50 &

100

YEA

RS A

GO

STRUCTURAL BIOLOGY

Dangerous liaisons on neuronsGiampietro Schiavo

Crystal structures show that botulinum toxins bind simultaneously to two sites on neurons. This dual interaction allows them to use a Trojan-horse strategy to enter nerve terminals, with deadly effect.

Botulinum neurotoxins (BoNTs) are some of the most deadly substances known to man-kind. By blocking nerve function, they cause botulism, a severe condition that may ulti-mately lead to muscular and respiratory paral-ysis. These sophisticated bacterial proteins owe their toxicity to their extraordinary specificity for neurons and to their enzymatic activity. In this issue, papers by Jin et al.1 and Chai et al.2 describe the mechanisms by which BoNT/B — a toxin that causes human botulism — rec-ognizes the surface of neuron junctions (syn-apses)*. This work provides insight into how other BoNTs may exert their lethal action, and describes a mode of binding that might be used by other biological compounds.

Once inside a neuron, a single molecule of BoNT is, in principle, capable of deactivating the whole synapse. BoNTs consist of two pro-tein segments, known as the heavy and light chains. It is the light chain that deactivates neuromuscular junctions — the synapses that connect muscles to their controlling neurons — by specifically inhibiting members of the SNARE protein family3. SNARE proteins are distributed over the membranes of all animal and plant cells and are also found on the mem-branes of synaptic vesicles, the bubble-shaped

organelles that store and release neurotrans-mitter chemicals at neuron terminals. SNARE proteins are essential for membrane fusion, during which vesicles merge with the cell mem-brane and release their load. Once the synaptic vesicles have done this, they are recycled by the neuron for further use.

So how do BoNTs enter neurons? The heavy chain is most likely to be responsible. One half of the heavy chain mediates binding to neurons by interacting with lipid molecules (polysialo-gangliosides, PSGs) in the cell membrane, and with either one of two integral membrane pro-teins — synaptotagmin I or synaptotagmin II — found in synaptic vesicles. A dual-receptor model for these toxins was proposed long ago4, but experimental validation of this theory has required a worldwide effort. The model pre-dicts that the interaction of BoNTs with both PSGs and protein receptors is necessary to explain their awesome potency3, with a differ-ent protein receptor being recognized by each BoNT.

Evidence for protein involvement in BoNT binding was scarce until it was discovered5 that BoNT/B binds to both PSGs and the part of synaptotagmins that lies inside synaptic vesi-cles, in the area known as the lumen. More recently, the specific regions of synaptotag-mins that bind BoNT/B have been identified6,

monochromaticism of any faulty colouring. Cosmetic enhancement comes at a price: here, the addition of new colours. So, in a second step, to restore the graph to its original hues of red, green and blue, we call upon error-correcting codes, the redundancy-adding devices invented by coding theorists to make signals immune to noise. This is no coinci-dence: just as an encoded string with not too many erroneously flipped bits can, through error correction, be restored to its original state, so a PCP proof with not too many errors can be seen to encode a correct proof. Indeed, any error that is not smeared widely enough to be easily spotted is de facto inconsequential.

If you think that no one besides children and cartographers has any interest in colouring graphs, think again: the Riemann hypothesis, protein folding, cryptography and most ques-tions in artificial intelligence can be reduced to the three-colourability of some graphs. That universality is another wonder in the comput-ing cosmos.

PCP, and its elementary proof by Dinur3, is the culmination of 40 years of research in the field of computational reduction. Keep in mind that this is all about verifying proofs, not about understanding them — with only three bits! — let alone discovering them. That must still be done the hard way. In the end, PCP boils down to one revolutionary insight: in the art of persuading others of the truth, the ultimate weapon is a set of dice. ■

Bernard Chazelle is in the Department of Computer Science, Princeton University, 35 Olden Street, Princeton, New Jersey 08540-5233, USA.e-mail: chazelle@cs.princeton.edu

1. Arora, S. & Safra, S. J. Assoc. Comput. Machin. 45, 70–122 (1998).

2. Arora, S., Lund, C., Motwani, R., Sudan, M. & Szegedy, M. J. Assoc. Comput. Machin. 45, 501–555 (1998).

3. Dinur, I. Proc. 38th Annu. ACM Symp. Theor. Comput. 241–250 (2006).

4. Håstad, J. J. Assoc. Comput. Machin. 48, 798–859 (2001). 5. Appel, K., Haken, W. & Koch, J. J. Math. 21, 439–567

(1977).

50 YEARS AGOThe recent passion for the progressive splitting up of genera has spread through most branches of zoology and latterly to botany as well. In fact, the present phase in taxonomy is to lump species and split genera… At present, names are very often proposed on flimsy grounds and the work of proving their validity or otherwise is handed down to future workers. New genera are established using merely the specific characters of the unique species included without any additional evidence. The result is that two species such as Cypraea tigris L. and C. pantherina Sol. are placed in different genera, whereas they are only just distinct species which actually hybridize. Such absurdities should have no standing in nomenclature. From Nature 22 December 1956.

100 YEARS AGO“Cutting a Round Cake on Scientific Principles” — Christmas suggests cakes, and with these the wish on my part to describe a method of cutting them that I have recently devised to my own amusement and satisfaction. The problem to be solved was, “given a round tea-cake of some 5 inches across, and two persons of moderate appetite to eat it, in what way should it be cut so as to leave a

minimum of exposed surface to become dry?” The ordinary method of cutting out a wedge is very faulty in this respect… The cuts shown on the figures represent those made with the intention of letting the cake last for three days, each successive operation having removed about one-third of the area of the original disc. A common India-rubber band embraces the whole and keeps its segments together.From Nature 20 December 1906.

*This article and the papers concerned1,2 were published online on 13 December 2006.

1019

NATURE|Vol 444|21/28 December 2006 NEWS & VIEWS

��������������������� �� �� ��������������

as have those bound by the closely related BoNT/G (ref. 7). Taken together, these results suggest that BoNTs enter neurons by stowing away inside recycled synaptic vesicles (Fig. 1), binding simultaneously to PSGs and to the luminal domain of synaptic-vesicle proteins. The toxins then escape from the vesicle lumen when the vesicles are acidified as they reload with neurotransmitters. Another BoNT family member, BoNT/A, has been shown to bind to the synaptic vesicle protein SV2 (refs 8, 9), in agreement with the prediction that each type of BoNT interacts with a different protein.

Jin et al.1 (page 1092) and Chai et al.2 (page 1096) now present crystal structures showing how BoNT/B binds to the luminal domain of synaptotagmin II. This domain is unstructured in solution, but upon binding to the heavy chain of BoNT/B it folds into a helix, which then enters a cleft adjacent to the PSG-bind-ing site of the toxin. The amino acids lining this cleft are very similar to those found in BoNT/G (ref. 10), but differ in the other toxin family members, which explains why differ-ent BoNTs recognize distinct protein receptors. Mutations in the synaptotagmin-binding cleft greatly reduce the toxicity of BoNT/B, more so than mutations in the PSG-binding pocket. Binding to synaptotagmin therefore seems to be central to BoNT/B toxicity, so blocking this interaction looks like a good strategy for pre-venting botulism.

Although these studies1,2 were performed in the absence of vesicle membranes, the bind-ing of PSGs and synaptotagmin II to adjacent sites on BoNT/B offers clues as to how the toxin might interact with these membranes. The two binding sites would firmly anchor the tip of BoNT/B to the vesicle’s inner sur-face, constraining the toxin’s mobility. The rigid character of this interaction might be further enhanced by the association of the

toxin’s heavy chain with nearby negatively charged lipid molecules, which play an accessory role in stabilizing the toxin on membranes3.

A difference between the binding modes proposed by Jin et al.1 and Chai et al.2 con-cerns the orientation of the remaining portion of the BoNT heavy chain. This section has two long helices that are thought to insert into the vesicle membrane, so helping the light chain to reach the cytoplasm. Jin and col-leagues suggest that these helices sit ortho-gonal to the plane of the membrane, where they resemble a plunger. But Chai et al. pro-pose a parallel orientation, which would place hydrophobic areas of the helices close to the inner surface of the vesicle, facilitating their insertion into the membrane. These two spa-tial arrangements could be snapshots of the journey that BoNT/B takes when close to the membrane of the neuron.

Another issue arising from this work con-cerns the partial overlap of the two binding sites. Chai et al. argue that complex PSGs could contact both BoNT/B and synaptotagmin. This mode of binding would lock the toxin to the neuronal surface, and could explain why BoNTs have a low tendency to detach from the membrane once in place. Furthermore, these studies1,2 do not define the sequence of events that leads to the membrane-binding of BoNT/B (Fig. 1). The simplest possibility is that BoNT/B binds to PSGs and synapto tagmin within the lumen of a synaptic vesicle that is fused to the neuron membrane. However, the recruitment of BoNT/B to the membrane could start outside the vesicle, if it binds first to PSGs and then transfers to a recycling vesicle that contains synaptotagmin. Alternatively, upon PSG binding, BoNT/B might interact with the fraction of synaptotagmin that is present on the surface of the neuron as a result

of inaccurate synaptic-vesicle recycling11. To identify the most relevant mechanism, it

is essential to assess whether PSGs are present on the luminal side of the synaptic-vesicle membrane, but conflicting data have been reported on this issue. Interestingly, tetanus toxin (TeNT) — which is closely related to BoNTs and has a similar mechanism of action — also requires PSGs for its neuronal binding. However, PSGs are excluded from the lumen of the organelles that allow TeNT to enter neuromuscular junctions12. This casts doubt on whether PSGs play a role in BoNT binding inside synaptic vesicles.

Such gaps in our knowledge serve only to heighten our interest in these fascinating neuro-toxins. BoNTs are powerful tools for biologists and find widespread use as therapeutics for the treatment of certain nervous-system diseases. For these reasons, the papers1,2 reported here are of tremendous value. ■

Giampietro Schiavo is at the Molecular Neuropathobiology Laboratory, Cancer Research UK, London Research Institute, 44 Lincoln’s Inn Fields, London WC2A 3PX, UK.e-mail: giampietro.schiavo@cancer.org.uk 1. Jin, R., Rummel, A., Binz, T. & Brunger, A. T. Nature 444,

1092–1095 (2006).2. Chai, Q. et al. Nature 444, 1096–1100 (2006).3. Schiavo, G., Matteoli, M. & Montecucco, C. Physiol. Rev.

80, 717–766 (2000).4. Montecucco, C. Trends Biochem. Sci. 11, 314–317

(1986).5. Nishiki, T. et al. FEBS Lett. 378, 253–257 (1996).6. Dong, M. et al. J. Cell Biol. 162, 1293–1303 (2003).7. Rummel, A., Karnath, T., Henke, T., Bigalke, H. & Binz, T.

J. Biol. Chem. 279, 30865–30870 (2004).8. Dong, M. et al. Science 312, 592–596 (2006).9. Mahrhold, S., Rummel, A., Bigalke, H., Davletov, B.

& Binz, T. FEBS Lett. 580, 2011–2014 (2006).10. Rummel, A. et al. Proc. Natl Acad. Sci. USA (in the press).11. Fernandez-Alfonso, T., Kwan, R. & Ryan, T. A. Neuron 51,

149–151 (2006).12. Deinhardt, K., Berninghausen, O., Willison, H. J., Hopkins,

C. R. & Schiavo, G. J. Cell Biol. 174, 459–471 (2006).

PSGs

Synaptic vesicle fused to neuronal membrane

Synaptotagmin I/II

BoNT/B

a b c

Neuronal membrane

Figure 1 | Possible binding sites of botulinum neurotoxin B (BoNT/B) on neurons. Crystal studies by Jin et al.1 and Chai et al.2 suggest that BoNT/B invades neurons by stowing away in carriers known as synaptic vesicles. It does this by forming a complex with lipid molecules (polysialogangliosides, PSGs) and with a vesicle protein (either synaptotagmin I or synaptotagmin II) in the neuronal membrane. The complex is stabilized by interactions with neighbouring acidic lipid molecules (orange). BoNT/B could enter open vesicles at the neuron’s membrane, in one of three possible sequences. a, BoNT/B enters the vesicle directly and forms the required complex. b, BoNT/B binds first to PSGs on the membrane, and is then transferred into a synaptic vesicle containing synaptotagmin. c, BoNT/B forms a full complex on the membrane, using synaptotagmin left behind by inaccurate vesicle recycling. It is then transferred into the lumen of the vesicle.

1020

NATURE|Vol 444|21/28 December 2006NEWS & VIEWS

�������������������� ������ ������ ������