Three types of binding assays were used to study the binding of Bacillus thuringiensis delta-endotoxin Cry1Ac to brush border membrane vesicle (BBMV) membranes and a purified putative receptor of the target insect Manduca sexta. Using hybrid proteins consisting of Cry1Ac and the related Cry1C protein, it was shown that domain III of Cry1Ac is involved in specificity of binding as observed by all three techniques. In ligand blotting experiments using SDS–PAGE-separated BBMV proteins as well as the purified putative receptor aminopeptidase N (APN), the presence of domain III of Cry1Ac in a hybrid with Cry1C was necessary and sufficient for specific binding to APN. Using the surface plasmon resonance (SPR) technique with immobilized APN, it was shown that the presence of domain III of Cry1Ac in a hybrid is sufficient for binding to one of the two previously identified Cry1Ac binding sites, whereas the second site requires the full Cry1Ac toxin for binding. In addition, the role of domain III in the very specific inhibition of Cry1Ac binding by the amino sugar N-acetylgalactosamine (GalNAc) was determined. Both in ligand blotting and in surface plasmon resonance experiments, as well as in binding assays using intact BBMVs, it was shown that the presence of domain III of Cry1Ac in a toxin molecule is sufficient for the inhibition of binding by GalNAc. These and other results strongly suggest that domain III of delta-endotoxins play a role in insect specificity through their involvement in specific binding to insect gut epithelial receptors.
During sporulation Bacillus thuringiensis (Bt) produces crystalline inclusions that consist of one or several insecticidal delta-endotoxins (Höfte and Whiteley, 1989). After ingestion by an insect the crystal is solubilized in the alkaline midgut, and the proteins are released in the form of protoxins. These are further processed by gut proteases to yield the active toxins that bind to specific receptors on the surface of the epithelial cells of the midgut brush border. Subsequently, through a still not well-understood mechanism, the bound toxins insert into the epithelial cell membranes to form pores that cause colloid-osmotic lysis of the cells and finally death of the insect (Knowles and Ellar, 1987Knowles and Dow, 1993).
The host range of the Bt delta-endotoxins, which form a large family with distinct insect specificities, is thought to be determined to a large extent by the specific interaction of the toxins with their receptors on the epithelial cell membranes (van Rie et al., 1990). Emphasizing the importance of this interaction, resistance to Bt toxins that occurs as a result of prolonged exposure of insect populations in the laboratory or in the field has been shown to be caused by changes in these receptors in several cases (Ferréet al., 1995). In the last few years several toxin-binding proteins have been purified to homogeneity, and their encoding genes have been cloned, thus allowing the detailed study of the toxin/receptor interaction at the molecular level. One Cry1Ac receptor of the tobacco hornworm, Manduca sexta, has been postulated to be an aminopeptidase N (APN) (Knight et al., 1994; Sangadala et al., 1994), bound to the epithelial cell membrane through a glycosyl-phosphatidylinositol anchor (Garczynski and Adang, 1995). Binding experiments using surface plasmon resonance subsequently showed the presence of two binding sites for Cry1Ac on the APN, one of which is shared with the related Cry1Aa and Cry1Ab toxins (Masson et al., 1995). Binding of Cry1Ac to both intact membrane vesicles as well as to isolated APN is inhibited by the sugar N-acetylgalactosamine (GalNAc), whereas binding of Cry1Aa and Cry1Ab is not affected (Garczynski et al., 1991; Knowles et al., 1991; Masson et al., 1995). As the APN is a glycoprotein that binds the galactosamine-specific lectin SBA (soybean agglutinin), it is thought that GalNAc or a structurally related sugar is part of the toxin binding site that is unique to Cry1Ac (Knight et al., 1994).
On the toxin side, elucidation of the three-dimensional structures of Cry3Aa (Li et al., 1991) and Cry1Aa (Grochulski et al., 1995) have revealed a three-domain structure. The N-terminal domain I is thought to be responsible for penetration of the epithelial cell membrane and pore formation. Results from mutagenesis studies strongly suggest that the highly variable domain II is involved in specific receptor recognition, thereby being one of the determinants of insect specificity of the toxin (reviewed in Dean et al., 1996). Experiments with hybrid toxins containing different combinations of domains (‘domain swapping’) have shown that domain III can also be a major determinant of specificity (Ge et al., 1991; Bosch et al., 1994; Masson et al., 1994). Domain III exchanges can result in new hybrid toxins with altered specificity and increased toxicity to a level not found in the original parental toxins (Bosch et al., 1994; de Maagd et al., 1996a). Using these hybrids as well as mutations in domain III, it was shown that, in addition to domain II, domain III also plays an important role in specific binding to putative receptors as visualized on ligand blots of BBMV (brush border membrane vesicle) proteins of target insects (Aronson et al., 1995; Lee et al., 1995; De Maagd et al., 1996a,b) as well as in binding of native APN to toxins on slot blots (Lee et al., 1995).
The purpose of the study described here was to use a combination of binding assay techniques to gain insight about the role of domain III in binding to SDS–PAGE-separated BBMV proteins, purified APN and intact BBMVs of M. sexta. Using this approach, we were able to show the direct involvement of domain III of Cry1Ac in insect epithelial gut protein binding and in inhibition of putative receptor binding by GalNAc.
Cry1 toxin and hybrid toxin binding on ligand blots with purified M. sexta aminopeptidase N
In previous experiments we have shown that Cry1Ac binds to a 120 kDa protein as well as to a 210 kDa protein on ligand blots of SDS–PAGE separated M. sexta BBMV proteins. Binding to the 120 kDa protein (assumed to be the putative Cry1Ac receptor APN) was shown to be dependent on domain III of Cry1Ac (de Maagd et al., 1996b). To confirm that domain III is indeed involved in binding to the APN, we incubated a ligand blot of purified APN with Cry1Ac, Cry1C and their hybrids H201 (domains I and II of Cry1C, domain III of Cry1Ac) and H130 (domain I and II of Cry1Ac, domain III of Cry1C). The results are shown in Fig. 1. Only Cry1Ac and hybrid H201, containing domain III of Cry1Ac, showed detectable binding to a 115 kDa band in the APN preparation (lanes 1 and 5 respectively). In addition, there was some binding of Cry1Ac and H201 to a slightly faster migrating band, possibly a degradation product of APN as a result of autolytic activity of the enzyme when stored in the absence of protease inhibitors (S. Sangadala, unpublished observation). Cry1C and the hybrid H130, containing domain I and II of Cry1Ac, did not bind (lanes 3 and 4 respectively). Therefore, the observed binding of Cry1Ac to APN on ligand blots was mediated through domain III. We did not observe binding of Cry1Ab to the APN (lane 2), although the presence of a common Cry1A binding site on the APN had been detected before in surface plasmon resonance (SPR) experiments (Masson et al., 1995). This suggests that under the conditions used here for ligand blotting, binding to this common Cry1A site is either too weak to detect, or the binding site is destroyed by the denaturation preceding the detection.
Binding of domain-swap hybrids to the aminopeptidase N detected by surface plasmon resonance
To study the role of domain III of Cry1Ac in binding to the non-denatured, purified APN we used the SPR technique. This technique measures the increase in mass of protein (toxin) binding to an immobilized receptor, in this case APN. It was shown previously that APN can bind twice as much Cry1Ac as it can bind Cry1Ab or Cry1Aa, and that this is probably due to the presence of one common binding site for all Cry1As and of one unique site for Cry1Ac (Masson et al., 1995). Cry1C did not bind to APN in these experiments. In this study, we tested binding of Cry1Ac and its hybrids H201 (1C/1C/1Ac) and H130 (1Ac/1Ac/1C) as well as that of Cry1Ab/Cry1C hybrids H04 (1Ab/1Ab/1C) and H205 (1C/1C/1Ab) to the purified APN. The results are shown in Fig. 2. Cry1Ac bound to the APN under these conditions, whereas Cry1C did not bind. The SPR response, in resonance units (RU), is linearly proportional to the surface protein concentration, therefore allowing the examination of the stoichiometry of the binding reaction. The amount of APN bound to the sensor chip corresponded to 900 RU, and the maximum amount of Cry1Ac bound also approached 900 RU. With the molecular weight of the activated toxin being approximately half of that of the receptor, it follows that the molar ratio of Cry1Ac/receptor is ≈2:1, as was found previously (Masson et al., 1995).
H201, which contains domain III of Cry1Ac, and H04 showed binding activity, although the latter to a much lesser extent. The other two hybrids, H205 and H130, did not bind to the immobilized APN (data not shown). For H201, binding levels reached ≈50% of that of Cry1Ac, suggesting that domain III of Cry1Ac is sufficient for binding to just one of the two Cry1Ac binding sites observed previously (Masson et al., 1995). The apparent binding affinity constant of H201 was calculated to be 5.07 × 10−7 M, which is approximately eight times higher (i.e. binding is weaker) than that reportedly previously for Cry1Ac (Masson et al., 1995). Most of this difference was caused by a substantially lower association rate of H201 [(7.54 ± 0.32) × 103 M−1 s−1 for H201 versus (9.0 ± 1.1) × 104 M−1 s−1 for Cry1Ac reported previously], whereas their dissociation rates were not significantly different [(3.82 ± 0.69) × 10−3 s−1 and (3.66 ± 0.13) × 10−3 s−1 respectively]. H04 (1Ab/1Ab/1C) showed binding with a very low affinity compared with Cry1Ac and H201 and did not reach saturation during the toxin injection period. After the end of toxin injection, the binding signal did not rapidly decrease, suggesting that rebinding of dissociated toxin readily takes place because of the availability of free receptor (Masson et al., 1995).
Cry1Ac domain III involvement in inhibition of binding to putative receptors on ligand blots by GalNAc
GalNAc has been shown to inhibit binding of Cry1Ac both to intact BBMVs of M. sexta (Knowles et al., 1991) as well as to the purified APN on ligand blots (Knight et al., 1994) and in SPR experiments (Masson et al., 1995). As shown in Fig. 3, GalNAc in increasing concentrations (0–80 mM) progressively inhibited binding of Cry1Ac to both the putative APN band at 120 kDa, as well as to the 210 kDa protein band on a ligand blot of M. sexta BBMV proteins (lanes 1–4). Equal concentrations of the structurally similar sugar GlcNAc had no inhibitory effect on binding to both BBMV proteins (Fig. 3, lanes 5–7). Binding of Cry1Ab to the 210 kDa protein was not affected by GalNAc, whereas binding to the Cry1Ab domain III-specific 250 kDa protein was slightly inhibited by 80 mM GalNAc (Fig. 3, lanes 8 and 9). These results show that binding of Cry1Ac to both the domain III-specific 120 kDa band as well as to the domain II-specific 210 kDa protein are specifically inhibited by the sugar GalNAc. Whereas the first part confirms earlier observations with purified APN, it is surprising that also domain I/II-specific binding of Cry1Ac to the 210 kDa protein is inhibited, although binding of Cry1Ab, which has a very similar domain I+II (only six amino acid residue differences), was not affected.
We tested further the role of domain III in the inhibition of binding by GalNAc described above by looking at the effect of this sugar on binding of the Cry1C/Cry1Ac hybrids H201 and H130 to ligand blots. Using these hybrids, we have shown previously that domains I and/or II of Cry1Ac are involved in the specificity of binding to the 210 kDa BBMV protein of M. sexta, whereas the presence of domain III of Cry1Ac in a toxin is necessary and sufficient for the specificity of binding to the 120 kDa protein (de Maagd et al., 1996b). The 210 kDa protein is a common binding protein for Cry1Aa, Cry1Ab and Cry1Ac; it has been observed by several groups (Vadlamudi et al., 1993; Martinez-Ramirez et al., 1994; de Maagd et al., 1996b; Keeton et al., 1998) and has been purified, cloned and shown to be a cadherin-like glycoprotein (Vadlamudi et al., 1995). In Fig. 4 (lanes 1–3), it is shown that increasing concentrations of GalNAc inhibited binding of H201 to the 120 kDa (APN–) band at least equally well and possibly more efficient than binding of the parental toxin Cry1Ac. On the other hand, in contrast to the observed inhibition for the parental Cry1Ac, the binding of hybrid H130 to the 210 kDa protein was not visibly inhibited by the highest GalNAc concentration used (Fig. 4, lanes 4–6).
Also on a ligand blot of a purified APN preparation (Fig. 5) both the binding of Cry1Ac (lanes 1 and 2) as well as that of hybrid H201 (lanes 3 and 4) was inhibited by GalNAc, confirming the results with separated BBMV proteins shown above.
Effects of GalNAc on Cry1Ac domain III-dependent binding to non-denatured APN in SPR experiments
In SPR experiments using the purified APN of M. sexta, Cry1Ac binding to both detected sites was inhibited by GalNAc, whereas binding of Cry1Aa and Cry1Ab to a single common site was not inhibited (Masson et al., 1995). In SPR experiments with the hybrid H201, which showed Cry1Ac domain III-mediated binding to the purified APN (see above), binding was increasingly inhibited up to ≈80% by GalNAc concentrations from 0 to 75 mM (Fig. 6). This inhibition level is similar to that previously observed for Cry1Ac (Masson et al., 1995). This shows that as in the experiments with ligand blots, also with the non-denatured purified APN, the presence of domain III of Cry1Ac in a toxin is sufficient for both binding to one binding site as well as for inhibition of this binding by GalNAc.
Binding of Cry1Ac and hybrid toxins to intact M. sexta BBMVs
Both ligand blot experiments and SPR experiments with purified APN have potential drawbacks when trying to elucidate the process of specific binding to target membranes and the role of the different toxin domains therein. Denaturation of BBMV membranes during ligand blotting may destroy important binding sites, or expose sites that are not taking part in binding of intact membranes. Even experiments with purified non-denatured APN can overlook the potentially important role of interactions between different membrane proteins and lipids for toxin binding. Therefore, we studied binding of parental and hybrid toxins to intact, non-denatured M. sexta BBMVs in a binding assay. Binding of biotin-labelled toxins in the presence or absence of various competitors was detected with streptavidin/peroxidase after blotting of the SDS–PAGE-separated BBMV proteins (including the bound, labelled toxin) to a nitrocellulose membrane. Results are shown in Fig. 7, where the bands shown represent the total amount of bound, labelled toxin. All four toxins tested, Cry1Ac, Cry1C and hybrids H201 and H130, showed specific binding, as can be deduced from comparison of the levels of binding in the absence (lanes 1) and in the presence of a 50-fold excess of homologous unlabelled competitor toxin (lanes 4). However, only binding of Cry1Ac and H201 could be inhibited by the presence of 100 mM GalNAc (lanes 2). The lack of inhibition of binding in the presence of the similar sugar GlcNAc again confirmed the specificity of the inhibition (lanes 3). These results show that also on intact membranes, although all tested toxins and hybrids do bind, the presence of domain III of Cry1Ac is required for inhibition of binding by GalNAc.
In this study, we have compiled data from three different types of binding assays to determine the role of domain III of Cry1Ac in binding to midgut epithelial membranes of its target insect M. sexta. The results of all three assays using natural and hybrid toxins in which domains were exchanged show that domain III of Cry1Ac plays an essential role in recognition of the putative receptor APN as well as in binding to intact membranes. Moreover, this role of domain III of Cry1Ac is tightly linked to a unique feature of Cry1Ac binding, i.e. inhibition by the amino sugar GalNAc. The combination of our results with others showing that GalNAc can decrease pore formation (Lorence et al., 1997) or swelling in BBMVs (Carroll et al., 1997) provides strong support for the involvement of domain III in in vivo binding and consequently toxicity.
Our experiments using toxin hybrids with ligand blots of separated BBMV proteins and of purified APN of M. sexta show that the presence of domain III of Cry1Ac in the toxin is sufficient for binding to the 120 kDa membrane anchored APN as well as its purified, GPI anchorless 115 kDa form. It must be stressed that this finding does not preclude a role for domains I and II in binding. It seems likely that other residues, especially in domain II, take part in (receptor) binding as well, yet in the experiments as described here specificity of much of the observed binding is determined by domain III of Cry1Ac. In those cases, domains I and/or II of Cry1Ac may play a role that can also be performed by the corresponding domains of Cry1C.
In contrast to the binding of Cry1Aa and Cry1Ab to APN observed by SPR (Masson et al., 1995), binding on ligand blots was not detected with either BBMV preparations (de Maagd et al., 1996b) or purified APN (this study). This suggests that the (common) binding site for Cry1A toxins postulated for M. sexta APN (Masson et al., 1995) may not survive the denaturation procedure that is part of the ligand blot protocol. In our SPR experiments we found that the presence of domain III of Cry1Ac is also sufficient for binding to one of the two Cry1Ac binding sites detected in this assay. However, the complementary hybrid H130 (1Ac/1Ac/1C) did not interact with immobilized APN, suggesting that for the binding to the second Cry1Ac site domain I and II of Cry1Ac are not sufficient but that a proper combination of domains (I+) II and III is required. Curiously, hybrid H04 (1Ab/1Ab/1C), which differs only in six amino acids from H130, did show some residual binding. Nonetheless, all hybrid toxins used in this study are active in M. sexta (data not shown). These results strongly suggest that domain III is participating in actual binding contacts to both binding sites, although alternative explanations remain possible.
GalNAc or a similar sugar seems to be part of a Cry1Ac binding site on APN of both M. sexta and L. dispar, because they both bind the lectin soybean agglutinin (Knight et al., 1994; Lee et al., 1996). Moreover, for these and other Cry1Ac-binding APN, GalNAc specifically inhibits Cry1Ac binding (Knight et al., 1994; Lee et al., 1996; Luo et al., 1997a,b; Valaitiset al., 1997). GalNAc also inhibits binding of Cry1Ac to intact membranes of M. sexta and, to a lesser extent, of T. ni (Haider and Ellar, 1987) as well as causing a reduction in pore formation in membrane preparations from these insects, indicating that this type of binding is essential for toxicity to insects (Carroll et al., 1997; Lorence et al., 1997). Our experiments have shown that the presence of domain III of Cry1Ac in a toxin is necessary and sufficient for this inhibitory effect of GalNAc on binding. This suggests that domain III of Cry1Ac is specifically interacting with this or a similar sugar on the receptor complex, and that this binding contributes to toxicity. In SPR experiments GalNAc has been shown to inhibit binding of Cry1Ac to both sites on the M. sexta APN (Masson et al., 1995), which again suggests the involvement of domain III in binding to both these sites. As GalNAc almost completely inhibited binding of Cry1Ac to intact M. sexta BBMVs, it is likely that domain III plays a specific binding role in vivo and, consequently, in toxicity. Recently it was shown that GalNAc-inhibited Cry1Ac pore formation activity only occurs in the posterior midgut of M. sexta, whereas both in posterior and in anterior midgut BBMVs there is a second, GalNAc-independent pore formation mechanism (Carroll et al., 1997). It is not known whether the second mechanism involves saturable, specific binding to the membrane. From our results we conclude that domain III of Cry1Ac is involved in at least the former of these two mechanisms.
The combined results of the different binding techniques used here and those found in the literature do not as yet allow a fully comprehensive model of Cry1Ac binding to M. sexta membranes. Clearly defined roles of the different putative receptors (APN and cadherin-like glycoprotein, and possibly others) and how they interact with the different toxin domains will require further experimentation. Binding studies with intact M. sexta membranes revealed a single high-affinity binding site for Cry1Ac, shared with Cry1Aa and Cry1Ab (van Rie et al., 1989). However, ligand blotting shows that these toxins can recognize and bind to different BBMV proteins through domain II and III (de Maagd et al., 1996b). In addition, SPR experiments show that a single putative receptor molecule, APN, may have two distinct binding sites (Masson et al., 1995). Although the purified APN shows a single band on a gel, the apparent ‘fuzziness’ of the APN band on ligand blots suggests a heterogeneity that could be caused by the association with lipids or by differences in the extent of glycosylation. However, the apparent 2:1 stoichiometry of the Cry1Ac/APN interaction in these experiments suggests that the APN preparation is likely to be homogeneous as far as the interaction with Cry1Ac is concerned (2 Cry1Ac binding sites per APN molecule). Using non-denaturing detergents, APN of M. sexta can be isolated as a complex containing various other GPI-anchored proteins and possibly lipids (Lu and Adang, 1996). The toxin-binding properties of this complex are different from those of the purified APN (Schwartz et al., 1997). Furthermore, it has been shown that these proteins are clustered with specific glycolipids when purified from brush border membranes, some of which can bind Cry1 toxins on thin-layer chromatography overlays (Adang et al., 1997). Although this complex does not necessarily represent the in vivo situation, this result does raise the possibility that receptors for different domains (II or III respectively) act as a single complex binding site in the intact membrane. Further experimentation using various techniques for studying binding and using toxin mutants specifically affected in domain II or domain III-dependent binding may in the future clarify this complex picture.
Toxin isolation and purification
All wild-type toxins and hybrid toxins were produced by expression in Escherichia coli strain XL-1 Blue, solubilization, trypsin activation of the protoxin, and FPLC purification of the mature toxin as described elsewhere (Bosch et al., 1994), except that 0.5 mM PMSF (phenylmethylsulphonyl fluoride) was added after incubation of protoxins with trypsin to prevent further degradation. Hybrid toxins H201 (domain composition: 1C/1C/1Ac), H130 (1Ac/1Ac/1C), H04 (1Ab/1Ab/1C) and H205 (1C/1C/1Ab) have been described previously (de Maagd et al., 1996a, b).
Isolation of brush border membrane vesicles
Brush border membrane vesicles (BBMVs) were isolated essentially as described by Wolfersberger et al. (1987) from dissected midguts of 7-day-old M. sexta larvae, except for the addition of 1 mM PMSF and 100 μM chymostatin in the MET (300 mM mannitol, 5 mM EGTA, 17 mM Tris-HCL pH 7.5) buffer during the initial homogenization step. Isolated BBMVs were resuspended in 0.5× MET buffer and stored at −80°C.
Proteins from isolated BBMVs were separated by SDS–PAGE. For ligand blotting, 50 μg of BBMV protein or 12.5 μg of purified APN from M. sexta was mixed with SDS–PAGE sample buffer, heated for 5 min at 100°C and loaded onto a 7.5% acrylamide gel in a continuous, 7.4-cm-wide sample slot. Prestained molecular-mass marker proteins were run alongside in a separate single slot. After electrophoresis, separated proteins were electrophoretically transferred to nitrocellulose (0.22 μm pore size, Schleicher and Schull). Strips (3 mm wide) were cut from the blots for incubation with toxins (2 μg of toxin per strip), followed by incubation with a Cry1 antiserum and detection with ECL reagent as described previously (de Maagd et al., 1996a).
Binding experiments on intact BBMVs
Toxins were labelled with biotin using biotin-N-hydroxysuccinimide ester (BNHS, Boehringer) as described previously (Bosch et al., 1994). For binding experiments, BBMVs (5 μg of protein) were mixed with 2 ng of biotin-labelled toxin and, if appropriate, 100 ng of unlabelled competitor toxin or 100 mM N-acetylaminosugar (-galactosamine or -glucosamine) in 100 μl of PBS/Tween (phosphate-buffered saline, pH 7.4, containing 0.1% Tween-20). After incubation for 1 h at room temperature, vesicles and unbound toxin were separated by centrifugation, and the pellet was washed once briefly with PBS/Tween. Subsequently, BBMV proteins (including bound, labelled toxin) were separated by SDS–PAGE and electroblotted onto a nitrocellulose membrane. Bound, labelled toxin on the blot was detected with streptavidin/peroxidase as described previously (Bosch et al., 1994).
Purification of M. sexta 115 kDa aminopeptidase N
Aminopeptidase N (120 kDa) was purified from M. sexta brush border membrane vesicles as described previously with a few modifications (Lu and Adang, 1996). Briefly, frozen BBMVs were thawed on ice and resuspended at 5 mg protein ml−1 in a buffer containing 20 mM Tris-HCl, pH 8.5, 100 mM NaCl, 5 mM EDTA, 1 mM PMSF and 1% CHAPS (3-((3-cholamidopropyl)dimethylammonio)-1-propanesulphonate) and stirred at 4°C overnight for solubilization. Under these conditions, the 115 kDa soluble form of aminopeptidase is released by endogenous phospholipase C from its lipid anchor. Insoluble material was removed by centrifugation at 100 000 × g for 1 h at 4°C. The supernatant was diluted fivefold with 20 mM Tris-HCl, pH 8.5; 0.05% CHAPS and syringe filtered through a 0.2 μm membrane for anion exchange chromatography using two consecutively connected 5 ml EconoQ anion exchange columns (Bio-Rad) linked to a FPLC system (Pharmacia). Buffer A contained 20 mM Tris-HCl, pH 8.5; 0.05% CHAPS, and buffer B contained 20 mM Tris-HCl, pH 8.5; 0.05% CHAPS; 1.0 M NaCl. Solubilized BBMV proteins (50 mg) were loaded onto the column and the bound proteins were eluted with a two-step salt gradient. The chromatographic run started at 1% min−1 gradient of buffer B at a flow rate of 2 ml min−1 for 40 min followed by a 3% min−1 gradient of buffer B for 20 min. Fractions (1 ml) were collected and analysed by SDS–PAGE and toxin binding. The soluble form of aminopeptidase (115 kDa) was eluted in fractions 5–8. These fractions were combined, concentrated and stored at −70°C until further use.
Surface plasmon resonance experiments
The binding of hybrid toxins to the purified soluble 115 kDa APN from M. sexta was studied by SPR using the BIAcore system and CM5 sensor chips, both from Pharmacia Biosensor. The aminopeptidase was immobilized by amine coupling, and binding analyses were performed exactly as described elsewhere (Masson et al., 1995). All sensorgram data transformations and analyses were performed with BIAEVALUATION software version 2.1. using non-linear least-squares curve fitting. Carbohydrate inhibition studies were performed by injecting a 1.5 μM solution of toxin diluted in HBS-P20 (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.05% BIAcore Surfactant P20) buffer possessing various concentrations of GalNAc. The highest response (in resonance units) obtained for each toxin/sugar combination at the end of an injection over the receptor surface after subtraction of a sugar blank (i.e. no toxin) was expressed as percentage inhibition using toxin binding in the absence of sugar as the 100% binding point.