Mechanical stress-activated integrin α5β1 induces opening of connexin 43 hemichannels

Abstract

The connexin 43 (Cx43) hemichannel (HC) in the mechanosensory osteocytes is a major portal for the release of factors responsible for the anabolic effects of mechanical loading on bone formation and remodeling. However, little is known about how the Cx43 molecule responds to mechanical stimulation leading to the opening of the HC. Here, we demonstrate that integrin α5β1 interacts directly with Cx43 and that this interaction is required for mechanical stimulation-induced opening of the Cx43 HC. Direct mechanical perturbation via magnetic beads or conformational activation of integrin α5β1 leads to the opening of the Cx43 HC, and this role of the integrin is independent of its association with an extracellular fibronectin substrate. PI3K signaling is responsible for the shear stress-induced conformational activation of integrin α5β1 leading to the opening of the HC. These results identify an unconventional function of integrin that acts as a mechanical tether to induce opening of the HC and provide a mechanism connecting the effect of mechanical forces directly to anabolic function of the bone.

Mechanical loading plays a critical role in maintaining skeletal integrity and remodeling of the bone (1). Osteocytes are dispersed throughout the mineralized matrix of the bone where, in addition to being the most abundant cell type, they function as mechanosensors. Mechanical forces applied to the bone cause fluid flow through the lacunar–canalicular network surrounding the osteocyte (2). These forces stimulate cellular responses that involve different types of receptors and multiple intracellular signaling pathways (3). Our laboratory and others have shown that signaling generated from fluid shear stress in osteocytes is likely to be transmitted between cells via gap-junction channels located at the tips of the connecting dendritic processes and through the hemichannel (HC) between the osteocyte cell body and dendrites and their lacunar–canalicular network (4, 5). The signaling cascade activated by mechanical forces leads to the expression and release of important bone anabolic molecules, such as prostaglandins and ATP, through connexin 43 (Cx43) HCs expressed on the cell surface (5, 6). Extracellular prostaglandins are critical anabolic modulators that act in an autocrine or paracrine manner to promote remodeling in response to mechanical stimulation (7). Therefore, the HC provides an important means for regulating the anabolic responses of osteocytes to mechanical stress.

Osteocytes interact with the extracellular matrix (ECM) in the pericellular space through integrins, focal adhesion proteins, and transverse elements that bridge osteocyte processes to the canalicular wall (8, 9). Integrins comprised of heterodimers of α and β subunits serve as the major receptors/transducers that connect the cytoskeleton to the ECM. Fibronectin (FN) in ECM is a ligand for integrin α5β1 recognized through arginine-glycine-aspartic acid (RGD) sequences in FN (10). Upon interaction, integrins frequently form focal adhesions where they recruit proteins such as vinculin and paxillin (11). In addition to focal adhesions, integrins also form fibrillar adhesions that are characterized by elongated/bead-like structures across the basal surface of the cell (12). Integrins are reported to be mechanical sensors on the cell surface (13) and have been proposed as candidate mechanosensors in bone cells (14, 15). Mechanical stimulation is thought to invoke various signaling pathways that are known to be activated by integrins (16). Integrin α5β1 is expressed in bone and cartilage and can induce responses to mechanical stimuli (17, 18). Hence, integrins not only provide support to the cell through focal and fibrillar adhesions but also function as mechanosensors. There is some evidence that integrins are involved in connexin expression and gap-junction communication (19–21). However, the association of connexins with integrins has not been reported, and the molecular mechanisms by which integrins regulate connexins to affect channel functions are also unknown.

Results

Cx43 Interacts Directly with Integrin α5β1, and Fluid Flow Enhances the Interaction.

Integrin α5 and Cx43 colocalized in osteocytic MLO-Y4 cells (Fig. 1A and Fig. S1 A and B) and in primary osteocytes (Fig. S1C). However, colocalization of α5 with vinculin or paxillin was not observed even under fluid flow (Fig. S2), suggesting that α5 was not present at the focal adhesions. Instead, α5 appeared to be located at fibrillar adhesions. Interaction between Cx43 and α5 was demonstrated by immunoprecipitation of α5 by Cx43 antibody (Fig. 1B, lane 3), but not by preimmune serum (Fig. 1B, lane 2). Likewise, reciprocal experiments showed immunoprecipitation of Cx43 (Fig. 1B, lane 5) and β1 (Fig. 1B, lane 8) by α5 antibody, but not by preimmune serum (Fig. 1B, lane 5). The cytoplasmic C terminus of Cx43 (Cx43CT) is known to interact with several intracellular proteins (22). The peptide spanning the entire C terminus of α5 was able to pull down the GST-fusion protein containing the C terminus of Cx43 (GST-Cx43CT), suggesting that the C termini of Cx43 and α5 interact (Fig. 1C). There is no interaction between Cx43 extracellular loop domains (E1 and E2) and α5 (Fig. S3). Further support for a direct interaction was obtained from surface plasmon resonance (SPR) experiments in which we detected a concentration-dependent binding of soluble α5 C-terminal peptide to immobilized GST-Cx43CT but not to GST (Fig. 1D, Upper). Using the GST-only surface as background, K_d estimated from both kinetic (Fig. 1D, Upper) and equilibrium (Fig. 1D, Lower) analyses were in good agreement (1.8 mM). The interaction with scrambled peptide was too weak to quantify. Fluid flow increased the interaction between α5 and Cx43 by approximately threefold (Fig. 1E). These results suggest that there is a direct, specific interaction between the C termini of Cx43 and integrin α5 and that shear stress enhances the interaction.

Fig. 1.

Integrin α5β1 interacts directly with Cx43, and fluid flow enhances the interaction. (A) Colocalization of α5 and Cx43 on the cell surface was detected by dual-immunolabeling of impermeable osteocytic MLO-Y4 cells. (Scale bar, 10 μm.) (B) Cell lysates (lanes 1 and 6) and immunoprecipitates (IP) with antibody to either Cx43 (lanes 2 and 3) or integrin α5 (lanes 4, 5, 7, and 8) were probed for integrin α5 (lanes 1–3), Cx43 (lanes 4 and 5), or integrin β1 (lanes 6–8). (C) The eluted fractions of GST-Cx43CT or GST from beads conjugated with α5 C-terminal or scrambled peptide were immunoblotted using anti-Cx43 (WB, Upper) or were stained with Coomassie Blue (CB, Lower). (D) A concentration series of α5 C-terminal peptide (0.015–2 mM) was injected at 100 μL/min in duplicate over a surface plasmon resonance sensor chip (CM5) to which GST-Cx43CT (specific surface) and GST (reference surface) were covalently attached. (Upper) Double-referenced data were plotted (black traces) and globally fit (red traces) to a 1:1 interaction model using Scrubber 2 software to determine the indicated kinetic and equilibrium constants (Inset). (Lower) Equilibrium analyses were performed by plotting the binding response at equilibrium as a function of α5 concentration and fitting the data to a hyperbolic equation with Scrubber 2 software. (E) Fluid flow increases the interaction. Preloaded cell lysates (Lower) and immunoprecipitates of α5 antibody (Upper) with (FF) or without fluid flow (C) were probed for Cx43 (Left), and band intensity was quantified (Right). P < 0.01, fluid flow (FF) versus control (C).

Integrin α5 and Its Interaction with Cx43 Are Required for the Opening of the HC.

The effect of fluid flow on the Cx43 HC function was studied by measuring the uptake of the tracer dye into the cells. siRNA-mediated knockdown of α5 (Fig. 2A, Right) completely abolished fluid flow-induced opening of the HC as detected by dye uptake (Fig. 2B), thereby indicating that integrin α5 is required for fluid flow-dependent opening of the Cx43 HC.

Fig. 2.

Integrin α5 and its interaction with Cx43 are critical for HC opening. (A) MLO-Y4 cells were transfected with α5 siRNA and siRNA from scrambled sequence or transfection reagent (vehicle). ***P < 0.001, α5 siRNA (60 or 90 nM) versus vehicle and scrambled siRNA. (B) Cells were subjected to static (C) or fluid-flow (FF) conditions, and LY dye uptake was quantified. Dye uptake was abolished in α5 siRNA (60 nM)-transfected cells (FF+α5 siRNA). ***P < 0.001, fluid flow versus FF+α5. (C) (Upper) Lysates of MLO-Y4 cells transfected with Cx43CT-GFP or vector (vehicle) were immunoblotted with anti-Cx43CT antibody (Cx43C). (Lower) α5 antibody primarily coimmunoprecipitated Cx43CT-GFP but not endogenous Cx43 in cells overexpressing Cx43CT-GFP. The immunoprecipitates were blotted with antibody to α5, Cx43C, or GFP. The asterisk indicates the heavy chain of IgG. (D) Overexpression of Cx43CT-GFP, but not GFP or vector (vehicle), in MLO-Y4 cells prevented the opening of the HC in response to fluid flow. The number of cells containing GFP taking up Alexa Fluor 350 dye under fluid-flow (FF) or static (C) conditions was quantified. ***P < 0.001, Cx43CT-GFP (static and fluid-flow conditions), GFP (static condition), and vehicle (static condition) versus GFP (fluid-flow condition) and vehicle (fluid-flow condition).

To determine if uncoupling of the interaction between Cx43 and integrin α5β1 compromised the HC opening, Cx43CT GFP-fusion protein (Cx43CT-GFP) was overexpressed in the cells (Fig. 2C, Upper). Overexpression of Cx43CT-GFP inhibited the association of endogenous Cx43 with α5 (Fig. 2C, Lower). In response to fluid flow, cells expressing the Cx43 C terminus failed to take up the dye, whereas nontransfected or GFP-expressing cells had normal HC function (Fig. 2D). These results suggest that interaction between integrin α5 and Cx43 through their cytoplasmic C termini is critical for the opening of the Cx43 HC in response to mechanical stress.

Association of Integrin α5 with its Substrate, FN, Is Not Required for HC Opening Induced by Fluid Flow.

We examined whether ligand engagement of integrin α5 with the FN substrate during mechanical stimulation is required for its function as a mechanosensor. First, we tested whether culturing the cells on different ECMs affects the HC opening in response to fluid flow. Compared with static conditions, dye uptake was similar in cells cultured on collagen and FN but not in cells cultured on polylysine (Fig. 3A), where minimal colocalization of α5 and Cx43 was observed (Fig. S4). Because α5 and FN normally bind through the RGD sequences, either soluble FN or RGD peptide was added to the cells after their attachment, and the dye uptake was observed. Neither FN nor RGD peptide had any effect on dye uptake induced by fluid flow (Fig. 3B), showing that α5-mediated opening of the HC does not require binding to its FN substrate. Incubation of the cells with the RGD peptide before plating significantly prevented cell attachment to the FN matrix, showing that the peptide was effective in binding to the integrins (Fig. 3C). To eliminate the possibility that the lack of cell attachment was caused by the dissociation of FN or RGD by fluid flow, the cells were incubated with FN-coated beads before application of fluid flow, and the cells attached to FN-coated beads could be visualized readily. Lucifer yellow (LY) dye uptake was similar in cells with and without bound FN-coated beads, further confirming that binding to FN or RGD is not required for HC opening induced by fluid flow (Fig. 3D). Although cells were cultured on collagen matrix, we could not exclude the possible involvement of FN secreted by the cells. Hence, we cultured the cells for 8 h, a time period during which minimal extracellular FN was accumulated (Fig. 3F). No change in HC opening was noticed when the cells were cultured for short time period (8 h) or for the usual 48 h (Fig. 3E). Furthermore, cells cultured for 8 h in an FN-depleted medium had a similar degree of HC opening (Fig. 3G). Together, these data exclude the involvement of FN, suggesting that the association of integrin α5 with its FN substrate is not essential for the fluid flow-induced opening of the HC.

Fig. 3.

The binding of FN substrate is not essential for the role of integrin in HC opening by fluid flow. (A) MLO-Y4 cells were cultured on collagen (Col), FN, or polylysine (poly-Lys) matrices and then were subjected to static (C) or fluid-flow (FF) conditions. ***P < 0.001, fluid flow (collagen or FN) versus the corresponding static condition (collagen or FN). (B) Neither FN nor RGD had any effect on the opening of the HC by fluid flow. ***P < 0.001, fluid flow (Ctrl, RGD, or FN) versus the corresponding static condition (Ctrl, RGD, or FN). (C) The number of cells attached and detached in the supernatant was quantified. ***P < 0.001, RGD (supernatant or attached) versus the corresponding static condition (supernatant or attached). (D) Dye uptake was analyzed, and only cells attached to FN beads were quantified. ***P < 0.001, fluid flow (FN-beads or control) versus the corresponding static condition (FN-beads or control). (E) MLO-Y4 cells were cultured on collagen-coated plates for 8 or 48 h and were subjected to static (C) or fluid-flow (FF) conditions. ***P < 0.001, fluid flow for 8 or 48 h versus the static condition for 8 or 48 h. (F) Nonpermeable MLO-Y4 cells cultured on collagen matrix for 8 or 48 h were coimmunolabeled with α5 and FN antibodies. (Scale bar, 50 μm.) (G) MLO-Y4 cells were cultured for 8 h on collagen-coated plates in FN-depleted (−FN) or normal (Norm) medium. ***P < 0.001, fluid flow in medium with or without FN versus the static condition in medium with or without FN.

Conformational Activation of Integrin α5β1 Through PI3K Activated by Mechanical Stimulation Opens the Cx43 HC.

To establish a role for direct perturbation of integrin α5β1 in a force-dependent regulation of the Cx43 HC, we used magnetic beads coated with either FN, which is the primary substrate for integrin α5, or an antibody against α5, because both bind directly to α5. The diameter of the HC determined by Thimm et al. (23) is around 1.8 nm when closed and 2.5 nm when open; thus the magnetic beads are likely to bind to multiple HCs. The HC opening observed through dye uptake could be a collective action of multiple HCs. Application of a magnetic field induced dye uptake in cells with attached magnetic beads coated with anti-α5 antibody or the α5 substrate FN (Fig. 4A). As controls, we examined beads coated with sheep IgG, anti-CD44 antibody, or polylysine. CD44, a receptor for hyaluronic acid, is expressed on the surface of MLO-Y4 cells as well as in primary osteocytes (24). Control beads failed to induce dye uptake (Fig. 4A, Right). These results indicate that forces on integrin α5, but not other cell-surface molecules, induce the opening of the HC. This effect could be a result of a change in conformation of α5, because force is known to trigger conformational activation of this integrin (13, 25). To examine this hypothesis, we treated cells with TS2/16, an integrin β1-activating antibody (26), in the absence of FN (FN-depleted serum was used) and observed Cx43 HC opening without mechanical stimulation (Fig. 4B). Indeed, α5β1 was activated from 5 min to 30 min of fluid flow as observed by the increase in the binding of GST-FNIII_9–11, a reporter that specifically binds the activated form of α5β1 (Fig. 4C) (27). We then examined whether fluid flow activates α5β directly or whether another component might influence the force that leads to the conformational activation of α5β1. PI3K has been reported to mediate activation of integrins after fluid flow in endothelial cells (27–29), and we demonstrated the activation of PI3K signaling by fluid flow in osteocytes (Fig. S5) (30). Fluid flow-induced Cx43 HC opening, as detected by dye uptake, was inhibited significantly by LY294002, a PI3K signaling inhibitor (Fig. 4D, Right). LY294002 also blocked the interaction between Cx43 and α5 (Fig. 4D, Lower). The activation of α5β1 was reduced effectively in the presence of the PI3K inhibitors LY294002 and wortmannin (Fig. 4E). These results suggest that fluid flow-activated PI3K mediates the activation of integrin α5β1, thereby leading to the opening of the HC.

Fig. 4.

Activation of integrin α5β1 by mechanical stimulation is necessary for HC opening, and fluid flow-activated PI3K signaling is involved. (A) MLO-Y4 cells were incubated with coated magnetic beads, followed by the application of a magnetic field. Dye uptake was quantified only in cells attached to magnetic beads. HC opening was increased significantly in cells attached to the beads coated with α5 antibody or FN but not in cells attached to the beads coated with polylysine (Poly-Lys), sheep IgG, or CD44 antibody. ***P < 0.001, FN and α5 antibody versus poly-lysine and CD44 antibody. (Scale bar, 50 μm.) (B) Cells cultured in serum without FN were incubated with mouse IgG or β1-activating antibody (β1 Ab) alone or were pretreated with either carbenoxolone (Carb) or HC-blocking Cx43 (E2) antibody before dye uptake. **P < 0.01, β1 antibody versus all others. (C) Fluid flow stimulates integrin activation. Lysates of fluid flow-treated cells incubated with GST-FNIII_9–11 were immunoblotted with anti-GST antibody for detection of bound GST-FNIII_9–11 or β-actin. (D) (Upper) Flow-induced HC opening. Compared with the static condition (C), the increased dye uptake in MLO-Y4 cells by fluid flow (FF) was decreased significantly with LY294002 (FF+LY294002). *P < 0.05. (Lower) The interaction between α5 and Cx43 is blocked by LY294002). (E) Fluid flow activates integrin α5β1 through PI3K. MLO-Y4 cells were pretreated for 30 min with the PI3K inhibitors LY294002 (LY, 10 μM) or wortmannin (WM, 100 nM) or were not treated as control (C) before fluid flow for 15 min. Cell lysates were immunoblotted for bound GST-FNIII_9–11 or β-actin. Bound GST-FNIII_9–11 was normalized with β-actin. **P < 0.01. (F) Schematic diagram of the role of integrin α5β1 in regulating Cx43 HC opening. (Left) In the absence of mechanical loading (static condition), the association between α5β1 and Cx43 exists through their C termini; however, the HCs remain closed. (Right) Upon fluid flow, PI3K signaling is activated, leading to activation of α5β1, promoting conformational change of its extended extracellular domain, and subsequently to the opening of the HC, which allows the passage of small bone anabolic factors, such as prostaglandin E₂ (5), that are essential for bone formation and remodeling.

Discussion

Mechanical forces regulate skeletal remodeling through a wide range of biochemical signals. Osteocytes, in particular, are well positioned in the bone to sense the magnitude of mechanical strain and are important for the skeleton's adaptive response to load (2). The shear stress activation of the Cx43 HC is a key element of this response. Here, we show that the integrin α5β1 interacts directly with the Cx43 HC through their respective C termini to promote HC opening in response to shear stress. The opening of the HC is likely a direct consequence of the activation and conformational change of the interacting integrin. Interestingly, recent data showed that mechanical force can trigger the conversion of integrin α5β1 to an activated state that is competent to signal (25). Here we give another dimension to the study by showing that integrin activation can induce Cx43 HC opening through a process that involves PI3K signaling stimulated by mechanical force. However, the process is independent of integrin binding to its extracellular substrate, FN. This study, therefore, establishes a physical and molecular mechanism involving integrin α5β1 in the regulation of mechanotransduction via direct manipulation of HC function, as summarized in Fig. 4F.

In our study, the role of integrin α5 was addressed using siRNA against α5, which abolished the opening of the Cx43 HC in response to fluid flow. Moreover, uncoupling of the interaction between α5 and Cx43 using the C-terminal domain of Cx43 inhibited fluid flow-induced HC opening. Although our results essentially show that Cx43 is not located at the focal adhesion, a previous study showed that Cx43 colocalizes and associates with vinculin, but not with β1, during the migration of cardiac neural crest cells (31). Our study demonstrates a direct and specific interaction between the C termini of Cx43 and α5. The interaction as quantified by SPR, although weak, is specific, because we clearly show that α5 requires activation to open the HC. However, the synthetic α5 C-terminal peptide used for the SPR experiment is not the activated form of α5 and hence might bind to Cx43 with less affinity than does the activated form. Of note, we are working with short protein fragments in solution rather than with native proteins confined to the 2D plasma membrane. Nevertheless, we do show here that mechanical stimulation strengthens the interaction of the two proteins, Cx43 and α5. Increased interaction upon mechanical stimulation could be caused by integrin activation that involves the separation of cytoplasmic tails of integrin α and β (32), thereby allowing the C terminus of α5 to be available for interactions with other proteins such as Cx43.

Bone cells assemble a biologically active ECM rich in collagen, FN, osteopontin, laminin, and other components. It is well known that integrin α5β1 is a receptor for FN (10) and that FN promotes cell attachment. As expected, we found that matrix attachment of osteocytes is necessary for proper HC function induced by fluid flow; however, there is no discernable difference in response to fluid flow between the cells cultured on collagen and cells cultured on FN matrices. Moreover, neither application of FN or RGD peptide to matrix-attached cells nor culturing cells in FN-free medium for a short period affected HC opening induced by fluid flow, suggesting that binding to FN is not essential for α5β1 to open the HC. In this context, it is plausible that α5 has an alternative function. By allowing other integrins to form attachments to the matrix, α5β1 molecules could “free” themselves to participate in other activities. Despite their apparent redundancy (because of the presence of eight integrins that bind to FN), some studies have shown that certain integrins may have unique functions. For example, studies have shown two integrins, αvβ1 and α5β1, both FN receptors with similar attachment functions, can have different activities and locations within the same cell (33). It is also noteworthy that gap-junction communication in cardiac neural crest cells is independent of the FN concentration in the matrix (31). The findings of the present study support the transduction of mechanical forces by integrins, whereas the association of integrin with its substrate appears not to be essential for the opening of the HC. PI3K signaling-dependent activation of integrin α5β1 is essential for the effect of mechanical stimulation on HC opening. PI3K signaling, rather than the integrin, appears to be the upstream element responsive to fluid flow that bears the force-induced signal and transmits it to integrin α5β1, leading to its conformational activation. It appears that conformational activation of integrins, regardless of the types of the mechanical stimulation (i.e., shear stress or nonshear stress), plays a predominant role in the opening of the HC. This idea is supported by previous published studies showing that fluid shear stress on endothelial cells and strain applied to integrins both trigger conformational activation of integrins (28, 34). Further investigation is warranted to elucidate the molecular mechanism governing the fluid flow-induced PI3K signaling and integrin activation. Our study establishes the critical link between mechanotransduction and its response on the anabolic activity of the bone via the collective effect of the integrin and connexin. This knowledge potentially could lead to therapeutics involving the role of the HC targeted at preventing bone loss caused not only by bone disease, such as osteoporosis, but also by the absence of mechanical loading during immobilization, space flight, or aging.

Materials and Methods

Cell Culture and Reagents.

MLO-Y4 osteocytic cells derived from murine long bones were cultured on rat tail collagen type I-coated surfaces and were grown in α-modified essential medium with 2.5% (vol/vol) FBS and 2.5% (vol/vol) BCS (24). FN from the serum was depleted using a gelatin Sepharose column (GE Healthcare). Antibodies against integrin α5 (CD49e; R&D Systems), FN (BD Biosciences), and RGD peptides (Biomol) were used in this study.

Immunofluorescence Labeling and 3D Scanning Confocal Fluorescence Microscopy.

Total cell labeling.

The cells cultured with FN or poly-lysine were fixed in 4% paraformaldehyde (PFA) for 10 min at room temperature, permeabilized with 0.25% Triton X-100, and blocked with 3% BSA. The cells were incubated overnight at 4 °C with affinity-purified antibodies against Cx43CT (1:300), integrin α5 (1:50), integrin β1 (1:500), vinculin (1:300), or paxillin (1:300) and for 1 h with the appropriate secondary antibody.

Surface labeling.

Cultured cells were washed in PBS followed by incubation for 1 h with α5 and FN antibody (1:50). The cells were fixed in 4% PFA and labeled with secondary antibodies in succession for 1 h each at room temperature. Confocal fluorescence with 3D z-stacking scanning was performed on mounted slides using a confocal laser-scanning microscope (Fluoview; Olympus Optical) at a thickness of 0.5 μm.

Immunoprecipitation, Protein Pull Down, and Immunoblotting.

Cultured MLO-Y4 cells were lysed in lysis buffer (5 mM Tris, 5 mM EDTA/EGTA, pH 8.0). Supernatants were incubated with Cx43 or α5 antibodies overnight at 4 °C, followed by incubation on beads for 2 h. Two biotinylated peptides (each 27 amino acids in length; >95% purity) were synthesized to cover the α5 C terminus (LGFFKRSLPYGTAMEKAQLKPPATSDA-biotin) and a scrambled peptide (LGKSATPYAQFGMLTKASELDPRFKPAK-biotin). Biotinylated peptides were conjugated to streptavidin-coupled Dynabeads and were incubated with GST-Cx43CT overnight at 4 °C. Bound proteins were eluted in 0.1% SDS buffer. Immunoprecipitates and elutes from pull down were immunoblotted with anti-Cx43CT (1:300) (35), anti-α5 (1:1,000), or anti-β1 (1:10,000) antibody.

Fluid Flow.

Fluid flow was created by parallel-plate flow chambers separated by a gasket of defined thickness with gravity-driven fluid flow using a peristaltic pump. The thickness of the gasket determined the channel height, which was adjusted along with flow rate to generate stress levels of 16 dyn/cm². The circulating medium was SMEM. The entire flow system was encased within a CO₂ incubator at 5% CO₂ and 37 °C.

Dye-Uptake Assay.

Cells were cultured on collagen type I, FN, and polylysine matrices for 48 h and were subjected to fluid flow at 16 dyn/cm² for 10 min. Dye-uptake experiments were performed as described previously (5). Briefly, cells were incubated with 0.2% LY (M_r ∼547 Da) and 0.2% rhodamine dextran (RD) (M_r ∼10 kDa) dye mixture for 5 min, and a ratio of fluorescent cells to total cells per image was determined using Image J software (National Institutes of Health). Cells cultured on collagen for 48 h were incubated with RGD peptide (1 mM)- and FN (10 μg)-coated Dynabeads for 30 min and 1 h and were subjected to fluid flow followed by dye uptake. Cells cultured on collagen for 8 h using FN-depleted medium or FN-containing medium were subjected to fluid flow for 10 min followed by dye uptake. Dye uptake also was analyzed after incubation with mouse IgG (50 μg/μL) or TS2/16 (50 μg/mL) diluted in SMEM for 30 min alone or after coincubation with carbenoxolone (100 μM) or HC-blocking Cx43 (E2) antibody (1:400). MLO-Y4 cells transfected with Cx43CT-GFP constructs were assayed for dye uptake using a mixture of Alexa Fluor 350 (1 mM) and RD (2%) after fluid flow.

SPR Analysis.

SPR experiments were performed on a Biacore T100 instrument using CM5 sensor chips (GE Healthcare), and the response was measured in resonance units. GST-Cx43CT and GST were immobilized on the chip surface. Binding analysis was performed by injecting in duplicate a concentration series of the purified α5 peptide (Genescript) over the chip, with concentrations ranging from 15 μM–2 mM (in increasing 2× increments), using PBS as a running buffer. All SPR studies were conducted in the Center for Macromolecular Interactions at the University of Texas Health Science Center, San Antonio, TX.

Integrin α5 siRNA.

MLO-Y4 cells were trypsinized and resuspended in antibiotic-free Opti-MEM (Invitrogen). Cells were transiently transfected with integrin α5 siRNA or scrambled siRNA (Ambion) using an siRNA transfection kit (Ambion). Cells were harvested after 48 h of transfection and were assessed for the expression of α5 and β-actin or for HC activity.

Generation of Cx43CT-GFP DNA Construct and Transfection into MLO-Y4 Cells.

The Cx43 C terminus encoding amino acids 231–384 was amplified by PCR and cloned into pcDNA3.1(a) containing GFP to generate Cx43CT-GFP. This construct and the GFP-alone pcDNA3.1 construct were used to transfect MLO-Y4 cells transiently. Cells positive for transfection were identified by GFP expression. For protein analysis, the Cx43CT-GFP construct was transfected into the MLO-Y4 cells using the Neon transfection system (Invitrogen), and the expression of endogenous Cx43 and transfected Cx43CT-GFP was determined by Western blotting with the anti-C terminus of Cx43 (Cx43C) antibody (35) (1:300) or anti-GFP antibody (1:1,000) (Abcam).

Application of Direct Force by Magnetic Field.

Magnetic beads were conjugated with goat anti-sheep IgG, goat anti-mouse IgG, goat anti-CD44, or goat anti-integrin α5 antibody and FN, respectively, for 1 h and were washed three times with PBS. Beads (4 × 10⁶) were added for 30 min to MLO-Y4 cells at low density. An electromagnetic tweezer then was used to apply a magnetic field to the beads. This device consists of ∼6,300 ft of AWG-30 magnet wire providing 648 Ω of electrical resistance spooled onto a core made of superparamagnetic stainless steel Magival (kindly provided by Jamey Duncan, Valbruna Stainless Steel, Milford, OH). A pole piece milled from mixed steel and soft iron composite was inserted into the Magival core. The electromagnet was positioned with a micromanipulator so that the pole piece was to the right of the acquired image. A magnetic field generating ∼50–200 pN of force per bead was applied to the cells for 30 min in the presence of LY and RD. Cells were fixed and observed under fluorescence microscopy, and the number of bead-attached cells with dye uptake was counted and calculated.

Integrin Activation Assay Using GST-FNIII_9–11.

MLO-Y4 cells cultured for 48 h were subjected to fluid flow at 16 dyn/cm². Integrin activation assay was performed according to the published method (27). Briefly, flow-treated cells were incubated for 30 min at 37 °C with GST-FNIII_9–11 (20 μg/mL) and were washed; lysates were immunoblotted with anti-GST (1:5,000) or anti-β-actin (1:1,000). To determine the involvement of PI3K signaling, cells were pretreated with PI3K inhibitors (LY294002 or wortmannin) for 30 min before fluid flow followed by incubation with GST-FNIII_9–11.

Statistical Analysis.

All data were analyzed using GraphPad Prism 5.04 software (GraphPad). One-way ANOVA and the Student–Newman–Keuls test were used for comparisons of more than two groups, and a paired Student's t test was used for comparisons between two groups. Unless otherwise specified in the figure legends, the data are presented as mean ± SEM of at least three determinations.