Identification of the C-terminal Part of BordetellaDermonecrotic Toxin as a Transglutaminase for Rho GTPases*

Bordetella dermonecrotic toxin (DNT) causes the deamidation of glutamine 63 of Rho. Here we identified the region of DNT harboring the enzyme activity and compared the toxin with the cytotoxic necrotizing factor 1, which also deamidates Rho. The DNT fragment (ΔDNT) covering amino acid residues 1136–1451 caused deamidation of RhoA at glutamine 63 as determined by mass spectrometric analysis and by the release of ammonia. In the presence of dansylcadaverine or ethylenediamine, ΔDNT caused transglutamination of Rho. Deamidase and transglutaminase activities were blocked in the mutant proteins Cys1292 → Ala, His1307 → Ala, and Lys1310 → Ala of ΔDNT. Deamidation and transglutamination induced by ΔDNT blocked intrinsic and Rho- GTPase-activating protein-stimulated GTPase activity of RhoA. ΔDNT deamidated and transglutaminated Rac and Cdc42 in the absence and presence of ethylenediamine, respectively. Modification of Rho proteins by ΔDNT was nucleotide-dependent and did not occur with GTPγS-loaded GTPases. In contrast to cytotoxic necrotizing factor, which caused the same kinetics of ammonia release in the absence and presence of ethylenediamine, ammonia release by ΔDNT was largely increased in the presence of ethylenediamine, indicating that ΔDNT acts primarily as a transglutaminase.

adhesion kinase and paxillin (10,12,13). Recent studies indicate that DNT causes deamidation of glutamine 63 of RhoA (10). Glutamine 63 is essential for GTP hydrolysis by Rho. Deamidation of glutamine by DNT inhibits the GTPase activity of Rho and renders the Rho protein constitutively active.
The same mechanism of Rho activation by deamidation was reported for the cytotoxic necrotizing factor CNF1 from Escherichia coli (14,15). Also CNF deamidates Rho at glutamine 63 and causes similar cytotoxic effects such as multinucleation of cells and stress fiber formation. CNF1 and DNT share a region of homology (amino acid residues 1250 -1351 of DNT) located at the C termini of the toxins (16). Other parts of the protein sequences are not significantly similar. Recently, it was shown that a C-terminal fragment of CNF1 (⌬CNF), covering the region of homology, causes the typical cytotoxic effects after microinjection and possesses full Rho-deamidating activity in vitro. In addition to deamidase activity, ⌬CNF possesses transglutaminase activity. However, this activity is observed only in the presence of high concentrations of primary amines and is apparently lower than the deamidase activity (17).
Here we attempted to identify the region of DNT that harbors the enzyme activity of the toxin and characterized its biological and biochemical activities. We report that ⌬DNT covering amino acid residues 1136 -1451 possesses full deamidating activity. Cysteine 1292, histidine 1307, and lysine 1310 are essential for enzyme activity. As found for ⌬CNF, the active fragment of DNT possesses transglutaminase activity. In contrast to CNF1, ⌬DNT exhibits a higher transglutaminase than deamidase activity, indicating that DNT acts preferentially as a transglutaminase. Another difference between ⌬CNF and ⌬DNT is the nucleotide dependence of the deamidation/transglutamination reaction. Whereas ⌬CNF modifies GDP-and GTP-loaded Rho proteins, ⌬DNT exclusively accepts GDPbound RhoA.

EXPERIMENTAL PROCEDURES
Materials-RhoA and p50 RhoGAP (obtained from A. Hall, London) were prepared from their fusion proteins as described. Dansylcadaverine and ethylenediamine were purchased from Sigma. Methanol and chloroform were of analytical grade, and trifluoroacetic acid and acetonitrile were of high pressure liquid chromatography grade.
The PCR product was purified from agarose gel (Jet sorb, Genomed) and amplified in the pCR TM II vector (Invitrogen) by means of TA cloning. From this vector the DNT fragment was cut with BamHI and EcoRI, purified, and ligated into the digested pGEX vector. The proper construct was checked by DNA sequencing. The vector was transformed into BL21 cells by heat shock at 42°C. Expression of the GST fusion protein in E. coli BL21 cells growing at 37°C was induced by adding 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside (final concentration) at OD 0.5. 6 h after induction, cells were collected and lysed by sonication in * This work was supported by the Sonderforschungbereich 388. 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.
Activation of FXIIIa-Activation of FXIII occurs through thrombin cleavage of the a-chains in the presence of calcium ions. 10 M human factor XIII a-chains (Centeon) were incubated with 2 g/l thrombin for 30 min at room temperature in reaction buffer containing 150 mM NaCl, 50 mM triethanolamine, and 8.5 mM CaCl 2 . Thrombin was then removed by incubation with benzamidine-Sepharose for 10 min at room temperature. The activity of FXIII was tested with fibronectin as a substrate.
Measurement of Ammonia-For qualitative measurement of ammonia a coupled enzymatic reaction was used that was based on the ammonia test combination for food analysis (Roche Molecular Biochemicals). NADH was diluted to give a concentration of 50 M with triethanolamine buffer containing 2-oxoglutarate, 20 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, and 1 mM EDTA. Ten units of GlDH and RhoA (final concentration 10 M) were added. After the addition of ⌬CNF1 or ⌬DNT (each 1 M), the decrease in NADH fluorescence was monitored in a Perkin-Elmer LS-50B luminescence spectrometer. The emission was measured at 460 nm with excitation at 340 nm.
For quantitative analysis of the ammonia release, Rho proteins (200 M) were incubated with ⌬CNF1 (1 M) or ⌬DNT (1 M) in a reaction buffer containing 20 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 8 mM CaCl 2 , 1 mM dithiothreitol, and 1 mM EDTA at 37°C. The reaction was stopped at different time points by incubation for 1 min at 95°C. Denatured proteins were removed by centrifugation, and ammonia produced was measured in the supernatant. Decrease in absorbance was measured following the instructions given for the ammonia test combination for food analysis (Roche Molecular Biochemicals).
Microinjection and Actin Staining-For microinjection, NIH3T3 cells were seeded subconfluently on glass coverslips (CELLocate, Eppendorf) and cultivated for 24 h in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum in 5% CO 2 at 37°C. After serum starvation GST-⌬DNT (2 mg/ml) or buffer was microinjected into NIH3T3 cells with a Microinjector 5242 (Eppendorf). 6 h after microinjection, cells were fixed with 4% formaldehyde and 0.1% Tween 20 in phosphate-buffered saline at room temperature for 10 min. For actin staining, formaldehyde-fixed cells were intensively washed with phosphate-buffered saline. The cells were then incubated with rhodamineconjugated phalloidine (1 unit/coverslip) at room temperature for 1 h, washed again, and applied for fluorescence microscopy (as bleaching preservative KAISER'S glycerol gelatin (Merck) was used).
Modification of GTPases by GST⌬-CNF1 or GST-⌬DNT-Small GTPases were incubated with GST-⌬CNF1 or GST-⌬DNT in the presence of monodansylcadaverine or ethylenediamine (50 mM) in transglutamination buffer (20 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 8 mM CaCl 2 , 1 mM dithiothreitol, 1 mM EDTA) for the indicated times at 37°C. As a control, RhoA was incubated without the toxins but in the presence of a cosubstrate. The molar ratio of toxin:RhoA was 1:20. Labeling of proteins with the fluorescent lysine analog dansylcadaverine was analyzed by fluorescence activity under UV light before staining and drying the gel.
GTPase Assay-Recombinant Rho proteins were modified by ⌬CNF1 or transglutaminase in the presence or absence of primary amines. The reaction was stopped by freezing in liquid nitrogen. After thawing the proteins were loaded with [␥-32 P]GTP for 5 min at 37°C in loading buffer (50 mM Tris-HCl, pH 7.5, 10 mM EDTA, 2 mM dithiothreitol). MgCl 2 (12 mM, final concentration) and unlabeled GTP (2 mM, final concentration) were added. For stimulation of GTPase activity by Rho-GAP, 50 nM p50 RhoGAP were added to 1 M Rho, and incubation was for 4 min at 37°C. GTPase activity was analyzed by filter binding assay as described (15).
Treatment of ⌬DNT with N-Ethylmaleimide-GST-⌬DNT was incubated with different concentrations of N-ethylmaleimide in 50 mM Tris-HCl, pH 7.5, for 30 min at room temperature. N-ethylmaleimide was than inactivated by adding dithiothreitol in a molar ratio of 10:1 (dithiothreitol:N-ethylmaleimide) for 10 min. For modification of RhoA, the GTPase was incubated with N-ethylmaleimide-treated or -untreated toxin in the presence of 50 mM ethylenediamine in transglutamination buffer (20 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 8 mM CaCl 2 , 1 mM dithiothreitol, 1 mM EDTA) for 30 min at 37°C.
Proteolytic Digestion in the Gel Matrix for Mass Spectrometric Analysis-The excised gel plugs of Rho A were destained for 1 h at 50°C in 40% acetonitrile, 60% hydrogen carbonate (50 mM, pH 7.8) to remove Coomassie Blue, gel buffer, SDS, and salts. The plug was subsequently dried in a vacuum centrifuge for 15 min. Thereafter, 30 l of digestion buffer with trypsin was added, and digestion was carried out for 12 h at 37°C.
Sample Preparation for Matrix-assisted Laser Desorption Ionization-Mass Spectrometry-4-Hydroxy-␣-cyanocinnamic acid (Aldrich) was recrystallized from hot methanol and stored in the dark. Saturated matrix solution of 4-hydroxy-␣-cyanocinnamic acid in a 1:1 solution of acetonitrile/aqueous 0.1% trifluoroacetic acid was prepared. 2 l of the proteolytic peptide mixture were mixed with 2 l of saturated matrix containing marker peptides (5 pmol of human ACTH (18 -39) clip (MW 2466, Sigma) and 5 pmol of human angiotensin II (MW 1047, Sigma), respectively) for internal calibration. Using the dried-drop method of matrix crystallization, 1 l of the sample matrix solution was placed on the matrix-assisted laser desorption ionization stainless-steel target and was allowed to air dry several minutes at room temperature resulting in a thin layer of fine granular matrix crystals.
Mass Spectrometry-Matrix-assisted laser desorption ionization/ time of flight-mass spectrometry was performed on a Bruker Biflex mass spectrometer equipped with a nitrogen laser (l ϭ 337 nm) to desorb and ionize the samples. Mass spectra were recorded in the reflector positive mode in combination with delayed extraction. External calibration was routinely used, and internal calibration with two points that bracketed the mass range of interest was additionally performed to consolidate peptide masses further. The computer program mass spectrometry-digest (Peter Baker and Karl Clauser, UCSF Mass Spectrometry Facility) was used for computer-assisted comparison of the tryptic peptide mapping data with the expected set of peptides.

RESULTS
The C-terminal Fragment DNT 1136 -1451 (⌬DNT) Is Sufficient for Deamidase and Transglutaminase Activity-Recently, it was shown that the CNF1 of E. coli activates Rho proteins by deamidating glutamine at position 63 of RhoA or 61 of Rac and Cdc42. Moreover, it has been reported that CNF1 possesses transglutaminase activity. CNF1 and DNT share a region of homology (amino acid residues 1250 -1351 of DNT) located at their C termini (16). Therefore, we studied whether the Cterminal fragment ⌬DNT (amino acid residues 1136 -1451 of the holotoxin) is sufficient for the enzyme activity.
To analyze the activity of GST-⌬DNT, we constructed the vector pGEX-⌬DNT, expressed the toxin fragment as a GST fusion protein, and purified it by affinity chromatography. Because the fusion toxin exhibited full activity and was not cleavable without degradation of ⌬DNT, we used the fusion toxin (which is termed ⌬DNT in the text) throughout the entire study. After incubation of RhoA with ⌬DNT for 15 min, the GTPase shifted to an apparent higher molecular mass in SDS-PAGE indicating deamidation of Rho (15) (Fig. 1). In the presence of the transglutaminase cosubstrate ethylenediamine, however, RhoA shifted slightly to an apparent lower molecular mass. Recently, we reported that a downward shift of RhoA in SDS-PAGE corresponding to transglutamination was obtained when the GTPase was incubated with ⌬CNF and ethylenediamine (17). Therefore, we analyzed tryptic peptides of ⌬DNTtreated RhoA by mass spectrometry. As shown in Fig. 2B, the mass analysis of the tryptic digest of the upper band of ⌬DNTtreated RhoA revealed the RhoA peptide Gln 52 -Arg 68 exhibiting a mass shift of one dalton in comparison to untreated RhoA ( Fig. 2A). Tryptic digest of the downward shifted band of RhoA resulted in identification of the same peptide (Gln 52 -Arg 68 ) but with a mass shift of 43 Da as compared with the control protein. This increase in mass indicated the transglutamination of Gln 63 by ethylenediamine (Fig. 2C). Also dansylcadaverine, a fluorescent primary amine (18), served as a cosubstrate for the transglutamination of RhoA by ⌬CNF1 (note that also ⌬CNF1 was used as the GST fusion protein) and ⌬DNT. To compare the transglutamination activity of the toxin fragments, RhoA was incubated with the enzymes in the presence of dansylcadaverine, and the amount of GTPase modified was analyzed in SDS-PAGE under UV light. As shown in Fig. 3, RhoA was dansylated by ⌬DNT to a larger extent than by ⌬CNF. To compare the transglutaminase activities of both toxins in more detail, kinetic studies were performed.
⌬DNT Is Preferentially a Transglutaminase-To compare kinetics of the deamidation and transglutamination reaction of ⌬CNF1 and ⌬DNT, the time course of ammonia release induced by the toxins was studied. The ammonia release assays were performed with a substrate concentration of 200 M RhoA and an enzyme (GST-⌬DNT or GST-⌬CNF1) concentration of 1 M. As shown in Fig. 4A, no difference in the production of ammonia was observed with or without ethylenediamine when RhoA was modified by ⌬CNF1. On the contrary, ⌬DNT released a higher amount of ammonia in the presence of ethylenediamine than in the absence of the primary amine (Fig. 4B). A similar result was obtained in the presence of increasing concentrations of ethylenediamine. As shown in Fig. 5A, with ⌬DNT the production of ammonia increased in an ethylenediamine concentration-dependent manner. In contrast, the addition of ethylenediamine at increasing concentration had no effect on ammonia production by ⌬CNF1. Thus, all these data indicate that ⌬DNT is preferentially a transglutaminase. Similarly, blood clotting factor FXIII, which is a mammalian transglutaminase (19), released a higher amount of ammonia in the presence of the primary amine than in its absence (not shown).
It is known that the activity of mammalian transglutaminases including FXIII is dependent on calcium ions (19). To test whether Ca 2ϩ ions affect the activity of ⌬DNT, we measured the ammonia release induced by the toxin fragment in the absence and presence of EGTA. Ammonia release caused by FXIII was dependent on the presence of Ca 2ϩ ions, whereas the presence of EGTA had no (5 mM) or a very small (10 mM) effect on the activity of ⌬DNT (not shown). In the presence of EGTA, both ⌬DNT and ⌬CNF1 modified RhoA by dansylation (not shown).
⌬DNT Modifies Gln 63 of RhoA and Gln 61 of Cdc42 and Rac-Blood clotting factor FXIII, which modifies various protein substrates such as fibronectin, actin, and casein, transglutaminates three of the five glutamine residues of RhoA, whereas CNF1 is specific for Gln 63 of RhoA and Gln 61 of Cdc42 and Rac1 (17). To investigate the specificity of ⌬DNT, we incubated RhoA, Rac1, Cdc42, the respective Q63E/Q61E mutants, actin, and casein in the presence of dansylcadaverine with ⌬DNT for 30 min at 37°C. Thereafter, transglutaminated proteins were analyzed by SDS-PAGE and UV light exposure. As shown in Fig. 6, ⌬DNT modified the wild-type GTPases RhoA, Cdc42, and Rac1 but not the Q63E/Q61E mutants, actin, or casein. In contrast, FXIII modified wild-type and mutant GTPases, actin, and casein (not shown). In line with the above observations, no ammonia was released during incubation of the Q63E mutant with the toxins (not shown).
Deamidation Kinetics-To compare kinetics of the deamidation/transglutamination reactions of RhoA, Rac1, and Cdc42 induced by the toxin fragments, we measured ammonia release in a time course. The reactions were performed with a protein substrate concentration of 200 M, an enzyme concentration of 1 M and 20 mM ethylenediamine. In Fig. 7, the time courses of ⌬DNT-induced ammonia release of RhoA, Cdc42, and Rac1 are shown as the mean of three independent experiments. All Rho proteins exhibited similar kinetics of ammonia release. Similarly, ⌬CNF1 did not show major differences in the kinetics of ammonia release between the three GTPases (not shown).
Effects of Transglutamination of RhoA-Gln 63 of RhoA is known to be important for the intrinsic and GAP-stimulated GTPase mechanism of the protein (20). To analyze whether transglutaminated protein is still able to hydrolyze GTP, we measured its p50 RhoGAP -stimulated GTPase activity. Fig. 5B illustrates the effects of ⌬CNF1 and ⌬DNT on the GTPase activity of RhoA in the presence of increasing concentrations of ethylenediamine. Similar as observed for the ammonia release (Fig. 5A), inhibition of the GTPase with ⌬CNF1 was independent of the ethylenediamine concentration, whereas the blockade of the GTP hydrolysis with ⌬DNT increased with increasing concentration of the primary amine. Thus, inhibition of GTPase activity and ammonia release induced by ⌬DNT correlated very well, indicating that the transglutamination inhibits GTPase activity of RhoA.
In Vivo Effects of ⌬DNT (Microinjection Experiments)-It has been shown by Horiguchi et al. (10) that treatment of cells with DNT leads to actin polymerization and stress fiber formation because of activation of RhoA. To investigate whether ⌬DNT possesses the same cytotoxic effect in intact cells, we microinjected the toxin fragment as a GST fusion protein into quiescent NIH3T3 cells. The toxin fragment caused formation of stress fibers after 6 h of incubation. However this effect was not as strong as observed with ⌬CNF1 (not shown). This may be because of instability of the GST-⌬DNT fusion protein, which significantly decreased in activity after a few days of storage at 4°C or after incubation for 30 min at 37°C.
Structure-Function Analysis of ⌬DNT-Recently, cysteine was identified to be a functionally essential residue in ⌬CNF1, which is most likely located in the active site of the enzyme. Like ⌬CNF, ⌬DNT contains a single cysteine residue in a protein region highly similar to ⌬CNF (Fig. 9). According to the findings with CNF, treatment of the toxin fragment with iodoacetamide or N-ethylmaleimide blocked the enzyme activity of ⌬DNT (not shown). Exchange of cysteine 1292 with serine or alanine largely decreased or completely inhibited the enzyme activity of ⌬DNT, respectively. Moreover the exchange of histidine 1307 with alanine blocked the enzyme activity of ⌬DNT in analogy to CNF1 (not shown). A nucleotide binding motif has been described for DNT (not present in CNF) covering residues 1304 -1311 (AFYHTGKS) with the consensus (A/ G)XXXXGK(S/T) (16). To study the relevance of this motif for ⌬DNT activity, we changed lysine 1310 to alanine. This mutation blocked ⌬DNT activity of the toxin fragment as already reported for the holotoxin (11).
We observed differences between the toxins in respect to the nucleotide dependence of the deamidation/transglutamination reactions. Fig. 8 shows the dansylation of V14RhoA previously loaded with GDP or GTP and of wild-type RhoA loaded with GDP or GTP␥S. Free nucleotide was removed by gel filtration before modification by the toxins. Whereas ⌬CNF catalyzed the deamidation reaction independently of the nucleotide bound, ⌬DNT accepted GDP-loaded RhoA or GDP-loaded V14RhoA as a substrate but did not modify GTP␥S-bound RhoA or GTPloaded V14RhoA. To test whether the activity of ⌬DNT was regulated by nucleotides via a direct interaction, the toxin was pretreated with nucleotides or nucleotides were added to the reaction mixture. In both cases, we were not able to obtain any evidence for an inhibition of ⌬DNT activity by a direct interaction of the enzyme with GTP␥S (not shown).

DISCUSSION
Recently Horiguchi et al. (10) showed that DNT from Bordetella modifies Rho GTPases by deamidation of Gln 63 . A similar deamidation of Rho at Gln 63 was reported for CNF1 from E. coli (14,15). DNT and CNF share a significant sequence homology in a rather small part of the proteins, suggesting that the deamidase activity is located in this region of the toxins. Therefore, we constructed ⌬DNT, which covered this homologous region (Fig. 9). This fragment consisting of amino acid residues 1136 -1451 possessed full deamidase activity and typically caused an upward shift of RhoA in SDS-PAGE. This change in migration in SDS-PAGE was not observed with the Q63E mutant of RhoA, confirming that exclusively Gln 63 was deamidated. Thus, the active fragment ⌬DNT exhibited the same biochemical properties as reported for the holotoxin DNT. Moreover, similarly as observed for the holotoxin but to a smaller extent microinjection of GST-DNT caused formation of stress fibers in fibroblasts.
Recently, we reported that CNF1 possesses transglutaminase activity and modifies Rho GTPases in the presence of primary amines (17). In the presence of ethylenediamine, transglutamination of RhoA by CNF caused a downward shift of the GTPase in SDS-PAGE. However, this activity of CNF was only observed at high concentrations of the primary amine and occurred slower than deamidation. Similarly as with CNF, we detected a downward shift of RhoA in SDS-PAGE after treatment with ⌬DNT in the presence of ethylenediamine. The transglutamination of Rho by ⌬DNT was verified by mass spectrometric analysis. To further characterize the enzyme activity of ⌬DNT in more detail and to compare it with ⌬CNF, we applied an ammonia release assay. Interestingly, we observed that the release of ammonia by ⌬DNT was largely dependent on the presence of ethylenediamine. Almost no ammonia was released in the absence of the primary amine. Increasing concentration of ethylenediamine also increased ammonia production. In contrast, CNF-induced ammonia release was hardly changed in the presence and absence of the primary amine. These data suggest that (at least under the conditions used) DNT acts preferentially as a transglutaminase, whereas CNF behaves preferentially as a deamidase. In fact, differences in the activities of CNF and DNT are obvious from studies in intact cells. Treatment of intact cells with CNF causes an upward shift of RhoA in SDS-PAGE indicating a deamidase reaction (15,21). By contrast, Horiguchi et al. (10) reported that DNT caused a downward shift of Rho after treatment of cells for 1-3 h. Longer incubation of cells with DNT (e.g. for up to 6 h) resulted in an occurrence of an additional upward shift. These data can be interpreted to indicate that DNT causes preferentially a transglutamination reaction also in intact cells. Because we did not succeed in the expression of a recombinant full-length DNT preparation, which was biologically active, we are at present not able to verify this hypothesis.
Because deamidation-or transglutamination-induced changes in migration of GTPases in SDS-PAGE are less pronounced with Rac and Cdc42, we used the ammonia release assay to study the substrate specificity of ⌬DNT. These data indicated that all Rho GTPases including Rac and Cdc42 are modified by ⌬DNT. We did not detect major differences in the ability of the various Rho proteins to serve as substrate for ⌬DNT. A similar substrate specificity was recently reported for CNF1 (17). However, we observed differences between ⌬CNF1 and ⌬DNT in respect to the nucleotide dependence of the deamidation/transglutamination reactions. Whereas ⌬CNF1 catalyzed the deamidation reaction with a similar velocity in the presence of GDP or GTP␥S-loaded RhoA, ⌬DNT accepted GDPloaded RhoA or GDP-loaded V14RhoA but did not modify GTP␥S-bound RhoA or V14RhoA that is not able to hydrolyze GTP. The slight modification of GTP␥S-loaded RhoA with ⌬DNT and the low modification of V14RhoA GDP may be because of an incomplete exchange of the nucleotides. As binding of nucleotides largely changes the conformation of the switch II region of the GTPases, these findings suggest that the structural requirements for modification by DNT are more restricted. Another possibility would be that free nucleotides interact with the enzyme to alter its activity. In fact, a nucle-otide binding motif has been described for DNT but not for CNF covering residues 1304 -1311 (AFYHTGKS) with the consensus (A/G)XXXXGK(S/T) (11,22). To study the relevance of this motif for DNT activity, we changed lysine 1310 to alanine. This mutation blocked ⌬DNT activity as reported earlier for the holotoxin (11). The role of lysine 1310 is not clear because this residue is not present with similar spacing in CNF1. It is conceivable that the loss in activity of the K1310A mutant is caused by structural changes of the toxin not directly involving catalysis, because K1310 is located in the vicinity of the catalytic important residue His 1307 . Although a putative nucleotide binding motif is present in DNT, we did not obtain evidence for a control of ⌬DNT activity by direct interaction of the enzyme with nucleotides.
All eukaryotic transglutaminases are characterized by a catalytic cysteine and histidine residue. Recently, we identified cysteine 866 in CNF1 as essential for deamidase and transglutaminase activity (17). Suggesting that a similar catalytic mechanism is functional in DNT, we changed cysteine 1292 of The reaction was stopped after 10 min by heating for 1 min at 95°C. Denatured proteins were removed by centrifugation, and the ammonia produced was measured in the supernatant. Shown is the ammonia produced at each ethylenediamine concentration as mean Ϯ S.D. of three independent experiments. B, GTPase activity. The reaction was stopped by freezing an aliquot of the proteins in liquid nitrogen. Toxin-treated RhoA was loaded with [␥-32 P]GTP. Thereafter, the GTPase activity was stimulated by adding p50 GAP . The hydrolysis of GTP was determined after 4 min by filter binding assay. Shown is the remaining bound radioactivity as percent of loaded radioactivity as mean ϩ S.D. of three independent experiments. ED, ethylenediamine.
DNT to serine or alanine. These mutations caused inhibition of the enzyme activity indicating an essential role in catalysis. Thus, as assumed from the amino acid sequence alignment of DNT and CNF1, cysteine 1292 of DNT is functionally equivalent to cysteine 866 of CNF1. In contrast to transglutaminases, like the blood clotting factor FXIII, the activity of ⌬DNT or ⌬CNF was not dependent on calcium ions. Another Ca 2ϩ -independent transglutaminase was recently cloned from Streptoverticillum (23).
The preferential transglutamination of Rho allowed studies on the GTPase activity of the cross-linked Rho protein. Gln 63 of RhoA is essential for the intrinsic and GAP-stimulated GTPase mechanism of the protein. Recent crystal structure analysis of Rho and Rho-GAP in a complex with a transition state analogue GDP-AlF 4 Ϫ explains the function of Gln 63 in stabilizing the transition state of GTP hydrolysis. To this end, the nitrogen of the carboxamide group of Gln 63 is bonded to the main chain carbonyl of Arg 85 of Rho-GAP and to one of the fluorides of AlF 4 Ϫ (20). If Gln 63 is deamidated (e.g. by CNF), this interaction with GAP is not possible resulting in the blockade of GAPstimulated GTP hydrolysis (14,15). After transglutamination of Gln 63 , however, the pivotal nitrogen residue is still present. Therefore, we were surprised that after transglutamination both intrinsic and GAP-stimulated GTPase activity of Rho were blocked. The reason for this inhibition is not entirely clear but may be based on structural changes that are the prerequi-site for catalysis of GTP hydrolysis. For example, binding of Rho-GAP and subsequent activation of Rho GTPase activity are accompanied by conformational changes to allow the introduction of the catalytic Arg 85 of RhoGAP into Rho (20). It is feasible that this interaction is hindered by transglutamination of Gln 63 . Further studies are underway to analyze the influence of smaller transglutaminase cosubstrates like methylamine on the GAP-stimulated and intrinsic GTPase activity of RhoA after modification with ⌬DNT.
In summary, we localized the enzyme domain of DNT to a C-terminal fragment covering amino acid residues 1136 -1451 with cysteine 1292, histidine 1307, and lysine 1310 as essential residues. This active fragment acts as a deamidase and/or transglutaminase to modify Gln 63 of Rho or Gln 61 of Rac and Cdc42, respectively, and to activate the GTPases. Kinetic analysis indicates that ⌬DNT acts preferentially as a transglutaminase. In contrast to ⌬CNF, which effectively modifies Rho proteins in the GDP-and GTP-bound form, GDP-bound Rho proteins are the preferred substrates of ⌬DNT. DNT and CNF1 consist of 1451 and 1014 amino acid residues, respectively. The ⌬DNT and ⌬CNF cover amino acid residues 1136 -1451 and 709 -1014, respectively. The regions of significant amino acid similarity are aligned. The sequence of ⌬DNT is 13% identical with ⌬CNF. The sequence identity is 45% in the aligned region of high similarity (dark bars). The catalytic essential amino acid residues are marked.