Neosynthesis and Activation of Rho by Escherichia coli Cytotoxic Necrotizing Factor (CNF1) Reverse Cytopathic Effects of ADP-ribosylated Rho*

Clostridium botulinum exoenzyme C3 inactivates the small GTPase Rho by ADP-ribosylation. We used a C3 fusion toxin (C2IN-C3) with high cell accessibility to study the kinetics of Rho inactivation by ADP-ribosylation. In primary cultures of rat astroglial cells and Chinese hamster ovary cells, C2IN-C3 induced the complete ADP-ribosylation of RhoA and concomitantly the disassembly of stress fibers within 3 h. Removal of C2IN-C3 from the medium caused the recovery of stress fibers and normal cell morphology within 4 h. The regeneration was preceded by the appearance of non-ADP-ribosylated RhoA. Recovery of cell morphology was blocked by the proteasome inhibitor lactacystin and by the translation inhibitors cycloheximide and puromycin, indicating that intracellular degradation of the C3 fusion toxin and the neosynthesis of Rho were required for reversal of cell morphology. Escherichia colicytotoxic necrotizing factor CNF1, which activates Rho by deamidation of Gln63, caused reconstitution of stress fibers and cell morphology in C2IN-C3-treated cells within 30–60 min. The effect of CNF1 was independent of RhoA neosynthesis and occurred in the presence of completely ADP-ribosylated RhoA. The data show three novel findings; 1) the cytopathic effects of ADP-ribosylation of Rho are rapidly reversed by neosynthesis of Rho, 2) CNF1-induced deamidation activates ADP-ribosylated Rho, and 3) inhibition of Rho activation but not inhibition of Rho-effector interaction is a major mechanism underlying inhibition of cellular functions of Rho by ADP-ribosylation.

Cell Culture-CHO-K1 cells were cultivated in tissue culture flasks at 37°C and 5% CO 2 in Ham's F-12/Dulbecco's modified Eagle's medium (1:1) containing 5% heat-inactivated (30 min, 56°C) fetal calf serum, 2 mM L-glutamate, 100 units/ml penicillin, and 100 g/ml streptomycin. Cells were routinely trypsinized and reseeded twice a week. Primary cultures of rat astroglial cells were set up as described previously (32,33). For cytotoxicity and regeneration assays, subconfluent cells (about 10 5 to 5 ϫ 10 5 cells/cm 2 ) in 3-cm plates containing coverslips or cells growing on Lab-Tec chamber slides were treated for different times with 200 ng/ml activated C2II together with 100 ng/ml C2IN-C3. CNF1 was used in a final concentration of 300 ng/ml medium. For regeneration assays cells were treated with the respective toxin for 3 h; then the cells were washed 5 times with prewarmed medium, new medium without any toxin was added, and the cells were incubated at 37°C.
Actin Staining-For fluorescence staining of F-actin, cells were fixed in 4% paraformaldehyde containing 0.1% Triton X-100 for 30 min. Thereafter, cells were briefly washed and incubated for 30 min with rhodamine-phalloidin (600 ng/ml).
ADP-ribosylation Assay of Rho-Cells were treated with the respective toxins as indicated. After incubation they were washed with cold PBS, harvested in 400 l of cold lysis buffer (2 mM MgCl 2 , 0.1 mM phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, 80 g/ml benzamidine in 50 mM HEPES, pH 7.4) and sonicated. Aliquots (100 g of protein) of the lysate were incubated with [ 32 P]NAD and 50 ng of C2IN-C3 fusion toxin at 37°C for 30 min, Laemmli buffer was added, the samples were heated for 3 min at 95°C and run on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). [ 32 P]ADP-ribosylated proteins were detected by autoradiography with a PhosphorImager from Molecular Dynamics (Krefeld, Germany).

SDS-PAGE and Western
Blotting -SDS-PAGE was performed according to Laemmli (34). For immunoblot analysis, the proteins were transferred onto a nitrocellulose membrane. The membrane was blocked for 30 min with 5% nonfat dry milk in PBS containing 0.05% Tween 20 (PBS-T) and the proteins were probed with anti-RhoA antibody (mouse, 1:2000 in PBS-T). After washing with PBS-T, the blot was incubated for 1 h with an anti-mouse antibody coupled to horseradish peroxidase (1:2000 in PBS-T). The membrane was washed again, and proteins were visualized using the ECL system according to the manufacturer's instructions.
Extraction and Filter Hybridization of RNA-Total RNA was extracted by the guanidine hydrochloride method (35). Seven-g aliquots were separated by denaturing gel electrophoresis (1.25% formaldehyde agarose gels; running buffer was 20 mM MOPS, 8 mM NaAc, and 1 mM EDTA) and electroblotted onto a nylon membrane. The following nonradioactive DIG filter hybridization and detection of RhoA mRNA was performed as described previously (36,37).
Fluorescence Assay for Nucleotide Exchange-All measurements were done on a LS50B spectrofluorometer from Perkin Elmer at 37°C. First RhoA was loaded with the fluorescent nucleotide mant-GDP by incubating 0.5 M GTPase with 2 M mant-GDP in buffer C (10 mM triethanolamine, pH 7.5, 150 mM NaCl, 2 mM MgCl 2 ) for 1 h. Thereafter, the exchange of Rho-bound mant-GDP for GTP was initiated by adding 100 M GTP and 0 -20 M LBC. The exchange of Rho-bound fluorescent mant-GDP for GTP was monitored by the decrease in fluorescence intensity upon the release of mant-GDP from RhoA. Exciting of the mant-fluorophor was at 357 nm; emission was measured at 444 nm. The rate constant for the release of mant-GDP was determined by fitting the experimental curve with a single exponential decay using SigmaPlot (Jandel Scientific).

C2IN-C3 Fusion Toxin Rapidly Disassembles Stress Fibers in
Rat Astroglial Cells-In primary cultures of rat astrocytes, C2IN-C3 fusion toxin caused changes in cell shape and destruction of stress fibers. These effects started 40 min after addition 1 The abbreviations used are: GST, glutathione S-transferase; mant-GDP, 2Ј-(3Ј)-O-(N-methylanthraniloyl)guanosine 5Ј-diphosphate; CHO, Chinese hamster ovary; PBS, phosphate-buffered saline; PBS-T, PBS plus Tween 20; PAGE, polyacrylamide gel electrophoresis; LBC, lymphoid blast crisis oncogene; MOPS, 4-morpholinepropanesulfonic acid. and 120 min (lane 6), respectively. After toxin treatment, the cells were lysed and lysate proteins (100 g) were subjected to an in vitro [ 32 P]ADP-ribosylation assay with C3 as described under "Experimental Procedures." Labeled proteins were detected by SDS-PAGE and autoradiography. C, anti-RhoA immunoblot analysis of 100 g of the same lysates described above. ADP-ribosylated RhoA is indicated by an asterisk.
of the toxin and were maximal after about 2 h (Fig. 1A). By 40 -60 min after addition of C3 fusion toxin to intact cells, radioactive [ 32 P]ADP-ribosylation of Rho was no longer observed in the cell lysate, indicating the complete modification of cellular Rho (Fig. 1B). The ADP-ribosylation of RhoA was confirmed by anti-RhoA immunoblot analysis (Fig. 1C). Twenty minutes after addition of the toxin, a fast moving (unmodified Rho) and a slower moving RhoA band (ADP-ribosylated Rho) were detected by immunoblotting. After 40 min, only the ADPribosylated RhoA was observed.
Effects of the C2IN-C3 Fusion Toxin Are Reversible-To test whether the effects of the fusion toxin were reversible, astroglial cells were incubated with the toxin for 3 h. Thereafter, the cells were extensively washed to remove extracellular toxin. Stress fibers reappeared 6 h after removal of the toxin, and the recovery was complete after 24 h ( Fig. 2A). Reorganization of stress fibers was accompanied by the reappearance of the polygonal cell morphology ( Fig. 2A). Removal of the toxin rapidly changed the proportion of ADP-ribosylated and non-modified Rho (Fig. 2B). Four hours after removal of the toxin, i.e. when the cells started to rebuild their stress fibers, a small amount of Rho was [ 32 P]ADP-ribosylated. Its level increased with the duration of toxin withdrawal (Fig. 2B). Accordingly, 4 h after removal of the toxin, non-ADP-ribosylated RhoA was detected by anti-RhoA immunoblotting and 24 h after toxin withdrawal only non-ADP-ribosylated RhoA was detected (Fig. 2C).
Reversal of the C3 effects was not cell type-specific. In CHO cells or HeLa cells, the C3 fusion toxin disassembled stress fibers within 3 h. Removal of the toxin resulted in the reorganization of stress fibers, adhesion to dish matrix, and restoration of cell morphology. The recovery started 2-3 h after addition of fresh medium (data not shown). Reversal of the cytopathic effects seemed to be specific for the fusion toxin. When CHO cells were treated with 10 g/ml wild-type C. limosum C3 exoenzyme for 24 h to induce the "C3 specific" morphology, cells did not recover after removal of the toxin by medium exchange.
Studies on the Duration of Action of the C2IN-C3 Fusion Toxin-Because the cytopathic effects disappeared rapidly after removal of C2IN-C3 from the medium, we studied whether the toxin was intracellularly degraded by proteases. When CHO cells were treated for 3 h with the fusion toxin and subsequently incubated for 16 h in fresh medium in the presence of the proteosome inhibitor lactacystin, cells did not regain their original morphology. In lysates of these cells, Rho was not radioactively labeled by ADP-ribosylation with [ 32 P]NAD and C3 (Fig. 3A, lane 5). As expected, only a slowly migrating RhoA band was detected by immunoblotting, indicating the presence of Rho ADP-ribosylated in intact cells (Fig. 3B, lane 5). Thus, lactacystin prolonged the cellular effects of the C3 fusion toxin.
Studies on the Recovery of RhoA-First, we tested whether recovery of non-ADP-ribosylated Rho was due to cleavage of the ADP-ribose moiety from Rho by a cellular hydrolase activity. CHO cells were treated with the C3 fusion toxin for 2 h and then incubated for 1 h in fresh medium without toxin. Cell lysates were prepared and incubated for 0, 3, and 6 h at 37°C. Thereafter, Rho protein was analyzed by immunoblotting (Fig.  4). Even after incubation of the lysates for 6 h at 37°C (Fig. 4, lanes 3 and 6), a significant proteolysis of RhoA was not detected. Furthermore, in lysates of cells that had been treated with C3 fusion toxin (lanes 4 -6), the slowly migrating band of ADP-ribosylated RhoA remained stable, while a rapidly migrating RhoA band did not appear. Taking these results together, we were not able to detect hydrolysis of the ADPribosylated RhoA in the lysates.
To test whether the reappearance of non-ADP-ribosylated RhoA after toxin withdrawal was due to newly synthesized protein, we blocked protein synthesis with cycloheximide or puromycin. Three hours after addition of the toxin, endogenous Rho was completely ADP-ribosylated (no [ 32 P]ADP-ribosylation, Fig. 5B, lane 2) and only the slowly moving ADP-ribosylated RhoA band was detected in the immunoblot (Fig. 5C, lane  2). When cycloheximide or puromycin was present after toxin withdrawal, regeneration of stress fibers (Fig. 5A) and [ 32 P]ADP-ribosylation of Rho was not observed (Fig. 5B, lanes  5 and 7). Moreover, the immunoblot did not show unmodified RhoA (Fig. 5C, lanes 5 and 7). All these data suggested that protein neosynthesis, not hydrolysis of ADP-ribosylated RhoA, was responsible for the reappearance of non-ADP-ribosylated RhoA.
Northern blot analysis of total RNA extracts from CHO cells was used to investigate whether the rapid neosynthesis was initiated by an increase in RhoA mRNA. Neither addition of the C3 toxin (added for 3 h) nor its removal had any effect on the amount of RhoA mRNA (Fig. 6). Apparently, neosynthesis of RhoA protein is due to ongoing protein synthesis.
C3-induced Destruction of the Actin Cytoskeleton Is Prevented and Rapidly Reversed by CNF1-Because a small amount of newly synthesized RhoA protein appeared to induce the complete reassembly of actin filaments and reconstitution of the cellular morphology, we tested whether this process could be enhanced by incubation of C3-treated cells with the E. coli cytotoxic necrotizing factor, CNF1, which activates Rho by deamidation (16,18). Therefore, actin stress fibers of CHO cells were destroyed by treatment (3 h) with the C3 fusion toxin (Fig.  7). The cells were subsequently incubated for 2 h with CNF1 in the presence of the C3 fusion toxin. Surprisingly, the cells started to form actin stress fibers as early as 60 min after CNF1 addition. The CNF1-induced rearrangement of the actin cytoskeleton was accompanied by a total reconstitution of "normal" cell morphology within about 2 h. CNF1 completely prevented the C3-induced breakdown of actin filaments when CNF was added 3 h prior to C2IN-C3 fusion toxin (data not shown). When both toxins were added together at the same time, a weak C3-induced disassembly of actin filaments was observed after 60 min. This effect was completely reversed by CNF1 within another 60 min. The same rapid recovery of actin stress fibers was observed when NIH 3T3 fibroblasts were exposed to the various toxin combinations (data not shown).
C3-induced ADP-ribosylation of Rho in the Presence of CNF1-Next we analyzed the extent of ADP-ribosylation of Rho after CNF1 treatment. As shown in Fig. 8, in C3-treated cells RhoA remained completely ADP-ribosylated even after treatment of cells with CNF1 for 3 h (lane 4) (note that at this time point all cells exhibited complete recovery of the actin cytoskeleton). Additionally, when C2IN-C3 and CNF1 were added together, RhoA was ADP-ribosylated (lane 5). After 3 h, the cells showed no disassembly of stress fibers or C3-typical cytopathic effects on cell morphology. Finally, preincubation of cells with CNF1 for 3 h did not prevent C3-induced ADPribosylation. Similar results were obtained with NIH 3T3 cells (data not shown).
The effects of CNF1 on the C3-induced alteration of cellular morphology occurred very rapidly, suggesting that the recovery of cell morphology was not caused by neosynthesis of Rho. However, to exclude that CNF1 accelerated the expression of Rho and to rule out that a very small amount of neosynthesized RhoA was sufficient for rebuilding of the actin cytoskeleton, we tested the toxin in combination with inhibitors of protein synthesis. Therefore, CHO cells were treated for 16 h with the C3 fusion toxin in the presence of cycloheximide. Subsequently, the cells were incubated with CNF1 for another 2 h in the presence of cycloheximide. Thereafter, cells were lysed and Rho was [ 32 P]ADP-ribosylated by C3. Fig. 9A shows that no radioactive labeling of Rho was detected in lysates of cells pretreated with C3 or C3 plus CNF1. In contrast, pretreatment of cells with CNF1 did not inhibit the radiolabeling of Rho. These findings indicated that Rho from cells treated with fusion toxin and additionally with CNF1 was completely ADP-ribosylated. The anti-RhoA immunoblot (Fig. 9B) confirmed the ADP-ribosylation of RhoA and showed an additional shift of Rho due to CNF1-induced modification. Taken together, these results demonstrated that, despite the ADP-ribosylation of cellular Rho, CNF1 treatment causes complete and rapid reorganization of the actin cytoskeleton, even under conditions excluding the de novo synthesis of Rho.
Effect of ADP-ribosylation on Nucleotide Exchange of RhoA Activated by the Exchange Factor LBC-The results described above (e.g. activation of ADP-ribosylated Rho by CNF1) suggested that inhibition of Rho function by ADP-ribosylation is not caused by a blockade of the interaction of the GTPase with its effectors but rather by inhibition of its activation. To test this hypothesis, we studied the nucleotide release of ADPribosylated Rho previously loaded with mant-GDP in the presence of the Rho-specific nucleotide exchange factor LBC (30). As shown in Fig. 10A, the release of mant-GDP of control RhoA was stimulated by GST-LBC by about 13-fold, whereas LBC stimulated the release of the nucleotide from ADP-ribosylated Rho only by about 1.5-fold. This dramatic reduction in the activation of GDP release was observed over a broad concentration range of LBC (Fig. 10B). DISCUSSION To study kinetics and functional consequences of Rho ADPribosylation in intact cells, we used the C2IN-C3 fusion toxin that uses the cell delivery system of the binary C. botulinum C2 toxin to enter target cells (25). C2IN-C3 caused rapid disassembly of stress fibers and major changes in morphology in cultured rat astroglial cells and in CHO cells. The effects started within 40 min and were maximal about 2 h after toxin addition. As reported for the effects of the C3 exoenzyme (38,39), these changes in morphology were preceded by a complete ADP-ribosylation of Rho. Surprisingly, the cells recovered within a few hours after removal of the toxin from the medium. Apparently, this recovery of the normal cell morphology depended on the occurrence of non-ADP-ribosylated RhoA. Concomitantly with the start of the rebuilding of stress fibers, a small amount of non-ADP-ribosylated RhoA was detected. Twenty-four hours after toxin removal, when reconstitution of Then GST-CNF1 (300 ng/ml) was added and the cells were further incubated. After 30 min (d), 60 min (e), 90 min (f), and 120 min (g), the cells were fixed and stained with rhodamine-phalloidin. For controls, cells without toxin (a) and cells treated for 2 h with GST-CNF1 (c) were used. Cells were fixed, and actin filaments of respective cells were stained with rhodamine phalloidin. Bar ϭ 50 m. the actin cytoskeleton and regeneration of cellular morphology were complete, only non-ADP-ribosylated Rho was observed. Subsequently, we tested whether reappearance of non-ADPribosylated Rho result from cleavage of the ADP-ribose moiety of modified Rho or from de novo synthesis of the GTPase. Cleavage of arginine-linked ADP-ribose is catalyzed by specific ADP-ribosylarginine hydrolases (40), but ADP-ribosylasparagine hydrolase has not been identified so far. Moreover, we were not able to obtain any evidence for a cleavage of the ADP-ribose moiety attached to Rho by hydrolase activity. In contrast, the translation inhibitors cycloheximide and puromycin blocked de novo synthesis of RhoA and prevented the recovery of cell morphology. This finding strongly suggested that Rho neosynthesis is crucial for reversal of cell morphology. Since Northern blot analysis did not show an increase in RhoA mRNA after treatment with C2IN-C3 or after its removal, the rapid de novo synthesis of RhoA was not due to enhanced transcription. Therefore, we conclude that the constitutive biosynthesis of RhoA is sufficient to restore RhoA functions. Morphological changes were reversed when neosynthesis had provided at approximately 10% of the normal cellular content of Rho but when ADP-ribosylated Rho was still present. Thus, our findings do not support a dominant negative role of ADPribosylated Rho as suggested from microinjection studies (39). This discrepancy may be explained by the high protein concentration usually used in such microinjection studies.
In addition to rapid neosynthesis of RhoA, the recovery of normal cell morphology depended on the intracellular degradation of the toxin. Since addition of the proteasome inhibitor lactacystin prevented the recovery of the normal cell morphology, the C2IN-C3 fusion toxin was probably degraded by the proteosomal pathway (41). Similarly, toxin degradation by the proteosomal pathway has been recently described for a fusion protein of anthrax toxin, which is related to C2 toxin (42). In the presence of lactacystin only ADP-ribosylated Rho was detected, indicating that the ADP-ribosylating activity persisted in the cytosol and immediately modified the neosynthesized Rho. The data suggested that a continuous delivery of the toxin into the cytosol occurs in the presence of extracellular C2IN-C3. Consequently, the ADP-ribosylation of RhoA persists as long as the toxin is supplied. In contrast, cells treated with the C. limosum C3 exoenzyme did not recover within 24 h after removal of the toxin (data not shown). This difference in cell recovery may be explained by the different structures of the toxins and/or by the different modes and extent of cellular uptake.
In addition, E. coli cytotoxic necrotizing factor CNF1 reversed the morphological changes induced by C2IN-C3. CNF1 catalyzes the deamidation of glutamine at position 63 of Rho (16,18). This modification blocks the GTPase activity of Rho and renders Rho constitutively active. The recovery of the flat cell morphology after removal of C2IN-C3 occurred much more rapidly in the presence than in the absence of CNF1. Surprisingly, recovery of the cell morphology and formation of stress fibers were induced by CNF1 even in the presence of C2IN-C3. Moreover analysis of Rho from cells treated in parallel with both toxins (e.g. CNF1 and C2IN-C3) revealed complete ADPribosylation of the GTPase. Thus, CNF1 was able to restore stress fiber formation despite the complete ADP-ribosylation of Rho. Recently, it has been described in Hep-2 cells that CNF1 prevented the cytopathic effects of C3B, a C3 fusion toxin consisting of C3 and the binding/translocation domain of diphtheria toxin (43). In this report it was suggested that CNF1 prevents ADP-ribosylation of Rho by C3 (43). In contrast, our data indicate that ADP-ribosylation and deamidation of Rho by CNF1 and C3, respectively, occur in parallel, but the functional consequences induced by CNF1 appear to be dominant. In addition to Rho, CNF1 was shown to deamidate and to activate Rac and Cdc42 (17). Also these GTPases are involved in organization of the actin cytoskeleton. However, there is no evidence that Rac and Cdc42 can substitute for Rho as regards the formation of stress fibers (2,3). Therefore, our findings suggest that CNF1 induced the formation of an activated form of ADPribosylated Rho, which still interacted with cellular targets and was capable of activating effectors necessary for the formation of stress fibers and focal adhesions.
Moreover, this observation provides new insights into the molecular mechanism underlying the inactivation of Rho by ADP-ribosylation. It was shown that ADP-ribosylation has no major effects on nucleotide binding as well as hydrolysis of bound GTP. GAP-stimulated GTPase activity is not affected either (39,44). Because the ADP-ribosylation at Asn 41 occurs near the effector region of Rho, it has been suggested that the ADP-ribosylation blocks the interaction of Rho with its effectors (22). However, this hypothesis was not supported by findings that ADP-ribosylated Rho interacts with effectors like protein kinase N (44) or phosphatidyl-4-phosphate 5-kinase (45). Based on the observation that ADP-ribosylated Rho binds with even higher affinity to phosphatidyl-4-phosphate 5-kinase than the native RhoA, it was speculated that the effector activation step but not the binding is blocked by ADP-ribosylation (45). Our findings suggested that rather the activation step of Rho is crucially altered by ADP-ribosylation. This hypothesis was supported by the in vitro studies with recombinant ADPribosylated RhoA, showing that the stimulation of the GDP release from Rho by the guanine nucleotide exchange factor LBC was inhibited after ADP-ribosylation. Because the inactivation mechanism of ADP-ribosylated Rho by GTP hydrolysis is fully functional, a blocked or diminished activation of Rho will shift the GTPase to its inactive state (Fig. 11). Thus, after inhibition of the GTPase activity of Rho by CNF1, ADP-ribosylated Rho becomes active because the equilibrium between the inactive and active state is again shifted to the active state (Fig. 11).