INTRODUCTION

Ras, the archetype of low molecular weight GTP-binding proteins, first was identified in mammalian cells through gain-of-function mutations in human tumors, corresponding to the transforming genes of Harvey and Kirsten murine sarcoma viruses (1, 2, 3). Ras subsequently was reported to function, more generally, as a molecular switch for signaling by growth factor receptors with tyrosine kinase domains (2), by certain heptahelical receptors (4), and, arguably, by receptor serine/threonine kinases (5) although the latter conclusion is disputed (6, 7, 8). Highly conserved during evolution, Ras is regarded as mediating diverse functions beyond proliferation (2, 9, 10), such as mating and sporulation in the fission yeast Schizosaccharomyces pombe, vulval development in Caenorhabditis elegans, photoreceptor determination in Drosophila, mesoderm induction in Xenopus laevis (5, 8), neuronal differentiation in pheochromocytoma cells (11, 12), and repression of myogenic differentiation (13). Despite numerous advances in identifying events that mediate Ras signal transduction and remarkable similarities among the downstream cascades common both to vertebrates and much simpler eukaryotes (1, 2, 3, 10), the exact mechanisms by which Ras exerts any given effect still are not completely understood. Activated Ras recruits Raf to the surface membrane, initiating a protein kinase cascade via mitogen-activated protein/extracellular signal-related kinase kinase and mitogen-activated protein kinase (10, 14); a second cascade leads directly from Ras to Jun N-terminal kinases (15), and a less characterized third pathway also exists for Ras-dependent phosphorylation of transcription factors (16). Ras also physically associates with mitogen-activated protein kinase kinase kinase (17) and with distinct other classes of proteins, such as the candidate effector phosphatidylinositol 3-kinase (PI-3-K)¹ (18), the Ral guanine nucleotide dissociation stimulator (19), and the Ras-GTPase activating protein, Ras-GAP (20), which is reported both to antagonize (21) and to mediate Ras effects (22, 23). Ultimately, Ras-dependent signals are relayed via multiple transcription factors including Fos/Jun (15, 16, 24), Ets/Elk proteins (25, 26), cyclic AMP response element-binding protein (27), and serum response factor (28). Accordingly, a wide array of growth-associated promoters are regarded as Ras-responsive, yet more generalized effects of Ras, including control over nominally constitutive viral and cellular promoters, also have been reported (29, 30).

Cardiac myocytes possess growth properties inherently different from most other lineages (31). Like other terminally differentiated cell types, ventricular muscle cells exit the proliferative cell cycle soon after birth, with little or no capacity for subsequent cell division; normal post-natal growth and adaptive growth in response to work load both proceed, instead, by cell enlargement (hypertrophy), with an increase in DNA content per cell (increased ploidy or karyokinesis, uncoupled from cytokinesis). Characteristically, cardiac hypertrophy is accompanied by induction of a "fetal" program of transcription, that is preferential expression of genes ordinarily associated with the embryonic ventricle. GTP occupancy on Ras is induced in cardiac myocytes by passive stretch and angiotensin II, consistent with a role in hypertrophy induced by these stimuli (32, 33). Hence, it has been of interest to determine by functional tests whether Ras-dependent pathways might mediate the signals that culminate in hypertrophic growth, the "fetal" program, or both. Microinjection of dominant-negative Ha-Ras protein was reported to block ₁-adrenergic induction of both morphological changes in myofibrillar structure and expression of atrial natriuretic factor, a member of the fetal gene ensemble (34). Conversely, an activated Ha-ras gene, targeted to myocardium in transgenic mice, elicited ventricular enlargement, atrial natriuretic factor expression, myofibrillar disarray, and impaired relaxation in diastole (35). However, despite this striking evidence for an aggregate effect of Ras activation in vivo, with many features similar to human cardiac hypertrophy, the exact circuit leading to ventricular growth can be obscured in any chronic model: are atrial natriuretic factor and hypertrophy itself relatively direct consequences of Ras or, instead, triggered as adaptations to the impaired mechanical performance and a resulting increase in wall stress? Indeed, evidence for an indirect mechanism is the occurrence of left atrial enlargement in this model, although the transgene is not expressed in atria, and absence of right ventricular enlargement, although the transgene is expressed in both ventricles (35). An analogous mechanism contingent on wall stress has been inferred to account even for familial cardiac hypertrophy caused by myosin heavy chain mutations (36).

Our own studies of Ras in cardiac signal transduction were prompted by the general question of whether Ras participates in signaling by type  transforming growth factor receptors, archetypes for the superfamily of receptors with serine/threonine kinase domains. TGF-dependent expression of skeletal -actin and plasminogen activator inhibitor-1 reporter genes in cardiac myocytes and non-cardiac cells, respectively, was suppressed by dominant-inhibitory S17N Ha-Ras (6). However, basal expression of these promoters in the absence of growth factors decreased in parallel, with identical dose-response relations. Thus, no indications were found for the hypothesis that TGF might signal through Ras. Moreover, dominant-negative, wild-type, or activated Ras cDNAs regulated all seven reporter genes tested, including those driven by nominally constitutive viral control regions, the Rous sarcoma virus promoter, and minimal herpes simplex virus thymidine kinase (HSV tk) promoter, and even a TATA-less adeno-associated virus initiator element (6). These broadly inclusive effects suggested the interpretation that Ras activity regulates an unexpectedly generalized or global set of genes or, conceivably, the basic transcriptional machinery (6).

Ras signaling activity is modulated by both positive and negative regulators (1, 2, 3). Guanine nucleotide releasing factor (GNRF) enhances the exchange of bound GDP for GTP, converting Ras to its active form, which in turn recruits physical association of an array of effector proteins. The dominant-negative mutation of Ras, S17N, has much higher affinity for GDP than GTP, is inactive with respect to signaling, and inhibits Ras-dependent pathways for proliferation and transformation at least in part by competing for and sequestering GNRF (37, 38). Ras mutations elsewhere, within the domain for binding effector proteins, have proven useful in addressing the requirement for the effector domain in different cellular functions: for example, amino acid substitutions that are unable to bind Raf-1 protein, such as D38N, also have reduced transforming activity (39, 40). Conversely, Ras GTPase-activating proteins (Ras-GAP, the NF1 gene product, neurofibromin, and GAP1^m) provoke the intrinsic catalytic activity of the Ras GTPase domain, cause hydrolysis of bound GTP, and restore Ras protein to its inactive GDP-bound form (3, 41). The effector domain mutation P34R discriminates between Raf and GAP, as it abolishes GAP but not Raf binding to Ras (42); accordingly, this form of Ras has diminished GTPase activity, increased occupancy by GTP, and augmented transforming activity, in accordance with the negative modulatory role envisioned for GAP. However, given the diversity of Ras effects and Ras-associated proteins, the extrapolation from transforming activity to transcriptional control is uncertain, particularly in the novel context of Ras as a potential governor of global gene expression.

This uncertainty is highlighted by dual actions proposed for Ras-GAP, first identified as a protein that suppresses transformation by Ras, yet later was shown to possess effector functions. Ras-GAP binds the Ras effector domain, in competition with a canonical effector such as Raf, and mimics oncogenic Ras in its inhibitory effect on K⁺ channel opening in isolated atrial cells (22). The non-catalytic, N-terminal Src homology (SH2-SH3) domain of GAP mediates this action (43) and mimics the impact of Ras on induction of the c-fos promoter (23, 44); other effector actions for this N-terminal region of GAP have been described (44, 45, 46). How these two opposing functions of GAP relate, and which predominates in a given biological context, is largely unknown.

In the present study, we have attempted to delineate the Ras-dependent mechanisms that govern the expression of both hypertrophy-associated and constitutive genes in cardiac muscle, utilizing mutations of Ras that are well-characterized with respect to transformation and to protein-protein interactions, and utilizing forced expression of full-length Ras-GAP versus GAP deletion mutants. Together, these investigations indicate that Ras regulates gene expression through differing downstream effectors contingent on cell type and that an effector function of Ras-GAP predominates in the cardiac background.

MATERIALS AND METHODS

Wild-type, S17N, and G12R human c-Ha-Ras-1 cDNAs, provided by L. A. Feig (47), were subcloned as XbaI-BamHI fragments into the eukaryotic expression vector, pSV-Sport-1 (Life Technologies, Inc.), as detailed previously (6). C186S mutations of Ras were engineered by site-directed mutagenesis using the polymerase chain reaction, as described previously (6). Replacement of cysteine 186 with serine renders S17N Ras proteins deficient for farnesylation and membrane localization (47).

The P34R, D38N, and E63K missense mutations of Ras were constructed using the polymerase chain reaction, using wild-type, S17N, or G12R Ras cDNAs in pSV-Sport-1 as template. P34R converts base 101 from C to G, and amino acid 34 from proline to arginine; D38N converts base 112 from G to A, and amino acid 38 from aspartic acid to asparagine; E63K converts base 187 from G to A, and amino acid 63 from glutamic acid to lysine. The initial polymerase chain reactions used primers harboring each mutation as the 3 primers; the 5 primer corresponded to the Sp6 promoter located 5 to the cloning site of the Ras expression vector. The resulting products were used as 5 primers for the second reaction, in concert with a 3 primer corresponding to the T7 promoter, 3 to the cloning site of the expression vector, to generate full-length Ras cDNAs that carry the designated mutations. The engineered Ras cDNAs were then subcloned into pSV-Sport-1.

Full-length Ras-GAP, the C-terminal catalytic domain (amino acids 705-1047), and N-terminal SH2-SH3 domain (amino acids 1-666), provided by Drs. C. J. Der and G. J. Clark (48), were subcloned from pZipNeo as BamHI fragments into pSV-Sport-1.

The skeletal -actin-luciferase reporter construct (49) comprises nucleotides 394 to +24 of the chicken skeletal -actin promoter, cloned between the SmaI and HindIII sites of the firefly luciferase reporter expression vector pXP1 (50). Nucleotides 105 to +51 of the HSV tk promoter (51) were subcloned into pXP2. The plasminogen activator inhibitor-1 (PAI-1) luciferase construct, containing nucleotides 800 to +75 of the human PAI-1 promoter, was obtained from D. J. Loskutoff (52).

Cardiac myocytes were isolated from 1- to 2-day-old Sprague-Dawley rats, purified by density centrifugation through a Percoll step gradient as previously described, and further depleted of non-muscle cells by preplating for 1 h (6). Myocytes were plated at a density of 1 × 10⁶ cells per 35-mm dish (Primaria, Falcon) and were cultured overnight in Dulbecco's modified Eagle's medium/Ham's nutrient mixture F12, 1:1, 17 mM HEPES, 3 mM NaHCO₃, 2 mM L-glutamine, 50 µg·ml¹ gentamicin, 5% horse serum. Under the conditions used here for preparation of the cardiac myocyte cultures, contamination by non-muscle cells, assessed by immunostaining for sarcomeric myosin heavy chains, is routinely less than 3%. Cells were transfected 24 h after plating by Lipofectamine following the manufacturer's recommendations (Life Technologies, Inc.), using 0.25 µg of the indicated luciferase reporter construct, and varying amounts of the SV40-driven expression vectors. For all comparisons, DNA and promoter content were kept constant using the identical SV40 vector, without an insert. Cells were incubated with the DNA-lipofectamine complex for 6 h. Cardiac myocytes were cultured overnight in the medium described above, which was then replaced by serum-free medium containing 5 µg·ml¹ transferrin, 1 nM Na₂SeO₄, 1 nM LiCl, and 25 µg·ml¹ ascorbic acid (49, 53, 54); insulin was omitted to obviate potential effects of the interventions on insulin- and insulin-like growth factor-I-dependent pathways. Mv1Lu mink lung epithelial cells were cultured as described (55) and were transfected using 150 µg/ml DEAE-dextran, 3.7 µM chloroquine, 0.5 µg of reporter, and the indicated concentrations of expression vector. Twenty-four hours after transfection, medium was replaced with the same serum-free medium described for cardiac myocytes.

For Western analysis of the Ras expression vectors, human 293 cells were transfected using a 5-h incubation with 20 µl of lipofectamine and 2 µg of plasmid per 60-mm dish; 293 cells are transfected at ~100% efficiency under these conditions versus ~2% for cardiac myocytes. Cells were lysed 48 h after transfection in 200 µl per 60-mm dish of RIPA buffer (130 mM NaCl, 15 mM Na₂HPO₄, 15 mM NaH₂PO₄, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate) plus protease inhibitors (0.7 µg/ml pepstatin, 2 µg/ml aprotinin, 0.5 µg/ml leupeptin, 1 mg/ml pefabloc; Boehringer Mannheim), and 50 µg of protein per lane was then loaded on a 10% sodium dodecyl sulfate-polyacrylamide gel. After electrophoresis, the proteins were then transferred electrophoretically to a 2-µm nylon membrane (Bio-Rad), using 0.1 M CAPS containing 10% methanol. The filter was rinsed with buffer D (500 mM NaCl, 20 mM Tris-HCl, pH 7.5, 0.3% Tween 20) and incubated for 1 h at room temperature with the primary antibody (polyclonal rabbit IgG reactive against human, mouse, and rat Ras, UBI 06-226), at a concentration of 0.1 µg/ml in buffer D plus 0.2% fatty acid-free bovine serum albumin. The membrane then was washed for 30 min at room temperature with 3 × 50 ml of buffer D containing 3% Tween 20. The filter was then incubated for 1 h at room temperature with the secondary antibody (horseradish peroxidase-conjugated donkey antibody against rabbit IgG, Amersham NA 934) at a concentration of 1:10,000 in buffer D plus 0.2% fatty acid-free bovine serum albumin and then was washed as above. Bound antibody was detected by chemiluminescence, using enhanced chemiluminescence reagents as directed (Amersham Corp.). For Western analysis of endogenous Ras-GAP, cardiac myocytes and Mv1Lu cells maintained in culture under the same conditions used for their transfection were lysed in RIPA buffer plus the protease inhibitors detailed above. GAP protein was detected by Western blotting using 0.2 µg/ml rabbit polyclonal antibody against amino acids 171-448 of human GAP (UBI 06-157).

Ras protein was immunoprecipitated from cardiac myocytes and Mv1Lu cells cultured under the same conditions used for transfection. Cells were lysed with RIPA buffer plus protease inhibitors as above. For each cell type, 0.5 mg of protein was incubated for 12 h at 4 °C with gentle agitation using 50 µl of Y13-259 rat monoclonal antibody against v-Ha-Ras, conjugated to agarose (1 mg of IgG/0.5 ml of agarose (Santa Cruz SC-35 A)). Agarose-bound immune complexes were precipitated by centrifugation and washed with 3 × 1 ml of RIPA buffer. The agarose was then resuspended in Laemmli loading buffer and boiled for 10 min to dissociate bound proteins. The agarose was precipitated by centrifugation and the supernatant analyzed by Western blotting for Ras protein as detailed above.

Cells were harvested and analyzed 48 h after transfection, as detailed previously (6, 49). Co-transfection with a constitutive lacZ gene was omitted because the discordant half-life of luciferase compared with LacZ can give rise to misleading interpretations (6). Luciferase activity was monitored as the oxidation of luciferin in the presence of coenzyme A (56), using an Analytical Luminescence model 2010 luminometer. Results (mean ± S.E.) are expressed relative to the activity of each promoter in parallel cultures of vector-transfected cells and were compared by analysis of variance using Scheffe's test and a significance level of p < 0.05.

RESULTS

Mutagenesis of the CAAX farnesylation motif disrupts the effects of S17N and G12R Ras on gene expression in both cardiac myocytes and Mv1Lu cells. To determine the mechanism for control of transcription by Ras, we generated a series of double mutations, affecting regions of the dominant-negative and activated Ras proteins that have been well-characterized with respect to transforming activity or protein-protein association. We first tested the requirement for C-terminal acylation, which mediates the targeting of Ras proteins to the surface membrane, using Mv1Lu mink lung epithelial cells, as we previously examined the farnesylation motif only in ventricular myocytes (6). The increase in PAI-1 promoter activity by G12R Ras and repression of promoter activity by S17N Ras were each abrogated by the C186S mutation of the CAAX motif (p = 0.0001 for each; Fig. 1, A and B). Thus, the G12R, S17N, and C186S mutations (which affect intrinsic GTPase activity, affinity for GDP versus GTP, and a post-translational modification, respectively) bear equally on transcriptional signaling by Ras in both cell backgrounds we have tested.

To test the prediction that the E63K substitution, which impairs the binding of guanine nucleotide exchange factors to Ras (37), would at least partially reverse the inhibition of gene expression by dominant-negative Ras, we next generated the S17N/E63K double mutation. At low concentrations of the Ras expression vectors, sufficient for 50% inhibition of the truncated HSV tk promoter by S17N Ras, no effect whatever was seen with S17N/E63K Ras (Fig. 1C; p = 0.0001). An inhibitory effect of the double mutation was detectable only at a concentration 10-fold or more higher that that required for repression by S17N Ras itself. These findings concur with the importance of E63 for Ras activation (37) and are consistent with the supposition that S17N Ras acts as a dominant-interfering protein by competing with endogenous wild-type Ras for GNRF. These data also confirm our prior observation that the nominally constitutive HSV tk promoter is suppressed under these conditions by S17N Ras to at least the same degree as the growth factor-inducible PAI-1 promoter (Fig. 1B) (6). Investigations of S17N/E63K Ras were not undertaken in cardiac muscle cells, where insufficient inhibition by S17N Ras was obtained at the low plasmid concentrations needed to discriminate between the single and double mutation.

The D38N mutation is known to abolish the binding of Raf and PI-3-K to Ras and inhibit the transforming capacity of the activated Ras (18, 39, 40). To test the effect of this substitution on gene induction by Ras, we engineered the double mutant G12R/D38N. In ventricular myocytes, up-regulation of both the skeletal -actin and HSV tk reporter genes by G12R Ras was abolished by the concomitant mutation at Asp-38 (p = 0.0001 for each; Fig. 2, A and B); analogous results were obtained for activation of PAI-1 and HSV tk promoters in Mv1Lu cells (p = 0.0001 for each; Fig. 2C, D). Conversely, the D38N mutation had little or no effect on repression by S17N Ras (not shown), consistent with the reported dependence of the dominant-negative mutation, instead, on sequestration of nucleotide exchange factors. These results thus confirm the necessity for an intact Raf/PI-3-K binding site for activation by Ras of both "inducible" and "constitutive" reporter genes in both cell backgrounds.

As a substitution of the effector domain more selective than D38N, the P34R mutation is refractory to inhibition by Ras-GAP and NF1, and, hence, favors the active, GTP-bound state (42). Although this mutation recently was shown to act as a transforming Ras protein, its potential consequences for transcription have not previously been explored. In complete concordance with effects of the G12R Ras mutation, whose intrinsic GTPase activity is defective, P34R Ras was 2-3-fold more active than wild-type Ras in Mv1Lu cells, increasing the transcription of both the PAI-1 and HSV tk promoters, with effects at least as great on the nominally constitutive viral gene (p = 0.0003 and 0.0001, respectively; Fig. 3, C and D). Remarkably, however, in cardiac myocytes, the P34R mutation abrogated the activation by Ras of reporter genes driven by the skeletal -actin and HSV tk promoters (p = 0.0002 and 0.0001, respectively; Fig. 3, A and B). Thus, a mutation of the Ras-GAP site has dichotomous effects, depending on cell background: the P34R substitution augments gene induction by wild-type Ras in Mv1Lu cells, yet prevents gene induction by wild-type Ras in cardiac myocytes. Provisionally, these results therefore suggest an effector function for Ras-GAP in ventricular myocytes but not Mv1Lu cells. The P34R mutation has been reported previously to destabilize G12R Ras (42), and, hence, no attempt was made to determine its effect on the activity of the constitutively active Ras protein.

Given that cell type-specific effects were confined to P34R, the one Ras mutation that is defective for GAP binding, studies with exogenous GAP proteins were performed to assess the prediction that GAP itself might function differently depending on cell background (Fig. 4). In Mv1Lu cells, the catalytic domain of Ras-GAP suppressed the HSV tk and PAI-1 reporter genes by up to 80% (p = 0.0001 for each versus vector-transfected cells), whereas neither full-length GAP nor the N-terminal domain produced a significant change (Fig. 4, D and E). As anticipated, inhibition by cGAP was reversed by co-introduction of G12R Ras, a mutation defective for GTPase activity but not by co-introduction of exogenous wild-type Ras (not shown). All three GAP constructs evoked antithetical effects in ventricular myocytes, concordant with the respective activities of P34R Ras in the two cell backgrounds: up to 9-fold activation of the HSV tk and skeletal -actin reporter genes by full-length GAP and nGAP (p = 0.0001 for each versus vector-transfected cells), with no significant change elicited by the catalytic domain (Fig. 4B, C). Pilot studies utilizing the catalytic domain of neurofibromin, the NF1 gene product, yielded similar overall conclusions, dose-dependent inhibition of the HSV tk reporter in Mv1Lu cells but little if any effect in the ventricular myocytes (not shown).

Western blot analysis confirmed the expression of the mutant proteins at comparable levels, albeit in a differing cell type in which the exogenous proteins could be detected, as well as changes in migration on sodium dodecyl sulfate-polyacrylamide gels reported previously for several of these substitutions (Fig. 5A). As differences in stoichiometry also might conceivably explain the dichotomous effects of Ras-GAP seen with differing cell backgrounds, Western blot analysis was performed to determine levels of endogenous Ras-GAP, and endogenous Ras was visualized by Western blotting after immunoprecipitation. p21^ras was detected as a doublet as seen in prior studies, ascribed to differences in electrophoretic mobility of processed and unprocessed Ras proteins (see e.g. Ref. 57). Only minor differences were seen between cell types in measurable Ras and Ras-GAP (Fig. 5B). As both were slightly more abundant in ventricular myocytes than in Mv1Lu cells, even less of a difference is expected between cell types in the relative levels of Ras versus Ras-GAP. Although the amino acid sequences of Ras and Ras-GAP are unknown for the mink, and levels of immunoreactive protein therefore should be interpreted cautiously, this short-coming is offset by the exceptional conservation of Ras and Ras-GAP within mammalian species, taken together with the use of polyclonal sera here to minimize susceptibility to changes at a single epitope. GTP occupancy on Ras likewise was similar for the ventricular myocytes and Mv1Lu cells (not shown), arguing against the level of endogenous Ras activation as the basis for the observed differences.

DISCUSSION

The inference that Ras activity is required for efficient expression of both growth factor-inducible and nominally constitutive genes was the conclusion of our prior investigations, implicating Ras as a governor of global gene expression in both cardiac myocytes and mink lung epithelial cells (6). As this action of Ras proteins was unexpected, and could not be reconciled with known effects of Ras-associated proteins, the present study tested the potential involvement of different Ras regulators, effectors, and functional domains in the control of global gene expression. With respect to growth inhibition (the principal end point previously characterized with double mutations of Ras) the form used here, S17N, functions by squelching exchange factors for Ras activation and, in turn, blocking activation of endogenous Ras; Ras residues 62, 63, 67, 69, and 75-78 mediate the interaction between these two proteins (37, 38). That transcriptional repression also was relieved by the double mutation S17N/E63K points to sequestration of the activator, GNRF, as a likely mechanism for inhibition of gene expression by S17N Ras here; the single and double mutations were expressed at comparable levels. Complete relief of inhibition was seen at the lowest doses of plasmid, with some degree of inhibition retained at higher doses, suggesting that affinity for binding GNRF is diminished, not abolished, by the E63K substitution. Alternatively, as no interaction between E63K and GNRF was measured in a yeast two-hybrid assay (37), it is conceivable that additional mechanisms underlie repression by S17N Ras at higher protein concentrations. Transcriptional activation by G12R Ras is not expected to be contingent on GNRF, as G12R, lacking GTPase activity, is predominantly GTP-bound, and the E63K mutation did not affect the constitutively activated Ras protein.

The array of proteins that bind the so-called Ras "effector" domain is expanding rapidly (1, 2, 3) but includes Raf (14), PI-3-K (18), GAP (20, 22), and Ral guanine nucleotide dissociation stimulators (19). Each interacts directly with Ras in a GTP-dependent fashion, although the functional significance of this interaction is not equally clear for all candidate effectors. Mutation of Asp-38 abrogates binding of Raf (39) and PI-3-K (18), impairs the functional interaction of GAP with Ras, and abolishes transformation by Ras (40). It is conjectural whether the control of global gene expression as measured here requires effectors identical to those for a more conventional end point, transformation by Ras. Hence, we also tested the potential impact of other effector domain mutations on gene induction by Ras. In both cell backgrounds (cardiac myocytes and Mv1Lu cells) this D38N substitution prevented Ras activation of the inducible and constitutive reporter genes equally. The D38N mutation does not discriminate between Raf and PI-3-K, however, and unidentified effectors also are a formal possibility. This distinction can potentially be addressed, in turn, by dominant inhibitors and constitutive forms of these two effectors themselves. Indeed, it has been proposed that Raf mediates gene activation in cardiac hypertrophy but not concurrent changes in myocyte morphology (58).

Dichotomous functions as both a negative modulator and an effector of Ras have previously been ascribed to the GTPase-activating protein, Ras-GAP, although inhibition of Ras-dependent signals is far more commonly observed (20, 22, 43, 59). How these opposing functions of GAP interplay in one cell type is poorly understood. Moreover, the functional significance of GAP as a Ras effector is uncertain, except in a context (coupling of muscarinic receptors to atrial K⁺ channels) that is neither representative nor necessarily predictive of other Ras-dependent events. For this reason, it was of interest to ascertain the transcriptional effects of P34R Ras, an effector domain mutation that selectively abolishes binding of Ras-GAP, but not Raf, to Ras, and generates a transforming Ras protein with high GTP occupancy (42). Notably, GAP is the only molecule identified thus far whose physical association to Ras is dependent on this site. Unlike all other Ras mutations we have investigated, which behave identically regardless of cell background, the P34R mutation enhances gene activation by Ras in Mv1Lu cells, with no discernible activity in cardiac myocytes. The loss of Ras-GAP binding implicates a predominantly negative effect of GAP on Ras in Mv1Lu cells, yet unmasks an apparent effector (or "effector-like") role of GAP in the cardiac muscle background. (By contrast to initial studies of D38N Ras, which suggested that binding of GAP to this mutation also was impaired, later evidence instead suggests a nearly normal affinity for binding but with failure to stimulate the intrinsic Ras GTPase (60).)

To substantiate this divergent role of GAP in the two cell types, and to exclude the contingency of a trivial, false-negative result in cardiac myocytes transfected with P34R, we systematically transfected each cell type, in turn, with full-length GAP, the C-terminal catalytic domain, and N-terminal SH2-SH3 domain. Results with all three GAP constructs confirmed the dichotomous role for GAP, as deduced from the divergent actions of P34R Ras. The catalytic domain, cGAP, previously has been shown to inhibit Ras-dependent transcription in NIH 3T3 cells (48), in agreement with suppressive effects seen here in Mv1Lu cells. cGAP is 20-fold less active than full-length GAP in stimulating the GTPase activity of Ras in vitro (61), which might contribute to the lack of discernible effect in cardiac cells. Antithetically, the net effect of nGAP and of full-length GAP in cardiac myocytes corresponded to that of activated Ras, in accordance with the ability of both GAP and nGAP to reproduce effects of Ras itself on atrial K⁺ channels (43), and consistent with the limited but provocative prior evidence for an effector role of the GAP SH2-SH3 domain in transcription (23, 44). By contrast to the latter studies in CHO9 and 3Y1 cells, moreover, it was unnecessary to remove the N-terminal Gly-Ala-Pro motif for gene induction in ventricular myocytes (23). GAP residues Arg-786, Lys-831, and Arg-925 are essential for binding to Ras·GTP (62); the isolated nGAP domain used here and by others (43, 48) lacks these amino acids, and the activity of nGAP is not thought to be contingent on physical association with Ras or on direct activation of Ras. In support of this inference, inhibition of cardiac K⁺ channels by GAP was reversed by anti-Ras antibody, whereas inhibition by nGAP was unaffected (43). The latter study proposes that the C-terminal Ras-binding catalytic domain interferes with the effector activity of the N-terminal Src homology domains and that binding of full-length GAP to Ras relieves this block, enabling the SH2-SH3 domain to interact with other proteins. Potential interference with endogenous GAP, causing a sustained increase in GTP-bound Ras, has been suggested as an alternative interpretation for the signaling activity of GAP Src homology domains (43), yet precisely concordant effects resulted in the present investigations using the P34R mutation of Ras itself. Hence, our provisional interpretation that the observed effects correspond to a bona fide effector function, and are not merely a phenocopy, draws strong support from this failure of P34R Ras to activate transcription in cardiac cells.

A direct corollary of the evidence for divergent effects of GAP contingent on cell background is the inference that Ras may employ different effectors, including GAP, for generation of Ras-dependent signals in a cell type-specific fashion. In part, the functional significance of this model is that differing Ras effectors must share certain actions in common (illustrated here by induction of the HSV tk promoter both through a GAP-dependent pathway in cardiac myocytes and through a GAP-independent pathway in non-muscle cells), yet may contribute to other, distinguishable effects that together confer the unique biological outcomes of Ras, dependent on cell context. Such a conclusion resembles, and extends, the consensus that precise consequences of Ras activation ultimately depend upon stage- and lineage-specific transcription factors present in a given cell (1, 2, 3). Conceivably, cell-specific utilization of Ras effector proteins could contribute to resolving a long-standing paradox, namely the divergent and even contradictory effects of Ras, e.g. enhanced DNA synthesis or transformation in most cell backgrounds, growth arrest and differentiation in neurons (11), dedifferentiation in skeletal muscle (13), and hypertrophy in cardiac myocytes (34, 35). Notably, GAP acts as a negative regulator of Ras signaling in PC12 cells, the other differentiated cell type for which such information is available (12). Loss-of-function mutations for GAP cause constitutive Ras activation in yeast and Drosophila, where GAP lacks the N-terminal Src homology domains (63, 64), and tumor predisposition (consistent with Ras activation) is the phenotype of mice that are heterozygous for disruption of NF1 (65). Interestingly, homozygous mice lacking NF1 die within 14 days of gestation, with hyperplasia in the sympathetic ganglia yet hypoplasia in ventricular muscle (65, 66).

The mechanism by which activated Ras allocates itself among potential effectors is not clear. Ras-GAP competes with Raf and PI-3-K for association with the effector domain of GTP-bound Ras, yet neither the relative abundance of GAP nor the apparent ratio of Ras to GAP provides a satisfactory explanation for differences between cell types in the present study; each was equivalent in the two cell types. Moreover, because the epithelial cells and cardiac myocytes both were growth-arrested under the conditions used, differences in the action of Ras-GAP and P34R Ras cannot be explained by mere differences in the cell cycle. A priori, potential mechanisms for the discrepancies related to Ras-GAP include relative proportions or modifications of Harvey-, Kirsten-, N-, and R-Ras; variable expression of proteins that compete for Ras effector domain (19); basal activity, affinity for Ras, subcellular localization, or phosphorylation of GAP; disparities in the GAP-associated protein p62 (67) or p190 (68); isoform differences in GAP, as shown for neurofibromin (69); and cross-talk with neurofibromin, Bcr, or other proteins homologous with Ras-GAP itself. A technical limitation of the present study is the meager efficiency for transfection into cardiac myocytes, less than 5% in our hands, which precludes our ability to compare the two lineages for GTP binding by Ras in the presence of exogenous GAP. No cardiac cell lines exist with fidelity to the ventricular phenotype, and the limited efficiency for transient transfection into ventricular myocytes presently obviates the use of several end points pertaining to signaling intermediaries or impact on endogenous genes.

By contrast to previous reports implicating Ras in the selective up-regulation of genes specifically associated with cardiac hypertrophy (34, 35), our investigations give consistent weight to a contrasting interpretation that the global increase in RNA and protein content, synonymous with hypertrophy, itself is contingent on Ras. The likely importance of endogenous Ras to hypertrophic growth, at least in myocardium, is suggested by the increased GTP occupancy on Ras activated by mechanical stretch or angiotensin II (32, 70), well-defined stimuli for cardiac hypertrophy, with rapid activation of Ras-dependent kinases (71). Whereas fetal gene induction has been ascribed to Raf and mitogen-activated protein kinase, neither was found to explain the changes in cell morphology provoked by hypertrophic signals (58, 72), which might reflect a role for Rho or Rac. Further evidence for dissociation between the fetal program and hypertrophic growth per se includes the ability of TGF to cause fetal cardiac gene expression without increasing protein content (54) and the reciprocal ability of rapamycin to block the increase in total protein and S6 kinase activity induced by angiotensin II, while sparing mitogen-activated protein kinase and fetal gene induction (73, 74). The Ras effector protein PI-3-K is reportedly sufficient and necessary for activation of p70^S6K (75); three other Ras effector proteins, Raf, mitogen-activated protein kinase kinase kinase, and JNK, impinge on Fos/Jun transcription factors that elicit fetal cardiac gene transcription (76, 77). Although negative modulation of Ras may outweigh other effects of Ras-GAP in most cell backgrounds, effector-like functions of Ras-GAP predominate in ventricular muscle cells, as measured by the activation of both growth factor-induced and constitutive reporter genes, and thus are credible as partial explanations for the global increase of RNA and protein synthesis in cardiac hypertrophy.

Whereas a formal possibility exists that certain effects of Ras or Ras-GAP mutants could result from levels of expression differing between cell types (indeed, there is a difference in the proportion of cells successfully transfected (1-3% for cardiac myocytes and ~25% for Mv1Lu cells)), several lines of evidence argue against this even as a contributory explanation for the observed results. First, wild-type Ras, dominant-negative Ras, constitutively activated Ras, the D38N effector mutation, and multiple double mutations of Ras (G12R/D38N, G12R/C186S, N17/C186S) each behave identically in the two cell backgrounds. Differences were exclusive to P34R. Thus, no difference in dose-response relations exists for any of seven other Ras proteins tested, arguing against mere disparities in the efficacy of Ras cDNA delivery. Second, relative to wild-type Ras, the P34R mutation confers a loss-of-function in cardiac myocytes but a gain-of-function in Mv1Lu cells. Thus, the potency of one Ras, relative to another, becomes systematically inverted on the basis of cell background. Third, full-length GAP and nGAP were highly active only in the cell type where transfection efficiency is less. Fourth, it is moot whether differences in transfection efficiency bear on levels of Ras or GAP expressed per cell, among those to which gene transfer was successful. Although the meager transfection efficiency in cardiac myocytes would confound a comparison among mutations or between cell types by Western blot analysis, adenoviral gene transfer is one means to overcome these inherent limitations. Given the efficacy for adenoviral gene transfer to cardiac muscle cells (78, 79), recombinant adenoviruses also could permit one to distinguish directly among mechanisms underlying the control of gene expression by Ras and Ras-GAP, even in primary culture, to achieve a molecular dissection of Ras-dependent responses and signaling events that would be infeasible at lower efficiencies for gene delivery.