Inactivation of raf-1 by a protein-tyrosine phosphatase stimulated by GTP and reconstituted by Galphai/o subunits.

A membrane-associated form of Raf-1 in v-Ras transformed NIH 3T3 cells can be inactivated by protein phosphatases regulated by GTP. Herein, a distinct protein-tyrosine phosphatase (PTPase) in membrane preparations from v-Ras transformed NIH 3T3 cells was found to be activated by guanyl-5′-yl imidodiphosphate (GMPPNP) and was identified as an effector for pertussis toxin (PTx)-sensitive G-protein α subunits. PTPase activation was blocked by prior treatment of cells with PTx. PTPase activation by GTP, but not GMPPNP, was transient. A GMPPNP-stimulated PTPase (PTPase-G) co-purified with G subunits during Superose 6 and Mono Q chromatography. PTPase-G activity in Superose 6 fractions from GDP-treated membranes was reconstituted by activated G, but not G, subunits. PTPase-G may contribute to GMPPNP-stimulated inactivation of Raf-1 in v-Ras cell membranes because Raf-1 inactivation was PTx-sensitive and PTPase-G inactivated exogenous Raf-1.

(reviewed in Ref. 4) or NIH 3T3 cells (5). Both tyrosine and serine phosphorylation of Raf-1 appear to be required for Raf-1 activation because Raf-1 isolated from Sf9 cells also expressing Ras alone 2 or Ras and Src Y527F (6) can be inactivated in vitro by treatment with either protein-serine/threonine or -tyrosine phosphatases.
Membranes from mammalian cells transformed with oncogenic Ras contain a portion of cellular Raf-1 in a constitutively active form (7). This membrane-associated form of Raf-1 can be inactivated by protein phosphatases regulated directly or indirectly by GTP (6). We performed experiments to characterize the protein phosphatase responsible for GTP-stimulated inactivation of Raf-1 and to define the mechanism of its regulation.

EXPERIMENTAL PROCEDURES
Materials-Guanine nucleotides were purchased from Boehringer Mannheim. NIH 3T3 cells transformed with Ha-Ras (G12V) were obtained from Dr. L. Feig (Tufts University, Boston, MA). Antibodies to G-protein subunits were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). PTx was a generous gift from Dr. E Hewlett (University of Virginia). The sources of other reagents have been described (4,6).
* This work was supported by the Howard Hughes Medical Institute, National Institutes of Health Grants DK41077 (to T. W. S.), DK-19952 (to J. C. G.), and GM35266 (to D. L. B.), and American Cancer Society Grant BE69D (to T. W. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ Postdoctoral fellow of the Juvenile Diabetes Foundation International.
‡ ‡ Supported by United States Public Health Service Training Grants DK07320 and DK07509.
Reactions were terminated by sequential additions of 200 l of cold 20% (w/v) trichloroacetic acid and 10 l of 10 mg/ml BSA with mixing. Samples were centrifuged, and 210-l supernatants were counted in scintillant. Data were expressed as percentage release of total 32 P in the assay.
Reversal of GMPPNP Activation by GDP-Portions (4 l) of Superose 6 fractions from GMPPNP-treated membranes were treated sequentially on ice by addition of 4 l of buffer A (15 min) to chelate magnesium and promote dissociation (9) of GMPPNP followed by addition of 4 l of either 6 mM GMPPNP or GDP containing 30 mM MgCl 2 (30 min) and then assayed for PTPase activity.
Reconstitution with G ␣i/o Subunits-A mixture of G ␣ proteins (principally G ␣i and G ␣o ) were purified from bovine brain and separated into G i/o , G ␣ , and G ␤␥ oligomers by phenyl-Sepharose chromatography essentially as described (10). The resolved G ␣i/o and G ␤␥ subunits were pooled separately and concentrated (50 g/ml) and were Ͼ85% pure. The G ␣i/o subunit preparation bound GTP␥S and was free of G ␤␥ as assessed by immunoblotting; the G ␤␥ preparation was free of G ␣ by the absence of GTP␥S binding. 4 Baculovirus-expressed G ␣ subunits were expressed and purified to near homogeneity essentially as described (11). Portions of each preparation were incubated on ice (15 min) with an equal volume of 12 mM guanine nucleotide, 60 mM MgCl 2 before use.

RESULTS AND DISCUSSION
Incubation of membranes from v-Ras cells with an antibody that binds the Ras effector domain did not prevent GMPPNPstimulated inactivation of membrane-associated Raf-1 (6), suggesting involvement of G-proteins other than Ras in Raf-1 inactivation (data not shown). To test for involvement of PTx-sensitive heterotrimeric G-proteins, we compared the abilities of guanine nucleotides to stimulate inactivation of endogenous Raf-1 in membranes from untreated v-Ras transformed cells in comparison with membranes from cells that were treated overnight with PTx (Table I). GTP or GMPPNP, but not GDP, stimulated inactivation of Raf-1 in membranes prepared from untreated cells, and inactivation was blocked by inclusion of microcystin-LR and vanadate to inhibit protein-serine/threonine and -tyrosine phosphatases. Raf-1 in membranes prepared from PTx-treated cells was not susceptible to guanine nucleotide-stimulated inactivation (Table I). PTx specifically catalyzes ADP-ribosylation of ␣ subunits of heterotrimeric G-proteins of the G i family (except G z ) (reviewed in Ref. 12). Thus, these results strongly suggested that GTP-stimulated inactivation of Raf-1 in the mixture of membranes from v-Ras cells was mediated by activation, directly or indirectly, of protein phosphatases by heterotrimeric G-proteins of the G i family.
Schally and co-workers (13) previously demonstrated that addition of somatostatin to membranes of pancreatic cancer cells promoted dephosphorylation of epidermal growth factor 4 M. A. Lindorfer and J. C. Garrison, unpublished data.  receptors autophosphorylated on tyrosine, implying activation of a PTPase by G-protein-coupled receptors. Stork and coworkers (14) extended this observation by demonstrating that GMPPNP addition to membranes could stimulate a PTPase activity, assayed using p-nitrophenyl phosphate as substrate, and that activation of PTPase activity by somatostatin was PTx-sensitive. Since Raf-1 deactivation was PTx-sensitive (Table I) and a purified PTPase can inactivate Raf-1 in vitro (6), we hypothesized that a G-protein-activated PTPase might be responsible for guanine nucleotide-stimulated inactivation of Raf-1.
To test this hypothesis, we determined whether loading of membranes isolated from v-Ras cells with GTP or GMPPNP stimulated PTPase activity, using as substrate [[ 32 P]Tyr]RCMlysozyme (Fig. 1). GTP significantly increased the rate of 32 P release in comparison to GDP (Fig. 1A). When membranes were reassayed after 30 min, the increase in PTPase activity due to GTP was absent (Fig. 1B, EOOE). Transient activation of PTPase activity by GTP is consistent with the transient activation of a G-protein ␣ subunit due to timed GTP hydrolysis (12). Reloading these same membranes with GTP partially recovered enhanced activity (Fig. 1B, E---E), proving that the decrease in activity was not due simply to protein lability. GMPPNP, which is not hydrolyzed by G ␣ subunits, caused a persistent elevation of PTPase activity (Fig. 1B). Activation also occurred when 100 M concentrations of guanine nucleotide were used (data not shown). Neither GTP nor GMPPNP caused an increase in PTPase activity in membranes from PTx-treated cells (data not shown). Together, these data demonstrate activation of a [[ 32 P]Tyr]RCM-lysozyme phosphatase by a PTx-sensitive G-protein.
In parallel experiments, GMPPNP or GTP stimulated PTPase activity in membranes isolated from parental NIH 3T3 cells with results similar to those in Fig. 1 (data not shown). Membranes from parental cells, however, contained 5-fold less GMPPNP-stimulated PTPase activity than membranes from v-Ras cells (data not shown). Membranes from cells transformed with a different oncogenic Ras, Ras Q61L , also showed significantly increased levels of GMPPNP-stimulated PTPase activity relative to membranes from parental cells. Thus, the greater specific activity of GMPPNP-stimulated PTPase is a consequence of Ras transformation.
The GMPPNP-stimulated PTPase was characterized by gel permeation chromatography in buffer containing 0.01% (v/v) Triton X-100 following solubilization of guanine nucleotidetreated membranes with 1% (v/v) detergent. Solubilization per se increased total PTPase activity 4 -5-fold, but a 2-3-fold stimulation of activity in GMPPNP-treated versus GDP-treated membranes was preserved (data not shown). The principal peak of PTPase activity observed with GDP-treated membranes from v-Ras cells eluted between the positions of the standards bovine ␥-globulin (150 kDa) and BSA (67 kDa) with an apparent mass of ϳ70 kDa ( Fig. 2A). Treatment of membranes with GMPPNP reproducibly caused the appearance of a peak of PTPase activity that eluted earlier, at ϳ100 kDa ( Fig.   FIG. 2. Characterization of PTPase-G by Superose 6 and Mono Q chromatographies. Membranes from v-Ras transformed NIH 3T3 cells were incubated with GMPPNP (å) or GDP (Ç), solubilized with Triton X-100, and fractionated for analyses (see "Experimental Proce- 2A). We refer herein to this GMPPNP-stimulated PTPase as PTPase-G. An additional species of GMPPNP-stimulated PTPase of higher M r (ϳ150,000) was occasionally observed as a smaller peak or leading shoulder (data not shown; see also Fig. 3B). The latter may correspond to a labile PTPase.
We tested whether PTPase-G could deactivate active Raf-1, using FLAG-Raf-1 purified from Sf9 cells co-expressing Ras and Src Y527F . A single Gaussian-shaped peak of activity causing deactivation of Raf-1 was detected and corresponded to PTPase-G (Fig. 2B). The deactivation of Raf-1 occurred in the presence of 2.4 M microcystin-LR and was abolished by 0.1 mM vanadate, consistent with the action of a PTPase (data not shown). Raf-1 deactivating activity was increased 5-7-fold by GMPPNP in comparison with GDP (data not shown). Thus, activation of PTPase-G can explain, at least in part, the ability of GMPPNP to cause inactivation of endogenous Raf-1 in membranes from v-Ras cells (6).
Our membrane preparations contain membranes derived from the endoplasmic reticulum in addition to plasma membranes and thus contain PTP1B (15). PTP1B was detected by immunoblotting (data not shown), and its elution corresponded to the major peak of [[ 32 P]Tyr]RCM-lysozyme phosphatase activity migrating with an apparent mass of ϳ70 kDa (centered at fraction 42) and not to the peak of GMPPNP-stimulated PTPase (centered at fraction 39). The peak of [[ 32 P]Tyr]RCMlysozyme phosphatase containing 50-kDa PTP1B (Fig. 2B) did not coincide with deactivation of exogenous Raf-1, suggesting that PTP1B is not PTPase-G. The COOH-truncated, 37-kDa form of PTP1B deactivated Raf-1 (6), but this may be due to its reactivity in vitro at high concentrations or the absence of regulatory sequences (15).
Using an antibody that recognizes G ␣i/o/t/z , we detected G ␣ subunit(s) in fractions containing GMPPNP-stimulated PTPase (Fig. 2C). Antibodies C-10 and K-20 to G ␣i1-3 and G ␣o also detected a protein of 41 kDa in these fractions (data not shown). No G ␤␥ subunits were detected in these fractions by immunoblotting with antibody T-20 that recognizes G ␤1-4 (data not shown). The profile of the elution of immunoreactive 41-kDa protein correlated well with the profile of GMPPNPstimulated PTPase activity (Fig. 2, compare A and C). G ␣ subunits were absent from corresponding fractions in profiles from GDP-treated membranes, indicating that co-elution of G ␣ and PTPase in these fractions required G-protein activation (data not shown). These results are consistent with activation of a PTPase by binding of G ␣ subunits and not G ␤␥ subunits. Superose 6 fractions encompassing the leading edge of the PTPase-G peak were pooled to reduce cross-contamination with the peak of PTP1B and subjected to ion exchange chromatography on Mono Q. PTPase-G was resolved into multiple forms that eluted in the middle of the gradient (Fig. 2D). The reasons for this heterogeneity are unknown. Each of these forms co-eluted with 41-kDa proteins recognized by antibodies to G ␣i/o/t/z (Fig. 2E) and also by antibodies to G ␣i1-3 and G ␣o (data not shown). PTP1B eluted in fractions 17-19 and did not co-elute with a GMPPNP-stimulated peak of PTPase activity or with G ␣ subunits. Recovery of PTPase-G activity after Mono Q chromatography was low, ϳ20 -30%, which may be indicative of disruption of a PTPase-G complex.
Reversibility of activation was demonstrated by substitution of GDP for bound GMPPNP in PTPase-G in fractions from Superose 6 chromatography (Fig. 3A). Portions of column fractions from GMPPNP-treated membranes were incubated with GDP or GMPPNP in the presence of Mg 2ϩ chelators to promote nucleotide dissociation (9) prior to readdition of Mg 2ϩ and assay. Replacement of GMPPNP with GMPPNP preserved the PTPase-G activity. Replacement of GMPPNP with GDP abolished the peak of PTPase-G, resulting in a profile after reassay nearly identical with the profile obtained by assay of fractions from GDP-treated membranes. Since G ␤␥ was not detected in the fractions containing PTPase activity, reversal of activation by GDP strongly supports regulation by G ␣ subunits.
To definitively test this hypothesis, we examined the ability of purified G-protein subunits to activate PTPase in fractions from GDP-treated membranes (Fig. 3, B and C). Preparations of brain G ␣ and G ␤␥ subunits were utilized to provide a range of subunits principally derived from members of the G ␣i family (10). Addition of brain G ␣ -GMPPNP, but not brain G ␣ -GDP, increased activity of PTPase that eluted in the leading shoulder of the constitutive ϳ70-kDa peak of PTPase activity (Fig. 3B). Brain G ␣ -GMPPNP also activated a species of M r ϳ150,000 that may correspond to the occasionally observed species of GMPPNP-stimulated PTPase activity noted above. Addition of a mixture of brain G ␤␥ subunits in the presence of GDP or GMPPNP did not alter PTPase activity (data not shown). Importantly, addition of activated recombinant G ␣o , G ␣i1 , and G ␣i2 also stimulated PTPase activity (Fig. 3C). G ␣o appeared to cause a greater activation than G ␣i1 and G ␣i2 suggesting specificity in the interaction with the effector PTPase, but detailed titration studies will be required to determine the relative potency and efficacy of various PTx-sensitive G ␣ subunits.
G-protein-coupled receptors may initiate both positive signals for MAP kinase activation via G ␤␥ subunits and negative signals via G ␣ subunits, depending on context and the specific G-protein coupled. For example, epitope-tagged MAP kinase expressed in Cos-7 cells is transiently activated by isoproterenol via endogenously expressed ␤-adrenergic receptors (16). The stimulatory signal is provided by G ␤␥ and is Ras-dependent. An inhibitory signal is provided by G ␣s -mediated elevation of cAMP concentration; elevation of cAMP in cells prior to stimulation blocks MAP kinase activation. In pheromone signaling in Saccharomyces cerevisiae, G ␣ stimulates an adaptive pathway that antagonizes G ␤␥ -mediated activation of the MAP kinase-related enzyme FUS3 (17). The signaling mechanisms for MAP kinase activation by mammalian G-coupled receptors are complex (18); it appears that specific G ␣ subunits may promote, inhibit, or possibly be indifferent to activation of MAP kinase by G ␤␥ . Complexity is also indicated by reports that overexpression of GTPase-impaired G ␣i2 caused transformation and activation of MAP kinase in Rat-1 but not NIH 3T3 cells (19), and that overexpression of GTPase-impaired G ␣o transformed NIH 3T3 cells (20).
Thus, a balance of positive and negative signals determines the extent of MAP kinase activation by G-protein-coupled receptors. PTPase-G may deliver a negative signal from receptors that couple to responsible subtypes of G ␣i/o to modulate the timing or extent of activation of MAP kinase. While our experiments reveal a negative modulation by PTPase-G of membrane-associated Raf-1 that has been already activated by v-Ras, we cannot exclude the possibility that PTPase-G may also act positively in other contexts or temporal sequences.
Identification of unambiguous effectors for G ␣i , as well as G ␣o , subunits has been elusive (reviewed in Ref. 21). Our findings strongly implicate PTPases as effectors for activated PTxsensitive G ␣ protein(s), fulfilling each of the classical biochemical criteria utilized to establish adenylate cyclase as an effector for G ␣s . Our findings also show that reconstitution with G ␣i/o subunits can serve as an assay for purification and identification of PTPase-G. Elucidation of the pathways regulated by G ␣ -regulated PTPases should provide insight into the mechanisms of mitogenic signaling and cell cycle control.