Inositol Hexakisphosphate-Induced Autoprocessing of Large Bacterial Protein Toxins

Summary

Large bacterial protein toxins autotranslocate functional effector domains to the eukaryotic cell cytosol, resulting in alterations to cellular functions that ultimately benefit the infecting pathogen. Among these toxins, the clostridial glucosylating toxins (CGTs) produced by Gram-positive bacteria and the multifunctional-autoprocessing RTX (MARTX) toxins of Gram-negative bacteria have distinct mechanisms for effector translocation, but a shared mechanism of post-translocation autoprocessing that releases these functional domains from the large holotoxins. These toxins carry an embedded cysteine protease domain (CPD) that is activated for autoprocessing by binding inositol hexakisphosphate (InsP₆), a molecule found exclusively in eukaryotic cells. Thus, InsP₆-induced autoprocessing represents a unique mechanism for toxin effector delivery specifically within the target cell. This review summarizes recent studies of the structural and molecular events for activation of autoprocessing for both CGT and MARTX toxins, demonstrating both similar and potentially distinct aspects of autoprocessing among the toxins that utilize this method of activation and effector delivery.

Introduction

Pathogenic bacteria frequently export protein toxins that target eukaryotic intracellular proteins to alter host cell function to the benefit of the infectious pathogen. Different exported toxins employ distinct strategies for translocation of their cytopathic effectors from the bacterium into the host cell. These strategies include direct injection, such as occurs using Type III, Type IV [1], and likely also Type VI secretion [2]. By contrast, some toxins are secreted or released from the bacteria and then bind to host cell surface receptors via a binding (B) component. The B component itself or a separate translocation component then transfers the catalytic subunit or domain across the plasma or endosomal membrane into the cytosol. In some toxins, the B component is a protein subunit assembled with the effector (A) subunit within the bacteria before export (such as cholera toxin [3]), while for other toxins, the B and A subunits are exported separately and then assembled at the surface of the target cell (such as anthrax toxin [4]). Still other toxins are expressed as a single polypeptide that is nicked to separate the A and B domains by endogenous bacterial proteases (such as botulinum toxin [5]) or by host cell proteases during translocation (such as diphtheria toxin [6]). All of these processes succeed in delivering the smaller active effector domains or subunits into the host cell, where they can then access their intracellular protein targets.

Yet, questions have remained as to how single polypeptide toxins that range in size from 250 to 600 kDa deliver their effector domains to the eukaryotic cytosol. A shared strategy for activation of autocatalytic processing upon binding of the eukaryotic signal molecule inositol hexakisphosphate (InsP₆) has recently been characterized for these toxins. This process represents a novel strategy for toxin activation and subsequent delivery of effectors to target cells.

Overview of Clostridial Glucosylating Toxins

Clostridial glucosylating toxins (CGTs), also known as large clostridial cytotoxins, are structurally and functionally related toxins produced by different Clostridium sp. that range in size from 250 to 308 kDa and have sequence identity from 26% to 76% [7], [8]. Clostridium difficile Toxin A (TcdA) and Toxin B (TcdB) are the major virulence factors of clinically important antibiotic-associated diarrheal infections and pseudomembranous colitis [9]. Recent studies revealed that, while some C. difficile strains produce both toxins, only TcdB is essential for virulence [10]. Other significant members of the CGT family are Lethal Toxin from C. sordellii (TcsL) and the α-toxin from C. novyi (Tcnα). These clostridia are more rare causes of disease, but have been associated with particularly severe invasive infections, including gas gangrene and toxic shock following abortions or gynecological procedures [11]–[14].

The CGTs are organized in a multidomain structure [15], including a biologically active effector domain, a middle translocation domain, and a C-terminal receptor-binding domain [8] (Figure 1A). To enter eukaryotic target cells, the secreted CGTs bind to extracellular receptors and follow the “short trip model” of exotoxin uptake [16]. After receptor-mediated endocytosis, a vesicular H⁺-ATPase leads to acidification of the early endosomes, inducing a conformational change and an increase in hydrophobicity [17]. A small hydrophobic region of the protein is proposed to form a pore through which the N-terminus-localized glucosyltransferase (GT) domain is translocated into the cytosol [18]–[20]. Using UDP-glucose (UDP-N-acetylglucosamine for Tcnα) as a co-substrate, the GT monoglucosylates Rho family GTPases. The covalent modification occurs at a specific threonine residue found within the GTP binding pocket (Thr37 in Rho, Thr35 in Rac), thereby preventing activation of the small GTPases by exchange of GTP for GDP. Ultimately, the accumulation of inactive GTPases results in reorganization of the cytoskeleton and other morphological changes [21].

Overview of Multifunctional-Autoprocessing RTX Toxins

Multifunctional-autoprocessing RTX (MARTX) toxins are larger toxins that range in size from 350 to 600 kDa [22]. The MARTX toxin of Vibrio cholerae (MARTX_Vc) has been linked to virulence, in which the toxin functions during early infection to promote colonization, possibly by inactivating cellular innate immunity [23]–[25]. The MARTX toxins from both human [26]–[29] and aquatic animal [30] infectious Vibrio vulnificus strains (MARTX_Vv) and the fish pathogen Vibrio anguillarum (MARTX_Va) [31] have likewise been associated with virulence. In addition, putative MARTX toxins have been identified in at least 13 other sequenced Gram-negative bacteria, including Proteus, Aeromonas, Yersinia, and Photorhabdus spp. [22], [32]–[36], suggesting that additional pathogens require these toxins as virulence factors.

Similar to CGTs, the MARTX toxins are modular in structure, but are typified by the presence of extensive repeats at both the N- and C-termini [22], [37]. These repeats are postulated to form the translocation structure for transfer of centrally located effector domains to the cytosol [22] (Figure 1B). For MARTX_Vc, cytopathic effects occur in the presence of inhibitors of endocytosis [38]–[40], suggesting that endocytosis is not required for MARTX toxin entry, as it is for the CGTs.

Among the various MARTX toxins, a total of ten potential effectors have been identified, although each independent toxin has an assortment of only one to five [22]. The best-characterized MARTX effector is the actin crosslinking domain (ACD), which introduces a Glu270-Lys50 isopeptide linkage between actin monomers by a mechanism similar to that for glutamine synthetases [41], [42]. Another MARTX effector inactivates RhoGTPases by an unknown mechanism [43], although a recent bioinformatics study suggested this domain is a thiol protease [44]. The remaining eight potential effectors are domains of unknown function, even though two share sequence homology with Photorhabdus luminescens and Pasteurella multocida toxins and one is conserved with the alpha/beta hydrolase family of proteins [22].

Both CGT and MARTX Toxins Undergo Autoprocessing by a Conserved Cysteine Protease Domain

Early studies of the CGTs postulated that they would undergo enzymatic processing after exposure to low pH [17], [45]. Subsequent in vivo studies demonstrated that only the 60-kDa GT effector of TcdB is delivered to the cytosol, while a larger C-terminal fragment remains in the membrane fraction [20]. This processing of TcdB also occurred in vitro after residue Leu543, with a strict dependence upon addition of eukaryotic cell lysate [46] (Figure 1A). These studies of TcdB were initially interpreted as indicating processing by a host cell–encoded protease, similar to the mechanism for maturation of diphtheria toxin and other bacteria toxins [6]. However, protein-free extracts also stimulated TcdB processing, indicating autocataytic cleavage [47].

Similarly, early studies of MARTX_Vc postulated that the ACD would need to be released by proteolysis to access the entire actin pool [40]. In fact, it has been demonstrated that MARTX_Vc is autoprocessed at four positions located before and after its three effector domains, resulting in the release of these domains from the holotoxin [48], [49] (Figure 1B).

The autoprotease domain responsible for MARTX_Vc processing was recognized first as a 25-kDa domain within MARTX_Vc that affected cell viability when ectopically expressed in eukaryotic cells [50]. This cytotoxicity was disrupted by mutation of a single Cys or His residue, and analysis of protein expression patterns revealed that the mutant proteins were the predicted size, while the wild-type protein was cleaved of its N-terminus. Studies with recombinant protein confirmed autoprocessing after Leu3428 and, similar to TcdB, processing was strictly dependent upon addition of protein-free cell cytosol extracts [50]. Mutation of the critical Cys in the full-length toxin significantly reduced the ability of the toxin to induce actin crosslinking, confirming autoproteolysis due to this cysteine protease domain (CPD) enhanced toxin function [50].

After its discovery within MARTX_Vc, the CPD was found to be conserved within all MARTX toxins and also in all CGTs, with common alignment of the processing sites and the catalytic Cys and His residues [50], [51]. Furthermore, for TcdB, mutation of the analogous Cys and His residues reduced cytotoxicity of the full-length toxin, and disrupted processing of recombinant CPD protein as well [51], [52]. Thus, it was recognized that both the CGTs and MARTX toxins share a common mechanism for autocatalytic processing inducible by protein-free eukaryotic cell cytosol and that autoprocessing is essential for optimal cytotoxicity.

InsP₆: The Inducer of Autoprocessing

To identify the molecule in cell cytosol required to induce CPD for autoprocessing, cell extracts that stimulated processing of TcdB were fractionated and analysis of active fractions by mass spectrometry supplied spectra with similarities to inositol phosphates [47]. Incubation of TcdB with several inositol phosphates indicated that InsP₆ induced the most efficient autoprocessing activity [47]. Similar studies indicated that InsP₆ is likewise the most effective stimulator of autoprocessing of the MARTX_Vc CPD [53].

As a signal molecule for the eukaryotic intracellular environment, InsP₆ (also known as phytic acid) is an excellent selection for bacterial toxins, since the molecule is ubiquitous in eukaryotes but absent in bacteria. Furthermore, InsP₆ is the most abundant of the inositol phosphates, is maintained within cells at relatively constant levels of 10–60 µM, and is generally freely soluble in the cytoplasm [54]. Within mammalian cells, InsP₆ may function as a high concentration storage molecule for phosphate as it does in plant seeds or as a highly charged buffer for cation- or protein-dependent processes. More recent studies have linked InsP₆ to numerous cellular processes, including vesicle recycling, mRNA transport out of the nucleus, and as a co-factor for a DNA-dependent protein kinase [54]. Regardless of its normal function, InsP₆ is a molecule constantly present in high concentrations in the eukaryotic cytosol, assuring that induction of CGT and MARTX CPDs occurs only after completion of translocation of effector domains to the cytosol, regardless of whether translocation requires endocytosis or transfer directly across the plasma membrane.

InsP₆ Binding to the CPDs

While both CGT and MARTX CPDs are induced for autoprocessing by InsP₆, there are differences in these proteins revealed by crystallography and InsP₆ binding studies that suggest slightly different mechanisms of activation. Mutational studies [53], [55]–[57] and analysis of four independent crystal structures [48], [49], [56], [57] revealed that binding of InsP₆ to the CPD involves contact of the six negatively charged phosphate groups within a positively charged pocket of the CPD (Figure 2). The most significant binding contacts of MARTX_Vc with InsP₆ involve Lys3482, Lys3611, and Lys3623. Other Lys, Arg, and positively charged residues that form the binding pocket are not essential for binding, but do contribute to the high affinity of the MARTX_Vc CPD for InsP₆ [53], [56]. In TcdB, Lys600 (analogous to Lys3481 of MARTX_Vc) is likewise essential for binding of InsP₆ [55], while other conserved Lys and Arg residues also contact InsP₆ [55], [57]. Interestingly, overlay of the structures of the CPD from MARTX_Vc and TcdA CPDs revealed that the orientation of InsP₆ in the binding pocket is not conserved [57] (Figure 2), which is a surprise since amino acid sequence alignments show strong conservation of the Lys and Arg residues that form the binding pocket [55], [57]. However, this difference in the structure of the binding pocket may in part account for variances revealed in studies of InsP₆ binding and CPD activation for the different CPDs.

Intramolecular processing of purified MARTX_Vc CPD was found to be optimal in the range of 0.001–1 µM InsP₆ [53], [56], and binding of InsP₆ occurred with affinities ranging from 0.2 to 1.3 µM InsP₆ [48], [53], [56]. The ability of MARTX_Vc CPD to complete autoprocessing in vitro at concentrations below the dissociation constant reflects the recycling of InsP₆ released from processed CPD back to predominantly unprocessed protein [53], since processed MARTX_Vc CPD has a 500-fold reduced affinity for InsP₆ [48].

By contrast, activation of purified TcdB CPD autoprocessing requires 2 µM InsP₆ [51], a concentration near the determined K_d of 2.3 µM [55]. Furthermore, recombinant TcdB CPD that mimics protein processed after Leu543 binds InsP₆ with a similar affinity as full-length TcdB and unprocessed recombinant CPD [55], suggesting that TcdB CPD, unlike MARTX_Vc CPD, does not have an altered affinity for InsP₆ after processing.

Although the dissociation constant has not as yet been determined for TcdA, processing studies indicate that its ability to bind InsP₆ may be less efficient than TcdB. Whereas full-length TcdB is cleaved to completion at InsP₆ concentrations of 2–10 µM [47], [51], full-length TcdA does not autoprocess at 10 µM InsP₆ and concentrations up to 10 mM have been used for experimentation [47], [51]. However, recombinant TcdA CPD does autoprocess at concentrations as low as 5 µM [57], suggesting that the affinity of TcdA CPD for InsP₆ could be near to that of TcdB, but not accurately reflected in holotoxin cleavage assays.

Despite the apparent differences in the structure of the InsP₆ binding pocket and binding kinetics, the high affinity of the unprocessed CPDs for InsP₆ indicates that autoprocessing of MARTX_Vc, TcdB, and TcdA would all proceed efficiently at InsP₆ concentrations of 10–60 µM that are found in the eukaryotic cell cytosol [54]. Thus, all of the studied CGT and MARTX toxins would be autoprocessed and effectively deliver their effector domains within the in vivo environment.

Structural Arrangement of the CPD Catalytic Site

The CPD catalytic dyad is composed of one His and one Cys residue separated by ∼6 Å in both MARTX_Vc and TcdA structures [48], [56], [57] (Figure 3A–3D). The distance between the catalytic residues indicates that the Cys is not activated by protonation from His, but rather suggests that the Cys is substrate-activated by close alignment of the scissile bond, while the His functions solely to protonate the leaving group [48], [57].

In addition, Asp and Glu residues play an essential function in proteolysis. Mutation of TcdB Asp567 or TcdA Asp589 disrupted autoprocessing [51], [57] and eliminated cytopathic effects when added to HeLa cells [55]. By contrast, mutation of the analogous Asp3469 in the MARTX_Vc CPD did not affect autoprocessing [53]. Analysis of the structural models indicates that this conserved Asp residue functions in both proteins to properly orient the catalytic His residue [48], [57] (Figure 3A–3D). However, in the MARTX_Vc CPD structure, this function is shared with residue Glu3467, such that only a double Asp/Glu mutant is defective for function [53].

The closest known cysteine proteases that share this structural arrangement of the catalytic site are caspase-1 and gingipain R [58], [59]. Similar to these other proteases [60], [61], the CPDs are resistant to cysteine protease inhibitor E64 [50], but sensitive to N-ethylmaleimide, iodoacetamide, or chloromethyl ketones [48], [50], [51]. Thus, the CGT and MARTX toxin CPDs are grouped together with caspase-1 and gingipain R in the CD clan of cysteine proteases, but form a new family, the C80 family (http://merops.sanger.ac.uk, [62]). The CPD proteases have also been incorporated into a larger CPD_adh family of putative bacterial and eukaryotic peptidase that are proposed to share a similar fold in the catalytic site [63].

Structure-Based Modeling of InsP₆-Induced Activation of the CPDs

As the binding site for InsP₆ occurs on the opposite side of the protein from the catalytic site (Figure 3E–3H), it was recognized that there must be a mechanism to transduce the binding signal across the entire protein structure [56]. Translocation of effector domains of both CGTs and MARTX toxins is predicted to involve transit through a pore for entry into the cytosol, and thus the CPD is likely partially unfolded when it is first presented to the InsP₆-rich environment of the cytosol [15], [22]. Consistent with this model, apo-CPD in the absence of InsP₆ for both MARTX_Vc and TcdA is highly sensitive to proteolysis [48], [57]. Nuclear magnetic resonance (NMR) studies of the TcdA CPD indicated that the apo-protein is folded, but undergoes a significant conformational reorganization upon InsP₆ binding [57]. This finding is consistent with an observed high negative enthalpy and entropy and a 14°C thermal stability shift upon binding of InsP₆ to MARTX_Vc CPD, suggesting that this protein undergoes a major structural rearrangement that also stabilizes the protein structure [48].

Upon binding InsP₆, the structure of the MARTX_Vc CPD adopts the stable conformation amenable to X-ray crystallography (Figure 3E–3G). The CPD is composed of a seven-stranded β-sheet with two α-helices flanking the sheet and one capping the sheet [56]. An additional five β-strands at the C-terminus form a subdomain, known as the β-flap, that is loosely attached to the core of the protease [56]. By contrast, the TcdA CPD is larger and consists of a nine-stranded β-sheet flanked by five α-helices (Figure 3G) [57]. In both proteins, the N-terminus is an unstructured strand wrapped around the outside of the protein and attached to the core structure by embedding of large hydrophobic residues [48], [57]. At the extreme N-terminus of the CPD, the P1 Leu residue found immediately before the scissile bond is buried in a hydrophobic pocket (Figure 3E) [48]. Mutagenesis studies and structural analysis [48], [49] have demonstrated that Leu is the only residue that can be accommodated at this position. On either side of the Leu, any residue can occur but there is a preference for small residues, creating a consensus sequence of small-Leu-small ([48], [49] and Table 1). The G₁ β-strand (also known as β8) forms part of the hydrophobic pocket [56], and conserved Leu and Val residues on this strand make direct contact with the P1 Leu before the scissile bond [48]. This G₁ strand is antiparallel to the G₂ β-strand (also known as β9), a strand that contributes positively charged amino acids that make contact with InsP₆ [56]. The current model for activation of the MARTX_Vc CPD proposes that binding of InsP₆ alters the structure of this antiparallel β-hairpin, resulting in stabilization of the N-terminus within the hydrophobic pocket [48] and possibly reorientation of the catalytic Cys [56]. The net effect is to orient the scissile bond between the catalytic Cys and His residues, resulting in substrate-activated autoprocessing [48] (Figure 4). A similar mechanism has been proposed for the activation of TcdA CPD [57].

Multisite Processing of MARTX_Vc

The CGT toxins require only a single processing event to release their GT effector domain (Figure 1A), and this may account for why activation apparently requires a structural change between two apparently stable conformations at approximately 5 µM InsP₆ [57]. By contrast, MARTX toxins must undergo processing at multiple sites to release each of the effectors independently [48], [49] (Figure 1B). A simplistic model of multisite processing would predict that activation of the CPD by binding of InsP₆ results in transition to a constitutive “on” conformation after which it processes all accessible sites immediately [56]. Yet, biochemical studies described above indicated that after processing of its own N-terminus, MARTX_Vc CPD converts to an inactive conformation with a 500-fold reduced affinity for InsP₆ (K_d = 100 µM) [48] (Figure 4). Since this concentration is above the upper limit of the in vivo concentration of InsP₆ [54], only a small fraction of processed CPD would bind InsP₆ in vivo, limiting the likelihood of multisite processing. However, it was found that reactivation of MARTX_Vc CPD for high affinity binding of InsP₆ occurs after insertion of a new substrate into the hydrophobic pocket, indicating cooperativity of substrate and InsP₆ binding [48]. Both binding studies [48] and crystallography [49] (Figure 3G) have shown that chloromethyl ketone and epoxide inhibitors bound to Leu can substitute for a new substrate to restore the protein to an active enzyme-substrate complex. Upon reactivation, the protein is able to process any other available processing sites [48], [49], although there is a preference for processing within the same molecule of MARTX_Vc, indicating there may be a physical association of the CPD with the effector domains [48].

Potential of CPD as a Target for Therapeutic Intervention

The discovery of InsP₆-induced autoproteolysis as a critical stage for activation and effector delivery for large bacterial toxins raises the potential for anti-toxin small molecules to be developed as therapeutics. TcdB is the most significant virulence factor of C. difficile [9], [10], and it is conceivable that specific TcdB anti-toxin drugs could be combined with antibiotic and anti-toxin antibody therapies for treatment of recurrent antibiotic-associated diarrhea [64]. In addition, the contribution of MARTX_Vv to V. vulnificus septicemic infection is significant [26]–[29], suggesting that anti-MARTX_Vv CPD therapeutics may be of interest, particularly since there are currently no anti-toxin treatments for these rapidly progressing fatal infections. By contrast, clinical intervention against any domain of MARTX_Vc during cholera disease is impractical since animal studies suggest that MARTX_Vc functions only during the earliest stage of infection, prior to the onset of symptoms [24], [25]. Indeed, classical V. cholerae strains responsible for severe cholera during the fifth and sixth pandemics have a natural deletion in the rtxA gene that encodes MARTX_Vc, demonstrating that it is dispensable for late stage infection [37].

For research purposes, potent small molecule inhibitors of the MARTX_Vc CPD activity have been identified. These include peptidyl (acyloxy)methyl ketone epoxide [49] and chloromethyl ketone [48] inhibitors in which the amino acid leucine is linked to the functional group independently or as part of a tripeptide. Both classes of inhibitor are cysteine reactive and become covalently linked to the catalytic cysteine (Figure 3D). However, analysis of the pre-processed form of MARTX_Vc CPD revealed the N-terminus is bound within the active site prior to InsP₆ binding, which occludes access of the catalytic Cys to protease inhibitors. Thus, inactivation of the CPD with Cys reactive inhibitors requires long incubation times of up to 30 minutes [48]. Yet, upon initial intramolecular processing immediately upstream of the CPD, the catalytic Cys is exposed, facilitating rapid inhibition of subsequent processing events that release effectors [48], [49]. Consistent with these in vitro findings, exogenous addition of the membrane permeant z-Leu-Leu-azaLeu-epoxide inhibitor JCP598 to culture cells reduced actin crosslinking in vivo [49], suggesting inhibitors could be useful at a critical point after CPD translocation.

Similar inhibition studies using the more clinically relevant TcdB remain to be performed. Since the CGTs are only processed one time (Figure 1A), there is a concern that cysteine reactive inhibitors would be ineffective if the N-terminus bound in the active site blocks access to the catalytic Cys. A structure of the enzyme-substrate complex of TcdA or TcdB CPD is not yet available and inhibition by N-ethylmaleimide has been performed only with 30 minutes of incubation [51]. Thus, it is unknown if the accessibility of the Cys will be blocked similar to MARTX_Vc CPD. As described above, binding of InsP₆ to the recombinant TcdB CPD protein with the P1 leucine removed has been measured and shown to be similar to that with the Leu attached [55]. These results thereby suggest the association of the N-terminus with the TcdB catalytic site and relevant exposure of the catalytic site to inhibitors may differ from MARTX_Vc CPD. Hence, the potential for inhibition of InsP₆-induced autoprocessing by CPDs as a therapeutic intervention against TcdB merits further exploration. If access to the cysteine is indeed found to be blocked, the CPD could still be a suitable target for therapeutics, but with molecules that mimic InsP₆ itself to promote processing outside of cells, potentially disrupting the entire translocation/activation process.