Membrane-induced 2D phase separation of the focal adhesion protein talin

Focal adhesions form liquid-like assemblies around activated integrin receptors at the plasma membrane. How they achieve their flexible properties is not well understood. Here, we use recombinant focal adhesion proteins to reconstitute the core structural machinery in vitro. We observe liquid-liquid phase separation of the core focal adhesion proteins talin and vinculin for a spectrum of conditions and interaction partners. Intriguingly, we show that binding to PI(4,5)P2-containing membranes triggers phase separation of these proteins on the membrane surface, which in turn induces the enrichment of integrin in the clusters. We suggest a mechanism by which 2-dimensional biomolecular condensates assemble on membranes from soluble proteins in the cytoplasm: lipid-binding triggers protein activation and thus, liquid-liquid phase separation of these membrane-bound proteins. This could explain how early focal adhesions maintain a structured and force-resistant organization into the cytoplasm, while still being highly dynamic and able to quickly assemble and disassemble.

Experiment on SLB with 8% PIP2, 1µM Tn2, 1µM Vn and 1µM actin.We performed a FRAP experiment on protein-lipid clusters on an SLBs in both the PIP2 and the Tn2 channel (FRAP region: magenta/blue circle).We see different recovery rates for the two channels, similar to Figure 3e.We also see different rates when comparing molecular exchange within the dense phase and molecular exchange between the dense phase and the dilute phase/volume above the membrane.We look into this, by using different reference areas for background correction.By using an area within the condensate (gray area) for background correction, the resulting curve (gray curves) tells us about exchange within the condensate.By using an area outside the condensate (light blue/magenta area) for background correction, the resulting curve (light blue/magenta curves) tells us about recovery through molecules from outside the cluster.The difference is particularly noticeable for the recovery of PIP2: Exchange within the cluster is relatively fast, so that with the low time resolution of this data (1 frame/20 seconds), the bleached area is not visible and the signal within the cluster is already homogenous in the first frame after bleaching.Hence a dip in fluorescence does not exist in the gray curve for PIP2.Recovery of the cluster itself (i.e.exchange with the dilute phase), is much slower, but fluorescence still fully recovers after a few minutes (light magenta curve).The two plots for Tn2 recovery are much more similar to each other.Given the very different nature of the molecules, this discrepancy between PIP2 and Tn2 is not surprising.PIP2 is entirely restricted to a 2D membrane, while Tn2 can exchange with the solution above.In fact, the observation that the curves for dilute phase-normalized Tn2 signal and cluster-normalized Tn2 signal are similar, could lead to the conclusion that recovery mostly happens through exchange with the solution above.However this does not agree with our microscopy images in which 2-dimensional gradients at the boundaries of the bleached region and at cluster boundaries can be observed, indicating that Tn2 also diffuses laterally on the membrane.It should also be noted that generally Tn2 recovery seems to be much slower compared to the recovery of the lipids underneath and compared to Tn2 recovery on SLBs without Vn (i.e.without condensate formation) (Figure S19), indicating that interactions within the dense phase reduce diffusivity of the proteins.Shown is data from a single FRAP experiment.
Image with higher zoom compared to (A).First panel shows merged channels of Tn2-SNAP647, TMR-PIP2 and actin-ATTO488, followed by single channel images for Tn2-SNAP647 and TMR-PIP2.Intensity and contrast of PIP2 signal in single-channel image (right) is increased compared to merged image (left).c The plot (right) shows the intensity profile along the white dotted line in the image on the left.The PIP2 lipids did not form clusters of high density that co-localize with the 2D Tn condensate above as clearly as for the condensates formed through nucleation (Figure 3d -g).This can be explained by the increased area of dense phase when compared to nucleated condensates, in combination with the limited reservoir of PIP2 compared to Tn2 (the latter being available from the solution above).However, when adjusting intensity and contrast, a slightly increased concentration of PIP2 could still be identified to co-localize with the Tn2 patterns, as shown in the intensity profile plot.b Some tethers were detected between anti-DIG beads with biotinylated DNA strands and silica beads coated with 5%PI(4,5)P2 lipid membranes, but at a lower frequency than in the presence of talin.Additionally, the distance at which rupture occurred distinguished these tethers from true talin-membrane interactions (see Figure S24).Three example curves are shown, rupturing at various forces, but all below 1um distance.c A streptavidin-coated silica bead is used to demonstrate a stereotypical force curve for a single DNA strand between two beads, and the melting event that occurs at a force of 60 pN.d Control experiment with a DNA-Tn2 tether and a membrane-coated silica bead, however without PIP2 in the lipid membrane.
. Talin1 requires higher concentrations to phase separate with Vn DR .a Representative image of Tn2-SNAP and Tn1-SNAP (10 µM) droplets that form upon mixing with Vn DR (10 µM) in the presence of a crowding agent.Each sample was done in triplicate, and incubated for one hour before imaging.All samples were mixed in our droplet buffer (10 mM imidazole, 50 mM KCl, 1 mM MgCl2, 1 mM EGTA, 0.2 mM ATP, pH 7.5) supplemented with 15 mM glucose, 20 µg/mL catalase, 100 µg/mL glucose oxidase, 1 mM DTT and 0.25% methylcellulose (4000 cp).b Tn1-Vn DR droplets have liquid-like properties, and return to a spherical shape after fusing.c Fluorescence recovery after photobleaching of Tn1-Vn DR and Tn2-Vn DR droplets.Bleaching was carried out 15-30 minutes after initial droplet formation.Error bars represent standard error; n = 12 droplets for Tn1,Vn DR , n= 6 droplets for Tn2,Vn DR .Scale bars are indicated for each figure panel. .FRAP experiment within Tn2 condensate.Fluorescence recovery after photobleaching a region within a large Tn2 condensate.Experiment with 2 µM Tn2, 2 µM Vn DR , 0.25% methyl cellulose.Fluorescence in the bleached region reaches ~67% of original intensity after 10 min.The boundary between bleached and unbleached area becomes gradually more blurry over time, i.e. the bleached area recovers from the outside in.This suggests a diffusionmediated process and thus indicates the liquid nature of the protein condensates.Shown is data from a single FRAP experiment.Supplementary Figure S5.Phase separation of talin is dependent on protein and crowding agent concentration.Representative images of droplet formation for separation of talin requires full-length, deregulated vinculin.a Additional conditions for Fig. 1f and quantification of the volume of phase separated material from confocal z-stacks.Disrupting the talinvinculin interaction in deregulated vinculin with the A50I mutation blocks phase separation of talin.Data are shown as mean values +/-SEM.For each condition, n ≥ 9 regions were analysed from 3 different samples.**** represents p < 0.0001.b Additional conditions for Fig. 1g and quantification of the volume of phase separated material from confocal z-stacks.Phase separation of talin with de-regulated vinculin is enhanced when autoinhibition of talin is weakened by the E1772A mutation.The Vn D1 domain, which contains the talin binding site, is insufficient for the formation of talin condensates.Data are shown as mean values +/-SEM.For each condition, n ≥ 8 regions were analysed from 3 different samples.**** represents p < 0.0001.Scale bar is 10 µm.Error bars represent standard error.Statistical test performed is an unpaired two-tailed t-test.
. Phase separation of talin with b1D peptide is concentration dependent.Representative confocal microscopy images of Tn2-b1D droplets formed at increasing concentrations of b1D integrin peptide.The amount of total condensate volume is quantified for the varying concentrations from confocal zstacks.Tn2 was mixed with b1D integrin peptide and incubated for 1 hour before imaging.Data are shown as mean values +/-SEM.For each condition, n ≥ 9 regions were analysed from 3 different samples.Scale bar is 10 µm.Supplementary Figure S9.Integrin b1D drives phase separation of Tn2, not Tn1.Additional conditions and quantification corresponding to Fig. 2b. a Tn2-b1D droplet formation requires the residue responsible for Tn2's higher affinity for the b1D integrin receptor, indicating a specific effect.Additionally, Tn1 does not form droplets with the b1D peptide, consistent with the importance of the talin2 S339L residue.3 µM TnX was mixed with 10 µM b1D integrin peptide and incubated for 1 hour before imaging.Data are shown as mean values +/-SEM.For each condition, n ≥ 9 regions were analysed from 3 different samples.b Additional images correspond to Tn2 + Vn DR (top) and b1D alone (bottom).Scale bar is 10 µm.reduces phase separation of Tn2-b1D, and does not phase separate with b1D alone.Representative images for data displayed in Fig. 2d. 3 µM Tn2 and/or 3 µM Vn were mixed with 10 µM b1D integrin peptide and incubated for 1 hour before imaging.Scale bar is indicated in the figure.Scale bar is 10 µm. .Talin1 phase separation can be induced by drastic reduction of salt in buffer.Previous study has shown with cryoEM (Dedden et al 2019) that 80% of talin1 exhibit open conformation in buffer with 500 mM salt (NaCl or KCl), while only 20% are open in buffer with 150 mM salt. a Upon dilution of talin1 pre-incubated at 500 mM salt (top panel) to low salt buffer (30 mM salt), phase separation occurs, relative to the amount of crowding agent (PEG) present.A similar trend is observed with talin incubated at 150 mM salt (bottom panel), although phase separation requires a much higher concentration of crowding agent, consistent with a smaller percentage of talin1 in an open conformation.b Talin1 elution from sizeexclusion column with buffer containing different salt concentrations.With 500 mM salt, talin1 elutes earlier than with 150 mM salt, consistent with a more open, less globular conformation.c Talin1 condensates formed in the presence of crowding agent can be dissolved with high salt that disrupts protein-protein interaction.
d FRAP experiment showing partial fluorescence recovery after 150 seconds.Fluorescence signal for Tn-SNAP647 is shown.e Patterns change dynamically and show coarsening.4 different examples with two frames each which are 165 seconds apart.Arrows highlight notable features that change.Fluorescence signal for Tn-SNAP647 is shown.Supplementary Figure S22.Tethering controls.a Tethers were not detected between anti-DIG beads with biotinylated DNA strands and uncoated silicon beads.

Supplementary Figure S6. Phase separation of talin is salt sensitive. The
amount of phase separated material decreases with increasing amount of salt.Above 150 mM KCl, very few droplets are observed.Experiment done in triplicate, each sample was incubated for one hour in the following buffer (10 mM imidazole, X mM KCl, 1 mM MgCl2, 1 mM EGTA, 0.2 mM ATP, pH 7.5) before imaging.Scale bar is 10 µm.