A Pathogenic PrP Mutation and Doppel Interfere with Polarized Sorting of the Prion Protein*

A hallmark of prion diseases in humans and animals is the conversion of the cellular prion protein PrP^C into an abnormally folded isoform, designated PrP^Sc, which is the major component of infectious prions (reviewed in Refs. 1–4). However, transgenic animal models revealed that misfolding or mistargeting of PrP^C can induce neurodegeneration in the absence of infectious PrP^Sc, indicating that the neurotoxic agent might be different from the infectious agent. For instance, mutations in the hydrophobic domain (HD)¹ of PrP, like A117V and AV3 (alanines at positions 112, 114, and 117 replaced by valines) lead to an aberrant folding of PrP during import into the endoplasmic reticulum (ER) and induce neurodegeneration in transgenic animals (5). Furthermore, neurodegeneration without the formation of PrP^Sc was induced by cytosolic expression of PrP^C (6). Of note, mistargeting of PrP to the cytosol was described for PrP145Stop and PrP160Stop, two mutations linked to inherited prion diseases in humans (7).

Spontaneous cerebellar neurodegeneration in certain strains of PRNP^0/0 mice led to the discovery of doppel (Dpl), a protein structurally related to PrP^C (8). Under physiological conditions, Dpl is not expressed in the brain; however, ectopic neuronal expression of Dpl induces Purkinje cell degeneration (9, 10). Dpl shows structural homology with the C-terminal globular domain of PrP^C but lacks the N-terminal octarepeats and the HD (11). Interestingly, expression of PrPΔF, a mutant devoid of the octarepeats and the HD (Δ32–134), induces cerebellar degeneration similarly to Dpl. Co-expression experiments in transgenic animals revealed that full-length PrP^C can antagonize both Dpl- and PrPΔF-induced neurodegeneration (10, 12).

Polarized sorting is an essential mechanism to establish and maintain the physiological functions of neurons and epithelial and endothelial cells. Selective trafficking is dependent on sorting signals within the cargo protein itself, on the lipid composition of membranes, and on adapter proteins of the transport machinery (reviewed in Refs. 13–16). Several proteins, which play a fundamental role in neurodegenerative diseases, like the β-amyloid precursor protein (APP) (17–20), amyloid β-peptide (Aβ) (17, 18), β-secretase (BACE) (21), and tau (22), undergo polarized sorting. Moreover, pathological conditions such as the familial AD-associated Swedish mutation of APP affect polarized sorting of BACE-generated soluble APP (23). While the cellular targeting of AD-associated proteins has been studied in great detail, little is known about sorting of PrP^C and its homologue Dpl.

In this study we investigated the trafficking of PrP^C and Dpl in polarized cells and found that the two proteins, which are both anchored to the plasma membrane via a glycosylphosphatidylinositol (GPI) anchor, are sorted differentially. We identified the HD of PrP^C as a dominant signal for basolateral sorting. Moreover, we demonstrate that pathogenic mutations within the HD and the expression of Dpl interfere with the physiological sorting of PrP^C.

EXPERIMENTAL PROCEDURES

cDNA Constructs—The coding region of mouse PRNP modified to express PrP-L108M/V111M was inserted into the pcDNA3.1/Zeo vector (Invitrogen), allowing detection by the 3F4 monoclonal antibody (24). Generation of the PrP mutants PrPmtGPI, PrP-M204S, and PrPΔHD was described previously (25). To generate a Dpl-expressing construct, the coding region of mouse PRND was amplified by PCR and subcloned into the pcDNA3.1/Zeo expression vector. For co-expression of PrP-3F4 and Dpl, the coding region of PrP-3F4 was amplified by PCR and subcloned into the pcDNA6/V5-HisA/Blasticidin expression vector (Invitrogen). PrP-N/Dpl and PrP-HD/Dpl were generated by transferring aa 1–133 or aa 113–133 of mouse PrP into the corresponding site in Dpl by PCR techniques.

Cell Culture and Transfection—Madin-Darby canine kidney (MDCK) cells (strain II) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, antibiotics, and glutamine. Stable transfections of MDCK cells were performed with FuGENE as described by the manufacturer. Cell lines expressing either PrP or Dpl from pcDNA3.1/Zeo were selected with 200 μg/ml zeocin; cell lines co-expressing PrP and Dpl were selected with 200 μg/ml zeocin and 2.5 μg/ml blasticidin. Cell lines with moderate expression levels and normal morphology were chosen for the study. To obtain polarized monolayers for surface biotinylation, cells were plated at confluence on 24-mm polycarbonate transwell filters. Media were changed every day, and cells were used for the corresponding experiment after 5 days. N2a cells were cultivated and transfected as described earlier (26).

Antibodies—Monoclonal anti-PrP antibody 3F4 (27) was purchased from Signet Pathology. The polyclonal anti-Dpl rabbit antiserum 2234 raised against recombinant Doppel was described earlier (10). Soluble APP (APPs) was immunoprecipitated with polyclonal antibody 5313 raised against the ectodomain of APP (28) and immunoblotted with monoclonal antibody 22C11 (Chemicon International).

Immunoprecipitation, Western Blotting, and Quantification Analysis—Secreted PrP mutants present in the cell culture medium were analyzed by immunoprecipitation (29) and subsequent Western blotting (30). Quantification was performed using AIDA 3.26 image analysis software (Raytest). The total amounts of protein located at/in the apical and basolateral cell surface/medium was set as 100%, and quantifications were based on at least four independent experiments.

Trypsin, Endo H, and PNGase F Digestion—Trypsin and Endo H digestions were described previoulsy (25). For PNGase F digestion, protein lysates were adjusted to 0.1% SDS, 0.9% Triton X-100, digested overnight at 37 °C and precipitated by trichloroacetic acid.

Surface Biotinylation—MDCK cell monolayers were washed with cold PBS and incubated with sulfo-NHS-LC-biotin (1 mg/ml), either apically or basolaterally for 1 h at 4 °C. 1% bovine serum albumin in PBS was added to the opposite chamber. The reaction was stopped by washing with PBS followed by quenching with 10 mm glycine in PBS. Cells were lysed in cold lysis buffer (0.5% Triton X-100 and 0.5% sodium deoxycholate in PBS) on ice for 30 min and scraped off the filter. Lysates were clarified by centrifugation for 20 min at 14,000 × g, and proteins were precipitated with streptavidin-Sepharose. Biotinylated proteins were detected by Western blotting. For the co-expression of PrP and Dpl, lysates were immunoprecipitated using the 3F4 or 2234 antibody, respectively. The immunopellet was analyzed by Western blotting using streptavidin-horseradish peroxidase conjugate. Results obtained by biotinylation were confirmed by indirect immunofluorescence experiments.

RESULTS

PrP^C and Dpl Are Differentially Sorted in Polarized Cells— MDCK cells have been used successfully to study polarized sorting of proteins linked to Alzheimer's disease. We now used this model system to study polarized trafficking of the mammalian PrP^C and its homologue Dpl. Individual clones of stably transfected MDCK cells expressing different levels of PrP^C or Dpl were selected (Fig. 1, A and B). Trafficking of PrP^C and Dpl was analyzed by cell surface biotinylation, which confirmed that PrP^C was selectively targeted to the basolateral membrane (Fig. 1C), corroborating earlier results (31). Basolateral sorting of PrP^C is unexpected, since GPI-anchored proteins are usually sorted to the apical surface (16, 32). Interestingly, sorting of Dpl was distinct from PrP^C sorting. Similarly to other GPI-anchored proteins, Dpl was sorted to the apical membrane of MDCK cells (Fig. 1D).

Co-expression of Dpl Targets PrP^C to the Apical Membrane— Neuronal expression of Dpl in PrP^C-deficient mouse strains has been shown to be responsible for cerebellar degeneration. Interestingly, co-expression of PrP^C can counteract Dpl-induced neurotoxicity (10, 33). An attractive model proposes that PrP^C and Dpl compete for a common ligand, and ligand binding to Dpl in the absence of PrP^C might initiate a neurotoxic cascade (12, 34). However, a neutralizing activity of PrP^C on Dpl-mediated neurodegeneration would be difficult to reconcile, since we found PrP^C and Dpl on opposite sites of MDCK cells.

To address this problem, we created cell lines stably expressing both PrP^C and Dpl (Fig. 2A) and analyzed if co-expression has an effect on the trafficking of PrP^C or Dpl. Co-expression did not interfere with the maturation of the individual proteins. Both Dpl and PrP^C were complex glycosylated and present at the cell surface (Fig. 2B). Polarized sorting of Dpl was not affected by the co-expression of PrP^C, since Dpl was still found predominantly at the apical membrane (Fig. 2C, α-Dpl). In contrast, Dpl expression had a profound effect on the sorting of PrP^C. Instead of being targeted to the basolateral site, PrP^C was found mainly at the apical membrane in the presence of Dpl (Fig. 2C, α-PrP). Notably, the effect of Dpl was selective for PrP^C, since in the presence of Dpl, APPs was still secreted from the basolateral membrane (Fig. 2C, α-APPs).

The Internal Hydrophobic Domain of PrP^C Acts as a Dominant Signal for Basolateral Sorting—To analyze polarized sorting of PrP^C mechanistically, we first addressed the role of the membrane anchor and of the complex N-linked glycans, as both post-translational modifications have been described to harbor possible sorting signals (15). PrP^C contains two partially sialylated complex N-linked glycans (35). Two different mutants, PrPmtGPI (ω site for GPI anchor attachment mutated) and PrP-M204S, are devoid of the GPI-anchor and are secreted into the cell culture medium as high mannose glycoforms (25). Similarly to wild type PrP^C, these PrP mutants were sorted basolaterally and secreted into the basolateral chamber (Fig. 2D).

In comparison to PrP^C, Dpl lacks the N-terminal unstructured region including the HD (Fig. 3A). Consequently, we asked whether this domain could contain information required for the basolateral sorting of PrP^C. Two PrP-Dpl chimeras were analyzed to address this question. PrP-N/Dpl contains the complete N-terminal domain up to amino acid 133 of PrP^C, while for the generation of PrP-HD/Dpl only the HD of PrP^C (aa 113–133) was transferred to Dpl (Fig. 3A). First, we demonstrated that PrP-HD/Dpl was present as a complex glycosylated protein at the cell surface, suggesting that folding and maturation was not impaired (Fig. 3B). PrP-N/Dpl showed the same biochemical behavior (data not shown). However, polarized sorting of the PrP-Dpl chimera was dramatically affected. Dpl containing the N-terminal domain of PrP^C was directed to the basolateral membrane (Fig. 3C, PrP-N/Dpl). Moreover, the HD of PrP^C comprising 20 amino acids was sufficient to mediate basolateral sorting of Dpl (Fig. 3C, PrP-HD/Dpl).

To confirm the role of the HD in the trafficking of PrP^C, we analyzed a PrP mutant lacking this internal stretch of hydrophobic amino acids. Deletion of the HD has no major effect on the maturation of PrP^C (25); however, PrPΔHD was found mainly at the apical membrane, supporting the pivotal role of the HD in basolateral sorting of PrP^C (Fig. 3C, PrPΔHD). This observation is consistent with the mistargeting of basolaterally targeted proteins after deletion of the basolateral sorting signal (see for example Ref. 17).

A Pathogenic Mutation Interferes with Physiological Sorting of PrP^C—After having identified the HD as a dominant sorting signal for basolateral trafficking of PrP^C, we wondered whether pathogenic mutations located within this domain would affect polarized sorting. Two mutations, A117V and AV3 (alanine residues at positions 112, 114, and 117 replaced by valines) (Fig. 3A), are linked to prion diseases in humans and transgenic mice, respectively (5). In previous studies the HD domain was identified as a putative transmembran domain, and it was shown that the AV3 mutation enhanced the formation of ^Ctm-PrP, a transmembrane topology of PrP with the C terminus facing the ER lumen (5). However, we could not detect a transmembrane topology of PrP-AV3 at the cell surface (data not shown). Strikingly, cell surface biotinylation and indirect immunofluorescence (data not shown) analysis revealed that in contrast to PrP^C, PrP-AV3 was predominantly found at the apical membrane (Fig. 3C, AV3). Thus, a pathogenic mutation within the HD affects polarized sorting of PrP.

DISCUSSION

Apical sorting of GPI-anchored proteins is thought to be mediated by the GPI anchor itself, its association with lipid microdomains, or N- and/or O-linked glycosylation (15, 32). Remarkably, PrP^C is to our knowledge the only GPI-anchored protein, which is sorted to the basolateral membrane in polarized MDCK cells. Our analysis showed that PrP^C sorting is independent of a membrane anchor and the glycosylation status; secreted high mannose glycoforms of PrP lacking the GPI anchor (PrPmtGPI, PrP-M204S) were efficiently sorted to the basolateral side as well.

A clue to the sorting determinants of PrP^C was obtained by analyzing the PrP^C homologue Dpl. This protein shows a structural organization similar to that of PrP^C, but it lacks the N-terminal region including the hydrophobic domain (11). Domain swapping experiments demonstrated that the HD of PrP^C confers basolateral sorting. Dpl containing either the whole N-terminal domain of PrP^C or the HD only was sorted basolaterally, indicating that this domain acts as a dominant sorting signal. Vice versa, PrP^C lacking the HD was found mainly at the apical surface of MDCK cells.

The atypical and unique sorting behavior of PrP^C might be of pathophysiological relevance. Previous studies indicated that expression of PrPΔF, a mutant devoid of the unstructured N-terminal region including the HD (Δ32–134), leads to cerebellar degeneration in transgenic mice. Importantly, the neurotoxic potential of PrP variants correlated with the disruption of the HD, which we identified as an important dominant sorting domain for polarized trafficking. PrP mutants lacking the N-terminal domain but containing the HD did not induce neurodegeneration (12). Moreover, we show in this study that a pathogenic mutation within the HD (PrP-AV3), which induces neurodegeneration in transgenic mice, alters the sorting of PrP^C as well. Inactivation of the sorting signal might be due to an alteration of either the primary sequence or the secondary structure.

Another interesting aspect of our study is the observation that the expression of Dpl has an impact on the sorting of PrP^C. Dpl induces cerebellar degeneration in transgenic mice, but its neurotoxic potential can be antagonized by wildtype PrP^C (10, 12). This effect is difficult to understand in the light of the differential sorting of PrP^C and Dpl. However, our co-expression studies may provide an explanation for the observations in transgenic mice. Co-expression of Dpl prevents basolateral sorting of PrP^C and directs PrP^C to the apical membrane, possibly by masking the dominant sorting signal within the HD. Thus, in cells expressing Dpl and PrP^C, both proteins are found at the same cellular locale, which could be a prerequisite for a functional interaction. Unfortunately, we could not demonstrate a direct interaction between PrP^C and Dpl. A general problem in PrP research is that no detergent conditions could be identified which allow efficient and reproducible co-immunoprecipitation experiments for either PrP^C or Dpl. It therefore will be important to employ other techniques, such as cross-linking in live cells or surface plasmon resonance with purified components, to further analyze a possible interaction of PrP^C and Dpl.

Our study provides experimental evidence that the HD is required for the physiological trafficking of PrP^C and reveals missorting of neurotoxic PrP mutants with a deleted or mutated HD. In transgenic animals expression of these mutants induces neurodegeneration, but the mechanism of neuronal cell death is still unknown. Based on our data, it is tempting to speculate that PrP might acquire a neurotoxic potential by missorting. However, it is equally plausible that a toxic gain of function is a consequence of an aberrant or missing protein interaction involving the hydrophobic domain.