Cell migration and spreading is driven by actin polymerization and actin stress fibers. Actin stress fibers are considered to contain α-actinin crosslinkers and nonmuscle myosin II motors. Although several actin stress fiber subtypes have been identified in migrating and spreading cells, the degree of molecular diversity of their composition and the signaling pathways regulating fiber subtypes remain largely uncharacterized. In the present study we identify that dorsal stress fiber assembly requires α-actinin-1. Loss of dorsal stress fibers in α-actinin-1-depleted cells results in defective maturation of leading edge focal adhesions. This is accompanied by a delay in early cell spreading and slower cell migration without noticeable alterations in myosin light chain phosphorylation. In agreement with the unaltered myosin II activity, dorsal stress fiber trunks lack myosin II and are resistant to myosin II ATPase inhibition. Furthermore, the non-contractility of dorsal stress fibers is supported by the finding that Rac1 induces dorsal stress fiber assembly whereas contractile ventral stress fibers are induced by RhoA. Loss of dorsal stress fibers either by depleting α-actinin-1 or Rac1 results in a β-actin accumulation at the leading edge in migrating and spreading cells. These findings molecularly specify dorsal stress fibers from other actin stress fiber subtypes. Furthermore, we propose that non-contractile dorsal stress fibers promote cell migration and early cell spreading through Rac1-induced actin polymerization.
The defining features of mesenchymal migration mode are polarized, fan-shaped cells having a prominent leading protrusion and a contractile rear (Friedl and Wolf, 2003; Friedl and Wolf, 2010). The migration and spreading of these cells need dynamic actin polymerization at the leading edge (Pollard and Borisy, 2003; Ponti et al., 2004) as well as actin stress fibers, which are considered to contain α-actinin crosslinkers and myosin II motors providing contractility (Friedl and Wolf, 2003; Pellegrin and Mellor, 2007). Noteworthy is that migrating cells may also contain other types of actin–myosin assemblies (e.g. Cramer et al., 1997; Svitkina et al., 1997). Studies of mesenchymal migrating or spreading cells describe distinct subtypes of actin stress fibers on the basis of their subcellular localization and termination sites (Hotulainen and Lappalainen, 2006; Small et al., 1998; Tojkander et al., 2012; Tojkander et al., 2011). The fiber subtypes are dubbed dorsal stress fibers, transverse arcs and ventral stress fibers. More recent studies show additional differences between actin stress fiber subtypes including their assembly mechanisms, dynamics, myosin II abundance and function (Ang et al., 2010; Burnette et al., 2011; Hotulainen and Lappalainen, 2006; Oakes et al., 2012; Tojkander et al., 2012; Tojkander et al., 2011).
To date the most well-characterized actin stress fiber subtypes are myosin IIB-rich ventral stress fibers (Ang et al., 2010), which localize on the ventral surface of migrating and spreading cells and terminate to focal adhesions on their both ends (Hotulainen and Lappalainen, 2006; Small et al., 1998; Tojkander et al., 2011) (Fig. 1A arrow). These fibers are thought to assemble via fusion of pre-existing transverse arcs and dorsal stress fibers (Hotulainen and Lappalainen, 2006). Ventral stress fibers promote rear establishment as well as formation of front-to-rear polarity axis during cell migration (Ang et al., 2010; Vicente-Manzanares et al., 2008).
Curve-shaped transverse arcs assemble on the lamellar region of migrating cells in parallel orientation with the leading edge (Hotulainen and Lappalainen, 2006; Small et al., 1998) (Fig. 1A open arrowhead). Transverse arcs do not directly associate with focal adhesions, and move constantly towards nucleus where they disassemble (Heath, 1983; Small et al., 1998). Transverse arcs appear to assemble through annealing of Arp2/3 nucleated actin fibers (Hotulainen and Lappalainen, 2006), and are abundant in myosin IIA (Burnette et al., 2011; Cai et al., 2010; Vicente-Manzanares et al., 2008).
In addition to transverse arcs, the lamellar region contains dorsal stress fibers (Fig. 1A filled arrowhead), which perpendicularly elongate from the leading edge focal adhesions and often attach with transverse arcs (Hotulainen and Lappalainen, 2006; Small et al., 1998). The assembly of dorsal stress fibers has been shown to be partially dependent on formin-family member mDia1 (Hotulainen and Lappalainen, 2006; Oakes et al., 2012). Recently, myosin II was reported to be incorporated into dorsal stress fibers after these fibers associate with transverse arcs (Hotulainen and Lappalainen, 2006). The myosin II incorporation to dorsal stress fibers was shown to occur through tropomyosin 4 recruitment (Tojkander et al., 2011). In addition, a recent study focusing on focal adhesion maturation reported that partial α-actinin-1 downregulation inhibits radial (dorsal) stress fiber assembly at focal adhesions but retain a contractile lamella (Oakes et al., 2012).
Interestingly, despite the increasing number of studies demonstrating stress fiber subtype specific properties, largely uncharacterized is the possible degree of molecular specificity between fiber subtypes as well as signaling pathways regulating their dynamics. With this motivation we initially addressed a question whether either of the two major nonmuscle actin stress fiber crosslinkers α-actinin-1 or α-actinin-4 could show specificity in crosslinking actin stress fiber subtypes. Both of these α-actinin isoforms are ubiquitously expressed and have shown to be associated with cancer cell migration (Otey and Carpen, 2004). The finding that α-actinin-1 is specifically required to assemble dorsal stress fibers provided us an unique tool to investigate the role of this actin stress fiber subtype in cell migration and spreading.
α-Actinin-1 and α-actinin-4 localization on dorsal stress fibers differ
In order to investigate whether α-actinin-1 or α-actinin-4 possess any specificity in crosslinking actin stress fiber subtypes, human osteosarcoma U2OS cells were selected due to their well-defined dorsal stress fibers, transverse arcs and ventral stress fibers (Hotulainen and Lappalainen, 2006; Vallenius et al., 2000) demonstrated by staining of focal adhesions and actin stress fibers (Fig. 1A). Prior to initiate subcellular localization studies we made an attempt to generate α-actinin-1 specific antibody (dubbed A1-341, see Materials and Methods), as the lack of it has been a major drawback in determining endogenous α-actinin-1 localization by immunofluorescence. Subsequent western blotting analysis from U2OS cells transiently expressing control, GFP-tagged α-actinin-4 or α-actinin-1 indicates that the generated A1-341 antibody only reacts with α-actinin-1 (Fig. 1B; A1-341), whereas a commercial BM-75.2 antibody reported to be specific for α-actinin-1 (Araki et al., 2000; Celli et al., 2006; Chen et al., 2006; Honda et al., 1998) recognizes both isoforms (Fig. 1B; BM-75.2), and ALX-210 is specific for α-actinin-4 (Chen et al., 2006) (Fig. 1B; ALX-210). The specificity of A1-341 and ALX-410 enabled to assess distributions of α-actinin-1 and α-actinin-4 on actin stress fibers by immunofluorescence analysis. Although α-actinin-1 and α-actinin-4 staining is noted on all actin stress fiber subtypes (Fig. 1C,D) an interesting differential staining pattern is observed on dorsal stress fibers. On dorsal stress fibers α-actinin-1 is abundant throughout the fibers (Fig. 1E), whereas α-actinin-4 staining is prominent only at the base of the fibers where it emerges from the focal adhesion (Fig. 1F; arrow). Separation of a dorsal stress fiber into a base and a trunk area as depicted in Fig. 1G,H [also referred to distal and proximal parts, respectively (Hotulainen and Lappalainen, 2006; Tojkander et al., 2011)] demonstrates a significant base selectivity of α-actinin-4 compared to α-actinin-1 (Fig. 1G,H). This difference is also noted between focal adhesions and dorsal stress fiber trunks (supplementary material Fig. S1A,B). Similar subcellular localization pattern of α-actinin-1 on actin stress fiber subtypes can be detected upon ectopic expression of GFP-tagged α-actinin-1 (supplementary material Fig. S1C). Although GFP-tagged α-actinin-4 overexpression results in notable changes in actin stress fiber morphology, it also mimics endogenous α-actinin-4 localization (supplementary material Fig. S1D). Moreover, endogenous α-actinin-1 localization identified with A1-341 antibody is undetectable in α-actinin-1-depleted cells (supplementary material Fig. S1E). These results indicate that the newly generated α-actinin-1 antibody specifically recognizes subcellular localization of α-actinin-1, which is abundant throughout dorsal stress fibers.
Assembly of dorsal stress fibers requires α-actinin-1
In order to investigate whether the observed localization differences correlate with α-actinin-1 or α-actinin-4 functions on dorsal stress fibers we performed small interfering RNA (siRNA) mediated downregulation of α-actinin-1, α-actinin-4 or both, which after optimization showed undetectable protein levels in U2OS cells (Fig. 2A; supplementary material Fig. S2B,C,E). The F-actin staining of control and α-actinin downregulated cells reveals that while α-actinin-4 depletion does not result in appreciable changes in stress fiber subtypes (Fig. 2B; si-A4), α-actinin-1-silenced cells show an almost complete loss of dorsal stress fibers at the leading edge without affecting other stress fiber subtypes (Fig. 2B; si-A1; see also Fig. 2E; supplementary material Fig. S3A). As expected simultaneous downregulation of α-actinin-1 and α-actinin-4 causes nearly undetectable actin stress fibers indicating that these two α-actinins are major actin stress fiber crosslinkers in U2OS cells (supplementary material Fig. S2D,E). The loss of dorsal stress fibers in α-actinin-1-depleted cells is accompanied by significantly smaller (50%, P<0.001) vinculin-positive adhesion sites at the leading edge, not noted at the trailing edge (Fig. 2C; compare red and yellow circles and the adjacent box-plot for quantification). Furthermore the smaller area of adhesion sites associates with an increase in adhesion numbers (63%, P<0.001; Fig. 2D). The observed change in adhesion sites and numbers at the leading edge but not at trailing edge suggests that α-actinin-1-crosslinked dorsal stress fibers regulate adhesion maturation at the leading edge.
In order to gain further insight into the mechanism resulting in loss of dorsal stress fibers in α-actinin-1-silenced cells, we followed actin stress fiber dynamics using time-lapse imaging of the available EGFP-CLP-36-expressing U2OS cells (Vallenius et al., 2004) (supplementary material Movies 1, 2). While in control cells prominent straight dorsal stress fibers form (Fig. 2E; filled arrowheads; supplementary material Movie 1) and attach to transverse arcs (open arrowhead), α-actinin-1-silenced cells occasionally display only a few weak nonlinear fibers that disappear abruptly (Fig. 2E; supplementary material Movie 2). Consistent with this, a focused analysis of dorsal stress fiber bases in α-actinin-1-silenced cells (not stably expressing EGFP-CLP-36) stained with F-actin reveals short fibers exhibiting significantly reduced amount of filamentous actin (57%, P<000.1, Fig. 2F; supplementary material Fig. S3A). The loss of dorsal stress fibers following α-actinin-1 silencing cannot be rescued by ectopic expression of α-actinin-4 (data not shown), although in murine lung fibroblasts α-actinin-1 was reported to rescue the studied α-actinin-4 functions (Shao et al., 2010a). This strongly proposes the specific requirement of α-actinin-1 in the dorsal stress fiber assembly. Furthermore, in agreement with U2OS cells, α-actinin-1 downregulation causes a loss of α-actinin-1 rich dorsal stress fibers in mouse embryonic fibroblasts (MEFs), which exhibit clearly identifiable actin stress fiber subtypes (supplementary material Fig. S3B,D). This can be also detected in spreading MEFs (supplementary material Fig. S3C). Collectively, these results demonstrate that the assembly of dorsal stress fibers requires α-actinin-1 in U2OS cells and in MEFs.
α-Actinin-1 depletion causes decreased cell migration without altering myosin II activity
The established function of actin stress fibers is to regulate cell migration (Ridley, 2001). Growing number of studies suggest that actin stress fiber subtypes regulate cell migration differently. For example ventral stress fibers promote rear contraction and formation of front-to-rear polarity axis (Ang et al., 2010; Vicente-Manzanares et al., 2008). To assess a possible role of dorsal stress fibers in cell migration, we selected to use a scratch wound-healing assay where cells at wound edge show directional migration. F-actin and vinculin co-staining indicates that cells at the wound edge exhibit evident dorsal stress fibers, transverse arcs and ventral stress fibers containing α-actinin-1 and are often fan-shaped (supplementary material Fig. S4A–C). Comparison of wound closure areas at 24-hour time point when the control cells are about to close their wounds, α-actinin-1-silenced cells show minor but significant delay, an average of 27% (n = 15; P<0.01) (Fig. 3A,B). Subsequent 24-hour time-lapse imaging of wound closure hourly (supplementary material Movie 3) revealed that control and α-actinin-1-silenced cells display similar migration pattern, but α-actinin-1-silenced cells migrate overall more slowly (in control cells wound closure rate on average 33.4 µm/h (24.4–39.9 µm) and in α-actinin-1-depleted cells 26.0 µm/h (20.4–32.7 µm). The observed wound closure rate in control cells is consistent with previous reports analyzing mesenchymal migration mode (Clarke et al., 2004; Friedl and Wolf, 2010). In agreement with the wound-healing assay, α-actinin-1 depletion decreases U2OS cell migration through Matrigel-coated transwells on average 28% (Fig. 3D,E). Compared to control cells, the cell number and total cell area is not significantly altered in α-actinin-1-silenced cells (supplementary material Fig. S4D,E). Collectively, these results implicate that the decreased migration reflects a loss of dorsal stress fibers following α-actinin-1 downregulation.
Interestingly, the decreased migration in α-actinin-1-silenced cells is opposite to what has been noted following silencing or inhibition of myosin IIA (Even-Ram et al., 2007; Sandquist et al., 2006), which is the major myosin isoform in transverse arcs in lamella region. Therefore, we next probed myosin II activity following α-actinin-1 silencing by western blotting analysis using phospho-myosin light chain (pMLC) antibody. While inhibition of myosin IIA with a ROCK kinase inhibitor (Y27632) leads to a dramatic drop in pMLC, α-actinin-1 downregulation does not significantly change pMLC levels (Fig. 3F). These results together with a loss of dorsal stress fibers in α-actinin-1-silenced cells suggest that dorsal stress fibers promote cell migration in myosin-II-independent manner.
Dorsal stress fiber trunks are non-contractile
The results above prompted us to investigate the status of myosin II on dorsal stress fibers. Interestingly, although pMLC staining of migrating U2OS cells clearly decorated ventral stress fibers and transverse arcs (Fig. 4A; open arrowhead), dorsal stress fibers are not evident in pMLC staining (Fig. 4A; filled arrowhead) except in regions close to transverse arcs. Consistently, there is a significantly lower pMLC/F-actin ratio in dorsal stress fibers compared to transverse arcs (Fig. 4B). To further investigate the unexpected lack of pMLC on dorsal stress fibers (Hotulainen and Lappalainen, 2006; Tojkander et al., 2011) we analyzed these fibers in cells overexpressing GFP-tagged myosin IIA. Vinculin and F-actin staining of these cells indicates that whereas GFP-myosin IIA localizes in a dotted-like pattern on transverse arcs (Fig. 4C; open arrowhead) and on ventral stress fibers (Fig. 4C; arrow), myosin IIA signal on dorsal stress fiber trunks is undetectable (Fig. 4C; filled arrowhead). Similar to myosin IIA, myosin IIB localizes to transverse arcs but is not detectable on dorsal stress fiber trunks (supplementary material Fig. S4F). These results strongly suggest that dorsal stress fiber trunks are non-contractile.
To further assay the role of myosin II on dorsal stress fibers, we treated U2OS cells with the ROCK kinase inhibitor (Y27632) or with the myosin II ATPase inhibitor blebbistatin for 30 minutes. As expected, a 30 minute inhibition with Y27632 (data not shown) or blebbistatin (Fig. 4D) leads to a dramatic decrease of ventral fibers and transverse arcs, but interestingly dorsal stress fibers are significantly more resistant to blebbistatin (Fig. 4D). It is also interesting to note that α-actinin-1 remains on dorsal stress fibers (Fig. 4D; zoom-in, lower bracket: dorsal sf) indicating that myosin II driven tension is not needed for α-actinin-1 localization on these fibers. In contrast blebbistatin treatment removes both α-actinin-1 and vinculin from focal adhesions as reported previously (Pasapera et al., 2010; Schiller et al., 2011). This is noted as a 5–10 µm gap from the leading edge (Fig. 4D; zoom-in, upper bracket, FA and Fig. 4D; vinculin). To summarize, the lack of myosin II on dorsal stress fiber trunks and the finding that these fibers are resistant to myosin II inhibitors strongly propose that dorsal stress fibers are non-contractile.
Delayed early cell spreading and β-actin accumulation at cell periphery following α-actinin-1 depletion
The apparent minor role of myosin II on dorsal stress fibers prompted us to assess whether functions of these fibers would be mediated through actin polymerization. Actin polymerization at cell periphery has a major role during early cell spreading, where myosin II activity is low (Cai et al., 2006; Johnston et al., 2008; Price et al., 1998). During early spreading phase radial actin fibers resembling dorsal stress fibers initiate to assemble from peripheral adhesions in fibroblasts (Cai et al., 2010). Therefore, we first analyzed actin stress fibers and α-actinin-1 localization from spreading control and α-actinin-1-depleted cells seeded on fibronectin-coated coverslips. The majority of control cells initiate to assemble radial fibers at the cell periphery within 30 minutes (Fig. 5A; si-NT) consistent with studies in fibroblasts (Cai et al., 2010). These fibers also associate with transverse arcs and contain abundant α-actinin-1 (Fig. 5A; si-NT) whereas α-actinin-4 staining is detectable only at bases of emerging fibers (arrowhead in supplementary material Fig. S5A). Interestingly, the α-actinin-1-depleted cells fail to form these radial fibers (Fig. 5A; si-A1). Occasional small radial fiber buds are detected similarly to the migrating cells lacking α-actinin-1 (supplementary material Fig. S3A). These results significantly extend the observed similarities between dorsal stress fibers in migrating cell and radial fibers in spreading cells, which we therefore suggest to be redefined as dorsal stress fibers.
To investigate the consequences of dorsal stress fiber loss in early cell spreading, we analyzed control and α-actinin-1-depleted cells using time-lapse imaging of EGFP-CLP-36 U2OS cells (Vallenius et al., 2004) enabling to follow spreading in concomitantly with actin fiber formation. In control cells protrusions emerge immediately from cell edge after plating and cells start spreading roughly 5 minutes afterwards. Dorsal stress fibers are noted at around 20 minutes following plating (supplementary material Movie 4). By contrast, in α-actinin-1-silenced cells spreading is delayed by ∼30 minutes and prominent dorsal stress fibers are not noted during the 2-hour follow-up. Consistent with the cell spreading delay, cell circumference at 20 minutes after plating is significantly smaller in α-actinin-1-silenced cells (Fig. 5B,C). Analysis of time-lapse images taken every 2 minutes after plating demonstrates constant spreading delay between 4 to 20 minutes time points (supplementary material Fig. S5B). Despite this, the α-actinin-1-silenced cells do gradually spread, and acquire a shape and size comparable to control cells (Fig. 5B; 120 minutes; supplementary material Movie 4). At the 120 minutes point they have a 1.3-fold larger circumference (three independent experiments; n = 20; P<0,001). This change is however transient based on the lack of detectable difference in cell area in continuously cultured cells (supplementary material Fig. S4E). During a two-hour follow-up significantly higher fraction of control cells (29%; 16/56) initiate migration compared to α-actinin-1-depleted cells (15%; 13/88), which is in agreement with wound-healing delay results (Fig. 3). Noteworthy is that early cell spreading delay is also observed in fixed U2OS cells not stably expressing EGFP-CLP-36 (data not shown). To summarize, the delay in early cell spreading together with the non-contractile nature of dorsal stress fibers imply that these fibers contribute to actin polymerization at the leading or spreading cell edge.
To investigate possible changes in actin isoforms associated with dorsal stress fiber assembly we analyzed subcellular localization of β-actin following α-actinin-1 silencing. Interestingly, both spreading and migrating α-actinin-1-silenced cells show an accumulation of β-actin at cell periphery (Fig. 5D,E). In agreement with the noted endogenous β-actin accumulation, GFP-tagged β-actin expression in α-actinin-1-silenced cells also accumulates at the leading edge of migrating cells (data not shown). This is specific for β-actin, as it is not observed in migrating cells stained for γ-actin (supplementary material Fig. S5C).
Rac1 induces assembly of dorsal stress fibers crosslinked by α-actinin-1
When considering the possible upstream signals regulating non-contractile dorsal stress fiber mediated functions, an interesting candidate is Rac1 as it is a critical regulator of actin polymerization at the leading edge (Ridley, 2011; Wu et al., 2009). To assess whether Rac1 promotes dorsal stress fiber assembly we expressed a GFP-tagged constitutively active Rac1 (Rac1 V12) or as a control constitutively active RhoA (RhoA V14) involved in contractility mediated cell migration (Parsons et al., 2010). Analysis of GFP signals indicate that Rac1 V12 localizes on focal adhesions (Fig. 6A,E) as reported previously (Nethe et al., 2010; ten Klooster et al., 2006), whereas RhoA exhibits only diffuse cytoplasmic pattern also consistent with previous studies (Ridley and Hall, 1992; Yonemura et al., 2004). Strikingly, F-actin staining revealed that Rac1 V12-expressing cells exhibit pronounced dorsal stress fibers (Fig. 6B; arrowhead) and only few ventral fibers (Fig. 6B; arrow) whereas in RhoA V14-expressing cells the reverse is noted regarding actin fiber types (supplementary material Fig. S6A). Quantification demonstrates a twofold increase in dorsal stress fiber numbers in Rac1 V12-expressing cells (Fig. 6C) and the fibers are also significantly longer (Fig. 6D). In contrast, RhoA V14 expression results in a significant increase in ventral stress fiber numbers and decrease in dorsal stress fiber numbers (Fig. 6C). These results suggest that Rac1 is a critical inducer of dorsal stress fibers. To investigate a possible interdependency of Rac1 and α-actinin-1 in dorsal stress fibers, we subsequently analyzed whether Rac1 could induce these fibers in the absence of α-actinin-1. Concomitant depletion of α-actinin-1 and ectopic expression of Rac1 V12 results in severe defects in dorsal stress fiber formation (Fig. 6E). Only few short fibers emerge from focal adhesions at the leading edge similarly as noted in α-actinin-1-depleted migrating cells (Fig. 2E; supplementary material Fig. S3A). Furthermore, subsequent analysis of dorsal stress fibers by F-actin staining following Rac1 downregulation indicates that Rac1 is required to assemble these fibers (Fig. 6F). Importantly, Rac1 silencing results in β-actin accumulation at the leading edge similarly as in α-actinin-1-depleted cells (supplementary material Fig. S6B). Overall, these results provide evidence that α-actinin-1 and Rac1 are critical molecules required for the assembly of dorsal stress fibers in migrating and spreading cells.
Our results demonstrate remarkable specificity of dorsal stress fibers both in molecular composition and in their regulation compared to ventral stress fibers and transverse arcs. α-Actinin-1 is required to crosslink linear dorsal stress fibers, as in its absence only rudiments of α-actinin-4 containing dorsal stress fibers emerge. α-Actinin-1 and α-actinin-4 are very similar in primary structure (87% identity), with less related regions in the third spectrin-like repeat and at the very N-terminus. Both of these regions interact with various proteins but their α-actinin isoform specificity has not been determined (Otey and Carpen, 2004). Interestingly, the N-terminal region of α-actinin-1, which associates with filamentous actin, has a focal adhesion kinase (FAK) site (Tyr12) (Craig et al., 2007; Izaguirre et al., 2001), which does not appear to be a major phosphorylation site in α-actinin-4 (Shao et al., 2010b). Moreover, this phosphorylation site was reported to promote tumor cell adhesion under pressure in an α-actinin-1-dependent manner suggesting a possible dorsal stress fiber-associated function. However, another in vitro study demonstrated that Tyr12 phosphorylation on α-actinin-1 decreased α-actinin-1 binding with filamentous actin (Izaguirre et al., 2001), and thereby future studies are needed to determine the role of Tyr12 phosphorylation.
The lack of dorsal stress fibers in α-actinin-1-silenced cells is accompanied with impaired adhesion maturation at the leading edge. This finding is consistent with observations made following α-actinin-1 downregulation in CHO.K1 cells, where α-actinin-1 is needed for the maturation of early stage adhesions in a myosin II-independent manner (Choi et al., 2008). In CHO.K1 cells myosin II independency was correlated with long-lived nascent adhesions whereas in U2OS cells or MEFs having short-lived nascent adhesions myosin II was reported to be needed for adhesion maturation (Choi et al., 2008). Another study on U2OS cells also utilizing α-actinin-1 downregulation showed that initial compositional maturation of focal adhesions is not tension dependent, but instead requires a stress fiber template (Oakes et al., 2012). The defective maturation of focal adhesions in α-actinin-1-downregulated cells was determined by the decreased phosphorylation of FAK (pY397) and paxillin (pY31) as well as smaller adhesion sites (Oakes et al., 2012), and these defects were not noted following the ROCK kinase inhibition. Here, we observe that the inhibition of myosin II ATPase activity results in compositional changes in focal adhesions determined by vinculin and α-actinin-1 disappearance from adhesion sites, which occurs at the presence of prominent dorsal stress fibers. This is in agreement with several recent studies including two large-scale proteomic purifications of focal adhesions showing a large set of proteins recruited to focal adhesions in myosin-II-dependent manner (Kuo et al., 2011; Pasapera et al., 2010; Schiller et al., 2011). In aggregate these studies suggest that dorsal stress fiber-mediated tension is required for focal adhesion maturation at the leading edge whereas initial adhesion formation apparently does not require tension. Presumably, the dorsal stress fiber mediated tension at the leading edge derives from lamellar retrograde flow including transverse arcs (Hotulainen and Lappalainen, 2006), which is coupled to focal adhesions through dorsal stress fibers.
The lack of myosin II on dorsal stress fiber trunks signifies another compositional specialization of stress fiber subtypes. This striking result challenges the widely presented view (Parsons et al., 2010; Pellegrin and Mellor, 2007; Vicente-Manzanares et al., 2009) that actin stress fibers including dorsal stress fibers (Hotulainen and Lappalainen, 2006; Tojkander et al., 2011) represent sarcomere-like structures containing myosin II. Interestingly, previous polarity studies demonstrate important differences between dorsal stress fibers and other stress fiber subtypes (Pellegrin and Mellor, 2007). While the actin filament polarity on ventral stress fibers and transverse arcs is alternating resembling that of muscle cell sarcomeres, dorsal stress fibers orient unipolarly, i.e. barbed ends at focal adhesions point to the leading edge and pointed ends towards the nucleus. Thereby, the generation of sarcomere-like fibers would not only require incorporation of myosin II molecules into dorsal stress fibers but also a drastic reorientation of actin filaments. Considering that myosin II appears to be incorporated following the association of dorsal stress fibers with transverse arcs (Hotulainen and Lappalainen, 2006) this association may trigger actin rearrangements and incorporation of myosin II at the tips of dorsal stress fibers following transverse arcs attachment.
This study additionally identifies that dorsal stress fibers are specifically induced by active Rac1. A wealth of literature highlights the importance of Rac1 as an inducer of Arp2/3 nucleated branched actin filaments pushing leading edge forward (Ridley, 2011). However, Rac1 also localizes to focal adhesions (Nethe et al., 2010; ten Klooster et al., 2006) and associates with mDia2 nucleating unbranched actin filaments (Ji et al., 2008; Lammers et al., 2008). Recently it was shown that mDia2 depletion abrogated formation of fibers termed lamellipodial filopodia in the presence of constitutively active Rac1 (Yang et al., 2007). Interestingly, these linear lamellipodial filopodia fibers oriented perpendicularly to the leading edge share characteristics with dorsal stress fibers. In U2OS cells mDia2 has been proposed to nucleate dorsal stress fibers (Oakes et al., 2012) (also our unpublished observation), although mDia1 is clearly additionally involved (Hotulainen and Lappalainen, 2006; Oakes et al., 2012), and associates with Rac1 (Lammers et al., 2008). In U2OS cells mDia2 also contributes to the nucleation of transverse arcs (Tojkander et al., 2011). Future experiments are needed to resolve links between Rac1 and mDia1, mDia2 as well as potential other fiber-specific nucleators (Ang et al., 2010).
The established role of Rac1 as a stimulator of cell migration and spreading through actin polymerization (Ridley et al., 2003) together with myosin II free dorsal stress fiber trunks propose that actin polymerization is important in dorsal stress fiber dependent cell migration and early cell spreading. The observed increase in β-actin at the leading edge of migrating and early spreading cells in the absence of dorsal stress fibers is consistent with this. Interestingly, β-actin deficient fibroblasts display a migration defect apparently associating with increased ventral stress fibers (Bunnell et al., 2011; Tondeleir et al., 2012) suggesting that β-actin is involved in dorsal stress fiber assembly. A possibility is also that dorsal stress fiber driven actin polymerization promotes cell migration and spreading by acting as ‘mother filaments’ for other lamellar actin network structures as suggested for lamellipodial filopodia as a way to increase nucleation points for the Arp2/3 complex (Yang et al., 2007).
Regarding cell migration the noted reciprocal assembly of dorsal stress fibers by Rac1 and ventral stress fibers by RhoA is interesting and demonstrates specificity in upstream regulators of stress fiber subtypes. The reciprocal activation of Rac1 and RhoA is often associated with temporal regulation such as when cell migration modes interconvert from amoeboid (RhoA) to mesenchymal (Rac1) (Sanz-Moreno et al., 2008), during cell protrusions (Machacek et al., 2009) and during neurite growth cone extension (Kozma et al., 1997). Our results suggest that spatial and temporal activation of Rac1 and RhoA results in assembly of distinct actin stress fiber subtypes, which coordinate leading edge protrusions and rear detachment, respectively.
Materials and Methods
Cell culture, reagents and antibodies
Human U2OS osteosarcoma cell line and immortal mouse embryonic fibroblasts (MEFs) were propagated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, penicillin and streptomycin and glutamine at 37°C in 5% CO2. To generate immortal MEFs, cells from passage 3 were infected with a retrovirus encoding residues 302 to 390 of p53 (Klefstrom et al., 1997) and grown as pools following selection with hygromycin (Invitrogen) at 0.2 mg/ml. To inhibit myosin II ATPase activity U2OS cells were treated for 30 minutes with 100 µM blebbistatin (Sigma) or with 10 µM ROCK kinase inhibitor (Y27632, Tocris Bioscience). The rabbit polyclonal anti-α-actinin-1 antibody was raised using the N-terminal α-actinin-1 sequence (DHYDSQQTND) as an immunogen, and subsequently affinity purified (EuroGentec). Other primary antibodies were: the rabbit polyclonal anti-α-actinin-4 (catalog no. ALX-210-356, Alexis Biochemicals), the mouse monoclonal anti-vinculin antibody (catalog no. V9131, Sigma), the rabbit polyclonal anti-phospho-MLC (Ser19) antibody (catalog no. 3671, Cell Signaling), the mouse monoclonal total anti-MLC antibody (catalog no. M4401, Sigma), the hybridoma antibody myosin IIB (CMII 23) was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biology, Iowa City, IA 52242, mouse monoclonal anti-β-actin antibody (catalog no. A1978, Sigma), mouse monoclonal anti-γ-actin antibody was kindly provided by Dr Christine Chaponnier (Geneva Faculty of Medicine, Switzerland) (Dugina et al., 2009), the rabbit polyclonal anti-CDK2 antibody (catalog no. sc-163, Santa Cruz), and rabbit polyclonal anti-GAPDH antibody (catalog no. 14C10, Cell Signaling). Filamentous actin was stained by using Alexa Fluor® 488, 546 or 647 phalloidin (Invitrogen). For secondary antibodies in western blotting anti-rabbit-HRP and anti-mouse-HRP (Chemicon International) were used and for immunofluorescence Alexa Fluor® anti-rabbit 594, Alexa Fluor® anti-mouse 647 or Alexa Fluor® anti-mouse 488 (Invitrogen). Other reagents used: 10 µg/ml fibronectin (catalog no. 734-0101, from VWR International) for coating coverslips.
Plasmid constructs and transfections
Myosin IIA (MYH9) was obtained from the human ORFeome library (Open Biosystems) as a Gateway pENTR223 entry clone, from where it was further transferred into the pcDNA6.2/N-EmGFP-DEST destination vector (Invitrogen) to generate GFP-tagged-myosin IIA using LR reaction kit (catalog no, 12538-013, Invitrogen). Human GFP-α-actinin-1 and GFP-α-actinin-4 were kindly provided by Prof. Carol Otey (University of North Carolina School of Medicine, USA) (Edlund et al., 2001), and GFP-RacV12 and GFP-RhoAV14 by Dr Anne Debant (CRBM- Macromolecular Biochemistry Research Center, France) (Bellanger et al., 2000). For overexpression studies Fugene6 transfection reagent (Roche) was used and for siRNAi-mediated gene silencing Lipofectamine2000 transfection reagent (Invitrogen). The human pooled siRNA oligonucleotides were L-011195 (siACTN1), L-011988 (siACTN4), L-003560 (siRAC1) and D-001206-13 (siNT) (Dharmacon) and mouse pooled siRNA oligos was L-066191 (si-actn1). In addition, two independent single oligos targeting ACTN1 (J-011195-05, J-011195-06, Dharmacon) were used. To obtain high knockdown efficiency RNAi-mediated silencing protocol was conducted on two consecutive days over a period of 72 hours following the first transfection. In all experiments knockdown efficiency was confirmed either using western blotting or immunofluorescence analysis.
Immunofluorescence and western blotting analyses
For immunofluorescence analysis U2OS cells seeded on glass or fibronectin (10 µg/ml) coated coverslips were fixed with 3.5% (w/v) paraformaldehyde (PFA) for 15 minutes and washed three times with TBS prior to permeabilization using 0.1% Triton X-100 for 5 minutes. Then washed three times with PBS, blocked with 5% goat serum in TBS for 30 minutes, labeled with primary antibody for 30 minutes, washed three times with TBS, labeled with secondary antibody for 30 minutes, washed three times with TBS, labeled with Hoechst, and mounted with Immu-mount (Thermo Scientific). Cells analyzed for β-actin antibody stainings were fixed in 1% PFA for 10 minutes and for 7 minutes with MeOH at −20°C followed by the staining protocol described above. For γ-actin antibody staining, cells were fixed as described (Dugina et al., 2009).
Stained coverslips were analyzed and imaged using upright Zeiss AxioImager M2 epifluorescence microscope equipped with 63x/Plan-Apochromat/1.40 Oil/M27 and AxioCam HRm camera, inverted Zeiss LSM 510 Meta and upright Leica TCS SP5 laser scanning confocal microscopes equipped with 63x/Plan-Apochromat/1.40/DIC and 63x/HCX APO/1.30/Corr (glycerol)/CS21 objectives, respectively. Final images were generated using Adobe Photoshop CS5, version 12.0 and Adobe Illustrator CS5, version 15.0.
For western blotting cells were lysed using SDS boiling buffer (2.5% SDS, 0.25 M Trizma base including 50 mM NaF, 10 mM β-glycerophosphate, 0.5 mM DTT, 0.5 mM PMSF, 2.5 µg/ml aprotinin and 1 µg/ml leupeptin), needled (25-gauge) and centrifuged at 20,800 g for 10 minutes at 4°C prior to protein concentration measurements using Bio-Rad DC protein assay (Bio-Rad Laboratories). 10–30 µg of lysates were run on SDS-PAGE gel, and for western blotting analysis TBS including 0.05% Tween 20 and 5% BSA was used for blocking and antibody incubations. Primary antibodies were incubated overnight and secondary antibodies for 30 minutes prior to detection with ECL reagents (SuperSignal West Femto, PIERCE).
Wound healing, transwell migration, cell spreading and live imaging
For the wound-healing assay 70,000 cells were initially seeded 24 hours before transfections and cell number was counted again following 72 hours of transfection. Wounds were generated to confluent monolayers of U2OS cells with a sterile tip 72 hours after transfection of pooled non-targeting or ACTN1 targeting siRNAs. Wound areas were marked to the bottom of each plate to enable image acquisition of the same area using Olympus CKX4 microscope equipped with UplanFL4x/0.13/PhP objective and Canon DS6041 EOS 300D digital camera. Images were taken 24 hours from the wound start when control cells almost reached closure. Similar transfection set-up as above was used to perform time-lapse imaging during wound closure, except cells were seeded on 6-well plates. Using Cell-IQ (CM Technologies) platform, phase-contrast images were taken hourly using 10×objective from cells cultured at 37°C in 5% CO2 for 24 hours. Data derives from five independent experiments. Subsequently all images were analyzed and quantified by Adobe Photoshop CS5, version 12,0 and the ImageJ software 1.46 calculating open areas in pixels.
Transwell migration assay was performed using Matrigel-coated (catalog no. E1270, Sigma) 24 mm Transwell invasion chambers with 8 µm pore polycarbonate membrane inserts (catalog no. 3428, Corning) according to manufacturer's instructions. Briefly, cells were pretreated with pooled non-targeting or ACTN1 targeting siRNAs for 72 h, trypsinized and resuspended into Dulbecco's modified Eagle's medium (DMEM) containing 0.1% BSA. 5×104 cells per well were then added to the upper chambers. DMEM containing 10% FBS was added to the lower chambers as chemoattractant. After 18 h, transmigrated cells on the membrane's undersurface were stained with hematoxylin, and counted using ImageJ software 1.46 from three independent experiments containing duplicates on average 4198 control and 3016 α-actinin-1-downregulated cells.
Time-lapse imaging of EGFP-CLP-36-expressing U2OS cells (Vallenius et al., 2004) was utilized in spreading assay and actin stress fiber dynamic analysis. For spreading assay 72 hours after transfection of pooled non-targeting or ACTN1 siRNA oligos cells were re-plated on fibronectin coated coverslips and allowed spreading for 30 minutes prior to 3.5% PFA fixation or were re-plated on fibronectin coated Lab-Tek chambers (catalog no. 155379, Nunc) for time-lapse imaging. For time-lapse imaging of actin stress fiber dynamics of pooled non-targeting or ACTN1 siRNA oligos was conducted as mentioned above but cells were allowed to spread on fibronectin coated Lab-Tek chambers for 3 h prior live imaging. For all time-lapse imaging Zeiss Stallion HSI microscope equipped with Zeiss AxioCam MRm camera and EC Plan-Neofluar 20x/0.50 Ph2 objective was used for spreading assay and EC Plan-Neofluar 63x/1.30 Oil Ph3 objective for actin stress fiber dynamic analysis. Images were acquired every 2 minutes for the period of 2 hours at 37°C and 5% CO2 by using phase contrast and GFP fields. SlideBook 4.2 software and QuickTime Player Pro 7.6 were used for generating movies.
Image analysis and quantifications
Dorsal stress fiber base and trunk areas were quantified by measuring the ratio between the integrated density of α-actinin-1 or α-actinin-4 antibodies and F-actin (phalloidin) by using ImageJ software 1.46. A base part of dorsal stress fiber was determined to be 1∶4 and a trunk 3∶4 of dorsal stress fiber length attaching with a transverse arc (Fig. 1G,H; supplementary material Fig. S1A,B). To measure F-actin intensity of nonlinear rudimental dorsal stress fibers following α-actinin silencing 1 µm area from adhesion was determined as the base of the dorsal stress fiber (Fig. 2E; supplementary material Fig. S3A).
Using ImageJ software 1.46 the vinculin positive leading edge adhesion sites were quantified from an area of 10 µm×10 µm, which was drawn by first determining the midpoint of nucleus, from where the vertical axis was drawn towards the highest adhesion amount of vinculin positive adhesions at the leading edge. For the trailing edge analysis a horizontal axis was determined perpendicular to the vertical leading edge axis and the boxed area of 10 µm×10 µm was drawn to include most of adhesions at trailing edge. For both leading and trailing edge adhesion analysis the area and amount of adhesions were quantified.
Circumference of spreading cells was measured from phase-contrast time-lapse images (Fig. 5B,C) or fixed cells (supplementary material Fig. S4E). For analysis cells reaching fan-shaped morphology during 120 minutes follow-up time without attaching neighboring cells were chosen, and cell circumference was analyzed by Image-Pro Analyzer or by ImageJ software 1.46 from three independent experiments.
Phospho-MLC (Ser19) staining along dorsal stress fibers and transverse arcs (determined by phalloidin staining) was quantified by measuring the ratio between the integrated density of pMLC (Ser19) antibody and F-actin (phalloidin) by utilizing ImageJ software 1.46.
Dorsal and ventral stress fibers from GFP-, GFP-Rac1 V12- and GFP-RhoA V14-expressing cells were counted based on their association with focal adhesions (vinculin and phalloidin staining). In addition, focal adhesion (vinculin staining) and dorsal stress fiber (phalloidin staining not overlapping with vinculin) length was measured using ImageJ software 1.46. Three independent experiments were included for both analyses.
Statistical analyses were performed using Student's t-test assuming a two-tailed distribution and unequal variance in all other figures except in Fig. 3C and in supplementary material Fig. S5B where a one-tailed distribution is used (analysis of migration hourly or of spreading every 2 minutes). In box-plot representations whiskers indicate the lowest datum still within 1.5 interquartile range (IQR) of the lower quartile, and the highest datum still within 1.5 IQR of the upper quartile. All outliers are indicated in box plot figures with grey asterisks (*), except in Fig. 2C, which contains adhesion sites from 20 cells. In this experiment outlier range at the leading edge are: si-NT 2.3-6.4; si-A1 1.1-3-1; si-A4 2.0-4.0 and at trailing edge: si-NT 2.2-7.6; si-A1 1.8-5.5; si-A4 2.1.-7.5. All outliers are included to Student's t-test calculations in all experiments.
We thank Prof. Carol Otey, Dr Anne Debant and Dr Christine Chaponnier for GFP-actinin constructs, GFP-Rac V12 and Rho V12 constructs and γ-actin antibody, respectively. Saana Ruusulampi and Niina Roine are thanked for excellent technical assistance, Lotta Hirvenkari for the initial α-actinin-1 antibody testing, Dr Niklas Ekman for a help with the generation of immortal MEFs, Dr Kari Vaahtomeri for statistical analysis, and the Biomedicum Imaging Unit, Light Microscopy Unit and Genome Biology Unit for their services. Prof. Pekka Lappalainen and Dr Kari Vaahtomeri are acknowledged for fruitful discussions and for commenting on the manuscript.
This work was supported by the Academy of Finland [grant numbers 258837 and 265170 to T.V., and 259278 and 262976 to T.M.], the Finnish Medical Foundation (T.V.), Biocentrum Helsinki start-up grant (T.V.), Finnish Cancer Foundation (T.M.) and Sigrid Juselius Foundation (T.M.). B.K. and J.T. are students of the Helsinki Graduate School in Biotechnology and Molecular Biology and T.V. is the Academy of Finland research fellow.