Integrin Activation and Matrix Binding Mediate Cellular Responses to Mechanical Stretch*

Mechanical tension is a critical determinant of cell growth, differentiation, apoptosis, migration, and development. Integrins have been implicated in sensing force but little is known about how forces are transduced to biochemical signals. We now show that mechanical strain stimulates conformational activation of integrin αvβ3 in NIH3T3 cells. Integrin activation is mediated by phosphoinositol 3-kinase and is followed by an increase in integrin binding to extracellular matrix proteins. Mechanical stretch stimulation of JNK was dependent on new integrin binding to extracellular matrix. These data define a molecular mechanism for the role of integrins in mechanotransduction.

It is well known that mechanical forces including tension and compression are critical to the normal growth and function of many tissues (1) and to pathological states such as cardiac hypertrophy (2) and atherosclerosis (3). Classical studies of cellular responses to mechanical stimuli using cells on elastic substrata demonstrated that fibroblasts orient along the axis of stretch (4), whereas cardiac myocytes elongate and orient perpendicular to the direction of stretch (5). Tension also induced neurite growth in the direction of applied force (6), whereas rearward tension applied to cells suppressed protrusive activity at the leading edge and caused alignment of actin microfilaments parallel to the tension (7). Gene expression in early Drosophila embryos can be modulated by both exogenously and endogenously applied mechanical strains (8). However, the molecular mechanisms by which cells transduce mechanical forces into biological responses are poorly understood.
Integrins are of particular interest in mechanotransduction because they both function as signaling receptors and physi-cally connect the cytoskeleton to the extracellular matrix, thus transmitting physical forces (9). Magnetic beads coated with a ␤1-integrin ligand showed force-dependent focal adhesion formation and stiffening, indicating that integrins can transfer external force through the plasma membrane (10). Mechanical tension locally applied to integrins induced the translocation of mRNA and ribosomes to points of force application (11), and a number of responses to strain are specific to particular ECM proteins (12). Mechanical stretch has been proposed to change the conformation or clustering of integrins (9) but this hypothesis has not been directly tested. We therefore sought to determine the role of integrin conformation and matrix binding in mechanotransduction.

EXPERIMENTAL PROCEDURES
Materials-Reagents were purchased from Sigma-Aldrich unless otherwise noted. WOW-1, LIBS-6, and LM609 were generous gifts from Drs. Sanford J. Shattil, Mark H. Ginsberg, and David A. Cheresh, respectively (Scripps Research Institute). Anti-FN antibodies 16G3 and 11E5 were generous gifts from Dr. Kenneth M. Yamada (National Institutes of Health). Anti-PI3K 1 p85 rabbit antibody was from Upstate biotechnology (Lake Placid, NY). Phospho-Akt (Ser 473 ) monoclonal antibody, phospho-JNK (Thr 183 /Thr 185 ) polyclonal antibody and JNK polyclonal antibody were from Cell Signaling Technology (Beverly, MA). Anti-Akt1 monoclonal antibody was from Santa Cruz Biotechnology (Santa Cruz, CA). PY20 was from BD Bioscience. Wortmannin and LY-294002 were from Calbiochem. Signal PIP TM kits was purchased from Echelon Biosciences Inc. (Salt Lake City, UT). Transparent silicone elastic membrane was purchased from Specialty Manufacturing, Inc. (Saginaw, MI).
Cell Culture and Stretch Procedures-NIH3T3 cells were maintained as described previously (13). The equibiaxial stretch devices were described previously (14). Briefly, a silicone membrane is attached to the membrane holder to form the bottom of the device. Indentation of the ring against the membrane results in a homogenous planar equibiaxial stretch of the membrane. Strain along circumferential and radial axes were equal in magnitude and homogenously distributed (14). The strain was applied over ϳ10 s, which is the time required to manually turn the rings (13). Strain remained constant for the duration of the experiments. Membranes were coated with either 20 g/ml fibronectin (FN) for 1h at 37°C, or with 10 g/ml fibrinogen (FBG) in the presence of 5 g/ml aprotinin for 16 h at 4°C.
Detection of Activated or Occupied Integrins-NIH3T3 cells were plated on silicone elastic membranes attached to an equibiaxial stretch device and coated with either FN or FBG, then incubated overnight in 0.5% FBS/DMEM. Cells were then left untreated or stretched to increase cell area by 15% and strain was maintained for the duration of the experiment. Cells were incubated with His-tagged recombinant WOW-1 Fab (30 g/ml) or LM609 (12 g/ml) for 30 min (15). Cells were washed in PBS and lysed in SDS sample buffer with 50 mM DTT. Lysates were sonicated for 10 -15 s to shear DNA and analyzed by Western blotting using His-probe (rabbit polyclonal antibody against His tag, horseradish peroxidase conjugate, Santa Cruz Biotechnology) to detect bound WOW-1. For the experiments with wortmannin or LY-294002, cells were preincubated with either 100 nM wortmannin or 5 M LY-294002 for 1 h at 37°C before stretching. For the experiments with intracellular delivery of phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ) (16), NIH3T3 cells were plated on FN-coated plastic dishes and incubated overnight in 0.5% FBS/DMEM. Cells were then incubated with WOW-1 and 5 M histone H1 as a carrier, with or without 3 M dipalmitoyl (di-C16)-PIP 3 for 30 min, and bound WOW-1 was quantified as above.
For detection of occupied integrins, NIH3T3 cells plated on FBG-or FN-coated silicone membrane were incubated overnight in 0.5% FBS. Cells were treated with or without 15% stretch, then incubated with LIBS-6 (5 g/ml) for 20 min at 37°C. Cells were rinsed, lysed with SDS sample buffer, and bound LIBS-6 detected by Western blotting using goat anti-mouse Ig-HRP conjugate (Jackson ImmunoReseach). Densitometry analysis was performed using Scion image software (Scion Corp.).
Activation Assay of Akt and JNK-NIH3T3 cells plated on FN-coated silicone membrane were incubated overnight in 0.5% FBS. Cells were treated with or without stretch, rinsed with PBS, and lysed in SDS sample buffer containing 50 mM DTT, and 1 mM concentration each of Na 3 VO 4 , NaF, and Na 2 P 2 O 7 . Lysates were analyzed by Western blotting with polyclonal antibodies against phospho-JNK (Thr 183 /Thr 185 ), JNK, and monoclonal antibodies against phospho-Akt and Akt1.
Activation Assay of PI3K-Cells were serum-starved overnight and treated with or without 15% stretch for the indicated times. Cells were lysed with 500 l of lysis buffer (50 mM Tris, pH 7.0, 0.5% Nonidet P-40, 500 mM NaCl, 1 mM MgCl 2 , 1 mM EGTA, 10 g/ml aprotinin/leupeptin, 1 mM concentration each of phenylmethylsulfonyl fluoride, Na 3 VO 4 , NaF, and Na 2 P 2 O 7 ). Lysates were precleared by adding 10 l of normal rabbit serum (Dako) and 30 l of protein A-Sepharose slurry (Amersham Biosciences), incubating for 1 h at 4°C, and centrifuging to remove the beads. Antibody against the p85 subunit of PI3K was added to the supernatant, and samples were incubated for 1 h at 4°C. Protein A-Sepharose slurry (30 l) was then added to each sample, they were incubated for 1 h at 4°C, and beads were collected by low speed centrifugation and washed three times with lysis buffer. Samples were analyzed by SDS-PAGE on 7.5% gels and Western blotting with PY20 anti-phosphotyrosine antibody or anti-p85 antibody.
Blocking of Unoccupied FN-NIH3T3 cells plated on FN-coated silicone membrane were starved overnight in 0.5% FBS/DMEM. Prior to the experiments, medium was aspirated, and cells were incubated with fresh DMEM containing 0.5% FBS for 2 h. Cells were washed with PBS and incubated as indicated with DMEM or DMEM containing 20 g/ml 16G3 or 11E5 for 15 min. Cells were washed with PBS twice, and fresh DMEM containing 0.5% FBS was added. Cells were then stretched by 15% for 10 min, washed with PBS twice, and lysed with SDS sample buffer with 50 mM DTT and 1 mM Na 3 VO 4 . Lysates were then applied to SDS-PAGE and detected by phospho-JNK (Thr 183 /Thr 185 ) or total JNK polyclonal antibody.

Mechanical Stretch Increases High Affinity ␣v␤3
Integrin-An engineered Fab fragment, WOW-1, reacts selectively with high affinity ␣v␤3 that is not bound to ECM ligands (17). Since WOW-1 is monovalent, it is relatively insensitive to changes in integrin clustering and reports mainly changes in affinity of ␣v␤3 integrin. To test whether mechanical strain alters affinity of integrin ␣v␤3, subconfluent monolayers of NIH3T3 cells plated on an elastic membrane coated with fibronectin were subjected to biaxial stretch by 15% or kept untreated. Strain was maintained for the duration of the experiment. Cells were incubated with WOW-1 for 30 min, rinsed, and lysed, and the amount of bound WOW-1 was quantitated by Western blotting (Fig. 1A). Strain increased WOW-1 binding by ϳ8-fold (Fig. 1A). Similar results were obtained with cells on fibrinogen (data not shown). As a control to test the specificity of binding, EDTA was added. As expected, EDTA abolished binding of WOW-1 to stretched cells (Fig. 1A). LM609, an antibody against ␣v␤3 integrin that is insensitive to integrin conformation, showed no change in binding after stretch (Fig. 1B), excluding changes in integrin surface expression. These results indicate that mechanical stretch induced conformational activation of integrin ␣v␤3.
Stretch Increases Integrin Binding to ECM-We then investigated whether stretch increases ␣v␤3 integrin binding to ECM proteins. The antibody LIBS-6 has higher affinity for ligand-occupied compared with unoccupied ␤3 integrins (18). LIBS-6 also stabilizes the ligand-occupied conformation of integrins and can therefore activate them, but at the low concentration and short times used in these experiments, it serves primarily as a read-out for integrin ligation to ECM (15). NIH3T3 cells plated on elastic substrata were stretched and the binding of LIBS-6 was assayed (Fig. 1C). Stretch increased LIBS-6 binding to NIH3T3 cells that had been plated on FBG (Fig. 1C) or FN (data not shown). Thus, integrin activation by stretch is associated with increased integrin-ligand binding.
Activation of PI3K/Akt Pathway by Stretch-In many cellu- lar systems, conversion of integrins to the high affinity state is triggered by activation of PI3K (19,20). Additionally, PI3K has been reported to be activated by stretch (21). We therefore measured PI3K phosphorylation on tyrosine following stretch in our system. The p85 subunit of PI3K was phosphorylated at early times (ϳ3 min) after stretch then declined to baseline ( Fig. 2A). Akt, which is downstream of PI3K, was also activated after stretch and its activity was more sustained (Fig. 2B). To test the functional significance of PI3K activity, NIH3T3 cells were preincubated with or without PI3K inhibitors, cells were stretched, and WOW-1 binding was assayed. Both wortmannin and LY-294002 significantly inhibited the stretch-induced increase of WOW-1 binding (Fig. 2C). Moreover, intracellular delivery of PIP 3 increased WOW-1 binding (Fig. 2D). Taken together, these results indicate that PI3K mediates stretchinduced activation of ␣v␤3 integrin.
New ECM Binding Mediates Stretch-induced JNK Activation-Previous studies found that static biaxial stretch of cardiac fibroblasts activated mitogen-activated protein kinases in an ECM-specific manner; ERK2 was stimulated only in cells plated on FN; c-Jun NH 2 -terminal kinase (JNK) was activated on FN, vitronectin or laminin; and cells on collagen did not activate either kinase (22). We confirmed that stretching 3T3 cells induced activation of JNK, as indicated by phosphorylation of p54 and p46 after 10 min of stretch, as reported (Fig. 3A) (22). p38 mitogen-activated protein kinase phosphorylation was not significantly affected by mechanical stretch (data not shown). We then tested whether the change in JNK activity was downstream of new integrin binding to ECM. NIH3T3 cells were plated on FN for 16 h to allow adhesion and spreading. To block unoccupied FN sites, the cells were incubated for 15 min with either function blocking (16G3) or non-blocking (11E5) anti-FN antibodies. This protocol did not noticeably induce cell rounding or alter cytoskeletal organization (data now shown). Cells were then stretched and JNK phosphorylation assayed (Fig. 3B). The blocking antibody 16G3 strongly inhibited the stretch-induced increase in JNK activity, whereas non-blocking antibody 11E5 had no effect (Fig. 3B). The effects of shear stress induced by washing out antibodies were negligible, since a "no shear" control in which the medium was not exchanged showed similar JNK activity to washed cells with no antibody or 11E5 (Fig. 3B). These results show that JNK activation by stretch is downstream of new integrin-ligand binding.

DISCUSSION
Previous studies have provided evidence that integrins are involved in cellular responses to mechanical forces (reviewed in Ref. 12). For example, some responses to mechanical strain depend on the ECM to which cells are adhered and inhibitors of focal adhesion signaling can block mechanotransduction. However, mechanisms of integrin-dependent mechanotransduction have not been elucidated. In this study, we demonstrate that integrin ␣v␤3 is activated by strain. This effect is due to affinity modulation as shown by increased binding to by the monovalent, conformation-sensitive Fab fragment, WOW-1. Stretch also induced an increase of LIBS-6 binding after stretch, evidence that conformational activation leads to new binding of integrins to the ECM. PI3K and Akt are both rapidly activated by strain, and PI3K inhibition decreases integrin activation. In addition, the delivery of PIP 3 into cells induces ␣v␤3 integrin activation. These data strongly suggest that PI3K mediates ␣v␤3 integrin activation. How PI3K stimulates integrin activation is unknown; however, it may be significant that the small GTPase Rap1, which has been shown to contribute to integrin activation in several cell lines, can be activated by a PI3K-dependent mechanism in platelets (23) and has been implicated in cell response to strain (24). Further investigation of a possible link may therefore be warranted. The blocking anti-FN antibody 16G3 strongly inhibited the stretch-induced activation of JNK under conditions where existing adhesions did not appear to be disrupted, indicating that new integrin-ECM is required for JNK activation.
Taken together, these data define a pathway by which early activation of PI3K, through induction of integrin activation and ECM binding, stimulates a cytoplasmic signaling pathway implicated in cellular responses. It is likely that this mechanism explains the observation that mechanotransduction in a number of systems is sensitive to short term treatment with integrin antagonists that would not be expected to disrupt existing integrin-ECM bonds (12,25,26). This mechanism is likely to be relevant to many systems where mechanical forces influence ECM accumulation and organization, cytoskeletal organization, gene expression, and cell motility. Furthermore, identification of PI3K as an early target of mechanical strain should facilitate future investigation into more proximal mechanisms of mechanotransduction.