Unique Substrate Recognition Mechanism of the Botulinum Neurotoxin D Light Chain*

Background: The mechanism of botulinum neurotoxin D light chain (LC/D) substrate recognition is not well defined. Results: A dual recognition strategy employed by LC/D was revealed, in which one site of VAMP-2 was recognized by two independent, functionally similar LC/D sites that were complementary to each other. Conclusion: LC/D utilizes a unique mechanism for substrate recognition. Significance: This study provides insights for LC/D engineering and antitoxin development. Botulinum neurotoxins are the most potent protein toxins in nature. Despite the potential to block neurotransmitter release at the neuromuscular junction and cause human botulism, they are widely used in protein therapies. Among the seven botulinum neurotoxin serotypes, mechanisms of substrate recognition and specificity are known to a certain extent in the A, B, E, and F light chains, but not in the D light chain (LC/D). In this study, we addressed the unique substrate recognition mechanism of LC/D and showed that this serotype underwent hydrophobic interactions with VAMP-2 at its V1 motif. The LC/D B3, B4, and B5 binding sites specifically recognize the hydrophobic residues in the V1 motif of VAMP-2. Interestingly, we identified a novel dual recognition mechanism employed by LC/D in recognition of VAMP-2 sites at both the active site and distal binding sites, in which one site of VAMP-2 was recognized by two independent, but functionally similar LC/D sites that were complementary to each other. The dual recognition strategy increases the tolerance of LC/D to mutations and renders it a good candidate for engineering to improve its therapeutic properties. In conclusion, in this study, we identified a unique multistep substrate recognition mechanism by LC/D and provide insights for LC/D engineering and antitoxin development.

Our data show that, similar to other serotypes, LC/D recognition of VAMP-2 occurs though multistep binding and, in particular, recognition of VAMP-2 sites by LC/D substrate recognition pockets. Interestingly, in contrast to all other BoNT LCs and metalloproteinases, LC/D employs a novel dual recognition mechanism, in which one VAMP-2 site is recognized by two independent LC/D sites that are complementary to each other. The dual recognition strategy increases the tolerance of LC/D to mutations and makes it a good candidate for engineering to improve its pharmacological properties.

EXPERIMENTAL PROCEDURES
Molecular Modeling-The structure of the LC/D⅐VAMP-2 complex was modeled and analyzed using SWISS-MODEL and refined using PyMOL as described previously (15).
VAMP-2 and LC/D Mutagenesis-The introduction of point mutations into LC/D and VAMP-2 genes was also performed using the QuikChange protocol as described previously (15,16). Plasmids were sequenced to confirm the mutations and that additional mutations were not present within the open reading frame of VAMP-2 and LC/D. Mutant proteins were produced and purified as described previously (15,16).
Linear Velocity and Kinetic Constants-Determination of the linear velocity and kinetic constants of LC/D and its derivatives was performed as described previously (14,15,19). Briefly, 10 M VAMP-2 or the indicated VAMP-2 derivatives were incubated with various concentrations of LC/D or its derivatives in 10 l of reaction buffer (10 mM Tris-HCl (pH 7.6) and 20 mM NaCl) at 37°C for 20 min. The reactions were stopped by adding an equal volume of SDS-PAGE sample buffer, boiled at 100°C for 5 min, and analyzed for the relative abundance of the substrate and cleaved product by SDS-PAGE (12% polyacrylamide gels). The amount of cleaved VAMP-2 was determined by densitometry. K m and k cat determinations were performed using the same assay, in which VAMP-2 concentrations were adjusted to between 1 and 20 M to achieve ϳ10% cleavage by LC/D or its derivatives. The amount of cleaved VAMP-2 substrate was determined by densitometry, and the velocity was determined by dividing the amount of substrate cleavage by the reaction time. The reaction velocity against the substrate concentration was fitted to the Michaelis-Menten equation, and kinetic constants were derived using GraphPad. At least five independent assays were performed to determine the kinetic constants for each protein.
Trypsin Digestion of LC/D and Its Derivatives-10 M LC/D and its derivatives were incubated with 2 mM trypsin in a 20-l reaction volume at 37°C for 30 min. The reactions were stopped by adding SDS-PAGE sample buffer, subjected to SDS-PAGE, and stained to visualize the partial trypsin digestion profiles.
Far-UV Circular Dichroism Analysis-LC/D and its derivatives were subjected to far-UV CD analysis. CD spectroscopy was performed at a wavelength range of 200 -250 nm at room temperature with a JASCO J-810 spectropolarimeter. Far-UV CD data were obtained with a 10-mm path length quartz cuvette containing 500 l of protein solution (0.1-0.4 mg/ml protein in 10 mM Tris-HCl (pH 7.9) and 20 mM NaCl) at a scanning speed of 50 nm/min and a 2-s response time. Each sample was measured in triplicate, the CD data were converted to molar ellipticity, and the spectrum was generated using GraphPad.

RESULTS
A previous study has shown that several residues of VAMP-2 contribute to substrate cleavage by LC/D, including Val 39 , Val 42 , Met 46 , Val 49 , Asp 53 , Lys 59 , Leu 60 , and Ser 61 (18). To better quantify the degree of contribution of these residues to LC/D substrate cleavage, linear velocity assays were performed to depict the impact of alanine mutagenesis of these residues. The results show that amino acid changes in VAMP-2, including V39A, V42A, M46A, V49A, D53A, D57A, K59A, L60A, and S61A, caused ϳ25-, 10-, 125-, 20-, 25-, 10-, 20-, 25-, and 20-fold reductions in LC/D cleavage, respectively. These data confirm that the residues in VAMP-2 contribute significantly to LC/D substrate recognition and cleavage. Analysis of the modeled structure of the LC/D⅐VAMP-2 complex identified three putative substrate recognition pockets in the active site of LC/D, S2Ј, S1Ј, and S3, which may specifically recognize the VAMP-2 P2Ј (Ser 61 ), P1Ј (Leu 60 ), and P3 (Asp 53 ) sites, respectively ( Fig.  1). In addition, we also predicted several other substrate-binding pockets distal to the active site of LC/D, including the B1-B5 binding sites (Fig. 1). To characterize the substrate recognition pockets and to confirm the specific recognition of VAMP-2 sites, the LC/D residues constituting the substrate recognition pockets were mutated to different amino acid residues to test the effects of these changes on substrate recognition. To exclude the possibility that the effect was due to the Left panel, view of the active site side; right panel, view after a 90°clockwise turn. LC/D is shown as a surface structure, and VAMP-2 is shown as a ribbon structure. The active site recognition and binding site interactions are highlighted. Negatively charged residues are shown in red, positively charged residues are shown in blue, hydrophobic residues are shown in gray, and polar residues are shown in green.

VAMP-2 Recognition by BoNT/D
overall conformational changes as a result of the mutations, we performed partial trypsin digestion on different mutant proteins, and the result indicated that all of the LC/D mutants had an identical digestion profile to WT-LC/D, suggesting that the mutations did not cause any conformational change in LC/D (data not shown). In addition, far-UV CD analysis of LC/D and its derivatives indicated that LC/D(I151D) had a slightly different spectrum compared with WT-LC/D, whereas all other LC/D derivatives had the same far-UV CD spectrum as WT-LC/D (Fig. 2). The curve vertexes of LC/D and its derivatives at ϳ240 nm looked different, which is probably due to the high degree of flexibility of LC/D owing to its relatively high number of turns and random coils. Therefore, the different curves at ϳ240 nm did not reflect the conformational changes of LC/D derivatives (Fig. 2). The CD spectra of the other LC/D derivatives were similar to that of WT-LC/D (data not shown).

Recognition of the P2 Site (Ser 61 ) of VAMP-2 by the S2 Pocket of LC/D
The P2Ј site (Ser 61 ) of VAMP-2 plays a certain role in LC/D substrate recognition, as the VAMP-2(S61A) mutation reduced LC/D substrate hydrolysis by ϳ20-fold. A S2Ј pocket in LC/D that recognized the P2Ј site of VAMP-2 at Ser 61 was identified through analysis of the modeled structure of the complex of LC/D and VAMP-2. The S2Ј pocket is composed of Arg 372 , and the R372A mutation resulted in an ϳ40-fold reduction of LC/D activity, with almost the same K m and an ϳ40-fold lower k cat compared with WT-LC/D ( Fig. 3a and Table 1). These data suggest that the S2Ј pocket (Arg 372 ) of LC/D may recognize the P2Ј site of VAMP-2 by forming a hydrogen bond between these two residues.

Dual Recognition of the VAMP-2 P1 Site by the S1 Pocket of LC/D
The P1Ј site of VAMP-2 was shown to be important for LC/D substrate recognition, as seen in other serotypes of BoNT (14,15,20,21). Mutation of the VAMP-2 P1Ј residue (L60A) reduced LC/D substrate hydrolysis by ϳ25-fold. A corresponding S1Ј pocket composed of two hydrophobic residues, Tyr 168 and Leu 200 , was identified in LC/D (Fig. 3a). The LC/D(Y168A) mutation had no effect on LC/D substrate hydrolysis, and LC/D(Y168D) affected substrate hydrolysis by only ϳ2-fold. The LC/D(L200A) mutation resulted in an ϳ2-fold reduction of LC/D substrate hydrolysis, whereas LC/D(L200D) affected substrate hydrolysis by ϳ8-fold ( Table 1). The complementary effect of Tyr 168 and Leu 200 was also examined, and we found that although LC/D(Y168A/L200A) resulted in only an ϳ2fold reduction of substrate hydrolysis, LC/D(Y168D/L200D) reduced substrate hydrolysis by 60-fold, with no effect on K m and an ϳ60-fold reduction of k cat . These data suggest that a hydrophobic S1Ј pocket is necessary to maintain the full recognition of VAMP-2 Leu 60 and that both hydrophobic residues in the S1Ј pocket, Tyr 168 and Leu 200 , play a complementary role in Leu 60 recognition.

The S3 Pocket Residue of LC/D Interacts with the P3 Residue (Asp 57 ) of VAMP-2
Asp 57 at the P3 site of VAMP-2 plays a certain role in LC/D substrate recognition, as the D57A mutation reduced LC/D substrate hydrolysis by ϳ20-fold. The S3 pocket of LC/D that specifically recognized the P3 site residue (Asp 57 ) of VAMP-2 contains Arg 63 . The R63A mutation had almost no effect on K m , but reduced substrate catalysis by ϳ13-fold ( Fig. 3a and Table  1). The charge reversal mutation LC/D(R63E) caused an ϳ50fold reduction of LC/D substrate hydrolysis, with no effect on K m and an ϳ50-fold reduction of k cat . These data suggest that a salt bridge or a side chain hydrogen bond is important for recognition of Asp 57 of VAMP-2 by Arg 63 of LC/D.

The Main Chain Oxygen Atom of LC/D Pro 64 Interacts with the VAMP-2 P1 Site Residue (Lys 59 )
There was no obvious residue or pocket that showed an interaction with the P1 site residue (Lys 59 ) of VAMP-2. The VAMP-2(K59A) mutation caused an ϳ20-fold reduction of LC/D substrate hydrolysis, suggesting a role for Lys 59 in LC/D substrate recognition. Structural analysis indicated that Lys 59 could potentially interact with the oxygen atom of LC/D Pro 64 through formation of a hydrogen bond. This interaction could not be tested through mutational analysis (Fig. 3a).

Recognition of VAMP-2 Val 53 by the LC/D B1 Binding
Site-The VAMP-2(V53A) mutation caused an ϳ25-fold reduction of LC/D hydrolysis. The B1 binding site of LC/D, which is formed by two residues, Phe 50 and Ile 191 , may interact with Val 53 (Fig. 3b). The LC/D(F50A) and LC/D(I191A) mutations had no effect on LC/D substrate hydrolysis, whereas LC/D(F50D) and LC/D(I191D) caused an ϳ4-fold reduction of VAMP-2 hydrolysis. To test for the complementary effect of these two residues on Val 53 recognition, the effects of double mutations were tested. The double mutation LC/D(F50A/ I191A) resulted in an ϳ60-fold reduction of substrate hydrolysis, with an ϳ2-fold increase in K m and an ϳ25-fold decrease in k cat , whereas LC/D(F50D/I191D) reduced substrate hydrolysis by ϳ400-fold, with an ϳ2-fold increase in K m and an ϳ200-fold decrease in k cat (Table 1). These data suggest that Phe 50 and Ile 191 have independent but complementary effects on Val 53 substrate recognition. The recognition of Val 53 might contribute mainly to the fine orientation of VAMP-2 for optimal substrate recognition in the active site of LC/D because this recognition site contributed mainly to substrate catalysis (k cat ), but not substrate binding (K m ).
Recognition of VAMP-2 Asn 49 by the LC/D B2 Binding Site-The VAMP-2(N49A) mutation was associated with an ϳ20fold reduction in cleavage efficiency of LC/D. The B2 binding site, which is composed of Arg 23 and His 132 of LC/D and may interact with Asn 49 , was revealed though analysis of the structure of the LC/D⅐VAMP-2 complex (Fig. 3c). The LC/D(R23A) and LC/D(H132A) mutations had no effect on LC/D substrate hydrolysis (Table 1). Surprisingly, the charge reversal mutation LC/D(R23D) still maintained the full activity on VAMP-2 as did WT-LC/D. The charge reversal mutation LC/D(H132D) AS-S1Ј P1Ј (Leu 60 ) a WT-VAMP-2 hydrolysis was measured as the ratio of the amount of LC/D derivatives needed to cleave 50% of WT-VAMP-2 to the amount of WT-LC/D needed to cleave 50% of WT-VAMP-2. b The numbers in parentheses are the S.E. of at least five independent experiments. c AS, active site; ND, not detectable. The mutant was too inactive to determine its kinetic constants in our experiments. d -, kinetic constants were not determined.

VAMP-2 Recognition by BoNT/D
became inactive in cleaving VAMP-2, whereas LC/D(H132Q) reduced substrate hydrolysis by ϳ100-fold (Table 1), suggesting that the formation of a hydrogen bond between LC/D His 132 and VAMP-2 Asn 49 may contribute to this recognition and that a negatively charged residue at position 132 may impair this interaction. However, the minimal effect of the LC/D(H132A) mutation on LC/D substrate hydrolysis may be due to the complementary effect of Arg 23 . To test this hypothesis, the effect of R23D and H132A double mutations on LC/D substrate hydrolysis was tested. LC/D(R23D/H132A) reduced K m by ϳ11-fold and k cat by ϳ3-fold (  (18), under our assay conditions, the VAMP-2(M46A) mutation reduced LC/D substrate hydrolysis by ϳ125-fold. Structural analysis also identified the B3 binding site, which is composed of two residues, Ile 151 and Val 148 , and which may interact with Met 46 (Fig. 3c). The point mutations associated with the LC/D(V148A) and LC/D(I151A) alterations did not show any effect on VAMP-2 hydrolysis. LC/D(V148D) also did not have any impact on VAMP-2 hydrolysis, but LC/D(I151D) reduced LC/D substrate hydrolysis by ϳ1000fold ( Table 1). The dramatic effect of the LC/D(I151D) mutation may be partially due to its conformational change based on our far-UV CD analysis. Similar to Asn 49 recognition, the minimal effect of the LC/D(I151A) mutation may be related to the complementary effect of Val 148 . To test this hypothesis, the effect of Ile 151 and Val 148 double mutations on LC/D substrate hydrolysis was tested. LC/D(I151A/V148D) reduced K m by ϳ10-fold and k cat by ϳ2-fold ( Table 1), suggesting that both Ile 151 and Val 148 play a role in LC/D substrate catalysis, with Ile 151 playing a dominant role in substrate recognition. This substrate recognition contributes significantly to substrate binding.
Recognition of VAMP-2 Val 42 by the LC/D B4 Binding Site-Val 42 of VAMP-2 is also important for LC/D substrate hydrolysis, and the VAMP-2(V42A) mutation affected LC/D cleavage of VAMP-2 by ϳ10-fold. Based on the modeled structure of the LC/D⅐VAMP-2 complex, LC/D Trp 315 in the B4 binding site was predicted to have a direct interaction with Val 42 (Fig. 3d). The LC/D(W315A) mutation reduced VAMP-2 hydrolysis by ϳ20-fold, with an ϳ4-fold increase in K m and an ϳ5-fold decrease in k cat , whereas LC/D(W315D) reduced VAMP-2 hydrolysis by ϳ40-fold, with an ϳ20-fold increase in K m and an ϳ2-fold decrease in k cat . These data suggest that LC/D Trp 315 and VAMP-2 Val 42 substrate recognition contributes significantly to substrate binding.
Recognition of VAMP-2 Val 39 by the LC/D B5 Binding Site-Another hydrophobic residue in VAMP-2 that contributes to LC/D substrate hydrolysis is Val 39 . The mutation VAMP-2(V39A) resulted in an ϳ25-fold reduction of LC/D hydrolysis. A hydrophobic pocket in LC/D, the B5 binding site, which is formed by Trp 44 , Ile 152 , and Pro 154 , was identified through analysis of the structure of the LC/D⅐VAMP-2 com-plex (Fig. 3d). The W44D, P152D, and P154D mutations were associated with ϳ2-, 2-, and 4-fold reductions of substrate hydrolysis, respectively (Table 1). These mutations had no effect on k cat , but caused a 4-fold increase in K m , suggesting a role for this pocket in substrate binding. Interestingly, the triple mutation LC/D(W44A/I152A/P154A) resulted in an ϳ20-fold increase in K m , but had no effect on k cat (Table 1), suggesting a complementary effect of these three residues on Val 39 recognition and VAMP-2 binding.

DISCUSSION
It was proposed that botulinum neurotoxins recognize their substrates through two separate regions, one that contains the scissile bond and the other distal to the scissile bond and containing the SNARE motif (22). A two-region substrate recognition model has been demonstrated in LC/A, LC/B, LC/E, LC/F, and the tetanus neurotoxin LC (14,15,20,21). The significance of the SNARE motif was not consistently proven in these toxins. However, for LC/D, the SNARE V1 motif ( 38 QVDEVVDIMR 47 ) was shown to be important for substrate recognition and hydrolysis. Three conservative hydrophobic residues in the V1 motif of VAMP-2, Val 39 , Val 42 , and Met 47 , are critical for efficient LC/D substrate hydrolysis. In addition, LC/D also utilizes other hydrophobic interactions to recognize the VAMP-2 substrate, such as the recognition of the P1Ј site (Leu 60 ) of VAMP-2 by the hydrophobic S1Ј pocket of LC/D and the recognition of Val 53 of VAMP-2 by the hydrophobic B1 binding site of LC/D. In contrast to the substrate binding contributed by the SNARE V1 motif (B3, B4, and B5), the B1 and B2 binding sites contribute more to LC/D substrate catalysis than substrate binding. The data suggest that the V1 motif plays a significant role in LC/D substrate binding, whereas the B1 and B2 binding sites may help more in fine-tuning the orientation of the substrate for specific recognition by the active site of LC/D rather than direct substrate binding.
In this study, we revealed the mechanism of LC/D substrate recognition and specificity. After internalization to the cytoplasm of neuronal cells, LC/D attacks the free form of VAMP-2 through interaction with and recognition of hydrophobic residues in the V1 motif of VAMP-2, including Val 39 , Val 42 , and Met 46 , by the substrate-binding regions B5, B4, and B3 of LC/D on the substrate-binding cleft, respectively. In particular, binding of Met 46 of VAMP-2 to the LC/D B3 binding site was suggested to be very important for LC/D substrate recognition. This binding facilitates further binding of VAMP-2 Asn 49 and Val 53 to the B2 and B1 binding sites located at the active site surface of LC/D. The recognition of VAMP-2 Val 53 by the LC/D Phe 50 /Ile 191 pocket further orientates and stabilizes VAMP-2 for subsequent recognition of its different P sites by the corresponding S pockets in the active site of LC/D. Active P site recognition includes the formation of a salt bridge between P3 (Asp 57 ) of VAMP-2 and S3 (Arg 63 ) of LC/D, a hydrogen bond interaction between P1 (Lys 58 ) of VAMP-2 and the main chain oxygen atom of Pro 64 , recognition of P1Ј (Leu 60 ) of VAMP-2 by the S1Ј pocket (Tyr 168 and Leu 200 ) of LC/D, and finally a hydrogen bond interaction between P2Ј (Ser 61 ) of VAMP-2 and the S2Ј pocket (Arg 372 ) of LC/D. The anchoring of VAMP-2 P sites to different S pockets in the active site of LC/D aligns the VAMP-2 scissile bond close enough to the active site zinc ion to facilitate peptide bond cleavage (Fig. 4).
Compared with substrate recognition by other serotypes of BoNT, LC/D possesses unique features of substrate recognition (14,15,20,21). First, hydrophobic interaction between LC/D and VAMP-2 plays an important role in substrate recognition. The interaction between VAMP-2 Met 46 and LC/D Ile 151 seems to be critical for LC/D substrate recognition. This may be the first step in substrate recognition, which may facilitate the conformational change in VAMP-2 from a double helix to a free loop confirmation, favoring the subsequent substrate binding and catalysis by different regions of LC/D. Further research may be needed to test this hypothesis. However, far-UV CD analysis showed that I151D displays a slightly different conformation compared with WT-LC/D, suggesting that the significant effect of I151D substrate recognition may be partially due to the conformational change in the whole protein, but not the loss of the Ile 151 site recognition. Second, in contrast to the recognition of one site of the substrate by one pocket of the LC for other serotypes of BoNT, LC/D utilizes two functionally similar residues to recognize one site of VAMP-2, such as the S1Ј pocket (Tyr 168 -Leu 200 ) of LC/D for recognition of the P1Ј site (Leu 60 ) of VAMP-2 in the active site of LC/D. In addition, dual recognition is also commonly employed at the substratebinding regions. VAMP-2 Val 53 is recognized by the LC/D pocket formed by Phe 50 and Ile 191 . Mutation to each residue did not have much effect (maximum of 4-fold) on substrate hydrol-ysis, whereas mutation to both residues resulted in a dramatic reduction (400-fold) of substrate hydrolysis. VAMP-2 Asn 49 and Met 46 are also recognized by pockets with the dual recognition mechanism. The pocket that recognizes Asn 49 of VAMP-2 is formed by Arg 23 and His 132 of LC/D. Although His 132 plays a dominant role in Asn 49 recognition, the H132A mutation can be complemented by Arg 23 . Similar to Asn 49 , the pocket that recognizes Met 46 of VAMP-2 is formed by Val 148 and Ile 151 of LC/D. Ile 151 plays a dominant role in Met 46 recognition, whereas Val 148 of VAMP-2 can play a complementary role when Ile 151 is mutated to alanine. Finally, the pocket that recognizes Val 39 of VAMP-2 is formed by three hydrophobic residues (Trp 44 , Ile 152 , and Pro 154 ) of LC/D. Mutation of each residue to alanine or asparagine had no effect or only a minor effect on substrate hydrolysis, whereas triple mutations to alanine resulted in a much stronger reduction of substrate hydrolysis, highlighting the complementary effects of the three residues forming this pocket. The presence of two or more functionally similar residues in the same substrate recognition pocket enables LC/D to tolerate mutations. This property of LC/D makes it a good candidate for further protein engineering.
Unlike BoNT/A, which is the most toxic botulinum neurotoxin and is implicated in human botulism, BoNT/D is responsible mainly for animal botulism, such as cattle botulism. However, our data indicate that LC/D and LC/A exhibit a similar potency in hydrolyzing their substrates under in vitro conditions (16). The role of BoNT/D as a human therapy or bioterrorism weapon remains to be investigated. Our data provide insights into the development of novel BoNT-based therapies and BoNT/D antitoxins.