Biochemical Comparisons of the Saccharomyces cerevisiae Bern2 and Bem3 Proteins DELINEATION OF A LIMIT Cdc42 GTPASE-ACTIVATING PROTEIN DOMAIN*

The Ben2 and Bem3 proteins, which appear to play tein (designated Cdc42Hs) (Shinju et al., 1990) as a model to roles in the regulation of bud site formation in Saccha-study the protein-protein interactions underlying the regula- romycee cerevisiae, show striking homology to a number tion of rho-subtype proteins. Although the role of Cdc42 in of proteins that compose a family of GTPase-activating mammalian cells is currently unknown, significant progress proteins (GW’s) for the rho-subSouP of ras-related has been made toward the identification of proteins that are GTP-binding Proteins- These members human capable of directly influencing the GTP binding and GTPase Platelet GAP for Cdc42Hs (the human Of a activities of the human Cdc42 protein. One of these, the DBL cerevisim GTP-binding Protein that regulates bud site oncogene product, stimulates the dissociation of GDP from brain protein chimerin, the 85-kDa subunit ity (Hart et al., 1991a). This putative exchange factor (or GTP GAP-binding protein (~190). A fusion protein composed of the glutathione S-transferase protein and the rho-amino

(ps5) Of the phosphatidylinositol 3-kinase, and the ras-dissociation shares a region of homology (of 238 stimulates the GTPase activity ofthe wild-type Cdc42Hs 1981). A second Cdc42Hs protein opposes or any ability to bind to the GTP-bound form of Cdc42Hs.
A third group of Cdc42Hs regulatory proteins are the GTPwe have taken advantage of the functional specificity ase-activating proteins (GAPS), which stimulate the GTP hyexhibited by Bern3 (versus Bem2) in using Bem2/Bem3 drolytic event, GAPS are especially interesting because they chimeras, as well as different deletion mutant versions are likely to represent a diversity of structures and because Of the Bem3 protein, to the limits Of a functhese regulatory proteins may serve as target/effectors as well tional Cdc42 GAP domain. The Of this study Inas negative regulators of Cdc42Hs and/or related GTP-binding dicate that the carboxyl-terminal "224 amino acids proteins. A Cdc42Hs GAP was first identified and purified from (which contain three regions of homology to the other members of the rho-GAP family) represent a "limit GAP." human platelets (Hart et al., 1991b). The apparent molecular The first two appear to be important for binding to mass of this GAP (25 kDa) is similar to that of the rho-GAP Cdc42Hs and for partial GAP activity.
(-30 kDa) that has been purified from the cytosol of spleen (Garrett et al., 1989) and bovine adrenal glands (Morii et al., 1991). Based on the available amino acid sequence from rho-GTp-binding proteins appear to play 'dorganizational roles,, in homology with rho-GAP were identified. One of these is the 1992; Ridley and ~~1 1 , 1992;. one example that is rearranged in the Philadelphia chromosome and fused is the Saccharomyces cerevisiae cdc42 protein (designated to the ab1 tyrosine kinase-encoding gene (Heisterkamp et al., The different members of the rho-subgroup of ras-related GAP, a number Of proteins (Of size) that share sequence different cytoskeleton-associated assembly processes (Hall, break point re@on protein (bcr), the product of a Cdc42Sc), a cell division cycle protein that mediates the assem-1985). The carboxyl-teminal third of bcr was &own to act as a bly of the bud site in yeast (Johnson and Pringle, 1990; Drubin, GAP for the protein (Diekmann et al.* lgg1)> and more 1991). We have used the human homolog of the Cdc42Sc pro-full-len@h bcr was reported to act as a GAP for Cdc42Hs . The brain protein chimerin (-30 * This work was supported by National Institutes of Health Grant kDa) contains a rho-GAP re@on and was shown GM47458. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby The abbreviations used are: GDI, GDP dissociation inhibitor; GAPS, solely to indicate this fact. marked ''advertisement" in accordance with 18 U.S.C. Section 1734 GTPase-activating proteins for low molecular mass GTP-binding proteins; bcr, break point cluster region protein; PAGE, polyacrylamide gel 8 Present address: Pfizer Central Research, Nagoya, Japan. electrophoresis.

24629
to serve as a GAP for racl (Diekmann et al., 1991). Two other proteins sharing homology with rho-GAP are the 85-kDa regulatory domain (p85) of the phosphatidylinositol 3-kinase and the ras-GAP-binding protein (p190) . The p190 protein was shown to have GAP activity for the rho-, rac-, and Cdc42Hs GTP-binding proteins (Settleman et al., 1992).
Recently, two S. cerevisiae gene products, designated Bema , that are suspected to be involved in the regulation of bud site formation (presumably acting downstream from Cdc42Sc) have been identified.2 These two yeast proteins contain sequences (near their carboxyl-terminal ends) that bear striking similarity to rho-GAP, bcr, chimerin, and p190. In this study, we compare the abilities of Bem3, bcr, p190, and platelet Cdc42Hs GAP to couple functionally to the human Cdc42 protein and take advantage of the functional differences between the Bern2 and Bern3 proteins in utilizing BemZiBem3 chimeras in combination with various deletion mutant versions of the Bem3 protein to delineate a limit catalytic Cdc42 GAP domain.
E~E R I M E N T~ PROCEDURES C o~~r u e~i o n of Exp~ssion Plasmids for Bern2 and Bem3 Mutants -The Escherichia coli vector pGEX-KG (for the expression of glutathione S-transferase fusion proteins) was generated from pGEX-2T (Pharmacia LKB Biotechnology Inc.) by insertion of a polylinker into the EcoRI cloning site (Guan and Dixon, 1991). The plasmids for expressing the Bem2 and Bem3 carboxyl-terminal polypeptides as glutathione Stransferase fusion proteins (GST-Bem2,  were constructed by digesting p~P / B E M 3 and p~P /~E M Z 3 with the indicated restriction enzymes, followed by treatment with the Klenow fragment of DNA polymerase I (when required for the generation of blunted ends) and ligation (using T4 DNA ligase) of the agarose gelpurified fragments into the corresponding in-frame sites on pGEX-KG HincII and HindIII to make GST-Bem2 (which contains residues 1652-2167 of the Bem2 protein), Ban1 and HindIII to make GST-BemS' (residues 1967-2167, and XfsoI and Hind111 to make GST-%em3 (which contains residues 751-1128 of the Bem3 protein). The segments of the BEM3 gene cDNA were further subcloned from pGEX-KG-BEM3 into the pGEX-KG vector to generate the deletion mutants: AEuI to HindIII to make Bem31 (residues 982-1128), NlaN to HindIII to make BemSz (residues 954-1128), XhoI to DpnI to make Bem33 (residues 751-1067), SspI to SspI to make Bem3, (residues 904-1090), and SspI to HindIII to make Bemafi (residues 904-1128). The chimeras of the Ben2 and Bem3 fusion proteins were made by ligating segments of the BEM3 gene to BEM2 using linkers that filled in any missing amino acid codons and then subcloning into the pGEX-KG plasmid: the HincII-HphI fragment of BEM2 was ligated to the DraI-Hind11 fragment of BEM3 and connected with a TAAAT"A1inker to make GST-Bem2,/3, (residues 1652-2065 of Bem2 and residues 1023-1128 of Bem3); the HincI1-EcoRI fragment of BEM2 was ligated to the NlaIV-Hind11 fragment of BEM3 with a CCCGGG linker to make the GST-Bem2J3z chimera and with a TC-CGGA linker to make the GST-Bem2,/3,, chimera; and the SspI-Bra1 fragment of BEM3 was ligated to the HphI-Hind111 fragment of BEM2 to make the GST-Bem3,/2, chimera. The protruding overhangs of each of these DNA fragments were blunt-ended with the Klenow fragment of DNA polymerase I.
Preparation of Cdc42Hs, GAPS, and Glutathione S-Zkansferase Fusion Proteins-The wild-type and mutant Cdc42Hs proteins were expressed in E. coli and prepared as described (Hart et aZ., 1991b). Human platelet GAP for Cdc42Hs was purified to near homogeneity through six chromato~aphy steps (Hart et at., 1991b). Holo-bcr was expressed in Sf9 insect cells (Maru and Witte, 1991) and immunoprecipitated by the anti-bcr amino-terminal antibody RB1. In some experiments, we also used insect cell lysates that overexpressed bcr (see Hart et al. (1992)) to estimate the approximate affinity of bcr for GTP-bound Cdc42Hs. Purified p190 was a kind gift from Dm. J. Settleman and R. A. Weinberg (Whitehead Institute). The p85 cDNA in the pGEX-2T vector was obtained from Dr. Lewis Cantley (Tub University). The E. coli-expressed ghtathione S-transferase fusion proteins were expressed as follows. DH5a cells carrying pGEX were grown at 37 "C to log phase and treated with 1 m~ isopropyl-1-thio-P-D-galactopyranoside for 1-2 h. The cells Y. Zheng, R. A. Cerione, and A. Bender, manuscript in preparation.
were collected; lysed in 20 m~ Tris-HC1 (pH 7.51, 100 m~ NaC1, 5 m M MgClz, 0.5% Nonidet P-40, 1 m M phenylmethyl~u~ony1 fluoride, 10 pdml leupeptin, and 10 pg/ml aprotinin by two rounds of 30-s sonication with a probe sonicator; and centrifuged at 10,000 x g for 15 min. The fusion proteins were purified from the supernatant using a glutathione-agarose affinity column (Koland et al., 1990). The purified proteins were visualized by SDS-PAGE with Coomassie Blue staining, and protein concentrations were determined by the method of Bradford (1976) with bovine serum albumin as a standard.
GTPase Activity Assay-A filtration assay (Hart et al., 1991b) was used to measure the GST-Bem2 and GST-Bem3 GAP activities. The apparent loss of [y-3"P]GTP or [a-32PIGTP bound to recombinant Cdc42Hs, as an outcome of the hydrolysis of the radiolabeled [y-y-"2PlGTP or the dissociation of tol-"PIGTP, respectively, was determined by measuring the [32PlGTP remaining on nitrocellulose filters at 22 "C in the presence or absence of the Bema or Bem3 protein. In Fig.  2 8 , [a-32P]GTP was substituted for [y-32PlGTP in order to be certain that the Bem3-stimulated loss of radiolabeled GTP was a specific monitor of (stimulated) GTP hydrolysis rather than GTP dissociation from Cdc42Hs. For assays comparing the relative GAP activities and affinities for GTP-bound Cdc42Hs, the GTPase reactions were terminated by precipitating the nonhydrolyzed PPIGTP with charcoal; the released 32Pi that remained in the supernatant was then measured (Brandt et ai., 1983).

Essential Functional Domains of Bern2 and Bem3 Proteins
Are Homologous to Members of rho-GAP Family-The S. cerevisiae BEM2 gene was first identified during a search for genes that interact with MSB1, a multicopy suppressor of temperature-sensitive ( T s ) mutations in Cdc42Sc Pringle, 1989,1991). Cells that lost BEM2 function appeared to be defective in bud formation (Bender and Pringle, 1991). BEM3 was then identified as a multicopy suppressor of the Tsgrowth defect of bem2 mutants (Bender and Pringle, 1991).2 The portions of the Bern2 and Bern3 proteins that are required for the suppression of the TSgrowth defect of bem2 mutants display similarity to rho-GAPS and to various other proteins that have been designated to be members of the rho-GAP family (Hall, 1992). These proteins include the bcr gene product (-160 kDa), the brain protein chimerin (-36 kDa), the ras-GAP-binding protein (~1901, and the 85-kDa regulatory subunit (p85) of the phosphatidylinositol 3-kinase (  Biochemical Studies of Cdc42 GAPs extent of amino acid identity occurs in box 1 (i.e. the homology region nearest the amino-terminal end), where the Bem2 and Bem3 proteins share -40% identity and where Bem3 is 44% identical to bcr, while Bem2 is 57% identical to bcr. The sequences within the middle region of homology (box 2) are -30% identical among the different proteins, whereas the sequences composing box 3 (nearest the carboxyl-terminal end) are 20-25% identical among the different proteins. Bem3 Protein Serves as GAP for Human Homolog of Cdc42 -The carboxyl-terminal portions of the BemP and Bem3 proteins have been expressed in E. coli as glutathione S-transferase fusion proteins using the pGEX-KG expression system.2 Two forms of the BEMZ gene cDNA (containing 600 base pairs and 1.7 kilobases) and one form of the BEM3 gene (containing 1.3 kilobases), which included all three boxes of the GAP homology region, were fused to the glutathione S-transferase gene under the control of the tac promoter. Induction by isopropyl-1-thio-P-n-galactopyranoside resulted in the appearance of a protein band of -70 kDa for GST-Bem3 and protein bands of -90 and 55 kDa for GST-Bem2 that are designated below (also see Fig. 2 A ) as GST-Bem2 and GST-Bem2', respectively. Fig. 2B shows that the recombinant GST-Bem3 protein is able to effectively stimulate the GTPase activity of the human Cdc42 protein t o an extent that is comparable to the stimulation obtained with Cdc42Hs GAP purified from human platelets (Hart et al., 1991b) (also see below). Neither the GST-Bem2 protein nor GST-Bem2' showed any ability to stimulate the GTPase activity of Cdc42Hs, i.e. the activities measured in the presence of these proteins were essentially identical to the basal rate of GTP hydrolysis and were indistinguishable from the GTPase activity for Cdc42Hs measured in the presence of glutathione S-transferase alone. The GST-Bem3 and GST-Bem2 or GST-Bem2' proteins also showed no effect on the binding of [ c Y -~~P~G T P to Cdc42Hs (relative to the binding of [a-"2P1GTP to Cdc42Hs in buffer alone), indicating that these glutathione S-transferase fusion proteins did not affect the GDP-GTP exchange activity of the GTP-binding protein.
One oncogenic form of H-ras has an amino acid alteration at position 12 (i.e. a valine residue is substituted for a glycine residue); this mutated ras-protein has an impaired (basal) GT-Pase activity that cannot be stimulated by ras-GAP (Trahey and McCormick, 1987). An analogous mutation in Cdc42Hs yields a similar result, i.e. a GTPase-defective protein that cannot be stimulated by platelet Cdc42Hs GAP (Hart et al., 1991b). As shown in Fig. 2C, the GST-Bem3 protein also is completely ineffective in stimulating the GTPase activity of the Cdc42HsV"'"" protein.
Comparisons of Activities of Different GAPs for Cdc42Hs -The abilities of different proteins to stimulate the GTPase activity of the human Cdc42 protein were assessed using a quantitative assay that measures 3zPi release (Brandt et al., 1983). The results of these studies, presented in Fig. 3, illustrate that the different proteins that serve as GAPs for Cdc42Hs vary in specific (GAP) activity over an -30-50-fold range. Human platelet Cdc42Hs GAP was the most potent stimulator of the GTPase activity, with half-maximal stimulation occurring at -5 nM. The p190 protein was slightly less effective, with half-maximal stimulation occurring at -10 nM. The GST-Bem3 protein was 10-fold less potent as a GAP (with half-maximal effects occurring a t -100 nM), and Spodoptera frugiperda-expressed, full-length bcr was the least potent, with half-maximal stimulation occurring a t >200 nM. It should be noted that although full-length bcr was capable of eliciting a measurable stimulation of the Cdc42Hs GTPase activity, assays performed with insect cell lysates expressing fragments of bcr typically showed a stronger ability to stimulate the GTPase activity, thereby suggesting that fragments of bcr (that contain GTP-bound Cdc42Hs proteins were assayed for the loss of the bound radioactivity over the indicated time a t 22 "C. C , the stimulations by GST-Bem3 of the GTPase activities of wild-type Cdc42Hs ( 0 ) and mut a n t C d c 4 2 H~~~l -'~ ( + ) were assayed as described above. Controls for the C d c 4 2 H~~~" '~ (El) and Cd~42Hs~'Y"~ ( 0 ) proteins were assayed with the glutathione S-transferase protein alone. the rho-GAP homology region) may have a higher specific activity than the full-length protein. Neither the GST-Bem2 protein nor an E. coli-expressed GST-p85 protein showed any ability to stimulate the GTPase activity of Cdc42Hs (even when the concentrations of these proteins exceeded 500 nM).

Deletion Mutations of Bem3 and Generation of Bem2iBem3
Chimeras: Zdentification of Limit Cdc42Hs GAP Domain-To delineate the domain necessary for the Cdc42Hs GAP activity, we have constructed a series of Bem3 deletion mutants and Bem2Bem3 chimeras (Fig. 4A). Each of these mutants was expressed as glutathione S-transferase fusion proteins and was affinity-purified and visualized by SDS-PAGE (Fig. 4B). We often have observed proteolytic fragments when expressing the different Bem3 proteins and Bem2Bem3 chimeras in E. coli;
however, adjustments were made in the amount of the recombinant parent proteins added to the GTPase assay mixtures so that all assays were performed using an identical amount of the parent Bem3 mutants or Bem2/Bem3 chimeras. We found that the NH2 terminus of Bem3 could be truncated as far as residue 904 (i.e. Bem35 in Fig. 4A) and still retain GAP activity. However, when the deletion was extended to residue 981 (Bem31) in the first GAP homology box (also see Fig. 11, no stimulation of the Cdc42Hs GTPase activity was detected. The truncation of the third homology box (residues 751-1067) (Bem33) rendered the fusion protein mostly insoluble (although a relatively high expression of this protein occurred) and perhaps resulted in an improper folding of the protein. Interestingly, the deletion of just 38 amino acids from the carboxyl-terminal end of Bem3 (residues 904-1090) (BernS4) also yielded an inactive GAP, even though this mutant contained virtually all of the residues that compose the three boxes of the GAP homology domain. Thus, the presence of the carboxyl-terminal 38 residues of Bema is at least required for the expression of a functional GAP; it seems possible that these residues, together with the residues composing the third GAP homology box, are necessary for proper protein stability and/or folding. None of the Bem3 deletion mutants that were ineffective as GAPs were able to compete with GST-Bem3 for Cdc42Hs, i.e. they did not serve as competitive inhibitors in the GAP assays even when examined at levels in high excess over GST-Bem3 (data not shown). Taken together, the results from the deletion analysis suggest that the regions in Bem3 required for GAP activity comprise the carboxyl-ter-minal224 residues of the protein and that all three of the GAP homology regions are necessary for full activity.
This suggestion is reinforced from the results of experiments with different Bem2/Bem3 chimeras. Four different Bem2/ Bem3 chimeras have been examined (Fig. 4A). Bem22/Bem31 joins residues 1652-2065 from the amino-terminal portion of Bema to residues 1023-1128 of Bem3, i.e. this chimera contains the first two GAP homology boxes of Bem2 and the carboxylterminal box (box 3) of Bem3. Bem32/Bem21 is composed of the first two GAP homology boxes of Bem3 (residues 904-1022) and the third homology box of Bem 2 (residues 2066-2167). Bem21/ Bem32 and Bem21/Bem32, were essentially identical, containing the first (amino-terminal) GAP homology box of Bema and the middle and carboxyl-terminal third GAP homology boxes of Bem3; the only difference between these two chimeras was that Bem21/Bem32, has a serine residue (position 952) from Bem3 at square checkered boxes for Bem3 and by small-square checkered boxes for Bem2. The Bern3 sequences between the first and second homology boxes and between the second and third boxes are depicted by rightward-slanted lined boxes. The corresponding sequences for Bema are depicted by leftward-slanted lined boxes. GAP activities were compared by assaying for 5 min at 22 "C using 1 pg of glutathione S-transferase fusion protein and 1 pg of the E. coli-expressed recombinant Cdc42Hs protein. B , analysis of the purified mutant proteins. The E. coli-expressed GST-Bem3 or GST-BemBBemS proteins were purified by glutathione-agarose affinity chromatography, and 1-2 pg of each protein were subjected to SDS-PAGE (10% acrylamide). the joining point (i.e. instead of the proline residue that is normally present in Bema; see Fig. 1). Bem22/Bem31, Bemal/ Bem32, and Bem21/Bem32p were all. completely inactive as GAPs, and none of these chimeras were able to compete with GST-Bem3 for the GTP-bound form of Cdc42Hs. However, the chimera Bem32/Bem21 was able to retain some GAP activity (i.e. -30% of the activity measured with GST-Bem3). These results suggest that the region responsible for GAP activity for Cdc42 primarily exists within the first two GAP homology boxes.

DISCUSSION
In S. cerevisiae, bud site formation appears to be regulated by the rho-subtype GTP-binding protein Cdc42Sc as well as by Cdc24 and the Beml/BemS proteins, which in turn may interact with or regulate Cdc42Sc (Adams et al., 1990;Johnson and Pringle, 1990;Drubin, 1991). Given that mammalian cells do not bud, it is interesting that the primary sequence of Cdc42 has been strongly conserved from yeast to humans. One approach that we have taken toward determining the role that Cdc42 plays in mammalian cells has been to identify and characterize mammalian proteins that regulate the GTP binding/GTPase cycle of this ras-like protein. Recently, we have shown that the DBL oncogene product (which shares homology with the S. cerevisiae CDC24 gene product) catalyzes the GDP-GTP exchange activity of the human Cdc42 protein (Cdc42Hs) (Hart et al., 1991a). This suggests a possible involvement of Cdc42Hs in pathways that regulate cell growth. Further support for this possibility comes from the findings that both bcr and the ras-GAP-binding protein (p190) are GAPs for Cdc42Hs. In addition to bcr and p190, a 25-kDa protein from human platelet membranes will serve as a GAP for Cdc42Hs (Hart et al., 1991b). It seems likely that platelet Cdc42Hs GAP will be similar, if not identical, to a 29-kDa GAP for the rho-GTP-binding protein that was first identified in spleen cytosol (Garrett et al., 1989).
The identification and biochemical characterization of Cdc42Hs GAPs could provide important clues as to the function of Cdc42Hs since GAPs may represent physiological targets for this GTP-binding protein. However, an important question concerns the physiological significance of the ability of these different proteins (bcr, p190, and the 25-kDa platelet protein) to serve as GAPs for Cdc42Hs in vitro, Le. are the in vitro activities representative of interactions between Cdc42 and the different GAPs that occur in vivo, or do they represent "crossreactivities" that are not reflective of in vivo interactions? This consideration makes the discovery that the BEM3 gene product serves as a GAP for Cdc42 especially important since there are strong indications that Cdc42 and Bem3 interact within a common biological pathway in yeast.2 Direct comparisons of the Cdc42Hs GAP activities for the platelet 25-kDa protein, bcr, p190, and Bem3 show that the specific GAP activities of these different proteins vary by as much as 30-50-fold. Human platelet GAP was the most potent stimulator of the Cdc42Hs GTPase activity, followed by the p190 protein, the yeast Bem3 protein, and then full-length bcr. The simplest explanation for these differences is that the different GAPs bind to the GTP-bound form of Cdc42Hs with different affinities. This is consistent with the finding that neither the GST-Bem2 nor GST-p85 protein shows any capability of binding to the GTP-bound form of Cdc42Hs based on their inability to compete with any of the GAPs. In the future, we hope to develop fluorescence spectroscopic read-outs to monitor Cdc42-GAP interactions in order to obtain direct determinations of the relative affinities of the different GAPS for the GTP-bound (and GDP-bound) forms of Cdc42Hs.
It is interesting that we have not yet identified any protein with Cdc42Hs GAP homology that will bind to the GTP-bound Cdc42Hs species, but not stimulate its GTPase activity. However, recently, we have demonstrated that a 28-kDa protein from brain that acts to inhibit the dissociation of GDP from Cdc42Hs (as well as from the rho-and rac-GTP-binding proteins (Ueda et al., 1990;Hiraoka et al., 1992)) and triggers the dissociation of Cdc42Hs from plasma membranes also elicits a strong inhibition of both the intrinsic and the GAP-stimulated GTPase activities of Cdc42Hs . The ability of GDI to inhibit the GAP-stimulated GTPase activity (which has been demonstrated both for 25-kDa human platelet GAP and full-length bcr) appears to be the outcome of a competition between GDI and GAPs for the GTP-bound Cdc42Hs protein.
Thus far, we have not found any significant sequence similarity between GDI and GAPs that might point to a structural motif that was important for the recognition of the GTP-bound Cdc42Hs species. This may be due to the differences in the specific modes of interaction of GDI and GAPs with Cdc42Hs. Specifically, GDI will only bind to the isoprenylated form of the GTP-binding protein, whereas none of the Cdc42Hs GAPs appear to discriminate between prenylated and non-prenylated forms of the GTP-binding protein.
We have taken advantage of the fact that the Bem3 protein is a GAP for Cdc42Hs, whereas the structurally related protein Bem2 is not, in the use of deletion analysis and chimeric approaches to obtain information regarding what constitutes a limit Cdc42Hs GAP domain. Comparisons of the sequences of the different members of the rho-subtype GAP family indicate that the general region of "GAP homology" can be subdivided into three subdomains (designated as homology boxes) of -25-30 amino acids each. It was the limited sequence that was initially obtained from the box 3 homology region of spleen rho-GAP (Diekmann et al., 1991) that led to the identification of bcr and chimerin as members of the rho-subtype GAP family. The results of our studies indicate that all three homology boxes are essential for full GAP activity. In fact, it is necessary that an -30-40-amino acid segment extend both from the amino-terminal end of the GAP homology region ( i e . from box 1) and from the carboxyl-terminal end (box 3); otherwise no GAP activity (or ability to bind to the GTP-bound Cdc42Hs species) is observed. Presumably, the requirement of these overhanging sequences is for the proper folding of the complete GAP domain since these sequences are not conserved among the different members of the GAP family.
Of the various Bem2Bem3 chimeras that we have examined, only the chimera that contains the first two homology boxes from Bem3 and the third homology box from Bem2 is a functional GAP. Interestingly, despite the fact that box 3 was instrumental in the identification of various members of the GAP family, a Bem2Bem3 chimera that contains the third homology box of Bem3 is inactive as a GAP for Cdc42Hs. These results suggest that the first two homology boxes may be the most important for the effective (functional) coupling of GAP to the rho-subtype protein. When comparing the first two homology boxes for the different proteins that serve as GAPs for Cdc42Hs versus the first two homology boxes of Bem2, which is not a GAP, one of the most obvious differences is that all of the Cdc42Hs GAPs contain a serine at the second position from the carboxyl-terminal residue of box 1, whereas Bem2 has a proline at this position. Nonetheless, the single change of a proline to a serine at this position does not result in Cdc42Hs GAP activity. Thus, it seems likely that like the situation for the ras-GAP family (Marshall et al., 19881, a small sequence of amino acids will not be sufficient to bind to the GTP-bound Cdc42Hs species and to catalyze its GTPase activity, but that the three homology boxes provide a specific tertiary conformation that forms the proper binding (and catalytic) pocket.