Kindlin-2 promotes rear focal adhesion disassembly and directional persistence during cell migration

Cell migration is an important process whose alteration is intimately associated with many human diseases (Avraamides et al., 2008; Etienne-Manneville, 2008; Mierke et al., 2008; Northey et al., 2017). Cell migratory behavior can be characterized by two independent quantities: linear cell locomotion speed and directional persistence (Lauffenburger and Horwitz, 1996). For efficient directional migration, migrating cells require establishment and maintenance of front-rear polarity (Horwitz and Parsons, 1999; Horwitz and Webb, 2003; Petrie et al., 2009). When undergoing morphological changes with front protrusions and rear retraction that involve dynamic actin cytoskeleton and actomyosin contraction, cells crawl in the intrinsic front-rear direction (Lauffenburger and Horwitz, 1996; Ridley et al., 2003; Bretscher, 2008; Insall and Machesky, 2009). At the leading edge, actin filament polymerization leads to protrusions, where adhesion to extracellular matrix (ECM) triggers the formation of cell-ECM adhesions (Hynes, 2002; Huttenlocher and Horwitz, 2011). These adhesions establish a link between ECM and the actin cytoskeleton and exert traction forces on the substratum, resulting in generation of net propulsive forces to drive forward cell movement (Ridley et al., 2003; Huttenlocher and Horwitz, 2011; Swaminathan and Waterman, 2016). Adhesions in distinct regions of the cell exhibit asymmetric dynamics during cell migration (Rid et al., 2005; Broussard et al., 2008; Mohl et al., 2012). In the front, the newly formed adhesions undergo turnover during active protrusions, or grow into focal adhesions (FAs) that reinforce cell-ECM adhesion (Dubash et al., 2009; Parsons et al., 2010). At the rear, FA sliding and disassembly are crucial for adhesive release, promoting trailing edge detachment and directional migration (Kirfel et al., 2004; Cramer, 2013; Chi et al., 2014). Actomyosin contractility appears to be a dominant force driving rear retraction when large forces are required (Cramer, 2013). Furthermore, myosin II plays an essential role in establishing migratory front-rear polarity and promoting rear retraction during cell migration (Vicente-Manzanares et al., 2007, 2008, 2011; Wilson et al., 2010).

FA is composed of integrins and a plethora of intracellular proteins (Dubash et al., 2009; Kanchanawong et al., 2010; Horton et al., 2015). Pulling forces generated by actomyosin contraction are required for FA assembly through mechanosensitive protein complexes (Gallant et al., 2005; Oakes and Gardel, 2014; Burridge and Guilluy, 2016). Compared to FA assembly, FA disassembly, however, is less well understood. Various mechanisms have been implicated in FA disassembly, including those involving actomyosin contractility (Broussard et al., 2008). Our understanding of proteins that control front-rear asymmetric FA dynamics during cell migration and their signaling mechanisms, however, remains incomplete.

Kindlins are a family of evolutionarily conserved proteins crucial for integrin-mediated cell-ECM adhesion and signaling (Tu et al., 2003; Montanez et al., 2008; Sun et al., 2019; Chronopoulos et al., 2020; Harburger et al., 2009; Huet-Calderwood et al., 2014). Mammalian kindlin family consists of three members, namely kindlin-1, -2 and -3 (officially known as FERMT1, FERMT2 and FERMT3, respectively). Kindlin-2 is expressed in a broad spectrum of cell types, whereas the expression of kindlin-1 and -3 is more restricted (Larjava et al., 2008; Plow et al., 2009; Rognoni et al., 2016). Genetic knockout (KO) and knockdown studies have shown that kindlin-2 is essential for embryonic development as well as homeostasis and functions of different types of tissue and organ (Sun et al., 2017; Rognoni et al., 2016; Zhang et al., 2016; Pluskota et al., 2011). At the cellular level, kindlin-2 has been implicated in regulation of actin cytoskeleton, cell spreading and migration (Shi et al., 2007; Theodosiou et al., 2016; Bottcher et al., 2017; Tu et al., 2003; Yasuda-Yamahara et al., 2018; Bledzka et al., 2016). However, how kindlin-2 functions in cell migration is not completely understood. In this study, we have investigated the function of kindlin-2 in cell migration as well as the underlying mechanism.

Cell migration involves coordinated changes of cell shape (Broussard et al., 2008; Oakes and Gardel, 2014). To begin to investigate the functions of kindlin-2 in cell migration, we assessed the effects of kindlin-2 and/or kindlin-1 deficiency on cell shape modulation. Using CRISPR/Cas9-mediated targeted-gene disruption (Fig. S1A), we knocked out kindlin-2 from human HT1080 cells (Fig. 1A, lane 3) and examined the effect on cell morphology (Fig. 1B). In parallel experiments, either kindlin-1 (Fig. 1A, lane 2), which was also expressed in HT1080 cells or both kindlin-1 and 2 (Fig. 1A, lane 4) were knocked out. On either fibronectin (FN) or collagen IV, kindlin-2 KO cells spread poorly, showing fewer and smaller lamellipodia and a highly elongated morphology compared to that of wild-type HT1080 cells (Figs 1B and 2A), suggesting that kindlin-2 is crucial for HT1080 cell shape modulation. No significant morphological and cell spreading defects were observed in kindlin-1 KO cells (Fig. 1B). Regarding kindlin-1 and -2 double KO cells, many of them failed to attach to FN even after prolonged incubation (e.g. overnight); however, those that did attach failed to spread and exhibited a rounded morphology (Fig. 1B). This suggests that the limited spreading observed in the kindlin-2 KO cells was probably supported by the remaining kindlin protein (i.e. kindlin-1) in kindlin-2 KO cells. To further test this, we re-expressed GFP-tagged kindlin-2 and -1, and GFP alone as a control in kindlin-2 KO cells. Western blotting showed that similar levels of GFP-tagged kindlin-2 and -1 were expressed in the corresponding cells (Fig. 1C, compare lanes 3 and 4). Analyses of cell spreading revealed that re-expression of GFP-kindlin-1, like that of GFP-kindlin-2 but not GFP alone, in kindlin-2 KO cells, restored normal cell morphology (Fig. 1E) and spreading (Fig. 2A,B). Thus, when expressed at a level similar to GFP-kindlin-2, GFP-kindlin-1 can functionally substitute GFP-kindlin-2 in promoting cell spreading.

Tryptophan (W) residue 619 in the kindlin-2 F3 domain, and residues leucine (L) 46 as well as lysine (K) 47 (hereafter referred to as LK47) in the F0 domain are crucial for binding to integrin (Li et al., 2017) and actin (Bledzka et al., 2016), respectively. To assess the role of integrin- and actin-binding in kindlin-2-mediated cell shape modulation we, respectively, expressed a GFP-tagged W619Q mutant defective in integrin-binding (Fig. 1D, lane 4), and a GFP-tagged actin-binding-defective mutant LK(47)/AA that comprises an L46A and K47A double substitution to alanine (Fig. 1C, lane 5), in kindlin-2 KO cells. Consistent with previous studies (Bledzka et al., 2016; Li et al., 2017), mutant LK(47)/AA but not W619Q localized to FAs (Fig. 1E). Furthermore, expression of W619Q, unlike that of wild-type kindlin-2 or LK(47)/AA, failed to restore normal cell morphology (Fig. 1E) and spreading (Fig. 2A,B), suggesting that integrin-binding but not LK47-mediated actin-binding is required for kindlin-2-mediated cell spreading.

Kindlin-2 KO cells exhibited a highly elongated morphology on either FN or collagen IV compared to wild-type cells (Fig. 2A). Quantification of the elongation index (i.e. the ratio of length to width of the cell) confirmed that kindlin-2-deficient cells were significantly elongated compared to wild-type cells (Fig. 2C). Re-expression of kindlin-1, wild-type kindlin-2 or LK(47)/AA but not W619Q in kindlin-2 KO cells restored the normal elongation index (Fig. 2C), confirming an essential role of integrin-binding in this process.

Consistent with the elongated cell shape, kindlin-2 KO cells showed smaller leading lamellipodia and drastically stretched trailing tails (Fig. 3A). A 4-fold decrease in the number of FAs per cell was observed in kindlin-2-deficient cells (Fig. 3B), which is consistent with the fact that kindlin-2 is crucial to integrin activation and cell-ECM adhesion (Ma et al., 2008; Harburger et al., 2009). Human HT1080 fibroblasts express multiple integrins including integrins α2, αV, β1 and β3 (Grenz et al., 1993). Consistent with a role of kindlins in the regulation of cell surface integrin levels (Harburger et al., 2009; Margadant et al., 2013, 2012; Kruger et al., 2008; Qu et al., 2011; Schmidt et al., 2011), flow cytometry analyses showed reduced cell surface levels of αVβ3 integrin but modestly increased levels of α2 and β1 integrins in response to loss of kindlin-2 (Fig. S1B). Furthermore, immunofluorescence staining showed clusters of β1 and αVβ3 integrins within many of the remaining FAs in kindlin-2 KO cells, although the level of αVβ3 integrin in some FAs appeared to be reduced (Fig. S1C,D).

It is worth noting that remaining FAs in kindlin-2 KO cells were large and peripheral rather than central FAs (Fig. 3A). In particular, the percentage of large (>15 µm²) FAs in kindlin-2 KO cells was much higher (∼14%) than that in wild-type cells (∼5%; Fig. 3C). Furthermore, ∼50% of the large FAs were >20 µm², and primarily present at the cell edge (Fig. 3A, arrowhead). By contrast, FAs of medium size (2–8 µm²) were decreased to 44% in kindlin-2 KO cells compared to 57% in wild-type cells (Fig. 3C). Collectively, these results suggest that kindlin-2 play a role in the regulation of FA dynamics.

We next investigated the effect of kindlin-2 deficiency on cell migration by using time-lapse imaging. Kindlin-2 KO cells initially spread well on FN, suggesting no obvious initial spreading defect. Approximately 20–25 min after adhering to FN, both wild-type and kindlin-2 KO cells displayed front-back polarity (Fig. 4A,C and Movies 1,2). However, kindlin-2 KO cells assumed the elongated morphology (Fig. 4A and Movie 1). Moreover, whereas wild-type cells migrated efficiently, showing lamellipodia formation at the front and trailing edge retraction at the rear, the majority (86%) of kindlin-2 KO cells were stretched, with ∼35% of them exhibiting abrupt rear-edge detachment (Fig. 4C, Movie 2). Some kindlin-2 KO cells moved forwards without detachment of the trailing edge, leaving behind a piece of cell body (Fig. 4D, arrowhead, Movie 3). Kindlin-2 deficiency-induced impairment of trailing edge detachment was observed in kindlin-2 KO HT1080 cells seeded on both soft and stiff matrices (Movies 4,5) as well as in other cell types, including human podocytes (Movie 6) and A549 lung adenocarcinoma cells (Movie 7), in which kindlin-2 had been knocked down or depleted (Fig. S2). Cell motility analysis revealed that both mean and maximum migration speed (Smean and Smax, respectively) were significantly increased in kindlin-2 KO cells (Fig. 4C, Fig. S2). Notably, directional persistence was significantly reduced in kindlin-2 KO cells (Fig. 4C, Fig. S2). Taken together, these results suggest that kindlin-2 is crucial for trailing edge detachment and directional persistence in migrating cells.

To further test this, we analyzed the migration of kindlin-2 KO cells re-expressing kindlin-1, kindlin-2, W619Q or LK(47)/AA all tagged to GFP (Fig. 1C–D), finding that kindlin-1, kindlin-2 and LK(47)/AA but not W619Q successfully restored rear retraction and coordinated migration in kindlin-2-deficient cells (Fig. S3, Movie 8). Collectively, these data identify an important role of kindlin-2 and kindlin-1 – when expressed at appropriate levels – in promoting rear retraction during cell migration and suggest that this regulation is dependent on kindlin-2 interaction with integrin.

Rear retraction is coupled to FA disassembly at the cell rear (Broussard et al., 2008). Thus, we analyzed the effect of kindlin-2 deficiency on FA dynamics by using time-lapse TIRF microscopy. Paxillin, a marker of FAs, tagged to Ruby2 was manually tracked, and FAs dynamics at the cell front and rear were analyzed (Fig. 5A) (Stehbens and Wittmann, 2014). Concurrent with rear retraction, FAs at the cell rear slid and were rapidly ripped off by rear retraction in wild-type cells (Movie 9). The disassembly rate (K_d) of rear FAs was approximately twice as high as that of frontal ones (Fig. 5B), whereas the FA assembly rate (K_a) in the front was similar to that at the rear (Fig. 5B). Thus, there is a front-rear polarization in terms of FA disassembly in wild-type cells. By contrast, FAs at the rear barely slid and appeared rather stable in kindlin-2 KO cells, preventing the rear edge from retraction (Movie 9). Here, the K_d was significantly decreased compared to that in wild-type cells, whereas the K_d of FAs in the front was not changed significantly (Fig. 5B). Thus, unlike wild-type cells, in which FAs at the rear disassembled faster than those at the front, kindlin-2 KO cells exhibited indistinguishable K_d values of front and rear FAs, suggesting that loss of kindlin-2 impairs the front-rear polarization of FA disassembly. We also tracked FA dynamics in the front and rear of kindlin-2 KO HT1080 cells that re-expressed GFP-tagged kindlin-1, kindlin-2 or LK(47)/AA by using GFP-tagged kindlin proteins as markers for FAs. Similar to wild-type HT1080 cells (Fig. 5A) – and consistent with the ability of kindlin-1, kindlin-2 or -LK(47)/AA to restore rear retraction and coordinated migration in kindlin-2 KO cells (Fig. S3, Movie 8) – these cells exhibited similar front-rear polarization of FA dynamics (i.e. faster FA disassembly at the cell rear) (Fig. 5C,D, Movie 10). This indicates that (1) kindlin-1 is functionally similar to kindlin-2 in this process when expressed at an appropriate level and (2) LK47-mediated actin-binding is not essential for establishment of this front-rear polarization of FA dynamics.

Because cellular contractile machinery has been implicated in FA disassembly and trailing edge detachment during cell migration (Ridley et al., 2003; Gupton and Waterman-Storer, 2006; Parsons et al., 2010), we sought to test whether loss of kindlin-2 affects actomyosin contractility at FAs. To do this, we employed TLN HP35, a FRET-based talin tension sensor that responds to actomyosin-dependent mechanical force at FAs (Austen et al., 2015). As it is known that serum stimulates stress fiber formation in serum-deprived cells, leading to increased actomyosin-dependent tension (Burridge and Guilluy, 2016), we tested whether kindlin-2 is required for serum-induced actomyosin contractility at FAs. Five minutes after serum stimulation, concurrent with stress fiber formation (Fig. 6A), a significant increase in the fluorescence lifetime of TLN HP35 (τ_TLN HP35=2.48±0.120 ns, n=7) was detected compared to the zero-force control (TLN CTL) (τ_TLN CTL=2.07±0.025 ns, n=10) (Fig. 6B, Fig. S2). This corresponded to a FRET efficiency of 14.3%, ∼50% reduction of FRET efficiency (27.8%) of the zero-force control (Fig. 6B), indicating that actomyosin-dependent tension at FAs developed upon serum stimulation. Fifteen minutes after serum stimulation, fluorescence lifetime of TLN HP35 (τ_TLN HP35=2.23±0.068 ns, n=9) was slightly decreased (FRET efficiency =22.2%) and remained virtually unchanged for the duration of measurements (τ=2.30±0.038 ns, FRET efficiency=19.1%, n=16, and τ=2.22±0.046 ns, FRET efficiency=22.2%, n=7 for 30-min and 60-min serum stimulation, respectively) (Fig. 6B, Fig. S2). Importantly, upon serum stimulation, fluorescence lifetime of TLN HP35 (τ_TLN HP35=2.13±0.028 ns, n=10) in kindlin-2 KO cells barely changed compared to that of zero-force control (τ_TLN CTL=2.12±0.041 ns, n=9) essentially showing the same FRET efficiency (28% and 28.3%) for TLN HP35 and TLN CTL, respectively (Fig. 6B, Fig. S4). This suggests that kindlin-2 is crucial for serum-induced actomyosin contraction at FAs.

In the absence of serum stimulation, the fluorescence lifetime of TLN HP35 (τ_TLN HP35=2.07±0.016 ns, FRET efficiency=23.4%, n=12) in HT1080 cells was only slightly higher than that of the zero-force control TLN CTL (τ_{TLN CTL}=1.95±0.012 ns, FRET efficiency=27.8%, n=11) (Fig. 6C), indicating a relatively low basal tension at FAs under this condition. Under the same conditions, in kindlin-2 KO cells the fluorescence lifetime of TLN HP35 (τ_TLN HP35=2.10±0.022 ns, FRET efficiency=23.4%, n=10) was also only slightly higher than that of the zero-force control TLN CTL (τ_TLN CTL=1.94±0.015 ns, FRET efficiency=29.5%, n=7) (Fig. 6C), suggesting that the relatively low basal tension at FAs was not noticeably affected by the loss of kindlin-2. Thus, although kindin-2 is crucial for serum stimulation-induced tension at FAs, it did not exert a significant effect on basal tension levels at FAs.

We next assessed the effect of serum stimulation on cell migration and FA turnover. After serum deprivation, wild-type cells appeared smaller and elongated (Fig. S5A, Movie 11), reminiscent of the shape of kindlin-2 KO cells (Fig. 1). During cell migration, the trailing edge of wild-type cells was stretched and cell motility was slow (Movie 11). Notably, serum deprivation markedly reduced FA disassembly at the cell rear (Fig. 6E,F, Movie 13), which was again similar to the phenotypes observed in kindlin-2 KO cells (Fig. 5A, Movie 9). Upon addition of serum to the culture medium, cell motility was increased (Fig. S5B, Movie 11) and serum deprivation-induced inhibition of FA disassembly at the cell rear was released (Fig. 6D,E, Movie 13), indicating a link between serum stimulation and increased FA disassembly at the cell rear. Kindlin-2 KO cells cultured under serum-deprived conditions were much smaller and even became rounded (Fig. S5A, Movie 12), and formed fewer FAs (data not shown). Serum stimulation promoted the motility of kindlin-2-deficient cells (i.e. Smean and Smax) with decreased directional persistence compared to that of wild-type cells (Fig. S5B, Movie 12), which was consistent with the results obtained under normal resting conditions in the presence of serum (Fig. 4). Taken together, these data indicate that serum stimulation promotes rear FA disassembly and cell migration.

We next sought to investigate the molecular mechanism by which kindin-2 functions in serum-induced regulation of actomyosin contraction. We have recently found that kindlin-2 forms a complex with myosin light chain kinase (MYLK, hereafter referred to as MLCK) (Guo et al., 2018), which is known to catalyze myosin regulatory light chain (MYL2, hereafter referred to as MLC) phosphorylation at Ser19 and Thr18 (pMLC^Ser19/Thr18) (Itoh et al., 1992) and regulate actomyosin contraction (Adelstein, 1982). Consistent with this, MLCK was readily immunoprecipitated together with GFP-kindlin-2 or GFP-kindlin-1, but not with GFP control (Fig. S6). Furthermore, serum stimulation of wild-type HT1080 cells significantly increased the association between kindlin-2 and MLCK (Fig. 7A) and the level of pMLC^Ser19/Thr18 (Fig. 7B, compare lanes 2 and 1). By marked contrast, no significant increase of pMLC^Ser19/Thr18 was detected in kindlin-2 KO cells in response to serum stimulation (Fig. 7B, compare lanes 4 and 3), suggesting that kindlin-2 is required for serum-induced phosphorylation of MLC. Furthermore, when seeded on soft matrices, which reportedly downregulates actomyosin contractility (Mih et al., 2012), serum-stimulation-induced MLC phosphorylation was also reduced (Fig. 7C, compare lanes 4 and 3). Together, these observations explain why kindlin-2-deficient cells were incapable to develop actomyosin-dependent tension at FAs in response to serum stimulation (Fig. 6).

Although we have shown previously that kindlin-2 forms a complex with MLCK (Guo et al., 2018), it was not known whether the interaction is direct or indirect. To test this, we expressed and purified glutathione S-transferase (GST)-tagged kindlin-2 together with maltose-binding protein (MBP)-tagged MLCK, and analyzed their interaction. The results showed that GST-kindlin-2 was readily pulled down by MBP-MLCK (Fig. 7D, lane 6) but not MBP (Fig. 7C, lane 5), suggesting that kindlin-2 directly interacts with MLCK.

Next, we sought to map the kindlin-2 domain that mediates MLCK-binding. GST pulldown assay demonstrated that kindlin-2 F0-F1 domains (Fig. 7E, lane 4), like kindlin-2 (Fig. 7E, lane 2), were sufficient to pull down MLCK, whereas F0, F2 or F2-F3 failed to do so (Fig. 7E, lanes 3, 5 and 6), suggesting that F1 is crucial for MLCK-binding. To further test this, we expressed a GFP-tagged kindlin-2 fragment containing amino acid (aa) residues 1–167, i.e. GFP-F0F1 (1–167), which encompasses the F0 and F1 N-terminal regions before the long inserted sequence (aa 168-217) in the F1 lobe (Li et al., 2017) and GFP as a control in HT1080 cells. Co-immunoprecipitation showed that GFP-F0F1 (1–167) but not GFP alone readily bound MLCK (Fig. 7F, compare lanes 4 and 3). Overexpression of GFP-F0F1 (1–167) significantly inhibited the interaction between endogenous kindlin-2 and MLCK (Fig. 7F, compare lane 4 with lane 3) and serum-induced MLC phosphorylation (Fig. 7H, compare lanes 3 and 1). Importantly, similar to loss of kindlin-2 (Fig. 4C; Fig. S2), overexpression of GFP-F0F1 (1–167) impaired rear retraction, reduced directional persistence and increased the speed of migration (Fig. 7G, Movie 14), although kindlin-2 and paxillin were detected in the retracting rear of these cells (Fig. 7I and J). Collectively, these results suggest that the interaction between kindlin-2 and MLCK is crucial for regulation of cell migration.

The results described above suggest that the interaction between kindlin-2 and MLCK promotes MLC phosphorylation and, consequently, actomyosin contraction, rear FA disassembly and retraction, as well as directional persistence. This implies that inhibition of MLCK results in migration defects similar to those caused by kindlin-2 deficiency. To test this, we treated HT1080 cells with the MLCK inhibitor ML-7 (Webb et al., 2002; Lou et al., 2015). As expected, the level of pMLC^Ser19/Thr18 was reduced upon ML-7 treatment (Fig. 8A). Compared to control cells, ML-7-treated cells showed decreases and increases in cell spreading area and elongation index, respectively (Fig. 8B), both being reminiscent of kindlin-2-deficient cells (Fig. 2). Furthermore, like kindlin-2-deficient cells (Movie 2), the majority (∼90%) of ML-7-treated cells showed stretched trailing tails resulting from impaired trailing edge detachment during migration (Fig. 8C,D, Movies 15,16). Similar to our observation in kindlin-2-deficient cells, Smean and Smax of migration were increased, and the directional persistence was reduced in response to inhibition of MLCK (Fig. 8E). Unlike wild-type HT1080 cells, kindlin-2 KO cells treated with ML-7 showed no changes in Smax and directional persistence, but a slight increase in Smean (Fig. 8E), suggesting that effects of ML-7 on cell migration largely depend on the presence of kindlin-2. Finally, FA disassembly at the cell rear was also significantly decreased in response to MLCK inhibition (Fig. 8F,G, Movie 17). Thus, the migration defects induced by inhibition of MLCK are reminiscent to those caused by loss of kindlin-2, i.e. increased migration speed, impaired rear retraction with slow FA disassembly at the rear and reduced directional persistence, which supports the notion that kindlin-2 functions in regulation of cell migration by – at least in part – facilitating MLCK function.

Cell migration is a dynamic and polarized process that requires adhesion at the cell front and adhesion release at the cell rear (Parsons et al., 2010; Cramer, 2013; Ladoux et al., 2016). This process is closely associated with continuous assembly and disassembly of FAs in the leading edge, and periodical sliding and disassembly of FAs at the trailing edge (Ridley et al., 2003; Broussard et al., 2008; Parsons et al., 2010). Siding and disassembly of FAs at the rear allow detachment of trailing edge, resulting in cell translocation and efficient migration. Therefore, the establishment and maintenance of front-rear asymmetric FA disassembly and sliding are crucial for efficient directional cell migration. Substantial evidence indicates that the front-rear polarization is elicited and achieved through regulation of cytoskeleton dynamics and orientation, FAs and the contractile machinery in response to various stimulations (Ladoux et al., 2016). However, the proteins that regulate this process are not yet completely understood. Non-muscle myosin II plays an essential role in establishing migratory front-rear polarity in mammalian cells and promoting rear retraction during cell migration (Vicente-Manzanares et al., 2007, 2008, 2011; Wilson et al., 2010). In this current study, we demonstrated that kindlin-2 – a broadly expressed and evolutionarily conserved component of FAs – is a key regulator of rear FA sliding and disassembly, and of rear retraction during cell migration.

By analyzing FA dynamics at the front and rear, we showed that FA disassembly is spatially polarized in migrating cells, i.e. at the rear, FAs exhibited adhesion sliding and a significantly higher disassembly rate as compared to those at the front, thereby facilitating effective trailing edge detachment and directional persistence (Fig. 5, Movie 9). Kindlin-2 deficiency impaired rear retraction (Movies 2–7), FA sliding and disassembly at the rear (Movie 9), resulting in loss of spatially polarized pattern of FA disassembly (Fig. 5). Kindlin-2-deficient cells exhibited a decrease in directional persistence, which depends on correct front-rear polarization (Fig. 4). These findings explain why the shape of kindlin-2-deficient cells is severely elongated (Fig. 2), and raise the interesting question as to how kindlin-2 exerts effects on front-rear polarization and directional persistence in migrating cells.

There is substantial evidence suggesting that actomyosin contractility plays crucial roles in establishing and maintaining front-rear polarity and rear retraction of the cell (Sanchez-Madrid and Serrador, 2009; Ladoux et al., 2016). We recently found that kindlin-2 associates with MLCK (Guo et al., 2018). In the current study, we have shown that kindlin-2 directly interacts with MLCK (Fig. 7C). Furthermore, we have assessed the role of kindlin-2 in regulation of localized tension at FAs. Serum is known to promote actomyosin-dependent forces (Burridge and Guilluy, 2016). Indeed, in response to serum stimulation, TLN HP35 in HT1080 cells demonstrated a significantly higher tension at FAs (Fig. 6B). This, however, was lost in kindlin-2-deficient cells (Fig. 6B), suggesting that kindin-2 is crucial for serum-induced actomyosin contractility at FAs. At the molecular level, serum stimulation promoted the interaction between kindlin-2 and MLCK, as well as the phosphorylation of MLC (Fig. 7A,B). In kindlin-2 KO cells, the phosphorylation of MLC in response to serum stimulation was impaired (Fig. 7B) – reminiscent of the reduced phosphorylation of MLC in wild-type cells grown on soft matrices (Fig. 7C), which are known to reduce actomyosin contractility (Mih et al., 2012). These data provide an explanation as to why kindlin-2-deficient cells failed to develop serum-induced actomyosin-dependent tension at FAs. Similar to myosin IIA-deficient fibroblasts (Vicente-Manzanares et al., 2007, 2008), kindlin-2-deficient cells showed defective rear retraction and impaired adhesion disassembly at the rear (Figs 4 and 5). Additionally, inhibition of MLCK in wild-type cells induced an elongated morphology with impaired rear-end detachment (Fig. 8). Furthermore, consistent with previous studies (Worthylake et al., 2001; Webb et al., 2002, 2004), inhibition of MLCK significantly reduced the rate of rear FA disassembly (Fig. 8). Of note, the effects chemical inhibition of MLCK has on cell migration largely depended on the presence of kindlin-2 (Fig. 8). Finally, similar to kindlin-2 deficiency, inhibition of the interaction between kindlin-2 and MLCK in response to GFP-F0F1 (aa 1–167) resulted in suppression of serum-induced myosin activation and cell migration defects (Fig. 7F–I). Collectively, these results suggest that kindlin-2 is crucial for the regulation of MLC phosphorylation and actomyosin contraction, rear retraction and directional persistence in cell migration. However, our preliminary experiments, using purified MLCK, failed to show a requirement of kindlin-2 in the biochemical activation of MLCK in vitro (Fig. S7). Furthermore, overexpression of constitutively active MLCK in kindlin-2 KO cells failed to reverse the migration defects induced by the loss of kindlin-2 (Fig. S8). Thus, the fact that loss of kindlin-2 impairs serum-induced MLCK-mediated MLC phosphorylation in cells probably reflects a function of kindlin-2 in facilitating MLCK-mediated MLC phosphorylation in cells – rather than a direct requirement of kindlin-2 in the biochemical activation of MLCK.

We found that, although loss of kindlin-2 did not significantly affect FA dynamics in the front, it markedly reduced FA sliding and disassembly at the rear of cells (Fig. 5B). How does kindlin-2 accomplish this spatially distinct regulation of FA disassembly during cell migration? One possibility is that kindlin-2 is only involved in regulating relatively fast FA disassembly – i.e. that at the cell rear – which is accompanied by fast forward sliding (Movie 10). Force-induced cell polarization mainly depends on FA sliding, which is linked to RhoA-driven actomyosin machinery (Goldyn et al., 2009). Kindlin-1 and -2, and the complex between ILK, PINCH and parvin are involved in the regulation of Rac and Rho activities (Has et al., 2009; Theodosiou et al., 2016; Bottcher et al., 2017; Sun et al., 2017; Yasuda-Yamahara et al., 2018; Chronopoulos et al., 2020; Kadry et al., 2018; Montanez et al., 2009). We recently found that RhoA activity is crucial for the formation of the kindlin-2–MLCK complex (Guo et al., 2018). However, we did not observe any significant alterations of the steady-state levels of Rac1 and RhoA activation in kindlin-2 KO HT1080 cells compared to those in wild-type HT1080 cells (Fig. S8C). We also analyzed RhoA activation in migrating cells by using a FRET-based RhoA activity sensor and did not observe a significant reduction of RhoA activation in kindlin-2 KO HT1080 cells compared to that in wild-type HT1080 cells (Fig. S8D). Loss of kindlin-2 appeared to slightly reduce serum stimulation-induced RhoA activation, although this was not statistically significant (Fig. S8E). Although the role of kindlin-2 in local – i.e. cell rear – activation of Rac1 and RhoA remains to be determined in future studies, previous research has shown that actomyosin contraction is a dominant force driving cell rear retraction when large forces are required (Cramer, 2013). Actomyosin contraction is increased in response to microtubule-induced release of Rho GEFs, resulting in increased FA disassembly (Zenke et al., 2004; Kwan and Kirschner, 2005). Thus, actomyosin contractility may help break adhesive interactions during cell migration (Jay et al., 1995), and an increase in actomyosin contractility is likely to be needed for FA disassembly and rear detachment (Broussard et al., 2008). In this regard, it is worth noting that kindlin-2 is crucial for serum-stimulated actomyosin contraction at FAs (Fig. 6B), although it is dispensable for the basal tension (Fig. 6C) under resting conditions. These results suggest that a locally increased actomyosin contractile activity, coupled with rear-edge retraction, is likely to be required to rip off rear FAs from the substratum in order to promote efficient cell migration (Webb et al., 2004; Lauffenburger and Horwitz, 1996). Interestingly, similar to kindlin-2 deficiency (Fig. 5), serum deprivation also selectively impaired FA disassembly at the cell rear, which was released after serum stimulation (Fig. 6D,E). The serum deprivation-induced impairment of rear FA disassembly is consistent with the notion that actomyosin activity is crucial for adhesion release at the rear to ensure efficient cell migration (Ladoux et al., 2016). Because spatially distinct FA dynamics correlate to directional migration (Mohl et al., 2012), this kindlin-2-mediated regulation of polarized FA disassembly provides an explanation as to why kindlin-2 deficiency significantly reduced directional persistence during cell migration (Fig. 4).

Cell migration is a complex process that requires the coordinated regulation of multiple processes. Although our study provides strong evidence that kindlin-2 exerts functions in promoting rear FA disassembly and directional persistence through regulation of actomyosin contraction, our results also support that other processes in which kindlin-2 is involved, such as regulation of cell surface integrin levels and cell substratum traction forces, are pertinent to cell migration as well. For example, consistent with a role of kindlins in the regulation of cell-surface integrin levels (Harburger et al., 2009; Margadant et al., 2013, 2012; Kruger et al., 2008; Qu et al., 2011; Schmidt et al., 2011), we found that, in response to loss of kindlin-2, cell surface levels of αVβ3 integrin were reduced, whereas those of α2 and β1 integrins were modestly increased (Fig. S1). Furthermore, loss of kindlin-2 markedly reduced the number of FAs (Fig. 3B), which also influences substratum traction, force exerted on the cell. These changes might also contribute to the migration defects found in kindlin-2-deficient cells. Indeed, migration speed, an important parameter of migratory behavior, is determined by both cellular contractile forces and substratum traction forces exerted on cell–ECM adhesions (Lauffenburger and Horwitz, 1996). Theoretically, the ratio of contractile force to substratum traction force influences migration speed in a biphasic manner (Dimilla et al., 1991). Maximal migration speed appears to correlate inversely with contractile force (Oliver et al., 1994). Kindlin-2 deficiency not only reduced directional persistence but also increased maximum migration speed (Fig. 4C, Fig. S1). Thus, the increased migration speed associated with kindlin-2-deficient cells might result not only from reduced contractile forces but also from altered cell surface integrin levels (Fig. S1) and weakened binding activity between integrin and ligand − as kindlin-2 is known to be crucial for integrin activation and cell-ECM adhesion (Shi et al., 2007; Larjava et al., 2008; Li et al., 2017; Ma et al., 2008).

Although they share considerable sequence similarity, kindlin-2 and -1 are encoded by different genes (Larjava et al., 2008). Furthermore, the tissue expression pattern and cellular functions of these two kindlins differ significantly (Rognoni et al., 2016; Plow et al., 2009; Bandyopadhyay et al., 2012). In this study, we also investigated the role of kindlin-1 in the regulation of cell migration. Unlike that of kindlin-2, depletion of kindlin-1 did not significantly affect cell shape and cell spreading in HT1080 cells (Fig. 1). This could be due to the difference in either functions or expression levels of these two kindlins. When GFP-tagged kindlin-1 was expressed in kindlin-2 KO cells at a level similar to GFP-kindlin-2, GFP-kindlin-1, like GFP-kindlin-2, could functionally rescue defects of kindlin-2 KO cells in cell spreading (Fig. 2), rear retraction (Movie 8) and rear FA disassembly (Fig. 5C, Movie 10). Moreover, MLCK can co-precipitate with either GFP-kindlin-1 or -2 (Fig. S6). Thus, the defects in cell spreading and migration caused by lack of kindlin-2 in HT1080 cells are probably due to the fact that kindlin-1 levels are insufficient to counteract this loss and to support these processes. Moreover, it appears that kindlin-1 and -2 share considerably functional redundancy in the regulation of these processes. Our findings, however, do not rule out that the functions of kindlin-1 or -2 may differ in regard to other processes. Thus, the vastly different consequences caused by genetic ablation of kindlin-1 or kindlin-2 are likely to reflect the differences regarding their function(s) in certain cellular processes but also their expression pattern in cells and tissues (Ussar et al., 2008; Postel et al., 2013; Huet-Calderwood et al., 2014; Lai-Cheong et al., 2008; Margadant et al., 2013; He et al., 2011; Harburger et al., 2009).

The CRISPR/Cas9 plasmid pSpCas9n (BB)-2A-GFP (PX461) was obtained from Addgene (#48140), deposited by Feng Zhang (Ran et al., 2013), to which the following oligonucleotides were annealed and ligated: kindlin-1 guide A1: 5′-caccgCTTCATTGGGATGGTCAACG-3′; kindlin-1 guide A2: 5′-aaacCGTTGACCATCCCAATGAAGc-3′; kindlin-1 guide B1: 5′-caccgACGTCACACTGAGAGTATC-3′; kindlin-1 guide B2: 5′-aaacGATACTCTCAGTGTGACGTc-3′; kindlin-2 guide A1: 5′-caccgCACCTCGCCGGTCACTCTCA-3′; kindlin-2 guide A2: 5′-aaacTGAGAGTGACCGGCGAGGTGc-3′; kindlin-2 guide B1: 5′-caccgTGGAGGCGTGATGCTTAAGCTGG-3′; kindlin-2 guide B2: 5′-aaacCCAGCTTAAGCATCACGCCTc-3′. Lowercase letters indicate overhang sequences for ligation into Bbs1-processed vector; uppercase letters indicate target site sequence.

Primers used to determine the disruption of human FERMT1 (kindlin-1) or FERMT2 (kindlin-2) are as follows:

kindlin-1 KO fw: 5′-TGGTCTTGGAGAAGTGGGTG-3′; kindlin-1 KO re: 5′-CCAGCCATTAGTGGTGATGC-3′; kindlin-2 KO fw: 5′-GAGTGACGCTTACGAGAGCC-3′; kindlin-2 KO re: 5′-GCATTTGCGAGATGCCAAGA-3′.

Constructs encoding respective GFP-tagged wild-type kindlin-2, W619Q mutant and LK(47)/AA kindlin-2 have previously been described (Li et al., 2017; Guo et al., 2018). GFP-tagged kindlin-1 was subcloned into pEGFP-C. The FRET-based talin tension-sensor system (TLN CTL, TLN Y, and TLN HP35) used to measure tension generated by actomyosin contractility was a kind gift from Dr Carsten Grashoff (Institute of Molecular Cell Biology, Münster, Germany) (Austen et al., 2015). The FRET-based RhoA sensor Clover-RhoA-PKN-mRuby2 was modified from a kind gift from Dr Michiyuki Matsuda (Department of Pathology and Biology of Diseases, Kyoto University, Japan) (Aoki and Matsuda, 2009).

The CRISPR/Cas9 system targeting kindlin-1 or kindlin-2 was transfected to human HT1080 fibroblasts using Lipofectamine 3000 transfection kit (Invitrogen) when the cell confluence had reached ∼80%. Twenty-four hours after transfection, single cells were sorted into 96-well plates according to expression of GFP. Single clones were isolated and examined to determine disruption of the targeted locus.

Wild-type, kindlin-2 KO or kindlin-1 KO HT1080 cells were cultured in minimum essential medium (MEM; CORNING) supplemented with 10% fetal bovine serum (FBS, PAN BIOTECH), 1× non-essential amino acid (NEAA) solution (Gibco), and penicillin-streptomycin (100 U/ml) solution (HyClone Laboratories). The cells were digested using 0.25% trypsin-EDTA (Gibco) and passaged at 1:8 and incubated at 37°C (5% CO₂) for 48 h. Human podocytes, mesenchymal stem cells and A549 lung adenocarcinoma cells were cultured as described previously (Guo et al., 2018, 2019; Sun et al., 2017). For serum stimulation assays, cells were cultured in serum-deprived medium for 24 h, and then were stimulated with 10% FBS for 5, 15, 30 or 60 min. For the ML-7 treatment, 15 or 30 μM ML-7 (Selleck, #S8388) was used to treat cultured cells for 2 h.

Western blotting was performed as previously described (Guo et al., 2018). In brief, whole-cell protein extracts were obtained using 1× SDS (50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol). The concentration of total proteins was determined using BCA protein assay kit (Pierce). 20–30 µg protein was loaded per lane. Proteins were separated by SDS-PAGE and transferred to PVDF membrane (Millipore). The membranes were blocked with 5% BSA in Tris-buffered saline (50 mM Tris-HCl and 150 mM NaCl pH 7.4) containing 0.1% Tween 20 for 1 h at room temperature, followed by overnight incubation at 4°C with specific primary antibodies against kindlin-1 (1:1000; Cell Signaling Technology, #36734S), kindlin-2 (1:1000; proteintech, #11453-1-AP), MLC2 (1:1000 in 5% BSA; Cell Signaling Technology, #3672S), pMLC2^Ser19/Thr18 (1:1000 in 5% BSA; Cell Signaling Technology, #3674S), MLCK (1:10,000; Sigma, #M7905), and GAPDH (1:10,000; ZSGB-BIO, #TA-08). After washing and incubation with the appropriate horseradish peroxidase (HRP)-conjugated anti-mouse (1:10,000; #711-005-152) or anti-rabbit (1:10,000; #715-005-151) secondary antibody (both Jackson ImmunoResearch), blots were developed using an ECL kit (Bio-Rad) and then exposed to X-ray film (Super RX-N-C, Fuji Film). The images were scanned using an imaging scanning system (L365, EPSON Scan).

Cells were lysed with immunoprecipitation (IP) lysis buffer (Beyotime, #P0013) containing 1 mM PMSF. After incubation on ice for 30 min and centrifugation for 15 min at 4°C, the supernatants were collected. Total protein (0.5 mg) from each sample was used for co-immunoprecipitation assay. The cell lysates were pre-cleaned by incubation with 15 µl of Protein A/G plus beads (sc-2003, Santa Cruz) at 4°C for 30 min, followed by addition of 1 µl anti-kindlin-2 antibody (3A3.5) or 3 µl of unspecific mouse IgG (Santa Cruz, sc-2025) as a negative control and incubation at 4°C for 2 h. The samples were then mixed with Protein A/G PLUS-Agarose beads (15 µl/sample), incubated overnight, followed by washing (5×) with lysis buffer. The precipitates were eluted with 2× Laemmli loading buffer and subjected to western blotting with antibodies as specified.

Direct association of kindlin-2 with MLCK was analyzed by using recombinant glutathione S-transferase (GST)-tagged to kindlin-2 and MBP-MLCK (also containing a His-tag) fusion proteins in pulldown assays as previously described (Guo et al., 2018). Briefly, 5 μg of purified MBP-MLCK fusion protein was incubated with 10 µg of GST-kindlin-2 [GST-F0 (aa 1–98), GST-F0-F1 (aa 1–280), GST-F2 (aa 281–565) and GST-F2-F3 (aa 281–680), respectively] or GST (as a negative control) bound to 20 µl of Glutathione Sepharose 4B (GE Healthcare) in the binding buffer (PBS supplemented with 1% TritonX-100; total mixture volume=1.5 ml). Reciprocally, 5 μg of purified GST-kindlin-2 fusion proteins were mixed with 6 µg of MBP-MLCK or MBP (as a negative control) bound to 20 µl of Amylose Resin (NEB) in the binding buffer (total mixture volume=1.5 ml). The respective mixtures were incubated at 4°C for 2 h, followed by washing once with binding buffer and twice with PBS. After the last wash, the beads were mixed with 25 µl of 2× sample buffer and boiled at 95°C for 5 min. MBP- (or GST-) tagged proteins bound to GST-kindlin-2 (or MBP-MLCK) were detected by western blotting with anti-His (or anti-GST) antibodies.

Cells were seeded in Nunc Lab-Tek II chambered coverglasses (Merck, Nalge Nunc, 3000 cells/chamber) and placed in a heated and air humidified chamber built in a Nikon inverted microscope TE2000E. Phase-contrast time-lapse imaging was carried out at 1.5-min or 2-min intervals for 1–3 h for HT1080 cells and at 4-min intervals for 1–3 h for human podocytes and A549 cells. Images were captured on the microscope with a 10× Ph1 objective (0.65 μm/pixel), perfect focus system (PFS), and Hamamatsu C-11440-22cU camera controlled by NIS-elements software (Nikon). Image stacks were processed and migrating cells were manually tracked using ImageJ software (NIH). The mean and maximum migration speed (Smean and Smax, respectively) as well as directional persistence were analyzed using the Chemotaxis and Migration Tool from ibidi.

Cells seeded on coverslips coated with 5 µg/ml fibronectin or 5 µg/ml collagen were fixed with 4% PFA+0.25% Triton-X 100 and washed 3× with PBS. 50 mM NH₄Cl was used to decrease the background fluorescence. After washing 3× with PBS and blocking with the blocking buffer (5% BSA with 0.1% Triton-X 100 in PBS), the cells were incubated with antibodies against paxillin (1:300; Cell Signaling Technology, #2542S) or kindlin-2 (1:500; clone 3A3.5), or Alexa Fluor 488-, 555- or 647-conjugated phalloidin for 1 h at room temperature. The cells stained with primary antibodies were then washed 3× with the blocking buffer and incubated with Alexa Fluor 488- or 647-conjugated goat anti-rabbit or mouse IgG (H+L) (Life Technologies) secondary antibodies. After washing 3× with blocking buffer and PBS, coverslips were mounted with Prolong Gold antifade reagent (Invitrogen, P36934). Images were acquired using a Nikon A1R confocal fluorescence microscope with a 20× dry objective, 100× oil objective, or a Nikon total internal reflection fluorescence (TIRF) microscope with a 60× or 100× oil objective.

Ferret's diameter along the long (length) and short (width) dimension of cells stained with Alexa Fluor 488- or 555-conjugated phalloidin (actin) was measured using image analysis software Image-Pro-Plus (Median Cybemetics). The elongation index was defined as the length-to-width ratio.

HT1080 wild-type or kindlin-2 KO cells stably expressing paxillin-mRuby2 were seeded into 8-well Nunc™ Lab-Tek™ II Chambered Coverglass (Thermofisher Scientific, Waltham, MA). Time-lapse TIRF imaging of FAs was performed as described (Stehbens and Wittmann, 2014). In brief, cells were imaged by using a Nikon TiE microscope equipped with a TIRF system using a CFI SR Apo TIRF 60xH (NA=1.49) oil immersion objective and an Andor iXon DU-897 EMCCD (Andor) under control of NIS-elements software (Nikon). Time-lapse fluorescence images of paxillin-mRuby2 excited by a 561 nm laser (or GFP-kindlin excited by a 488 nm laser) were collected every 45 s, and fluorescence intensity of mRuby2-paxillin or GFP-kindlin-2 at each FA was tracked using NIS-elements software (Nikon). Fluorescence intensity as a function of time was fitted with a sigmoidal and mono-exponential decay function during respective assembly and disassembly phase using SigmaPlot 12 (Systat Software Inc., San Jose, CA) and GraphPad Prism 6.0 (GraphPad Software, San Diego, CA) (Waters and Wittmann, 2014).

TLN HP35, the FRET-based talin tension sensor, has been described previously (Austen et al., 2015). In brief, the TLN HP35 probe contains the 35 aa-long villin headpiece peptide (HP35), an ultrafast-folding peptide that responds to mechanical forces, inserted into the unstructured linker region between the head and rod domains of mouse talin-1. A YPet/mCherry pair of fluorophores flanking the HP35 sequence allows FRET measurements, which indicate mechanical tension that talin bears. The fluorescence lifetime (τ) of YPet was measured to calculate the FRET efficiency at given conditions. For comparison, τ of two controls, i.e. the talin wild-type control TLN Y in which only YPet is C-terminally tagged to talin-1, and the zero-force control TLN CTL in which HP35 and the YPet/mCherry pair are tagged to the C-terminus of talin-1, were measured in parallel. Because forces are not exerted on HP35 tagged to the C-terminus of talin-1, the YPet/mCherry pair in TLN CTL exhibits FRET.

TCSPC-FLIM experiments were performed as described previously (Austen et al., 2015). In brief, HT1080 wild-type or kindlin-2 KO cells were transiently transfected with TLN CTL, TLN Y or TLN HP35, and incubated in MEM in the absence of FBS for 24 h. Cells were then activated with MEM supplemented with 10% FBS for the different time durations, i.e. 5 min, 15 min and 30 min. Cells were fixed with 4% PFA and fluorescence images were recorded using a Nikon A1R confocal microscope equipped with a 485 nm pulsed diode laser (PDL 800-D, PicoQuant), a photodetector (PMA Hybrid 40, PicoQuant), and a 100× oil objective. Detection was carried out over a time window of 44 ns after the excitation pulse with the TCSPC resolution of 25.0 ps.

For each group of cells, 7–16 cells were analyzed per experiment. Data analysis was performed using the SymPhoTime 64 software (PicoQuant). In brief, ‘n-Exponential Tailfit’ was selected as the fitting model, signals on FAs were selected as ROIs, and ‘calculated IRF’ was selected as initial fit to get the lifetime value (τ_amp). FRET efficiency was calculated as FRET efficiency=1-τ_DA/τ_D, where τ_D is the fluorescence lifetime of TLN Y and τ_DA is the fluorescence lifetime of TLN HP35 or TLN CTL.

Polyacrylamide hydrogels of different stiffness coated with collagen I (3 µg/ml) were prepared as published previously (Guo et al., 2018). Hydrogels containing 0.3% and 0.03% bis-acrylamide were regarded as stiff and soft matrices, respectively.

Cell surface expression levels of α2(CD49b), αVβ3(CD51/CD61) and β1(CD29) integrins in wild-type and kindlin-2 KO HT1080 cells were detected by using a BD FACS cytometer and analysed with FlowJo software as described previously (Theodosiou et al., 2016). Briefly, cells were plated on a fibronectin-coated (5 µg/ml) Petri dish overnight, trypsinized and washed with cold FACS buffer (30 mM Tris pH 7.4, 3% BSA, 180 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂, 3.5 mM KCl). 5×10⁵ cells were incubated with 3 µl fluorochrome-conjugated primary antibody (FITC anti-human CD49b, 1:33; FITC anti-human CD51/CD61, 1:33; or FITC Control IgG from eBioscience; Alexa Fluor 488 anti-human CD29, 1:33; and Alexa Fluor 488 control IgG from Biolegend) diluted in 100 µl of cold FACS buffer on ice for 1 h, and washed twice with FACS buffer before analysis.

Rac1 and RhoA G-LISA activation assays were performed following the manufacturer’s protocols (Cell Signaling). Additionally, the FRET sensor of RhoA activation (Raichu-RhoA-CR) from Michiyuki Matsuda was modified using Clover-RhoA-PKN-mRuby2, and was then utilized to assess RhoA activation in live cells as previously described (Guo et al., 2018).

In vitro MLCK kinase activity was assessed using recombinant MBP-MLCK and His-MLC2 (human RLC, NP291024.1) fusion proteins as previously described (Baumann et al., 2017). Briefly, 10 nM purified MBP-MLCK and 10 µM His-MLC2 were incubated with 500 µM ATP (Sigma, #A2382) in 20 mM HEPES pH 7.5, 50 mM NaCl, 10 mM MgCl₂, 1 mM DTT, with or without 40 nM calmodulin (CaM, Sigma)/1 mM CaCl₂ in the presence of 40 nM GST-tagged kindlin-2 or GST as control at 37°C for 5, 20 or 60 min. Reactions were quenched by addition of SDS sample buffer. Levels of phosphorylated and total MLC2 were analyzed by western blotting using antibodies against pMLC2 Ser19 (Cell Signaling Technology #3671) and His-tag (Transgen #AB102-02).

The effect of kindlin-2 on the interaction between MLCK and CaM was analyzed using MBP-MLCK pulldown assay following a previously described protocol (Guo et al., 2018). Briefly, 50 nM MBP-MLCK or MBP bound to amylose-resin was incubated with 400 nM GST-kindlin-2 or GST as control at 4°C for 60 min, followed by incubation with100 nM CaM (Sigma #C4874) and 10 mM CaCl₂ at 4°C for 30 min. Levels of CaM bound to MBP-MLCK in the presence or absence of GST-kindlin-2 were detected by western blotting with anti-CaM antibodies (Cell Signaling Technology, #35944).

Statistical data analysis for statistical significance was done using one- or two-tailed paired t-tests (GraphPad Prism version 5.00 for Windows, GraphPad Software, La Jolla, CA).

The authors are very grateful to Dr Carsten Grashoff for sharing the talin tension sensor system. The authors thank the SUSTech Core Research Facilities for excellent imaging assistance.