Translocation and dissemination to target neurons of botulinum neurotoxin type B in the mouse intestinal wall

Botulinum neurotoxins (BoNTs) are responsible for severe flaccid paralysis (botulism), which in most cases enter the organism via the digestive tract and then disseminate into the blood or lymph circulation to target autonomic and motor nerve endings. The passage way of BoNTs alone or in complex forms with associated nontoxic proteins through the epithelial barrier of the digestive tract still remains unclear. Here, we show using an in vivo model of mouse ligated intestinal loop that BoNT/B alone or the BoNT/B C‐terminal domain of the heavy chain (HCcB), which interacts with cell surface receptors, translocates across the intestinal barrier. The BoNT/B or HCcB translocation through the intestinal barrier occurred via an endocytosis‐dependent mechanism within 10–20 min, because Dynasore, a potent endocytosis inhibitor, significantly prevented BoNT/B as well as HCcB translocation. We also show that HCcB or BoNT/B specifically targets neuronal cells and neuronal extensions in the intestinal submucosa and musculosa expressing synaptotagmin, preferentially cholinergic neurons and to a lower extent other neuronal cell types, notably serotonergic neurons. Interestingly, rare intestinal epithelial cells accumulated HCcB suggesting that distinct cell types of the intestinal epithelium, still undefined, might mediate efficient translocation of BoNT/B.


Introduction
Clostridium botulinum produces potent neurotoxins [botulinum neurotoxins (BoNTs)] that are responsible for severe neuroparalytic illness (botulism) in man and animals resulting from inhibition of spontaneous and nerve-evoked acetylcholine (ACh) release at cholinergic nerve endings. BoNTs are synthesized as an inactive single protein (~150 kDa) that is proteolytically cleaved into a~100 kDa heavy chain (HC) and a~50 kDa light chain. Both chains remain linked by a disulfide bridge. The di-chain molecule constitutes the active neurotoxin. The half C-terminal part of HC (HCc) recognizes specific receptors on the surface of target neuronal cells and is involved in driving the toxin entry pathway into cells, whereas the N-terminal part permits the translocation of the L chain into the cytosol. Light chain catalyses a zinc-dependent proteolysis of one or two of the three proteins of the SNARE complex, which play an essential role in neurotransmitter exocytosis (Meunier et al., 2002;Poulain et al., 2008;Bercsenyi et al., 2013;Simpson, 2013;Rossetto et al., 2014).
Botulinum neurotoxins are divided into seven toxinotypes (A to G) according to their immunological properties based on neutralization with polyclonal antibodies. Each toxinotype is specifically neutralized only with the corresponding antibodies (Hill and Smith, 2013). A new toxinotype called H or rather a new hybrid F/A type has been recently reported but still needs further characterization (Dover et al., 2014;Gonzalez-Escalona et al., 2014). C. botulinum strains show genetic variations and are classified into four groups (I to IV) based on 16S rRNA gene sequences and biochemical characteristics. Most of the strains produce only one BoNT type. Genetic diversity is also observed in bont genes, and multiple subtypes (or genetic variants) have been identified in each BoNT toxinotype (reviewed in Hill and Smith, 2013).
Three forms of botulism are recognized in humans in natural conditions: foodborne botulism, infant botulism or botulism by intestinal colonization and wound botulism. Foodborne botulism is due to the ingestion of preformed BoNT in contaminated food, and it is the main form of botulism in adults. In contrast, in infant botulism and some adult cases, ingested C. botulinum spores develop in the intestinal content and produce the toxin in situ (Tacket and Rogawski, 1989;Sobel, 2005). In both forms, foodborne botulism and botulism by intestinal colonization, BoNT escapes the gastrointestinal tract to reach the target cholinergic nerve endings, possibly through the blood and lymph circulation (Maksymowych et al., 1999). Previous observations using experimental models of animal intoxination have shown that following oral administration, BoNT enters the blood stream and lymph circulation (Maksymowych et al., 1999). The upper small intestine was found to be the primary site of toxin absorption (Kitamura et al., 1969;Sugii et al., 1977;Bonventre, 1979;Fujinaga et al., 1997). However, BoNT can also be absorbed from the other parts of the digestive tract including the buccal cavity, stomach and colon, but to a lower extent than in the upper small intestine (Bonventre, 1979;Sakaguchi, 1983;Maksymowych et al., 1999). Therefore, BoNT translocation through the intestinal barrier and trafficking to the cholinergic nerve ending target represent the initial and critical steps of botulinum intoxination.
The mechanism of BoNT passage through the intestinal epithelial barrier remains partially unknown. Botulinum complexes formed by BoNT and associated nontoxic proteins (ANTPs) including haemagglutinin (HA) components have reported to pass across the epithelial barrier upon HA-mediated opening of intercellular junctions. Indeed, HA subunits have been shown to bind to epithelial cadherin (E-cadherin) and to disrupt the intercellular junctional complexes between epithelial cells (Jin et al., 2009;Sugawara et al., 2010;Sugawara and Fujinaga, 2011;Lee et al., 2014;Lam et al., 2015). However, purified BoNT has also been shown to undergo transcytosis through intestinal epithelial cells without the help of HAs (Fujinaga et al., 1997;Maksymowych and Simpson, 1998;Park and Simpson, 2003;Maksymowych and Simpson, 2004;Nishikawa et al., 2004;Ahsan et al., 2005;Couesnon et al., 2008). BoNT/A seems to preferentially use specific subsets of intestinal cells such as crypt enteroendocrine cells to enter the intestinal mucosa (Couesnon et al., 2012). In this study, we investigated the passage of BoNT/B in the mouse intestinal mucosa because it is a major cause of foodborne botulism and infant botulism in various countries (Fox et al., 2005;Peck, 2009;Malaska, 2014;Mazuet et al., 2014).

Entry of HCcB into the mouse intestinal mucosa
Recombinant HCc from BoNT/B (HCcB) labelled with Cy3 has been used to monitor BoNT/B entry into mouse intestinal mucosa. HCc domain of clostridial neurotoxins (BoNTs and tetanus neurotoxin), which specifically interacts with cell receptors, has already been extensively used to investigate internalization and intracellular trafficking of the neurotoxins in neuronal cells, as well as toxin translocation through intestinal epithelial cells (Lalli et al., 2003;Maksymowych and Simpson, 2004;Bohnert and Schiavo, 2005;Roux et al., 2006;Harper et al., 2011;Restani et al., 2012;Lam et al., 2015). Similar strategy was performed in the analysis of BoNT/A passage through the mouse intestinal mucosa (Couesnon et al., 2012). Therefore, fluorescent HCcB was injected into ligated ileum loops of anaesthetized mice. The animals were then killed at various time intervals, and the intestinal loops were washed, fixed and prepared for confocal microscopy observations. As shown in Fig. 1A (upper panel), limited HCcB fluorescence was observed in the intestinal lumen between the villi after 5 min incubation. Strikingly, rare epithelial cells of the villi accumulated fluorescent HCcB (Fig. 1A, middle panel). A few filamentous structures labelled with HCcB were also detected in some intestinal villi and in some submucosa areas (Fig. 1A, lower panel). After 10 and 20 min incubation, labelled HCcB was visualized in the inter-villi spaces, in the lumen of intestinal crypts, and on the surface of some intestinal villi, but no HCcB staining was evidenced in crypt epithelial cells (Fig. 1B,upper and middle panels;Fig. 1C,upper panel). Again, only very few cells of the villus epithelium were stained with HCcB ( Fig. 1B and 1C upper panels). Filamentous elements stained with HCcB were observed in the bottom of intestinal villi surrounding intestinal crypts (Fig. 1B,middle panel,and Fig. 1C,upper and middle panels). A dense network of filamentous structures in the musculosa was stained with fluorescent HCcB (Fig. 1B, lower panel; Fig. 1C, upper and lower panels). In addition, fluorescent filament structures were also evidenced inside the core of the villi (Fig. 1A, middle panel; Fig. 1B, upper panel). The most marked observations were the abundant filament labelling in the submucosa and to a larger extent in the musculosa after 10 and 20 min incubation of the ileal loops with fluorescent HCcB (Fig. 1B, lower panel; Fig. 1C, upper and lower panels). Albeit the in vivo experiments at different incubation time periods did not allow to determine a precise kinetics of the toxin passage through the intestinal barrier, they permitted to evidence the progressive translocation of HCcB lasting 5 to 20 min from the intestinal lumen to target structures in the submucosa and musculosa. This finding was further supported by the quantification of the HCcB fluorescence intensity showing HCcB accumulation in villi overtime, and entry into the submucosa/musculosa at 10-20 min (Fig. 1C). HCcB-Cy3 (100 μg) in Dulbecco's Modified Eagle's Medium was injected into ligated jejuno-ileal loop of anaesthetized mice. The intestinal loop was then introduced again into the mouse abdomen. After 5 (A), 10 (B) and 20 min (C), the intestinal loop was washed, fixed, sliced and prepared for fluorescent confocal microscopy. The preparations were co-stained with Hoechst (blue) to visualize nuclei. A. At 5 min incubation, HCcB was visualized (red labelling) on the surface and/or inside some villous epithelial cells, as well as in the inter-villous spaces. No or weak HCcB labelling was observed in the submucosa. B. HCcB was observed inside intestinal crypts, in a few villous epithelial cells, and in filamentous structures in the villi and around intestinal crypts. HCcB markedly stained filamentous structures in the submucosa and musculosa. C. At 20 min, the HCcB staining of filamentous structures in the submucosa and musculosa was prominent. Scale bars = 10 μm. D. Quantification of HCcB fluorescence in the villi and submucosa/musculosa of intestinal loops. Values are mean ± standard deviation from three independent experiments (three to five quantifications in each experiment). ***P < 0.001.

Dynasore impairs the translocation of HCcB into the intestinal mucosa
Botulinum neurotoxin A has been evidenced to cross the epithelial barrier through a transcytotic mechanism in the absence of ANTPs (Maksymowych and Simpson, 1998;Maksymowych and Simpson, 2004;Ahsan et al., 2005;Couesnon et al., 2008). Previous work also suggests that BoNT/B is able to use a similar transport (Maksymowych and Simpson, 1998). To investigate whether HCcB enters the mouse intestinal mucosa in in vivo conditions through a similar transcytotic mechanism, Dynasore, a specific inhibitor of endocytosis that blocks dynamin function (Harper et al., 2013), was injected (10 μg) into an ileal loop of anaesthetized mice 20 min prior the injection of fluorescent HCcB. Mice were then sacrificed 10 min after the administration of fluorescent HCcB. As controls, we checked that administration of Dynasore and/or HCcB did not induced morphological alteration of the intestinal mucosa as monitored by E-cadherin staining of intestinal epithelial cells ( Fig. 2A and 2B). In ileal loops, which only received fluorescent HCcB, an intense intracellular staining of rare intestinal epithelial cells and of filamentous elements in the mucosa and musculosa was observed ( Fig. 2A). In contrast, in the ileal loops pretreated with Dynasore, fluorescent HCcB showed a different staining pattern. Thus, HCcB staining was mostly observed on the surface of intestinal villi and in the inter-villi spaces, whereas only a few filamentous structures were labelled in the submucosa and musculosa (Fig. 2B). No morpholog-ical alteration of the basolateral epithelial cell membranes was evidenced as monitored by E-cadherin immunostaining ( Fig. 2A and 2B, middle panels). Quantification of fluorescence intensity in the submucosa and musculosa indicated that Dynasore impaired by almost 90% the transcytosis of fluorescent HCcB across the intestinal barrier (Fig. 2C). The significant Dynasore-induced inhibition of HCcB staining of target cells in the submucosa and musculosa strongly supports a transcytotic passage of HCcB through the intestinal epithelial barrier.

Endocytosis-dependent entry of BoNT/B into the mouse intestinal mucosa
Experiments were performed with purified BoNT/B to further assess that the results obtained with HCcB were representative of the trafficking of the whole toxin. For this, ligated mouse intestinal loops were injected with purified BoNT/B (100 μg) and were then processed for imaging studies as described earlier. After 10 min administration, BoNT/B was detected with immunopurified polyclonal antibodies directed against the HCc domain. BoNT/B was visualized on the surface of some villi after 10 min toxin administration as shown in Fig. 3A (upper panel) and decorated filamentous structures in the villi, submucosa and musculosa. BoNT/B was mostly localized in the submucosa and musculosa after a 20 min incubation time period (Fig. 3A, lower panels). Thereby, a similar pattern of staining was observed between HCcB and purified BoNT/B. HCcB-Cy3 (100 μg) was injected into a control ligated jejuno-ileal loop (A) or into a ligated intestinal loop pretreated with Dynasore (10 μg, 20 min) (B). After 10 min HCcB incubation, the intestinal segments were prepared as described in Fig. 1. The HCcB staining (red) of filamentous structures in the submucosa and musculosa (A) was prevented by pretreatment with Dynasore (B). E-cadherin staining (A and B middle panels, white) showed no morphological alteration of the intestinal epithelium. The preparations were costained with Hoechst (blue) and ECDD2 (white). C. Quantification of HCcB fluorescence in the submucosa and musculosa of intestinal loops pretreated or not with Dynasore and injected intraluminally with fluorescent HCcB. Fluorescence from intestinal loops treated with HCcB only was set to 100%. Values are mean ± standard deviation from three independent experiments (three to five quantifications in each experiment). ***P < 0.001. Scale bars = 10 μm.
Next, we checked whether BoNT/B like HCcB enters the intestinal mucosa through a transcytotic mechanism. Pretreatment of intestinal loop with Dynasore (10 μg) for 20 min significantly reduced the detection of BoNT/B in the submucosa and musculosa (Fig. 3C). No morphological alteration of the intestinal mucosa was observed in the intestinal loops treated with Dynasore and/or BoNT/B (not shown). Compared with intestinal loops treated with BoNT/B in the absence of Dynasore, only a few cells and filament structures were labelled in the submucosa and musculosa from Dynasore-pretreated intestinal loops (Fig. 3C). Dynasore induced an 80% decrease in the BoNT/B staining in both submucosa and musculosa (Fig. 3D), further supporting that similarly to HCcB, BoNT/B uses a transcytotic pathway to pass through the intestinal epithelial barrier. In addition, these results support the view that HCcB and BoNT/B use a similar traffic pathway in the mouse intestinal mucosa.

HCcB and BoNT/B target neuronal cells in mouse intestinal submucosa and musculosa
Imaging analyses with antibodies against neurofilaments (NF) were performed to identify the cells and filamentous structures recognized by fluorescent HCcB in the submucosa and musculosa. Fluorescent HCcB was localized on long cell extensions co-labelled with anti-NF antibodies ( Fig. 4A and B). However, not all but only certain cell extensions were recognized by both HCcB and anti-NF antibodies ( Fig. 4A and B). Partial colocalization between the two markers (Pearson's coefficient 0.38 ± 0.06) suggested that HCcB did not bind directly to NF, but rather bound to cells and cell extensions containing NF.
We took advantage of Thy1-yellow fluorescent protein (YFP) transgenic mouse in the C57BL6 genetic background, which has been widely used to analyse the morphology and function of neurons in the mouse brain (Feng et al., 2000;Fig. 3. BoNT/B enters the mouse intestinal mucosa and its entry is prevented by Dynasore. BoNT/B (100 μg) was injected into ligated jejuno-ileal loop of anaesthetized mice. After the indicated times, the intestinal segments were prepared as described in Fig. 1, and BoNT/B was detected with rabbit immunopurified immunoglobulins against HCcB and Alexa 594-anti-rabbit immunoglobulins. A. After 10 min incubation, BoNT/B was visualized in a few villous epithelial cells, as well as in filament structures inside the villi and to a lower extent in the submucosa. B. At 20 min incubation, a marked BoNT/B staining of filamentous structures in the submucosa and musculosa was evidenced. Epithelium brush border was visualized by actin staining with phalloidin (green) and nuclei with Hoechst (blue). C. Pretreatment with Dynasore (10 μg, 20 min) followed by 10 min incubation with BoNT/B significantly prevented BoNT/B staining in the submucosa and musculosa. D. Quantification of BoNT/B fluorescence in the submucosa and musculosa of intestinal loops pretreated or not with Dynasore and injected intraluminally with BoNT/B. Data are mean values ± standard deviation, from three independent experiments (three to five quantifications in each experiment). *** P < 0.001. Scale bars = 10 μm. Vuksic et al., 2008), to visualize neurons of the enteric nervous system (ENS). Thy1-YFP is mainly expressed in the Golgi apparatus of certain subpopulations of neurons (Feng et al., 2000). In the intestinal submucosa and musculosa of Thy1-YFP mice, neuronal cells were discontinuously labelled, to some extent because of the accumulation of YFP around the nuclei (Fig. 5A). Some neuronal cells showed an intense YFP expression in the cytosol in cell bodies and extensions. Fluorescent HCcB injected into the intestinal lumen of Thy1-YFP mice co-stained YFP-labelled cells and cell extensions after 10 min after injection (Fig. 5A). However, not all but only some subpopulations of YFP-expressing neurons were recognized by HCcB (Fig. 5A). The low Pearson's coefficient (0.241 ± 0.11) between YFP and Cy3-HCcB indicates that the two markers labelled distinct structures of the same cells.
Synaptotagmin (Syt) has been evidenced as the receptor protein part that in combination with gangliosides of the GD1b and GT1b series forms the high-affinity receptor of BoNT/B. Indeed, BoNT/B binds to the intraluminal domain of SytI and SytII through the HCc domain (Nishiki et al., 1996;Dong et al., 2003;Rummel et al., 2007;Rummel, 2013). Antibodies against SytII labelled cells and cell extensions in mouse intestinal submucosa and musculosa, which were colabelled with YFP in Thy1-YFP mice (Fig. 5A). Similar results were obtained with anti-SytI antibodies (not shown). Fluorescent HCcB specifically recognized cells and cell extensions expressing both Thy1 and SytII (Fig. 5A). This supports that HCcB targeted specific neurons expressing SytII in mouse intestine. The low Pearson's coefficient (0.24 ± 0.11) between HCcB staining and YFP expression also indicates that HCcB did not colocalize with YFP molecules but recognized cells expressing YFP-Thy1. In contrast, the high level of colocalization between HCcB and SytII (Pearson's coefficient: 0.51 ±0.11) supports that SytII is a specific BoNT/B receptor in intestinal neuronal cells.
Glial cells are abundant in the nerve tissues as well as in the ENS where they show multiple extensions, mainly in the submucosal and myenteric plexus, and in close contact with neuronal cells (Yu and Li, 2014). Glial cells from the mouse intestinal mucosa were specifically visualized with antibodies against the glial fibrillary acidic protein (GFAP) (Fig. 5B). Because glial cells are wound around neuronal cells, an apparent high Pearson's coefficient (0.35 ± 0.03) was observed between HCcB-Cy3-stained and GFAP-stained cells (Fig. 5B). Indeed, Fig. 4. HCcB stains neuronal cell extensions in mouse submucosa and musculosa. HCcB-Cy3 (100 μg) was injected into ligated jejuno-ileal loop, and after 10 min incubation, the intestinal samples were prepared as described in Fig. 1. Neuronal cell extensions were labelled with anti-neurofilament (NF) antibodies, and nuclei with Hoechst (blue). Scale bars = 20 and 10 μm for the magnification panels. glial cell extensions are in close contact with those of neuronal cells, but they are shorter. HCcB labelling pattern of cell extensions correlated with that of NF staining but was distinct from that obtained with GFAP staining (Fig. 5B). These findings suggest that HCcB entered the intestinal mucosa and targeted specific neuronal cells and extensions in the submucosa and musculosa and did not interact with glial cells in the mouse intestine.

HCcB targets cholinergic neurons in the mouse intestine
Botulinum neurotoxins are known to interact with cholinergic neurons and to block spontaneous and evoked quantal ACh release (Schiavo et al., 2000;Poulain et al., 2008;Rossetto et al., 2014). However, BoNTs are able to enter various neuronal cell types and to inhibit the release of a large range of other neurotransmitters such as ACh, glutamate and gamma-aminobutyric acid (GABA) (reviewed in Dolly et al., 2009;Popoff and Poulain, 2010).
First, imaging studies were carried out to test whether HCcB targets cholinergic neurons using antibodies directed again choline acetyltransferase (ChAT). Most of the ChAT-immunoreactive neuronal cell extensions from the submucosa and musculosa were co-labelled with HCcB (82.2% ± 3.8) 10 min after its administration in the intestinal loop (Fig. 6A, Table 1). ChAT neuronal cells labelled with HCcB also contained SytII (Fig. S1A). Similar results were obtained with whole BoNT/B injected into the intestinal lumen and detected with specific immunopurified antibodies (Fig. 7A). After 10 min incubation, a large proportion of ChAT-immunoreactive neuronal cells of the submucosa and musculosa were co-stained with anti-BoNT/B antibodies (87% ± 2.5) (Fig. 7A). However, it is noteworthy that HCcB or BoNT/B also labelled other neuronal cell types than ChAT neuronal cells, but to a lower extent.

Distinct neuronal cells recognized by HCcB in the mouse intestine
Binding of BoNT/B to the diverse neuronal cell types in mouse intestine was investigated by colocalization analysis of HCcB-Cy3 with specific neuronal cell markers. Only a low number of neuronal cells were stained with Fig. 5. HCcB recognizes neuronal cell extensions expressing synaptotagmin II but not glial cells in mouse intestinal submucosa and musculosa. HCcB-Cy3 (100 μg) was injected into ligated jejuno-ileal loop of transgenic Thy1-YFP mice. After 10 min incubation the intestinal samples were prepared as described in Fig. 1. A. HCcB stained neuronal cells and cell extensions expressing Thy1-YFP (green) and co-stained with anti-synaptotagmin II antibodies (white). B. No co-staining of glial cells monitored with anti-GFAP antibodies and fluorescent HCcB was evidenced. Nuclei were stained with Hoechst (blue). Scale bars = 20 and 10 μm for magnification panels.
anti-serotonin (5-hydroxytryptamine) antibodies in the submucosa and myenteric plexuses of the mouse intestine. After 10 min administration of HCcB or BoNT/B into the ileum lumen, a significant proportion (25-35%) of serotonin-immunoreactive cells was co-labelled with fluorescent HCcB or anti-BoNT/B antibodies ( Fig. 6B and 7B, Table 1). It is noteworthy that the serotoninimmunoreactive cells labelled with HCcB also expressed SytII (Fig. S1B).
The other neuronal cell types are underrepresented in ENS and were variably recognized by HCcB-Cy3. Indeed, a low proportion estimated to be about 10% of vasoactive intestinal peptide (VIP)-immunoreactive cells in the submucosa was co-stained with HCcB (Fig. 6C, Table 1). Furthermore, rare glutamate-immunoreactive and GABAimmunoreactive cells stained with antibodies against vesicular glutamate (V-GLUT) and GABA transporters (V-GAT), respectively, were visualized in the intestinal HCcB-Cy3 (100 μg) was injected into ligated jejuno-ileal loop, and after 10 min incubation, the intestinal samples were prepared as described in Fig. 1. The preparations were co-stained with anti-ChAT (marker of cholinergic neurons), anti-serotonin, anti-VIP, anti-V-GLUT (marker of glutamatergic neurons) and anti-V-GAT (marker of GABAergic neurons) (green). Enlarged views of the drawn squares are shown on the right side. Hoechst (blue), ECCD2 (white). Scale bars = 20 and 10 μm for magnification panels. submucosa, and most of them were partially co-stained with fluorescent HCcB (Fig. 6D-E, Table 1). Interestingly, HCcB labelled filamentous structures inside the core of villi, which were co-stained mostly with antibodies against ChAT (52.5 ± 2.9%) and to a lower extent with VIP antibodies (5 ± 2.5%) (Fig. 8). No serotoninimmunoreactive neuronal extensions were visualized in the intestinal villi. These structures recognized by HCcB are likely projections of intrinsic afferent neurons whose cell bodies are localized in the submucosa and which are mainly ChAT-immunoreactive and to a lower extent VIPimmunoreactive neurons (Furness et al., 2004).

Intestinal cells involved in the passage of BoNT/B
Subsequently to administration of fluorescent HCcB in the intestinal lumen, certain cells of villous epithelium were intensely fluorescent. Albeit HCcB was observed inside the lumen of most intestinal crypts, no crypt cells retained HCcB fluorescence (Fig. 1). Similar findings were observed with BoNT/B injected into the intestinal lumen and detected with BoNT/B (100 μg) was injected into mouse ligated jejuno-ileal loop, and after 10 min incubation, the intestinal samples were prepared as described in Fig. 1. BoNT/B was detected with rabbit immunopurified immunoglobulins against HCcB and Alexa 594-anti-rabbit immunoglobulins. The preparations were co-stained with anti-ChAT (A) and anti-serotonin (B). Enlarged views of the drawn squares are shown on the right side. Hoechst (blue). Scale bars = 20 and 10 μm for magnification panels.
anti-HCcB antibodies (Fig. 2, Fig. 3C and 3D). The question arises as to whether the rare cells from the villous intestinal epithelium accumulating HCcB fluorescence reflect the transcellular translocation of HCcB and BoNT/B or alternatively represent a site of toxin storage. Because these cells were mostly located in the upper part of intestinal villi, we first investigated whether they undergo apoptosis and thereby might non-specifically accumulate HCcB or BoNT/B. As shown in Fig. S2, no co-staining between fluorescent HCcB and activated Caspase-3 antibodies was evidenced, indicating that the fluorescent HCcB-stained cells were not apoptotic.
The inhibitory effect of Dynasore on the transport of HCcB or BoNT/B across the intestinal epithelium (Figs 2 and 3B) suggests an apical endocytic uptake and subsequent transcytotic delivery of HCcB or BoNT/B through the basolateral side. Colchicine, which is a tubulin filament disruption agent, impairs transcytosis in epithelial cells such as Caco-2 cells (Bose et al., 2007) and has been used to inhibit the translocation of Listeria monocytogenes across the intestinal epithelium in an intestinal ligated loop model (Nikitas et al., 2011). Pretreatment of mouse intestinal loops with colchicine in the same conditions as for the L. monocytogenes experiments (10 μg ml À1 , 20 min) (Nikitas et al., 2011) prior to the injection of fluorescent HCcB induced a significant increased number of villous epithelial cells accumulating HCcB ( Fig. 9A and 9B). In addition, the HCcB fluorescence in intestinal loops pretreated with colchicine was mainly localized at the periphery of the epithelial cells that accumulated HCcB, whereas in the nonpretreated intestinal loops, the epithelial cells that retained HCcB were uniformly stained, possibly resulting from a different traffic pathway than in colchicine-treated cells ( Fig. 9A and 9B). However, colchicine pretreatment did not significantly prevent HCcB staining of neuronal cells in the submucosa and musculosa (Fig. 10), indicating a delayed passage of HCcB through certain villous cells, independently of the one inhibited by colchicine.
The cells, which accumulated fluorescent HCcB, were not stained with phalloidin, a marker of actin filaments (Fig. 9A). This finding strongly suggests that HCcB accumulated in a subset of intestinal cells devoid of apical HCcB-Cy3 (100 μg) was injected into ligated jejuno-ileal loop, and after 10 min incubation, the intestinal samples were prepared as described in Fig. 1. Preparations were co-stained with antibodies anti-ChAT (cholinergic cells) or anti-VIP (green), and anti E-cadherin (ECCD2) (white) and Hoechst (blue). Scale bars = 20 and 10 μm in the magnification panels. brush border and raises the question whether BoNT/B uses specific cells for its transport through the intestinal barrier. To address this question, we investigated whether the HCcB-labelled cells belong to a specific subpopulation of the intestinal epithelium, using various markers of already characterized cell subsets of the intestinal epithelium: wheat germ agglutinin, which binds specifically to sialic acid and N-acetylglucosaminyl carbohydrate residues in mucous of goblet cells (Jang et al., 2004); Lectin Ulex europeus agglutinin type 1, which recognizes Paneth cells (Garabedian et al., 1997)  HCcB-Cy3 (100 μg) was injected into ligated jejuno-ileal loop with or without pretreatment with colchicine (10 μg ml À1 , 30 min). Albeit colchicine induced HCcB-Cy3 accumulation into certain intestinal epithelial cells, it did not completely impaired HCcB-Cy3 dissemination to neuronal cells in the submucosa and musculosa. Hoechst (blue). Scale bar = 10 μm. Fig. 9. HCcB accumulation in some cells of the intestinal epithelium. HCcB-Cy3 (100 μg) was injected into ligated jejuno-ileal loop, and after 10 min incubation, the intestinal samples were prepared as described in Fig. 1 enteroendocrine cells, Paneth, cells and goblet cells (Gerbe et al., 2011). Additional markers of goblet cells were used like antibodies specific of the intestinal trefoil factor or cytokeratin18 (Poulsom and Wright, 1993;Hesse et al., 2007). Double immunofluorescence studies using the various cell-specific antibodies or markers and fluorescent HCcB clearly showed that none of the markers tested colocalized or co-stained with HCcBlabelled cells in the intestinal epithelium (Figs S3 and S4). Furthermore, the rare cells of the villous epithelium exhibiting ChAT immunoreactivity were not targeted by HCcB (Fig. S3). Transcytosis through intestinal cells indicates that HCcB or BoNT/B recognizes specific cell surface receptor(s). Are BoNT/B receptors on intestinal cells the same as those on neuronal cells (gangliosides of the GD1b/GT1b series in combination with SytII)? An excess of GT1b has been found to prevent BoNT/B binding to synaptosomes (Atassi et al., 2014). A competition experiment was performed between HCcB and GT1b for binding and entry of toxin into the intestinal mucosa. As shown in Fig. 11, in contrast to control intestinal loop treated with only labelled HCcB, preincubation of HCcB with GT1b significantly prevented HCcB staining of intestinal villi, submucosa and musculosa. Only a faint HCcB staining was observed in certain villi, whereas no staining was visualized in the submucosa and musculosa. As control, GM1 was found to not impair HCcB entry into the intestinal mucosa (Fig. S5). Moreover, no labelling of intestinal epithelial cells with anti-SytII antibodies was evidenced, suggesting that BoNT/B uses GT1b but a distinct protein receptor from that of neuronal cells to enter intestinal cells.

Discussion
Translocation of BoNT through the intestinal barrier is a critical initial step in botulism resulting from the ingestion of toxin preformed in food or from intestinal colonization by C. botulinum. In in vitro cultures, food or intestinal content, BoNT is produced in complex forms by non-covalent association with ANTPs. Whether whole botulinum complexes or only BoNT passes through the intestinal barrier and the mode of its transport remain under debate. ANTPs play an essential role in the protection of BoNT against the stomach acidic pH and digestive proteases. The nontoxic non-HA component of C. botulinum type A is structurally related to BoNT/A and associates with BoNT/A in a pHdependent manner to form a medium-sized botulinum complex highly resistant to acidic pH and protease degradation (Gu et al., 2012;Gu and Jin, 2013). It is noteworthy that nontoxic non-HA is synthesized by all C. botulinum types and likely retains the same function in the various corresponding botulinum complexes. HAs, which are found in various BoNT complex types, have been evidenced to mediate the absorption of botulinum complex from the gut. HAs interact with oligosaccharides on the intestinal epithelial cells and possibly facilitate the toxin uptake into intestinal cells (reviewed in Fujinaga et al., 2013;Gu and Jin, 2013). The HA-mediated transport of BoNT/A complex through the intestinal barrier seems to preferentially occur via M cells (Matsumura et al., 2015). In addition, HAs bind to and disrupt E-cadherin-mediated intercellular junctions, thus allowing the passage of BoNT through the intestinal barrier via the paracellular route (reviewed in Fujinaga et al., 2013;Gu and Jin, 2013). However, HAs are not produced by C. botulinum E, F and certain type A strains, which are also involved in human botulism by the oral route (Simpson, 2013;Singh et al., 2014). These C. botulinum strains form BoNT complexes containing OrfX proteins instead of HAs (Hill and Smith, 2013;Popoff et al., 2013). However, OrfX proteins have not been found to interact with intestinal cells indicating that BoNT can itself cross the intestinal barrier without the help of additional protein. Therefore, when administrated into the small intestine, BoNT free of HAs and BoNT complexes are absorbed through the intestinal barrier in an equally efficient manner (Maksymowych et al., 1999). Moreover, purified BoNT was demonstrated to bind to intestinal cells, and to undergo receptor-mediated endocytosis, transcytosis, and subsequent release from the basolateral side (Maksymowych and Simpson, 1998;Maksymowych and Simpson, 2004;Ahsan et al., 2005;Couesnon et al., 2009). Here, we investigated the entry of BoNT/B into the intestinal mucosa in an in vivo mouse model using purified BoNT/B or the recombinant HCcB fragment. The corresponding C-terminal domain of BoNT/A has already been used to monitor the holotoxin trafficking in cultured cells or intact intestinal mucosa (Maksymowych and Simpson, 2004;Ahsan et al., 2005;Couesnon et al., 2012).
When administrated into a mouse jejuno-ileal ligated loop, a progressive passage of the fluorescent HCcB or BoNT/B from the intestinal epithelium to the submucosa and musculosa was visualized. Within 10-20 min, HCcB or BoNT/B targeted neuronal structures in the submucosa and musculosa. This rapid passage contrasts with the longer time (30-60 min) observed for the internalization of HCcA in the mouse intestinal mucosa (Couesnon et al., 2012). However, the experimental conditions used for HCcA investigation were slightly different. Mouse intestinal loops were excised, transferred into oxygenated culture medium and incubated at room temperature after intraluminal injection of HCcA (Couesnon et al., 2012). In the present work, the vascularization of the intestinal loops and the physiological mouse temperature were maintained. Thereby, in the more physiological conditions, a more rapid trafficking was observed indicating that BoNT/B translocation into the mouse intestinal mucosa is a rapid process. However, the difference in the experimental procedures does not fully exclude that BoNT types A and B may have a different way and kinetics of entry into the intestinal mucosa.
Interestingly, Dynasore, which is an inhibitor of dynamin-dependent endocytosis process (Macia et al., 2006;Harper et al., 2013), significantly prevented the entry of HCcB or BoNT/B into the mouse intestinal mucosa (Figs 2 and 3). The in vivo mouse model of ligated intestinal loop clearly supports that BoNT/B free of HAs or the receptor binding domain, HCcB, was able to pass through the intestinal barrier via an endocytic mechanism to target specific cells in the intestinal submucosa and musculosa. It should be noticed that the toxin amounts used in this study were higher than those expected in natural acquired botulism. Minimum amounts of fluorescent HCcB yielding detectable signal in intestinal tissues were around 10 μg per intestinal loop (Fig. S6). We selected to use higher amounts (100 μg/intestinal loop) to more efficiently visualize the cells mediating toxin trafficking. Because high HCcB or BoNT/B doses were efficiently inhibited by endocytosis inhibitor (Fig. 2 and 3) or in competition experiments with specific receptor such as GT1b (Fig. 10), and because only a restricted number of cell types were visualized (Figs 1 and 3), the physiological toxin trafficking was likely preserved in these experimental conditions, although lower amounts of HCcA (0.5 μg per intestinal loop) (Couesnon et al., 2012) or BoNT/A (0.6 μg orally per mouse) (Lam et al., 2015) were used. However, we were unable to reproduce the results with 0.6 μg BoNT/A per mouse. Moreover, it has to be taken into account that BoNT/B is less active than BoNT/A, about 10-fold less (Foran et al., 2003;Rasetti-Escargueil et al., 2009), and that only a low toxin fraction passes through the intestinal barrier based on in vivo experiments (reviewed in Popoff and Connan, 2014), therefore justifying the amounts used in this study. However, one cannot fully exclude that in the present work; HCcB or BoNT/B uses different entry pathways than in the in vivo situation.
HCcB or BoNT/B enters the mouse intestinal mucosa and specifically targets neuronal cell bodies and neuronal cell extensions in the submucosa and musculosa as visualized by co-staining of the filamentous structures with anti-NF antibodies and HCcB or BoNT/B (Fig. 4). Glial cells, which are abundant in the intestinal mucosa, were not recognized by HCcB (Fig. 5). These observations were further supported by the use of the transgenic Thy1-YFP mouse, which exhibits a nice labelling of the ENS, and co-staining of neuronal filaments expressing Thy1-YFP with HCcB (Fig. 5). It is noteworthy that HCcB or BoNT/B only recognizes certain neuronal cell extensions. Neuronal cell filaments identified by anti-NF antibodies or by Thy1-YFP expression were not all co-stained with HCcB (Fig. 5). Interestingly, HCcB and BoNT/B colocalized with SytII antibodies on intestinal neuronal extensions (Fig. 5, Fig. S1) in agreement with previous characterization of SytII as the BoNT/B receptor (Nishiki et al., 1996;Dong et al., 2003;Chai et al., 2006). ChAT and serotonin neurons labelled with HCcB also expressed SytII (Fig. S1), and likely the other neuronal cell types targeted by BoNT/B. Collectively the present findings support that BoNT/B uses SytII as specific receptor on neuronal cell extensions of the ENS.
Neuronal cells of the ENS are organized in the submucosa and myenteric plexuses, which project on the different layers of the intestinal mucosa. ENS neurons can synthesize a wide variety of neurotransmitters. More than 20 types of neurotransmitters have been identified in ENS, and most enteric neurons may produce and secrete several of them (Goyal and Hirano, 1996;Grundy and Schermann, 2006;Furness, 2012). Cholinergic neurons are the greater population of enteric neurons. In the submucosal plexus, cholinergic neurons (about 55% of the enteric neurons) regulate ion secretion and are involved in local sensory pathways in response to the luminal content composition. In addition, they interact with Peyer's patch follicles (Kulkarni-Narta et al., 1999;Furness et al., 2004;Harrington et al., 2010). Cholinergic neurons are preponderant in the myenteric plexus (about 80%) and have a main function in mediating muscle activity and controlling intestinal motility (Furness et al., 2004;Harrington et al., 2010). Serotonin is a major mediator of gastrointestinal function. Notably, serotonin controls motility, secretion and sensory responses. Serotonin is mainly produced by enteroendocrine cells from the mucosal epithelium of the intestine, and 2 to 20% of enteric neurons are serotonin immunoreactive (McLean et al., 2007;Mawe and Hoffman, 2013). VIP-immunoreactive neurons are the most preponderant non-cholinergic neurons of the ENS and constitute the main class of inhibitory interneurons. VIP-immunoreactive neurons are estimated to represent about 45% of neurons of the submucosal plexus in the ileum and jejunum. VIP modulates several functions including vasodilatation of intestinal vessels, intestinal smooth muscle relaxation and stimulation of electrolyte secretion and water (Brookes, 2001;Furness et al., 2004;Hernandes et al., 2004;Igarashi et al., 2011;Furness, 2012). Glutamate and GABA are neurotransmitters produced by a few number of ENS neuronal cells (Furness, 2000;Furness et al., 2004).
In in vivo experiments, HCcB or BoNT/B targeted certain but not all neuronal cells at the Syt binding sites. Cholinergic neurons of ENS were the main target of HCcB or BoNT/B, because more than 80% of the ChATimmunoreactive neuronal cells were co-stained with HCcB or BoNT/B (Table 1). This is in agreement with the fact that BoNT/B like the other BoNT types mainly targets peripheral cholinergic motor neuron endings (Simpson, 2013;Rossetto et al., 2014). HCcA was also found to preferentially localize to cholinergic neurons of the submucosa and musculosa of the mouse intestine (Black and Dolly, 1987;Couesnon et al., 2012). HCcB also targeted cholinergic extensions in the villi and around the intestinal crypts. Constipation is a frequent symptom (about 70%) in foodborne botulism and is a major and early event in botulism resulting from intestinal colonization (Arnon et al., 2001;Sobel, 2005;Brook, 2007;Mitchell and Tseng-Ong, 2008). Because cholinergic neurons are the main players of the control of intestinal motility and secretion (Furness, 2012), ChAT-immunoreactive targeting by BoNT likely accounts for the reduced intestinal peristalsis and secretion observed in botulism. During botulism by intestinal colonization, BoNT produced in the intestinal lumen might be absorbed in a higher local concentration able to enter the intestinal mucosa and inhibit the underlying cholinergic neurons compared with orally ingested toxin, which disseminates more broadly through the digestive tract.
HCcB or BoNT/B targeted other neuronal cell types than cholinergic neurons in the mouse intestinal mucosa but to a lower extent. Serotonin-immunoreactive neuronal cells were significantly co-stained with HCcB or BoNT/B (25 to 35%) (Table 1), whereas a lower proportion (about 10%) ( Table 1) of VIP-immunoreactive neurons was stained with HCcB. In the present work, only rare V-GLUTimmunoreactive and V-GAT-immunoreactive cells were visualized in the intestinal mucosa, and most of them costained with HCcB. However, the low number of identified cells did not allow estimating precisely the proportion of these sub-cell types targeted by HCcB.
The pathophysiological effects of BoNT on the noncholinergic neurons are still a matter of debate. Serotonin is an important neurotransmitter in the gastrointestinal tract, which is involved in multiple functions including intestinal motility and secretion. Although most part of serotonin is secreted by enteroendocrine cells, serotonergic neurons have an essential role in the gastrointestinal motility (Li et al., 2011;Gershon, 2013). BoNT/B effects on serotonergic neurons of the intestinal mucosa might contribute to the inhibition of intestinal motility in synergy with the blockade of ACh release at cholinergic nerve endings. The incidence of BoNT/B on the small part of VIPimmunoreactive neurons targeted by HCcB remains speculative. Because one of the main roles of VIP is to control gastrointestinal secretion (Igarashi et al., 2011;Furness et al., 2014), BoNT/B may exert its antisecretory effect partially through inhibition of VIP release. BoNT/B might use non-cholinergic neurons not only to contribute to the local paralytic effects but also to disseminate to other target neurons locally or at distance from the intestine. Indeed, BoNT/A has been shown to use a retrograde transport and transcytosis to migrate from the peripheral neurons to the central nervous system (Antonucci et al., 2008;Restani et al., 2011;Restani et al., 2012). Because the neurons of ENS are highly interconnected between them and with neurons of the central nervous system (Goyal and Hirano, 1996;Grundy and Schermann, 2006;Furness et al., 2014), one can speculates that BoNT/B uses non-cholinergic intestinal neurons for its transport to remote target neurons.
The present findings also raise an intriguing question: does BoNT/B use preferential cells to cross the intestinal barrier? In contrast to BoNT/A, which has been found to enter the intestinal mucosa preferentially via crypt enteroendocrine cells (Couesnon et al., 2012), no HCcB or BoNT/B labelling was detected in epithelial crypt cells, or enteroendocrine cells using anti-chromogranin A antibodies. Accumulation of HCcB or BoNT/B was observed in a few villous epithelial cells lacking apical actin staining (e.g. devoid of apical brush border). The fact that cells accumulating HCcB were not apoptotic cells rules out the possibility of non-specifically fluorescent protein accumulation. Using a wide panel of markers of intestinal epithelial and lymphocytic cells, the villous cells labelled with HCcB or BoNT/B could not be assigned to any already known subset of cell type. Whether these cells represent the toxin passage through the intestinal barrier remains questionable. A transcytotic passage supposes a continuous transcellular toxin passage over time. However, a transient accumulation in some intracellular compartment is not fully excluded. Time lapse monitoring the passage of fluorescent HCcB could be more informative. The in vivo experiments used in this study allowed collection of intestinal samples only at different time points. Pretreatment with colchicine, an inhibitor of microtubule-dependent transcytosis, evidenced an increased number of villous epithelial cells accumulating HCcB, but with a distinct pattern of fluorescent labelling than in colchicine-untreated intestinal loops (Fig. 9). HCcB was mainly observed on the basolateral peripheries of epithelial cells from pretreated intestinal loops with colchicine, whereas HCcB accumulated broadly into the intracellular space in non-pretreated loops. However, colchicine did not prevent significantly the HCcB passage to the underlying neuronal cells. Thus, it cannot be excluded that colchicine delayed, but not completely blocked, the transcytotic passage of HCcB in these subsets of villous cells. The absorption of BoNT from the digestive tract is a very low efficient process. In vivo experiments showed that less than 0.01% to 0.1% of intraduodenally administrated BoNT can be recovered in lymph of blood circulation (reviewed in Popoff and Connan, 2014). Thus, a low-toxin passage through the enterocytes or at least some of them may support the weak intestinal absorption of toxin, which has been observed in vivo.
In conclusion, the present study using an in vivo model of mouse ligated intestinal loop demonstrates that HCcB or BoNT/B free of HAs can pass through the intestinal barrier via an endocytosis-dependent mechanism because Dynasore, a specific inhibitor of endocytosis, significantly prevented their translocation. After 10-20 min incubation time, HCcB or BoNT/B specifically targeted neuronal cells and neuronal extensions in the submucosa and musculosa. Binding of HCcB or BoNT/B to neurons correlated with Syt expression, in agreement with the identification of Syt as the specific BoNT/B receptor (Nishiki et al., 1996;Dong et al., 2003;Rummel et al., 2007;Rummel, 2013). Cholinergic intestinal neurons were the main targets of HCcB or BoNT/B together with other neuronal cell types of the intestinal mucosa, notably serotonergic neurons and to a lower extent VIPimmunoreactive, V-GLUT-immunoreactive and V-GATimmunoreactive neurons. Although HCcB or BoNT/B accumulated in some intestinal cells, those that mediate the passage of BoNT/B remain to be identified.

Ethic statements
All experiments were performed in accordance with French and European Community guidelines for laboratory animal handling.
The protocols of experiments were approved by Pasteur Institute CETEA (Comité d'Ethique en Expérimentation Animale) with the agreement of laboratory animal use (N°2013-0118).

Animals
Adult Swiss mice (Charles River) and Thy-1-YFP transgenic mice (C57BL6 background, Jackson Research) were used.

Botulinum neurotoxin and recombinant HCcB protein production
Botulinum neurotoxin B was produced and purified as previously described (Shone and Tranter, 1995). Recombinant His-tag HCc fragment of BoNT/B was produced and purified from pET-28ac(+) vector containing DNA encoding for HCcB cloned into BamHI and SalI sites, as previously described (Tavallaie et al., 2004).
HCcB His-tag was labelled with Amersham Cy3 Mono-Reactive Dye Pack (Ge Healthcare) according to the manufacturer's recommendations. Free dye is removed from labelled protein using the de Zeba Spin Desalting Columns according to the manufacturer's recommendations (Thermo Scientific).

In vivo intestinal ligated loop experiment
Swiss mice (between 20 and 22 g) or Thy-1-YFP transgenic mice were fasted for 16 h before surgery. Mice were deeply anaesthetized with a mixture of ketamine (50 mg kg À1 body weight; Imalgene 1000; Merial) and medetomidine (0.5 mg kg À1 body weight; Domitor; Orion Corporation). A mouse laparotomy was performed, and a jejuno-ileal loop (approximately 4 cm long) was isolated. Three-hundred microlitre of labelled HCcB or purified BoNT/B containing 100 μg protein was injected into the intestinal lumen. After incubation times (5 to 20 min), mice were euthanatized, and the intestinal loops were harvested, opened longitudinally, washed repeatedly in 37°C Dulbecco's Modified Eagle's Medium (Invitrogen) and fixed flat (luminal side up) for 2 h in 4% paraformaldehyde at room temperature.
In competition assay, Cy3-HCcB (70 μg) was preincubated with a 10-fold excess of ganglioside GT1b from bovine brain (Sigma G3767) for 20 min at room temperature, and the mixture was injected into an intestinal loop. After 15 min incubation, the intestinal loops were processed for fluorescence microscopy.

Intestinal tissue preparations and indirect immunofluorescence
Fixed intestinal loops were embedded in 4% agarose, and sections (150 μm) were cut with a vibratome (Polyscience -Bangs Laboratories). Sections were permeabilized with 0.4% triton X-100 for 1 h at room temperature, washed in phosphate buffered saline (PBS) and then incubated for 1 h in PBS-bovine serum albumin (3%). Samples were then incubated or not with specific antibodies for 2 h at room temperature for primary antibodies and 1 h at room temperature for secondary antibodies in PBS-bovine serum albumin (0.5%) -Triton (0.2%). After three washes in PBS, sections were mounted in Fluoromount (FluorProbes), and samples were imaged using a Zeiss LSM700 confocal laser scanning microscope.

Image analysis and statistics
Image analysis was performed using IMAGEJ (National Institutes of Health). Colocalization test were performed using JacOp plugin from IMAGEJ. Values throughout the text are expressed as means ± standard deviations. Differences were assessed using unpaired Student's t-test. Statistical significance was assumed for P < 0.0001 (***) on at least three independent experiments for integrated density measurements and a 100 counted cells for the percentage of co-labelled structures.
(green) or anti-cytokeratin18 (green), as well as anti E cadherin (ECCD2) (white) and Hoechst (blue). Scale bars = 10 mm. Fig. S5. The ganglioside GM1 does not impair HCcB entrys into the intestinal mucosa. HCcB-Cy3 (70 mg) was injected into ligated jejunoileal loop (A) or incubated with a 10-fold excess of GM1 for 20 min at room temperature prior to injection into intestinal loop (B). After 15 min incubation the intestinal samples were prepared as described in Fig. 1.   Fig. S6. Visualization of fluorescent HCcB (10 mg) in the mouse intestine following its injection into the intestinal lumen. HCcB-Cy3 (10 mg) in Dulbecco's modified Eagle's medium was injected into ligated jejuno-ileal loop of anaesthetized mice. After 20 min incubation the intestinal samples were prepared as described in Fig. 1. Preparations were co-stained with antibodies anti E-cadherin (ECCD2) and Hoechst (blue). Scale bars = 10 mm.