ACD toxin–produced actin oligomers poison formin-controlled actin polymerization

See allHide authors and affiliations

Science  31 Jul 2015:
Vol. 349, Issue 6247, pp. 535-539
DOI: 10.1126/science.aab4090

A little toxin can do a lot

The actin cross-linking domain (ACD) is an actin-specific toxin produced by several bacterial pathogens. Heisler et al. discovered that ACD's pathogenic mechanism involves a highly unusual toxicity amplification cascade. Rather than directly inactivating the actin cytoskeleton, ACD blocks the activity of formins, actin regulatory proteins that play crucial roles in numerous cellular activities. ACD is exceptionally potent, even though its substrate is the most abundant protein of a eukaryotic cell: actin.

Science, this issue p. 535


The actin cross-linking domain (ACD) is an actin-specific toxin produced by several pathogens, including life-threatening spp. of Vibrio cholerae, Vibrio vulnificus, and Aeromonas hydrophila. Actin cross-linking by ACD is thought to lead to slow cytoskeleton failure owing to a gradual sequestration of actin in the form of nonfunctional oligomers. Here, we found that ACD converted cytoplasmic actin into highly toxic oligomers that potently “poisoned” the ability of major actin assembly proteins, formins, to sustain actin polymerization. Thus, ACD can target the most abundant cellular protein by using actin oligomers as secondary toxins to efficiently subvert cellular functions of actin while functioning at very low doses.

Bacterial toxins are the deadliest compounds on the planet. As little as a single molecule of a delivered toxin can compromise vital functions or even kill an affected host cell (1, 2). This is achieved by amplification of a toxin enzymatic activity by signaling cascades (e.g., by cholera, pertussis, and anthrax toxins) or by enzymatic inhibition of vital host complexes present in relatively few copies (e.g., Shiga and diphtheria toxins acting on ribosomes). Such efficiency is crucial because (i) the amount of a toxin produced early after infection is limited by an initially small number of bacterial cells; (ii) the host is protected by commensal bacteria; and (iii) the host immune system efficiently neutralizes toxins by means of adaptive (antibodies) and innate (e.g., defensins) (3) humoral defense factors.

Owing to its importance in multiple cellular processes, actin is a common target for bacterium- and parasite-produced toxins. Upon delivery to the cytoplasm of host cells by type I (as part of MARTX toxin) (4) or type VI (within VgrG1 toxin) (5) secretion systems, the actin cross-linking domain toxin (ACD) catalyzes the covalent cross-linking of Lys50 (K50) in subdomain 2 of one actin monomer with Glu270 (E270) in subdomain 3 of another actin monomer by means of an amide bond, which results in the formation of actin oligomers (6, 7). The actin subunits in the oligomers are oriented similarly to short-pitch subunits in the filament, except that a major twist of subdomain 2, required to accommodate such orientation, disrupts the normal intersubunit interface and precludes polymerization (6).

The currently accepted mechanism of ACD toxicity, by sequestering of bulk amounts of actin as nonfunctional oligomers, is compromised owing to the high concentration of actin (hundreds of micromolar) in a typical animal cell. Extrapolation of in vitro determined rates of the ACD activity (7) to cellular conditions suggests that a single ACD molecule per cell (i.e., ~1 pM) would require >6 months to covalently cross-link half of all cytoplasmic actin.

In contrast to these estimations, the integrity of the intestinal cell monolayers was disrupted when only a small fraction of cellular actin (2 to 6%) was cross-linked by ACD (Fig. 1, A to C, and fig. S1). To account for such dramatic effects, we hypothesized that the ACD–cross-linked actin oligomers are highly toxic because they can exert an abnormally high affinity to actin-regulatory proteins containing several actin-binding domains. To identify potential high-affinity partners of the actin oligomers, anthrax toxin delivery machinery was used to deliver ACD (8) into HeLa cells transfected with double-tagged (Twin-Strep–tag II and hemagglutinin) actin (SHA-actin) (fig. S2) and used for a pull-down assay. Several formins (DIAPH1, DIAPH2, DAAM1, and INF2) preferentially bound to the ACD–cross-linked actin oligomers (Fig. 1D). Treatment of epithelial monolayers with the formin inhibitor SMIFH2 affected the monolayer integrity similarly to ACD, whereas the Arp2/3 complex inhibitor CK-666 did not (fig. S3).

Fig. 1 Integrity of intestinal monolayers is compromised by low concentration of actin oligomers.

(A to C) Transepithelial electrical resistance (TEER) of small intestine epithelial cell (IEC-18) monolayers (A) was assessed after cytoplasmic delivery of a fusion protein of the N-terminal portion of a lethal factor from Bacillus anthracis and ACD (LFNACD) or a catalytically inactive ACD mutant as a control and correlated with the accumulation of ACD–cross-linked actin species by actin-specific antibody immunoblotting (B) and cell morphology (C). Additional actin-specific blots and quantification of cross-linked actin are presented in fig. S1. (D) SHA-actin pull-down. Lanes A: SHA-actin–transfected cells treated with inactive LFNACD (non–cross-linked actin). Lanes X: SHA-actin–transfected cells treated with active LFNACD (cross-linked actin). Lanes C: Nontransfected untreated cells used as a negative control. NaCl and FA label fractions eluted from Strep-Tactin beads with 0.5 M NaCl and 50% formamide, respectively. Samples were subjected to immunoblotting and probed with antibodies against hemagglutinin (HA) tag, actin, various formins, and profilin.

Formins are a major family of actin assembly factors involved in numerous actin-dependent cellular processes. The major functional domains of formins, formin homology domains 1 (FH1) and 2 (FH2), cooperate in nucleation and elongation of actin filaments. A noncovalent FH2/FH2 homodimer nucleates and remains at the polymerizing barbed end to facilitate processive filament elongation while protecting the filament from capping (9). Tandem poly(proline) (PP) stretches within the FH1 domains bind profilin-actin complexes and accelerate elongation as much as 10-fold (1012). FH1 domains of all formins that preferentially bound to the oligomers (Fig. 1D) contain 4 to 14 tandem PP stretches, which may contribute to strong profilin-mediated interaction with the oligomers.

To elucidate the mechanism of formin inhibition, we used constitutively active FH1-FH2 fragments of mDia1 and mDia2 (mouse orthologs of human DIAPH1 and DIAPH3, respectively) to monitor actin polymerization at the individual filament level by total internal reflection fluorescence microscopy (TIRFM) (Figs. 2 and 3 and fig. S4). In the presence of human profilin-1 (PFN1), the oligomers caused very prominent reversible blocks of elongation of formin-controlled, but not formin-free, actin filaments (Fig. 2, A to F; fig. S4, B and C; and movies S1 to S5). Formin-controlled filaments were identified by faster growth with a dimmer appearance (Fig. 2, A and E) (10) or by direct labeling of formin (Fig. 3A).

Fig. 2 Effects of ACD–cross-linked actin oligomers on polymerization of individual filaments controlled by mDia1(14PP).

(A) mDia1(14PP)-mediated polymerization from profilin-actin complexes in the absence (top) and presence (bottom) of actin oligomers (A-Oligo) was monitored by TIRFM. Arrowheads denote actin barbed ends: green, mDia1-controlled (dim and fast); yellow, mDia1-free (bright and slow); and red, mDia1-controlled stopped by the oligomers. (B and C) Quantification of (A): Filament elongation plots in the presence (B) or absence (C) of PFN1. Green and red curves describe filament elongation in the absence and presence of oligomers, respectively. Arrows denote the beginning and arrowheads indicate the end of elongation blocks caused by the oligomers on representative curves highlighted in black. (D) IC50 of oligomers determined by TIRFM as a percentage of stopped filaments (black) or growth rate inhibition (red curves). (E) Tethramethylrhodamine (TMR)–labeled actin (TMR-actin) (red) was polymerized in the presence of mDia1(14PP) and PFN1 without oligomers followed by flow of Oregon Green (OG)–labeled actin (OG-actin), oligomers, and PFN1. Arrowheads are as for (A). (F) Quantification of (E): Growth of mDia1-controlled filaments (green traces) and mDia1-free filaments (yellow traces). Better polymerization properties of OG-actin result in faster elongation at the formin-free ends.

Fig. 3 Effects of ACD–cross-linked actin oligomers on polymerization of individual filaments controlled by mDia2 and mDia1 with various FH1 lengths.

(A) OG-actin (green) polymerization in the presence of SNAP-tagged mDia2 labeled with SNAP-Surface 549 (SNAP-549–mDia2) (red) and PFN1 before and after the addition of oligomers (black arrow) was monitored by TIRFM. Red arrowheads indicate SNAP-549–mDia2 at an actin filament; white arrowheads indicate formin-free filament. Kymograph shows a stalled SNAP-549–mDia2–controlled filament upon addition of oligomers. (B) Effects of oligomers on formin-free filament elongation and elongation controlled by mDia2 and mDia1 formins with various FH1 lengths: 14PP, 2PP, and FH2 (no PP stretches). (C and D) Oligomer association (kon) (C) and dissociation (koff) (D) rates for mDia1(14PP) and mDia2.

In the presence of PFN1, the fraction of blocked mDia1 formin-associated filaments, as well as the inhibition of averaged growth rates, depended on the concentration of the added oligomers with an median inhibitory concentration (IC50) of 1.2 ± 0.6 (SEM) nM (Fig. 2D), in good agreement with the apparent equilibrium inhibition constant determined kinetically (appKi = koff/kon = 2.5 nM) (Fig. 3, C and D). After stops (oligomer dissociation), the filaments continued to polymerize with the rates characteristic for formin-controlled filaments (Fig. 2B and fig. S4A). In the absence of PFN1, the inhibition appeared to occur by a similar mechanism, but the overall effect was weaker, and the average duration of the blockage events was substantially shorter (Fig. 2, C and D). Although a profilin-mediated interaction of the oligomers with the PP stretches of FH1 was not absolutely required, it strongly amplified the efficiency of the inhibition at the elongation stage by contributing to multisite interaction with the oligomers. Thus, mDia1 constructs (fig. S5A) with either removed FH1 domains (FH2 only) or shortened from 14PP to 2PP stretches showed progressively lower response to inhibition by the oligomers in the presence of PFN1 (Fig. 3B). Similarly, the appKi of oligomers for mDia2 (containing 2PP) was 7.5 times that for mDia1 and depended on PFN1 (Fig. 3, B to D).

The inhibition of formin-mediated polymerization measured at the individual filament level correlated well with the inhibition observed in bulk pyrene assays (Fig. 4 and figs. S5 and S6). During spontaneous polymerization in the absence of PFN1, high concentrations (75 to 500 nM) of the oligomers mildly accelerated the polymerization, whereas mild inhibition was observed in the presence of profilin (Fig. 4, A and B). This is likely because of a low level of incorporation of the oligomers into the filaments (6) in the absence, but not in the presence, of PFN1 (fig. S5D), which leads to filament severing similar to that observed for actin species with impaired intersubunit surfaces (13).

Fig. 4 Actin oligomers inhibit mDia1-controlled actin polymerization in bulk pyrenyl-actin assays.

(A to D) Effects of actin oligomers (A-Oligo) on actin polymerization in the absence (A and B) or presence of mDia1(14PP) (C and D); without (A and C) or with PFN1 (B and D). Normalized FL, pyrene fluorescence expressed in percent of maximum. (E and F) Inhibition of profilin-dependent and independent actin polymerization controlled by various-length FH1 mDia1 constructs (14PP, 5PP, 2PP, or FH2 only) (see fig. S5, A and B, and fig. S6) assessed in the absence (E) and presence of PFN1 (F). (G) Apparent Ki for inhibition of mDia1(14PP) by the oligomers in the presence of PFN1 was calculated by measuring IC50 at two different concentrations of actin.

In contrast, the oligomers potently inhibited actin polymerization directed by mDia1 in the presence and, to a lesser extent, the absence of PFN1 (Fig. 4, C to F, and fig. S6). Fitting the inhibition of actin polymerization at 50% of maximum to a binding isotherm equation resulted in an IC50 for the mDia1(14PP) construct equal to 2.0 ± 0.2 nM and 4.8 ± 0.6 nM (SEM) in the presence and absence of PFN1 (Fig. 4, E and F). The ACD–cross-linked actin dimers purified to homogeneity (fig. S5B) inhibited the mDia1-controlled polymerization less efficiently than the mixture of higher-order oligomers (fig. S5, F to H), which suggests that the inhibition is amplified by multivalent interactions of the oligomers with mDia1. Accordingly, shortening the FH1 domain progressively decreased the efficiency of inhibition, with the IC50 values reaching ~30 and ~16 nM for the mDia1(FH2) constructs in the presence and absence of PFN1 (Fig. 4, E and F, and fig. S6).

Kinetic modeling (fig. S8) revealed that inhibition of both nucleation and elongation is required to accurately describe the effects of the oligomers on formin-controlled actin polymerization. Using experimentally determined parameter values for inhibition of elongation, we found good fits to the data (Fig. 4) by assuming that oligomers also inhibit nucleation by binding to free mDia1(14PP) formin with dissociation constants of 0.8 and 5 nM in the presence and absence of PFN1 (fig. S8, D and E). Inhibition of nucleation by the oligomers in the absence of PFN1 was also observed experimentally in filament seeding assays (fig. S7) and TIRFM experiments (fig. S4, D to G). Similar experiments in the presence of PFN1 were less conclusive owing to the overall lower nucleation ability of formins under these conditions (fig. S4, F and G, and fig. S7, G and H). To improve accuracy, modeling had to account for filament severing owing to incorporation of the oligomers in the absence of PFN1 (Fig. 4, A and C, and fig. S8, C and D).

Bacterial toxins are well known to disorganize the actin cytoskeleton acting on Rho family guanosine triphosphatase–controlled signaling pathways (14). Here, we found that toxins can not only exploit existing signaling pathways but also initiate a new toxicity cascade with de novo produced cross-linked actin species as “second messengers.” Because of a unique combination of properties that is not present in G- or F-actin (fig. S9A), these new actin species bind with high affinity to formins and adversely affect both nucleation and elongation abilities of these proteins, which causes their potent inhibition in both profilin-dependent and independent manners (fig. S9B). Thus, ACD creates toxic derivatives of actin with a disruptive “gain-of-function” mode of operation. We propose that the seemingly straightforward original assumption that ACD acts by the accumulation of bulk amounts of nonfunctional actin is inaccurate or at least incomplete. The toxin can be highly efficient at very low concentrations by acting on formins and, potentially, other actin regulatory proteins. This finding calls for the careful reevaluation of mechanisms used by other actin-related toxins, both of protein and small-molecule natures.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Table S1

References (15-45)

Movies S1 to S5

Modeling Program


  1. Supplementary materials are available on Science Online.
  2. Acknowledgments: We thank M. Vartiainen and T. Viita (University of Helsinki) for donating double-tagged SHA-actin plasmid. This work was supported by The Ohio State University startup funds (to D.S.K.), American Heart Association Innovative Research Grant (13IRG14780028 to D.S.K.), NIH (R01 GM079265 to D.R.K. and R01 GM098430 to D.V.). D.B.H. purified proteins, conducted pyrene actin and TIRFM experiments, analyzed data, and cowrote the manuscript; E.K. conducted cell culture, pull-down and immunoblotting experiments, analyzed data, and cowrote the manuscript; D.O.G. purified proteins and conducted pyrene actin assays; C.S., J.D.W., and D.R.K. conducted TIRFM and analyzed data; K.G.B., S.R.K., and N.L.P. conducted TEER; D.V. performed modeling; D.S.K. coordinated the study, analyzed data, and cowrote the manuscript. The authors declare no conflict of interest. Supplementary materials contain additional data.
View Abstract

Navigate This Article