Nε-Fatty acylation of Rho GTPases by a MARTX toxin effector

See allHide authors and affiliations

Science  27 Oct 2017:
Vol. 358, Issue 6362, pp. 528-531
DOI: 10.1126/science.aam8659

Bacterial toxin fatty-acylates lysine residues

A toxin produced by the bacterium that causes cholera has a catalytic activity that contributes to its effects on the cytoskeleton of host cells. Zhou et al. determined the protein structure of the Rho guanosine triphosphatase (GTPase)–inactivation domain of the toxin from Vibrio cholerae and found it to be similar to that of a human fatty acyltransferase. Indeed, the toxin peptide could catalyze fatty acylation of lysine residues of Rho-family GTPases, which regulate the actin cytoskeleton. Such covalent modification of lysine residues in mammalian proteins had been noted before, but the enzymes responsible were not known.

Science, this issue p. 528


The multifunctional autoprocessing repeats-in-toxin (MARTX) toxins are a family of large toxins that are extensively distributed in bacterial pathogens. MARTX toxins are autocatalytically cleaved to multiple effector domains, which are released into host cells to modulate the host signaling pathways. The Rho guanosine triphosphatase (GTPase) inactivation domain (RID), a conserved effector domain of MARTX toxins, is implicated in cell rounding by disrupting the host actin cytoskeleton. We found that the RID is an Nε-fatty acyltransferase that covalently modifies the lysine residues in the C-terminal polybasic region of Rho GTPases. The resulting fatty acylation inhibited Rho GTPases and disrupted Rho GTPase–mediated signaling in the host. Thus, RID can mediate the lysine Nε-fatty acylation of mammalian proteins and represents a family of toxins that harbor N-fatty acyltransferase activities in bacterial pathogens.

MARTX toxins are critical virulence factors of many bacterial pathogens (1). The secreted MARTX toxins insert into the host cell plasma membrane, where they can be autocatalytically cleaved by their conserved inositol hexakisphosphate–activating cysteine protease domain to release multiple effector domains into the cytosol of host cells (2) (fig. S1). The Rho guanosine triphosphatase (GTPase) inactivation domain (RID) was implicated in cell rounding by inducing a substantial decrease in the cellular levels of active GTP-bound Rho GTPases and disrupting the host actin cytoskeleton through an unknown mechanism (3).

The mutations Cys3022 → Ser (C3022S) and His2782 → Ala (H2782A) of RID of the Vibrio cholerae MARTX toxin (RIDvc) abolished its effect on the actin cytoskeleton (4) (Fig. 1A), which indicates that RIDvc may be an enzyme with the catalytic residues Cys3022 and His2782. We isolated proteins that associated with recombinant maltose-binding protein (MBP)–fused RIDvc C3022S in bovine brain cell extract. A specific protein bound by MBP-RIDvc C3022S was identified as a band of ~25 kDa molecular mass on an SDS–polyacrylamide gel electrophoresis (PAGE) gel (fig. S2). Mass spectrometry analysis of this band revealed it to be Rac1, a Rho family small GTPase that regulates the actin cytoskeleton. Consistently, ectopically expressed RIDvc C3022S interacted with endogenous Rac1 in 293T cells (Fig. 1B). Moreover, recombinant RIDvc protein directly interacted with Rac1 in vitro (Fig. 1C). The binding affinity of RIDvc to full-length Rac1 without prenylation was as high as 4 μM (fig. S2D). Thus, RIDvc may target the host Rac1.

Fig. 1 RID targets and modifies the host Rac1.

(A) Disruption of the host actin cytoskeleton by RIDvc. Wild-type (WT) RIDvc and its mutants H2782A and C3022S were cotransfected with the pEGFP-C1 vector into HeLa cells. The actin cytoskeleton was stained with rhodamine-labeled phalloidin. (B) Interaction between RIDvc C3022S and endogenous Rac1 in 293T cells via an immunoprecipitation (IP) assay. (C) Direct interaction between RIDvc C3022S and wild-type Rac1 in a glutathione S-transferase (GST) pulldown assay. (D) Modification of Rac1 Q61L by RIDvc in 293T cells. The Flag-tagged Rac1 in the samples was analyzed using Western blotting with antibody to Flag. The RIDvc-modified Rac1 is marked with an asterisk. (E) Inactivation of Rac1 by RIDvc in vivo. The GTP-bound active form of Rac1 in 293T cells was examined using PBD pulldown assays. (F) Disruption of the interaction between Rac1 Q61L and RhoGDI by RIDvc.

The recombinant wild-type RIDvc protein, but not the inactive C3022S mutant, caused a downward shift of endogenous Rac1 in 293T cell lysates on an SDS-PAGE gel (fig. S3A). Coexpression of wild-type RIDvc, but not the C3022S mutant, with wild-type Rac1 or the constitutively active Rac1 mutant Gln61 → Leu (Q61L) in 293T cells also caused a downward shift of Rac1 on SDS-PAGE gels (Fig. 1D and fig. S3B). These data indicate that Rac1 underwent an unknown modification by RIDvc.

When wild-type Rac1 was coexpressed with RIDvc in 293T cells, there was no exogenous active GTP-bound Rac1 pulled down by the Rac1/Cdc42 (p21)–binding domain (PBD) of the human p21-activated kinase 1 (PAK1) (Fig. 1E). However, RIDvc had only slight effects on the interaction between Rac1 Q61L and PBD, even though RIDvc also modified Rac1 Q61L (fig. S3C). This finding indicates that the RIDvc modification site on Rac1 appears not to be located in the switch regions that the PBD binds. RIDvc-modified Rac1 Q61L could not be bound by the Rho GDP dissociation inhibitor (RhoGDI) (Fig. 1F). All RIDvc-modified Rac1 was localized to the cell membrane (fig. S3D), which suggests that the RIDvc-catalyzed modification disrupts the cycling of Rac1 between the cell membrane and the cytosol.

We determined the crystal structure of the full-length RID from the V. vulnificus MARTX toxin (RIDvv), which has 87% sequence identity with RIDvc, at a resolution of 2.7 Å (table S1). The catalytic Cys2835 of RIDvv (which corresponds to Cys3022 in RIDvc) was mutated to Ala to avoid oxidation during crystallization. The overall structure of RIDvv is a twisted U-shape architecture with two lobes (Fig. 2A and figs. S4 and S5). The N-terminal lobe contains a membrane localization N2 domain (5) and can specifically interact with phosphatidylinositol 4,5-bisphosphate for RID membrane localization (fig. S4). The C-terminal lobe entirely consists of the catalytic domain, which has a permuted papain-like core structure (Fig. 2B and fig. S6) and a four-helix pair (4HP). The protein of known function that harbors the highest structural similarity to the catalytic domain of RIDvv is HRAS-like tumor suppressor 3 (HRASLS3), a human fatty acyltransferase (6) (Fig. 2C). In the structural comparison, the residues Ala2835 and His2595 of RIDvv adopt the same conformations as Cys113 and His23 in the catalytic dyad of HRASLS3, respectively. In addition, RIDvv exhibited a high binding affinity of 8.85 μM to palmitoyl–coenzyme A (CoA), a long-chain fatty acyl–CoA that is abundantly present in mammalian cells (fig. S7A). These findings indicate that RID may be a fatty acyltransferase that uses palmitoyl-CoA or other long-chain acyl-CoAs as its ligands to modify Rac1.

Fig. 2 Crystal structure of RID from V. vulnificus.

(A) Overall structure of RIDvv. The catalytic residues His2595 and Cys2835 [mutated to Ala (C2835A) in the structure] are indicated as sticks in yellow. A schematic diagram of domains in RIDvv is shown at the top. (B) The permuted papain-like fold of the catalytic domain of RIDvv. (C) Structural comparison between the catalytic core structure of RIDvv and HRASLS3 (6). The catalytic residues of HRASLS3 are shown as blue sticks.

We used alkyne-labeled stearic acid (alk-16) to metabolically label cells and detect fatty acylated proteins in cells via biotin tags (fig. S7B). RIDvc indeed catalyzed the fatty acylation of Rac1 in 293T cells (Fig. 3A). Recombinant wild-type RIDvc protein that was delivered into 293T cells by the N-terminal domain of anthrax lethal factor also modified Rac1 in vivo (fig. S8A). However, in an in vitro reaction system using purified recombinant proteins, RIDvc did not modify full-length Rac1 recombinant protein purified from Escherichia coli, indicating that the prenylation on the cysteine residue in the C-terminal CAAX motif of Rac1 may be required for RIDvc modification. We purified the Rac1 mutant protein Rac1LS (with the last leucine residue of Rac1 mutated to serine), which enabled prenylation of Rac1 by human farnesyltransferase (FTase) in vitro (7). When FTase and its ligand farnesyl diphosphate (FPP) as the prenyl donor were added to the reaction system, RIDvc exhibited a strong acyltransferase activity toward the prenylated Rac1LS and induced a downward shift of Rac1LS on an SDS-PAGE gel (Fig. 3B and fig. S8B). RIDvc fatty acylated both wild-type Rac1LS and Rac1LS Q61L, but did not modify the inactive Rac1LS mutant Thr17 → Asn (T17N) (Fig. 3C and fig. S8C). RIDvc also efficiently modified Rac1 during V. cholerae infection (Fig. 3D). RIDvc could modify Rac1LS using lauroyl-CoA and decanoyl-CoA but could not catalyze the modification using acyl-CoAs with chain lengths of fewer than nine carbon atoms as ligands (fig. S8D). Thus, RID is a long-chain fatty acyltransferase that modifies Rac1.

Fig. 3 Fatty acyltransferase activity of RID modifies host Rac1 GTPase.

(A) Fatty acylation of Rac1 Q61L by RIDvc in vivo. 293T cells were cotransfected with Rac1 Q61L and RIDvc and metabolically labeled with alk-16. Fatty acylation of Rac1 was detected via click chemistry reactions and Western blotting with a streptavidin-biotinylated horseradish peroxidase complex (fig. S7B). Rac1 Q61L was detected via antibody to Flag. (B) Fatty acylation of Rac1LS by RIDvc in vitro. Samples were analyzed via SDS-PAGE and Coomassie blue staining. (C) Fatty acylation of wild-type Rac1LS and Rac1LS Q61L, in contrast to inactive Rac1 T17N, by RIDvc in vitro. (D) Fatty acylation of Rac1 Q61L by RIDvc during V. cholerae infection. V.c denotes the wild-type V. cholerae strain; ΔE denotes a mutant V. cholerae strain in which all three effector domains of the MARTX toxin were in-frame deleted. ΔE+rid WT and ΔE+rid CS indicate the ΔE strains complemented with the wild-type RIDvc and inactive RIDvc C3022S, respectively.

After fatty acylation by RIDvc, Rac1 could not be cleaved by YopT, a Yersinia cysteine protease cleaving the C-terminal prenylated cysteine residues of Rho GTPases (8) (Fig. 4A), indicating that the modification site on Rac1 by RID is near the prenylated cysteine residue. The polybasic region (PBR) is adjacent to the prenylated cysteine residue in Rac1 (fig. S9A). The SUMO-tagged last 13 residues of Rac1LS, which include both the PBR and the prenylated cysteine residue, were indeed fatty acylated by RIDvc in vitro (Fig. 4B). The modification efficiency of the C-terminal fragment by RIDvc was less than that of the full-length Rac1LS. Consistently, RIDvc also interacted with the core G domain (residues 1 to 178) of Rac1 in yeast two-hybrid assays (fig. S2C). Therefore, RID specifically recognizes both the core G domain and the prenylated C terminus to fatty acylate Rac1 on its polybasic region.

Fig. 4 RID fatty acylation of lysine residues in the polybasic region of Rac1.

(A) Resistance of RIDvc-modified Rac1 to YopT digestion. The YopT-mediated cleavage induced an upward shift of Rac1 on SDS-PAGE gel. (B) Fatty acylation of the C-terminal 13–amino acid fragment of Rac1 by RIDvc in vitro. SUMO-Rac1LS C13 represents the last 13 residues of Rac1LS fused with an N-terminal SUMO tag. Fatty acylation was detected via in vitro click chemistry assays. (C) Mutagenesis analysis of the fatty acylation sites of Rac1 by RIDvc. 2RA denotes the double mutation of Arg185 → Ala and Arg187 → Ala; 4KA denotes the quadruple mutation of Lys183 → Ala, Lys184 → Ala, Lys186 → Ala, and Lys188 → Ala. (D) Mass spectrometry analysis of the molecular weight of Rac1LS that was palmitoylated by RIDvc. Each palmitoylation modification increases the molecular weight of the protein by 238 daltons. RIDvc transferred two, three, and four palmitoyl groups onto Rac1.

Double mutation of the two arginine residues in the PBR had no effect on the fatty acylation of Rac1 by RIDvc (Fig. 4C). The mutation of all four lysine residues in the PBR to alanine residues completely abolished the modification (Fig. 4C and fig. S9B). The single mutation of Lys183 or Lys184 had a slight effect on the Rac1 modification by RIDvc, whereas the single mutation of Lys186 or Lys188 severely inhibited the fatty acylation (Fig. 4C). We used mass spectrometry to analyze the recombinant Rac1LS protein that was modified by RIDvc with palmitoyl-CoA as its ligand. Because of the strong hydrophobic property of the fatty acyl and prenyl groups in the Rac1 C terminus, only an 11–amino acid peptide from the PBR, which contains Lys183 and Lys184, was identified in the secondary mass spectrometry results (fig. S9C). Both Lys183 and Lys184 were palmitoylated by RIDvc. We further measured the molecular weight of the RIDvc-modified Rac1LS protein. Mass spectrometry results indicated that Rac1 was modified with two, three, and four palmitoyl groups by RIDvc (Fig. 4D). Thus, all four lysine residues in the PBR can be modified by RIDvc. Fatty acylation of Rac1 at multiple sites is consistent with the finding that the RIDvc-modified Rac1 from the in vitro reaction system ran as a smeared band on SDS-PAGE gels (Fig. 3C).

We also observed fatty acylation by RIDvc of the lysine residues in the PBRs of other Rho GTPases: Rac2, Rac3, RhoA, and Cdc42 (figs. S10 to S12). Consistent with the absence of detectable interactions of RIDvc with Cdc42 or RhoA, modification of RhoA or Cdc42 by RIDvc was much less than that of Rac1. Thus, RIDvc can modify all members of Rho GTPases but has a substrate preference for Rac1. Like RIDvc, RIDvv has the same Nε-fatty acyltransferase activity toward Rho GTPases (figs. S13 and S14).

The fatty acylation by RIDvc inhibited Rac1 activation by different types of guanine nucleotide exchange factors (fig. S15), which is consistent with the observation that RIDvc induced a low level of active GTP-bound Rac1 in vivo. RIDvc-modified Rac1 also could not activate the downstream kinase PAK1 (fig. S16). During infection, RIDvc inhibited Rho GTPase–mediated cellular processes including phagocytosis, reactive oxygen species production, and migration of infected cells (fig. S17). Moreover, RID is required for the virulence of V. vulnificus during mouse infection (fig. S17D).

Our results show that RID is an Nε-fatty acyltransferase that modifies the lysine residues in the C-terminal polybasic regions to inactivate Rho GTPases. The catalytic domain of RIDvc shares 12 to 16% sequence identity with the type III effectors IcsB from Shigella species and BopA from Burkholderia species (4). Long-chain fatty acylation has been shown to occur on the lysine residues of some cytokines and histone proteins in mammalian cells (911). However, the enzyme for this modification has not been identified. Our study describes an enzyme that represents a family of bacterial toxins able to catalyze lysine Nε-fatty acylation on host proteins.

Supplementary Materials

Materials and Methods

Table S1

Figs. S1 to S17

References (1235)


Acknowledgments: We acknowledge F. Shao for sharing unpublished work with Y.Z. on the RID1 homolog IcsB from S. flexneri. We thank the staff in the beamline BL18U1 of SSRF and NCPSS for assistance in diffraction data collection, and Z. Chen, X.-H. Feng, J. Han, and K.-L. Guan for assistance in obtaining reagents. Supported by National Natural Science Foundation of China (NSFC) grants 81530068, 81322024, 31370722, and 81561130162, Zhejiang Natural Science Foundation grant LR13C050001, and Royal Society grant NA140239 (Y. Zhu); NSFC grant 81501717 (Y. Zhou); NSFC grant 31225002 and China Ministry of Science and Technology grants 2010CB835400 and 2012CB518700 (F.S.); NSFC grant 21622501 (X.L.); and the Fundamental Research Funds for the Central Universities.

Correction (13 April 2018): Citations for references not previously cited were missing in the caption of figures S4, S6, and S17 of the supplementary materials. The supplementary PDF file and the HTML reference list have been updated online.

View Abstract

Navigate This Article