|Home | About | Journals | Submit | Contact Us | Français|
Activated Ras has been found in many types of cancer. However, the mechanism underlying Ras-promoted tumor metastasis remains unclear. We demonstrate here that activated Ras induces tyrosine dephosphorylation and inhibition of FAK mediated by the Ras downstream Fgd1-Cdc42-PAK1-MEK-ERK signaling cascade. ERK phosphorylates FAK S910 and recruits PIN1 and PTP-PEST, which co-localize with FAK at the lamellipodia of migrating cells. PIN1 binding and prolyl isomerization of FAK cause PTP-PEST to interact with and dephosphorylate FAK Y397. Inhibition of FAK mediated by this signal relay promotes Ras-induced cell migration, invasion, and metastasis. These findings uncover the importance of sequential modification of FAK—by serine phosphorylation, isomerization, and tyrosine dephosphorylation—in the regulation of FAK activity and, thereby, in Ras-related tumor metastasis.
Ras is a small guanosine triphosphate-binding protein that plays an important role in signal transduction pathways that influence cellular proliferation, apoptosis, cytoskeletal organization, and other important biological processes. The three cellular ras genes encode four highly homologous 21-kDa proteins: N-Ras, H-Ras, K-Ras4A, and K-Ras4B. The two K-Ras proteins, of which only K-Ras4B is ubiquitously expressed, are formed from alternative splicing of a single transcript differing only at the C terminus (Malumbres and Barbacid, 2003).
Active GTP-bound Ras interacts with a variety of downstream effector proteins, which preferentially interact with the GTP-loaded form of Ras. The best characterized effectors of Ras are Raf kinases and phosphatidylinositol 3-kinase (PI3-K) (Malumbres and Barbacid, 2003; Schubbert et al., 2007). Other Ras effector proteins include certain guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs)—such as the Ral exchange factor RalGDS, the Rac exchange factor Tiam1, and p120RasGAP in association with p190RhoGAP—that couple Ras to GTPases of the Ral or Rho family (Malumbres and Barbacid, 2003; Schubbert et al., 2007). Rho family members, including Rho, Rac, and Cdc42, have been shown to be downstream effectors of Ras (Hingorani and Tuveson, 2003). The differential activation of distinct effector proteins results in execution of discrete cellular functions.
Aberrant activation of Ras proteins has been implicated in virtually all aspects of the malignant phenotype of cancer cells, including cellular proliferation, transformation, invasion, and metastasis (Campbell and Der, 2004). Activating ras mutations occur in approximately 30% of human cancers (Schubbert et al., 2007). Ras can be activated in tumors by loss of GAPs, which is exemplified by Ras activation in response to loss of NF1 (Schubbert et al., 2007). In addition to being activated by ras and ras-related gene mutation, Ras signaling pathways are persistently activated in breast, ovarian, and stomach carcinomas and other cancers that have overexpression of growth factor receptor tyrosine (Tyr, Y) kinases, such as epidermal growth factor (EGF) receptor (EGFR) and ErbB2, or other Tyr kinases, such as Bcr-Abl (Downward, 2003; Wells, 2000). Although the involvement of Ras in tumor initiation is well studied and this protein has been shown to be an important mediator of tumor cell invasion and metastasis caused by activation of receptor Tyr kinases, considerably less is known about how Ras promotes tumor cell migration, invasion, and metastasis (Campbell and Der, 2004).
Cell migration is a highly coordinated process of leading edge protrusion, turnover of focal adhesions, generation of tractional forces, and tail retraction and detachment, which involves precise regulation of cell-cell adhesion and cell-to-extracellular matrix (ECM) adhesion (Ridley et al., 2003). Functional regulation of the molecules involved in cell adhesion signaling is a key process in tumor cell motility. FAK is a nonreceptor protein Tyr kinase that localizes at focal adhesions, which are specific regions of cells that make close contact with the ECM through transmembrane integrin molecules. FAK is associated with integrin within focal adhesions, and integrin activation by ECM ligands is associated with increased Tyr phosphorylation and kinase activity of FAK (Parsons, 2003).
Activated FAK, marked by autophosphorylation at Y397, phosphorylates several substrate proteins, recruits a number of SH2- and SH3-domain-containing proteins, including c-Src, and plays an important role in diverse cellular functions, including integrin-mediated cell adhesion and spreading, growth factor signaling, cell cycle progression, and cell survival. The binding of c-Src to FAK is proposed to disrupt the intramolecular interaction between the c-Src SH2 domain and the negative regulatory carboxy-terminal Y529. Activated Src, in turn, phosphorylates FAK and further enhances FAK activity, thereby forming a positive feedback loop leading to the activation of downstream signaling molecules, such as ERK and PI3-K/AKT. FAK has been demonstrated to be a positive regulator of cell migration in several cell systems, especially in nonmalignant cells (Hanks et al., 2003; Parsons, 2003). However, the role of FAK in cancer cell migration remains controversial, and new insights suggest that under some conditions, FAK actually inhibits cell migration (Schaller, 2004).
Protein phosphorylation initiates an array of signaling events, including conformational changes in protein kinases themselves and their substrates (Pawson and Scott, 2005). PIN1 [protein interacting with NIMA (never in mitosis A)-1], a member of the parvulin subfamily of peptidyl-prolyl cis/trans isomerases (PPIases), is the only member of that family that specifically recognizes phosphorylated serine (S) or threonine (T) in pS/TP-peptide sequences (Lu and Zhou, 2007). PIN1 catalyzes the cis/trans isomerization of peptidyl-prolyl peptide bonds, thereby regulating the conformation of substrates after their phosphorylation and subsequently controlling protein functions (Lu and Zhou, 2007).
In the study reported here, we investigated the molecular events that mediate Ras-regulated cell mobility. We demonstrated that activated Ras induces dephosphorylation of FAK at Y397 and inhibition of FAK mediated by the Ras-induced Fgd1-Cdc42-PAK1-MEK-ERK signaling cascade. Ras-induced FAK phosphorylation at S910, which is mediated by ERK, creates an interacting motif for PIN1 binding. This binding recruits a protein-Tyr phosphatase (PTP), PTP-PEST, which dephosphorylates FAK at Y397 and subsequently inhibits FAK activity. The inhibition of FAK promotes Ras-induced cell migration, invasion, and metastasis.
The level of Tyr phosphorylation of FAK, particularly of the major autophosphorylation residue Y397, regulates its kinase activity (Parsons, 2003). To investigate the regulation of FAK by Ras, we stably expressed v-H-Ras (H-Ras V12) in 3Y1 rat fibroblasts or NIH3T3 cells. Expression of v-H-Ras in both 3Y1 cells (Fig. 1A) and NIH3T3 cells (data not shown) resulted in reduced phosphorylation levels of both total Tyr and Y397 of FAK. FAK dephosphorylation was dependent on v-H-Ras expression levels in NIH3T3 cells expressing a tetracycline (Tet)-inducible v-H-Ras (Tet-off) (Fig. 1B).
To examine whether FAK Tyr dephosphorylation is an immediate response to Ras activation rather than a secondary effect, we conditionally activated Sos, a guanine exchange factor for Ras, by inducing its membrane localization. Treatment with FK1012, a chemical inducer of dimerization, for 10 min results in binding of FKBP12-Sos fusion protein to the membrane-bound myristoylated FKBP12s and subsequent activation of Sos and Ras (Holsinger et al., 1995), as evidenced by phosphorylation and activation of the downstream effector ERK1/2 MAP kinase (Fig. S1A). This rapid activation of Ras sufficiently dephosphorylated FAK at Y397 (Fig. S1A) without detectable alteration of cell morphology and focal adhesion, as demonstrated by staining for actin and the focal adhesion protein vinculin (Fig. S1B). These results strongly suggested that FAK dephosphorylation is an immediate effect induced by Ras activation. In addition, more than 50% reduction of Y397 phosphorylation (quantified by scanning densitometry) was observed in adeno-v-H-Ras virus-infected human tumor cells, including BT549 breast cancer cells and U87 glioblastoma multiforme (GBM) cells (Fig. 1C), which do not harbor H-Ras or K-Ras mutation (Delmas et al., 2002; Eckert et al., 2004). That H-Ras activation led to FAK inhibition was further supported by the finding that autophosphorylation activity was lower for FAK immunoprecipitated from 3Y1-v-H-Ras cells, as detected by an in vitro kinase assay, than for FAK immunoprecipitated from 3Y1 cells (Fig. 1D).
Like activation of v-H-Ras, activation of a temperature-sensitive mutant of v-K-Ras reduced the level of FAK phosphorylation at Y397 (Fig. 1E), indicating that H-Ras and K-Ras have a similar regulatory effect on FAK. Paxillin and p130Cas, which associate with and are phosphorylated by FAK, are both important components of focal adhesions (Hanks et al., 2003). Activation of H-Ras resulted in a reduction of phosphorylation levels of both paxillin and p130Cas (Fig. 1F), but not FAK-related Pro-rich Tyr kinase-2 (PYK2) (Fig. 1G), indicating that activation of Ras specifically induces dephosphorylation of FAK and FAK-associated paxillin and p130Cas.
To examine whether the effects of Ras on FAK observed in cultured cells would also be observed in Ras-induced tumors, we subcutaneously injected athymic nude mice with MK14 cells, which are mouse melanoma cells with doxycycline-inducible v-H-Ras. Similar to previously published results (Chin et al., 1999), withdrawal of doxycycline shrank the tumor size (Fig. 1H). The degree of tumor shrinkage was directly correlated with repression of v-H-Ras, as detected by immunohistologic and immunoblotting analyses (Fig. 1I and 1J). In contrast, levels of FAK phosphorylation at Y397 were inversely correlated with v-H-Ras expression and tumor size.
Activation of Ras leads to activation of various downstream effectors, including Raf, Ral, PI3-K, and members of the Rho family of GTPases Rho, Rac, and Cdc42 (Hingorani and Tuveson, 2003). Expression of dominant-negative Raf 301-1 and Ral N28 mutants or a kinase-dead p110α subunit of PI3-K did not block Ras-induced FAK dephosphorylation at Y397 (Fig. 2A). These results indicated that Raf, Ral, and PI3-K are not involved in Ras-induced FAK Tyr dephosphorylation. In agreement with these results, expression of activated v-Raf, RalA Q72L, or the p110α catalytic subunit of PI3-K failed to induce efficient FAK Tyr dephosphorylation (Fig. S2A).
To examine the potential role of Rho family members in regulation of FAK, constitutively active mutants of Rac1 V12, Cdc42 V12, and RhoA V14 were transiently expressed in 293T human embryonic kidney cells. FAK dephosphorylation at Y397 was induced by the expression of activated Cdc42, but not activated Rac1 or RhoA (Fig. 2B), which increased the total activity of Rac1 or RhoA as measured by a GST pull-down assay using GST fusion proteins derived from the Rac/Cdc42-binding domain (PBD) of PAK1 or rhotekin RhoA-binding domain (RBD), respectively (Fig. S2B). Consistent with these findings, after induction of v-H-Ras expression in NIH3T3-v-H-Ras (Tet-off) cells, an increased level of active GTP-bound Cdc42 was detected (Fig. 2C), indicating that activated Ras results in Cdc42 activation. Furthermore, stable expression of a dominant-negative Cdc42 N17 mutant in both 3Y1-v-H-Ras (Fig. 2D) and NIH3T3-v-H-Ras cells (data not shown) or depletion of Cdc42 by expression of its shRNA largely blocked v-H-Ras-induced FAK dephosphorylation at Y397 (Fig. 2D). These data indicated that Ras-induced FAK dephosphorylation is mediated by Cdc42, but not by other Ras downstream effectors, such as Raf, Ral, PI3-K, Rac, or Rho.
To understand how activated Ras induced Cdc42-dependent FAK Tyr dephosphorylation, we expressed several GEFs, which are known to be responsible for the activation of Rho-family GTPases (Rossman et al., 2005). Expression of active mutant of faciogenital dysplasia protein Fgd1, but not Dbl or Asef, led to FAK dephosphorylation at Y397 (Fig. 2E). These results are consistent with the previous report that Fgd1 is a specific GEF for Cdc42 activation (Ayala et al., 2009; Olson et al., 1996). In addition, depletion of Fgd1, but not Asef, by expressing shRNA specific against Fgd1 or Asef blocked Ras-induced Cdc42 activation (Fig. S2C; data of Asef not shown) and abrogated the effect of Ras on FAK dephosphorylation (Fig. 2F). Although it remains unclear whether Ras activates Fgd1-Cdc42 by direct interaction with Fgd1 or by recruiting Fgd1 to Cdc42 in a specific cellular compartment, these results indicated that Fgd1 mediates Ras-induced Cdc42 activation and thereby Tyr dephosphorylation of FAK.
PAR6-protein kinase C ζ (PKCζ) and PAK1 are known downstream effectors of Cdc42 activation (Erickson and Cerione, 2001). To understand how Cdc42 regulates FAK Tyr dephosphorylation, constitutively active or dominant-negative kinase-dead mutants of PKCζ or Myc-tagged PAK1 were transiently transfected into 293T cells. The expression of constitutively active PAK1 T423E, but not the expression of dominant-negative PAK1 K299R, constitutively active PKCζ CA, or dominant-negative PKCζ DN, resulted in FAK dephosphorylation at Y397 (Fig. 2G). In addition, FAK Tyr dephosphorylation in 3Y1-v-H-Ras cells was largely abrogated by stable expression of dominant-negative PAK1 K299R (Fig. 2H), depletion of PAK1 by its shRNA (Fig. S2D), or treatment of such cells with a PAK inhibitor, PAK18 (Zhao et al., 2006), but not PAK18 control peptide (Fig. 2H). Similarly, co-expression of a constitutively active Cdc42 mutant with PAK1 K299R in 293T cells or treatment of such cells with PAK18 significantly blocked activated Cdc42-induced FAK dephosphorylation at Y397 (Fig. 2I). These results demonstrated that PAK1 is a downstream regulator of Ras-Cdc42 for FAK dephosphorylation at Y397.
Ras-induced FAK dephosphorylation and inhibition could result from at least two different mechanisms: inhibition of an upstream signaling molecule that activates FAK, or dephosphorylation of FAK by a PTP. To examine whether FAK dephosphorylation at Y397 is PTP dependent, we treated NIH3T3-v-H-Ras cells with pervanadate, a general PTP inhibitor. As shown in Fig. 3A, pretreatment with pervanadate rescued v-H-Ras-induced FAK dephosphorylation at Y397 in a dose-dependent manner. Several PTPs, including SHP1, SHP2, RPTPα, and PTP-PEST, have been shown to be involved in regulation of focal contact and cell migration (Angers-Loustau et al., 1999; Kumar et al., 1999; Manes et al., 1999; Zeng et al., 2003). Immunoprecipitated PTP-PEST, SHP1, SHP2, and RPTPα all demonstrated enzymatic activity toward 3,6-fluorescein diphosphate (Fig. S3A). Nevertheless, overexpression of SHP1, SHP2, or RPTPα did not dramatically affect FAK dephosphorylation at Y397 (Fig. 3B), whereas RPTPα and SHP2 efficiently dephosphorylated their substrates—c-Src at Y529 (Fig. S3B) and Gab2 at Y452 (Fig. S3C) (den Hertog et al., 1993; Gu et al., 1997), respectively. Overexpression of wild-type (WT) PTP-PEST moderately enhanced Src phosphorylation at Y529 (Fig. S3B), greatly reduced FAK phosphorylation at Y397 (Fig. 3B), and had minimal effect on phosphorylation of Gab2 (Fig. S3C); in contrast, overexpression of the catalytically inactive mutants PTP-PEST C231S and PTP-PEST D199A did not greatly reduce FAK phosphorylation at Y397 (Fig. S3D) (Garton et al., 1996). Furthermore, immunoblotting of immunoprecipitated PTP-PEST with an anti-FAK antibody showed a greater association between endogenous PTP-PEST and FAK in NIH3T3-v-H-Ras cells than in NIH3T3 parental cells (Fig. 3C, left panel). Enhanced binding of FAK to PTP-PEST in v-H-Ras-transfected 293T cells in contrast to 293T cell without v-H-Ras expression was also detected by a reciprocal immunoprecipitation (Fig. 3C, right panel). In addition, an in vitro phosphatase assay demonstrated that immunoprecipitated FLAG-tagged PTP-PEST dephosphorylated FAK at Y397, but did not dephosphorylate phosphorylated c-Src (Fig. 3D). These results indicated that whereas all examined phosphatases are active in vivo, PTP-PEST has specificity for FAK.
To examine whether PTP-PEST is required for Ras-induced FAK dephosphorylation at Y397, we stably or transiently transfected FLAG-tagged catalytically inactive PTP-PEST mutants into 3Y1-v-H-Ras or 293T cells. Expression of these mutants largely abolished FAK dephosphorylation at Y397 induced by activated H-Ras (Fig. 3E), Cdc42, or PAK1 (Fig. 3F). In addition, depletion of PTP-PEST by expression of its shRNA (Fig. 3G) inhibited FAK Tyr dephosphorylation resulting from expression of inducible v-H-Ras (Fig. 3H). Consistent with this finding, Ras-induced FAK dephosphorylation at Y397 was partially inhibited in PTP-PEST−/− cells (Fig. 3I), in contrast to PTP-PEST−/− cells with re-constituted expression of PTP-PEST (Fig. S3E).
To understand how the activated Ras-Cdc42-PAK1 cascade regulated FAK dephosphorylation at Y397, we analyzed proteins co-immunoprecipitated with FAK from 293T cells expressing v-H-Ras by mass spectrometry. The protein pool eluted from the immunoprecipitate with an anti-FAK antibody was digested with trypsin and analyzed by nano-LC-MS/MS on a linear ion-trap mass spectrometer. One of the proteins tentatively identified was PIN1, a peptidyl-prolyl cis/trans isomerase (Fig. S4A). Immunoblotting of immunoprecipitated FAK from NIH3T3-v-H-Ras (Tet-off) cells with an anti-PIN1 antibody revealed enhanced PIN1 binding to FAK in a v-H-Ras expression-dependent manner (Fig. 4A). Immunofluorescent analysis of 3Y1-v-H-Ras cells displayed a strong co-localization of FAK with PIN1 at lamellipodia, which are a characteristic feature at the front, leading edge of motile cells (Fig. 4B) (Quantification of protein co-localization is shown in Fig. S4B). Additional evidence that FAK might act as a substrate of PIN1 was obtained by GST-PIN1 pull-down assay. GST-PIN1 associated much more with FAK from 3Y1-v-H-Ras cells (Fig. 4C) and NIH3T3-v-H-Ras cells (data not shown) than with FAK from their nontransformed parental cells. In contrast, a GST-PIN1 WW domain mutant with substitutions at W11A, W34A, R14A, and R17A, which prevent its binding to a phospho-Ser or phospho-Thr substrate (Verdecia et al., 2000), significantly reduced PIN1 binding to FAK (Fig. 4C). Consistent with the increased association of PTP-PEST with FAK induced by v-H-Ras (Fig. 3C), the amount of PTP-PEST in FAK complex pulled down by GST-PIN1 WT was greater in 3Y1-v-H-Ras cells than in 3Y1 cells (Fig. 4C). To examine the role of PIN1 in PTP-PEST binding to FAK, we incubated purified His-PTP-PEST with immunoprecipitated FAK from v-H-Ras-expressing PIN1+/+ fibroblasts or PIN1−/− fibroblasts. Immunoblotting with a PTP-PEST antibody showed that PIN1 deficiency abrogated the binding of PTP-PEST to FAK, which could be rescued by inclusion of purified GST-PIN1 WT, but not a catalytically inactive mutant GST-PIN1 C133A (Fig. 4D). These results indicated that PIN1 is required for PTP-PEST to bind to FAK.
A number of phosphorylated Ser/Thr residues of FAK have been identified, and phosphorylation of these residues has been implicated in regulation of FAK function (Bianchi et al., 2005; Grigera et al., 2005; Hunger-Glaser et al., 2004; Xie et al., 2003). Analysis of FAK amino acid sequence revealed that FAK has several putative PIN1-binding Ser-Pro motifs at S722, S732, and S910, which are known sites of phosphorylation by GSK, Cdk5, and ERK, respectively (Bianchi et al., 2005; Hunger-Glaser et al., 2004; Villa-Moruzzi, 2007; Xie et al., 2003). Mutation of S910 but not S722 or S732 largely abrogated FAK association with PIN1 (Fig. 4E), indicating that phosphorylation at S910 is required for FAK interaction with PIN1. To examine the role of S910 phosphorylation in PTP-PEST binding to FAK, we incubated immunoprecipitated FLAG-FAK WT or FLAG-FAK S910A from 3Y1-v-H-Ras cells with purified His-PTP-PEST, in the absence or presence of purified GST-PIN1 WT or GST-PIN1 C133A mutant. Immunoblotting with a PTP-PEST antibody showed that PIN1 WT, but not its inactive mutant, promoted the binding of PTP-PEST to FAK WT but not FAK S910A (Fig. 4F), indicating that FAK S910 phosphorylation is required for PIN1-mediated PTP-PEST binding to FAK. Furthermore, immunofluorescent analysis detected co-localizations between FAK phosphorylated at S910 and PTP-PEST (Fig. 4G upper panel, Fig. S4C left panel) and between PIN1 and PTP-PEST (Fig. 4G lower panel, Fig. S4C right panel), but not between FAK S910A and PTP-PEST or PIN1 (Fig. S4D), at the lamellipodia of 3Y1-v-H-Ras cells. These results suggested that FAK phosphorylation at S910 is important for formation of a complex among FAK, PIN1, and PTP-PEST at lamellipodia.
In line with the finding of more FAK binding to PIN1 in v-H-Ras-transformed cells than in parental cells (Fig. 4A), FAK was highly phosphorylated at S910 in 3Y1-v-H-Ras cells or v-H-Ras-expressed BT549 or U87 cells compared to parental 3Y1, BT549, or U87 cells (Fig. 4H). Mutation of S910, but not S722 or S732, made FAK resistant to Ras-induced FAK dephosphorylation at Y397 (Fig. 4I), indicating that FAK phosphorylation at S910 is important for recruitment of PIN1 and PTP-PEST for subsequent FAK dephosphorylation at Y397.
ERK has been implicated in FAK phosphorylation at S910 (Hunger-Glaser et al., 2004). Inhibition of ERK by treatment of NIH3T3-v-H-Ras cells with the MEK inhibitors U0126 (Fig. 5A) and PD98059 (data not shown), which inhibited v-H-Ras-induced ERK activity as indicated by lowered ERK phosphorylation levels, abolished Ras-induced FAK phosphorylation at S910 and rescued Ras-induced FAK dephosphorylation at Y397. In addition, overexpression of constitutively active ERK2–MEK1 fusion protein, but not its kinase-dead mutant (KR-ERK2–MEK1), induced FAK phosphorylation at S910 and dephosphorylation at Y397 (Fig. 5B), whereas FAK S910A mutant was resistant to active ERK- and PTP-PEST-induced dephosphorylation of FAK Y397 (Fig. S5A). In agreement with an important role of ERK in regulation of FAK, GST-PIN1 pull-down assay revealed that inhibition of ERK activation significantly reduced the binding of FAK to PIN1 and PTP-PEST (Fig. 5C). In addition, immunoprecipitated FAK from U0126-treated 3Y1-v-H-Ras cells, which was incubated with purified GST-PIN1 and His-PTP-PEST, lost its ability to bind to PTP-PEST, in contrast to the FAK immunoprecipitated from untreated 3Y1-v-H-Ras cells (Fig. S5B). However, incubation of immunoprecipitated FAK from U0126-treated 3Y1-v-H-Ras cells with purified active ERK2 restored its ability to bind to PTP-PEST (Fig. S5B), implying an essential role of ERK-dependent phosphorylation for FAK binding to PTP-PEST. Furthermore, expression of dominant-negative ERK2 K52R mutant, which is catalytically inactive, reduced activated Ras-, Cdc42-, and PAK1-induced FAK dephosphorylation at Y397 (Fig. 5D). These results demonstrated that the activated Ras-initiated Cdc42-PAK1-MEK-ERK signaling cascade leads to FAK phosphorylation at S910 and subsequent recruitment of PIN1 and PTP-PEST for FAK dephosphorylation at Y397.
To examine the requirement of PIN1 isomerase activity for FAK dephosphorylation at Y397, we transiently transfected v-H-Ras, activated Cdc42, and activated PAK1 with two catalytically inactive mutants, PIN1 C113A and PIN1 R68/69A, into 293T cells. Expression of either catalytic mutant abrogated Ras-, Cdc42-, or PAK1-induced FAK dephosphorylation at Y397 (Fig. 5E). In line with this observation, activated Ras failed to induce Y397 dephosphorylation in PIN1−/− cells, which could be rescued by reconstitutive expression of PIN1 WT, but not Pin1 R68/69A mutant (Fig. 5F, G). In addition, an in vitro phosphatase assay demonstrated that PTP-PEST dephosphorylated immunoprecipitated FAK at Y397 from PIN1+/+ cells, but not from PIN1−/− cells (Fig. 5H). These results strongly suggested that FAK is a substrate of PIN1 and indicated that isomerase activity of PIN1 is required for Ras-induced FAK dephosphorylation at Y397.
Consistent with our observation that ERK inhibition abrogated Ras-induced FAK dephosphorylation at Y397 (Fig. 5A, D), treatment of NIH3T3-v-H-Ras cells with U0126 converted less-adherent, spherical, and spindle-shaped cells into adherent, well-spread, and flat cells that resembled parental nontransformed NIH3T3 cells (Fig. 6A). In a monolayer wound-healing assay, NIH3T3-v-H-Ras cells treated with U0126 (Fig. 6B) or expressing the ERK2 K52R mutant (data not shown) migrated into the wound much more slowly than did untreated cells (Quantification of cell migration is shown in Fig. S6A). Similarly, PIN1 deficiency (Fig. 6C, S6A) or inhibition of PTP-PEST activity by stable expression of PTP-PEST C231S mutant (Fig. 6F, S6A) reduced the migration rate of Ras-transformed cells, whereas the migration defect induced by PIN1 deficiency was restored by reconstitutive expression of PIN1 WT but not PIN1 R68/69A (Fig. 6C, S6A). Furthermore, inhibition of FAK by stable expression of dominant-negative FAK Y397F (data not shown) or FRNK (the C-terminal domain of FAK), as evidenced by reduced phosphorylation of the downstream substrate paxillin (Fig. 6D, E), promoted Ras-induced cell migration (Fig. 6F, S6A). In contrast, expression of Tyr dephosphorylation-resistant FAK S910A largely inhibited Ras-induced cell migration (Fig. 6D, F, S6A). We also observed reduced cell migration with expression of dominant-negative Cdc42 N17 (data not shown), which further supported a role of FAK inhibition in promotion of Ras-induced cell migration.
To examine whether the downregulation of FAK activity contributed to Ras-induced cell invasion, we performed a collagen-gel invasion assay. Whereas PIN1+/+ cells expressing v-H-Ras displayed less-adherent transformed morphology on the surface of collagen gel, v-H-Ras-expressed PIN1−/− cells, like cells with ERK inhibition (Fig. 6A), were well-adherent and much less transformed (Fig. 6G, top panel). Five days after seeding, more PIN1+/+ cells expressing v-H-Ras than PIN1−/− cells expressing v-H-Ras were able to penetrate into collagen gel (Fig. 6G, bottom panel). The invasion defect of PIN1−/− cells was rescued by reconstitutive expression of PIN1 WT, but not PIN1 R68/69A (Fig. 6G, graph). Given that we did not observe an obvious cell growth difference among these cell lines (Fig. S6B), the differences in cell invasion were most likely not caused by different cell proliferation rates. In line with the effects of PTP-PEST, FAK inhibition, and expression of FAK S910A on Ras-induced cell migration, 3Y1-v-H-Ras cells stably expressing PTP-PEST C231S, PTP-PEST D199A, or FAK S910A were less invasive than their parental cells, whereas cells expressing dominant-negative FAK Y397F or FRNK were much more invasive than their parental cells (Fig. 6H). Similar results were observed with a Matrigel invasion–transwell assay in which cells were plated for 12 h, during which time we did not detect any significant difference in cell growth among different cell lines with or without treatment (data not shown). As shown in Fig. S6C and S6D, inhibition of ERK by U0126 largely blocked v-H-Ras-induced cell migration and invasion, whereas the expression of FRNK or FAK Y397F significantly enhanced cell migration and invasion. These results indicated that FAK Y397 dephosphorylation mediated by ERK, PIN1, and PTP-PEST promotes Ras-induced cell migration and invasion.
To test whether tumor cells with downregulated FAK can potentiate tumor metastasis, we injected 3Y1-v-H-Ras cells or 3Y1-v-H-Ras cells expressing FAK Y397F, FRNK, or FAK S910A into the tail vein of athymic mice. Dissection of mice 3 weeks later revealed more metastatic nodules in the lungs of the mice injected with 3Y1-v-H-Ras cells expressing FAK Y397F or FRNK (Fig. 7A). In contrast, expression of Tyr dephosphorylation-resistant FAK S910A largely inhibited Ras-induced metastasis (Fig. 7A).
Activated Ras correlated with ERK activation has been detected in human GBM specimens (Feldkamp et al., 1999). We demonstrated that ERK activation is necessary and sufficient for Ras-induced FAK dephosphorylation at Y397 (Fig. 5B, D). To determine whether our findings had clinical relevance, we examined ERK activity and FAK phosphorylation at Y397 in serial sections of 41 human primary GBM specimens (grade 4) by immunohistochemical analyses. As shown in Fig. S7, the levels of ERK activity were inversely correlated with the levels of FAK phosphorylation at Y397. Quantification of the staining on a scale of 0 to 8 showed that this inverse correlation was significant (Fig. S7: r = −0.81, P < 0.0001). Moreover, a similar inverse correlation was detected in different regions of the same specimen (data not shown). This clinical evidence supported the role of ERK activation in dephosphorylation of FAK at Y397 in human GBM.
Ras has been found to be either mutated or activated in many types of human cancer and is recognized as a driving force for tumor development in mouse models (Campbell and Der, 2004; Chin et al., 1999). Notwithstanding, how Ras regulates focal contacts, thereby promoting tumor cell mobility, remains an enigma. We found that activated Ras induced both Tyr dephosphorylation and inhibition of FAK, which resulted in increased cell invasion and metastasis. We observed that the levels of FAK dephosphorylation induced by Ras varied in different cell types. This could be due to differences in cellular context between mesenchymal and epithelial cells and between normal and cancer cells. Indeed, in contrast to what we observed in nontumorigenic epithelial MCF10A cells, we detected a much reduced FAK Y397 phosphorylation level correlating with an enhanced phosphorylation of ERK1/2 and FAK S910 in BT549 breast cancer cells (Fig. S8), which may limit the effect of Ras on FAK Tyr dephosphorylation.
Ras-induced FAK dephosphorylation at Y397 required Fgd1 and Cdc42, but not other Ras downstream effectors, such as Raf, PI3-K, Ral, Rho, or Rac. Cdc42 promotes PAK1-mediated phosphorylation of MEK1, but not MEK2, which in turn activates MEK1-ERK kinases (Beeser et al., 2005; Park et al., 2007). Raf and PAK have been shown to regulate ERK via distinct mechanisms. In addition to regulating ERK by PAK-dependent Raf1 S388 phosphorylation (King et al., 1998), PAK directly phosphorylates MEK1 S298, which induces MEK1 autophosphorylation of S218/222 and increases kinase activity towards ERK (Park et al., 2007). In contrast, Raf phosphorylates S218/222 and induces MEK1 activation in a S298 phosphorylation- or MEK1 catalytic activity-independent manner (Park et al., 2007). Although integrin-dependent MEK1 S298 phosphorylation by PAK1 promotes growth factor receptor-Raf-induced full activation of MEK1, Raf is not involved in adhesion-dependent MEK1 S298 phosphorylation and activation by PAK1 (Slack-Davis et al., 2003). Inhibition of ERK (Fig. 5A and 5D), but not of Raf (Fig. 2A), blocked both Ras- and PAK1-induced FAK dephosphorylation at Y397, implying that ERK activation induced by PAK1, but not by Raf, plays a role in inhibition of FAK. Thus, PAK1 might stimulate the distinctly localized MEK1-ERK complexes, leading to spatially restricted activation of ERK, which is important for cell motility. This assumption is supported by our results revealing localized FAK phosphorylation at S910 at lamellipodia, which are migratory organelles and believed to be the actual motor that pulls the cell forward during the process of cell migration (Vignjevic and Montagnac, 2007). This observation is consistent with detection of both activated Cdc42 and PAK1 at lamellipodia and focal adhesions (Kurokawa et al., 2004; Sells et al., 2000), suggesting a spatially regulated Cdc42-PAK1-MEK1/ERK signal relay in serine phosphorylation of FAK.
Although how FAK inhibition promotes cytoskeletal changes at lamellipodia still needs to be defined, inactivation of FAK primed by ERK-dependent S910 phosphorylation might be required for turnover of focal adhesions at the front, moving edges of cells, which promotes cell migration. This possibility was supported by the finding that inhibition of ERK by U0126, which blocked Ras-induced FAK phosphorylation at S910 and dephosphorylation at Y397, largely inhibited Ras-induced cell migration and invasion, which was accompanied by an increase in cell adhesion and abrogation of transformed morphology.
SHP2 activity is required for FAK Tyr dephosphorylation in MCF7 cells in response to insulin-like growth factor (IGF)-1 signaling, which promotes cell invasion in an MAP kinase-independent manner (Manes et al., 1999). In contrast, we observed a moderate FAK Tyr dephosphorylation by expression of SHP2 or SHP1, which implies that FAK phosphorylation can be regulated by different phosphatases in response to distinct cell signaling in different cellular contexts. Overexpression of PTP-PEST, which associates with paxillin and p130Cas, was shown to dephosphorylate FAK and p130Cas (Davidson and Veillette, 2001; Garton et al., 1997), which is in agreement with our finding that inhibition of PTP-PEST blocked Ras-induced FAK Tyr dephosphorylation. Intriguingly, expression of PTP-PEST moderately enhanced c-Src phosphorylation at Y529 (Fig. S3B). Given that activated Src phosphorylates FAK and enhances FAK activity (Parsons, 2003) and that PTP-PEST cooperates with Csk to inactivate Src through its capacity to dephosphorylate the positive regulatory Tyr of Src family kinases (Cloutier and Veillette, 1999), PTP-PEST-induced reduction of FAK phosphorylation at Y397 may be a combined effect from direct FAK dephosphorylation and inhibition of Src activity (Parsons, 2003). The fact that Ras-mediated FAK dephosphorylation at Y397 was largely blocked in PTP-PEST−/− cells provides additional evidence supporting a pivotal role of PTP-PEST in regulation of focal adhesion proteins (Shen et al., 1998). In agreement with these results, PTP-PEST−/− cells spread faster on fibronectin than did their WT counterparts and had more focal adhesions, enhanced paxillin phosphorylation, and decreased motility on fibronectin (Angers-Loustau et al., 1999). Notably, deficiency of PTP-PEST, a PEST family member, blocked Ras-induced FAK Tyr dephosphorylation less significantly than PTP-PEST dominant-negative mutants did. Given that PTP-PEST, unlike its relatives PEP and PTP-HSCF, is expressed ubiquitously (Davidson and Veillette, 2001), PTP-PEST deficiency may cause a compensatory upregulation of the expression of other PEST family members or a compensatory effect from other non-PEST family phosphatases, which contributes to FAK Tyr dephosphorylation.
Like PTP-PEST deficiency, lack of PIN1 resulted in inhibition of Ras-induced FAK dephosphorylation at Y397 and cell migration and invasion. These results indicated that PIN1 acts as a positive regulator in Ras-related cell mobility by moderating FAK function, which is consistent with a positive role of PIN1 in tumor development, evidenced by the fact that inhibition of PIN1 reduced HER2- and Ras-induced transforming phenotypes (Lu and Zhou, 2007). Clinical evidence also showed PIN1 overexpression in a number of human cancers, and double PIN1/p53-knockout mice do not display the tumor formation induced by deficiency of p53 alone (Lu and Zhou, 2007; Takahashi et al., 2007).
FAK has been demonstrated to be a positive regulator of nonmalignant cell migration and cell survival (Hanks et al., 2003; Parsons, 2003). Nevertheless, paradoxical evidence has been reported regarding the role of FAK in tumor metastasis. Increased FAK expression has been detected in several types of human cancer (McLean et al., 2005). FAK overexpression in biopsy tissue of primary ovarian cancer was correlated with metastasis to lymph nodes and distant organs (Sood et al., 2004). In contrast, FAK expression was reduced in metastatic liver tumors compared to their matched primary human colorectal adenocarcinoma samples (Ayaki et al., 2001). In patients with cervical cancer or intrahepatic cholangiocarcinoma, lower levels of FAK are associated with higher rates of metastasis and significantly poorer overall survival (Gabriel et al., 2006; Ohta et al., 2006). In addition, a melanoma cell line derived from peripheral blood, but not five other melanoma cell lines isolated from solid metastases from the same patient, had lost FAK expression (Maung et al., 1999). Thus, FAK expression can be dynamically regulated, and FAK might have different roles in different tumors and at different stages of tumor progression.
Correlating with the clinical evidence that high FAK expression relates to tumor progression, reduced FAK expression decreased the expression of matrix metalloproteinases (McLean et al., 2005) and the lung metastasis rate of murine mammary gland carcinoma by inhibiting urokinase plasminogen activator expression (Mitra et al., 2006). In addition, conditional fak deletion in the epidermis suppressed chemically induced skin tumor formation, and this effect was linked to increased cell apoptosis. Nevertheless, fak deletion did not have an effect on wound re-epithelialization of mouse skin (McLean et al., 2004). In sharp contrast to the observation of a positive role of FAK in cell motility, new insights suggest that under some conditions, FAK inhibits cell migration (Schaller, 2004). Previously, we and other groups showed that activation of ErbB family members EGFR, ErbB2, ErbB3, and ErbB4, as well as IGF-1 receptor, induces Tyr dephosphorylation of FAK, p130Cas, and paxillin; inactivation of FAK; and increased tumor cell motility, invasion, and metastasis (Bose et al., 2006; Caceres et al., 2005; Guo et al., 2007; Guvakova and Surmacz, 1999; Lu et al., 2001; Manes et al., 1999; Vadlamudi et al., 2002). FAK has also been shown to be able to attenuate v-Src-induced oncogenic transformation, and FAK deficiency promotes v-Src-induced cell migration (Moissoglu and Gelman, 2003). Other evidence supporting a negative role of FAK in cancer cell migration includes the fact that colon cancer cells at the migrating front of monolayers exhibit reduced levels of both total and activated FAK; the fact that depletion of FAK by siRNA promotes cell migration (Basson et al., 2006); and the fact that expression of Helicobacter pylori CagA, which is associated with gastric adenocarcinoma, reduces Tyr phosphorylation of FAK (Tsutsumi et al., 2006). Although how inactivation of FAK contributes to tumor cell migration is not yet clear, a report revealed that FAK and paxillin inhibit cell migration by downregulating Rac activity, which appears to be necessary for maintenance of N-cadherin-based cell-cell adhesions in motile cells (Yano et al., 2004).
Considering the paradoxical evidence on the role of FAK in cell growth and migration, our proposed model is the following: overexpression of FAK in human tumor cells might contribute to malignancy by promoting survival under conditions that would normally lead to cell death. FAK is required for tumor development induced by some oncogenic proteins, which activate a positive FAK–c-Src feedback loop and subsequently tumor-promoting molecules, such as ERK and AKT. However, other tumors that have aberrant Ras, which does not require Src for their transforming ability (data not shown), can activate cell proliferation molecules such as ERK and AKT in a FAK-c-Src–independent manner. Activated Ras signaling inhibits FAK activity, leading to reduced cell adhesion and increased cell migration. Importantly, FAK activity is dynamically regulated during tumor development. When metastatic tumor cells re-adhere at foreign tissues or organs, activation of FAK by integrin promotes both the formation of new adhesions and the spread of metastatic tumor cells, which are necessary for the establishment of new metastatic deposits.
In summary, we have uncovered an important mechanism of tumor migration, invasion, and metastasis initiated by activated Ras, which is often mutated in human cancer and mutation of which has been correlated with a poor prognosis and negative clinical outcome (Weijzen et al., 1999). Activated Ras results in FAK dephosphorylation at Y397 and FAK inhibition mediated by downstream Fgd1-Cdc42-PAK1-MEK-ERK signal transmission. Activated ERK phosphorylates FAK at S910, leading to PIN1-dependent prolyl isomerization of FAK and subsequent recruitment of PTP-PEST (Fig. 7B). The downregulation of FAK activity mediated by PTP-PEST, which primarily occurs at lamellipodia of motile cells, promotes disassembly and turnover of focal contact, thereby promoting tumor cell migration, invasion, and metastasis. These findings may provide a molecular basis for treating activated Ras-related tumors by interfering with this Ras-induced signal transmission at multiple levels.
NIH3T3 mouse fibroblasts, NIH3T3-v-H-Ras cells, NIH3T3 cells expressing v-H-Ras (Tet-off), 3Y1 rat fibroblasts, 3Y1-v-H-Ras cells, MCF 10A cells, BT549 cells, U87 cells, rat thyroid epithelial cells transformed by a temperature-sensitive mutant of the Kirsten murine sarcoma virus (Colletta et al., 1983), MK14 cells, PIN1+/+, PIN1−/−, FAK−/−, and PTP-PEST−/− fibroblasts, and 293T cells were all maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (HyClone, Logan, UT).
pBJ5-SF3SosE, pAcActinOnly-MF3E, and pBJ5-SF3E plasmids and FK1012 were kindly provided by Gerald Crabtree (Stanford University). Polyclonal antibodies against Src and phospho-Src Y529 were from Signalway Antibody (Pearland, TX). Polyclonal antibodies against K-Ras (sc-522), Cdc42 (sc-87), GFP (sc-9996), ERK2 (sc-154), FAK (sc-558), p130Cas (Sc-860), and PKCζ (sc-216) were from Santa Cruz Biotechnology (Santa Cruz, CA).
Cells were plated at a density of 4 × 105 per 60-mm-diameter dish 18 h prior to transfection. Transfection was performed using HyFect reagents (Deveville Scientific) according to the vendor’s instructions. The transfection efficiency for 293T cells is about 90%. Stable cell lines were selected with G418 (800 μg/ml) or hygromycin (100 μg/ml) for 10–14 days at 37°C. After treatment with G418 or hygromycin, antibiotic-resistant colonies were picked, pooled, and expanded for further analysis under selective conditions.
MK14 cells [5 × 106 cells in 0.1 ml phosphate-buffered saline (PBS)/mouse] were injected subcutaneously into the right flanks of athymic nude mice. The mice were fed with water containing 2 mg/ml doxycycline. When the tumor size reached 0.8–1.2 cm in diameter, doxycycline was withdrawn from the water for 4 days. The tumor masses were removed and weighed. Six mice in each group were used, and the standard errors represent the variation in tumor weight. Statistical differences were evaluated by using the two-tailed Student’s t test. Differences were considered statistically significant at P<0.05.
3Y1-v-H-Ras cells and 3Y1-v-H-Ras cells expressing FRNK, FAK Y397F, or FAK S910A (1 × 105 cells in 0.1 ml PBS/mouse) were injected into the lateral tail vein of athymic nude mice. After three weeks, mice were sacrificed, and lung tissues were excised and fixed with Bouin’s solution. Lung surface tumors were counted under a low-power dissecting microscope. Six mice in each group were used, and the standard errors represent the variation in tumor number. Statistical differences were evaluated by using the two-tailed Student’s t test. Differences were considered statistically significant at P<0.05. Experiments were done twice.
We thank Drs. Tony Hunter, Lynda Chin, Ronald DePinho, Nicholas Tonks, David Foster, Anthony Means, Steven Hanks, David Schlaepfer, Steven Reed, Makoto Noda, Gerald R. Crabtree, Rakesh Kumar, Jeffrey Field, Bill Davis, Len Stephens, Richard Cerione, Giancarlo Vecchio, Zhizuang Joe Zhao, Kevin Pumiglia, Maria-Magdalena Georgescu, Seth Corey, Jean-Guy LeHoux, Alan Hall, Martin Alexander Schwartz, Yan Chen, Joseph Kissil, Kent Rossman, Jerome Gorski, Roberto Buccione, Michiyuki Matsuda, and Tetsu Akiyam for useful reagents. We thank Stephanie Deming for editing of this manuscript.
This work was supported by the Pediatric Brain Tumor Foundation (Z.L.), a Brain Tumor Society research grant (Z.L.), a Phi Beta Psi Sorority research grant (Z.L.), an institutional research grant from The University of Texas M. D. Anderson Cancer Center (Z.L.), National Cancer Institute grant 5R01CA109035 (Z.L.), and a grant from MOST of CHINA 2009CB918703 (Y.X.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.