramer multiple comparisons posthoc test applied. Group comparisons were determined to be significant for P < 0.05.

Fig. S1 shows the effects of actin polymerization inhibitors on F-actin content, distribution in spread cells, VLA-4 expression, and VLA-4 colocalization with F-actin. Fig. S2 shows the effects of actin polymerization inhibitors on fMLF-induced VLA-4 affinity up-regulation and accumulation of U937 cells in a flow chamber assay. Fig. S3 shows additional SEM images of leukocyte anchors, quantification of anchor dimensions, and supporting images of leukocytes transfected with Lifeact-GFP demonstrating anchor formation. Fig. S4 shows additional supporting images of U937 cells transfected with Rap1-GFP, PAK-PBD-YFP, dnRac-GFP, and dnCdc42-GFP. Fig. S5 shows the effects of inhibiting PI3K on VLA-4 affinity up-regulation, the extent of PI3K-γ knockdown with siRNA, and the colocalization of VLA-4 and F-actin. Videos 1 and 2 show that pretreatment of two different U937 cells with LatB results in F-actin extrusion from cells after exposure to fluid flow. Video 3 shows that stable U937 cell adhesion is disrupted upon addition of LatB to the flow buffer. Videos 4 and 5 show actin dynamics in representative Lifeact-GFP–transfected U937 cells that are adhered under static conditions to VCAM-1 (Video 4) or poly-l-lysine (Video 5). Video 6 shows the formation of upstream F-actin–enriched anchors upon introduction of fluid flow. Videos 7 and 8 show that anchors retract and that the upstream F-actin content diminishes when fluid flow is turned off in two different cells. Video 9 shows degradation of F-actin–enriched anchors when Wort is perfused onto of a stably adherent U937 cell. Online supplemental material is also provided in the JCB DataViewer.

We would like to thank Henry Hong and the imaging facility in the Department of Cell and Systems Biology at the University of Toronto for their expertise and assistance with EM imaging.

This work was supported by the Canadian Institutes of Health Research (grant MOP-89740 to M.I. Cybulsky and Canada Graduate Scholarship Doctoral Award to J. Rullo) and Heart and Stroke Foundation of Ontario (Career Investigator Award to M.I. Cybulsky). M.I. Cybulsky is a member of the Heart and Stroke/Richard Lewar Centre of Excellence at the University of Toronto.