Crk family adaptors, consisting of Src homology 2 (SH2) and SH3 protein-binding domains, mediate assembly of protein complexes in signaling. CrkI, an alternately spliced form of Crk, lacks the regulatory phosphorylation site and C-terminal SH3 domain present in CrkII and CrkL. We used gene silencing combined with mutational analysis to probe the role of Crk adaptors in platelet-derived growth-factor receptor β (PDGFβR) signaling. We demonstrate that Crk adaptors are required for formation of focal adhesions, and for PDGF-stimulated remodeling of the actin cytoskeleton and cell migration. Crk-dependent signaling is crucial during the early stages of PDGFβR activation, whereas its termination by Abl family tyrosine kinases is important for turnover of focal adhesions and progression of dorsal-membrane ruffles. CrkII and CrkL preferentially activate the small GTPase Rac1, whereas variants lacking a functional C-terminal SH3 domain, including CrkI, preferentially activate Rap1. Thus, differences in the activity of Crk isoforms, including their effectors and their ability to be downregulated by phosphorylation, are important for coordinating dynamic changes in the actin cytoskeleton in response to extracellular signals.
Growth factor receptors such as the PDGFβR regulate essential physiological processes in metazoans, including tissue morphogenesis, maintenance and repair. Dysregulated growth-factor-dependent signaling plays a role in a number of diseases such as cancer (Aaronson, 1991; Heldin et al., 1998; Wells, 2000). Cellular migration, which is regulated in part by growth factors, relies on the coordinated assembly and turnover of cellular structures. In particular, localized actin polymerization drives membrane protrusion, and the formation of focal adhesions (FAs) provides a physical link between the actin cytoskeleton and the underlying extracellular matrix that enables productive cell migration (Heldin et al., 1998; Ridley et al., 2003; Wells, 2000).
Activation of PDGFβR by ligand binding leads to the autophosphorylation of tyrosine residues in the cytosolic portion of the receptor. These phosphorylated sites bind to SH2 domains, thus recruiting a number of SH2-domain-containing proteins to the receptor. Among the specific downstream consequences is reorganization of the actin cytoskeleton. One of essential events in this pathway is thought to be activation of the small guanosine triphosphatase (GTPase) Rac1, which promotes Arp2/3-dependent actin polymerization by increasing the actin-nucleating activity of WAVE family proteins (Miki et al., 2000; Ridley et al., 1992). At present, the mechanisms by which the recruitment of SH2-domain-containing proteins to the activated PDGFβR influences the activities of actin cytoskeleton regulators remain largely unknown.
The Crk family of adaptor proteins consists of CrkI, CrkII and CrkL (Fig. 1A) (Matsuda et al., 1992; Mayer et al., 1988; Reichman et al., 1992; ten Hoeve et al., 1993). They are ubiquitously expressed in most tissues and are required for animal development (Guris et al., 2001; Park et al., 2006). CrkII and CrkL consist of one SH2 domain followed by two SH3 domains (designated hereafter as nSH3 and cSH3). In general, SH2 domains bind tyrosine-phosphorylated peptides whereas SH3 domains bind proline-rich sequences (Pawson and Nash, 2003). The SH2 and SH3 domains of CrkII and CrkL are highly similar in sequence (Reichman et al., 1992; ten Hoeve et al., 1993), and their interaction specificities are almost identical (Feller, 2001). This suggests that CrkII and CrkL have overlapping functions. Whereas the SH2 and nSH3 domains are known to function as interaction domains, no partners other than the nuclear exporter Crm1 have been reported to bind to the cSH3 domain of either CrkII or CrkL (Smith et al., 2002). This suggests that the cSH3 domain might not function as a typical SH3 domain.
In addition to these domains, CrkII and CrkL each have a regulatory tyrosine phosphorylation site (Y221 and Y207, respectively) located between the two SH3 domains. Phosphorylation of this site induces intramolecular binding between the phosphorylated tyrosine and the SH2 domain. This masks the binding surfaces of the SH2 and nSH3 domains, lowering their accessibility to other binding partners and shutting off Crk signaling (Kain and Klemke, 2001; Kobashigawa et al., 2007). CrkI is an alternately spliced form of CrkII that lacks the cSH3 domain and regulatory tyrosine phosphorylation site (Matsuda et al., 1992). Thus, unlike CrkII and CrkL, CrkI signaling cannot be downregulated by tyrosine phosphorylation (Kobashigawa et al., 2007), suggesting that the role of CrkI might be distinct from that of CrkII and CrkL.
Crk family adaptors have been implicated in the regulation of cellular adhesion, migration and proliferation (Feller, 2001). They exert their biological effects by interacting with various proteins through their SH2 and nSH3 domains. Key partners include the p130 Crk-associated substrate (p130Cas) and paxillin, both of which are FA proteins that bind the SH2 domains of Crk adaptors when phosphorylated (Birge et al., 1993; Sakai et al., 1994). Crk adaptors also bind via their nSH3 domains and activate several guanine nucleotide exchange factors (GEFs) for small GTPases, including Dock180 and C3G/GRF2 (Hasegawa et al., 1996; Knudsen et al., 1994; Tanaka et al., 1994). In addition, Crk family adaptors associate via their nSH3 domains with Abl family non-receptor tyrosine kinases, which in turn can phosphorylate CrkII and CrkL on the regulatory site (Knudsen et al., 1994). Many of these Crk family binding partners have been implicated in PDGFβR signaling (Cote et al., 2005; Plattner et al., 1999; Rivera et al., 2006; Voss et al., 2003), and CrkII and CrkL are rapidly tyrosine phosphorylated after PDGF stimulation of cells (Matsumoto et al., 2000; Sorokin et al., 1998). Despite these observations, the functional roles of Crk family proteins during PDGFβR signaling remain obscure.
Here, we show that Crk family adaptors are required for FA formation, and for PDGF-stimulated rearrangement of the actin cytoskeleton and cell migration. We also show that inactivation of Crk signaling following PDGF stimulation is important for proper coordination of FA disassembly and to prevent an extended period of actin-cytoskeleton rearrangement. Our results also reveal a novel role for the Crk cSH3 domain in promoting Rac1 activation, whereas Rap1 activation is favored in the absence of cSH3. These results highlight essential roles for Crk family adaptors in PDGFβR signaling, and differences in the biological activity of CrkI compared to CrkII and CrkL.
Crk family adaptors are required for PDGF-stimulated formation of circular dorsal membrane ruffles and cell motility
To determine whether Crk family adaptors play an essential role in PDGFβR signaling, CrkL, CrkI/II or both Crk family adaptor proteins were knocked down by vector-based siRNA in NIH-3T3 fibroblasts, and the responses to PDGF-bb (PDGF) stimulation examined. In these experiments, protein expression was decreased by approximately 90% for CrkI, 85% for CrkII and 99% for CrkL, without affecting the levels of PDGFβR (Fig. 1B).
We first examined rearrangements of the actin cytoskeleton, specifically the formation of circular dorsal ruffles. These are ring-shaped actin structures on the dorsal surface of the cell, which originate at the cell periphery and propagate in a wave-like fashion inwards, ultimately leading to membrane internalization (Fig. 1C) (Heldin et al., 1998; Orth and McNiven, 2006). When stimulated with PDGF for 10 minutes, CrkI/II- or CrkL-knockdown cells formed circular dorsal membrane ruffles to a similar extent as wild-type (wt) cells (Fig. 1C,D). By contrast, the percentage of CrkI/II/L-knockdown cells forming PDGF-induced circular dorsal ruffles was reduced by ∼65% (Fig. 1B,C). These results indicate that PDGF-stimulated formation of dorsal membrane ruffles is dependent on Crk family proteins, and that CrkI/II and CrkL have overlapping functions.
We next examined the role of Crk family adaptors in PDGF-stimulated cell migration, which is driven by reorganization of the actin cytoskeleton. Utilizing a wound-healing assay, we observed that the PDGF-stimulated increase in migration was significantly lower in CrkI/II- or CrkL-knockdown cells than in wt cells, despite no apparent reduction in basal motility (Fig. 1E,F). CrkI/II/L-knockdown cells displayed a complete loss of PDGF-enhanced cell migration as well as diminished basal motility. These results indicate that Crk family adaptors play a crucial role, not only in PDGF-stimulated cell motility, but also in regulation of cell movement in general. Taken together, our results indicate that Crk family adaptors are required for linking PDGFβR signaling to the actin cytoskeleton, and that in this role CrkI/II and CrkL are for the most part redundant.
PDGF stimulation induces downregulation of Crk adaptors by Abl family tyrosine kinases
PDGF stimulation of fibroblasts has been shown to induce tyrosine phosphorylation of CrkII and CrkL (Matsumoto et al., 2000; Sorokin et al., 1998). We next looked at which tyrosine kinase is responsible for this. The Abl family of tyrosine kinases, consisting of c-Abl and Arg, are known to phosphorylate CrkII and CrkL in various biological situations (Chodniewicz and Klemke, 2004). We therefore used RNA interference to test whether Abl kinases phosphorylate CrkII and CrkL upon PDGF stimulation. Simultaneous knockdown of Abl and Arg (each to 10-15% of controls) resulted in a significant decrease in PDGF-induced phosphorylation of CrkII and CrkL, whereas the knockdown of Abl or Arg alone had no apparent effect (Fig. 2A,B). Thus, phosphorylation of CrkII and CrkL induced by PDGF stimulation is chiefly controlled by Abl family kinases.
Previously, SH2 domains of Crk family adaptors have been shown to bind to the tyrosine-phosphorylated PDGFβR (Matsumoto et al., 2000). We re-examined this using an SH2-domain overlay assay (Nollau and Mayer, 2001; Rivera et al., 2006). The SH2 domains of the p85 PI-3 kinase subunit, which is known to interact with the activated PDGFβR (McGlade et al., 1992), bound strongly to the PDGFβR band co-migrating with the 182-kDa marker supplementary material Fig. S1A). By contrast, the SH2 domains of CrkI/II and CrkL bound preferentially to bands of approximately 120 and 70 kDa, probably corresponding to p130Cas and paxillin, respectively (Birge et al., 1993; Machida et al., 2007; Sakai et al., 1994). These interactions showed a biphasic pattern, characterized by a small increase in binding soon after PDGF stimulation followed by a rapid decrease to below the level of binding in unstimulated cells (supplementary material Fig. S1A).
Co-immunoprecipitation experiments also demonstrated a strong interaction of Crk with p130Cas but not with PDGFβR (Fig. 2C). We found that co-immunoprecipitation of p130Cas had decreased by more than 95% at 12 minutes after PDGF stimulation in both CrkII and CrkL immunoprecipitates (Fig. 2C, supplementary material Fig. S1B). This decrease is due to tyrosine phosphorylation of CrkII and CrkL, because when complex formation with p130Cas was examined for an unphosphorylatable form of CrkII (CrkII Y221F), no decrease in the interaction with p130Cas was observed after PDGF stimulation (supplementary material Fig. S1C). We also observed that the tyrosine phosphorylation of p130Cas decreased in concert with the dissociation of CrkII and CrkL after PDGF stimulation (Fig. 2C, supplementary material Fig. S1B). Presumably, p130Cas phosphotyrosine sites become exposed to phosphatases when the Crk SH2 domain disengages. Consistent with this, we have observed that overexpression of the Crk SH2 domain is sufficient to increase the tyrosine phosphorylation of p130Cas (our unpublished results).
Together, these results indicate that PDGF stimulation promotes the tyrosine phosphorylation of CrkII and CrkL, mediated by Abl family tyrosine kinases, leading to disruption of the Crk-p130Cas complex and decreased p130Cas phosphorylation. These results are consistent with a functional role for Crk family adaptors in PDGFβR signaling.
Crk family adaptors are required for FA formation and for PDGF-stimulated FA disassembly
Disassembly of FAs is important for cell motility (Ridley et al., 2003), and is observed after PDGFβR activation, in concert with the loss of actin stress fibers (Heldin et al., 1998; Ruusala et al., 2008). Crk family adaptors have been strongly implicated in FA formation (Antoku et al., 2008; Lamorte et al., 2003; Nievers et al., 1997). Therefore, we next examined the role of Crk in FA dynamics in PDGF-stimulated cells. FAs were visualized by immunofluorescent staining with antibodies to vinculin, and we observed that the prominent FAs seen in NIH-3T3 cells under serum starvation quickly disassembled following PDGF stimulation (Fig. 3A-C). We also consistently observed a decrease in pY397 Fak and to a lesser extent in pY118 paxillin, which are indirect biochemical indicators of FA formation, following PDGF stimulation (Fig. 2C, supplementary material Fig. S1B).
In contrast to wt cells, in CrkI/II/L-knockdown cells the number of FAs was drastically reduced in unstimulated cells (Fig. 3A). This indicates that Crk family adaptors are important for FA formation or maintenance in fibroblasts. Despite the much lower number of FAs present in CrkI/II/L-knockdown cells, quantitative image analysis (described in Materials and Methods) indicated that those FAs still disassembled in response to PDGF stimulation (Fig. 3C).
To explore the function of individual Crk domains and motifs in FA dynamics, we expressed RNAi-resistant Crk variants (fused to EYFP) in CrkI/II/L-knockdown cells (supplementary material Fig. S2A). In agreement with our result showing redundancy between endogenous CrkII and CrkL (Fig. 1), re-expression of either wt CrkII or CrkL was sufficient to rescue FA formation in CrkI/II/L-knockdown cells under serum starvation (Fig. 3A). When these cells were stimulated with PDGF, FAs disassembled (Fig. 3A,C) and the localization of CrkII and CrkL to FAs was lost (Fig. 3B, supplementary material Movies 1-4), as seen in wt cells. We also observed the disappearance of other FA proteins including paxillin, p130Cas and Nck2 from adhesion sites after PDGF stimulation (supplementary material Fig. S2C). We found that the SH2 and nSH3 domains of CrkII were required for restoring FA formation in unstimulated CrkI/II/L-knockdown cells (see CrkII SH2* or CrkII nSH3*, Fig. 3A,B). By contrast, the cSH3-domain mutant CrkII (CrkII cSH3*) rescued FA formation (Fig. 3A,B), indicating that the cSH3 domain is not required for this activity.
We previously showed that CrkII and CrkL are highly tyrosine-phosphorylated after PDGF stimulation, resulting in inactivation. Therefore, we next addressed the importance of Crk phosphorylation in PDGF-stimulated FA disassembly. The unphosphorylatable form of CrkII (CrkII Y221F), which cannot adopt the inactive conformation, rescued FA formation under serum starvation. However, in CrkII-Y221F-expressing cells, these FAs following PDGF stimulation persisted significantly longer than the ones from wt cells or CrkI/II/L-knockdown cells re-expressing wt CrkII (Fig. 3A-C). CrkI, the splicing variant of CrkII, cannot be regulated by tyrosine phosphorylation either and is unable to adopt the inactive conformation. Cells expressing CrkI showed a similar phenotype to those expressing CrkII Y221F, because FAs were formed under serum starvation and these FAs persisted longer after PDGF stimulation (Fig. 3A-C). Furthermore, EYFP fusion proteins of both CrkII Y221F and CrkI persisted in FAs longer than did wt CrkII after PDGF stimulation (Fig. 3B, supplementary material Movies 3, 5, 6). This is much more obvious in the movies of living cells than in fixed cells. Thus, the phosphorylation-dependent inactivation of CrkII and CrkL is important for the coordinated FA disassembly after PDGF stimulation
Tyrosine phosphorylation of CrkII and CrkL is required for `closure' of dorsal membrane ruffles induced by PDGF stimulation
Next, we extended the strategy of gene silencing combined with mutational analysis to study the Crk-dependent regulation of dorsal-membrane ruffle dynamics. Consistent with the functional compensation between endogenous CrkII and CrkL (Fig. 1), re-expression of either CrkII or CrkL rescued PDGF-induced formation of circular dorsal-membrane ruffles in CrkI/II/L-knockdown cells (Fig. 3A,D). In addition, re-expression of CrkI and CrkII cSH3* rescued circular dorsal-membrane ruffles after PDGF stimulation, indicating that the cSH3 domain is not required for this activity; however, neither SH2- or nSH3-domain mutants of CrkII could rescue formation of dorsal-membrane ruffles (Fig. 3A,D). Thus, the SH2 and nSH3 domains of Crk adaptors are strictly required for PDGF-stimulated actin rearrangements.
Cells expressing the unphosphorylatable form of CrkII (CrkII Y221F) exhibited cytoskeletal rearrangements in response to PDGF. However, actin puncta at the cell periphery (resembling the early stages of dorsal-ruffle formation) persisted for an extended period without progressing to the wave-like contraction and closure typical of circular dorsal ruffles (Fig. 3A,B, supplementary material Movie 6). This suggests that phosphorylation of CrkII and CrkL is important for the normal maturation of circular dorsal-membrane ruffles. Although CrkI cannot adopt the inactive conformation, cells re-expressing CrkI were similar to those expressing wt CrkII with respect to PDGF-stimulated formation of circular dorsal-membrane ruffles. The difference between CrkI and CrkII Y221F is the presence of the cSH3 domain. Therefore, we introduced a mutation in the cSH3 domain of CrkII Y221F, and its effects on the formation and structure of circular dorsal-membrane ruffles were examined. In CrkI/II/L-knockdown cells expressing this new mutant, CrkII Y221F cSH3*, circular dorsal ruffles with normal dynamics were observed (Fig. 3A,D). These results imply that although the cSH3 domain is not required for dorsal-membrane-ruffle formation, its presence in active (uninhibited) Crk prolongs actin cytoskeleton rearrangements and prevents normal maturation of circular dorsal ruffles.
We also examined the importance of Crk for biochemical signaling pathways downstream of PDGF stimulation. We found that in CrkI/II/L-knockdown cells, PDGF-dependent activation of the MAP kinases Erk1/2 was blunted, particularly at low levels of PDGF stimulation (supplementary material Fig. S3). By contrast there was no apparent effect of knockdown on PDGF-dependent activation of Akt. Thus, in addition to its crucial role in regulating FAs and actin rearrangements, Crk adaptors play a supporting role in regulating some pro-proliferative responses in PDGFβR signaling.
The cSH3 domain directs CrkII to activate Rac1 instead of Rap1
The nSH3 domain of Crk interacts with Dock180, a Rac1 GEF, and thus can promote an increase in active (GTP-bound) Rac1 (Hasegawa et al., 1996; Kiyokawa et al., 1998). Activation of Rac1 is thought to be required for PDGF-induced formation of dorsal ruffles (Ridley et al., 1992). Thus, we hypothesized that Crk family adaptors induce actin-cytoskeleton rearrangement through Rac1 activation in PDGFβR signaling. Indeed, using a pulldown assay, we could demonstrate that activation of Rac1 after PDGF stimulation was severely impaired in CrkI/II/L-knockdown cells (Fig. 4, supplementary material Fig. S3A). This indicates that Crk family adaptors are important for Rac1 activation in PDGFβR signaling.
To further analyze how Crk adaptors regulate Rac1 activity during PDGF stimulation, Rac1 activation was examined in CrkI/II/L-knockdown cells expressing wt or mutant Crk adaptors. In cells re-expressing either wt CrkII or CrkL, Rac1 was strongly activated 3 minutes after PDGF stimulation, and then returned nearly to basal levels at 15 minutes (Fig. 4). In cells expressing the CrkII SH2* or nSH3*mutants, which failed to rescue formation of PDGF-induced dorsal ruffles, Rac1 was only modestly activated 3 minutes after PDGF stimulation. Thus, CrkII appears to activate Rac1 by transducing signals through its SH2 and nSH3 domains to induce actin cytoskeleton rearrangements during PDGFβR signaling. The CrkII Y221F mutant, which induced `persistent' membrane ruffles (Fig. 3A, supplementary material Movie 6), also induced persistent Rac1 activation (Fig. 4). This suggests that inactivation of CrkII by tyrosine phosphorylation is required for termination of Rac1 signaling shortly after PDGF stimulation.
To our surprise, Rac1 activation was severely impaired in CrkI/II/L-knockdown cells expressing CrkI or CrkII cSH3*, both of which lack a functional cSH3 domain, although formation of dorsal ruffles proceeded normally (Fig. 4). As an alternative to Rac1, Crk adaptors can activate the Rap1 small GTPase through another nSH3-binding partner, C3G (Gotoh et al., 1995; Knudsen et al., 1994; Tanaka et al., 1994). Moreover, Rap1 is known to activate various downstream effectors including ARAP3, whose activation has been shown to promote membrane ruffles (Krugmann et al., 2006; Krugmann et al., 2004). Thus, we tested the hypothesis that CrkI and CrkII cSH3* induce membrane ruffles through Rap1 activation in a Rac1-independent manner. Strikingly, in knockdown cells expressing CrkI or CrkII cSH3*, Rap1 activation was four- to fivefold higher at 3 minutes after PDGF stimulation compared with approximately twofold in wt cells or in cells expressing CrkII mutants with a functional cSH3 domain (Fig. 4). These results suggest that the cSH3 domain enables Crk adaptors to selectively activate Rac1 instead of Rap1. Moreover, these results imply that in PDGFβR signaling, CrkI induces functional dorsal-membrane ruffles via Rap1 activation in the apparent absence of Rac1 activation.
In this study, the role of Crk family adaptors in PDGFβR signaling was investigated by a combination of gene silencing and mutational analysis. Our results demonstrate that Crk family adaptors play a funda mental role in FA formation and turnover in fibroblasts. We also demonstrated that Crk family adaptors are required for formation of circular dorsal-membrane ruffles and for promoting cell migration downstream of the PDGFβR. In this pathway, CrkII and CrkL showed overlapping functions and therefore dosage dependency. This redundancy among Crk family adaptors has been also observed in the integrin and reelin signaling pathways (Antoku et al., 2008; Matsuki et al., 2008; Park and Curran, 2008). In addition to positive roles for the Crk adaptors in the initiation of PDGFβR signaling, phosphorylation-mediated inhibition of CrkII and CrkL by Abl family tyrosine kinases is important for disassembly of FAs and for termination of reorganization of the actin cytoskeleton after PDGF stimulation. Furthermore, we uncovered a novel role of the cSH3 domain in directing Crk adaptors to activate Rac1 at the expense of Rap1 activation.
Regulation of FA dynamics by Crk adaptors
We have shown that Crk family adaptors are essential for normal FAs (Fig. 3A,B); the few FAs seen in Crk family knockdown cells are probably due to incomplete silencing of Crk expression. This observation is consistent with previous studies showing that overexpression of v-Crk or Crk adaptors promotes FA formation (Antoku et al., 2008; Lamorte et al., 2003; Nievers et al., 1997). We also previously showed that Crk family adaptors are required for FA formation by fibroblasts in the early stages of cell spreading onto fibronectin-coated surfaces (Antoku et al., 2008). After PDGF stimulation, however, FAs rapidly disassemble (Fig. 3A-C), in concert with the phosphorylation and functional inactivation of CrkII and CrkL by Abl family kinases (Fig. 2). By contrast, expression of an unphosphorylatable form of CrkII in CrkI/II/L-knockdown cells significantly slowed the disassembly of FAs stimulated by PDGF (Fig. 3A-C). Crk adaptors have been considered enhancers of FA formation (Feller, 2001; Nievers et al., 1997). Now, we reconsider this idea and propose that, rather than being merely enhancers, Crk adaptors are essential components of FAs and are crucial regulators of their assembly and disassembly. This model is supported by studies by us and others, showing that changes in tyrosine phosphorylation of Crk caused by modulating Abl activity tightly correlate with the number and size of FAs formed in cells (Antoku et al., 2008; Peacock et al., 2007).
Several lines of evidence suggest mechanisms whereby Crk adaptors could promote FA formation by mediating the assembly of protein complexes. Fibroblasts lacking the major Crk SH2-binding partner p130Cas exhibit slow assembly of FAs (Antoku et al., 2008; Honda et al., 1998). Fibroblasts lacking the Crk nSH3-binding partner C3G show severely reduced FAs (Voss et al., 2003). Moreover, formation of the p130Cas-Crk-C3G complex has been suggested to promote FA formation (Li et al., 2002). We show here that dissociation of p130Cas from CrkII and CrkL coincides with the disassembly of FAs that is induced by PDGF (Fig. 2C, Fig. 3A,C, supplementary material Fig. S1B,C). Alternatively, Crk adaptors interact with the paxillin-GIT2-β-PIX complex, which enhances formation of FAs (Lamorte et al., 2003).
One possible downstream effector of these complexes for regulation of FAs is the small GTPase RhoA. RhoA is a crucial regulator of FA dynamics, and its activity increases after PDGF stimulation, followed by a rapid decrease (Ruusala et al., 2008). ARAP3 is a GTPase activating protein (GAP) for RhoA and Arf6 (Krugmann et al., 2002). This GAP is regulated by C3G through another small GTPase, Rap1 (Gotoh et al., 1995; Krugmann et al., 2004). Thus, Crk complexes might ultimately regulate FA dynamics through a pathway that includes Rap1 and RhoA. However, we had technical difficulties quantifying RhoA activity by GTPase pulldown assay, and thus were unable to assess the involvement of RhoA in Crk-mediated FA dynamics. This remains to be clarified by future studies.
Regulation of actin reorganization by Crk adaptors
We showed that Crk adaptors are essential for the Rac1-dependent reorganization of the actin cytoskeleton that is induced by PDGF stimulation (Fig. 1B,C; Figs 3,4; supplementary material Fig. S3A). Interestingly, expression of CrkII Y221F in CrkI/II/L-knockdown cells resulted in prolonged dorsal ruffling and Rac1 activation (Fig. 3A,B; Fig. 4, supplementary material Movie 6). These results indicate that the functional inactivation of Crk adaptors (and thus downregulation of Rac1 activity) also plays an important role in coordinating rearrangement of the actin cytoskeleton. Furthermore, our data suggest that the activity of Crk adaptors is required only in the early period of PDGFβR signaling.
Previously, the SH2 domains of Crk adaptors have been shown to bind to tyrosine-phosphorylated PDGFβR (Matsumoto et al., 2000). However, in this study we did not detect specific interaction between Crk adaptors and PDGFβR using SH2-domain overlay assay and co-immunoprecipitation (Fig. 2C, supplementary material Fig. S1A). Instead, we observed strong binding of Crk adaptors to p130Cas (Fig. 2C, supplementary material Fig. S1B). p130Cas has 15 YxxP motifs, nine of which are YDxP sequences for which the Crk SH2 domain has strong affinity when phosphorylated (Shin et al., 2004). Thus, it is probable that Crk adaptors are recruited to the membrane after PDGF stimulation through binding of their SH2 domains to p130Cas. We have previously shown that the interaction of Nck family adaptors, via their SH2 domains, with p130Cas is essential for efficient dorsal-membrane ruffle induction by PDGF (Rivera et al., 2006). Like Crk SH2 domains, Nck SH2 domains have high affinity for YDxP motifs (Frese et al., 2006). Thus, p130Cas is likely to play a central role in PDGF-mediated actin reorganization by recruiting both Crk and Nck to the plasma membrane, where active cytoskeleton rearrangement takes place. During this step, downstream effector proteins bound to the Crk nSH3 domain are recruited to the complex. This is the key function for adaptor proteins, and explains why both SH2 and nSH3 domains are required for Rac1 activation and for rearrangement of the actin cytoskeleton during PDGF stimulation (Figs 3, 4). Nck is also required for the activation of Rac1 during PDGF signaling (Rivera et al., 2006), though its key downstream effectors are unlikely to be identical to those of Crk; possible candidates for Nck effectors include Abl kinases and WASP family promoters of actin nucleation (Antoku et al., 2008; Rivera et al., 2004).
A possible downstream effector for Crk adaptors is the Rac1 GEF Dock180, which is recruited to the plasma membrane after PDGF stimulation (Cote et al., 2005). In addition to the p130Cas-CrkII/L-Dock180 complex, the paxillin-CrkII-Dock180 complex has been shown to enhance cell migration through Rac1 activation (Valles et al., 2004). Thus, this complex might partly contribute to Rac1 activation in Crk-mediated PDGFβR signaling. Crk adaptors also bind another GEF, Sos, through their nSH3 domains (Feller et al., 1995). It is widely accepted that Sos is regulated by Grb2 and functions as a Ras GEF during growth-factor receptor signaling (Chardin et al., 1993; Nimnual and Bar-Sagi, 2002; Skolnik et al., 1993). However, Sos also has GEF activity toward Rac1 (Innocenti et al., 2002; Nimnual and Bar-Sagi, 2002), and therefore we cannot rule out the possibility that Sos contributes to Rac1 activation by Crk adaptors.
Unexpectedly, our mutational analysis also revealed that Crk adaptors lacking a functional cSH3 domain were unable to activate Rac1, but could still rescue formation of dorsal ruffles in response to PDGF (Figs 3, 4). We found that these forms of Crk strongly activated Rap1 instead of Rac1 during PDGF stimulation (Fig. 4). By contrast, Crk adaptors carrying a functional cSH3 domain did not dramatically increase Rap1 activity (Fig. 4). A previous study suggested that Rap1 activation itself results in Rac1 activation (Arthur et al., 2004). We did not detect Rap1-mediated Rac1 activation during PDGFβR signaling, though we cannot rule out localized increases in Rap1 activity in vivo. Fibroblasts lacking the Rap1 GEF C3G are deficient in formation of dorsal-membrane ruffles after PDGF stimulation (Voss et al., 2003), consistent with an important role for Rap1 in this process. Overexpression or knockdown of ARAP3, a downstream effector of Rap1, was shown to induce cytoskeletal changes including altered peripheral ruffles (Krugmann et al., 2006), suggesting it might play a role downstream of Rap1 in PDGF stimulated cells. It is interesting to note that sustained Rap1 activation (as induced by CrkI) is compatible with normal maturation of circular dorsal ruffles, whereas sustained Rac1 activation (as induced by CrkIIY221F) is not.
How might cSH3 direct activation of Rap1 instead of Rac1 in vivo? Biochemical and structural studies suggest two possible functions for the cSH3 domain of Crk adaptors. First, the cSH3 domain reduces overall affinity of the nSH3 domain towards its binding partners by lowering the accessibility of the binding pocket in the nSH3 domain (Kobashigawa et al., 2007; Sarkar et al., 2007). It is possible that this conformation favors binding of certain effectors, such as C3G, at the expense of others. Second, the cSH3 domain has been proposed to promote the homo-dimerization of Crk adaptors (Harkiolaki et al., 2006) and to promote nuclear export (Smith et al., 2002). Further studies will be required to resolve this issue.
The selective activation of Rap1 by CrkI raises an interesting point about the distinct functional role of CrkI relative to CrkII and CrkL. In cells, the expression level of CrkI is generally much lower than that of CrkII (Matsuda et al., 1992). However, unlike CrkII, CrkI signaling cannot be turned off and, as we have shown, CrkI preferentially activates Rap1 rather than Rac1. Thus, in resting cells CrkII will initially dominate signaling; however, when CrkII is locally inhibited by signal-induced phosphorylation, CrkI signaling will start to predominate at that site, albeit at lower levels. This probably serves as a mechanism for cells to switch from Rac1 to Rap1 signaling at specific sites, and might also play a role in shielding Crk SH2-binding proteins such as p130Cas and paxillin from total dephosphorylation.
Dynamic cell behaviors such as such as FA turnover and actin-cytoskeleton remodeling must be regulated precisely in time and space. For example, in order for a cell to move, adhesions must form at the leading edge, persist for a time and then disassemble. For a dorsal-membrane ruffle to form, contract and close, the zone of active actin polymerization must move in a wave-like fashion from the cell periphery towards the center. Crk adaptors are well suited to play a central role in such behaviors, because they can mediate the assembly of protein complexes in response to tyrosine phosphorylation and then be rapidly downregulated by their own phosphorylation. Presumably, engagement of the SH2 domain by phosphorylated targets not only recruits key downstream effectors, such as Dock180 and C3G, but also recruits the Abl family kinases that will locally shut down signaling after a short delay. Such delayed feedback loops are typical of biological systems with wave-like or oscillatory behavior (Tyson et al., 2003). Additional adaptors, such as those of the Nck family, can also recruit Abl kinases as well as different downstream effectors to these sites, and might further play a role both in downregulating the Crk signal and in switching effectors over time (Antoku et al., 2008; Rivera et al., 2006). The presence of low levels of unphosphorylatable CrkI further enriches the dynamic behavior of the system by providing a lower limit to activity, even after the most of the Crk signaling has been downregulated. Multiple species of Crk family adaptors, which differ in their activity and effectors, allow the cell to orchestrate precise spatiotemporal responses downstream of PDGFβR activation.
Materials and Methods
Antibodies directed against Abl (8E9, BD Pharmingen), CrkI/II (BD Transduction Laboratories), pY207-CrkL (Cell Signaling), CrkL (C-20, Santa Cruz Biotechnology), phosphotyrosine (P-Tyr-100, Cell Signaling Technology), vinculin (VIN-11-5, Sigma-Aldrich), pY397-Fak (44-624G, Invitrogen), Fak (C-20, Santa Cruz Biotechnology), pY118-paxillin (Cell Signaling), paxillin (BD Transduction Laboratories), p130Cas (BD Transduction Laboratories), Rac1 (23A8, Millipore), Rap1 (121, Santa Cruz Biotechnology), p-Erk1/2, pT202/pY204 (Cell Signaling), Erk2 (D-2, Santa Cruz Biotechnology), pS473-Akt (Cell Signaling), PDGFβR (958, Santa Cruz Biotechnology) EGFP/EYFP (FL, Santa Cruz Biotechnology) and actin (I-17, Santa Cruz Biotechnology) were purchased from commercial sources. Antibodies recognizing Arg (AR11) and pY221-CrkII were generous gifts from Peter Davies (Albert Einstein College of Medicine, Bronx, NY) and Michiyuki Matsuda (Kyoto University, Kyoto, Japan), respectively.
The short hairpin RNA (shRNA)-knockdown construct for CrkI/II/L was previously described (Antoku et al., 2008). The sequences for shRNA knockdown constructs targeting mouse CrkL (generous gifts from Nathaniel Allen and Jonathan Cooper, Fred Hutchinson Cancer Research Center, Seattle, WA), CrkI/II, Abl, Arg or Abl/Arg were; 5′-GGGCGAGCTTCTAGTGATAAT-3′ (CrkL); 5′-GTGGAGTGATTCTCAGGCA-3′ (CrkI/II); 5′-TGAGCTATGTGGACTCTAT-3′ (Abl); 5′-GGAGCCAAATTTCCTATTA-3′ (Arg); or 5′-GAGTACTTGGAGAAGAAGA-3′ (Abl/Arg), respectively. These shRNA sequences were inserted into the pSUPER.retro vector (Oligoengine). The cDNAs encoding human CrkI, CrkII, CrkL and Nck2; mouse paxillin; and rat p130Cas were previously described (Antoku et al., 2008; Rivera et al., 2006; Sharma et al., 2003). The cDNAs encoding EYFP and β-actin were from pEYFP-C1 actin (Clontech). The cDNAs encoding human CrkII Y221F; mCherry that was used for making pMSCV-puro (Clontech) mCherry-actin; and cerulean that was used for making pSUPER.retro.cerulean were generous gifts from Michiyuki Matsuda, Roger Tsien (UCSD, La Jolla, CA), and Dave Piston (Vanderbilt University, Nashville, TN) respectively. The shRNA-insensitive cDNAs were generated by changing nucleotides G381C, C387G, G540T, C549A for CrkI and CrkII; and G513T, T522A, C525G and T528G for CrkL. CrkII SH2*, CrkII nSH3* and CrkII cSH3* were generated by changing the amino acid sequence of shRNA-insensitive CrkII to R38L, W169L and W275L, respectively. All these mutations were generated by polymerase chain-reaction-based mutagenesis using Pfu DNA polymerase (Stratagene). All cDNAs were inserted into pMSCV-puro vectors containing N-terminal EYFP or pMIGR-1 vector containing N-terminal FLAG epitope (Antoku et al., 2008; Pear et al., 1998). pGEX4T-3 Ral-GDS GBD plasmid was obtained from Michiyuki Matsuda. pET-Pak1-PBD, pGEX-6P-1 Crk-SH2, pGEX-6P-1 CrkL-SH2 and pGEX-6P-1 PI3K-SH2 plasmids were previously described (Antoku et al., 2008; Machida et al., 2007).
Cell culture, transfection, viral infection and PDGF stimulation of cells
Maintenance of HEK293T and NIH-3T3 cells, virus production and viral infection were previously described (Antoku et al., 2008). For re-expression of Crk family adaptors in CrkI/II/L-knockdown cells, NIH-3T3 cells were first infected with pSUPER.retro.puro CrkI/II/L-derived virus. At 48 hours post-infection, the cells were replated. The next day, cells were super-infected with MSCV-puro-derived virus. At 24 hours post-super-infection, the cells were replated either on cell culture dishes or on coverslips coated with 0.1% (w/v) gelatin (Sigma-Aldrich) in cell-culture dishes, and serum-starved for further experiments. At 18-24 hours after serum starvation, the cells were stimulated with PDGF-bb (Millipore) for various times, and either harvested with lysis buffer or fixed for further experiments.
Cell lysis, immunoprecipitation, western blotting and far-western blotting were previously described (Antoku et al., 2008). Briefly, for immunoprecipitation, 0.35 μg anti-CrkI/II or anti-p130Cas, or 0.6 μg anti-CrkL antibodies were immobilized on Protein G- or Protein A-Sepharose (Pierce or GE Healthcare, respectively). The CrkI/II antibody was further chemically cross-linked to beads using disuccinimidyl suberate (DSS) (Pierce), following the manufacturer's instructions. The GTPase pulldown assay was performed as previously described (Azim et al., 2000). Immunoblot band intensities were quantified using ImageJ software (NIH, Bethesda, MD).
Immunofluorescent staining, assay of dorsal-membrane-ruffle formation and live-cell imaging
PDGF-stimulated cells were fixed on coverslips by adding one-tenth total volume of 10× PEM buffer (1 M PIPES pH 6.9, 10 mM MgCl2, 10 mM EGTA pH 8.0) and 37% (w/v) paraformaldehyde (Sigma-Aldrich) for 5 minutes, and permeabilized with phosphate-buffered saline (PBS) containing 0.5% (v/v) NP-40 for 5 minutes. The coverslips were blocked with PBS containing 1% (w/v) BSA for 1 hour at room temperature. Fixed Crk-knockdown cells were stained with phalloidin conjugated with Texas-red (Invitrogen), and Hoechst 33342 (Sigma-Aldrich). Alternatively, fixed Crk-knockdown cells expressing EYFP or EYFP fusion proteins were stained with anti-vinculin followed by goat-anti-mouse secondary antibody conjugated with Alexa Fluor 647 (Invitrogen), phalloidin conjugated with Texas-red, and Hoechst 33342. Photomicrographs were obtained on a Zeiss 510 laser-scanning inverted confocal microscope with 1.3 NA oil immersion objective lens. For assay of dorsal-membrane-ruffle formation the images of stained cells were examined for the number of cells exhibiting circular actin ring structures on the dorsal side and the total number of cells in each field. Each field contained approximately 15-25 cells. Percentages of cells with dorsal-membrane ruffles were calculated. For each sample, four to five fields were examined and the percentages were averaged. The quantification of FAs was calculated from images using NIH ImageJ software as the area occupied by vinculin staining divided by the total number of cells visualized by DNA and F-actin staining. For live-cell imaging, a coverslip was placed into an Attofluor chamber (Invitrogen) prior to PDGF stimulation. Before and after stimulation, cell images were taken using a Nikon TE2000 inverted spinning disc microscope using a Nikon Plan Fluor 40× 1.3 NA oil immersion objective lens.
After serum starvation, vertical and horizontal wounds were introduced using a P200 micropipet tip (USA Scientific) on a 35-mm tissue-culture dish whose surface was fully occupied by cells. The plate was then washed three times with PBS, and Dulbecco's modified Eagle's medium (DMEM) containing 0.1% (v/v) calf serum with or without PDGF was added. Immediately or 12 hours after placing in the medium, images for three locations of a marked wound were taken by a Nikon D50 digital SLR camera mounted on a Nikon Eclipse TS100 inverted phase contrast microscope with Nikon Plan Fluor 10× 0.3 NA objective lens. From the images, the number of cells that had migrated more than 200 μm into the wound were counted for each sample.
Statistical significance among samples for immunoblot quantification, wound healing assay and assay of dorsal-membrane-ruffle formation were determined by ANOVA followed by Tukey's test. Statistical significance of FA formation before and after PDGF stimulation was determined using the Student's t-test. All statistical values were obtained from three independent experiments.
We thank Gonzalo Rivera for critical reading of the manuscript; Nathaniel Allen and Jonathan Cooper for generously providing an unpublished reagent, the CrkL shRNA plasmid; Roger Tsien and Dave Piston for cDNAs of mCherry and Cerulean, respectively; Peter Davies for Arg antibody; and Michiyuki Matsuda for pY221-CrkII antibody, CrkII Y221F cDNA, and pGEX4T-3 Ral-GDS GBD plasmid. These studies were supported by grant CA82258 from the National Institutes of Health (to B.J.M.). Deposited in PMC for release after 12 months. This paper is dedicated by B.J.M. to the memory of Hidesaburo Hanafusa, deeply respected scientist and beloved mentor.