Cross-Kingdom Activation of Vibrio Toxins by ADP-Ribosylation Factor Family GTPases

Pathogenic Vibrio species use many different approaches to subvert, attack, and undermine the host response. The toxins they produce are often responsible for the devastating effects associated with their diseases. These toxins target a variety of host proteins, which leads to deleterious effects, including dissolution of cell organelle integrity and inhibition of protein secretion. Becoming increasingly prevalent as cofactors for Vibrio toxins are proteins of the small GTPase families. ABSTRACT Pathogenic Vibrio species use many different approaches to subvert, attack, and undermine the host response. The toxins they produce are often responsible for the devastating effects associated with their diseases. These toxins target a variety of host proteins, which leads to deleterious effects, including dissolution of cell organelle integrity and inhibition of protein secretion. Becoming increasingly prevalent as cofactors for Vibrio toxins are proteins of the small GTPase families. ADP-ribosylation factor small GTPases (ARFs) in particular are emerging as a common host cofactor necessary for full activation of Vibrio toxins. While ARFs are not the direct target of Vibrio cholerae cholera toxin (CT), ARF binding is required for its optimal activity as an ADP-ribosyltransferase. The makes caterpillars floppy (MCF)-like and the domain X (DmX) effectors of the Vibrio vulnificus multifunctional autoprocessing repeats-in-toxin (MARTX) toxin also both require ARFs to initiate autoprocessing and activation as independent effectors. ARFs are ubiquitously expressed in eukaryotes and are key regulators of many cellular processes, and as such they are ideal cofactors for Vibrio pathogens that infect many host species. In this review, we cover in detail the known Vibrio toxins that use ARFs as cross-kingdom activators to both stimulate and optimize their activity. We further discuss how these contrast to toxins and effectors from other bacterial species that coactivate, stimulate, or directly modify host ARFs as their mechanisms of action.

lipid droplet morphology and utilization (85). ARF1 further plays a critical role in facilitating vesicle fission, such that its activation is required to mediate actin assembly at the Golgi apparatus (46). Class II ARF4 and ARF5 also localize to the Golgi stacks, but they share only partially overlapping functions with class I ARFs in regulating its membrane trafficking (73).
In contrast to the other ARFs, ARF6 has no effect on processes at the Golgi apparatus but is instead active at the plasma membrane and regulates actin and endosomal membrane trafficking (86,87). It directs actin remodeling in circumstances where quick changes to cell morphology are necessary, modulating the formation of pseudopods, cell migration, phagocytosis, and membrane ruffling (88)(89)(90)(91). ARF6 also maintains the fluidity and integrity of adherens junctions in polarized epithelial cells (89). The role and localization of the different ARF isoforms is thus highly specific within cells.
Conformational shifts in ARF activation. ARF isoforms generally share a common structure with other Ras family small GTPases, in which they are composed of a typical G domain with two large extended loops designated switch I and switch II ( Fig. 2A). In addition, all ARFs have a signature N-terminal amphipathic helix required for membrane association (92). Stable membrane binding also requires the N-terminal glycine residue of ARFs to be cotranslationally myristoylated (92,93). Upon activation by GEFs and exchange of GDP for GTP, the conformation of the switch regions is remodeled. This conformation change drives their cellular membrane association and alters affinity of the ARFs for their AABPs. Thus, when AABPs are bound to ARFs, they are in close proximity to the lipid bilayer surface (13). ARF conformation and ability to interact with AABPs are dependent on its nucleotide-bound state.
Overall, ARFs are known to be critical to the control of many membrane-dependent cellular functions. Of particular note for this minireview on Vibrio toxins, many of the processes controlled by ARFs overlap cellular processes that are integral to defense of the host against bacterial infection. Thus, ARFs have been adopted not only as direct targets of bacterial toxins and effectors but also as activators of Vibrio toxins (Fig. 1).

CHOLERA TOXIN AND ITS ACTIVATION BY ARF GTPases
Structure and mechanism of action of cholera toxin. The primary virulence factor responsible for the serious diarrhea that typifies V. cholerae intestinal infections is cholera toxin (CT) (94). CT is an oligomeric 84-kDa toxin comprised of one CtxA subunit and five CtxB subunits. The CtxB pentamer binds to the host ganglioside receptor (95,96). CtxA is composed of A1 and A2 subunits joined by a single disulfide bond, where the A2 fragment functions as a tether to link the active A1 subunit to the CtxB pentamer (95,97). The holotoxin enters into the host cell and then traffics within vacuoles through the retrograde trafficking pathway. After release from the ER, reduction of the disulfide bond that joins A1 to A2 releases A1 into the cytosol, where it moves to the plasma membrane to access its cellular target (98-101) (Fig. 1).
Cell biological consequences of CT. G proteins are heterotrimeric proteins composed of ␣, ␤, and ␥ subunits. Similarly to the small GTPases, G proteins are active when GTP bound and inactive when GDP bound. The binding of GTP promotes dissociation of the ␣ subunit from the ␤ and ␥ subunits (102,103). Specifically, for the stimulatory G (G s ) protein, the disassociated ␣ subunit is freed to associate with and thereby activate the adenylyl cyclase complex that converts ATP to cyclic-AMP (cAMP) (104). When GTP on G s is hydrolyzed to GDP, the ␣, ␤, and ␥ subunits reassociate, and the adenylyl cyclase is downregulated (105).
In cells intoxicated with CT, the A1 subunit ADP-ribosyltransferase (ADPRT) activity catalyzes the transfer of ADP-ribose from NAD to arginine-201 of the ␣ subunit of the G s protein (6,106). As this arginine residue is essential for proper G s GTP hydrolysis, ADP-ribosylation by CT prevents deactivation of the GTPase and, in the presence of GTP, increases the activity of adenylyl cyclase (107)(108)(109)(110). CT thus promotes the disassociation of the ␣ subunit to constitutively activate the adenylyl cyclase complex and ultimately leads to increased cAMP levels in the cell (107). This activates a cAMPdependent protein kinase A, which phosphorylates ion channels and transporters, including the cystic fibrosis transmembrane conductance regulator, causing it to open and rapidly efflux chloride ions into the lumen (111,112) (Fig. 1). In an attempt to equilibrate the charge and osmolarity between the cell and the lumen, sodium ions and water are subsequently released into the lumen, causing the characteristic watery diarrhea associated with the disease (113).
Structural basis for CT activation by ARF6. Despite a clear biological activity in vivo, purified A1 subunit was initially found to have low ADPRT activity in vitro. However, its affinity for its substrates and its ADPRT activity were shown to be enhanced by ARFs from all three classes (114)(115)(116). Indeed, the ability to stimulate CT A1 ADPRT activity is where ARFs derived their unique name, prior to the discovery of their role at the Golgi apparatus and in vesicle trafficking (13).
This mechanism of activation of the A1 subunit has been worked out in structural and biochemical detail. After release of A1 from the A2 subunit, a conformational change in the A1 subunit occurs that allows for its interaction with ARFs. When bound to ARF6, residues in the activation loop of A1 form an amphipathic helix instead of the ordered coil structure and previously occluded residues are exposed (6) (Fig. 2B). The helical loop of A1 and the region directly preceding the active site open up revealing the ADPRT active site for NAD binding and to increase the affinity for the G s ␣ subunit (3,94).
The ARF6-GTP/A1 costructure shows ARF6 binds to A1 using 20 residues spanning both its switch I and switch II regions, as well as the interswitch region that is modestly conserved among the numerous ARFs (6). The A1 subunit alone is hydrophobic with low solubility (6), but it is soluble when bound to ARF6-GTP. The interface where the two proteins interact is mostly hydrophobic, where six water molecules are trapped at the periphery of their interface (6). Unlike ARF6-GTP, ARF6-GDP cannot bind A1. Structural analysis shows that when it is bound to GDP, switch I of ARF6 would sterically clash with the ␣7-␣8 loop and ␣8 helix of A1. The ␤4-␤5 and ␤6-␤7 loops, as well as the ␤7 strand of A1, would also be sterically hindered by switch II of ARF6. In addition, when GDP bound, the residues of ARF6 needed to make contact with A1 are buried (6).
The net effect of the cross-kingdom activation of CT A1 by ARFs only in the GTP-active form is that the toxin becomes active for ADRPT activity only upon close proximity with membranes. ARF6 in particular is present only in the plasma membrane and thus would be in close proximity of the target protein G s ␣ subunit. Thus, ARFs serve to help properly localize the toxin to its target and promote a favorable conformation for CT A1 to interact and modify its target. It is notable that this process is not limited to the Vibrio toxin. The heat-labile toxin of Escherichia coli, which is a toxin similar to CT that also ADP-ribosylates G s to stimulate adenylyl cyclase, is also activated by binding to ARFs, indicating that this cross-kingdom activation is shared across this family of toxins (117).

THE MARTX EFFECTOR MAKES CATERPILLARS FLOPPY (MCF) IS INDUCED TO AUTOPROCESS BY MULTIPLE ARF GTPases
MCF effectors of MARTX toxins. Multifunctional autoprocessing repeats-in-toxin (MARTX) toxins are secreted from bacteria as a long polypeptide composed of conserved N-terminal and C-terminal repeats, between which there is a cysteine protease domain (CPD) and two to five effector domains. Via a mechanism that has not been elucidated, the MARTX toxin repeats bind to eukaryotic cell membranes and organize to form a pore through which they transport the effector domains and the CPD. Once inside the cytosol, the CPD is activated by inositol hexakisphosphate to cleave the polypeptide of effectors, processing them into individual effector domains. Thus, MARTX toxins are effector delivery platforms for transfer of toxic effectors from bacteria to eukaryotic cells (118) (Fig. 1).
Among the nine different catalytic effectors that can be delivered by MARTX toxins (1,118), the makes caterpillars floppy (MCF)-like effector is the only effector present in all MARTX toxin variants of clinically relevant biotype I isolates of V. vulnificus (1). Furthermore, two nearly identical copies of MCF are found in a single MARTX toxin produced by the V. vulnificus biotype 2 strains that infect eels (4). This duplication and the conservation of MCF among strains highlights its importance during eukaryotic infection, as it has not been exchanged for other MARTX effectors by horizontal gene transfer, as commonly occurs for others (119).

MCF effector activation within host cells by autoprocessing and acetylation.
The MCF effector is 376 amino acids and is a member of the C58 family of cysteine peptidases (4). Experiments ectopically expressing the MCF effector domain along with part of the linker between it and the effector in front of it revealed that when it is recovered from cells, the effector is cleaved at its N terminus. Similarly to other members of the C58 peptidase family, MCF has a consensus catalytic triad formed by residues Cys148, His260, and Asp279 (4,120), and mutagenesis of these each of these residues resulted in defects in processing in vivo (4,120,121), demonstrating that MCF is an autoprocessing cysteine protease. Analysis of the cleaved sequence indicates that MCF recognizes the consensus sequence "X 1 -L-K-G-X 2 " (in which X 1 can be any small amino acid and X 2 can be any bulky hydrophobic residue) (120), with processing occurring between residues Lys15 and Gly16 (4,120). Subsequently to autocleavage, the new N-terminal glycine residue is acetylated (121).
As described above, MARTX effectors are typically released into the host cell cytosol as individual units by CPD processing. However, CPD is unable to cleave between MCF and the effector directly in front of it (120). This seemingly missed processing event produces "effector modules" that consist of two effectors tethered together instead of producing two discrete effectors. However, MCF is ultimately recovered from cells as an independent protein, indicating that MCF autoproteolytic activity is essential to complete its release from the MARTX toxin as a discrete toxic domain (120). Across the different MARTX toxin types of vibrios, MCF is typically found directly after either the alpha-beta hydrolase domain (ABH) or the actin cross-linking domain (ACD), such that the effector modules usually consist of ABH-MCF or ACD-MCF (120). These modules have been proposed to be important for coordinating MCF cellular localization and activation with the tethered effector (120) (Fig. 1).
Effects of MCF on host cell biology. Eukaryotic cells ectopically expressing autocleaved MCF become significantly rounded due to a loss of actin cytoskeleton structure (4). Single-amino-acid substitution of the catalytic cysteine completely disrupts this MCF-induced cell rounding, demonstrating that this cysteine is an essential residue for MCF cytotoxicity (4,122). Surprisingly, mutation of either His260 or Asp279, residues important for autoprocessing, did not eliminate cell rounding. In fact, 83 to 93% of cells ectopically expressing either mutant were still rounded. In contrast, alanine substitution at residues Arg147 or Asp149 adjacent to Cys148 completely abrogated the cellrounding phenotype, uncovering an Arg-Cys-Asp (RCD) tripeptide motif that is a signature of MCF and MCF-like effectors (4). The role of Arg147 and Asp149 in cell rounding and the discrepancy of why residues essential for autoprocessing are not essential for cell rounding remain to be resolved.
The effects of this toxin are not limited to cell rounding. MCF also decreases cell viability by disrupting the cellular metabolic activity of host cells and inhibiting proliferation by as much as 50% (4,122). In addition, MCF impairs cell growth and motility (122) and induces the intrinsic apoptotic pathway (122). The mitochondria of cells expressing MCF are shortened and fragmented, with their cristae condensed and disintegrated (121). This MCF-induced mitochondrial damage disrupts the mitochondrial membrane potential and stimulates the release of 40 to 90% of cytochrome c from the intermembrane space. The subsequent upregulation of the proapoptotic proteins Bax and Bak activates cleavage of caspases 9, 7, and 3, resulting in the processing of PARP-␥ and nuclear fragmentation (122).
In addition to damage to mitochondria, MCF causes dissolution of the Golgi apparatus (120,121). Transmission electron microscopy (TEM) shows that MCF induces extensive vesiculation of the Golgi stacks, causing a disintegration of the Golgi stacks and dispersion of Golgi-derived vesicles throughout the cytoplasm (121). Interestingly, catalytically inactive MCF colocalizes with the cis-Golgi marker GM130. However, if MCF is tethered to the ABH effector due to a defect in autoprocessing, MCF now localizes to the plasma membrane (120,121). This supports a model for stepwise MARTX effector processing, first by CPD and then by MCF, in proximity to the targeted subcellular compartment to release it from the tethered effector (120). It is important to note that ectopic expression of MCF alone does result in dissolution of the Golgi apparatus and cell rounding (4,120,121). This shows that, even without directed localization by tethering to either ABH or ACD, MCF is still cytotoxic and able to disrupt cellular metabolic activity, initiate apoptosis, and damage the Golgi apparatus and mitochondria.
MCF in bacterial pathogenesis. The impact of MCF on virulence is still under active investigation. There was no significant difference in the lethality of mice infected with V. vulnificus expressing a MARTX toxin with an internal deletion of mcf compared to a strain expressing the intact toxin (123). In contrast, a strain that delivers only MCF showed 40% increased survival when MCF was catalytically inactivated (120), although a different study showed no effect (123). This distinction could be due to interactions with other toxic factors, as the experiments were conducted in different genetic backgrounds and utilized different infectious routes. Therefore, the role that MCF plays during infection may depend on the route of infection and the context of expression with other effectors. Given the multidomain nature and complex delivery mechanisms of MCF from MARTX toxins, additional studies will be necessary to resolve the impact of MCF-induced Golgi stack dispersion and mitochondrial damage on pathogenesis.
MCF autoprocessing is stimulated by ARFs. Despite clear evidence of MCF autoprocessing within cells, MCF was initially found to not be processed in vitro unless combined with a host cell lysate, indicating that a cellular cofactor is essential for induction of autoprocessing (4). Indeed, autoprocessing can be stimulated in vitro by coincubating recombinant MCF with purified ARFs 1 to 6 (120,121). For stimulation, the first 17 residues of ARFs that confer the membrane targeting are not required (121). Isothermal titration calorimetry shows that catalytically inactive MCF binds tightly with ARF3, having a dissociation constant value of 3.79 nM (120). Recombinant MCF autoprocessing is stimulated in vitro equally well by ARF1 and ARF3, with cleavage occurring more rapidly and more completely when stimulated by GTP-bound ARF1 compared to GDP-bound ARF1 (120,121). These data suggest that MCF activation occurs predominantly by ARFs when they are membrane bound as opposed to soluble in the cytoplasm.
Although MCF binds to all ARF isoforms in vitro, one study found that MCF inactivated by alanine substitution of the catalytic cysteine (MCF CA ) consistently coimmunoprecipitates only with ARF1 when expressed in host cells (121). In contrast, another study showed that MCF inactivated by a serine substitution of the catalytic cysteine (MCF CS ) interacts with ARFs 1, 3, 4, 5, and 6 inside cells but not with other ARF-like proteins (120). This disparity is attributed to the difference in the catalytically inactive MCF mutant used between studies. While the MCF CA mutation may inhibit binding in vivo by ARFs other than ARF1, the MCF CS substitution does not completely abrogate MCF activity. In fact, the MCF CS mutant still causes 60% cell rounding, suggesting that it functions as a protein trap, keeping ARFs bound to it to disrupt its normal function, whereas MCF CA showed a more significant loss of function (4). However, this trapping activity could be only an artifact of overexpression of an inactive MCF protein in cells. In fact, expression of active MCF does not trap ARFs, but instead ARF is released after MCF is autoprocessed (4,121). Furthermore, similarly to CT, ARFs function only to stimulate the activation of MCF, but are not the direct target of the toxin. Binding to MCF does not result in ARF cleavage or posttranslational modification (121).
However, it has not been determined whether ARF is required for MCF to perform its toxic action on its unidentified direct target within a trimer complex from which ARF is released only after the target is modified. This is a possibility, particularly since a function of ARF binding may be to help stabilize the toxin to membranes, keeping it in proximity to its currently unknown target. In fact, MCF is capable of binding to membrane-bound phosphatidylinositol-5-phosphate only in its activated state (121). Additionally, the N-terminal acetylation of MCF may also help to stabilize its localization at the Golgi apparatus and nuclear membranes. Overall, similarly to CT, ARFs most likely function to activate MCF when in close proximity to its catalytic target at important organelle membranes, although the exact nature of the target and the context of interaction with the target remains to be worked out.
MCF conformational changes following activation. The structural mechanism for activation of MCF by association with ARFs has been investigated. The crystal structures of unprocessed MCF tethered to ABH and of the processed MCF CS in complex with ARF3 have both been solved (120) (Fig. 2C). MCF is not similar in structure to CT A1. It is composed of a helical bundle domain at its N terminus and has an ␣/␤-fold domain at its C terminus. The ABH-MCF structure did not capture the scissile bond in the MCF active site, suggesting that the site is closed in the absence of ARF activation.
Comparing the two structures reveals that MCF undergoes a significant conformational change upon binding ARF3. MCF binds ARF3 at its interswitch and switch II regions, aligning the ␣ helices in the C-terminal loop of MCF. Binding to ARF3 results in a shift of the N-terminal ␣ helices toward the ␣/␤-fold domain. Combined amino acid substitutions at residues G107L, R114L, Y178E, and E194G along the binding interface prevented MCF from binding with ARF3. In addition, the catalytic cysteine becomes exposed when bound to ARF3, and presumably for activation of its autocleavage activity (120). These structures demonstrate how, similarly to what was observed with CT, ARF binding to MCF leads to a conformational shift that activates MCF. However, unlike CT, the ultimate target of MCF that results in the Golgi stack dispersion and mitochondrial damage remains to be discovered.
This structural mechanism of induced autoprocessing stimulated by ARFs should apply to multiple other bacterial toxins of different species. MCF is a member of a broader class of cysteine protease effectors that share the signature RCD/Y catalytic motif. This includes a portion of Mcf1 and Mcf2 insecticidal toxins that cause limpness to the bodies of caterpillars infected with Photorhabdus luminescens (4). V. vulnificus MCF also has similarity to a domain of the Pseudomonas fluorescens FitD toxin and shares homology to MCFs in MARTX toxins of other species, including that from Aeromonas hydrophila (4,122). It remains to be tested if this cross-kingdom activation by ARFs is also conserved across this broader group of bacterial toxins, but the conservation of sequence strongly suggests this will be a shared mechanism across many bacterial species (4).

MARTX TOXIN DOMAIN X EFFECTOR IS ALSO ACTIVATED BY ARF GTPases
DmX is functionally similar to MCF. MARTX toxins are highly variable and regularly exchange effector domains by horizontal gene transfer (119). The domain X (DmX) effector is 22% identical with MCF, and amino acid sequence analysis also places DmX into the C58 family of cysteine peptidases (124). Across all bacterial species, 77% of MARTX toxins carry an effector that closely aligns with MCF, while 21% carry an effector closely aligned with DmX and less than 2% carry both (120). Remarkably, in MARTX toxins containing DmX, CPD also does not cleave directly between DmX and the effector in front of it, such that DmX is not fully released from the MARTX holotoxin (120). Similar to MCF, DmX is only fully released as an independent effector when stimulated to autoprocess at its N terminus within host cells (Fig. 1). Autoprocessing occurs at residues Val-Met-Lys, dependent on its catalytic cysteine residue (124). It is of note that, unlike MCF, DmX is not posttranslationally modified at its N terminus following the autocleavage event (121,124).
Cell biological studies reveal that DmX also shares cytopathic similarities to MCF. DmX causes cell rounding in various cell types. Furthermore, DmX causes Golgi stack dispersion and nuclear fragmentation. HEK 293 cells treated with V. vulnificus expressing DmX have significant defects in protein secretion (124). In contrast to the more dispersed localization of MCF, catalytically inactive DmX tightly localizes to the Golgi apparatus, suggesting that the function of DmX may be more specific for the Golgi apparatus compared to the broader effects of MCF, although more analysis will be necessary. No studies thus far have been conducted to test the functional role of DmX in pathogenesis.
DmX activity is stimulated by ARFs. As for CT and MCF, ARFs are the host factor required to stimulate the activation of DmX both in vitro and in host cells (124) (Fig. 1). For the V. vulnificus biotype 3 MARTX toxin, DmX is released from the adenylate cyclase ExoY effector domain immediately in front of it by autoprocessing when coincubated with purified ARF3 (120). DmX activation by ARFs does not require the first 17 residues at their N terminus or myristoylation. DmX autoprocessing is also preferentially stimulated by GTP-bound ARFs, suggesting that it interacts with ARFs in vivo after membrane association (124).
Cotransfection of DmX with ARFs, analogous to MCF, produces no visible changes to the size or abundance of ARF isoforms. Although less complete studies of DmX have been conducted thus far, ARFs are suggested to not be the target of this effector; instead, ARFs activate the effector, similarly to CT and MCF. As yet, there are no structures of ARFs bound to DmX, although residues along the ARF3 binding interface for MCF are among the regions conserved with DmX, suggesting that the activation by binding may be conserved as well (120).

ARF-BINDING EFFECTORS OF OTHER BACTERIA
Examples of bacterial effectors that interact with ARFs in different ways. Interaction with ARFs as cross-kingdom activators is emerging as a theme for Vibrio toxins that is shared with other bacterial species that translocate similar effectors. In addition, toxins and effectors of other bacterial species interact with ARFs in other ways, including directly altering ARF activity, acting as stimulators of ARF activity, and directly inactivating ARFs.
For example, the 27-kDa effector invasion plasmid antigen J (IpaJ) of the pathogen Shigella flexneri is a proteolytic enzyme that cleaves the N-myristoylated glycine from GTP-bound ARFs and closely related ARF-like proteins inside cells (125). IpaJ preferentially interacts with ARF1 and ARF5 at the Golgi apparatus to reduce the levels of these ARFs during infection. Without their myristoyl group, ARFs relocalize to the cytosol, ultimately leading to defects in cellular trafficking (125). Indeed, IpaJ causes severe Golgi stack fragmentation and decreases secretion, recycling, and endocytosis (126,127).
As another example, the type 4 secretion system effector RalF of Legionella pneumophila recruits ARF1 to promote vacuole formation. In fact, RalF acts as an ARF-GEF to activate ARFs 1, 3, 5, and 6, (128). The sustained activation presumably also overstimulates AABPs; however, the net impact of the RalF in pathogenesis is not yet known.
A final example is the EspG type 3 secretion effector of enteropathogenic and enterohemorrhagic E. coli strains (129). EspG binds ARF1-GTP and Rab1 in a trimer complex and prevents ARF GAPs from stimulating Rab1 GTP hydrolysis, locking the GTPase in its active state (130). By sustaining ARF1 activation, EspG promotes the recruitment of AABPs and downstream signaling and causes dysregulation of signaling pathways. Interestingly, ARF1 overstimulation by EspG also inhibits protein secretion (129)(130)(131). EspG binding of ARF6 can block ARF-GEF ARNO signaling to ARF1 (132). It is of note that both EspG functioning as a GEF to constitutively activate ARF1 and IpaJ inactivating ARFs result in Golgi apparatus dispersion.
ARF domains interacting with IpaJ, RalF, and EspG. CT A1 and MCF both interact with ARF along its switch I, interswitch, and switch II interface. Although a structure has not yet been determined for IpaJ, binding and mutagenesis experiments have indicated that residues Ile49 in the switch I region, Trp66 in the interswitch region, and Trp78 in the switch II region along the same interface are essential for IpaJ to interact with ARF1 (Fig. 2C) and further suggest that IpaJ must have a hydrophobic binding pocket to accommodate the myristoylated glycine (125). While the structure of RalF in complex with ARF has not been solved, the first 200 residues at the N terminus of RalF share 42% identity to the Sec7 domains of ARF GEFs and 70% similarity to ARNO (128,133). Modeling of the N terminus of RalF onto the structure of ARNO bound to ARF1 reveals that RalF could bind to switch I to facilitate GDP exchange for GTP (Fig. 2D). In addition to the Sec7 homology domain at the N terminus, RalF contains a discrete C-terminal domain connected by a long loop. The C-terminal domain has structural similarities to subunits important for the formation of vesicle coat complexes (133). It is composed of six helices, and six ␤ strands, forming a tight compact structure. The antiparallel ␤ strands and the four helices that surround them function as a cap, named the Sec7 capping domain (SCD), that occludes the active site of the Sec7 domain (133). The SCD extends into and interacts with the Sec7 domain via three glutamates, three lysines, and a tyrosine (133).
In contrast to the multiple contacts of the other effectors with ARF, the structure of EspG in complex with ARF1 shows that it binds solely to the switch I loop of ARF1. EspG comprises a central 6-strand sheet that is surrounded by a helical domain; its N terminus consists of a 4-strand ␤-sheet domain (134) (Fig. 2E). Furthermore, EspG has an arginine and glutamine finger motif that shares similarity to the motifs of the catalytic domains of endogenous host GAPs and is necessary for the activity of EspG to block binding of GAPs (134).
Overall, comparison of these other effectors with the Vibrio effectors shows that ARFs are critical to the cellular intoxication of cells by many pathogens. Furthermore, multiple faces of the GTPase are used for protein-protein interaction, demonstrating that multiple effectors have evolved very distinct strategies to utilize ARFs to promote bacterial pathogenesis.

ARF CONSERVATION AND IMPORTANCE IN CELLULAR FUNCTIONS PRODUCE AN IDEAL COFACTOR OR TARGET FOR BACTERIAL VIRULENCE FACTORS
A theme is emerging among bacterial toxins and effectors for either targeting or self-activation upon binding with ARFs. So why are effectors interacting with ARFs by so many distinct strategies, and specifically why the vibrios? One important reason is likely the ubiquitous nature of ARFs in hosts that are infected by vibrios. As vibrios exist in many environmental niches and interface with many different hosts, it is important to note there is a high degree of conservation of ARFs among eukaryotes. In fact, ARF homologs have been identified in Giardia lamblia, a protozoan parasite thought to be representative of the earliest diverging eukaryote lineage (20,21,135). The different classes of ARFs diverged early. While humans and other vertebrates express several isoforms from one class, other species such as flies and worms express only one member from each class (21). ARF homologs are found ubiquitously among a variety of species, including insects, fungi, mammals, trypanosomes, amphibians, yeast, and plants. Importantly, ARFs are expressed not only by humans but also by fish and prawns, all of which are infected by Vibrio species (11,136). In fact, ARF1 and ARF3 are identical in fish, amphibians, and humans (137). Considering their high degree of conservation, utilizing ARFs as activators of toxin function would be advantageous for bacterial survival in a broad range of hosts. Often, V. vulnificus infections arise from the consumption of contaminated seafood. In such a case the bacteria need to be able to seamlessly transition from one eukaryotic host to another for survival. Ensuring that its virulence factors are able to activate and function in both species is vital. Thus, ARFs can serve as important cues that can be used by toxins as signals for cross-kingdom activation in many species.
A second major reason why so many bacterial virulence factors have evolved to use ARFs as activators or targets is the critical function of ARFs in cells. ARFs are essential for maintaining cell survival, including regulating membrane stability, motility, and endocytic trafficking, and these processes often need to be disrupted during bacterial pathogenesis to prevent host immune response or antitoxin responses. Directly targeting ARF function can cause broad deleterious effects due to the extensive role in cell signaling and trafficking and secretion of proteins. Furthermore, depending on the isoform and function, they are found in many locations throughout the cell and can thus provide a means of regulation to activate toxin activity at the proper location or time depending on the needs of the bacteria.

CONCLUSION
A unique theme revealed in this minireview is that Vibrio species seem to have evolved at least three different toxin effectors that do not directly target host ARF function but which instead exploit the GTPases as cross-kingdom activators. This is in contrast to toxins and effectors from other bacteria that use effectors to directly modify or alter the activation state of ARFs. It is possible that vibrios have found a novel mechanism by which to use ARFs for directed and controlled activation, a strategy that works well across the many species that are hosts for vibrios. This strategy is well understood for the long-studied CT and the related E. coli heat-labile toxins, and future research to determine the targets of MCF and DmX and related toxins will certainly inform as to why they are specifically activated by ARFs as well. This research will then impact our understanding of many other similar effectors, as this mechanism of cross-kingdom toxin activation by ARFs is likely conserved across many vibrios and also other bacterial species.