Prion and doppel proteins bind to granule cells of the cerebellum

Abstract

We reported that expression of the cellular prion protein (PrP^C) rescues doppel (Dpl)-induced cerebellar degeneration in mice. To search for protein(s) that mediate this process, we fused the C-termini of mouse (Mo) PrP and Dpl to the Fc portion of an IgG. Although both MoPrP-Fc and MoDpl-Fc bound to many regions of the brain, we observed restricted binding to granule cells in the cerebellum, suggesting a scenario in which granule cells express a protein that mediates Dpl-induced neurodegeneration. Because granule cells do not express PrP^C, it seems unlikely that MoPrP-Fc binding reflects a ligand that is involved in the conversion of PrP^C into PrP^Sc, the disease-causing isoform. In contrast, the dominant-negative MoPrP(Q218K)-Fc not only binds to granule cells but also binds to neurons of the molecular layer where PrP^C is expressed. These findings raise the possibility that the cells of the molecular layer express an auxiliary protein, provisionally designated protein X, which is involved in prion formation and is likely to be distinct from the protein that mediates Dpl-induced degeneration. Although the binding of the dominant-negative MoPrP(Q218K)-Fc to cells in the molecular layer expressing PrP^C is consistent with a scenario for the binding of MoPrP(Q218K)-Fc to protein X, the absence of PrP^Sc deposition in the molecular layer requires that PrP^Sc, once formed there, be readily transported to the cerebellar white matter where PrP^Sc is found. Identifying the ligands to which PrP-Fc, Dpl-Fc, and dominant-negative PrP bind may provide new insights into the functions of PrP^C and Dpl as well as the mechanism of PrP^Sc formation.

The Prn gene family in mammals consists of two members: the prion protein (PrP) gene (Prnp) and the doppel protein (Dpl) gene (Prnd; ref. 1). PrP and Dpl genes encode paralogs: the two genes seem to have arisen from an ancient gene duplication. In mammals examined to date, the protein sequences of PrP and Dpl are ≈25% homologous, whereas the structures of the two proteins are quite similar (1, 2).

Both the cellular isoform of the prion protein (PrP^C) and Dpl are glycoproteins that are bound to the cell surface through a glycosylphosphatidylinositol (GPI) anchor (3–6). Neither the cellular function of PrP^C nor Dpl is known, but both proteins cause CNS degeneration but through different mechanisms. Before causing neurological disease, PrP^C undergoes a profound conformational change as it is converted into the disease-causing isoform, PrP^Sc (7). The accumulation of PrP^Sc to sufficiently high levels leads to CNS degeneration (8). In contrast, the expression of Dpl is restricted to the testes, but when Dpl is expressed in the CNS, it is neurotoxic (9).

Structural studies employing bacterially expressed PrPs have shown that a large portion of the N terminus is unstructured (10). The N terminus of PrP contains octarepeat sequences (PHGGGWGQ) that coordinate copper ions (11, 12), and the binding of copper has been reported to stimulate endocytosis of PrP (13). PrP-deficient (Prnp^0/0) mice have been reported to have diminished levels of copper in their brains and altered levels of superoxide dismutase activity (14, 15), but these findings could not be confirmed (16). Among other suggested functions for PrP^C are roles in neuritogenesis (17) and cell adhesion (18).

Initial studies with two lines of Prnp^0/0 mice revealed no abnormalities (19, 20), but a third line developed late-onset ataxia caused by Purkinje cell loss (21). The ataxia and Purkinje cell degeneration were prevented in the third Prnp^0/0 line by expressing a transgene encoding wild-type (wt) mouse (Mo) PrP in these mice (22). Subsequent studies suggested that the ataxia and Purkinje cell degeneration were likely caused by expression of Dpl in the cerebellum of these mice (1, 23). To test the hypothesis that expression of Dpl in the CNS causes neuronal degeneration, transgenic (Tg) mice expressing Dpl in the brain were constructed (9). The expression of Dpl in the CNS was inversely proportional to the age at which ataxia as well as Purkinje and granule cell degeneration were observed. Coexpression of Dpl and PrP in Tg mice rescued the mice from the Dpl-induced ataxic phenotype, arguing that expression of PrP neutralized the toxic effect of Dpl either by directly interacting with Dpl or through another protein (9). To explore these possibilities, we constructed PrP and Dpl fusion proteins. To the C-termini of MoPrP and MoDpl, the antibody fragment with the constant region (Fc) of human IgG was fused. In order for these fusion proteins to be glycosylated and the proper disulfide bonds to be formed, we expressed the proteins in mouse neuroblastoma (N2a) cells.

To assess the conformation of the PrP portion of MoPrP-Fc, we used a panel of recombinant chimeric human–mouse (HuM) antibody fragments (Fabs). Once we found that the immunoreactivity of MoPrP-Fc was similar to that of MoPrP^C, we used the fusion proteins as probes to stain cryostat sections of mouse brain. The staining procedure is similar to histoblotting, developed for anti-PrP antibodies (24). Unexpectedly, we found intense staining in the granule cell layer of the cerebellum with either wtMoPrP-Fc or wtMoDpl-Fc. This finding argues that both proteins bind either to the same ligands or to different ligands on granule cells. Because the granule cells did not stain with anti-PrP antibodies, we conclude that the ligand(s) could not be PrP.

To extend these findings, we created a PrP-Fc fusion protein with a dominant-negative mutation (25, 26). The Q218K mutation prevents MoPrP^C(Q218K) from being converted into mutant PrP^Sc; moreover, MoPrP(Q218K) acts as a dominant-negative in preventing wt MoPrP^C from being converted into wt PrP^Sc when both the wt and mutant proteins are coexpressed. This dominant-negative mutation was first reported as a common polymorphism in the Japanese population, who seem to be protected from developing Creutzfeldt-Jakob disease (27).

Because wtMoPrP-Fc bound to granule cells of the cerebellum, which do not express PrP^C, it seems unlikely that wtMoPrP-Fc binding reflects a ligand that is involved in the conversion of PrP^C into PrP^Sc. In contrast, the dominant-negative MoPrP(Q218K)-Fc not only binds to granule cells but also binds to the neurons of the molecular layer where PrP^C is expressed. These findings raise the possibility that the cells of the molecular layer express an auxiliary protein or proteins involved in prion formation, which is provisionally designated protein X and is likely to be distinct from a protein that may mediate Dpl-induced degeneration. The search for protein X has been frustrating because many proteins are known to bind to PrP^C, but none have been shown to participate in PrP^Sc formation (17, 18, 28–36). Although the binding of the dominant-negative MoPrP(Q218K)-Fc to cells in the molecular layer where PrP^C is expressed builds a case for the possibility that MoPrP(Q218K)-Fc is binding to protein X, the absence of PrP^Sc deposition in the molecular layer requires us to hypothesize that PrP^Sc formed in molecular layer is readily transported to the cerebellar white matter where PrP^Sc is found. Use of these Fc fusion proteins may provide new insights into the functions of PrP^C and Dpl as well as the mechanism of PrP^Sc formation.

Materials and Methods

Preparation of HuM-Fabs.

The preparation of HuM-Fabs has been described (37, 38).

Cell Culture.

Mouse N2a cells were cultured as described (38).

Cloning and Production of Fusion Proteins.

The mouse Prnp^a sequence was amplified by PCR and cloned between NdeI and XbaI sites of pSecTag plasmid (Invitrogen) containing the human IgG1-Fc region. The resulting construct, denoted wtMoPrP-Fc, contains amino acids 23–231 of MoPrP^C fused at the N terminus of the human IgG1-Fc region. For MoPrP(Q218K)-Fc, mutant PrP DNA was generated by following procedures of the QuickChange Site-Directed Mutagenesis Kit (Stratagene). wtMoDpl-Fc was constructed by cloning the MoDpl sequence 26–155 between NdeI and XbaI sites of pSecTag plasmid (Invitrogen) containing the human IgG1-Fc region.

N2a cells were transiently or stably transfected with appropriate DNA constructs (5 or 10 μg) by using the DOTAP DNA transfection kit (Boehringer Mannheim, Switzerland). Stably transfected N2a cells with wtMoPrP-Fc, MoPrP(Q218K)-Fc, or wtMoDpl-Fc constructs were maintained as described above, and the media was supplemented with zeocin at 200 μg/ml.

Cells were lysed in buffer T (10 mM Tris⋅HCl, pH 8.0/0.5% deoxycholate/0.5% Nonidet P-40/150 mM NaCl). Cell lysates or cell-conditioned media containing the fusion proteins were resolved by SDS/PAGE. Samples were blotted onto poly(vinylidene difluoride) (PVDF) membranes and blocked with 5% (wt/vol) nonfat milk protein in Tris-buffered saline with 0.05% Tween-20 (TBST). The PrP^C moiety was detected by using 1 μg/ml of either the HuM-D18 or HuM-D13 antibody in TBST. The Fc portion of the protein was detected by using anti-human Fc antibodies conjugated to horseradish peroxidase (HRP; Sigma) at various concentrations in TBST. Blots were developed with the enhanced chemiluminescence (ECL) reagent (Amersham Pharmacia) for 1 min and exposed to ECL Hypermax film (Amersham Pharmacia). PNGase F digestions were performed as suggested by the manufacturer (Glyco, Williamsport, PA).

ELISA for wtMoPrP-Fc and MoPrP(Q218K)-Fc.

A sandwich ELISA procedure was used for epitope-binding analysis, using recombinant anti-PrP HuM-Fabs. Briefly, 100 μl of goat α-human Fc at 20 μg/ml in 0.1 M NaHCO₃, pH 8.6, were plated overnight at room temperature (RT) in each well of a 96-well ELISA plate (Dynex, Franklin, MA). The plate was washed three times with TBST. For each well, 200 μl of a saturating solution (0.25% BSA/0.05% Tween-20 in PBS) was added and incubated for 1 h at RT, followed by five washes with TBST.

The PrP-Fc supernatants were mixed with 50 μl of the saturating solution containing 1 μg of the specific HuM-Fab, the solution was incubated for 15 min at RT, and then dispensed at 100 μl per well. The PrP-Fc/HuM-Fab solution was incubated for 2 h at RT, followed by five washes with TBST. To each well, 100 μl of INDIA-HRP (Pierce) was added at 1:1,000 dilution in 1% BSA in TBS for 1 h at RT. The plate was washed five times with TBST. Finally, 100 μl of TMB solution (Pierce) was added to each well and incubated for 15 min at RT. The developing reaction was terminated by adding 100 μl of 2 M H₂SO₄ to each well. The ELISA plate was read at 450 nm.

Neuropathology.

Animals were killed, their brains were removed and frozen at −80°C. Histoblots were performed on coronal brain sections (10 μm), transferred onto a nitrocellulose membrane, and processed for immunohistochemistry as described (24). HuM-D13 or HuM-D18 (1:1,000) monoclonal antibodies were used to immunostain PrP^C. Sites of ligand binding with wtMoPrP-Fc, MoPrP(Q218K)-Fc, and wtMoDpl-Fc were detected with an anti-human Fc polyclonal antibody conjugated to alkaline phosphatase (Sigma). Undiluted wtMoPrP-Fc and wtMoDpl-Fc in the culture media were incubated with the histoblot at 4°C overnight. The anti-human Fc secondary antibody with the attached alkaline phosphatase was incubated at 1:5,000 for 1 h at RT. Color was developed by standard procedures.

Results

wtMoPrP-Fc, MoPrP(Q218K)-Fc, and wtMoDpl-Fc Expression in N2a Cells.

Wild-type MoPrP(23–231) and wtMoDpl(26–155) sequences were inserted into an expression plasmid containing the Fc region of human IgG1 to produce the wtMoPrP-Fc and wtMoDpl-Fc constructs, respectively. Mutant MoPrP(23–231,Q218K) was inserted into the same expression plasmid to produce MoPrP(Q218K)-Fc.

The resulting plasmids were transfected independently into N2a cells, and the proteins were secreted into the media. Wt MoPrP-Fc, wtMoDpl-Fc, and MoPrP(Q218K)-Fc were expressed as dimers as detected by Western blot (data not shown). The dimerization was conferred by the Fc portion and not by either wtMoPrP or wtMoDpl. A control Fc protein was also secreted in the media of an N2a cell culture and detected by Western blotting (data not shown). PNGase F digestions showed that the fusion proteins were glycosylated as expected (data not shown).

Immunochemical Characterization of wtMoPrP-Fc by Using HuM-Fabs.

The wtMoPrP-Fc and MoPrP(Q218K)-Fc fusion proteins were characterized by using a panel of recombinant monoclonal HuM-Fabs, including HuM-E123, HuM-E149, HuM-D13, HuM-D18, HuM-R72, HuM-R1, and HuM-R2 (38–40). We used a sandwich ELISA assay that enables the dimeric fusion protein to retain its orientation and folding. Relative binding capacity at saturating concentrations of HuM-Fabs was used as a molecular signature for the folding of wtMoPrP within MoPrP-Fc and MoPrP(Q218K)-Fc. For wtMoPrP-Fc, as expected, HuM-D18, HuM-D13, and HuM-R1 resulted in the most avid binding (Fig. 1A), arguing that the epitopes to which these HuM-Fabs bind are exposed to the surface of the protein. HuM-R2 showed slightly less binding capacity (Fig. 1A), although its epitope [MoPrP(225–231)] is identical to that of HuM-R1 (Fig. 1A). At the N terminus, we found that HuM-E123, which recognizes the region of MoPrP(30–37), bound weakly, and that HuM-E149, a HuM-Fab specific to the octarepeat region, bound even more weakly than HuM-E123 (Fig. 1A). As a control, we used HuM-R72, which only binds to MoPrP when it is completely unfolded, suggesting that the epitope [MoPrP(152–163)] is buried in MoPrP^C. Not surprisingly, HuM-R72 did not bind to either wtMoPrP-Fc or MoPrP(Q218K)-Fc.

Fig 1.

ELISA measurements of the binding of various HuM-Fabs (x axis) to wtMoPrP-Fc (A) and MoPrP(Q218K)-Fc (B). Each HuM-Fab was read in triplicate (error bars are indicated). The control indicates background reading without HuM-Fab. Data shown are representative of at least five independent experiments.

The binding pattern found in the sandwich ELISA experiments was consistent with flow cytometry analysis experiments using the same panel of HuM-Fabs for wtMoPrP^C on the plasma membrane of N2a cells (38).

When we then analyzed the relative binding capacity of the HuM-Fabs to MoPrP(Q218K)-Fc, we found a change at the N terminus and not at the C terminus, as would be expected for the Q218K mutation. All HuM-Fabs reacted with the same avidity as they did with wtMoPrP-Fc, except for HuM-E123, which recognizes an N-terminal epitope and for which the binding was diminished nearly 50% (Fig. 1B). This diminution in the binding of HuM-E123 may be caused by the interaction of the N terminus of PrP with residues near the C terminus (10), which is enhanced by the Q218K mutation (41).

Localization of Ligands for wtMoPrP-Fc, MoPrP(Q218K)-Fc, and wtMoDpl-Fc.

Wild-type MoPrP-Fc and wtMoDpl-Fc were used as probes in the search for ligands in the CNS. An anti-human Fc polyclonal antibody conjugated to alkaline phosphatase was used to localize sites of binding in coronal sections of the brain. At the level of the hippocampus and thalamus, both bound to the neuropil in all gray matter regions; however, they did not bind to nerve cell bodies in those regions (Fig. 2 A and B). For example, the most intense signals with both probes were in the neuropil of the hippocampus. In contrast, there was weak binding to the cell bodies of neurons in the dentate gyrus and Ammon's horn. The MoPrP(Q218K)-Fc probe differed from the wtMoPrP-Fc probe by binding to nerve cell bodies in addition to the gray matter neuropil. Nerve cell body binding is seen best in the hippocampus (Fig. 2C). Staining with the HuM-D13 antibody demonstrates the location of PrP^C in the brain of an uninfected, wt FVB mouse (Fig. 2D). The location of PrP^C is similar to the sites of wtMoPrP-Fc and wtMoDpl-Fc binding; PrP^C is localized to the gray matter neuropil but is absent from nerve cell bodies in the hippocampus. HuM-D13 immunostained the white matter of the corpus callosum, whereas wtMoPrP-Fc, wtMoDpl-Fc, and MoPrP(Q218K)-Fc did not bind. The Fc control failed to show any binding (Fig. 2F). We also compared the localization of wtMoPrP-Fc, wtMoDpl-Fc, and MoPrP(Q218K)-Fc binding to PrP^Sc deposition in the brain of an FVB mouse killed 132 days after inoculation with RML prions. Immunohistochemistry of proteinase K-digested and guanidinium-denatured histoblot sections using HuM-D13 Fabs showed accumulation of PrP^Sc in white matter and selected regions of the gray matter, some of which are different from those bound by the Fc probes (Fig. 2E). Many of the regions in uninfected control brains that stained for PrP^C with HuM-D13 did not stain for PrP^Sc (Fig. 2 D and E). When wtMoPrP-Fc, wtMoDpl-Fc, and MoPrP(Q218K)-Fc were used to stain sections of RML-infected mouse brains, the staining was indistinguishable (data not shown) from that found in uninfected brains (Fig. 2 A–C).

Fig 2.

Histoblot of coronal sections of FVB mice at the level of the dorsal hippocampus/thalamus stained with wtMoPrP-Fc (A), wtMoDpl-Fc (B), MoPrP(Q218K)-Fc (C), or Fc (F). Shown is staining of sections of the hippocampus/thalamus with the anti-PrP HuM-D13 Fab from an age-matched FVB mouse (D) and an RML-infected FVB mouse at 132 days after inoculation (E). (Bar = 2 mm.)

The wtMoPrP-Fc, wtMoDpl-Fc, and MoPrP(Q218K)-Fc fusion proteins also were used to probe cryostat sections of the cerebellum in FVB mice, CD-1 mice (data not shown), and Prnp^0/0 mice, as well as Tg mice expressing high levels of MoPrP^C [designated Tg(MoPrP-A)4053/FVB mice; data not shown and ref. 42]. Binding of the wt Fc probes to coronal sections of the cerebellum in wt FVB mice is shown in Figs. 3 and 4 and in Prnp^0/0 mice in Fig. 5. Binding of wtMoPrP-Fc and wtMoDpl-Fc to cryostat sections of cerebellum was indistinguishable for four lines of mice described above, indicating that neither the absence nor the overexpression of PrP influenced the distribution of the PrP and Dpl ligand(s). PrP and Dpl ligand binding was confined to the granule cell layer of the cerebellar cortex (Figs. 3–5). In contrast, immunostaining with HuM-D18 was widespread throughout the CNS in both gray and white matter (Figs. 3D and 4C). Strikingly, the granule cell layer stained by wtMoPrP-Fc and wtMoDpl-Fc was devoid of staining for PrP^C; conversely, the molecular layer and white matter of the cerebellar cortex stained for PrP^C but was devoid of staining for PrP and Dpl ligands.

Fig 3.

Histoblots of coronal brain sections of FVB mice show ligand binding in the granule cell layer. Sections of the cerebellum/pons of FVB mice were probed with wtMoPrP-Fc (A), wtMoDpl-Fc (B), or Fc (C). As a control, staining of a section of the cerebellum/pons of an age-matched FVB mouse with the anti-PrP HuM-D18 Fab (D) is shown. (Bar = 1 mm.)

Fig 4.

Enlargements of the granule cell layer (GCL), molecular layer (ML), and white matter (WM) region of the coronal section of the cerebellum/pons area of FVB mice. Sections were stained by using wtMoPrP-Fc (A), wtMoDpl-Fc (B), and HuM-D18 (C). (Bar = 200 μm.)

Fig 5.

Histoblots of coronal brain sections of FVB/Prnp^0/0 mice show ligand binding in the granule cell layer. Sections of the cerebellum/pons of FVB/Prnp^0/0 mice were probed with wtMoPrP-Fc (A), wtMoDpl-Fc (B), or Fc (C). As a control, staining of a section of the cerebellum/pons of an age-matched FVB/Prnp^0/0 mouse with the anti-PrP HuM-D18 Fab (D) is shown. (Bar = 1 mm.)

Differential Binding of MoPrP(Q218K)-Fc and wtMoPrP-Fc in the Cerebellum.

The mutant MoPrP(Q218K)-Fc fusion protein was used to stain coronal sections of the cerebellum of FVB mice. MoPrP(Q218K)-Fc-ligand binding was different from that of wtMoPrP-Fc. In contrast to wtMoPrP-Fc, staining by MoPrP(Q218K)-Fc was not limited to the granule cell layer but included the molecular cell layer (Fig. 6A). The same pattern of MoPrP(Q218K)-Fc binding was observed in Prnp^0/0 mice, indicating that the absence of endogenous PrP^C does not influence the binding of this fusion protein (Fig. 6B). With HuM-D18 Fabs, the presence of PrP^C in the molecular layer and the white matter was demonstrated (Fig. 6C). We also compared the localization of MoPrP(Q218K)-Fc binding to PrP^Sc deposition in the cerebellum of an FVB mouse killed 132 days after inoculation with RML prions. Immunohistochemistry of proteinase K-digested and guanidinium-denatured histoblot sections using HuM-D18 Fabs showed accumulation of PrP^Sc in white matter, but both the granule cell and molecular layers were devoid of PrP^Sc (Fig. 6D).

Fig 6.

Enlargements of the granule cell layer (GCL), molecular layer (ML), and white matter (WM) region of the coronal section of the cerebellum/pons area of FVB mice (A and C), an age-matched FVB/Prnp^0/0 mouse (B), and an RML-infected FVB mouse at 132 days after inoculation (D). Sections were stained with MoPrP(Q218K)-Fc and HuM-D18 as indicated. (Bar = 250 μm.)

Discussion

Although many PrP ligands have been found, none have been demonstrated to participate in either the normal function of PrP^C or the conversion of PrP^C into PrP^Sc. The use of PrP as a ligand to identify binding partners has been problematic: first, recombinant PrP expressed in Escherichia coli lacks both the Asn-linked sugar chains and the glycolipid anchor found on PrP^C. Second, the glycolipid anchor of PrP^C makes it difficult to use this protein as a probe because it tends to form micelles both in the presence and absence of detergent. Third, in situ crosslinking studies may bind PrP^C to proteins that happen to inhabit the cholesterol-rich microdomains or rafts but do not specifically interact with PrP^C (18). In attempting to overcome these difficulties and continue our search for meaningful PrP ligands, we fused PrP to the Fc fragment. The advantages of the PrP-Fc protein as a probe are many, not the least of which is the solubility of this protein in aqueous buffers.

The discovery of a second member of the Prn gene family (1), Prnd that encodes Dpl, provided a second new approach to the search for PrP ligands. Studies on the rescue of mice from Dpl-induced CNS degeneration by expression of PrP^C demonstrated that Dpl and PrP must interact but most likely do so through another protein (9). Moreover, the low sequence homology between PrP and Dpl but the conserved tertiary structure (2) suggests that Dpl might be a useful tool in the search for meaningful PrP ligands.

Glycosylated PrP-Fc and Dpl-Fc Fusion Proteins.

By expressing the PrP-Fc and Dpl-Fc fusion proteins in N2a cells, we were able to produce probes that carry Asn-linked sugar chains. Both PrP and Dpl possess two consensus sites for Asn-linked glycosylation. In the glycosylated forms of PrP and Dpl, the Asn-linked oligosaccharides account for ≈30% of the molecular mass (5, 6). Both these glycosylated fusion proteins were secreted into the culture media from which they were recovered. The conformation of glycosylated PrP-Fc was monitored by using a panel of recombinant Fabs, which reacted in a manner that was indistinguishable from PrP^C on the plasma membrane of N2a cells (Fig. 1; ref. 38).

Studies on the structure of the N terminus of recombinant PrP from Syrian hamster and mouse have revealed the region to be unstructured (10, 43). In addition, PrP from two other species showed unstructured N termini (44, 45). In contrast, the C terminus composed of two α-helices is highly structured. In an earlier study, we fused alkaline phosphatase to the C terminus of chimeric MHM2PrP; the fusion protein was used to screen a λgt11 mouse brain cDNA expression library (34).

Although wtMoPrP-Fc expression in N2a cells produced a protein with a disulfide bridge formation and Asn-linked oligosaccharides as well as reactivity with a panel of HuM-Fabs similar to that found with PrP^C in situ, it is unknown how closely MoPrP-Fc resembles the structure of MoPrP^C. In contrast to PrP^C, MoPrP-Fc is a dimer covalently linked by disulfide bonds. Interestingly, there are many examples of receptor dimerization being required for signaling and signal transduction (for a review, see ref. 46). The ability of the MoPrP-Fc fusion protein used in this study to simulate oligomeric structures that might be required for PrP^C to function has been documented in studies of other GPI-anchored proteins using analogous strategies (47–49).

Dpl-Induced Cerebellar Degeneration.

Our results show that both wtPrP-Fc and wtMoDpl-Fc bind exclusively to the granule cells of the cerebellum in FVB and FVB/Prnp^0/0 mice (Figs. 3–5). This finding is of interest because several lines of Prnp^0/0 mice express Dpl in the cerebellum, which is accompanied by ataxia and Purkinje cell death (1, 50). Dpl expression in the CNS in Tg mice has been demonstrated to be responsible for cerebellar dysfunction with loss of both Purkinje and granule cells (9). Purkinje cells lie at the border between the granule cell and molecular layers and possess a wealth of synaptic connections between these two cell types; thus, compromised granule cell function might well be responsible for Purkinje cell death.

In Tg(Dpl)Prnp^0/0 mice and some lines of Prnp^0/0 mice that exhibit cerebellar degeneration, expression of PrP^C rescued the mice (9, 22). These findings provide clear evidence for an in vivo interaction between PrP and Dpl. The results raised two possibilities: (i) PrP binds to Dpl and prevents its neurotoxic effect, or (ii) PrP prevents Dpl from binding to a third, unknown protein, which, when bound to Dpl, produces neurodegeneration. Because neither the binding of wtPrP-Fc nor the binding of wtDpl-Fc differed by using cryostat sections from either Prnp^+/+ or Prnp^0/0 mice (Figs. 3 and 5), we conclude that PrP is not the receptor for the binding of these Fc fusion proteins to granule cells. Moreover, in Prnp^+/+ mice, anti-PrP antibodies did not bind to granule cells (Fig. 3). Whether a protein that mediates Dpl-induced degeneration can be retrieved from expression libraries using PrP-Fc and Dpl-Fc probes remains to be established.

Dominant-Negative Inhibition of PrP^Sc Formation.

In cell culture and Tg mice, the formation of PrP^Sc has been shown to be inhibited by mutant PrP molecules that act as dominant-negatives (25, 26). To explain these findings, the most likely scenario is the binding of dominant-negative PrP to an auxiliary factor, protein X, that participates in PrP^Sc formation. When dominant-negative PrP is bound to protein X, the factor is sequestered, making it unavailable to participate in the conversion of wt PrP^C into PrP^Sc.

Binding of Dominant-Negative PrP(218K).

Because dominant-negative PrP is thought to have a higher affinity for protein X than does wt PrP, we constructed the dominant-negative MoPrP(Q218K)-Fc fusion protein and used it to probe cryostat sections of mouse brain (25, 26). In contrast to wtMoPrP-Fc, MoPrP(Q218K)-Fc bound to both the granule cell and molecular layers of the cerebellum (Fig. 6 A and B). These findings argue that it is unlikely that the ligand for wtMoPrP-Fc on granule cells is the same protein to which MoPrP(Q218K)-Fc binds. The ligand in the molecular layer to which MoPrP(Q218K)-Fc binds is not PrP^C because the binding pattern was indistinguishable in both Prnp^+/+ and Prnp^0/0 mice (Fig. 6 A and B).

Although granule cells are devoid of PrP^C, the molecular layer has an abundance of PrP^C, and the white matter seems to have even more (Fig. 6C). PrP^Sc is deposited primarily in the white matter of the cerebellum, to which MoPrP(Q218K)-Fc does not bind (Fig. 6D). Because MoPrP(Q218K)-Fc does not bind to the white matter of the cerebellum where both PrP^C and PrP^Sc are found in the brains of FVB mice infected with RML prions, the most parsimonious conclusion is that MoPrP(Q218K)-Fc does not bind to protein X (Fig. 6). An alternative possibility is that PrP^Sc is formed in the molecular layer where both PrP^C and a ligand for MoPrP(Q218K)-Fc are found; once formed, PrP^Sc is readily transported to the white matter where it is deposited. In support of this hypothesis are earlier in situ hybridization studies indicating that PrP mRNA expression is confined to neurons, in which PrP^C is synthesized (51), and more recent investigations showing that PrP^Sc formed in brain grafts implanted in Prnp^0/0 mice travels along white matter tracts throughout the brain (52). In this scenario, the search for ligands that bind to MoPrP(Q218K)-Fc might yield protein X and thus, the screening of cDNA expression libraries with MoPrP(Q218K)-Fc seems prudent.

Conclusion.

In the studies described here, we report on the synthesis of PrP-Fc and Dpl-Fc fusion proteins. Although both PrP-Fc and Dpl-Fc bind to many regions of the brain, the binding of these fusion proteins to granule cells is perhaps the most informative. This restricted binding raises the possibility that these cells may express a protein that mediates neurodegeneration in mice expressing Dpl in the cerebellum. The rescue of Dpl-induced cerebellar degeneration by PrP expression suggests that through a single protein, Dpl may initiate neurodegeneration, whereas PrP inhibits the process.

Because wtMoPrP-Fc bound to granule cells of the cerebellum, which do not express PrP^C, it seems unlikely that wtMoPrP-Fc binding reflects a ligand that is involved in the conversion of PrP^C into PrP^Sc. In contrast, the dominant-negative MoPrP(Q218K)-Fc not only binds to granule cells but also to the neurons of the molecular layer where PrP^C is expressed. These findings raise the possibility that the cells of the molecular layer express an auxiliary protein involved in prion formation, i.e., protein X. The search for protein X has been frustrating because many proteins are known to bind to PrP^C, but none have been shown to participate in PrP^Sc formation as described above. Although the binding of dominant-negative MoPrP(Q218K)-Fc to cells in the molecular layer where PrP^C is abundant is consistent with the possibility that MoPrP(Q218K)-Fc is binding to protein X, the absence of PrP^Sc deposition in the molecular layer would require that PrP^Sc, once formed there, be readily transported to the cerebellar white matter where PrP^Sc is found. Use of these Fc fusion proteins to screen cDNA expression libraries may lead to the identification of protein X as well as another protein that mediates Dpl-induced neurodegeneration.

Abbreviations

• wt, wild type

• Mo, mouse

• Fc, antibody fragment with the constant region

• HuM, chimeric human–mouse

• Fabs, antibody fragments

• RT, room temperature

• Tg, transgenic