The activity of serum response factor (SRF), an essential transcription factor in mouse gastrulation, is regulated by changes in actin dynamics. Using Srf(−/−) embryonic stem (ES) cells, we demonstrate that SRF deficiency causes impairments in ES cell spreading, adhesion, and migration. These defects correlate with defective formation of cytoskeletal structures, namely actin stress fibers and focal adhesion (FA) plaques. The FA proteins FA kinase (FAK), β1-integrin, talin, zyxin, and vinculin were downregulated and/or mislocalized in ES cells lacking SRF, leading to inefficient activation of the FA signaling kinase FAK. Reduced overall actin expression levels in Srf(−/−) ES cells were accompanied by an offset treadmilling equilibrium, resulting in lowered F-actin levels. Expression of active RhoA-V14 rescued F-actin synthesis but not stress fiber formation. Introduction of constitutively active SRF-VP16 into Srf(−/−) ES cells, on the other hand, strongly induced expression of FA components and F-actin synthesis, leading to a dramatic reorganization of actin filaments into stress fibers and lamellipodia. Thus, using ES cell genetics, we demonstrate for the first time the importance of SRF for the formation of actin-directed cytoskeletal structures that determine cell spreading, adhesion, and migration. Our findings suggest an involvement of SRF in cell migratory processes in multicellular organisms.
The actin cytoskeleton directs many essential cellular processes, including cell–matrix and cell–cell adhesion, cell motility, cytokinesis, and endo-/exocytosis (Salmon, 1989; Gumbiner, 1996; Chen et al., 2000). F-actin and monomeric actin (G-actin) are balanced by an equilibrium reaction of actin incorporation at the barbed end of a filament and actin dissociation at its pointed end (“treadmilling”). The actin treadmilling cycle is temporally and spatially coordinated by signaling cascades that target actin binding proteins, such as ADF/cofilin and profilin (Beckerle, 1998).
Actin cytoskeletal structures include cortical actin, stress fibers, lamellipodia, and microspikes. Remodelling of preexisting actin filaments into such structures is mainly controlled by members of the Rho family of GTPases (Nobes and Hall, 1995). RhoA activation leads to the formation of actin stress fibers, consisting of long bundles of filaments traversing the cell (Ridley and Hall, 1992). Actin stress fibers are linked to integrins at the inner surface of the plasma membrane involving a multicomponent protein complex called focal adhesion (FA)*
Abbreviations used: ES, embryonic stem; FA, focal adhesion; FAK, FA kinase; IEG, immediate early gene; MLC, myosin light chain; MLCK, MLC kinase; RT, reverse transcription; SRE, serum response element; SRF, serum response factor; TCF, ternary complex factor.
Serum response factor (SRF) is a transcription factor of the MADS box family (Norman et al., 1988; Shore and Sharrocks, 1995). DNA binding sites for SRF (serum response elements [SREs] or CArG boxes) have been found in the promoters of ∼50 different genes so far (unpublished data), including immediate early genes (IEGs) like c-fos and Egr-1 (Herschman, 1991) and muscle-specific genes (e.g., the Actin genes, Dystrophin) (for review see Sobue et al., 1999). SRF exhibits important functions in concert with accessory proteins, e.g., ternary complex factors (TCFs) (Shaw et al., 1989) or other partner proteins. TCF-independent activation of SRF target genes has also been described (Graham and Gilman, 1991; Hill et al., 1994; Johansen and Prywes, 1994) and seems to be particularly important for the regulation of muscle-specific genes (Carnac et al., 1998; Wei et al., 1998). One TCF-independent signaling pathway targeting SRF involves RhoA (Hill et al., 1995), whereby downstream signaling components may include Rho kinase (Chihara et al., 1997), and possibly the transcription factors NF-κB and C/EBPβ as potential SRF partner proteins (Montaner et al., 1999). Recently, actin dynamics were demonstrated to activate SRF in fibroblasts (Sotiropoulos et al., 1999), regulating the endogenous Vinculin, Actin, and Srf genes largely in a RhoA-dependent manner (Gineitis and Treisman, 2001). On the other hand, in cardiac myocytes, an organized actin cytoskeleton was found to be essential for RhoA-dependent activation of SRF (Wei et al., 2001). SRF itself controls the expression of genes encoding cytoskeletal proteins, including vinculin and different actins (Frederickson et al., 1989; Sartorelli et al., 1990; Lee et al., 1991; Moiseyeva et al., 1993; Mack and Owens, 1999).
Functions of SRF in multicellular organisms were revealed by the Drosophila melanogaster mutations pruned and blistered, two alleles of the Drosophila SRF homologue (DSRF) (Affolter et al., 1994; Guillemin et al., 1996). Both mutant phenotypes are, at least in part, due to cytoskeletal defects. In mammals, SRF fulfills an essential function in mesoderm formation. Mouse embryos lacking SRF do not form any detectable mesoderm cells, apparently do not express mesodermal marker genes, show a drastic reduction in IEG expression, and die during early gastrulation (Arsenian et al., 1998).
To explore a potential role of SRF in regulating cytoskeletal activities, we used murine ES cells homozygous for an Srf null allele (Srf[−/−]). We provide evidence that SRF-deficient ES cells have reduced adhesive and migratory capacities, accompanied by the inability to assemble actin stress fibers and FAs. This reveals a direct functional link between the transcription factor SRF, the formation and remodelling of actin-based cytoskeletal structures, and the generation of functional FA complexes.
Srf( ES cells display impairments in cell spreading, substrate adhesion, and cell migration −/ −)
ES cells represent a highly informative in vitro cell system to study the phenotypic effects of gene disruption at the cellular level. Srf(−/−), Srf(−/+), and Srf(−/−)rescue murine ES cell lines have been described previously (Weinhold et al., 2000; Schratt et al., 2001). Plating of undifferentiated wild-type or heterozygous Srf(−/+) ES cells on gelatine-coated tissue culture plates in the presence of leukemia inhibitory factor (LIF) resulted in cell spreading of ∼30–40% of the cells after 2 h (arrows in Fig. 1
A, top). In contrast, cells of two independent Srf(−/−) ES cell lines remained rounded under these conditions, without any signs of cell surface protrusions (Fig. 1 A, bottom). ES cell spreading on different matrices after 2 h is quantified in Fig. 1 B. In contrast to plating on gelatine, only little or no difference between ES cells of different Srf genotype was observed regarding cell spreading on laminin and fibronectin. However, individual Srf(−/−) ES cells were spread to a lesser extent on these two matrices (unpublished data).
The ability to form cell projections and associated cell–matrix contacts is a prerequisite for cellular activities like substrate adhesion and migration. Therefore, we monitored adhesion properties and cell migration behavior of SRF-deficient ES cell lines. ES cells were plated for 1 h on fibronectin, laminin, or collagen IV, and the number of stably adhered cells was measured (Fig. 1 C). No differences were observed between different ES cell lines regarding adhesion to fibronectin. In contrast, adhesion to laminin and collagen IV was significantly reduced by ∼50% in Srf(−/−) ES cells as compared with Srf(−/+) ES cells. These results indicate that a subset of adhesion receptors that confers specific binding to laminin and collagen IV, but not to fibronectin, is affected by SRF deficiency.
To assess the effect of SRF deficiency on the migration behavior of ES cells, we performed Boyden chamber assays. A FBS gradient served as chemo-attractant. The amount of cells that migrated through the membrane toward higher FBS concentration was significantly reduced in both Srf(−/−) ES cell lines tested, as compared with wild-type, Srf(−/+), or Srf(−/−)rescue ES cells (Fig. 1 D). This difference did not arise as a consequence of altered adhesion properties, since Srf(−/−) ES cells adhere to a fibronectin matrix as strongly as wild-type cells (Fig. 1 C). In summary, these results show that SRF deficiency is associated with impaired cell surface activities, namely cell spreading, adhesion, and migration, which all depend on the proper establishment of FAs.
Deregulated expression of FA proteins in
Srf( ES cells −/ −)
Integrins and FA complexes link the extracellular matrix to the actin cytoskeleton. Since Srf(−/−) ES cells display impaired spreading and adhesion properties when plated on gelatine, we measured by Western blotting whether SRF deficiency affects the expression of the FA proteins α-actinin, FAK, β1-integrin, talin, vinculin, and zyxin (Fig. 2
A). Neither the expression of α-actinin nor FAK was affected by the absence of SRF. The β1-integrin antiserum recognized two proteins in the range of 120 kD in SRF-containing cell extracts. These proteins most likely represent the precursor (p) and the mature (m) β1-integrin chain (Belkin et al., 1997). In the two independent Srf(−/−) ES cell lines tested, the expression of the mature β1-integrin subunit was significantly reduced. In addition, a smaller protein of ∼60 kD was recognized by the β1-integrin antibody in Srf(−/−) ES cell extracts (asterisk in Fig. 2 A). These data suggest that the maturation of β1-integrin is impaired in the absence of SRF. In agreement with previous findings (Priddle et al., 1998), two talin isoforms were present in ES cell extracts (Fig. 2 A). Protein levels of both talin isoforms were severely reduced in Srf(−/−) ES cells. Furthermore, protein levels of the known SRF target vinculin were reduced to ∼50% of wild-type levels in Srf(−/−) ES cells. Interestingly, the full-length zyxin protein was nearly undetectable in Western blots of Srf(−/−) ES cell extracts (Fig. 2 A). Constitutive expression of wild-type SRF in an Srf null background completely restored the expression of β1-integrin, talin, vinculin, and zyxin, demonstrating that the observed defects in FA protein expression are a direct consequence of SRF deficiency.
To quantitate FA protein expression at the transcript level, we performed quantitative reverse transcription (RT)-PCR analysis with primers specific for β-actin, β1-integrin, Talin, Vinculin, and Zyxin. β1-Integrin mRNA levels did not significantly differ between ES cell lines of different Srf genotype (Fig. 2 B). Therefore, in light of the corresponding protein data (Fig. 2 A), we infer that SRF deficiency influences β1-Integrin expression at the posttranscriptional level. Similarly, in the case of Talin, mRNA levels in Srf(−/−) ES cells were reduced to ∼60% of wild-type levels, whereas talin protein was almost undetectable in Srf(−/−) ES cells (Fig. 2 A). Apparently, SRF-dependent posttranscriptional mechanisms also contribute to the regulation of talin expression. In contrast, SRF deficiency affects β-actin, Vinculin, and Zyxin expression at both the transcript and protein levels to similar extents (compare Fig. 2, A and B). We are currently analyzing whether SRF acts directly on the promoters of the Talin and Zyxin genes, in addition to regulating transcription of β-actin (Frederickson et al., 1989) and Vinculin (Moiseyeva et al., 1993; Gineitis and Treisman, 2001). The data presented here show that loss of SRF is associated with a reduction in the expression levels of several FA proteins. This dysregulation is seen at both the transcriptional and posttranscriptional levels and suggests a mechanism by which loss of SRF function interferes with cytoskeletal organization and FA activity.
Mislocalization of FA proteins in individual
Srf( ES cells −/ −)
Next, we tested the distribution of FA proteins in individual ES cells. Srf(+/+) and Srf(−/−) ES cells were trypsinized and replated on gelatine for 1 h. Subsequently, adherent cells were fixed and processed for indirect immunofluorescence with antibodies against α-actinin, FAK, β1-integrin, talin, vinculin, and zyxin, together with phalloidin to visualize F-actin. The majority of wild-type ES cells is not yet well spread after 1 h of cultivation. The small fraction of wild-type cells already spread after 1 h displays a characteristic punctate staining of FA proteins at the tips of actin stress fibers (unpublished data).
We first monitored the distribution of FA proteins in representative, individual Srf(+/+) and Srf(−/−) ES cells that were not spread. The distribution of α-actinin in isolated Srf(−/−) ES cells is similar to that in wild-type ES cells (Fig. 3, A and H)
. Membrane localization of FAK, however, is dramatically reduced in Srf(−/−) ES cells (Fig. 3, B and I), although overall expression levels of FAK are unchanged (Fig. 2 A). The epitope recognized by the β1-integrin antibody is localized preferentially at the cell margin in both ES cell lines tested (Fig. 3, C and J). Since protein levels of the mature β1-integrin subunit are reduced in Srf(−/−) ES cells (Fig. 2 A), we speculate that the immunofluorescence signal is generated by a COOH-terminally truncated β1-integrin variant that is still able to localize to the membrane. Talin is predominantly found at the membrane of Srf(+/+) and Srf(−/−) ES cells (Fig. 3, D and K), although the overall protein level is greatly diminished by SRF deficiency (Fig. 3 K). Staining for Vinculin and Zyxin revealed that these two proteins do not exclusively localize to the membrane in Srf(−/−) ES cells (Fig. 3, E, F, L, and M). Surprisingly, staining of SRF-deficient cells with antizyxin antibody gave a relatively strong signal (Fig. 3 M), although full-length zyxin protein was not detectable by Western blotting in Srf(−/−) cell extracts (Fig. 2 A). This may be explained by the presence of additional zyxin variants in Srf(−/−) ES cells (Murthy et al., 1999).
The cells investigated above were simultaneously analyzed for the localization of F-actin by phalloidin staining (Fig. 3, G and N). In wild-type ES cells, F-actin is predominantly found in the cell cortex (Fig. 3 G). All FA proteins tested colocalize with F-actin at the cell margins of wild-type ES cells (unpublished data). In agreement with earlier findings (Schratt et al., 2001), F-actin staining in Srf(−/−) ES cells is severely reduced (Fig. 3 N), with a frequent local concentration at structures that resemble membrane blebs (arrow). In summary, isolated Srf(−/−) ES cells display both reduction and mislocalization of a variety of proteins that have important functions in the attachment of the cortical actin network to the membrane and the transmission of signals from the ECM.
Impaired FAK activation in
Srf( ES cells −/ −)
The nonreceptor tyrosine-kinase FAK is not essential for the formation of FAs (Ilic et al., 1995a), but plays a critical role in integrin-stimulated cell migration (Sieg et al., 1999). Activation of FAK involves its recruitment to the membrane upon integrin-clustering and subsequent autophosphorylation at critical tyrosine residues. Phosphorylation of FAK at Tyr-397 creates a binding site for other kinases with known functions in cytoskeletal dynamics, such as Src, Fyn, and PI-3 kinase. We therefore investigated whether the observed FAK delocalization in isolated Srf(−/−) ES cells (Fig. 3 I) correlated with changes in FAK activity. Plating of wild-type ES cells on fibronectin led to a strong induction of FAK-Tyr397 phosphorylation after 10 min, indicative of FAK activation. The amount of Tyr-397–phosphorylated FAK stayed fairly constant in a 60-min time course (Fig. 4
A, top). A similar kinetic profile of FAK activation was observed for Srf(−/−) ES cells. However, the amount of Tyr-397 phosphorylated FAK was substantially reduced to ∼50% of wild-type levels at each measured time point. Western blotting with pan-FAK antibody was used to normalize for overall FAK recovery in wild-type and Srf(−/−) ES cells (Fig. 4 A, bottom). Transient overexpression of SRF-VP16 in Srf(−/−) ES cells was used to demonstrate the ability of activated SRF to restore FAK Tyr-397 phosphorylation (Fig. 4 B). The relative amount of Tyr-397–phosphorylated FAK was 2–3-fold higher in SRF-VP16 transfected as compared with mock-transfected 100 Srf(−/−) ES cells. A rescue of FAK activation was not observed in the ES cell line 100-2 Srf(−/−)rescue (unpublished data). This unexpected finding may reflect the fact that this cell clone expresses 3.5-fold more SRF protein than wild-type cells (Weinhold et al., 2000). It is therefore conceivable that a large fraction of SRF in these cells is not activated and competes with activated SRF on cytoskeletal promoters.
Our data indicate that impaired expression and mislocalization of FA components in Srf(−/−) ES cells correlate with defective transmission of signals from the ECM to important intracellular signal transducers like FAK. This finding offers a molecular link between the observed structural defects in FA complexes and reduced cell motility of Srf(−/−) ES cells.
Srf( ES cells are unable to assemble FAs and associated actin stress fibers −/ −)
Actin-based cytoskeletal structures, such as stress fibers and lamellipodia, are characteristically found with spread cells. We therefore analyzed wild-type, Srf(−/−), and Srf(−/−)rescue ES cells for their abilities to form actin stress fibers, lamellipodia, and associated FAs. For this assay, cells were plated for 48 h on either fibronectin (unpublished data) or gelatine. Under these conditions, only wild-type, Srf(−/+), and Srf(−/−)rescue cells were able to form individual, well-spread cells. Srf(−/−) ES cells, in contrast, were always distinctly nonspread and part of large colonies after 48 h of cultivation. Wild-type and Srf(−/−)rescue ES cells assembled numerous actin stress fibers that terminated in FAs. These cells stained positive for the FA marker proteins α-actinin, FAK, vinculin, and zyxin (Fig. 5
, left and right). Moreover, the formation of large lamellipodia could be observed in a subset of these cells (arrows in Fig. 5). In contrast, in Srf(−/−) ES cells, F-actin staining was limited to cell margins and was rather diffuse (Fig. 5, middle). Lamellipodia or stress fibers were never observed in Srf(−/−) ES cells. In addition, staining for α-actinin, FAK, and vinculin was limited to the cell margins and did not show the typical punctate pattern. In the case of vinculin, however, some punctate staining could be observed at the periphery of Srf(−/−) ES cell colonies. Zyxin staining was rather diffuse, and neither displayed a punctate pattern nor a preferential membrane localization. Fibronectin as substrate neither lead to an assembly of FAs nor of stress fibers in Srf(−/−) ES cells (unpublished data), as was previously seen with Talin(−/−) ES cells (Priddle et al., 1998). We conclude that undifferentiated Srf(−/−) ES cells are severely impaired in building up distinct cytoskeletal structures, such as actin stress fiber arrays, lamellipodia, and associated FA plaques. The defects in cytoskeletal organization are probably a major reason for altered spreading, adhesion, and migration of SRF-deficient ES cells.
Srf(− /− ) ES cells display reduced overall actin expression and an altered ratio of F- versus G-actin
The expression of genes encoding for several muscle- and nonmuscle actin isoforms is regulated by SRF (Chen et al., 1996; Frederickson et al., 1989; Sobue et al., 1999). In addition, we recently showed a two- to threefold reduction of overall actin protein levels in Srf(−/−) ES cells (Schratt et al., 2001). Given the lack of lamellipodia and actin stress fibers in Srf(−/−) ES cell colonies (Fig. 5), we investigated whether SRF deficiency also had an impact on the F/G-actin treadmilling equilibrium. Thus, we analyzed the actin polymerization state inside cells. First, we performed a sedimentation assay to monitor G- and F-actin levels in cell extracts. In agreement with previous results, total actin in cell extracts of two Srf(−/−) ES cell lines was reduced to ∼40% of wild-type levels (Fig. 6
A, lanes 1–4). The amount of nonpelletable G-actin in the supernatants of Srf(−/−) ES cell extracts after high-speed centrifugation was reduced by 20–40%, as compared with wild-type extracts (Fig. 6 A, lanes 5–8). In contrast, pelletable F-actin from Srf(−/−) ES cells was reduced by ∼80% (Fig. 6 A, lanes 9–12). SRF deficiency affecting F-actin levels more severely than G-actin levels was confirmed by FACS analysis. In this assay, a twofold reduction of G-actin levels was observed in Srf(−/−) ES cells as compared with Srf(−/+) cells (Fig. 6 B, top). In contrast, F-actin levels, as determined by phalloidin staining, were affected more than 3.5-fold (Fig. 6 B, bottom). Srf(−/−)rescue cells displayed a G-/F-actin ratio nearly identical to that of Srf(−/+) cells (unpublished data). Thus, both sedimentation and FACS analysis demonstrate that the F-actin pool is more severely affected by the absence of SRF than the G-actin pool. These results point to a role for SRF not only in the regulation of overall actin expression levels, but also in balancing F- versus G-actin. Reduced F-actin levels correlate with the impairment of Srf(−/−) ES cells in stress fiber assembly.
Overexpression of constitutively active RhoA fails to induce stress fibers in
Srf(− /− ) ES cells
The function of RhoA in the induction of FAs and associated actin stress fibers has been extensively characterized (Chrzanowska-Wodnicka and Burridge, 1996). Furthermore, RhoA has been implicated in regulating the actin treadmilling cycle more directly, either through inactivation of the actin depolymerizing factor ADF/cofilin (Maekawa et al., 1999) or through activation of the polymerization-promoting factor profilin (Watanabe et al., 1997). Finally, activated RhoA is able to drive SRF-dependent transcription (Hill et al., 1995). We therefore determined the importance of functional SRF for RhoA-mediated stress fiber formation. Constitutively active RhoA-V14 induces characteristic stress fiber bundles in wild-type ES cells (arrows in Fig. 7
B). This effect is Rho-specific, since expression of the control protein red fluorescent protein (RFP) had no effect on the cytoskeleton (unpublished data). In contrast, RhoA-V14 is not able to induce stress fiber formation in SRF-deficient cells (Fig. 7 E). Instead, RhoA-V14–transfected cells frequently display a large number of bleb-like structures (Fig. 7, D–F). Phalloidin staining revealed that F-actin is concentrated within these blebs (arrow in Fig. 7 E). Again, expression of RFP did not lead to cytoskeletal changes (unpublished data). These results demonstrate that RhoA-mediated stress fiber formation requires the expression of SRF-dependent genes downstream of RhoA. If such genes are not properly expressed due to SRF-deficiency, increased acto-myosin contractility in response to RhoA activation leads to membrane blebbing.
Constitutively active SRF (SRF-VP16) is sufficient to induce molecular and cell morphological changes in ES cells of
Srf(− /− ) genetic background
SRF activity is stimulated by effector kinases downstream of the RhoGTPase and MAP kinase pathways (Hill et al., 1995). SRF-VP16 fusion proteins were previously used to drive SRF-dependent transcription independent of the activity of upstream signaling cascades (Treisman and Ammerer, 1992; Johansen and Prywes, 1994; Hines et al., 1999). Such a constitutively active SRF variant, when introduced into Srf(−/−) ES cells, allowed us to test for a putative link between the SRF activation state and the establishment of cytoskeletal structures. To control for potential unspecific effects of the VP16 activation domain, we constructed SRFΔM-VP16, which lacks most of the MADS box. This mutant cannot dimerize nor bind to SREs, but still locates to nuclei (Fig. 8
A and unpublished data). Undifferentiated ES cells were transiently transfected with SRF-VP16 or SRFΔM-VP16 expression vectors, and RNA expression levels were measured by quantitative RT-PCR (Fig. 8 B). In an Srf(−/−) background, the mRNA levels for the IEGs Egr-1 (Fig. 8 B, top left) and c-fos (unpublished data) were increased ∼400-fold by the introduction of SRF-VP16, as compared with mock-transfected cells, confirming the capacity of SRF to drive expression of these genes. Importantly, introduction of SRFΔM-VP16 had no significant effect. Zyxin mRNA levels were specifically induced 3–6-fold by SRF-VP16 (Fig. 8 B, top right), similar to the known SRF-regulated genes Vinculin and β-Actin (Fig. 8 B, bottom). Therefore, SRF is identified to drive the expression of several cytoskeletal genes, including Zyxin. These results confirm previous findings that showed the ability of SRF expression in Srf(−/−) ES cells to restore vinculin and zyxin protein levels (Fig. 2 A). At the morphological level, introduction of SRF-VP16 into Srf(−/−) ES cells led to extensive cell flattening (arrows in Fig. 8 C, right) in ∼25% of cells, whereas cells transfected with the control construct looked unaltered (Fig. 8 C, left). Since the transfection efficiency was ∼20–30% in this experiment (unpublished data), flattened cells most likely represent cells that overexpress SRF-VP16. We speculated that SRF-VP16–induced cytoskeletal rearrangements are responsible for these changes in cell morphology. Indeed, staining for F-actin revealed the formation of massive arrays of stress fibers (Fig. 8 D, left) and frequent occurrence of lamellipodia (unpublished data) in each SRF-VP16–transfected Srf(−/−) ES cell. The cytoskeleton of SRFΔM-VP16–transfected Srf(−/−) ES cells did not undergo any detectable changes (Fig. 8 D, right). Both immunofluorescence staining with anti-VP16 antibody and Western blotting (unpublished data) confirmed that SRF-VP16 and SRFΔM-VP16 were expressed at comparable levels and that both proteins accumulated in the cell nuclei. Interestingly, the effects elicited by expression of constitutively active SRF on the expression of target genes (Fig. 8 B) and ES cell morphology (unpublished data) were much more pronounced in Srf(−/−) ES cells than in ES cells of wild-type or Srf(−/+) genetic background. Endogenous SRF protein present in the latter two cell lines, not being activated under these growth conditions, probably competes with SRF-VP16 for binding to cognate promoters. We note that transient overexpression of wild-type SRF in Srf(−/−) ES cells caused cell spreading only in a small fraction of transfected cells (unpublished data). In agreement with previous results (Schratt et al., 2001), introduction of SRF-VP16 into Srf(−/−) ES cells led to a strong increase in overall actin expression, accompanied by a strong elevation in F-actin and only a minor increase in G-actin, as judged by FACS analysis (Fig. 8 E). Clearly, a net shift of the actin equilibrium in favor of F-actin is seen. This effect was specific for SRF-VP16, since the expression of the control construct SRFΔM-VP16 had no significant influence on actin treadmilling. These data demonstrate conclusively that a constitutively active form of SRF is sufficient to regulate actin treadmilling and to elicit drastic cell morphological and molecular changes that are linked to remodelling of the actin cytoskeleton.
The targeted mutagenesis of ES cell genomes by homologous recombination offers a powerful tool to study gene function at the cellular level. We have used such ES cell genetics to study the physiological roles of the transcription factor SRF. In addition to generating null alleles at the Srf locus, we have used Srf(−/−) ES cells to reintroduce constitutively active variants of SRF and RhoA into these SRF-deficient cells. We uncovered a far reaching role of SRF in governing the structure and function of cytoskeletal structures, namely the actin cortex, actin stress fibers, and FA plaques. Consequently, cellular activities dependent upon these structures, such as cell spreading, adhesion, and migration, were found to be affected by SRF deficiency.
SRF regulates the actin treadmilling equilibrium
We have shown previously that expression of the SRE-containing β-actin gene was found to be significantly reduced in ES cells lacking SRF (Schratt et al., 2001). Furthermore, the promoters for the cell-type-specific α- and γ-Actins also contain functional SRF binding sites (Mohun et al., 1989; Kovacs and Zimmer, 1998). Interestingly, in addition to the reduced overall actin expression levels, the treadmilling equilibrium of G- versus F-actin is shifted in Srf(−/−) ES cells, thereby reducing F-actin levels disproportionately. It is conceivable that the altered G/F-actin ratio is simply due to mass action, i.e., due to reduced overall actin expression levels. However, increasing the pool of monomeric actin in Srf(−/−) ES cells by overexpressing HA-tagged actin did not result in a detectable increase in F-actin synthesis (unpublished observation). Overexpression of both actin and a constitutively active variant of RhoA (RhoA-V14) allowed some de novo actin polymerization (unpublished data). However, this treatment did not fully rescue stress fiber formation in Srf(−/−) ES cells, whereas SRF-VP16 elicited a very strong stress fiber response in these cells. Therefore, we postulate that SRF affects actin treadmilling by regulating the expression of additional proteins involved in actin turnover downstream of RhoA. Such candidates include members of the LIM kinase or mDia pathways, which have been described to influence actin polymerization via regulation of cofilin and profilin activation states, respectively (Watanabe et al., 1997; Maekawa et al., 1999). Overexpression of LIM kinase 2 alone or in conjunction with RhoA-V14, however, did not lead to enhanced F-actin levels (unpublished data). SRF-regulated zyxin/vinculin expression may also contribute to F-actin synthesis, since these proteins are able to recruit the actin polymerization machinery via binding to Ena/VASP family members (Laurent et al., 1999). Finally, talin, whose expression is partially dependent on SRF, has been reported to possess actin nucleation activity (Kaufmann et al., 1991). Treisman and coworkers recently demonstrated that, in fibroblasts, SRF itself is regulated by actin dynamics (Sotiropoulos et al., 1999). In that study, decreases in G-actin levels were argued to stimulate SRF activation. We describe here that SRF controls signaling components, yet to be identified (class Y in the model of Fig. 9)
, which act downstream of RhoA and influence the F-/G-actin equilibrium. Therefore, we propose that SRF is a central component of an autoregulatory feed-back loop controlling actin polymerization (Fig. 9). Such a mechanism may allow cells to adapt their F-actin synthesis to signal-induced cytoskeletal changes.
SRF is required for the formation of FA plaques and stress fibers
Several lines of evidence suggest that actin polymerization and actin organization, the latter referring to the formation of both stress fibers and FAs, are mediated by distinct effectors of RhoA (for review see Van Aelst and D'Souza-Schorey, 1997). This distinction is mirrored by the ability of SRF not only to regulate components acting downstream of RhoA but also to control the expression of cytoskeletal components, like the myosin light chain (MLC) gene (Papadopoulos and Crow, 1993). A potential mechanism for Rho-induced stress fiber formation involves phosphorylation of MLC by activated MLC kinase (MLCK) and subsequent increase in acto-myosin contractility (Burridge and Chrzanowska-Wodnicka, 1996). It is therefore possible that impaired MLC expression may contribute to the defect in stress fiber and FA formation in Srf(−/−) ES cells. Introduction of RhoA-V14 in Srf(−/−) ES cells lead to increased membrane blebbing which was also observed in cells with defective actin-membrane coupling in response to MLC phosphorylation (Coleman et al., 2001; Sebbagh et al., 2001). This indicates that the Rho-kinase/MLCK pathway can be activated, at least to some extent, in Srf(−/−) ES cells.
Regarding FAs, the failure of Srf(−/−) ES cells to assemble functional FAs may be associated with inadequate expression of essential FA components. Indeed, we find misexpression and/or mislocalization of several structural FA proteins, namely FAK, β1-integrin, talin, vinculin, and zyxin. Integrin clustering is the initial event in FA formation (Gumbiner, 1996). Therefore, impaired integrin expression and/or activation could account for the adhesion and spreading defects of Srf(−/−) ES cells. Since spreading is most severely reduced on a gelatin matrix, the function of the collagen receptors α1β1 and α2β1 may be compromised. We observe a reduced protein expression of the full-length β1 integrin subunit in Srf(−/−) ES cells, whereas expression of the α1 and α2 subunit is not affected at the mRNA level (unpublished data). β1-integrin is also important for adhesion to fibronectin (Fassler et al., 1995), but the upregulation of other β subunits (e.g., β3) may compensate for reduced β1-integrin expression in this case (Priddle et al., 1998). The regulation of β1-integrin expression by SRF seems to occur at the post-transcriptional level, since (a) β1-integrin mRNA levels are unaffected by SRF deficiency and (b) a truncated β1-integrin protein is enriched in Srf(−/−) ES cells. Recently, calpain-mediated cleavage of β1-integrin has been reported (Pfaff et al., 1999), and we are currently testing a putative link between SRF and calpain activity in ES cells.
Since FAK is able to bind β1-integrin, FAK delocalization in Srf(−/−) ES cells may be a direct consequence of β1-integrin cleavage. Reduced FAK activation may affect migration, differentiation, and survival of cells. Fibroblasts expressing the FAK Y397A mutant have migration defects (Sieg et al., 2000), indicating that recruitment to FAK of tyrosine kinases, such as Fyn and Src, is critical for cell migration. Also, FAK activation enables recruitment and activation of PI-3 kinase. The FAK/PI-3-kinase pathway counteracts apoptosis initiated by serum deprivation or deadhesion (Schlaepfer et al., 1999). Indeed, differentiating Srf(−/−) ES cells tend to undergo increased apoptosis (unpublished data), possibly due to inadequate activation of the FAK/PI-3 kinase survival pathway. We also present evidence that the scaffolding protein talin is regulated by SRF at the transcriptional level. Undifferentiated talin-deficient ES cells display striking phenotypic similarities to Srf(−/−) ES cells, including defects in cell shape, adhesion, and β1-integrin expression (Priddle et al., 1998). However, talin is dispensable for cytoskeletal organization in differentiated ES cells. Since we do not observe a well-established cytoskeleton in differentiated Srf(−/−) ES cells (unpublished data), we believe that reduced talin expression alone is not sufficient to explain the abnormal actin cytoskeleton in the absence of SRF. The same is true for another SRF-regulated protein, vinculin, that has been shown to be dispensable for FA formation (Xu et al., 1998). Finally, the expression of zyxin is strongly dependent upon SRF. Inhibitory peptides that block zyxin/α-actinin interaction, and thereby displace zyxin from FAs, predict a role for zyxin in cell spreading and motility (Drees et al., 1999). We do not know how the loss of zyxin affects FA formation, but downregulation of zyxin may contribute to the reduced motility of Srf(−/−) ES cells.
Together, we demonstrate that SRF is important for the expression of a variety of proteins that have known functions in the formation and/or stabilization of FAs. Furthermore, others have implicated SRF in the transcriptional regulation of both myosin light and heavy chains (Papadopoulos and Crow, 1993; Catala et al., 1995; Zilberman et al., 1998), α1-integrin (Sobue et al., 1999), tropomyosin (Toutant et al., 1994), and dystrophin (Galvagni et al., 1997). This strongly argues that SRF is a central determinant of the formation and activity of FAs. In addition, a recent study reported that, in myocytes, FA signaling cooperates with RhoA in SRF activation requiring a properly organized actin cytoskeleton (Wei et al., 2001). In light of these results, SRF-dependent regulation of β1-integrin maturation and FAK activity might also be part of a positive feed-back loop (Fig. 9). Such a mechanism may be important in muscle cells, where initial Rho-induced cytoskeletal changes have to be reinforced by increased transcription of SRF-dependent cytoskeletal genes.
Cellular behavior under SRF control
We have demonstrated that the cellular activities of spreading, adhesion, and migration are severely affected by SRF deficiency. The structural basis for these defects in Srf(−/−) ES cell behavior may rest jointly on unbalanced actin dynamics and impaired formation of FA plaques and stress fibers. Cytoskeletal defects might also contribute to the inability of Srf(−/−) mouse embryos to form mesoderm (Arsenian et al., 1998). Inductive events during embryogenesis require precise cell adhesion mechanisms, either to neighboring cells or to the extracellular matrix (Hynes, 1999). Inappropriate cell adhesion displayed by cells of the early Srf(−/−) embryo could therefore impair mesoderm induction and migration of newly formed mesodermal germ layer cells (Tam and Behringer, 1997). In summary, the Srf(−/−) ES cell system revealed a hitherto unrecognized and physiologically important role of SRF in determining the formation of specific cytoskeletal structures and, consequently, in governing cellular activities that require dynamic changes of the cytoskeleton. Such SRF-directed activities include cell adhesion and migration, and are suggested to be of relevance for more complex biological processes, like directed cell movement during embryogenesis, wound healing, and metastasis.
Materials And Methods
Reagents and antibodies
Laminin (mouse) and fibronectin (calf) were purchased by Harbor Bio Products and collagen IV was purchased by GIBCO BRL. Primary antisera were used which recognized the following antigens: FAK (Transduction Laboratories), phospho-FAK (BioSource International), vinculin, α-actinin, talin (all from Sigma-Aldrich), zyxin (gift of J. Wehland, GBF Braunschweig), β1-integrin (gift of R. Hynes, MIT), VP16, HA (both from Santa Cruz Biotechnology, Inc.), actin (gift of H. Knauth, MPI Tübingen), and myc (9E10 hybridoma supernatant).
To obtain pCS2+/hSRF-VP16, a 1.8-kb NcoI-XbaI fragment (NcoI site filled-in) was isolated from pKD24 (Dalton and Treisman, 1992), and subcloned into pCS2+ (gift of R. Rupp, MPI Tübingen) linearized with StuI-XbaI. SRFΔM-VP16 was generated from pCS2+/hSRF-VP16 by removing a fragment from the BglII to the BbsI site. The remaining vector sequences were religated with a double-stranded oligonucleotide of the following sequence: 5′-GG GGC GAA GAG ACA-3′ (top strand); 5′-G ATC TGT CTC TTC G-3′ (bottom strand). RhoA-V14 myc tag was excised from pUHD-RhoA-V14 myc-tag (a gift from M. Baehler, Muenster) and subcloned into pCS2+. Expression of the inserted cDNAs in pCS2+ is driven by the CMV enhancer/promoter.
ES cell culture conditions
General growth conditions.
The E14.1 ES cell line is derived from the 129/Ola mouse strain and has been described previously (Kuhn et al., 1991). ES cells were kept without feeders on gelatine-coated dishes in DME medium containing 4.5 g/liter glucose and 3.7 g/liter NaHCO3, supplemented with 2 mM l-glutamine, 100 U/ml penicillin, 100 g/ml streptomycin, 450 μM monothioglycerol, 15% FBS, and 100 U/ml LIF (referred to as complete medium) at 37°C in a humidified atmosphere at 5% CO2, and split every 2 d.
ES cells were seeded the day before transfection at 4 × 105 cells/10 cm dish or 105 cells per 6-well and grown overnight. Transfection was performed with LipofectAmin (GIBCO BRL) according to the manufacturer's protocol. A 10-fold excess of LipofectAmin over total DNA was used throughout the transfections. 10 μg of the described SRF expression plasmids were used per 10-cm dish, or 1 μg of all other expression plasmids per 6-well. Total DNA amount was kept constant at 10 μg per 10-cm dish or 2 μg per 6-well with pCS2+mt or pCS2+. 48–72 h after transfection, cells were either harvested for RNA preparation or fixed for immunofluorescence analysis.
96-well plates were coated with 25 μg/ml fibronectin, 25 μg/ml laminin or 50 μg/ml collagen IV at 4°C overnight. To avoid nonspecific binding, wells were washed once with PBS and blocked with 1% BSA for 1 h at 37°C. ES cells grown for 48 h in the presence of LIF were trypsinized to obtain a single-cell suspension, washed once in serum-free medium, and plated on the matrix-coated wells for 1 h at 37°C in serum-free medium. Nonadherent cells were washed away and adherent cells were fixed for 30 min with 10% formaldehyde. After washing with PBS, fixed cells were stained with crystal violet for 15 min. Residual dye was removed by extensive washing with PBS. Stained cells were air-dried, and the dye was extracted with 50 μl 2% SDS per well. Absorption of the solution was measured at 630 nm. The absorption measured from wells coated with BSA only served as a reference.
Migration assay (Boyden chamber assay)
Cell culture filter inserts (Becton Dickinson; pore size 8 μm) for 6-well plates were coated on the lower surface overnight with 20 μg/ml fibronectin at 4°C. After washing with PBS, the inserts were put into 6-well plates to obtain a modified Boyden chamber. The lower reservoir of the chamber was filled with complete medium. ES cells were trypsinized to a single-cell suspension, washed in serum-free medium, resuspended in serum-free medium to 2.5 × 105 cells/ml, and 2 ml of the suspension was filled into the upper reservoir. Cells were allowed to migrate through the filter for 22 h at 37°C. Cells on the upper surface were scraped off, and those on the lower surface were washed once with PBS and subsequently fixed with 10% formaldehyde. Staining was performed with crystal violet for 30 min, the bound dye was extracted with 1 ml 2% SDS, and the absorption was determined at 630 nm. The absorption measured from filter inserts coated with BSA only served as a reference.
Preparation of cell extracts and Western blotting were as described previously (Weinhold et al., 2000).
Total RNA preparation (RNeasy kit; QIAGEN) and first strand cDNA synthesis (Superscript II, GIBCO BRL) were done according to the manufacturer's protocols. 1 μg of total RNA treated with DNase I was used for RT. One twentieth of the RT reaction was included in a 25 μl PCR reaction. For a quantitative analysis, SYBR green PCR technology was used (PerkinElmer). Real-time detection of the PCR product was monitored by measuring the increase in fluorescence caused by the binding of SYBR green to double-stranded DNA with ABI PRISM 7700 Sequence Detector. To calculate relative quantification values, a threshold cycle (Ct), the cycle at which a statistically significant increase in fluorescence occurs, was derived from the resulting PCR profiles of each sample. Ct is a measure for the amount of template present in the starting reaction. To correct for different amounts of total cDNA in the starting reaction, Ct values of an endogenous control (Hprt) were subtracted from those of the corresponding sample, resulting in ΔCt. The relative quantitation value is expressed as 2-ΔCt. All RT-PCR data sets shown (Figs. 2 B and 8 B) provide representative results of several independent mRNA preparations and amplifications, as indicated in the figure legends.
Primers used: Hprt, F(gcctaagatgagcgcaagttg); B(tactaggcagatggccacagg); Egr-1, F(gccgagcgaacaacccta); B(tccaccatcgccttctcatt); Zyxin, F(ggctgctacaccgacactttg); B(ctcagcatgcggtcagtgat); Vinculin, F(ccaaggtcagagaagccttcc); B(cgtagctgttcaaggtctggtg); β-actin, F(ggcgcttttgactcaggatt); B(gggat-gtttgctccaaccaa); Talin, F(gctcgggcgttagcagtc); B(atggaatctgagacagtccgaga); β1-integrin, F(tggctttgatgcaatcatgc); B(tccgtggaaaacaccagca).
ES cells were collected, fixed in 4% formaldehyde for 10 min, permeabilized with 0.2% Triton X-100 for 5 min, and blocked in 1% BSA for 1 h. After three washes in PBS, the cells were incubated for 30 min in a mixture of 1 U Oregon green phalloidin and 0.3 μM Texas red DNase I (both from Molecular Probes) in 1% BSA. For a reference, incubation was performed in the absence of any fluorescent dye. For antibody staining, cells were incubated for 1 h with primary antibody, washed once with PBS, and incubated for an additional 30 min with secondary antibody solution. Actin staining reagents were included in the secondary antibody solution. Rabbit IgG served as isotype control. After washing, cells were subjected to FACscan analysis (Becton Dickinson). Dead cells were gated out and 104 cells were counted for each sample. Data analysis was performed with CellQuest software (Becton Dickinson).
F-actin sedimentation assay
Actin sedimentation assays were performed in principle as described (Jahraus et al., 2001). ES cell pellets were resuspended in 2 ml of lysis buffer (10 mM Hepes, pH 7.6, 100 mM KCl, 1 mM MgCl2, 0.1 mM EDTA, 1 mM DTT, 0.5 mM PMSF) and the cells broken using a probe sonicator. Nuclei were removed by centrifugation (3,000 g, 30 min) and the supernatant (low-speed extract) subjected to a high-speed centrifugation step (400,000 g, 1 h). The clear supernatant (high-speed extract) was collected and concentrated by using Centriprep-10 spin columns (Amicon Corp.). Pellets were subsequently solved in 1% Triton X-100. Equal amounts of low-speed extracts, high-speed extracts, and high-speed pellets were resolved on a 10% SDS-PAGE, blotted, and probed with a polyclonal rabbit anti-pan actin antiserum (gift of H. Knauth, MPI Tübingen). Relative actin levels in each sample were quantified by normalizing Western blot signals to Coomassie staining.
Cells were grown on gelatine-coated coverslips in complete growth medium for 48 h. The samples were then fixed in 4% formaldehyde and permeabilized in 0.2% Triton X-100. Nonspecific binding was blocked by incubating the cells for 1 h in 1% BSA in PBS at 37°C, before staining with primary antibody for 1 h at 37°C. Incubation with fluorescence-conjugated secondary antibodies (Molecular Probes; 1:200 in 0.2% BSA) was performed for 30 min at 37°C. To visualize filamentous actin, 1 U Oregon green phalloidin (Molecular Probes) was included into the mixture. Coverslips were washed four times with PBS, once in water, air-dried, and mounted in Moviol. Image acquisition was done with LSM 510 (ZEISS).
Embryonic stem cells were grown on gelatine-coated coverslips for 2 h. Samples were fixed with 1.6% glutaraldehyde in 20 mM Hepes, pH 7.6, 120 mM NaCl for 5 min at room temperature and for 1 h at 4°C, postfixed with 1% osmium tetroxide in PBS for 1 h on ice, washed with H2O, and treated with 1% aqueous uranyl acetate for 1 h at 4°C. For scanning electron microscopy, cells were dehydrated in ethanol and critical point dried from CO2. The samples were sputter-coated with 8 nm gold palladium and examined at 20 kV accelerating voltage in a Hitachi S-800 field emission scanning electron microscope.
Cells were either transfected as described or left untreated and grown in complete medium for 48 h. The medium was then exchanged and cells were kept for 18 h in complete medium lacking FBS. Cells were then detached by trypsinization, and the reaction was stopped by adding 0.5 mg/ml soybean trypsin inhibitor (Sigma-Aldrich). After washing once with serum-free medium, cells were kept in suspension in serum-free medium for an additional hour at 37°C. Then, the cells were replated on fibronectin-coated tissue plates for 10–60 min at 37°C. Preparation of cell extracts for Western blotting was as described (Weinhold et al., 2000).
This work was supported by the DFG through grants NO 120/7-4 and SFB 446 (B7), and the Volkswagen-Stiftung.