Neuron-recognizable characteristics of peptides recombined using a neuronal binding domain of botulinum neurotoxin

Recombinant peptides were designed using the C-terminal domain (receptor binding domain, RBD) and its subdomain (peptide A2) of a heavy chain of botulinum neurotoxin A-type 1 (BoNT/A1), which can bind to the luminal domain of synaptic vesicle glycoprotein 2C (SV2C-LD). Peptide A2- or RBD-containing recombinant peptides linked to an enhanced green fluorescence protein (EGFP) were prepared by expression in Escherichia coli. A pull-down assay using SV2C-LD-covered resins showed that the recombinant peptides for CDC297 BoNT/A1, referred to EGFP-A2ʹ and EGFP-RBDʹ, exhibited ≥ 2.0-times stronger binding affinity to SV2C-LD than those for the wild-type BoNT/A1. Using bio-layer interferometry, an equilibrium dissociation rate constant (KD) of EGFP-RBDʹ to SV2C-LD was determined to be 5.45 μM, which is 33.87- and 15.67-times smaller than the KD values for EGFP and EGFP-A2ʹ, respectively. Based on confocal laser fluorescence micrometric analysis, the adsorption/absorption of EGFP-RBDʹ to/in differentiated PC-12 cells was 2.49- and 1.29-times faster than those of EGFP and EGFP-A2ʹ, respectively. Consequently, the recombinant peptides acquired reasonable neuron-specific binding/internalizing ability through the recruitment of RBDʹ. In conclusion, RBDs of BoNTs are versatile protein domains that can be used to mark neural systems and treat a range of disorders in neural systems.


Results
Production of the neuro-recognizable recombinant peptides. Two different peptide A2s and RBDs of BoNT/A1, wild-type (ATCC3502; A2 and RBD) and CDC297 [having a same DNA sequence with CDC1903, CDC5328, and CDC51303 (ha − /orfX + A1 strains)]; A2ʹ and RBDʹ], were compared, and the better in terms of the neuro-recognizability was chosen. In particular, only three amino acid sequences at 1123, 1142, and 1156 of two RBDs are different [RBD: valine (V), serine (S), and arginine (R); RBDʹ: isoleucine (I), asparagine (N), and methionine (M)] (Fig. 1a). Only the third amino acids of A2 and A2ʹ were different as serine and asparagine, respectively. EGFP was connected to the N-termini of the peptides through a His6-tag (6 histidine) and a linker to mark these SV2C-LD-bindable peptides (Fig. 1b). The His6-tag and linker were used for purification and flexibility/folding, respectively, and a glutathione S-transferase (GST)-tag of SV2C-LD was adopted for the pull-down assay. Based on the designed constructs, expected molecular weights of His6-tagged EGFP, EGFP-A2, EGFP-A2ʹ, EGFP-RBD, EGFP-RBDʹ, and GST-tagged SV2C-LD are 27.77, 29.61, 29.64, 77.77, 77.78, and 38.88 kDa, respectively. Accordingly, the thick and dark bands with the corresponding sizes were observed clearly in all the sodium dodecyl sulfate (SDS)-PAGE gel lanes of the total and soluble recombinant peptides expressed (Fig. 1c). After purification using Ni 2+ -nitrilotriacetic acid-(Ni-NTA) and glutathione agarose resins, thick and dark target bands were observed clearly, while many non-target bands were disappeared or became vague (Fig. 1d). Unfortunately, some non-target bands remained in the gel lanes of EGFP-A2 and EGFP-A2ʹ, which might contribute to the proteolyzed residues. Consequently, the constructs of EGFP-A2 and EGFP-A2ʹ were optimized in the section below. On the other hand, the remaining peptides (EGFP, EGFP-RBD, EGFP-RBDʹ, and SV2C-LD) were used in further experiments with the present forms.

SV2C-LD-binding affinity of the neuro-recognizable recombinant peptides. The SV2C-LDs
were immobilized onto glutathione agarose resins via a GST-tag at the N-terminus to estimate the binding affinity of the fabricated recombinant peptides to SV2C-LD (Fig. 2a). The recombinant peptides were first incubated with the SV2C-LD-immobilized resins to allow time for binding, and the unbound recombinant peptides were washed out. After the washing step, the resins binding the recombinant peptides were loaded and electrophoresed in an SDS-PAGE gel (Fig. 2b). As shown in the gel image, the SV2C-LD bands were found not only in all the gel lanes, but the bands for the target recombinant peptides were also detected in the respective lanes of EGFP-A2, EGFP-A2ʹ, EGFP-RBD, and EGFP-RBDʹ, whereas none of the bands for EGFP was observed. This result shows none of the binding affinities of EGFP. In addition, the bands of EGFP-A2ʹ and EGFP-RBDʹ were www.nature.com/scientificreports/ thicker and darker than those of EGFP-A2 and EGFP-RBD, respectively, indicating the greater binding affinities of CDC297-originated peptides (EGFP-A2ʹ and EGFP-RBDʹ) than those of the wild-type peptides (EGFP-A2 and EGFP-RBD). This result is consistent with the relative intensity of the target bands (Fig. 2c). The values of EGFP-A2ʹ and EGFP-RBDʹ were ~ 3.43-and ~ 1.97-times larger than those of EGFP-A2 and EGFP-RBD, respectively, showing the greater binding affinities of the CDC297-originate peptides. In addition, the values of EGFP-RBD and EGFP-RBDʹ were ~ 9.66-and ~ 5.54-times higher than those of EGFP-A2 and EGFP-A2ʹ, respectively. This result suggests that the entire sequences of the RBDs bind better to SV2C-LD than the fragments (peptide A2s) of RBDs. In other words, the RBD sequences subtracted from the peptide A2s have a positive effect on the binding affinity. Consequently, EGFP-A2ʹ and EGFP-RBDʹ, the CDC297-originated peptides, were used in further experiments because of the greater binding affinities to SV2C-LD.  www.nature.com/scientificreports/ SV2C-LD-association and -dissociation of the neuro-recognizable recombinant peptides. The association and dissociation of the recombinant peptides to SV2C-LD were quantified by first immobilizing the GST-tagged SV2C-LDs onto well plates coated with anti-GST antibodies. While monitoring the bio-layer interferometric signals, the recombinant peptides were added to the SV2C-LD-immobilized wells to measure the association kinetics of the peptides and washed away with fresh buffer solutions to determine the dissociation kinetics (Fig. 3). The binding thicknesses of EGFP, EGFP-A2ʹ, and EGFP-RBDʹ were increased to 0.050, 0.071, and 0.657 nm within 120 s for the association, respectively. The values of EGFP, EGFP-A2ʹ, and EGFP-RBDʹ were then decreased to 0.015, 0.032, and 0.499 nm within the subsequent 120 s for the dissociation, respectively. The kinetic constants, k a , k d , and K D , were calculated based on the curves for the change in value ( Table 1). The k a values of EGFP-A2ʹ and EGFP-RBDʹ were 29.46-and 201.76-times larger than that of EGFP as 5.26 × 10 3 and 7.69 × 10 2 M −1 s −1 , respectively, which suggests their association properties specific to SV2C-LD. The k d value of EGFP-A2ʹ (6.56 × 10 −2 s −1 ) was 5.96-and 2.29-times larger than EGFP and EGFP-RBDʹ, respectively, showing the fastest dissociation. The K D value of EGFP-RBDʹ was the smallest as 5.45 × 10 0 μM, which was 33.87-and 15.67-times smaller than those of EGFP and EGFP-A2ʹ. This result shows that EGFP-RBDʹ has strong binding affinity to SV2C-LD, at least in protein-level. In contrast, EGFP-A2ʹ exhibited weak binding affinity, which is in accordance with the GST pull-down assay result.   S4b). However, for all recombinant peptides, the green cell signals of MDCK cells at 5 min were significantly smaller than those of differentiated PC-12 cells at 5 min (Fig. S5). This indicates the diffusional internalization non-specific to MDCK cells and indirectly implies the internalization of EGFP-A2ʹ and EGFP-RBDʹ specific to PC-12 cells. Unfortunately, differences in the green endocytic signals among the recombinant peptides are difficult to discriminate by the naked eye. Therefore, the changes in the endocytic green-fluorescent signal were quantified based on the CLFM images over time using the image processing technique, and the kinetic rate constants ( k m ) were analyzed using the regression fitting curves (Fig. 5a). The k m value of EGFP-RBDʹ was the largest at 0.26 s −1 , which showed 2.49-and 1.26-times faster endocytosis than those of EGFP and EGFP-A2ʹ (0.11 and 0.21 s −1 ), respectively. In addition to CLFM, flow cytometry was conducted to validate the endocytosis of the green-fluorescent peptides (Fig. S6a,b). The mean values of the Green-B channel fluorescence in the cells were acquired based on the histograms (Fig. S6c). The mean values subtracting the value of the control group were recorded as the Table 1. Equilibrium dissociation constant (K D = k d /k a ) and apparent association (k a ) and dissociation (k d ) rate constants for the interaction of SV2C-LD with EGFP, EGFP-A2ʹ, and EGFP-RBDʹ, determined from the biolayer interferometric result.  www.nature.com/scientificreports/ ' mean values' and analyzed using the regression fitting curves to determine the kinetic rate constants ( k m ) (Fig. 5b). Like the trend in CLFM, a k m value of EGFP-RBDʹ was the highest at 0.51 s −1 , indicating a 1.36-and 1.18-times faster endocytosis than those of EGFP and EGFP-A2ʹ at 0.38 and 0.43 s −1 , respectively. Despite the discrepancies in the individual value in the two measurements (CLFM and flow cytometry), the tendencies for endocytosis were the same regardless of the measurement. These are also in agreement with the biolayer interferometry above. Overall, both EGFP-A2ʹ and EGFP-RBDʹ can bind selectively to the SV2C-LDs present in the outer membrane of neurons and internalize them into the neurons. Moreover, the neuro-recognizable efficacy of EGFP-RBDʹ is greater than that of EGFP-A2ʹ.

Discussion
The crystal structure of RBD in complex with SV2C-LD reveals dominating backbone-backbone interactions between two short β-sheet, involving residues around E556-F563 of SV2C and R1140-N1147 of RBD. Additionally, the co-crystal structure suggests that a cation-π stacking interaction between F563 of SV2C and R1156 of RBD is crucial for binding 4 . In this concept, mutations in R1140-N1147 or R1156 of RBD can critically influence the binding affinity of RBD to SV2C-LD. Indeed, mutating R1156 of RBD (R1156E) significantly decreased the binding 7 . In present study, only three amino acid sequences at 1123, 1142, and 1156 are different between RBD (V, S, and R) and RBDʹ (I, N, and M). The difference between RBD and RBDʹ at 1123 (V1123I) is known neither as a key amino acid of ganglioside binding site nor protein receptor binding site 5 . Therefore, among three sequence differences between RBD and RBDʹ (V1123I, S1142N, and R1156M), S1142N and R1156M can greatly affect the binding affinity. With this respect, the greater binding affinity of RBDʹ than RBD was verified in this study. Through binding of the A2ʹ and RBDʹ residues, EGFP-A2ʹ and EGFP-RBDʹ exhibited specific binding affinity to SV2C-LD decorating the outer membrane of differentiated PC-12 cells, resulting in endocytosis. In particular, in EGFP-RBDʹ, the stronger SV2C-LD-binding affinity and greater endocytosis were verified compared to those of EGFP-A2ʹ. Therefore, EGFP-RBDʹ by itself could be a possible marker to various SV2C-occurring cells, such as dopaminergic, Purkinje, medium spiny, cholinergic, and motor neurons 13,[15][16][17] . In addition, if the EGFP residue of EGFP-RBDʹ is displaced with other residues having bioactivities in neurons, this recombinant construct could be applied to treat a range of neuronal malfunctions. For example, the removal of K63-linked ubiquitin chains on misfolded α-synuclein accumulates by deubiquitinase Usp8 may cause α-synuclein accumulation in Lewy body disease 34 . The RBDʹ constructs connected to the ubiquitination-upregulating or deubiquitination-downregulating enzymes might be a possible medication for the Parkinson's disease. In this manner, the RBDʹ might be a versatile neuro-recognizable residue for marking and targeting the SV2C-LD-containing cells. Moreover, the RBDʹ constructs can be produced easily and abundantly in E. coli, without animal ethics issues, using simple protein engineering techniques 31 , which can lower the production costs.
Synaptic vesicle-dependent endocytosis is well known for its fast procedures [35][36][37] . For the kiss-and-run, clathrin-mediated endocytosis, and bulk ultrafast endocytosis and endosomal budding models, the spending times for endocytosis are 1-2, 15-20, and ~ 0.1 s, respectively 38 . Therefore, the synaptic vesicle-targeting endocytosis has a great potential for neuron-specific delivery of various neuroactive molecules. This is also in accordance with the PC-12 endocytosis data of the RBDʹ recombinant peptides via fast and strong binding to SV2C-LD in the present study. In previous study, the K D value of wild-type RBD to SV2C-LD was 0.26 μM 7,39 . In present study, the greater binding affinity of RBDʹ over that of wild-type RBD was verified. Therefore, the K D value of RBDʹ was < 0.26 μM, which is a reasonable binding affinity. www.nature.com/scientificreports/ The K D value of RBDʹ was higher than K D values (1.1-5400 pM) of various monoclonal antibodies antitargeting proteins 40 . Moreover, the K D value was increase ~ 21-times to 5.45 μM because of the addition of EGFP-, polyhistidine-(His6-), and linker to the N-terminus of RBDʹ (EGFP-RBDʹ). Such binding affinity can be improved by editing the structure and some key amino acids of RBDʹ or recruiting other RBDs from other subtypes of BoNT/A. Fortunately, there are still synaptic vesicle proteins that can be targeted by the other RBDs, such as SV2A, SV2B, and synaptotagmins instead of SV2C 41 . In particular, synaptotagmins are ~ 9-times more incorporated in one average synaptic vesicle than the SV2 family in terms of the number of their copies 42 . After the recombinant peptides are optimized in terms of neuron-recognizibility, the endosomal escape of the peptides could be a further subject for the development of neuro-marking and -targeting systems. This study shows that the RBDs of BoNTs can be used as the block materials of recombinant peptides for marking nerve systems or treating various neuronal diseases.

Conclusions
Recombinant peptides specifically binding to SV2C-LD were fabricated by recombining EGFP to either peptide A2 (subdomain of RBD) or RBD of BoNT/A1s. The peptide A2 and RBD (A2ʹ and RBDʹ) of CDC297 (Swiss-Plot A2I2R4) had greater binding affinities to SV2C-LD than the wild-type peptide A2 and RBD (Swiss-Plot Q7B8V4), respectively. Among the all recombinant peptides prepared, the K D value of EGFP-RBDʹ was the smallest, suggesting the highest binding affinity to SV2C-LD. EGFP-RBDʹ exhibited the fastest and best endocytosis into simulated neurons owing to the relatively reasonable binding affinity. Overall, the recruitment of RBDʹ improved the neuron-specific binding and endocytosis to the recombinant peptides. In conclusion, RBDs of BoNTs are versatile block materials with the potential for marking neural systems and treating various disorders in the neural systems.  Table 2 lists the E. coli strain and plasmids used in the present study. Deoxyribose nucleic acid (DNA) sequences having > and ≤ 60 base pairs (bps) were cloned using a ligation-independent cloning method with T4 DNA polymerase (New England Biolabs Inc., Beverly, MA, USA) and a site-directed mutagenesis technique, respectively. Gene amplification using a polymerase chain reaction (PCR) with a Q5 site-directed mutagenesis kit (New England Biolabs Inc.) was repeated 30-times with the following cycles: denaturation (98 ℃, 3 s), elongation (72 ℃, 30 s kbp −1 ), and annealing (55-61 ℃, 3 s); the first denaturation and final . Detailed preparation procedures of the primers and plasmid constructs (pET28b-eGFP-His 6 -linker-A 2 , pET28b-eGFP-His 6 -linker-A 2 ʹ, pET28b-eGFP-His 6 -linker-Rbd, pET28b-eGFP-His 6 -linker-Rbdʹ, pET28b-His 6 -A 2 ʹ-linker-eGFP, pET28b-His 6 -eGFP-linker-A 2 ʹ, and pGEX-4 T-1-Sv2cLd) used in this study are described in the Supplementary Information (Table S1, Fig. S7). The recombinant peptide structures (Fig. S8) were optimized by SDS-PAGE of the purified peptides using the total harvested peptides and the soluble and insoluble (precipitated) peptides (50 μg) after the centrifugation. Herein, the quantity of peptides was quantified based on a standard of BSA using a Lowry assay (Protein DC kit; Bio-Rad, CA, USA). The intensities of the target bands were quantified under Image J (available as freeware from http:// rsb. info. nih. gov/ ij/) processing of the gel images and converted to the relative amounts divided by the intensity value of the total harvested His6-A2ʹ-EGFP. GST pull-down assays. The recombinant peptides to SV2C-LD were accessed using GST pull-down assays. First, purified 15 μM SV2C-LD was immobilized on 50 μL slurry of glutathione agarose resin and agitated with 20 μM of the recombinant peptides in a washing buffer (50 mM sodium phosphate buffer; total volume: 100 μL; pH 7.4, 10 min, 4 ℃). As a washing procedure, the resins binding the peptides were collected by centrifugation (13,200 rpm, 1 min), separated from the supernatants, and re-suspended in 1 mL of fresh washing buffer. The resins washed with three repetitions of this procedure were re-suspended in 50 μL of the washing buffer and mixed with 250 μL of the washing buffer containing 12% w/v SDS. The mixture solutions containing the peptide-binding resins were analyzed by SDS-PAGE. The intensities of the bands for the target recombinant peptides were quantified by Image J processing with the acquired gel image. The intensities of the target bands were normalized based on the concept that the amount of SV2C-LD in a single gel lane is the same as those in the other lanes. The normalized intensities were converted to the relative amounts by dividing by the normalized value of EGFP-RBDʹ.

Expression and purification
Bio-layer interferometry. The protein-level binding interactions between the recombinant peptides and SV2C-LD were measured by biolayer interferometry (BLItz system; FortéBio, Fremont, CA, USA). First, by monitoring the signal, 40 μL of GST-tagged SV2C-LD (40 μM) dispersed in the biolayer interferometry buffer (pH 7.4; 0.05% w/v Tween 20 and 1 mg mL −1 BSA in PBS) was dropped and immobilized onto anti-GST biosensors (FortéBio), which was then washed and equilibrated with fresh buffer. Subsequently, by monitoring the signal, 40 μL of the recombinant peptides (20 μM) dispersed in the buffer was dropped onto the SV2C-LD-covered biosensors, followed by washing with fresh buffer. The binding constants, including association rate (k a ), dissociation rate (k d ), and k d /k a (K D ) constants, were calculated from the association and dissociation curves, using the BLItz system software (FortéBio) according to the manufacturer's instruction.
Cell culture and differentiation. Before the culture, the bottom surface of the microplate-wells was coated with I-type collagen by adding 1 mL of a collagen solution (50 μg mL −1 collagen in an aqueous 20 mM The image processing was conducted with the Image J to quantify the green fluorescence of EGFP of the recombinant peptides adsorbed on the cells or internalized in the cells. The images merged were loaded on Image J, and only the green channel fluorescence in the images was collected using the macro (Supplementary Information). Subsequently, the outer line of the individual cell was selected manually based on the merged images. The mean values of the green fluorescence were then measured using the other macro (Supplementary Information) and fitted using regression curves [ mean = a 1 − b t ; a and b : constants; t : time (s)]. The increasing rate ( k m ) of the mean values was determined at the initial point of the incubation as −alnb.
Flow cytometry. In flow cytometry, while all the procedures before the incubation step with the recombinant peptides were the same as those for CLFM, the culture medium-based recombinant peptides (5 μM) were incubated for 2, 4, 8, and 16 min. After incubation, the cells were washed carefully three times with PBS and harvested by intensive pipetting > 100 times. The harvested cells were centrifuged (200 RCF, 3 min) and washed with PBS. The procedure was repeated three times. Subsequently, the cells re-suspended in PBS were loaded into a flow cytometer (guava easyCyte System; Millipore, MA, USA) to acquire the Green-B fluorescence signals of EGFP of the recombinant peptides within the cells. For data analysis, the obtained data files were loaded on FlowJo software (Version 10; FlowJo LLC, OR, USA). Based on the forward scatter-versus-side scatter plots, the signals only for cells (> 4000) were collected, and the signal histograms and mean values were acquired using the software. To quantify the recombinant peptides adsorbed on or internalized in the cells, the mean values acquired at 2, 4, 8, and 16 min were subtracted from a mean value of the cells and fitted using the regression curves [ mean = a 1 − b t ]. The increasing rate ( k m ) of the mean was determined at the initial point of the incubation as −alnb.
Statistical analyses. All data represents an average of at least three independent experiments or measurements. The results are reported as average ± standard deviation. The kinetic parameters and curves were determined and fitted using the regression iteration procedures in SigmaPlot 10.0 (IBM Co., Armonk, NY, USA). Statistical analyses were firstly started by examination of the data using the distribution normality test (Shapiro-Wilk test) and the variance homogeneity test (Levene's test). If normal distribution and homogeneous variance were guaranteed, further statistical analyses (Tukey's test) were conducted using SPSS Statistics (V23.0; IBM Co., Armonk, NY, USA). www.nature.com/scientificreports/