Structural differences in the minimal catalytic domains of the GTPase-activating proteins p120GAP and neurofibromin.

The kinetic properties for the enzymatic stimulation of the GTPase reaction of p21ras by the two GTPase-activating proteins (GAPs) p120GAP and neurofibromin are different. In order to understand these differences and since crystallization attempts have only been successful with truncated fragments, structure/function requirements of the catalytic core of these proteins were investigated. Differences in size of the minimal catalytic domains of these two proteins were found as determined by limited proteolysis. The minimal catalytic domain has a molecular mass of 30 kDa in the case of p120GAP and of 26 kDa in the case of neurofibromin. Both catalytic domains contain the homology boxes as well as the residues perfectly conserved among all Ras GAPs. The C termini of these fragments are identical, whereas the N-terminal part of the minimal p120GAP domain is 47 amino acids longer. These newly identified minimal catalytic fragments were as active in stimulating GTPase activity toward p21ras as the corresponding larger fragments GAP-334 and NF1-333 from which they had been generated via proteolytic digestion. Recently it was postulated that a fragment of 91 amino acids from neurofibromin located outside the conserved domain contains catalytic activity. In our hands this protein is unstable and has no catalytic activity. Thus, we believe that we have defined the true minimal domains of p120GAP (GAP-273, residues Met714-His986) and neurofibromin (NF1-230, residues Asp1248-Phe1477), which can be expressed via LMM fusion vectors in Escherichia coli and isolated in high purity.

The kinetic properties for the enzymatic stimulation of the GTPase reaction of p21 ras by the two GTPaseactivating proteins (GAPs) p120 GAP and neurofibromin are different. In order to understand these differences and since crystallization attempts have only been successful with truncated fragments, structure/function requirements of the catalytic core of these proteins were investigated. Differences in size of the minimal catalytic domains of these two proteins were found as determined by limited proteolysis. The minimal catalytic domain has a molecular mass of 30 kDa in the case of p120 GAP and of 26 kDa in the case of neurofibromin. Both catalytic domains contain the homology boxes as well as the residues perfectly conserved among all Ras GAPs. The C termini of these fragments are identical, whereas the N-terminal part of the minimal p120 GAP domain is 47 amino acids longer. These newly identified minimal catalytic fragments were as active in stimulating GTPase activity toward p21 ras as the corresponding larger fragments GAP-334 and NF1-333 from which they had been generated via proteolytic digestion. Recently it was postulated that a fragment of 91 amino acids from neurofibromin located outside the conserved domain contains catalytic activity. In our hands this protein is unstable and has no catalytic activity. Thus, we believe that we have defined the true minimal domains of p120 GAP (GAP-273, residues Met 714 -His 986 ) and neurofibromin (NF1-230, residues Asp 1248 -Phe 1477 ), which can be expressed via LMM fusion vectors in Escherichia coli and isolated in high purity.
The product of the ras protooncogene, p21 ras , is involved in a multitude of signal transduction pathways through which the binding of an extracellular signal molecule to a tyrosine kinase receptor is transmitted into the nucleus of the cell to induce transcription of specific genes. Like any other guanine nucleotide-binding protein, p21 ras exists in two different conformations, an active GTP-bound form and an inactive GDP-bound form (1,2).
In the cell the conformational state of guanine nucleotidebinding proteins is regulated by two kinds of interacting molecules, guanine nucleotide exchange factors and GTPase-activating proteins (called GAPs). 1 Three specific GAPs for p21 ras have been described. The first, p120 GAP , is the prototype of this class of proteins and was the first one to be isolated (3)(4)(5). The second is neurofibromin (NF1), which is the product of the neurofibromatosis gene (6) and has also been shown to stimulate the GTPase of p21 ras (7)(8)(9). This gene has been found to be frequently mutated in patients with the disease neurofibromatosis type I (10 -12) but also, albeit less frequently, in solid tumors (13). A mammalian homologue (GAP1 m ) of the Drosophila GAP1 gene (14) has been described as the third form of Ras GAP (15). Recently, a fourth member of the Ras GAP family (GAP1 IP4BP ) has been isolated, which performs GTPase-stimulating activities toward both Ras and Rap (16). p120 GAP and NF1 protein can be distinguished with respect to their catalytic properties (7,17,18). p120 GAP increases the GTPase reaction of p21 Ha-ras more than 10 5 -fold, k cat ϭ 19 s Ϫ1 , with a K m of 9.7 M for p21 ras ⅐GTP (18), whereas NF1 protein has been reported to have a lower k cat of 1.4 s Ϫ1 but a higher affinity (K m of 0.3 M; 17). In studies with N-Ras the difference in affinity was similar, but much smaller in k cat (19). Certain lipids were found to inhibit the GAP activity of NF1 protein at micromolar concentration without having an effect on p120 GAP activity (20,21), although the latter binds membranes in a Ca 2ϩ -dependent fashion in vitro (22).
Furthermore, it is likely that the biological roles of these proteins are different: p120 GAP , in addition to its catalytic domain, contains several other independent functional domains such as SH2, SH3, pleckstrin homology domain, and Ca 2ϩ -dependent phospholipid binding domain (for a review, see Refs. 23 and 24), all of which seem to be involved in interactions with other signaling molecules. However, no obvious sequence correlation with other signaling molecules outside the catalytic domain has yet been identified in the primary sequence of NF1. There is some evidence that p120 GAP might be more than just an inactivator of Ras and may in fact be an effector (25)(26)(27). NF1 and GAP1 m , on the other hand, seem to be more negative regulators of Ras, since, outside the catalytic domain, they display a high homology to the Saccharomyces cerevisiae IRA gene products and to the Drosophila melanogaster protein GAP1, which have genetically been found to be negative regulators of the Ras signal transduction pathway (14,28,29). In agreement with the concept of a negative Ras regulator, embryonic neurons from NF1 knockout mice survive in the absence of neurotrophin, which signals via Ras (30). In addition, the NF1 gene has characteristics of a tumor suppressor gene, as evidenced from the abnormal regulation of p21 ras activity in neurofibromatosis type I patients (31,32).
To obtain information on the mechanism of stimulation of GAP activity we have set out to crystallize the minimal catalytic domain of Ras GAPs. Originally, a 483-residue GAP-related domain of neurofibromin had been described to contain full Ras GAP activity (7). Later, smaller fragments of between 333 and 343 residues were shown to possess full or almost full Ras GAP activity (18,33,34). Recently, Maruta and co-workers (35) reported that 91-and 78-residue fragments from the NF1 protein located outside the conserved Ras GAP region (see Fig. 1) can still stimulate the GTPase activity of Ras and that an even smaller fragment of 56 amino acids from the same region still binds to Ras⅐GTP (36). These reports together with the results of crystallization trials prompted us to reinvestigate the properties of the catalytic domains of GAPs and to rigorously define their boundaries.

MATERIALS AND METHODS
Fragment Cloning-For the bacterial expression of NF1 fragments, the plasmid ptrcNF1-333, which directs the synthesis of the 333-amino acid fragment between positions Glu 1198 and His 1530 , was used as starting material (17). We have also created a new expression plasmid, pETNF1-333, which expresses the same fragment from a T7 promoter, by recloning a NcoI/HindIII fragment from ptrcNF1-333 into a modified pET-3d vector (37).
Using the polymerase chain reaction (PCR) we constructed various expression plasmids for the NF1 catalytic domain, which are further shortened at the N terminus (Ϫ18N, Ϫ32N, Ϫ46N . Fragments were isolated using Geneclean (Renner GmbH), digested with NcoI/HindIII, and ligated into the corresponding vectors (ptrc99A-and pET-3d-modified). For overproduction of the recombinant proteins from ptrc constructs we used the Escherichia coli strain XL1Blue (38) and (for pET-plasmids) strain BL21-DE3 (37).
Gene expression was induced with isopropyl-1-thio-␤-D-galactopyranoside for 16 h when using the ptrc constructs and 3 h for the pET constructs. The cells were lysed with lysozyme, and soluble extracts were prepared as described (17). GAP activity was measured directly in these extracts using equal amounts of total protein concentrations.
The 26-kDa fragment resulting from chymotrypsin treatment of NF1-333 was separated from the protease by chromatofocusing on polybuffer exchanger PBE 94 (Pharmacia), applying a pH gradient from pH 5-8.3 as described for NF1-333 (17). The elution maximum of the 26-kDa fragment was at pH 6.3. The protein fraction was dialyzed against buffer B, and the polybuffer was removed by gel filtration on Superdex as described for the clostripain fragment (see below).
The reaction mixture from preparative digestion of GAP-334 by proteinase K was applied to a Pharmacia Mono-Q HR 5/5 column and eluted with a 20-ml linear gradient of 0 -200 mM NaCl in buffer A at a flow rate of 1 ml/min. The major peak at 130 mM NaCl was pooled. Protein concentrations were determined following the method of Bradford (41) using bovine serum albumin as standard.
Protease Digestions-Purified NF1-333 or the fusion protein LMMNF1 (40) were treated with the indicated proteases at a substrate: protease ratio of 100:1 (w/w) at room temperature. The buffer for digestion with clostripain was 50 mM potassium phosphate (pH 7.8), 2 mM CaCl 2 , 2 mM DTE, and 0.04% NaN 3 , and for the other proteases it was 50 mM Tris/HCl (pH 8.0), 2 mM DTE, and 0.04% NaN 3 . For the digestion of the LMMNF1 fusion protein 600 mM KCl was included in the buffers, since it has been shown that LMM fusion proteins retain the property of intact LMM of being soluble only at high KCl concentrations. The digestion time was up to 2.5 h for NF1-333 and GAP-334 and up to 12 h for the LMMNF1 fusion protein. The reaction was terminated with 50 g/ml leupeptin for the trypsin reaction, 1 mM phenylmethylsulfonyl fluoride for chymotrypsin and proteinase K, and 20 mM EGTA for clostripain. The digestion of 3.4 M NF1-333 with proteinase K was done in the presence or absence of 11.5 M p21(Q61L)⅐Gpp(NH)p. Aliquots from the proteolysis reaction were withdrawn at the indicated time points, the reaction was terminated, and aliquots were analyzed either for GAP activity as described above or for protein content on SDS-polyacrylamide gels.
Protein Sequencing-Proteins and the proteolytic fragments thereof were sequenced on an Applied Biosystems A470 gas-phase sequencer, and the resulting phenylthiohydantoins were identified and quantified on line.
GAP Activity Measurements-GTPase assays were done as described previously (42,43) by loading p21 Ha-ras expressed in E. coli (44) with [␥-32 P]GTP. Excess nucleotide was removed on a NAP-5 gel filtration column (Pharmacia) in the GAP reaction buffer (20 mM Hepes/NaOH (pH 7.5), 1 mM DTE). In the presence of the appropriate amount of GAP/NF1 protein and 2 mM MgCl 2 the GTPase reaction was measured by following the decrease of [␥-32 P]GTP bound to p21 Ha-ras as determined by a nitrocellulose filter-binding assay. Under standard conditions, i.e. for the determination of the specific activities of different NF1 preparations, the starting p21 ras ⅐[␥-32 P]GTP concentration was 1 M. The decrease of radioactivity over time was fitted by a single-exponential decrease function using the program Enzfitter (Elsevier Biosoft). For the determination of the enzymatic properties (K m and k cat ) of NF1 fragments, a constant amount of NF1 was treated with increasing concentrations of p21 ras ⅐GTP using different reaction volumes such that similar numbers of radioactivity counts were retained on the filters. In the case of the 26-kDa NF1 fragment produced by chymotrypsin digestion 4 nM enzyme was added to 121, 242, 362, 604, 1210, and 1810 nM p21 ras ⅐[␥-32 P]GTP in GAP reaction buffer, respectively. In the case of NF1-230 produced by IgA-protease digestion, different concentrations of enzyme (0.5, 1, 2, and 5 nM) were reacted with 0.1, 0.2, 0.3, 0.5, 0.8, 1, 2, and 3 M p21 ras ⅐[␥-32 P]GTP in GAP reaction buffer. The initial linear portion of the reaction was used as the initial rate of the reaction. The error of these measurements was estimated to be 15%. For the determination of K m and k cat values, the initial rates measured with increasing concentrations of p21 ras substrate were fitted directly to the Michaelis-Menten equation using the program Enzfitter.

RESULTS
Deletion Analysis of the NF1 Catalytic Domain-We have shown earlier that a fragment of the NF1 protein containing 333 amino acids from position Glu 1198 to His 1530 (17) (see Fig.  1) can be expressed in E. coli using the trc promoter and that this recombinant protein, NF1-333, can be purified using a three-column procedure. From preparations of this catalytic fragment we obtained crystals that appeared to be suitable for x-ray crystallography. Gel electrophoretic analysis of these crystals showed, however, that they contained an NF1 fragment of a much lower molecular mass compared with the one representing the predominant species in the preparation used for the crystallization set-up (Fig. 2). The small fraction of this fragment that could be solubilized from crystals appeared to be catalytically active (not shown). This was surprising with respect to earlier results with p120 GAP , which showed that a fragment comprising 343 amino acids could not be truncated further without loss of catalytic activity or solubility after expression in E. coli (33). In contrast to this report, Nur-E-Kamal et al. (35) postulated Ras GAP activity for a 78-amino acid fragment of NF1 located outside the motifs conserved among RAS-GAPs. This prompted us to investigate more closely whether smaller fragments of NF1 and GAP could be generated that would retain full GTPase activating function toward p21 ras . Fig. 1 shows an alignment of various Ras GAP sequences. Boxes of homologous sequences reported earlier (9,45,46) are indicated. We have made various deletion constructs by shortening NF1-333 at the N and/or C terminus as indicated (i.e. Ϫ18N means that 18 amino acids have been deleted from the N terminus). The PCR-generated fragments were cloned into the NcoI/HindIII sites of ptrc99A, the vector used earlier for the expression of NF1-333. In parallel, we tested the expression vector pET-3d for some of the constructs, since higher protein yields for NF1-333 had been observed with the latter plasmid (37). GAP activities were tested in crude bacterial extracts. All of the extracts contained none or only weak (as in the case of constructs Ϫ8C and Ϫ22C) GAP activity on p21 ras (data not shown). This was surprising, considering our earlier observation that truncated NF1-333 extracted from the NF1 crystals had retained at least partial activity. However, the weak activities in extracts from bacteria with the Ϫ8C and Ϫ22C constructs (only 4% as compared with NF1-333) were apparently due to reduced expression, as we could demonstrate after partial purification of NF1-333(Ϫ22C) by Q-Sepharose chromatography and chromatofocusing. The enriched protein fraction had a specific GAP activity similar to NF1-333.
Proteolytic Fragments of NF1-Since the NF1 fragment in the crystals was found to have a molecular mass of approximately 26 kDa, it seemed likely that the lack of soluble GAP activity for practically all of the constructs shown above might have been caused by improper folding of the proteins in E. coli. We therefore tried to obtain a minimal, active NF1 domain from NF1-333 by proteolytic digestion, using purified NF1-333 or a fusion protein consisting of the LMM (light meromyosin) portion of rabbit muscle myosin and NF1-333, LMMNF1-333, which can be purified easily using a high salt-low salt extraction procedure analogous to that for complete rabbit muscle myosin (40). NF1-333 was subjected to independent digestions with the specific proteases clostripain, trypsin, chymotrypsin, and proteinase K. These proteases produced fragments of different sizes, which became smaller with the increasing number of recognition sites for the respective proteases (data not shown). Table I shows that in each case the core fragments obtained were close to fully active with respect to the specific activity of NF1-333.
The stable fragment after chymotrypsin digestion, whose apparent molecular mass of 26 kDa was close to that seen in the crystal, was further characterized after preparative proteolysis of LMMNF1. Purification via chromatofocusing and gel filtration removed the LMM portion, the smaller peptides as well as the protease, as described under "Materials and Methods." The purified fragment was subjected to protein sequencing. From the first 19 amino acids 17 were unequivocally identified as DSRHLLYQLLXNMFXKEVE, implying that chymotrypsin cleaved behind Phe 1247 and removes 50 amino acids from the N terminus of the NF1-333 protein as indicated (Fig. 1).
For the localization of the C-terminal end the molecular mass of the chymotrypsin fragment was measured by laser desorption mass spectroscopy from two independent preparations. In the first case a single mass peak of 25,961 Da was obtained, which is very close to the calculated molecular mass (25951 Da) of a 226-amino acid fragment starting at Asp 1248 and ending at His 1473 . In the second case three closely related mass peaks were found with a mean molecular mass of 26396 Da. This would indicate that Phe 1477 is the C-terminal amino acid rather than His 1473 . Cleavage after Phe 1477 is more likely, considering the specificity of chymotrypsin cleavage. The specific activity of the chymotryptic fragment amounted to 67% of the NF1-333 value (Table I). Its enzymatic properties were further analyzed by incubating the fragment with increasing concentrations of p21 ras ⅐GTP and measuring the rate of GTP hydrolysis as described for GAP-334 and NF1-333 (17,18). ) and similar to that of full-length NF1 (48).
The influence of complex formation was investigated by digesting NF1-333 (3.4 M) with the highly unspecific proteinase K in the absence or presence of saturating amounts of p21(Q61L)⅐Gpp(NH)p (11.5 M) (Fig. 3). The 34-kDa protein was digested with an estimated half-life of 20 min in the absence and of 2 h in the presence of p21(Q61L). In both cases a final product with the same molecular mass of 26 kDa was produced, similar to that of chymotryptic digestion. Comparison of the proteinase K fragment with that isolated from the crystal demonstrated very similar apparent molecular masses (Fig. 2, lane 4).
From the first 14 amino acids of the proteinase K fragment, 11 could be unequivocally identified as VXLFDXRXLLYQLL, which means that proteinase K cleaved NF1-333 behind Leu 1243 , as indicated in Fig. 1. Based on the molecular mass determined by SDS-gel electrophoretic analysis, the specific activity (in units/mg) of the proteinase K fragment represents FIG. 2. Results from crystallization experiments showing that the protein NF1-333 used for crystallization trials was accidentally digested to smaller fragments during the purification procedure, probably by contaminating proteases. An SDS-polyacrylamide gel containing various protein samples was silver-stained. Lanes 1-4 contain NF1-333 (1), mother liquor of crystallization setup (2), dissolved crystal (3), and the proteinase K fragment (4), the fragment from the experiment presented in Fig. 3, respectively. Lane M contains markers with molecular masses, in kDa, as indicated. a In these cases the specific activity of the proteolytic fragments was determined in the digested sample before further purification using the conditions as described under ''Materials and Methods.'' 81% of the wild type activity (Table I). Unfortunately, it turned out to be unsuitable for purification (and crystallization) because of continued proteolysis and a strong tendency to aggregate as seen on a native polyacrylamide gel.
Having established the minimal 230-residue fragment from NF1 by chymotrypsin cleavage we utilized improved LMM expression vectors with various specific protease sites 2 to test whether we could express this domain as a LMM fusion protein. Fig. 4A illustrates that the 230-residue fragment of NF1 (Asp 1248 to Phe 1477 ) could indeed be isolated using an LMM expression vector with an IgA protease cleavage site. The protein could be isolated as a fusion protein, released from LMM by IgAse proteolysis and isolated in large amounts (currently used to grow crystals of the catalytic domain of NF1). The enzymatic properties of NF1-230 were determined by Michaelis-Menten analysis (Fig. 5) to obtain a K m of 0.65 M and a k cat of 7.3 s Ϫ1 . In comparison with the activity of the chymotryptic product from NF1-333 the catalytic activity of the expressed NF1-230 is 8-fold higher. This can be explained by the improved preparation method of NF1-230 using the LMM vector. NF1-333 prepared from the corresponding vector has similar catalytic properties.
Minimal Fragment of p120 GAP -A 227-residue fragment from p120 GAP (Leu 761 -Asp 989 ), which according to sequence alignment corresponds to NF1-230 (Fig. 1), could be expressed as an LMM fusion protein in large amounts but was insoluble (data not shown). This prompted us to investigate the catalytic activity of GAP-334 after proteolysis. Fig. 6 shows the progressive digestion of GAP-334 with proteinase K. GAP-334 proteolysis produces a final fragment of an estimated 30-kDa molecular mass, which is stable against further degradation.
In parallel, GAP-334 was identically treated in the presence of saturating amounts of p21(Q61L)⅐Gpp(NH)p, which binds to GAP-334 with a dissociation constant of 5 M, 3 as compared with 19 M for wild type p21 ras . Thus, like full-length p120 GAP , GAP-334 has a higher affinity for the Gpp(NH)p complex of p21(Q61L), although the difference between wild type and mutant is more pronounced with full-length p120 GAP (0.1 M versus 4 M dissociation constants, respectively; Refs. 20 and 18).
As with NF1-333, the addition of p21(Q61L) considerably reduces the half-life of GAP-334 to about 2 h (Fig. 6B). The final fragments produced have identical molecular masses (30 kDa) in the presence and absence of complex formation with p21(Q61L)⅐Gpp(NH)p. Therefore, we conclude that in the GAP-334 complex, regions of GAP-334 are less accessible for proteolytic digestion and can only be cleaved after dissociation of the complex (Fig. 6).
The proteinase K fragment of GAP-334 was purified by ion exchange chromatography. By protein sequencing of the proteinase K fragment of GAP-334, 8 of the 14 N-terminal amino acids could be unequivocally identified and turned out to be identical to those of undigested GAP-334 (Fig. 1). Upon molecular mass determinations of the proteinase K by laser desorption mass spectroscopy, one major and two minor mass peaks were consistently found. This can be explained by heterogeneities in the original GAP-334 protein preparation, although the polypeptide subpopulations must have similar molecular weights, as they could not be visibly resolved on an SDSpolyacrylamide gel. The masses of the major and minor fragments were 31216, 32553, and 30178 Da, respectively. The dominant fragment mass is close to the calculated mass for the 273-amino acid-long GAP peptide with His 986 at the C terminus (Fig. 1). The minor peak at 32,553 Da probably corresponds to a protein that became visible in the SDS gel only after protease digestion (apparent molecular mass, 35,000 Da). It may be unrelated to GAP-334, but it co-chromatographed.
The 273-residue fragment from p120 GAP from Met 714 to His 986 , corresponding to the proteinase K cleavage product, could be expressed in E. coli as a LMM fusion protein. Fig. 4B shows that the protein was isolated as LMM fusion protein, cleaved from LMM by IgA-protease, and separated from LMM by dialysis against low salt buffer. It remained soluble after purification. Since the specific activity of GAP-273 (434 units/ mg) for catalyzed hydrolysis of p21-bound GTP is almost the same as that measured with GAP-334 (462 units/mg), we assume that the k cat and K m values are also similar.
NF1-91-Since it had been reported that an NF1 fragment of 91 residues (and an even smaller one of 78 residues) from amino acids Arg 1441 to Lys 1531 in the primary sequence ( Fig. 1) carried at least partial Ras GAP activity (35,36), and since the minimal domains of NF1 and p120 GAP determined by us only partially overlap with the 91-amino acid fragment, we tried to reproduce those results. We bacterially expressed the 91-residue fragment via different fusion expression systems (pLMM1, pMAL-c2, and pGEX-4T-1), but we could not detect any soluble fusion protein of the expected size or any Ras GAP activity in these bacterial extracts. DISCUSSION We have delineated the minimal domains of the Ras GAPs p120 GAP and neurofibromin that retain catalytic activity. In the case of p120 GAP this was established to be a 273-residue fragment (Met 714 -His 986 ) obtained by the proteolytic removal of 61 amino acids from the C terminus of GAP-334. In the case of NF1 we found that 50 and 53 residues could be removed from the N and C termini of NF1-333, respectively, resulting in a minimal catalytic domain of only 230 residues (Asp 1248 -Phe 1477 ). The cDNAs corresponding to these regions were expressed in E. coli, and the resulting polypeptides were isolated to high purity. 20-fold differences of affinities between Ras in the triphoshate form and the catalytic fragments of p120 GAP or NF1 have independently been found by different authors using direct or indirect techniques (17,49,50). For p120 GAP it has been shown that the enzymatic properties and the biological activity of the catalytic domain are modulated by other domains of the protein (17,55). For NF1, the 26-kDa fragment described here, NF1-333 (17,19), NF1-GRD with 483 amino acids (7), and the complete NF1 protein containing 2818 amino acids (48) have apparently similar enzymatic properties. By improving the purification procedure for NF1 fragments using LMM fusion vectors (40) 2 it now appears that the maximum rates of the GTPase reactions (k cat ) catalyzed by neurofibromin and p120 GAP may in fact be very similar, in agreement with data from Eccleston et al. (19). These authors used increasing concentrations of GAP-344 to determine k cat rather than increasing concentrations of substrate, whereby the measurement becomes independent of the quality of the GAP-344 preparation.
What is the basis of the different properties of p120 GAP and neurofibromin? Is there a recognizable structural difference between the catalytic domain of the two proteins? A partial answer to this question is given by the results presented in this paper concerning the dimensions and stability of the catalytic GTPase-activating domains of these proteins. Ballester et al. (9) and Wang et al. (45) applied the program MACAW (47) to locate, analyze, and assess the statistical significance of regions of local similarity between the sequences of the Ras GTPase-activating proteins (see also Ref. 46). We have extended these studies to include some recently cloned GAP genes (some of which are shown in Fig. 1). The sequences contain four blocks of similarity within the catalytic domains of these proteins. NF1, IRA1, and IRA2 display additional sequence similarities outside the catalytic domain (not shown). The sequence alignment depicts the 12 residues that are totally conserved and many other residues that are almost completely conserved. The proteolysis experiments together with the corresponding enzymatic activity measurements presented in this study show that the sequences in the homology boxes, and all the totally conserved amino acids are indeed necessary for structural and/or enzymatic integrity of at least the two Ras GAPs investigated here. Only two of the highly conserved residues in the N terminus of the catalytic domain (Ala 1226 and Leu 1243 of NF1) are missing in the minimal catalytic domain of NF1.
Our data show that the minimal, catalytically active domain of p120 GAP , which is reasonably stable against further proteolysis, is 47 amino acids longer than NF1-230 at the N terminus, while possessing the same C terminus. Since these additional 47 amino acids of GAP-334 cannot be cleaved from the catalytic domain, this part of the protein adopts a different conformation in p120 GAP and NF1. Another indication of the structural importance of the N-terminal amino acids of the catalytic domain of GAP-334 comes from the observation that GAP-227, which corresponds to NF1-230, cannot be folded correctly in bacteria and is therefore insoluble. It is very likely that the overall three-dimensional structures of the catalytic domains of p120 GAP and NF1 will be similar. The difference in structure may thus be confined to the N-terminal part of the polypeptide chain. This might either adopt a different secondary structure or, in the case of p120 GAP , might be characterized by stronger interactions with the rest of the domain, thereby preventing access to the protease.
It is interesting to note that the differences in enzymatic characteristics and structure are related to different biological roles for p120 GAP and NF1. Mostly due to the extensive sequence similarity with the S. cerevisiae IRA gene product and to the absence of SH2 and SH3 domains, the NF1 protein might be primarily considered a negative regulator that keeps the concentration of active GTP-bound p21 ras down to a low level in unstimulated cells. p120 GAP , on the other hand, is more likely to be primarily a downstream target molecule with respect to p21 ras because of its involvement, via its SH2 and SH3 domains, in the transmission of signals from activated tyrosine kinase receptors (25,26,(51)(52)(53)(54).
The region encompassing a 91-residue fragment from residues Arg 1441 to Lys 1531 reported to retain Ras GAP activity is shown in Fig. 1 (35). It does not contain any of the homology boxes or totally conserved amino acids that are found in all p120 GAP sequences as mentioned above. We have been unable to repeat those experiments because in our hands the 91residue fragment could not be isolated from E. coli as a stable protein domain. When testing the soluble bacterial extracts, not even residual GAP activity toward p21 ras was detectable. We do not have any explanation for this discrepancy.
The structural analysis of the Ras GAPs presented here has led to the crystallization of GAP-334, GAP-273, NF1-230, and a complex between NF1-230 and p21 ras ⅐Gpp(NH)p. The crystals will hopefully give us a detailed three-dimensional view of the p21 ras -GAP interaction, as it is critical for the signal transduction cascade via p21 ras .