Binding of Pk-Trisaccharide Analogs of Globotriaosylceramide toShiga Toxin Variants

Hailemichael O. Yosief,a Suri S. Iyer,b Alison A. Weissc

University of Cincinnati, Department of Chemistry, Cincinnati, Ohio, USAa; Georgia State University, Department of Chemistry, Center for Diagnostics and Therapeutics,Atlanta, Georgia, USAb; University of Cincinnati, Department of Molecular Genetics, Biochemistry, and Microbiology, Cincinnati, Ohio, USAc

The two major forms of Shiga toxin, Stx1 and Stx2, use the glycolipid globotriaosylceramide (Gb3) as their cellular receptor. Stx1primarily recognizes the Pk-trisaccharide portion and has three Pk binding sites per B monomer. The Stx2a subtype requiresglycolipid residues in addition to Pk. We synthesized analogs of Pk to examine the binding preferences of Stx1 and Stx2 subtypesa to d. Furthermore, to determine how many binding sites must be engaged, the Pk analogues were conjugated to biotinylatedmono- and biantennary platforms, allowing for the display of two to four Pk analogues per streptavidin molecule. Stx binding toPk analogues immobilized on streptavidin-coated plates was assessed by enzyme-linked immunosorbent assay (ELISA). Stx1,but not the Stx2 subtypes, bound to native Pk. Stx2a and Stx2c bound to the Pk analog with a terminal GalNAc (NAc-Pk), whileStx1, Stx2b, and Stx2d did not bind to this analog. Interestingly, the purified Stx2d B subunit bound to NAc-Pk, suggesting thatthe A subunit of Stx2d interferes with binding. Disaccharide analogs (Gal�1-4Gal, GalNAc�1-4Gal, and Gal�1-4GalNAc) didnot support the binding of any of the Stx forms, indicating that the trisaccharide is necessary for binding. Studies with monoan-tennary and biantennary analogs and mixtures suggest that Stx1, Stx2a, and Stx2c need to engage at least three Pk analogues foreffective binding. To our knowledge, this is the first study examining the minimum number of Pk analogs required for effectivebinding and the first report documenting the role of the A subunit in influencing Stx2 binding.

Shiga toxin is responsible for the life-threatening condition he-molytic-uremic syndrome (HUS). HUS is characterized by

thrombocytopenia, hemolytic anemia, acute renal failure, andneurologic symptoms (1–5). Shiga toxins are members of the AB5

toxin family, consisting of a single A subunit and a pentamer ofidentical noncovalently associated B subunits. The A subunit me-diates an RNA N-glycosidase activity that cleaves an adenine nu-cleotide from the 28S rRNA, leading to protein synthesis inhibi-tion in the cell and ultimately cellular death (6, 7). The B subunitis the binding part of the toxin; it binds to the cell surface receptor,typically the neutral glycolipid globotriaosylceramide (Gb3), andmediates delivery of the A subunit to the cytoplasm of the cell(8–10). The glycans on Gb3, termed the Pk-trisaccharide, play amajor role in Shiga toxin binding.

Two major forms of Shiga toxin, Stx1 and Stx2, share approx-imately 60% amino acid sequence identity but do not elicit cross-neutralizing antibodies. In addition, Stx2 has a number of sub-types that are over 90% identical. Recognized subtypes of Stx2include Stx2a through Stx2g (11–13). The highly potent prototypeform of Stx2 produced by Escherichia coli O157:H7 is now oftencalled Stx2a for clarity. Stx2a, Stx2c, and Stx2d are closely relatedand associated with human disease (Fig. 1). Strains of E. coli canexpress one or more Shiga toxin types, but epidemiological studieshave indicated that strains that produce Stx2a are more com-monly associated with HUS than are strains that produce Stx1(14–16). Strains producing only the Stx2c and Stx2d subtypeshave occasionally been implicated in the development of hemor-rhagic colitis and HUS (17). However, Shiga toxin subtypes differwidely in potency in both cellular and in vivo studies (12, 18–20).While the purified Stx2a and Stx2d subtypes are highly potent inmice (50% lethal doses [LD50s], �10 ng), Stx1, Stx2b, and Stx2care not potent (LD50s, �1,000 ng) (18). In two separate studies,baboons treated with Stx2a displayed more severe disease symp-toms than did animals treated with Stx1 (21, 22). Stx2b and Stx2e

to Stx2g are less closely related (18, 19) and are rarely isolated fromhuman clinical samples. Stx2e binds preferentially to globotetra-osylceramide, as opposed to Gb3, and it causes edema disease inpiglets (23–26).

Since Stx1 and Stx2a display indistinguishable RNA N-glyco-sidase activities in cell-free systems, differences in binding prop-erties have been proposed to account for the difference in toxicitybetween Stx1 and Stx2a (14, 27–30). However, studying the finedetails of receptor binding is challenging. Unlike typical protein-carbohydrate interactions, where a single binding site engages asingle glycan, binding of Shiga toxin is the sum of multiple inter-actions, originating from two distinct mechanisms, protein-gly-can interactions at the binding site and avidity, or the ability toenhance binding by simultaneously engaging multiple bindingsites on the toxin. For example, Stx1 incubated with a single Pk-trisaccharide displays a dissociation constant (Kd) in the millimo-lar range, and in some studies, binding to Stx2 was 10-fold lessavid (31–33). In contrast, both toxins bind to synthetic Pk glyco-clusters (e.g., Starfish and Daisy) (34), linear polymers of Pk-tri-saccharide (35), and the silicon-based dendrimers of the trisac-charide (36) in the micro- and nanomolar ranges. In practice,weak affinity at the binding site can be masked by the engagementof multiple binding sites, making it difficult to define the besttoxin receptor. Even though these synthetic receptor mimics have

Received 1 March 2013 Returned for modification 28 April 2013Accepted 12 May 2013

Published ahead of print 20 May 2013

Editor: S. R. Blanke

Address correspondence to Alison A. Weiss, alison.weiss@uc.edu.

Copyright © 2013, American Society for Microbiology. All Rights Reserved.

doi:10.1128/IAI.00274-13

August 2013 Volume 81 Number 8 Infection and Immunity p. 2753–2760 iai.asm.org 2753

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

been designed to act as competitive inhibitors and block Shigatoxin from binding to its natural cellular receptors, none have yetproven to be effective in clinical trials (37–39), possibly becausethe more potent toxin, Stx2a, does not have a high affinity for Pk.In addition, cellular toxicity occurs in the picomolar range (18),suggesting that in vitro systems do not replicate all aspects of invivo interactions.

More is known about how Stx1 interacts with its receptor thanabout how Stx2 interacts with its receptor. The crystal structure ofStx1 revealed 3 Pk-trisaccharide binding sites per monomer, for atheoretical total of 15 sites in the pentamer (40). However, each ofthe three binding sites displays a different affinity for Pk and it ispossible that some sites prefer a glycan other than Pk. This issupported by a recent study where Gb3 mixed with other neutralglycolipids bound Stx1 and Stx2a better than did Gb3 alone (33).Most studies examine binding to only a single Pk mimic and thusassess primarily the contribution of the five identical sites with thehighest affinity for the receptor mimic being examined, leavingunexplored the potential contribution of engagement of the otherbinding sites.

In contrast to Stx1, it is not clear how many binding sites arepresent on the Stx2 B monomer. Stx2 binds Pk-trisaccharideweakly, and recent studies suggest that Stx2a recognizes portionsof the ceramide of Gb3 in addition to Pk and prefers to bind toGb3 in the presence of cholesterol (33, 41, 42). In addition, whilemost, if not all, of the binding of Stx1 is thought to be mediated bythe B pentamer, the C terminus of the A subunit of Stx2 is dis-played on the binding face of the B pentamer and can influencebinding. For example, proteolytic removal of the two C-terminalamino acids (GE) of Stx2d has been reported to increase the po-tency of crude toxin extracts (20) and confer a modest increase inpotency on purified, elastase-treated Stx2d (18).

In this study, we synthesized a series of synthetic glycoconju-gates designed to independently assess the contributions of bind-ing site affinity and avidity (Fig. 2). Pk-trisaccharide and NAc-Pk-trisaccharide analogs of Gb3 can differentiate between Stx1 andStx2a (43). Di- and trisaccharides with and without an N-acetylgroup on the terminal or penultimate galactose unit allow forunderstanding of the contributions of individual glycans in thetrisaccharide. Mono- and biantennary constructs allow for under-standing of the role of avidity, with streptavidin display playing a

key role in avidity assessment. Streptavidin has four biotin bind-ing sites, but previous studies have demonstrated that a singlemolecule of Shiga toxin can access only Pk mimics immobilized inthe two adjacent biotin binding sites because of spatial constraints(44). Thus, monoantennary constructs immobilized on streptavi-din display two Pk analogues while biantennary constructs displayfour (Fig. 2A). In addition, mixing of different Pk analogues fordisplay on streptavidin can also provide insight into synergisticinteractions between binding sites with different binding prefer-ences. In this study, we used these constructs to examine the bind-ing preferences of the Shiga toxin subtypes commonly associatedwith human disease and their purified B subunits. We found thatStx2 subtypes display significant differences in glycoconjugatepreferences, and these studies have important implications for thedesign of receptor mimics for diagnostics and therapeutics.

MATERIALS AND METHODSGlycoconjugates. The synthesis of the biotinylated glycoconjugates usedin this study has been described previously (45). Biotinylated polyethyleneglycol (PEG; molecular weight, 3,400) was purchased from Laysan Bio(Arab, AL).

Antibodies. Rabbit polyclonal antibodies to Stx1 and Stx2 were ob-tained from Meridian Bioscience (Cincinnati, OH). Peroxidase-conju-gated goat anti-rabbit IgG was purchased from MP Biomedicals (Solo,OH). Chicken IgY against the Stx2 B subunit and peroxidase-conjugatedrabbit anti-chicken IgY were purchased from Lampire Biological Labora-tories (Ottsville, PA). A mouse monoclonal antibody against the Stx2 Asubunit (11E10) was obtained from the National Cell Culture Center, anda peroxidase-conjugated goat anti-mouse antibody was purchased fromICN Biomedicals (Aurora, OH).

Shiga toxin holotoxins. Shiga toxin-containing supernatants wereproduced as previously described (43, 46). Toxin concentrations weredetermined by quantitative Western blot assays with mouse monoclonalantibody 11E10 to the Stx2 A subunit and a rabbit polyclonal antibody toStx1. The purified Stx1 and Stx2 A subunits were used as standards inWestern blot assays. Toxin activities and concentrations were verified in aprotein synthesis inhibition assay with luciferase-conjugated Vero mon-key kidney cells.

Purified B subunit. The Stx2 B wild-type (WT), Stx2 B Q40L mutant,and Stx2d B WT subunits were purified as previously described (47).Protein purity was verified by a single band at 8 kDa, corresponding to themolecular mass of a B subunit monomer on a Coomassie-stained SDS-

FIG 1 Sequence alignment of Stx1 and Stx2 subtypes. Sequence alignment was done by using NCBI BLAST. Dots indicate identity, and dashes indicate absentamino acids. (A) Alignment of the C-terminal 20 of the 297 amino acids of the A subunit, corresponding to the region that traverses the B pentamer and influencesreceptor binding. (B) Alignment of the B subunits. Numbers correspond to the amino acid positions in the mature polypeptide of Stx2a.

Yosief et al.

2754 iai.asm.org Infection and Immunity

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

PAGE gel. The concentrations of the B subunits were determined by bi-cinchoninic acid protein assay.

Binding assays and ELISAs. For enzyme-linked immunosorbent as-say (ELISA) studies, biotinylated glycoconjugates were dissolved inphosphate-buffered saline (PBS) and added in excess (6.25 �M) tocommercially available high-capacity (125 pM) streptavidin-coatedmicrowell plates (Thermo Scientific), and binding was allowed to pro-ceed for 2 h at room temperature. The wells were washed three timeswith PBS (pH 7.4) containing 0.1% bovine serum albumin and 0.5%Tween 20 and incubated with Shiga toxin containing supernatants at37°C for 1 h. The maximum toxin concentration used was 1.40 �M.Toxin binding was detected with rabbit polyclonal antibodies (againstStx1 and Stx2) and peroxidase-conjugated goat anti-rabbit IgG.QuantaBlu fluorogenic peroxidase substrate (Pierce) was used to de-velop the plates, and the values were determined with a fluorometer.The same protocol was used in the purified Stx2B subunit bindingstudy, except that chicken IgY and peroxidase-conjugated rabbit anti-chicken IgY were used as the primary and secondary antibodies todetect binding. Binding curve determination and analysis were per-formed by Prism 5.0 (GraphPad Software, La Jolla, CA).

RESULTS AND DISCUSSION

Several studies have shown that Stx1 binds to the Pk-trisaccharide(Gal�1-4Gal�1-4Glc) portion of Gb3. However, some studieshave indicated that Stx2 can bind to the Pk-trisaccharide while inother studies it bound weakly or not at all (27, 32, 48, 49). Tofurther assess Shiga toxin binding preferences, we generated apanel of synthetic biotinylated glycoconjugates (Fig. 2). In previ-ous studies, while Stx1 bound to an analogue displaying the nativePk-trisaccharide, Stx2 did not bind (43, 44). Instead, Stx2 boundto a synthetic variant, NAc-Pk (GalNAc�1-4Gal�1-4Glc), whichdid not engage Stx1. In this study, synthetic disaccharides were

generated to examine the contributions of the individual glycansto binding site recognition. In addition, avidity (the additive effectof engaging multiple binding sites), which is important in Shigatoxin binding, was examined. The disaccharide and trisaccharidePk analogs were attached to either a monoantennary biotinylatedscaffold or a biantennary biotinylated scaffold and immobilizedon streptavidin-coated microtiter plates via the biotin moiety.Streptavidin is a tetramer with two closely spaced biotin bindingsites on each face of the molecule. While streptavidin has fourbiotin binding sites, we have demonstrated previously that a singlemolecule of Shiga toxin can access Pk analogues at only two adja-cent biotin binding sites because of spatial constraints (44). Thus,the biantennary glycoconjugates present four potential Pk ana-logues and the monoantennary glycoconjugates present two po-tential Pk analogues (Fig. 2A).

Holotoxin binding to biantennary trisaccharides. The bind-ing of Stx1 and Stx2 (Stx2a, -b, -c, and -d) to biantennary Pk andNAc-Pk was examined (Fig. 3, inserts). As reported previously(43, 44), Stx1 bound to Pk but Stx2a and Stx2c did not. Stx2b andStx2d did not bind to Pk either (Fig. 3A). In addition, consistentwith previous studies (43, 44), Stx1 did not bind to NAc-Pk butStx2a and Stx2c bound to NAc-Pk (Fig. 3B). To determine theapparent dissociation constants (Kds), the binding curves werefitted to a one-site-specific binding model using the Hill coeffi-cient. The apparent Kd of Stx1 binding to Pk was found to be 18nM (Fig. 3A), whereas the apparent Kds of Stx2a and Stx2c bindingto NAc-Pk were found to be 361 and 36 nM respectively (Fig. 3B).The apparent Kd values of the toxins for their preferred receptoranalogues fall within the same range (10�7 to 10�9 M) as the Kd

values in previous reports (33, 44, 48).

FIG 2 Biotinylated monoantennary and biantennary glycoconjugates. (A) Monoantennary (top) and biantennary (bottom) biotinylated constructs. Cartoonsrepresent glycoconjugate display on streptavidin. (B) Glycan structures. Cartoon representations use the standard symbol nomenclature proposed by theConsortium for Functional Glycomics: open circles, galactose (Gal); open squares, N-acetylgalactosamine (GalNAc); black circles, glucose (Glc).

Shiga Toxin Binding to Receptor Analogues

August 2013 Volume 81 Number 8 iai.asm.org 2755

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

Interestingly, neither Stx2b nor Stx2d bound to NAc-Pk glyco-conjugates (Fig. 3B). The B subunit of Stx2b displays several poly-morphisms compared to Stx2a, all of which involve the putative Pkbinding residues; thus, it was not surprising that Stx2b did not bind toNAc-Pk. On the other hand, the B subunit of the highly potent vari-ant of the Stx2d subtype characterized in this study differs from Stx2aby only a single, conservative amino acid change (I51V) (Fig. 1),which is present on the face of the B subunit away from the receptor-binding interface. However, both Stx2b and Stx2d possess an acidicglutamic acid residue at the C terminus of the A subunit, while Stx2aand Stx2c possess the basic amino acid lysine. The crystal structure ofStx2a shows that the terminal amino acid of the A subunit is displayedon the binding face of the B subunit (19), and this charge changecould affect binding.

Purified B subunit binding to biantennary trisaccharides. Todetermine if the C-terminal region of the A subunit interferes withStx2d binding, purified B subunits of Stx2a and Stx2d were assessedfor binding to NAc-Pk. Contrary to what was seen with the holotoxin,the B subunits of Stx2a and Stx2d bound almost equally to NAc-Pk(Fig. 4), suggesting that the C-terminal region of the Stx2d A subunitinterferes with holotoxin binding. Compared to the holotoxin, lessbinding was observed with the B subunits, likely because pentamer-ization is required for efficient binding, and the Stx2 B subunit formsa stable pentamer only at concentrations greater than 2 �M (47). Inprevious studies, the NAc-Pk analogue immobilized on gold nano-particles was shown to neutralize Stx2d in a cell-based assay (50);however, neutralization of Stx2d was much less efficient than neutral-ization of Stx2a. The density of glycoconjugate display was muchgreater on gold nanoparticles, suggesting that avidity could compen-sate for poor binding site recognition, although presentation couldalso play a role.

We also examined the binding of the Q40L mutant Stx2a Bsubunit described in a previous report (47). That study demon-

strated that the B pentamer of Stx1 is rather stable (Kd � 43 nM)and the B pentamer of Stx2a is much less stable (Kd � 2,300 nM).The stable Stx1 B pentamer has a hydrophobic leucine at the hy-drophobic interface of subunits at the B pentamer, and Stx2 insta-bility is due to the presence of a polar glutamine instead of aleucine at the hydrophobic interface. The Q40L mutant Stx2a Bsubunit, with the destabilizing glutamine substituted for the sta-bilizing leucine, displayed increased B pentamer stability (Kd �110 nM). The Q40L mutant form bound better than either theStx2a or the Stx2d WT B subunit (Fig. 4). Binding of the Q40Lmutant B subunit, but not WT Stx2a, is observed at around 1,200nM, which is well above the Kd for pentamer formation for theQ40L mutant form but below that for the WT Stx2a B subunit.These studies suggest that binding requires the multivalent effectafforded by the pentameric form.

Holotoxin binding to biantennary disaccharides. To investi-gate the contributions of the three sugar residues of Pk to binding,we examined the binding of a panel of biantennary disaccharides,including Gal�1-4Gal, Gal�1-4GalNAc, and GalNAc�1-4Gal(Fig. 2B). None of the disaccharides supported toxin binding(data not shown). The crystal structure of Stx1B in complex withthe Pk-trisaccharide demonstrates that most of the hydrogenbonding and the stacking interactions occur between the terminalgalactose units and the amino acid residues of all three bindingsites (40), and binding to glycoconjugates with terminal digalac-tose (Gal�1-4Gal) units was observed in previous studies (8).However, in studies using isothermal titration calorimetry (ITC),Stx1 was shown to bind to the Pk-trisaccharide but not to eitherGal�1-4Gal or lactose (31). Since ITC examined binding to thefree glycans, the contribution of avidity or the ability to improvebinding by engaging multiple binding sites was eliminated and thefailure to detect binding to the disaccharide by ITC strongly sug-gests that all three sugar residues, in the appropriate configuration(� or �), contribute to binding affinity. While less information isavailable about Stx2 binding to synthetic disaccharides, Lingwoodand his coworkers have demonstrated that both Stx1 and Stx2bind to galabiosyl ceramide on thin-layer chromatography (TLC)plates (41). The failure of Stx2 to bind to the biantennary disac-charide glycoconjugates could be due to the difference in the pre-sentation of sugars on the ELISA platform versus TLC plates ordue to the fact that fewer hydrogen bonding and hydrophobicinteractions can form with glycoconjugates that lack ceramidethan with glycoconjugates with galabiosyl ceramide. Nevertheless,failure to detect binding to the biantennary disaccharides indi-

FIG 3 Binding of Stx1 and Stx2 subtypes to biantennary Pk analogues. (A) Bind-ing to biantennary Pk. (B) Binding to biantennary NAc-Pk. Means � standarderrors of the means (error bars) of three independent experiments are shown.RFU, relative fluorescence units. For symbol definitions, see the legend to Fig. 2.

FIG 4 Stx2 B subunit binding to NAc-Pk. Means � standard errors of themeans (error bars) of three independent experiments are shown. RFU, relativefluorescence units. For symbol definitions, see the legend to Fig. 2.

Yosief et al.

2756 iai.asm.org Infection and Immunity

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

cates that Stx2 also requires three sugar residues for optimal bind-ing on an ELISA platform.

Holotoxin binding to monoantennary di- and trisaccha-rides. The contribution of avidity to binding was assessed by usingthe monoantennary glycoconjugates. Stx1 exhibited minimal

binding to monoantennary Pk, in contrast to the control bianten-nary Pk (Fig. 5), and failed to bind to any of the other monoan-tennary glycoconjugates. Stx2a and the other Stx2 subtypes failedto bind to any of the monoantennary glycoconjugates (data notshown). While the biotinylated biantennary glycoconjugates pro-vide four Pk-trisaccharide analogs for binding via the two strepta-vidin binding sites, monoantennary glycoconjugates provide onlytwo Pk-trisaccharide analogs for binding. The failure to see bind-ing with the monoantennary glycoconjugates suggests that theholotoxins need to engage more than two Pk-trisaccharide ana-logs for effective binding and this has been verified by bindingstudies with mixtures of glycoconjugates as described below.

Binding of holotoxins to glycoconjugate mixtures. Recentbinding studies have demonstrated that mixtures of glycoconju-gates can support better binding than a single glycoconjugate can(33, 44, 51). Stx1 possesses three nonequivalent binding sites forPk-trisaccharide per monomer of B pentamer (40), and glycocon-jugate mixtures could enhance overall binding by allowing eachbinding site to engage its preferred glycoconjugate. Alternatively,the presence of glycoconjugate mixtures could result in decreasedoverall binding because of competition between preferred and

FIG 5 Stx1 binding to monoantennary Pk analogues. Means � standard er-rors of the means (error bars) of three independent experiments are shown.RFU, relative fluorescence units. Biantennary Pk (Pk) served as the positivecontrol. For symbol definitions, see the legend to Fig. 2.

FIG 6 (A) Stx1 binding to biantennary Pk and mixtures. RFU, relative fluorescence units. (B) Pk analogues were mixed with each of the glycoconjugates shownin the bar graph, and percent binding was determined as a percentage of the binding of Stx1 to the indicated biantennary Pk analog in the absence of otheranalogues at the highest toxin concentrations tested (145 �M). Means � standard errors of the means (error bars) of three independent experiments are shown.Statistical analysis with a Student t test comparing each mixture with Pk was performed (*, P � 0.05). The zigzag line represents PEG. For symbol definitions, seethe legend to Fig. 2.

Shiga Toxin Binding to Receptor Analogues

August 2013 Volume 81 Number 8 iai.asm.org 2757

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

suboptimal receptor mimics at a single binding site. We assessedShiga toxin binding in the presence of its preferred glycoconjugatemixed with an equal amount of another glycoconjugate.

Binding preferences of Stx1. Biantennary Pk was mixed at a1:1 ratio with other glycoconjugates or PEG (as a negative con-trol), and binding to Stx1 was assessed (Fig. 6A and B). As ob-served previously (44), little to no binding was observed whenbiantennary Pk was mixed with PEG (Fig. 6A), which does notsupport binding by itself (34, 44). Similarly, we observed weakbinding to monoantennary Pk (Fig. 5), which also results in thepresentation of two Pk analogues, suggesting that two Pk ana-logues are not sufficient to support binding. However, mixtures ofbiantennary Pk with other glycoconjugates supported binding tovarious degrees. Binding to the glycoconjugate mixtures was cal-culated at the highest Stx1 concentration tested, by setting bindingto unmixed biantennary Pk at 100% (Fig. 6B). The most binding(72%) was observed when biantennary Pk was mixed with mono-antennary Pk, a mixture that should result in the presentation of,on average, three Pk analogues per streptavidin molecule. Theseresults suggest that the engagement of at least three Pk-trisaccha-rides is critical for Stx1 binding and the presence of the fourthPk-trisaccharides causes only a modest increase in binding.

While less effective than monoantennary Pk, NAc-Pk couldalso provide additional binding contacts when mixed with bian-tennary Pk (52%), and monoantennary NAc-Pk (51%) supportedequivalent levels of binding when mixed with Pk (Fig. 6A). Theability of biantennary NAc-Pk to support binding of Stx1 in the

presence of Pk was reported previously (44). This study suggeststhat only one NAc-Pk is engaged instead of two NAc-Pk receptormimics because there is no significant difference in Stx1 bindingto biantennary Pk mixed with biantennary or monoantennaryNAc-Pk.

The monoantennary disaccharides also seem to provide addi-tional binding contacts for Stx1 but are not as effective as the Pk orNAc-Pk-trisaccharide. Monoantennary Gal�1-4Gal (Fig. 6A and B,Gal-Gal mono) supported 46% binding. Two disaccharides with aGalNAc residue, GalNAc�1-4Gal (35%) and Gal�1-4GalNAc(35%), supported less binding than did the disaccharide Gal�1-4Gal,suggesting that galactose at both positions is preferable to GalNAc ateither position (Fig. 6A and B). These results are consistent with thecrystal structure of Stx1 with Pk, where contact is made primarilywith the two terminal galactose residues (40). Overall, these resultsdemonstrate that a minimum of three Pk analogues need to be en-gaged for the binding of Stx1. When two native Pk-trisaccharides arepresent, monoantennary trisaccharide and disaccharide Pk ana-logues can provide additional binding contacts with Stx1, with nativePk being more effective than other Pk analogues.

Binding preferences of Stx2 subtypes. The binding of Stx2a(Fig. 7A and B) and Stx2c (Fig. 7C and D) to biantennary NAc-Pkmixed with other glycoconjugates was also assessed. While thebinding preferences were similar, in all cases, Stx2c bound betterthan Stx2a. Similar to Stx1, the monoantennary form of its pre-ferred Pk analogue, NAc-Pk, provided an additional binding con-tact; however, only about 30% binding was recovered for Stx2a

FIG 7 Stx2 subtype binding to biantennary NAc-Pk and mixtures. Pk analogues were mixed with each of the glycoconjugates shown in the bar graph, and percentbinding was determined as a percentage of the value obtained for toxin binding to the indicated biantennary NAc-Pk analog in the absence of other analogues atthe highest toxin concentration tested (145 �M). (A, B) Stx2a binding to biantennary NAc-Pk and mixtures. (C, D) Stx2c binding to biantennary NAc-Pk andmixtures. Means � standard errors of the means (error bars) of three independent experiments are shown. Statistical analysis with a Student t test comparing eachmixture with NAc-Pk was performed (*, P � 0.05). Stx2a binding with the mixture of PEG and NAc-Pk was insignificant (data not shown). The zigzag linerepresents PEG. RFU, relative fluorescence units. For symbol definitions, see the legend to Fig. 2.

Yosief et al.

2758 iai.asm.org Infection and Immunity

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

(Fig. 7B), while 77% of Stx2c binding was recovered (Fig. 7D).These results suggest that, like Stx1, Stx2a and Stx2c require aminimum of three trisaccharides for binding.

The glycoconjugate preferences of the Stx2 subtypes differedfrom those of Stx1. For both Stx2a and Stx2c, monoantennaryNAc-Pk supported binding, suggesting that, like Stx1, these Stx2subtypes also need to engage three Pk analogues for binding.However, for Stx2a, only 30% binding was observed when bian-tennary NAc-Pk was mixed with monoantennary NAc-Pk (Fig.7B), while 77% of Stx2c binding was observed with this mixture(Fig. 7D). In addition, mixtures with native Pk, the nonpreferredanalogue, also supported binding. Furthermore, the biantennaryform supported more binding than the monoantennary form,suggesting that Stx2a and Stx2c could engage two molecules ofnative Pk. Unlike Stx1, none of the monoantennary or bianten-nary disaccharide Pk analogues supported the binding of eitherStx2a or Stx2c (data not shown).

In summary, each B subunit monomer of Shiga toxin possessesmultiple binding sites for Pk analogues, and each unique bindingsite is represented five times in the pentamer. Synthetic glycocon-jugates immobilized on streptavidin present a robust experimen-tal system to systematically dissect the roles of binding site prefer-ences and binding avidity. We have shown that each form of Shigatoxin displays distinct preferences for the Pk analogues examinedin this study. In addition, studies with the biantennary and mono-antennary Pk analogues demonstrated that Stx1 and Stx2 subtypes(Stx2a and -c) need to engage at least three Pk analogues for effec-tive binding. This information is critically important, since Pkanalogues have been developed to function as therapeutic agentsby blocking Shiga toxin binding to its natural cellular receptor(34). Unfortunately, they have not yet proven to be effective in-hibitors in clinical trials (37–39). While these synthetic analogsexhibited strong binding affinity for both Stx1 and Stx2a, theybound Stx1 better than the more potent form Stx2a, which couldexplain their failure in clinical trials.

In this report, we have shown that the different subtypes ofStx2 also display different binding preferences. Most surprisingwas the failure of the highly potent Stx2d holotoxin to bind toNAc-Pk, even though the purified B pentamer bound to NAc-Pk.This reveals that the C-terminal region of the A subunit couldinterfere with binding. In addition, there is no obvious relation-ship between recognition of the synthetic glycan and toxin po-tency. In previous studies, the lethal dose of either Stx2a or Stx2dfor mice was less than 10 ng, whereas the lethal dose of Stx2c wasgreater than 1,000 ng (18). These studies strongly suggest thatfactors other than glycan recognition could be important for de-termining potency. Future studies will examine whether Stx mayutilize protein coreceptors.

ACKNOWLEDGMENTS

This work was supported by National Institutes of Health grants R01 AI064893 (A.A.W.) and U01 AI 075498 (A.A.W.) and NSF CAREER CHE-0845005 (S.S.I.).

We thank Sayali Karve and Karen Gallegos for Shiga toxin prepara-tions.

REFERENCES1. Law D. 2000. Virulence factors of Escherichia coli O157 and other Shiga

toxin-producing E. coli. J. Appl. Microbiol. 88:729 –745.2. Obrig GT. 2010. Escherichia coli Shiga toxin mechanisms of action in renal

disease. Toxins 2:2769 –2794.

3. Keir SL, Marks DS, Kim JJ. 2012. Shigatoxin-associated hemolytic ure-mic syndrome: current molecular mechanisms and future therapies. DrugDes. Devel. Ther. 6:195–208.

4. Bauwens A, Bielaszewska M, Kemper B, Langehanenberg P, von BallyG, Reichelt R, Mulac D, Humpf HU, Friedrich AW, Kim KS, Karch H,Müthing J. 2011. Differential cytotoxic actions of Shiga toxin 1 and Shigatoxin 2 on microvascular and macrovascular endothelial cells. Thromb.Haemost. 105:515–528.

5. Obata F, Tohyama K, Bonev DA, Kolling LG, Keepers RT, Gross KL,Nelson TM, Sato S, Obrig GT. 2008. Shiga toxin 2 affects the centralnervous system through receptor Gb3 localized to neurons. J. Infect. Dis.198:1398 –1406.

6. O’Brien AD, Tesh VL, Donohue-Rolfe A, Jackson MP, Olsnes S, Sand-vig K, Lindberg AA, Keusch GT. 1992. Shiga toxin: biochemistry, genet-ics, mode of action, and role in pathogenesis. Curr. Top. Microbiol. Im-munol. 180:65–94.

7. Endo Y, Tsurugi K, Yutsudo T, Takeda Y, Ogasawara T, Igarashi K.1988. Site of action of a Vero toxin (VT2) from Escherichia coli O157:H7and of Shiga toxin on eukaryotic ribosomes. RNA N-glycosidase activityof the toxins. Eur. J. Biochem. 171:45–50.

8. Brown JE, Echeverria P, Lindberg AA. 1991. Digalactosyl-containingglycolipids as cell surface receptors for Shiga toxin of Shigella dysenteriae 1and related cytotoxins of Escherichia coli. Rev. Infect. Dis. 13(Suppl 4):S298 –S303.

9. Obrig TG, Moran TP, Brown JE. 1987. The mode of action of Shiga toxinon peptide elongation of eukaryotic protein synthesis. Biochem. J. 244:287–294.

10. Donohue-Rolfe A, Acheson DW, Keusch GT. 1991. Shiga toxin. purifi-cation, structure, and function. Rev. Infect. Dis. 13(Suppl 4):S293–S297.

11. Persson S, Olsen KE, Ethelberg S, Scheutz F. 2007. Subtyping methodfor Escherichia coli Shiga toxin (verocytotoxin) 2 variants and correlationsto clinical manifestations. J. Clin. Microbiol. 45:2020 –2024.

12. Johannes L, Römer W. 2010. Shiga toxins—from cell biology to biomed-ical applications. Nat. Rev. Microbiol. 8:105–116.

13. Scheiring J, Andreoli PS, Zimmerhackl BL. 2008. Treatment and out-come of Shiga-toxin-associated hemolytic uremic syndrome (HUS). Pe-diatr. Nephrol. 23:1749 –1760.

14. Shimizua T, Sato T, Kawakamic SO, Nodaa TM, Hamabata T. 2007.Receptor affinity, stability and binding mode of Shiga toxins are determi-nants of toxicity. Microb. Pathog. 43:88 –95.

15. Boerlin P, McEwen SA, Boerlin-Petzold F, Wilson JB, Johnson RP,Gyles CL. 1999. Associations between virulence factors of Shiga toxin-producing Escherichia coli and disease in humans. J. Clin. Microbiol. 37:497–503.

16. Orth D, Grif K, Khan BA, Naim A, Dierich PM, Würzner R. 2007. TheShiga toxin genotype rather than the amount of Shiga toxin or the cyto-toxicity of Shiga toxin in vitro correlates with the appearance of the he-molytic uremic syndrome. Diagn. Microbiol. Infect. Dis. 59:235–242.

17. Kawano K, Okada M, Haga T, Maeda K, Goto Y. 2008. Relationshipbetween pathogenicity for humans and stx genotype in Shiga toxin-producing Escherichia coli serotype O157. Eur. J. Clin. Microbiol. Infect.Dis. 27:227–232.

18. Fuller AC, Pellino AC, Flagler JM, Strasser EJ, Weiss AA. 2011. Shigatoxin subtypes display dramatic differences in potency. Infect. Immun.79:1329 –1337.

19. Fraser ME, Fujinaga M, Cherney MM, Melton-Celsa AR, Twiddy EM,O’Brien AD, James MN. 2004. Structure of Shiga toxin type 2 (Stx2) fromEscherichia coli O157:H7. J. Biol. Chem. 279:27511–27517.

20. Melton-Celsa AR, Kokai-Kun JF, O’Brien AD. 2002. Activation of Shigatoxin type 2d (Stx2d) by elastase involves cleavage of the C-terminal twoamino acids of the A2 peptide in the context of the appropriate B penta-mer. Mol. Microbiol. 43:207–215.

21. Siegler RL, Obrig TG, Pysher TJ, Tesh VL, Denkers ND, Taylor FB.2003. Response to Shiga toxin 1 and 2 in a baboon model of hemolyticuremic syndrome. Pediatr. Nephrol. 18:92–96.

22. Stearns-Kurosawa JD, Collins V, Freeman S, Tesh LV, Kurosawa S.2010. Distinct physiologic and inflammatory responses elicited in ba-boons after challenge with Shiga toxin type 1 or 2 from enterohemorrhagicEscherichia coli. Infect. Immun. 78:2497–2504.

23. Tyrrell GJ, Ramotar K, Toye B, Boyd B, Lingwood CA, Brunton JL.1992. Alteration of the carbohydrate binding specificity of verotoxinsfrom Gal alpha 1-4Gal to GalNAc beta 1-3Gal alpha 1-4Gal and vice versa

Shiga Toxin Binding to Receptor Analogues

August 2013 Volume 81 Number 8 iai.asm.org 2759

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

by site-directed mutagenesis of the binding subunit. Proc. Natl. Acad. Sci.U. S. A. 89:524 –528.

24. DeGrandis S, Law H, Brunton J, Gyles C, Lingwood CA. 1989. Globo-tetraosylceramide is recognized by the pig edema disease toxin. J. Biol.Chem. 264:12520 –12525.

25. Waddell T, Head S, Petric M, Cohen A, Lingwood C. 1988. Globotriosylceramide is specifically recognized by the Escherichia coli verocytotoxin 2.Biochem. Biophys. Res. Commun. 152:674 – 679.

26. Valdivieso-Garcia A, MacLeod DL, Clarke RC, Gyles CL, Lingwood C,Boyd B, Durette A. 1996. Comparative cytotoxicity of purified Shiga-liketoxin-IIe on porcine and bovine aortic endothelial and human colonicadenocarcinoma cells. J. Med. Microbiol. 45:331–337.

27. Head SC, Karmali MA, Lingwood CA. 1991. Preparation of VT1 andVT2 hybrid toxins from their purified dissociated subunits. Evidence for Bsubunit modulation of A subunit function. J. Biol. Chem. 266:3617–3621.

28. Lingwood CA. 1996. Role of verotoxin receptors in pathogenesis. TrendsMicrobiol. 4:147–153.

29. Marcato P, Vander Helm K, Mulvey GL, Armstrong GD. 2003. Serumamyloid P component binding to Shiga toxin 2 requires both a subunitand B pentamer. Infect. Immun. 71:6075– 6078.

30. Weinstein DL, Jackson MP, Perera LP, Holmes RK, O’Brien AD. 1989.In vivo formation of hybrid toxins comprising Shiga toxin and the Shiga-like toxins and role of the B subunit in localization and cytotoxic activity.Infect. Immun. 57:3743–3750.

31. St Hilaire PM, Boyd MK, Toone EJ. 1994. Interaction of the Shiga-liketoxin type 1 B subunit with its carbohydrate receptor. Biochemistry 33:14452–14463.

32. Kitova EN, Kitov PI, Paszkiewicz E, Kim J, Mulvey GL, Armstrong GD,Bundle DR, Klassen JS. 2007. Affinities of Shiga toxins 1 and 2 for uni-valent and oligovalent Pk-trisaccharide analogs measured by electrosprayionization mass spectrometry. Glycobiology 17:1127–1137.

33. Gallegos MK, Conrady GD, Karve SS, Gunasekera ST, Herr BA, WeissAA. 2012. Shiga toxin binding to glycolipids and glycans. PLoS One7:e30368. doi:10.1371/journal.pone.0030368.

34. Kitov PI, Sadowska JM, Mulvey G, Armstrong GD, Ling H, Pannu NS,Read RJ, Bundle DR. 2000. Shiga-like toxins are neutralized by tailoredmultivalent carbohydrate ligands. Nature 403:669 – 672.

35. Watanabe M, Matsuoka K, Kita E, Igai K, Higashi N, Miyagawa A,Watanabe T, Yanoshita R, Samejima Y, Terunuma D. 2004. Oral ther-apeutic agents with highly clustered globotriose for treatment of Shigatoxigenic Escherichia coli infections. J. Infect. Dis. 189:360 –368.

36. Nishikawa K, Matsuoka K, Kita E, Okabe N, Mizuguchi M, Hino K,Miyazawa S, Yamasaki C, Aoki J, Takashima S. 2002. A therapeuticagent with oriented carbohydrates for treatment of infections by Shigatoxin-producing Escherichia coli O157:H7. Proc. Natl. Acad. Sci. U. S. A.99:7669 –7674.

37. Mulvey GL, Marcato P, Kitov PI, Sadowska J, Bundle DR, Armstrong

GD. 2003. Assessment in mice of the therapeutic potential of tailored,multivalent Shiga toxin carbohydrate ligands. J. Infect. Dis. 187:640 – 649.

38. Trachtman H, Cnaan A, Christen E, Gibbs K, Zhao S, Acheson DW,Weiss R, Kaskel FJ, Spitzer A, Hirschman GH. 2003. Effect of an oralShiga toxin-binding agent on diarrhea-associated hemolytic uremic syn-drome in children: a randomized controlled trial. JAMA 290:1337–1344.

39. Nishikawa K. 2011. Recent progress of Shiga toxin neutralizer for treat-ment of infections by Shiga toxin-producing Escherichia coli. Arch. Immu-nol. Ther. Exp. 59:239 –247.

40. Ling H, Boodhoo A, Hazes B, Cummings MD, Armstrong GD, BruntonJL, Read RJ. 1998. Structure of the Shiga-like toxin I B pentamer com-plexed with an analogue of its receptor Gb3. Biochemistry 37:1777–1788.

41. Lingwood CA, Binnington B, Manis A, Branch DR. 2010. Globotriaosylceramide receptor function—where membrane structure and pathologyintersect. FEBS Lett. 584:1879 –1886.

42. Mahfoud R, Manis A, Lingwood AC. 2009. Fatty acid-dependent glo-botriaosyl ceramide receptor function in detergent resistant model mem-branes. J. Lipid Res. 50:1744 –1755.

43. Kale RR, McGannon CM, Fuller-Schaefer C, Hatch DM, Flagler MJ,Gamage SD, Weiss AA, Iyer SS. 2008. Differentiation between structur-ally homologous Shiga 1 and Shiga 2 toxins by using synthetic glycocon-jugates. Angew. Chem. Int. Ed. Engl. 47:1265–1268.

44. Flagler JM, Mahajan SS, Kulkarni AA, Iyer SS, Weiss AA. 2010. Com-parison of binding platforms yields insights into receptor binding differ-ences between Shiga toxins 1 and 2. Biochemistry 49:1649 –1657.

45. Yosief HO, Iyer SS, Weiss AA. 2013. Capture of uropathogenic Esche-richia coli using synthetic glycan ligands specific for the pap-pilus. Chem-BioChem 14:251–259.

46. Gamage SD, McGannon CM, Weiss AA. 2004. Escherichia coli serogroupO107/O117 lipopolysaccharide binds and neutralizes Shiga toxin 2. J. Bac-teriol. 186:5506 –5512.

47. Conrady GD, Flagler JM, Friedmann RD, Vander Wielen DB, KovallAR, Weiss AA, Herr BA. 2010. Molecular basis of differential B pentamerstability of Shiga toxins 1 and 2. PLoS One 5:e15153. doi:10.1371/journal.pone.0015153.

48. Nakajima H, Kiyokawa N, Katagiri YU, Taguchi T, Suzuki T, Sekino T,Mimori K, Ebata T, Saito M, Nakao H. 2001. Kinetic analysis of bindingbetween Shiga toxin and receptor glycolipid Gb3Cer by surface plasmonresonance. J. Biol. Chem. 276:42915– 42922.

49. Miura Y, Sasao Y, Dohi H, Nishida Y, Kobayashi K. 2002. Self-assembled monolayers of globotriaosylceramide (Gb3) mimics: surface-specific affinity with Shiga toxins. Anal. Biochem. 310:27–35.

50. Kulkarni AA, Fuller AC, Korman H, Weiss AA, Iyer SS. 2010. Glycanencapsulated gold nanoparticles selectively inhibit Shiga toxins 1 and 2.Bioconjug. Chem. 21:1486 –1493.

51. Liang CH, Wang SK, Lin CW, Wang CC, Wong CH, Wu CY. 2011.Effects of neighboring glycans on antibody-carbohydrate interaction. An-gew. Chem. Int. Ed. Engl. 50:1608 –1612.

Yosief et al.

2760 iai.asm.org Infection and Immunity

on Novem

ber 17, 2020 by guesthttp://iai.asm

.org/D

ownloaded from