The Dramatic Modulatory Role of the 2 ' N Substitution of the Terminal Amino Hexose of Globotetraosylceramide in Determining Binding by Members of the Verotoxin Family

Although globotetraosylceramide (Gb4) is only recognized by a single member of the verotoxin family namely, the pig edema disease toxin (VT2e), removal of the acetyl group from the terminal N-acetyl hexosamine of Gb4 to generate the free amino sugar containing species (aminoGb4) results in the generation of a glycolipid preferentially recognized by all members of the verotoxin family (i.e., VT1, VT2, VT2c, and VT2e). GT3, a site-specific mutant of VT2e, in which Gb4 recognition is lost but Gb3 binding is retained, also binds aminoGb4. We have now compared the binding of VT1, VT2, VT2e, and GT3 to a series of aminoGb4 derivatives using a TLC overlay technique. DimethylaminoGb4 is bound by VT1 and VT2 but not VT2e or GT3; formylaminoGb4 binds all toxins but poorly to VT2 and preferentially VT2e; trifluoroacetylaminoGb4 binds only VT2e and GT3; isopropylaminoGb4 binds VT1 and poorly to VT2; benzylaminoGb4 binds all four toxins. Thus, there is a marked distinction between the permissible amino substitutions for VT1 and VT2e binding. GT3 is a hybrid between these in that, according to the substitution, it behaves similarly either to VT1 or to VT2e. For each species, GT3 does not however, show a hybrid binding between that of VT1 and VT2e. Analysis of the binding as a function of pH shows OPEN ACCESS Chromatography 2015, 2 530 opposite effects for VT1 and VT2e: decreased pH increases VT1, but decreases VT2e receptor glycolipid binding.


Introduction
Verotoxins (Shiga toxins) are a family of E. coli elaborated subunit toxins involved in microvascular disease [1,2].VT1, VT2, and VT2c are produced by E. coli strains which colonize and affect humans, particularly children, whereas, VT2e is found in strains which infect pigs and is the cause of pig edema disease [3].The pentameric B subunit of these toxins bind to a glycolipid receptor, globotriaosylceramide (galα1-4-galβ1-4 glucosylceramide, Gb3) [4].VT2e, however, binds to globotetraosylceramide (galNAcβ1-3galα1-4-galβ1-4glucosylceramide, Gb4), the next homologue in the globo series of glycolipids, in addition to Gb3 [5].Although homologous, antibodies against VT1 or toxins of the VT2 series do not cross-neutralize cytotoxicity [6].VT1 and VT2 are approximately 60% identical, whereas VT2 and VT2e are approximately 84% identical.Site specific mutagenesis of amino acids within the B subunit of VT2e (binds Gb3 + Gb4) to the corresponding amino acid found in VT2 (binds Gb3 only), identified a double mutant (Gln64Glu/Lys66Gln,) termed GT3, in which the binding of Gb4 was lost but that of Gb3 retained [7].This mutant showed an in vivo pathology which was similar to that of VT1 (Gb3 binding-mediated) rather than that of VT2e [8].
Molecular modeling, in combination with an analysis of the binding of these toxins to deoxyGb3 analogues resulted in the identification of two potential Gb3 binding sites per B subunit monomer [9].One site (cleft site or site 1) was between adjacent B subunit monomers while the other was in a shallow depression on the B subunit pentamer surface which opposes the plasma membrane of the target cell (site 2).By homology modeling, the binding of Gb4 by VT2e and the lack of Gb4 binding by GT3, were explained in terms of binding within the cleft site.The lack of binding of Gb4 by the other verotoxins was consistent with binding in this site.In particular, the N-acetyl group of the terminal N-acetylgalactyosamine (galNAc) sugar of Gb4 was identified as a primary basis for the lack of Gb4 binding.An intramolecular H-bond from the 3OH of the galNAc to the acetyl oxygen prevented any reorientation of the NAc group to allow Gb4 binding.Chemical removal of the N-acetyl moiety [10] to give the free amino sugar containing GSL, resulted in a species (aminoGb4) which was highly effectively recognized by all members of the verotoxin family [9].Docking of aminoGb4 within the cleft site of VT1 demonstrated a salt bridge between the ammonium cation and the carboxylate anion of asparagine 16.The formation of such a stable linkage between the receptor and ligand provided a basis for the surprising observation of aminoGb4 binding by all verotoxins and provided support for cleft site Gb3 binding.
The co-crystal structure of the VT1 B subunit and a lipid-free Gb3 oligosaccharide analogue was solved [11].While the receptor density was found to correspond to the sites identified by modeling, the relative orientation of the sugar sequence within the binding sites, particularly site 2, were different from that modeled.In addition, density was associated with Trp34 (site 3).The structure of a similar VT1 complex was solved by NMR [12] which confirmed the preferential occupancy and sugar orientation of site 2 as in the crystal structure.Mutational studies provided evidence for the importance of all three sites [13] but discrepancies between lipid-free sugar binding and cytotoxicity were evident [14].Various globotriaose multimers binding in site 2, have proven effective in blocking cytotoxicity in vitro and in vivo [15][16][17].The import of site 3 remains less well established [18,19].This site is obstructed by the A subunit C terminus in VT2 [20] and empty in Gb3 sugar co-crystals [21].Similarly, this site was empty in the GT3/Gb3 sugar co-crystal [22] but was proposed as a potential secondary Gb4 binding site.Site 1 was partially occupied while site 2 was fully occupied in the co-crystal.The question of which site(s) bind membrane Gb3 glycolipid may relate to the conformational differences of the GSL sugar relative to the membrane [23,24].This is strongly influenced by the membrane cholesterol content [25], promoting a parallel conformation along the membrane surface, rather than a perpendicular orientation in which the carbohydrate protrudes from the membrane [26,27].Such distinct Gb3 conformers may bind in site 2 and site 1 respectively [28] as proposed [9].
Substitution/deletion of key residues in sites 1, 2, and 3 have shown all three sites to be important [13].Gb3 sugar binding only requires site 2, but membrane Gb3 glycolipid binding requires all sites [14].Recent studies show amino substitution within the terminal αgalactose of Gb3 can enhance VT2 binding [29,30], although such structures are not made by mammalian cells.The X-ray structure of VT2 [20] and the co-crystal structure with such an "aminoGb3" disaccharide derivative [21] supports binding in both sites 1 and 2.
Evidence indicates that the lipid moiety of membrane Gb3 has a major impact on the recognition of the Gb3 carbohydrate by the verotoxin B-subunit [24,31,32].Lipid-free oligosaccharides may represent an inadequate model of verotoxin/membrane Gb3 binding under physiological conditions.We now present evidence on the discriminatory binding of aminoGb4 derivatives consistent with site 1 occupancy.

Synthesis of NMe2, NHisoPr, or NHBz derivatives of Gb4•NH2
To a solution of Gb4•NH2 (2 mg, approximately 2 µmol) in dry chloroform (1 mL) were added approximately 3 equivalents (6 µmol) of the carbonyl reagent (formaldehyde, acetone, or PhCHO) and solid NaCNBH3 (6 mg, 0.1 mmol).After stirring the reaction mixture for 16 h at room temperature (25 °C), chloroform was removed under a stream of N2 and the remaining solid was then dissolved, by sonication, in 5 mL of distilled water.The resulting suspension was passed through a C-18 reverse phase cartridge, washed with 20 mL of water and eluted with 20 mL of ammonical methanol (0.2 M NH3).Reaction products were further purified by silica gel column (2 × 2 cm) chromatography using C:M:H2O; 60:30:2.Yields by TLC were greater than 85%.

Synthesis of Gb4•NHCOCF3
Acylating reagent trifluoroacetyl imidazole was prepared by adding a DCM solution of the anhydride ((F3CCO)2O, 0.85 g, 2.7 mmol) divided in 3 portions (1 mL) at 15 min intervals to an imidazole (0.41 g, 6.0 mmol) suspension in DCM (3 mL).The reaction mixture was stirred for 2 h and was assumed to be approximately a 0.5 M solution of the imidazole derivative.
A solution (20 µL, 10 µmol) of the imidazole derivative was added to a solution of Gb4•NH2 (1 mg) in methanol (1 mL) and the reaction was monitored by TLC (C:M:H2O S ; 60:35:8).Appearance of many orcinol positive products suggests some degree of acylation of OH groups.Once all the Gb4•NH2 was consumed, ammonia (0.1 mL of 2 M NH3 in EtOH) was added and the mixture was dried under a stream of N2.If a significant degree of O-acylation had occurred, the dried residue was treated with a solution of Et3N:M:H2O; 2:6:10 (0.5 mL/mg of GSL) at room temperature and monitored by TLC (C:M:H2O Salt ; 60:35:8 −0.25% KCl).Once the O-acyl group hydrolysis was complete the material was dried under a stream of N2, and purified on a silica gel column (0.5 × 2 cm) using C:M:H2O; 60:30:2.Yield 75%.

Mass Spectroscopic Analyses
Pure samples were dissolved in 9:1, M:H2O mixture containing 0.1% NaCl and the ES spectra were recorded on a SCIEX API III spectrometer.The acyl derivatives, natural Gb4 which has a terminal N-acetyl galactosamine or the corresponding formyl group of the trifluoro acetyl analogues predominantly give the sodium adduct.However the aminoGb4 and its alkylated derivatives predominantly gave the proton adduct.

TLC Overlay
Toxins were purified as previously described [34].The site-specific mutants were previously described [14].Toxin binding to GSLs was assayed by TLC overlay as previously described [35].GSL-bound VT1 and the VT1 B-subunit mutants were detected with mAb PH1 while VT2e, GT3, and VT2 were visualized using rabbit anti-VT2e [8] (a generous gift of Dr. Carlton Gyles, University of Guelph).
Receptor ELISA (RELISA) Glycolipids were coated on microtitre plate wells by evaporation from ethanolic solution.GSL aliquots were measured from stock solutions in dichloromethane/methanol (2:1, v/v) into screw-capped glass tubes, and the solvent was removed under a stream of nitrogen.Ethanol was added to the lipid residue, and a uniform glycolipid solution was prepared by immersing the tube for 30 s in a bath sonicator followed by 1 min of vortexing.A 100 µL aliquot of ethanolic GSL (0.5 µg/mL) was added per well in triplicate and the ethanol was allowed to evaporate at room temperature overnight.Once dry, the plates were stored in a desiccator at 4 °C.Wells were blocked with 200 µL of 0.2% (w/v) BSA in 10 mM sodium phosphate/150 mM NaCl, pH 7.4 (BSA-PBS) for 1-2 h, then the wells were washed twice with 150 µL/well 0.2% BSA-PBS.The following incubations were carried out in BSA-PBS at room temperature for 1 h.VT (200 ng/mL), primary antibody (mouse anti-VT1 mAb PH1, 1 µg/mL or rabbit anti-VT2e, 1/2000), goat anti-mouse or rabbit IgG-HRP conjugate (1/2000).Between incubations, wells were washed three times with 200 µL of 0.2% BSA-PBS.Finally wells were incubated with substrate solution, 0.5 mg/mL 2,2'-azino-bis-(3-ethylbenzthiazoline-6-sulphonic) diammonium salt (ABTS), 0.01% H2O2 in 0.08 M citric acid/0.1 M disodium phosphate, pH 4.0.After sufficient color had developed (usually 10-20 min), the absorbance of each well at 405 nm was measured using an ELISA plate reader.

Results
The structures of the aminoGb4 derivatives studied are shown in Figure 1.Mass spectrometry confirmed the structures of formylaminoGb4, TFAaminoGb4, and dimethyl-aminoGb4 (Tables 1 and 2).The isopropyl and benzyl species were found to be monosubstituted (Table 2).
TLC overlay assay of the binding of VT1, VT2e, GT3, and VT2 are shown in Figure 2. A distinct pattern of recognition of the aminoGb4 derivatives is seen for each toxin.As expected, Gb4 was bound only by VT2e, and aminoGb4 and Gb3 were bound equally by all toxins, except VT2 which preferentially bound aminoGb4.DimethylaminoGb4 and isopropylaminoGb4 were bound only by VT1 and VT2 but VT2 bound significantly less isopropylaminoGb4.FormylaminoGb4 bound all four toxins, but poorly to VT2 and preferentially VT2e, while benzylaminoGb4 bound all four toxins equally.TFAaminoGb4 was strongly bound by VT2e and GT3 but not at all by VT1 or VT2.The aminoGb4 derivative binding is thus divided into two groups: VT1/VT2 (bind dimethyl and isopropyl) and VT2e/GT3 (bind TFA but not dimethyl or isopropyl).VT1 and VT2 are distinguished by the reduced binding of VT2 to Gb3 and isopropylaminoGb4, while VT2e and GT3 are distinguished by the Gb4 and increased formylaminoGb4 binding of VT2e (Table 3).Effect of pH.The binding of VT2e and VT1 were then compared at different pHs.Glycolipid receptor binding was assayed by TLC overlay at pH 9.0, 6.0, and 4.5 (Figure 3A).For VT1, all species tested, Gb3, aminoGb4, dimethylaminoGb4, isopropylaminoGb4 and benzylaminoGb4 were less strongly bound at high pH.In contrast, VT2e binding to these aminoGb4 derivatives was reduced at low pH.Gb3, Gb4, and aminoGb4 were strongly bound by VT2e at pH 9.0.Significant VT2e binding to benzylaminoGb4 was seen but dimethylaminoGb4 and isopropylaminoGb4 were not recognized at any pH.Binding to Gb3, Gb4, and aminoGb4 and benzylaminoGb4 was reduced at pH 6.0 and further at pH 4.5.The effect of pH on aminoGb4 derivative binding by VT1, VT2e, and GT3 was also assayed by RELISA (Figure 3B).Consistent with TLC binding, the dimethyl and isopropyl species were only significantly bound by VT1.VT1 binding Gb4, not detected by TLC overlay, was however, significant by RELISA and remarkably, acid pH dependent, being greatly reduced at physiological pH, and eliminated at high pH.Gb4 RELISA binding was also pH sensitive for GT3, but not VT2e.VT1 and GT3 Gb3 binding were also reduced at increased pH.At physiological pH, a four-fold preference for Gb3 over Gb4 was maintained for VT1 but GT3 Gb3/Gb4 binding was equivalent.Of the aminoGb4 derivatives, only TFA-aminoGb4/VT1 binding was reduced at high pH.TFAaminoGb4 binding was not pH sensitive for VT2e or GT3.VT2e binding benzylaminoGb4 was reduced at low pH (as seen by TLC-Figure 3A).AminoGb4/VT2e binding was reduced at low pH but recognition of other species, notably Gb3 and Gb4 was unaffected.Only VT2e bound the pentahexoside Forssman glycolipid.This binding has not been previously reported.GT3 showed pH dependent Gb3 and Gb4 binding (similar to VT1), but did not bind dimethyl or isopropylaminoGb4 (similar to VT2e).Binding to the control GSLs, gangliotriaosylceramide (Gg3) and lactosylceramide (LacCer) was not seen at any pH, ruling out non-specific interactions induced by high or low pH treatment of the toxins.
Site specific mutants.Several site-specific VT1 mutants (and double mutants), in which a key amino acid within one of the three proposed receptor binding sites was deleted, were assessed for binding the aminoGb4 species by TLC overlay (Figure 4).Phe30 is a key residue in site 1 and, to a lesser extent, site 2 [14].Its mutation to alanine resulted in the loss of Gb3 and Gb4, TFAaminoGb4, and aminoGb4 binding (In this experiment, the formylaminoGb4 sample contained a trace of unreacted aminoGb4, below the limit of detection by orcinol but bound by wildtype VT1).Phe30Ala retains residual binding to dimethyl, isopropyl and benzyl-aminoGb4.This was consistent with RELISA (Figure 5) which showed similar loss of Gb3, aminoGb4 and formylaminoGb4 binding while dimethyl-and benzyl-aminoGb4 binding were retained.Glycine 62 is in site 2. For this mutant, binding to all species was deleted (Figure 4).However, in the combination double mutant, binding to dimethyl-, isopropyl-and benzyl-aminoGb4 was prevented and binding was similar to Phe30Ala.The loss of Alanine 56, also an important residue in site 2, showed a milder phenotype in which binding to all aminoGb4 derivatives, except TFA and formyl species, was retained.Asp 17 is key in site 1 [14].Mutation to glutamic acid deleted Gb3 and TFA and formyl-aminoGb4 recognition but dimethyl-, isopropyl-and benzyl-aminoGb4 binding were fully retained.Deletion of site 3 by mutation of tryptophan 34 which lines the B pentamer central pore [11] resulted in the loss of all binding.This phenotype was retained when combined with Asp17Glu.

Discussion
Following the original discovery that VT2e, unlike the other members of the verotoxin family, can bind to Gb 4 in addition to Gb3 [5], it was assumed it is the presence of the additional terminal sugar which prevents VT1, VT2 ,and VT2c from binding Gb4.Our observation that aminoGb4 is preferentially recognized by VT1, VT2, and VT2c, even compared with Gb3 [9], demonstrated this is not the case.Thus the cleft site of all verotoxins can accommodate a four sugar containing globo-series glycolipid.
We consistently observe a lack of VT1 or VT2 binding to Gb4 by TLC overlay, but other studies have reported a weaker receptor function for Gb4 [32,36,37].The current and previous studies show that VT1 binding to solid phase Gb4 can be induced under some conditions [38].We have also anecdotally found that 'old' toxin begins to show some Gb4 recognition.This could be interpreted as increased accommodation by the subunit cleft site (site1) due to some degree of pentamer long-term instability [39].
The basis for the lack of Gb4 binding by VT1, VT2, and VT2c is therefore subtle.In our early molecular modeling studies, we proposed that the pocket between the side chains of Asp16 and Asp17 in VT1 was not sufficient to accommodate the acetamido group of Gb4 and that the unfavorable orientation of the carboxyl group of the acetamido group towards the carboxyl group of Asp16 would prevent Gb4 binding.The N-acetyl group cannot reorient to accommodate this unfavorable interaction because of an intramolecular hydrogen bond from the 3' hydroxyl of the GalNAc residue to the carbonyl oxygen of the N-acetyl group [9].In VT2e, a pocket between Asp16 and Glu15 accommodates the acetamido group of the terminal GalNAc of Gb4 to allow binding.The NAc group does not stabilize the interaction in any way.The TFA derivative is larger than the NAc group and on this basis, should only bind VT2e and GT3 as observed.
The pH dependency of VT1 binding is consistent with an important role for the charge on the nitrogen of the aminohexose for binding, presumably in the formation of a salt bridge with the carboxyl group of Asp16 [9].The lack of VT1 recognition of TFAaminoGb4 and reduced recognition of formylaminoGb4 is consistent with the repulsion of the carbonyl oxygen of the acetyl group of the aminohexose of Gb4 by the carboxyl group of Asp16 as proposed to explain the lack of VT1/Gb4 binding [9].In formylaminoGb4, although the carbonyl oxygen will be restricted in the same way as in Gb4, the volume of the proton (of the formyl group) as opposed to the methyl group (of the acetyl group) will be less.This may allow some access to the trough between Asp17-16 side chains [9] to explain the low but significant VT1/formylaminoGb4 binding.Once this oxygen is removed, as in isopropylaminoGb4 and dimethylaminoGb4, binding to VT1 is allowed.From our published model [9], the amine of aminoGb4 in site 1 is directed towards solvent.Thus, when the constraint of the carbonyl oxygen is removed, there should be little size restriction for the amino substituent to retain binding.This would explain the isopropylaminoGb4/VT1 binding.
The current data indicates that the binding of aminoGb4 by VT2e does not involve the formation of a salt bridge.In contrast to VT1, the receptor binding of VT2e is reduced at lower pH.AminoGb4 is not bound by VT2e at pH4.5.This is not consistent with the involvement of a salt bridge in VT2e/aminoGb4 recognition.In VT2e, the trough between Asp16 and Glu15 is larger than that between Asp17 and Asp16 in VT1 and can accommodate the NAc group of Gb4 [9].Similarly, this trough could accommodate the TFA and formyl groups to explain the VT2e binding of TFAaminoGb4 and formylaminoGb4.The nitrogen of aminoGb4 docked in site I of VT2e is too far from the carboxyl group of Asp16, and may also be too far from that of Glu15, to form a salt bridge [9].Moreover, the region in VT2e corresponding to the 'Asp loop' of VT1 is less negatively charged and could be neutralized at acid pH.The absence of a salt bridge, could explain the pH dependency of VT2e binding.The VT2e B subunit has an overall charge of 2+ [40] which would disfavor aminoGb4 binding, particularly at low pH.However, the lack of VT2e binding to dimethylaminoGb4 and isopropylaminoGb4 cannot be explained without detailed modeling of these complexes.In the absence of a salt bridge, the amine protons, missing in dimethylaminoGb4, may be required for H-bonding to stabilize the VT2e complex, explaining the lack of VT2e recognition.It is also possible that this is a steric problem.Although we had planned to generate the dibenzylaminoGb4 derivative, mass spectrometry showed mono substitution had occurred.Thus an amide proton is available to explain the VT2e binding.However the amide proton is also available in isopropylaminoGb4 yet this is not bound by VT2e.The isopropyl group is considerably bulkier than the benzyl and this may account for the difference.
It is to some extent ironic that while VT2e is the only toxin to bind Gb4, it is the only toxin which will not bind the dimethyl or isopropyl derivatives.Indeed all the species substituted with hydrophobic groups are poorly recognized by VT2e but effectively bound by VT1.Those substitutions of aminoGb4 which have little effect on VT2e binding (formyl, TFA) remove VT1 recognition.
Our original model [9] explained the lack of binding of Gb4 by the double mutant GT3 and the fact that the single mutation in VT2e of either Gln64Glu or Lys66Gln alone has no effect on Gb4 binding.The modeling of Gb4 binding to VT2e, based on the co-crystal structure of Gb3 oligosaccharide in site 1 of VT1 [41], is consistent with the loss of Gb4 binding by GT3 since the terminal galNAc interacts with both VT2e Lys66 and Gln64.However, this study proposes two H-bonds from Lys13 and Glu16 to the NAc carbonyl oxygen which would be lost in the case of aminoGb4.Thus the binding of the lipid-free Gb 3 sugar in site 1 [11] as determined from the co-crystal, does not provide an explanation for the high affinity binding of aminoGb4.
Modeling of GT3 with Gb3 or aminoGb4 was not performed and thus, an explanation for the binding specificity of GT3 must be speculative.Overall, the present results show receptor binding is more complex than previously envisioned.In terms of the importance of the carbonyl oxygen on the amide nitrogen, GT3 behaves as VT1 and does not bind formylaminoGb4, consistent with our proposal that the trough between Asp16 and Glu15 is compromised in GT3 [9].However GT3 does bind TFAaminoGb4, unlike VT1, suggesting that this trough is distinct from that between Asp17 and Asp16 in VT1.Site 1 or site 2 modification blocks Gb3, aminoGb4, formylaminoGb4 recognition with little effect on dimethylor isopropyl-aminoGb4 binding, suggesting a different binding mechanism for these latter GSL analogues, likely related to the lack of binding of the latter group by VT2e and GT3.The resolution of such questions however, must await the crystal structure of GT3, VT2e and the modeling of their site 1 / 2 docked receptors.
In summary, we have documented a remarkable dependence on the amino substitution of aminoGb4 in terms of variable binding to VT1, GT3, and VT2e, as well as a distinct pH dependency for VT1 as opposed to VT2e with regard to glycolipid receptor binding.For the most part, these differences are consistent with our previous modeling of aminoGb4 binding in site 1 of the verotoxin B subunit pentamer.

Figure 4 .
Figure 4. Binding of VT1 mutants to aminoGb4 derivatives.The aminoGb4 derivatives were separated by TLC and chemically stained for carbohydrate with orcinol.The binding of these derivatives to wildtype VT1 was compared to that of the site deletion mutants Phe30Ala (site 1), Ala56Tyr (site 2), Gly62Ala (site 2), Trp34Ala (site 3), Arg17Glu (site 1), and the double mutant Phe30Ala/Gly62Ala by TLC overlay.

Figure 5 .
Figure 5. Phe30Ala VT1 B mutant retains recognition of an aminoGb4 derivative subset.Dose response RELISA shows while binding of F30A VT1 B-subunit to Gb3 is minimal, binding to aminoGb4 derivatives, except formylaminoGb4, is retained.

Table 1 .
Tabulation of major subspecies (i.e., having different acyl chains) of Gb4 and Gb4-acyl derivatives and their corresponding molecular ions determined by mass spectrometry.ND not detected.N = acyl chain carbon number.Gb4 = globotetraosylceramide.

Table 2 .
Major subspecies (i.e., having different acyl chains) of amino Gb4 derivatives and their corresponding molecular ions.

Table 3 .
Summary of GSL binding.