Most secreted bacterial toxins are produced as inactive precursors that become proteolytically activated upon entering a eukaryotic cell (1
). A select group of these toxins undergo autoproteolysis upon entry into the host cytosol, resulting in release of their effector domains (2, 3
). The Vibrio cholerae
RTX is a member of the M
utoprocessing RTX (MARTX) family of toxins, which all contain a cysteine protease domain (CPD) predicted to mediate autoproteolytic activation of the secreted protoxin upon entry into the eukaryotic cytosol (4
). Almost all clinical and environmental isolates of V. cholerae
produce RTX (5
), which enhances virulence and colonization in murine models of V. cholerae
infection (6, 7
). RTX autoprocessing is thought to release its actin disrupting effector domains from the target cell plasma membrane into the cytosol (4
) (fig. S1). Although autoproteolysis is essential for RTX toxin function (3
), the mechanism of RTX CPD activation is unclear.
It has recently been shown that the small molecules GTP and inositol hexakisphosphate (InsP6
) stimulate autoprocessing of the RTX CPD (3, 8
), while InsP6
activates a distantly related protease domain in Toxin B of Clostridium difficile
). We tested the ability of GTPγS, InsP6
, and InsP6
metabolites to activate V. cholerae
RTX CPD autocleavage in vitro. InsP6
potently activated autocatalysis, with a half-maximal autocleavage concentration (AC50
) of 0.9 nM versus 0.19 μM, 0.72 μM, and 240 μM for InsP(1,4,5,6)4
, and GTPγS, respectively ( and fig. S3). We confirmed the interaction between InsP6
and RTX CPD using surface plasmon resonance (SPR), which indicates InsP6
binds to the RTX CPD with an equilibrium affinity constant (KD
) of 1.3 ± 0.2 μM ().
Fig. 1 InsP6 activates the V. cholerae RTX cysteine protease domain (CPD) and the architecture of InsP6-CPD complex. (A) Activation of RTX cysteine protease domain autocleavage by InsP6. Recombinant RTX CPD (amino acids 3391-3650) was incubated with the indicated (more ...)
To gain structural insight into InsP6
-mediated activation of RTX CPD, we purified an autocleaved, minimal RTX CPD catalytic domain for co-crystallization with InsP6
). We determined the structure of the RTX CPD at 2.1 Å resolution, consisting of amino acids 5 through 203 (with residue 1 being the P1' alanine and P1' referring to the residue C-terminal to the scissile bond). The protease domain is comprised of a seven-stranded beta sheet, with five central parallel strands and two antiparallel capping strands (). Two helices flank one side of the sheet, lying diagonally in a groove created by a ~90° twist of the sheet, while a third helix caps the other side of the sheet. The Cys-His catalytic dyad (C140 and H91) lies at the C-terminal ends of the central D and E beta strands. The overall structure is reminiscent of the clan CD family of cysteine proteases, suggesting a common ancestor for this family of enzymes. Comparison with the two most closely related known protease structures, human caspase-7 (11
) and Porphyromonas gingivalis
) (fig. S4) reveals that two C-terminal helices have been replaced by a three-strand flap. This ‘β-flap’ structure forms a cleft in which we identified electron density for a single InsP6
molecule (fig. S5).
binding pocket is lined with basic residues, burying approximately 890 Å2
of surface area (). Nine of the twelve residues that directly interact with InsP6
via hydrogen bond contacts are positively charged, with a core of six residues (K54, R85, S136, R171, K183, and K200) forming the bottom of the pocket and six others surrounding the InsP6
molecule (T28, R29, H55, S169, and R182) ( and fig. S6). K195 covers the top of the InsP6
molecule, interacting with the C1, C5 and C6 phosphate groups. The protein–InsP6
interface is further stabilized by a network of water molecules between the InsP6
and β-flap. On the opposite side of the β-flap is the catalytic dyad and a large hydrophobic pocket for the P1 amino acid (the residue N-terminal to the scissile bond) (). The β-flap contributes three of the twelve hydrophobic amino acids that line the base of the P1 pocket, while helix 1 and two central beta sheets (D and E) contribute the remainder (). The surface properties and size of the P1 substrate pocket are consistent with the cleavage of RTX CPD after an N-terminal leucine (3
). The final N-terminal residue observed in our structure is I5 (the P5' position) (), which lies ~14 Å from the catalytic cysteine. No electron density was observed for the P1' through P4' positions, suggesting that they do not make strong contacts with the protease and may minimally contribute to substrate specificity.
Fig. 2 The InsP6 binding and active sites. (A) Electrostatic surface potential of the CPD as viewed from above the InsP6 binding site. Blue denotes positively charged surface; red denotes negatively charged surface. InsP6 is shown in the binding site as a stick (more ...)
The InsP6 binding site is structurally segregated from the active site, clearly indicating that InsP6 does not act as a co-factor for catalysis. Indeed, kinetic analyses revealed that InsP6 binding is independent of substrate binding to the active site and that the concentration of InsP6 does not alter the affinity of the InsP6-bound enzyme for its substrate (fig. S7). Since RTX CPD activity was strictly dependent on InsP6 in these analyses, we hypothesized that InsP6 binding may regulate exposure of the active site. We therefore tested the ability of a fluorescent maleimide derivative to alkylate the catalytic cysteine of wildtype CPD and a mutant CPD lacking two InsP6 interacting residues (R182Q/K183N). Whereas weak fluorescent labeling was observed for both wildtype and mutant CPD in the absence of InsP6, dose-dependent labeling of the active site cysteine was observed in the presence of InsP6 only for the wildtype CPD (fig. S8). Thus, productive binding of InsP6 is required to expose the active site cysteine of the protease to substrates and inhibitors. Consistent with this observation, pre-treatment of wildtype CPD immobilized on an SPR chip with N-ethylmaleimide (NEM) alone failed to block InsP6-induced autocleavage of the CPD from the chip (fig. S9). Simultaneous treatment of the CPD with InsP6 and NEM, however, inhibited InsP6-induced autocleavage, indicating that NEM can only react with the active site cysteine in the presence of InsP6. These results strongly suggest an allosteric mechanism of activation in which the active site is disordered or occluded in the absence of InsP6, a mode of regulation that likely protects the protease active site sulfhydryl until the toxin enters a eukaryotic cell.
Inspection of the structure suggested that the β-flap, which lines the side of the InsP6 binding cleft closest to the catalytic site, may contribute to enzyme activation by properly ordering the P1 pocket and active site. Among the many side chains that coordinate InsP6, the β-flap contains three interacting residues, R182, K183, and K195, that form a three-pronged ‘clamp’ above and below InsP6 ( and ). We tested the effect of the mutations R182Q, K183N, and K195N on both InsP6 binding (fig. S10) and catalytic activity (fig. S11) relative to wildtype CPD (). Mutation of K183 and K195 abrogated both InsP6 binding and autocleavage activity (, left table), while mutation of R182 only moderately reduced InsP6 binding but decreased autocatalysis by 340-fold relative to wildtype. Notably, R182 not only binds InsP6, it also engages in structurally-stabilizing hydrogen bonding interactions with D24 (fig. S12). This may explain the more significant impact of the R182Q mutation on catalysis rather than InsP6 binding, since R182 may primarily fine-tune the structure of the β-flap and contribute nominally to binding InsP6. Thus, diminution of InsP6 binding on one side of the flap clearly reduces catalysis mediated by residues on the opposite side of the flap.
Fig. 3 β-flap mutations decouple CPD autocatalysis and RTX activity from InsP6 binding. (A) Comparison of autocleavage efficiency (AC50) versus InsP6 binding (KD) measured by SPR for mutations in the InsP6 binding site (left table) and β-flap (more ...)
We sought to determine if mutation of non-InsP6 interacting residues in the β-flap might ‘decouple’ InsP6 binding affinity from the autocatalytic activity of the enzyme. We focused on W192 in strand G3, which makes van der Waals contacts with strands G1 and G2 (), and D178, which stabilizes the G1-G2 β-hairpin by hydrogen bonding with three backbone amide nitrogens and helps anchor the center of the G2 strand by forming a salt bridge with the side chain of H184 (). We made conservative (W192F, D178N) and more potent (W192A, D178A) mutations; as controls, we also mutated nearby solvent-facing residues D191 and E179 (D191A, E179A). The conservative mutations (W192F and D178N) minimally altered InsP6 binding affinity yet considerably reduced autocatalysis relative to wildtype and the controls D191A and E179A (, right panels). Furthermore, the more potent W192A and D178A mutations induced only a modest drop in InsP6 binding affinity yet a more severe defect in catalysis. The D178A mutant exhibited similar InsP6 binding and catalytic activity as the R182Q mutant, consistent with their similar positions in the β-flap. The W192A mutant was most dramatically affected: despite having moderate affinity for InsP6 (9.3 ± 0.6 μM), W192A had a catalytic defect that was twice as strong as that of the InsP6-binding mutant K195N. Thus, mutation of residues that apparently communicate InsP6 binding to the active site through structural rearrangement of the β-flap can decouple InsP6 binding from enzyme activation. The proposed structural rearrangement however appears to be subtle, as the circular dichroism (CD) spectra of the free- and InsP6-bound enzymes were nearly superimposable (fig. S13).
To assess the biological importance of InsP6 binding and β-flap integrity on RTX CPD function in vivo, we constructed V. cholerae strains containing CPD point mutations C140A (catalytic-dead), R182Q/K183N (InsP6-binding defective), and W192A (β-flap transition defective). Western detection of the CPD () showed that while most wildtype RTX was autoprocessed during growth of V. cholerae in LB media, RTX containing catalytic dead (C140A), InsP6-binding (R182Q/K183N), or β-flap (W192A) mutations were unprocessed (). Thus, even in the presence of high levels of InsP6 in LB media, all three mutations prevented CPD activation in the native RTX. When V. cholerae mutant strains or supernatants were incubated with human foreskin fibroblast (HFF) cells, intracellular actin crosslinking induced by mutant strains R182Q/K183N and W192A was severely reduced relative to wildtype, while C140A failed to induce actin crosslinking ( and fig. S14). Thus, InsP6 binding and an intact β-flap are required for RTX autocleavage and effector function.
-interacting residues in related MARTX cysteine protease domains are almost invariably conserved (fig. S2), strongly suggesting a shared mechanism for toxin activation in Gram-negative bacteria. Since InsP6
is exclusive to eukaryotes (13
) and is present at cytosolic concentrations >10 μM (14
), the responsiveness of the MARTX family to InsP6
through the evolution of a proteolytic biosensor seems an ingenious strategy for regulating the function of a secreted toxin.