Regulation of Sos Activity by Intramolecular Interactions

Abstract

The guanine nucleotide exchange factor Sos mediates the coupling of receptor tyrosine kinases to Ras activation. To investigate the mechanisms that control Sos activity, we have analyzed the contribution of various domains to its catalytic activity. Using human Sos1 (hSos1) truncation mutants, we show that Sos proteins lacking either the amino or the carboxyl terminus domain, or both, display a guanine nucleotide exchange activity that is significantly higher compared with that of the full-length protein. These results demonstrate that both the amino and the carboxyl terminus domains of Sos are involved in the negative regulation of its catalytic activity. Furthermore, in vitro Ras binding experiments suggest that the amino and carboxyl terminus domains exert negative allosteric control on the interaction of the Sos catalytic domain with Ras. The guanine nucleotide exchange activity of hSos1 was not augmented by growth factor stimulation, indicating that Sos activity is constitutively maintained in a downregulated state. Deletion of both the amino and the carboxyl terminus domains was sufficient to activate the transforming potential of Sos. These findings suggest a novel negative regulatory role for the amino terminus domain of Sos and indicate a cooperation between the amino and the carboxyl terminus domains in the regulation of Sos activity.

The Ras exchange factor Sos is critically involved in the coupling of growth factor receptors to Ras-dependent mitogenic signaling pathways (32). Mammalian cells contain two closely related and ubiquitously expressed Sos genes, Sos1 and Sos2. Their protein products consist of several defined domains each mediating a distinct function. The amino terminus domain of Sos is approximately 600 amino acids long and contains regions of homology to Dbl (DH) and pleckstrin (PH) domains. PH and DH domains are commonly found in signal transducting proteins, and several lines of evidence indicate that these domains are critical for their biological activity (4, 21, 26, 38). PH domains present in Sos proteins have been implicated in the regulation of their guanine nucleotide exchange activity (20, 24, 35) and ligand-dependent membrane targeting (9). The function of the DH domain of Sos is presently unknown. The catalytic activity of Sos is mediated by a central domain of approximately 420 amino acids that is highly conserved among different Ras exchange factors (3). The carboxyl terminus domain of Sos proteins is characterized by the presence of multiple proline-rich SH3 binding sites which mediate the interaction with the adaptor molecule Grb2 (5, 7, 17, 22, 23, 29).

The predominant mechanism by which Ras proteins are activated following receptor tyrosine kinase stimulation involves an increase in the rate of Sos-mediated guanine nucleotide exchange on Ras. This increase does not reflect the enhancement of the catalytic activity of Sos, as indicated by the observation that the guanine nucleotide exchange activity of Sos is not altered by growth factor stimulation (5, 18). Rather, it appears that the activation of Ras is achieved through the growth factor-dependent recruitment of Sos-Grb2 complexes to the activated receptor. This translocation event presumably serves to increase the local concentration of Sos in the plasma membrane where Ras is located. Another aspect of Sos regulation is represented by the growth-factor-induced phosphorylation of serine residues within its carboxyl terminus domain (11, 14, 29). This phosphorylation is mediated primarily by ERK mitogen-activated protein (MAP) kinase and results in the dissociation of the Grb2-Sos complex (10, 15, 37). The physiological significance of Sos phosphorylation remains to be determined, although it has been proposed that the phosphorylation-dependent disassembly of the Grb2-Sos complex might contribute to the downmodulation of Sos activity (36, 37).

In the present study, we sought to identify mechanisms that control the catalytic activity of Sos. We demonstrate that Sos truncation mutants lacking either the amino or the carboxyl terminus domain, or both, display an exchange activity that is significantly higher compared with that of the full-length protein. These results indicate that both the amino and carboxyl terminus domains of Sos impose constraints on the catalytic activity of Sos.

MATERIALS AND METHODS

Plasmids and expression vectors.

The amino acids corresponding to each human Sos1 (hSos1) construct are numbered as follows: hSos1, 1 to 1333; NCat, 1 to 1047; Cat, 601 to 1047; CatC, 601 to 1333; N, 1 to 614; and C, 1014 to 1333. hSos1 constructs were cloned into the mammalian expression vector pCGN (gift from Dr. M. Tanaka, Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y.). This vector contains the cytomegalovirus promoter and multicloning sites that allow the expression of genes fused 3′ to the hemagglutinin (HA) epitope. The glutathione S-transferase (GST)-HRas complex contains wild-type HRas (HRasWT) inserted into pGEX-2T (Pharmacia). HRasWT, HRasV12, and ERK2 were obtained by inserting the corresponding cDNAs into the pDCR mammalian expression vector. Plasmids pGADGH, pGBT10, pGADGH-SNF1, and pGBT9-SNF4, and the yeast strain YPB2 were described previously (7).

Cell culture and transfection assays.

COS1 cells were maintained in Dulbecco’s minimal essential medium (DMEM) (Gibco-BRL) supplemented with 5% fetal calf serum (FCS). Transfections of COS1 cells were performed with calcium phosphate as described by Wigler et al. (39). Briefly, DNA was diluted in Tris-EDTA buffer and 2M CaCl₂ was added to a final concentration of 125 mM. The mixture was then diluted in HEPES-Na₃PO₄ buffer and allowed to stand at room temperature for 30 min prior to being added to the culture medium. Twelve hours after transfection, the medium was aspirated and cultures were fed with DMEM supplemented with 5% FCS for 24 h and then serum starved in DMEM without FCS for 24 h before being harvested.

Yeast two-hybrid assays.

Yeast cells carrying a GAL4-LacZ reporter cassette were transformed with pairs of recombinant plasmids, one of which expressed a protein fused to the DNA binding domain of GAL4 and the other expressed a protein fused with the transcriptional activation domain of GAL4. LacZ expression was determined by using a paper filter assay as described previously (7).

Transformation assays.

NIH 3T3 cells were grown in DMEM supplemented with 10% calf serum. DNA transfections were done as described above with 0.2 μg of HRasV12, 1 μg of HRasWT, and 1 μg of each hSos1 construct. Transfected cells were maintained in DMEM supplemented with 5% calf serum, and the medium was changed every 3 days. Transformed foci were stained with Giemsa and quantitated after 14 to 16 days. The efficiency of transfection was measured by β-galactosidase assay (31).

Antibodies.

The anti-HA mouse monoclonal antibody 12CA5 (Babco) was used for immunoblotting and immunoprecipitation. The monoclonal antibody Y13-259 was produced from rat hybridoma Y13-259 cells. The antibody was precipitated with 45% (wt/vol) ammonium sulfate, and the pellet was resuspended in phosphate-buffered saline (PBS) and then dialyzed against PBS to yield a 2-mg/ml stock solution of antibody. Rabbit anti-rat immunoglobulin antibody was from Cappel.

Ras activation assay.

Cells were rinsed with phosphate-free DMEM and incubated in 4 ml of phosphate-free DMEM supplemented with 4 mCi of ³²P (Du Pont-NEN) for 3 to 4 h prior to harvest. At the end of the labeling period, the cells were washed three times with ice-cold PBS and lysed in 1 ml of cold buffer containing 150 mM NaCl, 10 mM Tris-HCl (pH 7.4), 1% Triton X-114, 1% deoxycholic acid, 0.1% sodium dodecyl sulfate (SDS), 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 mg of pepstatin ml, 50 mM NaF, 1% aprotinin, 10 μg of leupeptin ml, 1 mM sodium vanadate, 10 mM benzamidine, 10 μg of soybean trypsin inhibitor/ml, and 1 μM okadaic acid. Immunoprecipitations were done for 40 min with 3 μg per sample of the rat monoclonal anti-Ras antibody Y13-259 precoupled to protein A-Sepharose beads via rabbit anti-rat immunoglobulin antibody. Duplicate samples in which the Y13-259 antibody was omitted were used as controls. The immune complexes were collected by centrifugation and washed eight times with a buffer containing 50 mM HEPES (pH 7.4), 500 mM NaCl, 5 mM MgCl₂, 0.1% Triton X-100, and 0.005% SDS. Bound nucleotides were eluted with 2 mM EDTA, 2 mM dithiothreitol, and 0.2% SDS for 20 min at 68°C. Nucleotides were separated by thin-layer chromatography on polyethylenimine-cellulose plates in 0.75 M KH₂PO₄, pH 3.5. Plates were imaged and quantified with a PhosphorImager (Molecular Dynamics). A factor of 1/2 was applied to the counts obtained from GDP and a factor of 1/3 was applied to those obtained from GTP as a correction for their respective phosphate contents. Radioactivity in GTP or GDP from control samples did not exceed 10% of that measured in experimental samples.

ERK2 immune complex kinase assay.

COS1 cells were transiently cotransfected with HA-tagged hSos1 truncation mutants and HA-ERK2. After incubation for 24 h in serum-free DMEM, cells were lysed in immunoprecipitation (IP) buffer containing 150 mM NaCl, 10 mM Tris-HCl (pH 7.4), 1% Triton X-100, 10% glycerol, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 mg of pepstatin/ml, 50 mM NaF, 1% aprotinin, 10 μg of leupeptin/ml, 1 mM sodium vanadate, 10 mM benzamidine, 10 μg of soybean trypsin inhibitor/ml, and 1 μM okadaic acid. Fifty micrograms of protein from each lysate was electrophoresed on SDS–12.5% polyacrylamide gels and transferred onto nitrocellulose. Immunoblot analysis of the epitope-tagged transiently expressed proteins was carried out with anti-HA antibody by using enhanced chemiluminescence reagents. ERK2 was immunoprecipitated from lysates by using anti-HA antibody. The immune complexes were collected by centrifugation for 15 s at 1,000 × g and washed three times with IP buffer and twice with kinase buffer (25 mM Tris [pH 7.4] 20 mM MgCl₂, 2 mM MnCl₂, 1 mM Na₃VO₄, and 20 μM ATP). ERK2 kinase activity was assayed in 50 μl of kinase buffer containing 10 μCi of [γ-³²P]ATP and 0.2 mg of myelin basic protein (MBP)/ml. Reaction products were analyzed by SDS–7.5% polyacrylamide gel electrophoresis (PAGE) (2) and quantitated with a PhosphorImager.

In vitro binding assay.

GST fusion proteins were expressed in Escherichia coli by induction with 0.5 mM of isopropyl-1-thio-β-d-galactopyranoside (IPTG). The expressed GST fusion proteins were isolated from bacterial lysates by affinity chromatography with glutathione agarose beads for 1 h at 4°C. Nucleotide-free GST-HRasWT was prepared by incubating the beads in 20 mM Tris (pH 7.4)–1 mM dithiothreitol–5 mM EDTA for 30 min at 4°C (19). GST or GST-HRasWT beads were washed twice in PBS and twice in IP buffer containing 5 mM EDTA to ensure that the GST-HRasWT protein remained free of guanine nucleotides. Two micrograms of GST or GST-HRasWT was incubated for 1 h with COS1 cell lysates transfected with the different hSos1 constructs. Beads were pelleted and washed three times with IP buffer. Protein complexes were eluted with SDS sample buffer, separated by SDS–15% PAGE, and transferred to nitrocellulose paper. Bound Sos proteins were detected by immunoblotting with anti-HA antibody.

RESULTS

Effects of hSos1 truncation mutants on Ras signaling.

To characterize the roles of different domains of Sos in the regulation of its activity, we have generated a series of expression plasmids encoding for HA-tagged hSos1 truncation mutants (Fig. 1A). When transfected into COS1 cells, each of the expression plasmids gave rise to a polypeptide of the expected molecular weight, as determined by immunoblotting with anti-HA antibodies (Fig. 1B). In addition, all the hSos1 mutants were found to be soluble and predominantly localized to the cytosol as determined by subcellular fractionation experiments (not shown).

FIG. 1.

Description of hSos1 truncation mutants. (A) Schematic representation of hSos1 constructs. Shading denotes noncatalytic regions of hSos1. All constructs contained an amino-terminal HA epitope tag. (B) Expression of hSos1 truncation mutants. COS1 cells were transfected with expression vectors encoding HA-tagged hSos1 constructs. Cell lysates were separated on SDS–7.5% PAGE gels and subsequently analyzed by immunoblotting with anti-HA antibody.

To investigate the ability of the hSos1 mutants to activate Ras, we first compared their effects on ERK MAP kinase activation. COS1 cells were cotransfected with the hSos1 constructs and an expression vector encoding HA-tagged ERK2. Transfection conditions were adjusted to yield similar levels of expression of the various hSos1 constructs. ERK2 was immunoprecipitated with anti-HA antibody, and its kinase activity was measured by immunocomplex kinase assay using MBP as a substrate (13). As illustrated in Fig. 2, expression of the Cat domain induced a marked stimulation of ERK MAP kinase (10-fold). This stimulatory effect was abolished by the coexpression of the dominant inhibitory Ras mutant RasN17, indicating that the Cat-induced activation of ERK MAP kinase is mediated by Ras (not shown). In comparison, activation of ERK MAP kinase by full-length hSos1, NCat, and CatC was three- to fivefold lower, suggesting that both the amino and carboxyl terminus domains of hSos1 negatively regulate its activity. It should be noted that in a different study, it has been reported that a Cat construct derived from Drosophila Sos failed to activate Ras (24). The high degree of homology between human and Drosophila Sos proteins makes it unlikely that the apparent discrepancy between the two studies is due to structural variations between the proteins. It is possible that differences between the catalytic domain borders, as defined by the two studies, contribute to the observed differences in their respective activities.

FIG. 2.

Activation of ERK MAP kinase by hSos1 truncation mutants. COS1 cells were transiently cotransfected with HA-tagged ERK2 (0.5 μg) and the indicated HA-tagged hSos1 constructs. ERK2 activation was measured in serum-starved cells by immune complex kinase assay using MBP as a substrate. (Upper) MBP phosphorylation as visualized by autoradiography. The amount of ³²P incorporated into MBP was determined by phosphorimaging and is indicated under each lane in arbitrary units. Results shown are from a single representative experiment. Experiments were repeated three times with similar results. (Lower) Immunoblot with anti-HA antibody showing the level of expression of ERK2.

Ras binding activity of hSos1 truncation mutants.

To further investigate the functional properties of the various hSos1 domains, we compared the abilities of the different hSos1 truncation mutants to bind in vitro to HRasWT. Lysates prepared from COS1 cells expressing similar amounts of each truncation mutant (Fig. 3B) were incubated with nucleotide-free GST-HRasWT coupled to glutathione agarose beads. At the end of the incubation, the beads were washed and bound proteins were eluted and analyzed by Western blotting. All hSos1 constructs bound specifically to nucleotide-free HRasWT. This is expected based on the well-documented ability of guanine nucleotide exchange factors to bind tightly to the nucleotide-free form of their corresponding GTPases. However, the binding of the Cat domain to HRasWT was significantly stronger (∼50-fold increase) compared with full-length hSos1, NCat, and CatC (Fig. 3A). These results are consistent with the relative effectiveness of the hSos1 truncation mutants that we have observed with respect to the activation of ERK MAP kinase and provide further evidence that the amino and carboxyl terminus domains of hSos1 exert negative allosteric control on the interactions of the catalytic domain of hSos1 with Ras.

FIG. 3.

In vitro binding assay of hSos1 constructs to GST-HRasWT. (A) COS1 cells were cotransfected with the indicated hSos1 constructs. Following transfection, cells were grown in DMEM supplemented with 5% FCS for 24 h and serum starved for 24 h. Cell lysates were incubated with 2 μg of GST or 2 μg of nucleotide-free GST-HRasWT protein for 1 h at 4°C. Bound proteins were eluted in sample buffer and detected by immunoblotting with anti-HA antibody. (B) A proportion of cell lysates was separated on an SDS–7.5% polyacrylamide gel and was immunoblotted with anti-HA antibody.

Guanine nucleotide exchange activity of hSos1 truncation mutants.

Next, we examined the catalytic activity of the hSos1 mutants by measuring their effects on the state of Ras activation in vivo. To this end, COS1 cells were transiently transfected with the HA-tagged hSos1 constructs. Transfected cells were serum starved and then labeled with [³²P]orthophosphate, and Ras proteins were immunoprecipitated by using the monoclonal antibody Y13-259. The guanine nucleotides associated with Ras were resolved by thin-layer chromatography (Fig. 4A) and quantitated by phosphorimaging (Fig. 4B). In cells transfected with vector alone, we detected a very weak guanine nucleotide binding activity. This indicated that the level of endogenous Ras proteins in COS1 cells is too low to be measured and quantitated under the assay conditions that we used. To circumvent this problem cells were cotransfected with the hSos1 constructs and HRasWT. To ensure that only fully processed Ras molecules were being assayed, Ras immunoprecipitation was carried out in the Triton X-114-soluble fraction as previously described (6). This fractionation procedure has been shown to permit the selective isolation of fully processed Ras molecules.

FIG. 4.

Activation of Ras by hSos1 truncation mutants. (A) COS1 cells were cotransfected with HRasWT (0.2 μg) and the indicated hSos1 truncation mutants. Serum-starved cells were labeled with [³²P]orthophosphate, and the guanine nucleotide content of HRasWT was analyzed as described in Materials and Methods. GTP and GDP markers are labeled on the right. (B) Quantitation of GTP/(GDP + GTP) percentages. Results are the averages of four independent experiments. Error bars represent the standard deviations.

As illustrated in Fig. 4A and B, the expression of full-length hSos1 induced a twofold activation of HRasWT from 8 to 20% Ras-GTP. In comparison, hSos1 constructs lacking the amino or carboxyl terminus or both domains (CatC, NCat, and Cat, respectively) were twice as efficient in promoting guanine nucleotide exchange on HRasWT, leading to the accumulation of ∼40% of Ras-GTP. These results again indicate that both the amino and carboxyl terminus domains of hSos1 are involved in regulating its catalytic activity. It is important to note that while the NCat and CatC constructs were fully active in stimulating exchange on Ras, these constructs displayed a downregulated activity when tested for ERK MAP kinase activation (Fig. 2). These differences most likely reflect the differences between the assay conditions used in these experiments. The assay for the activation of ERK MAP kinase by the hSos1 mutants relied on endogenous Ras proteins, whereas the Ras activation assay employed ectopically expressed Ras. The increased levels of Ras expression in the latter assay might have obscured some of the affinity differences between the various hSos1 constructs. This interpretation is supported by our observations that in the presence of ectopically expressed Ras, the relative activation of ERK MAP kinase by the hSos1 constructs was in complete agreement with their ability to stimulate guanine nucleotide exchange on Ras (not shown).

The observations that under conditions of ectopic expression of Ras (Fig. 4) removal of either the amino or the carboxyl terminus domain of hSos1 is sufficient to cause an upregulation of its activity suggest that these domains might exert their inhibitory effect in trans. To test this possibility, we examined the ability of the amino-terminal and the carboxyl-terminal regions of Sos to interact with each other in the two-hybrid system. Pairwise combinations of the carboxyl terminus domain of Sos (CSos) fused to the GAL4 transcriptional activation domain and the amino terminus domain of Sos (NSos) or Grb2, as a positive control, fused to the GAL4 DNA binding domain were coexpressed in the yeast reporter strain YPB2, and their interaction was assessed by the induction of β-galactosidase. As illustrated in Fig. 5, whereas coexpression of Grb2 and CSos gave rise to a significant induction of β-galactosidase, no induction was detected when CSos and NSos were coexpressed, suggesting that the carboxyl-terminal and the amino-terminal domains of Sos do not interact directly.

FIG. 5.

Two-hybrid analysis of the interaction between the amino and carboxyl terminus domains of Sos. Yeast cells carrying a GAL4-LacZ reporter cassette were cotransformed with pairs of plasmids expressing proteins fused to the GAL4 binding domain or the GAL4 activation domain as indicated. The S. cerevisiae SNF1 and SNF4 fusion proteins were used as specificity controls. The interaction between the two fusion proteins is indicated by the induction of LacZ expression (dark color). Each patch represents an independent transformant.

Regulation of hSos1 catalytic activity by growth factors.

To gain insights into the significance of the negative regulation of hSos1 activity by its amino and carboxyl terminus domains, we asked whether growth factor stimulation could alleviate this autoinhibition. In COS1 cells coexpressing the epidermal growth factor (EGF) receptor and HRasWT, the increase in exchange activity following EGF stimulation was similar to that induced by the full-length hSos1 under serum-starved conditions (compare Fig. 6a with Fig. 4B). To ensure that in this setting Grb2 and Sos are not limiting, COS1 cells were cotransfected with expression plasmids for the EGF receptor, Grb2, hSos1, and HRasWT. For these experiments, transfection conditions were adjusted such that in the absence of growth factors the level of Ras-GTP remained at a basal state. As can be seen, the extent of EGF-induced Ras activation was not enhanced under these conditions (Fig. 6b). This did not reflect a specific property of the EGF signaling system, because the same level of activation was observed upon serum stimulation (Fig. 6c). Thus it appears that the catalytic activity of hSos1 is constitutively maintained in a downregulated state.

FIG. 6.

Regulation of hSos1 activity by growth factor stimulation. (a) COS1 cells were cotransfected with 0.2 μg of HRasWT and 0.2 μg of EGF receptor. Cells were serum starved, labelled with [³²P]orthophosphate and stimulated with 20 nM of EGF for 0, 1, 5, and 20 min. (b) COS1 cells were cotransfected with 0.2 μg of HRasWT, 0.2 μg of EGF receptor (EGFR); 0.2 μg of Grb2, and 0.005 μg of hSos1. Transfected cells were serum starved, labelled, and stimulated with EGF as described for panel a. (c) COS1 cells were transfected with 0.2 μg of HRasWT, 0.2 μg of Grb2, and 0.005 μg of hSos1. Transfected cells were serum starved, labelled with [³²P]orthophosphate, and serum stimulated for 0 and 5 min. Guanine nucleotide content of HRasWT was analyzed as described in Materials and Methods. GTP and GDP markers are labeled on the left.

Transforming activity of hSos1 truncation mutants.

We investigated the physiological importance of maintaining hSos1 activity under negative control by comparing the different hSos1 constructs with respect to their ability to synergize with HRasWT in the induction of cellular transformation. NIH 3T3 cells were cotransfected with the hSos1 constructs and HRasWT, and the formation of transformed foci was determined as described previously (12). As a positive control, the activated form of HRas, HRasV12, was used. In the presence of HRasWT, only the catalytic domain of hSos1 (Cat) was highly efficient in inducing cellular transformation, exhibiting focus-forming activity that is only two-fold lower than that induced by oncogenic HRasV12 (Fig. 7A and B). Moreover, cells transformed by Cat and HRasWT were morphologically indistinguishable from cells transformed by oncogenic Ras (Fig. 7C). The same results were obtained even when cells were transfected with higher amounts of hSos1 mutant constructs. None of the hSos1 mutants displayed transforming activity in the absence of HRasWT. Since the relative effectiveness of the hSos1 constructs in inducing NIH 3T3 transformation well correlates with their ability to activate endogenous Ras proteins (Fig. 2), we conclude that the transforming effects of Cat are mediated by both the ectopically expressed and endogenous Ras proteins.

FIG. 7.

Focus-forming activity of hSos1 truncation mutants. (A) NIH 3T3 cells were transfected with 1 μg of HRasWT alone, 0.2 μg of HRasV12 as a positive control, and 1 μg of HRasWT and either 1 μg of hSos1 or 1 μg of the Cat domain. After 14 days, the dishes were stained with Giemsa to visualize the transformed foci. (B) Quantitation of the focus formation assay performed with all of the hSos1 constructs. The data are averages of three dishes and are representative of three independent assays. (C) The morphological appearance of NIH 3T3 cells transfected with vector alone, HRasV12, or HRasWT and the Cat domain.

DISCUSSION

The use of intramolecular interactions as a mechanism for regulating the activity of molecules is widespread in biological systems. Among the best understood examples for this type of interaction is the regulation of the enzymatic activity of Src kinases by intramolecular interactions involving the SH2 and SH3 domains (25, 33, 40). In the present study we provide evidence implicating such a mechanism in the negative regulation of the guanine nucleotide exchange activity of Sos towards Ras. We demonstrate that the catalytic domain of Sos is negatively regulated by its amino and carboxyl terminus domains. Although the molecular determinants that might be involved in the autoinhibitory control of Sos activity remain to be determined, our observations that truncation of both the amino and carboxyl terminus domains lead to an increase in Sos catalytic and Ras binding activities suggest that both domains are involved in the downmodulation of Sos activity. Consistent with this idea are the findings that in Saccharomyces cerevisiae the catalytic domain of Sos could fully substitute for the Ras exchanger CDC25, while the full-length Sos was very weak in complementing CDC25 activity (reference 1 and unpublished data).

The role of the carboxyl terminus domain of Sos in the negative control of Sos activity has been already reported (1, 20, 24, 35). The carboxyl terminus domain of Sos is phosphorylated by MAP kinase on multiple sites following growth factor stimulation (8, 11, 14, 30). The phosphorylation of these sites does not appear to be functionally relevant for the negative regulatory role of the carboxyl terminus domain of Sos since a phosphorylation-deficient mutant of Sos (15) displayed the same exchange activity as the wild-type protein (15a). Since Sos truncation mutants lacking either the amino or the carboxyl terminal domain display reduced Ras binding affinity and catalytic activity in comparison with the catalytic domain, it is tempting to speculate that each domain may contribute independently to the negative regulation of Sos. This possibility is further supported by our findings that the amino and carboxyl terminal domains of Sos failed to interact with each other directly.

It has been argued that membrane localization of Sos is critical for its exchange activity towards Ras. The main evidence in support of this notion is that the Sos proteins that are constitutively targeted to the plasma membrane are potent activators of the Ras pathway, whereas their nontargeted counterparts are virtually inactive (1, 27). Our finding that the catalytic domain of hSos1 is extremely efficient in catalyzing guanine nucleotide exchange on Ras indicates that stable association of Sos with the membrane is not essential for its activity and that transient interaction between Sos and Ras might be sufficient for its catalytic activity in vivo. In this context, it should be emphasized that our findings do not refute earlier models for Sos activation which invoke the membrane recruitment of Sos as a critical regulatory event. Under conditions where endogenous Ras or Sos could be limiting with respect to either amounts or accessibility, the stable association of Sos with the membrane, directly and/or via interactions with activated receptors, might be essential for the efficient stimulation of exchange on Ras.

Many studies on Sos function and regulation have employed transient-transfection protocols in which Sos activity was analyzed with respect to ectopically expressed Ras. Our observations that the results obtained through assays involving endogenous Ras are quantitatively different from those obtained with ectopically expressed Ras underscore the importance of using both experimental conditions to define mechanisms that control Sos activity.

From a regulatory standpoint, the downregulated state of the full-length Sos can be viewed in two ways. It can represent a basal state which is upregulated upon signal. Alternatively, it may represent a constitutive state which serves the physiological needs of the cells. Our findings support the latter possibility in that only a fraction of the full catalytic potential of Sos appears to contributes to growth-factor-mediated Ras activation. The moderate levels of Ras activation generally detected in response to growth factor stimulation, typically not more than 20 to 25% GTP loading irrespective of receptor type or abundance, lend further support to the idea that Sos is constitutively maintained in the autoinhibited state. We have shown that, by uncovering the full potential of the catalytic activity of Sos through the removal of autoinhibitory domains, the protein acquires the ability to transform cells. In this context, it is of interest to note that oncogenic activation of Dbl family proteins which function as exchange factors for Rho GTPases also involves N-terminal truncation that presumably results in the upregulation of their catalytic activity (38). This along with the recent demonstration that quantitative differences in the extent of growth-factor-induced Ras activation can lead to qualitative differences in the signaling events that it triggers (16, 28, 34) underscores the physiological importance of tightly controlled exchange mechanisms. The identification of sequences that contribute to the repression of Sos activity should further the understanding of its signaling function within the Ras pathway.