Conversion of Mechanical Force into Biochemical Signaling*

Received for publication, June 21, 2004, and in revised form, September 20, 2004Published, JBC Papers in Press, October 14, 2004, DOI 10.1074/jbc.M406880200

Bing Han‡§, Xiao-Hui Bai‡, Monika Lodyga‡, Jing Xu‡, Burton B. Yang¶, Shaf Keshavjee‡,Martin Post�, and Mingyao Liu‡**

From the ‡Division of Cellular and Molecular Biology, Toronto General Research Institute, University Health Network;Toronto, Ontario M5G 2C4, Canada, the ¶Sunnybrook and Women’s College Health Sciences Center, Toronto, OntarioM4N 3M5, Canada, and the �Hospital for Sick Children, Toronto, Ontario M5G 1X8, Canada

Physical forces play important roles in regulating cellproliferation, differentiation, and death by activatingintracellular signal transduction pathways. How cellssense mechanical stimulation, however, is largely un-known. Most studies focus on cellular membrane pro-teins such as ion channels, integrins, and receptors forgrowth factors as mechanosensory units. Here we showthat mechanical stretch-induced c-Src protein tyrosinekinase activation is mediated through the actin fila-ment-associated protein (AFAP). Distributed along theactin filaments, AFAP can directly active c-Src throughbinding to its Src homology 3 and/or 2 domains. Muta-tions at these specific binding sites on AFAP blockedmechanical stretch-induced c-Src activation. Therefore,mechanical force can be transmitted along thecytoskeleton, and interaction between cytoskeletal as-sociated proteins and enzymes related to signal trans-duction may convert physical forces into biochemicalreactions. Cytoskeleton deformation-induced protein-protein interaction via specific binding sites may repre-sent a novel intracellular mechanism for cells to sensemechanical stimulation.

Sensing external and internal stimuli is a vital sign of livingorganisms. Physical forces, derived from or applied to cells, areessential signals in determining cellular structure, prolifera-tion, differentiation, and survival. Physical forces are involvedin the regulation of many physiological processes such as thematuration and branching of fetal lung (1), the formation ofventricles and atriums of the heart, angiogenesis (2), microvas-cular remodeling (3), and the maturation of bone (4) and car-tilages (5). Exposure to abnormal forces contributes to thepathogenesis of diseases such as ventilator-associated lunginjury (6), cardiac hypertrophy (7), atherosclerosis (8), etc. Overthe last two decades, mechanotransduction has become an im-portant area of research. How cells sense mechanical signalsand convert them into biochemical reactions for signal trans-duction is one of the fundamental questions.

Specialized mechanosensory structures have been developedthrough evolution, such as touch receptors in Caenorhabditis

elegans (9, 10) and bristle receptors in Drosophila (11, 12). Invertebrates, hair cells in the inner ear (13) and skin mechano-receptors (14) have been extensively studied. Recently, the cellmembrane proteins polycystin-1 and polycystin-2 have beenfound to mediate mechanosensation in the primary cilium ofkidney cells (15). In these structures, transduction channelsconnected to intracellular and extracellular anchors control theentry of ions, thus converting a mechanical stimulus into anelectrical signal, an alteration of membrane potential (13).

It has been believed that all cells in the body can respond tomechanical stimulation. How regular cells sense mechanicalsignals is unclear. Stretch-activated and -inactivated ion chan-nels have been found in many cell types (16). In response tomechanical forces, opening and closing of these channels canactivate intracellular signal transduction (17). However,Sawada and Sheetz used a Triton buffer to remove cytoplasmand apical cellular membrane and then applied a transientmechanical stretch to the Triton-insoluble cytoskeleton. Theydemonstrated a stretch-dependent binding of paxillin, focaladhesion kinase (pp125FAK), and p130CAS to the cytoskeleton(18). Therefore, in addition to ion channels there are othermechanisms that enable cells to sense physical forces. Cellsattach to the extracellular matrix (ECM)1 via integrins thatlink to the cytoskeleton. This complex provides a structuralconnection allowing the transmission of physical forces fromthe ECM to the cell interior. There is increasing evidence tosupport the belief that integrins are mechanosensors (2, 8), buthow physical forces are converted into biochemical signalsthrough the ECM-integrin-cytoskeleton complex is not ad-dressed by this model system. Recently, Tschumperlin et al.reported that compressive stress shrinks the lateral intercellu-lar space surrounding epithelial cells and triggers cellular sig-naling via autocrine binding of epidermal growth factor (EGF)family ligands to the EGF receptor (19). This observation againsupports the importance of cellular membrane proteins inmechanosensory processes.

We have previously found that mechanical stretch rapidlyactivates c-Src in fetal rat lung cells and that blocking stretch-induced activation of protein tyrosine kinases reduces stretch-mediated fetal lung cell proliferation (20). We also noted thatmechanical stretch increases the binding of c-Src to the actinfilament-associated protein (AFAP) (20). Based on the molecu-lar structure of c-Src and AFAP, we hypothesized that thecytoskeletal structure not only can transmit physical forces

* This work was supported by Canadian Institutes of Health Re-search Operating Grants MT-13270 and MOP-42546. The costs of pub-lication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked “advertisement”in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

§ Recipient of the Canadian Institutes of Health Research FellowshipAward.

** Recipient of the Premier’s Research Excellence Award from theOntario Government and to whom correspondence should be addressed:University of Toronto, Toronto General Hospital, Rm. MBRC5R422,200 Elizabeth St., Toronto, Ontario M5G 2C4, Canada. Tel.: 416-340-3495; Fax: 416-340-4768; E-mail: mingyao.liu@utoronto.ca.

1 The abbreviations used are: ECM, extracellular matrix; AFAP, ac-tin filament-associated protein; AFAP5Y, AFAP with mutants Y93F/Y94F/Y125F/Y451F/Y453F; AFAP71A, AFAP with mutant P71A;AFAP77A, AFAP with mutant P77A; cAFAP, chicken AFAP; EGF,epidermal growth factor; GFP, green fluorescent protein; GST, gluta-thione S-transferase; hAFAP, human AFAP; SH, Src homology; siRNA,small interfering RNA.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 279, No. 52, Issue of December 24, pp. 54793–54801, 2004© 2004 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 54793

intracellularly but can also convert physical forces into bio-chemical reactions for signaling via specific binding sites. Thepresent study tested this concept with supportive evidence thatmechanical signals can be sensed intracellularly through pro-tein-protein interaction.

EXPERIMENTAL PROCEDURES

Cell Culture, Transfection, and Mechanical Stretch—The pCMV1expression constructs for c-Src, chicken AFAP (cAFAP), and its mutants(AFAP71A, AFAP77A, and AFAP5Y) were gifts from Dr. Flynn (WestVirginia University) (21). The full-length human AFAP (hAFAP) codingsequence was subcloned into the pCMV1 vector by replacing the resi-dent KpnI-XbaI fragment of the pCMV-AFAP construct (21). ThepEGFP-C3 eukaryotic expression vector (Clontech) was used to expresscAFAP and its mutants fused to green fluorescent protein (GFP) byshuttling cAFAP, AFAP71A, AFAP77A, and AFAP5Y sequences frompCMV vectors to pEGFP-C3 vector, respectively, using EcoRI sub-cloning sites (22). Small interfering RNA (siRNA) was designed accord-ing to the nucleotides 311–331 of the human AFAP sequence (Gen-BankTM access number AF188700). The AFAP siRNA duplex with asense sequence of 5�-GCUCCGAAUACAUCACAU(dTdT)-3� and a non-specific control double-stranded RNA sequence of 5�-GAAUCCGCU-GAUAAGUGAC(dTdT)-3� were synthesized by Dharmacon (Lafayette,CO).

Cells were cultured in Dulbecco’s modified Eagle’s medium (Invitro-gen) supplemented with 10% fetal bovine serum in a humidified atmo-sphere at 37 °C and 5% CO2. For transfection, cells were seeded in6-well plates (2 � 105 cells/well) overnight. The cells were transfectedusing Lipofectamine reagent for DNA and Oligofectamine for RNAtransfections (Invitrogen) following the manufacturer’s protocols.Transfected cells were maintained in Dulbecco’s modified Eagle’s me-dium containing 5% fetal bovine serum and harvested 60 h later. Cells,with or without transfection, cultured on collagen I- or ProNectin (asynthetic peptide fragment of fibronectin)-coated BioFlex plates were

subjected to mechanical stretch (25% elongation, 60 cycles/min for 10min) with a Flexercell Strain Unit, FX-2000TM (Flexcell International,Hillsborough, NC). Cells cultured on the same type of plates served asnon-stretch controls.

Western Blotting Analysis and Immunoprecipitation—Cells were ly-sed with modified radioimmune precipitation assay buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EGTA, 2 mM EDTA, and 1% TritonX-100) containing 10 �g ml�1 each aprotinin, leupeptin, pepstatin, 1mM phenylmethylsulfonyl fluoride, 1 mM Na3VO4, and 10 mM NaF. Forisolating the Triton-insoluble cytoskeletal fraction, whole cell lysateswere centrifuged at 4 °C for 10 min. The insoluble pellets were dissolvedin Laemmli sample buffer. Protein content was determined by a stand-ard protein assay (Bio-Rad). Lysates containing equal amount of totalprotein were boiled in Laemmli sample buffer, subjected to SDS-PAGE,and then transferred to nitrocellulose membranes. Immunoblotting wasconducted with antibodies against protein phosphotyrosine (4G10), Src(GD11) (Upstate Biotechnology, Lake Placid, NY), Src phosphoty-rosines 416 and 527 (Biosource International, Camarillo, CA), and actin(AC-40) (Sigma) at 1:1,000 dilution. Polyclonal antibody against AFAP(F1) was a gift from Dr. Flynn and was used at 0.5 �g ml�1. Blots weredetected with horseradish peroxidase-conjugated secondary antibodiesand developed with an enhanced chemiluminescence detection kit (Am-ersham Biosciences). Membranes were stripped off and re-probed withother antibodies.

For immunoprecipitation, whole cell lysates containing 1 mg of totalprotein were adjusted to equal volume (1 ml) and incubated with des-ignated antibody (2 �g) at 4 °C overnight. Immune complexes wererecovered by incubating with 50 �l of protein A-Sepharose beads (20%;w/v) for polyclonal antibodies or 100 �l of protein G-Sepharose beads(10%; w/v) for monoclonal antibodies for 1 h at 4 °C under gentleagitation. The immunoprecipitates were washed twice with radioim-mune precipitation assay buffer. Proteins were eluted by boiling theprecipitates in sample buffer and then subjected to SDS-PAGE andimmunoblotted as described above.

c-Src Kinase Assay—After cell transfection, 10 �l of cell lysate from

FIG. 1. Mechanical stretch activatesc-Src. a, mechanical stretch increasedphosphorylation of c-Src Tyr-416 (SrcpY416) (a sign of c-Src activation) withoutchanging the phosphorylation of Tyr-527(Src pY527). Five cell lines were culturedon collagen I-coated BioFlex plates andsubjected to mechanical stretch (25% elon-gation at 60 cycles/min for 10 min). West-ern blotting was conducted with whole celllysates, and the membranes were subse-quently probed with the indicated anti-bodies. b, mechanical stretch induced c-Src translocation to the cytoskeleton.COS7 cells were subjected to mechanicalstretch or static culture. The expression ofSrc and actin in Triton-insoluble cytoskel-etal fractions of cell lysates was deter-mined by Western blotting. c, mechanicalstretch increased binding between c-Srcand AFAP. C3H10T1/2-5H cells were sub-jected to mechanical stretch or static cul-ture. Cell lysates were immunoprecipi-tated (IP) with antibodies against Src orAFAP, and the binding between c-Src andAFAP was examined by Western blotting.d, a model is proposed to illustrate howmechanical stretch-induced deformationof the cytoskeleton presents AFAP to c-Src. The high affinity binding betweenAFAP and c-Src via the SH3 and SH2domains leads to the change of c-Src con-figuration and activation.

Intracellular Mechanosensation54794

each sample was used for Src kinase assay as described previously (20). Todetermine whether AFAP can directly activate c-Src, an in vitro reconsti-tution assay was performed (23). Briefly, 20 ng of purified and C-terminalSrc kinase-inactivated c-Src (a gift from Dr. X. Huang, Cornell University)in Src kinase buffer (30 mM HEPES, pH 7.4, 5 mM MgCl2, 5 mM MnCl2,and 10 �M ATP) was incubated with 2 �g of Src substrate peptide (24) and10 �Ci of [�-32P]ATP with purified recombinant cAFAP or AFAP71A (22)in a total 20-�l reaction volume at 30 °C for 15 min. The reaction wasterminated by adding Laemmli sample buffer. After denaturing at 90 °Cfor 5 min, the samples were subjected to 20% SDS-PAGE to separate thesubstrate peptide. The gel was dried, autoradiographed, and quantifiedwith a GS-690 densitometer (Bio-Rad).

Immunofluorescent Staining and Microscopy—Immunofluorescentstaining was conducted at room temperature. After different treat-ments, cells were fixed in 3.7% formaldehyde for 10 min, permeabilizedwith 0.25% Triton X-100 for 5 min, and then stained with indicatedprimary antibody for 60 min and the proper secondary antibody at1:1,000 dilution (v/v) for 30 min in the dark. Slides were mounted withan anti-fading reagent, SlowFade, (Molecular Probes, Eugene, OR),followed by fluorescent microscopic examination (Nikon Canada,Toronto, Canada). The specificity of staining was determined by replac-ing the primary antibodies with nonspecific rabbit or mouse IgG(Sigma).

For detecting the distribution of endogenous AFAP, A549 orC3H10T1/2 cells were cultured on glass cover slips and stained withanti-AFAP antibody (F1; 5 �g ml�1) and fluorescein isothiocyanate-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch Laborato-ries, West Grove, PA). For confirming the intracellular distribution ofGFP-AFAP and its mutants, cells were cultured on glass coverslips andtransfected with either pEGFP vector alone or pEGFP vectors contain-ing fused cAFAP or its mutants (AFAP71A, AFAP77A, and AFAP5Y)for 48 h. Cells were co-stained with rhodamine-phalloidin (MolecularProbe) at 1:1,000 dilution (v/v) for 30 min to image actin filaments.

For evaluating the role of AFAP in stretch-induced c-Src activation,cells were cultured on collagen-coated BioFlex plates and transfectedwith either pEGFP vector alone or pEGFP vectors containing fused

AFAP71A, AFAP77A, and AFAP5Y mutants for 48 h and then exposedto mechanical stretch or static control, as described above. The trans-fected cells were identified by GFP. The stretch-induced activation ofSrc was shown by staining with anti-Src phosphotyrosine 416 antibodyfollowed by goat anti-mouse IgG conjugated with Alex 594 (MolecularProbe).

Statistical Analysis—Data are expressed as mean � S.D. from atleast three experiments and analyzed by one-way analysis of variancefollowed by a Student-Neuman-Keuls test with significance defined asp � 0.05.

RESULTS

Mechanical Stretch-induced c-Src Activation and c-Src/AFAP Interaction—C3H10T1/2, C3H10T1/2-5H, NIH3T3,COS7, and A549 cells were cultured on collagen I-coated flex-ible plates and subjected to mechanical stretch (25% elonga-tion, 60 cycles/min for 10 min) with a Flexcell Strain Unit.Autophosphorylation of Src Tyr-416 at the activation loop is acritical step leading to full activation of c-Src (25), which wasincreased in these cells by stretch regardless of the expressionlevels of total Src protein (Fig. 1a). Receptor protein tyrosinephosphatase � has been reported as a transducer of mechanicalforce on integrin-cytoskeleton linkage (26), which can activatec-Src by reducing its phosphorylation on the Tyr-527 residue.In the present study, the phosphorylation of c-Src Tyr-527 wasnot decreased after cell stretch (Fig. 1a), suggesting that thec-Src activation is not secondary to the activation of proteintyrosine phosphatases. Similar results were obtained whenthese cells were cultured on fibronectin fragment-coated plates(data not shown). Therefore, c-Src activation appears to be acommon phenomenon in different cell types responding to me-chanical stimuli on different ECMs.

FIG. 2. Molecular cloning and char-acterization of human AFAP. a, sche-matic diagram for Src SH3, SH2, and ac-tin binding motifs (Bmf.) in AFAP. b,differential expression of hAFAP in vari-ous human tissues. Manufacturer-mademembranes cross-linked with equalamounts of human RNA (�10% variation)were probed for expression of hAFAP byNorthern blotting. c, co-localization ofhAFAP with actin filaments. A549 cellswere stained with an anti-AFAP antibody(a) and rhodamine-phalloidin for the actinfilament (b). The merged image shows thealignment of hAFAP along with actin fil-aments (c). d, tyrosine phosphorylation ofhAFAP and its binding with c-Src. A549cell lysates were immunoprecipitated (IP)with the indicated antibodies, and theblots (IB) were probed with the indicatedantibodies. pTyr, phosphotyrosine; NRS,nonspecific rabbit serum. e, binding ofhAFAP to the Src SH2 and SH3 domains.A549 cell lysates were incubated with theGST protein alone or the GST proteinfused with the Src SH2, SH3, or SH3/SH2domains, respectively. The precipitateswere analyzed by immunoblotting withanti-AFAP antibody (top). For proteinloading control, after transferring pro-teins to the blotting membranes the gelwas stained with Coomassie Blue(bottom).

Intracellular Mechanosensation 54795

Using COS7 cells as an example, we demonstrated thatstretch for 10 min induced an increase of c-Src in a Triton-insoluble cytoskeletal fraction (Fig. 1b). AFAP was first clonedfrom chicken. It contains an actin-binding motif and is localizedto actin filaments, the cortical actin matrix, and along theleading edge of the cell (27). This distribution pattern suggeststhat AFAP is suitable for transmitting physical forces along theactin filaments and that it may mediate c-Src translocation tothe cytoskeletal fraction through protein binding. UsingC3H10T1/2-5H cells, we demonstrated that the binding of c-Srcand AFAP, as determined by co-immunoprecipitation followedby immunoblotting, increased after mechanical stretch (Fig.1c). Similar results were obtained with COS7 cells (data notshown).

Crystallographic structure studies have revealed that thekinase activity of c-Src is maintained at a low basal level by twomajor intramolecular interactions; one is the binding of its SH3domain to the linker between the SH2 domain and the kinasedomain, and the other is the binding of its SH2 domain to thephosphorylated tyrosine residue 527 in its C-terminal tail (28–30) (Fig. 1d). Interruption of these interactions with high af-finity ligands for either the Src SH2 or SH3 domain mayactivate the enzyme. AFAP contains Src SH3 and SH2 bindingmotifs (21). Therefore, it may competitively bind to the c-SrcSH2 and SH3 domains and induce an alteration of c-Src con-figuration, leading to c-Src activation. The interaction between

AFAP and c-Src may be an important mechanism for mechan-ical stretch-induced c-Src activation (Fig. 1d).

Cloning and Characterization of Human AFAP Protein—Todetermine the molecular structure of AFAP in mammaliancells, we cloned the hAFAP gene from human lung epithelial(A549) cells. The GenBankTM accession number of hAFAP isAF188700. The identity of peptide sequences between humanand chicken AFAP is 87%. All of the functional motifs anddomains described for chicken AFAP protein (31) are present inhAFAP in the same order (data not shown). The proline-richSrc SH3 binding motif, five putative tyrosines for Src SH2binding, and an actin binding region are presented in Fig. 2a.The hAFAP gene is localized on human chromosome 4p16, asconfirmed by PCR-based sequence-tagged sites serving as land-marks of genomic maps and by fluorescent in situ hybridization(data not shown). The steady state mRNA levels of hAFAP aredifferentially expressed in human tissues (Fig. 2b). The endog-enous hAFAP protein is co-localized with actin filaments inA549 cells (Fig. 2c), as determined by double staining withrhodamine-conjugated phalloidin to decorate actin stress fibersand the antibody for AFAP followed by fluorescein isothiocya-nate-conjugated anti-rabbit IgG to highlight hAFAP.

The endogenous hAFAP protein in A549 cells is tyrosine-phosphorylated and binds to c-Src as determined by immuno-precipitation and immunoblotting (Fig. 2d). To determine thebinding affinity of Src SH3 and SH2 domains to hAFAP, bac-

FIG. 3. AFAP directly activatesc-Src. a, co-expression of AFAP and c-Srcincreased total protein tyrosine phospho-rylation and Src Tyr-416 phosphoryla-tion. Constructs containing vector,hAFAP, cAFAP, or c-Src alone or hAFAP(or cAFAP) together with c-Src (or c-Src-KD, kinase inactive c-Src) were trans-fected into COS7 cells. The protein tyro-sine phosphorylation (pTyr) status, thephosphorylation of the Tyr-416 (Src-pY416) or Tyr-527 (Src-pY527) of c-Src,and the expression of c-Src and AFAP pro-tein levels were revealed by Western blot-ting. b, the Tyr-416 phosphorylation ofc-Src was significantly increased when co-expressed with either hAFAP or cAFAPin COS7 cells. Data were quantified bydensitometry from independent experi-ments. *, p � 0.05 versus other groups. c,the Src kinase activity was increased inhAFAP/c-Src or cAFAP/c-Src co-trans-fected cells. *, p � 0.05 versus c-Src alone.d, recombinant cAFAP directly activatesc-Src in vitro in a dose-dependent fashion.An in vitro reconstitution Src kinase as-say was conducted with C-terminal Srckinase-treated inactive c-Src and increas-ing doses of recombinant cAFAP. Thephosphorylation of the substrate peptidewas measured by densitometry. *, p �0.05; **, p � 0.01 versus control (nocAFAP).

Intracellular Mechanosensation54796

terial GST fusion proteins containing either or both Src SH3and SH2 domains and captured on glutathione beads wereincubated with A549 cell lysates, and the bound proteins wereresolved by SDS-PAGE and blotted with anti-AFAP antibody.hAFAP binds most efficiently to GST-Src SH2/SH3 beads, moreso to GST-Src SH2 than to Src SH3 beads, and fails to bind toGST-only beads (Fig. 2e).

Overexpression of AFAP Activates c-Src—To determinewhether AFAP can activate c-Src, either hAFAP or cAFAP wasco-expressed with c-Src in COS7 cells. Tyrosine phosphoryla-tion of multiple proteins (Fig. 3a, top section) and autophospho-rylation of c-Src at the Tyr-416 residue (Fig. 3a, second sectionfrom top, and b) were increased in the presence of both AFAPand c-Src. These effects are dependent on c-Src kinase activity,because the co-expression of c-Src-KD (a kinase inactive mu-tant of c-Src) with hAFAP or cAFAP did not have similar effects(Fig. 3a). The phosphorylation of c-Src Tyr527 was not reducedwhen hAFAP or cAFAP was co-expressed with c-Src (Fig. 3a,third section from top). The total Src protein levels in kinase-inactive c-Src-KD-transfected cells were higher (Fig. 3a, fourthsection from top). It has been shown that Src activation leads toits degradation via an ubiquitin-dependent mechanism (32).Thus, in kinase inactive c-Src-KD-transfected cells the lack ofkinase activity may stabilize c-Src. We further measured c-Srckinase activity with a specific Src substrate peptide, and thisactivity was significantly increased when hAFAP or cAFAPwas co-expressed with c-Src (Fig. 3c).

To determine whether AFAP-induced c-Src activation is adirect effect, various amounts of purified, recombinant cAFAPwere incubated with an inactivated c-Src protein, a Src sub-strate peptide, together with [�-32P]ATP. Phosphorylated sub-strate peptide was separated by 20% SDS-PAGE and exposed

to an x-ray film (23). Recombinant cAFAP increased phospho-rylation of the Src substrate in a dose-dependent fashion(Fig. 3d).

Effects of SH2 and SH3 Binding on AFAP-induced c-SrcActivation—Binding to SH3 domains is mediated through pro-line-rich sequences (33). Src-SH3-specific binding uses a se-quence of seven amino acids of the consensus RPLPXXP (34).There are two stretches of proline-rich amino acids in the Nterminus of AFAP. The first stretch (MPLPEIP) is a consensusSrc-SH3 binding motif (Fig. 4a). Mutation of the last proline toalanine at position 71 (AFAP71A) significantly decreased totalprotein tyrosine phosphorylation and c-Src Tyr-416 phospho-rylation (Fig. 4, b and c). In contrast, mutation from proline toalanine at position 77 (AFAP77A) in the second stretch of theproline-rich region had no such effects (Fig. 4, a–c). There arefive putative Src SH2 binding tyrosines in AFAP (21). Weco-expressed c-Src with a mutant (AFAP5Y), of which tyrosineswere replaced with phenylalanines at positions 93, 94, 125,451, and 453 (Fig. 4a). This mutant also decreased significantlyprotein tyrosine phosphorylation and the phosphorylation ofc-Src at Tyr-416 (Fig. 4, b and c). AFAP71A and AFAP5Y hadno significant effect on phosphorylation of the Tyr-527 of c-Src(Fig. 4b). Similarly to our finding, Lerner and Smithgall alsoshowed that human immunodeficiency virus-1 Nef protein-induced Hck (another Src family member) activation is notassociated with a decrease in the phosphorylation of the con-served C-terminal tyrosine (35). It is possible that SH3 and/orSH2 domain ligand-induced activation of Src family proteintyrosine kinases is a direct effect and does not require theactivation of protein tyrosine phosphatases. The binding ofc-Src with AFAP71A and AFAP5Y, but not with AFAP77A, wassignificantly reduced when co-expressed with c-Src in COS7

FIG. 4. Src SH3 or SH2 domain bind-ing plays an essential role in AFAP-induced c-Src activation. a, putativeSrc SH3 and SH2 domain binding sites oncAFAP. Positions of site-directed muta-tions are underlined for proline and bold-faced for tyrosine. b, the AFAP mutants(AFAP5Y and AFAP71A) reduced totalprotein tyrosine phosphorylation (pTyr)and Src Tyr-416 (Src pY416) phosphoryl-ation. Wild type cAFAP (Wt) and its mu-tants (AFAP5Y (5Y), AFAP71A (71A), andAFAP77A (77A)) were co-expressed withc-Src in COS7 cells. The total protein ty-rosine phosphorylation, the phosphoryla-tion status of c-Src at Tyr-416 and Tyr-527, and the expression levels of c-Src andcAFAP were revealed by Western blot-ting. c, densitometry analyses of the ef-fects of cAFAP mutants on Src Tyr-416phosphorylation. *, p � 0.05 versus othergroups. d, mutation at the SH3 bindingsite of AFAP (AFAP71A) abolished itsability to directly activate c-Src in vitro.The phosphorylation of the substrate pep-tide was determined by an in vitro recon-stitution Src kinase assay. *, p � 0.05versus other groups.

Intracellular Mechanosensation 54797

cells (data not shown). Furthermore, the recombinantAFAP71A protein did not significantly activate c-Src in vitro(Fig. 4d). These results suggest that AFAP activates c-Srcthrough these specific Src SH2 and SH3 binding motifs.

AFAP Mediates Mechanical Stretch-induced c-Src Translo-cation to Cytoskeleton—To determine the role of AFAP instretch-induced c-Src activation, the pEGFP eukaryotic expres-sion vector was used to express cAFAP and its mutants fusedwith GFP. The endogenous AFAP is co-localized with actinfilaments in C3H10T1/2 (Fig. 5a, top row), as is GFP-AFAP71A(Fig. 5a, second row from top). In contrast, GFP alone showeddiffuse distribution in the cell (Fig. 5a, third row from top).GFP-AFAP, GFP-AFAP77A, and GFP-AFAP5Y are also co-localized with actin filaments (Fig. 5a, bottom row). Similar re-sults were obtained with other cell lines (data not shown). Theoverexpression of GFP-AFAP activated c-Src in COS7 cells byincreasing its phosphorylation of Tyr-416 as determined by im-munoblotting, whereas overexpressing GFP-AFAP71A and GFP-AFAP5Y did not (data not shown). Therefore, these mutationsand the presence of GFP did not affect the specific intracellulardistribution of AFAP and its ability to activate c-Src.

Mechanical stretch-induced translocation of c-Src from cy-tosol to the Triton-insoluble cytoskeleton fraction in COS7 cellswas inhibited by the overexpression of GFP-AFAP71A but notby that of the GFP alone control (Fig. 5b). Overexpression ofthis mutant may compete with endogenous AFAP for binding tothe actin filaments and therefore inhibit stretch-induced c-Srctranslocation to the cytoskeleton.

Blocking AFAP-Src Binding Reduces Mechanical Stretch-induced c-Src Activation—Mechanical stretch activated c-Srckinase in C3H10T1/2-5H cells with increased fluorescent stain-ing of phosphorylated c-Src Tyr-416 (Fig. 6a). The basal level ofSrc Tyr-416 phosphorylation in static cultured cells was notaffected by the expression of GFP alone or GFP-AFAP71A.However, mechanical stretch-induced Src Tyr-416 phosphoryl-ation was much less in the GFP-AFAP71A-expressing (Fig. 6a,green) cells in comparison with the non-transfected surround-ing cells (Fig. 6a). GFP alone did not affect the stretch-inducedincrease in the Tyr-416 phosphorylation of c-Src (Fig. 6a). Thisinhibitory effect of GFP-AFAP71A was also observed fromother cell lines that we examined (Fig. 6b). To further confirmthe specific inhibitory effect of GFP-AFAP71A, we also used

FIG. 5. GFP-AFAP71A blocks stretch-induced c-Src translocation into cy-toskeletal fraction. a, GFP-fused AFAPand its mutants co-localized with actinfilaments. Endogenous AFAP (green) andtransfected GFP-AFAP71A, but not GFPalone, were co-localized with actin fila-ments (red) in C3H10T1/2 cells (yellow inmerged images). The same distributionpattern was observed in the cells trans-fected with GFP-AFAP, GFP-AFAP5Y, orGFP-AFAP77A (bottom row). b, overex-pression of GFP-AFAP71A blocked stretch-induced c-Src translocation to the Triton-insoluble cytoskeletal fraction. COS7 cellswere transfected with indicated plasmidsand subjected to mechanical stretch. TheTriton-insoluble cytoskeletal fraction wascollected from the cells and subjected toWestern blotting for c-Src and actin. Thec-Src to actin ratio was quantified by den-sitometry and expressed as the percent-age of non-stretched control. *, p � 0.05compared with static control.

Intracellular Mechanosensation54798

GFP-AFAP77A as a negative control that did not blockstretch-induced c-Src activation (Fig. 6c). These effects werefurther confirmed by Western blotting (Fig. 7a). To determinethe effect of Src SH2 domain on stretch-induced c-Src activa-tion, we expressed the GFP-AFAP5Y mutant, which alsoclearly attenuated the stretch-induced phosphorylation of c-Src Tyr-416 (Fig. 6c). These results suggest that althoughoverexpressed GFP-AFAP71A and GFP-AFAP5Y mutantscan perfectly associated with actin filaments, they cannoteffectively bind to the Src SH3 or SH2 domain and subse-

quently activate c-Src during mechanical stretch. Further-more, when we treated A549 cells with siRNA against humanAFAP at 100 nM for 60 h, it clearly reduced the endogenouslevel of AFAP as determined by Western blotting and inhib-ited mechanical stretch-induced c-Src activation (Fig. 7b). Incontrast, a nonspecific control double-stranded RNA had nosuch effects. Taken together, these results indicate that theinteraction between AFAP and c-Src could covert mechanicalforce into biochemical reaction of c-Src activation for furtherintracellular signaling.

FIG. 6. Expression of GFP-AFAP71A or GFP-AFAP5Y inhibits mechanical stretch-induced c-Src activation. a, overexpression ofGFP-AFAP71A blocked stretch-induced c-Src Tyr-416 phosphorylation (comparing GFP positive and surrounding cells). C3H10T1/2-5H cells weretransfected with GFP vector or GFP-AFAP71A and then subjected to mechanical stretch or static culture (Control). c-Src activation was determinedby immunofluorescent (red) staining with an antibody against the phosphorylated Tyr-416 of c-Src. Mechanical stretch-induced increase in c-SrcTyr-416 phosphorylation was blocked in the cells overexpressing GFP-AFAP71A but not GFP alone. b, similar blocking effects of GFP-AFAP71Aon the stretch-induced increase of c-Src Tyr-416 phosphorylation were seen in all tested cell lines (merged images shown). c, GFP-AFAP77A (anegative control of GFP-AFAP71A for Src SH3 domain binding) did not block stretch-induced c-Src activation, whereas GFP-AFAP5Y, a mutantfor Src SH2 domain binding, also blocked stretch-induced c-Src activation in C3H10T1/2 cells. Similar data were obtained from other cell lines (notshown).

Intracellular Mechanosensation 54799

DISCUSSION

Sensing mechanical stimuli is very important for perceivingthese valuable signals from external and internal environ-ments. For specialized mechanoreceptors in the body, ion chan-nels and their anchors are the common features of these mech-anosensors (13). For “non-specialized” cells, searching formechanosensors has also been focused on cellular membraneproteins such as stretch-activated ion channels (16, 17), inte-grins (2, 8), and growth factor receptors (19). It has been re-cently reported that a Z-disc complex in cardiomyocytes couldsense mechanical stretch in the heart (36). This suggests thatmechanical force can be sensed intracellularly. Based on re-sults from the present study, we propose that cytoskeleton-associated proteins such as AFAP could transmit and convertmechanical stimuli into biochemical reactions of signalingthrough specific binding sites for protein-protein interactions.This may represent a “generalized” intracellular mechanosen-sory mechanism.

There are several features for AFAP-mediated mechanore-ception. First, this protein has an actin-binding site (Fig. 2a).Its distribution pattern makes it a perfect candidate to trans-mit physical forces. Second, it has high affinity binding abilitiesto Src SH2 and SH3 domains, which allows it to bind compet-itively to c-Src and other Src family members and activatethese protein tyrosine kinases. pp125FAK (37) and pp130CAS(38) are also potential Src activators with Src SH2 and SH3binding motifs. Their distribution is also associated with cy-toskeleton. In response to mechanical stimulation, they mayactivate c-Src through mechanisms similar to those of AFAP.

We have demonstrated that the Src SH3 domain-binding sitein AFAP is critical for stretch-induced c-Src activation. A singleamino acid mutation (GFP-AFAP71A) blocked mechanical

stretch-induced c-Src activation in multiple cell lines. In contrast,mutation in another proline-rich site (GFP-AFAP77A) did nothave such inhibitory effect. This finding suggests that the spec-ificity of the Src SH3 domain binding is crucial for c-Src activa-tion. By mutating all putative tyrosines that may bind to c-Src,we demonstrated that Src SH2 domain binding is also involved inmechanical stretch-induced c-Src activation. Mechanical stretchinduced-cytoskeleton deformation may first increase the contactbetween AFAP and c-Src through the SH3 domain and activatec-Src. Activated c-Src can phosphorylate tyrosines on the SH2domain binding sites of AFAP, which may further increase thebinding of AFAP to more c-Src molecules and/or enhance thebinding affinity of AFAP to c-Src (Fig. 1d). Activated c-Src canalso phosphorylate other Src substrates at their tyrosine residuesand subsequently affect cellular functions.

The importance of protein-protein interactions in intracellu-lar signal transduction has been well accepted (39). The simi-larity of functional domains and motifs in proteins suggest thatthey might perform their interactions and functions as modules(39). However, this concept has not been addressed in mech-anotransduction. It has been speculated that physical distor-tion of the tissue or ECM would change force distributionsacross adhesion receptors and result in both local and globalrestructuring of the tensionally integrated cytoskeleton net-work (2). Indeed, mechanical stretch (40) or shear stress (41)can change the orientation of cells, but these processes areusually slow. Restructuring the cytoskeleton is a complicatedbiochemical process. It should be considered a result of mech-anotransduction, not a mechanism of mechanosensation. Hu etal. demonstrated that mechanical stretch activated the plate-let-derived growth factor receptor (42). They speculated thatphysical forces such as energy are absorbed by the receptor, by

FIG. 7. AFAP plays a critical role in mechanical stretch-induced c-Src activation. a, overexpression of GFP-AFAP71A blockedstretch-induced c-Src Tyr-416 phosphorylation (Src pY416). COS7 cells were transfected with GFP vector, GFP-AFAP71A, or GFP-AFAP77A for60 h and then subjected to mechanical stretch for 10 min or static culture (control). b, knockdown of endogenous AFAP expression with siRNAreduced stretch-induced c-Src activation. A549 cells were transfected with either siRNA against the human AFAP sequence or nonspecific controldouble-stranded RNA (100 nM for 60 h), and then subjected to mechanical stretch for 10 min. Equal amounts of proteins from the collected celllysates were subjected to Western blotting for phosphorylated Tyr-416 of c-Src, total Src, and AFAP levels. The blots from 3–4 separateexperiments were quantified by densitometry and expressed as a ratio of Try-416 phosphorylated Src to total Src. *, p � 0.05 compared with staticcontrol.

Intracellular Mechanosensation54800

non-protein or protein components of the plasma membrane, orby molecules at some other site of the cell followed by transferof the signal to the plasma membrane (42). How the absorbedphysical energy is converted into biochemical reaction is stillnot explained. Based on results from the present study, wesuggest that mechanosensory process through either integrinor the platelet-derived growth factor receptor might be alsomediated through protein-protein interactions via specificbinding mechanisms.

In summary, using AFAP and c-Src as an example, we dem-onstrated that mechanical forces could be converted into bio-chemical reactions related to intracellular signal transductionthrough specific interactions between adjacent and function-ally related proteins.

Acknowledgments—We are grateful to Daniel C. Flynn, Yong Qian,and Joseph M. Baisden for preparing recombinant cAFAP proteins andDNA constructs. We thank Xinyan Huang for providing the C-terminalSrc kinase (Csk)-treated c-Src and Eddy Xuan for making the illustra-tion of stretch-induced AFAP/Src interaction.

REFERENCES

1. Liu, M., and Post, M. (2000) J. Appl. Physiol. 89, 2078–20842. Ingber, D. E. (2002) Circ. Res. 91, 877–8873. Skalak, T. C., and Price, R. J. (1996) Microcirculation 3, 143–1654. Turner, C. H., Owan, I., and Takano, V. (1995) Am. J. Physiol. 269,

E438–E4425. Grodzinsky, A. J., Levenston, M. E., Jin, M., and Frank, E. H. (2000) Annu.

Rev. Biomed. Eng. 2, 691–7136. Pinhu, L., Whitehead, T., Evans, T., and Griffiths, M. (2003) Lancet 361,

332–3407. Epstein, N. D., and Davis, J. S. (2003) Cell 112, 147–1508. Shyy, J. Y., and Chien, S. (2002) Circ. Res. 91, 769–7759. Chalfie, M. (1997) Biol. Bull. 192, 125

10. Tavernarakis, N., and Driscoll, M. (1997) Annu. Rev. Physiol. 59, 659–68911. Kernan, M., Cowan, D., and Zuker, C. (1994) Neuron 12, 1195–120612. Walker, R. G., Willingham, A. T., and Zuker, C. S. (2000) Science 287,

2229–223413. Gillespie, P. G., and Walker, R. G. (2001) Nature 413, 194–20214. Price, M. P., Lewin, G. R., McIlwrath, S. L., Cheng, C., Xie, J., Heppenstall,

P. A., Stucky, C. L., Mannsfeldt, A. G., Brennan, T. J., Drummond, H. A.,Qiao, J., Benson, C. J., Tarr, D. E., Hrstka, R. F., Yang, B., Williamson,R. A., and Welsh, M. J. (2000) Nature 407, 1007–1011

15. Nauli, S. M., Alenghat, F. J., Luo, Y., Williams, E., Vassilev, P., Li, X., Elia,A. E., Lu, W., Brown, E. M., Quinn, S. J., Ingber, D. E., and Zhou, J. (2003)Nat. Genet. 33, 129–137

16. Yang, X.-C., and Sachs, F. (1989) Science 243, 1068–107117. Hamill, O. P., and Martinac, B. (2001) Physiol. Rev. 81, 685–74018. Sawada, Y., and Sheetz, M. P. (2002) J. Cell Biol. 156, 609–61519. Tschumperlin, D. J., Dai, G., Maly, I. V., Kikuchi, T., Laiho, L. H., McVittie,

A. K., Haley, K. J., Lilly, C. M., So, P. T., Lauffenburger, D. A., Kamm, R. D.,and Drazen, J. M. (2004) Nature 429, 83–86

20. Liu, M., Qin, Y., Liu, J., Tanswell, A. K., and Post, M. (1996) J. Biol. Chem.271, 7066–7071

21. Guappone, A. C., Weimer, T., and Flynn, D. C. (1998) Mol. Carcinog. 22,110–119

22. Qian, Y., Baisden, J. M., Zot, H. G., Van Winkle, W. B., and Flynn, D. C. (2000)Exp. Cell Res. 255, 102–113

23. Ma, Y. C., Huang, J., Ali, S., Lowry, W., and Huang, X. Y. (2000) Cell 102,635–646

24. Cheng, H. C., Nishio, H., Hatase, O., Ralph, S., and Wang, J. H. (1992) J. Biol.Chem. 267, 9248–9256

25. Xu, W., Doshi, A., Lei, M., Eck, M. J., and Harrison, S. C. (1999) Mol. Cell 3,629–638

26. von Wichert, G., Jiang, G., Kostic, A., De Vos, K., Sap, J., and Sheetz, M. P.(2003) J. Cell Biol. 161, 143–153

27. Flynn, D. C., Leu, T. H., Reynolds, A. B., and Parsons, J. T. (1993) Mol. Cell.Biol. 13, 7892–7900

28. Xu, W., Harrison, S. C., and Eck, M. J. (1997) Nature 385, 595–60229. Williams, J. C., Weijland, A., Gonfloni, S., Thompson, A., Courtneidge, S. A.,

Superti-Furga, G., and Wierenga, R. K. (1997) J. Mol. Biol. 274, 757–77530. Sicheri, F., and Kuriyan, J. (1997) Curr. Opin. Struct. Biol. 7, 777–78531. Qian, Y., Baisden, J. M., Westin, E. H., Guappone, A. C., Koay, T. C., and

Flynn, D. C. (1998) Oncogene 16, 2185–219532. Hakak, Y., and Martin, G. S. (1999) Curr. Biol. 9, 1039–104233. Pawson, T. (1994) Adv. Cancer Res. 64, 87–11034. Alexandropoulos, K., Cheng, G., and Baltimore, D. (1995) Proc. Natl. Acad. Sci.

U. S. A. 92, 3110–311435. Lerner, E. C., and Smithgall, T. E. (2002) Nat. Struct. Biol. 9, 365–36936. Knoll, R., Hoshijima, M., Hoffman, H. M., Person, V., Lorenzen-Schmidt, I.,

Bang, M. L., Hayashi, T., Shiga, N., Yasukawa, H., Schaper, W., McKenna,W., Yokoyama, M., Schork, N. J., Omens, J. H., McCulloch, A. D., Kimura,A., Gregorio, C. C., Poller, W., Schaper, J., Schultheiss, H. P., and Chien,K. R. (2002) Cell 111, 943–955

37. Thomas, J. W., Ellis, B., Boerner, R. J., Knight, W. B., White, G. C., andSchaller, M. D. (1998) J. Biol. Chem. 273, 577–583

38. Burnham, M. R., Bruce-Staskal, P. J., Harte, M. T., Weidow, C. L., Ma, A.,Weed, S. A., and Bouton, A. H. (2000) Mol. Cell. Biol. 20, 5865–5878

39. Pawson, T., and Nash, P. (2003) Science 300, 445–45240. Sai, X., Naruse, K., and Sokabe, M. (1999) J. Cell Sci. 112, 1365–137341. Langille, B. L. (2001) Microcirculation 8, 195–20642. Hu, Y., Bock, G., Wick, G., and Xu, Q. (1998) FASEB. J. 12, 1135–1142

Intracellular Mechanosensation 54801