Talin determines the nanoscale architecture of focal adhesions

Significance

Focal adhesions (FAs) mediate cell–extracellular matrix interactions and consist of integrin receptors linked to the actin cytoskeleton via multiprotein complexes organized into nanoscale strata. In this work, we sought to determine the molecular basis of FA nanostructure. Combining superresolution microscopy and protein engineering, we demonstrate the structural role of talin in regulating the nanoscale architecture of FAs. Talin specifies the dimension of the FA core, akin to a molecular ruler, in a remarkably modular manner. Our results define the molecular geometry of the integrin–talin–actin module that comprises the key mechanical linkage within FAs and elucidate how such interactions serve to integrate multiple cellular forces at adhesion sites.

Abstract

Insight into how molecular machines perform their biological functions depends on knowledge of the spatial organization of the components, their connectivity, geometry, and organizational hierarchy. However, these parameters are difficult to determine in multicomponent assemblies such as integrin-based focal adhesions (FAs). We have previously applied 3D superresolution fluorescence microscopy to probe the spatial organization of major FA components, observing a nanoscale stratification of proteins between integrins and the actin cytoskeleton. Here we combine superresolution imaging techniques with a protein engineering approach to investigate how such nanoscale architecture arises. We demonstrate that talin plays a key structural role in regulating the nanoscale architecture of FAs, akin to a molecular ruler. Talin diagonally spans the FA core, with its N terminus at the membrane and C terminus demarcating the FA/stress fiber interface. In contrast, vinculin is found to be dispensable for specification of FA nanoscale architecture. Recombinant analogs of talin with modified lengths recapitulated its polarized orientation but altered the FA/stress fiber interface in a linear manner, consistent with its modular structure, and implicating the integrin–talin–actin complex as the primary mechanical linkage in FAs. Talin was found to be ∼97 nm in length and oriented at ∼15° relative to the plasma membrane. Our results identify talin as the primary determinant of FA nanoscale organization and suggest how multiple cellular forces may be integrated at adhesion sites.

Cell adhesion to the ECM is a highly coordinated process that involves ECM-specific recognition by integrin transmembrane receptors, and their aggregation with numerous cytoplasmic proteins into dense supramolecular complexes called focal adhesions (FAs) (1). Actin stress fibers terminate at FAs where actomyosin contractility is transmitted to the ECM, generating traction (2–5). Mechanical tension impinging on each FA is implicated in key steps including the elongation, reinforcement, and maintenance of the FA structures (6). FA mechanotransduction is a major aspect of cellular microenvironment sensing with wide-ranging consequences in physiological and pathological processes (7–10). However, molecular-scale spatial parameters that specify FA nanoscale organization have been difficult to access experimentally. Nevertheless, these are essential to understand how mechanosensitivity arises 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 integrin and the actin cytoskeleton along the vertical (z) axis (16). The FA core consists of a membrane-proximal layer that contains signaling proteins such as FAK (focal adhesion kinase) and paxillin, an intermediate zone that contains force-transduction proteins such as talin and vinculin, and a stress fiber interfacial zone that contains actin-associated proteins such as VASP (vasodilator-stimulated protein) and α-actinin. Although such multilaminar architecture signifies a certain degree of compartmentalization within FAs that may serve to spatially constrain protein–protein interactions and dynamics, the structural connectivity, the molecular configuration and geometry of FA proteins, and the molecular basis of their higher-order organization remain unclear.

Proteomic and interactome analysis of the integrin adhesome have uncovered several direct and multitier connections between integrins and actin (17–20). This suggests that multiple highly interconnected protein–protein interactions could collectively self-organize into FA structures; such redundancy could also account for the remarkable mechanical robustness of FAs after cellular disruption or perturbation (21). Alternatively, a specific FA component may play a dominant role in regulating FA architecture. Aspects of both scenarios may also act cooperatively or function 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 cytoskeletal adaptor protein, adopts a highly polarized orientation in FAs (16), with the N terminus residing in the membrane-proximal layer and the C terminus elevated by z ∼30 nm to the FA/stress fiber interfacial zone. This led us to hypothesize that an array of 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) domains but with selective deletion of the multiple helical bundles within the central region of talin. By using a siRNA-mediated knockdown/rescue approach, we found that such talin analogs were able to support FA formation, clustering of activated integrins, and linkages to the actin cytoskeleton. By mapping the z-position of the FPs tagged at either the N or the C termini, we show that talin and its analogs are linearly extended and oriented in FAs, with their lengths regulating FA nanoscale organization. Chimeric-talin analogs with a 30-nm spacer insertion are also able to support FA assembly, facilitating the precise determination of talin geometry in FAs. Our results indicate that talin is oriented at 15° relative to the plasma membrane, measuring ∼97 nm end to end. FA nanoscale architecture in vinculin-null mouse embryonic fibroblasts (MEFs) retained its stratified organization and talin polarization similar to that in other cell types, suggesting that vinculin is dispensable for the specification of FA architecture. Our measurements demonstrate how the integrin–talin–actin module serves as the primary, and surprisingly modular, structural and tension-bearing core of FAs and geometrically define how such complexes could integrate multiple cellular forces at adhesion sites.

Results

Nanoscale Protein Stratification in FAs.

Two highly homologous talin genes, talin1 (tln1) and talin2 (tln2), have been identified; loss of talin1 is embryonic-lethal, whereas talin2-null mice are viable and fertile (22, 23). Nevertheless, both isoforms are expressed in most cell types (24) (Fig. S1 A–C), and talin2 is able to support FA assembly and cell spreading in talin1-null mouse fibroblasts (25). To study the structural role of talin in FAs, we therefore chose primary human umbilical vein endothelial cells (HUVECs), which express only talin1 (26) (Fig. S1 A–C). siRNA-mediated depletion of endogenous human talin1 results in impaired cell spreading and migration (Fig. S1 D–J), as well as FA assembly defects, which can be rescued by coexpression of a mouse talin1 GFP fusion as described previously (25, 26). The HUVEC platform thus allows substitution of endogenous talin1 with engineered talin constructs.

Fig. S1.

Expression and siRNA-mediated depletion of talin1 in HUVECs. (A) Immunolocalization of talin isoforms in HUVECs and human foreskin fibroblasts (HFF1). Top row, first and third columns, staining with a pan-Talin antibody (TLN1+TLN2, clone 8d4). Top row, second and fourth columns, staining with a talin2-specific antibody (clone 121A). Middle row, Paxillin staining. Bottom row, merged images (green: Talin, magenta: paxillin). (Scale bar: 10 μm.) (B) Immunoblot analysis of HUVECs (left) and HFF1 (right), using antibodies specific to (top to bottom) pan-Talin (1 + 2) (clone 8d4), talin1 (clone 93E12), talin2 (clone 121A), talin2 (clone 538), and GAPDH (loading control). (C) RT-PCR for talin1 and talin2 in HUVECs, Ea.hy926 (immortalized human umbilical vein cell line), and HFF1 cells. Primers used as described in SI Materials and Methods. (D and E) Flow cytometric analysis of HUVECs transfected with Alexa Fluor 647-conjugated control siRNA (D) or hTalin1 siRNA (E). Transfected cells (quadrant 1): 99.2% (D), 98.9% (E). (F) Time-course immunoblot analysis of siRNA-knockdown of human talin1 in HUVECs, 24–96 h. (G and H) Migration trajectory of HUVECs transfected with control siRNA + GFP (G) or hTalin1 siRNA + GFP (H) over 12 h on fibronectin; N_cells: 62 (SiCON) and 97 (SiTLN1). (I) Time-lapse phase contrast images of HUVECs transfected with control siRNA (Upper) or hTalin1 siRNA (Lower) spreading on fibronectin. (Scale bar: 20 μm.) (J) Statistical comparisons of cell migration parameters for control siRNA (blue) and hTalin1 siRNA (red).

We first sought to characterize FA architecture in HUVECs. Previously we reported that the primary organization axis for FAs is along the vertical (z) dimension (16). Here we used two techniques for the nanoscale measurement of FA protein z-positions. Interferometric photoactivated localization microscopy (iPALM) combines single-molecule localization for xy superresolution and multiphase interferometry for z superresolution, providing sub-20-nm resolution in 3D (27) (Fig. S2). We also used a surface-generated structured illumination technique—variable incidence angle fluorescence interference contrast microscopy (VIA-FLIC) (28) or scanning angle interference microscopy (SAIM) (29)—which map the ensemble z-position of fluorophores with high speed and a larger field of view, albeit with a diffraction-limited resolution in xy (Fig. S3). Although VIA-FLIC/SAIM are not true superresolving techniques per se, in that vertically overlapping objects are not resolved (28, 29), FA proteins were shown to organize in distinct strata (16). These are thin relative to the axial structured illumination pattern, and thus the z-position determined by VIA-FLIC/SAIM should closely correspond to the ensemble-averaged z-position of a given FA protein.

Fig. S2.

Three-dimensional superresolution microscopy by iPALM. (A) The iPALM instrument. 1 and 1′: Sample holder z piezo; 2 and 2′: upper and lower objective lenses (60×, N.A. 1.49); 3: TIRF excitation and photoactivation beampath; 4 and 4′: 22.5° slotted mirrors; 5: imaging sample (sealed chamber made from two no. 1.5 coverglasses); 6 and 6′: beampaths of fluorescence emission collected by objective lenses; 7: the Hess beamsplitter for three-way multiphase interferometry; 8′, 8′′, and 8′′′: beampaths of self-interfered emission exiting the beamsplitter; not shown: 400-mm tube lenses for each exiting beam, emission filter wheels, electron-multiplying CCD cameras, stage translator. (B) Schematic diagram of the Hess beamsplitter. Surface 1, 2:1 transmission/reflection; surface 2, 1:1 transmission/reflection; surface 3: 100% reflection with index-matching oil, mounted on a three-axis z-tip-tilt piezo pedestal. (C) iPALM calibration curve. Plasmonic gold nanoparticles immobilized on the coverglass are used as fiducial marks. The sample vertical position (z) is scanned by the piezoelectric stage (A: 1 and 1′). Fits to camera 1 (beam 8′′′, red); camera 2 (beam 8′′, green); camera 3 (beam 8′, blue). (D) Extracted z-position of fiducials (black) vs. the piezo stage position (red). (E and F) Localization precision in x (E) and z (F), estimated by repeatedly imaging a fiducial with photon numbers similar to typical Alexa Fluor 647 photon output.

Fig. S3.

Measurement of topographic Z map with nanoscale precision by surface-generated structured illumination. (A) Principles of surface-generated structured illumination (VIA-FLIC or SAIM) for high-precision mapping of fluorophore (green) z-position. (B) Intensity of excitation field as a function of incidence angle (θ_inc) and fluorophore height (z). (C) Schematic diagram of VIA-FLIC/SAIM measurements. (D) Angle dependence of fluorescence emission as a function of fluorophore z-position. The dramatic changes in lineshape as a function of z provide the basis for the nanoscale precision of the methods. (E) Topographic z map of rhodamine-fibronectin–coated substrate relative to SiO₂ surface. Z-positions are extracted pixel by pixel as described in SI Materials and Methods. Thermal SiO₂ thickness, 508.6 nm; excitation wavelength, 561 nm; color bar, 0–100 nm. (Scale bar: 5 μm.) (F) Histogram of z-position distribution (red) and fit to a Gaussian function (blue dashed line). Total number of pixels fitted and peak z-value ± SE shown.

We cultured HUVEC cells expressing FP-tagged FA proteins on a fibronectin-coated thermal SiO₂ surface of silicon wafers for VIA-FLIC/SAIM or on fibronectin-coated fiducialed coverglasses for iPALM. In VIA-FLIC/SAIM analysis, the topographic map of z-positions are extracted pixel by pixel for FA regions. These z_pixel values are offset by the average thickness of the fibronectin layer, which was determined to be ∼10 nm (Fig. S3). We used the median z-position from each region of interest (ROI) to represent the FA protein z-position (z_FA). We verified that total z_FA distributions are very similar to that of the total z_pixel distribution, and thus should be representative of the FA protein z-positions. We observed a hierarchical organization of FA proteins along the z axis (Fig. 1 A and B), similar to earlier measurements in other cell types (16). Integrin α_v cytoplasmic tail, FAK, and paxillin are located in close proximity to one another, forming a membrane proximal layer at ∼45 nm above the ECM, whereas actin regulatory or actin-associated FA proteins such as VASP, α-actinin, and zyxin are observed at higher z-positions, >70 nm. We next used iPALM to image the full-3D extent of actin organization (Fig. 1C) at the FA/stress fiber junctions (Fig. 1C). The vertical profile of actin (histogram, Fig. 1C, Lower) peaks at z ∼80–120 nm, below which the actin density rapidly declines, signifying the interface between the FA core and actin stress fibers. At higher z values, actin seems to taper rapidly into densely bundled stress fibers. The onset of F-actin compaction at z ∼120 nm coincides with the z-position of α-actinin, (Fig. 1B), consistent with its role in cross-linking F-actin (30).

Fig. 1.

Nanoscale architecture of FAs in HUVECs. (A) Topographic maps of FA protein z-positions (nanometers) in HUVECs. (B) Nanoscale stratification of FA protein z-positions. Notched boxes and histograms (bin size, 1 nm) for z_FA of indicated FA proteins: first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated: median z_FA (red), number of FA ROIs (black), and number of cells (blue). (C) iPALM 3D superresolution 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 sparse within FA core region, whereas talin adopts a polarized N–C orientation as described 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. Color bar 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.)

The nanoscale precision of our imaging techniques enables us to selectively map the positions of different ends of a protein, allowing inference of molecular orientation (16). Measurements of talin tagged 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 nm respectively, indicative of a polarized orientation. This was cross-verified by iPALM imaging using talin tagged with a photoactivatable FP, tdEos (31) (Fig. 1 D and E). The narrow z-position distributions of the N and C termini and their minimal mutual overlap confirm that most talin molecules adopt a highly polarized N–C orientation.

Because measurements of vinculin tagged at either termini also reveal a directional organization (Fig. 1B), we further investigated the role of vinculin in FA architecture by mapping the positions of FA proteins in vinculin-null MEFs (32). As shown in Fig. 2, we observed a similar hierarchical organization of FA proteins along the z-dimension, with the N–C polarized orientation of talin that seems to span the FA core, linking with actin stress fibers. These results suggest that vinculin is dispensable for the stratified nanoscale architecture of FAs, consistent with previous studies documenting FA assembly and maturation in the absence of vinculin (33, 34).

Fig. 2.

Nanoscale architecture of FAs in vinculin-null MEF. (A) Topographic map of FA protein z-positions (nanometers) in vinculin-null MEF. (B) Nanoscale stratification of FA protein z-positions. Notched box plots and histograms (bin size, 1 nm) for z_FA of indicated FA proteins: first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated: median z_FA (red), number of FA ROIs (black), and number of cells (blue). (C) iPALM 3D superresolution 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 projection of 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, and 250 nm in C, Lower.)

Probing the Structural Role of Talin by Recombinant Minitalin Analogs.

We next examined the hypothesis that talin may physically regulate FA architecture. The observed N–C polarization of talin (Fig. 1 D and E) suggests that its N terminus is positioned via binding of the talin FERM domain to integrin β cytodomains, whereas the C terminus is upshifted in z-position via the THATCH domain interacting with F-actin. In this view, integrin–talin–actin modules are aligned into a parallel, oriented array that comprises the tension-bearing, structural core of FAs, around which other FA proteins are organized. To test this and to further define the physical and molecular basis of such organization, we asked whether full-length endogenous talin could be substituted with recombinant analogs with altered lengths, and whether such analogs modulate the nanostructure of FAs.

Earlier superresolution imaging results suggested that both the talin FERM and THATCH domains are capable of targeting to their respective nanoscale compartments in FAs (16). We therefore created a series of recombinant mouse talin1 analogs that contain the FERM (residues 1–482) and THATCH (residues 2294–2541) domains but with selective deletion of the central talin rod domains. The talin rod region consists of 13 tandem α-helical bundles, forming a flexible chain of globular domains (Fig. S4A). Recent structural studies suggest that individual domains of the talin rod are highly modular (35), permitting us to generate “minitalins” analogs that vary in sequence from 29 to 58% of full-length talin. We refer to these by their nominal sequence lengths, e.g., T29 for the analog that contains 29% of the sequence of full-length talin (T100), and so on (Fig. 3 A and B). To allow nanoscale mapping of their positions and orientations, the constructs were tagged with GFP (or spectral variant) at either the N or the C terminus.

Fig. S4.

Recombinant talin analogs. (A) Talin1 structure. Models assembled from X-ray crystallography and NMR, adapted from ref. 45. Key domains, sequence number, and binding sites are indicated. (B) Schematic of talin analogs (Left) and structure-based models (Right), shorthand notation (T##) indicated. Note the slight deviations from the sequence-based and the structure-based lengths due to the topology of the R2–R3 four-helical bundles, e.g., T46. (C) Differences in the end-to-end (r_N–C) lengths between four-helix and five-helix bundles. Four-helix (Talin R2 domain, PDB ID code 2L7A); five-helix (Talin R6 domain, PDB ID code 2L10). (D) FilaminA domain structure and key binding sites. The IgFLNa1-8 region that we inserted into chimeric-talin is highlighted in yellow.

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 HUVECs cotransfected 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 cytoskeletal actin 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.)

We next tested whether the minitalins were able to support FA formation in place of endogenous human talin. HUVECs were cotransfected with a human talin1 siRNA and cDNAs encoding mouse minitalin FP fusions, cultured on fibronectin-coated substrates and imaged ∼96 h posttransfection to ensure optimal knockdown of endogenous talin1 (Fig. S1F). A monoclonal antibody that recognizes hTalin1 but not mTalin1 was used to assess knockdown efficiency by immunofluorescence microscopy (Fig. 3C). We observed that the minitalins rescued FA formation and colocalized with other FA proteins such as paxillin and activated integrins (Fig. 3C and Fig. S5A), forming linkages to F-actin bundles (Fig. 3A). GFP fluorescence was observed in FA-like clusters for all minitalins (Fig. 3C). With the larger minitalins (T41–T58), cell spreading was well-supported relative to full-length talin (T100), whereas a substantial reduction in cell area was observed with the smaller T29 and T36 constructs (Fig. S5C). Nevertheless, the latter still supported significant spreading compared with hTalin1 siRNA cells. Overall, these results indicate that the minitalins are able to support FA assembly, enabling further analysis of the nanoscale architecture of these modified FAs.

Fig. S5.

Effects on integrin clustering and cell areas by recombinant talin analogs. Epifluorescence micrographs of HUVECs cotransfected with hTalin1 siRNA (KD) or control siRNA (siCON), and mEmerald-tagged full-length talin (T100), GFP (control), and (A) minitalin constructs (T29–T58) or (B) chimeric-talin (XT29–XT58). GFP channel (green) indicates talin or minitalin localization to FAs. Efficient knockdown of endogenous talin was verified by hTalin1-specific antibody (cyan). Activated integrin was probed with 9EG7 antibody, and F-actin was probed with phalloidin (blue). (Scale bar: 25 μm.) (C) Effect of talin analogs on cell spread areas. Notched boxes: first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Number of cells indicated (blue).

Nanoscale Organization of Minitalins in FAs.

We next mapped the nanoscale z-positions of the N and C termini of the minitalins to characterize their organization in FAs (Fig. 4A). The N termini of minitalins were observed at z ∼40 nm, similar to that of the full-length talin (T100), indicating that their FERM domains remained in the membrane proximal layer (Fig. 4 B and D). In contrast, their C-termini z-positions were observed to be significantly downshifted relative to T100 (z = 66.4 nm), ranging from z = 54.8 nm (T58) to z = 43.1 nm (T29) (Fig. 4 C and D and Fig. S6). We found that the sequence lengths of these talin analogs correlate linearly with the C-terminal z-positions (R = 0.95). Slight deviations were observed for some constructs such as T46. This likely reflects topological variations among the rod domains; the N and C termini of four-helical bundles are in the cis configuration instead of trans in the case of five-helical bundles that comprise the majority of rod domains (Fig. S4 B and C). Thus, the end-to-end length of T46 (FERM–R1–R2–R3–THATCH) may be somewhat shorter than expected from its sequence length because the R2 and R3 domains are four-helical bundles.

Fig. 4.

Polarized orientation of recombinant minitalins. (A) Topographic map of talin analog z-positions (nanometers). N-terminal FP fusions (Upper) and C-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, median and confidence intervals; whiskers, 5th and 95th percentiles, for N-terminal probes (B), and C-terminal probes (C). (D) Comparison of N- and C-terminal z_FA positions. Numbers of FA ROIs in parentheses; median z_FA of the distribution indicated in red (B and C). (E) Polarized orientation of talin and minitalins is independent of calpain cleavage. Topographic map of C-terminal FP fusions containing L432G mutation (denoted by asterisk) that suppress calpain II cleavage in the linker 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 z_FA (red), number of FA ROIs (black), and number of cells (blue). (G) Comparison of the C-terminal z-positions of talin analogs with or without the L432G mutation. Statistical significance in D 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.)

Fig. S6.

Statistics of protein Z-positions in FAs. (A) Statistics of FA protein z-positions in HUVECs (Fig. 1 A and B). (B) Statistics of FA protein z-positions in Vinculin (^−/−) MEFs (Fig. S3). (C) Statistics of z-positions of recombinant talin analogs coexpressed with Talin1 siRNA in HUVECs. (D–F) Normalized histograms of pixel z-position (z_pixel) for recombinant talin analogs: minitalins (D), minitalins/L432G (E), chimeric-talin (F). (G) C-terminal z-positions of minitalins and chimeric-talin. Notched boxes: first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Full-length talin (T100) in orange; minitalins in blue; chimeric-tain in coral. ****P << 10⁻⁶, Mann–Whitney u test. (H) Statistics of FA protein z-positions in cells substituted with recombinant talin analogs in HUVECs. (I–K) ANOVA Tukey test for z-positions of recombinant talin analogs: N termini of minitalins (I), C termini of minitalins (J), C termini of minitalin/L432G (K).

We next sought to check whether the differences in the N- and C-termini z-positions observed above represent the intact talin molecule or that of N- and C-terminal talin fragments generated by calpain-mediated proteolysis in the linker between the FERM and rod domains (36). We therefore introduced the calpain-resistant L432G mutation in the linker (36) and mapped the C-terminal z-positions of talin and minitalin constructs. The C terminus of T100/L432G was again found at an elevated z-position (z = 62.6 nm), confirming that this accurately reflects the polarized orientation of intact talin. Likewise, the C-termini z-positions of the L432G minitalins again exhibited a downward shift relative to that of T100/L432G (Fig. 4 E–G). These scale linearly with the sequence lengths, with a slope similar to that of the original minitalin constructs (P = 0.404; SI Materials and Methods), but with a small but statistically significant downshift in z (Fig. 4G). The slightly lower C-terminal z-positions due to the L432G mutation may reflect the reduction in calpain-mediated cellular contractility documented previously (37). Consistent with this, the C-terminal z-position of wild-type T100 was similarly lowered upon pharmacological inhibition of calpain (z = 62.2 nm, Fig. S7 A and B). Altogether, these measurements establish that talin and minitalins are highly polarized in FAs, with the N terminus in the membrane proximal layer, and the C terminus at elevated z-positions. These also demonstrate the structural continuity of talin, which is required for mechanical force transmission.

Fig. S7.

(A) Topographic map of talin C-terminal z-positions in KD+T100 HUVECs with pharmacological inhibition of calpain II. Color bar, 40–80 nm. (B) Notched box plots and histograms (bin size, 1 nm) for z_FA of indicated FA proteins: first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated: median z_FA (red), number of FA ROIs (black), and number of cells (blue). ****P <<10⁻¹², Mann–Whitney u test. (C) Dual-channel z-position analysis of talin C terminus and F-actin. Topographic z map of talin C terminus (Left) and F-actin (Right). (D) Merged image (green: KD+T100, mEmerald) and F-actin (red: phalloidin, Alexa Fluor 568). (E) Pixel-based correlation between F-actin (x axis) and talin C terminus (y axis). Gray line, linear regression: y = mx + b, m = 0.461 ± 0.021, b = 19.713 ± 2.10. Spearman rank correlation ρ = 0.2416 (P <<10⁻¹², two-tailed test). (F) Topographic map of vinculin coexpressed in T100 or minitalin HUVECs. Color bars indicate z-position, 30–100 nm. (G) Box and whisker plot of vinculin z-positions. Notched boxes, histograms (1- nm bin): first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated in B, C, and E: median z_FA (red), number of FA ROIs (black), and number of cells (blue). (Scale bars in A, C, D, and F: 5 μm.) (H) Relationship between talin sequence lengths and the z-positions of vinculin (red) or stress fiber core (purple). Median of z_FA distribution (symbols), SD (error bands), and linear regression (solid line, y = mx + b): stress fiber (purple circle, b = 90.42 ± 0.78 nm, m = 0.063 ± 0.013), vinculin (red square, b = 67.7 ± 0.87 nm, m = 0.073 ± 0.015).

Probing Talin Geometry by Minitalin:FilaminA Chimeric Analogs.

The linear scaling between the C-termini z-positions and construct lengths implicates a linearly extended configuration (SI Discussion) and suggests that these talin analogs may be oriented with a similar contact angle (θ_Talin) relative to the plasma membrane. Our measurements thus provide an opportunity to determine the contact angle of how talin connects integrin to actin, which is a critical 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 talin rod to engineer in a “spacer” protein domain with a well-defined length. We chose the filaminA IgFLNa1–8 domains (residues 276–1061), because this Ig-like β-barrel chain has been shown to be monomeric, contains few binding sites for FA or actin-associated proteins, and to behave as a ∼30-nm linear elastic chain with a very high unfolding threshold of >60 pN (41, 42). This segment was inserted in lieu of the deletions present in the minitalins (Fig. 5A and Fig. S4D), and the resulting chimeric-talin constructs were named XT58, XT48, XT46, XT41, XT36, and XT29, in relation to their minitalin counterparts. As shown in Fig. 5 A–D, the chimeric-talin analogs expressed properly, localized to FAs, and supported FA formation in HUVECs in lieu of endogenous talin.

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, probed with anti-GFP antibody. (C) Epifluorescence micrographs of HUVECs cotransfected with hTalin1 siRNA (KD) or control siRNA (siCON), and mEmerald-tagged full-length talin (T100), chimeric-talins (XT29-XT58), or GFP (control). GFP channel (green) indicates talin or chimeric-talins localization to FAs. (D) Magnified view. Efficient knockdown of endogenous talin is verified by hTalin1-specific antibody (huTLN, cyan). FAs and the actin cytoskeletal organization are visualized 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): first and third quartiles, median and confidence intervals; whiskers, 5th and 95th percentiles. Numbers indicated in B, C, and E: median z_FA (red), number of FA ROIs (black), and number of cells (blue). (Scale bars: 25 μm in C, 12.5 μm in D, and 5 μm in E.)

Geometric Analysis of Integrin–Talin–Actin Contact Angle and Talin Extension.

We next mapped the nanoscale organization of the chimeric-talins in FAs (Fig. 5 E and F and Fig. S7). As predicted, their C termini were observed at elevated z-positions relative to the corresponding minitalins. Furthermore, these also scale linearly with construct lengths, with a slope virtually identical to that of the minitalins (P = 0.888; SI Materials and Methods). This suggests that the chimeric-talins are oriented in a polarized fashion, and with a similar contact angle (θ_Talin) relative to the plasma membrane. Because the IgFLNa should behave as a freely jointed-chain segment, the lengths of the chimeric-talins are expected to be ∼30 nm longer than their minitalin counterparts (SI Discussion). Thus, the vertical offset of z = 7.4 nm between the chimeric-talin and minitalin regression lines can be used to calculate the talin contact angle, yielding θ_Talin ∼15°. Sensitivity analysis suggests that this estimate is robust to within a few degrees given experimental uncertainties (SI Discussion).

Subsequently, the length of talin can be calculated, yielding ∼97 nm for full-length talin (T100) based on the N- and C-termini z-displacement of ∼25 nm (Fig. 4). Interestingly, the length of talin/L432G corresponds to ∼81 nm (21 nm/sin 15°), in good agreement with the recent estimate of the end-to-end length of talin in which all domains are folded (35). We surmise that the increased length of ∼16 nm observed in cells expressing wild-type talin could be due to their greater contractility relative to talin/L432G cells, which may lead to a partial unfolding of rod domains. These unfolded domains are likely the R2 or R3 four-helix bundles because they possess the lowest unfolding threshold (∼5 pN) (38), although further experiments would be necessary to confirm this.

Modulation of FA Architecture.

The coincidence of the talin C terminus with the onset of F-actin (Fig. 1 B and C) suggests that talin may physically define the z-position of the FA/stress fiber interface, and thus FA architecture. Therefore, we next investigated how minitalins affect the nanoscale organization of other FA components. We used iPALM to examine the organization of F-actin (Figs. 1C and 6A). In T100 cells, the sideview iPALM images (Figs. 1C and 6A) revealed that the FA/stress fiber interface resides at z >70 nm. In contrast, in HUVECs expressing minitalins, we observed the downshift of such interfaces by ∼10–20 nm compared with T100 (Fig. 6A). Owing to the variable shapes and relatively large sizes of stress fibers, we used (i) VASP, an actin-binding protein associated with F-actin barbed ends, to report on the FA/stress-fiber interface z-position (43) and (ii) F-actin, mapped using Alexa Fluor 568-phalloidin, to report the stress fiber core z-position. As expected, we observed downshifting of the z-positions of VASP FP-fusion coexpressed in the minitalin substituted cells (Figs. 6 B and C and 7A), although we were unable to locate VASP-containing T29 cells, suggesting that the interaction necessary for VASP localization may be absent in this minitalin. Likewise, the stress fiber core z-positions also exhibit a downshifting trend (Fig. 7A and Fig. S7H). To further characterize the relationship between talin length and FA nanostructure, we performed two-color nanoscale mapping experiments measuring the z-positions of F-actin and the talin C terminus. We observed a positive correlation between the talin C-terminal and F-actin z-positions at the pixel-to-pixel level (Fig. S7 C–E). Altogether, these results are consistent with talin physical length serving to regulate the FA vertical dimension.

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 size 1 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) Box and whisker plot of FA/stress fiber interface as marked by VASP. Notched boxes, 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 z_FA (red), number of FA ROIs (black), and number of cells (blue).

Fig. 7.

Molecular geometry of the integrin–talin–actin module. (A) Relationship between talin sequence lengths and the z-positions of talin analogs, or FA components. Median of z_FA distribution (symbols), SD (error bands), and linear regression (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 in FAs. Talin-membrane contact angle (θ_Talin) is calculated from ΔZ_FA, the z-offset between chimeric-talins and minitalins regression lines, and the ∼30-nm length of the IgFLNa1-8 domain, yielding θ_Talin ∼15 °. (C) Actin stress fiber geometry at FA interface. Measurements based on side view iPALM image of actin (Fig. 1C). FA/actin contact angles (θ_SF) ranges from 2° to 6°, for the distal and proximal FA edges, respectively (cell edge on left). Color scale, 0–150 nm relative to substrate surface. Histograms (bin size, 1 nm). (Scale bar: 250 nm.) (D) Geometry and force-balance relationship of the integrin–talin–actin module. Molecule sizes and orientation are approximately to scale. Talin dimer is omitted and some vectors are rescaled for clarity. Vectors indicate the direction of cellular forces, assuming mechanical equilibrium.

We note, however, that the slopes of the stress fiber core and FA/stress-fiber interface regression lines are relatively shallow compared with the talin C-termini regression lines (Fig. 7A), suggesting that the structural effect of talin length on FA architecture may be partially counteracted by other factors. Talin contains about 11 potential vinculin binding sequences (VBSs) with a cluster of five VBSs in the R1–R3 domains, the remainder being distributed throughout the rod (24). We determined vinculin z-positions by mapping vinculin-mCherry (N-terminal tag) coexpressed together with talin1 siRNA and GFP-tagged talin analogs. Vinculin in minitalin cells was located at lower z-positions relative to T100 (Fig. S7 G and H). We observed that vinculin in T58 cells was at an unusually lower z-position compared with the rest, implying that the VBS sites near the N terminus (likely the clusters of five VBSs in R1–R3 domains, Fig. S4A) may be activated to a greater extent. However, owing to the multiplicity of vinculin binding partners in FAs, vinculin nanoscale organization is highly complex, as shown in a recent study describing a phospho-paxillin-dependent positioning mechanism (44). Minitalin interactions with vinculin are perturbed by the deletion of various VBS (Fig. S4). Thus, a further study characterizing the properties of each VBS will be needed to fully account for the observed vinculin positioning effects. Interestingly, with T58 considered as an outlier, the vinculin z-position regression line exhibits a slope similar to that of the stress-fiber core (Fig. S7H), suggesting that although vinculin may not be required for FA nanoscale architecture (Fig. 2), when present it may play a role in the integration of FAs with the actin cytoskeleton (44).

We next further examined the physical geometry of force transmission in FAs by analyzing the iPALM sideview projection of F-actin (Fig. 7C). We observed that the stress fibers approach FAs at a very shallow angle: The contact angle of F-actin at FA/stress fiber interfaces (θ_SF) seems to range between 2° and 6° for the distal and proximal end of FAs, respectively. The shallow θ_SF angle compared with θ_Talin (∼15°) indicates that contractility conveyed by the stress fiber is not colinearly transmitted to talin. Such an angular mismatch (Fig. 7D) implies that an additional counterbalancing force must be present to maintain force balance. With the assumption that force generation primarily occurs via stress-fiber-associated myosin II contractility, 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 depicted in Fig. 7D.

SI Discussion

Molecular Extension of the Recombinant Talin Analogs.

The observed linear correlation between the molecular weights and the C-termini z-positions of talin, minitalins, and chimeric-talin analogs implies that they adopt an extended linear configuration—the end-to-end length of a polymer chain is expected to scale linearly with its mass (i.e., sequence length) for a linearly extended configuration, and with the power of 1/3 for a compact 3D-globular fold configuration. We note that there are minor deviations in the plot of C-terminal z-positions vs. sequence length; for example, the z-position of T41 is slightly higher than that of T46. These are expected based on the structure-based models for the minitalins, as depicted in Fig. S4. In T46, two out of four rod domains are four-helix bundles, which have a small end-to-end length owing to the juxtaposition of their N and C termini on the same end of the helical bundle (Fig. S4 A–C). In contrast, all three rod domains in T41 are five-helical bundles in which the N and C termini are at opposing ends (Fig. S4C), that is, ∼5 nm apart, favoring a more extended conformation. As a result, the end-to-end length of T41 is expected to be slightly longer than T46. Interestingly, for all three series of talin analogs we observed a similar “bump” at T46–T41 (Figs. 4 and 7A), consistent with the structure-based models (Fig. S4B). These deviations are relatively small and do not appreciably affect linear regression analysis. Nevertheless, this indicates that our measurements are likely sensitive enough to reflect a small conformational difference.

We next considered how the information on talin nanoscale organization can be extracted from the experimental data. Because the talin rod consists of compact globular domains joined by flexible linkers, we first consider the simplest physical model of a freely jointed chain model under tension F. Here, the end-to-end length X as a function of force F (in piconewtons) for a chain of N freely jointed rigid rods, each with a length L (in nanometers; total contour length, NL), is given by

$$X\left(F\right)=\hspace{0.17em}NL(\coth \left(\frac{FL}{k_{B}T}\right)-\hspace{0.17em}\frac{k_{B}T}{FL}\hspace{0.17em}),$$ [S1]

where coth is the hyperbolic cotangent function and k_BT is the thermal energy (Boltzmann’s constant and absolute temperature). We then considered how the end-to-end lengths of talin and minitalin vary with force, and whether a linear correlation can be expected for the end-to-end length and the sequence lengths. Talin (T100) is approximated by n = 14, L ∼5.7 nm, giving NL = 80 nm. A hypothetical minitalin T50 is approximated by n = 7, L ∼5.7 nm, and NL = 40 nm. The thermal energy, k_BT, is equal to 4.23 pN·nm at 37 °C. The freely jointed chain force-extension curves (F vs. X) are shown in Fig. S8A. It can be seen that across a wide range of F, the end-to-end lengths scale linearly with the contour lengths L₀. Furthermore, the extension X asymptotically converges to NL rapidly at X(F) >85% of NL at F = 5 pN, suggesting that the polymer assume a nearly linearized configuration even under low physiological force level.

Fig. S8.

Geometry-dependent talin-mediated force transmission model. (A) End-to-end length vs. force for a freely jointed polymer chain with contour lengths L₀ of 40 nm (green) and 80 nm (blue), with a subunit length of 5.7 nm to approximate talin and mini-Talin sizes. (B) End-to-end length vs. force for a semiflexible polymer according to the worm-like chain (Marko–Siggia) with contour lengths L₀ of 40 nm (green) and 80 nm (blue), and persistence length of 5 nm. Red line denotes the ratio between the extension for the 80-nm- vs. the 40-nm-long chains as a function of force. (C) Graphical representation of talin/actin balanced tension vectors in a case of talin/actin angular mismatch. Tensions depicted are actin stress fiber tension (blue), intramolecular Talin tension (green), in-plane tension at the talin–actin interface (parallel tension, magenta). Tension vectors at integrin–talin interface are omitted for clarity. Molecule size is approximately to scale. (D) Geometric relationship of three-vector force balance for C. All force vectors can be calculated from θ_Talin = 15°, θ_SF = 5°, and talin tension of 10 pN, using sine’s rule. The majority of tension is transmitted by the parallel component (magenta). (E) Graphical representation of talin/actin balanced tension vectors in a case of talin/actin near colinearity. Because θ_SF is constrained to small angle for cells on 2D substrate, this requires a significant lengthening of talin. (F) Geometric relationship of three-vector force balance for E, with θ_Talin nearly equal to θ_SF. Tension magnitudes are calculated with the assumption of similar actomyosin tension as in C and D. The majority of tension is transmitted through talin to integrin. The high intramolecular tension likely promotes talin lengthening.

We also consider the worm-like chain model in which talin and the minitalins are approximated by semiflexible polymers. Here the helical bundle length (∼5 nm) is used to approximate the persistence length P. The worm-like chain force-extension curve can be calculated by the Marko–Siggia equation (63):

$$\frac{FP}{k_{B}T}=\hspace{0.17em}\frac{1}{4}{(1-\frac{X}{L_0})}^{-2}-\frac{1}{4}+\frac{X}{L_0}.$$ [S2]

The force-extension curves for T100 and T50 (L₀ = 40 and 80 nm) are shown in Fig. S8B. It can be seen that the end-to-end length X asymptotically approach L₀ somewhat more slowly than the freely jointed chain case, owing to the entropic contribution of polymer flexibility. Nevertheless, across a wide range of F, the end-to-end lengths still scale linearly with the contour lengths L₀ (red line, Fig. S8B), as can be seen to be equal to 2 for T100 vs. T50. Altogether, these suggest that a linear correlation between the end-to-end length and the molecular weight is expected for a polymer under tension under both the ideal polymer (freely jointed chain) and the semiflexible worm-like chain models. Furthermore, this relationship holds for virtually all ranges of force, assuming that the polymers have a similar persistence length. In the context of our experiments, this indicates that talin and its analogs can be well-described as polymer chains held under tension.

Estimates of Talin Contact Angle from Z-Position Measurements.

The determination of talin contact angle by geometric analysis depends on the assumptions that talin and its analogs are oriented at a similar contact angle, and that the “spacer” IgFLNa segment of the chimeric-talin is extended to ∼30 nm in length. For the first, as described above, the observed linear correlation indicates that talin configuration can be described by a simple polymer chain under tension. In further support of this is the similar slope determined for the three talin analog regression lines. Additionally, the z-intercept of the L432G analogs, which have the shortest apparent lengths, corresponds very well to the z-position of the plasma membrane (z ∼30 nm) as determined earlier (16), further corroborating the validity of the first assumption.

For the second assumption, we note that the tension sustained by talin, F, is not directly measured, and thus the length of the spacer portion is not precisely known. We therefore consider how the estimate of θ_Talin is sensitive to variation in the spacer length X assuming a persistence length P of 5 nm, a contour length L₀ of 30 nm, and a worm-like chain behavior. In the limit of high talin tension, with end-to-end length X ∼ L₀, θ_Talin = sin⁻¹(ΔZ = 7.4 nm/X = 30 nm) = 14.3°. In the intermediate regime of moderate tension, X ∼ 0.9L₀, θ_Talin = sin⁻¹(ΔZ = 7.4 nm/X =27 nm) = 15.9°. In the limit of low tension, X ∼ 0.8L₀, θ_Talin = sin⁻¹(ΔZ = 7.4 nm/X = 24 nm) = 18.0°. Thus, our empirical estimate of θ_Talin ∼15° seems to be robust against a wide range of tension experienced by talins and their analogs.

Mechanical Force Balance and Geometric Relationship Between Tension Vectors.

By determining the geometry of talin, we also define how mechanical force is transmitted from the actin stress fiber to integrin. Therefore, we next considered the mechanical implications of the observed talin geometry, and how this may relate to other cellular force components. Because mechanical tension is transmitted via a structurally continuous link, this can be represented by a vector aligned along the force-bearing structure. At the cellular timescale mechanical quasi-equilibrium is assumed, and thus according to Newton’s first law, tension vectors at any given point should sum to zero. According to this premise, myosin II-generated tension within the actin stress fiber that pulls on talin in FAs must be balanced by net opposing vector(s). Our measurements show that actin stress fiber tension is geometrically mismatched with talin (Fig. 7C). Therefore, this requires the participation of an additional force vector with a membrane-parallel component. At the level of the individual integrin–talin–actin module, this could correspond to cross-linking by vinculin. The integration of such membrane-parallel vectors over many molecules could then collectively correspond to tension exerted on the cortical cytoskeleton by the stress fiber at the FA site. That is, the stress fiber contraction force is partitioned between the FA force, which is ultimately transmitted to the ECM as traction, and the membrane-parallel cortical component. Although it may seem surprising initially that a significant amount of tension is not transmitted to the ECM, we note the similarity between the membrane-parallel force and the notion of frictional forces commonly described in FA biophysical models (11, 14, 49). For example, the molecular clutch model of FA is conceptualized in terms of how actin retrograde flow is slowed down by “friction” exerted by FA proteins. Likewise, the motion coupling between FA proteins and retrograde flow is described in terms of frictional slippage between protein “layers” (60). In such dynamic situations, membrane-parallel force at FAs can be considered to relate to a kinetic friction in classic mechanics. By analogy, in our model the membrane-parallel force at FAs can probably be considered as a static friction.

The Newtonian force-balance relationship implies that the molecular geometry of force-transmitting structures determines force magnitude and direction. A balanced system involving three force vectors, such as the FA/stress fiber interface, can be graphically represented by a closed triangle. Here, the contact angles of actin stress fiber and talin determined from our measurements also constrain the other angle. Although we did not directly measure talin tension magnitude, an estimate can be made based on the in vitro force-extension relationship of the talin rod domain, which suggests an intermediate tension of between 5 and 15 pN for wild-type talin (38). Assuming a talin tension of 10 pN, we can then apply the sine rule to calculate the magnitude of tension in the actin stress fiber, yielding ∼30 pN, and the membrane-parallel tension, yielding ∼20 pN (Fig. S8). Note that because talin is a homodimer two scenarios are possible: Either one or both talin molecules are engaged and bear tension. If both talins are engaged, the tension on the actin stress fiber is expected to double to ∼60 pN, because two talin monomers are being pulled. Interestingly, the dimeric nature of talin is evocative of two-legged molecular motors such kinesin or Myosin V. Additional experiments will be necessary to probe the interplay of force and structural arrangement of the talin dimer within cells.

Although the talin rod domains seem to be linearly aligned, the neck region connecting the FERM domain to the rod domain is unstructured and likely flexible, which may permit the integrin–talin–actin module to adopt a wide range of contact angle. Based on the force-balance consideration, such variable geometry may thus permit a wide range of force transmitted to integrin. In other words, talin geometry may differentially regulate how much traction is being transmitted to the ECM. For example, if talin is aligned colinearly with the actin stress fiber, most of the tension would be transmitted to talin, and hence to integrin. This may account for the requirement of large (>50 pN) single-integrin tension, for initial spreading documented in a recent study (52). Furthermore, at such high force talin is expected to be highly stretched, and this could account for both vinculin recruitment during FA maturation and observed talin stretching (39) (Fig. S8). Hypothetically, during early spreading, talin in a rapidly growing adhesion could be aligned colinearly with actin stress fiber, such that large tension is transmitted to integrin (Fig. S8 E and F). Concurrently this would lead to a significant stretching and vinculin recruitment. Vinculin may then cross-link the talin rod to a nearby actin filament, and the differential motion of different actin filaments may result in a perpendicular tension that raises the talin contact angle. Subsequently, vinculin may bear increased tension loads, thereby reducing the tension on talin and integrin. This may signify the steady-state FAs as observed in our experiments. To directly visualize this would require a combined monitoring of tension and talin geometry, which may be possible with a properly designed biosensor. We also note that an analogous geometric situation may also apply at the talin–integrin–ECM interface, which may have interesting implications for traction force transmission.

Functional Modules in Talin Rod Domains.

Because the large size of talin (∼2,500 aa) is highly conserved (64), the ability of our significantly truncated recombinant analogs to support FA assembly is surprising, belying the complexity of talin function and regulation . The robust modularity of talin, however, is consistent with recent studies showing that multiple truncated talin2 isoforms are expressed in diverse tissues (24), and that talin proteolytic fragments function to reinforce cell–cell junction (65). Talin is negatively regulated by autoinhibitory interactions between the FERM and R9 domains (66) and thus talin analogs lacking R9 are expected to be constitutively active. Although our recombinant analogs are designed for probing biophysical parameters, some clues about their cellular functions may be gathered based on the absence or presence of the various domains. For example, the R4–R10 domains seem to be largely dispensable for FA assembly and cell spreading. Furthermore, because T48 (ΔR2–R10), T46 (ΔR4–R12), and T41 (ΔR1–R10) seem to support cell spreading to a significant extent, the contribution of R2–R3 (force-dependent vinculin binding sites) and R11–R12 (second integrin binding site) may be redundant. Because both T29 and T36 cells spread poorly, and these minitalins lack a binding site for RIAM (Rap1-GTP-interacting adaptor molecules), this suggests that talin–RIAM interaction may be necessary for efficient cell spreading (67). We note, however, that although minitalin localization to FAs seems to be robust, we observed a significant decrease in cell area, and a significant global reorganization of the actin cytoskeleton in cells expressing minitalins. Although generally considered to function primarily in FAs, talin has also been shown to localize to the actin cortex and has been implicated in cortical mechanotransduction in Dictyostelium discoideum independent of adhesion (68). Further experiments will be required to dissect the roles of individual domains within talin, the contribution of the physical length of talin to FA dynamics and cell migration, and the differential roles of FA-resident vs. non-FA-resident populations of talin.

We note also that whereas our study focuses on mature FAs whose protein positions can be measured reliably, the nanoscale organization within the smaller nascent adhesions and focal complexes remains unknown. A key challenge in studying early adhesion events is the small copy number of each protein (69), and the fact that several key FA proteins are also associated with the actin cortex (70), which makes it challenging to differentiate between cortex-associated clusters and nascent adhesion either in static superresolution images or from short-trajectory single-molecule data. One avenue to address this is to use dynamic information from long-trajectory 3D single-molecule tracking. However, at present this is limited by the lack of photostable FPs that can be observed on a long timescale compared with FA maturation time on the order of minutes (6, 71).

SI Materials and Methods

Cell Culture.

HUVECs were obtained from Life Technologies (C-015-10C, at passage 2) and cultured 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). Ea.hy926 (immortalized endothelial cells) was obtained from Evelyn Yim, National University of Singapore, and vinculin-null MEF was obtained from Michael Sheetz, Mechanobiology Institute, Singapore. These are cultured in a 5% CO₂, 37 °C humidified atmosphere in DMEM EmbryoMax media (SLM-120-B; Millipore), supplemented with 10% (vol/vol) FBS, sodium pyruvate, and penicillin/streptomycin. Immunofluorescence microscopy was performed to verify the absence of vinculin in vinculin-null MEFs. For HUVECs, cells of passage number <15 were used for experiments. Expression of talin1 and the absence of talin2 in HUVECs and Ea.hy926 were verified by immunofluorescence microscopy, immunoblotting, and RT-PCR (Fig. S1 A–C). For imaging, transfected cells were sparsely plated at a density of 7,000/cm² on fibronectin-coated coverslips/silicon wafer, or glass-bottom dishes (Iwaki) and imaged 16–20 h after replating.

siRNA-mediated knockdown of human Talin1 in HUVECs were performed as previously described (26), using custom stealth RNAi siRNA duplexes (oligo sequence: 5′-CCAAGAACGGAAA CCUGCC AGAGUU-3′; Life Technologies), transfected into HUVECs by electroporation (Neon; Life Technologies). For transfection, trypsinized HUVECs were resuspended at a density of 6 × 10⁶ cells/mL and mixed with siRNA and 0.5–2.0 μg of fusion protein expression vectors where required per each electroporation reaction. The transfection efficiency was ∼99% as determined by flow cytometry using fluorescently labeled siRNAs (Qiagen) (Fig. S1 D and E). Endogenous hTalin1 was depleted >90% by 72 h posttransfection as determined by immunoblot analysis (Fig. S1F). Similar to previous reports (25, 35), talin1-KD cells were able to spread initially during the actin polymerization-driven phase but collapsed afterward due to an inability to form FAs to sustain spreading (Fig. S1I). Talin1-KD cells also showed significantly reduced migration (speed, distance, length, and directionality) (Fig. S1 G, H, and J). Vinculin-null MEFs were transfected by Lipofectamine 2000 (Life Technologies) in Optimem media per the manufacturer’s instructions. For calpain inhibition experiments, the inhibitors (A6060 and A6185; Sigma-Aldrich) were incubated with HUVECs cells at 50 μM as described previously (36). Expression vectors for recombinant talin constructs were purified from Escherichia coli cultures by Endofree Plasmid MaxiPrep Kit (Qiagen) per the manufacturer’s protocol.

Immunofluorescence Microscopy.

Cells grown on no. 1.5 cleaned coverslips or glass-bottom dishes were fixed with 4% (vol/vol) PFA in PHEM buffer (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, and 2 mM MgSO₄, pH 7.0 by KOH) for 10 min at room temperature. After fixation, cells were washed with PBS and incubated for 3 min at room temperature with Hepes-based permeabilization buffer containing 300 mM sucrose and 0.2% Triton X-100, and then blocked for 30 min in a 4% (vol/vol) BSA in PBS. Cells were incubated with primary antibodies in 0.2% BSA in PBS for 1 h at room temperature and washed three times with 0.2% BSA in PBS. Cells were then incubated with secondary antibodies and/or dye-conjugated phalloidin in 0.2% BSA in PBS for 1 h, washed three times with 0.2% BSA in PBS, and then mounted in antifading mounting medium (Dako Technologies) for wide-field epifluorescence microscopy or kept in PBS for total internal reflection fluorescence (TIRF) microscopy. Fluorescence micrographs were acquired using a Nikon Eclipse Ti inverted microscope (Nikon Instruments) with a light emitting diode-based epifluorescence excitation source (SOLA; Lumencor), a motorized TIRF illuminator with fiber-coupled 60-mW 488-nm and 50-mW 561-nm solid-state lasers (Omicron Laserage), an ORCA-flash 4.0 sCMOS camera (Hamamatsu), and a 60× N.A. 1.49 objective lens (Nikon Instruments).

The following primary antibodies were used: rabbit anti-Paxillin (SC-5574; Santa Cruz), rabbit anti-Vinculin (AB73412; Abcam), mouse anti-Talin (clone 8d4) (T3287; Sigma), mouse anti-Talin2 (clone 121A) (AB105457; Abcam), and mouse anti-hTalin1 (clone TA205) (MAB1676; Millipore). Secondary antibodies used were Alexa Fluor 488 Goat anti-mouse IgG (A10680; Life Technologies), Alexa Fluor 488 Goat anti-rabbit IgG (A11008; Life Technologies), Alexa Fluor 568 Goat anti-mouse IgG (A21124; Life Technologies), Alexa Fluor 568 Goat anti-rabbit IgG (A11011; Life Technologies), Alexa Fluor 647 Goat anti-mouse IgG (A21235; Life Technologies), and Alexa Fluor 647 Goat anti-rabbit IgG (A21244; Life Technologies). F-actin was stained by CF405-conjugated Phalloidin (Biotium), Alexa Fluor 488 Phalloidin (A12379; Life Technologies), Alexa Fluor 568 Phalloidin (A12380; Life Technologies), or Alexa Fluor 647 Phalloidin (A22287; Life Technologies).

RT-PCR.

Total cellular RNA were isolated using Illustra RNAspin Mini kit (GE Healthcare) following the manufacturer’s protocol, and cDNA was obtained by retrotranscripting 1 μg of DNase-treated total RNA with Oligo(dT) primers (Life Technologies) and SuperScript III Reverse Transcriptase (Life Technologies). Nonquantitative PCR amplification with Talin1, Talin2, and GAPDH primers was performed in standard conditions at an annealing temperature of 60 °C as previously described (26). As control, nonretrotranscribed RNAs were tested by real-time PCR to ensure that no contaminating DNA was present in the RNA samples. Primer sequences: Talin1 forward, 5′-TCTCCCAAAATGCCAAGAAC-3′; Talin1 reverse, 5′-TGGCTATTGGGGTCAGAGAC-3′; Talin2 forward, 5′-CTGAGGCTCTTTTCACAGCA-3′; Talin2 reverse, 5′-CTCATCTCATCTGCCAAGCA-3′; GAPDH forward, 5′-AGGTCGGTGTGAACGGATTTG-3′; GAPDH reverse, 5′-TGTAGACCATGTAGTTGAGGTCA-3′.

Immunoblot Analysis.

Cells were lysed by radioimmunoprecipitation assay buffer. The lysates were centrifuged at 16,000 × g for 30 min and the pellets were discarded, and Laemmli buffer was added to the soluble lysates. The samples were heated at 95 °C for 5 min and then loaded onto precast 4–15% Tris⋅HCl gradient SDS/PAGE gels (Bio-Rad) and run at 10–15 V/cm. Alternatively, to resolve large-molecular-weight proteins (>200 kDa), precast 3–8% Tris-acetate PAGE gradient gels with Tris-acetate running buffer (NuPAGE Novex; Life Technologies) were used. Proteins from the SDS/PAGE or Tris-acetate PAGE gels were then transferred onto Immobilon-P membranes (Millipore) using a tankblot system (Bio-Rad) in blotting buffer (25 mM Tris, pH 7.5, 0.192 M glycine, and 20% methanol) at 350 mA for 1 h at 4 °C. Membranes were blocked for 1 h at room temperature in 5% (wt/vol) BSA-TBS-T (20 mM Tris, pH 7.5, 137 mM NaCl, and 0.1% Tween-20), washed in TBS-T, and incubated overnight at 4 °C with the primary antibody. After three additional washes in TBS-T, membranes were further incubated for 1 h with the secondary antibody conjugated to HRP and visualized by chemiluminescence, using the Luminata Classico western HRP substrate kit (Millipore). The immunoblot signals were visualized on a ChemiDoc MP imaging system (Bio-Rad). Densitometric quantification and analysis of the protein bands were performed using ImageJ. Antibodies used are as follows: mouse anti-Talin (T3287; Sigma), mouse anti-Talin1 (AB104913; Abcam), mouse anti-Talin2 (AB105457; Abcam), mouse anti-Talin2 (MCA4771; Serotec), mouse anti-GAPDH (AB9484; Abcam), mouse anti-Vinculin (V4139; Sigma), mouse anti-Paxillin (610569; BD Biosciences), and mouse anti-GFP (AB1218; Abcam).

FP Fusion Constructs.

All FP fusion constructs were based upon either N1- or C1- Clontech-style vectors with pCMV promoters. FP fusion constructs of integrin α_v, FAK, paxillin, vinculin, zyxin, VASP, and α-actinin were described previously (16). Vectors encoding wild-type mouse talin1 (National Center for Biotechnology Information ID no. X56123.1) tagged with tdEos (green-to-red photoconvertible fluorescent protein), EGFP, or mEmerald were generously provided by the M.W.D. laboratory. FilaminA construct were provided by the Stossel laboratory, Harvard University, Cambridge, MA. The rod-deletion constructs Δ(R4–R10) and Δ(R1–R10) (T58 and T41, respectively) were generated by N.B. as described previously (45).

The constructs are designed to retain the N-terminal head FERM domain (residues 1–482) and the C-terminal THATCH (R13 domain, residues 2294–2541) (Fig. 3A and Fig. S4), with the deletion or insertion of the rod domains as illustrated in Fig. S4. The recombinant talin constructs are linked with fluorescent proteins at either the N terminus or the C terminus with flexible linkers of 10, 18, or 22 aa: for C-terminal FP fusion of T58, T48, T46, T41, and T36, the 10-residue linker sequence is GSASRSPVAT; for N-terminal FP fusion of T58, T48, T46, T41, T36, and T29, the 18-residue linker sequence is SGLRSRAQASFEFAEAAT; and for C-terminal FP fusion of T29, XT58, XT48, XT46, XT41, XT36, and XT29, the 22-residue linker sequence is GSGSGSGQGSGAGSAS RSPVAT. To facilitate the combination of different talin domains without the need to introduce extraneous amino acid residues (72)_, site-directed mutagenesis using Phusion DNA polymerase (Thermo Scientific) was performed on mEmerald-tagged wild-type mouse talin1 vector to introduce silent mutations that remove 4 BsaI restriction sites (G255A, A621C, C4878T, and C5883G). Recombinant talin with partial deletion or duplication of rod domains can then be generated by ligating two PCR products into mEmerald N1 cloning vector. For example, T48 was created by ligating wild-type mouse talin1 nucleotides (1–1965) and (5920–7623) into mEmerald N1 cloning vector. The PCR primers sequence used are as follows. T48: FERM-R1 domain (nucleotide 1–1965), forward, 5′-AAAAAAGAATTCTAGCC ACC ATGGTTGCGC-3′, reverse, 3′- GGGCCCGGTCTCTTACCAA TTTGCTGCAACAGC TC-5′, restriction sites, EcoRI (5′) and BsaI (3′); R11-R12-THATCH domain (nucleotides 5620–7623), forward-5′- AAAAA AGGTCTCAGGTACCCA GGCCTGCATTAC-3′, reverse, 3′- AAAAAAAC CGG TGATCGGGAC GCGGAGC-5′, restriction sites, BsaI (5′), AgeI (3′). T46: FERM-R1-R2-R3 domain (nucleotides 1–2733), forward, 5′- AAAAAA GAATTCTAGCCACCATGGTTGCGC-3′, reverse, 3′- GGGCCCGGTCTCTTACCA ATTTGCTG CAA CAGC TC-5′, restriction sites, EcoRI (5′) and BsaI (3′);THATCH domain (nucleotides 6880–7623), forward -5′-AAAAAAGGTCTCAGAGTGGGTGGACC CAGAG -3′, reverse, 3′- AAAAAAACCGGTGA TCGG GACG CGGAGC -5′, restriction sites, BsaI (5′), AgeI(3′). T36: FERM-R1 domain (nucleotides 1–1965), forward, 5′- AAAAAA GAATTCTAGCCACC ATGGTTGCGC-3′, reverse, 3′- GGGCCCGGTCTCTTACC AATTT GC TGCAACAGC TC-5′, restriction sites, EcoRI (5′) and BsaI (3′); THATCH domain (nucleotides 6880–7623), forward-5′- AAAAAAGG TCTCA GAGTGGGTGGACCCAGAG -3′, reverse, 3′- AAAAAAACC GGT GATCGGGA CGCGGAGC-5′,restriction sites, BsaI (5′), AgeI (3′). T29: FERM domain (nucleotides 1–1443), forward, 5′- AAAAAA GAATTCTAGCCACCATGGTTGCGC-3′, reverse, 3′- AAAAAAGGTCT CTA CTCG TGCATCTGGCCACTGG TG -5′, restriction sites, EcoRI (5′) and BsaI (3′); THATCH domain (nucleotides 6880–7623), forward-5′- AAAAAAGGTCTCAGAGTGGGTGGA CCCAGAG -3′, reverse, 3′- AAAAAAACC GGTGATCGGGACGCGGAGC -5′, restriction sites, BsaI (5′), AgeI (3′).

The recombinant talin fusions with mEmerald tags (an EGFP variant with mutations designed to enhance folding) at the C terminus were generated first. Most C-terminal fusions were initially created with 10-aa flexible linkers between talin and EGFP or mEmerald. However, the expression level of the initial T29 construct was very low and thus a version with a 22-aa linker was used instead. The tdEos fusion constructs were created with 22-aa linkers, similar to that used in an earlier study (16). To generate the N terminus fusion, the recombinant constructs were subcloned into the C1 EGFP cloning vectors using EcoRI and AgeI restriction sites. To create calpain cleavage-resistant recombinant talin constructs, the site-specific mutation L432G is introduced using the following primers: forward, 5′-CCCAAAAA GTCAACA GTCGGTCAGCAGC AGT ACAACCG-3′; reverse, 5′-CGGTTGTACTGCTGCT GACCGA CTGTTG ACTT TTTGGG-3′. To create the chimeric talin-filaminA construct in which the filaminA IgFLNa 1–8 domain (residues 276–1061) are inserted into the talin rod regions, PCR products of filaminA were ligated into the middle of the talin recombinants using primers listed as follows. XT58: FERM-R1-R2-R3 domain (nucleotides 1–2733), forward, 5′- AAAAAAGAATTCTAG CCACCATGGTTGCGC-3′, reverse, 3′- AAAAAAGGTCTCATGG GCT TCTTGATG GCGTT CTGCG -5′, restriction sites, EcoRI (5′) and BsaI (3′); R11-R12-THATCH domain (nucleotides 5620–7623), forward, 5′- AAAAA AGGTCT CAG GTACCCAGGCCTGCATTAC -3′, reverse, 3′- AAAAAAACCGG TGATCGGGACGCGGAGC -5′, restriction sites, BsaI (5′), AgeI (3′), IgFLNa1-8 domain (FilaminA nucleotides 826–3183), forward, 5′- AAAAAAGGTCT CAC CCA AACTGAACCCGAAGA-3′, reverse, 3′- AAAAAAGGTCTCATACC CACAGCTTCCA GAGGAAAG-5′, restriction sites, BsaI(5′),BsaI (3′). XT48: FERM-R1 domain (nucleotides 1–1965), forward, 5′- AAAAAAG AATTCTAGCCACC ATGGTTGCGC-3′, reverse, 3′- AAAAAAGGTCTCATGGGAATTTGCTG CAACAGCTC -5′, restriction sites, EcoRI (5′) and BsaI (3′); R11-R12-THATCH domain (nucleotides 5620–7623), forward-5′- AAAAAAGGTCTCAGG TACCC A GGCCTGCATTAC -3′, reverse, 3′- AAAAAAACCGGTG ATCGGGACGCGGAGC -5′,restriction sites, BsaI (5′), AgeI (3′), IgFLNa1-8 domain (FilaminA nucleotides 826–3183), forward, 5′-AAAAAAGGTC TCA CCC AAACTGAACCCGAAGA-3′, reverse, 3′-AAAAAAGGTCT CATACCCACAG CTTCCAGA GGA AAG-5′, restriction sites, BsaI (5′),BsaI (3′). XT46: FERM-R1-R2-R3 domain (nucleotides 1–2733), forward, 5′- AAAAAAGAATTCTAGCCACC ATGGTTGCGC-3′, reverse, 3′- AAAAAAGGTCTCATGGGC TTCT TGA TGG CGTTCTGCG -5′, restriction sites, EcoRI (5′) and BsaI (3′); THATCH domain (nucleotide 6880–7623), forward-5′- AAAAAAGGTCTCAGAGTGGG TGGACC CAG AG -3′, reverse, 3′- AAAAAAACCGGTGAT CGGGACGCGGAGC -5′, restriction sites, BsaI (5′), AgeI (3′), IgFLNa1-8 domain (FilaminA nucleotides 826–3183), forward, 5′- AAAAAAGGTCTCACCCAA ACTGAACCCG AAGA-3′, reverse, 3′- AAAAAAGG TCTCATACCC ACA GCTTCCAGAGGAAAG-5′, restriction sites, BsaI (5′),BsaI (3′). XT41: FERM domain (nucleotides 1–1446), forward, 5′- AAAAAAGAATTCTAGCC ACCATGGTTGCGC -3′, reverse, 3′- AAA AAA GGTCTCATGG GTCGG TGCATCTGGCCACTGG -5′, restriction sites, EcoRI (5′) and BsaI (3′); R11-R12-THATCH domain (nucleotides 5920–7623), forward, 5′- AAAAAAGGTCTCAGGTACCCA GGCC TGCATTAC -3′, reverse, 3′- AAAAAAACCGGTG ATCGGGACGCGGAGC -5′, restriction sites, BsaI (5′), AgeI (3′), IgFLNa1-8 domain (FilaminA nucleotides 826–3183), forward, 5′- AAAAAAGGTCTCACCCAAACTGAA CCCGA AGA-3′, reverse, 3′- AAAAAAGGTCTCATACCCACAGCTTCCAGAGGAAAG -5′, restriction sites, BsaI (5′), BsaI (3′). XT36: FERM-R1 domain (nucleotides 1–1965), forward, 5′- AAAAAAGAATTCTAGCC ACCATGGT TGCGC -3′, reverse, 3′- AAAAAAGGTCTCATGGGAATTTGCTGCAACAGCTC -5′, restriction sites, EcoRI (5′) and BsaI (3′); THATCH domain (nucleotides 6880–7623), forward-5′- AAAAAAGGTC TCAGAGTGG GTGGAC CCAGAG -3′, reverse, 3′- AAAAAAACCGGTGATCG GGACG CGGAGC -5′, restriction sites, BsaI (5′), AgeI (3′), IgFLNa1-8 domain (FilaminA nucleotides 826–3183), forward, 5′- AAAAAA GGTCTCACCCAAACTGAA CCCG AAGA-3′, reverse, 3′- AAAAAAGGTC TCATA CCCACAGCT TCC AGAGGAAAG-5′, restriction sites, BsaI (5′),BsaI (3′). XT29: FERM domain (nucleotides 1–1443), forward, 5′- AAAAAAGAATTCTAGCCACCATGGTTGCGC-3′, reverse, 3′- AAAA AAGGTCTCATGGG GTG CATCTGGCCACTGGTG -5′, restriction sites, EcoRI (5′) and BsaI (3′); THATCH domain (nucleotides 6880–7623), forward, 5′- AAAAAAGGTCTCAGAGTGG GTGGACCC AGAG -3′, reverse, 3′- AAAAAAA CCG GTGATCGGGACGCGGAGC -5′, restriction sites, BsaI (5′), AgeI(3′), IgFLNa1-8 domain (FilaminA nucleotides 826–3183), forward, 5′- AAAAAAGGTCTCACC CAAACTGAACCCGAAGA-3′, reverse, 3′- AAAAAAGGTCTCATACCCACAGCTTCCAGAGGAAAG-5′, restriction sites, BsaI (5′), BsaI (3′).

All constructs were verified by sequencing. Localization to the FAs was verified by fluorescence microscopy. Proper molecular weights of the constructs were verified by Immunoblotting (Figs. 4B and 5B). Graphical diagrams of constructs were generated by DOG2.0 software (73).

Topographic Z Measurement by Surface-Generated Structured Illumination.

The imaging substrates for VIA-FLIC/SAIM were p-type (100)-orientation silicon wafers with ∼500 nm of thermal SiO₂ layer (Bondatek; Addison Engineering) and were prepared as previously described (28, 29). Briefly, the precise thickness of the oxide layer on each wafer was determined by a UV-visible variable angle spectroscopic ellipsometer (UV-VIS-VASE; J.A. Woollam) at A* IMRE, Singapore. Nine equally spaced spots on a 4-inch wafer were measured with a range of 300 to 800 nm in 5-nm steps, and the incident angle was 55°. The variations in oxide thickness across the wafers were typically less than 1.5 nm. The wafers were cut into 1.2 × 1.2 cm² by scoring them with a diamond-tip pen and subsequently cleaned in acetone and 1 M KOH for 20 min each. Wafers were then chemically activated with 0.5% (3-aminopropyl) trimethoxy-silane and 0.5% glutaraldehyde in PBS for 1 h each. Finally, the wafers were UV-sterilized for 20 min and coated with 10 μg/mL bovine plasma fibronectin (Sigma) for 40 min at 37 °C or overnight at 4 °C. After cell plating, samples were fixed in 4% (vol/vol) PFA for 12–15 min and then washed with PBS for 5 min. For two-color experiments, samples were then incubated with Alexa Fluor 568 Phalloidin in 0.2% BSA in PBS for 1 h and washed three times with 0.2% BSA in PBS.

Raw images acquisition was performed on a Nikon Eclipse Ti inverted microscope described above, using a motorized TIRF illuminator and either 488-nm or 561-nm laser excitation, fed through a polarization-maintaining optical fiber. The objective lens was 60× N.A. 1.49, which yielded a pixel size of 108.33 nm with the Orca-Flash 4.0 sCMOS camera (Hamamatsu). One of the ND filters in the TIRF illuminator is replaced by a linear polarizer (Thorlabs) to ensure that the excitation is s-polarized. The incidence angle is calibrated against the illuminator position by first determining the center position of the laser light (X_c). Thereafter, the motor position (X) to achieve a desired angle of incidence in aqueous samples were calculated according to Eqs. S3 and S4 below:

$$\sin {\theta }_{water}=\frac{n_{glass}}{n_{water}}sin{\theta }_{glass}$$ [S3]

$$\sin {\theta }_{glass}=\frac{2\cdot (X-X_c)\cdot M}{2\times {10}^5{n}_{glass}},$$ [S4]

where M denotes the objective magnification, n_glass the refractive index of glass, and n_water the refractive index of water. The tabulated motor positions (X) for the angle 0° to +56° were encoded into an acquisition script in Nikon NIS-Elements software to automate the acquisition.

For imaging, the silicon wafers were placed with the sample (and thermal oxide) side facing the glass surface of a 27-mm glass-bottom dish (Iwaki) filled with PBS (Fig. S3C), with a small thumb screw placed on the silicon wafer to maintain neutral buoyancy. A series of 14–28 images per cell and per fluorescence channel were obtained from 0° to +56° with a typical camera exposure times on the order of 100–200 ms. For two-color experiments, measurements were performed sequentially, using excitation at 488 nm to image GFP (or variants) and 561 nm to image AlexaFluor 568.

Analysis was performed using a custom-written software developed in Interactive Data Language (IDL; Exelis VIS), based in part on the MATLAB source code kindly provided by C. Ajo-Franklin, Lawrence Berkeley National Laboratory, Berkeley, CA (28). Binary masks for ROIs to analyze were defined either by simple thresholding or by Otsu thresholding using background-subtracted images. Topographic height (z) and other fit parameters were determined by fitting the theoretical angle-dependence curve to experimental angle-dependence intensity by the Levenberg–Marquadt nonlinear least square method, for each pixel in the ROI. To ensure exhaustive global optimum is attained, parameter space was searched at 10 initial z guess positions, placed at 15-nm intervals. The best-fit z-value determined for each pixel was then offset by the thickness of the ECM layer (Fig. S3E) to allow direct comparison with the iPALM coordinate system. Topographic Z values were displayed using color to encode z-position. Measurements were validated using adsorbed fluorescent nanobeads (100, 50, and 20 nm). For FAs, the median of the Z values for a given ROI (Z_median) is found to best represent to characteristic height of the structure. Alternatively, the histogram of the Z values can be used to visualize the Z distributions within the aggregated pixels, giving similar values to Z_FA analysis (Fig. S6 D–F). For two-color analysis, the scatter graph of z-position for each pixels of GFP (Talin-C) and AlexaFluor 568 (actin) were plotted, and the linear regression calculated using the software OriginPro.

Three-Dimensional Superresolution Microscopy by iPALM.

The iPALM 3D superresolution microscope system (Fig. S2A) was constructed as described previously (62), and controlled by custom-written LabVIEW software (National Instruments). Image processing and data analysis were performed using PeakSelector, a custom-written IDL-based software, developed by the Hess group, HHMI Janelia Farm. For imaging, cells were cultured on a no. 1.5 25-mm coverglass preembedded with metallic nanoparticles as fiducials for calibration and drift correction (Hestzig, LLC). For imaging of tdEos, the excitation is from a 200-mW 561-nm laser (Coherent) free-space coupled into the system. For actin imaging using Alexa Fluor 647, the laser excitation is from a 100-mW 642-nm laser (Coherent). Samples for F-Actin were prepared similarly to the sample for immunofluorescence microscopy described above, except for a higher concentration of phalloidin (1:20–1:40), and the use of imaging buffer containing 100 mM cysteamine in PHEM buffer. Photoactivation is provided by a 100-mW 405-nm laser (Coherent). Excitation intensity was controlled by an AOTF (A-A Optoelectronics). Objectives used were dual 60× N.A. 1.49 Apo TIRF (Nikon Instruments), equipped on a custom-machined Invar mounting, equipped with Picomotors piezoelectric actuators. The three-way beamsplitter (Fig. S3B) was custom-manufactured by Rocky Mountain Instruments, mounted in a custom-machined Invar mounting, and coupled with index matching to a broadband dielectric mirror mounted on a z-tip-tilt piezoelectric mount (Physik Instrument). Emission filters used for tdEos imaging were BrightLine FF01-588/21 (Semrock) and BrightLine FF01-676/37 (Semrock). Three electron-multiplying CCD cameras (Ixon Ultra; Andor) were used in frame transfer mode for acquisition of single-molecule raw images. Samples were mounted on a custom-designed sample holder equipped with dual piezoelectric scanners (Physik Instrument).

System alignment and z-coordinate calibration were carried out by optimizing the three-camera calibration curve (Fig. S2 C and D) before each cell was imaged using the piezoelectric sample holder. Between 25,000 and 75,000 image triplets were acquired for cells, with 50 ms per frame exposure time, yielding ∼10⁶ to 10⁷ localizations. Drift correction was performed by registration of the fiducial fluorescence signals. Z coordinates were extracted from the intensity ratio between the three cameras using Newton’s methods as described previously (16). To assess the resolution performance, the spread in lateral (x) and z coordinates fluorescent fiducials were analyzed (Fig. S2 E and F), showing FWHM of 8.97 nm for the x coordinates and 4.00 nm for the z coordinates, indicating a higher resolving power along the z axis due to multiphase interferometry, as described previously (27).

Statistical Analysis.

Nonparametric Mann–Whitney u test, plotting, and linear regression analysis were carried out in OriginPro or Microsoft Excel. Plots were generated with error bar representing the SD of z_FA. Student’s t test for the slopes of the regression line (Fig. 7A) were carried out by first testing using the F-test with the null hypothesis that each dataset has the same variance (74, 75). The null hypothesis cannot be rejected for all paired sets (P = 0.823 for minitalin vs. minitalin- L432G; P = 0.134 for minitalin vs. chimeric-talin), thus their variances are not statistically different, allowing the SE of the combined slope b to be calculated by

$$s_{b,combined}^2=\frac{\left(n_1-2\right)s_{b1}^2+(n_2-2)s_{b2}^2}{(n_1+n_2-4)},$$ [S5]

where n₁ and n₂ are the numbers in populations 1 and 2, and s_b1 and s_b2 are the SE of regression fits for slope b₁ and b₂, respectively. The t test statistics can be calculated by

$$t_{combined}=\frac{(b_1-b_{2)}}{s_{b,combined\sqrt{1/n_1\hspace{0.17em}+\hspace{0.17em}1/n_2}}}$$ [S6]

and compared with Student’s t distribution using the degree of freedom n₁ + n₂ − 4. It was found that the null hypothesis that the three regression lines have the same slope cannot be rejected (P = 0.404 for minitalin vs. minitalin-L432G; P = 0.888 for minitalin vs. chimeric-talin). For two-color z-position scatterplot, Spearman rank-correlation coefficient (r) was calculated by OriginPro. The Spearman correlation t score is calculated by

$$t=r\sqrt{\frac{n-2}{1-r^2}}$$ [S7]

and compared with Student’s t distribution using the degree of freedom n − 2 (74).

Discussion

In this study we sought to establish the molecular basis of FA organization and how FA components are geometrically organized into force-transmission modules. We combined superresolution microscopy with a molecular engineering approach such that information on molecular conformation, connectivity, and geometry could be gleaned from the measurements. Our results support the hypothesis that talin serves a physical role in regulating FA architecture, effectively setting the scale for the FA core, akin to a molecular ruler. The robust modularity of talin enables the geometric organization of the integrin–talin–actin module to be determined with high precision. This also provides empirical constraints suggesting that stress-fiber-borne actomyosin contractility and FA-borne tension may be mechanically 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 implications for both the force-mediated extension of FA proteins and how FA transmits force. First, we observed small but consistent and statistically significant elevation of the C-terminal z-positions of T100 and minitalin constructs containing the L432G calpain-resistant mutation compared with constructs lacking this mutation (Figs. 4 and 7A). Similar results were obtained with calpain inhibitors (Fig. S7 A and B). Based on our result of θ_Talin ∼15°, the ∼81-nm length obtained for T100/L432G agrees remarkably well with the expected length of talin based on NMR and X-ray crystallographic studies of individual domains (∼80 nm), suggesting that talin/L432G is oriented in a polarized manner but with all domains compact, likely due to relatively low (<5 pN) tension. The T100/L432G mutation and calpain inhibitors have previously been shown to inhibit cellular contractility (37). In contrast, wild-type talin is ∼16 nm longer. This could be due to the force-induced unzipping of the R2–R3 four-helical bundles, which has been shown in vitro to occur at ∼5–7 pN; further experiments using R2–R3 stabilizing mutations (45) may help to verify this. Because calpain-mediated proteolysis of talin is required for FA turnover and the up-regulation of promigratory signaling pathways such as Src, the increased length of wild-type talin likely reflects a relatively higher talin tension resulting from greater contractility (36, 37, 46, 47). In vitro force-extension experiments on talin rod domains 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 at that level of force.

The modulation of FA architecture by talin analogs implicates the integrin–talin–actin module as the primary force-transmitting unit, and thus the geometrical orientation of this module is important for how FA transmits force to the ECM. MyosinII-generated tension conveyed through the stress fiber must be balanced by the equivalent amount of net tension (Newton’s first law: All force vectors should sum to zero in an equilibrium). However, because we observed that talin does not align colinearly with the stress fiber, talin tension can only partially account for stress-fiber tension. This implicates the presence, at the talin–actin interface, of an additional tension vector with a significant membrane-parallel component (Fig. 7D). Because vinculin bridges the talin rod to actin and bears tension, we surmise that vinculin cross-linking may account for such counterbalancing force at the level of individual integrin–talin–actin modules (48). Collectively, the integration of such counterforces across numerous integrin–talin–actin modules in a given FA may then give rise to a membrane-parallel cortical tension. That is, the stress fiber may exert force on both the FA and the surrounding cortical cytoskeleton—the latter could be considered analogous to friction, a common parameter 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 geometry suggests that talin is pulling on the integrin β cytodomain at 15°. Previous superresolution microscopy measurements indicated that the ECM–plasma membrane separation is 20–30 nm, comparable to the ∼20-nm crystallographic length of activated integrin ectodomain (16, 27, 50). This suggests that the angle between the integrin 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, another membrane-parallel force vector, possibly the apparent membrane tension, is likely accounting for the force balance. In other words, talin may exert a pulling force on both the integrin and the surrounding plasma membrane. Consistent with this, a number of studies have documented how membrane tension may influence integrin–ECM interaction (51, 52). An interesting possibility is that such in-plane tensions may influence other FAs in the vicinity, thus providing a degree of local mechanical coordination. We suggest this could be a possible molecular-mechanical basis for how FA force transmission to ECM may be interdependent on the integration of multiple 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 experiments do not directly measure tension, the range of various forces can be estimated based on the measured geometry and in vitro talin force/extension data (38). For example, given θ_SF of 5°, and θ_Talin of 15°, a 5- to 15-pN range of talin tension would correspond 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 to talin-mediated force, our measurements would correspond to a single-integrin tension of ≤15 pN. Interestingly, this also implies that although a considerable tension is present in the stress fiber, it is significantly shunted away from the traction component and toward the membrane-parallel component by the talin–actin angle mismatch. Because talin contains a long flexible linker region (residues 401–481) that could act as a swiveling joint, a wide range of talin contact angles is likely possible during the course of FA maturation. We propose that variable talin geometry may differentially channel the actomyosin contraction force between the cortex and the ECM, and thus may help account for such a broad range of observed tensions (Fig. S8 and SI Discussion). For example, if talin is nearly colinear with the actin stress fiber, that is, θ_Talin ∼ θ_SF, the majority of the actomyosin contraction is expected to be channeled through talin. In particular, this may correspond to early cell spreading when large integrin tensions >50 pN have been implicated (52). Such high tension would also lead to talin becoming significantly extended, consistent with the observation of 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 FAs has recently been documented by high-resolution 3D traction force microscopy (57). Together with recent studies suggesting that FA traction force is spatially and temporally heterogeneous across a given FA structure, this would imply that talin geometry may be variable across FAs as well (54, 58). Current limitations to our methods are that the z-position measurements correspond to ensemble-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 heterogeneity may not be apparent. Whereas recent developments of cell-matrix tension biosensors have largely emphasized magnitude measurement, our analysis suggests that talin geometry would be an important quantity to probe simultaneously with tension, especially during different FA morphodynamic stages. Indeed, in an equilibrium of three force vectors, the measurement of two angles and one force magnitude will allow the calculation of all other force vectors by trigonometric identity (SI Discussion). Because the stress-fiber geometry should be relatively easy to estimate for cells on 2D substrate, these can simplify to one angle and one tension measurement. A number of strategies may be possible for the design of a single-molecule talin geometry reporter, such as a geometry-dependent FRET sensor (59) or by single-molecule localization of a pair of spectrally distinct FPs incorporated at well-defined sites within talin. Direct and simultaneous measurements of force and force-transmission structures in living cells will be instrumental in understanding how complex mechanosensitive responses originate in FAs.

Our study highlights the dominant structural role of talin in FAs, which contrasts with the dispensability of vinculin in specifying FA nanoscale organization. Nevertheless, vinculin has been shown to bear tension, playing important and complex roles in FA strengthening, stabilization, and actin retrograde flow engagement (44, 48, 60, 61). Altogether this suggests a structural-mechanical model in which the integrin–talin–actin module serves as the constitutive 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 be structurally sufficient for force transmission, cross-linking interactions mediated by vinculin and/or other proteins could help modulate the geometry of force transmission and thus regulate how cell-generated force is apportioned toward or away from ECM traction. At the nanoscale, regulation of mechanical cross-linking between adjacent integrin–talin–actin modules may also introduce 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 FA structural organization and cell mechanical functions and highlight the roles of molecular geometry, which should be measurable and testable in further live-cell studies.

Materials and Methods

Instrumentation.

Three-dimensional superresolution microscopy of F-actin was performed on a custom-built iPALM instrument, using a 561-nm laser for the excitation of tdEosFP, or a 642-nm laser for the excitation of AlexaFluor 647 phalloidin. Transfected cells were located by low-intensity excitation from a 488-nm laser. A 405-nm laser was used for photoconversion of tdEos or photoswitching of AlexaFluor 647. The theoretical foundation, operational principles, construction, and alignment of iPALM have been described previously (62). Topographic z mapping of protein position by surface-generated structured illumination techniques was performed on a Nikon Eclipse Ti inverted microscope (Nikon Instruments) equipped with a motorized TIRF illuminator, a 60× N.A. 1.49 Apo TIRF objective, a sCMOS camera (Orca Flash 4.0; Hamamatsu), and a fiber-coupled laser combiner (488 nm and 561 nm; Omicron Laserage). Fluorophore positions were measured using either a 488-nm laser (for mEmerald or EGFP) or a 561-nm laser for mApple, mCherry, or photo-converted mEos2. When needed, photoconversion of mEos2 was carried out using light-emitting diode excitation (SOLA; Lumencor) using the DAPI excitation filter set. A detailed description of the calibration, image reconstruction, and data analysis for both techniques is given in SI Materials and Methods.

Cell Culture and Sample Preparation.

HUVECs (passage number <15) were cultured 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 Talin1 in HUVECs was performed as previously described (26), using custom stealth RNAi siRNA duplexes (oligo sequence: 5′-CCAAGAACGGAAACCUGCCAGAGUU-3′; Life Technologies). To transfect cells with expression vectors and/or siRNA, cells were trypsinized and electroporated using the Neon platform (Life Technologies). For imaging samples, cells were replated at a relatively sparse density of 7,000 per square centimeter on fibronectin-coated fiducialed coverglass, silicon wafers, or glass-bottom dishes (Iwaki) and imaged 16–20 h after replating. Expression vectors for FA protein fusions were described previously (16, 27). Expression vectors for the recombinant talin analogs were created as detailed in SI Materials and Methods.