The Crystal Structure of Shiga Toxin Type 2 with Bound Disaccharide Guides the Design of a Heterobifunctional Toxin Inhibitor

Background: E. coli Shiga like toxin type 2 (Stx2a) is responsible for serious clinical outcomes.

Results: The crystal structure of Stx2a with bound disaccharide was solved and used to design a potent toxin inhibitor.

Conclusion: The primary binding site of Stx2a is able to accommodate extended structural elements.

Significance: Knowledge of the toxin binding site can guide discovery of therapeutics to treat E. coli food poisoning.

Abstract

Shiga toxin type 2 (Stx2a) is clinically most closely associated with enterohemorrhagic E. coli O157:H7-mediated hemorrhagic colitis that sometimes progresses to hemolytic-uremic syndrome. The ability to express the toxin has been acquired by other Escherichia coli strains, and outbreaks of food poisoning have caused significant mortality rates as, for example, in the 2011 outbreak in northern Germany. Stx2a, an AB₅ toxin, gains entry into human cells via the glycosphingolipid receptor Gb₃. We have determined the first crystal structure of a disaccharide analog of Gb₃ bound to the B₅ pentamer of Stx2a holotoxin. In this Gb₃ analog, α-GalNAc replaces the terminal α-Gal residue. This co-crystal structure confirms previous inferences that two of the primary binding sites identified in the B₅ pentamer of Stx1 are also functional in Stx2a. This knowledge provides a rationale for the synthesis and evaluation of heterobifunctional antagonists for E. coli toxins that target Stx2a. Incorporation of GalNAc Gb₃ trisaccharide in a heterobifunctional ligand with an attached pyruvate acetal, a ligand for human amyloid P component, and conjugation to poly[acrylamide-co-(3-azidopropylmethacrylamide)] produced a polymer that neutralized Stx2a in a mouse model of Shigatoxemia.

Introduction

Enterohemorrhagic Escherichia coli have acquired numerous genes that deftly orchestrate colonization of the epithelial surface of the terminal ileum and transverse colon in humans (1). This process leads to an exaggerated host immune response contributing ultimately to the development of hemorrhagic colitis, a self-limiting infection, from which otherwise healthy patients recover once they mount an adaptive humoral response to the offending bacteria. However, ∼10–15% of patients develop hemolytic-uremic syndrome (HUS),⁴ characterized by the clinical triad of thrombocytopenia, thrombotic microangiopathy, and hemolytic anemia that can result in acute kidney failure. HUS is caused by Shiga toxin producing E. coli such as E. coli O157:H7 (2, 3).

Shiga toxins are AB₅ proteins consisting of a homopentameric B₅ subunit that binds to the carbohydrate portion of the glycolipid Gb₃ to gain entry into the cell, where the catalytic A subunit cleaves a specific adenosine residue of 28 S rRNA, thereby stopping protein production and resulting in cell death (4–6). There are two families of Shiga toxins, type 1 (Stx1) and type 2 (Stx2), that share ∼56% amino acid sequence identity (7). Stx1 and Stx2 are phage-encoded, and several strains of E. coli have acquired the genes to produce them, such as E. coli O104:H4, the bacterium responsible for the outbreak in northern Germany in 2011 that resulted in 48 deaths (8, 9). Of the two toxins, a member of the Stx2 family, Stx2a, has been associated with the most serious pathology and clinical outcomes (10–12).

The largest body of structural and binding data has been obtained for Stx1. A crystal structure of the radially symmetric B₅ pentamer of Stx1 with bound Gb₃ (13) analog (methoxycarbonyloctyl glycoside of P^k trisaccharide (P^kMCO), Fig. 1A) showed that each B-subunit monomer contains three binding sites for the glycan component of Gb₃, referred to as P^k trisaccharide, α-d-Galp-(1→4)-β-d-Galp-(1→4)-β-d-Glcp-(1-O) (Fig. 1B). These sites are numbered 1–3, for a total of 15 binding sites per pentamer (Fig. 1C, Protein Data Bank (PDB) entry 1BOS) (14). In this structure, site 2 has the highest occupancy with electron density defining the position of the trisaccharide. However, electron density for ligands in both sites 1 and 3 defines di- or monosaccharide residues. In fact, site 1 is the one with the lowest occupancy; in many positions only an α-Gal unit is displayed, and the lactose fragment is disordered. At least a galabiose disaccharide is defined in each site 3. NMR spectroscopy showed that site 2 of Stx1 displayed higher occupancy than site 1, and site 3 was barely occupied in solution (15). Fourier transform ion cyclotron resonance electrospray mass spectrometry (MS) experiments established that the intrinsic affinity of Stx1 site 2 for the P^k trisaccharide was ∼10-fold higher than for site 1 (16). To date, the only crystal structure of Stx2a was solved without bound ligand and revealed a radial topology similar to that of Stx1 (17). Binding studies by Fourier transform ion cyclotron resonance-MS studies suggested the nature of mono- and oligovalent P^k ligand binding interactions with Stx1 and Stx2a are similar and that site 2 of Stx2a is the primary binding site (18).

FIGURE 1.

A, P^kMCO (MCO, methoxycarbonyloctyl) used for crystallization of the previously solved Stx1 complex 1BOS. B, globotriaosylceramide (Gb₃), the natural occurring ligand for Shiga toxins, consisting of P^k trisaccharide α-d-Galp-(1→4)-β-d-Galp-(1→4)-β-d-Glcp-(1-O) and ceramide. C, the Stx1B₅-P^kMCO complex. The five B subunits are named B1 through B5. The bound carbohydrates are shown as spheres with oxygen and nitrogen atoms colored red and blue, respectively; the carbon atoms of the carbohydrate ligands bound in sites 1, 2, and 3 are colored in gray, slate, and gold, respectively.

Although the intrinsic affinity of Stx1 and Stx2a binding to Gb₃ is millimolar at best and among the lowest measured protein-carbohydrate interactions (18), toxin binding to cells expressing Gb₃ is translated into nanomolar avidity due to the fluidity of glycolipids in the lipid bilayer. This permits lateral translation of receptors into an array that allows for optimally spaced multivalent binding via the numerous saccharide binding sites of the B₅ pentamer (19). STARFISH, a decavalent ligand, achieved nanomolar avidity by matching ligand-receptor spacing and exploiting these multivalency effects. A crystal structure of the STARFISH-Stx1 complex showed a face-to-face complex of two B₅ pentamers of Stx1 by engaging all 10 site 2 binding sites (PDB entry 1QNU) (20). Although this result demonstrates the importance of the highest affinity site 2, our theoretical treatment of solution binding considered and simulated simultaneous engagement of sites 1 and 2 by each bivalent arm of the dendrimer (21).

Periodic serious outbreaks of disease caused by food chain contamination with E. coli-expressing Shiga toxins have prompted diverse approaches to counter the effects of the toxins. One approach envisages blocking toxin entry into the cell by engaging the Gb₃ binding sites on the B₅ pentamer. To be successful such approaches have to overcome the millimolar intrinsic affinity for Gb₃ analogues and, at least for oligosaccharide based antagonists, the short half-life of such polar molecules in circulation. Several interesting approaches and ligand designs have been reported (22–29).

The ligand-induced protein dimerization observed in the Stx1-STARFISH crystal structure was exploited to capture Stx1 as a supramolecular complex mediated by a heterobifunctional ligand termed BAIT-P^k (Fig. 2A) (30, 31). The in-register overlap of site 2 of Stx1 with those of a similarly radially symmetric protein, human serum amyloid P-component (HuSAP) allowed a compact bivalent ligand to anchor the two proteins together (31, 32). A multivalent version of this heterobifunctional ligand, termed PolyBAIT-P^k (Fig. 2A), prolonged circulation half-life of the ligand and not only mediated the formation of an Stx1-HuSAP supramolecular complex that blocked toxin entry into cells, but by virtue of the properties of HuSAP, the complex was also cleared from circulation via the liver (32).

FIGURE 2.

A, BAIT-P^k and PolyBAIT-P^k. Both ligands consist of P^k trisaccharide with a terminal glucose residue modified to contain the 1,2-O-pyruvate moiety necessary for interaction with HuSAP. In addition, PolyBAIT-P^k also consists of a linker allowing for polymeric conjugation and presentation. B, methyl-P^kNAc (1), shows the terminal galactose unit of P^k trisaccharide as a 2-acetamido-2-deoxy-galactopyranose, found to bind Stx2a selectively over Stx1. C, structure of disaccharide methyl glycoside 2 α-d-GalNAcp-(14)-β-d-Galp-(1-O) used in Stx2a crystallization. D, polyBAIT-P^kNAc 3 multivalent heterobifunctional ligand consists of P^kNAc trisaccharide for Stx2a selective binding and a 1,2-O-pyruvate acetal for interaction with HuSAP. The monomeric ligand is conjugated to a methacrylamide polymeric scaffold with 5% incorporation (n:m ∼1:19).

Structural homology and gas phase binding studies (18) have encouraged inferences about the possible binding sites present in Stx2a and guided mutagenesis studies that point to similarities with the binding sites of Stx1 (33). However, the absence of a solved structure of Stx2a with bound carbohydrate ligand has prevented informed decisions for the rational design of inhibitors for this more deadly toxin.

Recently, Kale et al. (34) reported the discovery and synthesis of a modified P^k trisaccharide whereby the terminal galactose was replaced with 2-acetamido-2-deoxy-α-d-galactose moiety at the non-reducing end (Fig. 2B, represented as the methyl glycoside Methyl-P^kNAc, 1). A solid-phase binding assay found that this P^k trisaccharide derivative was capable of binding Stx2a selectively over Stx1.

We report here a solved crystal structure of Stx2a with bound disaccharide 2 (Fig. 2C) and the activity of inhibitor 3 (Fig. 2D) of Shigatoxemia in transgenic mice designed based on inferences drawn from this crystal structure.

EXPERIMENTAL PROCEDURES

Synthesis of Compounds

Syntheses of compounds 1, 2, 4, and 5 are described in the supplemental Methods. Synthesis of multivalent, heterobifunctional compound 3 was recently described (35).

Crystallization and Structure Determination of Stx2a Holotoxin in Complex with the α-d-GalNAcp-(1→4)-β-d-Galp-(1-OMe) 2

The holotoxin, Stx2a, at a concentration of ∼5 mg/ml was mixed with a range of concentrations (2–10 mm) of compound 2 (dissolved in water) and incubated for 30 min at room temperature (25 °C). Initial crystallization conditions were identified through a robotic screen (Robbins Scientific) in which equivolume mixtures (0.3 μl each) of the Stx2a-ligand complex and the reservoir solution (Hampton Research, Crystal and Crystal 2 screen, Index screens) were equilibrated against 80 μl of the reservoir solution in a sitting drop format on a 96-well tray (Intelliplate). Small crystals were observed in several conditions, but only one (4 m sodium formate) gave reproducible crystal growth and was adopted for further refinement. For the production of crystals of diffraction quality, we used the hanging drop method and a reservoir solution containing 3.8 m sodium formate, 50 mm Tris, pH 7.0, 2% ethylene glycol, and 1 mm 3-(1-pyridino)-1-propanesulfonate (PPS). PPS is an additive that was previously reported to be conducive for the crystallization of Stx2a (17).

Diffraction data were collected on the CMCF-1 beamline at the Canadian Light Source (Saskatoon, Saskatchewan, Canada). Raw data were indexed and integrated using the program HKL2000 (36). The Stx2a crystals belong to the space group P6₁ with unit cell dimensions of a = b = 146.25 Å and c = 60.31 Å, γ = 120°. The structure of the Stx2a-disaccharide complex was solved via the molecular replacement method using the program Phaser (37) and the published Stx2 structure (PDB code 1R4P) as the search model. The B₅ pentamer and the A monomer were used separately as search models. Manual fitting using XtalView (38) and further refinement using Refmac5 (39) completed the tracing of the protein chains. Solvent molecules were added and refined using programs Arp/Warp (40) and Refmac5, respectively. At this time, it was clear that there were ligands bound in two of the carbohydrate binding sites of Stx2a. XtalView and Refmac5 were used to fit the disaccharide ligand into these sites, and several rounds of refinement were carried out using Refmac5. The final model was validated by the programs procheck (41) and molprobity (42).

PDB Accession Codes

Coordinates and structure factors of the Stx2a-2 complex have been deposited in the PDB with the accession code 4M1U.

Stx2 Binding Assay

Polystyrene (or polypropylene) microtiter plates were incubated with one of the glycoconjugates 4 or 5 (10 μg/ml, 100 μl/well) at room temperature overnight, then washed (5×) with PBST (0.05% Tween 20 in phosphate-buffered saline (PBS)). The plates were treated with either P^kNAc-BSA 4 or P^k-BSA 5. Blocking solution (1% BSA in PBS, 100 μl/well) was added, and the plates were incubated for 1 h at room temperature, then washed (5×) with PBST. Shiga toxin type 2 was applied to the plates at decreasing concentrations (starting concentration 1 mg/ml, dilution factor 3.16, 100 μl/well). After incubation at room temperature for 2 h, the plates were washed (5×) with PBST, incubated with anti-Stx polyclonal mouse antibodies in PBST (ATCC ascites #1907, 1 μl/ml; 100 μl/well) for 1 h at room temperature, then washed (5×) with PBST. Horseradish peroxidase-labeled anti-mouse antibody (KPL, 0.2 mg/ml) was applied, and the plates were incubated for 30 min at room temperature, then washed (5×) with PBST. A peroxidase substrate, 3,3′,5,5′-tetramethylbenzidine (TMB) with H₂O₂, was added. After 15 min the reaction was quenched by the addition of H₃PO₄ (1 m, 100 μl/well). The plates were read at 450 nm.

Stx2a HuSAP Transgenic Mouse Challenge Experiments

All animal work was conducted according to the guidelines of the Canadian Council on Animal Care with approval of the University of Calgary Health Sciences Animal Care Committee (Protocol #M07054). The transgenic mice used for these experiments were C57BL/6-Tg(APSC)1Imeg mice, which exhibit liver-specific expression of HuSAP at a stable circulating serum concentration of 30–40 μg/ml. These mice were first created by Zhao et al. (43) and were bred in-house. Groups (n > 10) of mice were intraperitoneally injected with 225 pg/g body weight affinity-purified Stx2a combined with 300 ng/g body weight E. coli O55 lipopolysaccharide (LPS; Sigma) or 225 pg/g body weight Stx2a combined with 300 ng/g body weight LPS and then immediately injected with PolyBAIT-P^k or PolyBAIT-P^kNAc. Animals were monitored every 2–4 h and euthanized by CO₂ asphyxia if signs of toxemia (lethargy) became apparent.

Statistical Analysis

The significance of differences between treatment groups in the LPS-sensitized HuSAP transgenic mouse Stx2 challenge experiments was determined from Kaplan-Meier survival plots using the LogRank test.

RESULTS

The Preferred Ligand for Stx2a

When 2-acetamido-2-deoxy-galactose replaces the terminal galactose of P^k trisaccharide to create P^kNAc (α-d-GalNAcp-(1→4)-β-d-Galp-(1→4)-β-d-Glcp-(1-O), this oligosaccharide shows higher avidity for Stx2a (34). To probe the binding site preferences of Stx2a and develop a convenient method for the in vitro analysis of Stx2a binding events, two ligands were synthesized and conjugated to BSA to provide glycoconjugates P^kNAc-BSA 4 and P^k-BSA 5 (supplemental Schemes S1 and S2) (Fig. 3, A and B). The binding of Stx1 and Stx2a to these conjugates was measured by a solid phase assay where glycoconjugate-coated microtiter plates capture toxin, which was detected by toxin-specific antibody (Fig. 4). Stx2a bound with a 275-fold higher avidity to P^kNAc-BSA 4 than to P^k-BSA 5, whereas Stx1 bound with 50-fold higher avidity to 5 than to 4.

FIGURE 3.

A, P^kNAc-BSA conjugate 4 (Stx2 specific). B, P^k-BSA conjugate 5 (Stx1-specific).

FIGURE 4.

Solid-phase binding of Stx1 and Stx2 to microtiter plates coated with glycoconjugates 4 or 5. Red circles, P^kNAc-BSA 4 versus Stx2 (EC₅₀ = 9 ng/ml); blue triangles, P^kNAc-BSA 4 versus Stx1 (EC₅₀ = 1660 ng/ml); black diamonds, P^k-BSA 5 versus Stx2 (EC₅₀ = 2480 ng/ml); green squares, P^k-BSA 5 versus Stx1 (EC₅₀ = 33 ng/ml). Error bars represent S.D. for triplicates.

The Crystal Structure of Stx2a with Bound Ligand 2

Stx2a was co-crystallized with a P^kNAc disaccharide methyl glycoside 2 (Fig. 2C), and the structure of the complex was determined by x-ray crystallography (Table 1). The structure of the Stx2a-2 complex aligned very well to that of the apoStx2a (PDB entry 1R4P) holotoxin with an root mean square deviation of 0.4 Å over 624 C^α atoms. Two molecules of 2 were found to bind to two distinct sites in Stx2a in the asymmetric unit (Fig. 5, A and B, Table 2). For the residues making contacts with 2 (distance < 4 Å), the root mean square deviation is 0.1 and 0.2 Å for sites 1 and 2, respectively. This suggests that little conformational change in Stx2a is required for the binding of disaccharide 2. Similar conclusions could be drawn from the structures of the apo (PDB codes 1C48, 1CQF, 1DM0) and P^kMCO-bound Stx1 (PDB code 1BOS), which is hardly surprising given the high sequence identity between the protein sequences of the two families of AB₅ toxins. As shown in the Stx1B₅-P^kMCO complex, there exist three distinct carbohydrate binding pockets per B protomer of the B₅ pentamer; site 1 is formed within each protomer, whereas sites 2 and 3 involve residues from two adjacent protomers (Fig. 1C). Although the binding sites are lined with identical or similar amino acid residues in Stx1 and Stx2a, compound 2 in our Stx2a-2 complex occupied only 2 of the 15 potential sugar binding sites (Fig. 6, A and B).

TABLE 1.

Crystallographic statistics

Data collection
Space group	P61
Cell dimensions
a, b, c (Å)	146.25, 146.25, 60.31
α,β,γ (°)	90, 90, 120
Wavelength (Å)	0.97949
Resolution (Å)	50.0-1.56 (1.65-1.56)a
Rsym (%)	8.2 (97.8)a
I(σ(I))	8.8 (1.4)a
Completeness (%)	92.8 (89.9)a
Multiplicity	5.7 (2.9)a
Refinement
Resolution (Å)	40.0-1.56 (1.60-1.56)a
No. observed reflections	99,448
Rwork/R–free (%)	17.0/19.2
No. total atoms	5798
No. protein atoms	4,971 (12.9)b
No. ligand atoms	93 (35.8)b
No. water	734 (29.0)b
R.m.s. deviations
Bond length (Å)	0.01
Bond angle (°)	0.99

^a Values in parentheses are for the highest resolution shell. Data collected on a single crystal were used for structure determination.

^b Values in parentheses indicate the average value of temperature factors in Ų.

FIGURE 5.

Carbohydrate binding sites in the B₅ pentamers of Stx1 and Stx2a toxins. A and B, two structural overviews of the Stx2a-2 complex. The A and five B subunits are shown in blue, salmon, lime, light purple, yellow, and cyan, respectively. The two molecules of compound 2 and three molecules of PPS (3-(1-pyridino)-1-propanesulfonate) bound to Stx2a are shown in space-filling spheres and are distinguished by green and yellow carbon atoms, respectively.

TABLE 2.

Occupancy of carbohydrate ligands at the 15 potential carbohydrate binding sites in Stx2a-2 complex

*, These sites were formed between two neighboring β subunits. The assignment of the site is based on the subunit that provides the most residues forming the binding pocket.

^&, Torsion angles in degrees (°).

FIGURE 6.

A, the σ^A-weighted 2‖F_o| − |F_c‖ electron density contoured at 1 σ surrounding the bound disaccharide methyl glycoside 2 in sites 1 (left) and 2 (right) of Stx2a. Neighboring B subunits are shown in different colors and the oxygen, nitrogen and carbon atoms of 2 are colored red, blue, and yellow, respectively. B, quality of the electron density map used in the structure determination of the Stx2a-2 complex. A representative σ^A-weighted, 2‖F_o| − |F_c‖ electron density map surrounding (1.6 Å from any non-hydrogen atom) the 17-amino acid fragment (Ala-281 to Lys-297) of the C terminus of subunit A is shown in a blue-meshed stereogram.

Binding site 1 in Stx2a consists of residues Lys-12, Asn-14, Asp-16, Thr-18, Thr-20, Glu-27, Trp-29, and Gly-59, all of which are conserved in Stx1 except for Trp-29, which is substituted by Phe-30. The interactions between disaccharide 2 and the residues constituting Stx2a site 1 are similar to those observed between P^kMCO and Stx1B₅; Gal1-NAc is bound at the “bottom” of site 1, and the next sugar, Gal2, packs against the aromatic rings of Trp-29 or Phe-30 (Fig. 7A). The carbonyl oxygen of the N-acetyl moiety of compound 2 receives a weak H-bond from the ϵ-amino group of Lys-12 and is also involved in a solvent-bridged polar interaction with the backbone amide nitrogen of Glu-15. Compound 2 makes more contacts with the residues of site 1 of Stx2a (51 interactions within 4 Å) than those between the corresponding disaccharide in P^kMCO and Stx1 (37 interactions within 4 Å) (Table 3 and Fig. 7A); however, this may not directly translate into differences in binding affinity as the strengths of these interactions have not been taken into account. Altogether, site 1 may not contribute significantly to the different affinities of Stx1 and Stx2a toward P^k and P^kNAc.

FIGURE 7.

A, interactions between Stx2a and compound 2 at site 1. The aligned complexes of Stx2a-2 and Stx1B₅-P^kMCO (root mean square deviation 0.5 Å over 292 C^α atoms) are distinguished by the carbon atoms colored in cyan and salmon in one B subunit and in yellow and purple in the adjacent B subunit, respectively. The bound carbohydrate ligands (tubes) the interacting protein residues (sticks). Compound 2 and P^kMCO are distinguished by gray- and orange-colored carbon atoms, respectively. The solvent molecules bridging the interactions between Stx2a and compound 2 are shown as red spheres. Hydrogen bonds are depicted in black and red dashed lines for the Stx2a-2 and Stx1B₅-P^kMCO complexes, respectively. B, interactions between Stx2a and 2 at site 2. The graphical representation and color scheme are the same as in A. A solvent molecule bridging the interactions between Gal2 and Gly61 in Stx2a-2 complex is shown as a red sphere. Those atoms in Stx2a within van der Waals contact distance (4 Å) from the disaccharide ligand 2 are shown as spheres. C, a stereogram of the NAc-specificity determinant in the site 2 of Stx2a. The Stx2a-2 and Stx1B₅-P^kMCO complexes are colored as described in B. The semi-transparent spheres show the van der Waals radii of the two oxygen atoms in short contact in a hypothetical Stx1B₅-P^k-NAc complex at site 2 (see “Results” and “Discussion”).

TABLE 3.

Comparison of the interactions between carbohydrate ligands and Stx1 (PDB code 1BOS) and Stx2a (this report)

	Stx1-PkMCO	Stx2a-2
Site 1
H-Bonds		Gal1NAc:O7-N:Glu-15 (2.7, 2.9)a
		Gal1NAc:O7-Nζ:Lys-12 (3.4)
		Gal1NAc:O3-Nζ:Lys-12 (2.9)
		Gal1NAc:O4-Oϵ1:Glu-27 (2.7, 2.6)
		Gal1NAc:O4-Oγ1:Thr-20 (2.9)
		Gal1NAc:O6-Oϵ1:Glu-27 (2.9)
	Gal1:O5-Oγ1:Thr-21 (3.2)b	Gal1NAc:O5-Oγ1:Thr-20 (3.1)
	Gal2:O6-Oδ2:Asp-17 (2.5)	Gal2:O6-Oδ2:Asp-16 (2.6)
	Gal2:O3-O:Gly-60 (2.6)	Gal2:O3-O:Gly-59 (2.8)
VDWc	37	51
Site 2
H-bonds	Gal1:O2-Oδ1:Asp-16 (2.8)	Gal1NAc:O7-Nη2:Arg-32 (3.1)
	Gal1:O3-Nη2:Arg-33 (2.8)	Gal1NAc:O3-Nη2:Arg-32 (3.2)
	Gal1:O4-Oδ1:Asn-32 (2.7)	Gal1NAc:O4-Oγ:Ser-31 (3.2)
	Gal1:O4-Nϵ:Arg-33 (3.0)	Gal1NAc:O4-Nϵ:Arg-32 (3.0)
	Gal1:O4-Nη2:Arg-33 (3.5)	Gal1NAc:O4-Nη2:Arg-32 (3.3)
	Gal1:O6-Oδ1:Asn-32 (3.0)	Gal1NAc:O6-Oγ:Ser-31 (3.3)
	Gal1:O6-N:Asn-32 (3.1)	Gal1NAc:O6-N:Ser-31 (2.9)
	Gal1:O6-O:Phe-63 (2.9)	Gal1NAc:O6-O:Phe-62 (2.7)
		Gal2:O2-N:Gly-61 (3.2, 2.9)d
		Gal2:O3-N:Gly-61 (3.0, 2.9)d
		Gal2:O5-Oγ:Ser-54 (3.2)
		Gal2:O6-Oγ:Ser-54 (2.8)
	Gal2:O6-N:Asn-55 (3.2)	Gal2:O6-N:Ser-54 (2.9)
		Gal2:O6-Oϵ2:Glu-15 (3.3)
VDW	46	59

^a Solvent-bridged interaction between the ligand and the toxin (the H-bond distance between residue 1 and water, the H-bond distance between residue 2 and water).

^b Distance in Å between the two atoms forming the hydrogen bond.

^c VDW, number of van der Waals contacts between the ligand and the toxin.

^d These solvent-bridged interactions were also observed in Stx1-starfish complex (PDB code 1QNU).

Binding of carbohydrate ligands to site 2 in Stx2a involves more polar and van der Waals contacts than those occurring between P^kMCO and Stx1 and those occurring between compound 2 and Stx2a at site 1 (Table 3). Site 2 consists of residues Trp-29, Ser-31, Arg-32, Ser-53, Ser-54, Thr-55, Gly-61, Phe-62, Ala-63 of B5 and Glu-15 from the neighboring B1 subunit. The binding modes of compound 2 and P^kMCO in site 2 are similar, and the two small molecule ligands are virtually superimposable in the structurally aligned Stx2a and Stx1 (Fig. 7B). Arg-32, which forms bifurcated H-bonds with O3 and the acetyl oxygen of Gal1-NAc, adopts very similar conformations in the Stx1-P^kMCO and Stx2a-2 complexes. However, in contrast to site 1, there are significant differences between site 2 of Stx1 and Stx2a in addition to the substitution of Phe-30 by Trp-29. The residues in Stx1 that are equivalent to Glu-15, Ser-31, Ser-54, and Thr-55 in Stx2a are Asp-16, Asn-32, Asn-55, and Ala-56, respectively. The change from Glu-15 (Stx2a) to Asp-16 (Stx1) seems most relevant. In the Stx1B₅-P^kMCO complex, the side chain carboxylate oxygen O^δ1 of Asp-16 accepts a 2.8 Å H-bond from O₂ of Gal1; its location would most likely lead to a destabilizing short contact (1.5 Å) with the carbonyl oxygen if the 2-substituent is present as an N-acetyl function (Fig. 7C). On the other hand, the side chain of Glu-15 in the Stx2a-2 complex points away from the bound Gal1-NAc moiety and forms a weak H-bond (3.4 Å) with O6 of Gal2, which in turn forms an H-bond (2.8 Å) with the O^γ of Ser-54; both of these interactions have no counterparts in the Stx1B₅-P^kMCO complex. The major difference in the conformation of these two acidic residues lies in the placement of their side chains; the χ1 torsion angle of Glu-15 (Stx2a) adopts a gauche (−) conformation (−76 °), and the equivalent angle for Asp-16 (Stx1) adopts a trans conformation (−172 °). If the side chain of Asp-16 (Stx1) were to take on a gauche (−) conformation as observed for that of Stx2a Glu-15, its side chain carboxylate would fall short of making contact with the 6-OH of Gal2 in a bound P^kNAc-like ligand. It is interesting to note that the respective conformations of these two residues are preserved in all presently available crystal structures of Stx1 and Stx2a toxins with either ligand-bound or unoccupied site 2. In particular, Asp-16 of Stx1 is “locked” in the observed conformation through its interactions with Ser-64 and Arg-33; the main chain carbonyl of Asp-16 accepts two H-bonds from the guanidino group of Arg-33, and O^γ of Ser-64 donates an H-bond to O^δ1 of Asp-16 that receives another H-bond from N^η2 of Arg-33. The residue in Stx2a equivalent to Ser-64 in Stx1 is Ala-63; thus, only the H-bonds between the O of Glu-15 and the guanidino group of Arg-33 are conserved between Stx1 and Stx2a.

Specific Heterobifunctional Ligand That Sequesters Stx2a

The crystal structure data showing the orientation of disaccharide 2 in site 2 of Stx2a suggest that a heterobifunctional ligand similar to the BAIT-P^k molecule used to engage Stx1 with HuSAP (30) should also be a viable proposition for binding and neutralization of Stx2a. We synthesized the target compound PolyBAIT-P^kNAc 3, in which 2-acetamido-2-deoxy-α-d-galactose replaced the terminal α-Gal of P^k trisaccharide (35). It consists of three distinct features essential for supramolecular complex formation (Fig. 2D); an Stx2a binding P^kNAc trisaccharide modified with a 1,2-O-pyruvate acetal on the glucose residue as a ligand for HuSAP and a linker with a terminal alkyne group for attaching the monovalent ligand to a polymer. The polymeric format affords two essential benefits; it slows clearance of the agonist from circulation, and it provides additional avidity gains via multivalency (19, 32). Our prior work indicates linear polymers are preferred over dendrimeric counterparts as they allow for a peripheral rather than radial topology of binding, thereby potentially eliminating any unwanted steric interactions (30–32).

Compound 3 was tested in a modified HuSAP transgenic mouse model of Shigatoxemia that more accurately reflects Stx2a-initiated HUS in humans (Fig. 8). In this modified model, transgenic (HuSAP-tg) mice are pre-exposed to LPS to induce a sub-acute state of inflammation that ablates the confounding effects of HuSAP on toxicity of Stx2a (45). HuSAP-tg mice (C57BL/6-Tg(APSC)1Imeg) treated intraperitoneally with a lethal dose of Stx2a (225 pg/g mouse) combined with E. coli O55 LPS (300 ng/g mouse) were immediately given increasing intravenously concentrations of PolyBAIT-P^k (Fig. 2A) or PolyBAIT-P^kNAc 3 (Fig. 2D). Non-transgenic mice and HuSAP-tg mice pretreated with Stx2a and LPS, but not PolyBAIT-P^k or PolyBAIT-P^kNAc 3, succumbed to infection within 100 h post injection. Mice treated with PolyBAIT-P^k showed modest signs of protection albeit with ∼60% of mice surviving over the course of the experiment. In contrast, HuSAP-tg mice receiving PolyBAIT-P^kNAc 3 at two concentrations (100 or 31.5 μg/g mouse) showed >80% survival (p < 0.001) over 150 h. The difference (p = 0.00143) at the 31.5 μg/g body weight dose in protection between PolyBAIT-P^k and PolyBAIT-P^kNAc 3 can be attributed to selective binding of PolyBAIT-P^kNAc to Stx2a. The lowest effective concentration of PolyBAIT-P^kNAc was 31.5 μg/g mouse. In previous studies we showed that PolyBAIT-P^k protects HuSAP-tg mice against Stx1 at concentrations ∼10-fold lower (3.15 μg/g mouse), but treatment with PolyBAIT-P^k at a concentration as high as 100 μg/g mouse afforded only modest protection against Stx2a infection (32).

FIGURE 8.

Survival graph for in vivo mouse model of disease after challenge with a lethal dose of Stx2a. Gray line, wild-type mice, Stx2a, and O55 LPS, n = 15; black line, Stx2a, O55 LPS, and HuSAP, n = 21; blue line, Stx2a, O55 LPS, HuSAP, and PolyBAIT-P^k (100 μg/g), n = 15; solid red line, Stx2a, O55 LPS, HuSAP, and PolyBAIT-P^kNAc 3 (100 μg/g), n = 11; dotted red line, Stx2a, O55 LPS, HuSAP, and PolyBAIT-P^kNAc 3 (31.5 μg/g), n = 10; solid green line, Stx2a, O55 LPS, HuSAP, and PolyBAIT-P^kNAc 3 (10 μg/g), n = 5; dotted green line, Stx2a, O55 LPS, HuSAP, and PolyBAIT-P^kNAc 3 (3.15 μg/g), n = 5. n represents the number of HuSAP transgenic mice used for each set of experiments.

DISCUSSION

In vitro analysis of the synthesized glycoconjugates P^kNAc-BSA 4 and P^k-BSA 5 provided an interesting insight into Stx2 binding. P^kNAc-BSA 4 was selective for Stx2, and P^k-BSA 5 was selective for Stx1. This result is consistent with the suggestion of Flagler et al. (33) that a specific interaction involving the acetamido group of P^kNAc leads to differential binding between the two toxins.

The crystal structure of the Stx2-2 complex was solved by co-crystallization of Stx2 with the synthesized ligand 2. Analysis of the structure showed that only two molecules of disaccharide 2 were found in complex with Stx2; one ligand in site 1 and the other in site 2, whereas the remaining carbohydrate binding sites were unoccupied. The lack of carbohydrate binding to the remaining sites could be attributed to a variety of reasons: (a) steric hindrance from the C terminus of the A subunit of the holotoxin protruding from the central hole in the B₅ pentamer; (b) the presence of another small ligand used as a crystallization additive; (c) conformational changes in key residues forming the binding site (Trp-33); (d) crystal packing (Table 1). Except for some sites 3 as mentioned above in (a), the lack of observed binding of compound 2 to other sites is likely the fortuitous result of the crystallization conditions. It is possible that in solution or in vivo, these other sugar binding sites in Stx2a are capable of accommodating compound 2 or similar carbohydrate ligands.

Of the two binding sites found to possess ligand 2, site 2 was found to form more polar and van der Waals interactions with 2 than site 2 in either Stx1 or the previously solved Stx1-MCO complex. Most notably, Glu-15 was found to form a unique weak H-bond with O6 of Gal, which in turn forms an hydrogen bond with O^γ of Ser-54. These interactions are unique to Stx2a and are not found in either Stx1 or the Stx1-P^kMCO complex. Another residue in site 2, Ser-31 in Stx2a as opposed to Asn-32 in Stx1, may also play a role in the preferential binding of carbohydrate ligands by these two Shiga toxins. Flagler et al. (33) showed that point mutations N32S in Stx1 and S31N in Stx2a, residues that share the same approximate location in the binding sites, resulted in a reversal of selectivity; P^kNAc showed preferential binding to the mutated Stx1, whereas P^k preferentially bound to mutated Stx2a. However, the proposition that steric hindrance between an asparagine residue at this position and Gal1NAc might prevent the binding of P^k-NAc to site 2 is not convincingly supported by the structural data presented here. Furthermore, Stx2e, a shiga toxin causing edema in pigs, has a preference for Gal1NAc albeit possessing an asparagine at the equivalent position. Therefore, residue Glu-15, rather than Ser-31, of Stx2a seems to be the most important determinant of ligand specificity in site 2.

The heterobifunctional ligand designed to bind Stx2 selectively coined PolyBAIT-P^kNAc 3 (35) incorporates the P^kNAc trisaccharide to which is fused a 1,2-O-pyruvate acetal. When this entity is attached to a polymeric scaffold, the multivalent heterobifunctional ligand forms a supramolecular complex where the radially symmetric HuSAP acts as a template protein that sequesters Stx2a.

Evaluation of this polymeric inhibitor of Stx2a binding to verocell in vitro is masked by a document interaction of HuSAP with Stx2a. HuSAP binds to and neutralizes Stx2a in the absence of glycan ligands, and based on the observation that HuSAP transgenic mice were completely resistant to Stx2a, this interaction was initially considered to be of potential therapeutic value in preventing Stx2a-initiated HUS in E. coli O157:H7-infected patients (44). However, it was subsequently revealed that there was no correlation between the circulating concentration of HuSAP and the risk for developing HUS in humans (45).

In the presence of LPS, the interference of HuSAP with protection studies is ablated, and this allowed PolyBAIT-P^kNAc 3 to be evaluated in vivo. HuSAP-tg mice pre-exposed to LPS after challenge with a lethal amount of Stx2a were protect from the lethal effect of Stx2a at concentrations of 3 as low as 31.5 μg/g mouse.

We believe it is highly unlikely the protective effect of PolyBAIT-P^kNAc 3 in LPS-sensitized HuSAP transgenic mice is indirectly related to anything, i.e. anti-inflammatory activity, other than its ability to form supramolecular complexes involving HuSAP and Stx2a because the analogous PolyBAIT-P^k protects HuSAP transgenic mice from Stx1-mediated Shigatoxemia in the absence of inflammation. The only way for this to occur is for PolyBAIT-P^k to coordinate the formation of supramolecular complexes involving HuSAP and Stx1 because HuSAP does not bind to and neutralize Stx1 in vitro or in vivo (32).

The ability of 3 to protect HuSAP-tg mice after a lethal dose of Stx2a establishes that P^kNAc is a lead compound for the development of more active inhibitors of Stx2a. To our knowledge this represents the first and only report of an effective inhibitor based on structural evidence for the clinically more relevant toxin (Stx2) involved in E. coli O157:H7-mediated HUS.

It is important to emphasize that the work presented here establishes a base line for moving toward a therapeutic treatment of HUS, but it is not a drug discovery disclosure. Given our current understanding of the complexities of host-related factors in Stx2-initiated pathogenesis, it is quite possible that therapeutic potential of PolyBAIT-P^kNAc in preventing HUS from developing in E. coli O157:H7-infected subjects may be quite limited after these individuals present in the emergency room with hemorrhagic colitis. This is not to say, however, that administration of PolyBAIT-P^kNAc or another suitable compound to these patients would not ameliorate the spectrum of HUS symptoms, thereby shortening time-to-recovery and possibly the long term sequelae of the condition.

Moreover, it is well established that, due to the low infectious dose, person-to-person spread of the infection is quite common in sporadic occurrences of E. coli O157:H7 or other enterohemorrhagic E. coli serotypes (46). It is not unreasonable to suggest that a prophylactic application such as that evaluated in the LPS-sensitized HuSAP transgenic mouse experiments presented here to individuals at risk for developing the infection could be beneficial especially in large outbreak situations like the 2011 E. coli outbreak O104:H4 in Germany (8, 9).

Several possibilities evolve from this work. Although polyacrylamide was employed here, multivalent agonists employing a biocompatible polymer such as copovidone, approved for oral use as an excipient (47), could be employed to complex toxin released in a combination therapy with antibiotics. Perhaps even more attractive is the discovery of univalent low molecular weight antagonists after screening of focused libraries informed by the solved structure of the Stx2a-2 complex (48).