Low‐affinity nerve‐growth factor receptor (P75NTR) can serve as a receptor for rabies virus

Christine Tuffereau, Jacqueline Bénéjean, Danielle Blondel, Brigitte Kieffer, Anne Flamand

Author Affiliations

  1. Christine Tuffereau*,1,
  2. Jacqueline Bénéjean1,
  3. Danielle Blondel1,
  4. Brigitte Kieffer2 and
  5. Anne Flamand1
  1. 1 Laboratoire de Génétique des Virus, CNRS, 91198, Gif sur Yvette, Cedex, France
  2. 2 Laboratoire Protéines et Récepteurs Membranaires, UPR 9050 CNRS, ESBS, Parc d'innovation, Bld Sébastien Brand, 67400, Illkirch, France
  1. *Corresponding author. E-mail: ctuffer{at}
View Full Text


A random‐primed cDNA expression library constructed from the mRNA of neuroblastoma cells (NG108) was used to clone a specific rabies virus (RV) receptor. A soluble form of the RV glycoprotein (Gs) was utilized as a ligand to detect positive cells. We identified the murine low‐affinity nerve‐growth factor receptor, p75NTR. BSR cells stably expressing p75NTR were able to bind Gs and G‐expressing lepidopteran cells. The ability of the RV glycoprotein to bind p75NTR was dependent on the presence of a lysine and arginine in positions 330 and 333 respectively of antigenic site III, which is known to control virus penetration into motor and sensory neurons of adult mice. P75NTR‐expressing BSR cells were permissive for a non‐adapted fox RV isolate (street virus) and nerve growth factor (NGF) decreased this infection. In infected cells, p75NTR associates with the RV glycoprotein and could be precipitated with anti‐G monoclonal antibodies. Therefore, p75NTR is a receptor for street RV.


Rabies virus (RV) is a lyssavirus that belongs to the rhabdovirus family. It is a neurotropic virus usually transmitted through the bite of a rabid animal (Charlton, 1994; Dietzschold et al., 1996). RV penetrates either directly into nerve endings at the site of inoculation (Shankar et al., 1991) or after a limited multiplication in myotubes (Murphy et al., 1973a; Harrison and Murphy, 1978); it is then transported along axons (Tsiang, 1978) to the cell body of motor and sensory neurons, where replication takes place. Viral budding is observed mostly in internal compartments of infected neurons (Gosztonyi, 1994) and the virus is transported to synapses in vesicles. Within the nervous system (NS), propagation of RV between connected neurons occurs exclusively at the synapse. Late in infection, the virus eventually spreads to a few categories of non‐neuronal differentiated tissues, such as submaxillary salivary glands, taste buds, adrenal glands, pancreas, kidney, hair follicles and brown fat tissue (Murphy et al., 1973b). At this stage, classic rabies symptoms develop and death occurs rapidly.

Apart from the very beginning and end of the infectious process, RV multiplies and propagates exclusively inside neurons. This neuronal tropism in vivo is also observed in vitro with street RV isolates extracted from salivary glands or from the brains of rabid animals. In vitro, such isolates can only infect established cell lines of neuronal origin. However, viruses can be adapted (Kissling, 1958) and several passages are required for the virus to be adapted fully to the in vitro multiplication. Additional cycles of multiplication in non‐neuronal cells are necessary for the selection of fixed strains that would multiply in established cell lines such as BHK21, BSR and Vero cells (Wiktor et al., 1964; Schneider et al., 1971). Evelyn Rokitnicki Abelseth (ERA), Pasteur Virus (PV) or Challenge Virus Standard (CVS) are fixed RV strains that have been selected in the past according to this procedure and are used around the world for laboratory investigation. All have kept their specific tropism for neurons in animals and propagate in the nervous system like street viruses. Therefore, adaptation did not abolish neurotropism but rendered the virus able to grow in non‐neuronal cells. It is postulated, but not demonstrated, that adaptation is at least partly due to the capability of fixed strains of RV to use ubiquitous receptors present on every cell type investigated to date (Seganti et al., 1990). Ubiquitous receptors could be molecules such as phospholipids (Superti et al., 1984), gangliosides (Conti et al., 1986; Superti et al., 1986) or proteins (Wunner et al., 1984; Broughan and Wunner, 1995; Gastka et al., 1996). Recently, the neural cell‐adhesion molecule has been shown to be a receptor for RV laboratory strains (Thoulouze et al., 1998). Also, it has been proposed that the nicotinic acetylcholine receptor (nAChR) serves as a receptor for RV (Lentz et al., 1984, 1986; Hanham et al., 1993). Nevertheless, the fact that RV infects neurons that do not express nAChR (McGehee and Lorca, 1995) suggests the existence of other molecules mediating viral entry into neurons. It should be emphasized that the nAChR is located mainly on muscle cells and could account for the ability of street RV to multiply locally in myotubes at the site of inoculation (Burrage et al., 1985) which would facilitate subsequent penetration into neurons.

The RNA genome of RV is 11 930 nucleotides long and of negative polarity. It encodes five proteins. The unique glycoprotein (G) is organized in trimers which protrude from the viral envelope (Whitt et al., 1991; Gaudin et al., 1992). It is a type I integral transmembrane glycoprotein with an external domain of 439 amino acids (aa), a transmembrane region of 22 aa and a C‐terminal cytoplasmic domain of 44 aa. Extensive studies of the antigenicity of the protein have identified two immunodominant conformational sites, named sites II and III (Seif et al., 1985; Préhaud et al., 1988), one minor site (site a) (Benmansour et al., 1991) and several linear epitopes (Bunschoten et al., 1989; Raux et al., 1995; Lafay et al., 1996) on the external domain. The G protein is a major determinant of the viral neurotropism. Mutations in the glycoprotein reduce or abolish neuroinvasiveness without impairing the ability of the virus to multiply in cell culture. Replacement of Arg333, situated in site III of the G protein, results in the loss of virulence for adult animals (Dietzschold et al., 1983b; Seif et al., 1985; Tuffereau et al., 1989). The mutant virus is still able to infect peripheral neurons but is only transmitted to a few categories of second order neurons in the central nervous system (CNS) (Dietzschold et al., 1985; Coulon et al., 1989; Lafay et al., 1991). We have shown recently that the additional mutation of Lys330, also in site III, abolishes the penetration of the virus into motor and sensory neurons after intramuscular inoculation of the virus (Coulon et al., 1998).

Even if experimental evidence suggests the existence of as yet unidentified neuronal receptors for RV, the presence of ubiquitous receptors sufficient to mediate the penetration of RV into most, if not all, cell lines in vitro is a considerable limitation to the cloning of such receptors. We have recently demonstrated the existence of specific binding sites for the RV glycoprotein in neuronal cell lines of various origins. These sites mediate the fixation of Spodoptera frugiperda (Sf21) cells expressing the RV glycoprotein (G‐Sf21 cells) on their surface (Tuffereau et al., 1998). They are not present on the non‐neuronal cell lines tested so far. Mutations at positions 330 and 333 of the glycoprotein greatly reduce the binding. These sites are different from nAChR because antibodies directed at this receptor do not abolish binding to G‐Sf21 cells. In the present study, we have observed that the truncated form of the glycoprotein (Gs), which is cleaved and secreted naturally from infected cells (Dietzschold et al., 1983a; Morimoto et al., 1993), is also found in the supernatant of G‐Sf21 cells. Gs as well as G‐Sf21 cells were capable of binding to neuroblastoma cells (i.e. NG108 cells), but not to cell lines of non‐neuronal origin such as COS7 or BSR cells. This observation allowed us to use an expression‐cloning strategy to identify the protein responsible for Gs and G‐Sf21 binding to NG108 cells. We expressed an NG108 library in COS7 cells and obtained a cDNA clone encoding a protein capable of binding Gs. BSR cells stably expressing this protein could be infected by a non‐adapted field RV isolate originating from a fox. This virus was unable to grow on BSR cells not expressing the receptor. Sequencing of the isolated clone identified it as the mouse counterpart of the rat and human low‐affinity nerve‐growth factor receptor (p75NTR; Johnson et al., 1986; Radeke et al., 1987).


G‐Sf21 cells produce a soluble form of the RV glycoprotein

In the supernatant of RV‐infected BHK cells, a soluble glycoprotein, Gs, has been detected (Dietzschold et al., 1983a; Morimoto et al., 1993). Gs results from the cleavage of the mature G, releasing the ectodomain and the first 8 aa of the transmembrane domain in the supernatant. Gs is antigenically indistinguishable from the G protein in its native configuration (Dietzschold et al., 1983a). We analysed whether Sf21 cells that have been infected with a recombinant baculovirus carrying the RV glycoprotein gene also secreted Gs. Supernatants from four separate batches of infected lepidopteran cells were analysed by SDS–PAGE together with 3, 5 and 10 ng of purified G protein from RV. Proteins were transferred onto nitrocellulose membranes and revealed with an anti‐G monoclonal antibody (mAb). In the supernatants of G‐Sf21 cells (Figure 1A, lanes a–d), a major protein (Gs) and a minor protein (Gb) were detected. After ultracentrifugation, Gb was associated preferentially with the pellet fraction (Figure 1B, lane P), while Gs remained in the supernatant (Figure 1B, lane S). Therefore, it is probable that Gb is associated with heavier structures, such as membrane debris or vesicles, or that it is inserted in the envelope of recombinant baculoviruses, as already published (Barsoum et al., 1997). Gb produced in lepidopteran cells migrated faster than G produced in mammalian cells, due to the incomplete maturation of the sugar chains in insect cells (Jarvis and Finn, 1995) and due to the presence of bovine serum albumin in the cell medium. Comparison of the intensity of the bands between purified G and Gs suggested that the four batches of G‐Sf21 cells (Figure 1A, lanes a–d) secreted ∼0.5 μg/ml of Gs in the supernatant. Gs was recognized by mAbs directed to the known antigenic regions of the native G (data not shown) as described for Gs present in the supernatant from RV‐infected BSR cells (Dietzschold et al., 1983a).

Figure 1.

Analysis of the soluble form of the RV glycoprotein (Gs). (A) Three, 5 and 10 ng of purified G from RV together with 15 μl of four different G‐Sf21 cell supernatants were run on a 10% SDS–polyacrylamide gel (lanes a, b, c and d). Western blotting of the gel was performed and G was detected with the anti‐G mAb 17D2. (B) After ultracentrifugation of the G‐Sf21 culture medium, the pellet fraction (P) was resuspended in a volume equivalent to the initial volume and the supernatant fraction (S) was kept. Fifteen microlitres of each sample were analysed as in (A). Gb: migration of RV G expressed in Sf21 cells.

Gs attaches to neuroblastoma cells but not to COS7 cells

We have shown previously that G‐Sf21 cells attach to various neuroblastoma cell lines through interaction between the RV glycoprotein and neuronal cell‐surface molecules (Tuffereau et al., 1998). When neuroblastoma cells, such as NG108 cells, were treated with the supernatant of G‐Sf21 cells, they retained Gs at their surface. As a result, cells treated successively with a mixture of mAbs directed against various regions of the RV glycoprotein (see Materials and methods) and then with a β‐galactosidase‐labelled anti‐mouse antibody became blue after X‐gal staining. Anti‐G mAb 50AD1, which is specific for site III, did not recognize Gs when bound to the NG108 cells but recognized Gs in solution (data not shown). COS7 cells did not retain Gs and were not stained (Figure 2A). When a mixture of NG108 and COS7 cells was incubated with the G‐Sf21 supernatant, only NG108 cells, which could be clearly differentiated from COS7 cells by cell morphology, retained Gs (Figure 2B). When NG108 cells were differentiated with Na‐butyrate and incubated with the supernatant of G‐Sf21 cells, a high density of Gs bound to the nerve processes (Figure 2C). After ultracentrifugation of the G‐Sf21 culture medium, supernatant containing Gs gave a heavy staining when incubated with NG108 cells, while the resuspended pellet containing only Gb bound poorly to these cells (not shown). Consequently, the supernatant of G‐Sf21 cells was used without further purification. No blue staining of NG108 cells was observed after treatment with culture medium from non‐infected Sf21 cells or from Sf21 cells infected with a lacZ‐recombinant baculovirus (data not shown). Gs released in the supernatant of RV‐infected BSR cells was also able to bind specifically to NG108 cells (data not shown).

Figure 2.

Binding of Gs to neuroblastoma cells. COS7 cells (A), a COS7 and NG108 co‐culture (B) and NG108 cells differentiated for 4 days in the presence of 5 mM Na‐butyrate (C) were incubated with 3 ml of the supernatant from G‐Sf21 cells collected 2 days after infection and containing ∼1.5 μg of Gs. The cells were washed, fixed with paraformaldehyde, incubated with anti‐G mAbs, then with a β‐galactosidase‐labelled anti‐mouse conjugate and finally stained with the X‐gal substrate. Magnification: ×150.

Identification of a putative receptor for RV using Gs as a ligand

Since Gs binds specifically to NG108 cells, it could serve in a screening assay devised at cloning a putative RV receptor. We used a random‐primed cDNA library derived from mRNA of NG108 cells. This library contained 2.9×106 primary transformants, up to 85% of the clones had inserts, and half of the inserts were >1500 bp (Kieffer, 1991). This library was used successfully to clone the δ‐opioid receptor (Kieffer et al., 1992). In a preliminary experiment, we observed that the transfection of the whole library into COS7 cells resulted in a few cells that expressed a surface molecule able to retain Gs from G‐Sf21 cell supernatant (data not shown), indicating that the library contained the gene(s) of interest. It was divided into pools of 1000–1200 different recombinant bacterial colonies. Plasmid DNA from 150 of these pools were used to transfect 2×105 COS7 cells. One pool gave 30–40 cells that were light blue after successive incubation with the G‐Sf21 cell supernatant, anti‐G mAbs, the anti‐mouse antibody and the X‐gal substrate. This pool was subdivided twice into subpools of 200 and then eight plasmids. At this stage, 80% of the transfected cells stained blue and the intensity of the staining increased. Ultimate subcloning led to the isolation of a single plasmid (plasmid 8‐2). COS7 cells or BSR cells transfected with this plasmid bound Gs and the intensity of the staining was similar (Figure 3A and B). Transfected cells were not stained when the treatment with G‐Sf21 cell supernatant was omitted (Figure 3C).

Figure 3.

Binding of Gs to p75NTR‐expressing cells. BSR (A) or COS7 cells (B) transiently transfected with plasmid 8‐2 or BSR‐R5 cells (D) were stained as described in the legend to Figure 2. (C) Incubation of 8‐2‐transfected BSR cells with Gs was omitted. Magnification: ×150.

Plasmid 8‐2 encodes the murine low‐affinity nerve‐growth factor (NGF) receptor

The plasmid 8‐2 contained a 1.3 kb insert with a large open reading frame of 1251 bp corresponding to a sequence of 417 aa. The predicted sequence showed high homology with the rat (Radeke et al., 1987) and human (Johnson et al., 1986) low‐affinity NGF receptor, p75NTR. It was also homologous, although to a lesser extent, to the chicken receptor (Figure 4). The putative transmembrane protein contains a hydrophobic stretch of 21 aa at its N‐terminus, presumably acting as a signal peptide, an external domain of 220 aa containing four cysteine‐rich domains and a stalk rich in serine and threonine, a 22 aa transmembrane domain and a 154 aa cytoplasmic domain. The external domain of the protein has one conserved potential N‐glycosylation site and the stalk is highly O‐glycosylated. There are 14 aa differences between the rat p75NTR and our sequence, which are all located in the ectodomain (five in the cysteine‐rich domain and nine in the stalk of the ectodomain). Since the NG108 cell line is a hybrid between a murine neuroblastoma (N18) and a rat glioma (C6) cell line, sequencing of mouse genomic DNA amplified by PCR was performed to ensure the origin of our clone. Comparison between the mouse genomic sequence and the sequence from plasmid 8‐2 showed that we have actually cloned the murine p75NTR (data not shown).

Figure 4.

Alignment of the putative protein sequence encoded by plasmid 8‐2 to the rat (r), human (h) and chicken (c) p75NTR. The sequence of the chicken homolog was from Large et al. (1989). The transmembrane domain is underlined.

Since the size of the insert in plasmid 8‐2 was shorter (1300 bp) than the rat and human p75NTR mRNA (3700 bp), Northern blot analysis was performed. It indicated that the murine p75NTR mRNA in NG108 cells was ∼3700 bp (data not shown), which is in accordance with the mRNA size in rat PC12 cells (Radeke et al., 1987). The insert of the isolated clone, therefore, is missing the long 3′ untranslated sequence of the p75NTR mRNA. COS7 cells transfected with plasmid 8‐2 expressed a doublet of proteins of ∼75 kDa which was recognized by Western blotting with a rabbit polyclonal serum directed against the cytoplasmic domain of the human p75NTR (Figure 5A). This serum also recognized a doublet of the same molecular weight in PC12 cell extracts and did not react with COS7 cell extracts transfected with the pCDM8 control plasmid (Figure 5A). In addition, a minor band of low molecular weight was detected. This polypeptide corresponds to the cytoplasmic and transmembrane domains of the protein that have been described in PC12 cells after cleavage of the external domain (DiStephano and Johnson, 1988).

Figure 5.

Immunodetection of p75NTR. Cell extracts were prepared from PC12 cells or COS7 cells transfected with either plasmid pCDM8 or 8‐2 (A) or from the control line C12 and the p75NTR‐expressing cells R4 or R5 (B) and analysed by Western blotting using a polyclonal rabbit serum directed against the cytoplasmic domain of the human p75NTR. Truncated forms of p75NTR are indicated with arrow heads.

Stable expression of p75NTR in BSR cells

To isolate clones of BSR cells stably expressing p75NTR, BSR cells were cotransfected by plasmids 8‐2 and pSV2 Neo. After several cycles of multiplication, surviving cells were cloned and ∼30 colonies able to grow in the presence of geneticin were selected. Most of them were able to bind Gs when treated with the supernatant of G‐Sf21 cells, although the quantity of Gs attached, and consequently the intensity of staining, varied. The staining of the cells within individual clones was homogenous (Figure 3D). Two of these clones (R4 and R5) were shown to express p75NTR (Figure 5B). Control BSR clones also were isolated after transfection of BSR cells with plasmids pCDM8 and pSV2 Neo. None of these clones were able to bind Gs. One control clone (C12) was selected for further studies. As expected, it did not express p75NTR (Figure 5B). The attachment of G‐Sf21 cells on six clones stably expressing p75NTR (R4, R5, R7, R11, R12 and R13) was analysed. The level of fixation varied from clone to clone and remained stable during >30 passages of the cells (Figure 6A). For instance, R7 always showed a low level of G‐Sf21 binding, 3–5 times higher than the background, while binding to R5 was high. G‐Sf21 cells did not attach to the control BSR clone (C12) and Sf21 cells infected with a LacZ expressing baculovirus were not retained at the surface of receptor‐expressing cells (data not shown). We have demonstrated previously that mutations in positions 330 and 333 of the RV glycoprotein abolished the penetration of RV into motor and sensory neurons after intramuscular inoculation of adult mice (Coulon et al., 1998). We have shown also that these mutations greatly decreased the binding of G‐Sf21 cells to NG108 cells, although the quantity of mutated G at the surface of the insect cells was not reduced (Tuffereau et al., 1998). These mutations similarly reduced the binding of G‐Sf21 to p75NTR‐expressing clones (Figure 6B).

Figure 6.

Binding of G‐Sf21 cells to BSR lines expressing p75NTR. (A) The binding assay with Gcvs‐Sf21 was performed as described in Materials and methods. Two independent binding experiments were performed at passage 15 (hatched bars) or at passage 40 (open bars). For R13, only passage 40 was analysed. The binding efficiency was expressed as the ratio of the number of bound cells to the total number of insect cells added to the cell monolayers. Each bar represents the average of three determinations with standard deviations (SD). (B) Comparison of binding of G‐Sf21 cells expressing either the parental Gcvs glycoprotein (black bars) or the doubly mutated glycoprotein (K330N+R333M) (hatched bars) to BSR cells expressing p75NTR as described in Materials and methods. The binding efficiency was expressed as the ratio of the number of bound cells to the total number of insect cells added to the monolayers. The experiment was done at passage 43. Each bar represents the average of three determinations.

Cells expressing p75NTR are permissive for a field RV isolate

With the exception of neuroblastoma cells, field RV isolated from the brain of infected animals does not grow in cultured cells. The virus has to be adapted to growth in cultured cells by successive passages in neuroblastoma cells and then in cell lines of non‐neuronal origin. We infected p75NTR‐expressing cells with a fox isolate of RV that had been multiplied once in suckling mice and once in hamster. The hamster brain extract was titrated at 107 LD50/ml by intracerebral inoculation of adult mice. A 10‐fold dilution of this viral suspension was added to R4, R5, R13 and to the control clone C12. After 22 h, cells were fixed and stained with a fluorescein‐conjugated anti‐nucleocapsid antibody. On p75NTR‐expressing cells, numerous positive cells were counted (Table I) and the ratio between infectious units (I.U.) and LD50 in the inoculum was close to 0.1. Considering that the viral suspension was still contaminated heavily with brain material that could have an inhibitory effect on the development of the infectious cycle, it is likely that the ratio could be even higher. As expected, few infected cells were detected on C12 cells and the ratio between I.U. and LD50 in the inoculum was <0.002. The fox RV isolate could also propagate from cell to cell under agarose and gave foci which could be stained with fluorescent antinucleocapsid antibodies after permeabilization of the cell layer (data not shown). Therefore, the presence of p75NTR enabled the penetration of field RV and the development of an efficient infection into otherwise refractory cells. The fixation of field RV was inhibited by the mouse β‐NGF, a natural ligand of p75NTR (Figure 7). The inhibition was in the range of 52, 35 and 9% for doses of NGF equal to 100, 33 and 10 nM, respectively.

Figure 7.

Inhibition of the fixation of a field RV isolate to R5 by NGF. Cells were treated for 30 min at room temperature with 100, 33 or 10 nM of mouse NGF before infection with the G5H field RV as described in the legend to Table II. Cells were incubated at 37°C for 22 h, fixed and stained with anti‐nucleocapsid antibodies. Twelve to 15 fields were counted for each NGF concentration. Each bar represents the average of three determinations.

View this table:
Table 1. Infection of BSR cells stably expressing p75NTR with a RV street isolate

Interaction of the CVS glycoprotein with p75NTR

BSR cells are fully susceptible to the adapted strains of RV and the presence of p75NTR did not significantly increase their susceptibility to infection by these viruses. For instance, the number of CVS plaques counted on the p75NTR‐stable lines were only slightly higher than on the control cell line C12 (Table II). When viral infection was performed in the presence of 10% of serum, p75NTR‐expressing cells were 3–10 times more susceptible to CVS infection than control cells (data not shown), indicating that p75NTR can serve as a receptor for RV laboratory strains. Viral production of the p75NTR‐stable lines after infection by the CVS strain was equivalent or slightly lower than obtained from control cells (data not shown). When C12‐ or R5‐infected cells were labelled, solubilized with detergent and then incubated with anti‐p75NTR, some viral G coimmunoprecipitated with p75NTR (Figure 8A, lane R5+), thus indicating a direct interaction between G and its receptor. The inefficient 35S incorporation in p75NTR did not allow its detection by autoradiography when the immunoprecipitation was performed with anti‐G mAbs (Figure 8A). However, p75NTR was detected in Western blot analysis of the same immune complexes (Figure 8B). The receptor was not present in immune precipitates from non‐infected R5 cell extracts treated with anti‐G mAbs.

Figure 8.

Interaction between p75NTR and the RV glycoprotein. Stable lines C12 and R5 were either not infected (−) or infected (+) at a m.o.i. of 3 for 16 h with the CVS strain of RV. (A) Cells were labelled in the presence of 35S, then solubilized in lysis buffer. Immunoprecipitations were performed with anti‐G mAbs (30AA5 + 17D2) or with an anti‐p75NTR serum. Immune complexes were analysed by SDS–PAGE (8% acrylamide) followed by autoradiography. A longer exposure of the gel is shown for the immunoprecipitation performed with anti‐p75NTR. G from the CVS strain of RV and p75NTR migrate as doublets. (B) Immunoprecipitations of cold extracts with anti‐G mAbs were performed as described in (A). Immune complexes together with an extract from R5 cells were analysed by Western blotting with the anti‐p75NTR antibody after migration on an 8% SDS–polyacrylamide gel. The asterisk indicates the p75NTR immunoprecipitated by anti‐G mAbs.

View this table:
Table 2. Titration of the CVS laboratory RV strain on BSR cells stably expressing p75NTR


RV penetrates into peripheral neurons at the site of inoculation and propagates in the CNS by transynaptic transmission to connected neurons. Direct attempts to identify molecules which could mediate the entry of RV into nerve endings have been impaired by the fact that fixed strains of RV, the only ones which could be safely and easily manipulated, multiply in most mammalian cell lines investigated to date, with a few exceptions such as L cells, which are poorly infected by the CVS RV strain (Seganti et al., 1990; Thoulouze et al, 1998; D.Blondel, unpublished). On the contrary, field isolates of RV are unable to infect in vitro cell cultures with the exception of neuroblastoma cells. It is generally postulated that adaptation enables the virus to use ubiquitous receptors which would be present on continuous cell lines. These molecules complicate the identification of specific receptors, and procedures which have been successful for other viruses could not be adopted. Fortunately, we found that the soluble form of the viral glycoprotein, Gs, which is naturally cleaved in RV‐infected BHK cells (Dietzschold et al., 1983a; Morimoto et al., 1993) or in G‐expressing lepidopteran cells, bound exclusively to neuroblastoma cells (Tuffereau et al., 1998). Gs did not attach to non‐neuronal cells, thus reproducing at least some of the characteristics of field RV tropism.

Using an expression library derived from NG108 cells, neuroblastoma cells expressing a putative receptor for RV (Tuffereau et al., 1998) and Gs as a ligand, we were able to clone a gene whose gene product confers Gs binding capacity to transfected COS7 or BSR cells. Sequence analysis of the cloned gene showed that the putative receptor was the mouse homologue of rat (Radeke et al., 1987) and human (Johnson et al., 1986) low‐affinity NGF receptor p75NTR. BSR cells stably expressing the murine p75NTR could be directly infected by a non‐adapted fox isolate of RV which, as expected, multiplies poorly on normal BSR cells. Therefore, the protein enabling the penetration of field RV into otherwise non‐permissive cells behaves as a functional receptor for the virus. Pretreatment of cells with NGF reduced but did not abolish the penetration of field RV, probably because the virus competed efficiently with the ligand for fixation onto the receptor. Unlike street viruses, RV laboratory strains (as CVS) efficiently recognize other molecules such as the neural cell adhesion molecule as receptors (Thoulouze et al., 1998) and the presence of p75NTR did not drastically enhance the susceptibility of BSR cells to CVS infection. Nevertheless, the CVS glycoprotein was shown to interact with p75NTR in infected R5 cells since the receptor was precipitated together with the glycoprotein by anti‐G mAbs. This observation is consistent with the fact that CVS, although fully adapted to cell culture, has kept its tropism for neurons. Comparison of the sequence from laboratory strains and primary isolates has revealed several differences in the glycoprotein (Benmansour et al., 1992), but further investigations are necessary to identify the molecular basis of RV cell adaptation. These data support recent evidence that some viruses (such as echoviruses) recognize more than one receptor (Powell et al., 1998). It has also been reported that wild‐type strains of measles virus bind to a receptor different from CD46 which is used by vaccine strains (Bartz et al., 1998).

The p75NTR is a negatively‐charged transmembrane protein with four cysteine‐rich extracellular domains responsible for ligand binding and dimerization, a stalk rich in serine and threonine, and a cytoplasmic tail with a putative ‘death domain’ responsible for apoptosis (Chao, 1994; Liepinsh et al., 1997). Due to these characteristics, it has been included in the tumor necrosis factor receptor (TNFR) superfamily of proteins. The receptor has one N‐glycosylation site and is heavily O‐glycosylated on the stalk of the molecule. Interestingly, a few other viruses use proteins with cysteine‐rich domains as receptors. For instance, herpes simplex virus entry mediator (HVEM), an otherwise unknown protein of the TNF/NGF receptor family, was identified as a receptor for herpes simplex virus types 1 and 2 (Montgomery et al., 1996). Two other members of the same superfamily, CAR1 (Brojatsch et al., 1996) and Tva (Bates et al., 1993), were recently shown to serve as receptors for leukemia and Rous sarcoma viruses, respectively. CAR1, like p75NTR, contains a death domain in its cytoplasmic tail and the binding of leukemia virus (or of its glycoprotein) to CAR1‐expressing cells can trigger cytopathic effects. The role of p75NTR in the induction of apoptosis is under investigation and the results depending both on the ligand used and on the cellular context (Bredesen and Rabizaneh, 1997) are still controversial. In the case of RV infection, apoptosis has been described for some but not all infected brain areas (Jackson and Rossiter, 1997). Whether cell death occurring in the brains of RV‐infected mice is related to the interaction between G and p75NTR is an interesting question that needs to be addressed.

P75NTR is abundantly synthesized in neuronal and non‐neuronal tissues of young animals. In adults, its expression is more limited, but the presence of p75NTR was reported in several categories of neurons as well as in other tissues such as muscle, inner ear, testes or the submaxillary gland (Chao, 1994). In neurons, it is mainly located at synapses. The receptor binds NGF and several other neurotrophins (NT) such as the brain‐derived nerve factor (BDNF), NT‐3 and NT‐4/5 with a dissociation constant of ∼10−9 M. Neurotrophins are short molecules (dimers of two polypeptide chains of 120 amino acids), well conserved among vertebrates and showing 55–60% identity among themselves. Regions of the neurotrophins which interact with p75NTR have not been precisely delineated but it is known that a few basic amino acids are essential. For instance, mutation of three lysines in positions 32, 34 and 95 abolishes the interaction of NGF with p75NTR. Similarly, the binding of BDNF to the low‐affinity receptor depends on the presence of three lysines at position 95–97. The neurotrophins are generally positively charged, suggesting that ionic bonds could be implicated in the interaction with the cysteine‐rich domain of p75NTR (Ibàñez, 1995). No obvious sequence homology between neurotrophins and the RV glycoprotein could be detected, the only striking common feature being that a few positively charged amino acids control the RV neuroinvasiveness (Dietzschold et al., 1983b, 1985; Seif et al., 1985; Tuffereau et al., 1989). For instance, we have shown that Lys330 and Arg333 in antigenic site III are essential for viral penetration into neurons (Coulon et al., 1998) and are important for binding of G to p75NTR‐expressing BSR cells (Figure 5B). The substitution of a few other basic amino acids, for instance Arg184 and Lys198, results in the attenuation of the virus (Préhaud et al., 1988). Moreover, anti‐G mAb 50AD1 which specifically binds to site III (Lafay et al., 1994) is unable to recognize Gs when it is bound to p75NTR, suggesting that antigenic site III is then no longer accessible.

The fixation of a neurotrophin onto p75NTR can be followed by the association of the resulting heteropolymer with a member of the Trk family of proteins (TrkA, TrkB or TrkC, depending on the ligand), thus creating a high‐affinity binding site for the neurotrophin (Barbacid, 1994; Bothwell, 1995). Binding of neurotrophins to Trk involved more amino acids than the interaction with p75NTR (Ibàñez, 1995). In neurons, it was demonstrated that the complex p75NTR–NGF–TrkA is internalized in vesicles and retrogradely transported to the cell body (Riccio et al., 1997). Since we showed that p75NTR could serve as a receptor for RV, it is tempting to speculate that the interaction of the virus with its receptor is followed by the same succession of events. Alternatively, the fixation of RV to p75NTR could be sufficient to mediate the internalization of the virus, or its direct fusion at the synaptic membrane. In BSR cells which do not express Trk, we showed that the presence of p75NTR ensures the penetration of street RV, followed by a productive viral cycle supporting the conclusion that p75NTR is indeed a RV receptor.

Since p75NTR is present at synapses in several categories of neurons, we can postulate that it could be a neuronal receptor for RV. Nevertheless, this molecule is probably not the only one capable of mediating the penetration of the virus into nerve termini. First of all, p75NTR has not been found in every category of neurons permissive for RV. This is not a definitive argument because the molecule could be present below the level of detection. Secondly, it is also possible that RV binds to Trk in the absence of p75NTR, like neurotrophins do. Whether proteins of the Trk family could be alternative receptors for RV is under investigation.

Materials and methods


Cells from S.frugiperda (Sf21) were grown in TC100 medium plus 10% fetal bovine serum (FBS) at 28°C. NG108‐15, a hybrid of mouse neuroblastoma N18 and rat glioma C6 cells (Nelson et al., 1976), was cultured in Dulbecco‘s modified Eagle's medium (DMEM) plus 10% FBS. Cells were used between passages 20 and 35. COS7 cells were grown in DMEM plus 10% FBS. BSR cells, cloned from BHK‐21 cells, were grown in Glasgow's modified minimal essential medium (GMEM) plus 10% calf serum. Cells of the PC12 line, a rat adrenal pheochromocytoma cell line, were grown in DMEM plus 10% FBS.

Monoclonal antibodies

Seven anti‐G mAbs recognizing antigenic site II (30AA5), site III (50AD1), site minor a (9B4) and a linear epitope (8C3, 20C7, 21H8, 17D2) were obtained and characterized in our laboratory (Benmansour et al., 1991; Lafay et al., 1994, 1996; Raux et al., 1995).

Production and quantification of soluble glycoprotein

Monolayers of Sf21 cells were infected with a Gcvs‐expressing recombinant baculovirus (Préhaud et al., 1989) at a multiplicity of infection (m.o.i.) of 10 p.f.u./cell. Two days post‐infection, the supernatants were pooled, clarified at 3200 g for 15 min in a GDK‐Beckman centrifuge and directly used for cell binding. To determine the concentration of Gs in the supernatant, an aliquot (15 μl) was run on a 12% polyacrylamide gel together with 3, 5 and 10 ng of purified RV G obtained by virus solubilization using CHAPS {3‐[(3‐cholamydopropyl)‐dimethylammonio]‐1‐propanolosulfonate} according to the procedure described (Gaudin et al., 1992). Western blotting of the gel was performed and stained first with the anti‐G mAb 17D2, then with peroxidase‐labelled anti‐mouse antibody (Sigma). Detection was performed using the enhanced chemiluminescence protocol from Amersham.

Binding of Gs to cells

Monolayers of NG108 or COS7 cells were incubated with the Gcvs‐Sf21 cell supernatant for 2 h at room temperature. The cells were washed three times with phosphate‐buffered saline (PBS), then fixed for 15 min with cold 4% paraformaldehyde in PBS pH 7.4 and washed again three times. Ascites fluids from five anti‐G mAbs (21H8, 20 C7, 30AA5, 8C3 and 9B4) were pooled and diluted 500‐fold in PBS. Cell monolayers were incubated with 500 μl of this anti‐G mAbs dilution for 2 h at room temperature. After three washes with PBS, the cells were incubated for 1 h at room temperature with 300 μl of a β‐galactosidase‐labelled anti‐mouse antibody conjugate (Southern Biotechnology, Associates Inc.) diluted 100‐fold in PBS. After washing three times with PBS, the cell monolayers were incubated with PBS containing 0.02% X‐gal, 3 mM K4Fe(CN)6, 3 mM K3Fe(CN)6 and 2 mM MgCl2 for 3–5 h at 37°C (Ward et al., 1994).

Screening of the NG108 cDNA library

A random‐primed NG108 cDNA library cloned in the vector pCDM8 (Invitrogen) containing 1.5×106 clones was used (Kieffer, 1991). Aliquots of the DNA library were used to transform competent MC1061‐P3 bacteria and the resulting transformants were plated at a density of 1000–1200 colonies per dish. Colonies from each pool were collected by scraping and an aliquot was frozen as a glycerol stock. Plasmid DNA was prepared from each pool and one‐sixth of the DNA was transfected into COS7 cells in 35 mm culture dishes coated with 5 μg/cm2 of collagen type I (Collaborative Research) using the DEAE–dextran/chloroquine procedure. After 48 h, transfected cells were washed, incubated for 3 h at room temperature with 2 ml/well of Gcvs‐Sf21 cell supernatant. The staining was performed as described above. One positive pool was subdivided three times and re‐screened until a single clone (8‐2) could be isolated. The cDNA insert was sequenced using T7 DNA polymerase (Amersham) and the T7 primer (Promega) as well as other primers (MWG‐Biotech) which were generated as the sequence was obtained. The nucleotides sequences of the insert has been deposited in DDBJ/EMBL/GenBank under the accession No. AF105292.

Stably transfected BSR cells

Stable p75NTR‐expressing cell lines were produced by cotransfection of BSR cell monolayers with plasmid 8‐2 encoding p75NTR and plasmid pSV2 Neo (Southern and Berg, 1982) using lipofectin (Gibco‐BRL). After 48 h, the transfection medium was replaced with GMEM plus 10% FBS containing geneticin (500 μg/ml, Sigma). Surviving cells were transferred and expanded. Control BSR cell lines were generated in the same way with the pCDM8 and pSV2 Neo. After isolation, stable lines were maintained in presence of 250 μg/ml of geneticin.

Binding assays with Sf21 cells expressing the RV glycoprotein at their surface (Gcvs‐Sf21)

Recombinant baculovirus expressing the glycoprotein from the CVS strain of RV (Gcvs) was grown in Sf21 cells (Préhaud et al., 1989). The recombinant baculovirus expressing the glycoprotein containing two mutations in antigenic site III (K330N+R333M) was described previously (Tuffereau et al., 1998). Monolayers of stable lines expressing murine p75NTR or control cell lines were grown on poly‐l‐lysine (Sigma)‐treated 60 mm culture dishes for 2 days. Sf21 cells infected with recombinant baculoviruses at a m.o.i. of 3 were radiolabelled with 20–40 μCi of 35S (Promix, Amersham, 1000 Ci/mmol) for 20 h. Sf21 cells were collected 26–30 h after infection, spun and resuspended in DMEM plus 5 mM EDTA (1–2×106 cells/ml). Before use, the radioactivity of an aliquot (200 μl) of the cell suspension was measured in a β‐scintillation counter (Rackbeta 1211, Beckman). Two millilitres of the cell suspension were added dropwise to the BSR cell monolayers and incubated for 20 min at room temperature. Unbound Sf21 cells were removed by three washes, the bound Sf21 cells were scraped into 1 ml of PBS buffer plus 10 mM EDTA and the radioactivity of a 200 μl aliquot was measured. The binding efficiency was expressed as the ratio of the number of bound cells to the total number of insect cells added to the monolayer.

Virus infections

The CVS strain of RV was propagated and titrated on BSR cells as described previously (Gaudin et al., 1992). The street RV (GH5) was kindly provided by P.Perrin and N.Tordo (Pasteur Institute, Paris). It was isolated from salivary glands of a rabid fox. The virus was passaged once in the brain of a newborn mouse and once in the brain of a hamster. The hamster's brain had been homogenized to a 30% suspension and frozen in aliquots at −70°C. This virus suspension was titrated at 107 LD50 by intracerebral inoculation of adult mice (Perrin et al., 1988). Infection of cells was performed at 37°C with a 10‐fold dilution of the homogenized brain suspension; 22 h later the cells were fixed with 3.7% formaldehyde diluted in water and immunostaining was performed with a fluorescein‐labelled anti‐nucleocapsid antibody (Pasteur Institute, Paris). Infected cells were counted under a UV microscope (Olympus B×40).

For NGF inhibition experiments, cells were pre‐incubated with mouse β‐NGF (Sigma) diluted in DMEM supplemented with 10% FBS for 30 min at room temperature, then a 10‐fold dilution of the homogenized brain suspension was added and incubated for an extra 30 min. After removal of the inoculum and β‐NGF, incubation was continued at 37°C.

Immunodetection of murine p75NTR

Cell extracts were prepared by sonicating 3×106 cells in 300 μl of TD buffer (137 mM NaCl, 25 mM Tris–HCl pH 7.4, 0.7 mM Na2HPO4) in the presence of a cocktail of protease inhibitors (CLAPA: 2 μg/ml each of chymostatin, leupeptin, antipain and pepstatin, and 16 μg/ml of aprotinin). A 20 μl aliquot was analysed by Western blotting using a rabbit serum directed against the cytoplasmic domain of the human p75NTR (Promega), then with an anti‐rabbit peroxidase‐labelled antibody (Sigma).

Co‐immunoprecipitation of p75NTR and RV glycoprotein

Cells (3×106) were labelled for 3 h with 75 μCi of 35S (Promix, Amersham, 1000 Ci/mol) in a methionine and cysteine‐free medium (ICN). Cells were lysed with 1 ml of TD buffer plus 1% CHAPS in presence of a protease inhibitor mix (CLAPA) for 30 min on ice. The nuclei were then removed by centrifugation at 12 000 r.p.m. in an Eppendorf centrifuge for 5 min. One half of the extract was incubated with two different anti‐G mAbs (3 μl of each 30AA5 and 17D2) and the second with 3 μl of anti‐p75NTR antibody for 2 h at 4°C. Then, 15 μl of protein A–Sepharose was added and incubated for 1 h at 4°C, and the beads were washed three times with TD buffer plus CHAPS. Immune complexes were analysed by PAGE followed by autoradiography. Similar immunoprecipitations with anti‐G mAbs were performed on cold extracts. Then, immune complexes were analysed by Western blotting with the anti‐p75NTR antibody.


The authors would like to thank P.Perrin and N.Tordo (Pasteur Institute, Paris) for providing the street virus. The assistance of A.Bai during the library screening is fully appreciated. Characterization of antiglycoprotein mAbs 21H8 and 20C7 was performed by H.Raux; the pSV2 Neo plasmid was obtained from B.Delmas and COS7 cells from B.Seed. A.Benmansour, Y.Gaudin, F.Lafay, H.Laude and K.Kaelin are thanked for helpful discussions and critical reading of the manuscript. This work was supported by the CNRS (UPR 9053) and through a grant from MENRT (ACCSV No. 6).


View Abstract