Botulinum Neurotoxins A and E Undergo Retrograde Axonal Transport in Primary Motor Neurons

The striking differences between the clinical symptoms of tetanus and botulism have been ascribed to the different fate of the parental neurotoxins once internalised in motor neurons. Tetanus toxin (TeNT) is known to undergo transcytosis into inhibitory interneurons and block the release of inhibitory neurotransmitters in the spinal cord, causing a spastic paralysis. In contrast, botulinum neurotoxins (BoNTs) block acetylcholine release at the neuromuscular junction, therefore inducing a flaccid paralysis. Whilst overt experimental evidence supports the sorting of TeNT to the axonal retrograde transport pathway, recent findings challenge the established view that BoNT trafficking is restricted to the neuromuscular junction by highlighting central effects caused by these neurotoxins. These results suggest a more complex scenario whereby BoNTs also engage long-range trafficking mechanisms. However, the intracellular pathways underlying this process remain unclear. We sought to fill this gap by using primary motor neurons either in mass culture or differentiated in microfluidic devices to directly monitor the endocytosis and axonal transport of full length BoNT/A and BoNT/E and their recombinant binding fragments. We show that BoNT/A and BoNT/E are internalised by spinal cord motor neurons and undergo fast axonal retrograde transport. BoNT/A and BoNT/E are internalised in non-acidic axonal carriers that partially overlap with those containing TeNT, following a process that is largely independent of stimulated synaptic vesicle endo-exocytosis. Following intramuscular injection in vivo, BoNT/A and TeNT displayed central effects with a similar time course. Central actions paralleled the peripheral spastic paralysis for TeNT, but lagged behind the onset of flaccid paralysis for BoNT/A. These results suggest that the fast axonal retrograde transport compartment is composed of multifunctional trafficking organelles orchestrating the simultaneous transfer of diverse cargoes from nerve terminals to the soma, and represents a general gateway for the delivery of virulence factors and pathogens to the central nervous system.


Introduction
Through the years, bacterial and animal toxins have been the target of intense medical investigation due to their importance for human health [1]. As such, their structure-function relationships and mechanism of action have been extensively analysed, leading to the development of powerful vaccines and chemical inhibitors targeting either the toxin active site or key events in their cellular intoxication process [2][3][4][5]. Basic and clinical research into the mechanism of action of these toxins has been boosted by the inclusion of some of them, such as anthrax lethal factor and botulinum neurotoxins (BoNTs), among bioterrorism threats.
In spite of these warfare links, some of these protein toxins have recently acquired novel roles in human medicine, which go beyond their importance as vaccine products. The clearest example is represented by BoNTs, the causal agents of animal and human botulism. BoNTs are large proteins produced by different bacteria of the genus Clostridia and together with the closely related tetanus toxin (TeNT) form the clostridial neurotoxin family (CNT) [4,6,7]. Seven different serotypes named from A to G (BoNT/A-G) and more than 35 variants are presently known [8]. BoNTs and TeNT are produced as single polypeptides with an average molecular weight of 150 kDa. Single chain BoNTs are converted into fully active neurotoxins by bacterial and tissue proteases, yielding the di-chain fully active molecules. The active form is composed of a heavy (H) chain, in which the carboxy-terminal domain (H C ) mediates neurospecificity and high affinity binding to receptors present on the plasma membrane of neurons, and a light (L) chain, which is responsible for the intracellular activity of the neurotoxin [4,6,9]. The L chain is indeed a zinc-dependent endopeptidase specific for proteins belonging to the SNARE superfamily [10], which have an essential role in the fusion of synaptic vesicles with the presynaptic membrane [11]. Cleavage of synaptic SNAREs halts the release of neurotransmitters and is responsible for the long-lasting block of neuroexocytosis observed in cultured neurons [12] and the paralysis elicited by these neurotoxins in vivo [13].
With the exception of BoNT/C, a single substrate has been described for each CNT [6]. BoNT/B, D, F, G and TeNT cleave VAMP, a SNARE protein localised on synaptic vesicles, whilst BoNT/A and E target SNAP25, which is anchored to the plasma membrane at synaptic and extrasynaptic sites. BoNT/C cleaves both SNAP25 and syntaxin, another SNARE localised to the synaptic plasma membrane. BoNTs and TeNT cleave these SNARE proteins at a single peptide bond within their cytoplasmic domain, generating fragments that are unable to sustain membrane fusion and thus neurotransmitter release [10].
Despite their structural and mechanistic similarities, BoNTs and TeNT display striking differences in terms of the type of paralysis induced in vivo [5]. BoNTs cause a flaccid paralysis, which is the hallmark of clinical botulism and the basis of a main therapeutic effect observed upon injection of BoNTs in target tissues [7,10]. This type of paralysis has been attributed to the block of acetylcholine release at the neuromuscular junction, which, as a consequence, is unable to drive contraction of the target muscle [9,13]. In contrast, TeNT is known to cause a spastic paralysis, which has been attributed to the block in release of inhibitory neurotransmitters in spinal cord interneurons [14]. Several lines of evidence demonstrate that, upon entry into motor neurons, TeNT reaches the spinal cord via long-range axonal retrograde transport [15] followed by transcytosis into neighbouring inhibitory interneurons [16]. Therefore, a different trafficking pathway rather than a diverse intracellular mechanism of action has been suggested to be at the basis of the distinct physiological effects of BoNTs and TeNT.
Several experiments have recently challenged this paradigm demonstrating that BoNT/A is also capable of eliciting its activity in areas distant from the injection site [17][18][19][20], implying that BoNT/A undergoes long range trafficking in certain experimental conditions.
Here, we show that BoNT/A and BoNT/E are internalised by primary spinal cord motor neurons and undergo fast axonal transport in these cells. BoNT/A and BoNT/E are internalised in non-acidic axonal carriers that contain TeNT, following a process that is largely independent of membrane depolarisation. These results suggest the existence of a long-range transport pathway in motor neurons, which host receptors for several virulence factors and pathogens targeted to the central nervous system.

Results
The binding fragments of BoNT/A and BoNT/E are internalised in spinal cord motor neurons In the last decade, the recombinant binding fragment of TeNT (H C T) has been used extensively to monitor internalisation and axonal retrograde transport in many neuronal types [15], allowing the quantitative analysis of these trafficking processes both in vitro and in vivo [21][22][23][24].
To assess the ability of the binding fragments of BoNT/A (H C A) and BoNT/E (H C E) to bind and undergo internalisation in living neurons, we expressed them in bacteria as glutathione Stransferase (GST) fusion proteins containing a cysteine-rich tag previously described for H C T [21,25]. Since these GST fusion proteins were significantly more stable than the cleaved products, we decided to use uncleaved GST-H C A and GST-H C E for our studies. GST tagged with the same cysteine-rich peptide and labelled with a maleimide-based fluorophore was used as a control. To test for binding, spinal cord motor neurons were incubated at 4uC in the presence of fluorescent H C A and H C E. As shown in Figure 1, a specific signal was detectable on the surface of neurons incubated with H C A (15 nM) and H C E (7.5 nM), whereas fluorescent GST (15 nM) showed no detectable binding under the same conditions. BoNT/A and BoNT/E have been shown to rely on the synaptic vesicle protein SV2 for their binding and uptake in neurons [26][27][28]. Similarly, preincubation of H C A with an excess of a recombinant fragment of SV2C (residues 454-579) fused to GST (1:100; a kind gift of T. Binz and A. Rummel) compromised the binding and uptake of this fragment in motor neurons (data not shown). We then tested the extent of colocalisation of H C A and H C E with SV2C using an antibody raised against its cytoplasmic domain. Under the binding conditions used above, only limited colocalisation was however detected between the H C A and H C E fragments and SV2C on the plasma membrane of resting motor neurons ( Figure 1B). The relative low intensity of the signal observed upon incubation of H C A and H C E at 4uC, and its diffuse nature ( Figure 1B) prevented us to perform a reliable quantitative analysis on the extent of colocalisation of these binding fragments with SV2C under our experimental conditions.
BoNT/A and/E, as well as other CNTs, have been shown to interact with both polysialogangliosides and protein receptors on the surface of neuronal cells, and undergo both productive and unproductive binding [29]. To verify whether the binding of H C A and H C E is compatible with internalisation or is just a dead-end process (e.g. as a result of polysialoganglioside clustering), we incubated motor neurons with labelled H C A (15 nM) and H C E (7.5 nM) under resting conditions, followed by an acid wash to remove the H C still present on the plasma membrane. When neurons are kept at 4uC, no labelling is detectable upon acid wash (data not shown), as expected for conditions that are known to greatly reduce the rate of endocytosis. However, after incubation for 30 minutes at 37uC, H C A and H C E puncta resistant to acid stripping, which partially colocalise with SV2, were found in motor neurons (Figure 2A). Interestingly, H C A displays similar

Author Summary
Botulinum neurotoxins are the most toxic molecules known to mankind, and as a result, are currently listed among the top bio-threats. However, their ability to bind specifically to neurons and their inhibitory effects on regulated secretion prompted their clinical use in pathologies characterised by increased muscular tone, such as dystonia and various forms of spasticity, or abnormal secretion, such as drooling and excessive sweating, to cite a few. As a consequence, botulinum neurotoxin A, which is the serotype most commonly used in human therapy, has become the treatment of choice for an ever-expanding number of pathological and non-pathological (e.g. cosmetic) conditions. All current indications show that the systemic effects and toxicity of botulinum neurotoxin A are minimised by the specific route of administration (local injection) and the low diffusion of this molecule in tissues. However, recent reports suggest that in contrast to this common belief, botulinum neurotoxin A is able to reach distal sites in the body and may have previously unanticipated effects in the central nervous system. In this study, we demonstrate that botulinum neurotoxin A and E enter alternative endocytic pathway(s) in addition to synaptic vesicle recycling, and undergo long-range transport in a non degradative compartment in spinal cord motor neurons. Our results show that axonal retrograde transport is a common pathway for the dissemination in the central nervous system of pathogens and virulence factors important for human and animal health.
colocalisation levels with both SV2A and SV2C, whereas H C E shows a preference towards SV2C under the same experimental conditions (Mann-Whitney test; **, p,0.01, ***, p,0.001; Figure  S1). The finding that both H C A and H C E are internalised by resting motor neurons and display only a partial colocalisation with SV2 isoforms suggests that these neurotoxins may also exploit an alternative, synaptic vesicle-independent pathway to enter neuronal cells.
Since depolarisation has been shown to promote BoNT/A and/ E uptake [30], we then tested whether the endocytosis of H C A and H C E is enhanced by depolarising motor neurons with 60 mM KCl ( Figure 2B). Under these conditions, internalisation occurs with higher efficiency for both H C A and H C E, resulting in a statistically significant increase (Mann-Whitney test; ***, p,0.001; Figure 2C, D). However, the colocalisation observed between the H C fragments and SV2A and C does not significantly increase under depolarising conditions ( Figure 2B and Figure S1), suggesting that H C A and H C E do not only enter SV2-positive synaptic vesicles but also other endocytic organelles in primary motor neurons.
This unexpected behaviour was not due to the recombinant H C s, since similar results were obtained using full length BoNT/ A and BoNT/E. In this case, the depolarising conditions seem even less efficient in promoting the internalisation of the full length toxins and enhancing the colocalisation with SV2C ( Figure S2).

BoNT/A and/E are internalised in motor neurons by multiple endocytic routes
To further investigate the role of the synaptic vesicle cycle in the internalisation of H C A and H C E, we took advantage of the specific Motor neurons were incubated with 7.5 nM AlexaFluor488-GST-BoNT/E H C (H C E) and 7.5 nM AlexaFluor555-GST (GST) as control. Scale bar, 20 mm. No fluorescence signal was detectable for GST, whilst a punctate signal was present for H C A and H C E, indicating the ability of these binding fragments to bind to motor neurons. The signal is rather low, as expected in these experimental conditions (4uC). (B) Motor neurons were incubated with 15 nM H C A (top) or 15 nM H C E (bottom) for 15 min at 4uC, fixed, and stained for SV2C. Only limited colocalisation of H C A and H C E with SV2C was found in these conditions. Inset: high magnification of the indicated areas. This analysis was performed using two independent primary motor neuron cultures, and replicated at least twice for each motor neuron preparation. Shown are representative images for each condition. Scale bars, 10 mm (top); 20 mm (bottom). doi:10.1371/journal.ppat.1003087.g001 inhibition caused by BoNT/D on synaptic vesicle exocytosis. BoNT/D blocks the fusion of synaptic vesicles and their recycling by cleaving VAMP. We pre-incubated motor neurons with medium alone or containing 2 nM BoNT/D for 22 h, followed by the addition of H C A and H C E under depolarising conditions (60 mM KCl). As shown in Figure 3A and B, pre-treatment with BoNT/D did not prevent the internalisation of H C A and H C E, in spite of the complete cleavage of VAMP2 in treated neurons. Consistent with the results shown above, the colocalisation observed between H C A, H C E and VAMP2 at synaptic sites is incomplete ( Figure 3A, B; upper panels), but qualitatively similar to that detected with SV2C and SV2A ( Figure 2 and Figure S1), suggesting that these neurotoxins exploit multiple routes for their entry into motor neurons [31]. These routes extensively overlap for H C A and H C E, since these two binding fragments displayed an almost complete colocalisation when internalised together in cultured motor neurons under depolarising conditions in the presence or absence of BoNT/D ( Figure 3C).
Taken together, these experiments demonstrate that H C A and H C E are reliable tools to monitor the binding and uptake of their parental neurotoxins in primary motor neurons. Furthermore, the limited effect of BoNT/D on H C A and H C E internalisation provides a very good qualitative indication that synaptic vesicle endo-exocytosis is not the only mechanism responsible for the uptake of these binding fragments in motor neurons and that an endocytic pathway(s) largely independent of synaptic vesicle recycling is involved in BoNT/A and/E uptake in both resting and depolarising conditions. . Quantification of the uptake of H C A and H C E is shown in (C) and (D), respectively. Bars represent the mean 6 standard deviation (SD) of the fluorescence intensity determined from a representative experiment. Ten to thirty fields were analysed for each condition. Although the internalisation of H C A (Mann-Whitney test; ***, p,0.001), and H C E (Mann-Whitney test; ***, p,0.001), is significantly increased under stimulation, it is extensive also in resting conditions. doi:10.1371/journal.ppat.1003087.g002

H C A and H C E undergo retrograde transport in motor neurons
Our observation that the binding fragments of BoNT/A and/E enter motor neurons by parallel endocytic routes begs the question as to the fate of the organelles containing H C A and H C E. Whereas synaptic vesicles are known to recycle within the synapse and only occasionally transfer inter-synaptically [32,33], other endocytic compartments are targeted to distal sites via long-range transport pathways. Endogenous ligands, such as neurotrophins and their receptors, lectins (e.g. wheat germ agglutinin) and pathogens (e.g. several neurotropic viruses) enter axonal carriers that undergo microtubule-dependent retrograde transport to the cell body [15]. Since H C T is an established probe for this trafficking pathway [21][22][23][24], we sought to test whether BoNT H C s share the same axonal compartment as H C T.
We incubated motor neurons with 15 nM H C A and 40 nM H C T for 30 min at 37uC either in resting (5 mM KCl) or depolarising (60 mM KCl) conditions prior to shifting them on ice and acid wash. The latter two treatments were omitted when live imaging was performed. As shown in Figure S3, there was a good colocalisation between H C A and H C T in resting and depolarised neurons. Interestingly, a fraction of these H C A-and H C T-positive axonal puncta were mobile and underwent fast retrograde transport towards the soma ( Figure 4A). Colocalisation was not limited to moving carriers, but was also frequently observed at the level of stationary organelles (data not shown). Quantitative analysis of the retrograde transport of H C A displays a multimodal kinetics with an average speed of 0.8 mm/s ( Figure 4B). Interestingly, the speed distribution profiles of H C A and H C T largely overlapped, indicating, together with their colocalisation in axons ( Figure 4A) that these two binding fragments are transported by the same class of axonal organelles.
Whereas distal effects of BoNT/A have been described in multiple systems [17,20,34], such evidence for BoNT/E has been lacking [20]. Our finding that H C A undergoes retrograde transport in neurons raises the possibility that the different physiological effects of BoNT/A and/E are due to their differential ability to be recruited to this transport route once internalised in motor neurons. Therefore, we performed comparative assays to better understand whether the similarity between these two neurotoxins is only limited to binding and internalisation or if it extends to axonal transport.
Motor neurons were incubated with H C A and H C E and analysed by time-lapse confocal microscopy ( Figure 4C). Strikingly, the extensive colocalisation previously found between H C A and H C E was not only limited to stationary structures, but also involved organelles transported in the retrograde direction ( Figure 4C). Kinetic analysis of H C E transport revealed a mainly retrograde speed distribution profile, which overlaps with those of H C A and H C T ( Figure 4B).

H C A undergoes axonal retrograde transport in compartmented motor neuron cultures
The experiments presented so far were performed using spinal cord preparations enriched in motor neurons in mass cultures. Under these conditions, it is not always possible to provide an unequivocal identification of the type of neuron imaged at any given time, nor to assess the site of internalisation of the neurotoxin added to the medium. To overcome these technical shortcomings, we exploited a compartmentalised system based on microfluidic chambers (MFC), which allows the separation of cell bodies from axon terminals and is suitable for live imaging [35,36] and biochemical analyses [36]. To avoid diffusion between axonal and cell body compartments, the volume of medium in the latter compartment was maintained at higher levels at all times to ensure a laminar flow towards the axonal side. In a first set of experiments, we used as a source of motor neurons, an embryonic stem (ES) cell line stably transfected with a construct encoding green fluorescent protein (GFP) under the control of a motor neuron-specific promoter (HB9::GFP) [37]. Motor neurons differentiated from this ES cell line express GFP in their cytoplasm, allowing their unambiguous identification. Cells were also counterstained with the pan-neuronal marker bIII tubulin and the axonal marker SMI32. An example of a motor neuron axon positive for bIII tubulin, SMI32 and GFP, crossing the microgroove of the MFC is shown in the lower part of Figure 5A, whilst an axon belonging to a different type of neuron (or a motor neuron in which the HB9 promoter has already been switched off) not expressing GFP is visible in the upper part of the same panel.
To further prove that the axonal transport of H C A seen in mass cultures occurs in a retrograde direction, we incubated mouse primary motor neurons with fluorescently-labelled H C A alone (Movie S1) or together with an antibody directed against the neurotrophin receptor p75 NTR for 30 minutes at 37uC, adding the two probes only to the axonal side of the MFC. We then monitored axonal transport in the MFC microgrooves using timelapse microscopy. Representative stills of a movie (Movie S2) displaying retrogradely transported H C A and p75 NTR positive organelles moving retrogradely are shown in Figure 5B, together with the corresponding kymograph. Interestingly, several of the axonal carriers containing H C A are also positive for the neurotrophin receptor p75 NTR , an established marker of the axonal retrograde transport compartment [23,[38][39][40]. Altogether, these findings demonstrate that in motor neurons H C A and H C E undergo fast axonal retrograde transport in endosomal carriers, which are shared with H C T and endogenous cargoes, such as the neurotrophin receptor p75 NTR .

H C A and H C E are transported in non-acidic transport carriers
These results prompted us to further characterise the moving organelles containing H C A and H C E. Entry of TeNT and BoNTs in acidic compartments is indeed required for the translocation of the L chain into the cytoplasm and cleavage of its SNARE target with the resulting inhibition of neurotransmitter release [41].
To test the presence of H C A and H C E in acidic carriers, we incubated motor neurons cultures with H C A and H C E together with Lysotracker, a probe that accumulates only in acidic vesicles. Neither H C A ( Figure 6A; Movie S3), or H C E ( Figure 6B; Movie S4) were found in Lysotracker-positive organelles in the axons of motor neurons. This result was confirmed by the almost complete absence of yellow organelles in the kymographs corresponding to H C A ( Figure 6C; left panel) and H C E ( Figure 6C; right panel) and quantified in Figure 6D (H C A-Lysotracker positive carriers, 5.262.0%; H C E-Lysotracker positive carriers, 3.063.0%). The overlap between H C A-and H C E-positive compartments with acidic organelles was very limited in the soma as well ( Figure 6E), indicating that H C A and H C E are transported and sorted in nonacidic organelles in spinal cord motor neurons.

Full length BoNT/A and BoNT/E enter retrograde transport carriers with different efficiency
Prompted by the results obtained with H C A and H C E, we tested the ability of the full-length neurotoxins to undergo retrograde transport. To this end, spinal cord motor neurons were incubated with full length fluorescent BoNT/A or BoNT/E and imaged Quantitative kinetic analyses revealed similar speed distribution profiles for full length BoNT/A and BoNT/E, with a slight increase in the frequency of pauses and movements in the anterograde direction for the latter ( Figure 7C). Although both neurotoxins are retrogradely transported, this process seems to occur with different modalities for BoNT/A and/E. Whereas fast retrogradely-transported organelles showing progressive movements were detected with full length BoNT/A, BoNT/E-positive carriers displayed a less continuous motion toward the cell body, as demonstrated by a higher frequency of reversals ( Figure 7D). stationary double-positive carrier. Scale bar, 10 mm. (C, bottom) Kymographs of motor neuron axons correspond to the stills above. Note the high number of double-positive carriers, either moving towards the soma or oscillatory. This analysis was performed using three (H C A, H C T) or two (H C E) independent primary motor neuron cultures, and replicated at least twice. At least ten (H C A, H C T) or five (H C E) movies per conditions were used to assemble the speed distribution curves shown in (B). doi:10.1371/journal.ppat.1003087.g004  These transitory changes in direction of the carriers could account for the presence of a peak at 20.2 mm/s (anterograde direction) in the speed distribution profile of BoNT/E and a higher frequency of pauses ( Figure 7D). These differences in the transport kinetics of BoNT/A and BoNT/E might contribute to the lower efficiency of BoNT/E in cleaving SNAP25 in cell bodies compared to BoNT/A, when these neurotoxins are applied to axons in compartmentalised cultures of sympathetic neurons [18]. Altogether, these results suggest that full length BoNT/A and/E are retrogradely transported in primary motor neurons, albeit with different efficiency.

Long-distance effects of BoNT/A in spinal cord motor neurons
The next goal of our study was determining whether these neurotoxins reach the soma in an active form, since this would provide mechanistic insights on their long-range mode of action in vivo. We assessed the long-range effects of BoNT/A by monitoring the appearance of the cleaved fragment of SNAP25, which can be distinguished from the full-length protein using an antibody specific for the cleaved form [17,19,20,34].
Motor neurons grown in MFCs (DIV6) were treated with BoNT/A (10 nM) for 24 h at 37uC. Importantly, the neurotoxin was added only to the axonal compartment, which is microfluidically isolated from the somatic side. Therefore, appearance of the cleaved fragment of BoNT/A in the latter compartment would imply that full length BoNT/A underwent axonal transport to the cell body and translocated into the cytoplasm. Cells were washed, fixed, permeabilised and stained for BoNT/A-cleaved SNAP25 [17,19,20,34]. Strikingly, the cleaved fragment of SNAP25 was detected both in the axonal and somatic side only in MFCs treated with BoNT/A (BoNT was added only to the axonal compartment)( Figure 8A and B, left panels), but not in control chambers treated with vehicle ( Figure 8A and B, right panels), indicating that BoNT/A is not only retrogradely transported in motor neurons but it is also capable of eliciting its catalytic activity following transport. The presence of cleaved SNAP25 in the somatic side of MCFs treated with BoNT/A, but not in control MCFs, was further confirmed by western blot analysis of extracts obtained by pooling the content of three separate chambers ( Figure 8C).
To confirm long-distance SNAP25 cleavage in spinal cord motor neurons in vivo, we injected BoNT/A into the hind leg muscles of adult rats. Ten days after the delivery of the neurotoxin, lumbar samples of spinal cord were taken and processed for western blot. As shown in Figure 8D, detectable levels of cleaved SNAP25 were found in the spinal cord of BoNT/A-injected animals, but not in sham-treated controls, indicating long-distance neurotoxin action in vivo.

Different kinetics of BoNT/A and TeNT action account for their distinct pathophysiological effects in vivo
The demonstration that BoNT/A and TeNT shared a common retrograde transport pathway in motor neurons prompted the key question on why these two neurotoxins display such remarkable differences at pathophysiological level, inducing a flaccid vs. a spastic paralysis, respectively. To directly compare the effects of BoNT/A and TeNT in vivo, we chose to test their activity in facial motor neurons projecting to the whisker pad. This model system has two important advantages: i) peripheral neuroparalytic effects can be promptly monitored; and ii) facial motor neurons lack direct proprioceptive and sensory innervation [42], allowing the selective analysis of trafficking events in motor neurons in vivo.
We injected BoNT/A (1 nM) or TeNT (3 nM) into the whisker pad of rats and monitored the peripheral effects of the toxins on the behaviour of the animals at different time points. In a separate cohort of animals treated in parallel, we assessed the cleavage of SNAP25 in the ipsilateral brainstem facial nucleus, which contains the cell somas of the motor neurons innervating the whisker pad. On day one following the delivery of the toxin ( Figure 9A), we found that BoNT/A completely blocked whisker movements in the treated side. Phenotypically, vibrissae in the injected side were atonic and positioned backward (Movie S5; see an example of a control animal in Movie S6). However, BoNT/A-cleaved SNAP25 in the facial nucleus was only clearly detectable from day three ( Figure 9B), indicating a temporal shift between peripheral and central action of BoNT/A. Interestingly, we observed a build up of central cleaved SNAP25 over time ( Figure 9B), suggesting progressive cumulative effects. In striking contrast, TeNT action followed a different time course. Both the peripheral paralysis and central VAMP2 proteolysis (detected as decreased levels of intact VAMP2 in treated animals) occurred with very similar kinetics and were overt at day three ( Figure 9C and D). At this time point, whiskers appeared rigid and immobile, protruding at a right angle from the snout (Movie S7). This is consistent with a spastic paralysis, indicating TeNT action on inhibitory circuits after retrograde trafficking in brainstem motor neurons.
These results demonstrate that, in the case of BoNT/A, the peripheral action precedes the central one, whilst their kinetics overlap for TeNT. Thus, flaccid paralysis of injected muscle is maintained by the fast and robust blockade of acetylcholine release from peripheral nerve terminals, whilst neural circuits serving other (e.g. antagonistic) muscles can be significantly affected by centrally active BoNT/A.

Discussion
The textbook explanation of the differences between the clinical symptoms of tetanus and botulism is based on the distinct fates of TeNT and BoNTs once internalised in motor neurons [4,6]. Recent in vivo and in vitro evidence are now challenging this paradigm, suggesting that BoNTs might undergo long-distance axonal transport, especially at high doses. Early experiments with radiolabelled full length BoNT/A showed that the toxin is transferred to the ventral roots and adjacent spinal cord segments upon intramuscular injection in the cat gastrocnemius [43,44]. Similarly, Black and Dolly [45] observed radiolabelled BoNT/A within the axoplasm of myelinated axons after its peripheral injection in mice. A dose-dependent retrograde transport of BoNT/A in brainstem motor neurons was also shown by electrophysiological and ultrastructural experiments in cats [46,47]. In compartmentalised cultures of rat sympathetic neurons, BoNT/A moves retrogradely into cell bodies when applied at high concentrations into the distal compartments [18]. Finally, Antonucci et al. provided evidence for retrograde transport and transcytosis of BoNT/A in rat facial motor neurons after its injection into the whisker pad [20]. However, retrograde trafficking of BoNTs has been inferred mainly indirectly, i.e. by observing the appearance of radioactivity or BoNT-cleaved substrates away from the site of administration. Thus, the kinetics and intracellular pathways used by BoNTs for their long-range transport remains unclear.
Our work was designed to fill this gap, using differentiated motor neurons and both the binding domain and the full length forms of BoNT/A and BoNT/E. H C A and H C E were found to have comparable uptake and transport properties to the full length neurotoxins, implying that the H C domain carries the minimum determinant(s) for long-range transport [41,48]. We found that the uptake of H C A and H C E in motor neurons is enhanced under depolarisation, a condition that stimulates BoNT endocytosis into central neurons [27,30,49,50]. Thus, synaptic vesicle recycling, which is increased under depolarising conditions, plays an important role in toxin uptake in motor neurons. However, this is unlikely to be the only internalisation route exploited by BoNTs to enter these neurons. Pre-treatment with BoNT/D, which blocks exocytosis by selectively cleaving VAMP [6], does not completely prevent internalisation of H C A and H C E in these cells. Thus, an internalisation route independent of synaptic vesicle recycling should be considered for the entry of BoNTs in motor neurons. Such a pathway may involve the small fraction of SV2 that resides at steady state on the plasma membrane, together with other synaptic vesicle proteins [51,52]. However, the incomplete colocalisation of SV2A and C with H C s suggests the involvement of additional endocytic route(s) for BoNT/A and/E uptake in motor neurons.
Our time-lapse analyses in motor neurons in mass culture and in MFCs demonstrated that BoNT/A and BoNT/E, and their H C domains are retrogradely transported in neurons. These results provide direct evidence that at least a fraction of internalised BoNT is capable of fast long-range trafficking, consistent with previous data [18,20,43,44,46,47]. Determining the proportion of BoNTs entering the local (synaptic vesicle-based) versus distal (retrograde endosome-based) trafficking pathways is potentially very important to address the balance between peripheral and central effects of these neurotoxins. Although we cannot yet provide conclusive data to address this question, the quantification of the uptake of HcA and HcE under resting or depolarising conditions indicates that about 50% of these fragments are internalised in a stimulation-independent manner ( Figure 2C and  2D). Only a fraction of this pool is targeted to the axonal retrograde pathway, since direct comparison with H C T indicates that H C A and H C E are significantly less efficient in being recruited to this route, as exemplified by their higher proportion of stationary carriers and their overall lower frequency of transport (about 20-30% of H C T). In addition, movement of BoNT/E was less continuous compared to BoNT/A, as shown by the occurrence of stops and transitory changes in direction of these axonal endosomes. Based on these considerations, the overall proportion of distally targeted H C A would be between 10 and 15%, and even less for H C E. These values are in good agreement with the proportion of BoNT/A targeted to long-range axonal transport in the visual system (5 to 10%; LR and MC, unpublished results).
The high degree of colocalisation observed during uptake and transport indicates that H C A and H C E share the same organelles once internalised into neurons. A partial colocalisation of H C Aand H C T-positive axonal puncta was also apparent during transport. Since H C T and TeNT use a transport pathway shared with neurotrophins and their receptors [21,23], this result suggests that the same pathway is used by H C A and H C E for long-distance trafficking. This conclusion has been confirmed by our findings that p75 NTR and H C A enter the same axonal compartment in motor neurons. Retrogradely-moving BoNT-positive carriers display negligible colocalisation with Lysotracker, a marker of acidic vesicles. This is particularly important, since acidic pH is known to induce the translocation of the L chain of BoNTs into the cytosol [41,53]. Protection of BoNT-containing carriers from a drop in pH during transport may be key in order to entrap the neurotoxin into the lumen of the transport organelle during its transfer to the soma. Catalytically intact BoNT may eventually reach suitable release sites and become available for transcytosis to connected neurons in the network [19,20,34], as shown for TeNT [16,54,55].
Interfering with the acidification of the early endosomal compartment responsible for the synaptic uptake of H C A and H C E is a possible strategy to re-direct these fragments to the axonal retrograde pathway. However, this approach could not be pursued in our experimental system, since preventing endosome acidification by inhibiting the vATPase with concanamycin A or bafilomycin A impairs very early events in the sorting of H C T and halts its recruitment to retrogradely-transported signalling endosomes [22]. Moreover, a functional vATPase is required for other endocytic mechanisms linked to regulated secretion, such as the retrieval of secretory granule proteins stranded on the plasma membrane, which may be exploited by BoNT/A and BoNT/E to enter motor neurons upon binding to SV2 [56].
The similarity of trafficking mechanisms exploited by TeNT and BoNTs needs to be reconciled with the vastly different clinical symptoms of botulism and tetanus. Although alternative hypotheses have been previously suggested [57], our in vivo studies on facial motor neurons offer a sound explanation. Similar doses of BoNT/A and TeNT resulted in clearly distinct kinetics of the peripheral neuroparalytic effects, with TeNT inducing a delayed blockade of whisker movements (day three) compared to BoNT/A (day one). Conversely, the central proteolytic action displays a very similar onset for the two neurotoxins, starting after three days from injection. Whereas for TeNT the peripheral and central effects coincide, for BoNT/A the central effect lags behind the onset of peripheral symptoms. Consequently, the blockade of acetylcholine release at the neuromuscular junction ''masks'' any alteration in motor neuron firing caused by the central action of BoNT/A, thus leaving the injected muscle persistently flaccid. However, functional long-term consequences of BoNT/A acting at the level of central circuits [34], should not be overlooked. In this regard, any neurotoxin-induced modification in the strength of spinal cord synapses impinging onto motor neurons is bound to alterations in their output that in turn may affect firing of inhibitory interneurons, such as Renshaw cells [58,59], with a substantial impact on antagonistic and/or synergistic muscles. This may occur via the widespread projections of motor neuron recurrent collaterals to Renshaw cells impinging on motor nuclei supplying muscles acting at the same joint or trans-joint [60].
Although our results provide a rationale of the different neuroparalytic effects of TeNT and BoNTs, much more work is required to quantitatively assess the targeting of BoNT/A and BoNT/E to distal sites. Therefore, a major aim of future experiments is determining the dose dependence of the effects of these neurotoxins at synaptic sites and in the soma both in vitro and in vivo, a strategy that would indirectly address the proportion of BoNT/A, BoNT/E and TeNT taken up by the acidic pathway. This analysis is however complicated by the methodology used for the detection of the activity of these neurotoxins, which is based on antibodies recognizing the cleaved SNARE proteins. Indeed, this approach is highly dependent on variations in the enzymatic activity of the L chains of different serotypes, their efficiency of translocation into the cytoplasm, their post-translational modifications and the relative intracellular stability of both the L chain and the cleaved substrates, parameters for which our present understanding is very limited [6,41,61,62].
Several pathogens and virulence factors have been shown to exploit axonal retrograde transport pathways to spread into the central nervous system. We have recently shown that poliovirus [63] and canine adenovirus serotype 2 (CAV2) [40] enter the same transport carriers used by TeNT and BoNTs together with their physiological receptors. This result is surprising since these pathogens as well as several endogenous cargoes are taken up by neurons using different endocytic mechanisms. For example, even though the B subunit of cholera toxin (CTB) binds to the ganglioside GM1 and is internalised via a clathrin-independent route [64], it is co-transported together with H C T and p75 NTR , which are taken up by clathrin-mediated endocytosis [39,64]. Therefore, mechanisms operate at distal sites of neurons (e.g. the neuromuscular junction) to sort these diverse endosomal cargoes to common non-degradative organelles, which are then recruited to a long-range axonal transport route. This whole sequence of events serves to translocate endogenous ligands, pathogens and virulence factors from the periphery of motor and sensory neurons to the central nervous system. It is plausible that these transport carriers undergo another sorting step once they arrive in the soma, a process that would provide a novel regulatory mechanism in communicating information from nerve terminals to the cell body ( Figure 10). The presence of fewer types of retrograde transport organelles than anticipated could have profound effects on axonal homeostasis and the regulation of overall cargo flow. A limited number of carrier types is likely to streamline the mechanisms ensuring motor recruitment and cargo transfer, which would occur mainly at hubs positioned at distal nerve terminals and in the soma, thus simplifying the control of membrane flow in the axon. The presence of multiple receptors and ligands in these transport carriers would also enhance their plasticity in terms of signalling potential. Thus, the signal(s) generated by a single receptor/ligand complex could vary in amplitude, frequency and outcome based on the presence of other cargoes for a given retrogradelytransported organelle. Uncovering the determinants of these transport and sorting mechanisms will provide new insights on how long range communication is regulated and will identify new targets for the control of trafficking of pathogens in the nervous system.
Whereas this data indicates a robust retrograde transport of H C E and BoNT/E in cultured motor neurons, previous experiments based on unilateral BoNT/E delivery into the rodent brain failed to find evidence for propagation of the effects induced by BoNT/E to the contralateral hemisphere [20,65,66]. There are several possible reasons for this discrepancy, including the different experimental systems (spinal cord motor neurons in culture vs. central neurons in vivo). Importantly, the in vivo experiments used cleavage of SNAP25 as a detection method for long-range BoNT/ E trafficking, which requires not only axonal transport, but appropriate somatic sorting of these carriers, transcytosis and entry into a compartment which enables the translocation of the L chain into the cytoplasm. Previous work with neurotrophins has shown that some retrogradely transported cargoes, such as NGF, undergo lysosomal degradation, whilst others (e.g. BDNF) are released at synaptic sites, where they can affect second-order neurons [16,54]. It is conceivable that BoNT/A and BoNT/E might undergo differential sorting events at the cell soma, which could impact on their ability to undergo transcytosis. Another non-mutually exclusive possibility is the preferential degradation of the L chain of BoNT/E due to ubiquitination and proteasome targeting [62]. This is particularly relevant to the in vivo system, where BoNT/E trafficking has been examined in long-distance projecting neurons [20,65,66]. In this case, the rapid degradation of BoNT/E would not allow the accumulation of detectable amounts of truncated SNAP25 at distal sites. Conversely, the prolonged catalytic activity of BoNT/A [20,62,67,68] would enable the occurrence of longdistance effects and the detection of truncated SNAP25.
Our demonstration of retrograde transport of BoNT/A in spinal cord motor neurons may have implications for the analysis of the central effects of this neurotoxin in the clinic. BoNT/A and B are used for the treatment of many human pathologies characterised by hyperactivity of nerve terminals and hypersecretory syndromes [7,10]. The clinical benefits depend mainly on a localised neuromuscular blockade, but there is substantial evidence for central effects of BoNT/A, which could contribute to the overall therapeutic efficacy [69][70][71]. These central effects may depend either on neuronal plasticity, or on a direct BoNT/A activity on central synapses. Our data provide evidence in favour of such direct action.
The ability of H C A and H C E to undergo retrograde trafficking holds promise for the development of novel drug delivery vehicles for the targeting of therapeutics to the central nervous system. A similar approach was previously applied to H C T, which has been exploited for the delivery of various molecules to central neurons [72,73]. The translation of H C T derivatives to clinical practice is not straightforward however, due to the presence of circulating antibodies directed against TeNT in most individuals as a result of the widespread vaccination against tetanus in industrialised countries. In this context, the implementation of BoNT H C -based carriers might overcome this limitation and provide a novel class of drug delivery systems.

Ethics statement
All experiments were carried out following the guidelines of the H C A (residues 860-1296) and H C E (residues 820-1252) were expressed as GST-fusion proteins in E. coli BL21 [74] and after purification, dialyzed against 20 mM HEPES-NaOH pH 7.4, 1 M NaCl. H C T was prepared as previously described [22]. In selected experiments, a shorter version of H C T (H C T441, residues 875-1315) fused to an improved cysteine-rich tag [25], which has a longer shelf life, was used. BoNT/A was prepared and tested as previously described [19,20,34,75], whilst TeNT was from Lubio [76]. Purified full length BoNTs were labelled with AlexaFluor488 (green) or 568 (red) according to the manufacturer's instructions. The moles of dye per mole of BoNT averaged 6.0. Fluorescent H C s and BoNTs were dialysed against HEPES-NaOH 10 mM pH 7.4, 150 mM NaCl before use.

Neuronal cultures
Spinal cord motor neurons were prepared from 14.5 day old rat embryos (Sprague-Dawley, Charles River) [77] or 13.5 day old mouse embryos [78,79], and plated onto poly-L-ornithine and laminin-coated glass coverslips, MatTek dishes or MFCs, and cultured at 37uC and 7.5% CO 2 . Motor neurons were used for experiments starting from day in vitro 5 (DIV5) until DIV10. For binding studies, motor neurons were pre-cooled on ice, washed with 0.2% BSA in Hanks' buffer, and incubated with AlexaFluor-labelled H C s (7.5-15 nM) or full length BoNTs (15 nM) for 15 min. MNs were then washed in PBS and fixed in 4% paraformaldehyde (PFA) containing 20% sucrose. For endocytosis assays, motor neurons were incubated at 37uC with fluorescent H C s (7.5-15 nM) or full length BoNTs (15 nM) for 30 min in resting (NaCl 137 mM, KCl 5 mM, MgCl 2 1 mM, CaCl 2 2.5 mM, glucose 10 mM, HEPES-NaOH 5 mM, pH 7.4) or depolarising (NaCl 80 mM, KCl 60 mM, MgCl 2 1 mM, CaCl 2 2.5 mM, glucose 10 mM, HEPES-NaOH 5 mM, pH 7.4) conditions. Neurons were then cooled on ice, washed with acidic buffer (100 mM citrate-NaOH, 140 mM NaCl, pH 2.0) for 5 min at room temperature in order to remove the probe still bound to the cell surface, washed with PBS and fixed. In selected cases, motor neurons (DIV6) were pre-treated with 2 nM BoNT/D (Wako) for 22 h at 37uC, whereas controls were left untreated.
MFCs were produced using established methods [35,36]. Polydimethylsiloxane (Dow Corning) inserts were sterilised and fixed to 50 mm glass-bottomed WillCo dishes (IntraCel) using plasma cleaning. MFCs were blocked with 0.8% BSA in PBS overnight at 37uC and then coated with poly-L-ornithine and laminin. Motor neurons were plated in the somatic compartment of the MFC and left to adhere before the full medium was applied. Experiments were performed between DIV7-10 when axons crossed the microgrooves and reached the distal side of MFCs.
Colocalisation was quantified by applying a mask corresponding to the whole neuronal network obtained using an antibody against bIII-tubulin (Tuj1). Fluorescence intensity of H C s and SV2 isoforms was determined in resting and stimulating conditions in the area of the mask using ImageJ. The level of colocalisation was assessed by measuring the Mander's coefficient on randomly chosen fields. Statistical significance was calculated using Mann-Whitney test.

Axonal retrograde transport assays
Motor neurons plated onto MatTek glass-bottom dishes or onto MFCs were incubated with fluorescent H C A and E as described above. In selected experiments, motor neurons were co-incubated with 50 nM LysoTracker or with an antibody against the extracellular domain of p75 NTR receptor (1:1,000 dilution). After incubation for 30 min at 37uC under resting or stimulating conditions, cells were washed with E4 imaging medium containing 30 mM HEPES-NaOH, pH 7.4 and imaged by time-lapse confocal microscopy at 37uC. Images were acquired every 4 s over a total of up to 200 frames using a Zeiss LSM 510 confocal microscope equipped with a Zeiss 63X, Phase 3 Plan Apochromat oil-immersion objective and controlled by Zeiss LSM 510 software.
Carriers were tracked manually using Motion Analysis Software (Kinetic Imaging). Single-movements between two consecutive frames were measured to determine the speed of the carrier. Only moving carriers that could be followed for a minimum of four consecutive frames were analysed, and tracking was stopped when the organelle went out of focus or stopped for the remaining observation time. The distance covered by a carrier between two consecutive frames, termed single movement, was used to determine its instantaneous speed. A double-positive compartment was defined on the basis of the following criteria: i) the carrier was labelled in two different channels; ii) the morphology of the carrier was very similar in the two channels; and iii) its speed and direction was identical in the two channels for at least 4 time points in a time-series. Statistical analysis and curve fitting were performed using Kaleidagraph (Synergy Software). Kymographs were generated using MetaMorph (Molecular Devices) after rotation of the image stack to align the neuronal process vertically. Horizontal single line-scans through the thickness of each process were plotted sequentially for every frame in the time series.
Colocalisation of double-positive carriers was quantified by MetaMorph using ''manually-count objects'' options. For this, all H C -positive carriers were manually marked and automatically counted, then the other channel (i.e. Lysotracker) was overlaid and double-positive carriers were highlighted and counted. Student t-test was performed using Kaleidagraph.

In vivo experiments
Adult Long-Evans rats (35 in total) were kept on a 12 h light/ dark cycle and had access to food and water ad libitum. Animals were anaesthetized with isoflurane and injections of BoNT/A (1 nM, 0.5 ml; n = 12) or TeNT (3 nM, 0.5 ml, n = 11) were performed with a microsyringe on the right side of the snout at the centre of the whisker pad (i.e., between rows B and C of the vibrissae) [20,81,82]. Three naive control animals were also included. Brains were dissected out at 1, 3 or 10 d and 500 mm thick coronal sections were cut through the brainstem with a microtome (Leica). The facial nucleus ipsilateral to the injection site was microdissected and immediately frozen.
For the behavioural analysis, whisker movements were monitored for each animal before injection (baseline), and 1 and 3 d following toxin delivery (n = 6). Control naïve rats (n = 3) were monitored with an identical schedule. Each rat was placed in a clear plexiglas cylinder [83] and filmed for 3 minutes. Total time spent during whisking (injected side) was calculated offline for each movie. Statistical significance was calculated with two-way ANOVA followed by Holm-Sidak test.
Another group of animals were anaesthetized with isoflurane and a small skin incision was performed to expose the tibialis anterior and the gastrocnemius muscles [84]. Muscles were injected with BoNT/A (1 nM, 1 ml, n = 4 rats). After 10 d spinal cords were dissected and lumbar segments were taken and immediately frozen for western blotting. Figure S1 Quantification of the colocalisation of H C A and HcE with SV2 isoforms A and C under resting and depolarising conditions. H C A does not show any preference between SV2A and C in primary rat motor neurons (A). In contrast, H C E colocalises significantly more with SV2C in both resting and depolarising conditions (Mann-Whitney test; **, p,0.01, ***, p,0.001) (B). However, the colocalisation is not complete for both H C s and does not change upon depolarisation. The study reported in this figure was performed using at least two independent primary motor neuron cultures. At least ten fields were analysed for each conditions. Quantification reported here is from a representative experiment. Movie S1 H C A undergoes fast retrograde axonal transport in primary motor neurons in MFC. Motor neurons grown in MFC were incubated with fluorescent H C A for 30 min at 37uC in resting conditions, then washed and imaged by timelapse confocal microscopy. The toxin was added exclusively to the axonal side. Cell body is out of view on the right. Frames were taken every 3 s and the movie plays at 5 frames/s. This movie is a representative example of experiments performed on two independent primary motor neuron cultures, replicated three times. (MP4)

Supporting Information
Movie S2 H C A shares fast retrogradely transported organelles with the neurotrophin receptor p75 NTR . Motor neurons seeded in MFC were incubated for 30 min at 37uC with fluorescent H C A (red) and a fluorescently-labeled antibody against p75 NTR (green) in resting conditions, then washed and imaged. Both H C A and the antibody have been added to the axonal side only of the MFC. The two channels are merged in the movie. Yellow structures indicate double positive organelles. Cell body is out of view on the right. Frames were taken every 3 s and played at 5 frames/s. This movie is a representative example of experiments performed on two independent primary motor neuron cultures. (MP4) Movie S3 H C A undergoes fast retrograde transport in non acidic organelles. Motor neurons cultures were incubated with fluorescent H C A (green) and Lysotracker (red) for 30 min at 37uC under depolarising conditions (60 mM KCl). Cells were then washed and imaged. H C A undergo fast retrograde transport (from left to right) in non acidic organelles as demonstrated by the almost total lack of colocalisation with Lysotracker. Cell body is out of view on the right. Frames were taken every 4 s and played at 5 frames/s. This movie is a representative example of experiments performed on two independent primary motor neuron cultures, replicated twice. (MP4) Movie S4 H C E undergoes fast retrograde transport in non acidic organelles. Motor neurons cultures were incubated with fluorescent H C E (green) and Lysotracker (red) for 30 min at 37uC under depolarising conditions (60 mM KCl). Cells were then washed and imaged. H C E is retrogradely transported in organelles lacking Lysotracker. As discussed in the text, some of the carriers containing H C E (green) show a discontinuous transport in the retrograde direction (from left to right), whereas some others undergo anterograde transport. Cell body is out of view on the right. Frames were taken every 4 s and played at 5 frames/s. This movie is a representative example of experiments performed on two independent primary motor neuron cultures, replicated twice.