Identification of the Region of Rho Involved in Substrate Recognition by Escherichia coli Cytotoxic Necrotizing Factor 1 (CNF1)*

The Escherichia coli cytotoxic necrotizing factor 1 (CNF1) and the Bordetelladermonecrotic toxin (DNT) activate Rho GTPases by deamidation of Gln63 of RhoA (Gln61 of Cdc42 and Rac). In addition, both toxins possess in vitro transglutaminase activity in the presence of primary amines. Here we characterized the region of Rho essential for substrate recognition by the toxins using Rho/Ras chimeras as protein substrates. The chimeric protein Ras55Rho was deamidated or transglutaminated by CNF1. Rat pheochromocytoma PC12 cells microinjected with Ras55Rho developed formation of neurite-like structures after treatment with the CNF1 holotoxin indicating activation of the Ha-Ras chimera and Ras-like effects in intact cells. The Ras59Rho78Ras chimera protein contained the minimal Rho sequence allowing deamidation or transglutamination by CNF1. A peptide covering mainly the switch II region and consisting of amino acid residues Asp59 through Asp78 of RhoA was substrate for CNF1. Changes of amino acid residues Arg68 or Leu72 of RhoA into the corresponding residues of Ras (R68ARhoA and L72QRhoA) inhibited deamidation and transglutamination of the mutants by CNF1. In contrast to CNF1, DNT did not modify Rho/Ras chimeras or the switch II peptide (Asp59 through Asp78). Glucosylation of RhoA at Thr37 blocked deamidation by DNT but not by CNF. The data indicate that CNF1 recognizes Rho GTPases exclusively in the switch II region, whereas the substrate recognition by DNT is characterized by additional structural requirements.

The Escherichia coli cytotoxic necrotizing factor 1 (CNF1) and the Bordetella dermonecrotic toxin (DNT) activate Rho GTPases by deamidation of Gln 63 of RhoA (Gln 61 of Cdc42 and Rac). In addition, both toxins possess in vitro transglutaminase activity in the presence of primary amines. Here we characterized the region of Rho essential for substrate recognition by the toxins using Rho/Ras chimeras as protein substrates. The chimeric protein Ras55Rho was deamidated or transglutaminated by CNF1. Rat pheochromocytoma PC12 cells microinjected with Ras55Rho developed formation of neurite-like structures after treatment with the CNF1 holotoxin indicating activation of the Ha-Ras chimera and Ras-like effects in intact cells. The Ras59Rho78Ras chimera protein contained the minimal Rho sequence allowing deamidation or transglutamination by CNF1. A peptide covering mainly the switch II region and consisting of amino acid residues Asp 59 through Asp 78 of RhoA was substrate for CNF1. Changes of amino acid residues Arg 68 or Leu 72 of RhoA into the corresponding residues of Ras (R68ARhoA and L72QRhoA) inhibited deamidation and transglutamination of the mutants by CNF1. In contrast to CNF1, DNT did not modify Rho/Ras chimeras or the switch II peptide (Asp 59 through Asp 78 ). Glucosylation of RhoA at Thr 37 blocked deamidation by DNT but not by CNF. The data indicate that CNF1 recognizes Rho GTPases exclusively in the switch II region, whereas the substrate recognition by DNT is characterized by additional structural requirements.
Rho GTPases (e.g. Rho, Rac, and Cdc42) participate in the regulation of the actin cytoskeleton (1,2). Whereas Rho subtype proteins induce formation of stress fibers and adhesion complexes, Rac is involved in formation of lamellipodia and Cdc42 induces microspikes (3)(4)(5). Beside their roles in the organization of the actin cytoskeleton, Rho proteins act as molecular switches in various signal transduction processes (6,7).
Rho proteins are the preferred substrates for several bacterial protein toxins. Exoenzyme C3 from Clostridium botulinum and related C3-like transferases ADP-ribosylate RhoA, B, and C at Asn 41 thereby inhibiting the biological activity of the GTPases (8 -11). Rho proteins are monoglucosylated by members of the family of large clostridial cytotoxins (e.g. Clostrid-ium difficile toxins A and B) (12)(13)(14). The toxins modify RhoA at Thr 37 (Thr 35 of Rac and Cdc42), a modification which blocks the interaction of the GTPases with their effectors (15,16).
Rho family GTPases are also the targets for cytotoxic necrotizing factors (CNF) 1 1 and 2 from Escherichia coli and the dermonecrotizing toxin (DNT) produced by various Bordetella species. CNF and DNT are ϳ115 and ϳ165 kDa proteins which share a region of homology at their C termini harboring the enzyme domain of the toxins (17). In culture cells, the toxins induce actin polymerization and inhibit cytokinesis resulting in formation of multinucleated cells (18 -20). Recently it has been reported that CNF and DNT act on Rho GTPases by deamidation of glutamine 63 of RhoA, thereby inhibiting the GTPase activity of Rho. Because Gln 63 is essential for GTP hydrolysis, deamidation causes persistent activation of the GTPase resulting in strong formation of stress fibers of CNF-or DNT-treated cells. In addition to their deamidase activity, both toxins possess in vitro transglutaminase activity to attach primary amines onto Rho GTPases. Substrates of CNF and DNT are Rho subfamily members including Rac and Cdc42.
In the present communication the substrate recognition of Rho GTPases by CNF1 and DNT was studied. Using GTPase chimeras of RhoA and Ha-Ras which is not a substrate of the toxins, we identified the switch II region of Rho as being sufficient for recognition by CNF1. Accordingly, a peptide consisting of amino acid residues Asp 59 through Asp 78 of RhoA was deamidated and/or transglutaminated by CNF1. By contrast, the structural requirements for substrate recognition by DNT are more stringent.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant Proteins-For protein purification, E. coli strains carrying pGEX plasmids with the coding sequence for the respective GTPases, GTPase chimera, CNF1 (either as full-length or as the catalytic C-terminal part (⌬CNF1, amino acid residues 709 -1014), or DNT (the catalytic C-terminal part (⌬DNT, amino acid residues 1136 -1451) were grown in LB medium and induced with 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside at OD 0.5. The cells were harvested after 3-6 h and the glutathione S-transferase (GST) fusion proteins were purified by means of glutathione-Sepharose (Amersham Pharmacia Biotech). Unstable proteins were kept as GST fusion proteins (Rho115Ras, Ras59Rho115Ras, ⌬DNT, and full-length CNF1); otherwise, the proteins were subjected to thrombin cleavage to remove GST.
Construction of GTPase Chimeras-The GTPase chimeras were constructed using the splicing by overlap extension method described pre-viously (21). In brief, two polymerase chain reactions (PCR) were performed, each amplifying the sequences to be fused. The products of the two first PCRs were pooled and cycled in a third PCR, in which the overhanging sequences act as primers. We used pGEX2T-RhoA (wildtype, WT) and pGEX2TGL-Ha-Ras (wild-type) as templates with the different primers listed in Table I and pGEX primers either annealing 5Ј or 3Ј of the multiple cloning site of pGEX (Table I). The product of the second PCR was subjected to BamHI and EcoRI cleavage and subcloned into a pGEX2TGL vector. The sites of splicing are numbered by Rho nomenclature. The "sandwich" chimeras Ras55Rho70Ras and Ras59-Rho70Ras were constructed with splicing by overlay extension by using the primers for Ras55Rho and Ras59Rho, respectively, with Rho70Ras serving as template. Ras59Rho115Ras was constructed by using the primers for Ras59Rho with Rho115Ras as a template. All constructs were checked for proper sequences.
Mutagenesis-Mutants were constructed from pGEX2T-RhoA (wildtype) or pGEX2T-Ha-Ras (wild-type) by PCR in the presence of two primers (sense and corresponding antisense) carrying a base mismatch encoding the proper mutant. The parental DNA was eliminated using the restriction enzyme DpnI, which digests methylated DNA. The PCR products were transformed into Epicurean Blue XL-1 ultracompetent cells (Stratagene). After verifying the mutation by sequencing, the plasmids were transformed into E. coli BL21 for protein expression. The sense primers for the various mutants are listed in Table II.
Cell Culture and Microinjection-Rat pheochromocytoma cells (PC12) were cultivated on plastic dishes in Dulbecco's modified Eagle's medium supplemented with 10% horse serum, 5% fetal calf serum, penicillin (4 mM), and streptomycin (4 mM) and kept at 37°C in a humidified atmosphere with 10% CO 2 . The cells were microinjected with 0.5 g/l GST-Ras55Rho or buffer by means of a microinjector 5242 (Eppendorf) and subsequently treated for 48 h with full-length GST-CNF1 (300 ng/ml).
Transglutamination Assay-GTPase or chimeric protein (3-6 g) was incubated with or without CNF1 or DNT in a buffer containing 50 mM Tris-HCl (pH 7.4), 10 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, and 50 mM monodansylcadaverine for 15 min at 37°C. The proteins were separated by SDS-PAGE and the transglutaminated proteins were detected under UV light (206 nm). Subsequently, the gel was Coomassie-stained to check the amount of proteins in each lane.
Release of Ammonia-To determine the release of ammonia from the GTPases during incubation with CNF1, we utilized a coupled enzymatic reaction. In the presence of glutamate dehydrogenase and NADH, ammonia reacts with ␣-ketoglutarate to form L-glutamate, thereby oxidating the NADH. The amount of NADH oxidized in the reaction is stoichiometric to the amount of ammonia produced. RhoA, Ha-Ras, or Rho/Ras chimeras (each 10 M) were incubated in a buffer containing 5 mM Tris-HCl (pH 7.4), 10 mM MgCl 2,, 1 mM dithiothreitol, 1 mM EDTA and the components of the coupled enzyme reaction system (Roche Molecular Biochemicals, Ammonia Test Combination for Food Analysis): 140 M NADH, 7 mM ␣-ketoglutarate, and 10 units of glutamate dehydrogenase. The samples were equilibrated at 37°C prior to addition of CNF1 as indicated in the figure legends. The decrease in NADH concentration was measured fluorimetrically (excitation at 340 nm, emission at 460 nm) in a Perkin-Elmer LS 50B luminescence spectrometer.
Mass Spectrometry-RhoA switch II peptide (1 mM) was incubated for 3 h at 37°C with or without CNF1 (1 M) in the presence of 20 mM ethylenediamine in a buffer containing 25 mM triethanolamine (pH 7.5), 75 mM NaCl, 1 mM CaCl 2 , 2 mM dithiothreitol, 10 mM MgCl 2 , and 1 mM EDTA. To obtain a concentration of the RhoA switch II peptide of 10 M, the reaction mixture was diluted with TA (acetonitrile, 0.1% trifluoroacetic acid, 1:1). A saturated matrix solution of recrystallized 4-hydroxy-␣-cyanocinnamic acid (Aldrich) in TA was freshly prepared and marker peptides (ACTH 18 -39 clip, human, and angiotensin II, human) were added for internal calibration. Matrix and peptide solution were mixed in equal amounts. Using the dried-drop method of matrix crystallization, 1 l of the sample matrix solution was placed on the MALDI stainless steel target and allowed to air-dry several minutes at room temperature. MALDI/TOF-MS was performed on a Bruker Biflex mass spectrometer equipped with a nitrogen laser ( ϭ 337). Mass spectra were recorded in the reflector positive mode in combination with delayed extraction.

Rho But Not Ras Is Deamidated by CNF1-It has been
shown that Rho, Rac, and Cdc42 are deamidated by CNF1 (23). To test whether Ha-Ras is a substrate for CNF1, the release of ammonia occurring concomitantly with deamidation was determined in a coupled enzyme assay (see "Experimental Procedures"). Fig. 1 shows that addition of the active fragment of CNF1 (⌬CNF1) caused release of ammonia in the presence of RhoA but not in the presence of Ras indicating that the latter GTPase is not modified by CNF1. Because Rho and Ha-Ras differ in the amino acid residues flanking the site of modification in Rho (Fig. 2), we tested a Ha-Ras mutant with an aspartic acid residue instead of glutamic acid at position 63 of Ha-Ras. Like wild-type Ha-Ras, the mutant E63D Ras was not deamidated by ⌬CNF1 (not shown).
Identification of the Substrate Recognition Site of Rho-By exploiting the finding that Ha-Ras is not a substrate for ⌬CNF1, chimeras of Rho and Ha-Ras were constructed to identify the minimal amino acid sequence of Rho allowing the modification by ⌬CNF1. First, we constructed Rho115Ras and Ras55Rho chimeras (note that the Rho nomenclature is used), which both have the switch II region of RhoA. The chimeras were tested in a transglutaminase assay with monodansylcadaverine as a co-substrate. This primary amine contains a naphthalene group, which is excitable by UV light at 206 nm. Fig. 3 shows the proteins analyzed by SDS-PAGE and photographed under UV light. As observed for wild-type Rho, the Ras55Rho chimera was transglutaminated by ⌬CNF. Also the Rho115Ras chimera was substrate for ⌬CNF. This chimera was less stable than Ras55Rho. Therefore, we studied Rho115Ras as a GST fusion protein and compared it with GST fusion proteins of wild-type RhoA and Ha-Ras.
Hall and co-workers reported that amino acid residues 23 through 46 of Ras are sufficient to elicit Ras-dependent transformation in NIH3T3 cells (24). Constitutively active Ras is known to induce neurite formation in PC12 cells (25). Therefore, we tested whether CNF1 was able to induce neurite for- mation in PC12 cells microinjected with Ras55Rho prior to toxin treatment. PC12 cells were grown on plastic dishes, microinjected with 0.5 g/l GST-Ras55Rho, and treated with GST fusion protein of the CNF1 holotoxin (300 ng/ml) for 48 h. Whereas Ras55Rho-injected cells without CNF1 showed only small processes (Fig. 4A), injected cells developed long neurite extensions after CNF1 treatment (Fig. 4B) indicating an activation of the Ras chimera by CNF1.
Because experiments with the Ras55Rho and Rho115Ras chimeras suggested that neither the N terminus nor the C terminus of Rho are essential for substrate recognition by ⌬CNF1, we studied whether sandwich (Ras/Rho/Ras) chimeras are modified by the toxin. The sandwich chimera, Ras59Rho115Ras, was substrate for deamidation by ⌬CNF1 (not shown). Further reduction of the length of the RhoA insert in Ha-Ras resulted in the chimera Ras59Rho78Ras. We tested this chimera in the transglutamination assay at increasing concentrations of ⌬CNF1. The chimera was a substrate for transglutamination, however, whereas wild-type Rho was significantly transglutaminated at 10 nM ⌬CNF1, higher concentrations of CNF1 (about 10-fold) were required for transglutamination of Ras59Rho78Ras (Fig. 5). Chimeras containing very short Rho sequences of only 8 or 5 amino acid residues (Ras55Rho70Ras and Ras59Rho70Ras) did not serve as substrates of CNF1 (data not shown). To exclude that these results were due to incorrect protein folding, we tested the ability of the chimeras to bind nucleotide. Both chimeras (Ras55Rho70Ras and Ras59Rho70Ras) were able to bind 2Ј(3Ј)-O-(N-methylanthraniloyl)-GDP indicating a proper folding of the proteins (not shown). The results obtained with the different chimeras are summarized in Fig. 6.
Identification of Amino Acid Residues in RhoA Which Are Essential for CNF1 Substrate Recognition-Deduced from recent crystal structure analysis of Rho (26) and by means of the "Rasmol" program we identified four amino acids (Arg 68 , Leu 72 , Lys 98 , and His 105 ) between residues 59 through 115 of Rho which are surface exposed and are identical in Rho, Rac, and Cdc42. These amino acids were changed to the corresponding Ras residues. All mutant proteins were capable of binding 2Ј,3Ј-O-N-methylanthraniloyl-GDP in the nucleotide binding assay (not shown) indicating a correct protein folding. We tested the deamidation of the mutant RhoA proteins by ⌬CNF1 in the ammonia release assay. As shown in Fig. 7, A and B, whereas the K98Q and the H105R mutants were deamidated by ⌬CNF1, L72Q and R68A mutants did not serve as substrates or were only marginally modified by the toxin. The rate of the modification of the mutant K98Q varied between differ- ent experiments and was not consistently slower than the wild-type. To investigate whether arginine 68 and leucine 72 are sufficient for CNF1 substrate recognition, a Ha-Ras double mutant carrying an arginine residue at position 66 and a leucine residue at position 70 was constructed and tested for activity in the transglutaminase and ammonia release assay. However, this mutant was not modified by ⌬CNF1 (not shown).
A Peptide Corresponding to the RhoA Switch II Region Is Deamidated by CNF1-A peptide of the sequence 59 DTAG-QEDYDRLRPLSYPDTD 78 which covers the switch II region of RhoA was tested for its ability to serve as substrate for CNF1 in the ammonia release assay. The switch II peptide was compared with RhoA (10 M) at two different concentrations (10 and 100 M) in the presence of 1 M ⌬CNF1. As shown in Fig. 8, after addition of ⌬CNF1, ammonia was released from the peptide, however, at a slower rate than with RhoA. The difference in the modification rates was calculated as a 110-fold decrease for the peptide in comparison with the recombinant protein (Sigma Plot). A shorter peptide with the sequence 59 DTAGQEDY-DRLR 70 did not release ammonia after addition of ⌬CNF1 (not shown). To study whether the switch II peptide (Asp 59 -Asp 78 ) was modified by transglutamination, the peptide was treated with CNF1 in the presence of ethylenediamine. Thereafter, the sample was analyzed by MALDI-TOF mass spectrometry. As shown in Fig. 9, after ⌬CNF1 treatment a new peptide characterized by an increase in mass by 43 Da was detected indicating the attachment of ethylenediamine onto the switch II peptide.
Comparison of CNF1 with the Related Toxin DNT-Next we compared ⌬CNF with the related Bordetella deamidase DNT in respect to substrate specificity and substrate recognition. The active C-terminal part of Bordetella DNT (amino acids 1136 -1451, ⌬DNT) was applied as a GST fusion protein because GST-⌬DNT proved to be more stable than ⌬DNT. For comparison of the two enzymes, ⌬CNF1 was also used as a GST fusion protein.
To investigate whether DNT requires the same amino acid residues of Rho for substrate recognition as CNF1, the transglutaminase activities of both toxins were studied in the presence of monodansylcadaverine with the RhoA mutants (R68A, L72Q, K98Q, and H105R) mentioned above (Fig. 10). Like ⌬CNF1, ⌬DNT did not modify the mutants R68A RhoA and L72Q RhoA with alterations in the switch II region. K98Q RhoA and H105R RhoA were transglutaminated by both toxins.
We tested the Ras55Rho and Rho115Ras chimeras with GST-⌬CNF1 and GST-⌬DNT in the transglutamination assay. In contrast to GST-⌬CNF1, both chimeras were not modified by GST-⌬DNT (not shown). It has been reported that unlike CNF1, DNT modifies RhoA only in the GDP-bound form (27). To exclude the possibility that GTP binding caused inhibition  of modification by DNT, chimera Ras55R40 was loaded with GDP prior to toxin treatment. However, the Rho chimeras were not modified by DNT even after loading with GDP (not shown). The above findings suggested that the substrate recognition by DNT is more stringent than for CNF1. Because nucleotide binding causes major conformational changes in the switch I region of Rho, we studied the effects of glucosylation of RhoA at Thr 37 by C. difficile toxin B on modification by DNT. As shown in Fig. 11, prior glucosylation of Rho by toxin B inhibited the transglutamination by DNT but not by CNF1. By contrast, ADP-ribosylation of Rho at Asn 41 by exoenzyme C3, a modification that occurs downstream of the switch I region did not affect transglutamination by CNF1 or DNT. DISCUSSION In contrast to various bacterial protein toxins inactivating small GTPases including large clostridial cytotoxins and exoenzyme C3 (28), the E. coli cytotoxic necrotizing factors CNF1 and CNF2 activate Rho GTPases. The latter toxins deamidate Rho GTPases at Gln 63 of Rho (Gln 61 of Rac and Cdc42) to inhibit the endogenous and GAP-stimulated GTP hydrolysis thereby blocking the switch off reaction of the GTPase (29,30). Recently, we reported that CNF1 possesses in vitro transglutaminase activity in addition to deamidase activity (31,32). Here we attempted to identify the recognition site of Rho GTPases for CNF1.
Rozengurt and co-workers (33) reported that CNF1 does not activate the mitogen-activated protein kinases p42 mapk or p44 mapk in Swiss 3T3 cells, suggesting that Ras is not activated by the toxin. In line with this notion, we confirmed that recombinant Ras is not modified by CNF1. To get more information about the structural requirements for substrate recognition by CNF1, we constructed several chimeras consisting of Rho and Ras. The enzymatically active toxin fragment ⌬CNF1 (amino acid residues 709 -1014 of CNF1) deamidated and transglutaminated all chimeras consisting of Ras harboring the switch II region (Ala 61 through Asp 78 ) of Rho.
These results indicate that the switch II region of Rho is sufficient for substrate recognition by CNF1 and neither the switch I region nor the insert region (Asp 124 through Gln 136 ) are required for interaction with CNF1. This notion was corroborated by the findings that the peptide Asp 59 through Asp 78 covering the switch II region of RhoA was capable to serve as a substrate for the active fragment of CNF1. Similar as found for the chimera Ras59Rho78Ras, the peptide was a poorer substrate for CNF1 than wild-type RhoA, suggesting that additional residues, although not essential for modification by the toxin, increase the substrate properties of RhoA. With one exception (Asp 76 of Rho, which is glutamine in Rac and Cdc42), all Rho proteins including Rac and Cdc42 possess highly conserved amino acid sequences in the switch II region. Thus, the substrate requirements observed in this study are in agreement with recent findings that Rac and Cdc42 are modified by CNF1 (23) and it seems likely that all Rho GTPases, including those not studied so far, are also substrates of the toxin.
To identify amino acids which are possibly involved in interaction with CNF1, we selected surface-exposed amino acids (Arg 68 , Leu 72 , Lys 98 , and His 105 ) for site-directed mutagenesis by using the crystal structure data of RhoA (26). Replacement of Arg 68 or Leu 72 of RhoA with the corresponding amino acid in Ha-Ras (R68A and L72Q) prevented modification of the mutant proteins. By contrast, replacement of His 105 and Lys 98 , which are located outside the switch II region, with the corresponding residues of Ha-Ras did not alter the modification of RhoA, and the rate of ammonia release from H105R RhoA was greater than that from wild-type Rho. Although both Arg 68 and Leu 72 of RhoA are essential for recognition by CNF1, they are not sufficient for modification. This is concluded from the finding that replacement of the corresponding amino acids in Ras by arginine and leucine did not make it a substrate for CNF1.
The observation that the switch I region of Rho is not important for substrate recognition by CNF1 in vitro was confirmed in intact cells. Microinjection of the Ras55Rho chimera into PC12 cells and subsequent treatment with the CNF1 holotoxin, which in contrast to ⌬CNF1 (amino acids 709 through 1014) is able to enter cells, caused a typical Ras-like response, e.g. neurite outgrowth. Addition of CNF1 to buffer-injected cells did not lead to this phenotype (not shown). These results are explained by interaction of the Ras55Rho chimera with Ras effectors mediated by the intact switch I region of Ras and the deamidation of the chimera at Gln 63 (Rho nomenclature) thereby preventing the inactivation of the chimeric GTPase. These findings are in agreement with the recent report by Hall and co-workers (24) that a constitutively active Ras/Rho chimera, containing the switch I region of Ras caused the typical Ras effects such as cell transformation (24). Moreover the observed neurite outgrowth is clearly a Ras phenotype, because in contrast to microinjection of dominant active Ras (G12V) the injection of dominant active RhoA (G14V) does not induce neurite outgrowth in PC12 cells (not shown).
The Bordetella dermonecrotizing toxin DNT is another toxin which activates Rho GTPases. This toxin shares significant similarity at its C terminus (amino acids 1136 -1451) with the active region of CNF1 (17,34). Like CNF1, DNT catalyzes deamidation of Gln 63 of Rho (35). Moreover, DNT possesses high transglutaminase activity (36). We compared the structural requirements for substrate recognition by ⌬DNT and ⌬CNF1. As with ⌬CNF1, changes of Arg 68 and Leu 72 to the equivalent amino acids of Ras inhibited modification by ⌬DNT. However, in contrast to ⌬CNF1 (comparison between GST-⌬DNT and GST-⌬CNF1), ⌬DNT did not accept any of the Rho/Ras chimeras as substrates for deamidation or transglutamination. Accordingly, the peptide representing the switch II region of RhoA (Asp59 through Asp78) was not modified by ⌬DNT. Because neither the Ras55Rho nor the Rho115Ras chimera were modified by ⌬DNT, it can be concluded that structural determinants located both N-and C-terminal to the switch II region are essential for substrate recognition by DNT. Thus, it is obvious that the structural requirements for substrate recognition by DNT are more stringent than for CNF1. In line with this notion are recent reports that modification of Rho by DNT, but not by CNF1, depends on the nucleotide bound to Rho (27). Using the active fragment ⌬DNT, we confirmed that GDP, but not the GTP␥S bound form of Rho is a substrate for toxin-catalyzed deamidation and transglutamination. These findings suggest that in addition to the switch II region, the switch I region, which undergoes major changes upon nucleotide binding, is involved in substrate recognition by DNT. A role of the switch I region in substrate recognition by DNT is also supported by the finding that prior glucosylation of Rho by C. difficile toxin B impaired DNT-induced deamidation or transglutamination. Toxin B glucosylates Rho at Thr 37 which is located in the switch I region (13). ADP-ribosylation of Rho by C3 at Asn 41 , which is located outside the switch I region, did not alter DNT effects on Rho. By contrast, as substrate recognition by CNF1 depends almost exclusively on the sequence of the switch II region, it seems logical that nucleotide binding of Rho and glucosylation or ADP-ribosylation by toxins do not alter modification of the GTPase by CNF1.
In summary, the switch II region of Rho was identified as the substrate recognition site for CNF1. Minimal structural requirements for substrate recognition by CNF1 are provided by a peptide covering amino acid residues Asp 59 through Asp 78 of Rho. By contrast, the substrate recognition by DNT is much more complex requiring, in addition to the switch II region, further sites including the switch I region of Rho.