Botulinum Neurotoxin Is Shielded by NTNHA in an Interlocked Complex

See allHide authors and affiliations

Science  24 Feb 2012:
Vol. 335, Issue 6071, pp. 977-981
DOI: 10.1126/science.1214270


Botulinum neurotoxins (BoNTs) are highly poisonous substances that are also effective medicines. Accidental BoNT poisoning often occurs through ingestion of Clostridium botulinum–contaminated food. Here, we present the crystal structure of a BoNT in complex with a clostridial nontoxic nonhemagglutinin (NTNHA) protein at 2.7 angstroms. Biochemical and functional studies show that NTNHA provides large and multivalent binding interfaces to protect BoNT from gastrointestinal degradation. Moreover, the structure highlights key residues in BoNT that regulate complex assembly in a pH-dependent manner. Collectively, our findings define the molecular mechanisms by which NTNHA shields BoNT in the hostile gastrointestinal environment and releases it upon entry into the circulation. These results will assist in the design of small molecules for inhibiting oral BoNT intoxication and of delivery vehicles for oral administration of biologics.

The seven serotypes of botulinum neurotoxin (BoNT) (named A to G) are sequence-specific endopeptidases. They invade nerve cells at neuromuscular junctions, where they inhibit the release of the neurotransmitter acetylcholine by cleaving SNAREs (soluble N-ethylmaleimide–sensitive factor attachment protein receptors), and subsequently paralyze the affected muscles (1). Although BoNT poisoning often occurs through oral ingestion of tainted food products, the molecular mechanism by which BoNTs avoid destruction in the hostile environment of the gastrointestinal (GI) tract is unknown.

Naturally occurring BoNTs are protected within progenitor toxin complexes (PTCs), which are high molecular weight multiprotein complexes composed of the BoNT and several nontoxic neurotoxin-associated proteins (NAPs) (2). The NAPs include the 140-kD NTNHA (nontoxic nonhemagglutinin), which together with BoNT forms the minimally functional PTC (M-PTC), and three hemagglutinin (HA) proteins, which assemble with the M-PTC and have been suggested to facilitate BoNT transcytosis across the intestinal barrier (1, 3). The protecting function of the M-PTC is sufficiently effective that it reduces the oral median lethal dose (LD50) of BoNT 10- to 20-fold compared with free BoNT (4, 5). Intriguingly, the protected BoNT is released from the PTC upon transition from acidic to neutral pH, as occurs during absorption from the intestine into the bloodstream (6). Once in the circulation, free BoNT travels to neuromuscular junctions, where it invades neurons. The molecular architecture of PTC is largely unknown, with the exception of two low-resolution structures revealed by electron crystallography (30 Å) and electron microscopy (7, 8).

We focused our study on the M-PTC of serotype BoNT/A1 (referred to as BoNT/A) because it is a major cause of human botulism and is a concern for bioterrorism (9). We established a robust recombinant system in Escherichia coli to produce a genetically modified, catalytically inactive BoNT/A (~150 kD) that carries three mutations [Glu224→Gln224, Arg363→Ala363, and Tyr366→Phe366 (E224Q/R363A/Y366F)] in the catalytic site (10) and the corresponding full-length NTNHA-A1 (~140 kD, referred to as NTNHA-A) (11). BoNT/AE224Q/R363A/Y366F adopts a structure essentially identical to that of wild-type BoNT/A (12) and is referred to here as BoNT/Ai (fig. S1 and table S2). The free forms of BoNT/Ai and NTNHA-A are monomeric at pH = 6.0 and 7.5 (fig. S2) and assemble into a monomeric M-PTC at pH = 6.0 with a dissociation constant (Kd) of ~30.8 nM and 1:1 stoichiometry, as analyzed by analytic ultracentrifugation (AUC) and isothermal titration calorimetry (ITC) (fig. S2 and table S1). By contrast, no assembly occurs at pH = 7.5.

Free BoNT/Ai and NTNHA-A are inherently fragile, as shown by in vitro cleavage by digestive proteases such as trypsin and pepsin (Fig. 1A). Free wild-type BoNT/A is inactivated by trypsin or by incubation at pH = 3 or less, on the basis of ex vivo mice phrenic nerve hemidiaphragm (MPN) assays (Fig. 1B). In contrast, the structural integrity and activity of BoNT/A is protected in the in vitro–reconstituted M-PTC at low pH, but not at neutral or alkaline pH (Fig. 1, A and B). Thus, the recombinant M-PTC faithfully mimics the behavior of native M-PTC.

Fig. 1

The architecture of the M-PTC. (A) NTNHA-A protects BoNT/Ai against trypsin and pepsin digestion when they assemble into M-PTC in an acidic environment. (B) NTNHA-A protects active BoNT/A against low pH–mediated inactivation. It also protects BoNT/A against trypsin inactivation at pH = 6.0 but not at pH = 8.0. wt, wild type. (C) Cartoon presentation of the M-PTC. BoNT/Ai domains are blue (LC), orange (HN), and green (HC). NTNHA-A domains are yellow (nLC), cyan (nHN), and red (nHC). (D) Individual structures of BoNT/Ai and NTNHA-A in the M-PTC.

To understand the molecular mechanism underlying BoNT/A protection and the pH-dependent assembly of the M-PTC, we determined a 2.7 Å resolution crystal structure of the M-PTC facilitated by a BoNT/A-specific nanobody (13), as well as the structure of a nanobody-free M-PTC at 3.9 Å resolution (Fig. 1, C and D; figs. S3 and S4; and table S2). Nanobodies are the smallest antigen-binding fragments (~12 to 15 kD) of naturally occurring heavy-chain–only antibodies (VHH) present in camelids and have been exploited as inhibitors of BoNTs (13). The VHH binds on the surface of BoNT/Ai far from the NTNHA-A–interacting sites. It facilitates a more compact crystal packing and thus improves the diffraction quality of the crystals (fig. S5 and table S3). The two independently solved crystal structures of the M-PTC are indistinguishable, demonstrating that the structure presented here represents the physiological conformation of the M-PTC, independent of VHH or crystal packing (fig. S3B).

BoNT/Ai and NTNHA-A form an interlocked compact complex and bury an unusually large solvent-accessible area (~3600 Å2 per molecule) through multivalent interfaces (Fig. 1). BoNT/Ai is composed of a 50-kD light chain (LC, a Zn2+-endoprotease) and a 100-kD heavy chain (HC), which has two domains: The N-terminal domain (HN) mediates LC translocation across the endosomal membrane, whereas the C-terminal domain (HC) is the receptor-binding domain (1). Unexpectedly, the protective component NTNHA-A has a structure highly similar to BoNT/Ai despite a low sequence identity (~20%). NTNHA-A also displays three domains that, when compared pairwise with LC, HN, or HC, yield root mean square deviations of 2.2 Å (314 Cα pairs), 2.3 Å (300), and 1.9 Å (319), respectively, and thus are termed nLC, nHN, and nHC (“n” indicates NTNHA) (Fig. 1, C and D, and fig. S6). At the center of the complex is the HC fragment that is surrounded by all three domains of NTNHA-A (table S4). Notably, HC rotates about 140° around a linker connecting HN and HC, resulting in a distinct conformation in comparison to the free form of BoNT/A (Fig. 2) (12). Complementing this, HN also forms many polar interactions with nHN and nHC (Fig. 3, D and E). In contrast, LC does not bind NTNHA-A.

Fig. 2

The HC fragment of BoNT/A can rotate around a linker connecting HN and HC. (A) Superposition of the free and complex forms of BoNT/A based on Cα atoms in LC and HN. HC of free BoNT/A is violet; LC and HN are omitted for clarity. Linkers in the free and complex forms of BoNT/A are gray and red, respectively. (B) Cartoon model showing two distinct conformations of BoNT/A that bring the receptor-binding site in HC (red star) close to the opposite tip of the long helical HN.

Fig. 3

M-PTC is stabilized by extensive intermolecular interactions. (A to C) HC interacts with all three domains of NTNHA-A. The overall structure of the M-PTC is shown in (B), where NTNHA-A is in surface representation (gray) and BoNT/Ai is blue, orange, and green for LC, HN, and HC, respectively. Open-book views of the interfaces highlighted in the red and blue boxes are shown in (A) and (C), respectively. Residues in HC that form hydrogen bonds or salt bridges with nLC, nHN, or nHC are yellow, cyan, and red, respectively. C, Cys; I, Ile; V, Val. (D) Open-book view of the interface highlighted in the red box of (E). HN directly contacts nHC (interacting residues shown in red) and nHN (cyan), but not nLC. (F) Precipitation assays between full-length NTNHA-A (prey) and HC variants (baits). Bar graph shows mean ± SD, n = 3 (n is the number of assays). Binding of samples 8 to 10 at pH = 7.5 was not detectable.

Despite their similarities, a structural comparison between NTNHA-A and BoNT/Ai revealed several features specific to NTNHA-A (fig. S6). NTNHA-A does not possess most of the characteristic structural features of BoNT/A that are crucial to its functions: (i) nLC does not have the catalytic zinc-binding HExxH+E (H and x indicate His and any amino acid, respectively) signature motif that is conserved among all BoNT serotypes; (ii) NTNHA-A does not have the homologous long loop (Val431 to Leu453) that in BoNT/A connects LC and HC and is cleaved posttranslationally to activate BoNT/A; (iii) a functionally essential disulfide bond (Cys430–Cys454 in BoNT/A) that is involved in translocation of LC and is conserved in all BoNTs is not present in NTNHA-A (14); (iv) nHC maintains the core structure of HC but has deletions in many surface areas that are not directly involved in BoNT/Ai binding; and (v) none of the ganglioside-binding residues of HC, including the highly conserved ganglioside-binding motif (E+H+SxWY) (S and W indicate Ser and Trp, respectively) found in many BoNT serotypes, is conserved in NTNHA-A (15, 16).

NTNHA-A has a large insert in nLC (nGly116 to nAla148, termed the nLoop) that is not present in LC. Furthermore, the nLoop is conserved in NTNHA-A1, B, C, D, and G but is missing in NTNHA-A2, E, and F (fig. S7). Notably, many of the known spontaneous nicking sites in NTNHA are located in the nLoop (17), and, consistent with this, we observed that recombinant NTNHA-A was nicked between nLys133 and nLys134 during long-term storage. The nicking sites in NTNHA are masked in the HA-bound PTC, and the M-PTC that contains the nicked NTNHA can no longer assemble with HAs (17, 18). Moreover, NTNHA-A2, E, and F, which lack the nLoop, do not have accompanying HA proteins and only form the HA-negative M-PTC (1921). The crystal structure reveals that the nLoop is fully exposed on the M-PTC surface and has no visible electron density, presumably because of its high flexibility (Fig. 1C). Collectively, these data suggest that the nLoop could participate in the interaction with HAs to assemble the larger PTCs.

The conformation of BoNT/Ai in the M-PTC brings the receptor-binding site located in the C-terminal subdomain of HC close to the C-terminal boundary of the long rodlike HN (Fig. 2). This conformation contrasts with crystal structures of the free BoNT/Ai (fig. S1), free BoNT/A, and free BoNT/B serotype, where the receptor-binding site in HC points to the N-terminal boundary of HN (12, 2224). The HC reorientation is mediated by a linker between HN and HC (Leu845 to Thr876 in BoNT/A, referred to as the HN-HC linker), which adopts an essentially identical structure in free BoNT/A and BoNT/B but changes its conformation in the M-PTC (Fig. 2 and fig. S8). The conformational change is likely induced by NTNHA-A rather than pH because the same conformation is adopted by all structures of free BoNT/A or BoNT/B crystallized at pHs ranging from 5.0 to 7.0 (12, 2224).

We speculated that the flexible HN-HC linker may play a role in coordinating HC-mediated receptor binding and HN-mediated translocation, given that the membrane-anchored receptors impose some geometric restrictions on the position of HC with respect to the membrane surface (24, 25) and that the long helical HN needs to strategically orientate on the membrane to achieve efficient translocation of LC to the cytosol (2628). To test this, we produced two mutants of BoNT/A in which the structure of the HN-HC linker was altered by point mutations. BoNT/AR861A/E868P/K871P, containing the helix breaker proline (P) and where K indicates Lys, is expected to destabilize the helical linker, whereas BoNT/AL862K/L863E/T867K, with extra ion pairs and T indicating Thr, will likely stabilize the helical structure. The two BoNT/A mutants showed wild-type–like characteristics with respect to protein folding, receptor binding capability, Zn2+-endoprotease activity, and NTNHA-A binding (fig. S9). However, the toxicity of both mutants was significantly decreased by ~sixfold in the MPN assay. Although the exact conformations of the two mutants are not known, we conclude that the observed decrease in toxicity is due to the disrupted coordination between HN and HC, leading to decreased efficiency of LC translocation. These data further suggest that any disturbance in the structural integrity of the HN-HC linker may affect the optimal positioning of HC and HN during BoNT intoxication. Thus, the structural features of the HN-HC linker might represent a serotype-specific signature for the various BoNTs. For example, BoNT/E serotype has a more rapid LC translocation, which has been suggested to result from its intrinsically more flexible linker (fig. S8) (27, 29).

To validate the physiological relevance of the M-PTC structure, we performed systematic truncation studies. The isolated HC fragment bound to NTNHA-A with high affinity at pH = 6.0 (Kd ~ 48.3 nM) but not at pH = 7.5 and thus largely replicated the binding behavior of full-length BoNT/Ai. Consistent with this, the HC-deleted BoNT/Ai (LHN) no longer bound NTNHA-A (fig. S10). BoNT/Ai has a larger unfavorable binding entropy than HC (ΔΔS ~ –16.8 cal mol−1 K−1) (table S1), which could be due to a loss of conformational entropy of BoNT/Ai upon NTNHA-A binding and may be related to the reorientation of HC. Truncating nHC or the C-terminal subdomain of nHC (nHCC) in NTNHA-A abolished the protective function of NTNHA-A at pH = 2, suggesting that nHC is crucial for shielding the sensitive HC of BoNT/A (Fig. 1B).

Additional support for the M-PTC structure is provided by structure-based mutagenesis studies. Interactions between HC and NTNHA-A were gradually weakened as more intra-PTC charge-charge interactions were disrupted: The binding affinity at pH = 6.0 decreased in the order of HC-K1000A > HC-K1000A/K1039A > HC-K1000A/K1039A/K1121A (Fig. 3F and fig. S10). Even a single point mutation, HC-K1000A, significantly decreased the binding affinity by ~sevenfold (Kd ~ 337.5 nM). The HC-centered protection by NTNHA-A is biologically relevant. Although the integrity of HC is crucial to enrich BoNT/A on the neuron surface at the early stage of intoxication, it is much more sensitive to proteolysis than LC and HN in the free BoNT/A (30, 31).

To understand how a pH change can trigger the disassembly of the PTC, we focused on HC because it largely replicates the binding features of full-length BoNT/Ai. We specifically concentrated on residues located in the complex interface that are titratable in an acidic environment (e.g., His, Glu, and Asp). We found four acidic residues in HC, each of which is located in the vicinity of an intra-PTC salt bridge or a long-range electrostatic interaction: Glu982 is near Lys1000–nAsp808; Asp1037, Lys1039–nGlu810; Asp1118, Lys1121–nAsp455; and Asp1171, Arg1175–nAsp1110 (Fig. 4 and fig. S11). Our mutagenesis studies show that these charge-charge interactions are crucial for the assembly of the M-PTC (Fig. 3F). Analyzing the surface electrostatic potentials shows that BoNT/Ai is generally positively charged around these acidic residues, whereas the opposing NTNHA-A surface is largely negatively charged (fig. S11). Therefore, negatively charged BoNT/Ai residues in these areas pose an unfavorable force, which could potentially weaken the local electrostatic interactions in a pH-dependent manner.

Fig. 4

A pH-sensing mechanism. A close-up view of pH-sensing residues Glu982 and Asp1037, which are buried in the M-PTC in the vicinity of intra-PTC charge-charge interactions. Key residues in the interface are shown as ball-and-stick models. Hydrogen bonds and charge-charge interactions are indicated by dotted lines. A simulated-annealing omit map of the M-PTC contoured at 1 σ was overlaid with the final refined model.

To address the potential roles of these acidic residues in pH sensing, we mutated them individually to nontitratable residues. Single point mutants of the HC fragment (E982A, D1037A, D1118A, or D1171A, where D indicates Asp) all folded correctly (fig. S12) and bound NTNHA-A at pH = 6.0 to a similar extent as the wild-type HC. HC-E982A and HC-D1037A showed significant binding to NTNHA-A at pH = 7.5 (Fig. 3F and fig. S10). Thermodynamic studies showed that HC-E982A binds to NTNHA-A at pH = 6.0 similarly to the way the wild-type HC does, predominantly driven by enthalpy (Kd ~ 39.4 nM); it also binds to NTNHA-A at pH = 7.5, driven by both enthalpy and entropy (Kd ~ 279.5 nM) (table S1). Furthermore, a single mutation of E982A significantly increased binding between the full-length BoNT/Ai and NTNHA-A at pH = 7.5, as measured in a precipitation assay (~20% binding by mutant versus <5% by the wild-type protein). These data suggest that Glu982 and Asp1037 are important pH-sensing residues that modulate the pH-dependent assembly of the M-PTC.

The pKa (where Ka is the acid dissociation constant) of the ionizable groups in proteins may be substantially shifted from the intrinsic pKa depending on the microenvironment surrounding these groups (32). Glu982 and Asp1037 of BoNT/A are predicted (33) to have clearly increased pKa values upon binding to NTNHA-A, partly because of desolvation effect when they are buried in the context of the M-PTC. Thus, Glu982 and Asp1037 are likely protonated at pH = 6.0 as their pKa values increase during the BoNT/A–NTNHA-A interaction, leading to a stable M-PTC. In contrast, these residues are deprotonated in a neutral or alkaline environment, generating repulsive charge interactions with NTNHA-A to destabilize the M-PTC assembly. We replaced Glu982 and Asp1037 with the isosteric, nontitratable residues Gln and Asn (N), respectively, to mimic their protonated state. Both HC-E982Q and HC-D1037N showed significant binding with NTNHA-A at pH = 7.5, further supporting the pH-sensing role of these two residues (Fig. 3F and fig. S10). At the same time, thermodynamic studies revealed a loss of enthalpy on binding of HC-E982A to NTNHA-A at pH = 7.5 compared to pH = 6.0 (ΔΔH ~6.3 kcal mol−1) (table S1), suggesting that the protein-protein interaction network between HC and NTNHA-A is partly impaired at pH = 7.5. Thus, additional pH-sensing components might exist in HC that become fully engaged in binding with NTNHA-A only at pH = 6.0.

Collectively, these data suggest that the assembly of the M-PTC is dynamically regulated by key pH-sensing residues that switch protonation states in response to the environmental pH. These interaction sites would likely be the most malleable parts of the PTC and could be specifically targeted for the development of small-molecule inhibitors to break up the PTC in the acidic GI tract. A pharmacological approach such as this, which focuses on preventive countermeasures at the earliest stage of BoNT intoxication, would complement current efforts to design inhibitors that block the subsequent neuronal action of BoNT. Our results also suggest mechanisms for the development of an oral drug delivery system in which proteinaceous drugs could be conjugated to a BoNT fragment and protected from degradation with NTNHA.

Supporting Online Material

Materials and Methods

Figs. S1 to S13

Tables S1 to S4

References (3451)

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We thank A. Brunger and T. Binz for critical reading of the manuscript; the staff of beamline 9-2 at the Stanford Synchrotron Radiation Lightsource (SSRL) and the NE-CAT staff of the Advanced Photon Source (APS), particularly K. Rajashankar, for assistance in data collection; G. Yao, N. Krez, A. M. Kruel, and J. Tremblay for excellent technical assistance; and R. Liddington and A. Bobkov for assistance with ITC and AUC. This work was partly supported by an Alfred P. Sloan Research Fellowship (R.J.), by the Deutsche Forschungsgemeinschaft (DFG Exzellenzinitiative GSC 108 to S.R.), by the Robert-Koch-Institut (1362/I-979 to A.R.), and by grants from the National Institute of Allergy and Infectious Diseases (NIAID), NIH, Department of Health and Human Services (HHS), under award number U54 AI057159 (C.B.S.). Atomic coordinates and structure factors for the VHH-bound M-PTC, M-PTC, and BoNT/Ai have been deposited with the Protein Data Bank under accession codes 3V0A, 3V0B, and 3V0C, respectively. Sanford-Burnham Medical Research Institute has a pending patent application, titled “Botulinum neurotoxin protective complex delivery compositions,” that was filed in November of 2011. BoNT availability is subject to the restrictions that apply to HHS select agents and NIAID Category A pathogens.
View Abstract

Navigate This Article