Cryo-ET of actin cytoskeleton and membrane structure in lamellipodia formation using optogenetics

Summary

Lamellipodia are sheet-like protrusions essential for cell migration and endocytosis, but their ultrastructural dynamics remain poorly understood because conventional electron microscopy lacks temporal resolution. Here, we combined optogenetics with cryo-electron tomography (cryo-ET) to visualize the actin cytoskeleton and membrane structures during lamellipodia formation with temporal precision. Using photoactivatable-Rac1 (PA-Rac1) in COS-7 cells, we induced lamellipodia formation with a 2-min blue light irradiation, rapidly vitrified samples, and analyzed their ultrastructure with cryo-ET. We obtained 16 tomograms of lamellipodia at different degrees of extension from three cells. These revealed small protrusions with unbundled actin filaments, “mini filopodia” composed of short, bundled actin filaments at the leading edge, and actin bundles running nearly parallel to the leading edge within inner regions of lamellipodia, suggesting dynamic reorganizations of the actin cytoskeleton. This approach provides a powerful framework for elucidating the ultrastructural dynamics of cellular processes with precise temporal control.

Graphical abstract

Highlights

• •Integrating optogenetics with cryo-ET to visualize cell dynamics at specific times
• •Cryo-ET reveals actin and membrane ultrastructure during lamellipodia formation
• •Mini filopodia act as precursors driving lamellipodia membrane protrusion
• •Actin bundles align parallel to the leading edge during lamellipodia extension

Cell biology; Organizational aspects of cell biology

Introduction

Lamellipodia, thin, sheet-like cellular protrusions, are dynamic structures that emerge at the leading edge of migrating cells.^1,2 The actin cytoskeleton plays a fundamental role in cell shaping and motility,³ which is also essential for lamellipodia formation and retraction. The lamellipodia-like actin network contributes to cellular motility, including ameboid movement and endocytosis.^4,5 These processes are essential for the immune system,^6,7 early development,⁸ and cancer metastasis and invasion.⁹

The small GTPase Rac1 localizes at the periphery of the plasma membrane and serves as a major upstream regulator of lamellipodia formation.^10,11 Rac1 activates the SCAR/WAVE complex,¹² which in turn triggers the activation of the Arp2/3 complex.¹³ The Arp2/3 complex, an actin nucleation factor, binds to the sides of existing actin filaments, leading to the growth of new actin filaments at an angle of approximately 70°.¹⁴ This actin branching is believed to be a principal feature of lamellipodia. In addition to branched actin networks, lamellipodia also contain rib-like structures: microspikes, which do not protrude from the leading edge, and filopodia, which extend beyond it (Figure 1A). These structures consist of straight, bundled actin filaments, polymerized by the formin/mDia family and crosslinked by fascin.¹⁵

Figure 1.

PA-Rac1-induced lamellipodia formation

(A) Schematic representation of subdomains in lamellipodia.

(B) COS-7 cells expressing PA-Rac1 and Lifeact-mCherry on glass-bottom dishes. Images were captured every 10 s, with PA-Rac1 activated by a 458 nm laser from time 0 at 10-s intervals. The fluorescence images of Lifeact-mCherry are shown.

(C) Lifeact-mCherry image of COS-7 cells, highlighting the crescent-shaped cell edges in red.

(D–F) COS-7 cells expressing PA-Rac1 and Lifeact-mCherry on EM grids. Images were captured every 10 s from −5 min, with PA-Rac1 activated by scanning with a single 458 nm laser within 0–10 min at 10-s intervals.

(D and E) Representative time-lapse images. A merged image of DIC and fluorescence of Lifeact-mCherry (yellow) (D), and fluorescence images of Lifeact-mCherry (E) are shown. A yellow arrowhead indicates dotted actin structures within the interior of cells.

(F) Quantification of cell area changes. Data are presented as the means ± SD from 6 cells. Blue background indicates the activation period; red dotted line marks the timing of freezing. Scale bar = 20 μm. See also Figures S1–S4 and Videos S1 and S2.

The dynamics and structure of lamellipodia have been extensively studied over the past half-century using both light and electron microscopy (LM and EM). Super-resolution LM, total internal reflection fluorescence (TIRF) microscopy, and speckle microscopy have provided insights into protein localization and actin dynamics within lamellipodia.^{16,17,18,19,20,21} However, the dense actin network in lamellipodia makes it challenging to resolve single filament structures using LM. In contrast, EM has been instrumental in revealing the ultrastructure of lamellipodia.^22,23,24 Nonetheless, the chemical fixation and negative staining in conventional EM potentially introduce artifacts.

The recent development of cryo-electron tomography (cryo-ET) has overcome these limitations. Cryo-ET reconstructs three-dimensional structures through processing reconstruction of a series of tilted cryo-EM images. With advanced image processing, cryo-ET enables the visualization of native structures of proteins and protein complexes in fresh frozen, unstained cells in subnanometer resolution.^25,26,27 Given that lamellipodia are thin cellular structures (∼100–200 nm), they are ideal targets for cryo-ET. However, observing lamellipodia with intact cell membranes using cryo-ET remains challenging due to sample thickness and ice constraints. Consequently, most cryo-ET studies on lamellipodia have been conducted with chemically fixed and permeabilized cells.^28,29,30 Moreover, most studies have focused on highly motile cells, such as fish epidermal keratocytes, observing the structure of already established lamellipodia, which overlooks the early stage of lamellipodia formation. Recently, Chung et al. visualized lamellipodia with intact cell membranes in spreading cells by using galectin-8 for the extracellular matrix (ECM) to facilitate protrusion extension thinly enough for cryo-ET analysis.³¹ Their approach allowed the determination of the polarity of actin filaments and the high-resolution structure of the Arp2/3 complex within the plasma membrane. However, the protrusions formed during cell spreading may not fully mimic lamellipodia formation.

Here, we employed the optogenetic tool PA-Rac1 (photoactivatable-Rac1)³² to induce lamellipodia formation in COS-7 cells and examined their ultrastructure using cryo-ET. A key advantage of optogenetics is its ability to precisely control the timing of intracellular signaling activation. This feature is particularly crucial for studying lamellipodia formation, as conventional fixation methods fail to capture dynamic processes. By irradiating cells with blue light inside an automatic plunge freezer chamber, we allowed precise control over Rac1 activation timing, followed by rapid vitrification. Using cryogenic correlative light and electron microscopy (cryo-CLEM), we identified newly formed lamellipodia and examined their ultrastructure by cryo-ET. This approach enabled us to visualize the architecture of the actin cytoskeleton and plasma membrane during lamellipodia formation, providing additional perspectives on lamellipodia organization and the mechanisms driving cell motility.

Results

Induction of lamellipodia formation on EM grids using optogenetics

To study the ultrastructure of the actin cytoskeleton and membrane dynamics during lamellipodia formation, precise control of vitrification timing after stimulation is critical. In this study, we employed the optogenetic tool PA-Rac1, which enables temporal regulation of Rac1 activation via light irradiation, allowing controlled induction of lamellipodia formation.^11,32 COS-7 cells were selected due to their strong adhesion, efficient transfection, and ability to form well-defined lamellipodia on EM grids.

Initially, PA-Rac1-induced lamellipodia formation in COS-7 cells was characterized on glass-bottom dishes using confocal laser scanning microscopy (CLSM, Figures 1B and Video S1). Lifeact-mCherry was used as an F-actin marker.³³ Even before light activation, cells spontaneously formed lamellipodia (Figure 1B; 0:00). Upon activation, more prominent protrusions emerged from all around the cells, clearly distinguishing PA-Rac1-induced lamellipodia formation (Figure 1B; 2:30). Significant lamellipodia formation was observed along the cell periphery, primarily in regions where the cortex formed inwardly curved arcs over 20 μm in length, reminiscent of a crescent shape (Figure 1C). These “crescent-shaped edges” were characterized by abundant cortical actin bundles as indicated by the intense fluorescence of Lifeact-mCherry. In line with the hypothesis that lamellipodia formation is initiated by the Arp2/3 complex-mediated side branching from existing actin filaments oriented parallel to the plasmalemma,²⁸ these regions are reasonable sites of vigorous lamellipodia formation. The initial extension of lamellipodia was typically completed within 2 min of blue light irradiation, followed by cycles of retraction and re-extension. The extent of lamellipodia expansion varied among cells and within different regions of the same cell, with some extending over 5 μm. After lamellipodia formation, actin retrograde flow was observed (Video S1). Microspikes and filopodia were also observed within lamellipodia (Figure 1B). Interestingly, these structures exhibited lateral movement (Figure S1A), and their tips occasionally detached from the leading edge and gradually collapsed. Some of them were eventually incorporated into pre-existing cortical actin bundles (Figure S1B). These findings indicate that lateral flow²² also occurs in PA-Rac1-induced lamellipodia. The fluorescence intensity of Lifeact-mCherry was highest in the anterior region of the lamellipodia, comparable to cortical actin bundles, but significantly decreased in the central regions.

To investigate the characteristics of lamellipodia formed before and after light activation, we analyzed the localization of several actin-binding proteins (Figures S2–S4). The SCAR/WAVE complex subunit Abi1 was enriched at the tip of the lamellipodia (Figure S2), while Arp2 was distributed throughout the lamellipodia (Figure S3). In contrast, myosin light chain (MLC) was strongly associated with cortical actin bundles (Figure S4), consistent with previous studies.^23,34,35 Notably, these proteins exhibited similar localization patterns before and after light activation, suggesting that PA-Rac1-induced lamellipodia closely resemble endogenously formed lamellipodia in protein localization.

Next, lamellipodia formation with PA-Rac1 was induced on EM grids. Since lamellipodia formation and mobility are largely influenced by the ECM, optimizing the substrate for cryo-ET observations was essential. To ensure that lamellipodia spread thinly and remained close to the grid surface, EM grids were coated with poly-_L-lysine and laminin. Upon blue-light irradiation, lamellipodia formed around the cells, primarily emerging from the crescent-shaped cell edges, resembling those observed on glass-bottom dishes (Figures 1D and 1E, and Video S2). Actin retrograde flow, microspikes and filopodia formation, as well as their lateral movements and collapse were also observed (Figures S1C and S1D). These similarities suggest that lamellipodia formation on EM grids is comparable to that on glass-bottom dishes, at least based on CLSM observations.

A key characteristic of PA-Rac1-induced lamellipodia was their formation along the entire cell perimeter, distinguishing them from spontaneously formed lamellipodia. Another distinct feature was the appearance of punctate actin structures within the cell interior, approximately 1 μm in diameter (Figure 1E, yellow arrowhead), although their specific nature remains unclear. These two features served as reliable markers for confidently identifying PA-Rac1-induced lamellipodia after plunge freezing.

To determine the optimal timing for vitrification, we analyzed cell area changes following PA-Rac1 activation (Figure 1F). Within the first 4 min of blue-light irradiation, the cell area expanded significantly. Although lamellipodia persisted beyond this period, certain regions exhibited cycles of retraction and re-expansion (Video S2). Based on these observations, plunge freezing was performed 2 min after the start of light irradiation to capture nascent lamellipodia.

Acquisition of cryo-ET for PA-Rac1-induced lamellipodia

To analyze the ultrastructure of PA-Rac1-induced lamellipodia with an intact membrane, COS-7 cells expressing mVenus-PA-Rac1 and Lifeact-mCherry were cultured on the EM grids and subjected to plunge freezing immediately after 2 min of blue light irradiation (Figure 2). Cells with PA-Rac1-induced lamellipodia were selected based on their characteristic morphology (i.e., broad perimeter formation and punctate actin structures) using Leica Cryo-CLEM light microscopy. Subsequently, these grids were subjected to cryo-EM, and tilt series were collected while correlating the data with cryo-LM images. Since the plasma membrane tended to deform at the holes of the Quantifoil carbon film, likely due to the lack of substrate and mechanical stress during the blotting process (Figures S5A and S5B),^36,37 we avoided analyzing these areas to minimize potential artifacts and instead focused on regions located on the carbon film. While the signal-to-noise (S/N) ratio was lower on the Quantifoil support film, the contrast remained sufficient for segmenting cellular structures (Figures S5C and S5D). After several attempts at grid preparation and data collection, we successfully prepared two well-vitrified grids and measured them using cryo-EM. From these grids, 16 tomograms were obtained from three cells, covering regions from the leading edge to the rear of the lamellipodia at different degrees of extension (Figure 3). The white-boxed regions in Figure 3 indicate areas where tilt-series were acquired. Based on the tomograms from these regions, the cell membrane, actin filaments, and microtubules (MTs) were segmented, and the resulting images are presented in Figures S6–S8. The actin filaments are color-coded to indicate their orientation relative to the leading edge. Regions outlined with purple dashed squares in Figures S6–S8 were defined as lamellipodia in this study. The selection criteria for these regions included the following: They were located away from cortical actin bundles parallel to the plasma membrane and excluded grid hole areas to prevent artifacts.

Figure 2.

Experimental workflow from sample preparation to tomogram acquisition

Laminin-coated EM grids were placed on the 3D-printed grid holders in 4-well dishes, and cells were seeded on the EM grids. At the same time, plasmids encoded mVenus-PA-Rac1 and Lifeact-mCherry were transfected into the cells. The next day, the cells were rapidly frozen using a plunge freezer after inducing lamellipodia formation by activating PA-Rac1 with blue light irradiation, controlling the timing of the freezing. The frozen cells were observed using cryo-fluorescence microscopy to identify the cells that had formed lamellipodia. By correlating the fluorescence images with low-magnification EM images, the regions for tomography were determined. Subsequently, continuous tilt series images were acquired by cryo-EM from −70° to +70°, and tomograms were reconstructed.

Figure 3.

Cryo-CLEM images

Cryo-CLEM images of three cells from which tomograms were captured in this study. Green: mVenus-PA-Rac1, purple: Lifeact-mCherry. See also Figures S5–S8.

To compare PA-Rac1-induced lamellipodia with those described in previous studies, we quantified their overall structural characteristics (Figures 4 and S9). The average thickness of lamellipodia from three cells ranged from 50 to 250 nm (Figure 4A), and the filament density varied between 2,500 and 15,000 per μm³ (Figure 4B), consistent with previous studies.^31,38 The relatively low actin filament density in Cell 2 might be attributed to pronounced lamellipodia spreading.

Figure 4.

The organization of the actin cytoskeleton in PA-Rac1-induced lamellipodia

(A and B) The averaged thickness (A) and the filament density (B) in lamellipodia across individual cryo-ET.

(C) The orientation of the actin filaments relative to the leading edge (L.E.) in lamellipodia (D) Orientation distributions of actin filaments by length: 0–200 nm (left), 200–400 nm (center), >400 nm (right).

(E and F) Actin branches within lamellipodia. (E) Representative images of branched actin filaments. Scale bar = 25 nm

(F) The frequency of actin branches in lamellipodia across individual cryo-ET. See also Figure S9.

Across all analyzed lamellipodia, filaments oriented perpendicularly to the leading edge were relatively rare. The most frequent filament orientations were around 20°–30° and approximately 90° relative to the leading edge (Figure 4C). When analyzed separately for individual cells, filament orientation varied among cells (Figure S9A). The orientation of actin filaments in all tomogram regions, both within and outside the lamellipodia, is depicted in Figure S9B. We also explored the correlation between actin filament length and its orientation to the leading edge (Figure 4D). For filaments shorter than 200 nm, the peak orientation was between 15° and 25°. In contrast, longer filaments tended to align more parallel to the leading edge.

The Arp2/3 complex plays a crucial role in organizing the actin cytoskeleton in lamellipodia by mediating actin branch formation.^20,23,30 To investigate this process, we identified actin branches by meticulously reviewing tomograms in the 3dmod software, using the segmented actin filaments as references. The criteria for identifying branches included their location at filament ends, a branching angle of approximately 70°, and electron-dense features at the branch base, indicative of the Arp2/3 complex localization. Identified actin branches are represented as black dots in the segmentation images, with representative examples shown in Figure 4E. The average frequency of actin branching was one branch per 6.2 μm of filament length within lamellipodia (Figure 4F). In contrast, regions outside lamellipodia exhibited an average of one branch per 31.7 μm of filament length, indicating that actin branching occurs approximately five times more frequently within lamellipodia. These results underscore the essential role of actin branches in lamellipodia formation.

The architecture of the actin cytoskeleton and protrusive membrane structure of growing lamellipodia

Next, we analyzed the details of the actin cytoskeleton in lamellipodia at various degrees of extension. First, we examined a lamellipodium with a relatively low degree of extension, measuring approximately 0.5 μm (Figures 5A–5C and Video S3). The actin bundle on the left side of the tomogram is most likely the original cortical actin. Another example of a similarly extended lamellipodium is shown in Figure S8A. At the leading edge, the plasma membrane appeared slightly undulated. To systematically classify surface features, we first defined the geometric parameters of protrusions. The width (W) was measured as the distance between the two endpoints at their base, while the height (H) was defined as the perpendicular distance from this baseline to the protrusion apex (Figure 5D). Protrusions with a height of at least 20 nm were then manually identified, and their H/W ratio was calculated. Based on these measurements, we classified the protrusions into “Large protrusions” (H/W > 0.5) and “Minor protrusions” (0 < H/W ≤ 0.5). This threshold was chosen for practical purposes, as it provided a clear distinction based on the observed distribution of H/W values. The remaining areas, where H/W < 0, were classified as “Concave regions” (Figure 5D). In Minor protrusions, several actin filaments extended toward the leading edge (Figures 5E and 5F). Most of these actin filaments, located just beneath the protrusive plasma membrane, had their growing ends oriented toward the leading edge. In these Minor protrusions, the actin cytoskeleton was not well-organized, lacking cross-linked actin filaments, and slightly curved filaments were observed. Inside the lamellipodium, bundled actin filaments were not detected.

Figure 5.

Cryo-ET of PA-Rac1-induced lamellipodia in the low degree of extension

(A) A cryo-CLEM image of a PA-Rac1-induced lamellipodium in Cell 1, identical to the image shown in Figure 3. The lamellipodium extended approximately 0.5 μm. The yellow rectangles in the upper panels indicate the regions enlarged in the lower panels. Scale bars = 10 μm (upper) and 2 μm (lower), respectively.

(B and C) A cross-sectional (x-y) slice from cryo-ET (B) and a corresponding segmented image (C), derived from the area marked by a white box in A, corresponding to position 1 of Cell 1 in Figure 3. The plasma membrane is represented in cyan. Actin filaments are color-coded to indicate their orientation relative to the leading edge (L.E.) of the cell. Here, 0° represents perpendicular orientation relative to the leading edge, while 90° indicates parallel orientation. Branching points of actin filaments are marked with black dots. Scale bar = 200 nm. See also Video S3.

(D) Conceptual illustration of the measurement of membrane protrusions using their height (“H”) and base width (“W”). Schematic representations of a “Large protrusion” (W/H > 0.5) and a “Minor protrusion” (0 < W/H ≤ 0.5) are provided to explain the classification visually.

(E and F) Enlarged images of the protrusive structure at the leading edge are shown in the black dotted box in B and C. Scale bar = 50 nm.

Video S3. Cryo-ET of PA-Rac1-induced lamellipodia in the early stage of formation, related to Figure 5

Red: actin, Cyan: membrane, Yellow: actin branches. Scale bar = 100 nm.

Next, we studied a lamellipodium extended to approximately 1.5 μm (Figures 6A–6C and Video S4). In this case, two Large protrusions were identified at the leading edge. In one of these protrusions (Figure 6D), actin filaments accumulated near the plasma membrane, forming two clusters. These two clusters consisted of actin filaments extending along different edges of the protrusion, following the membrane contour. Their arrangement likely guided the protrusion along its paths. Another Large protrusion (Figure 6E) exhibited an arrangement where two or more straight actin filaments ran parallel, oriented toward the leading edge, resembling the organization seen in microspikes and filopodia, albeit on a smaller scale. The spacing between these parallel actin filaments, approximately 8 nm (Figure 6F), suggests fascin-mediated cross-linking, characteristic of microspike and filopodia formation.^39,40 Although fascin cross-linking was not directly visualized in this study, its involvement is inferred based on the filament spacing and previous studies. Due to the limited number of filaments and their relatively shorter length (Figure 6C), we refer to these structures as “mini filopodia”. These structures were observed in two tomograms (Figures 6C and S8B).

Figure 6.

Cryo-ET of PA-Rac1-induced lamellipodia extended to approximately 1.5 μm

(A) A cryo-CLEM image of a PA-Rac1-induced lamellipodium in Cell 1, identical to the image shown in Figure 3. The yellow rectangles in the upper panels indicate the regions enlarged in the lower panels. Scale bars = 10 μm (upper) and 2 μm (lower), respectively.

(B and C) A cross-sectional (x-y) slice from cryo-ET (B) and a corresponding segmented image (C) obtained from the area marked by a white box in A, corresponding to position 4 of Cell 1 in Figure 3. Red asterisks denote regions where the plasma membrane shows high intensity. Scale bar = 200 nm. See also Video S4.

(D and E) Enlarged images of the protrusive structure at the leading edge are shown in the black dotted box in B and C. Yellow dots mark branched points of actin filaments; red arrows point to locations where actin filaments attach to the plasma membrane. Scale bar = 50 nm

(F) Enlarged images of the actin filaments running parallel in E, with three filaments indicated by red arrows running parallel to each other. Scale bar = 10 nm

(G) Visualization of filaments with proximity-based pairing. Filaments within the lamellipodium are displayed based on their proximity and orientation. Red filaments indicate those that are part of a pair meeting the following criteria: a distance of ≤25 nm, an angular deviation of ≤7.5°, and at least 50% of their filament length within these thresholds. Black filaments represent those that do not meet these criteria. Scale bar = 200 nm.

Video S4. Cryo-ET of PA-Rac1-induced lamellipodia in the middle stage of formation, related to Figure 6B

Scale bar = 200 nm.

Notably, within these Large protrusions, localized high-density areas on the plasma membrane, which appeared to be darker regions in tomograms, were observed (Figure 6B, red asterisks). These dark regions likely indicate the presence of protein complexes or specialized membrane domains, offering a promising direction for further exploration.

Meanwhile, in the Minor protrusion at the leading edge (below box D in Figure 6C), several actin filaments extended toward the leading edge without forming bundles, similar to those in Figure 5. In concave regions, individual filaments extended separately toward the leading edge, while a few filaments extended parallel to the leading edge beneath the plasma membrane.

When examining the interior of the lamellipodium (Figure 6C), long actin filaments oriented parallel to the leading edge were prominent. Additionally, a few actin filaments, color-coded in blue at the lower left of the tomogram, appeared detached from the leading edge but exhibited an ordered structure resembling mini filopodia. For other actin filaments, the similar orientation of neighboring filaments (Figure 6G) suggests the formation of an organized actin network in this degree of extended lamellipodia.

To further investigate the relationship between actin filaments and plasma membrane dynamics, we measured the distance between the growing ends of actin filaments and the plasma membrane across the three identified groups, focusing on actin filaments oriented toward the leading edge (Figures 7A–7C). Actin filaments adhering to the plasma membrane (red arrows in Figures 6D, 6E, 7B, and 7C) were most frequently observed in Concave regions, accounting for approximately half of the cases (Figure 7D). In contrast, in both Large and Minor protrusions, only around 15% of filaments adhered to the membrane, with a notable number of filaments exhibiting gaps between their growing ends and the plasma membrane. These gaps predominantly ranged from 10 to 20 nm, although no consistent spacing pattern was observed (Figure 7D).

Figure 7.

The architecture of the plasma membrane at the leading edge and the associated actin network

(A–C) Gallery views of each protrusion type: “Large protrusion” (A), “Minor protrusion” (B), and “Concave regions” (C). Yellow dots mark branched points of actin filaments; red arrows point to locations where actin filaments attach to the plasma membrane; an orange asterisk denotes regions where the plasma membrane shows high intensity. The rightmost set in (C) corresponds to a region adjacent to and partially overlapping with that shown in Figure 5F. Scale bar = 50 nm

(D) Analysis of the distance between the membrane and the ends of actin filaments in the three types of membrane architectures. Only actin filaments oriented toward the leading edge were analyzed. The number of direct interaction points, relative to the total analyzed actin filaments and total number of analyzed membrane structures, are noted below the graph.

Analysis of actin architecture in well-extended lamellipodia

In our tomograms, the lamellipodium in Cell 2 was the most extended, measuring over 3 μm (Figure 8A). In the LM image, a few filopodia or microspikes were observed within the lamellipodia. A portion of a filopodium was observed in a tomogram (Figures 8B and 8C and Video S5), where bundles of over ten filaments extended linearly toward the leading edge. Upon magnification, these bundles were spaced approximately 8 nm apart (Figure 8D). Other actin bundles ran nearly parallel to the leading edge within the inner lamellipodia (e.g., white/black dotted boxes in Figures 8E and 8F and Video S6). These bundles, composed of three to five actin filaments, were also cross-linked at approximately 8 nm (Figures 8G and 8H). These findings suggest that these actin bundles, including filopodia and mini filopodia, are formed by the exact mechanism involving fascin-mediated cross-linking.

Figure 8.

Cryo-ET of PA-Rac1-induced lamellipodia in a high degree of extension

(A) A cryo-CLEM image of a PA-Rac1-induced lamellipodium in Cell 2, identical to the image shown in Figure 3, with an extension of over 3 μm. The yellow rectangles in the left panels indicate the regions enlarged in the right panels. Scale bars = 10 μm (left) and 2 μm (right), respectively.

(B, C, E, and F) Cross-sectional (x-y) slices from cryo-ET (B and E) and corresponding segmented images (C and F) obtained from the area marked by white boxes in A. Scale bar = 200 nm. See also Videos S5 and S6.

(D, G, and H) Enlarged images of actin bundles in filopodia (dotted boxes in B and C), and actin bundles running parallel to the L.E. within lamellipodia (dotted boxes in E and F). Scale bar = 10 nm

(I) Visualization of filaments with proximity-based pairing from tomogram (B), as shown in Figure 6G.

Video S5. Cryo-ET of PA-Rac1-induced lamellipodia in the late stage of formation, related to Figure 8B

Scale bar = 200 nm.

Video S6. Cryo-ET of PA-Rac1-induced lamellipodia in the late stage of formation, related to Figure 8E

Scale bar = 200 nm.

Our observations have primarily focused on bundled actin filaments within lamellipodia. However, other actin cross-linking proteins are also expected to contribute to the organization of the actin cytoskeleton in lamellipodia. For example, detailed analysis revealed variations in actin filament density, with some regions exhibiting densely packed filaments, while others displayed a more sparse distribution (Figures S6–S8). Furthermore, the observation that many actin filaments ran closely together in similar directions (Figures 6G and 8I) suggests the involvement of cross-linking proteins such as α-actinin⁴¹ in their organization.

In this cell, the protrusive membrane at the leading edge was not visible in the tomograms, likely due to the imaging location. In the Concave regions of the leading edge, individual filaments extended separately toward the membrane (Figures 8E and 8F), consistent with previous observations. Additionally, prominent actin filaments were oriented parallel to the leading edge beneath the plasma membrane in these regions.

Comparative analysis of actin dynamics and structure in anterior and posterior regions of lamellipodia

While many cryo-ET studies have primarily focused on the structure of lamellipodia leading edges, here we obtained tomograms from various regions spanning the anterior to posterior parts of lamellipodia. The anterior region, located approximately 1–1.5 μm from the leading edge, is characterized by exceptionally high actin polymerization and depolymerization activity.¹ Moreover, proteins that regulate actin dynamics, such as Ena/VASP, the SCAR/WAVE complex, and profilin, are concentrated in this region. Therefore, the anterior region is often analyzed separately from the posterior parts of the lamellipodia (Figure 1A).

In Cell 2, the fluorescence intensity of Lifeact-mCherry was high in the anterior regions and significantly lower intensity in the posterior regions of lamellipodia (Figure 3), suggesting structural differences between these regions. To compare these differences, we classified tomograms from all three cells based on their distance from the leading edge: “anterior” (near) and “posterior” (far) (Figure 9A).

Figure 9.

The orientation and the density of actin filaments relative to the leading edge (L.E.)

(A) The orientation of actin filaments relative to the L.E. Tomography numbers designated as “anterior region” are marked in red, while those identified as “posterior region” are noted in green.

(B) The orientation distributions of actin filaments in anterior regions (left: approximately 1.5 μm from the L.E.) and posterior region (right: more than 1.5 μm from L.E.).

(C) The density of actin filaments is relative to the L.E., where straight lines were drawn at 50 nm intervals from the L.E., and the number of intersecting actin filaments was counted. The data obtained from each tomography were plotted on a graph.

In the anterior region, the orientation of actin filaments was similar to the overall pattern (Figure 9B). In contrast, in the posterior region, the peak at 30°–40° was more pronounced, and another peak was observed around 90°. Although the density of actin filaments appeared to decrease toward the rear of the lamellipodia (Figure 9C), this reduction was surprisingly minor compared to the variation in Lifeact-mCherry fluorescence intensity (Figure 3, Cell 2).

Both the anterior and posterior regions contained bundled actin filaments. Despite differences in filament orientation, no significant structural differences in the actin cytoskeleton were observed between these regions, except for structures associated with the plasma membrane at the leading edge.

The ultrastructure of the convergence zone of lamellipodia and cortical region of cells

Finally, to further examine the structural organization of actin filaments and associated cellular components, we analyzed the convergence zone of lamellipodia and the cortical region of the cells.

The convergence zone of the lamellipodia, located near the original cortical region of cells (Figure 10A),^42,43 exhibited a predominantly disorganized actin cytoskeleton with minimal branching and a high degree of filament curvature (Figures 10B and S7F–S7H). The cortical actin bundles contained continuous filaments extending across entire tomograms, indicating that they are composed of actin filaments exceeding 1 μm in length (Figure 10B). Within these bundles, dozens of actin filaments twisted together with minimal branching. Additionally, MTs running parallel to the actin bundles suggested a coordinated structural arrangement.

Figure 10.

Cryo-ET of the convergence zone of lamellipodia and cell cortical region before lamellipodia formation

(A) A cryo-CLEM image of the same cell is shown in Figure 2, a cryo-ET area in a white box. Scale bar = 10 μm.

(B) Segmented cryo-ET of white box area in (A). Blue: ER and inner cellular membrane structure, Green: MTs, Red: actin filaments. Scale bar = 200 nm

(C and D) Magnified view of the area in B, focusing on the parallel arrangement of tubular ER and MTs. Includes a cryo-ET slice (C) and its segmentation (D). Scale bar = 200 nm.

Beyond these bundled actin filaments, tubular membrane structures were observed alongside MTs, suggesting the presence of the endoplasmic reticulum (ER; Figures 10C and 10D). This arrangement is consistent with previous LM studies,⁴⁴ highlighting the role of MTs and ER in cellular transport and the maintenance of structural integrity.

Discussion

Integrating cryo-ET and optogenetics for dynamic cellular ultrastructure analysis

In this study, we combined cryo-ET with optogenetics to overcome a major limitation of EM—the inability to capture temporal dynamics due to specimen fixation. Optogenetics enables the light-controlled manipulation of intracellular signaling molecules with high specificity and temporal resolution,^45,46 making it a powerful tool for precisely controlling cellular states before cryo-fixation. Previous studies have combined optogenetics with cryo-EM to study dynamic changes in neural cells^47,48 and purified proteins,⁴⁹ demonstrating the potential of integrating temporal control with high-resolution structural analysis. Unlike previous studies, this work applies in-cell cryo-ET without staining, providing detailed visualization of the ultrastructure in intact, living cells.

In this study, we applied this method to study lamellipodia formation, revealing the ultrastructure of the actin cytoskeleton and membrane organization. Although our study was limited to a single time point, the combination of optogenetic activation and rapid freezing could be extended in future studies to capture cellular events with precise temporal control, potentially enabling pseudo-time-lapse cryo samples for dynamic structural analysis.

Structural features of actin networks in PA-Rac1-induced lamellipodia formation

In this study, we used cryo-ET and optogenetic activation of PA-Rac1 to examine the organization and function of the actin cytoskeleton in lamellipodia formation. Previous studies have primarily examined fully developed lamellipodia in motile keratocyte cells,^28,29,50 limiting insights into their formation. By precisely controlling Rac1 activity, we investigated lamellipodia formation, providing new insights into its structural organization.

Overall, the density and orientation of actin filaments (Figures 4B and 4C) was consistent with previous studies,^22,31,38 validating our methodology. Additionally, the preservation of protrusive membrane structures and internal ER membranes (Figure 10) further supports the effectiveness of the freezing process.

One notable difference was the significantly lower frequency of actin branches (Figure 4F), approximately one-tenth of that reported in previous studies.^28,31,50 While some branches might not have been detected due to the inherent noise in cryo-ET images, this alone is unlikely to explain the 10-fold difference. Instead, this discrepancy could be attributed to differences in experimental conditions, including cell types and the protrusive activity of lamellipodia. The density of Arp2/3-mediated actin branching is thought to influence the stiffness of the actin cytoskeleton and the protrusion speed of lamellipodia.^51,52 Despite these discrepancies, the higher frequency of actin branches within lamellipodia underscores their critical role in the organization of the actin cytoskeleton during lamellipodia formation.

A hypothetical model for lamellipodia formation

In our structural analysis (Figures 5, 6, and 8), we examined lamellipodia with different extension lengths in a sequential manner, allowing us to compare their ultrastructural features at different degrees of extension. If lamellipodia extension represents a maturation process, their structural organization can be interpreted as undergoing distinct transitions. Based on this assumption, we propose a hypothetical model for their reorganization, as illustrated in Figure 11.

Figure 11.

A model of actin cytoskeleton reorganization during formation of lamellipodia

This figure illustrates a model for the reorganization of the actin cytoskeleton during lamellipodia formation, comprising three stages: Early Stage: Minor protrusions form, with unbundled actin filaments growing toward the leading edge; Middle Stage: mini filopodia emerge, with some actin filaments becoming cross-linked. Late Stage: mature filopodia are present, with cross-linked actin filaments running nearly parallel to the leading edge throughout the lamellipodia. See discussion for details—illustration created with BioRender.

Early stage

In the early stage, Minor protrusions emerged at the leading edge (Figure 5 and Video S3). Several actin filaments extended toward the leading edge but remained unbundled (Figures 5E and 5F). In Concave regions, actin filaments extended individually toward the leading edge. These observations suggest that actin polymerization is concentrated in specific regions, possibly involving liquid-liquid phase separation (LLPS) of actin-polymerizing factors.^53,54,55 At this stage, bundled actin filaments were not observed within the inner regions of lamellipodia.

Middle stage

The actin cytoskeleton became more structured and organized in the middle stage compared to the early stage (Figure 6 and Video S4). Large protrusions appeared, with 5–10 actin filaments growing toward the leading edge. Some of these filaments extended parallel to each other at consistent intervals, presumably cross-linked at approximately 8 nm. These structures, referred to as mini filopodia, resemble microspikes and filopodia but contain fewer and shorter filaments, making them detectable only with EM. Mini filopodia may arise through the elongation and bundling of Minor protrusions. As suggested earlier, Minor protrusions may form through LLPS, where actin-polymerizing factors such as Ena/VASP accumulate and promote actin assembly. Studies on LLPS have shown that actin filaments become bundled after nucleation.^41,55 Additionally, fascin-mediated actin bundling is reported to enhance Ena/VASP activity,⁵⁶ suggesting a cooperative mechanism in protrusion formation. Given their structural similarity to filopodia, it is plausible that mini filopodia are organized by both Ena/VASP and fascin.⁵⁷ These observations suggest that Minor protrusions could elongate and undergo actin bundling, potentially leading to the formation of mini filopodia.

These Large protrusions exhibited membrane regions with locally increased electron density (Figure 6B, red asterisks). Such regions may result from the local accumulation of specific lipids such as PIP2 and PIP3,⁵⁸ or from the clustering of proteins including Ena/VASP, Lamellipodin,⁵⁹ and Myo10,⁶⁰ which accumulate at the tip of filopodia and anchor to the membrane. Additionally, talin and integrins, recruited by these proteins, could contribute to clustering at these sites.⁶¹ At this stage, Minor protrusions remained observable, suggesting that these structures were undergoing turnover during lamellipodia formation.

A mini filopodia-like actin structure was observed slightly away from the leading edge, suggesting that it had formed at the leading edge and subsequently detached. Furthermore, some long actin filaments ran nearly parallel to the leading edge within the inner regions of lamellipodia.

Late stage

In the late stage, the tomograms we obtained did not capture protrusive structures at the leading edge. Instead, prominent microspikes and filopodia were observed in cryo-LM, and parts of these structures appeared in a tomogram (Figure 8). These microspikes and filopodia likely formed through the further maturation of mini filopodia. Conversely, many bundled actin filaments running nearly parallel to the leading edge were observed within the inner regions of lamellipodia (Figures 8G and 8H). These bundled actin filaments may have been initially part of mini filopodia and later detached from the leading edge. In living cells, microspikes and filopodia exhibit lateral movement due to lateral flow and occasionally collapse (Figure S1). A similar process may occur in mini filopodia, suggesting that mini filopodia either mature into filopodia or detach from the cell membrane to form bundled filaments running nearly parallel to the leading edge. Actin filaments running nearly parallel to the leading edge were often longer (Figure 4D). This observation suggests that retrograde actin flow pushes longer filaments into obstacles, where they remain aligned parallel to the leading edge. This mechanism is analogous to how longer logs in a river are more likely to encounter obstacles and align parallel to the flow. In cells, these obstacles may include other actin filaments and premature cellular adhesions.⁶²

Model summary

The mechanism of lamellipodia formation can be summarized as follows: The initiation of actin polymerization at the specific region of the plasma membrane at the leading edge leads to the formation of Minor protrusions. These Minor protrusions then develop into mini filopodia through the growth and bundling of actin filaments, which drive the plasma membrane forward. Some of these structures further mature into filopodia, while others may detach from the leading edge and remain within the lamellipodia, aligned parallel to the leading edge. The turnover of these structures contributes to lamellipodia extension. Our model complements the traditional role of Arp2/3-mediated branching in lamellipodia extension, providing an additional mechanism for actin network organization.

Actin filament-membrane interactions at the leading edge

In both Large and Minor protrusions, gaps were observed between the growing ends of actin filaments and the plasma membrane (Figure 7D). Such gaps are expected, as actin polymerization requires free space for the attachment of actin monomers to the filament tips. Such gaps may arise from the random movement of the plasma membrane, known as Brownian motion.⁶³ Although proteins such as Ena/VASP and mDia/formin are believed to regulate actin filament elongation at the growing plus ends, our cryo-ET did not provide sufficient contrast to identify electron-dense structures corresponding to these complexes. By contrast, in Concave regions, a greater proportion of actin filaments made direct contact with the plasma membrane compared to protrusive regions (Figures 7C and 7D), likely due to increased membrane tension. Moreover, some long actin filaments oriented parallel to the leading edge beneath the plasma membrane (Figures 6C and 8E), possibly helping to resist membrane tension.

The discrepancy between Lifeact-mCherry fluorescence and actin density in lamellipodia

Cryo-LM played a crucial role in identifying cells with PA-Rac1-induced lamellipodia and documenting their structural states. In particular, Cell 2, which exhibited the most extended lamellipodia, provided valuable insights. Tomograms No.1 and No.2 captured the anterior region (Figures S7A and S7B), with high Lifeact-mCherry fluorescence. In contrast, tomograms No.3 to 5 captured the posterior region (Figures S7C–S7E), where fluorescence was significantly lower. Despite this difference in fluorescence intensity, the density of actin filaments showed only minor variation (Figure 4B). This discrepancy may arise from two factors. First, at the tips of the lamellipodia, actin polymerization and depolymerization are active, with abundant G-actin.^16,64 Since Lifeact-mCherry binds not only to F-actin but also to G-actin,³³ its fluorescence intensity does not necessarily correlate with filament density. Second, actin retrograde flow could carry Lifeact-mCherry away from the posterior region.⁶⁵ Although actin filament orientation differed between these regions (Figures 9A and 9B), the limited number of tomograms prevented us from drawing statistically robust conclusions about these variations.

Conclusion

In conclusion, our study provides a detailed model for the reorganization of the actin cytoskeleton during lamellipodia formation by integrating cryo-ET with optogenetics. Our combined approach, which enables high-resolution structural analysis of intracellular dynamics, has the potential to expand the application of in-cell cryo-ET. While confirming the well-established role of Arp2/3-mediated branching in lamellipodia, our findings also reveal additional mechanisms, such as actin filament bundling and turnover, that contribute to lamellipodia extension. These insights not only refined our understanding of lamellipodia dynamics but also provided a framework for future studies to validate and expand upon this model.

Limitations of the study

The primary limitation of this study is the small sample size, consisting of only three cells and 16 tomograms. While we propose a model in Figure 11 based on the obtained data, the model incorporates various assumptions, such as the hypothesis that the degree of lamellipodia extension corresponds to distinct stages of lamellipodia formation. Further studies are necessary to validate these hypotheses. The small sample size reflects the challenges in preparing samples suitable for cryo-ET, which also introduces the risk of selection bias. The extended blotting process required for cryo-ET preparation can lead to cell desiccation and death. Indeed, we had to discard nearly half of the preparation grids because cryo-LM revealed that the cells had died. Moreover, even among surviving cells, many exhibited blebbing, indicating stress or damage (Figure S10). Consequently, when selecting grids containing healthy cells, only those with adequately thin ice layers for practical cryo-EM observation were chosen, comprising less than 20% of all prepared grids. This limitation resulted in a smaller sample size for analysis.

Additionally, several methodological limitations warrant attention. First, the blotting process may introduce artifacts into the actin cytoskeleton. Second, the accuracy of cryo-CLEM was limited by the absence of specific fluorescent labels for precise correlation. Instead, we relied on the positions of Quantifoil holes to align structures, which occasionally led to misalignments. Finally, we excluded actin filaments shorter than 35 nm and those oriented vertically at angles greater than 40° relative to the basal plane. While this threshold reduced false positives, it also limited our ability to fully analyze the complexity of the actin network.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Takao Nakata (info.cbio@tmd.ac.jp).

Materials availability

The present study generated several plasmids, which are available upon reasonable request.

Data and code availability

The reconstructed tomograms shown in Figures 5, 6, and 8 have been deposited in the EMDB under the following IDs: EMD-61088, EMD-61091, EMD-61092, and EMD-61093. The raw tilt images, prior to 3D reconstruction, as well as the segmentation data, have been deposited in the EMPIAR (ID: EMPIAR-12292).

This study used custom MATLAB scripts for data analysis. The code is available upon reasonable request.

Further information on data reported in this paper is available from the lead contact.

Acknowledgments

The authors are grateful to Dr. Florian Schur (Institute of Science and Technology (IST), Austria) for helpful discussion about analyzing actin network and 3D printing and Yuriko Sakamaki (the Research Core Center, Institute of Science Tokyo) for technical help. We thank Drs T. Ishii and T. Asano (Institute of Science Tokyo), S. Kikkawa (Kobe University) for the helpful discussion, and S. Nakamura, S. Kimura (Institute of Science Tokyo), K. Chin, Y. Sakihama, T. Shimizu (Kobe University), N. Makuta (Osaka University), and M. Nakamura (Mie University) for assistance, and other colleagues in Nakata lab and Nitta lab, the Research Core Center (Institute of Science Tokyo) for usage of the transmission electron microscopy and the carbon coater. This work was partly supported by the Nanotechnology Platform Program and “Advanced Research Infrastructure for Materials and Nanotechnology in Japan (ARIM)” of the Ministry of Education, Culture, Sports, Science and Technology (MEXT, Japan) proposal numbers JPMX09A18OS0055, JPMX09A19OS0046, JPMX09A20OS0002, JPMX09A21OS0020, and JPMX12220S0015. This research was also partially supported by the Research Support Project for Life Science and Drug Discovery (Basis for Supporting Innovative Drug Discovery and Life Science (BINDS)) from AMED under Grant Number JP22a.m.121001. This work was funded by JSPS KAKENHI (JP18K15002, JP20K16105, and JP22K06219 to H.I., JP18K19328 to H.I. and T.N., JP22K06809 to T.I., JP22K20680ZA and JP23K14178ZA to S.Y., JP23K23823 to T.K. and H.T., and JP23K06674 to H.G., JP21H05254 to R.N. and T.N.), by AMED-CREST from the Japan Agency for Medical Research and Development (AMED; JP21gm161003 to T.I.), by JST [FOREST (JPMJFR214K to T.I.), Moonshot R&D (JPMJMS2024 to R.N.)], by research grants from the Takeda Science Foundation (to H.I., T.I., and T.N.), and the Uehara Memorial Foundation (to H.I.).

Author contributions

Conceptualization: H.I., T.I., R.N., and T.N.; methodology: H.I., T.I., A.K., H.T., T.K., K.M., R.T., and T.K.; validation: H.I., T.I., R.N., and T.N.; investigation: H.I., T.I., A.K., S.Y., R.N., and T.N.; resources: H.I., T.I., T.K., H.G., K.M., R.N., and T.N.; data curation: H.I., T.I., K.A., H.T., T.K., M.R., R.N., and T.N.; writing – original draft: H.I.; writing–review and editing: T.I., S.Y., H.T., R.N., and T.N; visualization: H.I. and T.I.; supervision: R.N. and T.N.; project administration: H.I., T.I., H.G., R.N., and T.N.; funding acquisition: H.I., T.I., S.Y., H.Y., H.T., T.K., H.G., R.N., and T.N.

Declaration of interests

K.A. is affiliated with Thermo Fisher Scientific and declares no financial interests. All other authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies

During the preparation of this work, the authors used ChatGPT in order to improve the clarity and readability of the English text. Additionally, ChatGPT was utilized to develop MATLAB scripts for data analysis. After using this tool, the authors reviewed and edited the content and scripts as needed and took full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
poly-L-lysine	Nacalai Tesque	Cat# 28360-14
20 nm gold colloid	Sigma-Aldrich	Cat# 741965
laminin	Thermo Fisher Scientific	Cat# 23017015
Opti-MEMTM	Thermo Fisher Scientific	Cat# 31985070
ViafectTM Transfection Reagent	Promega	Cat# E4982
Cellmatrix Type I-C	Nitta Gelatin	Cat #631-00771
Deposited data
Raw tilt images and segmentation data	This paper	EMIPAR-12292
Reconstructed tomograms	This paper	EMD-61088; EMD-61091; EMDB-61092; EMD-61093
Experimental models: Cell lines
COS-7	RIKEN BRC, Japan	RCB0539; RRID:CVCL_0224
Recombinant DNA
pTriEx-mVenus-PA-Rac1	Wu et al.32	RRID:Addgene_22007
pTriEx-CFP-PA-Rac1	This paper	N/A
Lifeact-mCherry	Kakumoto and Nakata66	N/A
Lifeact-EGFP	This paper	N/A
pEGFP Abi1	Innocenti et al.35	RRID:Addgene 74905
mCherry-Abi1	This paper	N/A
mCherry-ARP2-N-14	Davidson lab – mCherry constructs, unpublished	RRID:Addgene 54980
mApple-LC-Myosin-C-10	Davidson lab – mApple constructs, unpublished	RRID:Addgene 54919
Software and algorithms
FV10-ASW	EVIDENT	RRID:SCR_014215
ZEN	Zeiss	RRID:SCR_013672
Leica Application Suite X	Leica microsystems	RRID:SCR_013673
Tomography 5	Thermo Fisher Scientific	N/A
IMOD	Kremer et al.67	http://bio3d.colorado.edu/imod; RRID:SCR_003297
Amira	Thermo Fisher Scientific	N/A
computational toolbox for ultrastructural quantitative analysis of filament networks in cryo-ET data	Dimchev et al.38	N/A
Fiji	NIH	http://fiji.sc; RRID:SCR_002285
MATLAB	MathWorks	RRID:SCR_001622
GraphPad Prism 10	GraphPad Software	RRID:SCR_002798
Adobe Photoshop	Adobe	RRID:SCR_014199
Adobe Illustrator	Adobe	RRID:SCR_010279

Experimental model and subject detail

Cell line

COS-7 cells (RCB0539: RIKEN BRC through the National BioResource Project of the MEXT/AMED, Japan) were used in this study. The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Nacalai Tesque #08458-65), supplemented with 10% fetal bovine serum (FBS, Biowest, #S1810) and incubated at 37°C with 5% CO₂. The cell line was neither authenticated nor tested for mycoplasma contamination; however, it was obtained from a reputable source (RIKEN BRC) and used immediately after acquisition exclusively for this study.

Method details

Plasmids

pTriEx-CFP-PA-Rac1 was constructed by replacing the mVenus fragment in pTriEx-mVenus-PA-Rac1³² (kindly gifted from Klaus Hahn; Addgene plasmid, #22007) with CFP. Lifeact-EGFP was generated by replacing the mCherry fragment in Lifeact-mCherry⁶⁶ with EGFP. mCherry-Abi1 was constructed by replacing the Abi1 fragment from EGFP Abi1³⁵ (kindly gift from Giorgio Scita; Addgene plasmid, #74905) into the BglII/SacII site of pmCherry-C1 (TaKaRa).

Cell preparation and transfection

Before seeding the cells, 200 mesh gold holey carbon grids (R1.2/1.3; Quantifoil Micro Tools) were coated with 0.01% (w/v) poly-_L-lysine (Nacalai Tesque, #28360-14), applied with 20 nm gold colloidal markers (Sigma-Aldrich, #741965). Next, a thin layer of a carbon was deposited using a carbon evaporation system, as described previously.³⁷ As a result, the holes of the Quantifoil grids were left uncovered by carbon. The grids were again coated with 0.01% (w/v) poly-_L-lysine for over 1 h and coated with 5 μg/mL laminin (Thermo Fisher Scientific, #23017015). A total of 1×10⁵ COS-7 cells were seeded onto the grids, which were placed in 3D-printed grid holders⁶⁸ in a four-well plate (SPL Life Sciences, #30004).

The transfection mixture, composed of 1 μg of pTriEx-mVenus-PA-Rac1³² and 0.02 μg of Lifeact-mCherry, 50 μL of Opti-MEM (Thermo Fisher Scientific #31985070), and 3.6 μL of ViaFect transfection reagent (Promega, #E4982), was incubated at RT for 15 min. This transfection mix was added to the cells just before seeding.

After transfection, the cells were incubated at 37°C with 5% CO₂ for 16–20 h before vitrification. Before vitrification, the expression of the fluorescent proteins was confirmed using either a BZ-X700 fluorescence microscope (Keyence) or an EVOS M5000 imaging system (Thermo Fisher Scientific).

For observing live cell imaging on the glass-bottom dishes, cells were cultured and transfected according to previously described methods.⁶⁹ Briefly, cells were seeded on glass-bottom dishes (Greiner Bio-one, #627870) coated with collagen (Cellmatrix Type IC; Nitta Gelatin, #631–00771) two days before the observations. The following day, cells were transfected with the plasmid described above, and analyses were conducted a day later.

For observing the localization of actin-binding proteins, cells were seeded on glass-bottom dishes coated with collagen. The following day, cells were transfected with 0.5 μg of ECFP-PA-Rac1, 0.02 μg of Lifeact-EGFP, and 0.1–0.5 μg of Abi1-mCherry, mCherry-Arp2-N-14 (kindly gifted from Michael Davidson; Addgene plasmid, #54980), or mApple-LC-Myosin-C-10 (kindly gifted from Michael Davidson; Addgene plasmid, #54919) using Viafect as described above. On a subsequent day, the cells were fixed with 4% paraformaldehyde solution (Nacalai Tesque #09154-85) after being exposed to blue LED light for 2 min or left unexposed. Blue light irradiation was performed using an optogenetics LED array system (LED array, 470 nm; BioResearch Center, #LEDA-B) controlled via an LED array driver (BRC Bio Research Center, #LAD-1) and an optogenetics stimulator (BRC Bio Research Center, #STOmk-2). The setting for the STOmk-2 was as follows: pulse width, 10 ms; interval, 100 ms; number of pulses, 10; rest interval, 1 s; amplitude, 5V. The amplitude of the LAD-1 was set to 13.5 V.

Live cell imaging

The grids with cultured cells were inverted so that the cell side faced downward and were then submerged in a glass-bottom dish (Violamo, #GBCD15) filled with Leibovitz’s L-15 (Thermo Fisher Scientific, #11415064). Cells were analyzed under the serum-starved condition in a sealed, heated chamber during live cell imaging at 37°C without CO₂ supplementation. Images were obtained using a confocal laser scanning microscope (FV1200, Olympus) on an IX83 microscope (Olympus) equipped with 40×/0.95 NA dry objective lenses and FV10-ASW software (Olympus). Photoactivation and imaging utilized standard setting for ECFP(C-Y-R), EYFP(C-Y-R), dsRed2(C-Y-R), and DIC (Argon laser power: 5% at 458-nm and 1% at 515-nm; diode laser power: 5% at 559-nm; DM 458/515/560 dichroic excitation, SDM510 and SDM560 emission filters; 475–500 nm emission window for ECFP, 530–540 nm emission window for EYFP, and 570–670 nm emission window for dsRed2; DIC images were obtained with dsRed2 settings) using the sequential line capturing mode (scan rate: 10 μs per pixel, pixel size 4.83 μm). The interval for both imaging and optoactivation was set to 10 s. PA-Rac1 activation was induced using the ECFP channels. mCherry images were acquired through the dsRed2 channel. Separately, mVenus images were captured with the EYFP channels only after live imaging to confirm the expression of mVenus-PA-Rac1.

Image analysis was conducted using ImageJ/Fiji software (NIH).⁷⁰ The changes in cell area were quantified using mCherry images. These images were binarized, and cell area was measured using the “Analyze Particles” function.

Adobe Photoshop 2024 and Illustrator 2024 (Adobe Systems) were used for the final figure preparation.

Localization analysis for actin-binding proteins

Images were obtained using a Zeiss Axio observer fluorescence microscope (Zeiss) equipped with the ApoTome 2 module, 63×/1.4 NA oil objective lenses, and ZEN software (Zeiss). The Colibri 7 light source and Axio Cam 506 camera (Zeiss) were used for imaging. For EGFP, band-pass filter No. 38HE (Zeiss, Ex 470/40, EM 605/70) was used, and for mCherry, band-pass filter No.43HE (Zeiss, Ex 550/25, Em 605/70) was applied. Z-sectioning was performed with ApoTome 2 at 0.52 μm steps. Images were processed with ImageJ/Fiji using Z-projection (Max intensity). Plot profiles were generated by drawing a line from the tips to the posterior regions of lamellipodia.

Blue light irradiation and vitrification

Samples were vitrified using either a Leica GP2 plunger (Leica Microsystems) or the Vitrobot Mark IV System (Thermo Fisher Scientific) set to 37°C and 95% humidity. Before transfer into the blotting chamber, the medium was manually blotted off, and 5 μL of PBS was added to the grids. Inside the blotting chamber, blue light irradiation was performed using a fiber-coupled LED (470 nm, 1000 mA; Thorlabs, #470F1), which was connected to a ferrule patch cable (ϕ400 μm core, 0.39 NA; Thorlabs, #M79L01). The output of LED was regulated by a T-Cube LED driver (1200 mA; Thorlabs, #LEDD1B). The cable was fed into the blotting chamber through an aperture reserved for sample application, with the blue light projecting directly from the front of the grid at approximately 1 cm distance, at maximum power. The irradiation timing was manually managed to ensure that blotting was completed, and the sample was plunged into liquid ethane precisely 2 min after the onset of irradiation. Vitrification of the grids in liquid ethane was following backside blotting for 10 s or bilateral blotting for 5 s, using No.2 filter paper and blotting sensor of the Leica GP2 or Vitrobot Mark IV System, respectively. Samples were stored under liquid nitrogen conditions until the time of imaging.

Cryo-fluorescence widefield microscopy

Fluorescence images of vitrified samples were captured using a THUNDER Imager EM Cryo CLEM widefield microscope (Leica Microsystems), which was equipped with a 50×/0.9 NA dry objective lens, a metal halide light source (EL6000), an air-cooled detector (DFC9000 GT), and a cryo-stage maintained at −190°C. Before observation, grids were placed into AutoGrids (Thermo Fisher Scientific). A 2 mm by 2 mm square area was imaged at the grid’s center using the LAS X Navigator software (Leica Microsystems). For each field of view, a symmetrical 20 μm z stack with 5 μm intervals was captured centered around the autofocus point to create maps. Subsequently, lamellipodia-forming cells were visually selected for further imaging. A 20 μm z stack with 0.75 μm intervals was acquired at these locations. The imaging utilized multiple channels: transmitted light brightfield, reflection, GFP (Ex: BP470/40, Em: 525/50), and Texas Red (Ex: BP560/40, Em: 630/75). Image stitching was executed using the LAS X software. Small Volume Computational Clearing (SVCC) was applied to the acquired image stacks to diminish blurring and enhance weaker or subdued signals. Images and mosaic tiles were exported in TIFF format. Additional image processing, including maximum intensity projection, flipping, cropping, and contrast adjustment, was conducted using ImageJ/Fiji and Adobe Photoshop.

Cryo-electron microscopy

Cryo-electron tomograms were acquired on a Titan Krios (Thermo Fisher Scientific) operating at an acceleration voltage of 300 kV and utilized a Cs corrector (CEOS, GmbH), a Volta phase plate, and BioQuantam K3 direct detector with energy filter (slit width of 20 eV) (Gatan). The microscope was controlled using the Tomography 5 software (Thermo Fisher Scientific). Initially, an atlas was acquired to identify the square where the target cells were located. Subsequently, the tomography acquisition position was determined based on the correlation with fluorescence images. Tilt series were collected using a continuous tilt scheme, starting at 0°, the sample was incrementally tilted to 70°, then to −70° in 2° steps. The magnification was ×19,500, resulting in a pixel size of 3.71 Å. The defocus was set to approximately −0.5 μm, and the spherical aberration (Cs) value was around 5 μm. The Volta phase plate spot was changed after every 2–3 tilt series. The electron dose per image was calculated as 1.44 e⁻/Ų, with the total dose for series approximately 100 e⁻/Ų.

Image processing

Before 3D reconstruction, poor quality tilt images caused by obstacles, such as grid bars blocking the beam at high tilt angles, were removed. The tilt series images were then reconstructed into 3D tomograms by using the IMOD software package.⁶⁷ Micrographs were aligned by cross-correlation, followed by alignment through tracking 20 nm gold fiducial beads coated on the grid. These aligned micrographs underwent reconstruction for visual analysis using IMOD SIRT (simultaneous iterative reconstruction technique, number of iterations: 8) at bin2. CTF correction was not performed on these reconstructions.

Segmentation of cellular components was initially performed using Amira software (Thermo Fisher Scientific). Before segmentation, the reconstructed images were binned 2 to 3 times (resulting in final bin4 and bin6 from the original images) and processed with a Gaussian filter. The membrane structure was manually segmented, while actin filaments and MTs were segmented using the XFiber module of Amira.⁷¹ Filament tracing was undertaken after high-contrast structures such as the membrane, grid edge, and outer cellular space were masked. Microtubule tracing utilized the following parameters, as described previously⁷²: cylinder length: 600 Å, angular sampling: 5, mask cylinder radius: 140 Å, outer cylinder radius: 125 Å, inner cylinder radius: 75 Å, and missing wedge according to the individual tilt series. For actin filaments, these parameters were used: cylinder length: 500 Å, angular sampling: 6 Å, mask cylinder radius: 45 Å, outer cylinder radius: 35 Å, inner cylinder radius: 0 Å. The Trace Correlation Lines module was configured with these parameter values: minimum seed correlation: 75–125 (tomogram dependent), minimum continuation quality: 60–90 (tomogram dependent), direction coefficient: 0.3, minimum distance: 70 Å, minimum length: 350 Å, search corn minimum step size (%): 10. Segments and point coordinates were extracted into separate Excel sheets from Amira and reformatted to IMOD/Etomo-style files using a MATLAB-script (“amira_reformat_to_coordinates.m” in “computational toolbox for ultrastructural quantitative analysis of filament networks in cryo-ET data”³⁸). Tracing errors were manually corrected in 3dmod in IMOD, and the branch points of actin filaments were also determined manually in 3dmod.

Visualization of the 3D model was performed in 3dmod. Color-coded maps illustrating the angular distribution of actin filaments relative to the leading edge direction were generated using the ultrastructural analysis toolbox. Three-dimensional images of membranes and MTs images created in 3dmod were meticulously merged with the color maps in Adobe Photoshop.

Quantification and statistical analysis

Quantitative analysis of actin filaments was conducted using the ultrastructural analysis toolbox.³⁸ The distance between the membrane and the tips of actin filaments was measured using ImageJ/Fiji.

To analyze the spatial arrangement of filamentous structures, a custom MATLAB script was developed to detect and visualize pairs of filaments based on specific geometric criteria. The raw coordinates were resampled at equal intervals of 20 nm along the filament’s trajectory to ensure consistent spatial resolution. For each filament pair, proximity and angular alignment were evaluated. The proximity criterion required the minimum distance between two filaments to be within 25 nm. The angular alignment was assessed by calculating the angle between the unit vectors of the two filaments, with a threshold of 7.5°. Pairs of filaments were considered valid if at least 50% of the sampled points met both criteria. Filaments involved in valid pairs were visualized in 3D plots, where they were highlighted in red, while filaments not meeting the criteria were shown in black.

Graphs were created using GraphPad Prism 10 (GraphPad Software) and MATLAB (MathWorks).

Published: April 24, 2025