Glycobiology

Glycobiology Advance Access originally published online on August 8, 2007
Glycobiology 2007 17(10):1127-1137; doi:10.1093/glycob/cwm081

This Article Abstract FREE Full Text (PDF) Supplementary Data All Versions of this Article: 17/10/1127    most recent cwm081v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (2) Request Permissions Disclaimer Google Scholar Articles by Kitova, E. N Articles by Klassen, J. S Search for Related Content PubMed PubMed Citation Articles by Kitova, E. N Articles by Klassen, J. S Social Bookmarking          What's this?

© The Author 2007. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org

Affinities of Shiga toxins 1 and 2 for univalent and oligovalent Pk-trisaccharide analogs measured by electrospray ionization mass spectrometry

Elena N Kitova², Pavel I Kitov², Eugenia Paszkiewicz², Jonghwa Kim², George L Mulvey³, Glen D Armstrong³, David R Bundle² and John S Klassen^1,2

² Department of Chemistry, University of Alberta, Edmonton, Alberta T6G 2 G2, Canada
³ Department of Microbiology and Infectious Diseases, Faculty of Medicine, University of Calgary, Alberta T2 N 4N1, Canada

---

¹ To whom correspondence should be addressed: Tel: +1 780 492 3501; Fax: +1 780 492 8231; e-mail: john.klassen{at}ualburta.ca.

Received on June 15, 2007; revised on July 20, 2007; accepted on July 22, 2007

	Abstract

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
The binding stoichiometry and affinities of the Shiga toxins, Stx1 and Stx2, for a series of uni- and oligovalent analogs of the Pk-trisaccharide were measured using the direct electrospray ionization mass spectrometry (ES-MS) assay. Importantly, it is shown that, for a given ligand, Stx1 and Stx2 exhibit similar affinities. The binding data suggest a high degree of similarity in the spatial arrangement and structural characteristics of the Pk binding sites in Stx1 and Stx2. The results confirm that both toxins recognize the -D-Galp(14)-ß-D-Galp(14)-ß-D-Glcp carbohydrate motif of the cell surface glycolipid Gb3. This, taken together with the results of the chemical mapping study, suggests that the nature of the Pk binding interactions with Stx1 and Stx2 are similar. The affinities of Stx1-B₅ and Stx2 for the multivalent ligands reveals that site 2 of Stx2, which shares the same spatial arrangement as site 2 in Stx1, is the primary Pk binding site and that site 1 of Stx1 and of Stx2 can also participate in Pk binding.

Key words: association constants / electrospray ionization mass spectrometry / protein-oligosaccharide complexes / Shiga toxins / stoichiometry

	Introduction

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
The Shiga toxins (Stx) 1 and 2 are bacterial enterotoxins produced by the enterohemorrhagic group of enterovirulent Escherichia coli (EHEC) (Karmali 2004). These organisms represent a sub-group of the Shigatoxigenic E. coli and are isolated from human subjects experiencing hemorrhagic colitis and, occasionally, the hemolytic-uremic syndrome (HUS). The Stx display classical AB₅ structures, in which a single A subunit (32 kDa) is associated with five identical B subunits (approximately 8 kDa each) that assemble into a doughnut-shaped structure. The central pore created by the B subunits hosts the C-terminus of the A subunit (Fraser et al. 2004; Karmali 2004). The amino acid sequences of the A and B subunits of Stx1 and Stx2 are 52% and 60% homologous, respectively (Calderwood et al. 1987; Strockbine et al. 1988). The A subunit is an N-glycosidase that cleaves a single adenine residue in the 28 S rRNA component of eukaryotic ribosomes and shuts down the protein biosynthesis machinery causing cell death. The B subunits are lectins that recognize Gb3 glycolipid (ceramide-linked Pk-trisaccharide, -D-Galp(14)-ß-D-Galp(14)-ß-D-Glcp (1O)-ceramide), the natural cell receptor that is located on the surface of target cells (Karmali et al. 1985; Lindberg et al. 1987; Jacewicz et al. 1986; Waddell et al. 1988). From the crystal structure of the Stx1-B₅ pentamer complexed with the Pk-trisaccharide analog, three receptor-binding sites (referred to as site 1, site 2 and site 3) in each B subunit were identified (Ling et al. 1998). The relative importance of these three sites for cellular recognition is not fully understood, although it is generally thought that site 2 provides the crucial recognition, with minor contributions from site 1 and site 3 (Ling et al. 1998; Shimizu et al. 1998). To our knowledge, there are no structural data available for the Gb3 receptor or the Pk-trisaccharide binding to the B subunits of Stx2.

Stx1 and Stx2 exhibit marked differences in biological activity; Stx2 being more potent than Stx1 in vitro and in vivo and more closely associated with the most severe consequences of EHEC infections in humans (Scotland et al. 1987; Ostroff et al. 1989; Kleanthous et al. 1990; Milford et al. 1990; Tesh et al. 1993; Louise and Obrig 1995). The origin of the differential toxicities of Stx1 and Stx2 is not fully known but may arise from differences in receptor recognition (Head et al. 1991). To address this question, a number of binding studies have been performed on Stx with Gb3 and analogs of the Pk-trisaccharide (Head et al. 1991; Nyholm et al. 1996; Kitov et al. 2000; Itoh et al. 2001; Nakajima et al. 2001; Miura et al. 2002; Watanabe et al. 2006). Because Gb3 is virtually insoluble in water, its binding to Stx has been investigated using solid-phase assays, in which the glycolipid was immobilized on a solid support. The affinities (K_a) determined from these studies typically range from 10⁶ to 10⁸ M^–1, with Stx1 generally exhibiting stronger binding to the receptor than Stx2, by about one order of magnitude (Head et al. 1991; Nakajima et al. 2001). The variability in the reported affinities appears to be due to differences in the nature of the assays used and the experimental conditions. Further, the nature of the aglycone of the Gb3 analogs has been shown to influence binding (Mylvaganam and Lingwood 1999; Binnington et al. 2002). The high affinities have been explained in terms of a significant multivalency effect, which operates in addition to the intrinsic interactions between the B subunits and the carbohydrate and lipid moieties of the receptor (Sandvig 2006).

It was suggested that the difference in the measured affinities of Gb3 for Stx1 and Stx2 may be related to the differential recognition of the Pk-trisaccharide by the B subunits of Stx1 and Stx2 (Fraser et al. 2004). To evaluate the strength of the individual interactions between the B subunits and the Pk-trisaccharide, as well as to evaluate the contribution of the multivalency effect to the affinity, binding measurements have been carried out on Stx1 and Stx1-B₅ with water soluble Gb3 analogs that are devoid of the ceramide aglycone. Measurements performed on univalent analogues of the Pk-trisaccharide revealed that the intrinsic interaction between the Pk-trisaccharide and the B subunit is very weak, K_a < 10³ M^–1 (St Hilaire et al. 1994; Soltyk et al. 2002). Significantly larger affinities were measured for oligovalent analogs. For example, subnanomolar inhibitory activity was measured for the oligovalent, water-soluble carbohydrate ligand named STARFISH using ELISA (enzyme-linked immunosorbent assay) (Kitov et al. 2000). The STARFISH ligand posssess five Pk-trisaccharide dimers at the tips of five appropriately oriented arms that radiate from a central glucose core which allows all five Pk dimers to be simultaneously engaged by the B-subunit pentamer in much the same way that the membrane surface binds the toxin. The high inhibitory activity, which was measured by competitive ELISA as a concentration of the inhibitor required to reduce binding of the toxin to the immobilized ligand by 50% (IC₅₀) and which is related to the reciprocal of K_a, is comparable with values measured for the natural glycolipid Gb3 (Kitov et al. 2000). A recently proposed thermodynamic model of multivalent interactions explains the dramatic increase in affinity in terms of an increase in the avidity entropy term (Kitov and Bundle 2003). To our knowledge, affinities for Stx2 with Pk-trisaccharide analogs have not been reported.

Here, we seek to gain a greater understanding of the interaction of Stx1 and Stx2 with the Pk-trisaccharide, the absolute affinities, the number and nature of the binding sites, and the influence of the aglycone moiety. To that end we have carried out the first quantitative comparative study of the binding of Stx1 and Stx2 with an array of water soluble uni- and oligovalent analogs of the Pk-trisaccharide (Figure 1). These measurements were carried out using the direct electrospray ionization mass spectrometry (ES-MS) assay, a technique pioneered by our laboratory for the quantification of protein-carbohydrate binding (Kitova et al. 2001; Wang et al. 2003a, 2003b; Daneshfar et al. 2004). The ES-MS assay has been shown to yield affinities that agree well with values obtained by other assays, including isothermal titration microcalorimetry (ITC) (Rademacher et al. 2007). The ES-MS assay is particularly well suited for the study of lectin–carbohydrate binding owing to the small sample requirements of the method, the absence of labeling or immobilization of protein or ligand, and the unique ability of the method to directly establish binding stoichiometry. Additionally, the ES-MS assay has distinguished itself in the analysis of very weak interactions 10³ M^–1 (Sun et al. 2006), which are often difficult to study by methods such as ITC (Dam and Brewer 2004).

View larger version (15K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 1 Structures of univalent (1, 2, 7–14) and oligovalent (3–6) ligands based on the Pk-trisaccharide.

 

	Results

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
NanoES mass spectra of Stx1 and Stx2
In order to implement the ES-MS assay to quantify protein–ligand binding, the concentrations of protein and ligand in solution must be accurately known. Because the Stx are noncovalent assemblies, it was necessary to first assess the quaternary structures of the toxins in solution. The results presented in Figures 2A and 2B are illustrative nanoES mass spectra acquired for aqueous solutions of Stx1 (nominal concentration of 4 µM) and Stx2 (6 µM), respectively, at pH 7 and 25°C. Ions corresponding to the protonated holotoxins of Stx1 and Stx2, i.e., AB₅ⁿ⁺ ions at charge states (n) ranging from +17 to +21, were detected. For Stx2, the AB₅ⁿ⁺ ions were the only toxin ions detected, while for Stx1, ions corresponding to the protonated B-subunit homopentamer, i.e., B₅ⁿ⁺ at charge states +12 to +15, and protonated B-subunit decamer, i.e., B₁₀ⁿ⁺ at charge states +18 to +20, were also detected. The assembly of the B subunits of Stx1 and Stx2, in the absence of A subunit, under a variety of solution conditions was recently investigated by our laboratory using temperature-controlled nanoES-MS (Kitova et al. 2005). It was found that, in the absence of the A subunit, the B subunits of Stx1 exists exclusively as homopentamers under a wide range of solution conditions. In contrast, the B subunit of Stx2 exists simultaneously in a number of different multimeric forms ranging from monomer to pentamer (i.e., B, B₂, ..., B₅). Furthermore, the distribution of Stx2 B subunit multimers is very sensitive to the solution conditions. At pH 7 and 25°C, the B₅ homopentamer is the dominant species only at relatively high subunit concentrations (>50 µM) and high ionic strength. Under these conditions, ions corresponding to the decamer, B₁₀, were also detected. However, it could not be ascertained whether the decamer originated from solution or whether it formed from nonspecific self-association of B₅ during the ES process. Similarly, the B₁₀ⁿ⁺ ions detected in the present study (Figure 2A) may originate in solution or from nonspecific binding. The dramatic difference in assembly of the B subunits of Stx1 and Stx2 was attributed to differences in the ionic interactions between subunits, as suggested by an analysis of the crystal structures of the Stx. Specifically, it was proposed that the lower stability of the B₅ pentamer of Stx2, compared to Stx1, is due to a repulsive interaction between Asp70 and Glu09 on adjacent subunits. In Stx1-B₅ there is no such repulsive interaction; instead analysis of the crystal structure suggests an additional stabilizing ionic interaction involving Arg69 and Asp10. The absence of ions corresponding to free B subunit in the mass spectra acquired for Stx2 in the present study reveals that the A subunit confers stability to the B₅ homopentamer.

View larger version (11K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 2 NanoES mass spectra of solutions consisting of 10 mM ammonium acetate (pH 7) and (A) 4 µM Stx1, (B) 6 µM Stx2 or (C) 20 µM Stx1-B subunit.

 
The abundant B subunit ions detected for the purified Stx1 sample likely arise from the simultaneous elution of holotoxin and free B subunit during affinity purification. Because of the contamination of the Stx1 sample with B subunit, the concentration of AB₅ holotoxin could not be reliably established. Consequently, the ligand binding measurements reported herein were performed using Stx1-B₅ instead of the holotoxin. Shown in Figure 2C is an illustrative ES mass spectrum acquired for a solution of Stx1 B subunit (20 µM). It can be seen that the only B subunit ions detected correspond to the homopentamer, B₅ⁿ⁺ at charge states +11 to +14. In view of the structural similarity of the B subunits with the Stx1 holotoxin and Stx1-B₅ complexed with Pk-trisaccharide analog, suggested by crystallographic data (Stein et al. 1992; Fraser et al. 1994), the absence of the A-subunit is not expected to have a significant effect on binding of the Pk analogs to the B subunits of Stx1.

Stx binding to univalent Pk-trisaccharide ligands (1–2, 7–13)
Binding measurements were performed for Stx1-B₅ and Stx2 with two Pk-trisaccharide analogs (1, 2) containing different aglycone groups at the reducing end (Table I), as well as a series of monodeoxy analogs (7–13) of 1 (Table II). As described below, Stx exhibits low affinities for these univalent ligands 10³ M^–1. The determination of reliable K_a values for such weakly interacting complexes presents challenges, regardless of the analytical method used. In ES-MS, the principle experimental challenge arises from the occurrence of nonspecific ligand binding to the protein during the ES process. Nonspecific ligand association is promoted by the high ligand concentrations needed to produce detectable levels of complex in solution (Wang et al. 2003b, 2005). Recently, our laboratory developed a quantitative method for correcting mass spectra for the occurrence of nonspecific protein–ligand binding (Sun et al. 2006). The method involves the addition of a reference protein, P_ref, to the ES solution. As shown previously, the nonspecific association of ligand to P_ref is identical, within experimental error, to that of the target protein (Sun et al. 2006). The importance of properly accounting for nonspecific ligand binding in ES-MS binding assays is illustrated below.

View this table: [in this window] [in a new window]   Table I Association constants (Ka, units of M–1) for binding of the univalent (1, 2) and oligovalent Pk-based ligands (3–6) with the B5 homopentamer of Stx1 and with the Stx2 holotoxin determined at 25°C and pH 7 by the direct ES-MS assaya

 

View this table: [in this window] [in a new window]   Table II Association constants (Ka, units of M–1) for binding of deoxy analogs of Pk-trisaccharide (7–13) with Stx1-B5 and Stx2 determined at 25°C and pH 7by the direct ES-MS assaya,b

 
Shown in Figures 3A and 3B are nanoES mass spectra measured for solutions of the trisaccharide 1 with Stx1-B₅ and Stx2, respectively. A reference protein was added to each solution to monitor nonspecific ligand association. Ions corresponding to unbound Stx1-B₅ and Stx2 and Stx1-B₅ and Stx2 bound to between one and three molecules of 1, i.e., (B₅ + q1)ⁿ⁺ and (Stx2 + q1)ⁿ⁺ ions where q = 0–3, were detected at the concentrations investigated (7 µM Stx1-B₅ and 100 µM 1; 6.5 µM Stx2 and 70 µM 1). Importantly, ions corresponding to unbound P_ref and P_ref bound to one molecule of 1 were also detected. The appearance of (P_ref + 1)ⁿ⁺ ions indicates that nonspecific binding of 1 to Stx1-B₅ and to Stx2 contributes to the signal for the (B₅ + q1)ⁿ⁺ and (Stx2 + q1)ⁿ⁺ ions. Shown in Figure 4 are the distributions of 1 bound to Stx1-B₅, Stx2 and P_ref, as determined from the MS data. Also shown are the distributions for Stx1-B₅ and Stx2 which have been corrected for nonspecific ligand binding following the procedure outlined in the Materials and methods section. The corrected distributions reveal that, at these concentrations, Stx1-B₅ and Stx2 bind at most two molecules of 1. Using the corrected distributions, K_a values of 360 ± 10 and 250 ± 130 M^–1 were determined for the binding of 1 to Stx1-B₅ and Stx2, respectively. Using the same procedure, K_a values of 810 ± 260 and 240 ± 30 M^–1 were determined for the binding of 2 to Stx1-B₅ and Stx2, respectively.

View larger version (17K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 3 NanoES mass spectra of solutions consisting of 10 mM ammonium acetate (pH 7) and (A) 7 µM Stx1-B5 and 1 at concentration of 100 µM or (B) 6.5 µM Stx2 and 70 µM of 1. A reference protein (Pref) was added to the ES solutions at a concentration of 5 µM to quantify the extent of nonspecific protein–ligand binding during the ES process. The designation q indicates the number of bound ligand molecules.

 

View larger version (12K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 4 Distributions of 1 bound to (A) Stx1-B5 and Pref, and (B) Stx2 and Pref, (labeled as initial and Pref) as determined from the ES-MS data shown in Fig. 3. Also shown are the distributions for Stx1-B5 and Stx2 (labeled as corrected) after correction for nonspecific ligand binding.

 
Chemical mapping was used to evaluate the binding epitope of the Pk-trisaccharide for the B subunits of Stx1 and Stx2. In this approach, which was developed by Lemieux and coworkers (Nikrad et al. 1992), the contribution of specific ligand OH groups to the stability of the protein-carbohydrate complex is assessed from changes in binding free energy (and K_a) resulting from deoxygenation of this site on the ligand. The elimination of OH groups that form energetically important hydrogen bonds will result in a significant decrease in the binding free energy (and K_a). To implement this method, binding measurements were performed on Stx1-B₅ and Stx2 with seven monodeoxy analogs of 1, and the K_a values are listed in Table II. Notably, with the exception of 8, the K_a values measured for the monodeoxy analogs are similar to the values determined for 1. These results suggest that each of these OH groups provides only a small contribution to the stability of the complex. In contrast, specific binding for ligand 8 to Stx1-B₅ and Stx2 is virtually absent, indicating that this OH group is critical for binding.

Because the success of the chemical mapping strategy described above depends on the precision of the ligand affinity measurements, the utility of this method can, in principle, be improved by directly evaluating the relative affinities of the ligands instead of comparing absolute affinities. Relative affinities can be determined from competition experiments in which ES-MS measurements are performed on solutions containing two or more ligands, which have different molecular weights (Wang et al. 2003b). Usually these measurements are carried out using equimolar concentrations of ligands, so that the affinities are directly proportional to the relative abundance of the protein-ligand complexes, as determined from the mass spectrum. Ideally, the competition experiments would be carried out on solutions containing Stx, trisccharide 1, which serves as the reference ligand, and one of the monodeoxy analogs. However, the difference in MW between 1 (MW 518 Da) and the monodeoxy analogs (MW 502 Da) is only 16 Da. This is not a sufficiently large difference to enable the ions of the complexes of Stx with 1 and with the monodeoxy analogs to be fully resolved, particularly given the presence of alkali metal ion (Na⁺ and K⁺) adducts of the Stx ions. Therefore, instead of using 1 as the reference ligand, measurements were performed using the trimethylsilylethyl Pk-trisaccharide glycoside 14. This ligand binds to Stx1-B₅ with a K_a (270 ± 80 M^–1) (Sun et al. 2006) that is indistinguishable from that measured for 1. The MW of 14 is 102 Da larger than the MW of the monodeoxy ligands, which is more than adequate to resolve the complexes of Stx with 14 and with the monodeoxy analogs. However, nanoES-MS analysis of solutions of Stx1-B₅, 14 and one of 7–13 revealed that the extent of nonspecific ligand binding varies significantly between 14 and the deoxy analogs. Shown in Figure 5 is an illustrative mass spectrum acquired for a solution of Stx1-B₅, 14, 8 and a reference protein. From the relative abundance of ions corresponding to P_ref bound to 8 and 14, the fraction of B₅ bound nonspecifically to 14 was calculated to be 4%, while for 8 it was significantly higher, 24%. The difference in the extent of nonspecific binding likely reflects the greater hydrophobicity of 14, compared to 8. It was shown previously that increased hydrophobicity reduces the extent of nonspecific binding (Wang et al. 2005). As a result of the differential nonspecific binding of the two ligands to the Stx, the correction for nonspecific binding must be performed for each ligand, which increases the error in the estimation of the relative binding constants. Consequently, in the present study competition experiments provide no advantage for evaluating the relative affinities of the monodeoxy analogs for the Stx.

View larger version (10K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 5 NanoES mass spectrum of solutions consisting of 10 mM ammonium acetate (pH 7), 7 µM Stx1-B5 and ligands 14 and 8 at equal concentrations of 140 µM. A reference protein (Pref) was added to the ES solutions at a concentration of 5 µM to quantify the extent of nonspecific protein–ligand binding during the ES process. The designation q indicates the number of bound ligand molecules.

 
Stx binding to oligovalent Pk-based ligands (3–6)
In an effort to compare the Pk binding sites (number and location) in Stx1 and Stx2, as well as to evaluate the influence of multivalency on avidity, binding measurements were performed for Stx1-B₅ and Stx2 with multivalent analogs of 1 and 2. Ligands 3 and 4 are divalent analogs of 2, consisting of two molecules of 2 bound through a covalent linker. The linkers were designed so as to allow the trisaccharide moieties to simultaneously interact with site 1 and site 2 in a given Stx1 B subunit. Ligands 5 and 6 are tri- and tetravalent analogs of 1 and were designed to allow for the simultaneous occupancy of site 2 on three or four of the B subunits within the B₅ homopentamer of Stx1, respectively. Illustrative ES mass spectra acquired for solutions of Stx1-B₅ (4 µM) with 3 (23 µM) or 4 (20 µM) and Stx2 (6 µM) with 3 (11 µM) or 4 (17 µM) are shown in Figure 6. At the concentrations investigated, ions corresponding to unbound Stx1-B₅ and Stx2 were detected, i.e., B₅ⁿ⁺ and Stx2ⁿ⁺, along with ions corresponding to Stx1-B₅ and Stx2 bound to one or two molecules of 3 or 4. The distributions of ligands bound to Stx1-B₅ and Stx2, after correction for nonspecific binding, are shown in Figure 7 and the corresponding K_a values are reported in Table I. Stx1-B₅ and Stx2 were found to exhibit similar affinities for ligands 3 and 4. Illustrative nanoES mass spectra of aqueous solutions of Stx1-B₅ (4 µM) with 5 (8 µM) or 6 (2 µM) and Stx2 (7 µM) with 5 (4 µM) or 6 (4 µM) are shown in Figure 8. At the concentrations investigated, ions corresponding to unbound Stx1-B₅ or Stx2 and Stx1-B₅ or Stx2 bound to a single ligand were detected. Assuming a single ligand-binding site, apparent K_a values for 5 and 6 were calculated from the MS data (Table I). For each ligand, the K_a values for Stx1-B₅ and Stx2 are identical within experimental error, 10⁵ (5) and 10⁶ M^–1 (6). Interestingly, at higher ligand concentrations, the attachment of a second molecule of 5 to Stx1-B₅ and to Stx2 was observed (Figure 9). In this case, the binding modes for the first and second ligand are necessarily different, with different K_a values. Values of 1.2 x 10⁵ (K_a,1) and 1.0 x 10⁴ M^–1 (K_a,2) were determined for the attachment of the first and second molecule of 5, respectively, to Stx1-B₅; values of 1.5 x 10⁵ (K_a,1) and 0.5 x 10⁴ M^–1 (K_a,2) were determined for Stx2. There was no evidence of the attachment of a second molecule of 6 to either Stx1-B₅ or Stx2 at the concentrations investigated.

View larger version (17K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 6 NanoES mass spectra of solutions consisting of 10 mM ammonium acetate (pH 7), 4 µM Stx1-B5 and (A) ligand 3 at a concentration of 23 µM and (B) ligand 4 at a concentration of 20 µM; 6 µM Stx2 and (C) ligand 3 at a concentration of 11 µM and (D) ligand 4 at a concentration of 17 µM. The designation q indicates the number of bound ligand molecules.

 

View larger version (15K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 7 Distributions of the relative abundance of unbound Stx1-B5 or Stx2 and their specific complexes with ligands 3 and 4. Each distribution represents the average of 15–30 measurements.

 

View larger version (14K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 8 NanoES mass spectra of solutions consisting of 10 mM ammonium acetate (pH 7) and (A) 4 µM Stx1-B5 and 2 µM of 5, (B) 4 µM Stx1-B5 and 2 µM of 6, (C) 7 µM Stx2 and 4 µM of 5 and (D) 7 µM Stx2 and 4 µM of 6.

 

View larger version (20K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 9 NanoES mass spectra of solutions consisting of 10 mM ammonium acetate (pH 7) and (A) 4 µM Stx1-B5 and 8 µM of 5, (B) 7 µM Stx2 and 15 µM of 5.

 

	Discussion

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
It is useful to begin by considering the affinities measured for Stx1-B₅ with the univalent and oligovalent ligands. Notably, the K_a value measured for 1 (360 ± 10 M^–1) is in reasonable agreement with a value of 208 ± 15 M^–1 [originally reported as a dissociation constant of 4.8 mM (Soltyk et al. 2002)] determined by surface plasmon resonance, and a value of 470 M^–1 (Kitov and Bundle 2001), estimated from the IC₅₀ value determined by competitive ELISA. The measured K_a value is also consistent with the estimated value of (0.5–1) x 10³ M^–1 determined by ITC (St Hilaire et al. 1994). The agreement between the K_a value determined for Stx1-B₅ and 1 in the present study and values obtained by different analytical methods offers additional evidence that the direct ES-MS assay can provide meaningful affinity data for protein-carbohydrate and, more generally, protein-ligand complexes.

Of the seven monodeoxy analogs of 1 investigated, only the 3''-deoxy analog (8) exhibited significantly weaker binding to Stx1-B₅ than 1 (Table II). This result indicates that the 3'' OH group is critical to binding. According to the hydrogen bond scheme suggested by the crystal structure of Stx1-B₅ and Pk-MCO, the 3'' OH group forms a hydrogen bond with side chain of Arg33 in binding site 2 but does not participate in hydrogen bonds at site 1 or site 3. Although precise K_a values could not be obtained for all of the remaining monodeoxy analogs, the present data suggest that the individual contributions of these OH groups are similarly small, less than 1 kcal mol^–1.

It is interesting that, for Stx1-B₅, replacement of the OMe group of 1 with NHAc (2) leads to an increase in the magnitude of K_a, by a factor of 2.5. However, replacement of the OMe group with the O(CH₂)₂Si(CH₃)₃ aglycone (14) has no significant effect on binding. As described above, site 2 of the B subunits of Stx1 is believed to be the dominant Pk-binding site (Shimizu et al. 1998). Analysis of the crystal structure for Stx1-B₅ complexed with Pk-MCO suggests that this carbohydrate-binding site is defined primarily by interactions between the Pk-trisaccharide and the amino acids, Asp16, Asn32, Arg33, Asn55 and Phe63. The superimposition of 2 with Pk-MCO in site 2 in the crystal structure suggests the possibility of a hydrogen bond between the side chain of Asn55 and the carbonyl oxygen of the NHAc aglycone. The formation of this additional interaction would account for the larger K_a value for 2, compared to 1. Although the O(CH₂)₂Si(CH₃)₃ aglycone found in 14 differs considerably from the ceramide group in Gb3, this aglycone is, nevertheless, hydrophobic. The similarity in K_a values for 1 and 14 is indirect evidence that the lipid portion of Gb3 does not contribute significantly to the affinity of the Stx1 B subunit for Gb3, at least for interactions involving site 2.

Ligands 3 and 4 are divalent analogs of 2, consisting of two molecules of 2 bound through a covalent linker, which were designed to allow each trisaccharide to simultaneously interact with site 1 and site 2 in the B subunit of Stx1 (Ling et al. 1998). The affinity amplification for ligands 3 [(1.0 ± 0.3) x 10³ M^–1)] and 4 [(4.0 ± 0.6) x 10³ M^–1)], compared to the monovalent ligand 2, with Stx1-B₅ confirms that site 1 can play a role in the toxin binding to the cell surface Gb3 receptors. The somewhat lower K_a value determined for 3, compared to 4, is consistent with the larger linker length in 3 (19 Å) compared to 4 (17 Å) (Kitov, Shimizu, et al. 2003). The greater flexibility of the linker in 3 translates into a larger loss of conformational entropy upon binding to the toxins (Kitov and Bundle 2003). The K_a values measured for Stx1-B₅ with the oligovalent ligands 5 [(1.1 ± 0.3) x 10⁵ M^–1)] and 6 [(1.4 ± 0.2) x 10⁶ M^–1)] are in good agreement with values obtained by competitive inhibition ELISA for binding of Stx1 holotoxin to 5 (2.8 x 10⁵ M^–1) and 6 (2.8 x 10⁶ M^–1) (Kitov and Bundle 2003). The agreement in K_a values lends support to the assumption that the influence of the A subunit on the binding properties of Stx1, at least with respect to the binding at site 2, is negligible. The increase of four and five orders of magnitude in K_a for 5 and 6, respectively, compared to 1 has been explained in terms of a multivalency effect (Kitov and Bundle 2003). Ligands 5 and 6 have pseudo-radial symmetry for three and four copies of the Pk-trisaccharide, respectively, that are connected to a core unit by flexible linkers. When bound to the toxin, each molecule of 5 and 6 can simultaneously engage site 2 of three or four B subunits, respectively. At concentrations of 5 much higher than 1/K_a, the addition of a second ligand occurs and the corresponding [B₅ + 2(5)]ⁿ⁺ and [Stx2 + 2(5)]ⁿ⁺ ions are detected (Figure 9). The large difference between K_a,1 10⁵ and K_a,2 10⁴ M^–1 can be explained by the significantly smaller multivalency effect (both statistical factor and specific interaction free energy) for the formation of the 1:2 complex compared to the 1:1 complex. The attachment of a second molecule of 6 to Stx1-B₅ does not occur, at least at the concentrations investigated. The absence of the 1:2 complex reflects the very small K_a,2 value, which is expected to be similar to the K_a for the univalent interaction between Stx1-B₅ and 1, i.e., 400 M^–1, because the other four binding sites are occupied by the first bound ligand.

Perhaps, the most significant finding of the present study is the similarity in the affinities determined for Stx1-B₅ and Stx2 for each of the univalent and oligovalent ligands considered, with the exception of 2. The similarity in the K_a values measured for Stx1-B₅ and Stx2 binding to 1 and the monodeoxy analogs confirms that both toxins recognize the Gal(14)ßGal(14)ßGlc carbohydrate motif of the cell surface glycolipid Gb3 and suggests similar interactions with the Pk-trisaccharide. Although the exact nature and location of the Pk binding sites in the B subunits of Stx2 are not known, a comparison of the structures of the B₅ homopentamers of Stx1 and Stx2 suggests the possibility that Stx2 possesses a binding site that is analogous to site 2 of Stx1 and comprises the amino acids Glu15, Ser31, Arg32, Thr55 and Phe62 (Fraser et al. 2004). Interestingly, the similar K_a values measured for 1 and 2 with Stx2 can be rationalized assuming that the putative binding site, which will be referred to as site 2 of Stx2, represents the primary interaction site and assuming similar conformations for the Pk-trisaccharide bound to Stx1-B₅ and to Stx2. In this case, the hydrogen bonding between the carbonyl oxygen of the NHAc aglycone and the B subunit suggested to be responsible for enhanced affinity for Stx1-B₅ and 2, compared to 1, cannot be established in Stx2. Even more compelling evidence favoring site 2 of Stx2 as the primary Pk binding site is found in the similar affinities determined for Stx1-B₅ and Stx2 with 5 (10⁵ M^–1) and with 6 (10⁶ M^–1). These results strongly suggest that the radial distances from the center of B₅ pentamer ring to the primary Pk binding sites in the B subunits are similar in Stx1 and Stx2. The similarity in the affinities measured for Stx1-B₅ and Stx2 with 3 (1–2 x 10⁴ M^–1) and with 4 (4 x 10⁴ M^–1) also suggests that the distances between the 2'-O positions of two Pk-trisaccharides residing in binding site 1 and site 2 are similar for the two toxins. Taken together, the binding data indicate a high degree of similarity in the spatial arrangement and structural characteristics of the Pk binding sites in Stx1 and Stx2.

Finally, it is important to note that the similarity in the binding affinities for Stx1 and Stx2 with the soluble uni- and oligovalent ligands contrasts the results of solid phase assays, in which the Pk-based ligands were immobilized (Nakajima et al. 2001; Head et al. 1991). According to the results obtained from the solid phase assays, Stx1 exhibits a greater affinity for Pk than Stx2, by an order of magnitude. The apparent discrepancy between the results obtained by ES-MS and the solid phase assays may reflect the contribution of site 1 and site 3 to Pk binding in the later studies. As shown here, the participation of site 1 in the binding of the divalent ligands 3 and 4 leads to a measurable increase in K_a. Small differences in the intrinsic affinities of the Pk-trisaccharide for site 1 and site 3 (if present) in Stx1 and Stx2 could account for the different affinities revealed by the solid-phase assay. Similar arguments may explain the differences in the relative contribution of individual OH groups to the binding of Gb3 to Stx1 and Stx2 suggested by the present study and by results obtained by the thin layer chromatography (TLC) overlay assay (Nyholm et al. 1996). From a comparison of the migration distances of the Stx in the TLC chromatograms, the relative contributions of individual OH groups within the Pk-trisaccharide to the binding of the toxins were found to have the following trend: 6'' 6' > 2' > 4'' > 2'' > 3' 3'' (Stx1) and 6'' 6' > 4'' 3'' > 2' > 2'' > 3' (Stx2) (Nyholm et al. 1996). Notably, the 3'' OH group was found to have the smallest contribution for binding of Stx1 and an intermediate contribution for Stx2. In contrast, in the present study, the 3'' OH group was found to be the most important of the OH groups investigated for both Stx1 and Stx2. As discussed above, analysis of the crystal structure of Stx1-B₅ complexed with Pk-MCO suggests that the 3'' OH group participates in intermolecular interactions only in site 2. Consequently, it is possible that the different trends observed here and in the previous studies reflects the differential contributions of site 1 and site 3 to Pk binding.

	Conclusions

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
Using the direct ES-MS assay, our laboratory has carried out the first quantitative, comparative study of the binding of Stx1 and Stx2 with water-soluble uni- and oligovalent ligands based on the Pk-trisaccharide. Analysis of the binding data reveals new information about the nature and location of the carbohydrate binding sites in both toxins. The most important finding of the present study is the similarity in the K_a values determined for Stx1-B₅ and Stx2 binding to 1 and its monodeoxy analogs, 7–13, and to the multivalent ligands, 3–6. Taken together, the binding data suggest a high degree of similarity in the spatial arrangement and structural characteristics of the Pk binding sites in Stx1 and Stx2. Specifically, the similarity in the K_a values measured for Stx1-B₅ and Stx2 binding to 1 confirms that both toxins recognize the -D-Galp(14)-ß-D-Galp(14)-ß-D-Glcp carbohydrate motif of the cell surface glycolipid Gb3. This, taken together with the results of the chemical mapping study, suggests that the nature of the Pk binding interactions is similar in Stx1 and Stx2. The affinities of Stx1-B₅ and Stx2 for the multivalent ligands 5 and 6 provide compelling evidence that the putative site 2 of Stx2, which shares the same spatial arrangement as site 2 in Stx1, is the primary Pk binding site. Finally, the participation of site 1 in Pk binding was shown to enhance ligand avidity in both Stx1 and Stx2.

	Materials and methods

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
Stx holotoxins, Stx1 B subunit and oligosaccharides
The compounds 1 (Garegg and Hultberg 1982), 2 (Kitov, Paszkiewicz, et al. 2003), 14 (Mogemark et al. 2003), 5 (Kitov and Bundle 2003) and 6 (Kitov and Bundle 2003) were obtained according to previously reported procedures. Complete synthesis of deoxy analogs of Pk-trisaccharide 7–13 will be described elsewhere (Kim 2007). Complete synthetic details for the preparation of compounds 3 and 4 are given in Supplementary Data.

Stx1 and Stx2 were expressed and affinity purified using Synsorb-Pk as described previously (Mulvey et al. 1998). The Stx2 A subunit and B pentamer were FPLC-purified from the holotoxin preparations as described by Head et al. (1991). SDS-PAGE analysis combined with the sensitive silver staining technique confirmed that these subunit preparations possessed no holotoxin contamination. This was also confirmed using the Vero cytotoxicity assay. To prepare stock solutions of Stx1, Stx2 and Stx1-B subunit for the ES-MS measurements, the protein samples were dialyzed against 50 mM ammonium acetate (pH 7) and stored at –20°C until needed.

Mass spectrometry
All measurements were performed using an Apex II 9.4 tesla Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (Bruker, Billerica, MA) equipped with a temperature-controlled nanoflow electrospray (nanoES) ion source. The nanoES solutions were prepared from aqueous stock solutions of protein and ligand with known concentrations. Unless otherwise indicated, aqueous ammonium acetate was added to the nanoES solution to yield a final buffer concentration of 10 mM. NanoES was performed using an aluminosilicate capillary (1.0 mm o.d., 0.68 mm i.d.), pulled to 4–7 µm o.d. at one end using a P-2000 micropipette puller (Sutter Instruments, Novato, CA). The electric field required to spray the solution was established by applying a voltage of +900 V to a platinum wire inserted inside the glass tip. The solution flow rate was typically 20 nL min^–1. The droplets and gaseous ions produced by nanoES were introduced into the mass spectrometer through a stainless steel capillary (i.d. 0.43 mm) maintained at an external temperature of 66°C. The ion/gas jet sampled by the capillary (±50 V) was transmitted through a skimmer (–2 V) and stored electrodynamically in an rf hexapole. Unless specified otherwise, a hexapole accumulation time of 5 s was used for all experiments. Ions were ejected from the hexapole and accelerated to 2700 V into the superconducting magnet, decelerated, and introduced into the ion cell. The trapping plates of the cell were maintained at a constant potential of ±2 V throughout the experiment. The typical base pressure for the instrument was 5 x 10^–10 mbar. Data acquisition was controlled by an SGI R5000 computer running the Bruker Daltonics XMASS software, version 5.0. Mass spectra were obtained using standard experimental sequences with chirp broadband excitation. The time domain signal, consisting of the sum of 50–100 transients containing 128 K data points per transient, were subjected to one zero-fill prior to Fourier transformation.

Determination of association constants by the direct ES-MS assay
The equilibrium equation for the association of a protein (P) and a single ligand (L) is given by Eq. (1):

(1)

The equilibrium concentrations, [PL], [P], and [L], can be calculated from the known initial concentrations of protein and ligand in solution, [P]₀ and [L]₀, and the relative abundance of the corresponding bound and unbound protein ions measured in the mass spectrum, i.e., PLⁿ⁺ and Pⁿ⁺ in positive ion mode. Assuming that the ionization and detection efficiencies for the PLⁿ⁺ and Pⁿ⁺ ions are similar (Wang et al. 2003b), the ratio (R) of the ion intensity (I) of the bound and unbound protein ions [I(PLⁿ⁺)/I(Pⁿ⁺)] determined from the mass spectrum is equivalent to the ratio of the concentrations in solution at equilibrium ([PL]/[P]). Since the nanoES process typically produces Pⁿ⁺ and PLⁿ⁺ ions with a distribution of charge states, the R value is obtained from the sum of the intensities for the complex ions divided by the sum of the intensities for the protein ions over all observed charge states. In FT-ICR MS the ion signal is proportional to the abundance and charge state (n) of the ions. Therefore, the measured ion intensities must be normalized for charge state. In this case, R should be calculated from Eq. (2) (Wang et al. 2003b):

(2)

The equilibrium concentration, [PL], can be determined from the value of R and [P]₀ using the following expression:

(3)

The equilibrium concentration [L] can be found from Eq. (4) and then K_a can be determined using Eq. (5):

(4)

(5a)

(5b)

If the protein (or protein complex) can bind multiple copies of molecules L, which is the case for the B₅ homopentamers of Stx1 and Stx2, the mathematical model must include all relevant equilibria. Shown below [Eqs. (6a)–(6N)] are the reactions and equilibrium expressions for sequential binding of L to P:

(6a)

(6b)

(6N)

The equilibrium concentrations, [P], [PL], ..., [PL_N], can be determined from the relative abundance of the corresponding ions observed in the mass spectrum and the equations of mass balance [Eq. (7a) and (7b)].

(7a)

(7b)

After substituting the equilibrium concentration ratios with the corresponding ion intensity ratios, the K_a values can be calculated from Eqs. (6) and (7). For example, the expressions for the apparent K_a values for the sequential binding of one and two ligand molecules, i.e., K_a,1 and K_a,2, to a protein are:

(8)

(9)

where R₁ = [PL]/[P] and R₂ = [PL₂]/[P]. If the binding sites are equivalent, the intrinsic K_a can be determined for each of the ligand binding reactions using the following general expression:

(10)

where p is the number of occupied binding sites and N is the total number of equivalent sites.

In the present study, the B₅ homopentamers were assumed to have five equivalent binding sites (one per subunit) for the univalent (1, 2, 7–13) ligands and the intrinsic K_a values were calculated from Eqs. (6), (7) and (10). The linkers between the trisaccharides in the divalent ligands 3 and 4 are only long enough to allow the trisaccharide units to simultaneously occupy two different types of binding sites, namely site 1 and site 2, on a given subunit of Stx-B₅ but not two sites on adjacent subunits. Consequently, to calculate their K_a values, ligands 3 and 4 were treated as univalent ligands. In contrast, the multivalent ligands 5 and 6 can simultaneously bind to as many as three or four subunits, respectively, in the B₅ pentamers. However, the B subunits that are not occupied by the first bound ligand are able to participate in binding with a second ligand molecule. As a result, two molecules of 5 and 6 can, in principle, bind to each B₅ pentamer. The apparent K_a,1 and K_a,2 values for the addition of the first and second ligand molecule, respectively, were calculated using Eqs. (8) and (9).

During the ES process nonspecific binding between the protein and the carbohydrate ligand may occur. This will influence the measured intensities for the unbound protein ions, Pⁿ⁺, and the ions of the specific protein-ligand complexes, PLⁿ⁺, PL₂ⁿ⁺, ... , PL_Nⁿ⁺. The errors in binding stoichiometry and affinity introduced by nonspecific binding are most significant for weakly interacting complexes, K_a 10⁴ M^–1, because relatively high concentrations of ligand are required to produce detectable levels of specific complex in solution. Recently, our laboratory reported a straightforward and quantitative method to account for the contribution of nonspecific protein–ligand binding in ES-MS (Sun et al. 2006). Briefly, this method involves the addition of a reference protein (P_ref), which exhibits no specific binding activity for the ligand(s) of interest, to the ES solution. The distribution of ligands bound nonspecifically to P_ref can be used to correct the intensities measured for the PLⁿ⁺, PL₂ⁿ⁺, ... , PL_Nⁿ⁺ ions for the contribution of nonspecific ligand binding according to the following general expression:

(11)

where I(Pⁿ⁺), I(PLⁿ⁺), ... , I(PLⁿ⁺_N) are the expected intensities for the PLⁿ⁺, PL₂ⁿ⁺, ... , PL_Nⁿ⁺ ions, I_app(Pⁿ⁺), I_app(PLⁿ⁺), ... , I_app(PL_Nⁿ⁺) are the measured ion intensities, and f_0,Pref, f_1,Pref, ..., f_N,Pref represent the fraction of P_ref bound nonspecifically to 0, 1, ..., N molecules of L.

	Funding

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
This work was generously supported by the Alberta Ingenuity Centre for Carbohydrate Science and the Natural Sciences and Engineering Research Council of Canada.

	Conflict of interest statement

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
None declared.

	Abbreviations

 
ELISA, enzyme-linked immuno sorbent assay; ES-MS, electrospray ionization mass spectrometry; FT-ICR, Fourier transform ion cyclotron resonance; ITC, isothermal titration calorimetry; nanoES, nanoflow electrospray ionization; Stx, Shiga toxin; TLC, thin layer chromatography

	References

Top Abstract Introduction Results Discussion Conclusions Materials and methods Funding Conflict of interest statement References

 
Binnington B, Lingwood D, Nutikka A, Lingwood CA. Effect of globotriaosyl ceramide fatty acid -hydroxylation on the binding by Verotoxin 1 and Verotoxin 2. Neurochem Res (2002) 27:807–813.[CrossRef][ISI][Medline]

Calderwood SB, Auclair F, Donohuerolfe A, Keusch GT, Mekalanos JJ. Nucleotide-sequence of the Shiga-like toxin genes of Escherichia coli. Proc Natl Acad Sci USA (1987) 84:4364–4368.[Abstract/Free Full Text]

Dam TK, Brewer CF. Multivalent protein-carbohydrate interactions: Isothermal titration microcalorimetry studies. Energetics of biological macromolecules. Methods Enzymol (2004) 379:107–128.[ISI][Medline]

Daneshfar R, Kitova EN, Klassen JS. Determination of protein-ligand association thermochemistry using variable-temperature nanoelectrospray mass spectrometry. J Am Chem Soc (2004) 126:4786–4787.[CrossRef][ISI][Medline]

Fraser ME, Chernaia MM, Kozlov YV, James MNG. Crystal-structure of the holotoxin from Shigella dysenteriae at 2.5 Å resolution. Nature Struct Biol (1994) 1:59–64.[CrossRef][ISI][Medline]

Fraser ME, Fujinaga M, Cherney MM, Melton-Celsa AR, Twiddy EM, O’Brien AD, James MNG. Structure of Shiga toxin type 2 (Stx2) from Escherichia coli O157: H7. J Biol Chem (2004) 279:27511–27517.[Abstract/Free Full Text]

Garegg PJ, Hultberg H. Synthesis of disaccharides and tri-saccharides corresponding to receptor structures recognized by pyelonephritogenic E. coli fimbriae (pili). Carbohydr Res (1982) 110:261–266.[CrossRef][ISI][Medline]

Head SC, Karmali MA, Lingwood CA. Preparation of Vt1 and Vt2 hybrid toxins from their purified dissociated subunits – evidence for B-subunit modulation of a subunit function. J Biol Chem (1991) 266:3617–3621.[Abstract/Free Full Text]

Itoh K, Tezuka T, Inoue K, Tada H, Suzuki T. Different binding property of Verotoxin-1 and Verotoxin-2 against their glycolipid receptor, globotriaosylceramide. Tohoku J Exp Med (2001) 195:237–243.[CrossRef]