Structure-Function Analysis of Inositol Hexakisphosphate-induced Autoprocessing of the Vibrio cholerae Multifunctional Autoprocessing RTX Toxin*

Vibrio cholerae secretes a large virulence-associated multifunctional autoprocessing RTX toxin (MARTXVc). Autoprocessing of this toxin by an embedded cysteine protease domain (CPD) is essential for this toxin to induce actin depolymerization in a broad range of cell types. A homologous CPD is also present in the large clostridial toxin TcdB and recent studies showed that inositol hexakisphosphate (Ins(1,2,3,4,5,6)P6 or InsP6) stimulated the autoprocessing of TcdB dependent upon the CPD (Egerer, M., Giesemann, T., Jank, T., Satchell, K. J., and Aktories, K. (2007) J. Biol. Chem. 282, 25314–25321). In this work, the autoprocessing activity of the CPD within MARTXVc is similarly found to be inducible by InsP6. The CPD is shown to bind InsP6 (Kd, 0.6 μm), and InsP6 is shown to stimulate intramolecular autoprocessing at both physiological concentrations and as low as 0.01 μm. Processed CPD did not bind InsP6 indicating that, subsequent to cleavage, the activated CPD may shift to an inactive conformation. To further pursue the mechanism of autoprocessing, conserved residues among 24 identified CPDs were mutagenized. In addition to cysteine and histidine residues that form the catalytic site, 2 lysine residues essential for InsP6 binding and 5 lysine and arginine residues resulting in loss of activity at low InsP6 concentrations were identified. Overall, our data support a model in which basic residues located across the CPD structure form an InsP6 binding pocket and that the binding of InsP6 stimulates processing by altering the CPD to an activated conformation. After processing, InsP6 is shown to be recycled, while the cleaved CPD becomes incapable of further binding of InsP6.

Vibrio cholerae is the etiologic agent of the acute intestinal infection cholera, that remains a world-wide problem with over 200,000 reported and an estimated 1 million actual cases each year (1,2). To cause illness, V. cholerae colonizes the small intestine, where it secretes its major virulence factor, the ADP-ribosylating cholera toxin, which elicits massive fluid secretion resulting in the profuse diarrhea that is the hallmark of cholera infection. Nearly all O1, O139, and non-O1/non-O139 clinical isolates of V. cholerae produce another secreted toxin that is the founding member of a new family of bacterial protein toxins called the multifunctional autoprocessing repeats-in-toxins (MARTX) 2 toxins (3)(4)(5)(6)(7). In V. cholerae, this toxin has recently been shown to contribute to virulence in mice and is among three secreted factors associated with the ability of V. cholerae to establish an intestinal infection that persists beyond 24 h (8,9). Hence, MARTX Vc is proposed to function during the earliest stages of human exposure to V. cholerae either to modify the intestinal tract allowing colonization to occur or to reduce the functionality of innate immune cells preventing clearance. The broad distribution of MARTX Vc among environmental isolates further suggests this toxin may have a role in extraintestinal survival (3,(5)(6)(7).
MARTX Vc is an unusually large protein of 4545 amino acids (aa) or Ͼ450 kDa. Most of its primary structure consists of 18 -20 aa glycine-rich repeats that are proposed to form a translocation structure that facilitates transfer of the central portion of the toxin across the eukaryotic plasma membrane (4). All known activities of MARTX Vc have been mapped to the central ϳ1700 aa of the protein. The RhoGTPase-inactivation domain (RID) causes conversion of activated GTP-bound Rho, Rac, and CDC42 to the inactive GDP-bound forms resulting in depolymerization of actin (10). The adjacent actin cross-linking domain (ACD) catalyzes the covalent cross-linking of monomeric G-actin resulting in the irreversible destruction of the cytoskeleton (7,11,12). Together these domains rapidly round eukaryotic cells without causing cell lysis.
To access their respective substrates, it was postulated that the RID and ACD are released into the eukaryotic cytosol after translocation by a 25-kDa cysteine protease domain (CPD) that is embedded within the central portion of MARTX Vc . Studies using recombinant CPD (rCPD) demonstrated that autoprocessing to the N-terminal side of rCPD is stimulated by a small molecule found in eukaryotic cell cytosol (13). Additional studies demonstrated that cleavage in vitro could be stimulated by binding of GTP, and particularly the non-hydrolyzable analog GTP␥ S . A mutation of the catalytic cysteine residue Cys-3568 in the holotoxin dramatically reduced both actin cross-linking and Rho inactivation activities, demonstrating that autoprocessing activity by the CPD is necessary for the cytopathic effects of MARTX Vc (13).
Recently, it has been reported that Clostridium difficile Toxin B (TcdB) also undergoes autocatalytic cleavage and that processing depends on a region of the toxin with sequence similarity to MARTX Vc CPD. The autoprocessing activity of TcdB was activated by dithiothreitol and inositol phosphate compounds with inositol hexakisphosphate (Ins(1,2,3,4,5,6)P 6 or InsP 6 ) functioning as the best stimulatory molecule (14,15).
Inositol phosphates (InsP x ) are commonly present in eukaryotic cells. Inositol triphosphate (Ins(1,4,5)P 3 ) plays an important role in signal transduction in eukaryotic cells. However, the highly phosphorylated inositol phosphates are the major inositol phosphates present in the cells (16) where the intracellular concentration of InsP 6 reaches up to 40 -60 M in several cell types (17)(18)(19). This range of concentrations was able to activate the cleavage of TcdB (14,15).
In this work, we focused on the molecular basis of MARTX Vc CPD autoproteolysis. We showed that InsP 6 is able to bind and activate the autoprocessing activity of MARTX Vc CPD. We have identified crucial amino acid residues within CPD that are involved in InsP 6 -induced processing of CPD. We also provide evidence for a molecular mechanism of InsP 6 -activated autoprocessing. Because of the high sequence similarity among the CPDs from a group of 24 known and putative bacterial toxins, it is likely that other proteins use a similar mechanism for autoprocessing.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Growth Conditions, and Reagents-Escherichia coli DH5␣, used for all DNA manipulations, and E. coli BL21(DE3), used for production of recombinant proteins, were routinely grown with shaking in Luria-Bertani (LB) medium supplemented with 100 g/ml of ampicillin at 37°C. myo-Inositol and sodium or potassium salts of phosphorylated D-myo-inositol compounds were obtained from Axxora or Sigma.
Site-directed Mutagenesis-Amino acid substitutions were introduced into the pHisCPD overexpression vector (13) using the Quikchange II XL mutagenesis kit (Stratagene, LA Jolla, CA). The mutations were generated with the suitable sense and antisense primers. Plasmid DNA was then prepared using the Qiaprep Spin Miniprep kit (Qiagen, Valencia, CA) and sequenced to confirm the incorporation of the desired mutation and the absence of secondary mutations. Sequences of primers and codon change introduced are found in supplemental Table S1.
Construction of Plasmid pMCSG7-cpd⌬51-Primers 5Ј-TACTTCCAATCCAATGCTTTAGCGGATGGAAAAAT-ACTCCA-3Ј and 5Ј-TTATCCACTTCCAATGTTAACCT-TGCGCGTCCCAGCTTAG-3Ј were used to amplify CPD from the pHisCPD overexpression vector (13) and then ligation-independent cloning was used according to the manufacturer's protocol for pET30-XA-LIC cloning (Novagen, Madison, WI) using modified vector pMCSG7 (20) as was done previously for pHisCPD overexpression vector (13). The protein expressed from this plasmid, rCPD⌬51, represents CPD from the cleavage site with 21 amino acids of plasmid-derived sequence and N-terminal His tag.
Production and Purification of Proteins-For the initial screening of mutant protein activities, proteins were purified from 50-ml cultures using TALON affinity chromatography (Clontech, Mountain View, CA) according to the manufacturer's recommendations. For InsP 6 binding assays, proteins were purified from 500-ml cultures by the method previously described (13) using a HisTrap HP column on an Á KTA purifier FPLC system (GE Healthcare). All proteins were eluted in buffer containing 250 mM imidazole, dialyzed into 20 mM Tris, 500 mM NaCl, pH 7.4, and after the addition of 10% glycerol were stored at Ϫ80°C. Protein concentration was determined using a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE), using extinction coefficients calculated by ProtParam.
In Vitro Cleavage Assay-Purified proteins were diluted into appropriate concentrations (as indicated in the figures) in a buffer of 20 mM Tris, 60 mM NaCl, 250 mM sucrose, 3 mM imidazole, pH 7.5. Reaction was initiated by addition of InsP 6 at concentrations indicated and incubated at 37°C for the time indicated. Reactions were stopped by the addition of SDS-PAGE loading buffer, and boiled for 1 min. Samples were separated by 15% SDS-PAGE gels and stained with Coomassie Blue R250. For semiquantitative determination of percent cleavage, gels were scanned, and the digital images were analyzed using NIH ImageJ 1.38 software. The percent cleavage was calculated using Equation 1, where P is the signal intensity of processed protein, and FL is the signal intensity of unprocessed protein, P 0 and FL 0 are the corresponding signal intensities of the control sample for the experiment. Fourier transform mass spectrometry of cleavage products was performed as previously described (13).
InsP 6 Binding Assay-Isothermal Titration Calorimetry (ITC) was performed with rCPD variants and InsP 6 using an MSC-ITC Calorimeter (Microcal, Inc.) at the Northwestern Keck Biophysics Facility. Purified proteins were dialyzed into 20 mM Tris, 150 mM NaCl, 1.5% glycerol, pH 7.5, and diluted to final concentrations of 15 M in the same buffer. The same buffer was used to dilute InsP 6 to a final concentration of 1.4 mM. Titrations were performed at 37°C by injecting 17 ϫ 2 l of ligand into the protein sample in the ITC cell. Resulting data were plotted with a nonlinear least-squares algorithm using Origin7 software (Microcal) and a model for a single class of binding. The binding constant (K b ϭ 1/K d ) was calculated from the curve.
Dynamic Light Scattering (DLS)-600 l of 0.5 mg/ml rCPD mutant proteins in 20 mM Tris, 150 mM NaCl, 1.5% glycerol, pH 8.0 (without InsP 6 ) were monitored for particle size in six measurements over 25 min at 25°C using a Zetasizer NanoS (Malvern Instruments, Westborough, MA). 6 Activate the Autoproteolytic Activity of rCPD-Recent studies showed that InsP 6 is a potent stimulator of autoprocessing of TcdB dependent upon a domain homologous to the MARTX Vc CPD (15). Based on this observation, it was tested whether InsP 6 is also a stimulator of the CPD from MARTX Vc . For in vitro experiments, rCPD, which represents the CPD of MARTX Vc plus 75-aa sequence upstream of the identified cleavage site with 6ϫ His tags at both the N and C terminus (13), was used. As shown in Fig. 1, addition of 100 M InsP 6 to 2 g (2.9 M) rCPD stimulated the autoprocessing activity of rCPD. A cleavage product of about 25 kDa was observed, which corresponds to the size of the rCPD cleavage product previously observed (13) after the incubation of rCPD with 1 mM GTP␥S (Fig. 1). To confirm that InsP 6 was inducing autoprocessing rCPD at the same site as GTP␥S, the rCPD cleavage products after InsP 6 treatment were analyzed by Fourier transform mass spectrometry. Products of 25,853.8 Da and 8320.9 Da were detected (data not shown), confirming that InsP 6 induces autoproteolytic cleavage of rCPD between Leu-3428 and Ala-3429, exactly as previously shown for GTP␥ S induction (13). Furthermore, rCPD with a Ser substitution at the catalytic Cys-3568 residue (rCPD C-S) failed to cleave upon addition of InsP 6 and processing was inhibited by N-ethylmaleimide (NEM), an inhibitor of cysteine protease activities (Fig. 1).

Physiological Concentrations of InsP
The induction of processing required the phosphate groups since myo-inositol did not stimulate processing ( Table 1). The ability of other lesser phosphorylated inositol compounds to stimulate rCPD was also investigated, and it was found that only the most highly phosphorylated InsP 6 stimulated maximal processing when added at concentrations as low as 1 M (Table  1). Altogether, these data demonstrate that autoprocessing of MARTX Vc CPD is specifically stimulated by InsP 6 dependent upon the catalytic cysteine residue.
Characterization of the InsP 6 -induced Autoprocessing of rCPD-To further characterize the cleavage reaction induced by InsP 6 , the kinetics of rCPD autoprocessing were investigated. As shown in Fig. 2, addition of 100 M InsP 6 resulted in autoprocessing with a reaction half-time of about 30 min under the given conditions and cleavage of nearly 100% was observed within 3 h. At 2 h, a wide range of concentrations of InsP 6 was able to activate the rCPD cleavage (Fig. 3). Even concentrations as low as 0.01 M were able to induce 35% cleavage by 2 h. By comparing the amount of InsP 6 added to the percent of rCPD cleaved, our data indicate that 1 molecule of InsP 6 was able to initiate the cleavage of ϳ100 molecules of rCPD indicating that InsP 6 is a potent stimulator of autoprocessing.
rCPD Cleaves in cis-Our observation that only one molecule of InsP 6 is able to stimulate autoprocessing of ϳ100 molecules of rCPD could be consistent with an in trans activation model wherein one activated CPD processes numerous other CPD proteins. To test a possible in trans cleavage model, we followed the approach of Reineke et al. (14) and used increasing concentrations of protein and limiting concentrations of InsP 6 . If cleavage occurs by an intermolecular mechanism, efficiency of processing should increase with increasing concentration of protein. If cleavage occurs by an intramolecular mechanism, then cleavage should remain constant with increasing protein concentration. As a positive control, a saturating InsP 6 concentration of 100 M InsP 6 was used to demonstrate maximum processing under these conditions at all dilutions ( Fig. 4). At lower InsP 6 concentrations, the recovered cleavage product remained constant with increasing protein concentration (Fig.

TABLE 1
Inositol compounds as stimulators of rCPD autoprocessing rCPD was tested for autoprocessing by addition of indicated compounds and incubated at 37°C for 2 h. Samples were then separated by SDS-PAGE. Autoprocessing was semiquantified by densitometric analysis of two experimental gels using ImageJ software.

Percent rCPD processed by indicated concentration of inositol compound
Ins(1,3,4,5,6)P 5 10 46 59 52 Ins(1,2,3,4,5,6)P 6 71 71 70 57 4). Thus, InsP 6 -activated cleavage is more efficient in a dilute sample compared with a concentrated sample, consistent with an in cis cleavage mechanism in vitro. This observation is consistent with previous observations in vivo where autoprocessing occurs predominantly by an intramolecular mechanism (13). Therefore, the ability of one molecule of InsP 6 to stimulate autoprocessing of ϳ100 molecules of rCPD cannot be explained by an in trans activation model. rCPD Binds InsP 6 -Therefore, to further understand mechanism for efficient activation of rCPD by InsP 6 , we more carefully characterized the interaction of InsP 6 with rCPD. Previous data showed that GTP␥ S bound to rCPD with the catalytic site mutation C3568S (rCPD C-S) with a K d of 1.3 ϫ 10 Ϫ4 M (13). As an initial test for InsP 6 binding, it was shown that a 30-min preincubation of rCPD C-S with InsP 6 inhibited binding of GTP␥S-BODIPY with a K i of 1.4 ϫ 10 Ϫ8 M (data not shown).
This result indicated that InsP 6 binds to rCPD at either the same or an overlapping site as GTP␥ S .
To further address binding of InsP 6 to rCPD, ITC was used to establish InsP 6 binding properties. ITC monitors and quantifies the heat absorption or release that accompanies binding of a ligand to a protein and allows the direct determination of a binding constant (K b ). A typical ITC experiment was designed wherein InsP 6 was titrated against a solution of uncleavable rCPD C-S in the ITC cell at 37°C. For these studies, the rCPD C-S mutant was used to prevent measurement of the heat change upon breaking of the peptide bond during processing in wild-type rCPD. The same buffer and temperature were used as in the previously established binding assay for GTP␥S-BODIPY and rCPD C-S (13) because a similar type of interaction was expected.
From the measured heat changes at each addition of InsP 6 (Fig. 5), a dissociation constant was calculated, K d , 6 ϫ 10 Ϫ7 M (0.6 M). When compared with the dissociation constant for GTP␥ S , K d , 1.3 ϫ 10 Ϫ4 M (130 M), the binding affinity of rCPD C-S for InsP 6 is much stronger. Importantly, the dissociation constant correlates to a concentration ϳ100-fold below the physiological InsP 6 concentration.
Processed rCPD No Longer Binds InsP 6 -To determine if rCPD can still bind InsP 6 after processing has occurred, 15 M rCPD was treated for 5 h with an InsP 6 concentration corresponding the first injection in an ITC experiment (ϳ2.15 M). Full cleavage of rCPD was confirmed by separating a portion of the reaction by SDS-PAGE (data not shown). The remaining cleaved rCPD was then measured by ITC for InsP 6 binding. As shown in Fig. 5, cleaved rCPD does not bind InsP 6 . To show that the 5-h incubation at 37°C was not the cause for the loss of binding, the same 5-h treatment was applied to rCPD C-S, and then ITC was done to test InsP 6 binding. The measured binding was the same as previously observed for rCPD C-S binding to InsP 6 except without the first InsP 6 injection (data not shown). These data suggest that after InsP 6 binds rCPD and activates cleavage, the protein undergoes a conformational change, which causes the release of InsP 6 molecule.
rCPD⌬51 Binds InsP 6 -The observation that the binding properties of rCPD to InsP 6 change after the cleavage raises the question as to whether the sequence of rCPD upstream of the cleavage site is involved in InsP 6 binding. The sequence upstream of rCPD cleavage site consists of 24 non-MARTX Vc aa from the cloning vector followed by 51 aa of MARTX Vc sequence. In the protein rCPD⌬51, the 51-aa part of MARTX Vc sequence is absent leaving only the plasmid sequence upstream of the rCPD cleavage site. Under the same reaction conditions, rCPD⌬51 was processed after addition of InsP 6 (data not shown) indicating that InsP 6 -induced processing does not require upstream MARTX Vc sequence. To confirm the binding properties remain the same when the 51-aa part of MARTX Vc sequence is absent, the C3568S mutation was introduced into rCPD⌬51 and binding of InsP 6 was measured by ITC. The same binding curve was measured giving the same K d as rCPD C-S (data not shown). These data showed that the MARTX Vc aa upstream of the Leu-3428 to Ala-3429 processing site do not contribute to InsP 6 binding or to the autoprocessing activity of CPD. Thus, the function of InsP 6 binding that induces cleavage  followed by structural changes is found solely within the CPD itself.
Site-directed Mutagenesis of Residues Involved in rCPD Autoprocessing-A previously performed bioinformatics analysis revealed 19 CPDs found in large bacterial proteins (13). A more recent search identified 24 CPDs (alignment shown in supplemental Fig. S1). From the sequence alignment, key residues were identified that included known and putative catalytic residues, positively charged residues that might interact with the negatively charged InsP 6 , and strongly conserved hydrophobic residues. Altogether, 23 residues were identified as potentially important for InsP 6 -induced autoprocessing (Fig.  6A). Site-directed mutagenesis was used to alter the identified codons on the rCPD overexpression vector pHisCPD (13), and then mutant proteins were purified and tested for autoprocessing at 37°C after addition of 0.1 M InsP 6 for 1 h (Fig. 6B). This concentration of InsP 6 under the experimental conditions stimulated the cleavage of ϳ60% of wild-type rCPD (Fig. 6B). This limiting concentration of InsP 6 thus allowed detection of mutants that were either totally or partially defective in autoprocessing activity.
Of the 23 residues targeted, 12 mutants showed no defects for in vitro autoprocessing activated by 0.1 M InsP 6 (Fig. 6B). The 11 defective mutants were then tested for autoprocessing with 10-fold increasing concentrations of InsP 6 . Of these, only one mutant, the catalytic cysteine mutant, rCPD C-S, was fully defective at the highest InsP 6 concentration of 100 M. All other mutants showed varying levels of partial defects. Four of these mutants were deemed significantly defective with at least 50% loss of function at 100 M InsP 6 compared with wild-type rCPD (Table 2) In addition, two proteins rCPD R3521A and R3593A showed increased cleavage with 100% processing at 0.1 M (Fig. 6B) and were nearly fully cleaved in concentrations of InsP 6 as low as 10 nM. Analysis of individual mutants is described in detail below.
Analysis of Putative Catalytic Site-Previously, it was shown that Cys-3568 and His-3519 residues were essential for CPD-EGFP autoprocessing in vivo after transient expression in epi-  thelial cells (13). The corresponding cysteine and histidine residues were essential also for autoprocessing of TcdB (15) and are 100% conserved among 24 putative CPDs (suppl. Fig. 1). These residues have thus been proposed to form the catalytic site for the cysteine protease. Consistent with this hypothesis, rCPD H3519A (Fig. 6B) and C3568S (Fig. 1) were both significantly defective for InsP 6 induced autoprocessing in vitro. While Cys-3568 was shown to be absolutely necessary for the activity at all InsP 6 concentrations, some residual cleavage of rCPD H3519A occurred at the highest concentration of InsP 6 suggesting that alteration to the structure by the mutation increased protein breakdown at or near the cleavage site.
Many cysteine proteases also require a negatively charged residue to form a catalytic triad. An aspartic acid residue of TcdB corresponding to D3469 of MARTX Vc CPD was previously proposed as a third part of the catalytic site because this residue was found to be essential for autoprocessing of an in vitro expressed N-terminal fragment of TcdB (15). Surprisingly, the rCPD mutation D3469A did not affect in vitro processing of MARTX Vc rCPD (Fig. 6B). An alternate negative charged residue Glu-3467, a residue that is not conserved in the CPD of TcdB (supplemental Fig. S1), was found nearby to Asp-3469 in the sequence of MARTX Vc CPD. However, the single mutation E3467A also had no influence on the processing activity of rCPD (Fig. 6B). Finally, it was considered that either Glu-3467 or Asp-3469 could serve as the essential negatively charge third catalytic residue. Indeed, the rCPD E3467A/D3469A double mutant abolished autoprocessing at 0.1 M InsP 6 , although the mutant was only partially defective at higher InsP 6 concentrations (Table 2). These results suggest that either Glu-3467 or Asp-3469 could play the role of the third part of the catalytic triad in the CPD but neither is absolutely essential.
Many cysteine proteases also require a glutamine residue near the negatively charged residue to form an oxyanion hole. Just to the N-terminal side of Glu-3467/Asp-3469, Gln-3465 was identified as a suitable candidate for the formation of an oxyanion hole. Despite the fact that Gln-3465 is 100% conserved among all 24 putative CPDs (supplemental Fig. S1), rCPD Q3465A was not defective for autoprocessing. A double mutation of nearby Q3461A in combination with Q3465A also showed no defect in autoprocessing (data not shown). Thus, if this protein requires an oxyanion hole, the essential residue is located elsewhere.
Putative InsP 6 Binding Residues-Because InsP 6 is a highly negatively charged molecule, strong basic CPD residues could participate in binding of InsP 6 . In total, 11 arginine or lysine residues including the 100% conserved Lys-3482 were changed to alanine and tested for a defect in autoprocessing. Of 11 mutant proteins tested, seven were defective for autoprocessing at 0.1 M InsP 6 (Fig. 6B) and two of these mutants, rCPD K3482A and K3611A, were also at least 50% defective at 100 M InsP 6 ( Table 2).
The inositol portion of InsP 6 could interact with hydrophobic residues to support the binding of InsP 6 . The most conserved rCPD hydrophobic residues were changed to alternate residues. Of the four residues targeted, two (Ala-3475 and Leu-3479) were important for processing at 0.1 M InsP 6 (Fig. 6B), but only rCPD L3479D was also defective at 10 and 100 M InsP 6 ( Table 2).
rCPD Residues Defective for InsP 6 Binding-The four residues critical for autoprocessing activity of rCPD, His-3519, Leu-3479, Lys-3482, and Lys-3611, and also the combination of Glu-3467/Asp-3469 were tested for InsP 6 binding. As shown in Fig. 5, rCPD K3482A was unable to bind InsP 6 . rCPD K3611A and L3479D, as well as these three mutations combined with a C3568S mutation, were also unable to bind InsP 6 (data not shown). Analysis of the rCPD L3479D protein by light scattering revealed this protein was aggregated in the absence of InsP 6 likely accounting for its loss of binding activity. By contrast, rCPD K3611A and K3482A were identical in size to rCPD C-S suggesting these proteins have a stable conformation and thus these two lysine residues could form the basis of binding pocket formed by two distant parts of CPD (data not shown). Other arginine and lysine residues that have reduced processing efficiency at low concentrations of InsP 6 may also assist in formation of this pocket.
The two putative catalytic site mutants, rCPD E3467A/ D3469A/C3568S (E3467A/D3469A mutations in combination with C3568S mutation to prevent cleavage) and rCPD H3519A were also found to have decreased binding of InsP 6 (Fig. 5). These proteins were not aggregated in the absence of InsP 6 as determined by light scattering although mutant E3467A/ D3469A/C3568S showed some drift in particle size over 25 min suggesting slight changes to conformation over time. At the concentration of protein used in the ITC, the K d for these mutants was outside the quantitative range of the assay and could not be determined although extrapolation of the data suggested the dissociation constant would be in the millimolar range.
The ability of proteins rCPD R3521A/C3568S and R3593A/ C3568S) to bind InsP 6 was examined to determine if their increase in processing efficiency was related to a lowered binding affinity. It was found that the binding constant was not significantly different than for rCPD C-S indicating that the enhancement to processing of these mutant is not related to an altered binding capacity of InsP 6 (data not shown).
Competitive Inhibition of rCPD Autoprocessing-To confirm all the results presented above, we designed a competitive processing experiment. rCPD was mixed with rCPD C-S, which binds InsP 6 , or with rCPD K3482A, which does not bind InsP 6 . Then 0.1 M InsP 6 (6-fold below the K d ) was added to initiate autoprocessing, and the mixture was incubated for 2 h. In the mixed reactions, the amounts of processed products were analyzed by SDS-PAGE.
In the rCPD/rCPD C-S mixed reaction (Fig. 7A, lane 7), the amount of processed product was not equivalent to that of the reaction with rCPD alone (lane 4) as would be expected from an in trans cleavage reaction, where rCPD would be able to cleave rCPD C-S. This result confirms our finding of an intramolecular processing mechanism. Indeed, rCPD C-S inhibited the autoprocessing of rCPD resulting in a lesser recovery of processed product then in rCPD alone (Fig. 7A, lanes 4 and 7). The inhibition of rCPD processing by rCPD C-S was overcome by increasing concentrations of InsP 6 (Fig. 7B), indicating that the basis for the inhibition is rCPD C-S competing for binding of a limiting amount of InsP 6 .
Similar results were obtained when rCPD C-S was mixed with the shorter protein rCPD⌬51, a reaction wherein the wildtype and mutant proteins can be distinguished by size. It was found that only rCPD⌬51 was processed, and this processing was inhibited by rCPD C-S at low InsP 6 concentrations, but was 100% processed when InsP 6 is present in excess (Fig. 7C, lanes 7  and 9).
Because inhibition of processing of both rCPD and rCPD⌬51 by rCPD C-S was observed at only 6-fold below the rCPD C-S dissociation constant, our results are consistent with the wildtype rCPD dissociation constant being close or the same as that measured for rCPD C-S. If the dissociation constant for the wild-type protein was significantly lower than for rCPD C-S, inhibition would not have been observed as InsP 6 would have preferentially bound to wild-type rCPD or rCPD⌬51 and rCPD C-S would thus fail to inhibit. Consistent with these conclusions, mutant rCPD K3482A, which does not bind InsP 6 (Fig. 5), does not compete with wild-type rCPD or rCPD⌬51 for InsP 6 resulting in no inhibition of rCPD or rCPD⌬51 cleavage (Fig. 7,  A, lane 8 and C, lane 8).

DISCUSSION
Recent studies have recognized a growing number of large bacterial protein toxins that are processed after translocation (13,21,22). These toxins fall into two families: the MARTX family toxins, including MARTX Vc (13) and MARTX Vv (22), and the clostridial glucosyltransferase toxins (CGTs), including TcdA, TcdB, TcsL, and Tcn␣ (14). All of these toxins have recently been shown to undergo an inducible autoprocessing event. For both TcdB and MARTX Vc , it is known that autoprocessing is essential for these toxins to induce cell rounding (13,15,23).
The signal to initiate autoprocessing and the catalytic domain responsible for the autoprocessing of TcdB has been the subject of much controversy. Rupnik et al. (24) originally proposed that the peptidase was a eukaryotic protease, but this model was negated when it was shown that protein-free cell extracts were able to stimulate processing in vitro (14). It was then found that inositol phosphates, in particular InsP 6 , were stimulatory factors for a protease activity embedded within the toxin itself. This group suggested the mechanism of processing involved an aspartyl protease located downstream of the TcdB hydrophobic region (14). About the same time, we published that MARTX Vc is autoprocessed, dependent upon a cysteine residue within the CPD (13). Egerer et al. (15) quickly corrected the processing model for TcdB and showed that InsP 6 -induced processing involved not an aspartyl protease, but a CPD homologous to the MARTX Vc CPD, and this CPD is similar located adjacent to the processing site. Although the mechanisms seemed similar, we had shown that GTP and particularly the non-hydrolyzable analog GTP␥ S , was a low affinity stimulator molecule for CPD activation (13). The poor efficiency of processing and binding that did not exactly correlate with the addition of eukaryotic cell cytosol suggested that MARTX Vc CPD could use an alternative activator in vivo or required a cytosolic co-activator for efficient stimulation of autoprocessing. Thus, we investigated whether MARTX Vc would respond to InsP 6 .
Here we show, that InsP 6 is able to activate the CPD of MARTX Vc . Moreover, very small concentrations as low as 10 nM InsP 6 were able to activate the CPD, whereas lesser phosphorylated inositol compounds required concentrations from 10 M to 1 mM to stimulate processing (Table 1). By comparing the binding properties and activation kinetics of InsP 6 with those of FIGURE 7. Competitive inhibition of rCPD autoprocessing by addition of rCPD mutants. All processing reactions were performed with indicated concentrations of protein and InsP 6 for 2 h at 37°C. In all gels, samples were separated by SDS-PAGE and stained with Coomassie Blue R250 to detect full-length (FL) or processed (P) proteins. A, 2 g of rCPD (wt) and/or rCPD C-S (or rCPD K3482A) were incubated in the absence or presence of 0.1 M InsP 6 . B, indicated concentrations of rCPD and/or rCPD C-S were incubated with or without indicated concentrations of InsP 6 . For both A and B, FL indicates fulllength unprocessed protein of both wt rCPD and/or mutant variants, while P indicates processed proteins. C, 2 g of wt rCPD⌬51 (wt⌬51) and/or rCPD C-S or rCPD K3482A variants were mixed and processing was activated by addition of the indicated concentration of InsP 6 . FL indicates the full-length of rCPD C-S or rCPD K3482A, while FL⌬51 marks the full-length size of rCPD⌬51 and P⌬51 marks the processed size of rCPD⌬51. As indicated, the processed form of rCPD C-S or rCPD K3482A would be expected to migrate similar to full-length rCPD⌬51. However, it is clear this processing does not occur as there is no decrease in the amount of FL rCPD C-S or rCPD K3482A. GTP␥ S , we propose that InsP 6 is the physiologically relevant compound. First, InsP 6 is widely present at relatively high concentrations in eukaryotic cells (up to 60 M) (17)(18)(19) but is not present in the extracellular spaces or the bacterial cytoplasm. Thus, toxin processing would only occur after translocation of the CPD to the eukaryotic cytosol and would not occur either within bacteria or before binding to a target cell. This model is consistent with our observation that rCPD can be purified from E. coli at its full, unprocessed size despite high concentrations of GTP in the bacterial cytoplasm. Second, the observed kinetics for the processing of rCPD induced by eukaryotic cell cytosol (13) closely follows that of InsP 6 (Fig. 2). Finally, we found that the affinity of InsP 6 for rCPD C-S is much stronger (K d , 6 ϫ 10 Ϫ7 M) than was described for GTP␥S (K d , 1.3 ϫ 10 Ϫ4 M). Moreover, the presence of 1 mM GTP␥ S did not change the K d of InsP 6 (data not shown), although preincubation with InsP 6 does inhibit binding of GTP␥ S -BODIPY. Under physiological conditions, GTP␥ S did not function as a co-activator with InsP 6 and a synergistic activity was not observed. An additive effect was observed only when a limiting concentration of 0.005 M InsP 6 was used (data not shown). A similar additive effect was observed with TcdB in presence of dithiothreitol and limiting concentrations of InsP 6 (15). Dithiothreitol has no effect on rCPD processing, likely because the protein has only one cysteine residue so no intramolecular disulfide bridges can be formed (data not shown). Overall, we conclude that InsP 6 is a potent in vivo stimulator of processing for both MARTX and CGT toxins and likely for all related bacterial toxins and large secreted proteins that contain a CPD (supplemental Fig. S1). GTP␥ S and GTP are likely successful mimics of InsP 6 and can interact with rCPD possibly through its negatively charged phosphate groups to activate processing when added at high concentration.
To study rCPD residues involved in InsP 6 -induced autoprocessing, scanning mutagenesis was performed targeting residues highly conserved among 24 putative CPDs. Over half of the mutations had no defect in processing when screened at low concentration of InsP 6 indicating they are not essential for processing. Concerning the catalytic site, we confirmed that Cys-3568 and His-3519 are necessary residues involved in autoprocessing. Surprisingly, while rCPD C3568S retains the ability to bind InsP 6 and indeed was used as a catalytically inactive mutant to solve the binding constant, rCPD H3519A had a dramatic loss in InsP 6 binding. This could indicate that the mutation causes a significant conformational change compared with the wild-type uninduced structure. Notably, a mutation of the nearby residues Arg-3521and Arg-3593 to Ala resulted in an increased cleavage activity (Fig. 6B) although the binding affinity for InsP 6 was not significantly increased (data not shown). Thus, we suggest that rather than affecting binding per se, this region is important for the conformation shift induced by binding of InsP 6 , which results in activation of the protease to initiate autoprocessing. Mutagenesis of the histidine could lock the protein in an open non-binding conformation while mutation of the arginines might set the protein closer to the activated conformation. This region may also be important for a second shift to a third conformation upon processing and release of the InsP 6 as the protein becomes inactive to further binding of InsP 6 .
Our further study of other residues that would contribute to the protease catalytic site conflicts with studies on the mechanism of TcdB processing. Egerer et al. (15) found that the Asp residue of TcdB equivalent to Asp-3469 was essential for processing. We found that both Asp-3469 and the nearby Glu-3467 needed to be mutated to block activity. Glu-3467 is not conserved in TcdB, which could explain why the single mutant was defective for TcdB. Interestingly, the CPD of a putative RTX toxin in Y. pseudotuberculosis has only the Glu-3467 residue and Asp-3469 is not conserved suggesting that in this protein, the glutamate might be absolutely essential for processing because the aspartate is absent. However, in MARTX Vc rCPD, processing of even the double mutant was lost only at low InsP 6 concentrations and light scattering measurements showed a slight drift in particle size of this protein over time. These observations suggest these residues may affect either protein structure leading to the loss of InsP 6 binding or the loss of the ability to transition to an active conformation except at high InsP 6 concentrations that stabilize the structure. Hence, if a third catalytic residue is essential for CPD enzyme action, it may require analysis of a crystal structure to identify the location of this residue. Furthermore, the location of residues that contribute to formation of an oxyanion hole may also require a tertiary structure for proper identification.
The site-directed mutagenesis also identified rCPD residues important for InsP 6 binding. The most highly conserved positively charged residues, Lys-3482 and Lys-3611, were identified as significantly defective for processing and defective for InsP 6 binding. An additional five lysine or arginine residues spread across the protein were also identified that are partially defective. Thus, it is proposed that in the folded structure, the peptide chains at both ends of the molecule interact to form a positively charged binding pocket for binding of the highly negatively charged InsP 6 . Analysis of other proteins that bind InsP 6 similarly showed that these proteins form a binding pocket of mostly lysine and arginine residues located throughout the protein (25,26).
Finally, to further understand the mechanism of rCPD autoprocessing, we needed to explain two observations. First, the minimal concentration of InsP 6 able to stimulate the autoprocessing was found to be 0.01 M, which is 60-fold below the determined dissociation constant (K d , 6 ϫ 10 Ϫ7 M). Second, we calculated that 1 molecule of InsP 6 activated the processing of up to 100 molecules of rCPD despite the fact that processing occurs by an intramolecular, not intermolecular, mechanism. During the course of a reaction, although occupancy of the InsP 6 binding pocket would be low under limiting concentrations of InsP 6 , proteins that do successfully bind InsP 6 would be processed but then become effectively inert and no longer compete with uncleaved protein for binding of InsP 6 . This suggestion was confirmed in the ITC binding assay wherein no binding of InsP 6 to the cleaved molecule was observed. Thus, the reaction could proceed, albeit more slowly, at concentrations below the binding constant and below a 1:1 ratio of protein to InsP 6 because the stimulator molecule can be recycled. This model is supported by our finding that a catalytic rCPD C-S mutant can inhibit processing of either wild-type rCPD or rCPD⌬51 by competing for low InsP 6 concentrations, but rCPD K3482A that cannot bind InsP 6 has no effect.
Based on the data obtained, we can propose a preliminary working model for the molecular mechanism of InsP 6 activated autoprocessing of the MARTX Vc holotoxin. After translocation into the cytosol, the CPD in its native conformation can bind InsP 6 with a K d well below the physiological concentrations of InsP 6 , into a binding pocket consisting of residues arranged across the entire length of the CPD. InsP 6 binding either induces proper protein folding or a conformation shift, possibly near the catalytic histidine, that then allows the catalytic site to form and/or the cleavage site to be inserted within the catalytic site. Alternatively, InsP 6 itself might help form part of the catalytic site. Once activated, the molecule undergoes autoprocessing. Following processing, the CPD portion of MARTX Vc is altered to a third conformation such that it releases InsP 6 and is no longer capable of binding InsP 6 . These observations suggest that the protease is inactivated after successful cleavage and thus does not likely go on to target other proteins within the eukaryotic cell.
Overall, in this study we examined the activation step for the autoprocessing activity of the MARTX Vc toxin of V. cholerae.
The domain corresponding to this activity, CPD, is conserved in other large bacterial toxins. Therefore, the obtained information could contribute to the understanding of the mechanism of activity of virulence factors of other pathogenic bacteria, such as C. difficile or Yersinia sp.