|Home | About | Journals | Submit | Contact Us | Français|
Enterohemorrhagic E. coli (EHEC) attaches to the intestine through actin pedestals that are formed when the bacterium injects the protein EspFU into host cells1. EspFU potently activates the host WASP (Wiskott-Aldrich syndrome protein) family of actin-nucleating factors, which are normally activated by the GTPase Cdc42, among other signaling molecules. Apart from its N-terminal type III secretion signal, EspFU consists of five and a half 47-amino-acid repeats. Here we find that a 17-residue motif within this EspFU repeat is sufficient for interaction with N-WASP. Unlike most pathogen proteins that interface with the cytoskeletal machinery, this motif does not mimic natural upstream activators: instead of mimicking an activated state of Cdc42, EspFU mimics an autoinhibitory element found within N-WASP. Thus EspFU activates N-WASP by competitively disrupting the autoinhibited state. By mimicking an internal regulatory element and not the natural activator, EspFU selectively activates only a precise subset of Cdc42-activated processes. Although one repeat is able to stimulate actin polymerization, we show that multiple-repeat fragments have dramatically increased potency. The activities of these EspFU fragments correlate with their ability to coordinate activation of at least two N-WASP proteins. Thus this pathogen has used a simple autoinhibitory fragment as a component to build a highly effective actin polymerization machine.
Many pathogens target their host’s cytoskeletal machinery to facilitate attachment, entry, or cell-to-cell spreading2, 3. EHEC (O157:H7) causes severe gastrointestinal disease, and infection is dependent on attachment of the bacterium to intestinal epithelial cells via actin pedestals1. Pedestal formation is mediated by the pathogenic protein EspFU (also known as TccP)4, 5, which is injected through a type III secretion system6 into host cells where it stimulates actin polymerization by activating host WASP proteins (Fig. 1a). WASPs are activators of the Arp2/3 complex, but under basal conditions this function is regulated by an autoinhibitory interaction between the GTPase-binding domain (GBD) and the “C” helix in the catalytic WCA region7. Normal activation of neuronal WASP (N-WASP) requires disruption of autoinhibition (Fig. 1b), involving coordinated action of multiple endogenous activators, including Cdc42, the phospholipid PIP2, and SH3-domain-containing proteins such as Nck8, 9 (Fig. 1c). EspFU can independently activate N-WASP with a potency that is orders of magnitude higher than single endogenous activators8 (Fig. 1c,d, Fig. S1).
A common strategy exploited by pathogens in targeting host cell processes is to produce a protein that mimics an endogenous activator of that process6, 10, 11. This mimic will generally be constitutively active compared to the endogenous protein, which shows regulated activity12, 13. Among pathogens that target actin polymerization, there are examples of effector proteins that mimic the active states of N-WASP14 or N-WASP activators like Cdc4215. Previous studies have shown that EspFU interacts with the GBD16, 17, the regulatory domain in N-WASP that participates in autoinhibition and binds to activated Cdc427. Thus, we set out to test the hypothesis that EspFU may mimic an activated state of Cdc42.
EspFU has a type III secretion signal followed by five and a half 47-amino-acid repeats (Fig. 1e) (a six-and-a-half-repeat isoform of EspFU has also been reported5). The repeats are nearly identical and share little sequence homology with other proteins, except for the related EHEC protein EspF. EspF also activates N-WASP18 but does not participate in pedestal formation4, 19 (EspF contributes to the disruption of epithelial tight junctions).
We first mapped the minimal interacting fragments of both EspFU and N-WASP using pulldown-binding assays. We found that the N-WASP GBD could bind to a single EspFU repeat. We tested other truncations and found that the first 17 amino acids of each repeat constitute the minimal fragment with high-affinity for the GBD (Fig. 1f, Fig. S2). This sequence corresponds to a predicted α-helix20 and it is repeated six times in the full-length protein. There are minor sequence differences among the six copies of this motif, but the repeats appear to be equivalent in binding and functional assays (data not shown).
The N-WASP GBD is a composite regulatory domain made up of two overlapping elements – a Cdc42-binding element and an autoinhibitory element that binds to the C helix7. Detailed mapping of the N-WASP side of the interaction showed that the C-terminal, autoinhibitory portion of the GBD interacts with EspFU and the N-terminal, Cdc42-binding portion does not. The minimal EspFU-binding fragment of N-WASP is residues 228–270 (Fig. 1f, Fig. S2), which form the core of the interaction with the C helix in the autoinhibited structure (Fig. 2a)7.
Since a small peptide from EspFU interacts with the autoinhibitory portion of the GBD, it seemed unlikely that EspFU was activating N-WASP by mimicking Cdc42. Instead, we postulated that this EspFU fragment might be mimicking the C helix and disrupting N-WASP autoinhibition by competitively binding to the GBD. We found that EspFU does indeed competitively disrupt the N-WASP GBD autoinhibitory interaction in a pulldown-binding assay (Fig. 2b). Sequence alignment of the C helix and the first 17 amino acids of the EspFU repeat revealed sequence homology at three positions where hydrophobic amino acids are known to be critical for interaction with the GBD (Fig. 2a,c)21. We carried out a systematic alanine scan of both the N-WASP and EspFU fragments and tested binding of the mutants to the GBD. The results showed that mutation of any of the conserved hydrophobic positions to alanine on either the C helix or the EspFU fragment completely disrupted interaction with the GBD (Fig. 2c).
Previous work has shown that the isolated WASP GBD is only partially folded and unstable, but that fusion of the C helix to the GBD leads to a stably folded, helical structure that approximates the autoinhibited state7. We made a fusion of the GBD to the minimal 17-amino-acid interacting fragment from EspFU found that this protein had very similar helical content and thermal stability to the GBD-C helix fusion (Fig. S3).
These results show that the N-terminal 17 amino acids of the EspFU repeat are mimicking the C helix (an autoinhibitory element) and this motif competitively disrupts autoinhibition of WASPs (Fig. 2d). A similar sequence motif is present in EspF and likely activates N-WASP using the same mechanism. We found that one EspFU repeat binds to the N-WASP GBD (amino acids 214–274) with a Kd of 18 nM (Fig. S4). This affinity is about 100-fold tighter than previously measured values for the corresponding WASP GBD-C helix interaction (in trans)7 and this higher affinity explains how an intermolecular interaction can competitively disrupt an intramolecular interaction. The C helix also contributes to N-WASP’s interactions with actin monomers and the Arp2/3 complex22. We found that EspFU can bind to these molecules as well (Fig. S4). In spite of this, EspFU cannot stimulate actin polymerization without N-WASP (data not shown).
Utilization of a competitive autoinhibitory fragment represents a novel strategy for a pathogen to use in hijacking host cell signaling. One possible advantage of this strategy is increased specificity: a constitutively active signaling mimic will activate all downstream endogenous signaling pathways, whereas a mimic of a regulatory element should only activate a small subset. To test this idea we investigated whether EspFU could activate the p21-activated kinases (PAKs) – another group of Cdc42-responsive proteins whose regulation is most closely related to WASPs’. Cdc42 activates PAKs by disrupting an autoinhibitory interaction (which is structurally homologous to that of WASP)23. However, the PAK homolog of the WASP C helix lacks the consensus sequence of three conserved hydrophobic amino acids. EspFU was unable to activate PAKs (Fig. 2e, Fig. S5), indicating specificity of EspFU towards WASP family proteins (Fig. 2d). This more targeted activation may be advantageous for the pathogen, given the broad range of downstream effects mediated by WASP activators like Cdc42.
We also investigated the functional significance of this autoinhibitory mimic being repeated six times in one EspFU molecule. We tested the relative abilities of different length EspFU fragments to stimulate actin assembly using an in vitro pyrene-actin polymerization assay. One repeat from EspFU can disrupt the N-WASP autoinhibitory interaction in a binding assay (Fig. S6), but has low ability to stimulate actin polymerization (comparable to a single endogenous activator) (Fig. 3a). A fragment of two repeats shows intermediate activity in the actin assay, but at least three repeats are necessary for high potency. Each additional repeat gives higher activity until full-length EspFU (residues 80–337, lacking the secretion signal), which has a maximal activity at the same level as unregulated N-WASP (Fig. 3a, Fig. S7). To see if this trend continued, we made an EspFU protein that has eight and a half repeats and found that it reaches the same maximal activity as the wildtype protein, but is even more potent at low concentrations (Fig 3a). At very high concentrations, EspFU proteins had a partial inhibitory effect on actin polymerization (data not shown), likely due to EspFU sequestering actin monomers and/or Arp2/3.
We constructed a minimized EspFU protein in which six copies of the 17-residue C-helix mimic are separated by glycine-serine linkers. This construct was still able to stimulate actin polymerization (albeit with lower potency than wildtype EspFU), showing that the C-terminal portion of each EspFU repeat is not essential for activity (Fig. S8).
Next, we tested whether multiple repeats were also required in more complex assays of actin assembly – namely a reconstituted bead motility assay and an in vivo clustering assay. In the first assay, EspFU fragments are chemically coupled to polystyrene beads and these beads are incubated with a mixture of actin and actin-regulatory proteins and examined by microscopy for the ability to stimulate actin assembly and motility24. Beads coated with full-length EspFU triggered formation of an F-actin shell that breaks symmetry and induces transient polarized bead movement (Fig. 3b). Beads coated with three EspFU repeats showed similar behavior (albeit with a slightly more diffuse actin shell), while two-repeat beads formed shells that were very diffuse, showed significantly reduced actin fluorescence, and dissembled within minutes. Beads coated with a single repeat showed minimal ability to polymerize actin (Fig. 3b). We varied the surface density of EspFU protein on the beads and found that this had little effect on the relative activities of the different fragments (data not shown). In particular, one-repeat beads had weak activity, even at maximal density. This shows that a high local concentration of a single repeat has minimal effect – multiple repeats in the same polypeptide chain are required for potent activation of actin polymerization.
To test the ability of EspFU fragments to form actin structures in cells, we used an antibody-induced clustering assay. Previous work has shown that N-WASP or N-WASP activators can form actin structures when clustered at the plasma membrane25. In our assay, the EspFU fragment is expressed as a fusion to the cytoplasmic tail of a transmembrane protein and then clustered using an antibody that recognizes the extracellular portion of the protein26. Upon clustering of full-length EspFU, we observed co-localization of N-WASP and bright actin foci to the clusters, showing that clustering of EspFU is sufficient to trigger localized actin assembly (Fig. 3c, Fig. S9, S10). Furthermore, we tested three-, two- and one-repeat fragments and observed the same trend: three repeats showed similar activity to full-length, two repeats had intermediate activity and one repeat had lower activity (Fig. 3c). We used a computational image analysis tool (D. Vasilescu, unpublished) to detect and correlate antibody and actin clusters (Fig. S11). These quantitative results show that clustering of EspFU fragments with more repeats causes formation of actin clusters with higher frequency and intensity (Fig. 3d).
The requirement of multiple repeats for activation of actin polymerization suggests that EspFU needs to simultaneously bind and activate multiple N-WASP proteins to be highly active. To address this, we investigated the stoichiometry of the EspFU – N-WASP complex using analytical ultracentrifugation (AUC). By AUC (data not shown) and binding assays (Fig. S12), we found that each repeat in an EspFU fragment could simultaneously bind an isolated N-WASP GBD (for example, a three-repeat fragment bound to three GBDs). However, the model that EspFU has high potency because it can activate multiple N-WASPs does not explain why a two-repeat fragment has lowered activity in our actin assays. We found a likely answer to this question through stoichiometric analyses of complexes of a two-repeat protein with a larger, autoinhibited N-WASP construct. This AUC data showed that while the two-repeat fragment can bind one N-WASP easily, binding of a second is less favorable (possibly due to steric effects) and only happens in large excess of N-WASP (Fig. S13). Larger fragments of EspFU, on the other hand, readily formed complexes with two or more autoinhibited N-WASP proteins (Fig. S13). Taken with our activation data, this suggests that activity of EspFU fragments correlates with their ability to coordinate the activation at least two N-WASP proteins (Fig. 3e). There are a number of other cases where oligomerization of WASPs or WASP activators leads to similarly increased activity – including the potent actin polymerization induced by GST-tagged N-WASP WCA27 (presumably due to GST dimerization) and by oligomerized SNX9 in endocytosis28.
We have shown here that the EHEC protein EspFU has evolved a unique mechanism to potently and specifically activate N-WASP-mediated actin polymerization through a repeated autoinhibitory mimic. This co-opting of a small internal regulatory unit for activation of a target protein represents a novel pathogenic strategy. Furthermore, multiple repeated copies of this motif are essential for potent activity, apparently because a single EspFU protein must coordinate simultaneous activation of multiple N-WASPs.
EspFU fragments were amplified by PCR from EHEC genomic DNA (ATCC). All cloning and expression was done in E. coli BLR(DE3) strains lacking recombinase A.
Arp2/3 was purified from bovine brain and actin was purified from rabbit muscle and pyrene-labeled, as previously described29. Actin polymerization was monitored as an increase in pyrene-actin fluorescence. Raw fluorescence data were normalized to the lower and upper baselines of each curve. Relative activities of EspFU fragments were determined through calculation of the time required to reach half-maximal fluorescence (t1/2).
Actin and Arp2/3 were purified from Acanthamoeba castellanii, as previously described30. EspFU proteins were coupled to polystyrene beads and mixed with actin, Arp2/3, N-WASP, capping protein and profilin. The assay was adapted from a protocol developed by Orkun Akin & R. D. Mullins24. Images were acquired with a Nikon TE300 Eclipse inverted epifluorescence microscope with constant imaging settings between experiments.
NIH 3T3 fibroblasts were co-transfected with actin-GFP and the corresponding CD16/7 fusion construct. Cells were treated with clustering antibodies, as previously described26. Fixed cells were imaged with an oil-immersion 60X NA 1.4 objective on a Zeiss LSM 510 Meta microscope. Imaging settings were constant between experiments, except for minor adjustments of detector gain to accommodate variation in expression level.
Quantitative detection of clusters was done with a computational tool developed by D. Vasilescu. Ten cell images from three independent experiments were quantified for each construct. Data were presented in two metrics: the percentage of antibody clusters that have associated actin clusters and the intensities of these associated actin clusters. The relative intensity value is the ratio of the sum of intensities of associated actin clusters to the sum of intensities of all antibody clusters.
We thank O. Akin for reagents and assistance with the bead motility experiments; J. C. Anderson, A. Chau, R. Howard, M. Lohse, A. Remenyi & L. Weaver for assistance; and members of the Lim lab for discussion. This work was supported by grants from the NIH (NIGMS and Nanomedicine Development Centers, NIH Roadmap), the NSF and the Packard Foundation. G.M.R. was supported by the American Heart Association.
Supplementary Information is linked to the online version of the paper at www.nature.com/nature.