Paxillin-Y118 phosphorylation contributes to the control of Src-induced anchorage-independent growth by FAK and adhesion

Autophosphorylation of the focal adhesion kinase, FAK, at Y397 upon integrin-mediated activation produces an SH2-mediated binding site for Src-family tyrosine kinases, or alternatively, other signaling proteins such as phosphatidylinositol 3-kinase (PI3K), Shc, phospholipase-Cγ or Grb7 [1, 2]. Once bound, Src can phosphorylate FAK on several residues including Y925 resulting in the further activation of FAK tyrosine kinase activity [3] and in the phosphorylation of many other cellular substrates. FAK also encodes ligands for multiple protein docking domains such as SH3, an N-terminal FERM domain that facilitates association with integrins and growth factor receptors, and a C-terminal domain that facilitates paxillin/talin binding and focal adhesion targeting (reviewed in ref. [1, 4, 5]). Indeed, the maximal tyrosine phosphorylation of many cellular substrates, such as paxillin, requires both Src and FAK activity [6]. Current thinking suggests that the mutual activation of Src and FAK in response to growth factors, chemotactic agents and cell adhesion leads to activation of a number of downstream pathways in a spatially and temporally controlled manner [7].

Both FAK and Src have been implicated as playing major roles in cancer progression, especially relating to metastatic potential. For example, activation of Src-family kinases has been reported in many primary cancers such as those affecting the colon, GI tract, breast and brain (reviewed in ref. [8]), and activated Src seems to play a critical role in the recruitment of endothelial cells to sites of tumor angiogenesis [9‐11]. FAK protein and activation levels are increased in many primary cancers and further increased in metastatic lesions [12‐23]. Moreover, the loss of FAK activity or expression suppresses metastatic progression in tumor xenograft models, underlining an important positive role for FAK in the development of malignancy [24].

Paradoxically, there are a growing number of studies correlating lower FAK levels with poorer patient survival [25, 26], suggesting that FAK may actually attenuate some of the lethal parameters of cancer progression.

As FAK and Src interact physically and functionally, it was expected that FAK-null cells would be defective for oncogenic transformation by v-Src. Thus, it was surprising that in the absence of FAK, v-Src could induce morphological transformation, cell motility, cell polarity and Matrigel invasiveness [27, 28], and even enhance the frequency of anchorage-independent growth 5- to 10-fold [27]. Indeed, a high-resolution one-dimensional anti-phosphotyrosine analysis of FAK+/+ versus FAK-/- fibroblasts expressing v-Src identified fewer than 10 varying substrates. Although the full phosphorylation of some cellular substrates, such as Endophilin A2, requires both FAK and Src [29], it is possible that such substrates are dispensable for oncogenic transformation. A further complication is that phosphorylation of some substrates, such as PI3K, varied only under conditions of anchorage independence [30]. Indeed, we demonstrated that the superactivation of PI3K was required and sufficient for the enhanced v-Src-induced anchorage-independent growth of FAK-/- fibroblasts [30].

Here, we analyze how FAK affects the phosphorylation of a panel of known Src substrates under conditions of anchorage-dependent and -independent growth. Whereas the majority of substrate phosphorylations neither required FAK for v-Src-induced phosphorylation nor were enhanced in the presence of FAK, we identified a small number of substrates whose phosphorylation was affected by FAK and/or adherence state. These variations suggest that FAK and integrins might play either positive or negative roles during the overall process of malignant progression.

Many studies have demonstrated physical and functional interactions between FAK and Src in response to integrin- and growth factor-mediated signals. However, most of the studies linking FAK/Src complexes with cell motility, proliferation and cell survival in cancer progression involve adherent cell populations. Yet, many parameters of cancer biology in vivo, especially metastasis, require anchorage-independent proliferation. To elucidate the roles of adhesion and/or FAK on Src-induced oncogenesis, we developed FAK+/+ or FAK-/- MEF expressing relatively similar protein and activity levels of the v-Src oncogene (Fig. 1A). As we reported previously [27], the relative level of v-Src autophosphorylation (poY416) under adherent conditions was consistently 2-fold higher in the FAK-/- background, and this correlated with a slightly higher overall level of cellular phosphotyrosyl proteins compared to the FAK+/+ background (Fig. 1B, left panel). Interestingly, overall v-Src-induced tyrosine phosphorylation of cellular substrates, as well as the total number of substrates, was slightly higher in v-Src cells kept in suspension, irrespective of FAK content (Fig. 1B, right panel). This was not due to relative increases in Src activation levels in the suspended cells (Fig. 1A). Although the gross level of Src-induced tyrosine phosphorylation is similar in the presence or absence of FAK, there are a small number of substrates whose relative phosphorylation level is either increased or decreased by FAK (Fig. 1B). Moreover, in adherent cells, there were more phosphorylated Src substrates in the absence of FAK (stars), whereas in the suspended cells, there were a roughly equal number of substrates favored in the presence (arrows) or absence (stars) of FAK. These data are consistent with the notion that FAK modulates the ability of activated Src to associate with and/or phosphorylate specific cellular substrates [27].

The possibility that variations in substrate choice by v-Src could be influenced by FAK and/or adhesion may have consequences on specific parameters of oncogenic transformation. For example, v-Src induces 5- to 10-fold higher anchorage-independent growth (AIG) in the absence of FAK, correlating with a concomitant increase in PI3K activation levels under AIG conditions [27, 30]. However, we showed previously that the twofold higher level of activated v-Src is not responsible for the enhanced AIG (eAIG) in the FAK-/- background inasmuch as varying v-Src levels only altered soft agar colony size [30]. In order to clarify roles for FAK and adhesion in regulating v-Src-induced phosphorylation of cellular substrates, FAK+/+, FAK-/- MEFs and their v-Src transformed derivatives, FAK+/+[v-Src] and FAK-/-[v-Src], were incubated in the presence of serum under conditions of adherence versus suspension, and then cell lysates were probed by IB for changes in the phosphorylation status of known Src substrates. Thus, IBs were probed with either phospho-specific substrate Abs (Table 1A), or alternatively, substrate proteins were immunoprecipitated and then analyzed by anti-PTyr IB using MAb-4G10 (Table 1B). In all cases, the change in the relative phosphorylation state of each substrate was normalized to apo-protein levels as well as to GAPDH, used as a loading control. One caveat is that GAPDH levels typically increase twofold in murine cells transformed with v-Src [30, 36], a finding consistent throughout the current study, and this was factored into the final normalization.

Table 2 shows that only a minority of the Src substrates studied were unaffected by FAK and/or adhesion. In fact, only the phosphorylation signals of Eps8, PLC-γ1 (at Y783), and Shc (at Y239/240) showed no change under these variables (Fig. 2). The phosphorylation of cortactin (at Y421) and paxillin (at Y31) was unaffected by FAK only in adherent cultures, whereas the phosphorylation of Sam68 and PKC-δ (at Y311) was unaffected by FAK only in suspension cultures. Interestingly, we observed a v-Src-induced increase in cortactin protein levels, and moreover, a slower-migrating form whose abundance was increased even more in FAK-/-[v-Src] cells. The phosphorylation of several Src substrates was enhanced by FAK, irrespective of adherence state. These include vinculin (at both the Y100 and Y1065 sites) and CAS. In contrast, the phosphorylation of p190RhoGAP, paxillin (at Y118) and Crk was favored in the absence of FAK, irrespective of adherence state. Lastly, although v-Src induced annexin II protein levels, the phosphorylation of annexin II seemed to be inhibited by v-Src in the absence of FAK.

The phosphorylation of a set of Src substrates was influenced by FAK depending on adhesion conditions. Specifically, the connexin43^poY265 signal was enhanced by FAK in suspension cultures, yet attenuated by FAK in adherent cultures (Table 2; Fig. 2). Conversely, the p120 catenin^poY228 (Fig. 3) and p56Dok-2^poY351 signals (Fig. 2) were enhanced by FAK in adherent cultures, yet attenuated by FAK in suspension cultures. The Src-induced phosphorylation of Sam68 was enhanced in the absence of FAK only in adherent cultures (Fig. 2). Lastly, the Src-induced phosphorylation of paxillin at Y118 in the absence of FAK is enhanced to an even greater extent in suspension cultures (10- versus 4-fold, respectively; Fig. 4). Taken together, our data indicate that the phosphorylation of most of the well-documented Src substrates is influenced by FAK and/or adhesion.

It is possible that the eAIG induced by v-Src in the absence of FAK is mediated by the enhanced phosphorylation of p120catenin or paxillin, Src substrates whose phosphorylation was consistently increased in FAK-/-[v-Src] cells. To address this, stable FAK+/+[v-Src] and FAK-/-[v-Src] cell clones were produced that express p120catenin^Y228F or paxillin^Y118F alleles. As a negative control, the cells were transfected with Shc^Y239/240F, whose phosphorylation is affected by neither FAK nor adherence status (Table 2). Figs. 5A–C show the expression of these exogenous proteins (probed for their epitope tags). As expected, the expression of these non-phosphorylatable mutants suppressed the levels of endogenous Shc^poY239/240, p120catenin^poY228 or paxillin^poY118. This effect was apparent in multiple cell clones, and, in FAK-/-[v-Src] cells transiently co-transfected with paxillin^poY118 and pEGFP (sorted for GFP-positive cells; data not shown), strongly suggesting that the Y->F mutants function as dominant-interfering alleles. We cannot rule out, however, that these mutants also affected the phosphorylation of other residues on their endogenous counterparts. Importantly, the ectopic expression of paxillin^Y118F, but not p120catenin^Y228F or Shc^Y239/240F, selectively decreased the enhanced AIG in FAK-null cells compared to those expressing FAK (Fig. 5, panels A-C). This was not due to changes in cell survival based on the similar clonogenic frequencies between FAK+/+ or FAK-/- v-Src clones expressing the paxillin^Y118F allele (Fig. 5A, right middle panel). As a further control, we showed that the stable expression of the WT-paxillin allele (expressed as a GFP-paxillin fusion) failed to alter eAIG in FAK-/-[v-Src] cells (Fig. 5A, bottom right). This strongly suggests that the enhanced Src-induced phosphorylation of paxillin at Y118 in the absence of FAK plays a significant role in inducing enhanced Src-mediated AIG. This contrasts with the finding that Src-induced phosphorylation of p120catenin on Y228, but not of Shc on Y239/240, was critical to Src-induced anchorage-independent growth irrespective of the FAK background (Fig. 5B &5C).

Given the artificial nature of our cell system, we sought to determine whether the relationship between loss of FAK, increased paxillin^poY118 and eAIG was manifest in human cancer cells lines. Colon cancer is marked by the activation of Src, mainly through overexpression although small percentages encode mutated Src known to cause upregulation of its intrinsic tyrosine kinase activity [37]. Because the knockdown of FAK in cancer cell lines using antisense oligonucleotides or siRNAs often induces apoptosis [27, 38‐42], we screened a panel of human colon cancer cell lines for those that failed to apoptose after treatment with FAK siRNA. Treatment of two such lines, HT-25 and RKO, with FAK-siRNA under conditions of adherent or suspension growth caused 3- to 4-fold decreases in FAK protein level (Fig. 6A). The loss of FAK marginally decreased cell survival as measured by clonogenic colony growth (Fig. 6C), yet resulted in significant increases in AIG as measured by soft agar colony-forming frequencies (Fig. 6B). Note that the loss of FAK did not affect the colony size formed by these cell lines (not shown). If the AIG frequency is normalized to cell survival rates, FAK knockdown induces an equivalent 1.75-fold increase in relative AIG in both HT-25 and RKO over control cells (GFP siRNA). Given that the loss of FAK in these cancer cell lines recapitulated the eAIG found in our FAK-/-]v-Src] MEF, we analyzed how the loss of FAK affected paxillin^poY118 levels under adherent or suspension conditions. Suspension cells treated with siFAK exhibit increased paxillin^poY118 levels, especially a fast-migrating isoform, whereas the loss of FAK in adherent cells resulted in decreased relative paxillin^poY118 levels (Fig. 6A). The protein levels of GAPDH (Fig. 6A) or paxillin (not shown) did not alter by these conditions. These data strengthen the notion that in HT-25 and RKO, FAK may antagonize AIG by inhibiting the phosphorylation of paxillin^Y118 by activated Src-family kinases.

Cancer progression, especially metastasis, is controlled by multiple stages, with varying dependencies on the proliferative, motility and survival signals mediated by cell adhesion and FAK [1, 43]. For example, primary and metastatic site tumors cells require growth conditions less dependent on integrin-mediated adhesion than individual metastatic cells require while intravasating from primary lesion and extravasating to peripheral sites [44]. Similarly, the motility-promoting activity of FAK would be more desirable to facilitate extravasation and intravasation of metastatic cells [45], yet once at peripheral sites, continued FAK motility signals would likely be suppressed in the growing metastatic lesion. Although the activation of Src-family kinases has been noted as an important event in early cancer progression- mostly involved in mediating proliferative signals- there is mounting data that Src signaling is especially critical for the metastatic process, specifically, to facilitate cancer cell survival and neovascularization of metastatic sites [8, 9, 11]. Our results indicate that in a majority of well-documented Src substrates, Src-induced phosphorylation is influenced by FAK and/or adhesion. Some are FAK dependent whereas as others are attenuated by FAK, irrespective of adherence state. Yet others are affected by FAK only under adherent or non-adherent conditions. Because our cells were derived in a p53-/- background [31], it would be interesting to understand how p53 might regulate these phosphorylations. However, the conditional loss of FAK in v-Src-transformed p53+/+ MEF (Lingqiu Gao and I.H. Gelman, unpublished data; Hilary Beggs, UCSF, personal communication) or the ectopic expression of WT-p53 in our FAK-/-[v-Src] cells (Sanjay Sachdev and I.H. Gelman, unpublished data) leads to rapid apoptosis, making this determination technically unfeasible at the present time. In sum, our data strongly suggest that FAK- and adhesion-signaling can modulate which substrates are favored (or disfavored) by Src, thereby influencing the various dynamic biologies that contribute to cancer progression.

In regards to mechanism, we speculated previously that regulation of cytoskeletal dynamics by FAK and/or adhesion-activated signals might affect co-localization of Src with its potential substrates. Alternatively, even if FAK/Src complexes are found at the subcellular sites of potential substrates, such as focal adhesion plaques, the preferential binding of FAK to the Src-SH2 domain through its poY397 site may shift substrates to other, weaker binding sites on Src, such as the SH3 domain, thereby resulting in lower levels of phosphorylation [30].

There are very few studies that identify an obligate role for FAK in enhancing Src-mediated phosphorylation of specific substrates. For example, Wu et al. [29] demonstrated that full phosphorylation of endophilin A2 by Src requires FAK. Brown et al. [46] show that Src and FAK cooperate to phosphorylate PKL during cytoskeletal remodeling. El Annabi et al. [47] describe a system in which insulin-mediated receptor activation and IRS-1 tyrosine phosphorylation is enhanced in cells adhering to fibronectin versus suspended cultures, and that phosphorylation can be restored in suspended cells by the overexpression of both FAK and Src. Ruest et al. [48] showed that maximal phosphorylation of p130CAS in COS-7 cells required co-expression of Src and FAK. Roy et al. [28] showed that the ability of v-Src to phosphorylate the FAK family member, Pyk2, is attenuated by FAK, whereas the Src-induced phosphorylation of CAS and paxillin is enhanced by FAK. Hsia et al. [49] and Moissoglu et al. [30] identify another example where a potential Src substrate, STAT3, required neither FAK nor adhesion for maximal phosphorylation.

The current study is the first to describe either positive or negative roles for FAK and adhesion in the Src-induced phosphorylation of PKC-δ, p190RhoGAP, Shc, p120 catenin, connexin-43, cortactin, p56Dok-2, Crk, Eps8, PLC-γ1, Sam68 and vinculin. Our data indicate that FAK and/or adhesion status affect the phosphorylation of most of these substrates.

Phosphorylation of PKC-δ at Y311 by Src is associated with caspase-mediated apoptosis in many cell types (reviewed in ref. [50]). Our data showing increased PKC-δ^Y311 phosphorylation in adherent FAK-/-[v-Src] cells, whereas phosphorylation in suspension cells is FAK-independent, suggests that the ability of FAK to suppress apoptosis [1‐4] is enhanced by integrin-mediated signals.

The Src-induced phosphorylation of Shc, PLC-γ1 and Eps8 was unaffected by FAK or adherence state. Phosphorylation of PLC-γ1 at Y783 is specifically induced by integrin-mediated adhesion, and mutation of this site antagonizes adhesion [51]. However, Tyorogov et al. [52] showed that adhesion-induced PLC-γ1^poY783 could occur in FAK-deficient MEF, in line with our findings that Src-induced PLC-γ1^poY783 is FAK-independent.

The Src-induced phosphorylation of p190RhoGAP and Crk is enhanced slightly in the absence of FAK irrespective of adherence state. Although p190RhoGAP can bind to and be phosphorylated by FAK [53, 54], our data indicate that the v-Src-induced phosphorylation of p190RhoGAP is not dependent on FAK or adhesion. p190RhoGAP phosphorylation by v-Src is known to activate its ability to suppress RhoA, thereby inhibiting the formation of actin stress fibers and focal adhesion complexes in transformed cells [55]. Moreover, v-Src facilitates increased complex formation between p190RhoGAP and p120RasGAP.

In contrast, FAK may play a critical role in regulating p190RhoGAP activity in untransformed cells. Specifically, Y31/118-phosphorylated paxillin has been shown to liberate p190RhoGAP from complexes with p120 RasGAP, thereby allowing p190RhoGAP to suppress RhoA-mediated cytoskeletal remodeling in endothelial cells [56]. Indeed, our current data show that the relative levels of paxillin^poY31/118 are decreased in FAK-/- versus FAK+/+ MEF (i.e.- not expressing v-Src, Fig. 3), suggesting that FAK is required for p190RhoGAP-mediated inhibition of RhoA.

The Src-induced phosphorylation of p130CAS and vinculin was enhanced by FAK but not by adherence state. For example, phosphorylation of the Crk-associated substrate, p130CAS, by v-Src was favored slightly in FAK-expressing cells even though its ability to be phosphorylated at lower levels by v-Src in adherent FAK-/- cells [27, 28] may be the result of adhesion-activated Pyk2 [57]. Our current results, that Src-induced phosphorylation of CAS at Y410 in adherent culture is enhanced by FAK, are similar to those of Roy et al. [28], though they looked at the overall tyrosine phosphorylation of CAS. In regards to biologic significance, the anchorage-independent tyrosine phosphorylation of CAS correlates with suppression of anoikis [58]. Src-mediated tyrosine phosphorylation of CAS is required for the invasive phenotype and metastasis formation but not for primary tumor growth [59]. Our present observation of enhanced phosphorylation of CAS in the presence of FAK, irrespective of adhesion, argues that FAK may positively influence the ability of Src to suppress anoikis and induce invasion and metastasis. Moreover, Patwardhan et al. [60] observed that CAS, independent of its phosphorylation status, enhances the ability of Src to promote anchorage-independent growth. Thus, taken with our data, we can conclude that differential CAS phosphorylation cannot be responsible for enhanced AIG we detect in the absence of FAK.

The FAK-dependent enhancement of Src-induced vinculin phosphorylation (at both Y100 and Y1065) irrespective of adherence state conflicts somewhat with the previous findings of Chang et al. [61] who showed that anoikis induced by FAK^Y397F expression in v-Src-transformed MEF only correlated with decreased CAS, but not vinculin, talin or paxillin protein levels. It should be stressed, though that the FAK^Y397F allele is not equivalent to a FAK-null condition in that i) FAK can bind v-Src (but not c-Src) through SH3-mediated interactions [62], and ii) FAK^Y397F may facilitate anoikis by acting as a sink for other survival factors.

The ability of FAK to either enhance or attenuate Src-induced phosphorylation of paxillin, p120catenin, connexin and p56Dok-2 was influenced by adherence state. We found that whereas Src-induced paxillin^Y118 phosphorylation is attenuated by FAK, FAK is required for optimal phosphorylation of paxillin^Y31 in suspension cultures. Interestingly, Roy et al. [28] and our lab [30] showed previously that the overall level of Src-induced paxillin tyrosine phosphorylation level is not affected by FAK, yet we showed that FAK-/-[v-Src] cells lacked some of the super-phosphorylated, slow-mobility paxillin isoforms found in FAK+/+[v-Src] cells. Our current work makes use of the site-specific phospho-Abs to show that FAK can either induce or attenuate paxillin phosphorylation at specific residues.

Our data show that FAK enhances Src-induced p120catenin^poY228 in adherent cultures yet attenuates Src-induced p120catenin^poY228 in suspension cultures. The tyrosine phosphorylation of p120catenin is associated with the loss of catenin-cadherin cell-cell adherens junctions in several systems [63, 64]. The preponderance of data identifies functions for FAK in focal adhesion complexes, and not in cell-cell junctions. Not surprisingly, only one study indirectly links FAK with cell-cell junctions: Irby and Yeatman [65] show that FRNK, a naturally-occurring dominant-interfering FAK allele, can restore some of the Src-suppressed cell-cell interactions. This suggests that FAK functions downstream of Src to facilitate catenin phosphorylation, a conclusion backed by our findings. Nonetheless, p120catenin^poY228 levels do not correlate with changes in cell-cell interactions under suspension conditions with our MEF. How FAK might associate with or antagonize the phosphorylation of catenins in the absence of adherence signaling is unclear at this time. One hint may come from the finding that p120catenin translocates from cell-cell junctions to the cytoplasm during epithelial to mesenchymal transition and that cytoplasmic p120catenin correlates with poor prognosis and lymph node metastasis in colon cancer [66].

Phosphorylation of p56Dok-2 at Y351 establishes a binding site for the SH2 domain of adaptor protein Nck, which is involved in the organization of actin cytoskeleton [67‐69]. Noguchi et al. [70] observed no appreciable difference between adherent versus suspension cultures of v-Src transformed rat fibroblasts in regards to Dok-1 phosphorylation level. However, the current study is the first to show that p56Dok-2 phosphorylation by Src is enhanced by FAK in adherent cells but attenuated by FAK in suspension cells. The significance of our observation remains unclear at this point but suggests that FAK can positively or negatively influence actin cytoskeletal dynamics through p56DOK-2 phosphorylation depending on adherence state.

We previously demonstrated that the AIG induced by Src in MEF was enhanced 5- to 10-fold in the absence of FAK [27]. The phosphorylation of two Src substrates, paxillin (at Y118), and p120catenin (at Y228), was increased in suspended FAK-/-[v-Src] cells relative to cells expressing FAK and growing in adherent conditions. Our data strongly suggest that paxillin^poY118, but not p120catenin^poY228, regulates the enhanced AIG phenomenon. This is the first identification of a required role for a specific paxillin phosphorylation event in AIG, and adds to recent evidence correlating paxillin phosphorylation at Y118 with AIG in colon cancer cells forced to express β₄GalNAc-T3 [71] and the requirement for paxillin for epidermal growth factor-induced AIG growth of JB6 Cl41 fibroblasts [72]. A central role for paxillin^poY118 in driving AIG is underlined by our finding that two human colon cancer epithelial cell lines, HT25 and RKO, exhibit increased AIG after siRNA-mediated FAK knockdown, correlating with increased relative paxillin^poY118 levels under suspension growth conditions.

We showed previously that PI3K/AKT is a major mediator of Src-induced enhanced AIG in the absence of FAK [30]. Thus, FAK seems to attenuate the ability of Src to bind to and/or phosphorylate paxillin at Y118 (this study) and the p85 regulatory subunit of PI3K [30], resulting in decreased AIG. It is unclear whether paxillin and PI3K control common AIG-inducing pathways; the possibility that there is crosstalk between these two mediators has not been described in the literature to date.

Our data identify FAK- and/or adhesion-regulated effects on the choice of substrates phosphorylated by Src. As oncogenic transformation and progression are multi-step processes, it will be interesting to determine how the FAK and adhesion effects we identified might correlate with the dynamic changes to the dependence on motility, cell survival and adhesion required at different points of cancer progression.