Recognition of RhoA by Clostridium botulinum C3 exoenzyme.

The C3-like ADP-ribosyltransferases exhibit a very confined substrate specificity compared with other Rho-modifying bacterial toxins; they selectively modify the RhoA, -B, and -C isoforms but not other members of the Rho or Ras subfamilies. In this study, the amino acid residues involved in the RhoA substrate recognition by C3 from Clostridium botulinum are identified by applying mutational analyses of the nonsubstrate Rac. First, the minimum domain responsible for the recognition by C3 was identified as the N-terminal 90 residues. Second, the combination of the N-terminal basic amino acids ((Rho)Arg(5)-Lys(6)), the acid residues (Rho)Glu(47) and (Rho)Glu(54) only slightly increases ADP-ribosylation but fully restores the binding of the respective mutant Rac to C3. Third, the residues (Rho)Glu(40) and (Rho)Val(43) also participate in binding to C3 but they are mainly involved in the correct formation of the ternary complex between Rho, C3, and NAD(+). Thus, these six residues (Arg(5), Lys(6), Glu(40), Val(43), Glu(47), and Glu(54)) distributed over the N-terminal part of Rho are involved in the correct binding of Rho to C3. Mutant Rac harboring these residues shows a kinetic property with regard to ADP-ribosylation, which is identical with that of RhoA. Differences in the conformation of Rho given by the nucleotide occupancy have only minor effects on ADP-ribosylation.

been overcome by the construction of chimeric C3 toxins, which exploit the cell entry domains of other toxins such as diphtheria or C. botulinum C2 toxin (11,12).
The RhoA, -B, and -C proteins belong to the Rho subfamily, which encompasses (in addition to Rho) Rac, Cdc42, RhoD, Rnd/RhoE, RhoG, and TC10. These low molecular mass GT-Pases are involved in the regulation of the actin cytoskeleton, membrane trafficking, cell cycle progression, cell transformation, and apoptosis. The function of the GTPases is tightly governed by regulatory proteins such as the exchange factors, the GTPase-activating proteins, and the guanine nucleotide dissociation inhibitors. Downstream signaling is mediated through interaction with effector proteins, which are serine/ threonine-kinases (Rho kinases, citron kinase, protein kinase N), lipid kinases (phosphatidyl inositol kinases), phospholipases (phospholipase D), and adapter proteins (rhotekin, rhophilin, p140Dia, myosin binding subunit) (13)(14)(15)(16). The finding, that C3-catalyzed ADP-ribosylation of RhoA in intact cells causes depolymerization of the actin filaments, led to the notion that ADP-ribosylation at Asn 41 inactivates Rho. ADPribosylation has no significant influence on nucleotide exchange, intrinsic and GTPase-activating protein-stimulated GTPase activity but clearly decreases the exchange activity of Lbc (17)(18)(19). Furthermore, the guanine nucleotide dissociation inhibitor-driven cycling between cytosol and membranes is blocked leading to sequestration of ADP-ribosylated Rho in the inactive guanine nucleotide dissociation inhibitor complex. 1 This sequestration seems to be an additional important functional consequence finally resulting in the inactivation of Rho.
The Rho proteins are also targets for other bacterial toxins. The Clostridium difficile toxins A and B glucosylate Rho at Thr 37 but also modify Rac and Cdc42 (21,22). A comparable substrate specificity exhibits the cytotoxic necrotizing factor from Escherichia coli, which catalyzes deamidation of the residue Gln 63 of RhoA (23,24). Among the Rho-modifying toxins, the C3-like exoenzymes exhibits the most confined substrate specificity to exclusively ADP-ribosylate RhoA, -B, and -C. The Rho-related GTPases Rac and Cdc42 also harbor the acceptor amino acid asparagine (Asn 39 in Rac and Cdc42) and are highly homologous in the vicinity of the acceptor amino acid but are not substrates for C3. We studied, therefore, which amino acids of Rho define the substrate specificity for C3 and whether it is possible to construct a mutant Rac that is a substrate for C3.

EXPERIMENTAL PROCEDURES
Materials and Chemicals-RhoA, Rac1, and the mutant forms were purified as GST 2 fusion proteins from E. coli followed by thrombin cleavage. Thrombin was removed by precipitation with benzamidine-* This work was supported by Deutsche Forschungsgemeinschaft Grant SFB 388 and by a German Israelian Foundation research grant. 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.
Sepharose. Oligonucleotides were obtained from MWG (Ebersberg, Germany), the pGEX-2T vector and the glutathione S-transferase gene fusion system were from Amersham Pharmacia Biotech, the Quickchange kit was from Stratagene (Heidelberg, Germany), and the restriction enzymes were from NEB (Schwalbach, Germany). The secondary anti-rabbit peroxidase-conjugated antibody was from Rockland (Gilbertsville, PA), the enhanced chemiluminescence detection kit was from Amersham Pharmacia Biotech. [ 32 P]NAD was from NEN (Belgium). All other chemicals were from commercial sources. C. botulinum exoenzyme C3 was prepared as described (25). Anti-C3 IgG was from rabbit (5).
Construction of Mutant Rac1 or RhoA Proteins-The RhoA or Rac1 mutants were constructed by site-directed mutagenesis with the pGEX2T-wild type human RhoA or the pGEX2T-wild type human Rac1 vector as templates and the respective oligonucleotides using the Quickchange kit according to the manufacturer's instructions.
The PCR reactions were carried out with PfuI polymerase according to the manual from Stratagene or with Taq-DNA polymerase (Roche Molecular Biochemicals) and the TOPO-TA vector system from Invitrogen (Leek, The Netherland) in a Gene Amp 2400 PCR system from Perkin-Elmer. DNA was sequenced with the cycle sequencing ready reaction kit (ABI PRISM) from Perkin-Elmer to verify the correct mutations.
Construction of RhoA90Rac1 Chimera-This chimera was created using the pGEX2T vectors mentioned above and a splicing by overlapping extension ("SOE") PCR reaction (26). The primers are: SOE RhoA, 5Ј-gac att ttc aaa tga atc agg gct gtc gat gga and SOE RhoA C, 5Ј-gaa aat gtt cga aga tcg t including a BstBI site, which are needed to amplify the first 90 amino acids of RhoA; and SOE Rac1, 5Ј-gac agc cct gat tca ttt gaa aat gtc cgt gc and Rac1 N, 5Ј-cca ttg ctg cag gca tc including a PstI site, which are needed to amplify the amino acids 89 -192 of Rac1. The products of each PCR were separated by agarose gel electrophoresis, purified, and put in a third PCR as template for each other. As a result, a 1.8-kilobase fragment was obtained and ligated in the TA vector system. Sequential digestion with BstBI and PstI resulted in mobilization of the PCR product, which can be ligated directly into a pGEX2T, cut with the same enzymes. The resulted plasmid was transformed in E. coli XL-1 Blue supercompetent cells (Stratagene).
Nucleotide Binding-Binding of N-methylanthraniloyl-GDP was used to prove structural and functional integrity of the mutant Rho and Rac proteins. The GTPases (0.5 M), dissolved in buffer C (10 mM Tris-HCl, pH 7.5, 2.5 mM MgCl 2, 150 mM NaCl) were warmed to 37°C, and the nucleotide exchange was initiated by the addition of N-methylanthraniloyl-GDP to give a final concentration of 2 M. The samples were measured in a fluorescence cuvette with a excitation wavelength of 357 nm and a emission wavelength of 444 nm. The binding of N-methylanthraniloyl-GDP was monitored as an increase in fluorescence. Wild type and mutant Rho/Rac proteins showed the same increase in fluorescence.
ADP-ribosylation Reaction-2 g of wild type RhoA, wild type Rac1, and the mutant proteins were incubated with 1 g/ml (43 nM) or 0.1 g/ml (4.3 nM) C3 (as indicated) in a buffer containing 50 mM HEPES (pH 7.3), 2 mM MgCl 2 , 1 mM dithiothreitol, 100 g/ml bovine serum albumin, 20 M [adenylate-32 P]NAD at 37°C for up to 15 min. For nucleotide-dependent ADP-ribosylation, wild type RhoA (15 M) was incubated with 3 mM GDP, GTP␥S, or Gpp(NH)p (a kind gift of Dr. C. Herrmann, Dortmund) in a buffer containing 10 mM EDTA, 50 mM HEPES (pH 7.3), 1 mM dithiothreitol, and 100 g/ml bovine serum albumin for 60 min on ice. The nucleotide exchange was terminated by the addition of 13 mM MgCl 2 , and the Rho proteins were subjected to [ 32 P]ADP-ribosylation. The reaction was stopped by the addition of Laemmli buffer, and the 32 P-labeled proteins were resolved by 12.5% SDS-PAGE followed by analysis with a PhosphorImager.
Binding of C3 to RhoA, Rac1, and the Mutant Forms-RhoA, Rac1, and the indicated mutant forms were used as GST fusion proteins immobilized to Sepharose. 5 g of the GST fusion proteins were incubated under mild agitation with 100 ng of C3, or as indicated, in a binding buffer containing 50 mM HEPES (pH 7.3), 2 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 0.3 mM phenylmethylsulfonyl fluoride (total volume of 100 l) at 4°C for 45 min. The Sepharose beads were washed once with 100 l of binding buffer followed by elution with 10 mM reduced glutathione. The eluted samples were separated by 12.5% SDS-PAGE followed by probing with anti-C3, and visualization was performed with enhanced chemiluminescence. The exposed films were evaluated by scanning and subsequent quantification. The nonspecific binding (binding to GST) was subtracted, and the remaining intensity was given as relative units.
To calculate the percentage of bound C3, 450 ng of C3 in the binding buffer was incubated with RhoA immobilized to Sepharose beads; thereafter the beads were processed as described above. Samples were drawn before the addition of RhoA from the supernatant after binding and from the RhoA-bound fraction. The amount of C3 in each sample was assessed by immunoblotting. NAD-Glycohydrolase-For detection of glycohydrolase activity, 200 mM [adenylate-32 P]NAD was incubated with or without 1 M of C3 and 5 M of the indicated GTPases in 10 mM HEPES (pH 7.3) and 2 mM MgCl 2 at 37°C for 3 h. Samples were filtrated through a 10-kDa Centricon, and the protein-free flow through was separated by TLC on Silica Gel 60 F 254 (Merck) with 66% 2-propanol and 0.33% ammonium sulfate. The amount of hydrolyzed [ 32 P]NAD was calculated from Phos-phorImager data.

RESULTS
To test whether the complete structure of Rho or only subdomains define the substrate recognition by C3 exoenzyme, chimeras between RhoA and Rac1 were constructed (Fig. 1) and tested for ADP-ribosylation. Only when the N-terminal part was Rho-like, the chimeras were modified by C3 (Fig. 1). Thus, the minimum part, which makes the difference between Rac and Rho with respect to ADP-ribosylation, is determined by amino acids 1-90. The non-ADP-ribosylatable chimeras were native in structure as tested by nucleotide binding (data not shown) except the Rho73Rac chimera, which was therefore excluded.
Sequence comparison of the N-terminal 90 amino acids of RhoA with Rac1 revealed several differences (Fig. 2). The basic stretch at the N terminus (position Rho Arg 5 -Lys 6 ), which is absent in Rac, and the acidic residues Rac Asp 28 , Asp 45 , Glu 47 , Glu 54 , Asp 76 , Asp 87 , and Asp 90 , which are not acidic in Rac, were inserted and exchanged, respectively. Mutant Rac1 proteins possessing the respective amino acid(s) of RhoA were then tested for ADP-ribosylation as depicted in Table I. Only the insert of the basic residues equivalent to Rho Arg 5 -Lys 6 or the exchange of Rac Met 45 and Rac Asn 52 to glutamic acid led to an increase in ADP-ribosylation, however, only to a minor degree. Therefore, combinations of those mutations identified to affect ADP-ribosylation were tested. The basic insert (residues Rac Arg 5 -Lys 6 ) together with Rac Glu 47 and Rac Glu 54 led to a significant increase in ADP-ribosylation of this mutant Rac (Table I; note that the insert of the two basic residues changes the numbering of the Rac1 residues). To compare the kinetics of the ADPribosylation reaction between wild type RhoA and Rac Arg 5 -Lys 6 -Glu 47 -Glu 54 , the concentration of C3 was decreased from 1 g/ml to 0.1 g/ml (4 nM) to extend the linear phase. Surprisingly, the mutant Rac protein ( Rac Arg 5 -Lys 6 -Glu 47 -Glu 54 ) was only faintly modified, indicating that the so far identified residues only partially contributed to the substrate recognition by C3 (Fig. 3).
As can be deduced from the crystal structure of RhoA, Asn 41 is surrounded by solvent-exposed lipophilic residues, such as   ADP-ribosylation (Fig. 4A), a finding supporting the importance of Rho Glu 40 and Rho Val 43 for Rho ADP-ribosylation. As expected, the exchange of Rho Trp 58 completely blocked ADPribosylation, whereas the exchange of Rho Phe 39 had no effect (data not shown). Surprisingly, the exchange of the respective amino acids in Rac1 (Asp 38 3 Glu, Ser 41 3 Val) did not significantly increase ADP-ribosylation of Rac (Fig. 4B), suggesting that Rho Glu 40 and Rho Val 43 are necessary but not sufficient for ADP-ribosylation. Therefore, those residues found to contribute to ADP-ribosylation were mixed in various combinations, and the multiple mutant Rac constructs were tested for ADPribosylation. As illustrated in Fig. 3, only Rac, which harbored the basic insert (Arg 5 , Lys 6 ) Rac Glu at positions 40, 47, and 54 and the lipophilic Rac Val 43 , was completely modified. In addition to complete ADP-ribosylation, this mutant Rac showed a kinetic behavior that was indistinguishable from that of wild type RhoA (Fig. 3). Thus, these residues are the essential ones allowing C3 to transfer the ADP-ribose moiety.
The residues identified to promote ADP-ribosylation by C3 were tested whether they participate in substrate recognition, i.e. binding to C3. To this end, the binding of multiple mutant Rho and Rac proteins to C3 was studied by applying a precipitation assay with immobilized mutant GST fusion proteins. Preliminary experiments showed that C3 binding was different from nonspecific binding. As illustrated in Fig. 5A, C3 bound to wild type RhoA but not to wild type Rac whose binding was comparable to control signal. The binding of the poor ADPribosylatable Rac Arg 5 -Lys 6 -Glu 47 -Glu 54 and Rac Glu 38 -Val 41 was comparable to C3 binding to RhoA. Furthermore, increasing concentrations of C3 resulted in increased binding but not in increased nonspecific binding. The amount of RhoA-bound C3 was about 25% (Fig. 5B). As shown in Fig. 5C, Rac Arg 5 -Lys 6 -Glu 47 -Glu 54 and Rac Glu 38 -Val 41 exhibited the same binding as the full substrate mutant Rac ( Rac Arg 5 -Lys 6 -Glu 40 -Val 43 -Glu 47 -Glu 54 ). Thus, the combination of all mutations, which contributed to ADP-ribosylation, did not further increase binding to C3. Consistent with this finding was that the exchange of Rho Glu 40 and Rho Val 43 did not alter the binding to C3. The exchange of Rho Trp 58 , which abolished ADP-ribosylation had no effect on C3 binding. Thus, there are two types of binding and only the combination makes Rac1 a substrate. It is conceivable that the residues involved in mere binding to C3 also participate in the correct formation of the ternary complex, consisting of RhoA, C3, and NAD ϩ . Especially, the residues Rho Glu 40 and Rho Val 43 , which are adjacent to the acceptor residue Asn 41 could be involved in the correct positioning of the catalytic site of C3 in respect to the acceptor amino acid Asn 41 . To prove this working hypothesis, we made use of the fact that the NAD glycohydrolase activity of C3 was increased by mutant Rho and deficient in the acceptor amino acid Asn 41 ( Rho Ile 41 ) (Fig. 6). This observation seems to be because of the fact that the formation of the ternary complex increases the rate of glycohydrolysis probably by decreasing the k m for NAD. Although both Rho Ile 41 and Rho Asp 40 -Ser 43 bound to C3 and were non-ADPribosylatable, only Rho Ile 41 was capable of increasing the glycohydrolase activity, whereas Rho Asp 40 -Ser 43 even blocked it (Fig. 6). The inhibitory effect of Rho Asp 40 -Ser 43 might be because of an incorrect interaction of C3 with RhoA not allowing the correct binding of NAD ϩ .
The acceptor amino acid Rho Asn 41 is located in the effector loop (switch I) of RhoA, which undergoes conformational changes upon GTP binding. To test whether nucleotide occupancy affects the accessibility of the amide moiety of Asn 41 , the nucleotide dependence of the ADP-ribosylation reaction was studied. To exclude GTPase activity to form GDP during the ADP-ribosylation reaction, the nonhydrolyzable GTP ana-logues Gpp(NH)p and GTP␥S were used. Fig. 7 shows the linear phase of the ADP-ribosylation reaction of RhoA loaded with either GDP, Gpp(NH)p, or GTP␥S. The rate of ADPribosylation of RhoA-GDP was about five times faster than that of RhoA-Gpp(NH)p/GTP␥S. The total amount of ADP-ribose incorporation, however, was not changed (Fig. 7, inset). Thus, the nucleotide occupancy had only minor effects on the kinetics of the transferase reaction.

DISCUSSION
The C3-like ADP-ribosyltransferases are characterized by a very selective substrate specificity to exclusively modify the Rho isoforms A, B, and C, which possess a homology of about 90%. This specificity is striking, because other Rho-modifying bacterial toxins exhibit less specificity. Furthermore, this spec- FIG. 5. Binding of C3 to wild type or mutant forms of Rho or Rac. A, wild type RhoA and Rac1 as well as multiple mutant Rac (each 1 g of protein) were incubated with increasing concentrations of C3. The bound fraction freed from nonspecific binding is shown. For details see "Experimental Procedures." B, to calculate the percentage of bound C3, 450 ng of C3 was incubated with RhoA immobilized to Sepharose beads; thereafter the beads were processed as described under "Experimental Procedures." Samples were drawn before the addition of RhoA, from the supernatant after binding, and from the RhoA-bound fraction. The amount of C3 in each sample was assessed by immunoblotting (shown). C, the GTPases (as indicated) immobilized to Sepharose beads were incubated with C3 followed by washing and glutathione elution. The released complexes were blotted and probed for C3 by immunoblot analysis. Lane 1, GST control; 2, RhoA wild type; 3, Rac1 wild type; 4, Rac1 R5K6 M47E N54E; 5, Rac1 R5K6 D40E M47E N54E; 6, Rac1 R5K6 D40E S43V M47E N54E; 7, RhoA E40D V43S; 8, Rac1 D38E S41V; 9, RhoA N41I; 10, RhoA W58V. ificity is the basis why C3 is an established and widely used tool in cell biology to selectively shut off the cellular functions of Rho. The first step to identify the regions of Rho involved in the interaction with C3 was to test chimeras between substrate RhoA and the nonsubstrate Rac1. The substrate recognition domain is at least restricted to the N-terminal part covering amino acids 1-90.
Comparison of this part from RhoA with the Rho isoforms B and C revealed that only RhoB has one single amino acid exchange. However, Rac shows differences, i.e. the absence of the basic N-terminal stretch and less acidic amino acids distributed over the N-terminal part. Single and combined mutational analysis of the nonsubstrate Rac revealed that the insertion of the basic stretch and the exchange to the respective residues of RhoA ( Rho Glu 47 and Rho Glu 54 ) clearly increased ADP-ribosylation of this multiple mutant Rac. However, neither the extent nor the time course of ADP-ribosylation were identical with the substrate RhoA. Therefore, we focused on the crystal structure of RhoA (27). In that study, the authors point out the unusual feature of the solvent exposed lipophilic amino acids Rho Val 38 , Phe 39 , Val 43 , and Trp 58 . However, only the position of Rho Val 43 is different in Rac, which is taken by a hydrophilic serine ( Rac Ser 41 ). Indeed, Rho Val 43 is important for ADP-ribose transfer because its exchange leads to a loss of ADP-ribosylation, whereas the conserved Rho Phe 39 is not in-volved in substrate recognition. The proposed function of the conserved Rho Trp 58 in the ADP-ribosylation reaction to facilitate the reaction through increase of the nucleophilicity of the acceptor moiety (27) was fully supported by mutagenesis; the exchange of Trp to a nonaromatic amino acid completely inhibited ADP-ribosylation. In addition, also the precursor residue of Rho Asn 41 , Rho Glu 40 , is different in Rac, which is Rac Asp 38 . Because the main-chain carbonyl group of Rho Glu 40 interacts with the side chain of Rho Asn 41 , the exchange to Asp (the side chain is one carbon atom shorter) should not affect ADP-ribosylation. However, exchange to Asp blocks ADP-ribosylation of Rho. The side-chain carbonyl group of Rho Glu 40 is solvent exposed, as can be deduced from the crystal structure, and the only explanation is that the acidic Glu finger is involved in binding to C3. Surprisingly, the introduction of the respective Glu and Val into Rac has only a minimum effect on ADP-ribosylation. Only the combination of these two residues together with the residues mentioned above ( Rac Arg 5 -Lys 6 -Glu 40 -Val 43 -Glu 47 -Glu 54 ) generates a mutant Rac protein, which meets the requirements for substrate recognition by C3. From the binding experiments of the wild type and mutant RhoA and Rac1, respectively, it became clear that the basic residues at the N terminus together with acidic residues ( Rho Glu 47 and Rho Glu 54 ) and the Rho Glu 40 / Rho Val 43 residues are independently responsible for the mere binding to C3. The combination, i.e. the generation of the 6-fold mutant, does not increase binding. The residues involved in binding of Rho to C3 are exclusively located in the N-terminal part; thus, it is unlikely that the C-terminal part of RhoA contributes to the binding.
However, the residues Rho Glu 40 / Rho Val 43 have a profound function in the transfer of the ADP-ribose moiety because their exchange results in abolished ADP-ribosylation. In the case of the NAD-glycohydrolase reaction, a binary complex is formed between C3 and NAD ϩ , and the rate of reaction is much slower (factor of 100) compared with the transferase reaction. The presence of non-ADP-ribosylatable Rho ( Rho Ile 41 ) increases the glycohydrolase activity most likely by increasing the affinity of NAD to the now existing ternary complex. In contrast, mutant Rho lacking Glu 40 and Val 43 , which still binds to C3, does block the NAD-glycohydrolase activity, most likely through its inability to form the correct ternary complex. Taken together, the residues Glu 40 and Val 43 are directly involved in the correct formation of the ternary complex formed by the exoenzyme C3, the protein substrate RhoA, and the co-substrate NAD ϩ .
The acceptor amino acid for ADP-ribosylation, Asn 41 , resides in the effector loop, as it can be concluded from the effects of partial loss of function mutants. The exchange of Rho Tyr 42 , the residue adjacent to Asn 41 , to Cys decreases PLD1 activation, whereas mutation of Asn 41 has no effect on stimulation (28). Also the exchange of Glu 40 , the residue just before Asn 41 , inhibits effector activation, namely the RhoA stimulation of Rho kinase (ROK) (29). Switch I undergoes conformational changes during nucleotide exchange but the solvent accessibility of the side chain of Asn 41 does not change, as can be deduced from the crystal structures of Rho-GTP and Rho-GDP (27,30). The nucleotide-dependent time course of the ADP-ribosylation revealed that Rho-GDP is the optimal substrate and that binding of the GTP analogues Gpp(NH)p or GTP␥S decreases the rate of reaction by a factor of five but does not change the total amount of ADP-ribosylation. Thus, conformational changes induced by nucleotide binding have only a minor effect on C3catalyzed ADP-ribosylation. However, this effect is only gradually compared with that of monoglycosylation of RhoA at Thr 37 by Clostridium novyi ␣-toxin and of H-Ras at Thr 35 by the Clostridium sordellii lethal toxin (31,32). Because the hydroxyl group of Rho Thr 37 ( Ras Thr 35 ) points to the inner core of RhoA (50 g/ml) was loaded with either GDP, GTP␥S, or Gpp(NH)p as described under "Experimental Procedures" followed by C3-catalyzed [ 32 P]ADP-ribosylation; samples from the linear phase of the reaction were analyzed by SDS-PAGE followed by phosphoimaging. Incorporation of ADP-ribose into the GTPase was given as mol/mol. The inset gives the ADP-ribosylation after 60 min. the protein (20,33), RhoA-GTP (Ras-GTP) is resistant to glucosylation. Therefore, the GDP form of RhoA is exclusively modified. Because the nucleotide occupancy has only effects on the kinetics but not the total amount of ADP-ribosylation, C3-catalyzed ADP-ribosylation is in fact not a reliable marker to determine the inactive fraction of Rho in cells.
In conclusion, it turned out that the basic stretch and at least three acidic amino acids are involved in the binding to C3. The N-terminal part of Rho has more acidic amino acids at the surface of the molecule than Rac. Thus, the specificity of Rho to serve as substrate for C3 is determined by several amino acids distributed over the N-terminal part of Rho. It is not possible to generate mutant Rac to be substrate of C3 by single or double exchange of amino acids. At least six residues are responsible for the interaction with C3 and thus define the selective interaction with Rho.