Talin determines the nanoscale architecture offocal adhesionsJaron Liua,1,2, Yilin Wanga,2, Wah Ing Goha,2, Honzhen Goha, Michelle A. Bairdb,3, Svenja Ruehlanda,4, Shijia Teoa,Neil Batec, David R. Critchleyc, Michael W. Davidsonb,d, and Pakorn Kanchanawonga,e,5

aMechanobiology Institute, Singapore, National University of Singapore, Republic of Singapore 117411; bNational High Magnetic Field Laboratory, TheFlorida State University, Tallahassee, FL 32306; cDepartment of Biochemistry, University of Leicester, Leicester LE1 7RH, United Kingdom; dDepartment ofBiological Sciences, The Florida State University, Tallahassee, FL 32306; and eDepartment of Biomedical Engineering, National University of Singapore,Republic of Singapore 117411

Edited by Vann Bennett, Duke University Medical Center, Durham, NC, and approved July 23, 2015 (received for review July 3, 2015)

Insight into howmolecular machines perform their biological functionsdepends on knowledge of the spatial organization of the components,their connectivity, geometry, and organizational hierarchy. However,these parameters are difficult to determine in multicomponent assem-blies such as integrin-based focal adhesions (FAs). We have previouslyapplied 3D superresolution fluorescence microscopy to probe thespatial organization of major FA components, observing a nanoscalestratification of proteins between integrins and the actin cytoskel-eton. Here we combine superresolution imaging techniques with aprotein engineering approach to investigate how such nanoscalearchitecture arises. We demonstrate that talin plays a key structuralrole in regulating the nanoscale architecture of FAs, akin to amolecular ruler. Talin diagonally spans the FA core, with its N terminusat the membrane and C terminus demarcating the FA/stress fiberinterface. In contrast, vinculin is found to be dispensable for specifi-cation of FA nanoscale architecture. Recombinant analogs of talin withmodified lengths recapitulated its polarized orientation but altered theFA/stress fiber interface in a linear manner, consistent with its modularstructure, and implicating the integrin–talin–actin complex as the pri-mary mechanical linkage in FAs. Talin was found to be ∼97 nm inlength and oriented at ∼15° relative to the plasma membrane. Ourresults identify talin as the primary determinant of FA nanoscale or-ganization and suggest howmultiple cellular forces may be integratedat adhesion sites.

superresolution microscopy | focal adhesions | talin | mechanobiology |nanoscale architecture

Cell adhesion to the ECM is a highly coordinated process thatinvolves ECM-specific recognition by integrin transmembrane

receptors, and their aggregation with numerous cytoplasmic pro-teins into dense supramolecular complexes called focal adhesions(FAs) (1). Actin stress fibers terminate at FAs where actomyosincontractility is transmitted to the ECM, generating traction (2–5).Mechanical tension impinging on each FA is implicated in key stepsincluding the elongation, reinforcement, and maintenance of theFA structures (6). FA mechanotransduction is a major aspectof cellular microenvironment sensing with wide-ranging conse-quences in physiological and pathological processes (7–10). How-ever, molecular-scale spatial parameters that specify FA nanoscaleorganization have been difficult to access experimentally. Never-theless, these are essential to understand how mechanosensitivityarises within such complex molecular machines (11–15).Previously 3D superresolution fluorescence microscopy has

unveiled the nanoscale organization of major FA components,whereby a core region of ∼30 nm interposes between the integrinand the actin cytoskeleton along the vertical (z) axis (16). TheFA core consists of a membrane-proximal layer that containssignaling proteins such as FAK (focal adhesion kinase) and pax-illin, an intermediate zone that contains force-transduction pro-teins such as talin and vinculin, and a stress fiber interfacial zonethat contains actin-associated proteins such as VASP (vasodilator-stimulated protein) and α-actinin. Although such multilaminar

architecture signifies a certain degree of compartmentalizationwithin FAs that may serve to spatially constrain protein–proteininteractions and dynamics, the structural connectivity, the molec-ular configuration and geometry of FA proteins, and the molec-ular basis of their higher-order organization remain unclear.Proteomic and interactome analysis of the integrin adhesome

have uncovered several direct and multitier connections betweenintegrins and actin (17–20). This suggests that multiple highlyinterconnected protein–protein interactions could collectively self-organize into FA structures; such redundancy could also accountfor the remarkable mechanical robustness of FAs after cellulardisruption or perturbation (21). Alternatively, a specific FAcomponent may play a dominant role in regulating FA architec-ture. Aspects of both scenarios may also act cooperatively orfunction at distinct stages of FA assembly and maturation.Superresolution microscopy of cells expressing fluorescent protein(FP)-tagged FA components has revealed that talin, a large cy-toskeletal adaptor protein, adopts a highly polarized orientation inFAs (16), with the N terminus residing in the membrane-proximallayer and the C terminus elevated by z ∼30 nm to the FA/stressfiber interfacial zone. This led us to hypothesize that an array of

Significance

Focal adhesions (FAs) mediate cell–extracellular matrix interac-tions and consist of integrin receptors linked to the actin cyto-skeleton via multiprotein complexes organized into nanoscalestrata. In this work, we sought to determine the molecular basisof FA nanostructure. Combining superresolution microscopy andprotein engineering, we demonstrate the structural role of talinin regulating the nanoscale architecture of FAs. Talin specifiesthe dimension of the FA core, akin to a molecular ruler, in aremarkably modular manner. Our results define the moleculargeometry of the integrin–talin–actin module that comprisesthe key mechanical linkage within FAs and elucidate howsuch interactions serve to integrate multiple cellular forces atadhesion sites.

Author contributions: J.L., W.I.G., and P.K. designed research; J.L., Y.W.,W.I.G., H.G., S.T., andP.K. performed research; H.G., M.A.B., S.R., N.B., D.R.C., and M.W.D. contributed new re-agents/analytic tools; J.L., Y.W., W.I.G., and P.K. analyzed data; and J.L., Y.W., W.I.G., D.R.C.,and P.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: Microscopy Unit, A*-STAR Institute of Medical Biology, Republic ofSingapore 138648.

2J.L., Y.W., and W.I.G. contributed equally to this work.3Present address: Laboratory of Cell and Tissue Morphodynamics, National Heart Lungand Blood Institute, National Institutes of Health, Bethesda, MD 20892.

4Present address: Faculty of Biology, Ludwig Maximilian University of Munich, 80539Munich, Germany.

5To whom correspondence should be addressed. Email: biekp@nus.edu.sg.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1512025112/-/DCSupplemental.

E4864–E4873 | PNAS | Published online August 17, 2015 www.pnas.org/cgi/doi/10.1073/pnas.1512025112

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

integrin–talin–actin linkages may vertically span the FA core,serving a structural role in determining FA architecture (16).To test this hypothesis, we sought to perturb FAs by substituting

endogenous talin with recombinant analogs having modified lengths.These were generated by retaining both the N-terminal FERM(band 4.1/Ezrin/Radixin/Moesin) and C-terminal THATCH (Talin/HIP1R/Sla2p Actin-tethering C-terminal Homology, or R13) do-mains but with selective deletion of the multiple helical bundleswithin the central region of talin. By using a siRNA-mediatedknockdown/rescue approach, we found that such talin analogs wereable to support FA formation, clustering of activated integrins, andlinkages to the actin cytoskeleton. By mapping the z-position of theFPs tagged at either the N or the C termini, we show that talin andits analogs are linearly extended and oriented in FAs, with theirlengths regulating FA nanoscale organization. Chimeric-talin ana-logs with a 30-nm spacer insertion are also able to support FAassembly, facilitating the precise determination of talin geometry inFAs. Our results indicate that talin is oriented at 15° relative to theplasma membrane, measuring ∼97 nm end to end. FA nanoscalearchitecture in vinculin-null mouse embryonic fibroblasts (MEFs)retained its stratified organization and talin polarization similar tothat in other cell types, suggesting that vinculin is dispensable forthe specification of FA architecture. Our measurements demon-strate how the integrin–talin–actin module serves as the primary,and surprisingly modular, structural and tension-bearing core ofFAs and geometrically define how such complexes could integratemultiple cellular forces at adhesion sites.

ResultsNanoscale Protein Stratification in FAs. Two highly homologoustalin genes, talin1 (tln1) and talin2 (tln2), have been identified;loss of talin1 is embryonic-lethal, whereas talin2-null mice are vi-able and fertile (22, 23). Nevertheless, both isoforms are expressedin most cell types (24) (Fig. S1 A–C), and talin2 is able to supportFA assembly and cell spreading in talin1-null mouse fibroblasts(25). To study the structural role of talin in FAs, we thereforechose primary human umbilical vein endothelial cells (HUVECs),which express only talin1 (26) (Fig. S1 A–C). siRNA-mediateddepletion of endogenous human talin1 results in impaired cellspreading and migration (Fig. S1 D–J), as well as FA assemblydefects, which can be rescued by coexpression of a mouse talin1GFP fusion as described previously (25, 26). The HUVEC platformthus allows substitution of endogenous talin1 with engineeredtalin constructs.We first sought to characterize FA architecture in HUVECs.

Previously we reported that the primary organization axis forFAs is along the vertical (z) dimension (16). Here we used twotechniques for the nanoscale measurement of FA proteinz-positions. Interferometric photoactivated localization microscopy(iPALM) combines single-molecule localization for xy super-resolution and multiphase interferometry for z superresolution,providing sub-20-nm resolution in 3D (27) (Fig. S2). We also useda surface-generated structured illumination technique—variableincidence angle fluorescence interference contrast microscopy(VIA-FLIC) (28) or scanning angle interference microscopy(SAIM) (29)—which map the ensemble z-position of fluorophoreswith high speed and a larger field of view, albeit with a diffraction-limited resolution in xy (Fig. S3). Although VIA-FLIC/SAIM arenot true superresolving techniques per se, in that vertically over-lapping objects are not resolved (28, 29), FA proteins were shownto organize in distinct strata (16). These are thin relative to theaxial structured illumination pattern, and thus the z-position de-termined by VIA-FLIC/SAIM should closely correspond to theensemble-averaged z-position of a given FA protein.We cultured HUVEC cells expressing FP-tagged FA proteins

on a fibronectin-coated thermal SiO2 surface of silicon wafers forVIA-FLIC/SAIM or on fibronectin-coated fiducialed coverglassesfor iPALM. In VIA-FLIC/SAIM analysis, the topographic map of

z-positions are extracted pixel by pixel for FA regions. These zpixelvalues are offset by the average thickness of the fibronectin layer,which was determined to be ∼10 nm (Fig. S3). We used the medianz-position from each region of interest (ROI) to represent the FAprotein z-position (zFA). We verified that total zFA distributions arevery similar to that of the total zpixel distribution, and thus should berepresentative of the FA protein z-positions. We observed a hier-archical organization of FA proteins along the z axis (Fig. 1 A andB), similar to earlier measurements in other cell types (16). Integrinαv cytoplasmic tail, FAK, and paxillin are located in close proximityto one another, forming a membrane proximal layer at ∼45 nmabove the ECM, whereas actin regulatory or actin-associatedFA proteins such as VASP, α-actinin, and zyxin are observed athigher z-positions, >70 nm. We next used iPALM to image thefull-3D extent of actin organization (Fig. 1C) at the FA/stress fiberjunctions (Fig. 1C). The vertical profile of actin (histogram, Fig. 1C,Lower) peaks at z ∼80–120 nm, below which the actin density rapidlydeclines, signifying the interface between the FA core and actin stressfibers. At higher z values, actin seems to taper rapidly into denselybundled stress fibers. The onset of F-actin compaction at z ∼120 nmcoincides with the z-position of α-actinin, (Fig. 1B), consistent with itsrole in cross-linking F-actin (30).

Fig. 1. Nanoscale architecture of FAs in HUVECs. (A) Topographic maps ofFA protein z-positions (nanometers) in HUVECs. (B) Nanoscale stratificationof FA protein z-positions. Notched boxes and histograms (bin size, 1 nm) forzFA of indicated FA proteins: first and third quartiles, median and confidenceintervals; whiskers, 5th and 95th percentiles. Numbers indicated: median zFA(red), number of FA ROIs (black), and number of cells (blue). (C) iPALM 3Dsuperresolution images of F-actin, (D) talin (N-terminal probe), and (E) talin(C-terminal probe). F-actin is highly enriched at z >100 nm but is sparsewithin FA core region, whereas talin adopts a polarized N–C orientation asdescribed earlier (16). C–E, Upper show top view (xy). C–E, Lower show side-view projection of boxed regions in upper panels, cell edges on left. Colorbar indicates z-position relative to ECM: 30–100 nm (A); 0–150 nm (C–E).(Scale bars: 5 μm in A, 2 μm in C–E, Upper, and 250 nm in C–E, Lower.)

Liu et al. PNAS | Published online August 17, 2015 | E4865

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

The nanoscale precision of our imaging techniques enables us toselectively map the positions of different ends of a protein, allowinginference of molecular orientation (16). Measurements of talintagged with FP at the N terminus (Talin-N) or the C terminus(Talin-C) reveals median z-positions of 41.4 nm and 66.4 nmrespectively, indicative of a polarized orientation. This was cross-verified by iPALM imaging using talin tagged with a photo-activatable FP, tdEos (31) (Fig. 1 D and E). The narrow z-positiondistributions of the N and C termini and their minimal mutualoverlap confirm that most talin molecules adopt a highly polarizedN–C orientation.Because measurements of vinculin tagged at either termini also

reveal a directional organization (Fig. 1B), we further investigatedthe role of vinculin in FA architecture by mapping the positions ofFA proteins in vinculin-null MEFs (32). As shown in Fig. 2, weobserved a similar hierarchical organization of FA proteins alongthe z-dimension, with the N–C polarized orientation of talin thatseems to span the FA core, linking with actin stress fibers. Theseresults suggest that vinculin is dispensable for the stratifiednanoscale architecture of FAs, consistent with previous studiesdocumenting FA assembly and maturation in the absence of vin-culin (33, 34).

Probing the Structural Role of Talin by Recombinant MinitalinAnalogs. We next examined the hypothesis that talin may physi-cally regulate FA architecture. The observed N–C polarization oftalin (Fig. 1 D and E) suggests that its N terminus is positioned viabinding of the talin FERM domain to integrin β cytodomains,whereas the C terminus is upshifted in z-position via the THATCHdomain interacting with F-actin. In this view, integrin–talin–actinmodules are aligned into a parallel, oriented array that comprisesthe tension-bearing, structural core of FAs, around which other

FA proteins are organized. To test this and to further definethe physical and molecular basis of such organization, we askedwhether full-length endogenous talin could be substituted withrecombinant analogs with altered lengths, and whether such an-alogs modulate the nanostructure of FAs.Earlier superresolution imaging results suggested that both the

talin FERM and THATCH domains are capable of targetingto their respective nanoscale compartments in FAs (16). Wetherefore created a series of recombinant mouse talin1 analogsthat contain the FERM (residues 1–482) and THATCH (resi-dues 2294–2541) domains but with selective deletion of thecentral talin rod domains. The talin rod region consists of 13tandem α-helical bundles, forming a flexible chain of globulardomains (Fig. S4A). Recent structural studies suggest that indi-vidual domains of the talin rod are highly modular (35), per-mitting us to generate “minitalins” analogs that vary in sequencefrom 29 to 58% of full-length talin. We refer to these by theirnominal sequence lengths, e.g., T29 for the analog that contains29% of the sequence of full-length talin (T100), and so on (Fig. 3A and B). To allow nanoscale mapping of their positions andorientations, the constructs were tagged with GFP (or spectralvariant) at either the N or the C terminus.We next tested whether the minitalins were able to support FA

formation in place of endogenous human talin. HUVECs werecotransfected with a human talin1 siRNA and cDNAs encodingmouse minitalin FP fusions, cultured on fibronectin-coated sub-strates and imaged ∼96 h posttransfection to ensure optimalknockdown of endogenous talin1 (Fig. S1F). A monoclonal anti-body that recognizes hTalin1 but not mTalin1 was used to assessknockdown efficiency by immunofluorescence microscopy (Fig.3C). We observed that the minitalins rescued FA formation andcolocalized with other FA proteins such as paxillin and activated

Fig. 2. Nanoscale architecture of FAs in vinculin-null MEF. (A) Topographic map of FA protein z-positions (nanometers) in vinculin-null MEF. (B) Nanoscalestratification of FA protein z-positions. Notched box plots and histograms (bin size, 1 nm) for zFA of indicated FA proteins: first and third quartiles, median andconfidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated: median zFA (red), number of FA ROIs (black), and number of cells (blue). (C) iPALM 3Dsuperresolution images of F-actin, indicating high density at z >80 nm, and low density within FA core region. (Upper) Top view (xy). (Lower) Side view projectionof boxed region, cell edge on left. Color bar indicates z-position relative to the ECM: (A) 30–100 nm; (C) 0–150 nm. (Scale bars: 5 μm in A, 2 μm in C, Upper, and250 nm in C, Lower.)

E4866 | www.pnas.org/cgi/doi/10.1073/pnas.1512025112 Liu et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

integrins (Fig. 3C and Fig. S5A), forming linkages to F-actinbundles (Fig. 3A). GFP fluorescence was observed in FA-likeclusters for all minitalins (Fig. 3C). With the larger minitalins(T41–T58), cell spreading was well-supported relative to full-lengthtalin (T100), whereas a substantial reduction in cell area was ob-served with the smaller T29 and T36 constructs (Fig. S5C). Nev-ertheless, the latter still supported significant spreading comparedwith hTalin1 siRNA cells. Overall, these results indicate that theminitalins are able to support FA assembly, enabling furtheranalysis of the nanoscale architecture of these modified FAs.

Nanoscale Organization of Minitalins in FAs. We next mapped thenanoscale z-positions of the N and C termini of the minitalins tocharacterize their organization in FAs (Fig. 4A). The N termini ofminitalins were observed at z ∼40 nm, similar to that of the full-length talin (T100), indicating that their FERM domains remainedin the membrane proximal layer (Fig. 4 B and D). In contrast, theirC-termini z-positions were observed to be significantly downshiftedrelative to T100 (z = 66.4 nm), ranging from z = 54.8 nm (T58) toz = 43.1 nm (T29) (Fig. 4 C and D and Fig. S6). We found that thesequence lengths of these talin analogs correlate linearly with theC-terminal z-positions (R = 0.95). Slight deviations were observedfor some constructs such as T46. This likely reflects topologicalvariations among the rod domains; the N and C termini of four-helical bundles are in the cis configuration instead of trans in thecase of five-helical bundles that comprise the majority of roddomains (Fig. S4 B and C). Thus, the end-to-end length of T46(FERM–R1–R2–R3–THATCH) may be somewhat shorter thanexpected from its sequence length because the R2 and R3 do-mains are four-helical bundles.We next sought to check whether the differences in the N- and

C-termini z-positions observed above represent the intact talinmolecule or that of N- and C-terminal talin fragments generated bycalpain-mediated proteolysis in the linker between the FERM and

rod domains (36). We therefore introduced the calpain-resistantL432G mutation in the linker (36) and mapped the C-terminalz-positions of talin and minitalin constructs. The C terminus ofT100/L432G was again found at an elevated z-position (z = 62.6 nm),confirming that this accurately reflects the polarized orien-tation of intact talin. Likewise, the C-termini z-positions of theL432G minitalins again exhibited a downward shift relative to thatof T100/L432G (Fig. 4 E–G). These scale linearly with the se-quence lengths, with a slope similar to that of the original minitalinconstructs (P = 0.404; SI Materials and Methods), but with a smallbut statistically significant downshift in z (Fig. 4G). The slightlylower C-terminal z-positions due to the L432G mutation mayreflect the reduction in calpain-mediated cellular contractilitydocumented previously (37). Consistent with this, the C-terminalz-position of wild-type T100 was similarly lowered upon pharma-cological inhibition of calpain (z = 62.2 nm, Fig. S7 A and B).Altogether, these measurements establish that talin and minitalinsare highly polarized in FAs, with the N terminus in the membraneproximal layer, and the C terminus at elevated z-positions. Thesealso demonstrate the structural continuity of talin, which is re-quired for mechanical force transmission.

Probing Talin Geometry by Minitalin:FilaminA Chimeric Analogs. Thelinear scaling between the C-termini z-positions and constructlengths implicates a linearly extended configuration (SI Discus-sion) and suggests that these talin analogs may be oriented with asimilar contact angle (θTalin) relative to the plasma membrane.Our measurements thus provide an opportunity to determine thecontact angle of how talin connects integrin to actin, which is acritical parameter for cell-ECM force transmission. However,because the rod domains could unfold under physiological force,the end-to-end lengths of talin and their analogs can be variable(38–40). To address this, we exploited the modularity of the talinrod to engineer in a “spacer” protein domain with a well-defined

Fig. 3. Recombinant minitalins support FA assembly. (A) Schematic diagram of minitalin constructs with selective deletions in the rod domain (see also Fig. S4).(B) Immunoblot analysis of mEmerald-tagged mouse talin and minitalin constructs (probed with anti-GFP antibody). (C) Epifluorescence micrographs of HUVECscotransfected with hTalin1 siRNA (KD) or control siRNA (siCON), and mEmerald-tagged full-length talin (T100), minitalins (T29-T58), or GFP (control). GFP channel(green) indicates talin or minitalins localization to FAs. Knockdown of endogenous talin is verified by hTalin1-specific antibody (huTLN, cyan). FAs and cytoskeletalactin organization are visualized by paxillin immunofluorescence (magenta) and phalloidin (blue), respectively. (D) Magnified view, GFP channel of C. (Scale bars:25 μm in C and 12.5 μm in D.)

Liu et al. PNAS | Published online August 17, 2015 | E4867

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

length. We chose the filaminA IgFLNa1–8 domains (residues276–1061), because this Ig-like β-barrel chain has been shown to bemonomeric, contains few binding sites for FA or actin-associatedproteins, and to behave as a ∼30-nm linear elastic chain with avery high unfolding threshold of >60 pN (41, 42). This segmentwas inserted in lieu of the deletions present in the minitalins (Fig.5A and Fig. S4D), and the resulting chimeric-talin constructs werenamed XT58, XT48, XT46, XT41, XT36, and XT29, in relation totheir minitalin counterparts. As shown in Fig. 5 A–D, the chimeric-talin analogs expressed properly, localized to FAs, and supportedFA formation in HUVECs in lieu of endogenous talin.

Geometric Analysis of Integrin–Talin–Actin Contact Angle and TalinExtension. We next mapped the nanoscale organization of thechimeric-talins in FAs (Fig. 5 E and F and Fig. S7). As predicted,their C termini were observed at elevated z-positions relative to thecorresponding minitalins. Furthermore, these also scale linearlywith construct lengths, with a slope virtually identical to that of theminitalins (P = 0.888; SI Materials and Methods). This suggests thatthe chimeric-talins are oriented in a polarized fashion, and with a

similar contact angle (θTalin) relative to the plasma membrane.Because the IgFLNa should behave as a freely jointed-chainsegment, the lengths of the chimeric-talins are expected to be ∼30nm longer than their minitalin counterparts (SI Discussion). Thus,the vertical offset of z = 7.4 nm between the chimeric-talin andminitalin regression lines can be used to calculate the talin contactangle, yielding θTalin ∼15°. Sensitivity analysis suggests that thisestimate is robust to within a few degrees given experimentaluncertainties (SI Discussion).Subsequently, the length of talin can be calculated, yielding

∼97 nm for full-length talin (T100) based on the N- andC-termini z-displacement of ∼25 nm (Fig. 4). Interestingly, thelength of talin/L432G corresponds to ∼81 nm (21 nm/sin 15°),in good agreement with the recent estimate of the end-to-endlength of talin in which all domains are folded (35). We surmisethat the increased length of ∼16 nm observed in cells expressingwild-type talin could be due to their greater contractility rela-tive to talin/L432G cells, which may lead to a partial unfold-ing of rod domains. These unfolded domains are likely theR2 or R3 four-helix bundles because they possess the lowest

Fig. 4. Polarized orientation of recombinant minitalins. (A) Topographic map of talin analog z-positions (nanometers). N-terminal FP fusions (Upper) andC-terminal FP fusions (Lower). (B and C) Polarized orientation of minitalins in FAs. Notched boxes and histograms (bin size, 1 nm): first and third quartiles, medianand confidence intervals; whiskers, 5th and 95th percentiles, for N-terminal probes (B), and C-terminal probes (C). (D) Comparison of N- and C-terminal zFApositions. Numbers of FA ROIs in parentheses; median zFA of the distribution indicated in red (B and C). (E) Polarized orientation of talin and minitalins is in-dependent of calpain cleavage. Topographic map of C-terminal FP fusions containing L432Gmutation (denoted by asterisk) that suppress calpain II cleavage in thelinker between the talin FERM and rod domains. (F) Notched boxes and histograms (bin size, 1 nm): first and third quartiles, median and confidence intervals;whiskers, 5th and 95th percentiles. Numbers indicated in B, C, and E: median zFA (red), number of FA ROIs (black), and number of cells (blue). (G) Comparison ofthe C-terminal z-positions of talin analogs with or without the L432G mutation. Statistical significance inD and G: ****P << 10−6; *P < 0.05, Mann–Whitney u test.Color bars indicate z-position, 40–80 nm (A and E). (Scale bars: 5 μm.)

E4868 | www.pnas.org/cgi/doi/10.1073/pnas.1512025112 Liu et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

unfolding threshold (∼5 pN) (38), although further experimentswould be necessary to confirm this.

Modulation of FA Architecture. The coincidence of the talin C ter-minus with the onset of F-actin (Fig. 1 B and C) suggests that talinmay physically define the z-position of the FA/stress fiber in-terface, and thus FA architecture. Therefore, we next investigatedhow minitalins affect the nanoscale organization of other FAcomponents. We used iPALM to examine the organization ofF-actin (Figs. 1C and 6A). In T100 cells, the sideview iPALM images(Figs. 1C and 6A) revealed that the FA/stress fiber interface re-sides at z >70 nm. In contrast, in HUVECs expressing minitalins,we observed the downshift of such interfaces by ∼10–20 nmcompared with T100 (Fig. 6A). Owing to the variable shapes andrelatively large sizes of stress fibers, we used (i) VASP, an actin-binding protein associated with F-actin barbed ends, to report onthe FA/stress-fiber interface z-position (43) and (ii) F-actin,mapped using Alexa Fluor 568-phalloidin, to report the stressfiber core z-position. As expected, we observed downshifting ofthe z-positions of VASP FP-fusion coexpressed in the minitalin

substituted cells (Figs. 6 B and C and 7A), although we were unableto locate VASP-containing T29 cells, suggesting that the in-teraction necessary for VASP localization may be absent in thisminitalin. Likewise, the stress fiber core z-positions also exhibit adownshifting trend (Fig. 7A and Fig. S7H). To further characterizethe relationship between talin length and FA nanostructure, weperformed two-color nanoscale mapping experiments measuringthe z-positions of F-actin and the talin C terminus. We observed apositive correlation between the talin C-terminal and F-actinz-positions at the pixel-to-pixel level (Fig. S7 C–E). Altogether,these results are consistent with talin physical length serving toregulate the FA vertical dimension.We note, however, that the slopes of the stress fiber core and

FA/stress-fiber interface regression lines are relatively shallowcompared with the talin C-termini regression lines (Fig. 7A),suggesting that the structural effect of talin length on FA ar-chitecture may be partially counteracted by other factors. Talincontains about 11 potential vinculin binding sequences (VBSs)with a cluster of five VBSs in the R1–R3 domains, the remainderbeing distributed throughout the rod (24). We determined vinculin

Fig. 5. Recombinant chimeric-talins support FA assembly and adopt a polarized orientation. (A) Schematic diagram of recombinant chimeric-talin constructs.The IgFLNa1-8 domain (yellow) is inserted in place of the deletion in minitalins. (B) Immunoblot analysis of mEmerald-tagged chimeric-talin fusions, probedwith anti-GFP antibody. (C) Epifluorescence micrographs of HUVECs cotransfected with hTalin1 siRNA (KD) or control siRNA (siCON), and mEmerald-taggedfull-length talin (T100), chimeric-talins (XT29-XT58), or GFP (control). GFP channel (green) indicates talin or chimeric-talins localization to FAs. (D) Magnifiedview. Efficient knockdown of endogenous talin is verified by hTalin1-specific antibody (huTLN, cyan). FAs and the actin cytoskeletal organization are visu-alized by paxillin immunofluorescence (magenta) and phalloidin (blue), respectively. (E) Topographic map of chimeric-talin z-position (C-terminal FP fusion).Color bars indicate z-position, 40–80 nm. (F) Box and whisker plot of chimeric-talin C-terminal z-positions. Notched boxes and histograms (bin size, 1 nm): firstand third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated in B, C, and E: median zFA (red), number of FAROIs (black), and number of cells (blue). (Scale bars: 25 μm in C, 12.5 μm in D, and 5 μm in E.)

Liu et al. PNAS | Published online August 17, 2015 | E4869

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

z-positions by mapping vinculin-mCherry (N-terminal tag) coex-pressed together with talin1 siRNA and GFP-tagged talin analogs.Vinculin in minitalin cells was located at lower z-positions relativeto T100 (Fig. S7 G and H). We observed that vinculin in T58 cellswas at an unusually lower z-position compared with the rest, im-plying that the VBS sites near the N terminus (likely the clusters offive VBSs in R1–R3 domains, Fig. S4A) may be activated to agreater extent. However, owing to the multiplicity of vinculinbinding partners in FAs, vinculin nanoscale organization is highlycomplex, as shown in a recent study describing a phospho-paxillin-dependent positioning mechanism (44). Minitalin interactions withvinculin are perturbed by the deletion of various VBS (Fig. S4).Thus, a further study characterizing the properties of each VBS willbe needed to fully account for the observed vinculin positioningeffects. Interestingly, with T58 considered as an outlier, the vinculinz-position regression line exhibits a slope similar to that of thestress-fiber core (Fig. S7H), suggesting that although vinculin maynot be required for FA nanoscale architecture (Fig. 2), whenpresent it may play a role in the integration of FAs with the actincytoskeleton (44).We next further examined the physical geometry of force

transmission in FAs by analyzing the iPALM sideview pro-jection of F-actin (Fig. 7C). We observed that the stress fibersapproach FAs at a very shallow angle: The contact angle ofF-actin at FA/stress fiber interfaces (θSF) seems to range between2° and 6° for the distal and proximal end of FAs, respectively.The shallow θSF angle compared with θTalin (∼15°) indicatesthat contractility conveyed by the stress fiber is not colinearlytransmitted to talin. Such an angular mismatch (Fig. 7D) impliesthat an additional counterbalancing force must be present tomaintain force balance. With the assumption that force genera-tion primarily occurs via stress-fiber-associated myosin II con-tractility, stress-fiber tension should have the greatest magnitude,and thus the counterbalancing tension at FA is expected to have a

significant plasma membrane parallel component, as depictedin Fig. 7D.

DiscussionIn this study we sought to establish the molecular basis of FAorganization and how FA components are geometrically orga-nized into force-transmission modules. We combined super-resolution microscopy with a molecular engineering approachsuch that information on molecular conformation, connectivity,and geometry could be gleaned from the measurements. Ourresults support the hypothesis that talin serves a physical role inregulating FA architecture, effectively setting the scale for theFA core, akin to a molecular ruler. The robust modularity oftalin enables the geometric organization of the integrin–talin–actin module to be determined with high precision. This alsoprovides empirical constraints suggesting that stress-fiber-borneactomyosin contractility and FA-borne tension may be mechan-ically coupled in a talin-geometry-dependent manner.By defining the physical geometry of the integrin–talin–actin

module, our measurements raise a number of interesting impli-cations for both the force-mediated extension of FA proteins andhow FA transmits force. First, we observed small but consistentand statistically significant elevation of the C-terminal z-posi-tions of T100 and minitalin constructs containing the L432Gcalpain-resistant mutation compared with constructs lacking thismutation (Figs. 4 and 7A). Similar results were obtained withcalpain inhibitors (Fig. S7 A and B). Based on our result of θTalin∼15°, the ∼81-nm length obtained for T100/L432G agreesremarkably well with the expected length of talin based onNMR and X-ray crystallographic studies of individual domains(∼80 nm), suggesting that talin/L432G is oriented in a polarizedmanner but with all domains compact, likely due to relatively low(<5 pN) tension. The T100/L432G mutation and calpain inhibitorshave previously been shown to inhibit cellular contractility (37). In

Fig. 6. Modulation of FA architecture. (A) iPALM 3D superresolution images of F-actin in HUVECs expressing full-length talin (T100) or minitalin constructs.(Upper) Top-view (xy) image with color encoding z-position. (Lower) Side-view projection of boxed regions in upper panel, cell edge on left. Histograms (bin size1 nm) of z-position shown on left. Color bar, 0–150 nm relative to substrate surface. Gray line, z = 50 nm. (Scale bars: 2 μm in top view and 250 nm in side view.)(B) Topographic map of vertical (z) positions of VASP in HUVECs expressing T100 or minitalins. Color bars indicate z-position, 30–100 nm. (Scale bar: 5 μm.) (C) Boxand whisker plot of FA/stress fiber interface as marked by VASP. Notched boxes, histograms (bin size, 1 nm): first and third quartiles, median and confidenceintervals; whiskers, 5th and 95th percentiles. Numbers indicated in B, C, and E: median zFA (red), number of FA ROIs (black), and number of cells (blue).

E4870 | www.pnas.org/cgi/doi/10.1073/pnas.1512025112 Liu et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

contrast, wild-type talin is ∼16 nm longer. This could be due to theforce-induced unzipping of the R2–R3 four-helical bundles, whichhas been shown in vitro to occur at ∼5–7 pN; further experimentsusing R2–R3 stabilizing mutations (45) may help to verify this.Because calpain-mediated proteolysis of talin is required for FAturnover and the up-regulation of promigratory signaling pathwayssuch as Src, the increased length of wild-type talin likely reflects arelatively higher talin tension resulting from greater contractility(36, 37, 46, 47). In vitro force-extension experiments on talin roddomains allow this tension to be estimated, suggesting that wild-type talin sustains greater than 5 pN but less than 15 pN of tension,because another unfolding (hence lengthening) transition occurs atthat level of force.

The modulation of FA architecture by talin analogs implicates theintegrin–talin–actin module as the primary force-transmitting unit,and thus the geometrical orientation of this module is important forhow FA transmits force to the ECM. MyosinII-generated tensionconveyed through the stress fiber must be balanced by the equivalentamount of net tension (Newton’s first law: All force vectors shouldsum to zero in an equilibrium). However, because we observed thattalin does not align colinearly with the stress fiber, talin tension canonly partially account for stress-fiber tension. This implicates thepresence, at the talin–actin interface, of an additional tension vectorwith a significant membrane-parallel component (Fig. 7D). Becausevinculin bridges the talin rod to actin and bears tension, we surmisethat vinculin cross-linking may account for such counterbalancingforce at the level of individual integrin–talin–actin modules (48).Collectively, the integration of such counterforces across numerousintegrin–talin–actin modules in a given FA may then give rise to amembrane-parallel cortical tension. That is, the stress fiber may exertforce on both the FA and the surrounding cortical cytoskeleton—thelatter could be considered analogous to friction, a common param-eter in biophysical models of FAs (14, 49).We note also that a similar force-balance situation may apply

at the talin–integrin interface. The measured talin geometrysuggests that talin is pulling on the integrin β cytodomain at 15°.Previous superresolution microscopy measurements indicatedthat the ECM–plasma membrane separation is 20–30 nm, compa-rable to the ∼20-nm crystallographic length of activated integrinectodomain (16, 27, 50). This suggests that the angle between theintegrin tension and the talin tension is probably also mismatched,because otherwise an extension of integrin β ectodomain to >75 nm,that is, 20 nm/sin−1(15°), would be required. Therefore, anothermembrane-parallel force vector, possibly the apparent membranetension, is likely accounting for the force balance. In other words,talin may exert a pulling force on both the integrin and the sur-rounding plasma membrane. Consistent with this, a number ofstudies have documented how membrane tension may influenceintegrin–ECM interaction (51, 52). An interesting possibility is thatsuch in-plane tensions may influence other FAs in the vicinity, thusproviding a degree of local mechanical coordination. We suggest thiscould be a possible molecular-mechanical basis for how FA forcetransmission to ECM may be interdependent on the integration ofmultiple cellular forces.Several recent studies have focused on single integrin tension

measurements at FAs, reporting a wide range of force magnitude,from a few piconewtons to >50 pN (52–56). Although our ex-periments do not directly measure tension, the range of variousforces can be estimated based on the measured geometry and invitro talin force/extension data (38). For example, given θSF of 5°,and θTalin of 15°, a 5- to 15-pN range of talin tension would cor-respond to a stress-fiber tension of 15–45 pN, and a membrane-parallel tension of 10–30 pN, per each integrin–talin–actin module(SI Discussion). Assuming that integrin tension is due largely totalin-mediated force, our measurements would correspond to asingle-integrin tension of ≤15 pN. Interestingly, this also impliesthat although a considerable tension is present in the stress fiber, itis significantly shunted away from the traction component andtoward the membrane-parallel component by the talin–actin anglemismatch. Because talin contains a long flexible linker region(residues 401–481) that could act as a swiveling joint, a wide rangeof talin contact angles is likely possible during the course of FAmaturation. We propose that variable talin geometry may differ-entially channel the actomyosin contraction force between thecortex and the ECM, and thus may help account for such a broadrange of observed tensions (Fig. S8 and SI Discussion). For ex-ample, if talin is nearly colinear with the actin stress fiber, that is,θTalin ∼ θSF, the majority of the actomyosin contraction is expectedto be channeled through talin. In particular, this may correspondto early cell spreading when large integrin tensions >50 pN havebeen implicated (52). Such high tension would also lead to talin

Fig. 7. Molecular geometry of the integrin–talin–actin module. (A) Relation-ship between talin sequence lengths and the z-positions of talin analogs, or FAcomponents. Median of zFA distribution (symbols), SD (error bands), and linearregression (solid line, y = mx + b): minitalins (blue square, b = 34.2 ± 2.66 nm,m = 0.326 ± 0.048); minitalins-L432G (blue circle, b = 31.7 ± 1.08 nm,m = 0.309 ±0.019); chimeric-talins (orange triangle, b = 41.6 ± 4.37 nm, m = 0.332 ±0.099);VASP (red inverted triangle, b = 65.16 ± 0.91 nm, m = 0.176 ± 0.015); F-actin(purple square, b = 90.42 ± 0.78 nm, m = 0.063 ± 0.013). (B) Talin geometry inFAs. Talin-membrane contact angle (θTalin) is calculated from ΔZFA, the z-offsetbetween chimeric-talins and minitalins regression lines, and the ∼30-nm lengthof the IgFLNa1-8 domain, yielding θTalin ∼15 °. (C) Actin stress fiber geometry atFA interface. Measurements based on side view iPALM image of actin (Fig. 1C).FA/actin contact angles (θSF) ranges from 2° to 6°, for the distal and proximal FAedges, respectively (cell edge on left). Color scale, 0–150 nm relative to substratesurface. Histograms (bin size, 1 nm). (Scale bar: 250 nm.) (D) Geometry andforce-balance relationship of the integrin–talin–actin module. Molecule sizesand orientation are approximately to scale. Talin dimer is omitted and somevectors are rescaled for clarity. Vectors indicate the direction of cellular forces,assuming mechanical equilibrium.

Liu et al. PNAS | Published online August 17, 2015 | E4871

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

becoming significantly extended, consistent with the observationof talin molecules spanning >200 nm in cells (39).At a mechanical equilibrium, in addition to the force balance,

torque balance is also required. Indeed, torque exertion by FAshas recently been documented by high-resolution 3D tractionforce microscopy (57). Together with recent studies suggestingthat FA traction force is spatially and temporally heterogeneousacross a given FA structure, this would imply that talin geometrymay be variable across FAs as well (54, 58). Current limitations toour methods are that the z-position measurements correspond toensemble-averaged values, that they represent a static snapshot,and that the majority of FAs are likely mature and stable FAs.Therefore, both intra-FA and inter-FA spatiotemporal heteroge-neity may not be apparent. Whereas recent developments of cell-matrix tension biosensors have largely emphasized magnitudemeasurement, our analysis suggests that talin geometry would bean important quantity to probe simultaneously with tension, es-pecially during different FA morphodynamic stages. Indeed, in anequilibrium of three force vectors, the measurement of two anglesand one force magnitude will allow the calculation of all otherforce vectors by trigonometric identity (SI Discussion). Becausethe stress-fiber geometry should be relatively easy to estimate forcells on 2D substrate, these can simplify to one angle and onetension measurement. A number of strategies may be possible forthe design of a single-molecule talin geometry reporter, such as ageometry-dependent FRET sensor (59) or by single-molecule lo-calization of a pair of spectrally distinct FPs incorporated at well-defined sites within talin. Direct and simultaneous measurementsof force and force-transmission structures in living cells will beinstrumental in understanding how complex mechanosensitiveresponses originate in FAs.Our study highlights the dominant structural role of talin in

FAs, which contrasts with the dispensability of vinculin in speci-fying FA nanoscale organization. Nevertheless, vinculin has beenshown to bear tension, playing important and complex roles in FAstrengthening, stabilization, and actin retrograde flow engagement(44, 48, 60, 61). Altogether this suggests a structural-mechanicalmodel in which the integrin–talin–actin module serves as theconstitutive structural component of the FA molecular clutch,whereas vinculin may function as a regulatable clutch element.In particular, although the integrin–talin–actin module may bestructurally sufficient for force transmission, cross-linking inter-actions mediated by vinculin and/or other proteins could helpmodulate the geometry of force transmission and thus regulatehow cell-generated force is apportioned toward or away fromECM traction. At the nanoscale, regulation of mechanical cross-linking between adjacent integrin–talin–actin modules may alsointroduce additional layers of cooperativity to allow complex

regulatory control and adaptive behaviors such as durotaxis (58).In conclusion, our measurements suggest a molecular basis for FAstructural organization and cell mechanical functions and highlightthe roles of molecular geometry, which should be measurable andtestable in further live-cell studies.

Materials and MethodsInstrumentation. Three-dimensional superresolution microscopy of F-actinwas performed on a custom-built iPALM instrument, using a 561-nm laser forthe excitation of tdEosFP, or a 642-nm laser for the excitation of AlexaFluor647 phalloidin. Transfected cells were located by low-intensity excitationfrom a 488-nm laser. A 405-nm laser was used for photoconversion of tdEos orphotoswitching of AlexaFluor 647. The theoretical foundation, operationalprinciples, construction, and alignment of iPALM have been described pre-viously (62). Topographic z mapping of protein position by surface-gener-ated structured illumination techniques was performed on a Nikon Eclipse Tiinverted microscope (Nikon Instruments) equipped with a motorized TIRFilluminator, a 60× N.A. 1.49 Apo TIRF objective, a sCMOS camera (Orca Flash4.0; Hamamatsu), and a fiber-coupled laser combiner (488 nm and 561 nm;Omicron Laserage). Fluorophore positions were measured using either a488-nm laser (for mEmerald or EGFP) or a 561-nm laser for mApple, mCherry,or photo-converted mEos2. When needed, photoconversion of mEos2 wascarried out using light-emitting diode excitation (SOLA; Lumencor) using theDAPI excitation filter set. A detailed description of the calibration, imagereconstruction, and data analysis for both techniques is given in SI Materialsand Methods.

Cell Culture and Sample Preparation. HUVECs (passage number <15) werecultured in a 5% CO2, 37 °C humidified atmosphere in Medium 200 (M-200-500; Life Technologies) supplemented with large vessel endothelial factors(A14608-01; Life Technologies). siRNA-mediated knockdown of human Talin1in HUVECs was performed as previously described (26), using custom stealthRNAi siRNA duplexes (oligo sequence: 5′-CCAAGAACGGAAACCUGCCAGA-GUU-3′; Life Technologies). To transfect cells with expression vectors and/orsiRNA, cells were trypsinized and electroporated using the Neon platform (LifeTechnologies). For imaging samples, cells were replated at a relatively sparsedensity of 7,000 per square centimeter on fibronectin-coated fiducialed cov-erglass, silicon wafers, or glass-bottom dishes (Iwaki) and imaged 16–20 h afterreplating. Expression vectors for FA protein fusions were described previously(16, 27). Expression vectors for the recombinant talin analogs were created asdetailed in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Caroline Ajo-Franklin (Lawrence BerkeleyNational Laboratory) for sharing MATLAB codes for variable incidence anglefluorescence interference contrast microscopy analysis, Gleb Shtengel andHarald Hess (Howard HughesMedical Institute, Janelia Farm Research Campus)for help with iPALM instrumentation and analysis and for discussion. We thankAlexander Dunn (Stanford University) and Jacques Prost (Institut Curie) forhelpful discussion and Min Wu and Ronen Zaidel-Bar (National University ofSingapore) for critical reading of the manuscript. P.K., J.L., Y.W., W.I.G., H.G.,and S.T. are supported by Singapore National Research Foundation FellowshipNRF-NRFF-2011-04 (to P.K.). S.R. is supported by the summer internshipprogramme of the Mechanobiology Institute.

1. Geiger B, Bershadsky A, Pankov R, Yamada KM (2001) Transmembrane crosstalk be-

tween the extracellular matrix–cytoskeleton crosstalk. Nat Rev Mol Cell Biol 2(11):

793–805.2. Bershadsky AD, Balaban NQ, Geiger B (2003) Adhesion-dependent cell mechano-

sensitivity. Annu Rev Cell Dev Biol 19:677–695.3. Ridley AJ, et al. (2003) Cell migration: Integrating signals from front to back. Science

302(5651):1704–1709.4. Parsons JT, Horwitz AR, Schwartz MA (2010) Cell adhesion: Integrating cytoskeletal

dynamics and cellular tension. Nat Rev Mol Cell Biol 11(9):633–643.5. Burridge K, Wittchen ES (2013) The tension mounts: Stress fibers as force-generating

mechanotransducers. J Cell Biol 200(1):9–19.6. Gardel M, Schneider I, Aratyn-Schaus Y, Waterman CM (2009) Mechanical integration

of actin and adhesion dynamics in cell migration. Annu Rev Cell Dev Biol 26:3151–333.7. Engler AJ, Sen S, Sweeney HL, Discher DE (2006) Matrix elasticity directs stem cell

lineage specification. Cell 126(4):677–689.8. McBeath R, Pirone DM, Nelson CM, Bhadriraju K, Chen CS (2004) Cell shape, cyto-

skeletal tension, and RhoA regulate stem cell lineage commitment. Dev Cell 6(4):

483–495.9. Paszek MJ, et al. (2005) Tensional homeostasis and the malignant phenotype. Cancer

Cell 8(3):241–254.10. Mammoto T, Mammoto A, Ingber DE (2013) Mechanobiology and developmental

control. Annu Rev Cell Dev Biol 29:27–61.

11. Giannone G, Mège RM, Thoumine O (2009) Multi-level molecular clutches in motile

cell processes. Trends Cell Biol 19(9):475–486.12. Shemesh T, Geiger B, Bershadsky AD, Kozlov MM (2005) Focal adhesions as mecha-

nosensors: A physical mechanism. Proc Natl Acad Sci USA 102(35):12383–12388.13. Chan CE, Odde DJ (2008) Traction dynamics of filopodia on compliant substrates.

Science 322(5908):1687–1691.14. Danuser G, Allard J, Mogilner A (2013) Mathematical modeling of eukaryotic cell

migration: Insights beyond experiments. Annu Rev Cell Dev Biol 29:501–528.15. Macdonald A, Horwitz AR, Lauffenburger DA (2008) Kinetic model for lamellipodal

actin-integrin ‘clutch’ dynamics. Cell Adhes Migr 2(2):95–105.16. Kanchanawong P, et al. (2010) Nanoscale architecture of integrin-based cell adhe-

sions. Nature 468(7323):580–584.17. Zaidel-Bar R, Itzkovitz S, Ma’ayan A, Iyengar R, Geiger B (2007) Functional atlas of the

integrin adhesome. Nat Cell Biol 9(8):858–867.18. Humphries JD, et al. (2009) Proteomic analysis of integrin-associated complexes

identifies RCC2 as a dual regulator of Rac1 and Arf6. Sci Signal 2(87):ra51.19. Schiller HB, Friedel CC, Boulegue C, Fässler R (2011) Quantitative proteomics of the

integrin adhesome show a myosin II-dependent recruitment of LIM domain proteins.

EMBO Rep 12(3):259–266.20. Kuo JC, Han X, Hsiao CT, Yates JR, 3rd, Waterman CM (2011) Analysis of the myosin-

II-responsive focal adhesion proteome reveals a role for β-Pix in negative regulation

of focal adhesion maturation. Nat Cell Biol 13(4):383–393.

E4872 | www.pnas.org/cgi/doi/10.1073/pnas.1512025112 Liu et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1

21. Zaidel-Bar R, Geiger B (2010) The switchable integrin adhesome. J Cell Sci 123(Pt 9):1385–1388.

22. Monkley SJ, et al. (2000) Disruption of the talin gene arrests mouse development atthe gastrulation stage. Dev Dyn 219(4):560–574.

23. Chen NT, Lo SH (2005) The N-terminal half of talin2 is sufficient for mouse develop-ment and survival. Biochem Biophys Res Commun 337(2):670–676.

24. Critchley DR (2009) Biochemical and structural properties of the integrin-associatedcytoskeletal protein talin. Annu Rev Biophys 38:235–254.

25. Zhang X, et al. (2008) Talin depletion reveals independence of initial cell spreadingfrom integrin activation and traction. Nat Cell Biol 10(9):1062–1068.

26. Kopp PM, et al. (2010) Studies on the morphology and spreading of human endo-thelial cells define key inter- and intramolecular interactions for talin1. Eur J Cell Biol89(9):661–673.

27. Shtengel G, et al. (2009) Interferometric fluorescent super-resolution microscopy re-solves 3D cellular ultrastructure. Proc Natl Acad Sci USA 106(9):3125–3130.

28. Ajo-Franklin CM, Ganesan PV, Boxer SG (2005) Variable incidence angle fluorescenceinterference contrast microscopy for z-imaging single objects. Biophys J 89(4):2759–2769.

29. Paszek MJ, et al. (2012) Scanning angle interference microscopy reveals cell dynamicsat the nanoscale. Nat Methods 9(8):825–827.

30. Choi CK, et al. (2008) Actin and alpha-actinin orchestrate the assembly and matura-tion of nascent adhesions in a myosin II motor-independent manner. Nat Cell Biol10(9):1039–1050.

31. Wiedenmann J, et al. (2004) EosFP, a fluorescent marker protein with UV-induciblegreen-to-red fluorescence conversion. Proc Natl Acad Sci USA 101(45):15905–15910.

32. Saunders RM, et al. (2006) Role of vinculin in regulating focal adhesion turnover. Eur JCell Biol 85(6):487–500.

33. Thievessen I, et al. (2013) Vinculin-actin interaction couples actin retrograde flow tofocal adhesions, but is dispensable for focal adhesion growth. J Cell Biol 202(1):163–177.

34. Dumbauld DW, et al. (2013) How vinculin regulates force transmission. Proc Natl AcadSci USA 110(24):9788–9793.

35. Goult BT, et al. (2013) Structural studies on full-length talin1 reveal a compact auto-inhibited dimer: Implications for talin activation. J Struct Biol 184(1):21–32.

36. Franco SJ, et al. (2004) Calpain-mediated proteolysis of talin regulates adhesion dy-namics. Nat Cell Biol 6(10):977–983.

37. Kotecki M, Zeiger AS, Van Vliet KJ, Herman IM (2010) Calpain- and talin-dependentcontrol of microvascular pericyte contractility and cellular stiffness. Microvasc Res80(3):339–348.

38. Yao M, et al. (2014) Mechanical activation of vinculin binding to talin locks talin in anunfolded conformation. Sci Rep 4:4610.

39. Margadant F, et al. (2011) Mechanotransduction in vivo by repeated talin stretch-relaxation events depends upon vinculin. PLoS Biol 9(12):e1001223.

40. del Rio A, et al. (2009) Stretching single talin rod molecules activates vinculin binding.Science 323(5914):638–641.

41. Nakamura F, Osborn TM, Hartemink CA, Hartwig JH, Stossel TP (2007) Structural basisof filamin A functions. J Cell Biol 179(5):1011–1025.

42. Chen H, et al. (2011) Differential mechanical stability of filamin A rod segments.Biophys J 101(5):1231–1237.

43. Bear JE, Gertler FB (2009) Ena/VASP: Towards resolving a pointed controversy at thebarbed end. J Cell Sci 122(Pt 12):1947–1953.

44. Case LB, et al. (2015) Molecular mechanism of vinculin activation and nanoscalespatial organization in focal adhesions. Nat Cell Biol 17(7):880–892.

45. Goult BT, et al. (2013) RIAM and vinculin binding to talin are mutually exclusive andregulate adhesion assembly and turnover. J Biol Chem 288(12):8238–8249.

46. Huang C, et al. (2009) Talin phosphorylation by Cdk5 regulates Smurf1-mediated talinhead ubiquitylation and cell migration. Nat Cell Biol 11(5):624–630.

47. Cortesio CL, et al. (2008) Calpain 2 and PTP1B function in a novel pathway with Src toregulate invadopodia dynamics and breast cancer cell invasion. J Cell Biol 180(5):957–971.

48. Grashoff C, et al. (2010) Measuring mechanical tension across vinculin reveals regu-lation of focal adhesion dynamics. Nature 466(7303):263–266.

49. Oakes PW, Gardel ML (2014) Stressing the limits of focal adhesion mechanosensitivity.Curr Opin Cell Biol 30:68–73.

50. Luo BH, Springer TA (2006) Integrin structures and conformational signaling. CurrOpin Cell Biol 18(5):579–586.

51. Ferraris GM, et al. (2014) The interaction between uPAR and vitronectin triggers li-gand-independent adhesion signalling by integrins. EMBO J 33(21):2458–2472.

52. Wang X, Ha T (2013) Defining single molecular forces required to activate integrinand notch signaling. Science 340(6135):991–994.

53. Liu Y, et al. (2014) Nanoparticle tension probes patterned at the nanoscale: Impact ofintegrin clustering on force transmission. Nano Lett 14(10):5539–5546.

54. Blakely BL, et al. (2014) A DNA-based molecular probe for optically reporting cellulartraction forces. Nat Methods 11(12):1229–1232.

55. Morimatsu M, Mekhdjian AH, Adhikari AS, Dunn AR (2013) Molecular tension sensorsreport forces generated by single integrin molecules in living cells. Nano Lett 13(9):3985–3989.

56. Morimatsu M, Mekhdjian AH, Chang AC, Tan SJ, Dunn AR (2015) Visualizing the in-terior architecture of focal adhesions with high-resolution traction maps. Nano Lett15(4):2220–2228.

57. Legant WR, et al. (2013) Multidimensional traction force microscopy reveals out-of-plane rotational moments about focal adhesions. Proc Natl Acad Sci USA 110(3):881–886.

58. Plotnikov SV, Pasapera AM, Sabass B, Waterman CM (2012) Force fluctuations withinfocal adhesions mediate ECM-rigidity sensing to guide directed cell migration. Cell151(7):1513–1527.

59. Meng F, Sachs F (2012) Orientation-based FRET sensor for real-time imaging of cel-lular forces. J Cell Sci 125(Pt 3):743–750.

60. Hu K, Ji L, Applegate KT, Danuser G, Waterman-Storer CM (2007) Differentialtransmission of actin motion within focal adhesions. Science 315(5808):111–115.

61. Galbraith CG, Yamada KM, Sheetz MP (2002) The relationship between force andfocal complex development. J Cell Biol 159(4):695–705.

62. Shtengel G, et al. (2014) Imaging cellular ultrastructure by PALM, iPALM, and cor-relative iPALM-EM. Methods Cell Biol 123:273–294.

63. Bustamante C, Marko JF, Siggia ED, Smith S (1994) Entropic elasticity of lambda-phage DNA. Science 265(5178):1599–1600.

64. Senetar MA, McCann RO (2005) Gene duplication and functional divergence duringevolution of the cytoskeletal linker protein talin. Gene 362:141–152.

65. Zhang F, Saha S, Kashina A (2012) Arginylation-dependent regulation of a proteolyticproduct of talin is essential for cell-cell adhesion. J Cell Biol 197(6):819–836.

66. Goksoy E, et al. (2008) Structural basis for the autoinhibition of talin in regulatingintegrin activation. Mol Cell 31(1):124–133.

67. Lee HS, Lim CJ, Puzon-McLaughlin W, Shattil SJ, Ginsberg MH (2009) RIAM activatesintegrins by linking talin to ras GTPase membrane-targeting sequences. J Biol Chem284(8):5119–5127.

68. Luo T, Mohan K, Iglesias PA, Robinson DN (2013) Molecular mechanisms of cellularmechanosensing. Nat Mater 12(11):1064–1071.

69. Bachir AI, et al. (2014) Integrin-associated complexes form hierarchically with variablestoichiometry in nascent adhesions. Curr Biol 24(16):1845–1853.

70. Biro M, et al. (2013) Cell cortex composition and homeostasis resolved by integratingproteomics and quantitative imaging. Cytoskeleton (Hoboken) 70(11):741–754.

71. Patterson G, Davidson M, Manley S, Lippincott-Schwartz J (2010) Superresolutionimaging using single-molecule localization. Annu Rev Phys Chem 61:345–367.

72. Engler C, Kandzia R, Marillonnet S (2008) A one pot, one step, precision cloningmethod with high throughput capability. PLoS One 3(11):e3647.

73. Ren J, et al. (2009) DOG 1.0: Illustrator of protein domain structures. Cell Res 19(2):271–273.

74. Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1996) Numerical Recipes in C(Cambridge Univ Press, Cambridge, UK).

75. Meier PC, Zünd RE (2000) Statistical Methods in Analytical Chemistry (Wiley, NewYork), 2nd Ed.

Liu et al. PNAS | Published online August 17, 2015 | E4873

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Oct

ober

21,

202

1