Bacterial factors exploit eukaryotic Rho GTPase signaling cascades to promote invasion and proliferation within their host

Abstract

Actin cytoskeleton is a main target of many bacterial pathogens. Among the multiple regulation steps of the actin cytoskeleton, bacterial factors interact preferentially with RhoGTPases. Pathogens secrete either toxins which diffuse in the surrounding environment, or directly inject virulence factors into target cells. Bacterial toxins, which interfere with RhoGTPases, and to some extent with RasGTPases, catalyze a covalent modification (ADPribosylation, glucosylation, deamidation, adenylation, proteolysis) blocking these molecules in their active or inactive state, resulting in alteration of epithelial and/or endothelial barriers, which contributes to dissemination of bacteria in the host. Injected bacterial virulence factors preferentially manipulate the RhoGTPase signaling cascade by mimicry of eukaryotic regulatory proteins leading to local actin cytoskeleton rearrangement, which mediates bacterial entry into host cells or in contrast escape to phagocytosis and immune defense. Invasive bacteria can also manipulate RhoGTPase signaling through recognition and stimulation of cell surface receptor(s). Changes in RhoGTPase activation state is sensed by the innate immunity pathways and allows the host cell to adapt an appropriate defense response.

Introduction

Actin cytoskeleton is a major cell structure which is involved in multiple functions and is one of the main targets of bacterial pathogens. Indeed, the actin cytoskeleton controls cellular function critical for eukaryotic life including locomotion, endocytosis, exocytosis, trafficking of intracellular organelles as well as maintenance of intercellular connections. Since intercellular junctions are essential in building cell barriers, like the epithelia and endothelia that protect an organism from the environment as well as delineate internal compartments, pathogen manipulation of host via the actin cytoskeleton provides an effective means to breach a host's containment capabilities and an access to essential nutrients from more appropriate niches throughout the body. Pathogens also manipulate the host actin cytoskeleton to enter non-phagocytic target cell and then to escape to immune defenses, or in contrast to avoid phagocytosis by macrophages. Actin cytoskeleton is a highly dynamic structure which is regulated by numerous proteins and factors including GTPases from the Rho family which play a key role. Due to long interactive life of bacteria with higher organisms, pathogens have engineered various strategies to attack or to survive in hostile eukaryotic cell environment. Certain pathogens secrete potent toxins which diffuse in the surrounding environment and induce tissue destruction allowing bacterial growth and dissemination. Other bacteria inject directly into the target cells virulence factors, which hijack the actin cytoskeleton machinery for their own profit, notably to escape immune defenses. Excellent publications have described the multifaceted interactions between bacterial effectors and RhoGTPases.^1-10 The bacterial effectors which manipulate the eukaryotic GTPases are also called modulins.¹¹ This review is focused on the diversity of RhoGTPase manipulations by bacterial pathogens to disseminate and survive in host.

Rho-GTPase Manipulation by Bacterial Toxins and Alteration of Cell Barriers

Toxigenic bacteria release in their environment potent toxins, which interact with the host tissues and are responsible for the lesions and symptoms of the disease caused by these microorganisms. Various bacterial toxins manipulate the actin cytoskeleton either by interfering directly with actin monomers or by modulating RhoGTPases. Indeed, clostridial toxins from the Iota and C2 toxin families are proteins, which ADP-ribosylate actin monomers at Arg177 and thus impair actin monomer/monomer assembly. This results in loss of cellular actin filaments.^12,13 Other toxins target specifically RhoGTPases that they covalently modify. This triggers an amplified effect on the actin cytoskeleton compared with the toxins active on the actin monomers, since RhoGTPases are upstream regulators of actin polymerization. Moreover, these toxins also alter other cellular functions controlled by RhoGTPases.

Enzymatic modification of RhoGTPases by bacterial toxins results either in activation or inactivation of these molecules. Toxins, which inactivate RhoGTPases, add a cumbersome moiety (AMP, ADP-ribose, glucose, or N-acetyl-glucosamine) to a residue of switch I or in a close area, whereas activating toxins of RhoGTPases catalyze the ADP-ribosylation or deamidation of a critical glutamine of switch II in glutamic acid. RhoGTPases are active in the GTP-bound form and inactive in the GDP-bound form. These proteins adopt a conformational change localized in two regions called switch I and switch II, when they are in GTP- or in GDP-bound form. The transition from GDP- (inactive) to GTP-bound form (active) is enhanced by an exchange factor (GEF) in response to an external signal transduced by a membrane receptor. Switch I interacts with the downstream effectors and switch II is involved in the GTPase activity that is stimulated by a GTPase activating protein (GAP). GTPases act as molecular switches between a membrane receptor activated by an external factor and downstream effectors. RhoGTPases bound to GDP are localized in the cytosol in association with a protein called guanine dissociation inhibitor (GDI), and they translocate to the membrane after dissociation of the Rho-GDI complex possibly mediated by ERM proteins. At the membrane, Rho is activated by GEFs and interacts with its effectors^14-16 (Fig. 1). Indeed, modification of residues in switch I impair RhoGTPase binding to their effectors and are blocked in their inactive state, whereas alteration of functional residue in switch II prevents their GTPase activity yielding permanent active molecules (Fig. 2 and 3).

Figure 1. RhoGTPase cycle. In the inactive GDP-bound form, RhoGTPases are localized in the cytosol in complex associated with a guanine nucleotide dissociation inhibitor (GDI), which prevents nucleotide exchange. Upon a cell signaling, a guanine nucleotide exchange factor (GEF) induces the release of GDP. Since the GTP concentration in the cytosol is 100 fold higher than that of GDP, RhoGTPases load GTP and become active. The active GTP-bound RhoGTPase adopts a conformational change of two surface loops named switch I (I) and switch II (II). Switch I is the main region for interactions with effector molecules and switch II plays an important role in GTP catalysis. The intrinsic low GTPase activity of RhoGTPase is enhanced by GTPase-activating proteins (GAPs), thereby inactivating RhoGTPases. Guanine-nucleotide-dissociation inhibitor s(GDIs) stabilize the GDP, thereby the inactive, forms of RhoGTPases in the cytosol.

Figure 2. Inactivation of RhoGTPase signaling by bacterial toxins and virulence factors. Toxins inactivating RhoGTPases modify a residue of switch I by addition of a cumbersome moiety (ADPR by C3, glucose by large clostridial glucosylating toxins (LCGT) or P. asymbiotica toxin (PaTox), AMP by VOPs or IbpA) which impairs the interaction with downstream effector, or proteolytically cleaves RhoGTPase C-terminal part preventing its docking to the membrane and subsequent interaction with downstream effector. Virulence factors modify the RhoGTPase cycle signaling by mimicry eukaryotic regulatory proteins (GAP, GDI, GEF inhibitor).

Figure 3. Activation of RhoGTPase signaling by bacterial toxins and virulence factors. Toxins activating RhoGTPases modify a residue of switch II (deamidation, ADPribosylation) which impairs the intrinsic GAP activity thus yielding permanent active molecules. Bacterial virulence factors mimic GEF thus activate RhoGTPases.

RhoGTPase inactivating toxins and endothelial barrier permeability

C3 exoenzyme is produced by Clostridium botulinum type C and D, and C3 related exoenzymes are also synthesized by Clostridium limosum, Bacillus cereus, Bacillus thuringiensis and Staphylococcus aureus in which it is called epidermal cell differentiation inhibitor (EDIN).^1,17 C3 from C. botulinum was the first toxin, which has been found to interact with Rho proteins and was of a great interest to elucidate their function on the control of actin polymerization. All C3 exoenzymes recognize RhoA, B and C, and in addition, EDIN also modifies RhoE.¹⁸

C3 exoenzymes are small proteins (about 28 kDa) which only possess a catalytic domain and lack the binding and translocation domains permitting their entry into cells. The crystal structure shows that C3 consists of a core structure of five antiparallel β-strands packed against a three-stranded antiparallel β-sheet, and flanked by four consecutive α-helices.^19,20 Interestingly, the C3 structure is similar to that of the catalytic domain of the actin ADP-ribosylating toxins such as C. perfringens Iota toxin and Bacillus vegetative insecticidal protein (VIP).^13,19,21 Although there is no significant overall sequence homology with other ADP-ribosylating toxins, C3 retains the conserved NAD binding site and catalytic pocket which consists of an α-helix (α3 in C3) bent over the two antiparallel β-sheets forming a central cleft. The amino acid (Glu214) that has an essential role in ADP-ribosylation, is conserved.^19,22,23

C3 ADP-ribosylates RhoA at Asn-41 which is localized on an extended stretch close to the switch I. Rho-GDP is a preferential substrate for C3 as Rho-Asn41 is solvent accessible in the GDP structure.²⁴ In contrast, the Asn41 residue of Rho found in a Rho-GDI-complex is hidden and thus resistant to C3-mediated ADP-ribosylation.²⁵ ADP-ribosylation of Rho-Asn41 by C3 does not impair GDP/GTP exchange, does not affect intrinsic and GAP-stimulated GTPase activity, and does not impinge upon Rho interaction with its effectors.^26-28 However, C3 prevents GEF activation of Rho.²⁹ In addition, ADP-ribosylated Rho reassociates more efficiently with GDI than unmodified Rho, thus causing an accumulation of inactive Rho in the cytosol and preventing its translocation to the membrane and subsequent activation by GEFs as well as interaction with its effectors.^30,31 Thereby, ADP-ribosylated Rho is trapped in a permanent inactive form in the cytosol, and subsequently degraded by the proteasome complex²⁹

C3 ADP-ribosylates the three isoforms RhoA, B and C. Most of the cellular effects described with this enzyme are related to RhoA. The first evidence that Rho is involved in the actin cytoskeleton organization comes from the initial study of C3 on Vero cells in which the effects are characterized by a cell rounding up and destruction of actin filaments.³² Since then, the effects of C3 on the actin cytoskeleton and related cellular functions are well documented. C3 induces a disorganization of the actin stress fibers, cell morphology change, alteration of epithelial and endothelial barrier function (mainly by perturbing tight junctions), impairment of endocytosis, exocytosis, phagocytosis, cytokinesis, neuronal plasticity, inhibition of cell cycle progression and migration of immune cells, as well as induction of apoptosis (rev in^33-35). However, the role of C3 in natural disease such as botulism, is not known. C. botulinum can grow and produce toxins in the environment including contaminated food or in the intestinal lumen, and the passage of botulinum neurotoxin through the intestinal barrier and trafficking to the target motorneurons are responsible for the neurological symptoms of paralysis. C3 does not enter cells actively, since receptor binding and translocation domains are lacking. But, C3 enzymes are selectively internalized into macrophages and monocytes via acidic endosomes.³⁶ Since C3 can inhibit Rho-mediated phagocytosis in macrophages,³⁷ it may play an immunosuppressive role. In addition to its ADP-ribosylation activity, C3 exerts ADP-ribosylation-independent effects. C3 binds to RalA a GTPase from the Ras family, via a site adjacent but distinct from the catalytic site. C3 binding results in a stabilization of RalA in its GDP-bound and thus inactive conformation preventing its interaction with downstream effectors.^38,39 The biological effects of C3 interaction with Ral remains to be elucidated.

In contrast, EDIN which is produced by certain S. aureus strains, is considered as an important virulence factor, which is notably involved in impetigo, diabetic foot ulcers, and other skin infections.^40-43 S. aureus can invade eukaryotic cells and release EDIN intracellularly, which contributes to actin cytoskeleton disorganization and tissue destruction.⁴⁴ Thus, EDIN facilitates bacterial dissemination through the altered tissues. Indeed, in a mouse model EDIN promotes increased infection foci in deep tissues.⁴⁵ An original effect triggered by EDIN and related C3 enzymes consists in the perturbation of endothelial permeability by formation of transcellular channels. Thereby, EDIN-mediated RhoA inactivation in endothelial cells induces a reorganization of the actin cytoskeleton resulting in the formation of transient macroapertures termed large transendothelial cell macroaperture tunnels (TEMs).^4,46,47 The increased endothelial permeability facilitates the binding of S. aureus to extracellular matrix proteins and its dissemination from the blood stream to underlying tissues.

RhoGTPase inactivating toxins and epithelial/endothelial barrier permeability as well as inflammatory response

The manipulation of epithelial barrier permeability by RhoGTPase inactivating toxins is well illustrated with the large clostridial glucosylating toxins (LCGTs). These toxins are 250–300 kDa proteins encompassing Clostridium difficile toxins A and B (TcdA and TcdB), Clostridium sordellii lethal toxin (TcsL), and hemorrhagic toxin (TcsH)), Clostridium novyi αlpha-toxin TcnA), and C. perfringens TpeL (toxin C. perfringens large cytotoxin) (Table 1).

Table 1. Examples of bacterial toxins that modulate Rho/RasGTPase signaling and alter epithelial/endothelial barriers.

Toxin	Pathogen	Target	Biochemical activity	Effects	Reference
Toxins inhibiting RhoGTPases
C3 exoenzymes
C3bot	Clostridium botulinum C and D	RhoA, B, C	ADP-ribosylation at N41 Cosubstrate NAD	Actin filament depolymerization, macroaperture	32
C3lim	Clostridium limosum				169 , 170
C3cer	Bacillus cereus				171
C3stau-1, -2, -3 or EDIN-A, -B, -C	Staphylococcus aureus	RhoA, B, C, E			18 , 172 - 174

Large clostridial glucosdylating toxins
TcdA, TcdB	Clostridium difficile	RhoA, B, C, G Rac, Cdc42 TC10, TCL	Glucosylation at T37 of RhoA, T35 of Rac1 (and equivalent) Cosubstrate UDP-glucose	Actin filament depolymerization Alteration of intercellular junctions Apoptosis, necrosis	53 - 55
TcdBF	Clostridium difficile 1470	Rac, Cdc42, R-Ras, Ral, Rap			175
TcsL	Clostridium sordellii	RhoG, Rac Cdc42 (variable) Ras, Rap, Ral TC10, TCL R-Ras1, 2, 3			53 , 56 , 176
TcsH	Clostridium sordellii	RhoA, Rac, Cdc42			177
TcnA	Clostridium novyi	RhoA, Rac, Cdc42	N-acetyl-glucosamination at T37 of RhoA Cosubstrate UDP-N-acetylglucosamine		178
TpeL	Clostridium difficileperfringens	Ras, Ral, Rap	N-acetyl-glucosamination and glucosylation at T35 of Ras	Apoptosis	179 , 180
PaTox	Photorhabdus asymbiotica	RhoA,B,C Rac1,2,3, Cdc42	N-acetyl-glucosamination at Y34 of RhoA, Y32 of Rac and Cdc42	Actin cytoskeleton disorganization Anti-phagocytosis Toxicity toward insect larvae	84

Multifunctional-autoprocessing RTX toxins
MARTX	Vibrio cholerae, Vibrio sp.	Rho, Rac, Cdc42	Inactivation (non yet defined mechanism)	Actin filament depolymerization Intestinal colonization, escape of immune cells	86 , 87

Toxins activating RhoGTPases
Deamidating toxins
Dermonecrotic toxin (DNT)	Bordetella pertussis Bordetella parapertussis Bordetella bronchiseptica	Rho, Rac, Cdc42	Deamidation, transglutamination at Q63 or Q61	actin filament polymerization activated-Rho ubiquitin-mediated proteosomal degradation	181
CNF1, CNF2, CNF3	Escherichia coli		Deamidation at Q61 or Q63		182 , 183
CNFY	Yersinia pseudotuberculosis				184

LCGTs are single protein chains containing at least four functional domains. The one third C-terminal part exhibits multiple repeated sequences (31 short repeats and 7 long repeats in TcdA), which are involved in the recognition of a cell surface receptor. The central part contains a hydrophobic segment and probably mediates the translocation of the toxin across the membrane. The enzymatic site which is characterized by the DxD motif surrounded by a hydrophobic region, and the substrate recognition site are localized within the 543 N-terminal residues forming the enzymatic domain. The overall structure of this domain in TcdB, TcsL and TcnA is conserved and consists of a β-strain central core (about 235 amino acids) forming an active center pocket surrounded by numerous α-helices.⁴⁸ The first aspartic residue of the DxD motif binds to ribosyl and glucosyl moieties of UDP-glucose and the second aspartic residues binds to divalent cation (mainly Mn²⁺) which increases the hydrolase activity and/or the binding of UDP-glucose.⁴⁹ In addition, a cysteine protease domain (543–567) with DHC motif, lies in the vicinity of the autocleavage site.⁴⁸

LCGTs enter cells by receptor-mediated endocytosis and release the enzymatic domain into the cytosol from acidified endosome by an auto-proteolytic activity stimulated by inositol hexakisphosphate.^50-52 LCGTs catalyze the glucosylation of Rho- and/or Ras-GTPases from UDP-glucose, except TcnA, which uses UDP-N-acetylglucosamine as co-substrate. TcdA and TcdB glucosylate Rho, Rac and Cdc42 at Thr-37, whereas TcsL glucosylates Ras at Thr-35, Rap, Ral and Rac at Thr-37.^14,53 The large glucosylating clostridial toxins cleave the cosubstrate and transfer the glucose moiety to the acceptor amino acid of the Rho proteins.^54-56 The conserved Thr, which is glucosylated, is located in switch I. Thr37/35 is involved in the coordination of Mg²⁺ and subsequently to the binding of the β and γ phosphates of GTP. The hydroxyl group of Thr37/35 is exposed to the surface of the molecule in its GDP-bound form, which is the only accessible substrate of glucosylating toxins. The nucleotide binding of the glucosylated G-protein Ras by TcsL is not grossly altered, but the GEF activation of GDP forms is decreased.⁵⁷ Glucosylation of Thr35 completely prevents the recognition of the downstream effector, blocking the G-protein in the inactive form.⁵⁷ The crystal structure of Ras modified by TcsL shows that glucosylation prevents the conformational change in the GTP state of the Ras effector loop, which is required for the interaction with the effector Raf.⁵⁸ Similar results were found with RhoA glucosylated by TcdB.²⁷ In addition, glucosylation of GTPase slightly reduces the intrinsic GTPase activity, completely inhibits GAP-stimulated GTP hydrolysis,⁵⁷ and leads to accumulation of the GTP-bound form of Rho to the membrane where it is tightly bound.⁵⁹

LCGTs by inactivating Rho proteins induce cell rounding, with loss of actin stress fibers, reorganization of the cortical actin, disruption of the intercellular junctions and thus increase in cell barrier permeability. Rho is a major regulator of actin polymerization as well as of tight junction function, whereas Rac is mainly involved in the control of cortical actin and E-cadherin-dependent adherens junctions.¹⁰ Rac inactivation by LCGTs seems to be major player in actin cytoskeleton disorganization.⁶⁰ TcdA and TcdB, which inactivate Rho, Rac and Cdc42, depolymerize apical and basal actin filaments and subsequently disorganize the ultrastructure and component distribution (ZO-1, ZO-2, occludin, claudin) of tight junctions, as well as perturb the organization of adherens junctions resulting in disruption of epithelial barrier function.^61-64 In addition, TcdA and TcdB induce cell detachment, cell death by apoptosis and necrosis, and a severe inflammatory response of the intestinal mucosa.^65-68 Indeed, in intestinal epithelial and immune cells, TcdA and TcdB stimulate the secretion of proinflammatory cytokines (TNF-α, IL-1β, IL-6, IL-8, and many other mediators) mainly via activation of p38 MAP kinase.^65,67,69-71 The activity of the C. difficile RhoGTPase inactivating toxins results in an increased intestinal barrier permeability, fluid secretion, and destruction of the intestinal epithelium. The inflammatory response exacerbates the mucosal damage elicited by the C. difficile toxins. The rupture of the intestinal barrier integrity is a major initial mechanism in the C. difficile infections and facilitates the migration of neutrophils and other immune cells into the intestines. Toxigenic C. difficile colonizes first the intestine of susceptible patients mainly subsequently to a perturbation of the intestinal microbiota induced by antimicrobial agents⁷² and produces toxins which attack the intestinal barrier. Then, the resulting necrotic intestinal mucosa constitutes a favorable site of further C. difficile growth and subsequent release of toxins.

LCGTs are responsible for severe diseases in man and animals, the most common of which are the C. difficile infections. Toxigenic C. difficile strains are the causative agent of pseudomembranous colitis and about 30% of the postantibiotic diarrhea, which are the most frequent nosocomial intestinal diseases.⁷³ TcdA, which experimentally induces necrotic and hemorrhagic intestinal lesions was considered as the main virulence factor.^2,68 However, in the hamster disease model, TcdB was found to be the essential virulence factor by using genetically modified C. difficile strains.⁷⁴ Since C. difficile strains producing both TcdA and TcdB or only TcdB cause enteric disease in humans, TcdB might be also an important enterotoxin.^2,68 Both TcdB and TcdA participate in the alteration of the intestinal barrier and in the recruitment of inflammatory cells, which are abundant in the lesions. C. sordellii and C. novyi are involved in gangrene. C. sordellii induces sporadic cases of fulminant toxic shock syndrome which accompanies local infections, mainly of uterus.⁷⁵ It is also an agent of hemorrhagic enteritis and enterotoxemia in cattle.¹⁷

TcsL of C. sordellii, which only modifies Rac among the Rho proteins, also alters the permeability of epithelium, but through a slightly different way than that of C. difficile toxins. Indeed, TcsL mainly causes a redistribution of E-cadherin whereas tight junctions are not significantly affected.⁷⁶ C. sordellii can cause necrotic and hemorrhagic enteritis, probably in a similar manner than C. difficile colitis, but this pathogen is also responsible for sporadic cases of bacteraemia and toxic shock without myonecrosis of skeletal muscle.⁷⁷ The portal of entry is presumed to be the gastrointestinal tract. Local increased intestinal barrier permeability due to C. sordellii toxins probably facilitates the translocation of bacteria into the blood stream. C. sordellii is also associated with toxic shock syndrome (hypotension, hemoconcentration, pleural effusion, sero-sanguineous ascite) consecutive to local infection of uterus or perineum.^78-82 Diffusion of TcsL in the blood circulation and alteration of vascular endothelial barrier are the main features of the pathogenesis. In experimental mice, TcsL causes a marked edema in the cardio-respiratory system by altering E-cadherin junctions between lung endothelial cells in a Rac modification-dependent manner and a rapid death.⁸³

A novel enzymatically inactivation of RhoGTPases has been recently identified consisting in N-acetyl-glucosamination of a switch I tyrosine. Photorhabdus asymbiotica toxin (PaTox) (334 kDa) contains a C-terminal domain with a conserved glucosyltransferase motif (DxDD), and catalyzes mono-glucosylation of RhoA at Y34 as well as Rac and Cdc42 at Y32 using UDP-N-acetylglucosamine as co-substrate. PaTox induces actin filament disruption and cell rounding, inhibits macrophage phagocytosis, and is toxic for insect larvae.⁸⁴ Modification of the conserved tyrosine in switch I impairs downstream signaling of RhoGTPases as well as activation by RhoGEFs. In addition, PaTox exhibits a second enzymatic activity related to that of RhoGTPase activating toxins (see below). Indeed, PaTox deamidates a crucial glutamine residue of switch II involved in GTP hydrolysis of heterotrimeric G proteins (Gα_q/11 and Gα_i).⁸⁴ The PaTox effects of heterotrimeric G protein activation remain to be elucidated. They might counterbalance the inhibitory activity of RhoGTPases in a spatiotemporal manner during the host infection. P. asymbiotica is an entomopathogen, and also an emerging pathogen for human causing localized soft tissue infection or disseminated bacteremic infection.⁸⁵ Vibrio cholerae is a potent intestinal pathogen, which secretes in addition to the cholera toxin, multiple virulence factors which are involved in the early colonization step by preventing bacteria killing via the innate immune cells like neutrophils and macrophages. Among the accessory toxins, MARTX (multifunctional autoprocessing repeats-in-toxin) is a large type I-secreted protein (350–560 kDa), which alters the actin filaments via two mechanisms mediated by two distinct domains. The actin cross-linking domain is responsible for covalent cross-linking of actin monomers, whereas the Rho-inactivation domain (RID) block Rho, Rac and Cdc42 in their GDP inactive form leading to actin filament depolymerization. MARTX do not directly modify RhoGTPases, but rather interact with regulatory proteins of RhoGTPases. RID shows homology with the cysteine proteases of the Clan CE family, but the exact mechanism of action remains to be identified.^86,87 V. cholerae MARTX toxin carries two effector domains which attack the actin cytoskeleton by two mechanisms, one of which involves a downregulation of RhoGTPases.

RhoGTPase activating toxins

Another class of bacterial toxins enzymatically modify RhoGTPases, but by blocking these molecules in their active form. This toxin family encompasses DNT (dermonecrotic factor) from Bordetella sp, CNF (cytotoxic necrotizing factor) from Escherichia coli, and Yersinia pseudotuberculosis (Table 1). consists of several isoforms which are highly related at the amino acid level. CNF1-producing E. coli strains are pathogens for humans, in which they represent about 30% of uropathogenic strains, and are also involved in enteritis and septicaemia in animals.^88-90 CNF2 strains are mainly isolated from calf and piglet, and CNF3 strains from sheep and goat.^5,91 DNT is more distantly related to CNF and shares sequence identity mainly in the catalytic domain.

CNF is the prototype of the RhoGTPase activating toxins and consists of a single chain proteins (about 110 kDa) containing three functional domains: an N-terminal domain (amino acids 1–299), which is involved in the recognition of a cell surface receptor, a central domain (amino acids 299–720) containing two hydrophobic regions which have been proposed to translocate the toxin across the cell membrane, and a C-terminal (720–1014) catalytic domain. The C-terminal domain of CNF1 has an original protein fold formed by a central β-sandwich, that is composed of two mixed β–sheets, surrounded by α-helices and extensive loop regions.⁹² CNF1 seems to be secreted by a Type I secretion system in the extracellular medium similarly to α-hemolysin.⁵ Then, the toxin recognizes laminin receptor (LPR) on the surface of target cells and it is endocytosed in a clathrin-independent pathway. The enzymatic domain is then released from acidic late endosomes to the cytosol.⁹³

CNF1 catalyzes the deamidation of Gln63 in Rho and Gln61 in Rac and Cdc42 to glutamic acid. Gln63/Gln61 are located in the switch II region of the Rho protein, which has an important function in GTP hydrolysis. Thereby, CNF1 blocks the RhoGTPases in their active form linked to GTP and thereby stimulates actin filament polymerization. Indeed, in fibroblasts like Vero cells, CNF1 causes dense actin stress fibers and focal contact point formations, whereas in epithelial cells (Hep2) the formation of lamellipodia and filopodia predominate. In both cell types, CNF1 leads to cell spreading resulting from the increase in actin filament formation at the leading edge and anchorage of acto-myosin filaments to focal contact points.⁹⁴ However, RhoGTPase activation by CNF1 is only transient, since deamidation increases the sensitization of RhoGTPases to ubiquitylation and subsequent proteasome degradation.⁹⁵ One of the primary CNF-induced disturbances involves modification of intestinal or urinary tract cell barriers. Activation of RhoGTPases results in changes in the dynamic of the actin cytoskeleton and subsequently to alteration of intercellular junction assembly. In addition the increased cell motility even within monolayers, and ubiquitin-mediated degradation of RhoGTPases facilitate the disorganization of intercellular junctions.⁵ Thereby, CNF1 decreases trans epithelial resistance and enhances paracellular permeability from the basal to apical surface of polarized cell monolayers. In T84 cell monolayers, CNF1 induces profound alterations of tight junction components. Occludin, ZO-1, claudin-1, and JAMs are highly redistributed from the apical ring to the cytoplasm, leading to a complete disruption of the continuity of TJ integrity between neighboring cells.⁹⁶ Disruption of the integrity of the epithelium barrier favors the dissemination of bacteria in the underlying tissues. Another aspect of CNF-dependent RhoGTPase activation concerns the stimulation of bacterial internalization into epithelial cells. Thereby, CNF1 triggers bladder cell invasion by uropathogenic E. coli, mainly through Rac activation and increased membrane ruffling as well as a crosstalk between Rac and beta1 integrin.⁹⁵ Thus, bacteria can escape the innate immunity and inflammatory defense, and bacterial persistence into bladder cells might be responsible for recurrence of urinary tract infections.⁵

Bacterial Pathogens Manipulate RhoGTPase Signaling via Injected Virulence Factors to Enter Cells and Colonize Host Tissues

Pathogenic bacteria can invade the host organism at different levels: multiplication in the natural cavities, colonization of the surface of the mucosa, invasion of the extracellular space of tissues, or cell invasion and intracellular life. Here, the term “invasive bacteria” refers to bacteria, which enter and develop inside cells. Invasive bacteria manipulate differently the RhoGTPase pathways to enter the target cells. Instead to secrete toxins into the extracellular medium, they directly inject virulence factors into cells. Thereby, Gram-negative invasive bacteria have developed sophisticated delivery systems allowing the internalization of multiple effectors into a same cell in a temporally and spatially coordinated manner. These delivery systems consist of multicomponent apparatus forming a needle-like structure, which traverses both bacterial envelopes and eukaryotic cell membranes. The effector proteins are delivered through the central pore into the target cell, and in most cases via chaperone proteins which regulate proper folding and translocation. According to the structure and mode of delivery, several types of secretion system have been described. The most commonly used system by invasive bacteria is the type III secretion system, and to a lower extent type IV and type VI. Type III secretion system, also called injectisome, shows similarity with bacterial flagella, whereas type IV secretion system which is also used to deliver DNA has probably evolved from the conjugation machinery. Type VI secretion system retains structural similarity with the T4 bacteriophage tail spike.^97-99

The bacterial effectors delivered into the target cell play multiple functions in bacterial infection. A central role concerns the modulation of the actin cytoskeleton mediating the internalization of bacteria into non-phagocytic cells. Type III secretion system is a common mechanism used by various pathogens to inject virulence factors, which modify the RhoGTPase signaling cascade and thus trigger the bacterial invasion.¹⁰⁰ The virulence factors involved in bacterial invasion do not enzymatically modify RhoGTPases, but mimic protein regulators which activate or inactivate RhoGTPases. One of the best-documented examples of RhoGTPase manipulation by invasive bacteria is that of Salmonella.^98,101-105

Salmonella can bind to intestinal epithelial cells via fimbriae and then enter by a trigger mechanism that induces the formation of large membrane ruffles engulfing the bacteria. The subsequent rearrangements of the actin cytoskeleton and plasma membrane are reminiscent of lamellipodia and filopodia responses stimulated by various agonists such as growth factors, hormones, or activated oncogenes. It has been demonstrated that Rac and Cdc42 are involved in the Salmonella dependent cytoskeletal rearrangements. These effects are mediated by two virulence factors secreted by type III secretion system, SopE and SopE2. Like GEFs, SopE activates Rac1, Rac2, Cdc42, RhoG, and also but to a lesser extent RhoA by catalyzing the exchange of GDP for GTP,¹⁰⁶ whereas SopE2, an isoform of SopE, interacts with Cdc42 but not with Rac1.¹⁰¹ Rac1 plays a central role in actin remodeling and membrane ruffling that mediate Salmonella uptake in non-phagocytic cells. Salmonella exploits the Rac signaling pathway mainly via IQGAP, a scaffold protein which stabilizes Rac and Cdc42 in their active form and promotes actin assembly by interacting with N-WASP and Arp2/3. IQGAP also activates the MAPK/ERK pathway, which contributes to efficient Salmonella uptake, resulting in synergistic effect on bacterial infection.¹⁰⁷ Since Cdc42 activates Rac, a crosstalk between the two RhoGTPase coordinates the actin cytoskeleton rearrangement mediating Salmonella entry.¹⁰³

SopE binds to the switch I and switch II regions of GDP-loaded RhoGTPases like other RhoGEFs, ultimately leading to switch I and II rearrangements in a conformation of weak GDP affinity binding. This results in GDP release. Then, RhoGTPases interact with GTP, which is in higher concentration (200–500 μM) in cells vs. GDP, leading to the conformational change in the active GTP-RhoGTPase structure.^108,109 This mechanism is similar to that used by eukaryotic GEFs which belong to the Dbl family of proteins containing a Dbl homology domain (DH) and a plekstrin homology domain (PH), like the exchange factor Tiam1 (T-lymphoma invasion and metastasis)-1 protein.¹¹⁰ However, the catalytic domain of SopE has a different structure than that of Tiam1 and interacts with the switch regions via a GAGA motif to reorient switches I and II, whereas the catalytic core of other RhoGEFs consists of an α-helix containing a critical Lys at the active site.¹¹⁰ SopE shows neither amino acid sequence nor structure homology with Dbl proteins, but both types of proteins trigger a similar mechanism of nucleotide exchange. Bacterial and eukaryotic GEFs are examples of convergent evolution.

Salmonella produces additional type III secretion factors which trigger redundant SopE effects on the actin cytoskeleton such as SopB, which is involved in the actin cytoskeleton rearrangement including membrane ruffling and bacterial uptake into cells. The effects of SopB are mediated by RhoG and Cdc42 activation. SopB is not a bacterial GEF, it is an inositol phosphate phosphatase which indirectly stimulates Cdc42 by activation of Vav2, a Cdc42-GEF, as well as RhoG via SGEF (SH3-containing guanine nucleotide exchange factor, an exchange factor for RhoG) activation, mediated by the inositol phosphate metabolism.^103,111,112

In addition to their effect to facilitate bacterial uptake, SopE, SopE2, and SopB alter the tight junctions and increase the intestinal barrier permeability in a RhoGTPase activation-dependent manner.¹¹³

Moreover, one report shows that SopE also acts as a GEF for Rab5 and mediates the recruitment of Rab5 in its GTP form to phagosomes containing Salmonella. This promotes the fusion of these phagosomes with early endosomes, preventing their transport to lysozomes and subsequent destruction.¹¹⁴ Thereby, SopE facilitates intracellular survival of Salmonella.

When internalized, Salmonella secretes two additional virulence factors (SifA and SifB) belonging to the WxxxE family of effectors described by Alto.^115-117 The WxxxE family which includes IpgB1 and IpgB2 from Shigella flexneri, MAP and EspM2 from Escherichia coli, and EspT from Citrobacter rhodentium (Table 2) have been described as RhoGTPase mimics. Indeed, IpgB2 has been found to interact and activate RhoA effectors such as Rock and mDia, and thus promoting actin filament polymerization.¹¹⁵ However, IpgB2 and IpgB1 previously identified as RhoA and RhoG mimic, respectively,^115,118 have been found to activate RhoGTPases through a GEF activity.^118-120 SifA and SifB contain the WxxxE motif, but also two three helix bundles forming a V-shapped structure which is highly similar to that of SopE, suggesting that they could are GEFs.¹¹⁶ However, no GEF activity of SifA toward RhoA has been evidenced.¹²¹ The precise mechanism of action of SifA and SifB leading to Rho or other RhoGTPase activation remains to be elucidated. SifA mediates the formation of filaments which maintain the integrity of vacuoles containing Salmoenlla and coordinates the equilibirum between kinesin and dynein in the control of membrane dynamics.^117,122 Thereby, SifA by controling the membrane integritiy of vacuoles containing Salmonella, plays a role in the virulence of this pathogen.

Table 2. Examples of bacterial effectors injected by type III or IV secretion system that modulate RhoGTPase signaling by mimicking regulators and mediate bacterial entry into target cells.

Bacterial effectors	Pathogen	Target	Mode of action	Main effects	Reference
Bacterial effectors activating Rho-GTPases
SopE, SopE2	Salmonella typhimurium	Rac, CDC42	GEF	Actin filament polymerization, membrane ruffling, bacterial cell invasion	101 , 185
SopB		Cdc42, RhoG	Inositol phosphatase VAV2 (Cdc42 GEF) activation SGEF (RhoG GEF) activation	Actin cytoskeleton rearrangement, membrane ruffling	103 , 111 , 112
SifA, SifB		RhoA	GEF?	Membrane integrity of Salmonella-containing vacuoles	116 , 117
IpgB1	Shigella flexneri	Rac, Cdc42	GEF	Formation of actin-rich membrane ruffles	118 , 186
IpgB2		RhoA, Rac, Cdc42	GEF	Formation of actin stress fibers	119 , 120
EspM1, EspM2, EspM3	Enteropathogenic , Enterohemorrhagic Escherichia coli Citrobacter rodentium	RhoA	GEF	Formation of actin stress fibers	121
EspT	Enteropathogenic Escherichia coli Citrobacter rodentium	Rac, Cdc42	GEF	Formation of actin-rich membrane ruffles	187 , 188
EspG1, EspG2	Enteropathogenic Escherichia coli	Arf RhoA	Arf GTPase inhibiton PAK activation GEF-H1	Microtubule depolymerization Disruption of tight junctions	189 , 190
Map	Enteropathogenic, Enterohemorrhagic Escherichia coli	Cdc42	GEF	Formation of actin-rich filopodia	115 , 191
BopE	Burkholderia pseudomallei	Rac, Cdc42	GEF	Formation of actin-rich membrane ruffles	192
Tarp	Chlamydia trachomatis	Rac, Cdc42?	Activation of Rac GEF (Vav2, Sos1/Abi1/Eps8)	Actin rich pedestal at the site of entry	129 , 130

Bacterial effectors inhbiting Rho-GTPases
SptP	Salmonella typhimurium	Rac, Cdc42	GAP	Reversion of SopE-induced actin filament polymerization, bacterial internalization	125
YopE	Yersinia pseudotuberculosis Yersinia enterocolitica Yersinia pestis	Rac, RhoG, RhoA, Cdc42	GAP	Actin filament depolymerization Ant-iphagoctytosis	193 - 195
ExoS, ExoT	Pseudomonas aeruginosa	Rho, Rac, Cdc42	GAP	Actin filament depolymerization, anti-phagocytosis	149 , 196
EspH	Enteropathogenic, enterohemorrhagic Escherichia coli	RhoGEFs	Inhibition of eukaryotic RhoGEFs	Actin filament depolymerization	128
AexT	Aeromonas salmonicida	Rho, Rac, Cdc42	GAP	Actin filament depolymerization	147
YopO/YpkA	Yersinia pseudotuberculosis Yersinia enterocolitica Yersinia pestis	Rho, Rac	GDI	Actin filament depolymerization Anti-phagocytosis	152
CT166	Chlamydia trachomatis	Rac	Rac glucosylation	Actin filament depolymerization Restoration of the actin architecture	132

In addition, activation of Cdc42 and Rac by SopE leads to stimulation of p21-activated kinase (PAK) and subsequent activation of JNK, the MAP kinase pathways and a number of transcriptional factors, as well as caspase-1 activation resulting in secretion of proinflammatory cytokines.^106,123 Indeed SopE promotes a host inflammatory response which changes the microbiota composition allowing intestinal colonization by the pathogen Salmonella.¹²⁴ Therefore, SopE in involved in cellular invasion of Salmonella but also in tissue dissemination.

After entry into cell, Salmonella restores the original cytoskeleton architecture by injecting another type of virulence factor (SptP) which inactivates Rac and Cdc42 and thus antagonizes the effects of SpoE/E2 and SopB. SptP contains a N-terminal domain which mediates GAP activity toward Rac and Cdc42, and a C-terminal tyrosine phosphatase domain. Despite the absence of structural homology with eukaryotic GAPs, SptP retains a similar mechanism of stimulation of the GTP hydrolysis reaction via an Arg residue, called Arg finger, supporting a convergent evolution between bacterial and eukaryotic GAPs.¹²⁵ Thereby, the remodeling of the local actin cytoskeleton mediated by SopE/E2 and SopB allowing the uptake of Salmonella is transient, since this effect is reverted by SptP. A temporal coordination between the bacterial GEFs and GAP is required to induce an efficient bacterial internalization into cells. Indeed, SopE and SptP are equally delivered during the initial step of infection. But SopE is polyubiquitinated and rapidly degraded by the proteasome, whereas SptP is more stable allowing a longer activity and thus a restoration of the actin cytoskeleton.¹²⁶

Enteropathogenic and enterohemorrhagic E. coli use a different strategy to coordinate RhoGTPase activity by injecting into cells an inhibitor of eukaryotic Rho-GEFs (EspH) and bacterial GEFs (EspT, EspM2, Map). EspH binds to the DH-PH domain of eukaryotic RhoGEFs thus preventing their interaction with RhoGTPases. Thereby, EspH inactivates RhoGTPases and induces focal adhesion disassembly, actin filament depolymerization, cell detachment and cytotoxicity similarly to TcdB. However, EspH is not active on bacterial GEFs, which are structurally different from eukaryotic RhoGEFs.^127,128 These invasive pathogens trigger a subtle control of RhoGTPase activity by inhibiting eukaryotic RhoGEFs and by translocating bacterial GEFs insensitive to EspH, which modulates endogenous RhoGTPase for their own benefit.

Manipulation of the RhoGTPase signaling by bacterial virulence factors which are injected trough type III secretion system and which mimic eukaryotic Rho-GEFs is used by various enteroinvasive bacteria such as Salmonella, Shigella, enteropathogenic and enterohemorrhagic E. coli, and also other invasive bacteria like Burkholderia pseudomallei (Table 2). Activation of Rac and Cdc42 and subsequent formation of membrane ruffles are key features induced by the bacterial GEFs allowing invasion into non-phagocytic cells and intracellular survival. Indeed, Shigella IpgB1, E. coli EspT, and B. pseudomallei BopE are GEFs having an equivalent role than that of SopE in the local actin remodeling and subsequent bacterial uptake and internalization into vacuoles (Table 2). Manipulation of RhoGTPases by bacterial GEFs support additional functions in the invasion process including stimulation of host immune response via activation of the MAPK and/or NF-κB signaling pathways, and disorganization of the intercellular junctions leading to increased epithelial barrier permeability (reviewed in¹⁰⁸).

Chlamydia trachomatis, an obligate intracellular pathogen, also uses a balanced control of the actin cytoskeleton through manipulation of RhoGTPase signaling to invade host cells. The elementary bodies, which are the extracellular infectious forms of C. trachomatis, attach to cells and induce the formation of an actin-rich pedestal at the site of entry, which promotes the internalization into endocytic vesicles. The remodeling of the local actin cytoskeleton is Rac-dependent and is mediated by a type III secretion system factor called, TARP. The role of TARP is not yet fully understood. Phosphorylated TARP interacts with GEFs (Vav2, and Sos1/Eps8/Abi1), which activate Rac. TARP exhibits no GEF activity but is a scaffolding protein, which also recruits the PI3-kinase, thus leading to the activation of endogenous Rac GEFs and subsequently of WAVE2, Abi1, and Arp2/3 complex.^129-131 Actin filament polymerization is counterbalanced by an additional factor, CT166, secreted inside cells. CT166 contains in its N-terminal part the conserved enzymatic DxD motif of LCGTs and inactivates Rac1 but not RhoA thus leading to actin filament depolymerization.¹³² Thereby, Chlamydia controls the host actin cytoskeleton to mediate its internalization by activating Rac through TARP, which stimulates eukaryotic Rac GEFs and by reversing the reorganization of actin filaments by inactivating directly Rac via glucosylation.

Manipulation of RhoGTPase Signaling by Bacterial Adhesins

In addition to specific factors interacting with RhoGTPase signaling, numerous bacterial pathogens use membrane proteins or adhesins to trigger signaling pathways which facilitate bacterial entry into cell and which directly or indirectly involve RhoGTPase activation. Bacterial pathogens have adopted complex strategies to interact with the host. Some bacterial factors induce unique effect on cells, but many other manipulate complex networks of signaling pathways. Examples are shown in Table 3 to illustrate the activation of RhoGTPase signaling by surface bacterial components involved in bacterial entry into cells. This is the case of CadF, a Campylobacter jejuni outer membrane protein, which binds to fibronectin, a major extracellular matrix component. CadF as well as C. jejuni flagella, which are constituted of two major proteins, FlaA and FlaB, mediate membrane ruffling, filopodia formation, and engulfment of bacteria in a RhoGTPase, mainly Cdc42 and Rac, activation-dependent manner. The proposed signaling pathway leading to Cdc42 activation and bacterial invasion includes clustering and activation of integrin β1 receptor, phosphorylation of FAK and Src, activation of EGF and PDGF receptors, PI3-kinase, Vav2 and Cdc42 activation.^133,134 FAK can also stimulates DOCK180/TIAM-1, which are GEFs activating Rac1.¹³⁵

Table 3. Examples of bacterial adhesins and membrane proteins which trigger signaling pathways including activation of RhoGTPase and mediate pathogen internalization into cell.

Membrane protein	Pathogen	Activated RhoGTPase	Signaling pathway	Effects	Reference
CadF, Flagella (FlaA, FlaB)	Campylobacter jejuni	Cdc42 Rac	integrin β1 receptor, FAK and Src, PI3-kinase, VAV2 integrin β1 receptor, FAK, DOCK180/TIAM-1	Actin cytoskeleton rearrangement, membrane ruffling,	157 135
Fibronectin binding protein-A	Staphylococcus aureus	Rho, Rac, Cdc42	Integrin signaling	Actin cytoskeleton rearrangement, bacterial invasion	136
Pneumococcal surface protein C (PspC)	Streptococcus pneumoniae	Cdc42	plgR, PI3-kinase, Akt	Actin cytoskeleton rearrangement, bacterial invasion	137
Invasin	Yersinia pseudotuberculosis	Rac1	Integrin signaling N-WASP, Arp2/3	Bacterial uptake into epithelial cells	138
?	Brucela abortus	Rho, Rac, Cdc42	?	Bacterial uptake	139
?	Bartonella bacillifomis	Rho, Rac, Cdc42	PAK MAPK, SAPK/JNK	Formation of lamellipodia and filopodia Bacterial uptake	140

Numerous other invasive bacterial pathogens use the RhoGTPase pathway to induce local actin cytoskeleton rearrangement and their subsequent entry into cells. Indeed, RhoGTPase inhibition significantly prevents the bacterial uptake. However, the molecules, membrane proteins or translocated effectors, which activate the RhoGTPase signaling pathway as well as the mechanism of RhoGTPase activation remain to be defined for many pathogens. For example, the S. aureus adhesin called fibronectin binding protein-A induces β1 integrin clustering and subsequent signaling leading to reorganization of the actin cytoskeleton and staphylococci invasion in a Src tyrosine kinase and RhoGTPase activation-dependent manner including N-WASP and Arp3/3 complex (Table 3).¹³⁶ The pneumococcal surface protein C (PspC), the major adhesin of Streptococcus pneumoniae binds to polymeric immunoglobulin receptor (pIgR) and promotes bacterial invasion via Cdc42, PI3 kinase and Akt activation.¹³⁷ Activation of β1 integrin is also achieved by invasin, an outer membrane protein of Y. pseudotuberculosis, leading to Rac activation and actin cytoskeleton remodeling via N-WASP and Arp2/3 complex.¹³⁸ Brucella abortus and Bartonella bacilliformis are intracellular pathogens which also enter cells in a RhoGTPase dependent-manner (Table 3), but the molecules that trigger this cell signaling have not yet been identified.^139,140 However, the actin cytoskeleton remodeling by these microorganisms is triggered upon contact between bacteria and target cells suggesting that recognition and activation of a cell surface receptor is probably a required early event in the stimulation of the RhoGTPase cascade.

Bacterial Pathogens Manipulate RhoGTPase Signaling via Injected Virulence Factors to Avoid Phagocytosis

Several pathogens replicate in phagocytic cells, such as macrophages, and thus invade the organism or proliferate in local necrotic tissues avoiding phagocytosis. Bacteria do not require a specific equipment to enter phagocytic cells, but in contrast they have an absolute necessity to escape or survive in hostile cells. For this purpose, they inject into phagocytic cells an array of virulence factors which block the phagocytosis or prevent their killing in phagocytic cells. Numerous virulence factors which induce an anti-phagocytic effect, act by modulating the RhoGTPase signaling. RhoGTPases are key players in phagocytosis,^37,141 and down modulation of RhoGTPases results in an efficient strategy to impair the phagocytosis. Bacterial virulence factors modulate the RhoGTPase signaling cascade either by mimicking a GAP/GDI activity or by inducing a biochemical modification which blocks the RhoGTPases in their inactive form (Table 4).

Table 4. Examples of bacterial effectors injected by type III or IV secretion system that modulate RhoGTPase signaling by biochemical modification and mediate bacterial entry into target cells or anti-phagocytosis.

Bacterial effector	Pathogen	Target	Mode of action	Effects	Reference
Bacterial effectors activating Rho-GTPases
TccC5	Photorhabdus luminescens	Rho, Rac, Cdc42	ADP-ribosylation at Gln61/Gln63	Aggregation of actin filaments Anti-phagocytosis	157

Bacterial effectors inhbiting Rho-GTPases
YopT	Yersinia enterocolitica	Rho, Rac, Cdc42	Proteolysis at Cys 200 (RhoA)	Actin filament depolymerization Anti-phagocytosis	197
LopT	Photorhabdus luminescens	Rho, Rac	Proteolysis at Cys200 (RhoA)	Anti-phagocytosis	156
VopS	Vibrio parahemolyticus	Rho, Rac, Cdc42	AMPylation at Thr35/Thr37	Actin filament depolymerization Intestinal invasion	158
IbpA	Histophilus somni	Rho, Rac, Cdc42	AMPylation at Tyr32/Tyr34	Actin filament depolymerization Bacterial dissemination, escape of immune defense	160 , 161
PfhB2	Pasteurella multocida	Rho, Rac, Cdc42	AMPylation at Tyr32/Tyr34	Actin filament depolymerization	162

Mimicking host regulatory RhoGTPases

YopE from Yersinia is secreted by type III secretion system and exerts a GAP activity toward Rac1, RhoG, RhoA and Cdc42 resulting in actin filament depolymerization and inhibition of phagocytosis. YopE shares sequence and structure similarities with the GAP N-terminal domain of SptP and Pseudomonas ExoS (see below), but not with eukaryotic GAPs. However, YopE retains en Arg finger motif like eukaryotic GAPs, which is critical for the activity.¹⁴² YopE contains a N-terminal domain (amino acids 54–75), termed membrane localization domain, which mediates the binding to membrane close to its host targets. The subcellular localization of YopE at the membrane seems to drive its specificity toward target RhoGTPases.¹⁴³ Indeed, Rac is more rapidly inactivated than RhoA in Yersinia-infected cells.¹⁴⁴ A key central player in Yersinia invasion has been addressed to RhoG, which is an upstream regulator of RhoGTPases. Indeed, active RhoG stimulates the GEF activity of the complex Elmo.Dock180 toward Rac1. Yersinia adherence to cell promotes RhoG activation through the bacterial surface protein, Invasin, and clustering of β1 integrin, thus facilitating cell invasion. Then, RhoG and subsequently Rac1 are downregulated by YopE GAP activity.¹⁴⁵ YopE is one of the most efficient virulence factors produced by Yersinia pestis, Yersinia pseudotuberculosis and Yersinia enterocolitica which are involved in the impairment of phagocytosis by macrophages and neutrophils allowing to these pathogens to invade the host tissues.¹⁴⁶ YopE which inhibits immune defenses, is a major determinant of Yersinia virulence.

Other pathogens, like Pseudomonas and Aeromonas, exploit the RhoGTPase signaling by injecting effectors, which mimic eukaryotic RhoGAPs to escape phagocytic cells (Table 2). Interestingly, Pseudomonas ExoS and ExoT as well as Aeromonas AexT have a double activity contributing to their antiphagocytic effects. These effectors contain a N-terminal GAP domain and a C-terminal ADP-ribosylating domain. ExoS ADP-ribosylates numerous substrates including Ras and Rab proteins, whereas ExoT modifies CRK proteins and AexT targets actin monomers.^147-150 ADPribosylating activity of ExoS toward Rab5 plays an essential role in phagocytosis inhibition.¹⁵¹ ADPribosylation of CRK proteins prevents their interaction with focal adhesion proteins and/or with DOCK180 which is a GEF for Rac, thus inhibiting Rac-dependent phagocytosis.¹⁵⁰

Instead to mimic GAP function, YopO from Y. enterocolitica and YpkA from Y. pseudotubercuosis and Yersinia pestis contain a C-terminal domain, which interacts with RhoA and Rac to mimic RhoGDI thereby sequestering the RhoGTPases in their inactive form. Indeed, The YpkA C-terminal domain interacts with Rac adopting a structural conformation similar to that of RhoGDI and blocks nucleotide exchange and subsequent activation.¹⁵² The N-terminal domain of YpkA/YopO is a kinase domain which is activated by monomeric actin and which phosphorylates Gαq.¹⁵³ YpkA induces actin filament disruption and inhibits Rac-dependent Fcγ receptor- but not complement receptor-3-dependent phagocytosis through its RhoGDI-like activity and has a critical role in Yersinia virulence.^144,152,154

Biochemical modification of RhoGTPases

Several invasive pathogens use virulence factors which enzymatically modify RhoGTPases and block them in their inactive form (Table 4). Thereby, Yersinia produce a cysteine protease, YopT, which is an additional virulence factor involved in the inhibition of phagocytosis. YopT is structurally related to the papain-like cysteine proteases and removes specifically the C-terminal cysteine of RhoGTPases to which the geranyl-geranyl group is attached. The cleavage of the isoprenoid moiety results in the RhoGTPase release from the membrane and from GDI leading to their inactivation. YopT recognizes Rho, Rac and Cdc42 in their isoprenylated form but independently of their GDP- or GTP- bound state. However, RhoA seems to be the preferred substrate in living cells. YopT disrupts the actin cytoskeleton and prevents the formation of the phagocytic cups thus inhibiting phagocytosis of opsonized and non-opsonized Yersinia by macrophages and neutrophils.^144,155 However, Yersiniae use a combination of Yops to interact with the immune cells and the role of YopT alone during infection remains to be defined.¹⁴⁴

Other pathogens also use a cysteine protease specific of RhoGTPases as virulence factor. Thereby, the entomopathogenic bacterium, Photorhabdus luminescens, delivers the type III secretion effector, LopT, which induces proteolytic removing of RhoA and Rac from eukaryotic cell membrane similarly to YopT. LopT plays an essential role in Photorhabdus virulence by preventing phagocytosis from insect macrophages.¹⁵⁶ P. luminescens produces additional virulence factors which interfere with the actin cytoskeleton like TccC5 and TccC3. TccC5 ADP-ribosylates RhoA at Gln63 as well as Rac and Cdc42 at Gln61, thus leading to the loss of the GTPase activity and subsequent RhoGTPase activation similarly to the deamidation of Gln61/63 by CNF.¹⁵⁷ TccC3 is also an ADP-ribosyltransferase, but which modifies actin monomers at Thr148 preventing the interaction with thymosin β4, a blocker of actin-actin association. The combined effects of TccC5 and TccC3 results in increased actin polymerization but leading to a disorganization of the actin cytoskeleton and impairment of phagocytosis.¹⁵⁷ Indeed, Tcc5 and TccC3 induce aggregation of actin filaments which are not assembled in long and continuous stress fibers but form patches of short filaments. These opposite effects on actin polymerization support that an appropriate balance between actin monomers and filaments is required for a properly organized and functional actin cytoskeleton. P. luminescens uses a set of toxins which target the actin cytoskeleton but which exploit different modes of action including activation and inactivation of RhoGTPases to fully prevent the phagocytosis.

An additional biochemical mechanism of RhoGTPase inhibition consists of modification of a residue of switch I by adenylation (or AMPylation), which prevents interaction with downstream effectors by steric hindrance. Indeed, VopS, a type III secretion system effector produced by V. paraheamolyticus, catalyzes the adenylation of Rho, Rac, and Cdc42 at Thr35/37. Vops and kinases use the same cosubstrate which is ATP. However, VopS transfers the AMP part to an acceptor amino acid, whereas kinases use the γ-phosphate to modify the target residue.¹⁵⁸ VopS induces a disruption of actin filament in HeLa cells by inactivating RhoGTPases and contributes with additional effectors like VopQ, a PI3K-independent autophagy inducer, to the suppression of the NLRC4-dependent inflammasome response. Indeed, VopS via inactivation of Cdc42 inhibits NLRC4-dependent caspase-1 activation.^158,159 Escape of intestinal NLRC4 expressing macrophages confers an advantage to the invading V. parahemolyticus. VopS contains the adenylation core motif HPExxGNGR, which is called Fic (filamentous induced by c-AMP) domain. This domain is conserved in other bacterial virulence factors and in one human protein.¹⁶⁰

IbpA (immunoglobulin-binding protein A) is secreted by a two-partner system by Histophilus somni which is responsible for pneumonia and septicemia in cattle. IbpA secreted at the bacterial surface is then internalized into cells by a yet undetermined pathway.¹⁶¹ IbpA catalyzes adenylation of Rho, Rac and Cdc42, but at a distinct residue of switch I than VopS, Tyr32/34 instead of Thr35/37.¹⁶² IbpA is critical in H. somnus virulence and induces HeLa cell rounding and actin cytoskeleton disruption. It is proposed that cell retraction leads to an increased paracellular permeability of pulmonary alveolar barrier allowing the transmigration of bacteria through the alveolar epithelial cell monolayers and subsequent septicemia.¹⁶¹ The role of IbpA in pathogenesis is not fully understood but seems important for the bacterial escape from the immune defense.¹⁶⁰ It is noteworthy that the adenylating factors including VopS and IbpA recognize equally both GTP- and GDP-bound RhoGTPase forms, in contrast to the other bacterial virulence factors which preferentially interact with RhoGTPases in their GDP state.¹⁶² In addition PfhB2 (Pasteurella filamentous hemagglutinin B2) produced by Pasteurella multocida also modifies RhoGTPases similarly to IbpA.¹⁶² Although IbpA and PfhB2 are not significantly related at the amino acid sequence level, they share similar structure. The role of PfhB in P. multocida virulence is not yet well defined. P. multocida genome contains two genes encoding PfhB1 and PfhB2, which show a high homology with Bordetella pertussis filamentous hemagglutinin (FhaB), and which retain a C-terminal domain similar to that of the serum resistance protein p76 of H. somnus. FhaB mediates bacterial adherence to host cells, and p76 induces bacterial resistance to opsonization.¹⁶³ PfhB1 and PfhB2 could have similar functions in P. multocida and PfhB2 possibly triggers additional effects linked to its ability in modifying RhoGTPases.

Activation of RhoGTPases and Innate Immunity

Host can detect the presence of pathogens and can develop an appropriate defense response trough the innate immunity. RhoGTPases are involved in various signaling pathways controlling innate immune functions such as Toll-like receptor (TLR) signaling, leukocyte chemotaxis and migration, phagocytosis and NADPH oxidase activation.¹⁶⁴ Among the various innate immunity molecules, NODs (nucleotide-binding and oligomerization domain) are intracellular receptors able to detect invasive bacteria by sensing cytosolic microbial products. NOD1 is expressed by many cell types including intestinal epithelial cells whereas NOD2 is restricted to monocytes, macrophages, dendritic cells and Paneth cells.¹⁶⁵ Peptidoglycan fragments in the cytosol can activate NOD1 or NOD2 which forms a complex with receptor-interacting protein (RIP2) and other proteins called nodosome. This later activates the NF-κB and mitogen-activated protein (MAP) kinase pathways leading to proinflammatory and antimicrobial responses. Recently, RhoGTPases have been evidenced as components of NOD1 nodosomes and to control their activity. Indeed, activation of Rac and Cdc42 by bacterial virulence factors like SopE triggers the NOD1 signaling pathways and induces an inflammatory response.^166,167 Similarly, CNF1 which activates Rho, Rac and Cdc42, triggers host immune response in a Rac2 activation-dependent manner and subsequent Rip (receptor-interacting protein kinase) pathway.¹⁶⁸ Inversely, inactivation of RhoGTPases by toxins or virulence factors is probably detected by the NOD signaling pathway. Thereby, changes in RhoGTPase activation state by bacterial virulence factors or toxins are sensed by NOD1, which elicits an adaptive response toward pathogens, thus conferring to host cell an efficient mechanism monitoring the presence of invasive bacteria or pathogens able to attack an epithelial barrier through manipulation of RhoGTPases.

Concluding Remarks

Modulation of the actin cytoskeleton via manipulation of RhoGTPase signaling is a common mechanism used by numerous pathogens to invade and disseminate in host organism. It is noteworthy that regarding the complex pathway of regulation of actin polymerization, bacterial pathogens mainly target RhoGTPases and to a lesser extent actin monomers. Thereby, RhoGTPases are the preferred targets of toxins and virulence factors which modulate the actin cytoskeleton. Bacterial pathogens have developed numerous strategies to interfere with RhoGTPases including direct interaction and subsequent biochemical modification, or mimicry of RhoGTPases or proteins involved in their regulation. Targeting RhoGTPases instead of actin molecules offer two main advantages. First, interacting on an upstream step of the regulation cascade of actin polymerization affords an amplified effect. Second, targeting distinct sets of RhoGTPases allows a focused activity on only certain pools of actin like localized cortical actin thus leading to specific modification of actin structure such as phagocytic cups. Based on the mode of secretion of virulence factors, two main mechanism of pathogenicity can be distinguished: secretion of toxins or injection of virulence factors. Toxins are exported and diffuse in the external bacterial medium to all the surrounding cells. Toxins enter target cells using eukaryotic endocytosis machinery and induce biochemical modification of RhoGTPases via their enzymatic activity. This often results in a brutal and global effect of toxins on the actin cytoskeleton leading to severe lesions like destruction of epithelial and endothelial barriers (tissue necrosis, hemorrhage). In contrast, the virulence factors are selectively injected into the target cells via specialized apparatus (type III or IV secretion systems) and preferentially control the RhoGTPase signaling cycle by regulatory protein mimicry avoiding cell destruction but hijacking cell function for the own pathogen benefit (bacterial internalization, anti-phagocytosis, escape of immune defense). Invasive bacteria can also manipulate the RhoGTPase signaling through recognition and activation of cell surface receptor(s) (Fig. 4). The distinct modes of interactions of bacterial effectors with RhoGTPases or eukaryotic protein mimicry by bacteria likely represent different modes of adaptation of pathogens with their respective host. However, the mechanisms of evolution resulting from the co-habitation between bacteria and eukaryotes remain enigmatic. If RhoGTPases are a preferred target for many invasive an toxigenic bacteria, these molecules also constitutes a sentinel sensed by the host cell to monitor the presence of pathogens.

Figure 4. Main effects of manipulation of RhoGTPase signaling by toxigenic and invasive bacteria on pathogen dissemination. Toxigenic bacteria secrete toxins, which upon biochemical RhoGTPase modification alter the actin cytoskeleton and subsequently epithelial and endothelial barriers opening a breach for bacterial dissemination. Invasive bacteria modulate the RhoGTPase signaling mainly by injecting virulence factors which mimic regulatory proteins of the RhoGTPase cycle or by triggering a cell surface receptor mediated pathway which activates RhoGTpases leading to bacterial invasion in epithelial cells or to prevention of phagocytosis by professional phagocytic cells. Activation of RhoGTPase state is sensed by the NOD1 pathway of innate immunity which controls a proinflammatory defense response.