AMPylation of Rho GTPases by Vibrio VopS Disrupts Effector Binding and Downstream Signaling

See allHide authors and affiliations

Science  09 Jan 2009:
Vol. 323, Issue 5911, pp. 269-272
DOI: 10.1126/science.1166382


The Vibrio parahaemolyticus type III effector VopS is implicated in cell rounding and the collapse of the actin cytoskeleton by inhibiting Rho guanosine triphosphatases (GTPases). We found that VopS could act to covalently modify a conserved threonine residue on Rho, Rac, and Cdc42 with adenosine 5′-monophosphate (AMP). The resulting AMPylation prevented the interaction of Rho GTPases with downstream effectors, thereby inhibiting actin assembly in the infected cell. Eukaryotic proteins were also directly modified with AMP, potentially expanding the repertoire of posttranslational modifications for molecular signaling.

Vibrio parahaemolyticus is a Gram-negative halophilic bacterium that causes foodborne illness worldwide, and infections with it lead to acute gastroenteritis (1). Sequencing of a pathogenic strain of V. parahaemolyticus revealed the presence of a previously well-characterized virulence factor called the thermostable direct hemolysin (TDH) and two type III secretion systems (T3SSs) (2, 3). These bacterial T3SSs deliver proteins, called effectors, into the cytosol of host cells during infection (4). The first system on chromosome one, T3SS1, uses a multifaceted mechanism to cause the cytotoxicity of host cells, involving the induction of autophagy, cell rounding, and cell lysis (5). These events occur in parallel, using multiple effectors to induce cytotoxicity.

The T3SS1 effector VopS (VP1686) was implicated in cell rounding by its inactivation of Rho family guanosine triphosphatases (GTPases), including Rac, Rho, and Cdc42 (6). After 1 hour of infection of HeLa cells with the V. parahaemolyticus POR3 strain, which contains a functional T3SS1 and inactivating mutations for both TDHs and T3SS2 (7), we observed the presence of active GTP-bound Cdc42, as measured by the ability of Cdc42 to interact with a glutathione S-transferase (GST) fusion of its downstream effector, the p21-activated kinase 3 Cdc42-protein binding domain (GST-PAK PBD) (Fig. 1A). Beyond 1 hour after infection, the presence of VopS in the POR3 strain was consistent with the inactivation of Cdc42, because the population of active GTP-bound Cdc42 was no longer observed (Fig. 1A). In cells infected with a VopS deletion strain (POR3ΔvopS), levels of Cdc42 in the active GTP-bound state persisted, but all Cdc42 was eventually inactivated by 4 hours, suggesting the presence of another factor contributing to the destabilization of Rho family GTPases (Fig. 1A). Reconstitution of the POR3ΔvopS strain with wild-type VopS caused a loss of active Cdc42 1 hour after infection, similar to that seen in the POR3 strain (Fig. 1A). During infection with all three strains, the total level of Cdc42 was reduced at the later time points of infection. However, cell rounding was delayed in the POR3ΔvopS strain occurring at 2 hours after infection (Fig. 1A and fig. S1A).

Fig. 1.

The expression of VopS leads to inactivation of Cdc42 and cell rounding. (A) HeLa cells uninfected (U) or infected with POR3, POR3ΔvopSvopS), or POR3ΔvopS+VopS (ΔvopS+VopS) for 4 hours were tested for active GTP-bound Cdc42 by means of GST-PAK PBD pulldown assays followed by immunoblot analysis with antibodies to Cdc42 and ß-actin. (B) HeLa cells transfected with pSFFV-eGFP and pcDNA3 empty vector, pcDNA3-VopS, pcDNA3-VopS-H348A, pcDNA3-VopSΔ30, or pcDNA3-VopSΔ30-H348A and visualized with confocal microscopy for green fluorescent protein (green) to identify transfected cells and Hoechst stain to identify nuclei (blue). Scale bar, 10 μm. (C) Multiple sequence alignment of representative Fic sequences labeled by species and colored by superkingdom: bacteria (black), eukaryote (blue), and archaea (red). Residue positions are highlighted by conservation: hydrophobic (yellow), small (gray), and polar (black). Residue numbers are depicted to the left of the alignment, and omitted residues are in parentheses. Observed secondary structure (SS) elements consistent in all Fic domain structures are indicated below the alignment (cylinder for helix and arrow for strand) and are colored from N to C terminus in rainbow succession. An asterisk above the alignment marks the VopS H348 mutation. Alignments refer to the following proteins in the order listed: VopS, GeneInfo Identifier 88192876 (gi|88192876), gi|151567990, gi| 42794620, gi|17544594, gi|24582217, and gi|21228708.

The expression of full-length VopS induced a severely rounded phenotype in transfected HeLa cells, probably due to a VopS-dependent disruption of Rho GTPase signaling (Fig. 1B). VopS includes a C-terminal domain of unknown function called Fic (filamentation induced by cAMP). Fic domains are found in a variety of species and contain an invariant histidine (H) within a conserved motif [HPFX(D/E)GNGR] (8, 9) (Fig. 1C). Mutation of the Fic domain at this conserved histidine (VopS-H348A) abrogated the VopS-mediated cell rounding (Fig. 1B). Because VopS is an effector secreted by T3SS1 during infection, we determined whether the secretion signal sequence was required for cell-rounding activity. We observed that tranfection of VopSΔ30, a mutant in which the putative signal sequence was deleted, caused rounding of HeLa cells similar to that produced by wild-type VopS (Fig. 1B). As expected, VopSΔ30-H348A did not cause any obvious changes in the actin cytosketon of transfected cells (Fig. 1B). Thus, VopS cell-rounding activity is independent of its signal sequence but requires a wild-type Fic domain.

To elucidate the biochemical mechanism used by VopS to inhibit Rho family GTPases, we produced soluble recombinant GST-VopSΔ30 and GST-VopSΔ30-H348A that lacks the largely hydrophobic 30–amino acid signal sequence. Although 35S-radiolabeled Rac (35S-Rac) did not interact with GST-VopSΔ30, it did interact with GST-VopSΔ30-H348A (Fig. 2A). In addition, GTP loading of the 35S-Rac increased the amount of 35S-Rac that interacted with recombinant VopSΔ30-H348A (Fig. 2A), reminiscent of the conformation-dependent interaction of GTP-bound Rac with its downstream effector PAK (10). When we loaded 35S-Rac with GTP, more 35S-Rac interacted with GST-PAK PBD (Fig. 2A). Many type III effectors use a catalytic mechanism to alter eukaryotic signaling pathways, and in some cases the catalytically inactive form of the effector can act as a substrate trap (11). The VopSΔ30-H348A mutant might act as a trap by binding its substrate, the active GTP-bound Rac, but failing to release a product.

Fig. 2.

VopS inhibits in vitro binding of the Rho GTPase Rac to its downstream effector PAK. (A) GST pulldown of in vitro translated 35S-Rac and GTP-γ-S–loaded 35S-Rac with GST (G), GST-PAK PBD (P), GST-VopSΔ30 (S), or GST-VopSΔ30-H348A (H/A). (B) 35S-DA-Rac was pre-incubated with VopSΔ30 or VopSΔ30-H348A, followed by a GST pulldown with GST-PAK PBD. (C) 35S-DA-Rac was pre-incubated with decreasing concentrations of VopSΔ30 or VopSΔ30-H348A (125 to 0.125 pmol) or left untreated (U), followed by a GST pulldown with GST-PAK PBD or GST alone (G). (D) 35S-DA-Rac was pre-incubated over time with 0.25 pmol of VopSΔ30 or VopSΔ30-H348A, followed by a pulldown assay with GST-PAK PBD. Samples were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed by autoradiography. Inputs (I) correspond to 5% of material in pulldown.

To further analyze the possibility that GTP-bound Rho family GTPases might be substrates for VopS, we produced the 35S-radiolabeled constitutively active form of Rac (35S-DA-Rac). We pre-incubated 35S-DA-Rac with 25 pmol of purified recombinant VopSΔ30 or VopSΔ30-H348A and then tested whether the 35S-DA-Rac was able to interact with its downstream effector GST-PAK PBD. Although 35S-DA-Rac pre-incubated with the mutant VopSΔ30-H348A interacted with GST-PAK PBD, 35S-Rac pre-incubated with VopSΔ30 was unable to interact with GST-PAK PBD (Fig. 2B). We then pre-incubated 35S-DA-Rac with serial 10-fold dilutions of purified recombinant VopSΔ30 (125 to 0.125 pmol) for 15 min and tested whether the 35S-DA-Rac could bind to GST-PAK PBD (Fig. 2C). Incubation of 35S-DA-Rac with as little as 1.25 pmol of VopSΔ30 prevented the binding of 35S-DA-Rac to GST-PAK PBD (Fig. 2C). As expected, the incubation of DA-Rac with decreasing amounts of VopSΔ30-H348A had no effect on the ability of DA-Rac to bind to GST-PAK PBD (Fig. 2C). Next, we incubated 35S-DA-Rac with a limiting amount of VopSΔ30 (0.25 pmol) over 1 hour. By 40 min, no 35S-DA-Rac interacted with GST-PAK PBD (Fig. 2D). Thus, VopS uses an enzymatic activity to inhibit the interaction of Rho family GTPases with their respective downstream effectors.

To determine the activity that VopS exerts on the Rho family GTPases, we expressed histidine-tagged DA-Rac either alone (DA-Rac) or with active GST-tagged VopSΔ30 (DA-Rac/VopS). Both forms of recombinant DA-Rac could be loaded with 35S-radiolabeled GTP-γ-S, a nonhydrolyzable form of GTP, eliminating the possibility that VopS inhibited the Rho family of proteins by preventing GTP binding (fig. S2A). To determine whether DA-Rac was altered by coexpression with VopS, we measured the mass of recombinant DA-Rac by mass spectrometry. Although DA-Rac had the expected molecular weight, DA-Rac/VopS had an increase in molecular weight of 329 daltons (fig. S2, B and C). As expected, DA-Rac coexpressed with the mutant VopS did not show an increase in molecular weight. The increase of 329 daltons is consistent with the mass of adenosine 5′-monophosphate (AMP). To identify where this putative covalent modification occurred on Rac, we analyzed tryptic and AspN peptides using liquid chromatography followed by tandem mass spectrometry (LC-MS/MS). When the samples were digested with AspN, peptide A [amino acids 11 to 37, mass-to-charge ratio (m/z) = 1488.7 when z = 2] was observed for DA-Rac (Fig. 3A). For DA-Rac/VopS, only the modified form of peptide A was observed [amino acids 11 to 37, with a mass increase of 329.1 daltons, m/z = 1653.3 when z = 2] (Fig. 3B). Peptide A contained a covalent modification of 329 daltons on Thr35 (Fig. 3C), a conserved residue located in the effector loop of the switch I region of Rho family GTPases that plays a role in GTP, Mg2+, and effector binding (Fig. 3D) (10, 12). Thus, VopS modifies the Rho family GTPases with AMP, which prevents Rho GTPases from binding to downstream effectors by steric hindrance.

Fig. 3.

DA-Rac-VopS is AMPylated on Thr35 in the switch I region. (A and B) Extracted ion chromatogram (XIC) of the wild-type peptide A (m/z = 1488.7) and the modified peptide A (m/z = 1653.3) for DA-Rac (A) and DA-Rac/VopS (B). The XIC indicates the intensity of the ion as a function of time during the LC-MS/MS process. (C) Electrospray ionization MS/MS spectra of modified peptide A. The b and y ions are marked on the MS/MS spectra. Ions corresponding to products of internal fragmentation are marked in brackets. Compared to the MS/MS of the wild-type peptide A (see supporting online material), two forms of mass shift were detected, either with a mass decrease of 18 daltons or with a mass increase of 329 daltons. Ions with a mass shift are marked with an asterisk. The mass shift started with y3, indicating that the modification site is Thr35. A further cut of peptide A with AspN resulted in an ion corresponding to peptide B (amino acids 32 to 37) with a mass increase of 329 daltons, and the MS/MS spectrum of modified peptide B further confirmed that Thr35 is the modification site. (D) Alignment of the effector loop region corresponding to residues 32 to 50 of RhoA and 30 to 48 of RacI and Cdc42. Conserved residues are shown in red and the switch I region is underlined. The asterisk denotes the AMPylated Thr residue.

To test directly whether VopS functions as an enzyme to modify Rho family GTPases with AMP, we performed an in vitro labeling assay with 32P-α–labeled adenosine triphosphate (ATP). Incubation of purified recombinant VopSΔ30 with recombinant DA-Rac and 32P-α–labeled ATP resulted in VopS-dependent modification of DA-Rac (Fig. 4A). VopSΔ30 did not modify DA-Rac-T35A, confirming that the AMP modification is specific for Thr35 (Fig. 4A). As expected, DA-Rac incubated with VopS-H348A was not modified (Fig. 4A). To confirm that VopS modifies other members of the Rho family GTPases, we repeated the in vitro labeling assay using Rho, Rac, and Cdc42. In the presence of VopS, all of the GTPases were modified with AMP, whereas they were not modified in the presence of 32P-α–labeled ATP alone or by VopS-H348A (Fig. 4B). Thus, VopS modifies Rho GTPases with AMP. We now refer to this activity as AMPylation and the enzyme as an AMPylator. VopS uses this posttranslational modification of AMPylation to hinder signaling between Rho GTPases and their downstream effectors by blocking the effector binding site on the switch I region of the GTPase with AMP.

Fig. 4.

In vitro AMPylation of Rho family GTPases by VopS. (A) Recombinant VopSΔ30 and VopSΔ30-H348A were incubated with or without DA-Rac or DA-Rac-T35A in the presence of 32P-α–labeled ATP. (B) Recombinant VopSΔ30 or VopSΔ30-H348A were incubated with GTP-Rho, GTP-Rac, or GTP-Cdc42 in the presence of 32P-α–labeled ATP. (C) S100 HeLa cell lysates or boiled, denatured S100 HeLa cell lysates were incubated with 32P-α–labeled ATP or 32P-γ–labeled ATP in the presence or absence of VopSΔ30 (125 pmol). Samples in lane 4 and 8 were treated with λ-phosphatase (λ-Ppase, 400 units) in the last 5 min of incubation. Samples were separated by SDS-PAGE and analyzed by autoradiography.

Both VopS and protein kinases use ATP to modify substrates, but the phosphate attached to the substrate is distinct. Kinases use the γ phosphate of ATP to modify their substrates on tyrosine, threonine, and serine residues, whereas VopS uses the α phosphate linked to adenosine to modify its substrate on a threonine residue. This type of posttranslational modification on eukaryotic proteins has not previously been observed. However, it has been observed for bacterial glutamine synthetase, albeit autocatalytically on a tyrosine residue, resulting in the sensitization of end-product inhibition (13, 14). Because bacterial type III secreted effectors often mimic eukaryotic mechanisms, the observation of AMPylation by a bacterial effector prompted us to investigate whether eukaryotes use this posttranslational modification. Incubation of S100 HeLa cell lysates with 32P-γ–labeled ATP predictably revealed many phosphorylated protein substrates (Fig. 4C). This modification, phosphorylation, was labile in the presence of a phosphatase (Fig. 4C). To test whether the same type of experiment would reveal AMPylated protein substrates, we incubated S100 lysate with 32P-α–labeled ATP. A number of radiolabeled proteins were observed but were insensitive to phosphatase treatment (Fig. 4C). The addition of purified recombinant VopSΔ30 to the reaction using 32P-α–labeled ATP, but not 32P-γ–labeled ATP, specifically increased labeling at the predicted size of the Rho GTPases (Fig. 4C). Thus, VopS is not a promiscuous AMPylator but rather targets the Rho family of GTPases. Consistent with this observation, phosphorylation and AMPylation did not occur in the presence of denatured protein (Fig. 4C). Thus, eukaryotic proteins can use ATP to modify proteins by AMPylation.

VopS contains a C-terminal Fic domain, and mutation of an invariant histidine residue within this domain led to the discovery of the catalytic activity of modifying proteins with AMP. The conserved histidine is critical for the AMPylation activity. The limited eukaryotic distribution of Fic resembles that of other components of signal transduction machinery and might support a role for AMPylation by eukaryotic Fic domains in signaling. Structures of Fic domains place the conserved polar residues of this motif within a cleft that could represent an active site, with conserved side chains (from E and N) forming polar contacts with a phosphate in one structure (fig. S3A) (15). A β hairpin located near the motif binds peptide in another structure, placing a side chain of the peptide within van der Waals contact of the motif histidine (fig. S3B). Although enzymes, such as an activated E1, form AMP-bound covalent enzyme intermediates to drive chemical ligation reactions (16), AMP has not previously been shown to be used as a stable posttranslational modification for a protein. This activity represents an ideal posttranslational modification because it (i) uses a highly abundant high-energy substrate, ATP; (ii) results in the formation of a reversible phosphodiester bond; (iii) is bulky enough to bind to an adaptor protein and be used in dynamic multidomain signaling complexes; and (iv) alters the activity of the protein it modifies. It is intriguing that we observed this modification on threonine because this residue is used in many other modifications that might compete with AMPylation. The identification of the substrates and enzymes involved in eukaryotic AMPylation will undoubtedly add a new layer to the expanding complexity of our information about cellular signal transduction.

Supporting Online Material

Materials and Methods

Figs. S1 to S3


References and Notes

View Abstract

Navigate This Article