Myosin-X and talin modulate integrin activity at filopodia tips

Summary Filopodia assemble unique integrin-adhesion complexes to sense the extracellular matrix. However, the mechanisms of integrin regulation in filopodia are poorly defined. Here, we report that active integrins accumulate at the tip of myosin-X (MYO10)-positive filopodia, while inactive integrins are uniformly distributed. We identify talin and MYO10 as the principal integrin activators in filopodia. In addition, deletion of MYO10’s FERM domain, or mutation of its β1-integrin-binding residues, reveals MYO10 as facilitating integrin activation, but not transport, in filopodia. However, MYO10’s isolated FERM domain alone cannot activate integrins, potentially because of binding to both integrin tails. Finally, because a chimera construct generated by swapping MYO10-FERM by talin-FERM enables integrin activation in filopodia, our data indicate that an integrin-binding FERM domain coupled to a myosin motor is a core requirement for integrin activation in filopodia. Therefore, we propose a two-step integrin activation model in filopodia: receptor tethering by MYO10 followed by talin-mediated integrin activation.


In brief
Cells utilize cellular protrusions such as filopodia to explore their environment as they migrate. Filopodia assemble unique integrin-adhesion complexes to sense the extracellular matrix. Miihkinen et al. show that MYO10 and talin regulate integrin activity at filopodia tips and propose a two-step integrin activation model in filopodia.

INTRODUCTION
Filopodia are actin-rich ''antenna-like'' protrusions that are responsible for constantly probing the cellular environment composed of neighboring cells and the extracellular matrix (ECM). As such, filopodia contain cell-surface receptors, such as integrins, cadherins, and growth factor receptors, that can interact with and interpret a wide variety of extracellular cues . Filopodia are especially abundant in cells as they migrate in 3D and in vivo, where they contribute to efficient directional migration by probing and remodeling the surrounding ECM (Jacquemet et al., 2013(Jacquemet et al., , 2017Paul et al., 2015).
Filopodia have a unique cytoskeleton composed of tightly packed parallel actin filaments with barbed ends oriented toward the filopodium tip (Mattila and Lappalainen, 2008). This organization allows molecular motors, such as unconventional myosin-X (MYO10), to move toward and accumulate at the tips (at approximately 600 nm/s) (Kerber et al., 2009). By doing so, these molecular motors are thought to transport various proteins, including integrins, along actin filaments to the tips of filopodia (Jacque-met et al., 2015;Arjonen et al., 2014;Berg and Cheney, 2002;Hirano et al., 2011;Zhang et al., 2004). In particular, MYO10 is known to bind directly to the NPxY motif of the b-integrin cytoplasmic tail via its FERM (protein 4.1R, ezrin, radixin, moesin) domain (Zhang et al., 2004). At filopodia tips, integrins assemble a specific adhesion complex that tethers filopodia to the ECM (Alieva et al., 2019;Jacquemet et al., 2019;Gallop, 2020). Filopodia adhesions contain several adhesion proteins, including talin, kindlin, and p130Cas, but are devoid of the nascent adhesion markers focal adhesion kinase (FAK) and paxillin , indicating that filopodia adhesions are distinct in their molecular composition from other adhesion types. The subsequent maturation of these filopodia adhesions into nascent and focal adhesions can promote directional cell migration (Hu et al., 2014;Jacquemet et al., 2016Jacquemet et al., , 2019. Integrin functions are tightly regulated by a conformational switch that modulates ECM binding, often referred to as activation. Integrin extracellular domain conformations can range from a bent to an extended open conformation, where the integrin's ligand affinity increases with a stepwise opening (Conway and Jacquemet, 2019;Sun et al., 2019;Askari et al., 2009). For Article ll OPEN ACCESS b1-integrin, this unfolding can be viewed using activation-specific antibodies (Byron et al., 2009). Mechanistically, integrin activity can be finely tuned, from within the cell, by multiple proteins that bind to the integrin cytoplasmic tails (Conway and Jacquemet, 2019;Sun et al., 2019;Askari et al., 2009;Bouvard et al., 2013). For instance, talin (TLN), a key integrin activator, can bind to the conserved membrane-proximal NPxY motif of the b-integrin cytoplasmic tail leading to the physical separation of the integrin ɑ and b cytoplasmic tails and integrin activation. Kindlin, another critical regulator of integrin activity, binds to membrane distal conserved NxxY motif in b-integrin cytoplasmic tails, where it cooperates with talin to induce integrin activation (Sun et al., 2019). Although it is clear that integrins and integrin signaling are key regulators of filopodia function (Lagarrigue et al., 2015;Jacquemet et al., 2016Jacquemet et al., , 2019Gallop, 2020), how integrin activity is regulated within filopodia is not fully understood.
Here, we observed that active (high-affinity) integrin accumulates at filopodia tips, while inactive (unoccupied) integrin localizes throughout filopodia. We find that integrin activation in filopodia is locally regulated by talin and MYO10. Contrary to previous assumptions, the FERM domain of MYO10 is not required to transport integrins to filopodia but instead functions to activate integrins at filopodia tips. Because MYO10 contributes to integrin activation at filopodia tips, but MYO10-FERM alone does not directly activate integrins, our data support a two-step integrin activation model in filopodia. In this model, MYO10 enables integrin receptor tethering at filopodia tips, which is then followed by talin-mediated integrin activation.

RESULTS
Integrin activation occurs at filopodia tips independently of cellular forces and focal adhesions We and others have previously described the formation of integrin-mediated ECM-sensing adhesions at filopodia tips (Shibue et al., 2012;Jacquemet et al., 2019;Lagarrigue et al., 2015;Alieva et al., 2019;Gallop, 2020). To gain further insights into how integrin activity is regulated in MYO10 filopodia, we first assessed the spatial distribution of high-affinity and unoccupied b1-integrin (termed active and inactive integrin, respectively, for simplicity) in U2-OS cells overexpressing fluorescently tagged MYO10 using structured illumination microscopy (SIM) (Figures 1A-1C) and scanning electron microscopy (SEM) ( Figure 1D). We focused on b1-integrin because antibodies recognizing the active and inactive forms of this receptor are well characterized (Byron et al., 2009). The average distribution of the b1-integrin species along filopodia was mapped from the SIM and the SEM images revealing enrichment and clustering of active b1-integrins at filopodia tips ( Figures 1B-1E). In contrast, inactive b1-integrins were more uniformly distributed along the entire length of the filopodium (Figures 1A-1E). Importantly, this pattern of integrin localization was also recapitulated in endogenous filopodia forming in actively spreading cells (in the absence of MYO10 overexpression) ( Figures 1F and 1G).
Previous work reported that forces generated by the actomyosin machinery are required for integrin-mediated adhesion at filopodia tips (Alieva et al., 2019). In addition, we observed that filopodia often align with the force generated by focal adhesions (Stubb et al., 2020). Therefore, we investigated whether cellular forces generated by the cell body and transmitted at focal adhesions were responsible for integrin activation at filopodia tips. U2-OS cells overexpressing fluorescently tagged MYO10 and adhering to fibronectin were treated with DMSO, a myosin II inhibitor (10 mM blebbistatin), or an established focal adhesion inhibitor (CDK1 inhibitor, 10 mM RO-3306) (Robertson et al., 2015;Jones et al., 2018). As expected, inhibition of myosin II or CDK1 led to rapid disassembly of focal adhesions (Figures 1H and S1A). Blebbistatin treatment promoted longer and more numerous filopodia, in line with our earlier report (Stubb et al., 2020), while treatment with the CDK1 inhibitor increased filopodia numbers, but not filopodia length (Figures S1B and S1C). However, no decrease in filopodial integrin activation could be observed when myosin II or CDK1 was inhibited (Figures 1H and 1I). In contrast, CDK1 inhibition led to an increase in the amount of active integrin at filopodia tips (Figures 1J and S1D). Altogether these data indicate that integrin activation at filopodia tips is regulated independently of cellular forces and focal adhesions. Nevertheless, cellular forces are likely required to induce filopodia adhesion maturation into focal adhesions and for efficient ECM sensing (Alieva et al., 2019;Jacquemet et al., 2019). Active integrins accumulate at filopodia tips independently of the cellular forces generated at focal adhesion (A-C) U2-OS cells expressing mScarlet-MYO10 or EGFP-MYO10 were plated on fibronectin (FN) for 2 h, stained for active (12G10 and HUTS21) or inactive (4B4 and mAb13) b1-integrin and F-actin, and imaged using structured illumination microscopy (SIM). Representative maximum intensity projections (MIPs) are displayed; scale bars: (main) 20 mm; (inset) 2 mm. (B) Heatmap highlighting the sub-filopodial localization of the proteins stained in (A) based on their intensity profiles. (C) The preferential recruitment of active and inactive b1-integrin to filopodia tips or shafts was assessed by calculating an enrichment ratio (averaged intensity at filopodium tip versus shaft). Results are displayed as Tukey boxplots. (B and C) MYO10, n = 623 filopodia; F-actin, n = 623; filopodia; HUTS21, n = 538 filopodia; 12G10, n = 329 filopodia; 4B4, n = 413 filopodia; mAb13, n = 369 filopodia; three biological repeats). (D and E) U2-OS cells expressing EGFP-MYO10 were plated on FN for 2 h, stained for active (12G10) or inactive (4B4) b1-integrin, and imaged using a scanning electron microscope (SEM). (E) Representative images of single filopodia are displayed. The upper row was acquired using a secondary electron detector (SED) and the lower row using a backscattered electron detector (vCD). The distance of the two b1-integrin pools from the filopodia tip was measured, and the results are displayed as a density plot (n > 175 gold particles). (F and G) U2-OS cells were plated on FN for 20 min, stained for active (F, 12G10) or inactive (G, 4B4) b1-integrin, and imaged using SIM. Representative MIPs are displayed; scale bars: (main) 20 mm; (inset) 1 mm. (H-J) U2-OS cells expressing EGFP-MYO10 were plated on FN for 1 h and treated for 1 h with 10 mM blebbistatin, 10 mM RO-3306, or DMSO. Cells were stained for active b1-integrin (12G10) and imaged using SIM. (H) Representative MIPs are displayed; scale bars: (main) 20 mm; (inset) 2 mm. (I) Heatmap displaying the subfilopodial localization of active b1-integrin in cells treated with DMSO, blebbistatin, or RO-3306. (J) The average intensity of 12G10 at filopodia tips measured in (I) are displayed as boxplots (I and J; n > 483 filopodia; three biological repeats; ***p < 0.001). For all panels, p values were determined using a randomization test. See also Figure S1.

OPEN ACCESS
Talin is required to activate b1-integrin at filopodia tips The enrichment of active b1-integrin at filopodia tips ( Figure 1) indicates that b1-integrin activation is likely to be spatially regulated by one or multiple components of the filopodium-tip complex. We and others have previously reported that several proteins implicated in the regulation of integrin activity, including the integrin activators talins and kindlins, as well as the integrin inactivator ICAP-1 (ITGB1BP1), accumulate at filopodia tips, where their function remains largely unknown (Lagarrigue et al., 2015;Jacquemet et al., 2016). In addition, we previously reported that enhanced integrin activity often correlates with increased filopodia numbers and stability (Jacquemet et al., 2016). Therefore, we set up a microscopy-based small interfering RNA (siRNA) screen to test the contribution of 10 known integrin activity regulators on filopodia formation. Each target was silenced with two independent siRNA oligos in U2-OS cells stably overexpressing MYO10-GFP ( Figure 2A). The effect on MYO10-positive filopodia was scored, and the silencing efficiency of each siRNA was validated by qPCR ( Figure S1E) or western blot (Figures S1F and S1G). Of the 10 integrin regulators, only talin (combined TLN1 and TLN2) silencing significantly reduced filopodia numbers. Because kindlin-2 (FERMT2) is a major regulator of integrin activity (Theodosiou et al., 2016) and FERMT2 localizes to filopodia tips , we were surprised that FERMT2 silencing did not impact filopodia. To validate this further, we imaged filopodia dynamics in cells silenced for both FERMT1 and FERMT2 (over 90% silencing efficiency). There was no effect on filopodia number or dynamics, suggesting that kindlins are not directly required to support filopodia formation or adhesion under the conditions tested (Figures S1H and S1I). Talin is a critical regulator of integrin activity, known to localize to and modulate filopodia function (Lagarrigue et al., 2015;Jacquemet et al., 2016), and has been predicted by us and others to trigger integrin activation at filopodia tips Lagarrigue et al., 2015). To validate this notion, we plated cells silenced for TLN1 and TLN2 on fibronectin and stained for active b1-integrin ( Figure 2B). Reduced talin expression did not affect filopodia length ( Figure 2C) but was sufficient to decrease active b1-integrin localization at filopodia tips, as well as the percentage of filopodia containing active b1-integrin at their tips ( Figures 2D-2F). Altogether, our data demonstrate that talin is required for integrin activation at filopodia tips.
The FERM domain of MYO10 is required for integrin activation, but not localization, at filopodia tips We previously observed that FMNL3-induced filopodia rarely contain active b1-integrin . A careful reanalysis of these data, using intensity profile mapping, indicates that active b1-integrin can be detected in only 23% of FMNL3induced filopodia ( Figures S2A-S2D). However, this is not due to an absence of b1-integrin because all FMNL3-induced filopodia are strongly positive for inactive b1-integrins ( Figures S2A-S2D). Because integrin activation is a prominent feature of MYO10-positive filopodia (Figure 1), we hypothesized that MYO10 could functionally contribute to integrin activation in filopodia tips.
MYO10 directly binds to integrins via its FERM domain (Hirano et al., 2011;Zhang et al., 2004). In this context, MYO10 is thought to transport integrins and other cargo to filopodia tips actively. We assessed the contribution of the MYO10 FERM domain to integrin localization in filopodia by creating a MyTH4/FERM domain deletion construct (MYO10 DF ) ( Figure 3A). We carefully designed this construct by considering the previously reported MYO10-FERM domain structures (PDB: 3PZD and 3AU5) (Wei et al., 2011;Hirano et al., 2011). MYO10 DF was overexpressed in U2-OS cells, which express low endogenous MYO10 (Young et al., 2018;Jacquemet et al., 2016). Deleting the MYO10-MyTH4/FERM domain led to a small but significant reduction in filopodia number and filopodia length, in line with previous reports (Zhang et al., 2004;Watanabe et al., 2010) (Figures 3B-3D). Strikingly, the majority of MYO10 DF filopodia (80%) were devoid of active b1-integrins at their tips ( Figures 3E-3H), while the uniform distribution of inactive b1-integrins along the filopodium length remained unaffected ( Figures 3E-3H). In line with these results, MYO10 DF -induced filopodia were much more dynamic and seemingly unable to stabilize and attach to the underlying ECM ( Figure 3I; Video S1). Taken together, these findings demonstrate that MYO10 and its MyTH4/FERM domain are required for integrin activation at filopodia tips, but not for b1-integrin localization to filopodia tips (Figures 3 and S2).
Because these findings challenge the model of the MYO10 MyTH4/FERM domain acting as a cargo transporter of integrin to filopodia tips, we tested whether the presence of inactive b1-integrins in MYO10 DF filopodia could be because of the low endogenous MYO10 present in these cells. We expressed wild-type (WT) or MYO10 DF in MYO10-silenced U2-OS cells (90% silencing efficiency with a 3 0 UTR-targeting RNA oligo) and analyzed b1-integrin distribution using SIM ( Figure S3A). Inactive b1-integrin localization in MYO10 DF filopodia was not affected by the silencing of endogenous MYO10, further validating that the MYO10 MyTH4/FERM is not required to localize b1-integrin to filopodia ( Figures S3B-3E). Interestingly, silencing of endogenous MYO10 led to a small decrease in the percentage of MYO10 filopodia that contain active integrin at their tips, suggesting that integrin activation at filopodia tips by MYO10 may be dose dependent ( Figure S3D).

MYO10-MyTH4/FERM deletion does not influence the localization of established filopodia tip components
Because the MYO10 MyTH4/FERM domain is thought to be the cargo binding site in MYO10 (Wei et al., 2011), we hypothesized that the lack of integrin activation at the tip of MYO10 DF filopodia would be caused by the absence of a key integrin activity modulator. We co-overexpressed six established filopodia tip components , TLN1, FERMT2, CRK, DIAPH3, BCAR1, and VASP, with either MYO10 WT or MYO10 DF . SIM sub-filopodial localization of the indicated proteins based on their intensity profiles (n > 799 filopodia; three biological repeats, siTLN1 #3 and siTLN2 #3). (E) The average intensity of 12G10 at filopodia tips as measured in (D) is displayed as boxplots (***p < 0.001). (F) Bar chart highlighting the percentage of filopodia with detectable levels of active b1-integrin in CTRL or siTLN cells (E and F: n > 545 filopodia; three biological repeats). For all panels, p values were determined using a randomization test. Article ll OPEN ACCESS microscopy revealed that the localization of these proteins was unaffected by MYO10-FERM domain deletion ( Figure S4). Interestingly, VASP has been previously described as an MYO10-FERM cargo, but its localization at filopodia tips was unaffected by MYO10-FERM deletion (Young et al., 2018;Tokuo and Ikebe, 2004;Lin et al., 2013). Altogether, our results demonstrate that the recruitment of key filopodia tip proteins, including TLN1, is independent of the MYO10 FERM domain and suggest that MYO10-FERM may regulate integrin activity via another mechanism than cargo transport.  Figure S5C). Importantly, MYO10 DF2F3 and MYO10 DF3 filopodia displayed low amounts of active b1-integrin at their tips, indicating that the MYO10 F3 subdomain is required to activate integrin at filopodia tips ( Figures  S5D-S5F). These data also indicate that the MyTH4, F1, and F2 subdomains are not directly required to modulate integrin activity at filopodia tips. As others have shown that the MYO10 F3 subdomain contains the b1 integrin binding site (Zhang et al., 2004), our results led us to speculate that MYO10 needs to interact with integrin directly to promote integrin activation.
Although the site where b1-integrin binds to MYO10-FERM remains unknown, the integrin binding site has been mapped in talin-FERM. Despite some controversy regarding the full talin-FERM structure, superimposition of talin and MYO10 FERM domains revealed that both adopt a similar fold in the b-integrin tail binding subdomains ( Figure 4A; Figure S6A) (Zhang et al., 2020;Elliott et al., 2010). Therefore, we can predict mutations likely to disturb the MYO10-integrin interaction (S2001_F2002insA and T2009D; Figure 4B). The introduction of these mutations in MYO10-FERM (FERM ITGBD ) led to a 64% reduction in the ability of b1-integrin tail peptides to pull down GFP-tagged MYO10-FERM domains from cell lysate, indicating that these mutations can impede the interaction between MYO10 and integrins (Figure 4C). Cells expressing full-length MYO10 with the integrin-binding mutation (MYO10 ITGBD ) generated filopodia to the same extent as cells expressing MYO10 WT ( Figures 4D and 4E), but MY-O10 ITGBD filopodia were shorter than MYO10 WT filopodia (Figure 4F). Notably, only 25% of MYO10 ITGBD filopodia contained detectable levels of active b1-integrin at their tips ( Figures 4G-4I). Thus, we conclude that an intact integrin binding site within MYO10-FERM is required for MYO10 to activate b1-integrin at filopodia tips efficiently.
Unlike Talin-FERM, the MYO10 MyTH4/FERM domain is not able to activate integrins The talin-FERM domain is necessary and sufficient to activate integrins Lilja et al., 2017). Given our data indicating that MYO10-FERM is required to activate integrin at filopodia tips (Figures 3 and 4), we tested whether MYO10-FERM could modulate integrin activity similarly to talin-FERM. We employed a flow cytometric assay to measure active cell-surface integrins relative to total cell-surface integrins (Lilja et al., 2017) (Figures 5A-5C). As expected, overexpression of the talin-FERM domain significantly increased integrin activity ( Figure 5A). In contrast, overexpression of the MYO10-FERM domain failed to activate integrins and instead led to a small but highly reproducible decrease in integrin activity in CHO and U2-OS cells (Figures 5A and 5B). Similar data were obtained in U2-OS cells overexpressing full-length MYO10 ( Figure 5B). Conversely, silencing of MYO10 increased integrin activity in MDA-MB-231 cells, where mutant p53 drives high endogenous MYO10 levels (Arjonen et al., 2014), and this was reversed by the reintroduction of full-length MYO10 ( Figure 5C and S6B). Consistent with decreased integrin activation, MYO10-FERM expression attenuated cell adhesion/spreading on fibronectin over time (Figures 5D-5F) . Altogether, our data indicate that, even though the MYO10-FERM domain is necessary for spatially restricted integrin activation at filopodia tips, the MYO10-FERM domain alone cannot activate integrins.
Unlike Talin-FERM, MYO10-FERM binds to both aand b-integrin tails Despite being homologous domains with high structural similarity, the functional difference between MYO10-FERM and Talin-FERM domains prompted us to compare their binding affinities to integrin cytoplasmic tails. Recombinant MYO10-and talin-FERM were expressed in bacteria, purified ( Figure S6C), and their binding affinity to integrin a and b tails was measured using microscale thermophoresis (Figures 6A and 6B; see STAR Methods for details) (Jerabek-Willemsen et al., 2014). As expected, talin-FERM interacted with the b1-integrin tail (measured affinity of 4.7 mM), but not with a-integrin tails .
(H) Bar chart highlighting the percentage of MYO10 WT and MYO10 DF -induced filopodia with detectable levels of active (12G10) and inactive (mAb13) b1-integrin (H and G; n > 250 filopodia; three biological repeats). (I) U2-OS cells expressing EGP-MYO10 WT or EGFP-MYO10 DF were plated on FN and imaged live using an Airyscan confocal microscope (scale bar: 25 mm; Video S1). MYO10 spot lifetime is displayed as boxplots (three biological repeats; n > 33 cells; ***p < 0.006). For all panels, p values were determined using a randomization test. See also Figures  This result agrees with measurements done by others using the same method (Haage et al., 2018). Interestingly, MYO10-FERM bound to the b1-integrin tail with a slightly lower affinity than talin-FERM (measured affinity of 25.1 mM) ( Figures 6A and 6B). This result indicates that talin may be able to outcompete MYO10 for integrin binding.
Unexpectedly, our results indicated that, in contrast with talin-FERM, a-integrin tails also interact with MYO10-FERM in vitro ( Figures 6A and 6B) and with endogenous MYO10 in cell lysate ( Figure 6C). The ability of MYO10 to interact with both aand b-tail peptides appeared to be specific because the clathrin adaptor AP2m, a known a2-integrin tail-specific binder (De Franceschi et al., 2016), was pulled down only with the a2-integrin tail ( Figure 6C). The MYO10-a-tail interaction was dependent on the highly conserved membrane-proximal GFFKR motif, present in most integrin a tails (De Franceschi et al., 2016). Mutation of the motif in the a2-integrin tail (FF/AA mutation, named ITGA2 GAAKR ) abolished the binding of recombinant MYO10-FERM in vitro (Figure 6D) and in pull-downs with full-length MYO10 ( Figure 6E). Importantly, AP2m recruitment was unaffected by the mutation (AP2m binds to a separate motif in the a2-tail) ( Figure 6E). Together, these experiments demonstrate that MYO10 binds to . Samples were then analyzed by flow cytometry, and the integrin activity index was calculated (see STAR Methods; *p = 0.012, **p = 0.0062, one-sample t test; n = 7 of biological repeats).
(D and E) CHO or U2-OS cells transiently expressing EGFP or EGFP-MYO10 FERM were left to adhere to FN, and their spreading was monitored over time using the xCELLigence system. The cell index over time is displayed; gray areas indicate the 95% confidence intervals. The cell index at 60 min is also displayed as a bar chart (***p < 0.001, Student's two-tailed t test; D, n = 4 biological repeats; E, n = 3 biological repeats).
(F) U2-OS cells transiently expressing EGFP or EGFP-MYO10 FERM were seeded on FN and allowed to spread for 40 min prior to fixation. Samples were imaged using a confocal microscope and the cell area measured (***p < 0.001, randomization test; n > 188 cells; 3 biological repeats; scale bars: 16 mm). integrin b tails, in line with previous reports (Zhang et al., 2004;Hirano et al., 2011), revealing a previously unknown interaction between MYO10-FERM and the GFFKR motif in integrin a tails. Binding to both integrin tails has been demonstrated as a mechanism for Filamin-A-mediated integrin inactivation (Liu et al., 2015) and, thus, may be the underlying reason for the inability of MYO10-FERM alone to activate integrins.
To test the relevance of the GFFKR a-integrin tail motif in filopodia induction, we overexpressed full-length WT ITGA2 and ITGA2 GAAKR in CHO cells (these cells lack endogenous collagenbinding integrins) and investigated MYO10 filopodia formation on collagen I ( Figure 6F). ITGA2 GAAKR localizes to the plasma membrane and is expressed at similar levels to WT in CHO cells (Alanko et al., 2015). ITGA2 GAAKR -expressing cells generated fewer filopodia than cells expressing WT ITGA2, indicating that the GFFKR motif in the ITGA2 tail contributes to filopodia formation. We could not directly assess the relevance of the MYO10-a-integrin interaction to filopodia functions because the MYO10 ITGBD construct also displayed reduced binding toward ITGA2 ( Figure 6G).

MYO10-FERM domain fine-tunes integrin activity at filopodia tips
To further investigate how MYO10-FERM regulates integrin activity in filopodia and the functional differences between talin and MYO10 FERM domains, we created a chimera construct, where the FERM domain from MYO10 was replaced by the one from TLN1 (MYO10 TF ) ( Figure 7A). Both MYO10 WT and MY-O10 TF strongly accumulated at filopodia tips (Figures 7B and 7C). Interestingly, in a small proportion of cells (below 1%), MY-O10 TF also localized to enlarged structures connected to stress fibers that are reminiscent of focal adhesions ( Figure 7C).
Cells overexpressing MYO10 TF generated filopodia to the same extent as cells expressing MYO10 WT ( Figure 7D). MYO10 TF filopodia were slightly shorter than MYO10 WT filopodia but of comparable dynamics ( Figures 7E and 7F). These results show that the talin-FERM can replace the MYO10-FERM domain, and highlight an unanticipated level of interchangeability between integrin-binding FERM domains in regulating filopodia properties. Importantly, active b1-integrin accumulated more efficiently at the tips of MYO10 TF filopodia, and MYO10 TF filopodia were more likely to contain active b1-integrin at their tips than MYO10 WT filopodia ( Figures 7G-7J). Silencing of TLN1 and TLN2 still impeded MYO10 TF filopodia formation, indicating that talin-FERM fused to the MYO10 motor is insufficient to substitute for the lack of endogenous full-length talin ( Figures S6C and S6D). The increased amount of active b1-integrin at the tip of MYO10 TF filopodia is likely due to the ability of talin-FERM to activate integrin directly ( Figure 5) or because talin-FERM binds to integrins with a higher affinity than MYO10-FERM ( Figure 6). Altogether, our data indicate that an integrin-binding proficient FERM domain coupled to a myosin motor is required to activate, but not to transport, integrin in filopodia (Figures 2 and 5).

DISCUSSION
Here, we observed that active integrin accumulates at filopodia tips, while inactive integrin localizes throughout filopodia shafts. We find that integrin activation in filopodia is uncoupled from focal adhesions or the actomyosin machinery but is instead regulated by talin and MYO10. Contrary to previous assumptions, MYO10 is not required to localize integrin to filopodia, but its integrin-binding FERM domain is required for integrin activation at filopodia tips. We find, however, that, unlike talin-FERM, MYO10-FERM itself does not promote integrin activation. MYO10 and integrins also localize and modulate other cellular structures, including retraction fibers, invadopodia, growth cone filopodia, and neuronal spines (Schoumacher et al., 2010;Lin et al., 2013;Lilja and Ivaska, 2018;Pelá ez et al., 2019). Here, we focused on the role of MYO10 in modulating integrin in filopodia. Still, it is tempting to speculate that MYO10 may also regulate integrin activity in these other actin-rich protrusions.
We find that MYO10-FERM interaction with integrins is required to localize active integrin to filopodia tips. The simplest assumption would be that MYO10, in its typical capacity as a myosin motor, specifically transports active integrin to filopodia tips. However, our data suggest otherwise as (1) the MYO10 FERM domain alone inactivates integrins, and therefore integrins would not be in an active state during transport; (2) talin is required to localize active integrins at filopodia tips; and (3) integrin activation is thought to be a fast and tightly regulated process (Sun et al., 2019), with all evidence pointing to an on-site integrin activation mechanism in filopodia tips. In addition, direct transport of integrin by MYO10 to filopodia tips has yet to be formally observed. Our data do not exclude the possibility that MYO10 can directly transport integrin in filopodia. Testing this would require performing two-color, single-molecule imaging of MYO10 and integrin to see if they move toward filopodia tips together. However, we find integrins abundantly in filopodia regardless of the MYO10 status. Altogether, we propose that inactive integrins localize along the filopodia plasma membrane via membrane diffusion and are activated at filopodia tips in a two-step process by MYO10 and talin. In this model, MYO10 could tether integrins at filopodia tips because of its motor domain and provide resistance against the actin retrograde flow present in filopodia (Bornschlö gl et al., 2013;Lidke et al., 2005) allowing sufficient time for talin-mediated activation. The precise mechanisms favoring integrin binding to MYO10 or talin in filopodia remain to be elucidated. One possibility is that talin-FERM outcompetes MYO10-FERM. Indeed, our in vitro experiments indicate that talin-FERM has, in solution, a higher affinity for integrin b tail compared with MYO10-FERM. In addition, talin affinity for b-integrin tails will be even stronger in cells because of the presence of negatively charged membrane phosphoinositides that interact with talin-FERM (Chinthalapudi et al., 2018;De Franceschi et al., 2018), and which are known to accumulate at filopodia tips . Interestingly, although MYO10 and talin FERM domains structurally adopt a very similar fold, we find that these two FERM domains are functionally distinct. MYO10-FERM is not capable of directly activating integrin and can interact with both integrin tails. Yet, remarkably, swapping MYO10-FERM with talin-FERM fully supported filopodia function and integrin activation at filopodia tips, suggesting unanticipated interchangeability between these FERM domains in spatially regulating integrin activation in filopodia. Other FERM domain-containing myosins, including MYO7 and MYO15, also localize to filopodia tips Arthur et al., 2019), where their roles are mostly unknown; future work will examine the contribution of these unconventional myosins to filopodia functions.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Guillaume Jacquemet (guillaume.jacquemet@abo.fi).

Materials availability
Several of the plasmids generated in this study have been deposited to Addgene: EGFP-MYO10 FERM (catalog number: 145140), EGFP-MYO10 DF (catalog number: 145816), mScarlet-I-MYO10 DF (catalog number: 145139), EGFP-MYO10 TF (catalog number: 145141). The other plasmids generated in this study will also be available on Addgene soon.

Data and code availability
The authors declare that the data supporting the findings of this study are available within the article and from the authors upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. The ImageJ macro as well as the R code used to generate the filopodia maps were previously described and are available on GitHub (https://github.com/guijacquemet/FiloMAP).

METHOD DETAILS
Plasmids and transfection U2-OS, MDA-MB-231, and CHO cells were transfected using Lipofectamine 3000 and the P3000TM Enhancer Reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. The U2-OS MYO10-GFP lines were generated by transfecting U2-OS cells using Lipofectamine 3000 (ThermoFisher Scientific), selected using Geneticin (ThermoFisher Scientific; 400 mg.ml -1 final concentration) and sorted for green fluorescence using a fluorescence-assisted cell sorter (FACS). All cell lines tested negative for mycoplasma.

Plasmids
The construct encoding the EGFP-tagged MYO10-FERM domain (EGFP-MYO10 FERM ) was designed using the boundaries from the MYO10-FERM crystal structure (Wei et al., 2011). The MYO10 coding region 1480-2053 was amplified by PCR (primers: 5 0 -ATT AGA GAA TTC AAC CCG GTG GTC CAG TGC-3 0 , 5 0 -ATT AGA GGT ACC TCA CCT GGA GCT GCC CTG-3 0 ), and the resulting PCR products were ligated into pEGFP-C1 using the EcoRI and KpnI restriction sites. Article ll OPEN ACCESS GCG GCC GCA CCG ATC GAC ACC CCC AC, 5 0 -ATT AG AGA ATT CTC ACC TGG AGC TGC CCT G) and introduced in pET151 using the NotI and EcoRI restriction sites. The MYO10 MyTH/FERM deletion construct (EGFP-MYO10 DF ) was generated by introducing a premature stop codon in full-length EGFP-MYO10 (boundaries 1-1512 in MYO10) using a gene block (IDT). The gene block was inserted in EGFP-MYO10 using the PvuI and XbaI restriction sites.
The mScarlet-I-MYO10 DF construct was created from EGFP-MYO10 DF by swapping the fluorescent tag. The mScarlet-I (Bindels et al., 2017) coding sequence, acquired as a gene block (IDT), was inserted in EGFP-MYO10 DF using the NheI and KpnI restriction sites.
The EGFP-MYO10 ITGBD construct was generated by replacing the wild-type MYO10-FERM domain (boundaries 1504-2056 in MYO10) with a MYO10 FERM domain containing the required mutations (S2001_F2002insA/T2009D) using a gene block (IDT). The gene block was inserted in EGFP-MYO10 using the PvuI and XbaI restriction sites.
The EGFP-MYO10 DF2F3 and EGFP-MYO10 DF3 constructs were generated by replacing the wild-type MYO10-FERM domain (boundaries 1504-2056 in MYO10) with truncated MYO10 FERM domains where the F2-F3 or F3 FERM lobes are deleted using gene blocks (IDT). The gene blocks were inserted in EGFP-MYO10 using the PvuI and XbaI restriction sites. The final boundaries compared to full-length MYO10 are 1-1794 for MYO10 DF2F3 and 1-1951 for MYO10 DF3 .

siRNA-mediated gene silencing
The expression of proteins of interest was suppressed using 83 nM siRNA and lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer's instructions. All siRNAs used were purchased from QIAGEN. siMYO10 #7 targets the 3 0 UTR of the MYO10 mRNA and therefore does not affect the expression of MYO10 constructs.

SDS-PAGE and quantitative western blotting
Purified proteins or protein extracts were separated under denaturing conditions by SDS-PAGE and transferred to nitrocellulose membranes using Trans-Blot Turbo nitrocellulose transfer pack (Bio-Rad, 1704159). Membranes were blocked for 45 min at room temperature using 1x StartingBlock buffer (Thermo Fisher Scientific, 37578). After blocking, membranes were incubated overnight with the appropriate primary antibody (1:1000 in PBS), washed three times in TBST, and probed for 40 min using a fluorophore-conjugated secondary antibody diluted 1:5000 in the blocking buffer. Membranes were washed three times using TBST, over 15 min, and scanned using an Odyssey infrared imaging system (LI-COR Biosciences).
siRNA screen 96-well glass-bottom plates (Cellvis, P96-1.5H-N) were first coated with a solution of poly-D-lysine (10 mg/ml in PBS, Sigma-Aldrich, A-003-M) at 4 C overnight. Plates were then washed with PBS and coated with a solution containing 10 mg/ml of bovine fibronectin (in PBS) also at 4 C overnight. Excess fibronectin was washed away with PBS.
U2-OS cells stably expressing MYO10-GFP were silenced for the gene of interest using a panel of siRNAs (QIAGEN flexiplate, 1704159) using Lipofectamine 3000 (Thermo Fisher Scientific, L3000075). 48 h post silencing, cells were trypsinized and plated on both fibronectin-coated 96-well glass-bottom plates and 96-well plastic-bottom plates in full culture medium. Cells plated in the plastic-bottom plates were allowed to spread for two hours before being lysed using an RNA extraction buffer. RNAs were then purified and the silencing efficiency of each siRNA was validated by qPCR analysis.
Cells plated in the glass-bottom plates were allowed to spread for two hours and fixed with a warm solution of 4% paraformaldehyde (PFA; Thermo Scientific, 28906). After washing, the samples were incubated with a solution of 1 M glycine (30 min, in PBS) and then for one hour in a solution containing phalloidin-Atto647N (1/400 in PBS, Thermo Fisher Scientific, 65906) and DAPI (0.5 mg/ml in PBS, Thermo Fisher Scientific, D1306). The 96-well glass-bottom plates were then imaged using a spinning-disk confocal microscope equipped with a 40x objective. Images were analyzed using Fiji (Schindelin et al., 2012). Briefly, images were opened and, after background subtraction and normalization, MYO10 spots were automatically detected using Michael Schmid's 'Find maxima' plugin. As inactive MYO10 is known to accumulate in rab7 vesicles (Plantard et al., 2010), to obtain an accurate number of filopodiaspecific MYO10 spots, intracellular MYO10 spots were excluded from the analysis. Intracellular MYO10 spots were automatically filtered by masking the cells using the F-actin staining. The remaining spots per field of view were counted.
RNA extraction, cDNA preparation, and Taq-Man qPCR Total RNA extracted using the NucleoSpin RNA Kit (Macherey-Nagel, 740955.240C) was reverse transcribed into cDNA using the high-capacity cDNA reverse transcription kit (Applied Biosystems,Thermo Fisher Scientific, according to the manufacturer's instructions. The TaqMan primer sequences and associated universal probes were generated using ProbeFinder (version 2.53, Roche). The primers themselves were ordered from IDT, and the TaqMan fast advanced master mix (Thermo Fisher Scientific, 4444557) was used to perform the qPCR reactions according to the manufacturer's instructions. qPCR reactions were analyzed with the 7900HT fast RT-PCR System (Applied Biosystems), and the results were analyzed using the RQ Manager Software (Applied Article ll OPEN ACCESS Biosystems). Relative expression was calculated by the 2 -DDCT method. GAPDH mRNA levels were used to normalize data between experiments and conditions.
Generation of filopodia maps U2-OS cells transiently expressing the constructs of interests were plated on high tolerance glass-bottom dishes (MatTek Corporation, coverslip #1.7) pre-coated first with Poly-L-lysine (10 mg/ml, 1 h at 37 C) and then with bovine plasma fibronectin (10 mg/ml, 2 h at 37 C). After 2 h, samples were fixed and permeabilized simultaneously using a solution of 4% (wt/vol) PFA and 0.25% (vol/vol) Triton X-100 for 10 min. Cells were then washed with PBS, quenched using a solution of 1 M glycine for 30 min, and, when appropriate, incubated with the primary antibody for 1 h (1:100). After three washes, cells were incubated with a secondary antibody for 1 h (1:100). Samples were then washed three times and incubated with SiR-actin (100 nM in PBS; Cytoskeleton; catalog number: CY-SC001) at 4 C until imaging (minimum length of staining, overnight at 4 C; maximum length, one week). Just before imaging, samples were washed three times in PBS and mounted in Vectashield (Vector Laboratories).
To map the localization of each protein within filopodia, images were first processed in Fiji (Schindelin et al., 2012) and data analyzed using R as previously described . Briefly, in Fiji, the brightness and contrast of each image was automatically adjusted using, as an upper maximum, the brightest cellular structure labeled in the field of view. In Fiji, line intensity profiles (1-pixel width) were manually drawn from filopodium tip to base (defined by the intersection of the filopodium and the lamellipodium). To avoid any bias in the analysis, the intensity profile lines were drawn from a merged image. All visible filopodia in each image were analyzed and exported for further analysis (export was performed using the ''Multi Plot'' function). For each staining, line intensity profiles were then compiled and analyzed in R. To homogenize filopodia length; each line intensity profile was binned into 40 bins (using the median value of pixels in each bin and the R function ''tapply''). Using the line intensity profiles, the percentage of filopodia positive for active b1 at their tip was quantified. A positive identification was defined as requiring at least an average value of 5000 (values between 0-65535) within the bins defining the filopodium tip (identified using MYO10 staining). The map of each protein of interest was created by averaging hundreds of binned intensity profiles. The length of each filopodium analyzed was directly extracted from the line intensity profiles.
The preferential recruitment of active and inactive b1 integrin to filopodia tips or shafts was assessed by calculating an enrichment ratio where the averaged intensity of the b1 integrin species at the filopodium tip (bin 1-6) was divided by the averaged intensity at the filopodium shaft (bin 7-40). This enrichment ratio was calculated for each filopodium analyzed and the results were displayed as Tukey boxplots.

Quantification of filopodia numbers and dynamics
For the filopodia formation assays, cells were plated on fibronectin-coated glass-bottom dishes (MatTek Corporation) for 2 h. Samples were fixed for 10 min using a solution of 4% PFA, then permeabilized using a solution of 0.25% (vol/vol) Triton X-100 for 3 min. Cells were then washed with PBS and quenched using a solution of 1 M glycine for 30 min. Samples were then washed three times in PBS and stored in PBS containing SiR-actin (100 nM; Cytoskeleton; catalog number: CY-SC001) at 4 C until imaging. Just before imaging, samples were washed three times in PBS. Images were acquired using a spinning-disk confocal microscope (100x objective). The number of filopodia per cell was manually scored using Fiji (Schindelin et al., 2012).
To study filopodia stability, U2-OS cells expressing MYO10-GFP were plated for at least 2 h on fibronectin before the start of live imaging (pictures taken every 5 s at 37 C, on an Airyscan microscope, using a 40x objective). All live-cell imaging experiments were performed in normal growth media, supplemented with 50 mM HEPES, at 37 C and in the presence of 5% CO 2 . Filopodia lifetimes were then measured by identifying and tracking all MYO10 spots using the Fiji plugin TrackMate (Tinevez et al., 2017). In TrackMate, the LoG detector (estimated bob diameter = 0.8 mm; threshold = 20; subpixel localization enabled) and the simple LAP tracker (linking max distance = 1 mm; gap-closing max distance = 1 mm; gap-closing max frame gap = 0) were used.
The structured illumination microscope (SIM) used was DeltaVision OMX v4 (GE Healthcare Life Sciences) fitted with a 60x Plan-Apochromat objective lens, 1.42 NA (immersion oil RI of 1.516) used in SIM illumination mode (five phases x three rotations). Emitted light was collected on a front-illuminated pco.edge sCMOS (pixel size 6.5 mm, readout speed 95 MHz; PCO AG) controlled by SoftWorx.
The confocal microscope used was a laser scanning confocal microscope LSM880 (Zeiss) equipped with an Airyscan detector (Carl Zeiss) and a 40x oil (NA 1.4) objective. The microscope was controlled using Zen Black (2.3), and the Airyscan was used in standard super-resolution mode.
MDA-MB-231 and U2-OS cells detached using Hyclone HyQTase (Thermo Fisher Scientific, SV300.30.01) were fixed with 4% PFA (in PBS) for 10 min and stained for active (antibody 9EG7) and total b1 integrin (antibody P5D2) overnight at 4 C. Cells were then stained with the appropriate Alexa Fluor 647-conjugated secondary antibody (45 min at RT, 1:200, Thermo Fisher Scientific) and the fluorescence was recorded using FACS. Data were gated and analyzed using the Flowing Software (https://bioscience.fi/ services/cell-imaging/flowing-software/) and the integrin activity (IA) was calculated as indicated below where F 9EG7 and F P5D2 are the signals intensities of the 9EG7 and P5D2 stainings, respectively. F 2nd Ab corresponds to the signal intensity recorded when the cells are stained with only the secondary antibody.

Cell spreading assay
The xCELLigence RTCA instrument (Roche) was used to measure cell adhesion on fibronectin in real-time . The RTCA instrument uses gold-bottom electrode plates to measure the impedance between two electrodes. This is expressed as an arbitrary cell index value. The xCELLigence 96-well plates (Acea Biosciences, E-Plate VIEW 96 PET, 00300600900) were coated with a solution of 20 mg/ml of poly-D-lysine (in PBS) for 1 h at 37 C, washed with PBS, and coated with a solution of 10 mg/ml fibronectin (in PBS) for 1 h at 37 C. Plates were then blocked using a solution of 1% BSA (in PBS) for 1 h in RT. After 2 PBS washes, 15000 cells were seeded into each well in a serum-free culture medium. The cell index was recorded over time.

Recombinant protein expression and purification
The E. coli BL-21(DE3) strain was transformed with IPTG inducible, His-tagged expression constructs, and the transformed bacteria were grown at 37 C in LB media supplemented with ampicillin (1 mg/ml) until OD600 was 0.6-0.8. Protein expression was then induced using IPTG (0.5 mM), and the temperature was lowered to 25 C. Cells were harvested after 5 h by centrifugation (20 min at 6000 g). Bacteria were then resuspended in a resuspension buffer (1x TBS, cOmplete protease inhibitor tablet (Roche, cat. no. 5056489001), 1x AEBSF inhibitor, 1x PMSF, RNase 0.05 mg/ml, DNase 0.05 mg/ml). To lyse the bacteria, a small spoonful of lysozyme and 1x BugBuster (Merck Millipore, were added, and the suspension was agitated for 30 min at 4 C. Cell debris was pelleted using a JA25.5 rotor at 20000 rpm for 1 h. His-tagged proteins were batch purified from the supernatant using a Protino Ni-TED 2000 column (Macherey Nagel,cat. no. 745120.25) according to the manufacturer's instructions. Proteins were eluted using the elution buffer provided with the kit supplemented with 1 mM AEBSF. For each purified protein, several 1 mL fractions were collected, ran on a 4%-20% protein gel (Bio-Rad Mini-PROTEAN TGX, #4561093), stained with InstantBlueâ (Expedeon, ISB1L), and the fractions abundant in tagged protein were combined. Imidazole was removed in a buffer exchange overnight at 4 C and 1 mM AEBSF was added to the imidazole-free protein. Proteins were stored at 4 C for up to one week.
Whole-mount immuno-SEM U2-OS cells expressing MYO10-GFP were plated for 2 h on fibronectin-coated coverslips and fixed with a solution of 4% PFA (in 0.1 M HEPES, pH 7.3) for 30 min. After washing and quenching with 50 mM NH 4 Cl (in 0.1 M HEPES), non-specific binding was blocked with a buffer containing 2% BSA (in 0.1 M HEPES). Samples were then labeled using the appropriate primary antibody (1:10 in 0.1 M HEPES) for 30 min, washed, and labeled with a gold conjugated secondary antibody (1:50 in 0.1 M HEPES, 30 nm gold particles, BBI solutions, EM.GAF30) for 30 min. After immunolabeling, the samples were washed, and post-fixed with a solution of 2.5% glutaraldehyde and 1% buffered osmium tetroxide prior to dehydration and drying using hexamethyldisilazane. The dried samples were mounted on SEM stubs and sputter-coated with carbon. The micrographs were acquired with FEI Quanta FEG 250 microscope with SE and vC detectors (FEI Comp.) using an acceleration voltage of 5.00 kV and a working distance ranging from 7.7 to 10.9 mm.
To compare the distribution of active and inactive integrin from EM images, we manually measured the distance between each detected gold particle and the filopodium tip using Fiji. Results were then plotted as a probability density function where the area under the curve represents 100% probability. A bootstrap version of the univariate Kolmogorov-Smirnov test was then used to assess statistical significance (using Rstudio). Importantly, filopodia length was not normalized in these analyses.

Microscale thermophoresis
Eq.1 Alternatively, binding was also expressed as a change in MST signal (normalized fluorescence DFnorm). This is defined as a ratio: Where F0 is the fluorescence prior and F1 after IR laser activation. All binding data were analyzed using MO.Control and MO.Affinity software (NanoTemper).