|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: JEZ JEN JCB PW SB ATB. Performed the experiments: JEZ JTH PW. Analyzed the data: JEZ JTH OD JCB RG PW ATB. Contributed reagents/materials/analysis tools: JTH JEN RG. Wrote the paper: JEZ OD JCB PW SB ATB.
The botulinum neurotoxin serotype A light chain (BoNT/A LC) protease is the catalytic component responsible for the neuroparalysis that is characteristic of the disease state botulism. Three related peptide-like molecules (PLMs) were designed using previous information from co-crystal structures, synthesized, and assayed for in vitro inhibition against BoNT/A LC. Our results indicate these PLMS are competitive inhibitors of the BoNT/A LC protease and their Ki values are in the nM-range. A co-crystal structure for one of these inhibitors was determined and reveals that the PLM, in accord with the goals of our design strategy, simultaneously involves both ionic interactions via its P1 residue and hydrophobic contacts by means of an aromatic group in the P2′ position. The PLM adopts a helical conformation similar to previously determined co-crystal structures of PLMs, although there are also major differences to these other structures such as contacts with specific BoNT/A LC residues. Our structure further demonstrates the remarkable plasticity of the substrate binding cleft of the BoNT/A LC protease and provides a paradigm for iterative structure-based design and development of BoNT/A LC inhibitors.
Botulinum neurotoxins (BoNTs), secreted by Clostridium botulinum , provide invaluable treatments for a range of medical conditions , , , , , , , , ,  and cosmetic purposes , , , , . Paradoxically, BoNTs are also the most potent biological toxins known by causing the disease state botulism. As a result, these enzymes are classified as category A biothreat agents by the Centers for Disease Control and Prevention (http://emergency.cdc.gov/agent/agentlist-category.asp), with the clandestine contamination of liquids and/or food stuffs being plausible scenarios , .
The seven known BoNT serotypes are designated A – G. Post secretion, they undergo proteolytic processing to provide the bioactive (i.e., poisonous) holotoxin . The holotoxin is composed of a 100 kDa heavy chain (HC) subunit and a 50 kDa light chain (LC) subunit; these two components are tethered by a disulfide bridge , , . Mechanistically, the HC binds to specific motor neuron receptors and induces endosomal internalization . The LC (BoNT/LC) is a zinc (Zn)(II) metalloprotease that is released from the holotoxin into the neuronal cytosol . Once inside the neuronal cytosol, the LC cleaves specific peptide bonds (depending on the serotype) of proteins composing the neuronal SNARE complex: the synaptosomal-associated protein of 25 kDa (SNAP-25), the vesicle-associated membrane protein (VAMP), also referred to as synaptobrevin, and syntaxin , . Botulinum neurotoxin serotypes A, C, and E cleave SNAP-25 , , ; serotypes B, D, F, and G cleave VAMP , , , , and BoNT serotype C also cleaves syntaxin . The BoNT/LC mediated proteolytic cleavage of any one of the three SNARE proteins prevents acetylcholine-filled vesicles in the neuron from fusing with the active zone at the synaptic cleft . This inhibits the transmission of motor nerve impulses, and as indicated above, results in the flaccid paralysis that is characteristic of botulism .
At present, the only treatments available for BoNT intoxication involve antitoxin administration , followed by critical care mechanical respiration. However, this treatment would not be practical for treating even a modest number of poisoned individuals: antitoxin administration is ineffective after BoNT internalization (and it is likely that victims would seek medical attention only after the paralysis manifestation). Critical care mechanical respiration is costly and the small number of medical facilities in the U.S. equipped with such devices would more than likely be overwhelmed. Thus, there is an urgent need for the development of small-molecule inhibitors of BoNT LCs.
Of the seven BoNT serotype LCs, the BoNT serotype A LC (BoNT/A LC) possesses the longest duration of action in the neuronal cytosol . Hence, there continues to be a significant effort to identify and develop both peptidic and small, drug-like molecule inhibitors  of this particular serotype LC. Previously, we identified and developed BoNT/A LC inhibitors involving the simultaneous identification, design, and generation of both small molecule, non-peptidic, inhibitors (SMNPIs) , , , ,  and peptide-like molecules (PLMs) , . Such PLM design complements SMNPI development, as BoNT/A LC:PLM co-crystal structures form the bases for: 1) the design and synthesis of more potent, drug-like peptidomimetics, 2) the rational, structure-based modification of existing SMNPIs to improve inhibitory efficacies, and 3) the discovery and development of novel SMNPIs via database mining (employing PLM binding modes as search query templates). For example, the conformation and chemical contacts of a PLM bound to the BoNT/A LC can be used to generate three-dimensional (3D) search queries to discover new SMNPI chemotypes via the database mining of virtual small molecule libraries.
Here, we describe three new PLMs (Figure 1) that were designed to explore the BoNT/A LC substrate cleft based on the inhibitor-protease interactions found in a previously published co-crystal structure of BoNT/A LC with the inhibitor I1 . The three new PLMs possess Ki values in the nM range which, together with I1, place them among the most potent BoNT/A LC inhibitors characterized to date. One of the PLMs, JTH-NB72-39, was co-crystallized in complex with the BoNT/A LC protease, confirming the interactions aimed by our design strategy while revealing new, unforeseen inhibitor:enzyme contacts that will preface future studies to design more potent PLM and SMNPI inhibitors.
Previously, we reported several nanomolar (nM)-range PLM inhibitors resembling the cleavage site sequence of SNAP-25 . The seven-residue P1-P6′ sequence QRATKML (residue positions 197–203 of human SNAP-25) was used to design the PLMs. Of these, a co-crystal structure of the BoNT/A LC with PLM I1 (Figure 1) was determined and its binding contacts and mechanism of inhibition (with respect to the LC's active site) were studied in detail . Based on a general design strategy (Figure 1) and the BoNT/A LC:I1 co-crystal structure, we attempted to increase the inhibitory potency of the PLM I1 ‘template’ by replacing and incorporating different components to increase both ionic and hydrophobic contacts with residues in the enzymes's binding cleft, and to stabilize the 310 helical conformation of I1. We hypothesized that such conformational stabilization of the otherwise flexible inhibitor would decrease the binding entropy of the resulting inhibitors, and thus increase affinity. Along these lines, we restrained the conformation of I1, as it is bound to the BoNT/A LC , and thus attempted to reduce the inhibitor's binding entropy, by introducing an aminoisobutyric acid (Aib) residue (Figure 1), which is known to favor the type II' β-turn repeat in a 310 helix , . In addition, we replaced the redox-active DNP-DAP functional group of I1  with an Arg residue, which, as Kumaran et al. ,  demonstrated with co-crystal structures, provides direct electrostatic contacts with anionic residues in the BoNT/A LC substrate cleft (in contrast to the DNP-DAP residue of I1 ). Furthermore, we increased the hydrophobic nature of the C-terminus of I1 by replacing the DAB residue  with an Ala residue (as found in the P4′ position of SNAP-25). Finally, we allowed for combinatorial exploration of the aromatic hydrophobic interactions of the I1 Trp residue (Figure 1), mainly by substituting this position with two sterically and electronically diverse benzyl and naphthylene methylene substituents. The resulting PLM designs JTH-NB72-35, JTH-NB72-38, and JTH-NB72-39 are shown in Figure 1.
The synthesis of the PLMs used manual microwave assisted solid phase peptide synthesis using Fmoc protected amino acids and Rink amide SS resin. After swelling the resin in dichloromethane solvent for 30 min, a stepwise synthesis was initiated by removal of the Fmoc protecting group from the Rink amide resin with a solution of 20% piperidine in DMF. The newly formed free amine was then coupled to the activated, protected amino acid corresponding to the C-terminus of the desired PLM. Initial attempts to activate the protected amino acids for coupling using PyBop and HOBt provided unsatisfactory yields; however, a brief screening of activating agents revealed that Goodman's reagent (DEPBT) provided the desired PLMs in good yield , . With the first amino acid successfully coupled to the Rink resin, the Fmoc group was again removed, and the subsequent, activated, amino acid was coupled to the freshly deprotected peptide chain. This process of deprotection and coupling was repeated until the amide-terminal residue of the desired PLM was appended. Following the final Fmoc deprotection, the PLMs were cleaved from the solid support by stirring in a modified version of Reagent K (87.5% TFA, 3.6% thioanisole, 2.3% EDT, 3.7% phenol, 1.8% H2O, 1.1% triisopropylsilane) for 2 h at RT. Precipitation with diethyl ether provided crude products, which were subsequently purified by preparative RP HPLC (34–42% yield).
Using the methods described below, we obtained Ki values in the nM range for the JTH-NB72-35, JTH-NB72-38, and JTH-NB72-39 PLMs (Figure 1), although none of them were as potent as I1. Therefore, co-crystallization experiments were conducted in order to collect any structural information that might explain this unexpected result.
Of the co-crystallization experiments conducted with the three PLMs, only BoNT/A LC:JTH-NB72-39 produced diffracting crystals. We obtained a co-crystal structure of this complex at 2.4 Å resolution (Table 1). The structure was determined by molecular replacement using the structure of BoNT/A LC as the search model (PDB reference code 3DSE ), but omitting the inhibitor coordinates, water molecules, and other ligands (i.e., Zn(II) and Ni(II) ions) from the search model. Significant electron density for the PLM emerged next to the catalytic Zn(II) around the binding cleft defined by loops 70, 250 and 370 in the LC protease (Figure 2).
The electron density for the first six residues of the PLM inhibitor is well-defined (i.e., visible at a contour level of 2.0 σ in the Fo-Fc difference electron density map), but is weaker for the last Leu residue. As discussed in detail below, most of the specific interactions observed between JTH-NB72-39 and the BoNT/A LC are mediated by the first four residues of the PLM. Briefly, JTH-NB72-39 also possesses the electrostatic contacts reported for the RRGC, RRGI, RRGM, and RRGL tetrapeptides, as well as for the RRATKM PLM. Moreover, our design resulted in some of the same hydrophobic interactions previously observed between I1 and BoNT/A LC , but to a lesser degree.
The carbonyl oxygen of the JTH-NB72-39 P1 residue (Arg) coordinates the enzyme's catalytic Zn(II) ion (distance is 2.4 Å) and also engages in a hydrogen bond with the hydroxyl group of residue Tyr 366 (which is known to directly stabilize the tetrahedral intermediate formed during SNAP-25 catalysis ), while the amino terminal group of the P1 Arg also coordinates the enzyme's catalytic Zn(II) (Figure 2C). Furthermore, the conformation of the backbone atoms of this residue is determined, in part, by electrostatic interactions between its side-chain and BoNT/A LC residues. Specifically, and in contrast to I1, the P1 Arg of the inhibitor engages, via a salt bridge, with the LC's Glu 164 residue while it also shares a hydrogen bond with the carbonyl oxygen of the LC's Cys 165 residue (Figures 3A and and4).4). A similar ionic contact with Glu 164 has been proposed for Gln 197 in the P1 position of SNAP-25 substrate during the formation of the enzyme:substrate complex , . Moreover, the orientation of the JTH-NB72-39 P1 Arg resembles that of the P1 Arg residues found in previously reported tetrameric inhibitors RRGI, RRGC, RRGM, and RRGL and RRATKM  (Figure 3B), but is in closer contact with the enzyme's Glu 164, and it is the only PLM that interacts with Cys 165 of the BoNT/A LC (Figure 4). In contrast to the binding of the P1 DNP-DAP component of I1 , the JTH-NB72-39 P1 Arg does not interact with BoNT/A LC residue Ser 259.
In the P1′ position of JTH-NB72-39 there is a second Arg residue (Figure 3). In SNAP-25, a corresponding P1′ Arg plays a key role in facilitating hydrolysis by specifically interacting with Asp 370 in the S1′ pocket. This native interaction is also exploited by all structurally known peptidic inhibitors, as all BoNT/A LC:PLM co-crystal structures reported to date posses an Arg residue at the P1′ position which provides optimal binding in the enzyme's S1′ pocket , , . This Arg side chain adopts one of two rotamers, and both conformations are stabilized by the formation of a salt bridge with the side-chain carboxylate of BoNT/A LC residue Asp 370 (Figure 3). The conformation of the side-chain of the P1′ Arg of JTH-NB72-39 closely resembles that observed for the corresponding P1′ Arg residue in PLMs RRGI, RRGL, and RRATKM , . Additionally, the guanidinium group of the P1′ Arg engages in a cation-π interaction with Phe 194 of the LC's substrate cleft (not shown). This is a contact that is consistently observed in other BoNT/A LC:PLM co-crystal structures, and mutations of Phe 194 have been reported to diminish the catalytic efficiency of the BoNT/A LC by ~100-fold . Overall, the observed P1′-S1′ Arg:Asp 370/Phe 194 interactions appear to be key for general PLM inhibitory potency.
While looking for additional contacts further down the sequence of the JTH-NB72-39 PLM, we identified an interaction never observed before for any other BoNT/A LC inhibitor. The amide nitrogen of the JTH-NB72-39 P2′ Phe residue engages in a water mediated interaction with the guanidinium group of BoNT/A LC residue Arg 363 (Figure 5A–C). In other BoNT/A LC:PLM complexes, such as RRATKM, and those of the tetrameric peptides RRGI, RRGL, RRGM, and RRGC , , it is the carbonyl oxygen of the PLM's P1′ that directly interacts with the enzyme's Arg 363 side-chain guanidinium group (Figure 5A). For JTH-NB72-39, this carbonyl group is rotated 180° relative to its orientation in the tetrameric peptides (Figure 5A). However, by virtue of this water-mediated interaction with Arg 363, the JTH-NB72-39 P2′ Phe amide nitrogen replaces this direct interaction observed for other PLMs , . This is relevant, as the BoNT/A LC Arg 363 is proposed to be critical for the binding and hydrolysis of the SNAP-25 substrate, presumably by maintaining proper geometry and charge distribution around the active site; mutation of this residue results in a 80-fold decrease of the catalytic rate of SNAP25 hydrolysis by BoNT/A LC , . In addition, the side chain rotamers of Arg 363 are similar for all BoNT/A LC:PLM complexes, but differ from the rotamer observed in the unbound form of BoNT/LC, indicating that Arg 363 undergoes significant conformational changes upon PLM and substrate binding. This water-mediated contact between the JTH-NB72-39 and Arg 363 is not observed in the I1-bound BoNT/A LC complex (Figure 5B).
Another novel interaction observed in this position of the PLM is a stabilizing, intra-molecular hydrogen bond formed between the carbonyl oxygen of the P2′ Phe and the amide nitrogen of the P5′ Met residues (Figure 5D), which is not present in the I1-bound complex. There are also hydrophobic interactions between the aromatic ring of the JTH-NB72-39 P2′ Phe and BoNT/A LC residues previously found to form a hydrophobic pocket for binding by the larger, indol P2′ Trp moiety of I1  (Figure 6). For JTH-NB72-39, the BoNT/A LC side-chains of Leu 367 and Phe 369, together with the aliphatic portions of Asn 368, contribute to the formation of this hydrophobic pocket. Additionally, the aliphatic side-chain of Leu 256 interacts with the JTH-NB72-39 P2′ Phe side-chain (Figure 6), but its electron density is weaker than observed in the BoNT/A LC:I1 complex, suggesting that the larger I1 P2′ Trp is better suited than the JTH-NB72-39 P2′ Phe for stabilizing this residue and forming a non-polar binding site (Figure 6). This observation partially explains the lower potency of these three PLMs (Figure 1) relative to I1.
A new PLM component incorporated into the design of JTH-NB72-39 also present in JTH-NB72-35 and JTH-NB72-38 is the gem-dimethyl-glycine residue, Aib, in the P3′ position (Figure 1). The rationale for incorporating this component was to stabilize the PLMs' observed 310 helical conformation, as inferred from the I1 binding mode , and to decrease the binding entropies of the new designs. In the present co-crystal structure, Aib engages in favorable intermolecular contacts (via its hydrophobic gem-dimethyl groups) with the side-chain of BoNT/A LC residue Val 70, and an intramolecular, stabilizing, hydrogen bond with the amide nitrogen of its P6′ Leu (not shown). In addition to these interactions, our structure reveals unexpected conformational restrains introduced by this Aib residue that affect neighboring residues in the PLM. The backbone atoms of the more potent I1 PLM superimpose well with corresponding atoms found in JTH-NB72-39 up to the P2′ position (Figure 6B). However, the backbone of JTH-NB72-39 abruptly contracts forming a sharp bend at the P3′ position which results from the distinct conformational effects of the Aib residue. Based on this observation, we conclude that, in addition to possessing a smaller volume than the P2′ Trp group of I1, and the concurrent absence of an ionic indol group, the P2′ Phe of JTH-NB72-39 is hindered in its orientation by the geometrical restraints imposed on the its backbone by the adjacent P3′ Aib residue.
The P4′ Ala residue of JTH-NB72-39 engages in intramolecular, hydrophobic interactions with the side-chain methylenes of the PLM's P1 Arg and P6′ Met, as well as with the backbone amides of the PLM's P1′ Arg and P3′ Aib, while the P5′ Met engages in intermolecular interactions with BoNT/A LC residues Glu 257, Val 258, Ser 259, and Glu 262, and intra-molecular contacts with the P1 Arg, P2′ Phe, P4′ Ala, and P6′ Leu. The alkyl chain of the outermost residue in the inhibitor (i.e., JTH-NB72-39 P6′ Leu), although solvent exposed, engages in favorable intramolecular interactions with the hydrophobic surfaces of its neighboring P2′ Phe and P3′ Aib PLM residues, as well as an inter-molecular interaction with the BoNT/A LC Phe 369 side-chain phenyl.
Silvaggi et al. observed structural plasticity in the BoNT/A LC substrate cleft upon binding to three different hydroxamate derivatives . This plasticity has also been documented by the distinct binding contacts identified in subsequent complexes of the BoNT/A LC with other PLM inhibitors , . Our co-crystal structure of JTH-NB72-39 bound to the BoNT/A LC further underscores this plasticity. Importantly, the complexes of BoNT/A LC with JTH-NB72-39 and with I1 (PDB reference code 3DS9, ), and the unbound crystal structure (PDB reference code 3DSE ) crystallized in the same space group with very similar cell dimensions. Thus, we can make comparisons of conformations between these specific crystal structures without the potential influence of crystal packing contacts.
The BoNT/A LC's overall fold is similar in the bound and unbound forms: the JTH-NB72-39-bound form superimposes to the unbound form with a 0.58 Å r.m.s.d. for all Cα atoms. In the BoNT/A LC:JTH-NB72-39 structure, the three LC residues that directly coordinate the Zn(II) ion in the active site, i.e., His 223, His 227 and Glu 262, maintain the same geometry and conformation observed in the unbound form of the LC protease. However, Glu 224 no longer associates with the Zn(II) ion via a ‘catalytic water’ molecule, as observed in all crystal structures of the unbound form of the enzyme; instead it interacts with the amino terminal nitrogen atom of the PLM via a H-bond (Figure 2C). Additional major differences between the bound and unbound structures are observed in the 20, 200, and 250 loops. Specifically, backbone atom differences observed for the 370 loop suggest JTH-NB72-39-induced conformational changes around residues Asn 368, Phe 369, and Asp 370, since similar arrangements of the symmetry mates near the 370 loop are observed for both bound (JTH-NB72-39 and I1) and unbound forms of the LC. Residue Phe 369 moves closer to JTH-NB72-39 than observed for any of the reported tetrameric peptides or RRATKM , . This interaction is even more pronounced in the BoNT/A LC:I1 co-crystal structure. Additionally, I1 is also in closer proximity to BoNT/A LC residue Leu 256  than JTH-NB72-39 (Figure 6). These observations reinforce the hypothesis that the P2′ Trp residue in inhibitor I1 is a more favorable ‘binding anchor’ than the corresponding P2′ Phe of JTH-NB72-39.
In the unbound form of BoNT/A LC, residues 64–70 adopt a loop conformation by packing against the β strand formed by residues 415–420. This loop is unaltered in the I1- and JTH-NB72-39-bound structures (Figure 7A). However, for the crystal structures of the complexes with the hydroxamate derivatives and the CRATKML peptide , no electron density was observed for these residues (Figure 7A). Binding of the tetrapeptides, and of the QRATKM and RRATKM PLMs results in a significant displacement of the backbone in the 70 loop away from the active site (Figure 7B).
Electron density for the 250 loop of the BoNT/A LC is only observed for structures of the LC complexed with PLMs containing either no side chain (i.e. a Gly component) or an Ala in the P2′ position (Figure 7C). In contrast, no electron density is observed for various residues within this loop for either the unbound form of the BoNT/A LC, or when bound to arginine hydroxamate (ArgHX) (residues 245–256), I1 (residues 250–253), or JTH-NB72-39 (residues 247–255) (Figure 7C).
In all BoNT/A LC:inhibitor complexes described thus far, including the co-crystal described herein, there is a conformational change in the 367–372 residue segment of the 370 loop that is associated with rotamer changes in enzyme residues Phe 369 and Asp 370 (Figure 7D). The latter engages in the salt bridge (described above) with the Arg sidechain in the P1′ position of the inhibitors, a key specific interaction for binding of the SNAP-25 substrate and substrate-analog inhibitors. As for Phe 369, its side chain projects away from the binding cleft in the unbound form of the BoNT/A LC. By contrast, in the JTH-NB72-39 complex, the center of this ring moves by ~3.5 Å towards the binding cleft, facilitating the formation of a hydrophobic pocket that accommodates the Phe aromatic ring of the PLM. This conformational change for Phe 369 is even more pronounced in the BoNT/A LC:I1 complex due to this PLM's larger P2′ Trp component (Figure 7D BoNT/A LC). In all other inhibitor complexes, Phe 369 adopts a conformation that differs when compared with either the unbound form of the enzyme, the JTH-NB72-39-, or the I1-bound complexes. Taken together, these conformational changes observed in the BoNT/A LC protease upon binding with different inhibitors reveal a highly ‘plastic’ binding cleft.
Including the present structure, there are now five co-crystal structures of PLM-based inhibitors in complex with the BoNT/A LC that are longer than four residues. A common structural feature found in all five bound PLMs is a 310 helical turn in the inhibitor's backbones (Table 2). As indicated above, JTH-NB72-39 was designed to further stabilize this helical turn by introducing an Aib residue in the P3′ position. Indeed, there is a 310 helical conformation for the backbone atoms of this PLM, although the P3′ Aib residue slightly distorts the helical turn and deviates from the canonical 310-helix conformation adopted by I1 (Figure 8A), likely by virtue of the unusual geometric constrains of this residue, i.e, the gem-dimethyl effect . The electron density suggests that the helical pathway also includes the Met residue in the P5′ position, resembling the 310 helix observed for residues P2′-P5′ in I1.
A helical backbone conformation is also observed in the Zn-chelating CRATKML PLM, similarly encompassing positions P2′ throughout P5′ (Figure 8A). Remarkably, only PLMs containing a Leu in the P6′ position adopt a right-handed 310 α-helical conformation, in contrast to two closely related hexapeptide PLMs lacking a terminal Leu residue (Table 2) . Instead, the and ψ torsion angles for QRATKM and RRATKM do not correspond to a canonical right-handed helical conformation (Figure 8B and Table 2). The backbone atoms of both peptides undergo a geometrical ‘bend’ which is most conspicuous along the backbone of the Thr residue in QRATKM (Table 2 and Figure 8B). Whereas inhibitors CRATKML, I1 and JTH-NB72-39 display a common right handedness (or positive chirality) of canonical α-helices, the QRATKM and RRATKM PLMs display unusual negative values in their dihedral α angles (i.e. negative helical chirality) (Table 2). The helical segments observed for inhibitors I1, JTH-NB72-39, and CRATKML are of similar length and entail the same residue positions (i.e. P2′-P5′); nevertheless, the CRATKML “helix” is slightly shifted relative to the other two PLMs due to Zn-coordination by its thiol group in the P1 Cys. Interestingly, PLMs QRATKM and RRATKM, which do not adopt a canonical helical conformation, are less potent (Ki values =133 µM and 95 µM, respectively) than the ‘helical’ PLMs JTH-NB72-39, I1, and CRATKML all of which possess nM range Ki values (Table S1). Hence, it is likely that the P6′ Leu residue in the helical PLMs not only stabilizes their helical conformations, but also increases inhibitory potency via increased occupation of the enzyme's substrate binding cleft. Together with the favorable hydrophobic interactions of the substituents at the P2′ position of JTH-NB72-35, -38, -39, and I1, the presence of a seventh residue might account for the higher potency of these PLMs over the hexa- or tetrapeptides reported by other groups (Table S1) , .
The PLM JTH-NB72-39 shares the same inter-molecular contacts that facilitate the mechanism of inhibition with other PLMs (Figure 5B), in particular I1, as previously described . Specifically, all PLMs possessing a free amino terminus orient this substituent such that it engages in contacts with the proton shuttle Glu 224 side-chain carboxylate (Figure 5A–C). In this way, Glu 224 is no longer capable of ionizing the ‘catalytic’ water molecule and, as a consequence, it is hampered from using the protons from this water to catalyze the final cleavage of the scissile bond. In addition, all reported PLMs exhibit simultaneous, substrate-like interactions between their P1 residue backbone carbonyl oxygens and both, the side-chain hydroxyl group of Tyr 366 and the Zn(II) ion (Figure 5A–C). Furthermore, BoNT/A LC residue Arg 363 engages in electrostatic contacts with the peptidic backbones of reported PLMs (Figure 5A). The RRGL, RRGI, RRGM, RRGC, and RRATKM peptides ,  consistently contact Arg 363 via their P1′ carbonyl oxygen (Figure 5A). JTH-NB72-39, on the other hand, forms this interaction indirectly, via a bridging water molecule (Figure 5A). This difference is most likely due to the geometric restraints imposed by the aromatic group in the P2′ position of JTH-NB72-39 (which is either a Gly or an Ala in all other PLM inhibitors , ).
The CRATKML inhibitor conformation deviates from those of all the other PLMs. It is likely that the Zn-coordinating geometry of the P1 Cys of CRATKML shifts the other interactions in the complex (Figure 5C) . For its first residue, P1 Cys, the terminal amino group of this PLM is not proximal to the enzyme's proton shuttle - Glu 224, but rather, this P1 Cys engages in contacts with BoNT/A LC residue Arg 363 through its P1 carbonyl oxygen, as opposed to the P1′ carbonyl oxygen as observed for other PLMs ,  (Figure 5A and 5C). As a result, CRATKML also lacks the hydrogen bond with Tyr 366 that is detected in all other PLM co-crystal structures (Figure 5C).
The PLMs described here were designed in an attempt to improve potency displayed by previously reported PLM I1 via the incorporation of a P1 position Arg residue (to engage in direct electrostatic interactions with BoNT/A LC acidic residues), and the incorporation of an Aib residue (to stabilize the inhibitors helical conformations) . However, neither JTH-NB72-35 (Ki=315.5±28.6 nM), JTH-NB72-38 (Ki=990.5±116.9 nM), nor JTH-NB72-39 (Ki=638±92.0 nM) (Figure 1) are as potent as I1 (Ki=41 nM) . The co-crystal structures of JTH-NB72-39 and I1 in complex with BoNT/A LC explain this surprising result. The higher Ki value for JTH-NB72-39 compared to I1 is, in part, due to fewer favorable hydrophobic contacts provided by the PLM's Phe component in the P2′ position versus the larger P2′ Trp of I1. Specifically, the I1 Trp component is more efficient in coalescing the non-polar side-chains of BoNT/A LC residues Leu 256 and Phe 369 in their common binding site (Figure 6). Additionally, while the aliphatic side-chain of Leu 256 does contact the JTH-NB72-39 P2′ Phe, the electron density for the side-chain of this BoNT/A LC residue is weaker than that observed in the BoNT/A LC:I1 complex . Hence, the I1 P2′ Trp residue is more efficient for inducing the formation of this non-polar pocket than the P2′ Phe residue of JTH-NB72-39. Another unexpected result from our co-crystal structure is that the Aib residue introduces a slight deformation of the canonical 310 helical backbone conformation observed for I1 (Figure 7A).
Our co-crystal structure also explains the different potencies observed for the other two PLMs (Figure 1). In particular, the most potent of the three reported PLMs (i.e., JTH-NB72-35) possesses a P2′ naphthalene methylene substituent, which would more efficiently bring together, and engage in more favorable hydrophobic contacts with, the non-polar side-chains of BoNT/A LC residues Leu 256 and Phe 369 (versus JTH-NB72-39), as well as with the aliphatic portions of the side-chain of residue Asn 368 at the enzyme binding site. Finally, it is likely that the same hypothesis applies to JTH-NB72-38; however, while the JTH-NB72-38 P2′ Trp indole does provide a larger ring system for binding in the indicated BoNT/A LC hydrophobic pocket (see above), the polar pyrrole nitrogen atom is unable to engage in a favorable hydrogen bond with the backbone carbonyl of BoNT/A LC residue Glu 257, an interaction that was previously observed in the BoNT/A LC:I1 complex (Figure 9A) . Instead, modeling of a Trp side chain in the P2′ position of the JTH-NB72-39 PLM revealed that none of the rotamers available for the - CH2-indole side chain positions it so that it engages in a hydrogen bond with any residue in the enzyme (Figure 9B). Indeed, our analyses indicated that the polar nature of the JTH-NB72-38 pyrrole nitrogen results in unfavorable hydrophobic-polar clashes in the hydrophobic pocket indicated above. Hence, this analysis provides a rational basis for explaining the higher Ki compared of this PLM versus those of JTH-NB72-35, JTH-NB72-39, and I1.
The decreased inhibitory efficacies of all three JTH PLMs compared to Il also appears to be partially due to the fact that they all incorporate an Arg residue at the P1 versus the P1 – P2′ DNP-DAP component of I1. In particular, the rigidity of the DNP phenyl and the solvation of its two nitro functional groups may be necessary for stabilizing this PLM's helical structure upon binding in the enzyme's substrate cleft. This, in turn, would decrease I1′s binding entropy. By comparison, the flexible side-chain of the P1 Arg of the JTH PLMs does not provide the same helix-stabilizing character. Additionally, the JTH PLMs lack the DAB component found in I1. In the BoNT/A LC:I1 co-crystal structure, this cationic component engages in a hydrogen bond with the side-chain amide of BoNT/A LC residue Gln 162 . Comparatively, in the BoNT/A LC:JTH-NB72-39 co-crystal structure, the PLM's non-polar P4′ Ala residue cannot engage in such a favorable H-bond. Finally, the initial weak electron density observed for the C-terminal residues of the JTH PLMs indicates that these residues may be destabilizing the overall binding modes of these inhibitors via entropic contributions, and therefore, other components that engage in more definitive hydrophobic and/or polar contacts with the enzyme cleft would be necessary for improving future PLM potencies.
Based on available structural and mechanistic data, future PLM designs will focus on optimizing the P1 - P2′positions, while simultaneously introducing changes to terminal PLM residues/components to decrease their entropic contributions. Moreover, future designs will also incorporate peptidomimetic features that will increase the drug-like character of the PLMs.
The design and synthesis of three new PLM inhibitors, which are pivotal for guiding the development of peptidomimetics and SMNPIs, have been presented. In order to characterize the binding modes for the PLMs, a co-crystal structure of one, JTH-NB72-39, was determined, which possesses components that have been independently reported to directly interact with the active site of the BoNT/A LC , , . Based on comparisons between the binding mode determined for JTH-NB72-39 with other non-Zn-chelating BoNT/A LC:PLM co-crystal structures , , , a consistent inhibition mechanism has emerged . Specifically, we provide a unifying PLM-based mechanism of action: in all cases the presence of a P1 amino terminal residue is key for effectively ‘arresting’ the proteolytic activity of the BoNT/A LC. This discovery explains why SNAP-25 substrates N-terminally extended beyond the P1 position are cleaved by the BoNT/A LC; in such peptides, due to their participation in a peptide bond with the P2 residue, the P1 amino group becomes an amide, rendering it non-competent for ‘locking’ the Glu 224 carboxylate group . This observation emphasizes the requirement of a highly ionizable group in this position in order to strongly interact with Glu 224 – to effectively inhibit the BoNT/A LC protease and will be important for guiding the rational design of new PLMs and peptidomimetics, as well as for the discovery of new SMNPIs and the synthetic optimizations of existing SMNPIs.
The JTH-NB72-39-BoNT/A LC co-crystal structure presented here, and its comparison with all other SNAP-25-derived PLM inhibitor co-crystal structures known to date, highlights the importance of the BoNT/A LC 370 loop for substrate binding and cleavage specificity. It is this loop that contains the Asp 370 residue, which is pivotal for substrate discrimination. Also, BoNT/A LC residue Phe 369 is located in this region, and according to the structure presented here and the co-crystal structure with I1 , forms part of a hydrophobic pocket that efficiently anchors non-polar groups located in the P2′ position of these PLMs. Future designs will involve PLM components that can further stabilize a helical backbone orientation without interfering with binding, as well as the incorporation of bulkier non-polar components at the P2′ position.
N,N-Diisopropylethylamine was sequentially distilled from ninhydrin then KOH and stored under argon. Piperidine was distilled from CaH2 and stored under argon. Phenol was purified by dissolving the solid in diethyl ether, washing with a saturated aqueous solution of NaHCO3 (3x), extracting with aqueous NaOH (0.1 M) (3x), acidifying the aqueous extracts with 0.1 N HCl, extracting with Et2O (3x), concentrating the ethereal extracts under reduced pressure, and the dry solid was stored under argon. N,N-Dimethylformamide was purchased from Alfa Aesar as anhydrous and amine free in 4 L quantities and stored in 1 L Amber bottles (dried overnight in an oven at 140°C) over activated 4 Å molecular sieves under argon. Trifluoroacetic acid (biochemical grade, 99.5+% pure) was purchased from Alfa Aesar and used as received. Methanol (HPLC grade), water (HPLC grade), and thioanisole (99% purity) were purchased from Aldrich and used as received. Triisopropylsilane (99% purity) was purchased from Acros and used as received. 1,2-ethanedithiol (>98% pure) was purchased from Fluka and used as received. All natural Fmoc-protected amino acids and 3-(diethoxy-phosphoryloxy)-3H-benzo[d], , -triazin-4-one were purchased from either Peptides International or Advanced Automated Peptide Protein Technologies (AAPPTEC) and used as received. Unnatural Fmoc protected amino acids and Rink Amide Resin SS, 100–200 mesh, 1% DVB were purchased from Advanced Chemtech and used as received. BD Falcon BlueMax 50 mL graduated tubes and 25 mm syringe filters with a 0.45 µm nylon frit were purchased from Fischer Scientific.
The Fmoc-solid phase peptide syntheses were performed on a CEM Discover manual microwave peptide synthesizer fitted with a fiber-optic temperature probe. Solid phase peptide syntheses were performed in a 25 mL polypropylene reaction vessel. The 25 mL polypropylene reaction vessel was constructed by inserting a Teflon ring (0.4 mm height, 2.1 mm outer diameter, 1.8 mm inner diameter) into a capped 25 mL SPE reservoir purchased from Grace Davison Discovery Science (Catalogue #: 210425) containing a frit purchased from Grace Davison Discovery Science (Catalogue #: 211416) (Figure S1).
Preparative reverse phase HPLC purifications were performed on a Gilson HPLC system with 220 and 254 nm UV detection, using a Phenomenex Luna 5µ C18(2) 100 Å, AX (75×30.0 mm) column at a flow rate of 10 mL/min. Unless otherwise noted, all preparative runs used linear gradients of 30–60% buffer B in A (A: water containing 0.1% TFA, B: CH3CN containing 0.1% TFA) over 30 min. Analytical HPLC traces of final products were performed on a Gilson HPLC system with 220 and 254 nm UV detection, using a Varian Microsorb 100-3 C18 (100×4.6 mm) column at a flow rate of 0.7 mL/min. Unless otherwise noted, all analytical runs used linear gradients of 30–100% buffer B in A (A: water containing 0.1% TFA, B: MeOH) over 70 min. CD Spectra were recorded on a Jasco J-815 Circular Dichroism Spectrometer. Unless otherwise noted, all CD spectra were recorded in MeOH at a concentration of 0.5 mM, at 298 K, over a range of 300–200 nm, at a scan rate of 50 nm/min. Mass spectra were obtained using MALDI TOF/TOF with 2,5-dihydroxybenzoic acid as the matrix in the positive ion mode. Lyophilization was accomplished using a Labconco FreeZone 4.5 liter bench top freeze dry system. Centrifugation was accomplished using a Sorvall RT-7 Plus bench top centrifuge.
Proton and carbon NMR spectra were recorded using a Bruker Avance spectrometer at 600 MHz/150 MHz (1H NMR/13C NMR) in D2O (298 K), unless otherwise noted. Chemical shifts (δ) are reported in parts per million (ppm) using MeOH solvent peaks as an internal reference (referenced to 3.34 ppm (1H) and 49.5 ppm (13C)). 1H NMR data are reported as follows: chemical shift, multiplicity (s=singlet, d=doublet, t=triplet, q=quartet, m=multiplet, dd=doublet of doublets, dt=doublet of triplets, td=triplet of doublets, qd=quartet of doublets), coupling constants (J) in Hertz (Hz), and integration. 13C NMR spectra were obtained using a proton-decoupled pulse sequence with d1 of 6 sec, and are tabulated by observed peak.
A stock solution of the coupling base was prepared by dissolving DIPEA (1.74 mL, 1.00 mmol) in DMF (5.00 mL) to give a 0.148 M solution. A stock solution of the Fmoc-cleavage base was prepared by dissolving piperidine (1.00 mL, 10.1 mmol) in DMF (4.00 mL) to give a 2.02 M solution. A stock solution of the resin cleavage cocktail was prepared by combining TFA (5.07 g, 44.5 mmol), PhSCH3 (0.210 g, 1.69 mmol), PhOH (0.215 g, 2.28 mmol), TIPSH (0.0620 g, 0.392 mmol), 1,2-EDT (0.135 g, 1.43 mmol) and H2O (0.100 g, 5.56 mmol). All stock solutions were freshly prepared prior to use.
To a 25 mL polypropylene reaction vessel charged with a Teflon stir bar (10×3 mm) was added the Rink Amide resin (0.143 g, 0.100 mmol, loading 0.700 mmol/g, 1.00 equiv). The resin was washed with MeOH (2×5 mL), CH2Cl2 (3×10 mL) and DMF (3×10 mL), suspended in CH2Cl2 (5 mL) and allowed to swell at room temperature for 30 min. The resin was filtered and washed with DMF (3×10 mL). The Fmoc group was cleaved by heating the resin in the Fmoc-cleavage base stock solution (1 mL) in the microwave (35 W, 78°C, 3 min). The resin was filtered and washed with DMF (3×10 mL), CH2Cl2 (3×10 mL) and DMF (3×10 mL). The first Fmoc protected amino acid was coupled to the resin by heating in a pre-mixed solution of amino acid (0.350 mmol, 3.50 equiv), DEPBT (0.105 g, 0.350 mmol, 3.50 equiv), DMF (0.80 mL), and Fmoc-coupling base stock solution (0.75 mL) in the microwave (25 W, 80°C, 5 min). The resin was filtered and washed with DMF (3×10 mL), CH2Cl2 (3×10 mL) and DMF (3×10 mL). The Fmoc group was cleaved as previously described, and the next amino acid was coupled. This process of Fmoc cleavage and amino acid coupling was repeated for each additional amino acid. After the final Fmoc cleavage, the resin was washed with DMF (30 mL) and CH2Cl2 (20 mL). The protecting groups were cleaved by treatment of the dry resin with the resin cleavage cocktail stock solution (2.50 mL) for 2 h at room temperature with vigorous stirring. The resin was filtered and rinsed with the remaining resin cleavage cocktail stock solution (1.50 mL) and TFA (1.50 mL), collecting the filtrate and rinses in a 50 mL BD Falcon tube. The sample was concentrated to a heterogeneous mixture (approximately 0.2 mL) under a stream of argon for 30 min. Cold diethyl ether (45 mL) was added to precipitate the crude peptide. The sample was centrifuged (3200 rpm, −8°C, 15 min) and the supernatant was discarded. The crude peptide was transferred to a 20 mL scintillation vial with approximately 5 mL of a mixture of H2O/CH3CN (91) and lyophilized overnight. The crude peptide was dissolved in H2O containing 0.1% TFA (5.00 mL) and filtered through a 0.45 µm nylon syringe filter. The filtrate was purified by preparative RP HPLC.
Prepared according to general procedure A utilizing the following amino acid sequence: Fmoc-L-Leu-OH (0.124 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Met-OH (0.130 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Ala-OH (0.115 g, 0.350 mmol, 3.50 equiv), Fmoc-Aib-OH (0.114 g, 0.350 mmol, 3.50 equiv), Fmoc-L-1-Nal-OH (0.153 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Arg(Pbf)-OH (0.227 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Arg(Pbf)-OH (0.227 g, 0.350 mmol, 3.50 equiv). JTH-NB72-35 (0.0399 g, 40%) was obtained as a white powder: The product was characterized by 1H NMR (Table S2); 13C NMR (Table S2); DEPT-135; COSY; HMBC; HMQC; HPLC RT 5.7 min, HRMS (MALDI+) m/z calcd for C43H71N14O7S [M+H] 927.5351, Found 927.5355. Figures S2 and S3 provide the CD spectrum and HPLC trace, respectively, for this PLM.
Prepared according to general procedure A utilizing the following amino acid sequence: Fmoc-L-Leu-OH (0.124 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Met-OH (0.130 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Ala-OH (0.115 g, 0.350 mmol, 3.50 equiv), Fmoc-Aib-OH (0.114 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Trp(Boc)-OH (0.185 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Arg(Pbf)-OH (0.227 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Arg(Pbf)-OH (0.227 g, 0.350 mmol, 3.50 equiv). JTH-NB72-38 (0.0415 g, 41%) was obtained as a white powder: The product was characterized by 1H NMR (Table S3); 13C NMR (Table S3); DEPT-135; COSY; HMBC; HMQC; HPLC RT 5.7 min, HRMS (MALDI+) m/z calcd for C41H70N15O7S [M+H] 916.5303, Found 916.5461. Figures S4 and S5 provide the CD spectrum and HPLC trace, respectively, for this PLM.
Prepared according to general procedure A utilizing the following amino acid sequence: Fmoc-L-Leu-OH (0.124 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Met-OH (0.130 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Ala-OH (0.115 g, 0.350 mmol, 3.50 equiv), Fmoc-Aib-OH (0.114 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Phe-OH (0.136 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Arg(Pbf)-OH (0.227 g, 0.350 mmol, 3.50 equiv), Fmoc-L-Arg(Pbf)-OH (0.227 g, 0.350 mmol, 3.50 equiv). JTH-NB72-39 (0.0318 g, 34%) was obtained as a white powder: The product was characterized by 1H NMR (Table S4); 13C NMR (Table S4); DEPT-135; COSY; HMBC; HMQC; HPLC RT 5.6 min, HRMS (MALDI+) m/z calcd for C39H68N14NaO7S [M+Na] 899.5014, Found 899.5021. Figures S6 and S7 provide the CD spectrum and HPLC trace, respectively, for this PLM.
The HPLC-based assay used to calculate PLM inhibition constants has been published extensively , , , , , . In brief, the assay utilizes an N-terminal acetylated, C-terminal aminated, synthetic peptide identical in sequence to residues 187–203 of SNAP-25. Substrate hydrolysis is determined by HPLC separation of the products from the substrate, followed by measurement of the peak areas. Assay mixtures consisted of 40 mM HEPES–0.05% Tween (pH 7.3), recombinant BoNT/A LC, peptide substrate, 0.5 mg/ml Bovine Serum Albumin, and various PLM concentrations. Assays were run at 37°C, quenched by the addition of TFA, and analyzed by reverse-phase HPLC. To eliminate Zn chelating agents, the assay is run in the presence of excess Zn (50 µM). Ki values were calculated by measuring PLM mediated inhibition at different substrate concentrations and treating the kinetic data by the method of Dixon. Inhibition constants (i.e. Ki values) were extracted from the slopes of Dixon plots: Ki = Km/(slope x Vmax x S), where S is the substrate concentration. All reported values are averages of at least three independent determinations using nine PLM concentrations.
Details of the bacterial expression and purification of the active form of wt BoNT/A LC used in this study have been previously described. Stock solutions of wt BoNTA-LC containing 20 mM HEPES, pH 7.4 were adjusted to a final 150 µM protein concentration. Lyophilized JTH-NB72-35, -38, and -39 inhibitors were resuspended in distilled water to a final 10 mM concentration. Individual mixtures of BoNT/A LC and each of the three PLMs were prepared by mixing both stock solutions to attain a final 50 µM BoNT/A LC and 1 mM PLM concentrations.
Crystals were obtained by using the hanging drop vapor diffusion method at 20°C. Briefly, 3 µL of a mixture of 50 µM BoNT/A and 1 mM of each PLM inhibitor were mixed with 1.5 µL of the mother liquor containing 14% PEG MME 2000, 10 mM NiCl2, and 100 mM HEPES pH 8.5. A layer of a 11 mixture of paraffin:silicon oil was overlaid onto the mother liquor present in the well. Crystals for the BoNT/A LC:JTH-NB72-39 mixture appeared after approximately five days of incubation and they were directly transferred into a cryo-solution containing 25%(v/v) PEG600, 0.14 X PEG MME 2000, 10 mM NiCl2, and 100 mM HEPES pH 8.5, and then flash-frozen in liquid nitrogen. Only microcrystals were observed for the other two PLMs, and further manipulation did not result in any improvement of their size. The diffraction data were collected at beamline 11.1 of the SSRL (Stanford Synchrotron Radiation Laboratory) at a wavelength of 1 Å, and at a temperature of 100 °K. The diffracted crystals belonged to the P21212 group. Integration, indexing, and scaling of the diffraction data was performed using the HKL2000 suite of programs .
The coordinates in the 1XTF pdb file were used as the search model to determine the structure of the wt BoNT/A LC:JTH-NB72-39 complex by molecular replacement using the PHASER module in CCP4i . The initial values for the Rwork and Rfree of the generated model were 27.1% and 31.3%, respectively. The σA-weighted mFo-Fc electron density map clearly indicated the presence of JTH-NB72-39 in the vicinity of the active site (Fig. 2). The coordinates of the JTH-NB72-39 inhibitor were then added to those of the BoNT/A LC in the structure of the complex using Coot. Final refinement and modeling was performed using Phenix . The quality of the final structure was assessed using MolProbity. Ramachandran analysis showed that the BoNT/A LC:JTH-NB72-39 structure had 97.14% residues in the favored region with no outliers. The coordinates and structure factors have been deposited in the PDB (ID 3NF3).
Diagram of assembled 25 mL polypropylene reaction vessel.
(0.11 MB DOC)
CD spectrum of JTH-NB72-35 (0.5 mmol) in MeOH.
(0.38 MB DOC)
Analytical HPLC trace of JTH-NB72-35 using a linear gradient of 30–100% buffer B in A (A: water containing 0.1% TFA, B: MeOH) over 70 min with UV detection at 220 nm.
(0.14 MB DOC)
CD spectrum of JTH-NB72-38 (0.5 mmol) in MeOH.
(0.37 MB DOC)
Analytical HPLC trace of JTH-NB72-38 using a linear gradient of 30–100% buffer B in A (A: water containing 0.1% TFA, B: MeOH) over 70 min with UV detection at 220 nm at a flow rate of 0.7 mL/min.
(0.14 MB DOC)
CD spectrum of JTH-NB72-39 (0.5 mmol) in MeOH.
(0.37 MB DOC)
Analytical HPLC trace of JTH-NB72-39 using a linear gradient of 30–100% buffer B in A (A: water containing 0.1% TFA, B: MeOH) over 70 min with UV detection at 220 nm at a flow rate of 0.7 mL/min.
(0.14 MB DOC)
Potencies of structurally characterized BoNT/A LC inhibitors.
(0.07 MB DOC)
1H and 13C NMR Data for JTH-NB72-35 (Figure 1) (600 MHz/150 MHz) in D2O (298 K) with MeOH as an internal reference (referenced to 3.34 ppm (1H) and 49.5 ppm (13C)).
(0.13 MB DOC)
1H and 13C NMR Data for JTH-NB72-38 (Figure1) (600 MHz/150 MHz) in D2O (298 K) with MeOH as an internal reference (referenced to 3.34 ppm (1H) and 49.5 ppm (13C)).
(0.14 MB DOC)
We thank Dr. Yuan-Ping Pang for his useful comments on the geometry of the Zn(II) coordination, and Dr. Tim Fenn for discussions.
Competing Interests: The authors have declared that no competing interests exist.
Funding: This work was supported by the Howard Hughes Medical Institute and the Department of Defense (proposal number 3.10024_06_RD_B to A.T.B.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.