Reconstituted nascent adhesion condensates drive actin polymerization on supported lipid bilayers

Abstract

Nascent adhesions are early integrin-based assemblies that couple the extracellular matrix to the actin cytoskeleton and mature into focal adhesions. Many nascent-adhesion proteins interact through weak, multivalent contacts, suggesting that liquid-like organization may contribute to adhesion assembly. However, how phase separation shapes actin polymerization and organization remains unclear. Here, we compare two vasodilator-stimulated phosphoprotein (VASP)–recruiting adaptor proteins, zyxin and vinculin, to determine how adaptor identity tunes condensate properties and actin coupling. Both zyxin-VASP and vinculin-VASP assemblies drive integrin clustering and support actin filament growth. Notably, zyxin-VASP condensates remain fluid and redistribute along newly formed actin bundles, whereas vinculin-VASP condensates are more rigid and fail to spread along actin despite sustaining polymerization. These results suggest that differential VASP recruitment can modulate condensate properties and actin architecture, providing a potential mechanism for the maturation of nascent adhesions into focal adhesions.

Adhesion protein condensates with distinct properties differentially shape actin organization.

INTRODUCTION

Focal adhesions (FAs) are multiprotein assemblies that form at the plasma membrane to mechanically and biochemically couple the extracellular matrix to the actin cytoskeleton (1, 2). Their core components are integrins, heterodimeric α/β transmembrane receptors whose activation and ligand binding require conformational changes stabilized by talin and kindlin binding to the integrin β cytoplasmic tail (1, 3–7). These interactions initiate the recruitment of the adhesome (8), a large network of adaptor and signaling proteins that reinforces cytoskeletal coupling and enables mechanotransduction (9, 10). Integrin adhesion assembly begins with small, short-lived nascent adhesions (NAs) at the leading edge of migrating cells. While most NAs disassemble, a subset matures into larger, long-lived FAs (9). Integrin clustering within NAs is driven by weak, multivalent protein-protein interactions. Such interactions are a hallmark of liquid-liquid phase separation (LLPS), which is a form of reversible organization of protein assemblies within cells (11). Consistent with NAs being highly dynamic structures, multiple integrin adhesion components display liquid-like behavior. Recent in vitro studies have shown that several FA proteins—including talin, kindlin, paxillin, LIMD1 (LIM domain containing protein 1), TNS1 (tensin 1), vinculin, zyxin, FAK (focal adhesion kinase), p130Cas, and VASP (vasodilator-stimulated phosphoprotein)—can phase separate (12–20). In addition, a subset of these proteins has been shown to cluster integrin tails, particularly in the presence of lipid membranes enriched in phosphoinositides such as phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] (12, 13, 15, 16). Together, these findings suggest that membrane-assisted phase separation contributes to NA assembly. However, how the resulting phase-separated assemblies trigger actin polymerization and regulate cytoskeletal organization at different stages of adhesion maturation remains poorly understood.

Actin polymerization is essential for NA maturation, force transmission, and stress fiber formation (1, 21). Coupling integrin adhesions to the actin cytoskeleton requires recruitment of actin-binding proteins such as talin and vinculin (22), which can directly bind and bundle polymerized actin (23). At the same time, spontaneous actin polymerization is suppressed in cells by capping and monomer-sequestering factors such as CapZ and profilin (24, 25), necessitating localized actin polymerization factors at adhesion sites. VASP is a prominent actin polymerization factor in NAs (21). VASP promotes filament elongation by competing with CapZ for barbed ends and by incorporating profilin-actin complexes (26). Notably, VASP itself can undergo phase separation in solution or upon tethering to lipid bilayers, where it nucleates and polymerizes actin (14, 20, 27). In cells, however, VASP is targeted to adhesion sites via adaptor proteins, among them zyxin and vinculin, both of which bind VASP and actin and are critical for adhesion function (28–32). Zyxin and vinculin display differential behavior at different stages of adhesion maturation: Vinculin is present in NAs (22, 33, 34), whereas zyxin arrives later and becomes increasingly important as traction forces rise and adhesions grow (35–37), raising the question of their effects in NA assembly and maturation. The material properties of phase-separated assemblies are known to depend on their molecular composition (11). We therefore hypothesize that recruiting VASP through different adaptor proteins could tune properties of adhesion condensates and, in turn, regulate actin polymerization and network architecture during adhesion assembly. Although both zyxin and vinculin can recruit VASP to integrin adhesions, it is unknown whether they differentially modulate VASP condensate mechanics and actin organization.

Here, we reconstitute a minimal NA-like system by inducing phase separation of key adhesion proteins on supported lipid bilayers (SLBs). This reconstitution recapitulates two defining features of NAs: integrin tail clustering and localized actin polymerization. Using this system, we directly compare zyxin- and vinculin-mediated VASP assemblies and show that they form condensates with distinct physical properties and actin network geometries. Zyxin-based assemblies remain fluid and promote actin bundling, whereas vinculin-based assemblies are more rigid and support filament growth without bundling. On the basis of these findings, we propose that differential tuning of condensate material properties by adhesion adaptor proteins contributes to distinct stages of integrin adhesion assembly.

RESULTS

Zyxin and vinculin undergo phase separation with VASP and polymerize actin

We reconstitute the minimal NA-like complexes in vitro to compare the role of zyxin and vinculin in integrin adhesion assembly and actin polymerization. To achieve this, we first incubate SLBs containing 5 mol % PI(4,5)P₂ with β1 integrin tails (β1 tails), 1 μM talin-1, and 1 μM kindlin-2. Subsequently, we introduce 0.5 μM zyxin-mCherry (zyxin) combined with 0.5 μM VASP or 0.5 μM vinculin-Alexa Fluor 555 (AF555) (vinculin) combined with 0.5 μM VASP (Fig. 1A; see Materials and Methods). Upon mixing, zyxin-VASP and vinculin-VASP complexes undergo rapid phase separation in solution (movies S1 and S2), leading to the formation of biomolecular condensates that sediment and adhere to the membrane (Fig. 1B and movies S3 and S4). Zyxin-VASP attachment to the membrane is likely mediated by nonspecific interactions between the condensate and the membrane or membrane-associated proteins, as no direct interactions between zyxin or VASP and kindlin-2, β1 integrin, talin-1, or lipids have been reported. In contrast, for vinculin-VASP condensates, membrane association can be explained by known interactions of vinculin with PI(4,5)P₂ and talin (23, 38, 39). In our system, diffusion of β1 tails and associated kindlin-2/talin-1 is largely confined to the membrane plane, whereas zyxin, vinculin, and VASP diffuse in solution. In the cellular context, likely diffusion both in the cytosol and at the membrane plane upon interaction is relevant for adhesion assembly. Talin has been demonstrated to be recruited to FAs from the cytoplasm, while kindlin is recruited to FAs through membrane diffusion (40). Most of the condensates formed by the zyxin-VASP and vinculin-VASP pairs adopt an area size smaller than 0.25 μm² (Fig. 1C). The observed small condensates and their slow kinetics are probably due to the two-dimensional membrane confinement, which has been observed in the decreased cluster growth of RNA-protein complexes on the membrane (41). Such condensate characteristics are also observed across additional protein concentrations and stoichiometries (figs. S1 and S2).

Fig. 1. Zyxin and vinculin undergo phase separation with VASP on reconstituted membranes and cluster β1 tails.

(A) Schematic of the experimental setup for NA-like complex reconstitution on SLBs. (B) Representative epifluorescence images showing zyxin-VASP and vinculin-VASP condensates forming on the SLB and clustering β1 tails. Scale bars, 10 μm. (C) Quantification of condensate surface area (area per condensate) for zyxin-VASP and vinculin-VASP condensates after 10-min incubation. Zyxin is visualized with zyxin-mCherry, vinculin is visualized with 10% vinculin-AF555, and β1 tails are visualized with β1-ATTO647. Condensates and β1 tails are imaged by epifluorescence microscopy.

To explore the functionality of reconstituted NAs, especially their ability to support actin polymerization, we assess the effect of zyxin-VASP and vinculin-VASP condensates in a pyrene-actin polymerization assay (Fig. 2A). While pyrene-actin alone polymerizes rapidly, the addition of 50 nM CapZ and 13 µM profilin–two key regulators of actin dynamics–suppresses polymerization. In contrast, inclusion of 0.5 µM zyxin and 0.5 µM VASP or 0.5 µM vinculin and 0.5 µM VASP restores polymerization, albeit at a reduced rate, demonstrating that both condensates actively promote actin filament formation. CapZ and profilin mediate polymerization inhibition to prevent uncontrolled actin growth in cells. CapZ caps filament barbed ends, preventing profilin-actin from binding. Profilin binds G-actin and promotes ADP-to-ATP exchange to prime it for polymerization (42) while it simultaneously blocks its incorporation at free barbed ends. CapZ and profilin have been shown to be inactivated near membranes through interaction with PI(4,5)P2 (24, 25, 43). This implies reduced local concentrations of active CapZ and profilin at membranes containing PI(4,5)P2.

To examine how PI(4,5)P₂-containing membranes influence condensate-mediated actin polymerization, zyxin-VASP or vinculin-VASP condensates are incubated on model membranes functionalized with β1 tails, talin-1, and kindlin-2. After 10 min, an actin polymerization mix (2 μM G-actin, 50 nM CapZ, and 13 μM profilin) is added (Fig. 2B; see Materials and Methods). Upon addition, we observe rapid actin accumulation within the condensates (Fig. 2, C and D), driven by actin binding of kindlin-2, talin-1, VASP, and zyxin or vinculin (44–48). Enrichment of G-actin within condensates is characteristic of a general feature of phase-separated systems—preferential sequestration of further components—which locally increases their effective concentrations (11). Beyond actin, both zyxin-VASP and vinculin-VASP condensates also partition FAK (fig. S3), a key signaling factor implicated in NA assembly and maintenance (9). In addition, CapZ accumulated in condensates when added 10 min after the onset of actin polymerization (fig. S4).

Fig. 2. Zyxin and vinculin undergo phase separation with VASP and promote actin polymerization.

(A) Pyrene-actin polymerization assay in the absence or presence of profilin, CapZ, and/or zyxin-VASP or vinculin-VASP condensates. Error bars represent the standard deviations of three independent experiments. (B) Schematic of the experimental setup for condensate-mediated actin polymerization on SLBs. (C and D) Representative images showing G-actin partitioning into zyxin-VASP (C) and vinculin-VASP (D) condensates. Scale bars, 10 μm. (E) Time-lapse image series showing actin polymerization by zyxin-VASP and vinculin-VASP condensates. Scale bars, 1 μm. (F) Representative images of actin filaments nucleating outside condensates; arrows indicate filament origin sites, and circles mark condensate positions. Scale bars, 1 μm. (G) Box plot quantifying the fraction of actin filaments originating from condensates compared to those originating outside condensates. Boxes indicate the 25th to 75th percentiles, the median is shown as a line, and whiskers represent 1.5× the interquartile range; P < 0.05 is reported. Significance was tested with a Mann-Whitney U test. Each condition contains n measurements from three independent experiments. Zyxin is visualized with zyxin-mCherry, and actin is visualized with 12.5% actin-ATTO488. Condensates are imaged by epifluorescence microscopy and actin filaments by TIRF microscopy.

Localized actin enrichment enhances the nucleation and polymerization potential inside the condensates. VASP, a well-established actin polymerizing factor (49), is able to outcompete CapZ at filament barbed ends and use profilin-actin to promote filament elongation (Fig. 2E and fig. S5) despite the enrichment of CapZ in the condensates (fig. S4 and movies S5 and S6). In contrast, when the polymerization mix is added to PI(4,5)P₂-containing membranes lacking condensates, polymerization is fully suppressed, indicating that profilin and residual CapZ alone are sufficient to inhibit filament formation under these conditions (movie S7), similar to the observations in the pyrene actin assay.

Although both zyxin-VASP and vinculin-VASP condensates promote actin assembly, the filament origin differs. For zyxin-VASP, most filaments appear to polymerize from the condensates, whereas in the case of vinculin-VASP condensates, filament formation predominantly occurs outside the condensates (Fig. 2, F and G, and movies S8 and S9). At the same time, both types of condensates bind preexisting filaments that are being further elongated upon binding to the condensates. Whether filaments nucleate inside condensates or nucleate outside and then elongate rapidly after binding remains unclear. In cells, NAs form near the actomyosin cortex and branched actin networks in lamellipodia (50). Condensate affinity for filaments may be essential for connecting NAs to the actomyosin complex, while actin polymerization from condensates may reinforce these connections.

We hypothesize that the variation in actin polymerization arises from differences in how effectively VASP is able to compete with CapZ and to displace it from the actin barbed end. Both zyxin and vinculin bind actin filaments and recruit VASP. Zyxin-bound VASP may more efficiently outcompete CapZ and therefore drive robust polymerization from within condensates. In contrast, vinculin-bound VASP may be less effective, likely due to the additional barbed-end capping activity of vinculin-talin complexes (51), which inhibits polymerization inside condensates. Outside the condensates, capping is reduced because of CapZ inactivation by PI(4,5)P₂ and its partition into the condensates. This, together with residual free VASP, promotes filament nucleation in solution, allowing newly formed filaments to become incorporated into condensates. In conclusion, linking VASP to the membrane via zyxin or vinculin enables condensate-driven actin polymerization.

Zyxin-VASP and vinculin-VASP condensation and β1 tail clustering depend on membrane functionalization

To dissect individual roles of VASP, zyxin, and vinculin in condensate formation, we incubate them individually or in combinations on SLBs containing PI(4,5)P₂ and β1 tails preincubated with 1 μM talin-1 and 1 μM kindlin-2 (see Materials and Methods). Incubation of 0.5 μM VASP-AF532 alone neither induces phase separation nor clusters β1 tails, consistent with the lack of known interactions between VASP, β1, kindlin-2, talin-1, and PI(4,5)P₂ (Fig. 3A). Furthermore, no phase separation of VASP occurs in solution under our experimental conditions.

Fig. 3. Zyxin-VASP and vinculin-VASP condensate formation and β1 tail clustering is talin-1– and kindlin-2–dependent.

(A) Zyxin, vinculin, or VASP alone neither forms condensates nor clusters β1 tails, whereas combinations of these proteins do. (B and C) Box plots quantifying the number of zyxin-VASP (B) and vinculin-VASP (C) condensates formed on membranes as a function of PI(4,5)P₂, talin-1, and kindlin-2. Boxes indicate the 25th to 75th percentiles, the median is shown as a line, and whiskers represent 1.5× the interquartile range; “ns” denotes not significant, significance was tested with a Mann-Whitney U test, and P < 0.05 are reported. Error bars in (B) and (C) represent the standard deviations of n measurements from three independent experiments. (D and E) β1 tail partitioning into zyxin-VASP (D) and vinculin-VASP (E) condensates after 10 min, comparing conditions with or without talin-1 and kindlin-2. (E) β1 tail partitioning to the vinculin-VASP condensates after 10 min depending on the presence or absence of talin-1 and kindlin-2. (F and G) FRAP-derived diffusion parameters for β1 tails in the presence of zyxin-VASP and vinculin-VASP condensates as a function of PI(4,5)P₂, talin-1, and kindlin-2: (F) fast half-time of recovery ($t_{1/2\_\text{fast}}$) and (G) mobile fraction. Data are shown as the means ± standard deviation. Zyxin is visualized with zyxin-mCherry, vinculin is visualized with 10% vinculin-AF555, and β1 tails are visualized with β1-ATTO647. Condensates and β1 tails are imaged by epifluorescence microscopy. Error bars in (F) and (G) represent the standard deviations from three independent experiments. Scale bars, 10 μm (all panels).

When incubated individually, both 0.5 μM zyxin-mCherry (zyxin) and 0.5 μM vinculin-AF555 (vinculin) adsorb to the membrane but do not induce phase separation or β1 tail clustering. In contrast, combining either zyxin or vinculin with VASP leads to rapid phase separation in solution and the formation of condensates that sediment onto the membrane (Figs. 1B and 3A).

To assess the role of talin-1 and kindlin-2 in condensate attachment and β1 tail clustering, zyxin-VASP or vinculin-VASP condensates are incubated on β1 tail–containing membranes with either protein alone or both. Condensates attach to the membrane under all conditions (Fig. 3, B and C), with zyxin-VASP forming more condensates than vinculin-VASP. For zyxin-VASP, attachment increases in the presence of either talin-1 or kindlin-2 and is maximal when both are present, whereas vinculin-VASP condensate numbers increase only when both proteins are present. Removal of PI(4,5)P₂ reduces the number of membrane-associated condensates for both systems (Fig. 3, B and C).

Both zyxin-VASP and vinculin-VASP condensates promote β1 clustering, with the strongest effect observed in the presence of both talin-1 and kindlin-2 (Fig. 3, D and E), consistent with previous findings in which phase separation stabilized β1 tail complexes with talin-1 and kindlin-2 (15). To test this hypothesis, we use fluorescence recovery after photobleaching (FRAP) to assess the effect of talin-1 and kindlin-2 on β1 tail diffusion in the presence of zyxin-VASP and vinculin-VASP condensates (Fig. 3, F and G). In our experimental setup, β1 tail diffusion is confined to the membrane plane because of its coupling to lipids. This approximates membrane-based diffusion of full-length integrin receptors in cells. FRAP measurements reveal that kindlin-2 substantially reduces β1 tail diffusion and mobile fraction, whereas talin-1 alone has little effect; the strongest reduction occurs when both proteins are present (Fig. 3, F and G). Together, these results indicate that kindlin-2 exerts a stronger stabilizing effect on β1 tails than talin-1, while their combined presence displays an additive strengthening of the interactions between the membrane and condensates, as well as an additive increase in β1 tail clustering.

Zyxin-VASP and vinculin-VASP condensates generate different actin networks

While both zyxin-VASP and vinculin-VASP condensates promote actin polymerization, the resulting actin architectures exhibit notable differences (Fig. 4, A and B, and movies S10 and S11). First, actin filaments originating from zyxin-VASP networks display zipping when they encounter each other (Fig. 4C and movie S12), while actin filaments within vinculin-VASP networks cross over each other without merging and bundling (Fig. 4D and movie S13). Second, actin filaments originating from zyxin-VASP condensates bundle at a steady rate, while actin connected to vinculin-VASP condensates does not bundle and remains in the form of single filaments (fig. S6). As a result, zyxin-VASP generates interconnected networks primarily composed of bundles, while vinculin-VASP condensates generate a network of single filaments with a low number of bundles (Fig. 4E). As shown above, VASP is necessary for phase separation and actin polymerization. In addition, VASP is known to bundle actin filaments, a process that requires free movement of VASP along the growing filament (14). When VASP is part of a phase-separated condensate, its ability to move along the filaments depends on the condensate’s fluidity (20). Specifically, more fluid condensates allow more bundling. The actin polymerization and presence/absence of bundle wetting phenomena are also observed across additional concentrations and stoichiometries (figs. S7 and S8).

Fig. 4. Zyxin-VASP and vinculin-VASP condensates generate distinct actin network architectures.

(A) Representative image of an actin network generated by zyxin-VASP condensates after 20 min, showing prominent actin bundles. Scale bar, 10 μm. (B) Representative image of an actin network generated by vinculin-VASP condensates after 20 min, showing predominantly single filaments. Scale bar, 10 μm. (C) Time-lapse image series showing zyxin-VASP–generated filaments forming a bundle upon filament-filament encounter; arrows mark the interaction point. Scale bar, 1 μm. (D) Time-lapse image series showing vinculin-VASP–generated filaments crossing over each other upon encounter; arrows mark the interaction point. Scale bar, 1 μm. (E) Box plot quantifying actin bundle line density (number of bundles intersecting a line). Boxes indicate the 25th to 75th percentiles, the median is shown as a line, and whiskers represent 1.5× the interquartile range; P < 0.05 is reported. Significance was tested with a Mann-Whitney U test. Each condition contains n measurements from three independent experiments. Actin is visualized with actin-ATTO488. Actin images are imaged using TIRF microscopy.

To confirm condensate fluidity, we investigate condensate behavior during actin polymerization. Consistent with liquid-like properties, condensates formed on actin filaments and redistributed along them in a manner reminiscent of wetting (14, 20, 52). As actin polymerizes, zyxin-VASP condensates extend along the growing filaments. As these filaments bundle, the condensates elongate further and wet the bundles (Fig. 5A and movie S14). By contrast, vinculin-VASP condensates maintain a constant shape and show neither actin wetting nor filament bundling (Fig. 5B and movie S15). To quantify these shape changes, we measured condensate eccentricity as a measure of how much the condensate shape deviates from a circle before and after polymerization. Consistent with the qualitative observations, the eccentricity distribution of zyxin-VASP condensates shifts toward more elongated shapes (Fig. 5C), whereas the distribution for vinculin-VASP condensates remains unchanged (Fig. 5D). In the experiments with CapZATTO488 added after actin polymerization, CapZ partitions into the vinculin-VASP condensates, indicating that most of barbed ends are found within a condensate. At the same time, CapZ partitions both to the zyxin-VASP condensates and to the bundles wetted by zyxin-VASP (fig. S4). This indicates that zyxin-VASP complexes can track the polymerizing barbed end against the surface tension of the condensate to allow bundle formation. This leads to a deformation of the condensate area to decorate the actin bundle. Together, the results suggest that condensate fluidity plays a critical role in controlling actin polymerization.

Fig. 5. Actin-dependent remodeling and dynamics of zyxin-VASP and vinculin-VASP condensates.

(A and B) Time-lapse image series illustrating changes in condensate shape upon actin polymerization. Scale bars, 5 μm. (C and D) Histograms showing changes in condensate eccentricity as a function of actin polymerization. (E) Representative FRAP image series of zyxin-VASP and vinculin-VASP condensates. Scale bars, 1 μm. (F) Normalized FRAP recovery curves and extracted parameters [fast half-time of recovery $\left(t_{1/2\_\text{fast}}\right)$ and mobile fraction] for zyxin-VASP and vinculin-VASP condensates. Data are shown as the means ± standard deviation. Error bars represent the standard deviations of three independent experiments. (G) Representative image of an actin network generated by zyxin-VASP and subsequently contracted by NMMII in the presence of α-actinin after 60 min, showing clustered actin filaments and condensates. Scale bar, 10 μm. (H) Time-lapse image series showing actin network contraction and concomitant clustering of zyxin-VASP condensates in the presence of NMMII and α-actinin. Scale bar, 5 μm. (I) Box plot quantifying the number of zyxin-VASP condensates before and after actin network contraction in the presence of NMMII and α-actinin. Boxes indicate the 25th to 75th percentiles, the median is shown as a line, and whiskers represent 1.5× the interquartile range. Significance was tested with a Mann-Whitney U test. Each condition contains n measurements from three independent experiments. Zyxin is visualized with zyxin-mCherry, vinculin is visualized with 10% vinculin-AF555, and actin is visualized with actin-ATTO488. Condensates are imaged using epifluorescence microscopy, and actin filaments are imaged using TIRF microscopy.

To determine whether the difference in bundling behavior depends on the condensate properties, we perform FRAP measurements. After bleaching, both types of condensates show fluorescence recovery, indicating a presence of a mobile fraction characteristic of LLPS (Fig. 5E). However, vinculin-VASP condensates demonstrate four times slower recovery than zyxin-VASP, suggesting lower fluidity within the vinculin-containing condensates (Fig. 5F). This reduced fluidity resembles a more cross-linked or gelated phase separation system rather than the rapid dynamics typical for phase separation in solution (53).

Given the high network connectivity and condensate fluidity observed for zyxin-VASP, we tested the force responsiveness of these condensates, as zyxin and other LIM-domain proteins are mechanosensitive and mechanical tension promotes the maturation of NAs into stable focal complexes and FAs (9, 54, 55). In addition to the forces generated by the myosin activity, the presence of a cross-linker such as α-actinin was demonstrated to be important for the NA maturation (56, 57) as well as to establish elastic connections that allow myosin II motors to generate contractile stresses within the network (58). To test the force responsiveness of zyxin-VASP condensates, we introduce 0.2 μM α-actinin1b (α-actinin) and 0.5 μM nonmuscle myosin IIa (NMMII) into actin networks polymerized by zyxin-VASP (see Materials and Methods). When both α-actinin and NMMII are present, the actin network undergoes pronounced contraction. Zyxin-VASP condensates detach from the membrane while remaining connected to the actin network (Fig. 5G). As contraction progresses, condensates cluster together (Fig. 5, H and I, and movie S16), suggesting that NMMII-driven contractility can promote condensate fusion.

Our findings suggest that the observed differences may stem from how VASP interacts with either zyxin or vinculin. Vinculin directly binds PI(4,5)P₂ (38) and the talin-1 rod domain (39) and engages VASP through a single FP4 motif (59). In contrast, zyxin has not been reported to interact with lipids or talin-1/kindlin-2, yet it binds VASP via four FP4 motifs (60). We therefore hypothesize that vinculin-VASP provides stronger membrane anchoring, confining condensates to the membrane plane and limiting their mobility. Conversely, zyxin-VASP may remain more weakly coupled to the membrane and, thus, more fluid. This hypothesis would require further testing through the generation of interaction domain mutants and systematically varying them.

DISCUSSION

We reconstitute minimal, functional NA-like complexes anchored to lipid bilayer membranes and investigate how zyxin and vinculin affect VASP phase separation and resulting actin polymerization. In our assays, NA-like condensates form in solution through the LLPS of zyxin-VASP or vinculin-VASP and then attach to the lipid membrane via interactions with membrane-bound proteins and/or lipids. Once on the membrane, the condensates partition β1 tails and G-actin, and actin partitioning results in condensate-driven actin polymerization. The two condensate types exhibit distinct properties: Vinculin-VASP condensates behave more solid-like, whereas zyxin-VASP condensates are fluid. This difference correlates with the actin architecture: Zyxin-VASP condensate fluidity supports the growth of actin bundles, while vinculin-VASP condensates generate single actin filaments. Moreover, the zyxin-VASP network displays myosin-driven condensate fusion and actin clustering. When interpreted in the context of existing models of NA formation and maturation (9, 61), these findings motivate a hypothetical mechanism for NA assembly driven by membrane-assisted phase separation (Fig. 6).

Fig. 6. Hypothetical model for NA assembly and maturation via LLPS.

(A) Integrin activation and stabilization: Talin and kindlin bind to β1 integrin tails at the membrane, promoting integrin activation and stabilizing the active dimeric state. (B) NA formation through LLPS: Multivalent interactions between NA proteins, such as vinculin and VASP, drive condensate formation and association with the membrane. (C) Integrin clustering and active integrin filtering: Activated integrins preferentially partition into the condensate through specific protein interactions, while inactive integrins exchange back into the surrounding membrane, leading to integrin enrichment and clustering. (D) Actin coupling to the NA: The phase-separated NA engages the actin cytoskeleton, promoting actin polymerization and interaction with branched actin networks, facilitated by cross-linkers and myosin engagement. (E) Zyxin-VASP recruitment and actin bundling: Zyxin-VASP condensates are recruited to the NA, where they promote actin filament elongation, bundling, and condensate wetting along actin bundles, supporting NA transition into FA. (F) Force-dependent maturation and stress fiber formation: Contractile forces generated by NMMII drive the fusion of zyxin-VASP–rich adhesion condensates, promoting adhesion growth, reinforcing actin bundles, and facilitating stress fiber formation during maturation toward FAs.

In the first step, activated integrins, together with the PI(4,5)P₂-dependent recruitment of kindlin-2 and talin-1, can render the membrane surface favorable for the formation of early NA condensates (Fig. 6, A and B). Consistent with this hypothesis, we observe robust formation of zyxin-VASP and vinculin-VASP phase-separated condensates that readily attach to the lipid bilayer. In line with prior experimental observations, the initial NA assembly is reported to be independent of myosin II–generated contractile forces and substrate stiffness (22, 50) and is instead likely driven by lipid-protein and protein-protein interactions. Interactions between phase-separating components and the membrane can increase local protein concentrations and promote phase separation at lower bulk concentrations (62). Beyond vinculin and VASP, which are present early in NAs, additional phase-separating proteins—including paxillin, FAK, and p130Cas—are also recruited at the onset of NA formation in the lamellipodium (63–65). Consistent with their early appearance, kindlin, talin, FAK, paxillin, vinculin, and p130Cas have been shown to phase separate on lipid bilayers in vitro (12, 13, 15, 16). In line with this, we find that zyxin-VASP and vinculin-VASP condensates partition FAK, a key signaling factor in FA assembly and maintenance (9). Together, on the basis of our membrane-attached vinculin-VASP condensates and FAK partitioning and on these prior in vitro and cellular observations, we hypothesize that membrane-assisted phase separation can seed early NA condensates under conditions created by activated integrins and the PI(4,5)P₂-dependent recruitment of early NA components.

In the second step, as integrins diffuse within the membrane, activated integrins partition into condensates and are trapped because of specific interactions between NA proteins and activated integrin tails. In our study, this corresponds to the observed partitioning of integrin tails and the slowdown of their dynamics (Fig. 6C). Supporting this framework, it has been demonstrated that integrins are trapped in islands within FAs (66) and integrins within clusters are activated (67). As two primary integrin binders, talin and kindlin were demonstrated to play complementary roles in stabilizing integrin-tail binding (68) and integrin-tail clustering (15). On the basis of our observed tail partitioning and slowed dynamics together with these prior findings, we hypothesize that talin- and kindlin-rich condensates may act as filters for activated integrins, stabilizing and prolonging their partitioning compared to inactive integrins. This hypothesis requires further investigation because the β1 tails used here artificially mimic constitutively active integrin tails.

In the third step, the phase-separated NA establishes a physical linkage to the actin cytoskeleton (Fig. 6D). In our reconstituted system, we find that vinculin-based condensates are inefficient at nucleating new actin filaments but efficiently capture preexisting filaments. This may resemble the formation of an initial actin-NA link, as was suggested before (69). In cells, NAs form in the lamellipodium (50), where an Arp2/3 (actin-related protein 2/3)–generated branched actin network is present (70), and filament capture could therefore facilitate formation of the initial integrin-talin-vinculin-actin linkage of the molecular clutch (10). It was demonstrated that while NA formation is myosin II–independent (22, 50), retrograde flow applies force to the integrins (71). On the basis of our filament-capture behavior and the solid-like condensate properties we observe for vinculin-based condensates, we speculate that these properties may help stabilize early load-bearing connections between NA components and actin. In agreement with the importance of vinculin in force transmission, cells deficient in vinculin display reduced adhesion strength and decreased traction forces (34).

In the fourth step, rising tension is expected to drive the recruitment of additional mechanosensitive components and promote maturation toward FAs (Fig. 6E). Zyxin is often viewed as a later-stage, force-responsive component that is absent from early adhesions (33, 35, 37). In our experiments, however, zyxin-VASP condensates can form and interact with membranes even in the absence of externally applied force, suggesting that local concentration and binding interactions may contribute to its recruitment. Once present, zyxin and VASP promote robust actin polymerization and bundling, and the fluid nature of zyxin-VASP condensates allows them to deform and wet growing actin bundles. Together with the view of zyxin as a later-stage component, we therefore speculate that zyxin-VASP–driven actin remodeling may be more relevant at later stages of adhesion growth when linear actin bundles and stress fibers emerge and reinforce the adhesion (36).

As tension increases, NMMII-driven contraction of cross-linked actin (e.g., by α-actinin) may recruit and concentrate additional zyxin-VASP to phase separate and fuse with existing complexes, thereby promoting adhesion growth (Fig. 6F). In our experiments, zyxin-VASP condensates form a network of separate condensates interconnected through actin filaments and bundles, and upon myosin-driven contraction, separate condensates fuse together to form larger condensate clusters. In cells, FAs display lateral heterogeneity, with paxillin, vinculin, and integrins located in islands embedded in larger FAs, and these islands display slow dynamics with proteins being trapped (72). On the basis of our observation of fusing zyxin-VASP condensates during contraction, we hypothesize that such a lateral structure could arise from smaller, sturdier islands (for example, vinculin-based phase-separated entities) that become connected through incorporation of new phase-separating entities such as zyxin-VASP that fuse with existing phase-separated structures as a result of force application. Mature FAs also exhibit vertical stratification, with integrin-interacting proteins (such as kindlin, talin, and vinculin) in membrane-proximal layers and actin-regulatory proteins (such as zyxin and VASP) in more distal layers (73, 74). Our in vitro system does not recapitulate this three-dimensional architecture. Layer stratification was suggested to result from force application and additional protein interactions with stretched talin (73). One possible speculation is that a rigid, membrane-anchored vinculin-rich layer stabilizes the membrane-proximal interface, while a more fluid zyxin-VASP–rich layer remains dynamically coupled to actin to support filament elongation and bundling, with this separation likely emerging from force-dependent protein interactions such as talin stretching that exposes hidden binding sites for different FA components.

The integrin adhesion–like complex presented in this study does not capture the full complexity of the cellular environment. In particular, it lacks upstream signaling pathways, full-length integrins, and extracellular matrix engagement, all of which are required for force-mediated coupling between the extracellular matrix and the actin cytoskeleton, as well as for integrin clustering. In addition, protein stoichiometry and concentrations in our assays likely deviate from physiological concentrations in cells. These limitations constrain direct extrapolation of our in vitro results to the cellular setting. However, the value of our minimal reconstitution system is that it allows to dissect how phase separation could help organize NAs by locally enriching key components and promoting coupling to actin. In this view, differences in condensate dynamics might generate adhesion heterogeneity and tune actin organization and force transmission during maturation, suggesting that phase separation may be a broader organizing principle of integrin adhesion assembly.

MATERIALS AND METHODS

Materials and reagents

All standard chemicals were of analytical grade and purchased from Sigma-Aldrich or Carl Roth, unless otherwise stated. Pyranose oxidase from Coriolus sp. (P4234), catalase from bovine liver (C9322), creatine phosphokinase from rabbit muscle (C3755), benzonase nuclease (E1014), and creatine phosphate (237911) were purchased from Sigma-Aldrich. Lysozyme (A4972) and deoxyribonuclease I (A3778) were purchased from PanReac AppliChem. 1,2-Dioleoyl-sn-glycero-3-phosphocholine (850375P), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-[4-(p-maleimidomethyl)cyclohexane-carboxamide] (sodium salt) (18:1 PE-MCC) (780201P), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (ammonium salt) (18:1 PEG2000-PE) (880130P), 1,2-dioleoyl-sn-glycero-3phospho-(1′-myo-inositol-4′,5′-bisphosphate) (ammonium salt) [18:1 PI(4,5)P₂] (850155P), and 1-oleoyl-2-(6-((4,4-difluoro-1,3-dimethyl-5-(4-methoxyphenyl)-4-bora-3a,4a-diaza-s-indacene-2-propionyl)amino)hexanoyl)-sn-glycero-3-phosphoinositol-4.5-bisphosphate (ammonium salt) [TMR-PI(4,5)P₂] (810384) were purchased from Avanti Research. cOmplete, EDTA-free Protease Inhibitor Cocktail (4693132001) and cOmplete His-tag Purification Resin (5893682001) were purchased from Roche. HiLoad 16/600 Superdex 200 pg (GE28-9893-35), Superdex 200 Increase 10/300 GL (GE28-990944), HisTrap HP (GE17-5248), HiTrap Desalting Column (GE17-1408), HiTrap Heparin HP (GE17-0407), Superose 6 Increase (GE29-0915-98), HiLoad 26/600 Superdex 200 (GE28-9893-36), and HiTrap Q HP (GE29-0513-25) were purchased from Cytiva. AF555 NHS-ester (A20109), AF532 C5 Maleimide (A10255), and 1,2-dihexadecanoyl-sn-glycerin-3-phosphoethanolamine (Oregon Green DHPE) (O12650) were purchased from Invitrogen. Zeba Spin Desalting Columns, 7000 Da molecular weight cutoff (MWCO) (89882), were purchased from Thermo Fisher Scientific. ATTO488 NHS-ester (FP201-488) was purchased from Jena Bioscience. ATTO647 NHS-ester (AD 647) was purchased from ATTO-TEC. Escherichia coli (DE3) Rosetta competent cells (71400-M) were purchased from Novagen. E. coli BL21 CodonPlus DE3-RIPL competent cells (230280) were purchased from Agilent Technologies. Amicon Ultra Centrifugal Filters were purchased from Millipore. CapZ (nonmuscle, human recombinant) (8322) and vinculin (smooth muscle, turkey) (8307) were purchased from Hypermol. Microplates, PS, 96-well, PS, F-bottom (chimney well) μClear, black, nonbinding (655906), used in pyrene assay were purchased from Greiner Bio-One.

Buffers

KMEI buffer (pH 7.5) contained 50 mM KCl, 1 mM MgCl₂, 1 mM EGTA, and 10 mM imidazole. After incubation of β1 tails on SLBs, KMEI was supplemented with 1 mM dithiothreitol (DTT). For experiments involving zyxin, KMEI and actin polymerization buffers were additionally supplemented with 2.5 μM ZnCl₂ to stabilize zyxin LIM domains (75). Citric acid buffer (pH 4.85) contained 20 mM citric acid, 50 mM KCl, 0.1 mM EDTA, and 0.1 mM NaN₃.

Protein expression, purification, and labeling

ATTO647-labeled N-terminal cysteine β1 integrin cytoplasmic tail peptides were synthesized and purified by the Max Planck Institute of Biochemistry Core Facility, and the peptide identity was verified by high-resolution intact mass spectrometry. See the Supplementary Materials for the peptide sequence.

Rabbit skeletal muscle actin was purified from acetone powder using method previously described in (76). No rabbits were directly used in this study. Monomeric actin was stored at 4°C in G-buffer [2 mM tris (pH 8.0), 0.2 mM adenosine 5′-triphosphate (ATP), 0.2 mM CaCl₂, 0.2 mM DTT, and 0.005% NaN₃]. For fluorescent actin, surface-exposed lysines were labeled with ATTO488 NHS-ester or ATTO647 NHS-ester. For pyrene assays, actin was labeled with N-(1-pyrenyl)iodoacetamide.

Murine His-tagged kindlin-2 was expressed in E. coli Rosetta 2 (DE3) and induced with 0.2 mM isopropyl-β-d-thiogalactopyranoside overnight at 18°C. Cells were harvested by centrifugation and lysed using a constant cell disruption system in immobilized metal affinity chromatography (IMAC) running buffer [25 mM tris (pH 7.8), 500 mM NaCl, and 1 mM tris (2-carboxyethyl)phosphine (TCEP)] supplemented with 2 mM MgCl₂, a spatula tip of deoxyribonuclease I, lysozyme (40 μg/ml), and one tablet of cOmplete, EDTA-free protease inhibitor cocktail. After clarification, the lysate was loaded onto a 5-ml HisTrap HP Ni-NTA IMAC column, washed with IMAC running buffer, and eluted with IMAC elution buffer [25 mM tris (pH 7.8), 500 mM NaCl, 1 mM TCEP, and 500 mM imidazole]. Elution fractions were pooled, concentrated using a 30-kDa molecular weight cutoff Amicon Ultra centrifugal filter, and further purified by size-exclusion chromatography on a HiLoad 16/600 Superdex 200-pg column equilibrated in 25 mM tris (pH 7.8), 200 mM NaCl, 5% (v/v) glycerol, and 1 mM TCEP. All purification steps were performed at 4°C.

Human His-tagged talin-1 was expressed and purified as described earlier (16). Briefly, the protein was expressed in E. coli (DE3) Rosetta 2 using ZYM autoinduction medium. After cell lysis, the protein was captured by Ni-NTA IMAC. The protein was then concentrated and buffer exchanged using a HiTrap Desalting column, purified further on a HiTrap Heparin HP column, and subjected to final size-exclusion chromatography on a Superose 6 Increase column.

Human His-tagged zyxin-mCherry was expressed and purified as described earlier (16). Briefly, the protein was expressed in E. coli BL21 (DE3). After cell lysis, the protein was captured using Ni-NTA IMAC. The protein was further purified by size-exclusion chromatography on a HiLoad 16/600 Superdex 200-pg column.

His-tagged human VASP containing the high-affinity GAB from Dictyostelium discoideum was expressed and purified as described earlier (14). Briefly, the protein was expressed in E. coli BL21 (DE3) and purified by Ni-NTA IMAC followed by size-exclusion chromatography. VASP was labeled with AF532 C5 maleimide, as described in (14).

Mouse profilin-2a was expressed and purified as described earlier (14). Briefly, the protein was expressed in E. coli BL21 (DE3) as a glutathione S-transferase fusion protein and captured on glutathione sepharose. The glutathione S-transferase tag was cleaved, and the protein was dialyzed to remove the cleaved tag.

Turkey smooth muscle vinculin was purchased from Hypermol and labeled with AF555 NHS-ester (vinculin-AF555). The vinculin solution was buffer exchanged into phosphate-buffered saline (PBS) using a Zeba Spin Desalting Column preequilibrated with PBS. AF555 NHS-ester (dissolved in dimethyl sulfoxide) was added at a 2.5-fold molar excess relative to vinculin, and the reaction was incubated for 1 hour at room temperature. Unreacted dye was removed using a Zeba Spin Desalting Column preequilibrated with 50 mM tris-HCl (pH 7.5), 200 mM NaCl, and 1 mM DTT.

Human nonmuscle myosin IIa (NMMII) was purified as described earlier (16). Briefly, human thrombocytes were pelleted and lysed in PBS. Actomyosin was collected by centrifugation and depolymerized. NMMII was precipitated by ammonium sulfate fractionation, clarified, and further purified by size-exclusion chromatography on a HiLoad 26/600 Superdex 200 column. Purified NMMII was precipitated again with ammonium sulfate and stored.

Ca²⁺-independent human α-actinin1b was purified similarly to VASP. Briefly, expression was induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside in E. coli BL21 (DE3) at 26°C for 12 hours. Cells were harvested by centrifugation and resuspended in lysis buffer [30 mM tris-HCl (pH 8.0), 250 mM NaCl, 20 mM imidazole, 5% glycerol (v/v), and 2 mM DTT] supplemented with lysozyme (40 μg/ml), 0.5 tablet cOmplete protease inhibitor, and 5 μl of benzonase. Cells were lysed, and the lysate was clarified by centrifugation. The supernatant was filtered (0.2 μm) and loaded onto a HisTrap FF column; washed with lysis buffer, lysis buffer containing 300 mM NaCl, and lysis buffer; and eluted with lysis buffer containing 500 mM imidazole. Elution fractions were concentrated, clarified (15,000 rpm, 10 min), and further purified by size-exclusion chromatography on a Superdex 200 Increase 10/300 column equilibrated in 20 mM TES [N-tris(hydroxymethyl)methyl-2-aminoethanesulfonic acid]-HCl (pH 8.0), 150 mM KCl, and 1 mM DTT.

Proteins derived from different species were used in the reconstitution assays together as the protein function and interfaces involved in actin regulation are highly conserved. His tags were used for purification purpose only.

Small unilamellar vesicle (SUV) production

A lipid mixture containing 87.5 mol % of 1,2-dioleoyl-sn-glycero-3-phosphocholine, 5.0 mol % of PI(4,5)P₂, 5.0 mol % of 18:1 PE-MCC, and 2.5 mol % of PEG2000-PE was prepared in chloroform. Lipid films were formed by drying the mixture under a gentle stream of nitrogen, followed by vacuum desiccation for at least 2 hours. Films were then hydrated in citric acid buffer (77) and resuspended by sonication for 30 min to a final total lipid concentration of 0.5 mM. Small unilamellar vesicles (SUVs) were produced by extruding the lipid suspension 20 times through 100-nm-pore-size membrane filters using a miniextruder set. SUVs were stored at 4°C and used within 72 hours. For membrane fluidity control experiments, vesicles containing 0.05 mol % of TMR-PI(4,5)P₂ were prepared (fig. S9) and analyzed using FRAP.

Flow chamber preparation

Flow chambers (final volume of ~40 μl) were assembled by sealing 22- by 22-mm coverslips to 25- by 75-mm microscope slides using three layers of parafilm as spacers. Coverslips were sonicated for 30 min in 3 M NaOH, followed by 30 min in 2% (v/v) Hellmanex solution and rinsed thoroughly with Milli-Q water. They were then treated with piranha solution (2:1 H₂SO₄:H₂O₂) for 10 min to render the surface hydrophilic, rinsed extensively with Milli-Q water, and stored in Milli-Q water until use. Microscope slides were sonicated for 30 min in 2% (v/v) Hellmanex solution, rinsed with Milli-Q water, and stored in ethanol. Coverslips and slides were used within 1 week of cleaning.

Model membrane system reconstitution

SLBs were formed in flow chambers by adding SUVs to a final lipid concentration of 0.167 mM and allowing them to rupture on the glass surface for 20 min at room temperature. The resulting SLBs were washed with 2 × 0.8 ml of citric acid buffer (77), followed by 2 × 0.8 ml of KMEI buffer to remove excess vesicles. SLBs were then incubated with 10 μM β1-ATTO647 integrin tail containing terminal cysteine in KMEI buffer for 20 min and washed with 2 × 0.8 ml of KMEI buffer. During this step, the terminal cysteine reacts with the maleimide group of 18:1 PE-MCC, yielding covalent attachment of β1-ATTO647 to the membrane; unreacted β1-ATTO647 was removed by washing, leaving only membrane-bound β1-ATTO647. β1-functionalized SLBs were incubated with 1 μM talin-1 and 1 μM kindlin-2 in KMEI buffer for 20 min, followed by washing with 2 × 0.8 ml of KMEI buffer. Individual proteins, or defined protein combinations, were subsequently introduced and incubated for 10 min before initiating actin polymerization. Actin polymerization was performed in KMEI buffer containing 50 nM CapZ, 13 μM profilin, 2 μM G-actin (12.5% ATTO488-labeled), 1 mM DTT, and 1 mM ATP. For the condensate eccentricity assay, actin was polymerized in KMEI buffer containing 10 nM CapZ, 13 μM profilin, 1 mM DTT, and 1 mM ATP. For actin-network contractility experiments, actin was polymerized in KMEI buffer containing 10 nM CapZ, 13 μM profilin, 1 mM DTT, and 1 mM ATP. Contractility was then induced by introducing 0.2 μM α-actinin1b, 0.1 μM NMMII, 0.1 mM ATP, 0.4% methylcellulose, 1 mM DTT, and an ATP regeneration system [creatine phosphokinase (1200 U/ml) and 1 mM creatine phosphate]. In all measurements, an oxygen-scavenging system [catalase (1700 U/ml), pyranose oxidase (26 U/ml), and 36 mM glucose] was included for photoprotection. Imaging of the chamber was started immediately afterward.

Pyrene assay

Actin polymerization kinetics was quantified using a pyrene-actin fluorescence assay under the following conditions: (i) actin alone, (ii) actin +50 nM CapZ +13 μM profilin, (iii) actin +50 nM CapZ +13 μM profilin +0.5 μM zyxin-mCherry +0.5 μM VASP, and (iv) actin +50 nM CapZ +13 μM profilin +0.5 μM vinculin-AF555 + 0.5 μM VASP. Protein components were diluted in KMEI buffer supplemented with 1 mM DTT and 1 mM ATP, and reactions were assembled in Greiner black, clear-bottom 96-well plates. Polymerization was initiated by addition of 2 μM G-actin (20% pyrene-labeled). Pyrene fluorescence was monitored at 25°C using a SpectraMax iD3 plate reader controlled with SoftMax Pro 7.1 (Molecular Devices). Kinetic traces were recorded for 1 hour at 20-s intervals (excitation, 365 nm; emission, 407 nm) using medium photomultiplier tube sensitivity and a 200-ms integration time.

Imaging and data acquisition

Fluorescence imaging was performed on a Leica DMi8 microscope equipped with an HC PL APO 100×/1.47–numerical aperture oil-immersion objective and an ORCA-Flash4.0 complementary metal-oxide semiconductor camera (C13440-20CU; Hamamatsu). Time-lapse images were acquired continuously to monitor membrane, condensate, and actin dynamics. Dynamics of β1 integrin tails, zyxin-mCherry, vinculin-AF555, and VASP-AF532 were recorded by epifluorescence microscopy. Actin polymerization (ATTO488-actin) was imaged by total internal reflection fluorescence (TIRF) microscopy using the same 100×/1.47–numerical aperture oil-immersion objective. FRAP experiments were performed using a Leica Infinity Scanner unit.

Image analysis

Acquired images were analyzed in Fiji/ImageJ (78). Segmentation masks for condensate size distributions, condensate counts, and eccentricity measurements were generated in ilastik (79). For condensate area and eccentricity analyses, only condensates larger than the effective resolution limit of the microscope were included. Eccentricity was defined as $e=\sqrt{1-(b^2/a^2)}$, where $a$ is the major axis, and $b$ is the minor axis. Eccentricity was quantified before polymerization and again 30 min after polymerization onset. The actin bundle line density was quantified 30 min after polymerization by counting the number of intensity peaks along a horizontal line region of interest spanning the field of view. Before peak detection, images were preprocessed by thresholding intensity values to retain bundled structures while excluding single filaments. Data visualization and plotting were performed in Python 3.

FRAP experiments

FRAP was performed by bleaching a circular region (5-μm diameter) at a 100% laser power, followed by acquisition of fluorescence images for 80 s. A region of the SLB outside the bleached area was used for background subtraction. Fluorescence intensities within the bleached region of interest were exported using LAS X software (Leica), normalized to the prebleach intensity, and fitted with a double-component exponential recovery function: $F\left(t\right)=y_0+A_{\text{fast}}·e^{-t∕{\mathrm{\tau }}_{\text{fast}}}+A_{\text{slow}}·e^{-t∕{\mathrm{\tau }}_{\text{slow}}}$, where $F\left(t\right)$ is the relative fluorescence intensity over time, $\frac{-(A_{\text{fast}}+A_{\text{slow}})}{1-(y_0+A_{\text{fast}}+A_{\text{slow}})}$ is the mobile fraction, ${\mathrm{\tau }}_{\text{fast}}$ is the fast recovery time constant, and ${\mathrm{\tau }}_{\text{slow}}$ is the slow recovery time constant. $t_{1/2\_\text{fast}}={\mathrm{\tau }}_{\text{fast}}·\text{ln}2$ and $t_{1/2\_\text{slow}}={\mathrm{\tau }}_{\text{slow}}·\text{ln}2$ are the fast and slow half recovery times, respectively. Model fitting and analysis were performed in Python 3.

Statistical analysis

Differences in the fraction of filaments originating from condensates (Fig. 2G), the effects of membrane and membrane-bound components on condensate attachment (Fig. 3, B and C), and the line density of actin bundles (Fig. 4E) were assessed using the Mann-Whitney U test. Analyses were performed in Python using SciPy (version 1.11.1) with two-tailed tests, and statistical significance was defined as P < 0.05.

Supplementary Materials

The PDF file includes:

Supplementary Methods

Figs. S1 to S9

Legends for movies S1 to S16

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S16