Structural basis for the interaction of Shiga toxin 2a with a C-terminal peptide of ribosomal P stalk proteins

The principal virulence factor of human pathogenic enterohemorrhagic Escherichia coli is Shiga toxin (Stx). Shiga toxin 2a (Stx2a) is the subtype most commonly associated with severe disease outcomes such as hemorrhagic colitis and hemolytic uremic syndrome. The catalytic A1 subunit (Stx2A1) binds to the conserved elongation factor binding C-terminal domain (CTD) of ribosomal P stalk proteins to inhibit translation. Stx2a holotoxin also binds to the CTD of P stalk proteins because the ribosome-binding site is exposed. We show here that Stx2a binds to an 11-mer peptide (P11) mimicking the CTD of P stalk proteins with low micromolar affinity. We cocrystallized Stx2a with P11 and defined their interactions by X-ray crystallogra-phy. We found that the last six residues of P11 inserted into a shallow pocket on Stx2A1 and interacted with Arg-172, Arg-176, and Arg-179, which were previously shown to be critical for binding of Stx2A1 to the ribosome. Stx2a formed a distinct P11-binding mode within a different surface pocket relative to ricin toxin A subunit and trichosanthin, suggesting different ribosome recognition mechanisms for each ribosome inactivat-ing protein (RIP). The binding mode of Stx2a to P11 is also conserved among the different Stx subtypes.

Shiga toxin-producing Escherichia coli (STEC) and Shigella dysenteriae are foodborne pathogens that cause intestinal and extraintestinal disease associated with severe morbidity and mortality. The emergence of new virulent and multidrug-resistant enterohemorrhagic E. coli and Shigella strains (1-3) highlight the public health and economic impact of these pathogens. Therapeutics effective at preventing and/or treating STEC infection are actively being sought, but as of now only supportive care is available. Shiga toxins (Stxs) are the major virulence factors of STEC that can cause hemorrhagic colitis and hemolytic uremic syndrome (HUS) (4). HUS is the most common cause of renal failure in infants and young children in the United States (5). Stxs are AB5 toxins, which belong to a family of type II RIPs, consisting of an enzymatically active A subunit that associates with a pentamer of identical B subunits. Stxs have two immunologically distinct types, Stx1a and Stx2a, and there are several subtypes within each of them. Although Stx1 and Stx2 are similar in structure and function, Stx2 subtypes are more toxic and are particularly associated with progression to severe disease including HUS (6)(7)(8). Stx1a differs from Shigella Stx only in one amino acid in the A subunit, whereas Stx2a has 55% homology to Shigella Stx in the A subunit and 57% in the B subunit. The B subunits bind to the glycolipid globotriaosylceramide (Gb3, CD77) receptor (9)(10)(11). After binding, the holotoxin is internalized by endocytosis and traffics retrograde through the Golgi apparatus, where the A subunit is proteolytically cleaved into a catalytic A1 subunit and an A2 fragment, which remain linked via a disulfide bond (12,13). In the endoplasmic reticulum, the A1 subunit is released from the A2-B5 complex by reduction of the disulfide bond and undergoes retrotranslocation from the endoplasmic reticulum into the cytosol (14). The A1 subunit is an N-glycosidase that removes an adenine from the highly conserved sarcin/ricin loop (SRL) in the 28S rRNA, resulting in inhibition of protein synthesis (15). We refer to the A1, A2, and B subunits of Stx2a as Stx2A1, Stx2A2, and Stx2B, respectively.
Several RIPs exploit the P stalk-dependent recruitment mechanism of translation factors to facilitate depurination. Stx2A1 (16) and Stx1A1 (16,17) bind to the ribosomal P stalk to depurinate the SRL (Fig. 1). The P proteins are critical for their depurination activity and cytotoxicity (16,18). In eukaryotic ribosomes the P stalk is composed of two heterodimers of P1 and P2 bound to the uL10 protein (previously known as P0) (19,20). The last 17 amino acids of these five P proteins are conserved in human ribosomes and the last 11 amino acids (SDDDMGFGLFD) are universally conserved among all eukaryotes (21). The crystal structure of the C-terminal domain (CTD) of P proteins has not been solved because of their intrinsic flexibility. The NMR structure of the human P1/P2 heterodimer showed that the C-terminal tails are flexible and can extend away from the dimerization domain (22). The main function of the P protein CTD is to bind to the translational GTPases, such as the elongation factors, and to recruit them to the SRL (23)(24)(25)(26). The P proteins are also critical for ricin toxin A subunit (RTA) to access the SRL (27)(28)(29)(30). Peptides corresponding to P protein CTD interact with RTA (31), trichosanthin (TCS) (32,33), maize RIP (MOD) (34), and Stx1A1 (17,35). Peptide mimics of the P protein CTD reduce the depurination activity of RTA by inhibiting ribosome binding (31) and This article contains supporting information. * For correspondence: Xiao-Ping Li, xpli@sebs.rutgers.edu; Nilgun E. Tumer, tumer@sebs.rutgers.edu.
rescue Stx1A1-mediated inhibition of protein synthesis in vitro (35). RTA binds the ribosome by a two-step mechanism, which involves first slow and non-stalk-specific electrostatic interactions with the ribosome, followed by fast electrostatic and hydrophobic interactions with the P stalk (27,29,36,37). The X-ray crystal structure of the complex between a peptide mimicking the last 11 amino acids of P proteins (P11) with RTA and TCS has been solved (33,38,39). Both toxins bound to P11 at a similar position, but the conformation and the orientation of the peptide on the two toxins were different. Only the last six residues (GFGLFD) (P6) bound to a hydrophobic pocket on RTA (38,39). Interaction with the acidic residues at the N terminus of P11 (SDDDM) was not defined even though this motif contributed to interactions with RTA (38). In contrast, most of the P11 peptide was present in the TCS structure bound to P11 (33).
The structure of the ribosome-binding site of Stxs has not been characterized. We previously showed that Stx holotoxins bind ribosomes, but do not depurinate the SRL because their active sites are blocked (40). Although Stx2a holotoxin produced in E. coli is not toxic, the A1 subunits expressed in E. coli are toxic because they can attack bacterial ribosomes (41). Hence it has been difficult to purify high levels of recombinant A1 subunits. Here we present the first biochemical and structural characterization of the complex between Stx2a holotoxin and the P stalk CTD. Stx2a binds P11 with low micromolar affinity. We cocrystallized P11 with Stx2a and solved the crystal structure. Our results demonstrate that P11 contacts Stx2a at a location different from the binding site of P11 on TCS or RTA. The binding mode of Stx2a to P11 is conserved among the different Stxs, providing a new target for the discovery of inhibitors.

Results
The Stx2 holotoxin interacts with P11 with low micromolar affinity The ribosome-binding site of Stx1A1 and Stx2A1 is distant from the active site ( Fig. 1) (18). Unlike ricin holotoxin, Stx1a and Stx2a holotoxins can interact with the ribosome because their ribosome-binding site is exposed, whereas their active sites are blocked (40). To determine whether Stx2a interacts with P11 ( Fig. S1A), we measured the binding kinetics using surface plasmon resonance. Stx2a was immobilized on a CM5 sensor chip of a Biacore T200 and P11 was passed over the surface at different concentrations. As shown in Fig. 2, P11 interacted with Stx2a with fast on and fast off kinetics. The equilibrium dissociation constant (K D ) was determined by fitting the steady-state binding levels at different peptide concentrations. The K D of the Stx2a-P11 interaction was 66 6 6 mM (Fig. 2), which was similar to the K D of the RTA-P11 interaction (79 6 29 in PBS-P buffer).

Structure of the Stx2a-P11 complex
To obtain a detailed molecular understanding of the interaction between Stx2a and the ribosomal P stalk, we solved the Xray crystal structure of Stx2a in complex with P11. The Stx2a-P11 structure was solved at 1.9 Å resolution in the P2 1 2 1 2 space group and formed a 1:1 stoichiometric complex with one Stx2a-P11 complex in the crystallographic asymmetric unit (Fig. 3A). After molecular replacement calculations were performed placing the Stx2a holotoxin, the resulting phase information was used to calculate electron density maps utilized to manually build residues 6 to 11 of the P11 peptide (Fig. 3B). Table S1 contains the crystallographic and refinement data for the Stx2a-P11 structure demonstrating a refined molecular model with agreement to the crystallographic data, as well as excellent geometry. This includes the P11 peptide which has an RMSD bond length of 0.0034 Å, an RMSD bond angle of 0.54°, along with no Ramachandran plot outliers. The Stx2a holotoxin is comprised of one enzymatic Stx2A subunit and five lectinlike Stx2B subunits that assemble into a pentameric arrangement (42,43). The Stx2a-P11 structure was very similar to the apo form of the Stx2a holotoxin (PDB ID: 1R4P) with a rootmean-square deviation (RMSD) of 1.4 Å between 622 C-a atoms within each molecule upon superposition, indicating no major conformational changes of Stx2a when bound to the P11 peptide relative to the unbound form of Stx2a. In the Stx2a-P11 structure Stx2A2 residue Glu-259 is buried in the active site of Stx2A1 ;2.7 Å away from active site residue Tyr-77 (Fig. 3C). The position of Glu-259 deep within the active site likely blocks interaction of Stx2a with the SRL rendering the holotoxin catalytically inactive. The Stx2a-P11 complex structure demonstrates that the P protein-binding pocket is not sterically hindered, indicating that the holotoxin is capable of ribosome interaction.

Interaction between Stx2a and P11
The first five residues of the P11 peptide (SDDDM) were not observed in the electron density maps likely because of the inherent flexibility of these residues from their relatively diminished interaction with Stx2a (Fig. 3B). The last six residues of P11 (GFGLFD) bind to a shallow pocket on the surface of Stx2A. The P11-binding pocket, located on the opposite face of Stx2A relative to the active site (Fig. 3A), is formed by a-helices A, F, and I along with b-strands b and c and the loop between these two b-strands within Stx2A. Altogether, the interaction between Stx2A and P11 buried a total surface area of 681 Å 2 generating four hydrogen bonds and two salt bridges that possessed interaction distances of 3.5 Å or less. The P11 residues that interact substantially with Stx2A were Phe-7, Phe-10, and Asp-11. Phe-7 and Asp-11 sit in positively charged segments of the P11-binding pocket (Fig. 4A) lined by Stx2A residues Arg-172, Arg-176, and Arg-179. Phe-7 p-stacks with Stx2A residue Arg-179 whereas the two salt bridges in this complex form between the Asp-11 side chain and Arg-172 and the C-terminal carboxylate of Asp-11 with Arg-179 (Fig. 4B). Arg-176 generated one of the four hydrogen bonds in the Stx2-P11 complex with the main chain carbonyl oxygen of Phe-10 (Fig. 4C). The other three hydrogen bonds in the Stx2a-P11 complex were also formed in this region of the pocket between the Asp-11 side chain with Asn-18 and Gln-175 and the C-terminal carboxylate of Asp-11 with Gln-175 (Fig. 4C). Phe-10 sits in a small nonionic portion of this Stx2A pocket (Fig. 4A) lined by residues Gln-33, Thr-36, Ser-229, and the hydrophobic residue Leu-232 (Fig. 4D). Interaction of P11 with Stx2a. Stx2a was immobilized on Fc2 of Biacore T200 at 4364 relative units. Fc1 was activated and blocked as reference. P11 was passed over the surface at six different concentrations (5-1215 mM). The association and dissociation times were for 30 s at a flow rate of 50 ml/min. The running buffer was PBS-P with 2% DMSO. The data were solvent corrected and fit with Biacore T200 software 3.0. The analysis was repeated three times. Figure 3. Structure of Stx2a-P11 complex. A, structure of Stx2a with Stx2A (green) and the Stx2B pentamer (yellow, magenta, cyan, salmon, and light gray) depicted as a ribbon diagram in complex with the P11 peptide (blue) drawn as sticks. Stx2A1 active site residue Tyr-77 is drawn as sticks and colored red. B, original 2F o 2 F c (blue mesh) and F o 2 F c (red mesh) electron density maps of the P11 peptide drawn as sticks at the binding pocket of Stx2a with carbon and nitrogen atoms blue and oxygen atoms red. The 2 F o 2 F c electron density map is presented at 1.0 s level, whereas the F o 2 F c electron density map is presented at 3.0 s level. The maps were calculated before P11 was built into the density maps. The P11 primary sequence depicted in bold text. C, close-up of the Stx2A1 active site occluded by Stx2A2. Stx2A1 (green) active site residue Tyr-77 is drawn as sticks and colored red. Stx2A2 residue Glu-259 is color coordinated to the main chain color with oxygen atoms red. The proximity of Tyr-77 and Glu-259 is highlighted by blue dashes.
Different P11-binding modes of Stx2A, RTA, and TCS The Stx2a-P11-binding site is distinctly located relative to the P11-binding sites on RTA (PDB ID: 5GU4) and TCS (PDB ID: 2JDL) which are similarly positioned (Fig. 5A). RTA and TCS share the same P11-binding pocket with the P11 peptide backbone rotated clockwise ;90 degrees relative to each other (Fig. 5A). Despite the overall structural similarity of the core region of Stx2A residues 1-206 with RTA residues 7-214 (RMSD of 1.8 Å for Ca atoms) and TCS residues 1-198 (RMSD of 1.9 Å for Ca atoms) when they are bound to P11, Stx2A has a distinct structural feature compared with RTA and TCS within the P11-binding pocket that contributes to the unique interaction with P11. b-strands b and c along with the loop region between these two b-strands (loop b-c) in Stx2A are more extended relative to the analogous region in RTA or TCS. This relatively elongated segment in Stx2A forms an appreciable part of the P11-binding surface, contributing to the unique P11-binding mode of Stx2A (Fig. S2, A and B). Additional primary sequence differences within the Stx2A P11-binding pocket further preclude RTA and TCS binding to P11 within the same pocket as Stx2A.
The primary sequence deviations among RTA, TCS, and Stx2A include three positively charged residues within the P11binding pocket (Arg-172, Arg-176, and Arg-179) of Stx2a. Each of these residues made key interactions, including H-bonds and salt bridges with P11. The equivalent residues to Arg-172 and Arg-176 in RTA are Gln-182 and Gly-186, respectively (Fig.  S1B). Replacement of the positively charged Arg-172 with Gln-182 in RTA would preclude salt bridge formation with Asp-11. Gly-186 in RTA, instead of Arg-176 in the equivalent position in Stx2A would prevent the H-bond of Stx2A with the main chain carbonyl oxygen of Phe-10. In TCS, Gln-169 and Gly-172 are replaced by Stx2A residues Arg-176 and Arg-179, respectively. Replacement of the positively charged Arg residue with Gln-169 in TCS would preclude salt bridge formation with Asp-11, whereas replacing the Arg residue with Gly-172 would prevent formation of p-stacking interaction with Phe-7, as well as the salt bridge formed with the C-terminal carboxylate of Asp-11 (Fig. S1B). Collectively, the absence of residues in RTA and TCS that form key interactions within the Stx2a-P11 complex further underscores why RTA and TCS bind P11 at a distinct site relative to Stx2a. Structure of Shiga toxin 2 with P stalk protein C termini RTA and TCS are more structurally similar to each other than to Stx2A with an RMSD of 1.2 Å between 233 Ca atoms within each molecule upon superposition. RTA (residues 215-251) and TCS (residues 196-236) form a b-sheet comprised of two b-strands (b-strands i and j), which is not found in Stx2A. This structurally unique b-sheet in RTA and TCS formed an appreciable portion of the P11-binding pocket, contributing to the different interaction mode of RTA and TCS with P11 relative to Stx2a (Fig. S2, C and D). In RTA, Arg-234 and Arg-235 within loop i-j established an electropositive region (Fig. 5B) not present in TCS (Fig. 5D) that interacted with the negatively charged C-terminal residue Asp-11 side chain and carboxylate moiety in P11 forming one salt bridge between Asp-11 and Arg-235 (Fig. 5C). TCS replaced these positively charged Arg residues within loop i-j with Asn-217 and Ala-218 removing the electropositive charge that facilitated contact with the negatively charged C terminus of P11. Consequently, P11-bound TCS rotated approximately 290 degrees compared with the RTA-P11 complex. The different P11 configuration in TCS positioned Asp-11 in a nonpolar environment whereas Asp-2, Asp-3, and Asp-4 were positioned in a positively charged region within this pocket (Fig. 5D) forming three H-bonds with TCS residues Gln-169 and Gln-170 and one salt bridge between Asp-4 and Lys-173 (Fig. 5E). Five additional H-bonds formed between TCS residues Gln-219, Gly-231, Val-232, and Asn-236 with P11s Met-5, Leu-9, Phe-10, and Asp-11, respectively. Despite the different orientation of P11 when bound to RTA and TCS, Phe-10 anchored each peptide in the same hydrophobic pocket within the P11-binding cavity of RTA and TCS.

Discussion
The crystal structure of Stx2a-P11 determined in this study shows that the active site of Stx2A is sterically occluded by residues within Stx2A2, whereas the binding pocket for P proteins is unhindered, indicating that the holotoxin can interact with the ribosome. Because of the structurally impeded active site of Stx2a, the holotoxin is catalytically inactive. Because the structure of Stx2A1 is similar to RTA and TCS, it has been suggested that the CTD of ribosomal P proteins would bind at a similar site on these three RIPs (33,44). The structure of P11 with RTA and TCS confirmed that RTA and TCS similarly bind P11 within the same pocket with the respective peptides in a different orientation relative to each other (38,39). We show here that key structural differences resulted in a different binding mode for P11 that involved a completely different binding pocket on the surface of the Stx2a holotoxin. The P11-binding pocket constituted a-helices A, F, and I along with bstrands b and c of the A subunit. The b-strands b and c along with the loop region between these two b-strands, loop b-c are more extended in Stx2A relative to the analogous region in RTA or TCS. This comparatively elongated segment in Stx2A formed an appreciable part of the P11-binding pocket, indicating why Stx2a formed an exclusive P11-binding mode relative to RTA and TCS (Fig. S2).
The structure of Stx2a bound to P11 is consistent with the mutation analysis of Stx2A1 indicating that Arg-172, Arg-176 and Arg-179 are important for ribosome binding, depurination activity, and cytotoxicity of Stx2A1 (18). These residues are also conserved in Stx1A1 and play an important role in ribosome binding, depurination activity, and cytotoxicity of Stx1A1 (18). Arg-172, Arg-176, and Arg-179 established a positively charged region within the P11-binding pocket that interacted with the negatively charged C-terminal Asp-11 while forming several key H-bonds and salt bridges with P11. Yeast two-hybrid studies showed that alanine substitutions in Arg-172, Arg-176, Arg-179, and Leu-233 (Leu-232 in Stx2A1) abolished the interaction of Stx1A1 with P2 protein and with P11 and reduced the translation inhibitory activity of Stx1A1 (35). Amino acid changes at the three highly conserved arginines within the P11-binding pocket of Stx2A1 would prevent the formation of several key noncovalent interactions with P11, such as the salt bridges between Arg-172 and Arg-179 and Asp-11, the Hbond between Arg-176 and P11 as well as the p-stacking interaction between Phe-7 and Arg-179, ultimately precluding the interaction with P11. The P11-binding pocket of Stx2a also has a nonpolar patch lined by the strictly conserved Thr-36 and Leu-232, which interact with Phe-10 of P11. Although mutation analysis of Stx2A1 indicated that P11 binds Stx2A1 in the same position as in Stx2a, further studies will address if the P protein-binding site on Stx2A1 is the same as on Stx2a.
The P11-binding mode of Stx2a is likely similar to other Stxs from E. coli, such as Stx2e, as well as Stx from S. dysenteriae. Stx2A has a similar fold to the A subunit of Stx2e (PDB ID: 4P2C) with an RMSD of 0.7 Å over 296 Ca atoms and a similar fold to the A subunit of Shigella Stx (PDB ID: 1R4Q) with an RMSD of 1.1 Å over 275 Ca atoms (Fig. S3A). Stx2A has a primary sequence identity of 92.6% and 61.7% to the A subunits of Stx2e and Stx, respectively (Fig. S1C). The structural similarity combined with the high primary sequence identity between the A subunits of Stx2a, Stx2e, and Stx indicates that many of the residues within the P11-binding surface of Stx2a are also present in Stx2e and Stx, suggesting a similar P11-binding mode. The residues responsible for the electropositive surface on Stx2A (Arg-172, Arg-176, and Arg-179) that interact with the C-terminal Asp-11 and form other key interactions with P11 are strictly conserved in the A subunits of Stx2e and Stx (Fig.  S1C). The nonionic residues within the nonpolar portion of the P11-binding pocket that interact with the hydrophobic residue Phe-10 of P11 are also highly conserved in the A subunits of Stx2e and Stx, including the strictly conserved Thr-36 and Leu-232 (Fig. S1C). Moreover, the superpositioned P11 peptide from the Stx2a-P11 structure onto the Stx structure (PDB ID: 1R4Q), displays a very similar Stx-binding pocket for P11 with an equivalent overall landscape and electrostatic surface distribution supporting that P11 interacts in a comparable fashion with Stx and Stx2e (Fig. S3B).
Docking analysis suggested that Stx1A1 may interact with P11 with charge-charge interactions at the N-terminal end and with hydrophobic interactions at the C-terminal end (33,44). However, electron density was observed only for the last six residues (GFGLFD) of P11 in the structure of the Stx2a-P11 complex. The absence of electron density for the first five acidic residues at the N terminus of P11 suggests a weaker affinity for these residues that ultimately results in their absence in the electron density maps. In contrast, the last six residues of P11 were readily evident in the electron density maps, indicating a relatively stronger interaction, underscoring the importance of the last six residues of P proteins in binding of Stx2a to the P stalk. Alanine-scanning mutagenesis of P11 corroborated these results and showed that the last five amino acids (FGLFD) are critical for binding to Stx1A1 (35).
The first five residues of P11 were also missing in the structure of RTA bound to P11 (PDB ID: 5GU4) possibly because (SDDDM) motif of P11 was less tightly bound to RTA because of higher disorder. In contrast, most of the P11 peptide was present in the TCS structure bound to P11 (TCS-P11) (PDB ID: 2JDL) with only the N-terminal serine residue missing (33). However, the electron density for residues 2 to 5 of the TCS-P11 structure was qualitatively diminished relative to residues 6 through 11. P11 residues 2 to 5 in the TCS-P11 complex had comparably higher B-factors at ;60 Å 2 relative to P11 residues 6 to 11 with B-factors ranging from 20 to 30 Å 2 , further indicating a more limited interaction with the N-terminal region of P11. The positively charged residues, Lys-173, Arg-174, and Lys-177 of TCS formed electrostatic interactions with the DDD residues of P11 and the hydrophobic motif of P11 docked into a hydrophobic pocket on TCS (33). The maize RIP (MOD) interacted with the P stalk CTD through yet another mechanism, Structure of Shiga toxin 2 with P stalk protein C termini which involved four positively charged lysines, Lys-143-Lys-146 interacting with the negatively charged DDD motif of P11 (34,46).
We show here that Stx2a binds P11 at a different site and in a distinct configuration compared with RTA or TCS (Fig. S4), providing evidence that the P stalk CTD interacts with RIPs by distinct mechanisms. The P stalk-binding sites have been shown to differ between the translational GTPase families as well (47). It has been suggested that P stalk proteins do not recognize specific sequences and positions in translational GTPases, rather they recognize a hydrophobic grove (48). The residues responsible for P protein interaction are not conserved among the RIPs. The structures of the Stx2a-P11, RTA-P11, and TCS-P11 complexes indicate that the flexible CTD of P stalk proteins can adopt distinct orientations and interaction modes depending on the structural differences between the RIPs to facilitate depurination of the SRL.
Recent results identified small molecules that bind to the ribosome-binding site of RTA and inhibit its activity, establishing ribosome-binding site of RTA as a new target for therapeutic intervention (49). The ribosome-binding site of Stxs has not been previously explored as a target for therapeutic development. If the A1 subunits bind ribosomes in a similar manner as Stx2a, the structure of the Stx2a-P11 complex could be exploited to design novel small molecule inhibitors that can bind to the toxin more potently than P11. During the development of HUS, STEC adhere to the intestinal mucosa and release Stx holotoxins, which are absorbed into the circulation and are targeted to the renal cells (50). Therefore, inhibitors that bind to Stx2a could neutralize the holotoxin released into the circulation before it distributes to target tissues.

Toxin and peptide
Stx2a was purchased from the Phoenix laboratory (Tufts University, Boston, MA). The P11 peptide was obtained from GenScript (Piscataway, NJ, USA). The sequence of the peptide was confirmed by HPLC and MS analysis.

Interaction of Stx2 with P11
The affinity of P11 for Stx2a was determined by surface plasmon resonance using a Biacore T200. Stx2a was immobilized on a CM5 chip to 4364 RU with amine coupling. The reference channel was activated and blocked the same way as the active channel. P11 was passed over both surfaces at six different concentrations (5,15,45,135,405, 1215 mM) for 30 s and dissociated for another 30 s. The running buffer was PBS-P (20 mM phosphate buffer, pH 7.4, 137 mM NaCl, 2.7 mM KCl, 0.05% surfactant P20) with 2% DMSO. The affinity (K D ) was determined by fitting the binding levels at steady state using the Biacore T200 Evaluation Software 3.0.

Crystallization and data collection
The Stx2a-P11 crystal was grown by sitting drop vapor diffusion at 20°C using a protein to reservoir volume ratio of 1:1 with total drop volumes of 0.2 ml. Crystals of the Stx2a-P11 complex were produced using crystallization buffer containing 100 mM Na acetate, pH 4.5, 100 mM NaCl, and 30% PEG 200. All crystals were flash frozen in liquid nitrogen after a short soak in the appropriate crystallization buffers supplemented with 20-25% ethylene glycol. Data were collected at the NYX beamline 19-ID at the National Synchrotron Light Source II, Brookhaven National Laboratory. All data were indexed, merged, and scaled using HKL2000 (51) then converted to structure factor amplitudes using CCP4 (52).

Structure determination and refinement
The Stx2a structure was solved by molecular replacement using the program Phaser (53). Molecular replacement calculations were performed using the coordinates of the Stx2a holotoxin (PDB ID: 1R4P) as a search model for Stx2a. The resulting phase information from molecular replacement was used to manually build residues 6 to 11 of the P11 peptide using the molecular graphics program COOT (54). Structural refinement of the Stx2a-P11 coordinates was performed using the PHENIX package (45). Three independent datasets were merged into the final dataset used for structural refinement of Stx2a-P11. During refinement a crossvalidation test set was created from a random 5% of the reflections. Data collection and refinement statistics are listed in Table S1. Molecular graphics were prepared using PyMOL (Schrodinger) (DeLano Scientific LLC, Palo Alto, CA, USA).

Data availability
The structure generated in this study was deposited in the Protein Data Bank under accession number 6X6H as described in Table S1.