The GTPase-activating NF1 Fragment of 91 Amino Acids Reverses v-Ha-Ras-induced Malignant Phenotype*

The human neurofibromatosis type 1 gene encodes a Ras GAP (GTPase-activating protein) of 2818 amino acids called NFl. This GAP contains a domain of 338 amino acids (residues 1194-1531) called NF1-GRD, which shares 26% sequence identity with the C-ter- minal domain (GAPlC, residues 709-1044) of another Ras GAP called GAP1. Both NF1-GRD and GAPlC activate normal Ras GTPases but not oncogenic mutants such as v-Ha-Ras. Any attempt to reverse the v- Ha-Ras-induced malignant transformation by the GAPlC has, so far, been unsuccessful. However, we have found that when the NFl-GRD is overexpressed in v-Ha-Ras-transformed NIH/3T3 cells, it greatly re- duces their ability to form colonies in a soft agar, the property tightly associated with their malignancy. This is, so far, the first demonstration that a Ras- binding protein can act as a potent antagonist of the oncogenic Ras mutants in mammalian cells. In an at- tempt to further screen the smallest anti-oncogenic fragment derived from the NF1-GRD, we have pre- pared a series of its deletion mutants and examined their interaction with Ras first by monitoring their GAP activity (ability to activate the normal GTPase). The deletion analysis has revealed that the N-terminal 247 amino acids (residues 1194-1440) of NFl-GRD are not required for its GAP activity, suggesting nontransfected 0); 2, sense NF338-transfected (clone 8); 3, sense NF338-transfected (clone 22); 4, anti-sense NF338-transfected (clone 12). The steady-state levels of both NFI-GRD and v-Ha-Ras mRNAs the parental and were were

bound G T P is hydrolyzed to GDP by their low intrinsic GTPase activity. The majority of oncogenically mutated Ras shows much lower intrinsic GTPase activity than the normal Ras. Furthermore, the normal Ras GTPases, but not oncogenic mutants, are activated by two distinct GAPS,' i.e. GAPl of 1044 amino acids (also called Ras GAP of 120 kDa) and NF1 of 2818 amino acids (also called neurofibromatosis type 1 gene product or neurofibromin), in mammalian cells (2)(3)(4)(5)(6)(7). Thus, the oncogenic Ras mutants such as v-Ha-Ras would remain in the GTP-bound (active) forms for much longer periods than the normal Ras in the cells. This appears to be the major reason why the Ras mutants are much more oncogenic than the normal Ras for the majority of the target cells, in particular pancreatic and colonic cells, where v-Ha-Ras and other Ras mutants most frequently cause development of malignant tumors in human (8).
For more than a decade since the first discovery of oncogenic Ras mutants in retroviruses (v-Ha-Ras and v-Ki-Ras) and human carcinomas (l), various attempts have been made to attenuate the oncogenic Ras signals in the tumor cell lines such as v-Ha-Ras/v-Ki-Ras-transformed NIH/3T3 cells with a hope to find a cure for Ras-associated cancers. The major approach in the field of molecular Ras oncology has been to screen the tumor suppressor genes that are able to reverse the v-Ha-Ras/v-Ki-Ras-inducedmalignant transformation by the DNA-mediated transfection and overexpression. Several distinct tumor suppressor genes have been reported to reverse the malignancy caused by the Ras oncogenes (9-15) as follows: the full-length human Rap1 (also called Krevl) encoding a Ras-related G protein of 184 amino acids (9), which tightly binds GAPl but without any stimulation of its intrinsic GTPase activity (16); a truncated v-Jun gene (10) encoding only the C-terminal domain of 150 amino acids (residues 147-296), which lacks the N-terminal transactivating domain; the full-length murine Thy-1 gene encoding a cell surface glycoprotein of 142 amino acids, which is covalently linked to a glycophosphatidyl inositol (11); a human c-Ets-2 DNA (12) encoding the C-terminal DNA-binding domain of 133 amino acids (residues 333-466), which binds a Ras-responsive DNA element in enhancers of several Ras-transactivated genes; a rat b-Myc DNA (15) encoding the N-terminal DNA-binding domain of 120 amino acids, which acts as a c-Myc antagonist. Interestingly, none of these gene products binds directly Ras GTPases.
Any attempt with either the full-length GAPl or its Cterminal GTPase-activating domain (GAPlC, residues 709-1044) to reverse the malignant transformation caused by the oncogenic Ras mutants has been unsuccessful so far, although the GAPlC (but not the full-length GAP1) is able to reverse the malignant transformation caused by overexpression of normal Ras genes such as c-Ha-Ras (17). Shortly after the ' The abbreviation used is: GAP, GTPase-activating protein.

22331
Anti-v-Ha-Ras Action of NF1 Fragments in Mammalian Cells NF1-GRD of 338 amino acids, i.e. a GAPlC-related domain (GRD, residues 1194-1531) of NF1, which also activates the normal Ras GTPases (18,19), was shown to bind Ras much more tightly than the GAPlC (6), and a similar, but slightly larger, NF1-GRD fragment was reported to reduce the v-Ha-Ras-induced heat-shock susceptibility of yeasts (20), we have started to examine the possible anti-v-Ha-Ras action of the NF1-GRD and its truncated mutants in mammalian cells.
In this paper, we demonstrate, for the first time, that when either the NF1-GRD or its short fragment of 91 amino acids (NF91, residues 1441-1531) are overexpressed, each reverses the v-Ha-Ras-induced malignant transformation of NIH/3T3 cells. The NF91 is, so far, the smallest anti-tumor polypeptide that has ever shown the potent anti-v-Ha-Ras action in mammalian cells.

MATERIALS AND METHODS
Construction of the Plasmids Expressing NFl -GRD (in Sense and Anti-sense Orientations) and NF91 (in Sense Orientation)-For the NF1-GRD (NF338+ or NF338-) expression, a HindIII DNA fragment of 1.1 kilobases containing at 5' end a Kozak sequence (GCC GCC ACC ATG) followed by codons 1194-1531 of human NF1 (type I) and at 3' end a termination codon (TAA) was prepared by polymerase chain reaction and subcloned into the retroviral vector pMV7 expressing also a neomycin resistance selectable marker (21). The orientation of the insert was determined by EcoRI digestion, as one EcoRI site is located 100 base pairs upstream of the HindIII insertion site and the other at the codon 1355 of the NF1 (for detail, see Fig.  1). For the NF91 expression, an EcoRI/HindIII polymerase chain reaction fragment of 0.3 kilobases containing at 5' end the Kozak sequence followed by codon 1441-1531 of the human NF1 and at 3' end the termination codon was prepared and inserted into the EcoRI/ HindIII sites of the vector pMV7. The resultant plasmid called NF91 as well as the other two plasmids called NF338+ (sense) and NF338-(anti-sense) were purified by CsCl density gradient centrifugation and used for transfection.
Assay for Colony-forming Ability in a Soft Agar of u-Ha-Ras-transformed Cells Expressing the NF1-GRD or NF91-v-Ha-Ras-transformed NIH/3T3 cells (22) were transfected with either NF338+, NF338-, or NF91 (1 pg) as complexes with liposomes (35 pg) as described previously (23). The resultant G418-resistant transfectants were cloned in the presence of 400 pg/ml G418 (a neomycin analogue) as described previously (22, 23). Colony-forming ability of the parental clone and each transfectant was examined in a soft agar by incubating 1000 cells/plate at 37 "C for 3 weeks under the standard culture conditions (22).
Northern Blot Analysis of NFI-GRD and u-Ha-Ras Expression-5 pg of poly(A) tail-containing mRNAs, affinity-purified on an oligo (dT)-cellulose column from each clone, was separated by electrophoresis on 1.2% agarose gel and transferred onto a nitrocellulose filter paper as described previously (23). To prepare the hybridization probes, either NF1-GRD or v-Ha-Ras DNAs were labeled with [a-32P]dATP using a random labeling kit (Bresatec LTD, Australia) according to the supplier's instruction. The specific radioactivity of each DNA probe was approximately lo9 cpmlpg. The Northern hybridization was performed by blotting the filter with either 32Plabeled NF1-GRD or v-Ha-Ras probes at 42 "C in the presence of 50% formamide, and the filter was subsequently washed at 30 "C with 2 X SSC (24) plus 0.1% SDS for 15 min (twice) and at 40 "C with 0.1 X SSC plus 0.1% SDS for 15 min (twice). The remaining probes on the filter were then scanned by a PhosphorImagerTM-400 series (Molecular Dynamics, California) or radioautographed.
Preparation of Recombinant NFl -GRD Mutants and Ras GTPase-By subcloning the corresponding polymerase chain reaction DNA fragments into the bacterial expression vector pGEX-2TH (22), a NF1-GRD, human type I1 NF1-GRD (residues 1355-1531 plus the series of N-and C-terminal deletion mutants of the human type I insert of 21 amino acids), and murine c-Ha-Ras GTPase (residues 1-189) were produced in Escherichia coli as glutathione S-transferase fusion proteins and affinity-purified as described previously (19,22).
Assay for GAP (Ras GTPase-stimulating) Activity of the NFl -GRD Mutants-The hydrolysis of [-y-32]GTP bound to the c-Ha-Ras GTPase was determined by incubating at 25 "C for 20 min in the presence or absence of each NF1-GRD mutant preparation in 100 pl of a buffer containing 50 mM Tris-HC1, pH 7.5, 15 mM MgC12, 2.5 mM EDTA, 3 mM ATP, 1 mM dithiothreitol, and 0.5 mg/ml bovine serum albumin, precipitating the unhydrolyzed [Y-~'P]GTP with charcoal, and, finally, measuring the radioactivity of the inorganic phosphate in the supernatant as described previously (22). Protein concentrations were determined by the Bradford method with bovine serum albumin as standard (22).

RESULTS AND DISCUSSION
Anti-u-Ha-Ras Action of NF1 -GRD in Mammalian Cells-To examine whether the NF1-GRD is able to reverse the oncogenic action of v-Ha-Ras in mammalian cells, we have subcloned a DNA encoding the NF1-GRD of 338 amino acids (residues 1194-1531) into the retroviral vector pMV7, in both sense (+) and anti-sense (-) orientations, which has the neomycin-resistant selectable marker. The detail of the resultant constructs called NF338+ and NF338-is shown in Fig. 1. The NF1-GRD was overexpressed under the control of the LTR promoter in the v-Ha-Ras-transformed NIH/3T3 cells in both orientations as describedpreviously (22,23). The transfectants were cloned in the presence of a neomycin analogue (G418), and the reversion of their transforming phenotypes, i.e. their morphological changes and loss or reduction of the soft agar colony-forming ability, were examined as described previously (22,23). Unlike the normal NIH/3T3 cells, the majority of v-Ha-Ras-transformed cells expressing the sense NF1-GRD did not look very flat (data not shown). However, overexpression of the sense NF1-GRD in the transfectants greatly reduced their soft agar colony-forming ability ( Fig. 2 and Table I). The clones 22 and 8 formed only 3 (small size) and 37 (medium size) colonies, respectively, whereas the parental v-Ha-Ras-transformed cells formed 735 (large size) colonies. However, the clone 12, which overexpressed the antisense NF1-GRD, did not significantly differ from the parental cells in either their soft agar colony-forming ability (Table I)   * The number of colonies formed in soft agar whose average size is indicated in parentheses (large, more than 100 cells; medium, approximately 25 cells; small, less than 10 cells/colony (see Fig. 28). The relative soft agar colony (SAC)-forming ability of each clone. Each presented value was the average of the data from three independent experiments, and the standard deviation in each case was less than 5%. a good correlation between the sense NF1-GRD mRNA level and the extent of reduction in the soft agar colony-forming ability ( Fig. 3A and Table I). As judged by Northern analysis, the NF1-GRD mRNA level of the clone 22 was much higher than that of the clone 8. Furthermore, Northern analysis of v-Ha-Ras mRNA confirmed that all transfectants still express the v-Ha-Ras as the parental cells (Fig. 3B), indicating that the reduction of their oncogenicity is not due to the loss of v-Ha-Ras expression but instead due to overexpression of the NF1-GRD. This is the first demonstration of the NF1-GRD's interference with the oncogenic action of v-Ha-Ras in mammalian cells.
How does the NF1-GRD interfere with the oncogenic action of v-Ha-Ras? Unlike the c-Ha-Ras GTPase, the v-Ha-Ras GTPase is not activated by the NF1-GRD. Thus, it is unlikely that the conversion of GTP/v-Ha-Ras to GDP/v-Ha-Ras is the basis for the anti-v-Ha-Ras action of the NF1-GRD. Instead, as the NF1-GRD binds Ras much more tightly than GAPl (6), it is more likely that the NF1-GRD competitively inhibits the binding of GAPl or other Ras targets to the v-Ha-Ras, and, therefore, blocks their oncogenic Ras-effector actions. It has been shown previously that either v-Ha-Ras. GAPl complex or the N-terminal SH2 domain of GAPl alone block the muscarinic K+ channels of atrial myocytes (26, 27), indicating that GAPl serves as an effector of v-Ha-Ras, as well as an attenuator of the normal Ras signals. Interestingly, Arshavsky and Bownds (28) have recently demonstrated that another GTPase called transducin is activated by the y-subunit of retinal rod cGMP phosphodiesterase, indicating that the phosphodiesterase-y of only 87 amino acids serves as a transducin GAP. This observation has raised a distinct possibility that, unlike the GAPlC, only a small region of the NF1-GRD might be sufficient to bind Ras GTPases and thereby stimulate the intrinsic GTPase activity of the normal Ras or interfere with the transforming ability of oncogenic Ras mutants. In an attempt to find the smallest Ras-binding fragment of the NF1-GRD, which would potentially be useful for treatment of Ras-associated cancers, as well as to more precisely understand the nature of Ras-NF1 interaction(s), we have recently begun to screen such a candidate among the short GTPase-activating fragments derived from the NF1-GRD.
The NFl Residues Required for GTPase Activation-We have shown previously that the GAPlC, residues 720-1044 of bovine GAP1, is the minimal domain that is still able to activate Ras GTPases (22). Further deletion of either Tyr7*0 or the last 7 amino acids of the GAPlC completely abolished the Ras GTPase activation.' Interestingly, the corresponding domain of NF1, residues 1205-1531 (NF327), is also able to activate Ras GTPases (18, 19). Neither GAPl nor NF1 activates Rap1 or Rho GTPases. The specificity for both GAPl and NF1 was predominantly determined by Gln6* of Ras GTPases, and the sensitivity to both GAPl and NF1 was maximized by G l d 3 (19). * H. Maruta, unpublished observation.
To determine the minimal GTPase-activating domain of the NF1 and further screen the smallest anti-oncogenic NF1 fragment, we have prepared a series of its N-or C-terminal deletion mutants (see Fig. 4) as glutathione S-transferase fusion proteins in E. coli as described previously (19, 22). As shown in Table 11, the deletion of either its N-terminal 236 amino acids (residues 1205-1440) or its C-terminal 35 amino acids (residues 1497-1531) did not abolish the Ras GTPase activation, although these mutations significantly reduced the efficiency of the GTPase activation; the NF1 fragment of 91 amino acids (NF91, residues 1441-1531) still activates Ras GTPase but 20 times less efficiently than the NF1-GRD. However, the further deletion of the N-terminal 11 amino acids (residues 1441-1451) completely abolished the GTPase activation. These observations suggested that ( i ) the 56 amino

FIG. 4. Nand C-terminal deletion mutantsof the NF1-GRD (type I).
A series of the NF1-GRD mutants containing the indicated residues were produced as glutathione S-transferase fusion proteins and affinity-purified as described under "Materials and Methods." Their GAP (Ras GTPase-stimulating) activity is shown in Table 11.
The number following NF of each mutant indicates the total amino acid residues included. The type 11-specific insert of 21 amino acids is indicated by the number 21 Fig. 4.
Activation of Ras GTPase by 20 pg/ml/ml NF1-GRD mutants. The NF1-GRD concentrations required for 50% hydrolysis of GTP bound to Ras. Each presented value was the average of the data from four to six independent experiments, and the standard deviations in each case was less than 5%. acids (residues 1441-1496) of NF1 might be sufficient for GTPase activation, and (ii) some of the 11 amino acids (residues 1441-1451) are essential for its action.
However, the glutathione S-transferase fusion protein containing only these 56 amino acids of NF1 (residues 1441-1496, NF56) was no longer able to activate the Ras GTPases ( Fig. 4 and Table 11). Instead, the slightly larger NF1 fragment of 78 amino acids (residues 1441-1518) linked to glutathione S-transferase was able to activate the Ras GTPase ( Fig. 4 and Table 11). These observations indicate that (i) the 78 amino acids, but not the 56 amino acids, of NF1 are sufficient for GTPase activation, and (ii) some of the 22 amino acids (residues 1497-1518) are functionally replaceable with the Nterminal 236 amino acids (residues 1205-1440) for the proper folding of NF56 or stability of its active site. This N-terminal domain alone did not show any GTPase activation ( Table 11), suggesting that its possible role is the structural rather than the functional.
Since the deletion of the N-terminal 236 amino acids significantly reduced the efficiency of the GTPase activation, we have been identifying the residue(s) that play the key role in maximizing the GTPase activation. Interestingly, deletion of Phe''05 (or its replacement by Ser) alone significantly reduced the GTPase activation of the NF327 (Fig. 4 and Table 11), suggesting that the Phe'205 plays an important role in maximizing the GTPase activation. In this context, it is of interest to note that the corresponding residue (Tyr"") of the GAPlC also appears to play the key role in the GTPase activation; although replacement of Tyr7" by either Phe or Leu does not affect the GTPase activation, either Gh7" or Gly7'0 mutations almost completely abolish the GTPase activation.' These observations suggest that a hydrophobic residue (Tyr, Phe, or Leu) is required at position 720 of the GAPlC or at position 1205 of the NF1 for the full activation of the Ras GTPases. Further deletion of 235 amino acids (residues 1206-1440) from the NF326 also significantly reduced the GTPase activation ( Fig. 4 and Table 11), indicating that some of these amino acids are involved in stimulation of the NF91 and activation of the NF56, which alone is inactive.
It is of interest to note that the GTPase-activating NF1 fragment of 78 amino acids (NF78) shares only 19% sequence identity with the corresponding domain of GAP1, which is totally inactive. Surprisingly, this sequence identity is even lower than that (26%) between the active GAPlC (residues 720-1044) and the corresponding NF1-GRD. Furthermore, although the GAPlC is completely inactivated by either freezing/thawing or storage at 4 "C for 6 months, the GAP activity of both the corresponding NF1-GRD and even the NF78 is not significantly affected by these treatment/storage conditions. These observations suggest that the NF1-GRD and NF78 are significantly different from the corresponding domains of the GAP1 in either the overall conformation or stability of the active site.

Anti-u-Ha-Ras Action of 91-Amino Acid Fragment Derived from NFI-GRD-We
have shown that deletion of the Nterminal 247 amino acids (residues 1194-1440) from NF1-GRD still allows the remaining domain of 91 amino acids (NF91) to activate the normal Ras GTPases, although it significantly reduced (20-fold) its GAP activity. This observation indicates that the NF91 still binds to the Ras GTPases. Thus, to examine whether the N-terminal 247 amino acids are required for the anti-oncogenicity of NF1-GRD against the v-Ha-Ras, we have overexpressed its Ras-binding domain of 91 amino acids (NF91) in the v-Ha-Ras-transformed NIH/ 3T3 cells by means of the vector pMV7. Unlike the parental cells, many NF91-transfected clones became morphologically flat as the normal NIH/3T3 fibroblasts (Fig. 5). Some flat revertants, such as clone 17, were no longer able to form any colony in soft agar, while other flat revertants such as clone 7 formed only a few small colonies in soft agar (see Table I).
These observations suggest that, like the NF1-GRD, the NF91 is anti-oncogenic toward the v-Ha-Ras-transformed NIH/3T3 cells. By Northern blot analysis, we have compared the steady-state levels of v-Ha-Ras and NF91 mRNAs between the two flat revertants and the parental cell line. As shown in Fig. 6, both flat revertants still expressed the v-Ha-Ras gene as the parental cell line, whereas only the flat revertants overexpressed the NF91 gene (slightly higher expression in clone 17 than clone 7). These observations clearly indicate that their loss or reduction of the malignant phenotype is due to overexpression of NF91 gene and not due to a loss of the v-Ha-Ras gene expression. Thus, the N-terminal 247 amino acids of NF1-GRD are not essential for its anti-oncogenicity toward the v-Ha-Ras. It is of interest to note that the NF91 is significantly smaller than the b-Myc N-terminal domain of 120 amino acids, which, so far, has been the smallest among the proteins that reverse v-Ha-Ras-induced malignancy (15).

Relationship between Anti-oncogenicity and GAP Activity of
NFl-It is worth noting that, unlike the NF91 revertants, none of the NF1-GRD revertants is flat, although both revertants form little colony in soft agar, suggesting that the NF91 is more potent than the NF1-GRD as a v-Ha-Ras antagonist in this sense. Conversely, as a GAP (Ras GTPase-activating protein), the NF1-GRD is 20 times more active than the NF91 (see Table 11). These observations have suggested, if not proved as yet, that the poorer as a GAP, the more potent as an anti-oncogene, as long as it still binds the v-Ha-Ras. Interestingly, deletion of the C-terminal 13 amino acids from the NF91 reduces its GAP activity by 2.5-fold (see Table 11). Thus, it is of interest to examine whether the remaining domain of 78 amino acids (NF78) is more potent as a v-Ha-Ras antagonist than the NF91. More importantly, deletion of the C-terminal 35 amino acids (residues 1497-1531) from NF1-GRD still allows the remaining domain to activate the normal Ras GTPases (see Table 11), suggesting that this C-terminal domain is not essential for the Ras binding of NF1-GRD. Unfortunately, the remaining domain of 56 amino acids (NF56) alone, produced by deletion of nonessential N-and C-terminal domains (residues 1194-1440 and 1497-1531), is no longer able to activate the Ras GTPases (see Table 11).
However, it is still conceivable that the NF56 does bind tightly the v-Ha-Ras and blocks the Ras oncogenicity, perhaps more strongly than the NF91 and NF78. Our preliminary results indicate that both NF78 and NF56 are anti-oncogenic as NF91 at the cell culture levels," suggesting that NF56 still binds v-Ha-Ras. We are currently examining whether these NFSl/NF78/NF56-induced flat revertants develop tumors in nude mice or not.
The Role of Lys1423 and the Type 11-specific Insert of 21 Amino Acids Located Outside the GTPase-activating Domain of the NFl-It has been reported previously that replacement of LYS'~'~ by either Glu or Gln in the NF1 almost completely abolishes its GTPase activation (29), suggesting that this residue plays an important role in GTPase activation. However, the Met'423 mutation only converts the NF1 to a temperature-sensitive (ts) molecule but does not significantly affect the GTPase activation (18), suggesting t h a t L Y S '~*~ may be involved in stabilization of the NF1, but is not essential for GTPase activation. We have demonstrated here that either NF78 or NF91 (residues 1441-1518 or 1531), which lack LYS'~'~, are still able to activate Ras GTPases, clearly indicating that this Lys residue is localized outside the GTPase-activating domain of the NF1.
It has been shown recently that there are a t least two different NF1 isoforms (types I and 11) in human cell lines (30). The type I is identical to that previously reported (4, 7), whereas the type I1 contains an additional insertion of 21 amino acids between residues 1370 and 1371 of the type I. Interestingly, the expression of the type I1 is associated with neuronal differentiation (30). In SH-SY5Y neuroblastoma cells, the type I is predominantly expressed. However is suppressed, and the type I1 expression is highly induced along the neuronal differentiation. These observations have implied the distinct roles of the types I and I1 isoforms of the NF1 in cell proliferation and neuronal differentiation, respectively. Our deletion analysis has revealed that this insertion in the type I1 occurs outside the GTPase-activating domain (NF78).
To examine further whether the insertion of 21 amino acids affects the GTPase activation, we have assayed the GAP activity of the NF1 fragment (residues 1355-1531) in the NFI-GRD (pglml) FIG. 7. GAP activity of the NF1-GRD in the presence (0) or absence (0) of the type 11-specific inert. The NF1-GRDs (residues 1355-1531) derived from type I (minus insert) or type I1 (plus insert) were affinity-purified as glutathione S-transferase fusion proteins, and their GAP activities were compared at indicated concentrations as described under "Materials and Methods." presence or absence of the insert. As shown in Fig. 7, the insertion resulted in only 50% reduction of the NFl-dependent c-Ha-Ras GTPase activation. Interestingly, like the type I NF1-GRD, the type I1 NF1-GRD is able to reduce the v-Ha-Ras-induced susceptibility of yeast to heat shock when it is overexpressed (31). Thus, it is not clear whether this insert alone is sufficient for playing any important role in its regulation of Ras GTPase activation or its anti-v-Ha-Ras action in vivo. It is also conceivable that its interaction with the Nterminal non-GTPase-activating domain of the NF1 or with other molecule(s) plays an essential role in the induction of neuronal differentiation. We are currently identifying the possible target(s) of the type 11-specific insert.
Another NF1 cDNA has recently been cloned from rat, which encodes a third isoform of the NF1 (32). This isoform, called type 111, is identical to both types I and I1 in the Nterminal half (residues 1-1370) but differs from type I, as it contains the additional insert of 21 amino acids as type 11.
The type I11 differs from type I1 as it lacks entirely the Cterminal half (residues 1371-2818) including the GTPaseactivating domain (NF78). Thus, it is most likely that the type I11 is unable to activate Ras GTPases. It is intriguing to identify the possible function of the N-terminal non-GTPaseactivating half of the NF1 (type 111) in mammalian cells.