Palladin promotes assembly of non-contractile dorsal stress fibers through VASP recruitment

Actin filaments assemble into diverse protrusive and contractile structures to provide force for cellular processes, such as endocytosis, morphogenesis and migration. Three types of actin structures – lamellipodia, filopodia and stress fibers – are linked to cell migration. Force in lamellipodia and filopodia is produced through the polymerization of actin against the plasma membrane, whereas stress fibers are contractile structures where the force is generated by the sliding of bipolar bundles of myosin II along actin filaments (Pollard and Cooper, 2009; Bugyi and Carlier, 2010). In animal tissues, stress fibers are especially prominent in endothelial and epithelial cells, where they contribute to resistance against fluid shear and developmental processes, such as dorsal closure during Drosophila embryogenesis (Wong et al., 1983; Jacinto et al., 2002; Sato and Ohashi, 2005; Millán et al., 2010). Furthermore, stress fibers are involved in cell adhesion, migration, dermal wound closure, the regulation of receptor cluster dynamics in T-lymphocytes, and serve as precursors for the assembly of myofibrils in developing muscle cells (Bershadsky et al., 2006; Pellegrin and Mellor, 2007; Sanger et al., 2009; Sandbo and Dulin, 2011; Tojkander et al., 2012; Yi et al., 2012). In cultured animal cells, stress fibers also play an important role in mechanotransduction (Albiges-Rizo et al., 2009).

Stress fibers can be divided into three categories: dorsal stress fibers, transverse arcs and ventral stress fibers (Khatau et al., 2009; Tojkander et al., 2012). Ventral stress fibers are anchored to focal adhesions at each end and contain myosin II, which enables them to contract. Transverse arcs also contain myosin II and are contractile but they do not directly associate with focal adhesions. Dorsal stress fibers are anchored to focal adhesions at their distal ends but, unlike other stress fiber categories, they do not contain myosin II and, thus, cannot contract (Heath, 1983; Pellegrin and Mellor, 2007; Tojkander et al., 2012). Furthermore, certain animal cells exhibit another type of contractile actomyosin bundle, perinuclear actin cap, which covers the interphase nucleus (Khatau et al., 2009). Importantly, different stress fiber categories are dynamically interlinked. Ventral stress fibers are generated from a pre-existing network of dorsal stress fibers and transverse arcs. Dorsal stress fibers assemble through actin polymerization at focal adhesions, whereas arcs are generated by endwise assembly of Arp2/3-nucleated and formin-nucleated myosin-II-containing actin filaments at the lamellipodium (Hotulainen and Lappalainen, 2006; Shemesh et al., 2009; Fan et al., 2010; Burnette et al., 2011; Tojkander et al., 2012).

More than twenty functionally distinct proteins have been shown to associate with stress fibers, but the exact functions of the majority of the stress fiber components have remained obscure (Tojkander et al., 2012). Palladin is a large protein that, depending on the splice variant, is composed of three to five Ig-like domains and one to two proline-rich regions (Goicoechea et al., 2008). Palladin is critical for embryonic development, and inactivation of palladin results in defects in the development and motility of smooth muscle cells, primary neurons and fibroblasts (Parast and Otey, 2000; Boukhelifa et al., 2001; Jin et al., 2007; Liu et al., 2007). Palladin has also been linked to the invasive migration of cancer cells, but whether it promotes (Goicoechea et al., 2009) or inhibits (Chin and Toker, 2010) migration is a matter of controversy. In cultured cells, palladin colocalizes with the actin crosslinking protein α-actinin in stress fibers, and inactivation of palladin leads to disruption of the stress fiber network (Parast and Otey, 2000; Mykkänen et al., 2001; Rönty et al., 2004; Liu et al., 2007; Rönty et al., 2007). Palladin possesses actin filament crosslinking activity and, additionally, interacts with other regulators of actin dynamics, such as VASP, profilin 1 and 2a, α-actinin, ezrin, CLP36 (also known as PDLIM1), Src, and Eps8 (Mykkänen et al., 2001; Boukhelifa et al., 2004; Haikarainen et al., 2009; Rönty et al., 2004; Boukhelifa et al., 2006; Goicoechea et al., 2006; Rönty et al., 2007; Dixon et al., 2008; Maeda et al., 2009). However, the biological relevance and possible cellular roles of these various protein interactions have remained obscure. Consequently, whether palladin promotes stress fiber assembly by crosslinking actin filaments or by acting as scaffolding protein is not known. Furthermore, whether palladin contributes to the assembly of a specific stress fiber category, or whether it is involved in formation of all stress fiber types, has not been reported.

To elucidate the mechanism by which palladin contributes to the assembly of contractile actomyosin bundles, we examined its role in stress fiber formation in U2OS osteosarcoma cells. These cells provide an excellent model system for studying stress fiber assembly in vivo owing to the presence of a dynamic stress fiber network, where the assembly of all three stress fiber categories can be readily visualized (Hotulainen and Lappalainen, 2006; Tojkander et al., 2011; Oakes et al., 2012). We reveal that depletion of palladin specifically disrupts non-contractile dorsal stress fibers. We also discovered that palladin displays highly dynamic association–dissociation dynamics with stress fibers and promotes actin filament assembly in dorsal stress fibers by recruiting the actin polymerizing protein VASP. Thus, palladin functions as a dynamic scaffolding protein that promotes actin filament assembly at dorsal stress fibers in complex with VASP.

Previous studies have demonstrated that depletion of palladin in mouse embryonic fibroblasts and U251 cells leads to a disruption of the stress fiber network (Liu et al., 2007; Rönty et al., 2007). Because non-contractile dorsal stress fibers and contractile transverse arcs both contribute to the generation of the mature stress fiber network, it was important to determine whether palladin specifically contributes to the assembly of only a particular stress fiber subtype. We, therefore, used a human osteosarcoma cell line (U2OS) to elucidate the localization of palladin in the three types of stress fiber (Fig. 1). Western blot analysis revealed that at least three palladin splice variants (with apparent molecular masses of ∼90 kDa, ∼140 kDa and ∼200 kDa) are expressed in cultured U2OS cells (Fig. 2A). In addition, two other spice variants (with apparent molecular masses of 120 kDa and 175 kDa) might also be expressed in U2OS cells, based on western blot analysis using three other antibodies (supplementary material Fig. S1A,B). Immunofluorescence microscopy demonstrated that endogenous palladin was especially prominent in dorsal stress fibers but that it also localized to the two other categories of stress fiber – transverse arcs and ventral stress fibers (Fig. 1A; supplementary material Fig. S1C). During dorsal stress fiber assembly, palladin localized along the dorsal stress fibers and to focal adhesions at their distal ends (Fig. 1B). However, palladin was generally excluded from large mature focal adhesions (Fig. 1B). Thus, palladin is a component of all three types of stress fiber in U2OS cells but displays maturation-controlled localization to focal adhesions.

To examine the role of palladin in the assembly and maintenance of the stress fiber network, we depleted it from U2OS cells by using RNA interference (RNAi). The transfection of cells with siRNA oligonucleotides that are specific to palladin resulted in an ∼80% reduction of the three major palladin isoforms at the protein level after 3 days of transfection, whereas cells that had been transfected with control siRNA oligonucleotides did not display detectable alterations </emph>in the amount of each of the palladin isoforms (Fig. 2A; supplementary material Fig. S1A,B). Compared with non-transfected control cells, the stress fibers in cells in which palladin had been knocked down were thin and organized into an abnormal network, as visualized by using conventional light microscopy methods. Particularly, the focal-adhesion-attached dorsal stress fibers were diminished in length and thickness (Fig. 2B). Super-resolution analysis of wild-type and palladin knockdown cells, by using stimulated emission depletion microscopy (STED), revealed that palladin depletion resulted in abnormal connections between the remaining dorsal stress fibers and transverse arcs. Instead of T-shaped connections between the arcs and dorsal stress fibers, which were prominent in control cells, palladin knockdown cells often displayed intersections between multiple arcs and dorsal stress fibers (Fig. 2C; supplementary material Fig. S2A). These features might result from the diminished elongation and stability of dorsal stress fibers, but further studies are required to reveal the exact mechanistic basis of these defects (supplementary material Movies 1, 2).

Because of the large variation in cell shapes and stress fiber network architectures, it has traditionally been difficult to obtain quantifiable information concerning the specific roles of various stress fiber components. To examine the effects of palladin depletion on the stress fiber network more precisely, cells were plated on crossbow-shaped fibronectin micropatterns. Using these micropatterned chips, all cells are induced to exhibit roughly identical shapes that enabled quantification of the densities of different stress fiber types (supplementary material Fig. S2B). Both wild-type cells and palladin knockdown cells contained comparable amounts of transverse arcs (Fig. 2D,F). However, the amount of dorsal stress fibers was diminished by ∼70% in knockdown cells compared with control cells (Fig. 2D,E). Taken together, these data demonstrate that, in U2OS cells, palladin is specifically involved in the assembly of dorsal stress fibers.

The existence of focal adhesions and stress fibers in a three-dimensional (3D) extracellular environment has been a subject of controversy (Fraley et al., 2010), although recent studies have presented evidence that such structures exist in cells that display mesenchymal migration in 3D collagen matrix (Kubow and Horwitz, 2011; Petrie et al., 2012). Furthermore, the role of stress fibers in migration and in the morphogenesis of cells that have been embedded in a 3D extracellular matrix has remained obscure. Therefore, we examined the appearance and organization of stress fibers and focal adhesions in U2OS cells that had been embedded in 3D collagen matrix, and we then tested how the depletion of palladin affects these features. The visualization of actin filaments by using fluorescent phalloidin, and the labeling of myosin II with an antibody revealed that wild-type U2OS cells that had been embedded in a 3D collagen matrix, indeed, displayed actin- and myosin-II-containing bundles that resembled stress fibers (supplementary material Fig. S3). These actomyosin bundles were attached to vinculin-rich punctae at their ends (Fig. 3A; supplementary material Movie 3), which were reminiscent of the focal adhesion structures that have been observed previously in cells migrating in a 3D environment (Kubow and Horwitz, 2011). Similar to the cells migrating in a 2D environment, palladin that had been tagged with green fluorescent protein (GFP) localized to the actomyosin bundles (supplementary material Fig. S3).

Because palladin is important for the maintenance of proper stress fiber architecture in 2D conditions (Fig. 2), we examined the possible effects of palladin depletion in cells that had been cultured in a 3D collagen matrix. In contrast with control cells that had been transfected with scrambled siRNA oligonucleotides, the actin filament network in palladin-depleted cells was severely disorganized and stress-fiber-like actomyosin bundles were absent. Furthermore, vinculin localized to small punctae at the plasma membrane and did not form focal-adhesion-like structures at the tips of the cells (Fig. 3A; supplementary material Movie 4). Importantly, palladin knockdown cells that lacked actomyosin bundles exhibited a rather ellipsoid shape, whereas control cells that had been transfected with scrambled siRNA oligonucleotides were typically elongated with stress fibers running parallel to the long cell axis (Fig. 3B; supplementary material Fig. S4A). Furthermore, palladin knockdown cells displayed reduced contractility in a 3D collagen matrix (Fig. 3C; supplementary material Fig. S4B). Thus, U2OS cells also display stress fibers in a 3D environment. The formation of the focal-adhesion-attached actomyosin bundles is dependent on palladin, and these structures are crucial for cell morphogenesis.

Palladin crosslinks actin filaments in vitro and, additionally, interacts with many regulators of cytoskeletal dynamics (Mykkänen et al., 2001; Boukhelifa et al., 2004; Rönty et al., 2004; Boukhelifa et al., 2006; Goicoechea et al., 2006; Dixon et al., 2008; Maeda et al., 2009). It has been shown that palladin overexpression generates thick actin bars (see Materials and Methods; Boukhelifa et al., 2003; Rönty et al., 2004), which resembles the phenotype that is induced by the overexpression of proteins that are involved in actin filament nucleation and elongation, such as the formin-homology protein Dia1 and VASP (supplementary material Fig. S5A–D). Importantly, the actin filament bars that were induced by palladin overexpression were typically connected to vinculin-rich focal adhesions, from at least one end, but they did not accumulate myosin II, suggesting that they are unable to contract (Fig. 4; supplementary material Fig. S5E). These data suggest that the actin bars present in palladin-overexpressing cells are not thick transverse arcs or ventral stress fibers, instead, they resemble dorsal stress fibers, which are non-contractile and associate with focal adhesions at their distal ends (Tojkander et al., 2012).

Because palladin overexpression induced thick actin filament bars that were comparable to those induced by the overexpression of VASP (supplementary material Fig. S5), and because palladin has been previously shown to co-immunoprecipitate with VASP in Swiss 3T3 fibroblasts and to interact with VASP in vitro by using a blot overlay assay (Boukhelifa et al., 2004), we examined whether palladin could regulate the localization of VASP in cells. In U2OS cells, VASP colocalized with palladin along the non-contractile dorsal stress fibers and also, occasionally, in contractile transverse arcs and ventral stress fibers (supplementary material Fig. S6A). Upon overexpression of palladin, VASP also accumulated on thick actin filament bars (Fig. 4). Importantly, the depletion of palladin resulted in the severely diminished localization of VASP along the remaining dorsal stress fibers (Fig. 5A,B); however, VASP still localized to a subset of focal adhesions in palladin knockdown cells (Fig. 5A), which suggests that palladin recruits VASP, especially to dorsal stress fibers. It is important to note that, in these experiments, the intensity of the VASP staining in the dorsal stress fibers was normalized against the intensity of F-actin staining to ensure that the alterations in the stress fiber thickness did not affect the interpretation of the data.

To elucidate the relevance of VASP recruitment by palladin in the assembly of dorsal stress fibers, we examined the role of VASP in the assembly of dorsal stress fibers in U2OS cells. Transfection of U2OS cells with specific RNAi oligonucleotides led to the efficient depletion of VASP at a protein level (Fig. 5C). Similar to the palladin-depleted cells (Fig. 2), the VASP knockdown cells that had been plated on crossbow-shaped fibronectin micropatterns exhibited transverse arcs and ventral stress fibers, whereas the amount of dorsal stress fibers was diminished by ∼70% compared with control cells (Fig. 5D,E). The decrease in the amount of dorsal stress fibers did not result from the lack of focal adhesions. VASP-depleted cells contained similar amounts of vinculin-rich adhesions compared to control cells, although the adhesions appeared smaller in size in the absence of VASP (data not shown). Furthermore, VASP-depleted cells displayed similar abnormalities in a 3D collagen matrix compared to palladin knockdown cells (Fig. 5F). Taken together, these data suggest that palladin recruits VASP to dorsal stress fibers to promote the assembly of these stress fiber precursors.

Palladin harbors a proline-rich region, which has been shown to interact with VASP in vitro (Boukhelifa et al., 2004). To examine whether an interaction with VASP is required for palladin to promote actin filament assembly in cells, we first monitored the formation of thick actin bars upon overexpression of palladin. For these experiments, we generated a palladin mutant, where the two proline-rich stretches that are implicated in interactions with VASP were inactivated by replacing proline residues 81–84 (in the amino acid sequence 80-FPPPP-84) and proline residues 173, 175 and 176 (in the amino acid sequence 172-FPLPP-176) with alanine residues. The mutated residues are located outside of the domains that mediate the interactions of palladin with actin and α-actinin (Otey et al., 2005). Immunofluorescence microscopy and a co-immunoprecipitation assay revealed that these mutations specifically impaired the interaction between VASP and palladin in vivo (Fig. 6A) without affecting the localization of the protein to stress fibers (Fig. 6B). Importantly, the palladin mutant construct was significantly less efficient at inducing the formation of thick actin bars in cells, as compared with wild-type palladin that was expressed at similar levels (Fig. 6B,C). Both wild-type and mutant palladin were unable to efficiently generate thick actin bars in VASP knockdown cells (Fig. 6C), suggesting that palladin was capable of inducing the formation of thick actin bars only in the presence of VASP.

To examine whether an interaction between palladin and VASP is necessary for the assembly of dorsal stress fibers, we performed an RNAi rescue experiment that used wild-type palladin or the mutant palladin that was unable to interact with VASP. Expression of wild-type palladin that was refractory to siRNA-mediated knockdown efficiently rescued the loss of dorsal stress fibers in palladin knockdown cells that had been plated on crossbow-shaped fibronectin micropatterns, whereas rescue cells that expressed the mutant palladin contained a low amount of dorsal stress fibers, similar to palladin knockdown cells that did not express a rescue construct (Fig. 7). Thus, palladin recruits VASP to dorsal stress fibers and promotes actin filament assembly in these structures, at least partially through VASP.

The experiments described above suggested that palladin is essential for the recruitment of VASP to dorsal stress fibers. To elucidate whether these two proteins incorporate into stress fibers as a complex with each other, we examined the dynamics of palladin and VASP in dorsal stress fibers by using fluorescence recovery after photobleaching (FRAP). Previous studies have demonstrated that actin, myosin II and α-actinin-1 display distinct kinetics in stress fibers, α-actinin-1 being the most dynamic stress fiber component (Hotulainen and Lappalainen, 2006). FRAP experiments using GFP–VASP, GFP–α-actinin-1 and GFP–palladin revealed that palladin and VASP display nearly identical association and dissociation dynamics to dorsal stress fibers (t_½ = 1.28 s for VASP and t_½ = 1.18 s for palladin), and that the dynamics of their association with and dissociation from stress fibers are more rapid than that of α-actinin (Fig. 8A,B). Thus, palladin and VASP display highly dynamic interactions with stress fibers and could associate with and dissociate from dorsal stress fibers as a complex with each other.

Palladin is a large, multidomain protein, which has an important role in the assembly of actin-rich structures, such as podosomes, invadopodia and stress fibers (Parast and Otey, 2000; Goicoechea et al., 2009; Goicoechea et al., 2013). Consequently, palladin regulates cell migration to drive tumor invasion and metastasis (Goicoechea et al., 2009; Goicoechea et al., 2010); however, the mechanisms by which palladin contributes to the assembly and/or dynamics of cellular actin filament arrays has remained elusive. Here, we approached this question by investigating the mechanism by which palladin contributes to the generation of stress fibers in motile cells.

Although palladin localizes to all three stress fiber categories in U2OS cells, it is specifically required for the assembly of dorsal stress fibers. This is evidenced by experiments where palladin depletion resulted in the specific disappearance of dorsal stress fibers, whereas the arcs and ventral stress fibers were still present in palladin knockdown cells. Furthermore, the localization of myosin II to stress fibers is unaffected in palladin-deficient cells (supplementary material Fig. S6B), suggesting that arcs and ventral stress fibers, which are myosin-II-containing contractile actin structures, are not affected by palladin depletion. By contrast, dorsal stress fibers are non-contractile and contain myosin II only in the points of connection with arcs (Tojkander et al., 2011). It is important to note that although arcs and dorsal stress fibers serve as precursors of ventral stress fibers (Hotulainen and Lappalainen, 2006), the latter category of stress fibers could still form when dorsal stress fiber assembly was compromised by palladin depletion. Therefore, we propose that the remaining thin dorsal stress fibers in palladin knockdown cells can still link arcs to focal adhesions, although the dynamics and architecture of the transverse-arc–dorsal stress fiber connections are abnormal in the absence of palladin (Fig. 2C). The presence of proper dorsal stress fibers appears to become more important when cells are migrating in a 3D environment. This is because, in U2OS cells that had been cultured in 3D collagen matrix, palladin depletion resulted in nearly complete loss of the stress fiber network and consequent defects in cell elongation. However, it is important to note that palladin and its interaction partner VASP also contribute to the assembly and maintenance of other actin filament structures in cells. Thus, the effects of depleting palladin and VASP on cell elongation in a 3D collagen matrix might, in part, result from defects in the assembly of these other actin-dependent structures.

Several lines of evidence suggest that palladin promotes the assembly of dorsal stress fibers by recruiting VASP to these structures. First, earlier biochemical studies have demonstrated that palladin binds to VASP (Boukhelifa et al., 2004). Second, palladin depletion resulted in the loss of VASP in dorsal stress fibers (Fig. 5A,B). Third, VASP depletion led to a similar diminishment of dorsal stress fibers compared with palladin depletion (Fig. 5D–F). Fourth, VASP localized to thick actin filament bars that had been induced by palladin, and a palladin mutant that was unable to interact with VASP did not rescue the dorsal stress fiber phenotype that was induced by palladin depletion (Figs 4 and 7). VASP and palladin also displayed nearly identical rapid dynamics in dorsal stress fibers, suggesting that they might associate with dorsal stress fibers in a complex with each other (Fig. 8). U2OS cells also express at least one other VASP family protein, Mena, which localizes to focal adhesions and dorsal stress fibers, and interacts with the proline-rich region of palladin (supplementary material Fig. S7A,C). Depletion of Mena resulted in a modest decrease in the number of dorsal stress fibers (supplementary material Fig. S7B–D); thus, other Ena/VASP family proteins also interact with palladin and contribute to the assembly and maintenance of dorsal stress fibers. However, at least in U2OS cells, VASP appears to be the most important member of the family with respect to stress fiber generation.

Ena/VASP family proteins promote actin filament assembly through a complex mechanism that involves processive filament elongation at the barbed end and blocking the binding of capping proteins (Breitsprecher et al., 2011; Hansen and Mullins, 2010). Furthermore, VASP can serve as an actin filament bundling protein (Schirenbeck et al., 2006). Although it is possible that VASP functions as a filament crosslinker in dorsal stress fibers, we believe that it contributes to dorsal stress fiber integrity by promoting actin filament assembly. This is because the overexpression phenotypes of VASP and palladin are more similar to those of actin filament assembly factors than to those of actin filament crosslinkers. Furthermore, VASP and palladin display a highly dynamic association with stress fibers that is more consistent with actin filament assembly than crosslinking. However, further studies are required to reveal the exact mechanism by which the palladin–VASP complex contributes to dorsal stress fiber assembly and to elucidate the possible interactions of palladin with other Ena/VASP family proteins.

Although VASP recruitment along dorsal stress fibers depends on palladin, VASP can associate with the leading edge of cells and with focal adhesions in the absence of palladin. Thus, other VASP binding partners appear to be sufficient for recruiting VASP to these actin-rich structures. Indeed, recent studies have demonstrated that zyxin is important for VASP recruitment to focal adhesions, as well as for the stretch-induced accumulation of VASP at ventral stress fibers (Hoffman et al., 2006; Smith et al., 2010; Hoffman et al., 2012). Therefore, VASP can be targeted to specific cellular structures by at least two independent pathways, which involve interactions with palladin and zyxin. We, thus, propose that VASP contributes to the formation of the stress fiber network by promoting actin filament assembly both at focal adhesions (in complex with zyxin) and along the dorsal stress fibers (in complex with palladin).

Our findings bring new light to our previous proposal that the assembly of dorsal stress fibers is a complex process that requires actin filament assembly both at focal adhesions and along the actin filament bundle of the dorsal stress fiber. Our previous FRAP studies have revealed that dorsal stress fibers elongate through ‘vectorial’ actin polymerization at focal adhesions, but that they also undergo actin dynamics along their entire length. Dia1 drives actin filament nucleation and assembly in U2OS cells to promote dorsal stress fiber elongation at focal adhesions (Hotulainen and Lappalainen, 2006), whereas the present study provides evidence that the palladin–VASP complex is involved in regulating actin dynamics along dorsal stress fibers. In the absence of palladin, dorsal stress fibers are thin and their dynamic association with arcs is defective (Fig. 8). Thus, palladin–VASP-promoted actin dynamics appear to be important for the maintenance of dorsal stress fibers, as well as for their ability to make proper connections with transverse arcs during the maturation of the stress fiber network. α-actinin-1, in addition to formin proteins, palladin and VASP, also contributes to the formation of dorsal stress fibers in U2OS cells (Oakes et al., 2012; Kovac et al., 2013). Palladin interacts with α-actinin-1 and CLP36, and these interactions are essential for palladin recruitment to the stress fiber network (Rönty et al., 2004; Maeda et al., 2009). Because α-actinin-1 is specifically enriched in the dorsal stress fibers of U2OS cells (Kovac et al., 2013), this might provide a possible mechanism for the enrichment of palladin in dorsal stress fibers.

It is important to note that in addition to VASP, actin and α-actinin, palladin interacts with many other proteins (Goicoechea et al., 2008). Although our studies provide evidence that interactions with VASP are important for the role of palladin in the assembly and dynamics of dorsal stress fibers, it is also possible that other activities of palladin contribute to this process. Thus, it will be important to identify mutations in palladin that impair its actin filament crosslinking activity and interactions with other ligands to elucidate the possible roles of these interactions and activities in the assembly of dorsal stress fibers. Additionally, it is possible that other palladin interactions are important for different actin-dependent processes, such as podosome or invadopodia assembly and dynamics; thus, future studies are needed to reveal the interactome of palladin in different actin-based cellular processes.

Human osteosarcoma (U2OS) cells were grown at 37°C under a humidified atmosphere (5% CO₂) in Dulbecco's modified Eagle's medium (DMEM) that was supplemented with 10% fetal bovine serum (Gibco), 2 mM l-glutamine, penicillin and streptomycin (Sigma-Aldrich). Lipofectamine 2000™ (Invitrogen) was used, according to the manufacturer's instructions, for transient transfections. After transfection, the cells were incubated for 18–24 h and either fixed with 4% paraformaldehyde (PFA) or used for live-cell imaging. In order to deplete specific proteins of interest, the cells that had been plated onto 35-mm dishes were transfected with 2100 ng of preannealed 3′ Alexa-Fluor-488 or -647-labeled oligonucleotide duplexes by using GeneSilencer's siRNA transfection reagent (Gene Therapy Systems) according to the manufacturer's instructions. Cells were incubated for 72 or 96 h to allow for efficient depletion of the target protein.

After transient transfection, the cells were incubated for 24 h and re-plated before imaging on 10-µg/ml fibronectin-coated glass-bottomed dishes (MatTek Corporation). The time-lapse images were acquired by using a 3I Marianas imaging system (3I intelligent Imaging Innovations), consisting of an inverted spinning disk confocal microscope Zeiss Axio Observer Z1 (Zeiss) and a Yokogawa CSU-X1 M1 confocal scanner. A ×63/1.2 W C-Apochromat Corr WD = 0.28 M27 objective (Zeiss), appropriate filters, a heated sample environment (37°C) and CO₂ control were used. The SlideBook 5.0 software (3I intelligent Imaging Innovations) was used for image acquisition. A Neo sCMOS (Andor) camera was used for image recording. Deconvolution of the time-lapse videos was performed using AutoQuant AutoDeblur 2D blind Deconvolution (AutoQuant Imaging).

To analyze the kinetics of α-actinin, palladin and VASP in stress fibers, cells were transfected with the GFP or enhanced (E)GFP constructs and cultured for 24 h. For imaging, the cells were plated onto plastic dishes (Nunc plates, ThermoScientific) that had been coated with 10-µg/ml fibronectin. Confocal imaging was performed on a TCS SP5 Leica microscope equipped with Leica Confocal Software (LasAF 2.5.1) under a heated sample environment (37°C) and CO₂ control. For imaging of GFP, a 488-nm laser line and a HCX APO L 63x/0.90 W Lbd. bl. objective was used. After three pre-bleach images, three bleaching scans (0.113 s each) with 100% intensity of 476-nm (15 mW), 488-nm (70 mW) or 496-nm (15 mW) laser lines over the region of interest (5×15 µm) were performed. The fluorescence recovery was followed for 548.36 s after the bleaching but only the first 10 s are shown in the image because of the rapid recovery rates. The intensity of a neighboring non-bleached stress fiber was used to normalize the intensity of the bleached area. The bleached and control areas used for measurements during data analysis were outlined so that the diffuse soluble protein around stress fibers was excluded. Because only a narrow area containing the stress fiber (where the proteins were enriched) was included in the fluorescence recovery quantification analysis, the possible contribution of the ‘free cytoplasmic pool’ of these proteins at the selected region should be very small. To calculate the mean scatter plots, 10–14 different FRAP experiments were used. Finally, data were fitted with SigmaPlot 11.0 to double f = a×[1−exp(−b×x)]+c×[1−exp(−d×x)] exponential equations. t_½ was calculated as previously described (Tojkander et al., 2011).

To obtain full-length EGFP–palladin, the first 21 base pairs from the full-length palladin sequence (corresponding to NM_016081) were cloned into an EGFP-palladin vector that encoded amino acid residues 8–772 of palladin (previously described by Rönty et al., 2004) by using the following primers: 5′-CCGGCGGGCGAGGTATAAAGCCCGATACCTGCCCC-3′ and 5′-GTGACTCAGGCATGAATTCGAAGCTTGAGCTCGA-3′. Full-length palladin was cloned into mCherry-C1 by amplifying the sequence from the EGFP–palladin construct with the following primers: 5′-GCGCGAATTCAATGCCTGAGTCACCCGGCGGGCGAGGTATAA-3′ and 5′-GCGCGGATCCCAGGTCCTCACTTTCTACCAAGGCAGTATT-3′. To examine the role of the palladin–VASP interaction, a mutant palladin protein containing alanine replacements in the two polyproline stretches was generated. To mutate the first polyproline stretch of the palladin sequence, the EGFP-palladin vector was used as a template for the following primers: 5′- GCAGCGCGCCGCCCTCGCCCCCCGCCGCGGCGGCGGCCGCCTTCCCCGAGCTCGCGGC-3′ and 5′- GCCGCGAGCTCGGGGAAGGCGGCCGCCGCCGCGGCGGGGGGCGAGGGCGGCGCGCTGC-3′. The second stretch was mutated by using the primers: 5′- GCCTTCCCGGTGCCCGACGTGTTCGCACTGGCGGCGCCACCACCGCCGCTCCCGAGCCCG-3′ and 5′- CGGGCTCGGGAGCGGCGGTGGTGGCGCCGCCAGTGCGAACACGTCGGGCACCGGGAAGGC-3′. Human non-muscle α-actinin-1 cDNA that had been cloned onto the 5′-end of the humanized S55T version of GFP was a gift from Carol Otey (University of North Carolina School of Medicine, Chapel Hill, NC). Human EGFP–VASP was a gift from Frank Gertler (David H. Koch Institute for Integrative Cancer Research, MIT, Cambridge, MA). FLAG–mDia1 (constitutively active), human GFP–β-actin and pCherry-β-actin plasmids were gifts from Martin Bähler (Westfalian Wilhelms University, Münster, Germany). The VASP siRNA Dharmacon ON-TARGETplus Smartpool (catalog number L-019763-01, lot number 121105) was used to deplete VASP. The target sequence of the oligonucleotide that was used to deplete palladin was 5′-AATCACTACACCATTCAAAGA-3′. This oligonucleotide was designed against exon 15 of the palladin sequence and should, therefore, deplete all previously characterized palladin splice variants. The palladin siRNA oligonucleotide duplexes were labeled with Alexa Fluor 488 or 647 and purchased from Qiagen. The AllStars negative control siRNA (Qiagen) was used as a control siRNA.

For immunofluorescence, the cells were plated either onto glass coverslips or a chip with a medium sized pattern that contained fibronectin-650-coated crossbow micropatterns according to the manufacturer's instructions (Cytoo). After 20–24 h, the cells were fixed with 4% PFA, washed three times with 0.2% BSA in Dulbecco's PBS and then permeabilized using 0.1% Triton X-100 in PBS. Immunofluorescence staining was performed as described previously (Hotulainen and Lappalainen, 2006; Tojkander et al., 2011). A microscope (Axio Imager.M2, Zeiss) equipped with a charge-coupled device camera (AxioCam HRm, Zeiss), AxioVision Rel. 4.8 software (Zeiss) and a PlanApo ×63/1.40 (oil) objective (Zeiss) were used for image acquisition. The following antibodies were used for immunofluorescence staining: polyclonal antibody Ab-3Ig for palladin detection (Rönty et al., 2006), a monoclonal mouse antibody against α-actinin (Sigma-Aldrich), a monoclonal antibody against vinculin (Sigma-Aldrich) as a focal adhesion marker, VASP was visualized by using either a monoclonal mouse (Enzo Life Sciences) or polyclonal rabbit antibody (Sigma-Aldrich), Mena was visualized by using a polyclonal rabbit antibody (Sigma-Aldrich), a rabbit monoclonal antibody against non-muscle myosin (Covance) was used to stain myosin, the FLAG tag was detected by using a mouse monoclonal antibody (Sigma-Aldrich). F-actin was visualized by using phalloidin conjugated to Alexa Fluor 488, 568 or 647 (Invitrogen) or Atto-647N–phalloidin (Sigma-Aldrich) for STED imaging. Secondary antibodies were conjugated to Alexa Fluor 488, 555 or 568 (Invitrogen). Rescue experiments were performed as previously described, but cells were plated on CYTOOchips™ before fixation (Tojkander et al., 2011). To quantify the dorsal stress fiber density per µm, the number of dorsal stress fibers was counted along a line parallel to the cell edge and set within 3 µm of the cell edge, and the number was then divided by the line length. A dorsal stress fiber was defined as an actin bundle starting at a focal adhesion at the cell front and terminating in a transverse arc. To quantify the abundance of thick actin bars in cells that overexpressed palladin, images were obtained with a PlanApo ×20/0.8 objective for a very short excitation time in the GFP channel, which ensured that only the signal from cells overexpressing palladin was visualized. The exposure times were kept constant for all samples. The abundance of cells containing thick actin bars (actin fibers with a at least 50% larger diameter as</emph> compared with the thick ventral stress fibers of non-transfected cells) were then counted from the corresponding cells in the channel detecting phalloidin–Alexa-568-stained actin.

The cells were rinsed 24 h after transfection with cold PBS and then covered with cold immunoprecipitation buffer (150 mM NaCl, 10 mM Tris HCl [pH 7.4], 1 mM EDTA, 1 mM EGTA, 0.2 mM PMSF, 1% Triton X-100, 0.5% Nonidet P-40 and protease inhibitor cocktail tablets [Roche]) and incubated for 30 min at 4°C. The cells were scraped and pre-cleared by using a short centrifugation. The supernatants were incubated first with a mouse monoclonal antibody against GFP (Sigma-Aldrich) that was followed by incubation with Protein-G magnetic Dynabeads (Invitrogen). The beads were washed extensively, the proteins were eluted in SDS loading buffer and loaded into gels for SDS-PAGE and western blotting. To test the efficiency of siRNA knockdown, cell lysates were prepared and western blotted as previously described (Tojkander et al., 2011). The antibodies used were: the polyclonal palladin antibodies Ab-3Ig, Ab-4IgMo, Ab-4IgHu, a polyclonal palladin antibody (Sigma-Aldrich), a monoclonal mouse antibody against α-tubulin (Sigma-Aldrich), a polyclonal rabbit antibody against VASP (Sigma-Aldrich) and a polyclonal rabbit antibody against Enah (Sigma-Aldrich). Horseradish-peroxidase-linked secondary antibodies were used and chemiluminescence was measured after applying enhanced chemiluminescence western blotting reagent (Amersham, GE Healthcare).

The levels of VASP that had incorporated into dorsal stress fibers in palladin-depleted cells were analyzed by using the line profile tool from ImagePro Plus 5.1. (Media Cybernetics) and were then compared with the neighboring control cells. The VASP levels were normalized against the corresponding phalloidin staining to achieve comparable VASP intensities between cells that had been treated with control siRNA and siRNA against palladin. Data are represented as mean±s.e.m., n = 30 fibers from 15 cells (e.g. two fibers per cell were used for the quantification).

We transfected U2OS cells with siRNA as previously described. After trypsinization, the cells were counted, and 5×10⁴ cells that had been treated with the control siRNA or the siRNA against palladin were resuspended in FBS and mixed with 150 µl of collagen matrix that had been prepared as previously described (Tojkander et al., 2011). For the quantification of the phenotype of VASP-depleted cells in 3D collagen matrices, the control cells were labeled with green 5-chloromethylfluorescein diacetate dye and the knockdown cells with CellTracker Red CMTPX dye (Life Technologies). Because the VASP siRNA oligonucleotides were non-fluorescent, we were able to distinguish between control- and VASP-siRNA-transfected cells when these two types of cells were placed together in the same matrix. After the matrix was polymerized in a MatTek dish, it was detached with a needle from the glass and covered with 2 ml medium. After 20–24 h, the matrices were washed with PBS and fixed with 4% PFA, washed extensively with 0.2% Dulbecco/BSA and permeabilized with 0.2% Triton X-100 in PBS. The matrices were stained with the primary antibodies overnight, washed thoroughly and then stained with the secondary antibodies overnight. After thorough washing, the matrices were mounted in Mowiol-Dabco solution and imaged by using a Leica TCS SP5 or Leica TCS SP5 MP SMD with an HCX APO ×63/1.30 Corr (glycerol) CS 21 objective. Alexa Fluor 488 was excited with optically pumped semiconductor lasers at 488 nm/270 mW on the SP5 microscope or with an argon 488 nm/35 mW laser line on the SP5 MP SMD microscope. Alexa Fluor 568 was excited by using a diode-pumped solid-state laser at 561 nm/20 mW and Alexa Fluor 647 was excited at 633 nm/12 mW by using a helium-neon laser on both the SP5 and SP5 MP SMD microscopes. Single harmonic generation images were obtained with an upright Leica TCS SP5 MP microscope that was equipped with a MaiTai HP pulsed infrared femtosecond TiSa laser (Newport, RI), a HCX APO ×63/1.30 Corr (glycerol) CS 21 objective and a 1.3 NA oil condensor. Images of the collagen single harmonic generation signal were obtained by using transmitted non-descanned detection with a 445/20 nm filter (Semrock, Rochester, NY). The laser was tuned to 890 nm. To quantify cell shape, only cells that had at least one protrusion were considered. The length of the cell was measured as the distance between opposite cell ends by using the longest protrusion. The cell width was measured at the thickest part of the cell body, which was usually around the nuclear region. The ability of cells to contract the collagen matrix was measured as previously described (Tojkander et al., 2011). The diameter of the collagen matrices were imaged and measured as a percentage of the glass area of the MatTek dish at two timepoints, 24 and 48 h. Data are represented as mean±s.e.m., n = 6.

Super-resolution images were taken by using a commercial Leica TCS STED microscope. The Atto-647N-labeled samples were excited with a 632-nm excitation laser and depletion was set to 760 nm. Using these settings, the microscope has a lateral resolution of ∼50 nm, according to calibration measurements of sub-resolution beads.

We thank Kimmo Tanhuanpää from the light microscopy unit for the support in two photon imaging and data analysis. Anna-Liisa Nyfors is acknowledged for excellent technical assistance. We thank Martin Bähler, Frank Gertler and Carol Otey for providing valuable reagents for this study; and Pirta Hotulainen, Minna Poukkula and Maria Vartiainen for critical reading of the manuscript.