Rabies Virus Glycoprotein (RVG) Is a Trimeric Ligand for the N-terminal Cysteine-rich Domain of the Mammalian p75 Neurotrophin Receptor*

Rabies virus glycoprotein (RVG) is a trimeric and surface-exposed viral coat protein that has been shown to interact with the murine p75 neurotrophin receptor. We have investigated binding of RVG to p75 and describe several features that distinguish the p75-RVG interaction from conventional neurotrophin binding to p75. RVG binds mammalian but not avian p75 and does not bind to any of the Trk neurotrophin receptors. The mammalian p75 specificity of RVG binding may partly explain the phyletic specificity of rabies infection. Radioiodinated nerve growth factor (NGF) and RVG both bind to rat p75 but do not compete with each other's binding site. Although neurotrophins bind to the second and third cysteine-rich domains (CRD) of p75, RVG specifically interacts with high affinity (K d 30–35 pm) with the first CRD (CRD1). Substitution of Gln33 in p75-CRD1 by Glu completely abolishes RVG binding. Our data therefore firmly establish RVG as a trimeric high affinity ligand for a non-neurotrophin binding site on p75. Interestingly, the CRD1 in another TNF/NGF family receptor was recently shown to be involved in the binding of the herpes virus glycoprotein gD, suggesting that the CRD1 of TNF/NGF family members may be a widely used binding domain for viral glycoproteins.

dae (1). It is an enveloped virus with a single type I glycoprotein G (RVG) of 65 kDa inserted in its membrane. RVG is organized as a trimer and is responsible for the binding of the virus to the target cells and for the fusion between viral and cell membranes during the endocytosis of the virus. RV and other lyssaviruses are highly neurotropic and can cause fatal encephalomyelitis in mammals (2). Infection of both wild type and laboratory strains of RV is mainly targeted to neurons in vivo. In contrast in vitro infections with laboratory strains can target a range of cell types, and only wild type RV still shows specific neurotropism under these conditions. Specific interactions between RVG and neuronal cell surface molecules have been demonstrated (3), suggesting the existence of specific neuronal receptor(s). An expression cloning approach using soluble RVG to screen a neuroblastoma cell library led us to the identification of the murine p75 neurotrophin receptor as a candidate RVG receptor (4). Subsequent examination of six other lyssavirus genotypes revealed one additional viral glycoprotein (from European Bat lyssavirus 2) that can also bind p75 (5).
The p75 neurotrophin receptor is highly conserved in vertebrates, and its expression is developmentally regulated (6,7). During embryogenesis, p75 is present in neuronal as well as in non-neuronal cells, but in the adult its expression is essentially restricted to the peripheral and the central nervous system (8). p75 contains four cysteine-rich domains (CRD) in the N-terminal ectodomain and a type II death domain in its cytoplasmic C-terminal segment (8), both features characteristic of the TNF receptor superfamily. However, in contrast to the trimeric organization of all TNF family ligands, the NGF family of neurotrophin ligands of p75 are dimers. Neurotrophins bind to p75 with nanomolar range affinities, and their binding site has been mapped to a region encompassing both CRD2 and CRD3 (9 -11), a characteristic shared by most members of the TNF family. The physiological role of this binding in neurons is still under debate, with reported effects ranging from enhancement of axonal outgrowth through modulation of cell survival or active cell death (12)(13)(14)(15)(16)(17). Some aspects of this discussion might be resolved by the recent discovery that pro-neurotrophins also bind to p75 with higher affinities than the mature domain alone (18 -20). In addition, a number of other molecules with no sequence homology to neurotrophins have been reported to bind p75, including cysteine-rich neurotrophic factor (CRNF) from the mollusk Lymnaea (21) and aggregation-prone domains such as ␤-amyloid peptide (22)(23)(24) and a peptide from the scrapie prion protein (PrP sc ) (25). Interaction domains have not been mapped for any of these non-neurotrophin ligands, and their physiological significance is unknown.
Although cells expressing p75 become susceptible to infec-* This work was supported by CNRS funding from the Ministère de l'Education Nationale de la Recherche et de la Technologie and by research Grant QLRT-1999-00573 from the 5 th Framework Program of the European Union and the Israel Science Foundation Grant 647/01. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  tion by a wild type RV isolate, the role of p75 in natural RV infection is still a matter of debate. The p75 exonIIIϪ/Ϫ mice established by Lee et al. (26) are susceptible to RV infection (27). The significance of this finding is debatable because it was recently reported that these mice retain expression of a novel splice variant of p75 consisting of CRD1 linked to the transmembrane and intracellular domains (28). Nonetheless, a number of studies have proposed other candidate neuronal RV receptors, including the nicotinic acetylcholine receptor (29) and neuronal cell adhesion molecule (NCAM) (30). We have analyzed structural determinants of p75 binding of RVG with the hope of shedding light on the role of p75 in rabies infectivity and with the goal of determining the potential of RVG as a novel pharmacological probe of p75. Our results establish that RVG is a high affinity ligand of the first CRD of p75, and as such provides a natural trimeric ligand for a non-neurotrophin binding site on the receptor.
RVG and VSVG Preparations-Viral glycoproteins from RV and VSV were purified using slight modifications of a previously described protocol (32). Briefly, purified viruses were treated with 1% CHAPS in the presence of a protease inhibitor mixture (2 g each of chymostatin, leupeptin, antipapain, and pepstatin per ml and 16 g of aprotinin per ml) for 30 min at 37°C. Solubilized virions were overlaid on the top of a 20% sucrose cushion in PBS buffer containing 1% CHAPS and centrifuged for 1 h (SW41, 35,000 rpm, 4°C). Supernatants containing mainly glycoproteins were concentrated on a Centricon 30 (Amicon) then loaded on a 5-20% sucrose gradient in 10 mM NaCl, 10 mM Tris, pH 8, buffer containing 1% CHAPS and centrifuged for 16 h at 35000 rpm in a SW41 rotor (Beckman). Fractions containing trimeric glycoprotein corresponding to a 9 S sedimentation coefficient were pooled and loaded onto a DEAE-Trisacryl column (IBF, Biotechnics) in NaCl 10 mM, Tris 10 mM, pH 8, plus 1% CHAPS. Glycoproteins were eluted by 250 mM NaCl, 20 mM Tris, pH 7.5, plus 1% CHAPS. Extensive dialysis against PBS was performed to remove the detergent, and the protein concentration was determined by Bradford quantification. The final protein preparations consisted of pure rosettes of RVG or VSVG, averaging 10 trimers per rosette and at concentrations ranging from 0. 8 to 1 mg/ml.
Antibodies and Plasmids-A rabbit polyclonal antibody raised against the intracellular domain of p75 (Promega) was used for Western blot detection. For immunofluorescence staining, p75 was detected with either the REX polyclonal antibody, kindly provided by Louis Reichardt (University of California at San Francisco, San Francisco, CA) or with the mouse monoclonal ME20.4 (ATCC). Trk receptors were detected with a rabbit polyclonal pan-Trk antibody (PC-31, Oncogene Research Products).
Human and chicken p75 plasmids were kindly provided by B. Hempsteadt (Cornell University, NY) and G. Dechant (Max Planck Institute, Munich), respectively. Plasmids encoding human TrkA, murine TrkB, or swine TrkC receptors kinase were kindly donated by M. Barbacid (Centro Nacional de Investigaciones, Madrid). The p75 and the Trk coding sequences were sub-cloned into pcDNA1 vector (Invitrogen).
p75 Constructs-CRD deletion mutants were obtained by a recombinant PCR technique using a human p75 cDNA cloned in pcDNA1 as a template and specific primers to introduce CRD deletions within the extracellular domain of the receptor. Briefly, the CRD1, CRD1-2, and CRD1-3 mutants were constructed by PCR amplifying the N-terminal encoding segment (from the ATG to the corresponding 3ЈCRD) with the following primers: upper primer RA1 (GCGGGATCCCGGGCGATGG-GGGCAG) versus lower primers GCCAGGGATCTCCTCACACACGGT-CTGGTT (for CRD1), GCCAGGGATCTCCTCGCACACGGCGTCGTC (for CRD1-2), or GCCAGGGATCTCCTCGCACACGGTGTTCTG (for CRD1-3). These PCR products were then annealed to the end-overlapping C-terminal encoding segment (transmembrane and intracellular domains) constructed by PCR with lower primer RB5 (CGCGGATCC-TCATCACACCGGGGATGTGGC) versus upper primers AACCAGACC-GTGTGTGAGGAGATCCCTGGC (for CRD1), GACGACGCCGTGTGC-GAGGAGATCCCTGGC (for CRD1-2), or CAGAACACCGTGTGCGAG-GAGATCCCTGGC (for CRD1-3). The annealed full construct was then PCR amplified with primers RA1 and RB5, digested, and subcloned. For the CRD2-4 mutant a slightly modified strategy was used, namely amplification of the whole construct from the N-terminal of CRD2 to the stop codon, using RB5 as a lower primer versus a specific upper primer (TCCCTTGGAGGTGCCTGCCTGGACAGCGTG). This product was then annealed to a short segment encoding the ATG and signal sequence segment made with RA1 as upper primer versus a specific lower primer (CACGCTGTCCAGGCAGGCACCTCCAAGGGA). The full construct was then amplified with primers RA1 versus RB5. All constructs were subcloned into pCDNA1 and sequence-verified on both strands.
Point mutations in the CRD1 mutant were introduced using the QuikChange TM Site-Directed Mutagenesis Kit (Stratagene). The mutations were introduced in the gene by amplification using PfuTurbo TM DNA polymerase and two synthetic oligonucleotide primers bearing the desired mutation. Incorporation of the oligonucleotide primers generates a mutated plasmid containing staggered nicks. After the amplification process, the parental DNA template (methylated or hemi-methylated) was digested for 1 h at 37°by the DpnI endonuclease C, and the digestion product was used to transform MC1061 bacteria. Recombinant plasmids were then sequenced.
Protein Expression Analysis-Cos 7 cells were transiently transfected with the pcDNA1 derived plasmids using the DEAE-dextran method as described previously (4).
Indirect Immunofluorescence Staining-Transfected Cos 7 cells were washed with PBS 48 h post-transfection and incubated with REX antibody (1/1000) for 2 h at room temperature. After three washes, cells were fixed with cold paraformaldehyde 4% in PBS for 15 min and washed twice before incubation with a goat anti-rabbit rhodamineconjugated secondary antibody (Jackson ImmunoResearch) at 1/100 dilution for 1 h at room temperature. For Trk expression, rabbit antipan Trk antibody was diluted (1:100) in PBS containing 0.1% Triton X-100 and applied on paraformaldehyde-fixed cells.
Cell Extracts-Transiently transfected Cos 7 cells were analyzed for protein expression. Forty-eight hours post-transfection, they were lysed in a Tris buffer (25 mM Tris-HCl, pH 7.4, 137 mM NaCl, 0.7 mM Na 2 HPO4, 5 mM KCl)-CHAPS 1% buffer containing a protease inhibitor mixture for 30 min on ice. After clarification by centrifugation, an aliquot of the cell extract was analyzed by SDS-PAGE, and the mutated p75 proteins were detected by Western blot using an antibody to the p75 cytoplasmic domain (Promega).
Cell Surface Protein Biotinylation Labeling-Cell surface proteins from transiently transfected cells were labeled with EZ-Link TM Sulfo-NHS-LC-Biotin (Pierce) at 1 mg/ml in PBS for 30 min at room temperature. The cells were washed three times prior to lysis by 1% CHAPS. Cell lysates were centrifuged, and biotin-labeled proteins were precipitated using agarose streptavidin beads (Pierce) and then separated by SDS-PAGE. Biotinylated p75 molecules were detected on Western blot using the Promega anti-cytoplasmic domain antibody.

SDS-Polyacrylamide Gel Electrophoresis (PAGE) and
Immunoblotting-Proteins were separated on a polyacrylamide SDS gel in Tris-glycine-SDS buffer and transferred by electrophoresis onto nitrocellulose membranes (Hybond ECL, Amersham Biosciences). The membranes were blocked for 1 h in TBS buffer (50 mM Tris HCl, pH 8, 150 mM NaCl, 0.05% Tween 20) containing 2.5% (w/v) dried nonfat milk, and then incubated with 1/3000 dilution of the primary antibody in TBS buffer containing 0.5% milk. After washes in TBS buffer, the membrane was incubated with a horseradish peroxidase-conjugated donkey antirabbit secondary antibody (Sigma) (1/3000). The membrane was then washed three times and proteins were detected by enhanced chemiluminescence (ECL) (Amersham Biosciences).
Binding of RVG by the ␤-galactosidase Assay-Transfected Cos 7 cells were treated with 10 g of purified RVG in 1.5 ml of DMEM plus 10% fetal bovine serum for 2 h at room temperature. Cells were washed three times and fixed by 4% cold paraformaldehyde. Cells were then incubated with specific anti-RVG monoclonal antibodies and then with a ␤-galactosidase-conjugated goat anti-mouse antibody (Southern Biotechnology Associates), using X-Gal substrate as previously described (4).
Cross-linking Experiments-Cos 7 cells were tranfected with the hp75 plasmid, and 40 h post-transfection RVG (10 g/ml) was added for 2 h at room temperature. Cells were rapidly washed and then incubated with 2 mM DTSSP (3,3Ј-dithiobis(sulfosuccinimidylpropionate)) (Pierce) in phosphate buffer, pH 7.5, for 30 min on ice. The reaction was stopped by addition of 20 mM Tris, pH 8, for 15 min, and the cells were then washed with phosphate buffer before lysis in ice cold radioimmune precipitation assay buffer (150 mM NaCl, 50 mM Tris, pH 7.4, 0.5% sodium deoxycholate, 0.1% SDS, and 1% Nonidet P-40) containing 20 mM N-ethylmaleimide and a protease inhibitor mixture. Lysates were cleared by quick centrifugation, mixed with PAGE loading buffer with or without reducing agents, and resolved on an 8% SDS-PAGE gel run in cold buffer at low voltage. After blotting, the nitrocellulose membrane was probed with anti-p75 monoclonal antibody (ME20.4) or polyclonal anti-cytoplasmic antibody (Promega).
Iodination of RVG and Binding Assays with 125 I-RVG-Iodination was done by the lactoperoxidase method to an average specific activity of 7.5 ϫ 10 7 cpm/g. Iodinated RVG was purified by size-exclusion chromatography through a Sephadex G25 column. For steady-state competitive binding assays, cells were suspended in Eppendorf tubes with 20 ng/ml 125 I-RVG (ϳ30 pM, assuming one available binding trimer per rosette) in PBS supplemented with 1 mg/ml bovine serum albumin, 1 mM MgCl 2 and 0.5 mM CaCl 2 , in the presence or absence of cold competitors. Binding was allowed to occur with gentle rocking at 4°C for at least 2 h, and the bound ligand was separated from free ligand by rapid pelleting of the cells in an Eppendorf microcentrifuge at 4°C. Background binding was determined using a 400-fold excess of unlabeled RVG.
Homology Modeling-A structural model for human p75 was built with the program Modeler (33) using a high resolution extracellular domain structure of TNFR (PDB code 1ext) (34) as a template. Sequence alignment was performed with PSI-BLAST (35) and manually adjusted before input into Modeler.

RVG Binds Specifically to Mammalian p75 but Not Mammalian Trks-We first investigated if RVG interacts with p75
from different species (human, mouse, and chicken). Cos 7 cells were transiently transfected, and the cell surface expression of p75 was verified by indirect immunofluorescence (Fig. 1A). In all cases, p75 was detected at the cell surface; however, the intensity of the fluorescence was weaker in the case of chicken p75 (cp75) than in the case of human (hp75) or mouse (mp75). Binding of RVG to the different p75 species was assessed by immunostaining of RVG-challenged cells with an anti-RVG monoclonal, followed by a ␤-galactosidase-coupled secondary antibody. This protocol was previously shown to be specific (4), and no blue staining was observed when RVG or anti-RVG monoclonal antibodies were omitted. Only cells expressing human or mouse p75 had a blue appearance (Fig. 1), whereas RVG binding could not be detected on cells expressing chicken p75.
We then investigated if RVG was also able to bind to the high affinity tyrosine kinase receptors for neurotrophins (TrkA, TrkB, or TrkC). Trk expression in transiently transfected Cos 7 cells was confirmed with a pan-Trk antibody, and RVG binding was then measured as described above. As shown in Fig. 1, although all three Trk receptors were expressed, no RVG bind-ing was observed. In addition no RVG binding could be observed on stable cell lines expressing murine TrkB or TrkC (data not shown).
RVG Binds Rat p75 at High Affinity and Does Not Compete with NGF Binding-Our next step was to quantify the affinity of RVG binding to p75 in PC12 cells. RVG was iodinated to a specific activity of 7.5 ϫ 10 4 cpm/ng by the lactoperoxidase method, and its binding to PC12 cells was measured on its own or in the presence of excess cold RVG, NGF, or VSVG. Parallel experiments were performed with iodinated NGF for comparison. Both ligands bound to PC12 cells with good ratios of total to nonspecific binding ( Fig. 2A), and the control VSVG protein did not displace either ligand. Interestingly, excess cold NGF could not compete with 125 I-RVG binding, nor could excess cold RVG displace 125 I-NGF. Similar results were obtained with PC12-nnr5 cells, which express p75 but not Trk (data not shown), suggesting that the RVG binding observed in PC12 is to the p75 receptor. RVG binding on PC12 cells was found to be a high affinity interaction, with a K d of 35 Ϯ 5 pM (Fig. 2B), assuming one binding trimer per rosette of RVG.
Different Oligomeric States of p75-We then investigated the oligomerization state of p75 in the presence or absence of RVG because it was recently suggested that p75 might form trimers in the absence of ligand (23). In p75-overexpressing cells, both monomeric and oligomeric forms of the receptor were observed, with the latter migrating as apparent trimers (Fig. 3A). The putative trimer was present in all extraction conditions used and in transiently transfected Cos 7 cells its proportion increased with time (data not shown). After cross-linking an additional band migrating as an apparent dimer was detected in non-reducing conditions (Fig. 3B) and was absent in reducing conditions while some of the apparent trimers were still observed (Fig. 3C). Upon cross-linking after incubation with RVG, all three p75 bands were diminished, and an additional high molecular weight complex that does not enter the gel was observed (Fig. 3B). This unresolved complex is consistent with the high molecular mass of the RVG rosette, thus unfortunately not allowing definitive identification of a preferred oligomeric species of p75 for RVG binding.
The RVG Binding Site Is on the CRD1 of p75-To map the binding site for RVG on p75, we constructed different deletion mutants of the CRDs in human p75 (hp75). To maintain correct folding and processing of the mutants, the integrity of each individual CRD was preserved. The deletion mutants con-tained one, two, or three CRDs connected via the O-glycosylated linker to the transmembrane and cytoplasmic domains of hp75 (Fig. 4A). Cos 7 cells were transfected with the different plasmids, and expression of the mutants was analyzed by Western blot with an anti-cytoplasmic p75 antibody (Fig. 4B). The different deletion mutants migrated on SDS-PAGE according to their predicted molecular mass. In the case of the CRD1only mutant, the protein migrated as a doublet. Cell surface expression of all the mutants was confirmed by immunofluo-  3. Different oligomeric states of p75. A, SK-N-BE (SK) cells stably expressing p75 (SKp75) and Cos 7 cells transiently transfected with hp75 plasmid for 40 h were lyzed in radioimmune precipitation assay buffer (lanes 1-4) containing 20 mM N-ethylmaleimide. Cell lysates were mixed with Laemmli loading buffer without reducing agents, resolved on SDS-PAGE at low voltage in the cold, and transferred to nitrocellulose for detection of p75 with the ME20.4 antibody. Cos 7 cells transiently transfected with hp75 plasmid were incubated with RVG (10 g/ml) prior to cross-linking with 3,3Ј-dithiobis(sulfosuccinimidylpropionate) for 30 min on ice and then lyzed in radioimmune precipitation assay buffer. B, samples were analyzed as in Fig. 3A. C, cell lysates were mixed with loading buffer containing reducing agent, and the blot was treated with the anti-p75 polyclonal antibody (Promega).
Glu 33 in CRD1 Is Essential for RVG Binding-To further map RVG binding on p75, a structural homology model of p75 (Fig. 5A) was built based on the TNFR-I x-ray structure (36). The extracellular domains of the two molecules share a sequence identity of only 25% over 152 residues. However, the conservation of more than 20 cysteines in the TNFR family, and the use of a position-specific alignment algorithm (35) allowed a reliable alignment, which is a prerequisite for accurate homology modeling (33). Although the relative orientations of the CRDs are likely to be different in p75 and TNFR-I, the overall structure of each CRD can be assessed with confidence. Previous modeling studies of p75NTR (37, 38) have focused on CRD2, CRD3, and CRD4, being concerned with the possible interactions with neurotrophins. A closer view of CRD1 is presented here (Fig. 5B), showing that several polar or charged residues (Glu 14 , Lys 17 , the N-glycosylation site Asn 32 , and Gln 33 ) cluster on the same side of the molecule near Ala 31 , a residue that is a valine in chicken p75 (Fig. 5C). Point mutations were introduced at those positions and at those differing between chicken and human p75 (Ala 3 , Gly 7 , Ala 26 , Ala 31 , and Ala 26 ϩ Ala 31 ). The sequences of the different mutants were confirmed, and their expression was verified by cell surface biotinylation as well as Western blot analysis using an anti-cytoplasmic p75 antibody (Fig. 6A). Two bands were detected, the major one represents the mature protein migrating around 46 kDa, and the minor one (noted by an asterisk) represents the immature form of the protein. The different mutant proteins were all expressed at the cell surface (Fig. 6A), albeit at somewhat lower levels for the mutants A3K, G7K, and A31V. For the N32Q mutant, good cell surface expression was demonstrated by efficient surface biotinylation, despite the lack of N-linked glycosylation. Cell surface expression of the mutants was also confirmed by immunofluorescence as shown for four examples (E14Q, K17Q, N32Q, and Q33E) in Fig. 6B. RVG binding was assessed by the ␤-galactosidase method on cells expressing CRD1 mutants (Fig. 6C). Most of the point mutations had little or no effect on RVG binding, apart from the Q33E mutant, which totally lost RVG binding. Apparently reduced ␤-galactosidase staining was observed for E14Q and K17Q. None of the chicken-based substitutions, even the double A26VϩA31V, were able to abolish RVG binding. To quantify the apparent differences in RVG binding observed for some of the mutants, we conducted steady-state competitive binding assays with 125 I-RVG. Specific binding was observed to wild type CRD1 and to all the point mutants tested except for Q33E (Fig. 7A). Thus the only mutant that conclusively lost RVG binding capacity is at the Gln 33 position. As shown in Fig. 6B, RVG bound to the CRD1-only deletion construct of human p75 with high affinity (K d ϭ 32 pM Ϯ 6, n ϭ 3) and in good correlation with the affinity previously determined for endogenous full-length rat p75 on PC12 cells (Fig. 2B). The three mutants with apparently changed binding in the ␤-galactosidase assay turned out to have remarkably similar affinities for RVG as wild type CRD1 (Fig. 7B). The only parameter with significant differences in the Scatchard plots is the observed B max , indicating lower expression levels for E14Q and K17Q. This lower expression most likely accounts for the reduced staining originally observed in the ␤-galactosidase assay. DISCUSSION The results reported here show that RVG is a high affinityspecific trimeric ligand of the N-terminal CRD of the p75 neurotrophin receptor and as such differs from all previously characterized p75 ligands. The first obvious difference is in the domain of the receptor to which RVG binds. Neurotrophins are thought to bind to CRD2 and CRD3 (9 -11), and a similar binding pattern has been observed for most TNF family ligands (39). In contrast, CRD1 is both necessary and sufficient for RVG binding (Figs. 4, 6, and 7). The difference in the binding sites of RVG and NGF is confirmed by the lack of competition between the two ligands (Fig. 2). This observation is especially striking given the large size of the viral protein rosette. Although the precise binding site of cysteine-rich neurotrophic factor, prion peptide, or ␤-amyloid peptide on p75 has not been determined, all these ligands compete with NGF (21)(22)(23)(24)(25), indicating that their binding site is also likely to be distinct from that of RVG.
The second and extremely striking difference lies in the high avidity of RVG rosettes for p75, with K d of 32-35 pM (Figs. 2  and 6). The apparent affinities of most other p75 ligands meas- ured to date are above 2 nM, and even the recently described high affinity binding of pro-NGF to p75 (18) is 5-10-fold less avid than that of RVG. We should note that all our RVG binding affinities are calculated based on the assumption of one binding trimer per rosette. This assumption is based on steric considerations of accessibility between RVG in the rosette and p75 on the cell surface and is supported by independent estimations of the affinity of RVG for p75 using soluble monomeric RVG in surface plasmon resonance assays. 2 The third intriguing aspect of RVG as a p75 ligand is that it is naturally trimeric. Although all members of the TNF ligand family are trimers, and their receptors signal as trimerized complexes, p75 is thought to signal as a dimer when bound by neurotrophins or in a complex with Trk. This signaling is typically rather weak as compared with TNFR signaling (19), perhaps due to a non-optimal conformation of dimerized p75. Indeed, our cross-linking data suggests that a dimeric form of p75 might not be the most stable oligomer of this receptor (Fig.  3). Aggregation of NGF on glass beads has been reported to induce more robust activation of p75, perhaps due to forcing of higher order oligomers (40). Induced trimerization of the p75 death domain was previously attempted in the context of a Fas-p75 chimera (41). Although this chimeric construct did not reveal enhanced signaling, this could be due to a requirement for the p75 extracellular domain for optimal trimerization. Although a recent study (23) purports to show trimers of p75, and we also observe oligomeric forms migrating as apparent trimers (Fig. 3), these observations do not provide definitive proof for physiological trimerization of p75. Use of RVG as a natural trimeric ligand for p75 may provide an interesting avenue to address this issue in the future.
Our data also show that RVG binds to mammalian (mouse, rat, and human) but not to avian (chicken) p75 (Fig. 1) 6. RVG binding to CRD1 mutants. A, point mutations were introduced in the CRD1 plasmid, and protein was expressed in transfected Cos 7 cells. Cells were surface-biotinylated, lysed, and extracts were precipitated with streptavidin beads, followed by p75 detection by Western blot. B, surface immunofluorescence for the mutants K14Q, E17Q, N32Q, and Q33E. C, RVG binding assayed with the ␤-galactosidase assay. proposed as receptors for rabies virus, including nicotinic acetyl chain receptor (29) and neuronal cell adhesion molecule (30). Furthermore, the p75 exonIIIϪ/Ϫ mice established by Lee et al. (26) were recently shown to be susceptible to RV infection (27). However, the significance of the latter finding is not clear, because these mice retain expression of a novel CRD1-containing splice variant of p75 (28), which we now have shown to be sufficient for RVG binding (Fig. 4). A definitive assessment of the importance of p75 in rabies infection will have to await the results of additional studies, including experiments using wild type virus and the recently described p75 exonIVϪ/Ϫ mice (28) or normal mice and RV mutants deficient in p75 binding.
Site-directed mutagenesis of specific amino acid residues in CRD1 showed that replacement of Gln 33 by a glutamic acid completely abolishes RVG binding. In contrast, substituting the adjacent Asn 32 by a glutamine had no effect, highlighting that it is specifically the nature of the residue at position 33 that is crucial for binding. Gln 33 was previously flagged by Chapman and Kuntz (37) as one of two side chains for potential hydrogen bonding on the surface of CRD1, the other being Lys 17 . By contrast, Asn 32 and Ala 31 would be at the very tip of CRD1, pointing away. Likewise, Ala 26 would be on the same side of CRD1 as Gln 33 but further down this module. These results thus pinpoint a site on the surface of p75-CRD1 that is essential for RVG binding. They also establish that positions 26 and 31 are not involved in RVG binding nor are any of the other residues that differ between human and chicken p75 in this region. We can only surmise that differences in the rest of the extracellular domain change the relative orientation of the CRDs and make an identical surface inaccessible in chicken p75. Additional studies will be required to uncover the molecular basis of the differences between human and chicken p75 regarding RVG binding. Interestingly, the residue corresponding to Gln 33 in TNFR-I lies just before the dimerization surface observed for the parallel TNFR-I dimer as in Ref. 34 (the equivalent of Lys 17 being part of the interface). If p75 forms a similar dimer at the membrane surface, a slight shift of the orientation of CRD1 may greatly reduce its accessibility to the bulky RVG.
It is interesting to note that p75 and RVG are not the only examples of a virus ligand for a TNFR family receptor. Two additional examples have been reported in the literature, namely binding of the gD herpes virus glycoprotein to the TNFR-I-related HVEM (herpes virus entry mediator) receptor (42) and binding of the TRAIL-related CAR1 receptor (cytopathic avian leukosis-sarcoma virus Receptor) to the envelope protein of retrovirus avian leukosis virus subgroups B and D (43,44). Another common parameter is that the receptor CRD1 domain is involved in binding of all these viral glycoproteins (44 -46). The structure of the ectodomain of herpes simplex virus glycoprotein D bound to HVEM (47) shows interactions of the viral protein mostly with the CRD1 of the receptor, with three residues from CRD2 completing the interface. Interestingly, the region of HVEM that forms the bulk of the interactions with gD corresponds to residues 17-39 of p75. Most contacts are through the main chain of HVEM, and the side chain of the equivalent of Gln 33 of p75 (a threonine in HVEM) points away from the viral glycoprotein. Thus, viral glycoproteins bind on the outer-most domain of TNFR family receptors, whereas endogenous ligands target residues located in the inner CRD2 and 3 domains. This observation can be explained by the fact that viral glycoproteins are much bulkier than endogenous TNF or NGF family ligands and that their anchorage in the viral envelope places additional steric constraints on their ability to interact with non-exposed residues on the cell surface. Finally it should be noted that it has been suggested that CRD1 acts as a pre ligand assembly domain (PLAD) for TNF family receptors (47). Binding of viral glycoprotein to CRD1 might therefore disrupt receptor oligomers preassembled through their PLAD, thus interfering with basal signaling events. The oligomerization state of gD is still not established (48), but the envelope proteins of both RV (32) and ALV (49) form trimers at the viral surface.
To conclude, we have shown that RVG is a specific high affinity ligand for a non-neurotrophin binding site on the p75 receptor. The existence of a specific high affinity trimeric ligand for p75 will be useful for future work on the pharmacology and physiology of this receptor.