Polymerizing Actin Fibers Position Integrins Primed to Probe for Adhesion Sites

See allHide authors and affiliations

Science  16 Feb 2007:
Vol. 315, Issue 5814, pp. 992-995
DOI: 10.1126/science.1137904


Migrating cells extend protrusions, probing the surrounding matrix in search of permissive sites to form adhesions. We found that actin fibers polymerizing along the leading edge directed local protrusions and drove synchronous sideways movement of β1 integrin adhesion receptors. These movements lead to the clustering and positioning of conformationally activated, but unligated, β1 integrins along the leading edge of fibroblast lamellae and growth cone filopodia. Thus, rapid actin-based movement of primed integrins along the leading edge suggests a “sticky fingers” mechanism to probe for new adhesion sites and to direct migration.

The first steps of mammalian cell migration are extension of the leading edge and attachment to the surrounding extracellular matrix (1). Extension is driven by actin polymerization (2), with the position (3), speed, and persistence (4) of the leading edge regulated by actin-binding proteins. However, extension is not uniform; adjacent regions of many broad, flat protrusions such as fibroblast lamellae and neuronal growth cones advance and retract independently, and which region moves forward is continually redefined by transverse movement (5, 6). Observations on protrusion dynamics led to the hypothesis that these protrusion variations are used to search for permissive sites to form new adhesions (7, 8). However, no known direct relation between protrusion dynamics and the ability of adhesion receptors to probe the matrix has been identified.

We studied protrusion dynamics during the initial matrix exploration. Using high-contrast differential interference contrast (DIC) optics, we observed small ripple protrusions traversing the outermost edge of fibroblast lamellipodia, the leading 1 to 2 μm of the lamella (Fig. 1, A and B). The ripples did not move sideways at the same speed (0.10 ± 0.06 μm/s, n = 34) that actin network treadmilling (0.028 ± 0.013 μm/s, n = 51) (9) moved microspikes and filopodia precursors laterally (Fig. 1D) (10, 11). Instead, the ripples moved at the same speed that actin polymerization elongates growth cone filopodia (12) and rockets Listeria bacteria (13). Although the frequency of ripples decreased in response to the actin-depolymerizing agent cytochalasin D (Fig. 1C), the associated actin dynamics are difficult to visualize with green fluorescent protein (GFP)–actin because of the density and rapid turnover (9). We therefore exploited the properties of photoactivatable GFP (PaGFP) (14) to establish the threshold of visible actin and to restrict visualization to ∼40% of the expressed PaGFP-actin (15). Limiting activation revealed localized increases in actin intensity that moved sideways with the ripples (Fig. 1D and movie S1). The increases did not represent thickness variations or membrane lifting, because they persisted after normalization by the cotransfected monomeric red fluorescent protein (mRFP) signal, and the ripples remained within a single DIC optical section (15). The intensity increases are local actin polymerization at crests of ripples traversing the leading edge (82% colocalization, n = 198) (Fig. 1D and movie S1) (15).

Fig. 1.

Local increases in actin polymerization occur at small ripples traversing the leading edge of migratory fibroblasts. (A) DIC and PaGFP-actin images showing NIH3T3 fibroblast lamella (l) and lamellipodium (lp). Region of cell is shown in inset, and black box indicates region in (D). Scale bar, 5 μm. (B) Histogram of the ripple duration or wave period (n = 106, 21 cells) (15). Kymographs from DIC images show the duration of the transverse ripples (white bracket) and local forward protrusions (black bracket). Horizontal scale bar, 20 s; vertical, 0.5 μm. (C) Cytochalasin D (CD) decreases the frequency of the ripples. (D) At the tips of ripples, there are localized increases in PaGFP-actin fluorescent intensity (red arrowheads). The ripples move sideways faster than either the lateral movement of filopodia (vertical yellow arrowheads) or the rearward movement of actin (horizontal yellow arrowheads) due to lamella network treadmilling. Dashed lines mark initial positions of yellow arrowheads. Scale bar, 5 μm.

Restricting the region of photoactivation revealed that activated PaGFP-actin localized at the tip of a nascent ripple and accumulated along the elongating ripple (Fig. 2A). In contrast to this asymmetric accumulation of PaGFP-actin, photoactivation of PaGFP alone produced a symmetric signal that spread throughout the cell and was consistent with simple diffusion (fig. S1). We also found that the vasodilator-stimulated phosphoprotein (VASP), which protects barbed ends from capping (4), localized to actin elongating across the leading edge (Fig. 2B and movie S2). The elongating VASP-tipped actin reshaped and redirected the larger regions of forward protrusion (movie S2). Thus, it seems that the transverse ripples represent local sites of preferential actin elongation along the edge of advancing protrusions.

Fig. 2.

β1 Integrins move synchronously with actin polymerizing at the crests of ripples. (A) Activated PaGFP-actin (fluorescence) initially localizes to the tip of a nascent ripple (DIC) and accumulates along the entire ripple as it elongates. Arrows slanted toward direction of ripple movement. Scale bar, 1 μm. (B) VASP, which enhances actin polymerization (4), is localized at the tips of actin fibers elongating along the leading edge. Scale bar, 5 μm. (C) Histogram of ripple speed (n = 34, 27 cells). (D) A sequence from movie S3 shows enhanced green fluorescent protein (EGFP)–actin and β1 integrin moving synchronously with the tip of an elongating ripple (arrows). Scale bar, 1 μm. (E) The first image of movie S3 shows EGFP-actin and β1 integrins have corresponding increases in intensity. The yellow contour indicates the single-pixel-wide line used to obtain the intensity traces shown in (F). Scale bar, 5 μm. (F) Fluorescent intensity of EGFP-actin (red) and β1 integrin (green) for image in (E); arrowheads correspond to those in (E). (G) Cross-correlation analysis shows a strong correlation between the intensity traces of EGFP-actin and β1 integrin at the edge and essentially no correlation behind the edge. (H) A high correlation between EGFP-actin and β1 integrins is maintained along a 20 μm length of the leading edge throughout the ∼3-min monitoring period (correlation coefficient 0.53 ± 0.13 at the leading edge, n = 54, and 0.01 ± 0.14 behind the lamellipodia in a region devoid of adhesive contacts, n = 54). (A to C) are NIH3T3 cells and (D to H) are human foreskin fibroblasts.

To test if the localized increases in actin polymerization are involved in matrix probing, we compared them with the location of integrin adhesion receptors. Using a fluorescent, non-perturbing antibody, we observed clustering of β1 integrins that was temporally and spatially synchronized with actin polymerization at the crests of ripples (Fig. 2, D to F, and movie S2). The β1 integrins and actin remained colocalized during an ∼3-min monitoring period (correlation coefficient 0.50 ± 0.12, n = 217, five cells). In contrast, β1 integrins did not colocalize with actin behind the leading edge (correlation coefficient 0.01 ± 0.16, n = 140, five cells) (Fig. 2, G and H, and movie S3) (15). Similar correlations were obtained for positive and negative controls, respectively (fig. S3). Thus, β1 integrins move synchronously with polymerizing actin along the leading edge.

To test whether these β1 integrins were capable of binding extracellular matrix, we examined their conformational state and found activated β1 integrins colocalized to the tips of actin along the leading edge (correlation coefficient 0.63 ± 0.10, n = 14) (Fig. 3, A and B, figs. S2 to S5). To visualize activation and to determine where the transition to a high-affinity state occurred, we developed a live-cell assay using a fluorescent antibody targeting high-affinity β1 integrins (9EG7). We initially labeled all of the activated β1 integrins on the cell surface so that any subsequent labeling would reveal newly activated β1 integrins. These newly labeled integrins appeared only at the leading edge (Fig. 3C and movie S4).

Fig. 3.

Activated but unligated β1 integrins are located at the tips of actin fibers at the leading edge. (A) Conformationally activated β1 integrins are spatially concentrated at the tips of polymerizing actin at the leading edge of human foreskin fibroblasts. Scale bar, 5 μm. (B) Fluorescent intensity of F-actin (red), activated β1 integrin (blue), and total (active + inactive) β1 integrin (green) along the yellow line in (A). In this example, the correlation coefficient is 0.69 between total β1 integrin and F-actin, and 0.75 between activated β1 integrin and F-actin. The average correlation coefficients for multiple cells are shown in fig. S3 and Fig. 2D (n = 14). Total and active integrin distributions are compared in fig. S2D. Scale bar, 5 μm. (C) Sequential fluorescent (top) and DIC (bottom) images showing the activation of β1 integrins on NIH3T3 cells. Different color arrowheads differentiate individual activation-specific antibody binding sites at the leading edge that subsequently move rearward. Vertical scale bar, 2 μm. (D) Boxplot shows activated β1 integrins do not cross-correlate with endogenous fibronectin (n = 8) but do correlate with added FN120kD (n = 17) and can be blocked by inhibitory antibodies (n = 8) (P < 0.001).

The activated β1 integrins were not ligated to fibronectin but could bind fibronectin fragments (FN120kD), specifically, with the α5β1 integrin heterodimer (Fig. 3D and fig. S2), which was consistent with an increased avidity for fibronectin at the leading edge (16, 17). Although unligated high-affinity αvβ3 integrins, which can bind vitronectin or fibronectin, transiently localize to the edge of spreading cells (18), we did not find β3 integrin localization (fig. S3) or increased vitronectin avidity at the leading edge of migrating cells (17, 19). Moreover, the observed spatial correlation between high-affinity β1 integrins and actin was not restricted to a single cell type, antibody, or extracellular matrix (Fig. 4 and fig. S3). Activated, but unligated, fibronectin adhesion receptors at the leading edge may thus have a direct role in sampling the extracellular matrix during migration.

Fig. 4.

Localization of activated β1 integrins at the leading edge depends on availability of polymerizing actin ends. (A) NIH3T3 cells stained for activated β1 integrins (green) and F-actin (red and grayscale) show a decrease in activated β1 integrin density when actin is capped (cytochalasin D, n = 19) and an increase when actin is stabilized to increase the number of free ends (jasplakinolide, n = 22). Inhibiting myosin II ATPase (blebbistatin, n = 18) did not change the number of actin ends or the integrin concentration. Scale bar, 5 μm. Boxplot of activated β1 integrin density normalized to control (n = 27). (B) Activated β1 integrins are also localized to the tips of polymerizing growth cone filopodia (NG108 cells). Scale bar, 5 μm.

We tested the functional relation between actin polymerization and activated β1 integrins by altering actin dynamics (Fig. 4A and fig. S4). Decreasing the number of polymerizing actin ends by capping with cytochalasin D decreased the density of activated β1 integrins, while increasing the number of polymerizing ends with the stabilizer jasplakinolide (9) increased the activated β1 integrin density at the cell periphery. Moreover, increasing actin turnover without changing the number of barbed ends, with the myosin II adenosine triphosphatase (ATPase) inhibitor blebbistatin, (15, 20) did not change the activated β1 integrin density. Consistent changes in activated β1 integrin density were observed with siRNA (small interfering RNA) knockdown of the cofilin phosphatase slingshot and capping protein β2 (fig. S4). Despite the changes in density, the correlation between activated β1 integrins and polymerizing actin remained high [correlation coefficient 0.49 to 0.54 with no statistical difference (P > 0.05) between cytochalasin D (n = 23), jasplakinolide (n = 27), blebbistatin (n = 34), and control (n = 25)] (15). The changes in activated β1 integrin density thus correspond to the number of polymerizing actin ends.

To further test whether β1 integrins interact with polymerizing actin, we demembranated cells with Triton X-100 to remove the cell membrane and membrane-associated proteins but leave behind the cytoskeleton and cytoskeletally associated proteins (15). Activated β1 integrins quantitatively colocalized with the tips of polymerizing actin at the leading edge in the demembranated cells (fig. S5A). The leading edge localization of activated and total β1 integrins was decreased by siRNA knockdown of talin 1, a direct physical linker between actin and β1 integrins (21) that is associated with tips of actin fibers along the leading edge (22). Additionally, the localization was not changed by knockdown of α-actinin 1, another direct linker (23) that is not similarly concentrated along the leading edge (fig. S5C). These data suggest that talin contributes to the linkage between β1 integrins and polymerizing actin at the leading edge.

We also found activated β1 integrins localized at the tips of growth cone filopodia (Fig. 4B). Growth cones guide migration by sampling the matrix for ligands and relative stiffness (24), and their actin-based filopodia are necessary for correct steering in vivo (25). The localization of activated β1 integrins at filopodia tips suggests a role in growth cone path-finding. Additionally, we tested whether the activated adhesion receptors at the cell perimeter regulated migration force to match matrix stiffness by using an optical trap assay (17, 24, 26). We found connections formed between fibronectin and the adhesion receptors at ripple protrusions were more likely to regulate migration force (85%, n = 13) than connections formed at other regions along the leading edge (25%, n = 12). This suggests that activated β1 integrins are localized at the ends of actin fibers to probe for ligands, to guide migration direction, and to regulate migration force.

Polymerization of actin fibers directs the movement of a variety of protrusions ranging from rocketing Listeria (27) to elongating filopodia (12). Our data indicate that preferential localized polymerization and elongation of actin filaments or fibers along the leading edge steer local forward protrusions. The polymerizing actin moved and clustered β1 integrins to spatially position integrins primed to probe the matrix at the very front of cell protrusions, creating “sticky fingers” along the leading edge. Transverse variations in actin polymerization along the advancing perimeter allow the primed integrin receptors to sample more of the extracellular matrix than simple forward and backward movement. This might allow cells with well-developed adhesive contacts to make small changes in migration direction without severing and reestablishing their matrix connections under the cell body. Thus the extremely dynamic local variations in actin polymerization probe the matrix for adhesion sites by positioning primed adhesion receptors along the advancing cell perimeter during migration.

Supporting Online Material

Materials and Methods

Figs. S1 to S5


Movies S1 to S4

References and Notes

View Abstract

Navigate This Article