Nidogens are therapeutic targets for the prevention of tetanus

See allHide authors and affiliations

Science  28 Nov 2014:
Vol. 346, Issue 6213, pp. 1118-1123
DOI: 10.1126/science.1258138


Tetanus neurotoxin (TeNT) is among the most poisonous substances on Earth and a major cause of neonatal death in nonvaccinated areas. TeNT targets the neuromuscular junction (NMJ) with high affinity, yet the nature of the TeNT receptor complex remains unknown. Here, we show that the presence of nidogens (also known as entactins) at the NMJ is the main determinant for TeNT binding. Inhibition of the TeNT-nidogen interaction by using small nidogen-derived peptides or genetic ablation of nidogens prevented the binding of TeNT to neurons and protected mice from TeNT-induced spastic paralysis. Our findings demonstrate the direct involvement of an extracellular matrix protein as a receptor for TeNT at the NMJ, paving the way for the development of therapeutics for the prevention of tetanus by targeting this protein-protein interaction.

A potential peptide to prevent tetanus?

Tetanus (TeNT) and botulinum (BoNT) neurotoxins represent a family of powerful bacterial protein toxins that cause tetanus and botulism in humans and animals. The molecular mechanisms responsible for the entry and axonal retrograde transport of these toxins have been the subject of intense research. However, tetanus and botulism remain incurable, at least in part because of their high-affinity binding to synapses. Although the receptors for BoNT have recently been characterized at the molecular level, no protein receptor for TeNT at the neuromuscular junction has been identified. Bercsenyi et al. now suggest that TeNT exploits nidogen-1 and -2 for its binding to motor neurons. This binding is required for TeNT's internalization and axonal retrograde transport. Nidogens are extracellular matrix proteins that engage in multiple protein-protein interactions essential for the integrity of several tissues, including the nervous system. Interfering with the interaction between nidogens and TeNT by administering short nidogen-derived peptides blocked toxin binding to the neuromuscular junction and protected mice from tetanus.

Science, this issue p. 1118

Tetanus neurotoxin (TeNT) is composed of two subunits, which perform specific functions necessary for targeting this toxin to the central nervous system (CNS) and its high potency. The heavy (H) chain mediates high-affinity binding and entry into neurons, whereas the light (L) chain causes synaptic silencing of inhibitory interneurons, which normally suppress motor neuron activity, and thereby induces spastic paralysis (1). After internalization into motor neurons, TeNT and the carboxyl-terminal fragment of its H chain (HCT) are sorted to signaling endosomes, which undergo axonal retrograde transport toward the motor neuron cell body (2). We have previously characterized the proteome of these organelles using HCT-conjugated magnetic nanoparticles (2). This affinity purification approach identified a list of proteins associated with signaling endosomes, some of which were predicted to be involved in the binding and/or internalization of HCT. We analyzed the sequences of these proteins for the presence of the tripeptide Tyr-Glu-Trp (YEW), which was previously shown to interact with the sialic acid binding site (R site) of HCT, a region implicated in polysialoganglioside and acidic lipid binding (3, 4). The corresponding region of botulinum neurotoxins (BoNTs) interacts with the synaptic proteins SV2A-C or synaptotagmin I/II (511), raising the possibility that the R site of TeNT also binds to a protein receptor.

To identify putative TeNT receptors, we selected candidates containing variants of the YEW peptide, where this motif is present within the lumen of signaling endosomes, a localization topologically equivalent to the extracellular domain of these proteins when localized to the plasma membrane (table S1). A total of 35 nine-residue peptides containing the YEW motif or closely related sequences were tested for binding to HCT by using direct fluorescent binding, enzyme-linked immunosorbent assay (ELISA), and peptide pull-down assays (Fig. 1A and fig. S1, A and B). The interacting peptides were then assessed for their ability to compete for HCT binding on primary motor neurons (fig. S1C). Peptides N1 and N2, which correspond to short sequences in nidogen-1 and nidogen-2, respectively (table S1 and fig. S1D), were selected for further investigation because they significantly reduced HCT binding to neurons (Fig. 1, C and D, and fig. S1C). The YEW-like motif of nidogen-1 (YQW) and nidogen-2 (WSY) and their flanking residues were responsible for this interaction because single alanine mutants showed reduced binding to HCT (fig. S1, E and F), and the double alanine mutant (N1AA) (table S1) failed to block the binding of HCT to motor neurons (Fig. 1, C and D). Furthermore, when the N1 peptide, but not N1AA, was added to motor neurons, either together with HCT or 10 min afterwards, it still significantly inhibited HCT binding to these neurons (Fig. 1, E and F).

Fig. 1 Nidogen peptides bind to HCT and prevent its interaction with full-length nidogens in vitro.

(A) Assay for peptides that bind fluorescent HCT. The percentage of maximal binding was plotted for each peptide. The results were tested for significance between individual peptides and Tris-buffered saline (TBS) control (n = 3 independent experiments; *P < 0.05; error bar, SD). (B) Pull-down of HA-HCT preincubated with full-length recombinant nidogen-1 (top) or nidogen-2 (bottom). “Empty beads” corresponds to samples in which HA-HCT was omitted, and “DMSO” (dimethyl sulfoxide) to samples treated with vehicle control. Both nidogen-1 and -2 bind directly to HA-HCT, and pretreatment with the N1 or N2 peptide blocks this interaction. In contrast, mutant peptides (N1AA and N2AA) are ineffective in preventing this interaction (n = 3 independent experiments for both proteins). The bands detected in the empty bead controls are due to the heavy chain of the capture antibody, which comigrates with HCT. (C and E) Vehicle control (HCT alone), N1, or N1AA peptides were (C) preincubated, (E) added together (0 min), or added after a 10-min delay (10 min) to primary mouse motor neurons treated with AlexaFluor555-HCT in the absence or the presence of exogenous nidogen-1 (C) before fixing, immunostaining for βIII tubulin, and imaging. Scale bar, (C) 50 μm; (E) 10 μm. (D and F) Quantification of the data shown in (C) and (E). Twenty images (D) or three-tile scans of 4 by 4 fields (F) were taken for each condition, and fluorescence intensity of AlexaFluor555-HCT bound to neurons as defined by a βIII tubulin mask was quantified. Results were tested for statistical significance (n = 3 independent experiments; NS, nonsignificant; **P < 0.01; ***P < 0.005; ****P < 0.001; error bar, SD).

We then tested whether HCT interacts directly with full-length nidogens by incubating immobilized hemagglutinin (HA)–tagged HCT with full-length recombinant nidogen-1 or nidogen-2 and demonstrated that both proteins directly bind to HCT (Fig. 1B). The N1 and N2 peptide sequences of nidogen-1 and -2 played a key role in this interaction because preincubating HCT with an excess of these peptides, but not their respective mutants (table S1), abolished this binding (Fig. 1B).

The two mammalian nidogens play partially overlapping roles in basement membrane (BM) formation and maintenance (1214). Although nidogen-2 knockout (KO) mice do not display any overt phenotype, animals lacking nidogen-1 develop a progressive hindlimb paralysis, indicating a role for this protein in motor neuron survival (15). Nidogens are tightly integrated within the BM, which renders them difficult to access for exogenous ligands. To determine whether the amount of cell surface–associated nidogens is the limiting factor for HCT binding to motor neurons, we added soluble full-length nidogen-1 together with HCT to cultured motor neurons. This resulted in the binding of exogenous nidogen-1 to motor neurons (fig. S2) and augmented the amount of cell surface–bound HCT (Fig. 1, C and D), suggesting that the addition of exogenous nidogen-1 increased the number of HCT binding sites on the neuronal surface. This effect was dependent on the YEW-like motif of nidogen-1 because preincubation of HCT with the N1 but not the N1AA peptide prevented the increased cell surface binding of HCT (Fig. 1, C and D). Coimmunoprecipitation of HCT and endogenous nidogen-2 from motor neurons incubated with HCT for 5 min at 37°C (fig. S3) suggested that the interaction between nidogens and HCT is preserved upon entry of HCT into endosomal compartments.

Inspection of the domain structure of nidogens revealed that the N1 and N2 peptides reside within the globular G2 domain (fig. S4A and movie S1). Molecular modeling (fig. S4, B to D) revealed that the N1 peptide, both in isolation and within the intact G2 domain, is able to dock to the R site of HCT, where sialic acid, disialyllactose, and the YEW motif bind (4). The molecular surface concealed by the N1 peptide-HCT complex constitutes more than 40% of the total peptide surface (fig. S4C), suggesting that this interaction, which is driven by the potential formation of five hydrogen bonds and other polar interactions (table S2), is very strong.

The G2 domain of nidogen-1 fits into a large crevice at the top of the trefoil domain of HCT (fig. S4B and movie S1), concealing 1114 Å2 and 1106 Å2 of the molecular surface of HCT and the G2 domain of nidogen-1, respectively. Complex formation is likely to be promoted by complementary electrostatic surfaces and several potential hydrogen bonds (table S3 and fig. S4D). To confirm the validity of this structural model, we mutated several residues of the R site of HCT that are predicted to play an important role in the formation of this complex. As shown in fig. S5, we found that mutations of R1226, T1146, and Y1229, which form potential H-bonds with nidogen-1 (tables S2 and S3, respectively), as well as of other R site residues (P1212 and G1215) (16), reduced the binding of both N1 and N2 peptides. Similarly, alanine scanning of the YEW-like motif and flanking residues in the N1 and N2 peptides strongly decreased their interaction with HCT (fig. S1, D to F), which strongly supported the validity of the model shown in fig. S4. This mode of interaction is likely to be conserved through evolution because the human N1 and N2 peptides show a high degree of homology to the corresponding mouse sequences and also bound HCT (fig. S1D).

Crucially, this model predicts that the association between HCT and nidogen-1 is unlikely to interfere with the interaction between HCT and polysialogangliosides, which is required for TeNT binding to the neuronal surface (3, 17). Thus, nidogens and polysialogangliosides have the potential to bind TeNT simultaneously and thereby fulfill the requirements for the proposed dual receptor model (18, 19), which predicts the concurrent interaction of lipid and protein receptors with TeNT and BoNTs for their high-affinity binding to neuronal membranes.

Nidogen-2 is highly enriched at the NMJ (20), and we sought to investigate whether it is localized to the specialized presynaptic BM of this structure. We thus injected the lower hindlimb of wild-type mice with HCT and then immunostained cryosections of the extensor digitorum longus (EDL) muscle for choline acetyltransferase (ChAT), a motor neuron marker, and nidogen-2. This analysis revealed that HCT binds to the presynaptic motor neuron nerve terminal, which is in close apposition to the postsynaptic acetylcholine receptor, labeled by α-bungarotoxin (α-BTx) (Fig. 2A). Increased HCT binding was detected in areas where nidogen-2 was most abundant (Fig. 2B and movie S2)and was absent from NMJs that did not express nidogen-2 which revealed a strong correlation between the nidogen-2 content and HCT binding at the NMJ (Fig. 2C). We then assessed whether these two proteins are internalized and transported together in signaling endosomes. HCT and an antibody against nidogen-2 were added to primary motor neurons at 37°C and allowed to internalize. After a mild acid wash to remove all probes still bound to the neuronal surface, the internalized α-nidogen-2 was detected by using a fluorescently tagged secondary antibody. The extensive colocalization between α-nidogen-2 and HCT throughout the axonal network and cell bodies (66%) strongly indicated that these two proteins were internalized together and cotransported by the same organelles (Fig. 2D).

Fig. 2 The nidogen-2 content of the neuromuscular junction correlates with HCT binding and uptake.

(A and B) The anterior part of the lower hindlimb of adult mice was injected with HA-HCT. After 30 min, animals were transcardially perfused; the muscles were dissected, post-fixed, sectioned, and stained for HA-HCT, choline acetyltransferase (ChAT), or nidogen-2 and for AlexaFluor555-α-bungarotoxin (α-BTx, in gray) to label NMJs. HCT (red) binds to the presynaptic site of the NMJ, where ChAT (green) is localized (A). Bright HCT puncta are found in regions where nidogen-2 (green) accumulates (B). Scale bar, 10 μm. (C) HCT and nidogen-2 fluorescence intensities were quantified in 50 NMJs. There is a significant correlation between the integrated signal intensity of both probes (P < 0.001, Spearman coefficient 0.923). (D) Primary motor neurons were incubated with AlexaFluor555-HCT and a rabbit antibody to nidogen-2 (for 45 min at 37°C), acid-washed, fixed, and stained with an antibody to rabbit immunoglobulin G. HCT and nidogen-2 colocalization was quantified, and the corresponding Mander’s coefficient is shown. Scale bar, 10 μm (n = 3 independent experiments).

To further test the importance of nidogens in TeNT intoxication, we determined the susceptibility of nidogen KO mice to TeNT-induced paralysis and assessed the ability of HCT to bind neurons and tissues derived from these mice. Compared with wild-type motor neurons, nidogen-2 KO motor neurons exhibited a significantly reduced ability to bind HCT in vitro (Fig. 3, A and B). This was further verified in vivo by demonstrating that NMJ uptake of HCT was severely impaired in levator auris longus (LAL) muscles isolated from both nidogen-1 and nidogen-2 KO mice (Fig. 3, C and D). In stark contrast, the capacity of the corresponding binding fragment of BoNT/A (HCA) to bind to nidogen null NMJ was largely unaffected (fig. S6A). Crucially, the addition of full-length recombinant nidogen-1 restored the ability of HCT to bind NMJs lacking nidogen-1 (fig. S6B). This rescue was determined by the recruitment of recombinant nidogen-1 to mutant NMJs (fig. S6B), suggesting a replenishment of HCT acceptor sites at the synaptic cleft.

Fig. 3 Neurons lacking nidogens display reduced binding to HCT.

(A and B) Motor neurons derived from E13.5 wild-type and nidogen-2 knockout (KO) mouse embryos were incubated with HA-HCT (gray). Cells were fixed and stained for HA-HCT, and fluorescence intensity was quantified in 14 images by using a βIII tubulin mask (green). Scale bar, 20 μm. The value of nontreated cells was subtracted, and data were analyzed [(B), n = 3 experiments—shown is a representative experiment; ****P < 0.001; error bar, SD]. (C) LAL muscles of wild-type, and nidogen-1 and -2 KO animals were incubated with HA-HCT, fixed, and stained for HA (gray) and α-BTx (green) to identify NMJs. Scale bar, 5 μm. (D) Both nidogen-1 and -2 KO NMJs show a significant decrease in their ability to bind and internalize HCT (n = 3 animals per group, 20 images per animal; ****P < 0.001; error bar, SD). (E) Nidogen-1 and -2 DKO and wild-type E11.5 embryonic hindbrains were tested for HCT binding and uptake. Scale bar, 1 mm. (F) Mean intensity of HCT staining was measured and analyzed for significance (n = 3 animals per group; *P < 0.05; error bar, SD).

In the genetic background used in our experiment, nidogen-1 and -2 double knockout (DKO) mice show early embryonic lethality (21), precluding the derivation of motor neuron cultures. To overcome this limitation, we assayed HCT uptake into isolated hindbrains of DKO embryos taken at embryonic day 11.5 (E11.5). In contrast to control hindbrains, which displayed robust HCT staining, hindbrains from DKO littermates were unable to bind HCT (Fig. 3E), further indicating that nidogen-1 and -2 are essential for HCT binding in vivo (Fig. 3F).

Because the N1 peptide inhibited HCT binding to NMJs in whole LAL muscle preparations (Fig. 4A), we investigated the ability of the N1 peptide to prevent uptake of TeNT in vivo, using three independent assays. First, full-length TeNT was injected locally into the triceps surae muscle either alone or in combination with the N1 or N1AA peptides, and the effect on spastic paralysis was assessed in vivo by means of footprint analysis 24 hours after injection (Fig. 4, B and C). Administration of TeNT alone caused a severe gait abnormalities, with the characteristic inability to place the affected hindpaw in the former position of the ipsilateral forepaw, as occurs in normal mice (Fig. 4, B and C). At a later stage, TeNT caused a permanent plantar-flexion of the affected hindpaw (movie S3). Coadministration of the N1 peptide completely abolished these TeNT-induced gait abnormalities (movie S3), whereas the N1AA peptide was ineffective (Fig. 4, B and C).

Fig. 4 The N1 peptide blocks tetanic paralysis by inhibiting TeNT binding at the NMJ.

(A) LAL muscles were incubated with HA-HCT, which had been pretreated either with vehicle control (HCT alone), the N1 peptide, or the N1AA mutant in motor neuron culture medium. Muscles were then immunostained for HA and α-BTx to identify NMJs. The ratio of HCT-positive to HcT-negative NMJs was quantified and analyzed for significance (n = 3 independent experiments, six images per condition; NS, nonsignificant; **P < 0.01; error bar, SD). (B) Footprint analysis of control mice, or of mice after injection into the triceps surae muscle of a sublethal dose of TeNT, or TeNT preincubated with the N1AA or N1 peptides. Black circles show the paw prints from the noninjected left hindlimb and right frontlimb, and red circles mark the paw print of the TeNT-injected right hindlimb. In control and TeNT+N1-injected mice, the black and red circles overlap, whereas in TeNT- and TeNT+N1AA-injected mice, they are separated, indicating impairment in coordination due to local tetanus. (C) The distance between the paw prints on the injected side was measured and tested for statistical significance (n = 2 independent experiments; data from two animals are shown; NS, nonsignificant; *p < 0.05; **p < 0.01; error bar, SD). (D) The TA muscle was injected with TeNT preincubated with either DMSO (vehicle control), N1AA, or N1 peptides. After anesthesia, the tendon of this muscle was attached to a tension sensor, and the peroneal nerve was stimulated. Maximal muscle force was measured, normalized to muscle weight, and tested for statistical significance in control versus injected muscle. The N1-treated group retained a significantly higher muscle force than that of the vehicle- or N1AA-treated groups. (n = 6 mice per group; **P < 0.01; error bar, SEM). (E) Age- and weight-matched nidogen-2 KO and wild-type mice were injected intraperitoneally with TeNT and monitored for up to 96 hours. The time of termination due to loss of righting reflex was plotted against the TeNT dose. Generalized tetanus in nidogen-2 KO animals was delayed at lower doses (1.5 ng/kg) compared with that in wild-type controls (n = 2 experiments, minimum two mice per group; **P < 0.01; error bar, SD).

To corroborate this finding, we quantified the protective effect of the N1 peptide by co-injecting it with TeNT and measuring the maximal force generated by the tibialis anterior (TA) muscle using in vivo isometric muscle tension physiology. The contractile force of the TeNT-injected muscle was diminished (3.11 ± 0.88% of noninjected control) (Fig. 4D and fig. S7A), as was the case for muscles co-injected with TeNT and the N1AA peptide (2.14 ± 1.44% of noninjected control) (Fig. 4D and fig. S7B). In contrast, coadministration of the N1 peptide significantly abrogated the TeNT-mediated decline in contractile force compared with that of the other two groups (47.46 ± 10.31% of noninjected control) (Fig. 4D and fig. S7C). Similar results were obtained by using phrenic nerve-hemidiaphragm preparations (fig. S7D) (22). Together, these results confirmed the ability of the N1 peptide to prevent symptoms of tetanus in vivo.

In order to assess the susceptibility of nidogen-2 KO mice to systemic TeNT toxicity in vivo, we injected wild-type and nidogen-2 KO mice intraperitoneally with different doses of TeNT and monitored the appearance of tetanus. Nidogen-2 KO animals showed a significant delay in the onset and progression of tetanic symptoms at lower TeNT doses (Fig. 4E). Unfortunately, the possibility to extend these assays to mice lacking nidogen-1 was precluded by the severe epileptic and paralytic phenotype affecting these animals (15).

Some of the symptoms displayed by nidogen-2 KO mice injected with TeNT, such as keratoconjuctivitis sicca (23), resembled those of botulism. This suggests that in the absence of one of its physiological receptors, a proportion of TeNT may exploit alternative routes of entry at the NMJ, which are shared with BoNTs. Cross-talk between the two modes of entry and intracellular transport pathways taken by TeNT and BoNTs has been previously suggested (24) and is further supported by the botulism-like symptoms shown by mice injected with high doses of TeNT (25).

Because BoNTs use several SV2 isoforms as receptors (911, 22) and SV2A has been proposed as a TeNT receptor in central neurons (26), we quantified the colocalization of HCT with SV2A and SV2C in motor neurons and tested whether preincubation of HCT with recombinant nidogen-1 modifies the association of HCT with these synaptic vesicle proteins. In the absence of exogenous nidogen, SV2A showed robust colocalization with HCT internalized for 45 min at 37°C (46%) (fig. S8, A and B), whereas the codistribution of HCT with SV2C under these conditions was relatively low (fig. S8, C and D). Therefore, at least a subpool of TeNT may be internalized via synaptic vesicle recycling at the relatively high concentrations used in this assay. However, the addition of recombinant nidogen-1 caused a significant reduction in the colocalization of HCT with SV2A (22%) (fig. S8, A and B), but not with SV2C (fig. S8, C and D). However, despite the addition of recombinant nidogen-1 causing an increased intracellular pool of HcT, the proportion of this probe colocalizing with SV2A was significantly reduced (fig. S8, A and B). Taken together, these data suggest that in the presence of nidogen-1, HCT is preferentially targeted to an endocytic pathway linked to axonal retrograde transport to the cell body and is not taken up by synaptic vesicle recycling.

We have identified the protein receptor responsible for TeNT binding and internalization at the NMJ as well as a potential mechanism to explain the overlap between the entry routes of TeNT and BoNTs (fig. S9). Engineered growth factors with higher affinity for the extracellular matrix (ECM) display enhanced biological activity (27), indicating that differential binding of such ligands to ECM components is crucial for their signaling and biological function (28). TeNT might exploit a similar strategy based on a very efficient capture mechanism at specialized NMJ sites rich in nidogens, which may function to concentrate TeNT as well as physiological ligands, such as neurotrophic factors, to facilitate their uptake and sorting to axonal transport organelles. At these sites, TeNT in complex with nidogens may interact with surface receptors known to bind nidogens, such as the protein phosphatase LAR (29, 30). This specialized capture mechanism is likely to be indispensable to the host cell, and this enables TeNT to be lethal at extremely low concentrations. Our study suggests that nidogens are prime therapeutic targets for suppressing the uptake of TeNT at the NMJ and its access to the CNS, preventing its lethal effects.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 to S4

References (3138)

Movies S1 to S3

References and Notes

  1. Acknowledgments: We thank S. Kjaer (Cancer Research UK London Research Institute) for help with the Octet RED96 system and Kd determination. N. O’Reilly and members of the Protein and Peptide Chemistry Laboratory (Cancer Research UK London Research Institute) for peptide synthesis, purification, and analysis; F. Giribaldi (Cancer Research UK London Research Institute) for help with footprint analyses; K. Deinhardt (University of Southampton) for mass spec analysis of axonal signaling endosomes; S. Swaminathan (Brookhaven National Laboratory) for experiments aiming at the crystallization of the N1 peptide-HCT complex; A. Rummel and T. Binz (Medizinische Hochschule Hannover) and N. Fairweather (Imperial College London) for providing access to published HCT mutants; I. Koxholt (University of Cologne) for help with nidogen-1 and -2 KO mice; and J. N. Sleigh (University College London) for help with statistical analyses. We also thank C. Montecucco (University of Padua) and S. Novoselov (University College London) for constructive comments. This work was supported by Cancer Research UK (K.B., N.S., M.G., M.W., and G.S.), by the Medical Research Council (M.W.), by the University of Padua (P.C. and G.Z.), and by the Deutsche Forschungsgemeinschaft (SFB 829) through the Collaborative Research Center at the University of Cologne and the “Köln Fortune Programm” (R.N.). L.G. is the Graham Watts Senior Research Fellow, supported by the Brain Research Trust. The data presented in this manuscript are found in the main paper and the supplementary data and materials and methods. The authors declare that they have no conflict of interest.
View Abstract

Navigate This Article