Construction of hybrid proteins that migrate retrogradely and transynaptically into the central nervous system

Abstract

The nontoxic proteolytic C fragment of tetanus toxin (TTC peptide) has the same ability to bind nerve cells and be retrogradely transported through a synapse as the native toxin. We have investigated its potential use as an in vivo neurotropic carrier. In this work we show that a hybrid protein encoded by the lacZ–TTC gene fusion retains the biological functions of both proteins in vivo—i.e., retrograde transynaptic transport of the TTC fragment and β-galactosidase enzymatic activity. After intramuscular injection, enzymatic activity could be detected in motoneurons and connected neurons of the brainstem areas. This strategy could be used to deliver a biological activity to neurons from the periphery to the central nervous system. Such a hybrid protein could also be used to map synaptic connections between neural cells.

Tetanus toxin is a potent neurotoxin of 1,315 amino acids that is produced by Clostridium tetani (1, 2). Tetanus toxin prevents the inhibitory neurotransmitter release from spinal cord interneurons by a specific mechanism of cell intoxication (for review, see ref. 3). This pathological mechanism has been demonstrated to involve retrograde axonal and transynaptic transport of the tetanus toxin. The toxin is taken up by nerve endings at the neuromuscular junction but does not act at this site; rather, the toxin is transported into a vesicular compartment and travels along motor axons for a considerable distance until it reaches its targets. The transynaptic movement of tetanus toxin was first demonstrated by autoradiographic localization in spinal cord interneurons after injection into a muscle (4). However, previous studies of transynaptic passage of tetanus toxin from motoneurons were limited by the rapid development of clinical tetanus and death of the experimental animal (4–6).

The C fragment of tetanus toxin obtained by protease digestion, the TTC fragment, has been shown to be transported by neurons in a similar manner to that of the native toxin without causing clinical symptoms (7–10). A recombinant TTC fragment was reported to possess the same properties as the fragment obtained by protease digestion (11). The fact that an atoxic fragment of the toxin molecule was able to migrate retrogradely within the axons and to accumulate into the central nervous system (CNS) led to speculation that such a fragment could be used as a neurotrophic carrier (12). A TTC fragment chemically conjugated to various large proteins was taken up by neurons in tissue culture (13) and by motor neurons in animal models (12, 14, 15). In a more recent in vitro study, the human CuZn superoxyde dismutase, SOD-1, fused to the TTC fragment was internalized by neurons and retained some of its biological functions (16).

In this report, we demonstrate that the hybrid protein produced from a lacZ–TTC gene fusion can be retrogradely transported across a synapse in vivo. The recombinant TTC fragment could therefore be used as an efficient carrier to deliver foreign biological activities from the periphery into neurons of the CNS. This strategy could also be exploited to analyze synaptic activity in neural networks and allow their in vivo mapping.

MATERIALS AND METHODS

Plasmid Constructions.

TTC cloning. Full-length TTC DNA was generated from the genomic DNA from the Clostridium tetani strain (a gift from M. Popoff, Pasteur Institute) using PCR. Three overlaping fragments were synthesized: PCR1 of 465 bp (primer 1, 5′-CCC CCC GGG CCA CCA TGG TTT TTT CAA CAC CAA TTC CAT TTT CTT ATT C-3′; and primer 2, 5′-CTA AAC CAG TAA TTT CTG-3′), PCR2 of 648 bp (primer 3, 5′-AAT TAT GGA CTT TAA AAG ATT CCG C-3′; and primer 4, 5′-GGC ATT ATA ACC TAC TCT TAG AAT-3′), and PCR3 of 338 bp (primer 5, 5′-AAT GCC TTT AAT AAT CTT GAT AGA AAT-3′; and primer 6, 5′-CCC CCC GGG CAT ATG TCA TGA ACA TAT CAA TCT GTT TAA TC-3′). The three fragments were sequentially introduced into pBluescript KS+ (Stratagene) to give pBS:TTC plasmid. The upstream primer 1 also contains an optimized eukaryotic ribosome binding site and translational initiation signals. The DNA sequence of all PCR products was identical to that of native TTC DNA (11).

pGEX:lacZ–TTC:

pGEX:lacZ was obtained by cloning a SmaI/XhoI lacZ fragment from the pGNA vector (a gift from H. Le Mouellic, Pasteur Institute), into pGEX 4T-2 (Pharmacia). PCR was used to convert the lacZ stop codon into an NcoI restriction site. Two primers (upstream, 5′-CTG AAT ATC GAC GGT TTC CAT ATG-3′; and downstream, 5′-GGC AGT CTC GAG TCT AGA CCA TGG CTT TTT GAC ACC AGA C-3′) were used to amplify the sequence between NdeI and XhoI, generating pGEX:lacZ(NcoI) from pGEX:lacZ. pGEX:lacZ–TTC was obtained by insertion of the TTC NcoI/XhoI fragment into pGEX:lacZ(NcoI), fusing TTC immediately downstream of the lacZ coding region and in the same reading frame.

Purification of the Hybrid Protein.

The Escherichia coli strain SR3315 (a gift from A. Pugsley, Pasteur Institute) transfected with pGEX:lacZ–TTC was used for protein production. An overnight bacterial culture was diluted 1:100 in Luria–Bertani medium containing 100 μg/ml ampicillin, and grown for several hours at 32°C until an OD of 0.5 was reached. Induction from the Ptac promoter was achieved by the addition of 1 mM isopropyl β-d-thiogalactoside and 1 mM MgCl₂ and a further 2 hr of incubation. The induced bacteria were pelleted by centrifugation for 20 min at 3,000 rpm, washed with PBS, and resuspended in lysis buffer containing 0.1 M Tris (pH 7.8), 0.1 M NaCl, 20% glycerol, 10 mM EDTA, 0.1% Triton X-100, 4 mM DTT, 1 mg/ml lysosyme, and a mixture of anti-proteases (100 μg/ml Pefablok/1 μg/ml leupeptin/1 μg/ml pepstatin/1 mM benzamidine). After cell disruption in a French Press, total bacterial lysate was centrifuged for 10 min at 30,000 rpm. The resulting supernatant was incubated overnight at 4°C with the affinity matrix glutathione-Sepharose 4B (Stratagene) with slow agitation. After centrifugation for 5 min at 3,000 rpm, the matrix was washed three times with the same lysis buffer but without lysosyme and glycerol, and then three times with PBS. The resin was incubated overnight at 4°C with Thrombin (10 units/ml; Sigma) in PBS to cleave the β-galactosidase (β-gal)–TTC fusion protein from the glutathione S-transferase sequence and thereby elute it from the affinity column. Concentration of the eluted fusion protein was achieved by centrifugation in centricon X-100 tubes (Amicon; 100,000 molecular weight cutoff membrane).

Purified hybrid protein was analyzed by Western blotting after electrophoretic separation in 8% acrylamide SDS/PAGE under reducing conditions followed by electrophoretic transfer onto nitrocellulose membranes (0.2 mm porosity; Bio-Rad). Immunodetection of blotted proteins was performed with a Vectastain ABC-alkaline phosphatase kit (Vector Laboratories) and diaminobenzidine color development. Antibodies were used as follows: rabbit anti-β-gal antisera (Cappel Laboratories), dilution 1:1,000; rabbit anti-TTC antisera (Calbiochem), dilution 1:20,000. A major band with a relative molecular mass of 180 kDa corresponding to the β-gal–TTC hybrid protein was detected with both anti-β-gal anti-TTC antibodies.

Binding and Internalization of Recombinant Protein in Differentiated 1009 Cells.

The 1009 cell line was derived from a spontaneous testicular teratocarcinoma arising in a recombinant inbred mouse strain (129 × B6) (17). The 1009 cells were grown in DMEM containing 10% fetal calf serum and passaged at subconfluence. In vitro differentiation with retinoic acid and cAMP was performed as described (18). Eight days after retinoic acid treatment, cells were used for the internalization experiments with either the hybrid protein or β-gal.

Binding and internalization of the β-gal–TTC fusion were assessed using a modified protocol (16). Differentiated 1009 cells were incubated for 2 hr at 37°C with 5 μg/ml of β-gal–TTC or β-gal protein diluted in binding buffer (0.25% sucrose/20 mM Tris acetate/1 mM CaCl₂/1 mM MgCl₂/0.25% BSA in PBS). The cells were then incubated with 1 μg/ml Pronase E (Sigma) in PBS for 10 min at 37°C, followed by washing with proteases inhibitors diluted in PBS (100 μg/ml Pefablok/1 mM benzamidine).

The cells were fixed with 4% formalin in PBS for 10 min at room temperature and then washed extensively with PBS. β-Gal activity was detected on fixed cells by an overnight staining at 37°C in 5-bromo-4-chloro-3-indolyl β-d-galactoside (X-Gal) solution (0.8 mg/ml X-Gal/4 mM potassium ferricyanide/4 mM potassium ferrocyanide/4 mM MgCl₂ in PBS). For electron microscopy, the cells were further fixed in 2,5% glutaraldehyde for 18 hr and then processed as described (19).

For immunohistochemical labeling, cells were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature (RT), then washed extensively with PBS, followed by a 1-hr incubation at RT with 2% BSA/0.02% Triton X-100 in PBS. Cells were coincubated in primary antibodies diluted in 2% BSA/0.02% Triton X-100 in PBS for 2 hr at RT. Antibodies used were a mouse anti-neurofilament antibody (NF, 200 kDa; dilution 1:50; Sigma) or the rabbit anti-TTC antibody (dilution 1:1,000). The labeling was visualized using fluorescent secondary antibodies: Cy3, goat anti-rabbit IgG (dilution 1:500; Amersham) or anti-mouse IgG with extravidin–fluorescein isothiocyanate (dilution 1:200; Sigma). Cells were mounted in moviol and visualized with epifluorescence.

In Vivo Recombinant Protein Injections.

Fourteen-week-old B6D2F1 mice were obtained from Iffa-Credo. The animal’s tongue muscle was injected with 3% Avertin (15 μl/g of animal) using an Hamilton syringe (20 μl per animal) while under general anesthesia. The protein concentration was 0.5–5 μg/μl in PBS; therefore mice received approximately 10–100 μg per injection. Animals were kept alive for 12–48 hr postinjection to permit migration of the injected protein, and in no case were any tetanus symptoms detected. The mice were killed by intracardiac perfusion with 4% paraformaldehyde in PBS while under deep anesthesia. Brains were harvested, rinsed in PBS, and incubated in 15% sucrose overnight at 4°C, then mounted in tissue-tek before sectioning into 15-μm-thick slices using a cryostat.

Histology, Immunohistology, and X-Gal Staining.

For in toto X-Gal staining of the dissected brain and tongue, mice (10 animals) were killed and fixed as described above. The brain was further cut with a scapel along a median plane and directly incubated for 12 hr in X-Gal solution.

For immunohistology, sections were incubated in a 1:5,000 dilution of anti-TTC antibody in 2% BSA/0.02% Triton X-100 in PBS overnight at 4°C after nonspecific antibody binding sites were blocked by a 1-hr incubation in the same buffer. Antibody detection was carried out using the Vectastain ABC alkaline phosphatase kit with diaminobenzidine color development. For X-Gal staining, sections were incubated in X-Gal solution and counterstained for 30 sec with hematoxylin [1:5 (vol/vol)] in PBS. Histology on adjacent sections was done after X-Gal staining, using a 30-sec incubation in hematoxylin/thionin solution. All sections were mounted in moviol before light microscopy analysis.

RESULTS

Internalization of the β-Gal–TTC Fusion Protein by Neurons in Vitro.

Differentiation of 1009 cells with retinoic acid and cAMP in vitro yields neuronal and glial cells (18, 20). X-Gal staining or immunolabeling were performed after incubation with the β-gal–TTC fusion protein or with either the β-gal or TTC proteins alone. Only when the hybrid protein was incubated with differentiated 1009 cells was a strong X-Gal staining detected in cells having a neuronal phenotype. No signal was detected when β-gal alone was incubated under the same conditions (Fig. 1 A and B). A similar X-Gal staining pattern was obtained after pronase treatment of the cells to remove surface bound proteins, indicating that the hybrid protein had been internalized. The intracellular localization of the hybrid protein was further confirmed by electron microscopic analysis of X-Gal-stained cells (Fig. 1C). Furthermore, the enzymatic activity observed in axons seemed to be localized in vesicles associated with filaments (Fig. 1C), which is in agreement with previous work on TTC fragment or native tetanus toxin (14, 21, 22). Colabeling with anti-TTC and anti-neurofilament antibodies revealed that β-gal activity colocalized with TTC fragment in neuronal cells (Fig. 1D). No glial cells were labeled with either antibody.

Figure 1.

In vitro uptake of β-gal–TTC hybrid protein and control β-gal by 1009 neural cells. Eight days after the start of retinoic treatment, differentiated 1009 cells were incubated at 37°C for 2 hr with 5 μg/ml of each protein and fixed. X-Gal staining was performed after incubation with β-gal–TTC (A) or β-gal (B). Strong labeling was specifically detected in neural cells. (C) Electron micrograph of X-Gal-stained neural cell axon showing intracellular localization of the fusion protein: arrows indicate the precipitate probably associated with filaments. After β-gal–TTC incubation, immunodetection was performed with anti-neurofilament and anti-TTC antibodies (D): colabeling showing that β-gal–TTC was only uptaken in neuronal cells. [Bars = 100 μm (A, B, and D) and 1 μm (C).]

Retrograde Transport of the Hybrid Protein in Vivo.

To study the behavior of the β-gal–TTC protein in vivo, we chose to test the hybrid protein in a well-characterized neuronal network, the hypoglossal system. After intramuscular injection of β-gal–TTC protein into the mouse tongue, the distribution of the hybrid protein in the CNS was analyzed by X-Gal staining. Various dilutions of the protein were injected and sequential time points were analyzed to permit protein transport into hypoglossal motoneurons (XII) and its further transneuronal migration into connected second order neurons.

A well-defined profile of large, apparently retrogradely labeled neurons was clearly evident in the hypoglossal structure, analyzed in toto at 12 hr postinjection (Fig. 2 A and B). A strong labeling was also apparent in the hypoglossal nerve (XIIn) of the tongue of the injected mice (Fig. 2 C and D). At the level of muscle fibers, button structures were observed that might reflect labeling of neuromuscular junctions where the hybrid protein was internalized into nerve axons (Fig. 2E). These data demonstrate that the β-gal–TTC hybrid protein can migrate rapidly by retrograde axonal transport as far as motoneuron cell bodies, after prior uptake by nerve terminals in the tongue. This specific uptake and the intra-axonal transport are similar to the properties that have been described for the native toxin (6, 21, 23).

Figure 2.

Retrograde labeling following β-gal–TTC intramuscular injection in the tongue of mice. A total of 100 μg of protein was injected, and observations were made 12 hr (A, B, C, D, and E) and 18 hr (F, G, and H) postinjection. In toto X-Gal staining on brain clearly showed the hypoglossal structure (XII) with its two nuclei retrogradely labeled (A and B). The hypoglossal nerve (XIIn) was also intensively stained (C), as was its arborization (E, see arrows). Putative neuromuscular junctions (D) with button-like structure (black arrows) and the terminal arborization of axons (white arrows) are visible. (F and G with different magnification) Histological analysis of brain slices after X-Gal staining. The hybrid protein is localized in XII motoneurons cytoplasm. (H) Immunodetection with anti-TTC antibody showing a colocalization with the β-gal activity. [Bars = 1 mm (A and C), 100 μm (B and D–H).] cb, Cerebellum; sc, spinal cord; MN, motoneuron.

Transport of the hybrid protein was examined in greater detail by analyzing X-Gal-stained brain sections. Motoneurons of the hypoglossal nucleus became labeled rapidly, with 12 hr being the earliest time point examined. Most of the label was confined to neuronal somata, the cell nuclei being unlabeled. The intensity of the labeling depends upon the concentration of the β-gal–TTC protein injected: when 10 μg of protein was injected, only the hypoglossal somata were detected, whereas with 25–50 μg of protein, a fuzzy network of dendrites was visualized; transynaptic transfer was detected with 100 μg of hybrid protein. An identical distribution of label was observed when brain sections were immunostained with an anti-TTC antibody, demonstrating that β-gal and TTC fragment colocalize within cells. Finally, injection of β-gal alone did not result in labeling of the hypoglossal nuclei and therefore confirms that transport of the hybrid protein is TTC-dependent. In our hands, labeling with an anti-TTC antibody was less informative than detection of β-gal activity; for instance, the nerve pathway to the brain could not be visualized by anti-TTC immunostaining (Fig. 2H). At 18 hr postinjection, labeling was observed in the hypoglossal nuclei: all motoneuron cell bodies and the most proximal part of their dendrites were very densely stained (Fig. 2 F–H). In contrast, no labeling was ever detected in glial cells adjoining XII motoneurons or their axons. Our results are in accordance with others who reported an identical pattern of immunolabeling after injection of the TTC fragment alone (9). Transneuronal transfer is detectable after 24 hr. An additional 24 hr and beyond did not yield a different staining.

Transneuronal Transport of the Hybrid Protein.

Second-order interneurons, as well as higher-order neurons that synapse with the hypoglossal motoneurons, have been extensively analyzed using conventional markers, such as the wheat germ agglutinin–horseradish peroxidase complex or neurotropic viruses such as α-herpes (24) and rhabdoviruses (25). An exhaustive compilation of regions in the brain that synaptically connect to the hypoglossal nucleus has also been described (25). In our study, the distribution of the β-gal–TTC fusion depended on the initial concentration of protein injected into the muscle and the time allowed for transport after injection. Up to 24 hr postinjection, labeling was restricted to the hypoglossal nuclei. After 24 hr, the distribution of second-order transneuronally labeled cells in various regions of the brain was consistent and reproducible; however, we have not yet analyzed higher-order connections, such as in cortical areas. Even at longer time points (e.g., 48 hr), labeling of the hypoglossal nucleus remained constant. At higher magnification, a discrete and localized staining of second-order neurons was observed, suggesting that the hybrid protein had been targeted to vesicles within cell somata, synapses, and axons. A similar patchy distribution was previously described for tetanus toxin and TTC fragment alone (14, 21, 22).

Intense transneuronal labeling was detected in the lateral reticular formation (LRF), where medullary reticular neurons have been reported to form numerous projections onto the hypoglossal nucleus (26, 27). β-Gal activity was detected bilaterally in these sections. Labeled LRF projections formed a continuous column along the rostrocaudal axis, beginning lateral to the hypoglossal nucleus, with a few neurons being preferentially stained in the medullary reticular dorsal (MdD) and the medullary reticular ventral (MdV) nuclei (Fig. 3 A, B, and A′). This column extends rostrally through the medulla, with neurons more intensely labeled in the parvicellular reticular nucleus (PCRt, caudal and rostral in Fig. 3 B, C, and B′). After 48 hr, cells in MdD and PCRt were more intensely stained. A second bilateral distribution of medullary neurons projecting to the hypoglossal nucleus was detected in the solitary nucleus (Sol in Fig. 3B), but the labeling was less intense than in the reticular formation, presumably because relatively few cells of the solitary nucleus project onto the hypoglossal nucleus (26). However, no labeling was found in the spinal trigeminal nucleus (Sp5), which has also been shown to project onto the hypoglossal nucleus (26). Transynaptic transport of the β-gal–TTC protein was also detected in the pontine reticular nucleus caudal (PnC), the locus coeruleus (LC), the medial vestibular nucleus (MVe), and in a few cells of the inferior vestibular nucleus (IV). These cell groups are known to project onto the hypoglossal nucleus (25), but their labeling was weak, probably because of the greater length of their axons (Fig. 3 C, D, and E). We have observed a few labeled cells in the dorsal paragigantocellular nucleus (DPGi), the magnocellular nucleus caudal (RMc), and the caudal raphe nucleus (R) (Fig. 3 C, D, and E); their connections to the hypoglossal nucleus have also been reported (25). Finally, labeled neurons were detected bilaterally in midbrain projections, such as those of the mesencephalic trigeminal nucleus (Me5 in Fig. 3 D, E, and C′), and a few neurons were stained in the mesencephalic central gray region (CG in Fig. 3E). These latter nuclei have been typed as putative third order cell groups related to the hypoglossal nucleus (25).

Figure 3.

Transneuronal labeling following intramuscular injection of β-gal–TTC into the mouse tongue. A total of 100 μg of protein was injected, and observations were made 24–48 hr postinjection. (A–E) Distribution of β-gal-positive neurons in different brainstem sections, also summarized in Table 1. One dot represents one labeled neuron; six animals were analyzed. (A′–E′) Examples of labeled neurons: labeling in medullary reticular dorsal (MdD) area (A′), arrows showing second order stained neurons; labeling in parvicellular reticular nucleus (PCRt) area (B′ and B" with different magnification) showing a β-gal-positive column of neurons; labeling in Me5 area (C′), arrows show second-order stained neurons; labeling in N7 (D′) and Mo5 areas (E′), arrows show β-gal-positive neurons. [Bars = 100 μm (A′, B′, B", D′, and E′) and 50 μm (C′).] AP, area postrema; Aq, Aqueduct (Sylvius); 4V, 4th ventricule; Gi, gigantocellular reticular nucleus (nu); IO, inferior olive; KF, Kölliker–Fuse nu; LRt, lateral reticular nu; PB, parabrachial nu; Pr5, principal sensory trigeminal nu; RtTg, reticulotegmental nu pons; SO, superior olive; Sp5C, spinal trigeminal nu, caudal part; Sp5I, spinal trigeminal nu, interpolar part; Sp5O, spinal trigeminal nu, oral part; see Table 1 for other abbreviations.

Neurons in the motor trigeminal nucleus (Mo5 in Fig. 3 D, E, and E′) and the accessory trigeminal tract (Acs5 in Fig. 3E) were also labeled, along with a population of neurons in the facial nucleus (N7 in Fig. 3 C and D′). However, interpretation of this labeling is more ambiguous because it is known that motoneurons in these nuclei also innervate other parts of the muscular tissue, and diffusion of the hybrid protein might have occurred at the point of injection. Conversely, these nuclei may have also projected to the tongue musculature via nerve XII, since neurons in N7 have been reported to receive direct hypoglossal nerve input (28). This latter explanation is consistent with the fact that labeling in these nuclei was detected only after 24 hr; however, this point was not further investigated.

Altogether, the data summarized in Table 1 and Fig. 3 clearly establish transneuronal transport of the β-gal–TTC fusion protein from the hypoglossal neurons into several connected regions of the brainstem.

Table 1.

Transneuronal transport of the lacZ-TTC fusion from the XII nerve: labeling of different cells types in the central nervous system

Cell groups	12–18 hr	24–48 hr
First order neurons
First category
XII, hypoglossal motoneurons	++	+++
Second category
N7, facial nu	−	++
Mo5, motor trigeminal nu	−	++
Acs5, accessory trigeminal nu	−	+
Second order cell groups
MdD, medullary reticular nu, dorsal	−	++
MdV, medullary reticular nu, ventral	−	+/−
PCRt, parvicellular reticular nu, caudal	−	++
PCRt, parvicellular reticular nu, rostral	−	++
Sol, solitary tract nu	−	+
DPGi, dorsal paragigantocellullar nu	−	+/−
PnC, pontine reticular nu, caudal	−	+
RMc, magnocellular reticular nu	−	+/−
R, caudal raphe nu	−	+/−
MVe, medial vestibular nu	−	+
IV, inferior vestibular nu	−	+/−
LC, locus coeruleus	−	+
Me5, mesencephalic trigeminal nu (*)	−	+
CG, mesenphalic central gray (*)	−	+/−

(∗), Second-order cell groups that also contain putative third order neurons (see text); −, no labeling; + to +++, increased density of label; +/− weak labeling. Sixteen animals were analyzed for the 12- to 18-hr postinjection data; 6 animals were analyzed for the 24- to 48-hr postinjection data. nu, Nucleus. 

DISCUSSION

In this study, we show that a β-gal–TTC hybrid protein retains the biological activities of both proteins in vivo. Therefore, the hybrid protein can undergo retrograde and transneuronal transport through a chain of interconnected neurons, as traced by its enzymatic activity. Our results are consistent with those of others who used chemically conjugated TTC, or TTC fused to other proteins (12–15). In these in vitro analyses, the activity of the conjugated or hybrid proteins was likewise retained or only weakly diminished. Depending on the nature of the TTC fusion partner, different types of potential applications can be envisioned. For example, one might be able to deliver a biologically active protein into the CNS for therapeutic purposes. Such hybrid genes could also be used to analyze and map synaptically connected neurons if reporters such as lacZ or the green fluorescent protein (GFP; ref. 29) gene were fused to TTC.

The retrograde transport of the hybrid protein is best demonstrated by the data in Fig. 2. When injected into a muscle, β-gal activity rapidly localized to the somata of motoneurons that innervate the muscle. The arborization of the whole nerve, axon, somata, and dendrites can easily be visualized. However, in comparison to the neurotropic viruses, the extent of retrograde transneuronal transport of the hybrid protein from the hypoglossal neurons indicates that only a subset of interconnected neurons is detected, although most areas containing second-order interneurons have been identified by the β-gal–TTC marker. Transneuronal uptake is mostly restricted to second-order neurons. In such experiments, when a limited amount of a neuronal tracer is injected into a muscle or cell, only a fraction will be transported through a synapse, thereby imposing an experimental constraint on its detection. Presently, the most efficient method, in terms of the extent of transport, relies on neurotropic viruses. Examples include: α-herpes viruses, such as herpes simplex type 1, pseudorabies virus, and rhabdoviruses (24, 25). Viral methods are very sensitive because each time a virus infects a new cell, it replicates, thereby amplifying the signal and permitting visualization of higher order neurons in a chain. Ultimately, however, one wants to map a neuronal network in an in vivo situation such as a transgenic animal (see below). Here, the disadvantage of viral labeling is its potential toxicity. Most viruses are not innocuous for the neural cell, and their replication induces a cellular response and sometimes cell degeneration (24). Furthermore, depending on experimental conditions, budding of the virus can occur leading to its spread into adjoining cells and tissues.

Differences in mechanisms of transneuronal migration could also account for the restricted number of neurons labeled by β-gal–TTC. Matteoli et al. (22) have provided strong evidence that the intact tetanus toxin crosses the synapses by parasitizing the physiological process of synaptic vesicle recycling at the nerve terminal. The toxin probably binds to the inner surface of a synaptic vesicle during the time the lumen is exposed to the external medium. Vesicle endocytosis would then presumably provide the mechanism for internalization of the toxin. Because the TTC fragment is known to mimic the migration of the toxin in vivo, it could therefore direct the fusion protein along a similar transynaptic pathway. If this hypothesis is confirmed, it would strongly suggest that synaptic activity is required for the transneuronal transport of β-gal–TTC. Therefore, only active neuronal circuits would be detected by the hybrid protein. The possible dependence of β-gal–TTC on synaptic vesicle exocytosis and endocytosis could be further investigated, since techniques are now available to record synaptic activity in neural networks in vitro (30). In contrast, the transneuronal pathway of neurotropic viruses has not yet been elucidated and could be fundamentally different, involving virus budding in the vicinity of a synapse. Finally, the transneuronal transport of the hybrid protein might depend on a synaptic specificity, although the tetanus toxin is not known to display any (7, 23); it is therefore likely that a virus would cross different or inactive synapses. In summary, the restricted spectrum of interneuronal transport, in addition to its nontoxicity, make the β-gal–TTC hybrid protein a novel and powerful tool for analysis of neural pathways.

In our view, the one advantage of the fusion gene that we describe for neuronal mapping is that it derives from a single genetic entity that is amenable to genetic manipulation and engineering. Several years ago, a technique based on homologous recombination in embryonic stem cells was developed to specifically replace genes in the mouse (31, 32). This method generates a null mutation in the substituted gene, although in a slightly modified strategy, a dicistronic messenger RNA can also be produced (33, 34). When a reporter gene, such as E. coli lacZ, is used as the substituting gene, this technique provides a means of marking the mutated cells so that they can be followed during embryogenesis. Thus, this technique greatly simplifies the analysis of both the heterozygote expression of the targeted gene as well as the phenotype of null (homozygous) mutant animals.

Neural cells establish specific and complex networks of interconnected cells. If a gene were mutated in a given neural cell, we would expect this mutation to have an impact on the functions of other, interconnected neural cells. With these considerations in mind, a genetic marker that can diffuse through active synapses would be very useful in analyzing the effect of the mutation. In heterozygous mutant animals, the cells in which the targeted gene is normally transcribed could be identified, as could the synaptically connected cells of a neural network. In a homozygous animal, the impact of the mutation on the establishment or activity of the neural network could be determined. The feasibility of such an in vivo approach depends critically on the efficiency of synaptic transfer of the fusion protein, as well as its stability and cellular localization. We are currently generating various transgenic animals based on this type of gene fusion to evaluate the application of this technique to the study of CNS development.

Another extension of our approach is to gene therapy applied to the CNS. We have shown that a nontoxic, enzyme–vector conjugate is taken up by axon terminals and conveyed retrogradely to brainstem motoneurons. A selective retrograde transynaptic mechanism subsequently transports the hybrid protein into second-order connected neurons. Such a pathway, which bypasses the blood–brain barrier, could be exploited to deliver macromolecules to the CNS. In fact, pathogenic agents such as tetanus toxin and neurotropic viruses are similarly taken up by nerve endings, internalized and retrogradely transported to the nerve cell somata. In such a scenario, the lacZ reporter would be replaced by a gene encoding a protein that provides a necessary or interesting activity and/or function. For example, the human CuZn superoxide dismutase (SOD-1) and the human enzyme β-N-acetylhexosaminidase A have been fused or chemically coupled to the TTC fragment (13, 16), and their uptake by neurons in vitro was considerably increased and their enzymatic functions partially conserved. Combined with the in vivo experiments described here using β-gal–TTC, a gene therapy approach based on TTC hybrid proteins appears to be a feasible method of delivering a biological function to the CNS. However, ways have to be found to target the TTC hybrid proteins, which are likely to be sequestrated into vesicles, to the appropriate subcellular compartment. Such a therapeutic strategy could be particularly useful for treating neurodegenerative and motoneuron diseases, such as amyotrophy lateral sclerosis (35), spinal muscular atrophies (36, 37), or neurodegenerative lysosomal storage diseases (38, 39). Injection into selected muscles, even in utero, could help to specifically target the appropriate neurons. In addition, such an approach would avoid the secondary, and potentially toxic, effects associated with the use of defective viruses to deliver a gene (40, 41).

ABBREVIATIONS

CNS

central nervous system

TTC

tetanus toxin fragment C

β-gal

β-galactosidase

X-Gal

5-bromo-4-chloro-3-indolyl β-d-galactoside