Toxin-induced activation of the G protein p21 Rho by deamidation of glutamine

Abstract

Pathogenic Escherichia coli are responsible for a variety of diseases, including diarrhoea, haemolytic uraemic syndrome, kidney infection, septicaemia, pneumonia and meningitis. Toxins called cytotoxic necrotizing factors (CNFs) are among the virulence factors produced by uropathogenic (CNF1)¹ or enteropathogenic (CNF2)² E. coli strains that cause diseases in humans and animals, respectively. CNFs induce an increase in the content of actin stress fibres and focal contacts in cultured cells³,⁴. Effects of CNFs on the actin cytoskeleton correlated with a decrease in the electrophoretic mobility of the GTP-binding protein Rho⁴,⁵ and indirect evidence indicates that CNF1 might constitutively activate Rho⁶. Here we show that CNF1 catalyses the deamidation of a glutamine residue at position 63 of Rho, turning it into glutamic acid, which inhibits both intrinsic GTP hydrolysis and that stimulated by its GTPase-activating protein (GAP). Thus, this deamidation of glutamine 63 by CNF1 leads to the constitutive activation of Rho, and induces the reorganization of actin stress fibres. To our knowledge, CNF1 is the first example of a bacterial toxin acting by deamidation of a specific target protein.

Main

Incubation of Vero cells with CNF1 induces actin stress fibre accumulation and cell spreading (Fig. 1a, b), with a concomitant decrease in the electrophoretic mobility of the Rho protein on SDS–PAGE (Fig. 1c). Incubation of recombinant RhoA with purified CNF1 in vitro gives the same mobility shift as that observed after in vivo treatment (Fig. 2). This result indicates that CNF1 directly modifies Rho without the need for cellular cofactors. Incubation with CNF1 did not modify the electrophoretic mobilities of Rac, Cdc42, Rab6 or Ras either in vivo or in vitro (data not shown). In vitro induction of the mobility shift by CNF1 was blocked by heat denaturation of either RhoA or CNF1 (Fig. 2). To analyse the modification induced by CNF1, we eluted RhoA treated with CNF1 in vitro, untreated RhoA, or a mixture of both from gels and digested the eluate with trypsin. Separation of the tryptic peptides on a DEAE C-18 HPLC column yielded three major peptides (A,B,C); peptide B in CNF1-treated RhoA eluted slightly later than peptide B from untreated RhoA (Fig. 3a, top and middle panels). In the sample containing both CNF1-treated and untreated RhoA, peptide B is duplicated (Fig. 3a, lower panel). Peptide B1 elutes as peptide B from the untreated RhoA sample, whereas peptide B2 elutes as peptide B from CNF1-treated RhoA (Fig. 3a). Amino-acid sequencing of peptide B from untreated RhoA yielded the sequence QVELALWDTAG Q EDYDR, corresponding to amino acids 52–68 of RhoA⁷, whereas in CNF1-treated RhoA the sequence was QVELALWDTAG E EDYDR, which differs only by having a glutamic acid residue at position 63 instead of glutamine. Accordingly, in the CNF1-treated and untreated RhoA mixture, the amino-acid sequence of the B1 peptide corresponded exactly to RhoA 52–68, whereas the B2 peptide had the single Q63E modification of CNF1-treated RhoA. Apart from Gln 63, RhoA contains two additional glutamine residues in positions 29 and 52 (ref. 7) which were not modified by treatment with CNF1. Thus, CNF1 acts as a deamidase that specifically modifies Gln 63 of Rho to glutamic acid.

Figure 1: Induction of actin stress fibres in Vero cells by CNF1 and modification of the electrophoretic mobility of the Rho GTP-binding protein by CNF1 in vivo.

a, b, Vero cells were treated with 10⁻¹⁰ M CNF1 for 18 h, fixed with 4% paraformaldehyde, permeabilized with 0.05% Triton X-100 and then stained for F-actin with FITC–phalloidin. a, Control cells; b, cells incubated with CNF1. c, Mobility-shift of Rho in control and CNF1-treated Vero cells.

Figure 2: CNF1-induced mobility-shifting of RhoA in vitro.

RhoA was incubated with CNF1 and processed for electrophoresis. As a control, RhoA or CNF1 was denaturated by treatment at 95 °C for 15 min, then incubated with either native CNF1 or RhoA.

Figure 3: Deamidation of Rho glutamine 63 by CNF1.

a, Peptide chromatograms of RhoA incubated with (T) or without (U) CNF1, and of a mixture of both (U + T). The amino-acid sequence of the RhoA tryptic peptide spanning amino acids 52–68 is indicated by arrows. Arrowhead indicates the direction of peptides elution. b, Time course of CNF1 deamidase effect on RhoA, as studied by formation of peptide 52–68 of RhoA with glutamic acid at position 63.

CNF1 must work enzymatically on Rho because the mobility shift induced by CNF1 is blocked by heat denaturation of Rho or of CNF1 (Fig. 2), and because the amount of Q63E RhoA formed depends on the incubation time of Rho with CNF1 (Fig. 3b). Rho deamidation in vitro is relatively slow, with only 60% of Q63 RhoA being converted after 1 h (Fig. 3b). It is known that cells need to be incubated with CNF1 for a relatively long time before actin stress fibres accumulate and cell spreading is seen³,⁴,⁶. The kinetics of deamidation by CNF1 of Rho in vitro were not significantly affected by including cytosolic factors or altering our reaction. Mutations at position 61 of Ras (corresponding to Rho position 63) also modify its apparent electrophoretic mobility⁸,⁹.

We next studied the effects of CNF1 on the rate of nucleotide dissociation, the rate of intrinsic GTP hydrolysis, and the rate of GTP hydrolysis stimulated by the Rho GTPase-activating protein (rhoGAP)¹⁰, and compared them with those of a Q63E RhoA mutant. The electrophoretic mobility of Q63E RhoA was identical to CNF1-modified RhoA (Fig. 4a). Dissociation of GTP-γS at 1mM Mg²⁺from both CNF1-treated RhoA and mutated Q63E RhoA was slightly increased compared to RhoA (Fig. 4b), with the GTP-γS ‘off’ rate being k = 0.04 min⁻¹ for both CNF1-modified RhoA and Q63E RhoA, and 0.023 min⁻¹for unmodified RhoA. rhoGAP-stimulated GTP hydrolysis was abolished on Q63E RhoA compared with unmodified RhoA (Fig. 4c). A small residual activity could be detected on CNF1-treated RhoA (Fig. 4c), probably due to the presence of trace amounts of unmodified RhoA. The intrinsic GTPase activity was also lost in the mutant (Fig. 4c). CNF1-modified RhoA and the Q63E mutant have comparable nucleotide affinities and both lack GTPase activity, indicating that they could be identical.

Figure 4: Electrophoretic mobility shift, time course of [³⁵S]GTP-γS dissociation, and determination of the intrinsic and rhoGAP-stimulated GTP hydrolysis of RhoA wild type, Q63E RhoA, and CNF1-treated RhoA.

a, Electrophoretic mobilities of: 1, RhoA; 2, CNF1-treated RhoA; 3, RhoA; 4, Q63E RhoA. b, Dissociation of [³⁵S]GTP-γS in 1 mM MgCl₂ from RhoA (white circles) Q63E RhoA (black circles) and CNF1-treated RhoA (triangles). c, Time course of intrinsic and rhoGAP-stimulated GTP hydrolysis. Reactions are described in Methods. Equal loading of the different GTP-binding proteins with [³²P]GTP was verified by filtration, as described for the dissociation assay. MgCl₂ was added 1 min after the onset of the reaction to initiate GTP hydrolysis (arrowhead). GST–rhoGAP was added 3 min after the onset of the reaction (arrow); final concentration, 10 nM. Line a, RhoA minus rhoGAP; line b, RhoA plus rhoGAP; line c, Q63E RhoA plus or minus rhoGAP; line d, CNF1-treated RhoA minus rhoGAP; line e, CNF1-treated RhoA plus rhoGAP. Results are given in c.p.m. of free ³²P_i. Times are shown in minutes after the start of the reaction.

To show that CNF1-treated RhoA and Q63E RhoA behave as activated forms of RhoA, we microinjected both into Vero cells. As wild-type RhoA can induce stress fibres at concentrations close to those required by the permanently active form of RhoA, V14 RhoA¹¹, we determined the minimum concentrations of RhoA, CNF1-modified RhoA and Q63E RhoA needed to induce formation of actin stress fibres as a result of microinjection into Vero cells. Microinjection of Vero cells with RhoA at 50 μg ml⁻¹induced maximal stress fibre formation (Fig. 5a), but at 10 μg ml⁻¹there was no effect (Fig. 5b). CNF1-modified RhoA (Fig. 5c, d) or Q63E RhoA (Fig. 5e, f) had maximal effect at 25 μg ml⁻¹and at 10 μg ml⁻¹these effects were still evident (Fig. 5d, f). Thus, CNF1-modified RhoA and Q63E RhoA are both equally effective at inducing the formation of stress fibres. Our results show that CNF1 activates Rho by deamidating Gln 63 to glutamic acid, thereby impairing the intrinsic and rhoGAP-stimulated GTPase. CNF1-modified Rho behaves as the other constitutively active forms of Rho that are mutated at residues 14 or 63.

Figure 5: Microinjection into Vero cells of CNF1-treated RhoA, Q63E RhoA and RhoA.

a, b, Cells microinjected with RhoA (50 μg ml⁻¹in a; 10 μg ml⁻¹in b). c, d, Cells microinjected with CNF1-treated RhoA (c, 25 μg ml⁻¹; d, 10 μg ml⁻¹). e, f, Cells microinjected with Q63E RhoA (e, 25 μg ml⁻¹; f, 10 μg ml⁻¹). Arrows, microinjected cells. Scale bar, 10 μm.

The Rho family proteins play a key role in actin remodelling in response to growth factors¹². In Swiss 3T3 cells Rho controls actin reorganization into stress fibres and the formation of focal contacts¹³, Rac regulates membrane ruffling¹⁴, and Cdc42 is responsible for the formation of filopodia¹⁵. We have previously shown that CNF1 induces a phagocytic behaviour through activation of membrane ruffling in HEp-2 cells¹⁶. We have now shown that in Vero cells CNF1 induces actin stress fibres but not ruffling (Fig. 1) and decreases the electrophoretic mobility of Rho but not of Rac or Cdc42 (although we cannot rule out the possibility that the PAGE system might not resolve the CNF1-modified forms of Rac and Cdc42). Membrane ruffling may be induced by Rho (in addition to Rac) only in some cell types such as HEp-2 or KB cells when they are stimulated by hepatocyte growth factor or phorbol esters¹⁷. We and others have shown that the invasive bacterium Shigella flexneri enters HeLa cells as a result of membrane ruffling triggered by Rho¹⁸,¹⁹. The dermonecrotic toxin from Bordetella bronchiseptica (DNT)²⁰, which shares amino-acid sequence homology with CNF1 (ref. 21), also promotes actin reorganization in cultured cells²². Furthermore, DNT induces a mobility shift of Rho²², suggesting that it also deamidates Gln 63 of Rho. Enzymatic deamidation is a previously undescribed activity of bacterial toxins acting inside host cells; others include ADP-ribosylation, depurination, adenylate cycling, metal-dependent proteolysis and glycosylation. We have now demonstrated that the deamidase activity of CNF1 is responsible for the cytopathic effects of this toxin, and it is likely that this modification is used by other microbial or eukaryotic cell factors to modify regulatory proteins in host cells.

Methods

Rho electrophoretic mobility-shift assay. Highly purified CNF¹⁶ was used. Recombinant RhoA was made from the pGEX 2T GST–F25N RhoA plasmid²³ (gift from A. Hall). The GST–RhoA fusion protein was cleaved by thrombin and incubated for 2 h at 37 °C (0.5 μg CNF1 for 2 μg RhoA) in 20 mM Tris-HCl buffer, pH 7.5, containing 1 mM MgCl₂, 10 mM dithiothreitol (DTT) and 100 μM GDP. Samples were boiled for 1 min, electrophoresed on 12% SDS–polyacrylamide gels and stained with Coomassie blue.

Trypsin digestion of gel-eluted Rho, peptide separation and amino-acid sequencing. RhoA was incubated for 2 h at 37 °C with CNF1 (0.5 μg CNF1 for 2 μg RhoA) in 20 mM Tris-HCl buffer, pH 7.5, containing 1 mM MgCl₂, 10 mM DTT and 100 μM GDP, and electrophoresed on a 12% SDS–polyacrylamide gel which was stained with amido-black. Bands corresponding to the Rho protein were cut out of the gel, eluted and digested with 1 μg trypsin in 200 μl 100 mM Tris-HCl buffer (pH 8.8) containing 0.01% Tween 20, for 18 h at 35 °C. Tryptic peptides were separated by HPLC on a DEAE-C18 column using an acetonitrile/trifluoracetic acid gradient. Peptides were sequenced using an Applied Biosystems microsequencer.

Preparation of the Q63E RhoA mutant. The Q63E mutation was introduced by polymerase chain reaction (PCR) and the plasmid checked by sequencing. Q63E RhoA was expressed in E. coli as a GST fusion protein which was then cleaved by thrombin.

GTP-γS dissociation assay. Q63 RhoA, CNF1-treated RhoA and Q63E RhoA (1 μM final) were loaded with 25 μM [³⁵S]GTP-γS in a low-magnesium buffer (50 mM Tris-HCl, pH 7.5, 100 mM KCl, 1 mM DTT, 1 mM MgCl₂, and 2 mM EDTA) at 37 °C for 1 min. The final concentration of free Mg²⁺was adjusted to 1 mM and 1 mM GTP-γS was added to initiate [³⁵S]GTP-γS dissociation. Samples of 25 μl were filtered at the indicated times and radioactivity counted²⁴.

GTPase assay. Q63 RhoA, CNF1-treated RhoA and Q63E RhoA (1 μM final) were loaded with 25 μM [γ-³²P]GTP as described for [³⁵S]GTP-γS. MgCl₂(1 mM free Mg²⁺) was added to start the intrinsic GTPase. For measurements of rhoGAP-stimulated GTP hydrolysis, 10 nM of a 30K C-terminal fragment of GST–rhoGAP (gift from A. Hall²⁵) was added 2 min after MgCl₂. At the indicated times, aliquots of 25 μl were removed and the ³²P_i release was measured by the charcoal method²⁶.

Cell microinjection, immunofluorescence and ADP-ribosylation of Rho. F25N RhoA²³ was used as it is threefold more potent, upon microinjection, in inducing stress fibres than F25 RhoA¹⁵. CNF1 (12 μg) was incubated with 50 μg recombinant GST–RhoA bound to glutathione beads (Sigma) for 1 h at 37 °C in 50 mM Tris-HCl buffer, pH 7.5, containing 5 mM MgCl₂, 50 mM NaCl, 10 mM DTT, 150 mM KCl (buffer A) supplemented with 100 μM GDP (final volume, 1 ml). The mixture was then packed into a mini-column which was washed with 30 ml buffer A. GST–RhoA was eluted from the glutathione beads with buffer A containing 20 mM reduced glutathione, in 100 μl (containing 0.5 mg ml⁻¹GST–RhoA) and tested for CNF1-induced mobility shift by SDS-PAGE. GST–RhoA (100 μl) was cleaved with 1 μl thrombin (1 U μl⁻¹) (Sigma) for 1 h at room temperature then dialysed against 50 mM Tris-HCl buffer, pH 7.5, containing 5 mM MgCl₂, 150 mM KCl, 0.1 mM DTT (microinjection buffer (MB) to remove glutathione. GST was then separated from Rho by binding to glutathione beads and incubating with 10 μl p-aminobenzamidine beads (Sigma) to remove thrombin. Before microinjection, the concentration of active GTP-binding proteins present in the RhoA, RhoA Q63E and CNF1-modified RhoA preparations was determined by binding to [³⁵S]GTP-γS as described²⁴. Preparations of the different GTPases were found to be 60% active. RhoA or Q63E RhoA was diluted with protein A-purified rabbit non-immune antibodies (0.5 mg ml⁻¹) (to determine the efficiency of microinjection) in MB (final volume 20 μl) and microinjected (Transjector 5246 system, Eppendorf) into serum-starved (0.2% FCS, overnight) Vero cells. Microinjections were done in DMEM, which contained 25 mM HEPES buffer to maintain the pH at 7.3. Cells were then incubated at 37 °C for 20 min, fixed with 4% paraformaldehyde, processed for F-actin immunofluorescence with FITC–phalloidin together with rhodamine-labelled anti-rabbit antibodies (Amersham) and examined and photographed with a fluorescence-equipped photomicroscope.

Semi-confluent Vero cells (2 × 10⁶) were incubated or not with 10⁻¹⁰ M CNF1 for 18 h then lysed, and lysates were radioactively ADP-ribosylated with C. botulinum exoenzyme C₃ as described²⁷.