Cdc42 and Ras Cooperate to Mediate Cellular Transformation by Intersectin-L*

Cdc42, a Ras-related GTP-binding protein, has been implicated in the regulation of the actin cytoskeleton, membrane trafficking, cell-cycle progression, and malignant transformation. We have shown previously that a Cdc42 mutant (Cdc42(F28L)), capable of spontaneously exchanging GDP for GTP (referred to as “fast-cycling”), transformed NIH 3T3 cells because of its ability to interfere with epidermal growth factor receptor (EGFR)-Cbl interactions and EGFR down-regulation. To further examine the link between the hyperactivation of Cdc42 and its ability to alter EGFR signaling and thereby cause cellular transformation, we examined the effects of expressing different forms of the Cdc42-specific guanine nucleotide exchange factor, intersectin-L, in fibroblasts. Full-length intersectin-L exhibited little ability to stimulate nucleotide exchange on Cdc42, whereas a truncated version that contained five Src homology 3 (SH3) domains, the Dbl and pleckstrin homology domains (DH and PH domains, respectively), and a C2 domain (designated as SH3A-C2) showed modest guanine nucleotide exchange factor activity, whereas a form containing just the DH, PH, and C2 domains (DH-C2) strongly activated Cdc42. However, DH-C2 showed little ability to stimulate growth in low serum or colony formation in soft agar, whereas SH3A-C2 gave rise to a much stronger stimulation of cell growth in low serum and was highly effective in stimulating colony formation. Moreover, although SH3A-C2 strongly transformed fibroblasts, it differed from the actions of the Cdc42(F28L) mutant, as SH3A-C2 showed little ability to alter EGFR levels or the lifetime of EGF-coupled signaling through ERK. Rather, we found that SH3A-C2 exhibited strong transforming activity through its ability to mediate cooperation between Ras and Cdc42.

Cdc42, a Ras-related GTP-binding protein, has been implicated in the regulation of the actin cytoskeleton, membrane trafficking, cell-cycle progression, and malignant transformation. We have shown previously that a Cdc42 mutant (Cdc42(F28L)), capable of spontaneously exchanging GDP for GTP (referred to as "fastcycling"), transformed NIH 3T3 cells because of its ability to interfere with epidermal growth factor receptor (EGFR)-Cbl interactions and EGFR down-regulation. To further examine the link between the hyperactivation of Cdc42 and its ability to alter EGFR signaling and thereby cause cellular transformation, we examined the effects of expressing different forms of the Cdc42-specific guanine nucleotide exchange factor, intersectin-L, in fibroblasts. Full-length intersectin-L exhibited little ability to stimulate nucleotide exchange on Cdc42, whereas a truncated version that contained five Src homology 3 (SH3) domains, the Dbl and pleckstrin homology domains (DH and PH domains, respectively), and a C2 domain (designated as SH3A-C2) showed modest guanine nucleotide exchange factor activity, whereas a form containing just the DH, PH, and C2 domains (DH-C2) strongly activated Cdc42. However, DH-C2 showed little ability to stimulate growth in low serum or colony formation in soft agar, whereas SH3A-C2 gave rise to a much stronger stimulation of cell growth in low serum and was highly effective in stimulating colony formation. Moreover, although SH3A-C2 strongly transformed fibroblasts, it differed from the actions of the Cdc42(F28L) mutant, as SH3A-C2 showed little ability to alter EGFR levels or the lifetime of EGFcoupled signaling through ERK. Rather, we found that SH3A-C2 exhibited strong transforming activity through its ability to mediate cooperation between Ras and Cdc42.
Cdc42 is a Ras-related GTP-binding protein that functions in a wide range of cellular activities. Like other members of the Rho subfamily, Cdc42 was first implicated in the regulation of the actin cytoskeletal architecture. In particular, the microinjection of activated forms of Cdc42 caused filopodia formation and the generation of microspikes, whereas activated Rac1 stimulated the formation of lamellipodia and membrane ruffling, and activated RhoA promoted actin stress fibers (1)(2)(3)(4).
However, subsequently, a number of lines of evidence have also implicated Cdc42, as well as other Rho-family proteins, in the control of normal cell growth and, when aberrantly regulated, in tumorigenesis and metastasis. Perhaps foremost among the lines of evidence linking these GTP-binding proteins to cell proliferation was the initial discovery that members of the Dbl (for diffuse B cell lymphoma) family of guanine nucleotide exchange factors (GEFs), 1 either when point-mutated or truncated, were capable of potently transforming mouse fibroblasts (5,6). Mutations in Cdc42 which mimic the functional activity of Dbl proteins and cause the constitutive exchange of GDP for GTP (called "fast-cycling" mutants) are transforming (7)(8)(9). Both Rac1 and Cdc42 have also been linked to cell invasiveness (10 -12), and another Rho family member, RhoC, has been strongly implicated in metastasis (13). Moreover, the activation of Cdc42, as well as Rac1 and RhoA, is essential for Rasinduced malignant transformation (14 -16).
We have recently obtained some important insights into the mechanistic basis by which fast-cycling mutants of Cdc42 transform fibroblasts (17). Somewhat surprisingly, we discovered that cells expressing Cdc42(F28L) exhibited significantly higher levels of epidermal growth factor receptors (EGFRs) compared with control fibroblasts and showed a markedly extended lifetime for EGFR-coupled signaling. These observations demonstrated the ability of activated Cdc42 to negatively regulate the interactions of the EGFR with one of its binding partners, the adaptor protein c-Cbl. Various lines of evidence have shown that the degradation of EGFRs occurs via a cascade of ubiquitination enzymes that culminates in the E3 ligase activity catalyzed by c-Cbl (18 -21). Activated forms of Cdc42, by associating with the p85Cool-1/␤-Pix protein, a Cbl-binding partner, are able to sequester c-Cbl away from the EGFR. This prevents the EGFR from phosphorylating Cbl, which has been suggested to activate its E3 ligase activity (19). Because the fast-cycling Cdc42(F28L) mutant is constitutively active, it is able to persistently inhibit EGFR-Cbl interactions, thereby leading to a significant reduction in receptor degradation. This, in turn, causes EGFRs to accumulate and exhibit sustained signaling, resulting in cellular transformation.
These findings raised the question of whether the negative regulation of EGFR degradation was a general feature of Cdc42-mediated cell growth control and cellular transformation. We were especially interested in seeing whether activated Dbl family GEFs that were specific for Cdc42 led to the same effects on EGFR lifetime and signaling as the fast-cycling Cdc42(F28L) mutant. To address this question, we examined the cellular effects of different forms of intersectin-L, a Cdc42specific GEF (22). We thought that intersectin-L would be especially interesting to study in the context of possible effects on EGFR degradation because a shorter splice variant form of the protein, designated intersectin-S, has been implicated in endocytosis (23). In addition, the overexpression of intersectin-S has been shown to exhibit weak transforming activity, perhaps in part because of a disruption of the normal endocytosis of growth factor receptors and because of a reported ability to input into the Ras-signaling pathway via an interaction with the Ras-GEF, Sos (Son-of-sevenless) (24). Here, in an attempt to directly correlate the activation of Cdc42 by intersectin-L with cellular transformation, we have compared the abilities of different versions of intersectin-L to act as Cdc42-GEFs. However, unexpectedly, we found that the most potent transforming activity was induced by an intersectin-L construct that only moderately activated Cdc42, whereas a form of intersectin-L that more strongly activated Cdc42 was less effective in transforming fibroblasts. Maximal transformation by intersectin-L required its ability to activate both Cdc42 and Ras. Unlike the case for fast-cycling Cdc42(F28L), transformation by intersectin-L was not the result of significant changes in EGFR levels or signaling to ERK but, rather, required the combined effects of Ras and Cdc42 to signal to the c-Jun NH 2terminal kinase (JNK) and the phosphoinositide 3-kinase (PI3K). Overall, these results provide an interesting example of how Cdc42 can cooperate with Ras to elicit cellular transformation through signaling pathways that are distinct from those used by fast-cycling Cdc42 mutants.
Plasmids and Stable Cell Lines-Mouse intersectin-L was a gift from Dr. Sean E. Egan (The University of Toronto), and the plasmid encoding HA-tagged Ras was obtained from Dr. John P. O'Bryan (National Institutes of Health). Myc-tagged SH3A-C2 and DH-C2 were generated by PCR using wild-type intersectin-L as a template. These two constructs were subcloned into the KpnI and NotI sites of pcDNA3 and into the SmalI and KpnI sites of the pJ4H mammalian expression vector for transient and stable expression, respectively. SH3A-C2(E1237A), which contains a single point mutation in the DH domain, and SH3A-C2(WS), which contains point mutations at the conserved tryptophan residues 777 and 1031 in the SH3A and SH3C domains, respectively, were produced using the QuikChange TM site-directed mutagenesis kit (Stratagene). For selection of stable cell lines, NIH 3T3 cells were transfected with pcDNA3-Neo, together with pJ4H-intersectin-L, pJ4H-SH3A-C2, pJ4H-DH-C2, pJ4H-SH3A-C2(WS), or pJ4H-SH3A-C2(E1237A). Transfected cells were maintained in DMEM supplemented with 5% calf serum and 700 g/ml G418 (Invitrogen). After 10 -14 days, G418-resistant colonies were selected.
Cell Culture, Transfection, Immunoprecipitation, and Immunoblotting-NIH 3T3 cells were cultured in DMEM supplemented with 10% calf serum (Invitrogen). COS-7 or HEK 293 cells were maintained in DMEM supplemented with 10% fetal bovine serum (Invitrogen). For transient transfections, plasmid DNA was introduced into NIH 3T3, COS-7, or HEK 293 cells using the Lipofectamine transfection kit (Invitrogen). Two days after transfection, the cells were lysed in 20 mM Hepes, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 20 mM ␤-glycerol phosphate, 1 mM sodium orthovanadate, 20 mM NaF, 10 g/ml leupeptin, and 10 g/ml of aprotinin. The lysates were cleared by centrifuga-tion at 13,000 rpm for 10 min, incubated with Myc antibody for 2 h, and then incubated with protein G beads for 1 h. The denatured samples were resolved by SDS-PAGE and detected by Western blotting with the indicated antibodies.
Low Serum, Saturation Density, and Soft Agar Assays-For low serum assays, stable cell lines were plated at a density of 10 5 cells in 12-well dishes in DMEM supplemented with 10% calf serum. After 5 h, the medium was changed to DMEM supplemented with 1% calf serum and then changed every other day for 6 days. At the indicated times, the cells were trypsinized and counted with a hemocytometer. For the saturation density assays, 10 5 cells were seeded in 12-well dishes in DMEM supplemented with 10% calf serum. The medium was changed every other day for 6 days, and then the cell density was determined by counting the cells, as described above, on day 6. For the soft agar assays, 10 4 cells stably expressing the different constructs were mixed with DMEM supplemented with 10% calf serum and 0.3% agarose and plated on top of a solidified layer of DMEM supplemented with 0.5% agarose and 10% calf serum. Cells were fed weekly by adding 1 ml of DMEM supplemented with 10% calf serum and 0.3% agarose. After 2 weeks, colonies larger than 50 m were scored.
[ 3 H]GDP Dissociation Assays-COS-7 cells were transfected with cDNAs encoding Myc-tagged intersectin-L, SH3A-C2, SH3A-C2(E1237A), and DH-C2. Forty-eight hours after transfection, cells were lysed in mammalian cell lysis buffer. The lysates were immunoprecipitated with anti-Myc antibody and then incubated with protein G beads for 1 h at 4°C. The immunoprecipitated complex was incubated with 1 g of Cdc42 preloaded with [ 3 H]GDP in 140 l of reaction buffer at room temperature, and aliquots (23 l) were diluted into 1.5 ml of ice-cold termination buffer (20 mM Tris-HCl, 10 mM MgCl 2 , and 100 mM NaCl, pH 7.4) at various time points. The percent [ 3 H]GDP remaining on the filters was detected by scintillation counting.
Assaying the Cellular Activation of Cdc42 and Ras-To assay the cellular activation of Cdc42, COS-7 cells were transiently transfected with the cDNAs encoding Myc-Cdc42, together with either Myc-tagged intersectin-L, SH3A-C2, or DH-C2. Forty-eight hours after transfection, cells were lysed in mammalian cell lysis buffer and incubated with 40 g of recombinant glutathione S-transferase (GST)-PBD (for p21-binding domain from PAK). The GST-PBD was precipitated with glutathione-agarose beads, and the beads were then washed three times with lysis buffer, subjected to SDS-PAGE, and immunoblotted with anti-Myc antibodies. When assaying the cellular activation of Ras, HEK 293 cells were transiently transfected with the cDNA encoding HA-Ras together with different forms of Myc-SH3A-C2 or the vector control. Thirty-six hours after transfection, the cells were starved overnight and then, in some cases, stimulated with EGF for 5 min. The cell lysates were incubated with 40 g of recombinant GST-Raf-RBD (for Ras-binding domain). The GST-RBD was precipitated with glutathione-agarose beads, similar to GST-PBD, and then subjected to SDS-PAGE and immunoblotted with anti-HA antibody.

Comparisons of the Abilities of Different Forms of Intersec-
tin-L to Activate Cdc42-It was reported previously that intersectin-L functions as a specific GEF for Cdc42 (22). As a first step toward developing a correlation between the extent of Cdc42 activation and its ability to transform NIH 3T3 fibroblasts, we assayed the Cdc42-GEF activities of different forms of intersectin-L. In particular, the GEF activity of full-length intersectin-L, which contains two EH domains, five SH3 domains, DH and PH domains, and a C2 domain (Fig. 1, ITSN-L) was compared with those of a shorter version (SH3A-C2), which is composed of the five SH3 domains and everything downstream, and a still shorter form that simply contains the DH, PH, and C2 domains (DH-C2). These different intersectin-L constructs were expressed in COS-7 cells as Myc-tagged proteins, immunoprecipitated with anti-Myc antibody ( Fig. 2A Fig. 2A  (top panel) show that the DH-C2 construct exhibited the strongest GEF activity, whereas SH3A-C2 showed moderate activity, and full-length intersectin-L (designated Myc-Intersectin in Fig. 2A, top panel, and Myc-ITSN in Fig. 2A, bottom  panel) was completely inactive. These results are consistent with the idea that full-length intersectin-L is susceptible to an autoinhibition, perhaps because the SH3 domains fold over and block access to the DH domain (25). We next examined whether the different forms of intersectin-L showed the same relative abilities to activate Cdc42 in cells. We performed this examination using an assay that takes advantage of the binding of activated Cdc42 to the Cdc42 limitbinding domain from the Cdc42/Rac target, PAK3, fused to GST (GST-PBD). The different forms of Myc-tagged intersectin-L were expressed together with Myc-tagged Cdc42 in COS-7 cells, and the relative abilities of the intersectin-L constructs to promote the activation of Myc-Cdc42 were compared by incubating lysates from the different transfectants with GST-PBD followed by precipitation with glutathione-agarose. cell lysates. Similar to the results obtained from the in vitro nucleotide exchange assays ( Fig. 2A), the DH-C2 construct was the most effective at stimulating Cdc42 activation in cells. In fact, the extent of Cdc42 activation promoted by DH-C2 yielded a signal in the PBD assay that was comparable with the positive control obtained from cells expressing constitutively active, GTPase-defective Myc-Cdc42(Q61L). The SH3A-C2 construct gave rise to a weaker cellular activation of Cdc42, whereas full-length intersectin-L was again ineffective at activating Cdc42.
When we compared the abilities of the different forms of intersectin-L to stimulate the formation of microspikes in NIH 3T3 cells, we were not able to distinguish between SH3A-C2 and DH-C2, as both constructs induced this phenotype in the majority of cells examined. Fig. 2C (upper left panel) shows a representative example for cells expressing SH3A-C2. Expression of full-length intersectin-L did not induce these cellular changes (not shown). This suggests that although Cdc42 is less effectively activated by SH3A-C2 compared with DH-C2, the SH3A-C2 construct is able to stimulate sufficient activation of the endogenous Cdc42, such that complete microspike formation is observed.
For the case of the founding member of the Dbl family (i.e. oncogenic Dbl), it has been shown that changing glutamic acid 502 to an alanine yields a Dbl protein that is still capable of binding to Cdc42 but is unable to stimulate nucleotide exchange (26). Sequence alignments indicate that glutamic acid 1237 within the DH domain of intersectin-L corresponds to glutamic acid 502 in oncogenic Dbl. When we changed glutamic acid 1237 in the intersectin-L proteins to alanine, we completely eliminated their ability to exhibit Cdc42-GEF activity. This is demonstrated for SH3A-C2 both in the in vitro nucleotide exchange assay ( Fig. 2A) and when monitoring microspike formation (Fig. 2C, upper right panel). As expected (and as will be discussed further below), mutating conserved tryptophan residues to serines within the SH3 domains (e.g. SH3A and SH3C, designated SH3A-C2(WS)) had no effect on Cdc42-GEF activity (see Fig. 2A).
Comparisons of the Abilities of Different Forms of Intersectin-L to Induce Cellular Transformation-Given the distinct capabilities exhibited by the different forms of intersectin-L to promote the activation of Cdc42, we were interested in whether they showed the same relative abilities to induce the transformation of fibroblasts. Cell lines stably expressing each of the intersectin-L constructs were generated and analyzed for their ability to grow under different conditions (Western blots of the different constructs are shown in Fig. 3A, lower panel; H and L denote different cell lines exhibiting higher or lower stable expression, respectively, of a particular construct). The upper panel in Fig. 3A shows the results of cell growth assays per- H and L refer to cell lines that exhibit higher and lower levels, respectively, of the stable expression of the particular protein of interest. B, appearance of the different NIH 3T3 clones described in A after being cultured in low serum for 6 days. C, saturation density assays. The indicated stable cell lines were cultured in DMEM supplemented with 10% calf serum for 6 days, trypsinized, and counted. The data represent the average of three independent experiments. D, soft agar assays. Cells stably expressing the indicated constructs were mixed with DMEM supplemented with 0.3% agarose and 10% calf serum and plated on top of DMEM supplemented with 0.5% agarose and 10% calf serum. Colonies were scored after 14 days of growth. The data shown are the average of three independent experiments. 100% represents 500 cells counted. E, photomicrographs of colony formation in soft agar, as described in D (magnification is ϫ70). formed in low serum. As reported previously, NIH 3T3 cells expressing the constitutively active, fast-cycling Cdc42(F28L) mutant were extremely effective at growing in low serum (7), whereas control fibroblasts were unable to grow under these conditions. Fig. 3B demonstrates the striking difference in morphology between cells expressing Cdc42(F28L) versus the vector control in low serum conditions. However, surprisingly, we found that cells expressing SH3A-C2 were significantly more effective in their ability to grow in low serum and exhibited a more pronounced transformed appearance compared with cells expressing DH-C2, despite the fact that the latter construct was much better at stimulating Cdc42 activation.
Similar results were obtained when examining the relative ability of the intersectin-L constructs to enable fibroblasts to increase their saturation density and to form colonies in soft agar (Fig. 3, C-E). In each case, SH3A-C2 elicited a stronger transforming phenotype than DH-C2, with the extent of transformation being dependent on the relative levels of expression of SH3A-C2. However, interestingly, SH3A-C2 was even more effective than the Cdc42(F28L) mutant in stimulating colony formation in soft agar (Fig. 3D). The latter assay has been regarded typically as the most reliable indicator of malignant transformation. Given that SH3A-C2 was only moderately effective at stimulating Cdc42 activation in cells (i.e. compared with DH-C2), this then raised the question of how it was able to potently transform fibroblasts.
What Are the Downstream Signals Necessary for SH3A-C2induced Cellular Transformation?-We had shown previously that an essential aspect of Cdc42(F28L)-induced cellular transformation is the sustained signaling by EGFRs because of activated Cdc42 interfering with EGFR-Cbl interactions (17). However, whereas the stable expression of Cdc42(F28L) in NIH 3T3 cells resulted in EGF-stimulated ERK activation that was still detected after 2 h (Fig. 4A, top panel), cells expressing SH3A-C2 showed the transient profile for EGF-stimulated ERK activation typical of control cells, such that the stimulation of ERK activity was no longer detectable within 15-30 min of treatment with the growth factor (Fig. 4A, middle panel). Moreover, although Cdc42(F28L)-induced cellular transformation, as assayed by the ability of fibroblasts to grow in low serum, was completely blocked by inhibiting ERK activation with the MEK inhibitor PD98059, SH3A-C2-stimulated growth in low serum was unaffected by treatment with PD98059 (Fig.  4B). These results clearly demonstrated that transformation by SH3A-C2 proceeded through a mechanism distinct from that responsible for cellular transformation by the fast-cycling Cdc42 mutant.
We then examined the potential involvement of other downstream signaling activities. As shown in Fig. 5A (top panel), both Cdc42(F28L) and SH3A-C2 augmented EGF-stimulated PI3K activity, as assayed by the phosphorylation of Akt. Fig.  5B shows that Cdc42(F28L)-induced growth in low serum was in fact inhibited by the PI3K inhibitor, LY294002. Likewise, the ability of SH3A-C2 to promote cell growth under these conditions required PI3K activation.
Both Cdc42(F28L) and SH3A-C2 also enhanced the EGFstimulation of JNK 1 and 2 (Fig. 5A, middle panel). However, it was interesting that although Cdc42(F28L)-induced transformation was not dependent on JNK activation, SH3A-C2-mediated transformation was completely blocked when cells were treated with the JNK inhibitor, SP600125 (Fig. 5C). Thus, these results further indicate that distinct signaling pathways contribute to the ability of constitutively active Cdc42, versus the intersectin-L construct SH3A-C2, to transform cells.
SH3A-C2 Promotes the Activation of Ras-It has been reported that the short form of intersectin (intersectin-S), which lacks the DH and PH domains and is incapable of acting as a Cdc42-GEF, is able to bind through one of its SH3 domains (SH3A) to the Ras-GEF, Sos (24,26). Thus, we were interested in determining whether intersectin-L was capable of forming a complex with Sos and promoting the activation of Ras. Indeed, the results presented in Fig. 6A (top panel) demonstrate that endogenous Sos1 can be co-immunoprecipitated with Myctagged SH3A-C2 from cells. We then examined the ability of Sos to bind to a double point-mutated SH3A-C2 construct, in which the conserved tryptophan residues in the first and third SH3 domains (i.e. Trp-777 in SH3A and Trp-1031 in SH3C) were changed to serines (this construct was designated SH3A-C2(WS)). SH3 domains A and C were mutated because these regions have been most often implicated in intersectin-binding interactions (25,27). As shown in Fig. 6A, SH3A-C2(WS) was completely ineffective in binding to Sos1, but it was still able to bind to N-WASP (neuronal-enriched Wiscott-Aldrich Syndrome protein) (Fig. 6B, top panel), a Cdc42-specific target that has also been shown to interact with intersectin-L through its SH3 domains and to promote its Cdc42-GEF activity (22). As already indicated in Fig. 2A, the SH3A-C2(WS) mutant was also capable of stimulating the [ 3 H]GDP-GTP␥S exchange activity of Cdc42 to the same extent as SH3A-C2.
We then examined whether SH3A-C2 was capable of stimulating Ras activation in cells, as this might contribute to its transformation activity. Ras activation was detected by using a GST pull-down assay where the limit-binding domain from the Ras-target Raf was fused to GST (GST-RBD) (28). Fig. 6C (top  panel) shows that the treatment of cells expressing HA-tagged (wild-type) Ras with EGF promoted the cellular activation of HA- Ras (compare lanes 3 and 4). Cells expressing SH3A-C2 were also able to effectively stimulate the activation of HA-Ras, even in the absence of EGF treatment (Fig. 6C, lane 1), as were cells expressing the SH3A-C2(E1237A) mutant (lane 5), which is defective for Cdc42-GEF activity. However, SH3A-C2(WS) FIG. 4. Effects of SH3A-C2 on EGF-stimulated ERK activity. A, time course for the EGF-dependent activation of ERK in NIH 3T3 cells expressing Cdc42(F28L), SH3A-C2, and vector control. Following serum starvation overnight, cells were stimulated with EGF (100 ng/ml) for the indicated times. Anti-phospho-p44/42 ERK antibody was used to detect activated p44/42 ERK. B, growth profiles (1% serum) for cells stably expressing the indicated constructs in the presence or absence of the ERK inhibitor (PD98059, 5 m). Data represent the average of three independent experiments. was completely ineffective (Fig. 6C, compare lanes 1 and 2). These results then raised the possibility that the ability of SH3A-C2 to induce cellular transformation relied on its ability to activate Ras as well as Cdc42.
Both Ras and Cdc42 Are Necessary for SH3A-C2-induced Cellular Transformation-The results presented in Fig. 7 show that mutations blocking either Ras or Cdc42 activation are sufficient to inhibit completely SH3A-C2-induced cellular transformation. For example, both the Cdc42-GEF-defective mutant SH3A-C2(E1237A) and the double point mutant SH3A-C2(WS), which is unable to bind Sos and stimulate Ras activation, were ineffective in stimulating the growth of NIH 3T3 fibroblasts in low serum (Fig. 7A) and did not induce a transformed morphology (Fig. 7B). Likewise, these mutants were unable to increase saturation density (Fig. 7C) or allow fibroblasts to form colonies in soft agar (Fig. 7, D and E).
Given the indications that SH3A-C2 needed to activate both Ras and Cdc42 to transform cells, it was of interest to examine which of these GTP-binding proteins was responsible for signaling to JNK and PI3K, as both are essential for transformation (Fig. 5, B and C). As shown in Fig. 8, the ability of SH3A-C2 to activate Ras was necessary for enhancing EGF-stimulated PI3K (top panel) and JNK (second panel from top) activity, as both of these EGF-stimulated activities were significantly reduced in cells expressing SH3A-C2(WS) compared with cells expressing SH3A-C2. Apparently, the SH3A-C2-mediated activation of Cdc42 was only necessary for achieving full activation of PI3K, because maximal EGF-stimulated activation of JNK still occurred in cells expressing SH3A-C2(E1237A). DISCUSSION A number of reports have implicated members of the Rho subfamily of the Ras-related GTP-binding proteins in malig-nant transformation (10 -13). To further probe the role of Cdc42 in cell growth regulation, we examined the cellular consequences of expressing a Cdc42 mutant that was capable of constitutive GDP-GTP exchange but also was able to hydrolyze GTP back to GDP (this was referred to as a fast-cycling mutant). Unlike dominant-active, GTPase-defective forms of Cdc42, which we have found to mainly inhibit cell growth, fast-cycling Cdc42 mutants induce the transformation of NIH 3T3 fibroblasts and are particularly effective in stimulating the formation of colonies in soft agar (7)(8)(9). Recently, we uncovered an important clue regarding the underlying mechanisms responsible for these Cdc42-induced transformation phenotypes (17); namely, the constitutive activation of Cdc42 results in the negative regulation of EGFR interactions with the E3 ligase, Cbl, and thereby blocks receptor ubiquitination and degradation. The resulting accumulation of EGFRs gives rise to excessive mitogenic signaling and cellular transformation.
Given these findings, we might predict that the aberrant expression or regulation of Cdc42-GEFs would also induce cellular transformation by altering the normal down-regulation of EGFRs. We were particularly interested in examining such a possibility using the Cdc42-specific GEF, intersectin-L, because a shorter version of this protein, intersectin-S, is implicated in endocytic events and has some capability for transforming fibroblasts (23,29). We suspected that the deregulation of the Cdc42-GEF activity of intersectin-L would result in the hyperactivation of Cdc42, as well as the potential disruption of EGFR endocytosis, and could thus be especially potent in stimulating cellular transformation. Although indeed it turned out that a truncated version of intersectin-L lacking the EH domains (SH3A-C2) was very effective at transforming fibroblasts, as indicated by its ability to stimulate colony for- mation in soft agar and to promote growth in low serum, the mechanism responsible for its transforming activity was not what we had originally expected and, in fact, turned out to be rather interesting and novel.
Particularly surprising was our finding that the ability of different intersectin-L constructs to transform NIH 3T3 cells did not directly correlate with their ability to promote the activation of Cdc42. For example, although an intersectin-L construct comprising just the DH, PH, and C2 domains (DH-C2) was very effective in stimulating the GDP-GTP exchange activity of Cdc42 and promoting Cdc42 activation in cells, it was rather ineffective in stimulating the transformation of fibroblasts. On the other hand, SH3A-C2, an intersectin-L construct containing all five SH3 domains together with the DH, PH, and C2 domains, was only capable of modest Cdc42-GEF activity both in vitro and in cells, but it was extremely effective in inducing the transformation of fibroblasts as assayed by colony formation in soft agar. Moreover, although the underlying mechanism responsible for transformation by constitutively active, fast-cycling Cdc42 mutants was the accumulation of EGFRs and prolonged EGF-coupled signaling, this was not the case for SH3A-C2-induced transformation. Cells expressing this intersectin-L construct did not show increased levels of EGFRs or a sustained EGF-coupled signaling to ERK. Nonetheless, SH3A-C2-induced transformation was dependent on the Cdc42-GEF activity of SH3A-C2, as mutating the DH domain and blocking this activity completely inhibited its transforming activity. Thus, although it was necessary that SH3A-C2 increased the levels of GTP-bound Cdc42 to stimulate colony formation in soft agar, apparently the amounts of activated Cdc42 generated in SH3A-C2-expressing cells were not sufficient to adequately sequester Cbl and reduce EGFR degradation. This then suggested that the SH3A-C2-induced activation of Cdc42, together with another SH3A-C2-promoted activity, was necessary to give rise to the potent stimulation of growth in soft agar.
One interesting possibility regarding an additional SH3A-C2-promoted activity seemed to be Ras, as previous studies have shown that intersectin-S, which lacks the DH and PH domains and thereby is ineffective in promoting the activation of Cdc42, is capable of binding via its SH3 domains to Sos (27). Likewise, we found that SH3A-C2 was capable of forming a stable complex with Sos and promoting the activation of Ras in cells, whereas Ras activation was blocked by mutating conserved tryptophan residues in the first and third SH3 domains (i.e. SH3A and SH3C). It is worth noting that mutating the conserved tryptophan residues within these domains did not inhibit the binding of WASP, a well known interaction partner of intersectin-L that has been shown to help activate the Cdc42-GEF activity of the full-length intersectin protein, ap- parently by reversing an autoinhibitory intramolecular interaction between the SH3 and DH domains (22). The mutations of the conserved tryptophan residues in SH3A and SH3C also did not have any effect on the ability of SH3A-C2 to activate Cdc42. However, importantly, mutations that prevented SH3A-C2 from functionally coupling to Sos and Ras also completely blocked the transforming activity of this intersectin-L construct.
Thus, intersectin-L appears to be capable of coordinating the activation of both Ras and Cdc42, the former occurring through interactions with Sos and the latter being a result of directly functioning as a specific GEF for Cdc42. Presumably, the actions of intersectin-L are tightly regulated, such that its ability to activate Ras-and Cdc42-coupled pathways occurs when the proper growth-promoting signals are received. Apparently, the SH3A-C2 construct is only partially effective in its ability to FIG. 7. Mutations within the DH or SH3 domains of SH3A-C2 abolish its ability to transform NIH 3T3 fibroblasts. A, NIH 3T3 cells stably expressing the indicated constructs were cultured in low serum. Cell numbers were counted on days 2, 4, and 6. Data represent the average of three independent experiments. The lower images show Western blots comparing the relative expression in whole cell lysates of the indicated constructs. H and L refer to cell lines that exhibit higher and lower levels, respectively, of the stable expression of the particular protein of interest. B, photomicrograph of the different stable cell lines cultured in low serum for 6 days. C, saturation densities of different stable cell lines cultured in DMEM supplemented with 10% calf serum for 6 days. Data are the average of three independent experiments. D, soft agar assays for the indicated stable cell lines. Data are the average of three independent experiments. E, photomicrograph of the colonies formed in soft agar. Original magnification is ϫ70.
FIG. 8. Mutations within the DH or SH3 domains of SH3A-C2 affect its ability to activate PI3K and JNK. NIH 3T3 cells stably expressing the indicated constructs were cultured in low serum. Following serum starvation overnight, the cells were stimulated with EGF (100 ng/ml) for 5 min or were untreated, lysed, and Western blotted with phospho-Akt or phospho-JNK antibodies.
activate Cdc42 (at least when compared with a construct that lacks the SH3 domains (i.e. DH-C2)); still, this construct is nonetheless able to generate sufficient amounts of activated Cdc42 as well as activated Ras via interactions with Sos, so that these GTP-binding proteins can work together to elicit a transforming signal (Fig. 9). As discussed above, excessive signaling to ERK is not responsible for SH3A-C2-induced cellular transformation. Although SH3A-C2 at best only weakly activates JNK and PI3K, both of these signals are essential for transforming activity. It is also worth noting that EGF-stimulated JNK and PI3K activities are reduced in cells expressing the SH3A-C2(WS) mutant that is defective for binding Sos and activating Ras. This suggests that some portion of the EGFmediated stimulation of JNK and PI3K occurs via endogenous intersectin (24), such that the Ras activation-defective SH3A-C2(WS) construct has a dominant-negative inhibitory effect on these signals.
Overall, these findings suggest a novel and somewhat unexpected mechanism by which Ras and Cdc42 cooperate to derail the normal control of cell growth and raise some interesting implications regarding other situations where it has been observed that Ras transformation relies on the activation of Cdc42 (16). A puzzling question that emerges from these studies is why DH-C2, which effectively activates Cdc42 in cells, does not mimic the actions of the constitutively active Cdc42(F28L) mutant and give rise to cellular transformation. The answer may lie in the apparent differences in the cellular localization of DH-C2 versus Cdc42(F28L) or SH3A-C2. We have found that the predominant location of both Cdc42(F28L) and SH3A-C2 is perinuclear, whereas DH-C2 is diffusely distributed throughout the cell. 2 Thus, it may be that at least some of the Cdc42-signaling activities necessary for transformation need to originate from the Golgi (30). Future studies will be directed toward addressing this and a number of other interesting questions that now arise from these findings, such as how full-length intersectin-L is regulated normally so that excessive signaling through Ras or improper cooperation between Ras and Cdc42 does not occur.