Surface-induced phase separation of reconstituted nascent integrin clusters on lipid membranes

Significance

Integrin adhesion complexes assembled at the inner leaflet of the plasma membrane mediate cell adhesion to the extracellular matrix proteins and activate a range of signaling pathways. Kindlin, talin, paxillin, and focal adhesion kinase are among the “early” focal adhesion proteins assembled in nascent adhesions. While liquid–liquid phase separation has been proposed as a mechanism for inducing the assembly of nascent adhesions, how membrane composition shapes the phase separation to occur at physiological conditions remains unexplored. Here, we demonstrate that phosphoinositides containing lipid membranes induce minimal nascent adhesion condensates. Furthermore, we show that membrane surfaces set the effective local solvent condition and the stability of multivalent protein interactions. This effect enables specific membrane-associated phase separations necessary for adhesion complex formation.

Abstract

Integrin adhesion complexes are essential membrane-associated cellular compartments for metazoan life. The formation of initial integrin adhesion complexes is a dynamic process involving focal adhesion proteins assembled at the integrin cytoplasmic tails and the inner leaflet of the plasma membrane. The weak multivalent protein interactions within the complex and with the plasma membrane suggest that liquid–liquid phase separation could play a role in the nascent adhesion assembly. Here, we report that solid-supported lipid membranes supplemented with phosphoinositides induce the phase separation of minimal integrin adhesion condensates composed of integrin β1 tails, kindlin, talin, paxillin, and FAK at physiological ionic strengths and protein concentrations. We show that the presence of phosphoinositides is key to enriching kindlin and talin on the lipid membrane, which is necessary to further induce the phase separation of paxillin and FAK at the membrane. Our data demonstrate that lipid membrane surfaces set the local solvent conditions for steering the membrane-localized phase separation even in a regime where no condensate formation of proteins occurs in bulk solution.

Integrin-containing multiprotein complexes mediate adhesion between cells and the extracellular matrix (1, 2). Integrins are heterodimers consisting of α and β subunits. They have large ectodomains, which mediate ligand binding, single-span transmembrane domains, and short cytoplasmic tails, which indirectly associate with the contractile actomyosin cytoskeleton (3, 4). Nascent adhesions are the first integrin adhesion complexes that assemble at the leading edge of protruding membranes. Integrin-associated proteins such as kindlin, talin, paxillin, and focal adhesion kinase (FAK) are present in and regulate the assembly of nascent adhesions while rapidly exchanging between the integrin adhesion complexes and the cytoplasm (5–8). Protein interactions have been identified between these integrin-associated proteins such that kindlin interacts Zn²⁺-dependently with paxillin (8–10), and FAK interacts directly with paxillin, kindlin, and talin (11–13). Despite all detailed insights into these molecular interactions, the physical mechanism of the nascent adhesion assembly is still not fully understood. Nevertheless, the weak multivalent protein interactions suggest that phase separation could play a role.

Liquid–liquid phase separation driven by such weak interactions between multivalent molecules has appeared as an important mechanism to facilitate the formation of biomolecular condensates (14–16). Indeed, kindlin, paxillin, and FAK have been shown recently to spontaneously phase separate in bulk and be bound by integrin tails to phosphoinositides-free membranes (17). Such biomolecular condensates assemble in bulk when the bulk concentration exceeds a critical concentration or upon change of solvent conditions, thereby partitioning into dense and dilute phases of the proteins (18). These condensates often exhibit properties of liquid-like droplets and dynamically exchange molecular constituents with the surrounding environment (14, 15). Bulk-formed condensates can be subsequently localized to lipid membranes via transmembrane and membrane-associated proteins, further inducing their condensation (17, 19–23). In vitro reconstitutions have shown that the concentration threshold at which phase-separated condensates form when associated with membranes can be lower than in bulk solutions (24, 25). The association with membranes is critical as many in vitro condensates form in bulk only at high protein concentrations, low concentrations of salt, and often only in the presence of a crowding agent. However, the mechanism of how membrane composition shapes the binodal decomposition to occur at more physiological conditions remains unexplored. The important role of phosphoinositides in cellular adhesion (26–28) triggered us to evaluate their role in the formation of membrane-associated nascent-like integrin adhesion condensates.

Here, we demonstrate that phosphoinositides containing lipid membranes induce minimal nascent integrin adhesion condensates composed of β1 tails, kindlin, talin, paxillin, and FAK at high ionic strengths and nanomolar protein concentrations, mimicking physiological conditions. We show that the presence of phosphoinositides is key to enriching kindlin and talin on the membrane, forming first complexes with β1 tails, which are necessary to nucleate condensates. Our results show that membrane surfaces set the effective local solvent condition and enable specific membrane-associated phase separations.

PIP₂-Dependent Emergence of Integrin Clusters on Membranes

We measure the solution turbidity of four focal adhesion proteins at high (1 μM kindlin, 1 μM talin, 1 μM paxillin, and 200 nM FAK) and low (0.2 μM kindlin, 0.2 μM talin, 0.2 μM paxillin, and 40 nM FAK) concentrations (C_Proteins) with different ionic strengths to examine the spontaneous occurrence of phase separation in bulk solution (Fig. 1A). At high C_Proteins, the observation of solution turbidity indicates the formation of protein condensates. The turbidity decreases with increasing KCl concentration, while no turbidity is observed at low C_Proteins (Fig. 1A). The concentration- and KCl-dependence of solution turbidity confirms that the phase separation of the tested set of focal adhesion proteins is driven by protein concentrations and electrostatic interactions in bulk solutions.

Fig. 1.

PIP₂ enhances β1 integrin clustering on membranes. (A) Solution turbidity measurements with high (1 μM kindlin, 1 μM talin, 1 μM paxillin, and 200 nM FAK) and low (0.2 μM kindlin, 0.2 μM talin, 0.2 μM paxillin, and 40 nM FAK) proteins concentrations (C_Proteins) as a function of KCl concentration in the buffer. Error bars represent the standard deviations from five measurements. (B) Schematic of the reconstituted membranes containing 0% and 5% PIP₂. (C and D) Epifluorescence images of β1 after 90-min incubation with low and high C_Proteins; 50 and 150 mM of KCl on 0% PIP₂ (C) and 5% PIP₂ (D) membranes. (E) Time series of the formation of β1 clusters with high C_Proteins at high ionic strength (150 mM KCl) on 0% PIP₂ (Left) and 5% PIP₂ (Right) membranes. (F) Fluorescence recovery after photobleaching (FRAP) images of a β1 cluster formed with high C_Proteins at high ionic strength (150 mM KCl) on 0% PIP₂ (Top) and 5% PIP₂ (Bottom) membranes. (G) The density of β1 clusters formed after 90-min incubation with high C_Proteins on lipid membranes as indicated. Each condition contains 20 measurements. Box plots indicate the median (line); 25th and 75th percentiles (box); 0th and 100th percentiles (whiskers). (H) Distributions of β1 cluster area formed after 90-min incubation with high C_Proteins at high ionic strength (150 mM KCl) on 0% PIP₂ (black, bin width = 0.5 μm²) and 5% PIP₂ (red, bin width = 0.5 μm²) membranes. (Scale bars, 10 μm).

As β integrin cytoplasmic tail represents the minimal domain required to examine kindlin- or talin-dependent integrin clustering (29, 30), we reconstitute 0 and 5 mol% phosphatidylinositol 4,5-bisphosphate (PIP₂) supported lipid bilayer membranes with attached β1 integrin tails (β1) (Materials and Methods, Fig. 1B). After incubating with the kindlin, talin, paxillin, and FAK at high C_Proteins, β1 clusters emerge on both 0 and 5% PIP₂ membranes with both low ionic strength (50 mM KCl) and physiological ionic strength (150 mM KCl) (Fig. 1 C and D). This suggests that the protein condensates formed at high C_Proteins settle on the membrane and are coupled to the β1 through kindlin and talin (17), which promotes β1 clustering on membranes (Fig. 1E). Fluorescence recovery after photobleaching (FRAP) measurements demonstrate that β1 clusters are dynamic assemblies in which β1 is free to diffuse and exchange on the membrane (Fig. 1F). While β1 clusters emerge at high C_Proteins, the density of β1 clusters depends on the ionic strength and PIP₂ (Fig. 1G). In the absence of PIP₂, the β1 cluster density decreases two-fold in the presence of 150 mM KCl compared to 50 mM KCl. This difference suggests that β1 clustering on membranes lacking PIP₂ is regulated by the ionic strength in bulk solutions, which controls the phase separation of tested focal adhesion proteins. In sharp contrast, the β1 cluster density on 5% PIP₂ membranes is not influenced by the ionic strength up to 150 mM KCl and increases five-fold compared to the one on membranes lacking PIP₂ with high ionic strength. The β1 cluster density on 5% PIP₂ membranes only decreases at higher nonphysiological ionic strengths (SI Appendix, Fig. S1). In the absence of attached β1, protein condensates also form with labeled kindlin or FAK on lipid membranes at high protein concentrations with the same dependence on PIP₂ concentration but at a lower density (SI Appendix, Figs. S2 and S3). Altogether, our findings indicate that β1 tail and the lipid membrane composition enhance the binding of the preformed protein condensates to the membrane, with the membrane composition being the dominating factor.

The increase in the density of β1 clusters on PIP₂-supplemented lipid membranes indicates that PIP₂ facilitates the recruitment of protein condensates to lipid membranes. In the presence of PIP₂, the density of β1 clusters on lipid membranes at low and high ionic strength is similar, indicating that β1 clustering and β1-tail-associated condensates binding are determined by the lipid membrane surface rather than merely by bulk solution conditions. Small β1 clusters are observed more frequently on 5% PIP₂ membranes than on 0% PIP₂ membranes at high C_Proteins and ionic strength (Fig. 1H). While PIP₂ enhances β1 clustering on lipid membranes, such β1 clusters still depend on the preformed proteins condensates in bulk. Such preformed protein condensates at high protein concentrations do not necessarily capture the formation of nascent adhesions with low, endogenous protein concentrations at high salt concentrations, as found in cells (31). To mimic these physiological conditions, we investigate next the pairwise interaction between β1 and kindlin, talin, paxillin, and FAK on membranes at high ionic strength (150 mM KCl).

Diffusion of β1 Slows Down in the Presence of β1-Binding Proteins

To address the individual role of kindlin, talin, paxillin, and FAK mixtures for β1 clustering, we next incubate each protein separately with the model membranes (Fig. 2A). Since β1 distributes and diffuses uniformly on the fluid membrane, the diffusion of β1 reflects the interaction between β1, the membrane (32), and focal adhesion proteins. The diffusion of β1 is determined by FRAP measurements (Materials and Methods, Fig. 2 B and C), with which we compute the half-recovery time. A longer half-recovery time corresponds to slower diffusion of the β1 tails on the lipid membrane (Materials and Methods, Fig. 2 D and E). The diffusion of uniformly distributed β1 on the membrane depends on the incubated focal adhesion proteins and PIP₂ (Fig. 2F and SI Appendix, Figs. S4 and S5). In the absence of the focal adhesion proteins, β1 diffusion remains unaffected by PIP₂, excluding a potential interaction of the β1 tail with the PIP₂-containing lipid membrane. While incubation with 0.2 μM FAK or 1 μM paxillin does not affect the diffusion of β1 tails on the lipid membrane, the incubation of 1 μM kindlin, or 1 μM talin slows down the diffusion of β1 tails.

Fig. 2.

Diffusion of β1 integrin tails on membranes. (A) Schematic of the incubation and flush experiment steps. (B and C) FRAP images of β1 (Top) and β1 after 30 min incubation with 1 μM kindlin and 1 μM talin (Bottom) on 0% PIP₂ (B) and 5% PIP₂ (C) membranes. (Scale bars, 10 μm.) (D and E) Normalized FRAP measurements of β1 on 0% PIP₂ (D) and 5% PIP₂ (E) membranes. Solid curves show the double component exponential recovery fittings. (F) Diffusion of β1 (t_{1/2_fast}) after 30-min incubation with different integrin-associated proteins on 0% PIP₂ and 5% PIP₂ membranes. (G and H) Diffusion of β1 as a function of kindlin concentration after 30-min incubation and after flush on 0% PIP₂ (G) and 5% PIP₂ (H) membranes. (I and J) Diffusion of β1 as a function of talin concentration after 30-min incubation and after flush on 0% PIP₂ (I) and 5% PIP₂ (J) membranes. (K) Diffusion of β1 after 30-min incubation with 1 μM kindlin and 1 μM talin, and after flush on 0% PIP₂ and 5% PIP₂ membranes. Error bars in (F–K) represent the SDs from 15 measurements. (L) Schematic of β1 diffusion on 5% PIP₂ membranes decreases with increasing β1-binding proteins.

This slowdown of β1 tail diffusion points to the binding of β1 tails to kindlin and talin at the membrane. Furthermore, this slowdown is enhanced in the presence of PIP₂-supplemented lipid membranes (Fig. 2F). On 5% PIP₂ membranes, β1 tails incubated with 1 μM kindlin diffuses 2.6 times slower than β1 tails alone, and compared to 0% PIP₂ membranes, the diffusion of β1 tails is reduced by 1.5 times in the presence of 1 μM kindlin. Upon incubation of 1 μM kindlin and 1 μM talin, β1 tails diffuse 3.6 times slower, and the mobile fraction of β1 tails decreases to 0.83 on 5% PIP₂ membranes (Fig. 2C, E, and F), suggesting that ternary β1–kindlin–talin complexes form on PIP₂-containing membrane. It has been shown that slower molecule diffusion on membranes leads to less cluster growth for RNA–protein complexes (25). Similarly, the slower diffusion of β1 when kindlin and talin are present on 5% PIP₂ compared with 0% PIP₂ membranes (Fig. 2D–F) would lead to the smaller β1 clusters observed on 5% PIP₂ membranes with the same protein mixture (Fig. 1H). This indicates that the coarsening is also slowed down. The PIP₂-dependent slowdown of β1 tail diffusion on the membrane is the consequence of kindlin and talin binding to the phosphoinositide in lipid membranes via their PH and FERM domains, respectively (26, 33, 34). In the presence of PIP₂, kindlin and talin are recruited to the lipid membrane, effectively increasing their local concentration and binding probability to β1 tails. Conversely, despite the ability of FAK to bind PIP₂ (27, 28), when incubated with FAK alone or FAK and paxillin, the diffusion of β1 remains unaltered due to no direct binding of FAK to β1 tails (Fig. 2F).

To further estimate the off-rate of kindlin (Fig. 2 G and H) and talin (Fig. 2 I and J) to β1 tails, we compare the diffusion of β1 tails after 30 min incubation and then after flushing with buffer. In the first set of experiments, 0.1 to 2 μM of kindlin or talin is incubated individually with the model membranes. On 0% PIP₂ membranes, the diffusion of β1 tails after flushing is approximately the same as the diffusion of β1 tails alone, indicating that buffer flushing removes most kindlin or talin from the solution. In contrast, on 5% PIP₂ membranes, the diffusion of β1 tails after flushing only decreases slightly and is slower than the diffusion of β1 tails alone. Therefore, the recruitment of kindlin and talin to 5% PIP₂ membranes is higher than 0% PIP₂ membranes (SI Appendix, Fig. S6). Overall, on 5% PIP₂ membranes, all diffusion times of β1 incubated with kindlin or talin are significantly longer than on 0% PIP₂ membranes. This indicates that already kindlin, as well as talin alone, can form stable complexes with β1 on the 5% PIP₂ membranes. The diffusion times remain constant over two hours, pointing to the high stability of the complexes (SI Appendix, Fig. S7). The simultaneous incubation with kindlin and talin leads to the formation of larger complexes with β1 tails in the presence of PIP₂, as can be seen by the increased diffusion time after flushing (Fig. 2K). These complexes lead to 3.2 times slower diffusion of β1 tails compared to the diffusion of β1 tails in the absence of kindlin or talin on 5% PIP₂ membranes (Fig. 2K). Also, the dimerization of kindlin and talin (30, 35) could play a role in the formation of the β1 complexes. Nonetheless, these complexes are still too small and mobile to be resolved optically. The formation of β1-kindlin, β1-talin, and β1-kindlin-talin complexes on 5% PIP₂ membranes (Fig. 2L) prompts us to explore how PIP₂ membrane can induce the recruitment of paxillin and FAK from the solution to the membrane to form minimal integrin adhesion complexes.

Protein Assembling Sequence Determines Integrin Clustering on Membranes

So far, there is no evidence in the literature that paxillin or FAK are able to directly bind β1 tails, which is in line with our measurements from above. Yet, paxillin and FAK are associated with a preassembled multiprotein building block for the adhesion sites in the cytoplasm (36) and interact with both kindlin and talin (8–11, 13). To test whether the interaction of FAK and paxillin with lipids is sufficient to induce condensate formation on the surface of the membrane only, we incubate lipid membranes with 200 nM paxillin and 100 nM FAK at high ionic strength (150 mM KCl). In bulk, we observe by turbidity measurements that paxillin and FAK are unable to form aggregates by phase separation (Fig. 3A). Only at low ionic strength (50 mM KCl), protein condensates of paxillin and FAK form in bulk (Fig. 3A).

Fig. 3.

Protein assembling sequence regulates the emergence of β1 clusters. (A) Solution turbidity measurements with 200 nM paxillin and 100 nM FAK as a function of KCl concentration in the buffer. Error bars represent the standard deviations from five measurements. (B) Different incubation procedures lead to the emergence or absence of β1 clusters after 90-min incubation at high ionic strength (150 mM KCl) on 5% PIP₂ membranes. (C) Time series of the formation of β1 clusters with the III procedure in (B). (D) The density of β1 clusters formed after 90-min incubation with the three procedures in (B). Each condition contains 20 measurements. Box plots indicate the median (line); 25th and 75th percentiles (box); 0th and 100th percentiles (whiskers). (E) FRAP images of a β1 cluster formed with the III procedure in (B). (Scale bars, 10 μm).

To determine whether the condensate formation of paxillin and FAK can be induced on membrane surfaces, we reconstitute the β1-kindlin, β1-talin, and β1-kindlin-talin complexes on 5% PIP₂ membranes via incubation and following flush steps (Fig. 3B). After the incubation of 200 nM paxillin and 100 nM FAK at a high ionic strength of 150 mM KCl, β1 tail clusters emerge spontaneously only on lipid membranes containing ternary β1-kindlin-talin complexes (Fig. 3C). The emergence of β1 tail clusters suggests that the phase separation of paxillin and FAK is induced by the local biochemical environment of the membrane and gives rise to β1 tail clustering, even in conditions where no spontaneous phase separation occurs in bulk. Barely any β1 tail cluster appears on lipid membranes with either β1-kindlin or β1-talin complexes alone (Fig. 3 B and D). Additionally, no condensates are observed on β1-free membranes with labeled FAK at this concentration (200 nM paxillin and 100 nM FAK). FRAP measurements confirm that β1 clusters induced by such assembling sequence are dynamic assemblies in which β1 tails are free to diffuse and exchange on the lipid membrane (Fig. 3E). Our observations show that both kindlin and talin need to be assembled on the membrane before paxillin and FAK to promote phase separation on the lipid membrane under high conditions.

Discussion

While the function of talin and kindlin for integrin activation, adhesion, and integrin-dependent signaling is firmly established, the initiation of the adhesion complex and the role of lipid membrane compositions are less clear. To clarify this issue, we established an in vitro reconstitution essay based on solid-supported lipid membranes. We demonstrate the spontaneous formation of nascent integrin clusters at 150 mM KCl and nanomolar protein concentrations at lipid membranes, close to reported physiological conditions (31), without the need for any crowding agents. The presence of PIP₂ in the lipid membrane is required to recruit kindlin and talin to the lipid membrane. The binding correlates with a slow down of β1 tails diffusion within the lipid membrane. The complex formation of both proteins with β1 tails on the lipid membrane induces the recruitment and phase separation of paxillin and FAK. The phase separation occurs spontaneously and solely at the membrane. Under these conditions, no spontaneous phase separations in the bulk solution occur. Thus, the presence of PIP₂ in the lipid membrane is crucial for the formation and localization of the reconstituted minimal adhesion complexes.

This is in accordance with the observation in cells, where phosphoinositides binding motifs in kindlin and talin are required to form focal adhesion sites (13, 37). Furthermore, PIP₂ prompts the recruitment of kindlin and talin from the cytosol to the plasma membrane (26, 33, 34). Following PIP₂ binding, talin undergoes a conformational change, which seems critically important for integrin activation in fibroblasts(13). Without PIP₂ binding domain, talin does not bind to the plasma membrane, and no subsequent adhesion complex formation occurs (13). In addition, the phosphoinositides binding motif in kindlin is essential for kindlin recruitment to focal adhesions sites and subsequent cell spreading (37). The synergistic effect of kindlin and talin reported for cell adhesion (2, 8, 13, 37) is confirmed by our in vitro reconstitution with minimal focal adhesion molecules. In the reconstituted in vitro system described here, the PIP₂-induced membrane localization of kindlin and talin further recruits paxillin and FAK from the bulk solution. This suggests that the local PIP₂ concentration in the plasma membrane triggers the membrane localization of adhesion complex formation by activating and recruiting key focal adhesion proteins from the cytosol to the plasma membrane. This mechanism ensures exclusive localization and formation of the multiprotein condensate at the PIP₂-enriched sites of the plasma membrane rather than within the cytosol. It still needs to be determined how the exact membrane composition in vivo regulates the emergence of integrin β1 tail clusters. In vivo, it has been shown that in the hierarchical structure of adhesion sites (36, 38), paxillin and FAK bind only indirectly to the integrin tails via kindlin and talin (8–11, 13), which perfectly agrees with the sequential binding of the involved proteins reported in our current study.

Nascent adhesions typically consist of several tens of integrins and promote the early adhesion of cells to the extracellular matrix (7, 39). They are also associated with early-stage biochemical signaling and the mechano-response, serving as platforms for the recruitment and activation of proteins to induce membrane protrusions and build mature focal adhesions (5, 40, 41). In cells, integrin activation is believed to precede nascent adhesion formation. It has been shown that integrin activation commences with kindlin binding to the β-integrin tail, followed by talin binding (6). Talin can recruit PIPKlγ, which locally produces high levels of PIP₂ (42), further recruiting kindlin and talin that induce the assembly of a nascent adhesion. Our in vitro reconstitution on PIP₂ membranes captures the formation of nascent adhesions with low protein concentrations, similar to their endogenous levels found in cells (31). The formation time scales with the focal adhesion proteins analyzed in this study are relatively slow, indicating that additional focal adhesion proteins accelerate the formation of nascent adhesion complexes in vivo. For instance, additional focal adhesion proteins such as p130CAS can promote phase separation and accelerate the formation of adhesion complexes in vivo (17). The presented approach is ideally suited to address the intricate balance between kinetics and membrane-induced phase separation. Although nascent adhesions are sufficient for cell attachment, their efficient linkage to actomyosin and subsequent growth is a prerequisite for firm anchorage to substrates, cell spreading, and locomotion (2, 8). This linkage to actomyosin needs to be addressed in future studies by including the actin cytoskeletal proteins with the solid-supported membrane assay presented here.

It has been shown that protein phase separation in solution depends on the concentration of multivalent interacting proteins and the bulk solution conditions such as ionic strength, pH, and temperature (Fig. 4). Importantly, complex condensates containing dozens or hundreds of components span a multidimensional phase space with a strong interdependency: The phase boundary for one component depends on the concentration of the other components (43). Thus, phase separation can either be induced by a modification in solvent conditions or by a change in the concentration of the components. Here, we demonstrate that the composite surface of lipid membranes can drive the emergence of membrane-associated condensates (Fig. 4). Membrane surfaces effectively set the local solvent conditions for protein solutions. This offers a unique strategy to regulate a solution’s stability locally by controlling lipid composition. Such surface-induced phase separation has been reported in studies on the phase stability of polymer mixtures (44–46). The presence of a surface may alter the phase stability by breaking the translational and rotational symmetry of the components in the system. Consequently, the preferential attraction of the surface for one of the components in a binary polymer mixture can lead to the phase separation of the system (44, 45).

Fig. 4.

Membrane surfaces steer the phase separation. Schematic of the phase diagrams in the absence and presence of membrane surfaces.

Localizing components of phase-separating systems to membranes enables cells to take advantage of local concentration enhancements to drive the condensate formation and function in specific locations at membranes. This interplay of local membrane composition, signaling via lipid or protein phosphorylation, and sequential protein binding set a framework for understanding the control of adhesion complex formation. The intricate balance of these different pathways ensures the precise control of the formation of nascent adhesion sites, which needs further exploration.

Materials and Methods

Materials and Reagents.

Isopropyl-β-D-thiogalactopyranoside (IPTG), tris, tris(2-carboxyethyl)phosphine (TCEP), hydrogen peroxide (H₂O₂), sulfuric acid (H₂SO₄), imidazole, sodium hydroxide (NaOH), 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), potassium chloride (KCl), zinc chloride (ZnCl₂) sodium azide (NaN₃), citric acid, ethylenediaminetetraacetic acid (EDTA), Atto 647 NHS-ester, dithiothreitol (DTT), glucose, pyranose-oxidase (PO), and catalase (C) were purchased from Sigma-Aldrich. 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] (nickel salt) (DGS-NTA-Ni), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (ammonium salt) (PEG2000-PE), and 1,2-dioleoyl-sn-glycero-3-phospho-(1’-myo-inositol-4’,5’-bisphosphate) (ammonium salt) (PI(4,5)P₂) were purchased from Avanti Polar Lipids. 1,2-Dihexadecanoyl-sn-Glycerin-3-Phosphoethanolamin (Oregon Green DHPE) was purchased from Thermo Fisher Scientific.

Buffers.

HEPES buffer (pH 7.4) contains 50 mM HEPES, 1 mM DTT, 2.5 μM ZnCl₂, and 50/100/150 mM KCl. The buffer contains 2.5 μM ZnCl₂ to ensure the binding of kindlin and paxillin (8–10). Citric buffer (pH 4.85) contains 20 mM citric acid, 50 mM KCl, 0.1 mM EDTA, and 0.1 mM NaN₃.

Protein Expression and Purification.

Atto643-labeled his₆-β1 integrin cytoplasmic tail peptides were synthesized and purified by the Max Planck Institute of Biochemistry (MPIB) Core Facility. The correct peptide sequence was controlled by high-resolution intact mass spectrometry.

Kindlin-2 and paxillin were expressed and purified as described earlier (9). Briefly, an N-terminal his-sumo tag was added to express the proteins soluble in E. coli (DE3) Rosetta 2. After cell lysis, the proteins were captured with a Ni-NTA-immobilized metal ion affinity chromatography (IMAC); the sumo tag was removed by SenP2 digest and subsequent pull-down with Ni-NTA beads followed by a final size exclusion chromatography.

His-tagged talin-1 was expressed and purified as described earlier (47) with minor changes. The protein was expressed in E. coli (DE3) Rosetta 2 upon induction with 0.2 mM IPTG overnight at 18 °C. The cells were harvested by centrifugation and lysed by sonication in IMAC running buffer (25 mM Tris (pH 7.8), 500 mM NaCl, and 1 mM TCEP) supplemented with 2 mM MgCl₂, 40 μL DNaseI (obtained from MPIB Core Facility), and a spatula tip of lysozyme. After clearing the lysate, the sample was loaded on a 5-mL HisTrap HP Ni-NTA IMAC column (17524801; Cytiva), washed with IMAC running buffer, and eluted with a step-wise gradient 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, and further purified with a Superose 6 Increase 10/300 GL size exclusion chromatography (29091596, Cytiva) using 20 mM Tris, pH 7.8, 200 mM NaCl, and 1 mM TCEP as running buffer.

Murine focal adhesion kinase carrying an N-terminal his-sumo tag was cloned using Gateway cloning into PB-T-Rfa to establish a stable FAK-expressing HEK293T cell line (48). The cells were cultured in FreeStyle 293 Expression Medium (12338018, Gibco) until reaching a cell density of 10⁶ cells per mL before inducing protein expression by doxycycline addition (1 μg/mL) for three days. Cells were collected by centrifugation, resuspended in IMAC running buffer [25 mM Tris (pH 7.5), 500 mM NaCl, 10% glycerol, and 1 mM TCEP] supplemented with a cOmplete^TM, EDTA-free Protease Inhibitor tablet and 40 μL Benzonase (obtained from MPIB Core Facility) and lysed by douncing on ice. Centrifugation-cleared lysate (60 min, 58,000g, 4 °C) was sterile-filtered, applied to the Ni-NTA IMAC column (HisTrap SP, 5 mL, Cytiva), washed with IMAC running buffer and eluted with a step-wise gradient with IMAC elution buffer [25 mM Tris (pH 7.5), 500 mM Imidazole, 500 mM NaCl, 10% glycerol, and 1 mM TCEP]. The protein was diafiltered using a 30-kDa cutoff Amicon Ultra 15 (UFC903024; Merck Millipore) filter against IMAC running buffer and the his-sumo-tag was cleaved using sumo protease (obtained from MPIB Core Facility) overnight at 4 °C. Cleaved protein was further purified with size exclusion chromatography to remove any protein aggregates using a Superdex 200 Increase 10/300 GL column (28-9909-44; Cytiva) in phosphate-buffered saline supplemented with additional 150 mM NaCl and 1 mM TCEP.

All protein preparations were controlled for their integrity and purity using SDS-PAGE, high-resolution intact mass spectrometry, UV/Vis spectrum, and dynamic light scattering.

Turbidity Measurements.

Proteins were diluted in HEPES buffer and incubated for 30 min. The solution was transferred to a quartz cuvette, and the absorbance at 350 nm was measured in a UV/Vis spectrometer (Lambda 25; PerkinElmer).

Small Unilamellar Vesicles Production.

Lipid mixtures of 0% PIP₂ (94.3 mol% DOPC, 5 mol% DGS-NTA-Ni, 0.5 mol% PEG2000-PE, and 0.2 mol% Oregon Green DHPE) and 5% PIP₂ (89.3 mol% DOPC, 5 mol% DGS-NTA-Ni, 5 mol% PI(4,5)P₂, 0.5 mol% PEG2000-PE, and 0.2 mol% Oregon Green DHPE) were mixed in chloroform. Lipid films were prepared by drying the lipid mixtures under a stream of nitrogen and placing under a vacuum for at least 2 h. Lipid films were hydrated and resuspended by sonication for 30 min in HEPES buffer (0% PIP₂ mixture) or Citric buffer (5% PIP₂ mixture) at a final lipid concentration of 0.5 mM. Small unilamellar vesicles (SUVs) were then created by extruding lipid suspensions 20 times through 100-nm-pore membrane filters (Whatman) using the miniextruder (Avanti Polar Lipids). SUVs were stored at 4 °C and used within 72 h.

Flow Chamber Preparation.

Flow chambers (≈40 μL) that consist of coverslips (22 mm×22 mm; Carl Roth) fixed to microscope slides (25 mm×75 mm; Carl Roth) by three-layer parafilm were used for the assays. The coverslips were sonicated for 30 min in 3 M NaOH and rinsed with MilliQ H₂O before being cleaned in piranha solution (2:1, H₂SO₄/H₂O₂) for 10 min to render the surface hydrophilic. The piranha-cleaned coverslips were then rinsed and stored in MilliQ H₂O. The microscope slides were sonicated for 30 min in 2 wt% Hellmanex aqueous solution (Hellma) and rinsed with MilliQ H₂O before being stored in ethanol. The coverslips and microscope slides were used within 72 h after cleaning.

Model Membrane Systems Reconstitution.

Supported lipid bilayers (SLBs) were prepared in flow chambers. SUVs were added to chambers at a final lipid concentration of 0.167 mM and allowed to rupture and form an SLB on the glass surface for 20 min at room temperature. Afterward, 0% PIP₂ SLB was washed with 1.6 mL HEPES buffer, and 5% PIP₂ SLB (49) was first washed with 0.8 mL citric buffer, and then 0.8 mL HEPES buffer to remove excess liposomes. Oregon Green DHPE was used to assess the membrane’s fluidity; the fluidity of 0% and 5% PIP₂ membranes was identical (SI Appendix, Fig. S8). SLBs were incubated with 1 μM his₆-tagged integrin β1 tails in HEPES buffer for 20 min and then washed with 0.8 mL HEPES buffer. The β1-bound SLBs were incubated with different protein mixtures depending on the assays. The protein mixtures contain 8 U/mL PO, 1.7 kU/mL C, and 36 mM glucose to function as an oxygen-scavenging system to prevent protein denature and photobleaching during fluorescence imaging.

Imaging and Data Acquisition.

The Leica DMi8 microscope with an HC PL APO 100x/1.47 oil immersion objective was used to perform the epifluorescence imaging using an ORCA-Flash 4.0 CMOS camera (C13440-20CU; Hamamatsu). Images were recorded continuously to follow the condensate dynamics. A Leica Infinity Scanner unit was used to perform fluorescence recovery after photobleaching (FRAP) experiments. Fiji/ImageJ was used to analyze the density and size of β1 clusters (50).

FRAP Experiments.

A circle with a diameter of 5 μm was bleached with a 638-nm laser, and fluorescence images were acquired for 80 s. A region of the SLB outside the circle was used for background subtraction. Fluorescence intensity values within bleached regions were exported using LAS X software. Data were normalized to prebleach levels and fitted to a double-component exponential recovery function, F(t)=y₀ + A_fast ⋅ e^−t/τ_fast + A_slow ⋅ e^−t/τ_slow, where F(t) is the relative fluorescence intensity over time, $\frac{-(A_{\mathit{fast}}+A_{\mathit{slow}})}{1-(y_0+A_{\mathit{fast}}+A_{\mathit{slow}})}$ is the mobile fraction, τ_fast is the fast recovery time constant, and τ_slow is the slow recovery time constant. t_{1/2_fast} = τ_fast ⋅ ln2 and t_{1/2_slow} = τ_slow ⋅ ln2 are the fast and slow half recovery time, respectively. Fitting was performed using Python3.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Image data have been deposited in Zenodo (https://doi.org/10.5281/zenodo.7598352) (51).