The extraordinary potency of botulinum neurotoxins (BoNTs) is mediated by their high neurospecificity, targeting peripheral cholinergic motoneurons leading to flaccid paralysis and successive respiratory failure. Complex polysialo gangliosides accumulate BoNTs on the plasma membrane and facilitate subsequent binding to synaptic vesicle membrane proteins which results in toxin endocytosis. The luminal domain 4 (LD4) of the three synaptic vesicle glycoprotein 2 (SV2) isoforms A–C mediates uptake of the clinically most relevant serotype BoNT/A1. SV2C-LD4 exhibits the strongest protein–protein interaction and comprises five putative N-glycosylation sites (PNG sites). Here, we expressed human SV2C-LD4 fused to human IgG-Fc in prokaryotic and eukaryotic expression systems to analyse the effect of N-glycosylation of SV2C on the interaction with BoNT/A1. Mass spectrometric analysis of gSV2CLD-Fc demonstrates glycosylation of N534, N559 and N565, the latter two residing at the BoNT/A interface. Mutational analysis demonstrates that only the N559-glycan, but not N565-glycan increases affinity of BoNT/A for human gSV2C-LD4. The N559-glycan was characterised as a complex core-fucosylated type with a heterogeneity ranging up to tetra-antennary structure with bisecting N-acetylglucosamine which can establish extensive interactions with BoNT/A. The mutant gSV2CLD-Fc N559A displayed a 50-fold increased dissociation rate kd resulting in an overall 12-fold decreased binding affinity in surface plasmon resonance (SPR) experiments. The delayed dissociation might provide BoNT/A more time for endocytosis into synaptic vesicles. In conclusion, we show the importance of the complex N559-glycan of SV2C-LD4, adding a third anchor point beside a ganglioside and the SV2C-LD4 peptide, for BoNT/A neuronal cell surface binding and uptake.
- botulinum neurotoxin A
- HEK cell expression
- surface plasmon resonance
- synaptic vesicle glycoprotein 2C
The family of botulinum neurotoxins (BoNTs) consists of seven established serotypes termed A to G (BoNT/A–G) which are the aetiological agents of the neuromuscular disease botulism. Their extreme potency, the ubiquitous presence of Clostridium botulinum spores and the high mortality rate of untreated botulism led to the declaration of BoNTs as category A select agents [1,2] and as a potential weapon of bioterrorism. Nevertheless, especially BoNT/A and B are successful, licensed drugs to treat an ever-growing number of medical and cosmetic conditions which are characterized by hyperactivity of peripheral synapses . The BoNTs are bacterial protein toxins of the AB-type that are synthesized as a single polypeptide chain which becomes proteolytically activated by bacterial or host proteases to form a 100 kDa heavy chain (HC) and a 50 kDa light chain (LC). In the active di-chain toxin, both chains remain associated by a single disulfide bond . The LC, a Zn2+-dependent metalloprotease, represents the enzymatically active part A and specifically cleaves members of the SNARE-family of proteins which are indispensable for the neurotransmission at synapses. Their cleavage halts the neuroexocytosis of acetylcholine at the neuromuscular junction (NMJ), which leads to a flaccid paralysis of the affected muscle and hence to symptoms of botulism. Since the HC mediates the binding and uptake of the LC into target cells, it represents the binding part B. It consists of the 50 kDa amino-terminal translocation (HN) and the 50 kDa carboxyl-terminal cell-binding domain (HC). HN is supposed to form a channel in the membrane of the acidifying endosome, which allows the transport of the LC through the endosomal membrane into the target cell cytosol [5,6]. The HC mediates binding to target cells and consists of the 25 kDa amino-terminal HCN and the 25 kDa carboxyl-terminal HCC domain .
The extreme potency of BoNTs is predominantly due to their specific interaction with the presynaptic membrane at NMJs. It was shown that complex polysialo gangliosides accumulate the toxin molecules via their HCC domain in the plane of the plasma membrane with sub-micromolar affinity . In the case of BoNT/A, a single ganglioside binds with its terminal GalNAc-Gal moiety into a ganglioside binding site (GBS) comprising a residue motif also largely conserved in BoNT/B, E, F, G and tetanus neurotoxin (TeNT) [8–10]. Enriching the BoNT in the 2D plane of the presynaptic membrane considerably facilitates the subsequent interaction with a second receptor which leads to an activity-dependent, receptor-mediated uptake of the toxin into small synaptic vesicles. Accordingly, BoNT/B, G and mosaic type BoNT/DC employ the luminal domain (LD) of the synaptic vesicle transmembrane proteins synaptotagmin-I and -II as protein receptor [11–14]. BoNT/A interacts with the fourth large luminal domain (LD4) of the synaptic vesicle glycoprotein 2 isoforms A–C (SV2A–C). The SV2s are integral, rare (1–2 copies per SV ) proteins residing in the membrane of synaptic vesicles. They are predicted to span the vesicle membrane 12 times (Figure 1A). The LD4 between transmembrane regions seven and eight is approximately 125 amino acids (AA) long, glycosylated and becomes extracellularly exposed upon neuroexocytosis (Figure 1A) . Also BoNT/E binds the LD4 of SV2A and SV2B as second receptor [17–19]. The exact molecular mechanism of how BoNT/D, F and the closely related TeNT use SV2A–C during their uptake process [10,20–22] is unknown, because no direct interaction between SV2 and BoNT/D, F or TeNT has yet been demonstrated. Initially, TeNT binds the extracellular matrix proteins nidogen-1 and -2 (also known as entactins) as receptors at the NMJ . Above all, no protein receptor has been identified for BoNT/C, but it was proposed instead that this serotype enters neurons only by binding to two ganglioside molecules .
It was demonstrated that BoNT/A binds Escherichia coli-expressed, non-glycosylated SV2C-LD4 much stronger compared with non-glycosylated SV2A- and SV2B-LD4 [17,19,25]. Recently, the crystal structure of HC of BoNT serotype A (HCA) in complex with non-glycosylated human SV2C-LD4 was solved . Here, SV2C-LD4 forms a right-handed, quadrilateral β-helix which associates with HCA mainly via backbone–backbone interactions and only a few side chain interactions at the opposite site of the single GBS in HCC close to the HCN (Figures 1B–1C). In parallel, mutational analysis also allocated the SV2 binding site opposite the single GBS in HCCA, but additionally identified functional residues outside the protein/protein interface  which could interact with posttranslational modifications of SV2. It is known that the LD4 of SV2 is glycosylated  and that the LD4 of SV2A, B and C contain three, three and five putative N-glycosylation (PNG) sites, respectively, based on the motif N-X-S/T (X can be any AA except proline). It was demonstrated that BoNT/E, in contrast to BoNT/A, only exploits glycosylated SV2A and B as cellular receptors. In particular, BoNT/E strictly requires glycosylation of N573 in SV2A to be able to bind and to enter target cells [18,28]. A similar role of N573 in SV2A for binding of BoNT/A was postulated, but has not been unambiguously demonstrated . No work about the contribution of N-glycans in SV2C-LD4 to the interaction with BoNT/A has been published so far.
Here, we determined the binding kinetics of BoNT/A1 to glycosylated SV2C-LD4, identified the N559-glycan in glycosylated SV2C-LD4 as the main factor mediating the high affinity interaction with BoNT/A, and characterized the N559-glycan as mainly a tetra-antennary, core-fucosylated structure partly equipped with a bisecting N-acetylglucosamine (GlcNAc).
Chemicals and reagents
Methanol and acetonitrile (LC–MS grade) were purchased from Merck, formic acid (MS grade) from Fluka, DTT, iodoacetamide (IAM) and ammonium bicarbonate from Sigma–Aldrich, porcine trypsin of sequencing grade from Promega and water from a Milli-Q system, Merck.
Construction of plasmids
The plasmid pIg+H6thSV2CLDS, facilitating the eukaryotic expression and secretion of the glycosylated human SV2C-LD4 (gSV2CLD) fused C-terminally to human IgG1-Fc (gSV2CLD-Fc), was generated by cloning a synthetic E. coli codon usage optimized DNA sequence encoding hSV2C-455–579  equipped with N-terminal His6- and C-terminal Strep-tag (IBA GmbH) to allow tandem affinity purification into the EcoRV and XbaI sites between the CD33 signal peptide and the hIgG1-Fc sequence of the plasmid pIg+ (R&D Systems). The plasmid pH6thSV2CLDSFc, facilitating the prokaryotic expression of non-glycosylated SV2CLD-Fc, was generated by amplification of the corresponding DNA sequence omitting the CD33 signal peptide and His6-tag sequences by means of PCR using pIg+H6thSV2CLDS as template and appropriate oligonucleotides. The resulting DNA fragment was cloned into the BamHI and HindIII sites of pH6thSV2C 455–579 . All mutations were introduced using the GeneTailor Mutagenesis System (life technologies) and the nucleotide sequences of all constructs and mutants were verified by DNA sequencing (GATC Biotech).
Purification of recombinant proteins
Recombinant HCA1 comprising a C-terminal Streptag and full-length, tag-free, di-chain BoNT/A1 have been expressed and purified as previously described [8,29]. gSV2CLD-Fc was purified from cell culture supernatants of transiently transfected HEK293T cells. Briefly, 1 day before transfection 18 × 106 cells were seeded in 300 cm2 culture flasks using growth medium (DMEM, Sigma, D5671) supplemented with 10% FCS (PAN Biotech, P30-1502), 10 units·ml−1 penicillin, 100 μg·ml–1 streptomycin and 2 mM L-glutamine. Transfection was performed by exchanging the growth medium of the culture and incubating 32 μg pIg+H6thSV2CLDS and 92 μl branched polyethylenimine (1 mg·ml−1, Sigma, #408727) in 4 ml DMEM without serum. Following 20 min of incubation at room temperature the mixture was added to the culture. Transfection efficiency was estimated by replacing one-tenth of the amount of the DNA with the plasmid pEGFP-N2 (BD Biosciences Clontech) and visual inspection of the green fluorescence of the cells 1 day after transfection. Following this protocol the transfection efficiency was typically approximately 60%. Cell culture supernatants were collected 2, 4 and 6 days after transfection. Non-glycosylated SV2CLD-Fc was purified from lysates of transformed E. coli strain M15 [pREP4] (Qiagen), following 16–18 h of induction with 0.2 mM IPTG at 21°C.
gSV2CLD-Fc and SV2CLD-Fc were consecutively affinity purified on Talon matrix (Clontech) and Streptactin-Superflow (IBA GmbH) columns according to the manufacturers’ instructions. Both fusion proteins were eluted using 100 mM Tris/HCl (pH 8.0), 150 mM NaCl and 10 mM desthiobiotin. Fractions containing high amounts of protein were pooled, shock-frozen in liquid nitrogen and kept frozen at −70°C until usage. Protein concentrations were determined by densitometry following SDS-PAGE and Coomassie Blue staining using various known concentrations of bovine serum albumin (BSA) as standards. Identity of gSV2CLD-FC proteins was verified by semi-dry Western blotting employing rabbit polyclonal anti-SV2C antibody (Synaptic Systems, Göttingen, Germany, dilution 1:5000) and goat anti-rabbit antibody conjugated to HRP (Thermo Scientific, Darmstadt, Germany, dilution 1:5000).
SV2C binding assay
Binding of wild-type and mutant gSV2CLD-Fc and SV2CLD-Fc to HCA and BoNT/A, respectively, was carried out in binding buffer (20 mM Tris/HCl, pH 7.4, 150 mM NaCl and 0.5% Triton X-100). A total of 10 μl protein G-sepharose beads (GE Healthcare) coated with 100 pmol of gSV2CLD-Fc or non-glycosylated SV2CLD-Fc was incubated with the indicated concentrations of HCA or BoNT/A, respectively, for 2–3 h in a total volume of 200 μl. Following incubation, beads were collected by centrifugation at 4500 × g and washed three times using binding buffer. Washed pellet fractions were incubated in SDS sample buffer for 5 min at 99°C and analysed by SDS-PAGE. Proteins were visualized by means of Coomassie Blue staining.
LC–MS and LC–MS/MS analysis
gSV2CLD-Fc protein samples were precipitated using ice-cold methanol as described elsewhere . The precipitated proteins were reduced by addition of 20 μl 10 mM DTT/80 mM ammonium bicarbonate solution (60°C, 60 min) followed by alkylation in darkness (30°C, 40 min) after addition of 2 μl 150 mM IAM solution (final conc. 15 mM, ∼pH 8). Sample digestion (45°C, 60 min) was performed after adding 2 μl of a 0.2 μg·μl−1 trypsin solution. The obtained peptides were acidified, diluted and analysed by LC–MS and MS/MS on a Waters Nano-Acquity UPLC system connected to a Waters QTOF Ultima mass spectrometer equipped with a nano-electrospray ion source (Waters Corporation). The peptides were separated on a 150 mm BEH C18 nanoAcquity UPLC column (Waters) using a water:acetonitrile gradient containing 0.1% formic acid from 3 to 40% over 25 min at a flow rate of 400 nl·min−1. Samples were screened for glycopeptide content in LC–MS mode using collision energies (CEs) ranging from 6 to 40 eV with argon as the collision target. Glycopeptides were analysed in terms of sugar content using LC–MS/MS of selected precursors.
Surface plasmon resonance (SPR) measurements
Binding kinetics and affinity were determined on a Biacore X100 unit (GE Healthcare) at 25°C using phosphate buffered saline (PBS, pH 7.3) supplemented with 0.5% Triton X-100 as running buffer. To this aim, a mouse anti-human IgG-Fc-specific monoclonal antibody was covalently coupled to a CM5 sensor chip according to manufacturer's instructions (Human IgG Capture Kit; GE Healthcare). gSV2CLD-Fc wild-type and N559A mutant (10 μg·ml−1) were immobilized for 120 s at a flow rate of 5 μl·min−1 on flow cell 2 by a capture approach via the C-terminal human IgG1-Fc-portion to a mean surface density of 317 resonance units (RU). Recombinant HCA was injected for 60 s over both flow cells at a flow rate of 30 μl·min−1 to monitor binding association followed by 300 s injections of running buffer to monitor binding dissociation. Between measurements, the sensor chip was regenerated by injections of 3 M MgCl2 for 30 s at 20 μl·min−1. Each measurement cycle consisted of injections of 1:3 serial dilutions of recombinant HCA ranging from 1800 nM to 22.2 nM with duplicate injections of the highest concentration and triplicate injections of buffer only. The kinetic association rate constants ka, the dissociation rate constants kd and the equilibrium binding constants KD were determined by fitting a 1:1 Langmuir interaction model with RI set to 0 and Rmax fitted globally to the double referenced  binding curves using the Biacore Evaluation Software 2.01 or the BIAevaluation Software 4.1.1 (GE Healthcare).
Luminal domain 4 of SV2C comprises five putative N-glycosylation sites
Sequence comparison of the LD4s of all three SV2 isoforms of mouse, rat and human origin revealed the presence of three conserved PNG sites (Figure 1 and Supplementary Figure S1). In human SV2C (hSV2C), these sites are N484, N534 and N559. However, compared with the SV2A and B isoforms SV2C displays two additional unique PNG sites at N480 and N565 (Figure 1 and Supplementary Figure S1), suggesting that SV2C contains two additional N-glycans. Indeed, gSV2CLD-TD migrates at higher molecular weight in SDS-PAGE than gSV2ALD-TD .
Previous studies demonstrated that the glycosylation of the third conserved N573 in SV2A is a prerequisite for the interaction with BoNT/E . Interestingly, our inspection of the crystal structure of HCA in complex with non-glycosylated hSV2C-LD4  revealed that the corresponding third conserved PNG site in hSV2C (N559) as well as the non-homologous PNG site at N565 are localized in very close proximity to HCA whereas the remaining three PNG sites either point away (N534) or are too far from the HCA surface (N480, N484; Figure 1B). Therefore, it was reasonable to assume that most likely the sugar moieties of N-glycans attached to N559 or N565 in SV2C would mediate additional interactions between SV2C and HCA. These additional interactions were hypothesized to translate into a higher affinity and/or different binding kinetics of HCA to glycosylated compared with non-glycosylated SV2C.
gSV2CLD-Fc is secreted as soluble protein
To test our hypothesis, we expressed hSV2C-LD4 fused to human IgG-Fc in HEK cells to obtain glycosylated gSV2CLD-Fc (Figure 2A). Subsequently, several mutations were introduced to disable glycosylation of N559 and/or N565, respectively. In order to unambiguously discriminate between effects elicited by the exchange of the amino acid side chains and effects caused by the loss of the corresponding N-glycan, we additionally expressed the fusion proteins in a bacterial expression system to generate glycan-free SV2CLD-Fc.
Sufficient amounts of soluble gSV2CLD-Fc fusion proteins (Figure 2A) were obtained from culture supernatants of transiently transfected HEK293T cells, indicating that the gSV2CLD-Fc becomes readily secreted into the culture medium. Affinity purified gSV2CLD-Fc appeared as a diffuse array of bands with an apparent molecular weight of around 60 kDa in SDS-PAGE (Figure 2B), which clearly exceeds the calculated molecular weight of approximately 46 kDa for the SV2CLD polypeptide chain. Hence we conclude that the proteins are glycosylated upon expression in HEK cells. The masses of N-glycans range from 892 Da of the common core structure subunit GlcNAc2Man3 to >3500 Da of a complex sialylated, core-fucosylated, tetra-antennary structure. On average, the 15 kDa molecular weight increase agrees well with the presence of 4–6 N-glycans in gSV2CLD-Fc. The diffuse appearance of the protein bands is probably caused by the observed micro heterogeneity of some of the N-glycans. Both the mutation of the fourth (N559A, N559Q and S561A) or the fifth (N565Q) PNG of gSV2CLD-Fc resulted in an increased mobility of the mutant in SDS-PAGE compared with gSV2CLD-Fc wild-type, indicating that both sites are N-glycosylated (Figure 2B). Furthermore, the double mutant gSV2CLD-Fc N559Q/N565Q showed an even stronger increase in mobility, confirming the simultaneous glycosylation of both sites in HEK cells. In addition, N-glycosylation of N534, N559 and N565 in SV2CLD and of N297 in the IgG-Fc peptide EEQYNSTYR was confirmed by LC-MS analysis of trypsin digestion derived peptides of gSV2CLD-Fc wild-type. The identified N-glycosylated SV2C peptides are summarized in Table 1 and the corresponding data for single mutants gSV2CLD-Fc N559Q and N565Q are included as supplementary information (Supplementary Table S1). Non-glycosylated SV2CLD-Fc constructs were purified from E. coli cell lysates and, as expected, migrated as a discrete band corresponding to ∼46 kDa. Here, no differences in mobility were visible between SV2CLD-Fc wild-type and SV2CLD-Fc mutants (Figure 3B).
The N559-glycan increases affinity of gSV2CLD-Fc for BoNT/A
To analyse the influence of the N-glycans attached to N559 and N565 on the interaction of SV2C with BoNT/A, purified gSV2CLD-Fc and glycan-free SV2CLD-Fc fusion proteins were immobilized to protein G-sepharose beads to precipitate isolated HCA or BoNT/A from solution. BoNT/A had to be used as an alternative ligand for hSV2CLD-Fc because HCA exhibits a similar mobility to SV2CLD-Fc which would have impeded the analysis of the bound proteins by SDS-PAGE.
Wild-type gSV2CLD-Fc and SV2CLD-Fc showed robust binding to HCA and BoNT/A respectively (Figures 3A and 3B). Also the single mutant gSV2CLD-Fc N565Q exhibited binding to HCA indistinguishable from gSV2CLD-Fc wild-type, indicating that neither the N565-glycan nor the amino acid N565 itself contributes to the interaction with BoNT/A (Figure 3A, left panel). In contrast, the single mutant gSV2CLD-Fc N559Q displayed clearly reduced affinity for BoNT/A (Figure 3A). Accordingly, simultaneous mutation of N559Q and N565Q did not reduce binding any further compared with gSV2CLD-Fc N559Q, again confirming that neither the N565-glycan nor N565 itself contributes to the interaction with BoNT/A (Figure 3A, left panel). Hence, mutant N565Q was excluded from further analysis, whereas mutant SV2CLD-Fc N559Q was additionally expressed in E. coli to verify the effect of the elongated side chain of Q559 (Figures 1C and 3B) on the binding to BoNT/A. Surprisingly, also SV2CLD-Fc N559Q displayed decreased affinity for BoNT/A compared with SV2CLD-Fc wild-type indicating a negative influence of the elongated amide side chain. In silico substitution of N559Q reveals that the H-bonds between the amide group of N559 and T1145/Y1149 of BoNT/A  are lost and an interference of the amide group of Q559 with the phenyl ring of F953 might arise due to insertion of the methylene group (Figure 1C). Since the glycosylated and non-glycosylated N559Q mutants both show a reduced binding, the observed negative effect can be attributed to the perturbation by glutamine as well as to the missing N559-glycan. Hence the N559Q mutation is not suitable to discriminate between the effect caused by the missing N559-glycan and the effect on protein–protein interaction.
Therefore we tried to prevent glycosylation of N559 by exchanging S561 of the PNG motif N-X-S/T for alanine. Indeed, the gSV2CLD-Fc S561A mutant exhibited an increased mobility in SDS-PAGE compared with wild-type, indicating the absence of the N559-glycan (Figure 2B). However, gSV2CLD-Fc S561A as well as SV2CLD-Fc S561A displayed reduced binding to HCA (Figure 3A) and BoNT/A (Figure 3B) respectively. Here, lack of N559-glycan as well as lack of the S561 hydroxyl group establishing water mediated H-bonds with the amide and carbonyl groups of R1294 and the carboxyl group of E1293 of HCA (Figure 1C) impair binding. Hence, S561 cannot be modified without affecting the protein–protein interaction and also does not allow analysis of N559 glycosylation on BoNT/A binding.
Although the amide group of N559 displays H-bonds to T1145 and Y1149 of BoNT/A  we then generated the mutant SV2CLD-Fc N559A (Figure 2). Surprisingly, SV2CLD-Fc N559A exhibited no effect on binding to BoNT/A confirming previous biochemical data by Benoit et al.  (Figure 3B), but also suggesting that N559 backbone interactions are by far more important than its side chain H-bonds. Like gSV2CLD-Fc N559Q also gSV2CLD-Fc N559A showed an increased mobility in SDS-PAGE indicating absence of the N559-glycan. Finally, gSV2CLD-Fc N559A exhibited a strong reduction in binding to HCA (Figure 3A, right panel) compared with gSV2CLD wild-type, demonstrating the importance of N559-glycan for high affinity binding to BoNT/A.
N559-glycan drastically decreases dissociation of HCA from gSV2CLD-Fc as shown by surface plasmon resonance
To validate the increased binding of BoNT/A to gSV2CLD-Fc and to determine the influence of the N559-glycan on binding affinity and kinetics, we employed surface plasmon resonance (SPR) measurements. To mirror the natural presentation of the SV2C-LD4, gSV2CLD-Fc wild-type and the three gSV2CLD-Fc mutants N559A, N559Q and S561A served as receptors and were captured employing the Fc-portion via an immobilized anti-human-IgG capture antibody. Recombinant HCA was injected as analyte in a concentration range from 1800 nM to 22.2 nM and kinetic binding rates were determined by fitting the measured interactions to a 1:1 Langmuir binding model. Only mutant gSV2CLD-Fc S561A required fitting to the heterogeneous binding model (Table 2; Figure 4 and Supplementary Figure S2).
In agreement with the binding assay, we also observed much stronger binding of HCA to gSV2CLD-Fc wild-type as compared with all three gSV2CLD-Fc mutants, e.g., the overall affinity is 12-fold higher than for gSV2CLD-Fc N559A mainly due to a ∼50-fold decreased dissociation rate kd of HCA from gSV2CLD-Fc wild-type (Table 2). On the other hand, the approximately 4-fold reduced association rate of HCA to the gSV2CLD-Fc wild-type might be explained by the spatial constraint introduced by the additional interaction of the N559-glycan requiring the carbohydrates to correctly orient for interaction with HCA. gSV2CLD Fc N559A and N559Q display identical binding kinetics (fast association and dissociation typical for constructs lacking the N559-glycan) and similar binding constants (Figure 4 and Supplementary Figure S2). gSV2CLD Fc S561A exhibits a heterogeneous binding of HCA, but again the dominating component describes fast association and dissociation of HCA which is characteristic of a transient SV2C peptide interaction (Supplementary Figure S2).
In conclusion, SPR measurements delivered a mechanistic explanation for the increased affinity of BoNT/A binding to glycosylated SV2C by showing that the additional interaction with the N559-glycan stabilizes the complex by introducing an additional interaction site.
N559 comprises a tetra-antennary, core-fucosylated N-glycan with a bisecting GlcNAc
The gSV2CLD-Fc N559Q and gSV2CLD-Fc N565Q mutants were initially used to identify the N-glycans present in gSV2CLD-Fc wild-type. The glycan identification in the single mutants made use of the glycan heterogeneity in the expressed proteins, including a small fraction of non-glycosylated tryptic peptides. Thus, the non-glycosylated Q559C*SFFHNK566 (C*=carbamidomethylated cysteine) and N559C*SFFHQK566 peptides were identified in the mutants gSV2CLD-Fc N559Q and gSV2CLD-Fc N565Q respectively. Supplementary Figure S3 displays LC–MS chromatograms of these non-glycosylated peptides together with the wild-type analogue. Glycosylation of peptides has a limited effect on their hydrophilic properties, and consequently the non-glycosylated peptides indicated the retention time of interest for corresponding glycopeptides. The retention times were used in combination with LC–MS analysis at high collision energy to uncover the corresponding glycopeptides Q559C*SFFHNglycoK566 and N559glycoC*SFFHQK566 in gSV2CLD-Fc N559Q and gSV2CLD-Fc N565Q, respectively. Diagnostic carbohydrate fragment ions (m/z 204.1 and 366.1) produced at high collision energy  were used for the glycopeptide identification. Both of the two peptides were N-glycosylated and showed a high degree of glycan heterogeneity. The glycan moiety at N559 in the gSV2CLD-Fc N565Q mutant was expected to suppress the trypsin digestion at the N-terminal K558/N559-trypsin cleavage site and indeed the intensities of the two almost identical Q-mutant peptides were very different (results not shown). This suppression of K558/N559-cleavage increased the intensity of the longer F552IDSEFKNglycoC*SFFHQK566 peptide and the corresponding double glycosylated peptide F552IDSEFKNglycoC*SFFHNglycoK566 was identified in the gSV2CLD-Fc wild-type construct. Thus, the N559- and N565-glycans of gSV2CLD-Fc wild-type were confirmed in both the short N559glycoC*SFFHNglycoK566 peptide (Table 1) and in the longer F552IDSEFKNglycoC*SFFHNglycoK566 peptide (Figure 5, Table 1) produced by one missed trypsin cleavage site. A summary of identified peptides and glycopeptides within the sequence of interest is shown in Table 1 (for gSV2CLD-Fc wild-type) and in Supplementary Table S1 (for gSV2CLD-Fc N559Q and gSV2CLD-Fc N565Q mutants).
LC–MS/MS analysis of the glycosylated F552IDSEF-KNglycoC*SFFHQK566 peptide derived from the gSV2CLD-Fc N565Q mutant which showed BoNT/A binding indistinguishable from gSV2CLD-Fc wild-type revealed the detailed N559-glycan structure. At intermediate collision energy the losses of terminal galactoses (Gals) and GlcNAcs were detected (Figure 6, top panel). At higher collision energy, subsequent fragmentation of tri-mannose (Man3), fucose (Fuc) and GlcNAc core structures were observed (Figure 6, lower panel). Based on the data the presence of a tetra-antennary, core-fucosylated N-glycan type was concluded. In addition, LC-MS data of the peptide verified a bisecting GlcNAc (Supplementary Table S1). LC–MS analysis of the gSV2CLD-Fc wild-type showed glycan heterogeneity at N559 and N565. Both sites carry core-fucosylated N-glycans ranging from dual- (∼15%), tri- (∼39%) to tetra-antennary (∼45%) structures of the total glycan pool. Approximately 13% of the tetra-antennary structures also contain a bisecting GlcNAc (Figure 5).
BoNTs exploit the recycling pathway of synaptic vesicles to gain access to the cytosol of peripheral cholinergic motoneurons. To this end, BoNTs interact with luminal segments of transmembrane proteins of synaptic vesicles which temporarily become exposed to the synaptic cleft upon exocytosis of neurotransmitter . The luminal segments of most of these proteins including synaptotagmin (Syt) and synaptic vesicle glycoprotein 2 are known to be multiply glycosylated [16,33,34]. Nonetheless, the interaction of BoNT/A with SV2C was first discovered and characterized using E. coli-expressed non-glycosylated LD4 of SV2C [17,19], demonstrating sufficient affinity of BoNT/A for the bare SV2C peptide to detect robust binding in vitro. However, expressing the LD4 of human SV2C fused to human IgG1-Fc in HEK cells, we were able to isolate secreted glycosylated gSV2CLD-Fc. MS analysis could detect N-glycosylation at the three C-terminal of the five PNG sites. Subsequently, we unambiguously demonstrated that the N559-glycan strengthens the binding of BoNT/A to SV2C 12-fold.
N-glycan patterns can vary from tissue to tissue, between cell types or even depending on the localization/organelle or the developmental stage of the organism. Few studies (e.g. [35,36]) analysed sialylated N-glycans of brain homogenate. However, apart from the use of CNS tissue instead of PNS a complex mixture of different neurons and other accompanying cells was also analysed and does not reflect the true situation BoNT/A will face at the motoneuron. Since we expressed SV2C in non-neuronal HEK cells, we could not expect to obtain the exact, yet unknown, N-glycosylation pattern of native SV2C which might even differ between SV2C being expressed in central or peripheral neurons. Nevertheless, our HEK cell-expressed gSV2CLD-Fc supported high affinity binding of BoNT/A with a KD of 110 nM. In contrast, mutant gSV2CLD-Fc N559A lacking the N559-glycan displayed a 12-fold reduced affinity. Looking at the co-crystal structure of E. coli-expressed non-glycosylated SV2C in complex with HCA , the side chain of N559 points towards a crevice in HCA formed by the amino acids G1292, F953, Y1149 and T1145 which could accommodate the proximal core sugars of the N559-glycan. In contrast, residue N565, also close to the HCA surface, points away from the binding interface and is not offered a similar crevice on HCA, thereby explaining the observation that mutation of this residue did not alter the BoNT/A-SV2C interaction at all. Mutants gSV2CLD-Fc N559Q and S561A also devoid of the N559-glycan showed similarly decreased binding to HCA, respectively, but the bacterially expressed, non-glycosylated SV2CLD-Fc N559Q and S561A also exhibited reduced affinity for BoNT/A preventing a clear-cut conclusion about the role of the N559-glycan in BoNT/A recognition. However, SV2CLD-Fc N559A bound BoNT/A indistinguishably from SV2CLD-Fc wild-type thereby unambiguously demonstrating the important role of N559-glycan in BoNT/A recognition. The function of the N559-glycan is reminiscent of the BoNT/E-SV2A interaction which is facilitated by the presence of an N-glycan attached to the homologous residue N573 . Whether the N573Q mutation in SV2A used by Dong et al. negatively influences the already weak SV2A-BoNT/E protein–protein interaction cannot be revealed, because it is impossible to detect binding of BoNT/E to E. coli-expressed non-glycosylated SV2ALD in vitro [17,19]. However, conclusions drawn that the N573-glycan also contributes to BoNT/A binding and uptake by using the SV2A N573Q mutant  are ambiguous because we demonstrated a negative impact on the BoNT/A-SV2C protein–protein interaction by a glutamine at the homologous position 559.
Kinetic analysis demonstrates that N559-glycan clearly increases affinity of SV2C towards BoNT/A. The known affinity of BoNT/A towards bacterial SV2C suggests that the protein–protein interaction controls the overall binding mechanism by initiating the binding process. Subsequently, the N559-glycan is able to bury in the neighbouring crevice via multiple non-covalent interactions which on one hand decreases the association rate ka 4-fold, but on the other hand reduces the dissociation rate kd of HCA 50-fold. In other words, the residence time of BoNT/A, bound in a trivalent fashion to a single ganglioside, the SV2C protein backbone and its N559-glycan, is sufficiently extended to allow its efficient endocytosis upon recycling of the SV. A similar mechanism can be postulated also for the other SV2 binders BoNT/E and F. However, in their case N-glycosylation is much more crucial since the protein–protein interactions between BoNT/E and F with SV2A-C are very weak, even preventing the demonstration of binding to bacterially expressed SV2A-C LD4 [17,19]. The observation that BoNT/E exhibits an expanded bipartite interface for the interaction with SV2A  which differs significantly from the SV2C binding site of BoNT/A  might compensate the weak protein–protein interaction.
Previously, the identification and characterization of the SV2C binding site of BoNT/A by means of site-directed mutagenesis  identified residues which retrospectively do not directly interact with amino acids of SV2C according to the SV2C-HCA co-crystal structure . Explicitly, mutation of the amino acids N905, F917, T1063, H1064, D1108, V1287 and G1292 interfered with the neurotoxicity of BoNT/A in the mice phrenic nerve hemidiaphragm assay, e.g. mutant G1292R showed a 350-fold reduction in neurotoxicity . A structural explanation for this observation might be that the large arginine residue blocks the core structure of the N559-glycan from entering its binding site. The other residues are located further away from the crevice near G1292 up to a groove on the surface of HCCA. Aromatic residues like F917 and H1064 are well known to interact with the hydrophobic face of the carbohydrates, whereas polar residues like N905, T1063 and D1108 might form H-bonds with the sugars’ hydroxyl groups. ND2 of N559 in SV2C is 18 Å (1 Å=0.1 nm), 21 Å and 29 Å away from F917, N905 and D1108 respectively. The distance between the ND2 of N314 in human IgG and its terminal Man in the common core structure subunit ranges from 18 to 20 Å . The subsequent GlcNAc adds another 5 Å. Complex N-glycan types are expected to extend up to 30 Å out from the N559 residue and will occupy a large volume as calculated for similar glycan structures . Hence this groove is probably also occupied by several carbohydrates of the N559-glycan und prompted us to analyse the glycan type at N559 by MS. Our LC–MS/MS analysis revealed a complex tetra-antennary core-fucosylated N559-glycan (45% abundance) of which 13% comprises a bisecting GlcNAc comprising terminal Gal and GlcNAc added to the GlcNAc2(Fuc1)Man3 core structure. Hence, the terminal Gal and GlcNAc can easily reach N905 and D1108 whereas the core Fuc might add additional hydrophobic interactions with either F953, H1064 and/or Y1066. In summary, the N559-glycan can establish extensive interactions within the groove built up i.a. by N905, H1064 and D1108 which most probably causes the slow dissociation rate kd and results in the high affinity interaction with BoNT/A.
Limited knowledge about neuron specific N-glycan types is available and nothing is known about the type of N-glycans found in synaptic vesicle proteins. Therefore a comprehensive analysis of native SV proteins with respect to N-glycans should be performed. However, since our HEK cell-expressed gSV2CLD-Fc supports high affinity binding of BoNT/A we assume that mainly the GlcNAc2(Fuc1)Man3 core structure of the N559-glycan contributes to recognition of BoNT/A whereas sugar moieties more distal to N559 probably protrude from the surface of HCA and therefore do not or only transiently participate in the interaction. In contrast, the observed expanded bipartite interface in BoNT/E for the interaction with SV2A points towards a different glycan recognition by BoNT/E . Ultimately, N559/573-glycan types of native SV2C and SV2A remain to be elucidated by LC–MS/MS methods and the interacting amino acids by structural approaches.
In conclusion, by means of recombinant eukaryotic expression of glycosylated SV2CLD we were able to demonstrate that BoNT/A exhibits a 12-fold higher affinity for glycosylated compared with non-glycosylated SV2C and identified the N559-glycan as being responsible. Furthermore, this work provides insights into the mode of binding of BoNT/A to N-glycan structures. This detailed knowledge will be helpful for the development of multivalent binding inhibitors to prevent BoNT/A intoxications as well as of high affinity peptides to be employed in innovative detection systems to capture or enrich BoNT/A e.g. out of complex matrices.
Stefan Mahrhold, Tomas Bergström, Daniel Stern, Crister Astot, Brigitte Dorner and Andreas Rummel conceived and designed the experiments. Stefan Mahrhold, Tomas Bergström and Daniel Stern performed the experiments. Stefan Mahrhold, Tomas Bergström, Daniel Stern, Crister Astot, Brigitte Dorner and Andreas Rummel analysed the data. Stefan Mahrhold and Andreas Rummel wrote the paper with input from all other authors.
This work was supported by the Bundesministerium für Bildung und Forschung [grant numbers FK 031A212A (to A.R.) and FK 031A212B (to B.G.D.)]; and the Swedish Civil Contingencies Agency.
We thank Nadja Krez for excellent technical assistance and Dr Jasmin Weisemann for critical discussions and proof reading.
Abbreviations: AA, amino acid; BoNT, botulinum neurotoxin; Fc, homodimer of second and third constant domains of IgG heavy chains (fragment crystallisable); GBS, ganglioside binding site; HC, C-terminal fragment of the HC; HC, heavy chain; HCA, HC of BoNT serotype A; HCC, C-terminal domain of HC; HCE, HC of BoNT serotype E; HCN, N-terminal domain of HC-fragment; IAM, iodoacetamide; LC, light chain; NGS, N-glycosylation site; NMJ, neuromuscular junction; PNG, putative N-glycosylation site; SPR, surface plasmon resonance; SV2, synaptic vesicle glycoprotein 2; Syt, synaptotagmin; TeNT, tetanus neurotoxin
- © 2016 The Author(s). published by Portland Press Limited on behalf of the Biochemical Society