Structural and mechanistic aspects influencing the ADAM10-mediated shedding of the prion protein

The following constructs were used for transient transfection of cells. Detailed descriptions of the constructs can be found in the corresponding references: PrP-WT, PrP^C glycomutants PrP-G1, PrP-G2, PrP-G3 and anchor-mutant PrP^GPI-Thy1 [61], PrP-TM (PrP-CD4 [62, 63]). All PrP constructs contained the 3F4 tag [64]. The N-terminally truncated PrP-C1 construct was cloned from the plasmid pcDNA3.1(+)/Zeo containing the murine Prnp gene. The sequence coding for the N-terminal part of PrP^C (aa23–110) was deleted by use of the restriction enzymes XbaI and HindIII and the resulting construct (Δaa23–110; i.e. PrP-C1) was verified by DNA sequencing.

The sPrP^G228 antibody for the specific detection of shed PrP was generated by use of an anti-peptide approach and the classical 87-day polyclonal protocol (Eurogentec, Belgium). Briefly, based on the sequence information of murine PrP^C and previous determination of the cleavage site for ADAM10 [47], a recombinant peptide NH₂-C-QAYYDG-COOH (in which G-COOH represents G228 as the new C-terminus of shed PrP exposed after cleavage (Fig. 1a)) was produced and N-terminally coupled to Megathura crenulata keyhole limpet hemocyanin (KLH) as carrier protein. This peptide was used as immunogen and injected into rabbit at days 0, 14, 28 and 56 of the programme. Bleedings were done at days 0, 38 and 66 to investigate the success of the immunization process by standardized ELISA tests. Animals were sacrificed and final bleeds were obtained at day 87. Standardized quality measures and affinity purification were performed at Eurogentec. Importantly, a second peptide (NH₂-C-KESQAYYDGRRS-COOH) mimicking the C-terminus of fl-PrP (without the GPI anchor) was produced, coupled to a resin and served as a “negative control” to eliminate all antibodies from the polyclonal serum that would otherwise bind to fl-PrP.

Use of animal material in this study was in strict compliance with the Guide for the Care and Use of Laboratory Animals and ethics guidelines of the responsible local authorities. Frozen forebrain samples from wild-type C57BL/6, prion protein deficient (Prnp^0/0 [65]), prion protein overexpressing (tga20 [66]) mice as well as from mice with conditional knockout of ADAM10 in forebrain neurons (A10 cKO and wild-type littermate controls [67]), with transgenic overexpression of dominant negative ADAM10 (A10 d.n. and wild-type controls [68]), with depletion of sortilin-1 (Sort1 KO and wild-type controls [69]) or with a knock-in of 3F4-tagged PrP^C (PrP^3F4KI and controls; both had a 192S4 background [70]), and from a rat and a rabbit were used to prepare 10% (w/v) homogenates in RIPA buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1% NP-40, 0.5% Na-Deoxycholate, 0.1% SDS) freshly supplemented with Complete EDTA-free protease (PI) and PhosStop phosphatase inhibitor cocktails (Roche) on ice. Samples were homogenized with 30 strokes using a dounce homogenizer and incubated on ice for 20 min, shortly vortexed and incubated for another 20 min before centrifugation at 12,000 g at 4 °C for 12 min. Total protein content was assessed by Bradford assay (BioRad). Supernatants were either further processed for SDS-PAGE or stored at − 80 °C.

Murine neuroblastoma cells (N2a) and mouse embryonic fibroblasts (MEF; [71]) were maintained at 37 °C under an atmosphere of 5% CO₂ in Dulbecco’s modified Eagle’s medium (DMEM; Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Thermo Scientific Fisher). N2a PrP-KO cells were generated using the TALEN approach and characterized in detail before [72]. N2a PrP-KO cells were transfected using Lipofectamine 2000 (Thermo Fisher Scientific) following the manufacturer’s instructions. For stable overexpression of PrP^3F4 in N2a PrP-KO (used for the glycopattern analysis shown in Fig. 1h) cells were kept for 3 weeks in selection media (Zeocin 400 μg/ml; Thermo Fisher Scientific) and single resistant clones were selected for amplification.

Treatments of cells were performed by adding the following compounds (and concentrations) to the cell culture media: Resveratrol (20 μM), Tamibarotene/Am80 (1 μM), GI254023X (3 μM), Tunicamycin (2.5 μg/ml), Swainsonine (5 μg/ml), Leupeptin (200 μg/ml). All compounds were purchased from Merck. These treatments were carried out in 6-well plates with 1 ml OptiMEM for 18 h overnight. In the case of Tunicamycin and Swainsonine cells were pretreated for 8 h. Treatment with GM6001 (25 μM) or Batimastat (10 μM) was for 10 h.

Two days post-transfection cells (grown in 6-well plate format) were incubated with 0.5 U/ml Phospholipase C (PI-PLC; Sigma-Aldrich) in 1 ml OptiMEM for 2 h at 37 °C, 5% CO₂ in order to cleave GPI-anchor structures and release GPI-anchored proteins from the cellular surface. Supernatants were subsequently harvested and further processed while cells were lysed as described below.

For removal as well as for investigations on processing and maturation of N-linked glycans attached to PrP^C, cell lysates and/or supernatants were digested with either PNGase F or Endo H (New England Biolabs) according to the manufacturer’s protocols.

N2a cells were washed with PBS and lysed with RIPA buffer, incubated on ice for 15 min before centrifugation at 12,000 g for 12 min at 4 °C. The protein content of the resulting supernatant was determined by Bradford assay. Prior to SDS-PAGE, cell lysates or brain homogenates (see above) were mixed with 4× loading buffer (including β-mercaptoethanol) and denatured for 6 min at 96 °C.

For the analysis of cell culture supernatants, experiments were carried out with serum-free media (OptiMEM). Supernatants were precipitated with trichloroacetic acid (TCA). For this, supernatants were collected and immediately incubated with already dissolved protease inhibitor cocktail, cleared from dead cells and debris by mild centrifugations at 500 g and 5.000 g for 5 min each. 1/100 volume of 2% sodium deoxycholate (NaDOC) was then added and each sample was shortly vortexed. After 30 min incubation on ice, samples were mixed with 1/10 volume of 100% TCA and again incubated for 30 min on ice. After centrifugation at 15,000 g for 15 min at 4 °C, the supernatant was aspirated, and the air-dried pellet was dissolved in 1× loading buffer and boiled for 6 min at 96 °C.

For labelling and purification of proteins at the cell surface, a surface biotinylation assay was performed as described earlier [50] prior to cell lysis.

For SDS-PAGE, denatured samples were loaded on either precast Nu-PAGE 4–12% Bis-Tris protein gels (Thermo Fisher Scientific) or self-made 10% or 12% SDS-gels. After electrophoretic separation, proteins were transferred to nitrocellulose membranes (BioRad) by wet-blotting and membranes were subsequently blocked for 1 h with 5% skimmed dry milk dissolved in TBS-T (containing 0.1% Tween-20) and incubated with primary antibody diluted in 5% skimmed dry milk in TBS-T overnight at 4 °C on a shaking platform. For detection of full length PrP^C (fl-PrP), mouse monoclonal antibodies POM1 (1 μg/ml), POM2 (0.6 μg/ml) [73] or, in the case of the sortilin-1 knockout mouse brains (Fig. 5j), SAF61 (0.2 μg/ml; Bertin Pharma) were used. Proteolytically shed PrP^C was detected with our new rabbit polyclonal sPrP^G228 antibody (0.2 μg/ml) characterized in detail herein. Moreover, we used anti-ADAM10 (0.4 μg/ml; abcam), anti-mouse β-amyloid antibody for detection of sAPPα (1 μg/ml; BioLegend), anti-actin antibody clone C4 (MAB1501, 1:1000; Merck) and anti-Flotillin-1 clone 18 (0.25 μg/ml; BD Biosciences). Membranes were subsequently washed with TBS-T and incubated for 1 h with respective HRP-conjugated secondary antibodies and subsequently washed 6× with TBS-T. After incubation with Pierce ECL Pico or Super Signal West Femto substrate (Thermo Fisher Scientific), chemiluminescence was detected with a ChemiDoc imaging station (BioRad) and densitometrically quantified using Image Studio Lite software version 5.2 (LI-COR).

N2a cells were grown on glass coverslips. After washing with PBS, living cells were incubated for 20 min on ice (to avoid endocytosis) with the primary antibody dissolved in 2% BSA/PBS. Surface PrP^C was detected with POM1 antibody (10 μg/ml). After several washes with PBS, cells were incubated with suitable secondary antibodies for 20 min on ice, subsequently fixed in 4% paraformaldehyde for 20 min at room temperature and mounted on glass slides with DAPI Fluoromount G (Southern Biotech). Analysis was performed using a TCS SP5 confocal microscope (Leica).

Sampling, formalin fixation, paraffin embedding, hematoxilin and eosin (H&E) staining as well as immunostaining with anti-prion protein antibody SAF84 (Caiman Chemical) of murine brain samples has been described earlier [50]. Immunostaining of shed PrP was likewise performed in one run using a Benchmark XT machine (Ventana) to allow for best comparability. In brief, deparaffinated brain sections were boiled for 1 h in citrate buffer (CC1 Cell Conditioning Solution, Ventana) for antigen retrieval and then incubated for 1 h with the sPrP^G228 primary antibody (7 μg/ml; in antibody diluent solution (Zytomed) with 5% goat serum). Detection with anti-rabbit secondary antibody (Nichirei Biosciences) and Ultra View Universal DAB Detection kit (Ventana), as well as blue counterstaining were performed by the machine following standardized protocols.

N2a cells were cultured in OptiMEM for 18 h. For the harvest of extracellular vesicles (here further referred to as exosomes), cell culture supernatants were first complemented with PI and centrifuged for 10 min at 1000 g and further at 7500 g for 15 min at 4 °C, followed by filtration through a 22 μm membrane to clear it from dead cells and debris. Exosomes were then pelleted by ultracentrifugation at 100,000 g for 70 min at 4 °C in an Optima L-100 XP using a SW40Ti rotor (Beckman Coulter, Inc.) and subsequently resuspended in PBS containing PI. For quantification and characterization, a NanoSight LM14 (Malvern Instruments) equipped with a 638 nm laser and a Marlin F033B IRF camera (Allied Vision Technologies) was used. For each sample, 10 videos of 10 s length were recorded with a camera intensity setting of 16 and analysed to assess average size and concentration of exosomes using the batch processing function of the software. For normalized western blot analysis, 5 × 10¹⁰ exosomes per sample were used.

For experiments using mouse brain samples, n refers to the number of biological samples (i.e. mice) per experimental group. For cell culture-based data, n stands for the number of independent experiments. Statistical comparison of western blot quantifications was performed using Student’s t-test and significance was considered with p-values as follows: *p < 0.05, **p < 0.005, ***p < 0.001.

ADAM10 is the relevant sheddase of PrP^C releasing a soluble form (shed PrP, sPrP) from the plasma membrane [47, 48]. Since membrane-bound full length (fl) PrP^C and its shed form only differ in three amino acids (murine sequence) and the GPI-anchor, it is hard to reliably discriminate between both in most approaches. Based on available sequence information and the previous identification of the cleavage site for ADAM10 [47], we therefore generated an antibody specific for sPrP (sPrP^G228) being directed against the newly generated carboxy-terminus at Glycine 228 (G228) exposed after cleavage (Fig. 1a).

To characterize this antibody in detail, we analyzed forebrain homogenates of age-matched Prnp^0/0 mice (as negative control), recently described mice with neuron-specific (CamKIIα-driven) depletion of ADAM10 (A10 cKO; to control for specificity) as well as wild-type littermate controls, and PrP^C-overexpressing tga20 mice (as positive controls) by western blot. As expected, detection with our new antibody consistently revealed no signal in Prnp^0/0 samples, basal levels in wild-type mice and strongly increased signal intensity in tga20 mice (Fig. 1b). Though we expected significantly reduced levels of shed PrP in A10 cKO mice, to our surprise we could not detect any signal in these samples. Besides supporting the specificity of the antibody, this indicates that no other cell types or proteases contribute to this cleavage in brain. Re-probing the same blot with an antibody against fl-PrP revealed an increase in total PrP^C levels in A10 cKO mouse brains (Fig. 1b), a finding that has been made earlier and can be attributed to the lack of shedding [50]. Moreover, while this blotting strategy demonstrated the overlapping banding pattern (as well as the masking of sPrP signals by excess amounts of fl-PrP using common PrP antibodies), it also revealed a small shift in the molecular weight of sPrP corresponding to the lack of three amino acids and the GPI-anchor (Fig. 1b and Additional file 1). We also investigated the species specificity of the new antibody using mouse, rat and rabbit brain samples. As expected for the different C-terminal PrP^C sequences, the sPrP^G228 antibody only detected sPrP in brain homogenates of mice and rat (Additional file 2).

Though not being in the focus of this study, we were also interested in the applicability of the antibody in morphological analyses and performed immunohistochemical staining of paraffin-embedded mouse brain sections. As exemplified for the hippocampal area in Additional file 3, no signal was obtained in a Prnp^0/0 mouse, whereas a diffuse staining was found in wild-type and, with higher intensity, in tga20 brain as could be expected for a soluble fragment distributed in the brain parenchyma.

Although, structurally, all three glycoforms of PrP^C can be shed (as demonstrated in tga20 brain (Fig. 1c)), a strong predominance of the diglycosylated form of sPrP was obvious in all of our biochemical analyses. To investigate this in more detail, we analyzed the glycopattern of sPrP compared to cell-associated fl-PrP in brain homogenates of wild-type mice (Fig. 1d) and found a clear shift and a drastic increase in the proportion of diglycosylated sPrP (mean: 97 ± 1%; compared to 60 ± 4% for fl-PrP; n = 3; ±SD) with only little mono- (3 ± 1%; fl-PrP: 33 ± 3%) and almost no unglycosylated sPrP (0.07 ± 0.03%; fl-PrP: 6.8 ± 0.8%).

As a model for downregulation of ADAM10-mediated cleavage events, we investigated PrP shedding in forebrains of mice overexpressing a dominant negative form of ADAM10 (A10 d.n.) in addition to the endogenous protease [68] (Fig. 1e). When referring the sPrP to the respective fl-PrP signal, we found a ~ 50% reduction (mean sPrP/fl-PrP ratio: 0.51 ± 0.05; n = 4; ±SEM) in A10 d.n. mice compared to matched controls (set to 1.00 ± 0.13).

Since for main parts of this study we used N2a cells transfected with murine versions of PrP^C containing the 3F4 tag in the middle of the protein sequence, we first had to show that this modification does not influence the shedding event. This is even more important as it is known, that the course of prion diseases is altered by this tag [70, 74]. We therefore decided to investigate shedding in the best possible model, i.e. in PrP^3F4 knock-in (KI) mice expressing levels of 3F4-tagged PrP identical to PrP^C levels in wild-type mice (Additional file 1) [70]. No significant differences in sPrP levels were observed between controls (set to 1.00 ± 0.23; n = 3; ±SD) and PrP^3F4 KI mice (0.85 ± 0.12) thus ruling out an impact of this modification on PrP shedding as could be expected from its intramolecular distance to the membrane-proximate shedding site.

We next employed the new antibody in cell culture-based experiments. Given that manipulation of PrP^C shedding may become a therapeutic option in different pathologies, we investigated how pharmacological stimulation and inhibition of ADAM10 affect sPrP production in N2a cells (Fig. 1f,g). Among others, the stilbenoid resveratrol and the synthetic retinoid tamibarotene (Am80) have been successfully used to increase ADAM10-mediated cleavage events [75, 76]. We also found elevated levels of sPrP in supernatants of N2a cells treated with these substances compared to solvent-treated controls (Fig. 1f). In contrast, shedding was abolished upon treatment with the ADAM10-selective inhibitor GI254023X (GI) [77]. Of note, upon re-probing the “supernatant blot” with another PrP antibody (POM2), a strong signal was obtained under GI-treatment indicative of a release of fl-PrP by alternative routes when shedding is blocked (as discussed later). Fittingly, cell-associated PrP^C levels (in lysates) remained rather unaffected by the different treatments further supporting existence of compensatory mechanisms regulating PrP^C homeostasis in N2a cells (discussed later). We also assessed the metalloprotease inhibitors GM6001 and batimastat (Fig. 1g). These drugs likewise abolished the shedding of PrP^C at the cell surface yet did not significantly alter production of N1 and C1 fragments resulting from the α-cleavage of PrP^C. There is controversy regarding the involvement of ADAMs in the α-cleavage (reviewed in [45, 78]). However, due to the lack of membrane permeability of the inhibitors used here, this finding cannot count as an argument against ADAMs as potential “α-PrPases”, given that α-cleavage is thought to occur mainly within the secretory pathway [79]. Again, levels of cell-associated fl-PrP did not appear to be altered by these treatments.

Lastly, consistent with our findings in mouse brain (Fig. 1d), we also observed a changed glycopattern of sPrP compared to fl-PrP in N2a cells (Fig. 1h; diglycosylated: 84.2 ± 4.4% (sPrP) vs. 67.4 ± 0.9% (fl-PrP); monoglycosylated: 15.6 ± 4.2% (sPrP) vs. 29.6 ± 1.0% (fl-PrP); unglycosylated: 0.22 ± 0.18% (sPrP) vs. 2.9 ± 0.3% (fl-PrP); n = 3; ±SD) though relatively more monoglycosylated sPrP is found in N2a cells (15.6 ± 4.2%) than in brain (3 ± 1%; Fig. 1d). To clarify whether our findings indicate a real preference for the shedding of diglycosylated PrP or rather reflect the availability of different glycoforms at the plasma membrane, we performed cell surface biotinylation and glycopattern analysis in N2a cells (Additional file 4). Though relatively more diglycosylated PrP is indeed available at the plasma membrane (compared to total PrP levels in cell lysates; Additional file 4B), our data still argues in favor of a preference for diglycosylated PrP given the strong predominance of this form among shed PrP (Fig. 1d, h). In summary, we have generated a sensitive and highly specific antibody to discriminate between shed and fl-PrP in mouse brains and cell culture supernatants. ADAM10 on neurons seems to be the dominant (if not exclusive) PrP sheddase. ADAM10-mediated shedding of PrP^C can be modulated by various means, and our shedding-specific antibody is a useful read-out tool for such experiments. Though all glycoforms can in principle be shed, diglycosylated PrP by far represents the major substrate for ADAM10.

Glycosylation of PrP^C impacts on its biology and role in prion disease [62, 80‐83]. Given the predominance of diglycosylated sPrP under normal conditions (Fig. 1), we wondered how shedding would be affected if only specific glycoforms of PrP^C are present in cells. To this end, we transfected PrP^C-depleted N2a cells (PrP-KO; generated using the TALEN strategy and described earlier [72]) with either wild-type PrP or PrP glycomutants carrying a mutation in either the first (N180Q mutant; PrP-G1), the second (N196Q mutant; PrP-G2) or both (N180Q/N196Q mutant; PrP-G3) N-glycosylation sites and, thus, giving rise to mono- (G1 and G2) or unglycosylated (G3) PrP^C. Using these glycomutants, we could previously demonstrate a relevant impact of the N-glycans on the sorting of PrP^C in polarized cells [61]. Despite differences in transfection efficiencies, western blot analysis revealed the typical banding pattern for the glycomutants as observed for similar mutants in transgenic mice (Fig. 2a) [83]. Like fl-PrP, the membrane-attached C1 fragment resulting from α-cleavage of PrP^C also presents with a three-banding pattern with the diglycosylated C1 overlapping with unglycosylated fl-PrP. For a better characterization, deglycosylation was performed to only obtain unglycosylated fl-PrP and C1 fragment and showed that all mutants undergo α-cleavage (Fig. 2b). No effect was observed upon treatment of lysates with Endo H indicating a correct processing of the glycans and trafficking out of the ER and to the cellular surface for all transfected mutants (Fig. 2c). Immunofluorescence analysis of surface PrP^C further supported a correct biosynthesis and showed that all mutants are readily expressed at the plasma membrane (Fig. 2f) confirming previous results in polarized MDCK cells [61]. Of note, analysis of sPrP in media supernatants revealed that not only fl-PrP but also the truncated C1 fragment can be shed for all glycoforms (Fig. 2d; deglycosylation of supernatants for confirmation of bands is shown in Fig. 2g). Quantification (Fig. 2e) of sPrP referred to total PrP (to correct for different transfection efficiencies) showed a moderate decrease in sPrP for the monoglycosylated mutants, yet a significant reduction to ~ 50% was observed for the unglycosylated PrP-G3 (mean: 0.49 ± 0.11 compared to PrP-WT set to 1.00 ± 0.04; n = 3; ±SEM). Interestingly, as indicated by asterisks in Fig. 2a and d, for the unglycosylated PrP-G3 mutant we consistently detected a significant difference in the ratio of PrP and C1 between membrane-associated (fl-PrP: 61%; C1: 39%; ±4.5% SEM; n = 3) and shed forms (sPrP: 21%; shed C1: 79%; ±6.2%). This increase in the proportion of C1 (from 39% in lysates to 79% among the shed PrP forms) in the absence of N-glycans may relate to the longer half-life of C1 at the plasma membrane and a fast re-internalisation of unglycosylated fl-PrP [84‐86] which seems to be disfavored as a substrate (Fig. 2i). In contrast, for normally glycosylated PrP-WT, there appeared to be a preference for the shedding of (diglycosylated) fl-PrP over the (diglycosylated) C1 fragment when comparing fl-PrP and C1 ratios in PNGase F digested lysates (Fig. 2b) and media supernatants (Fig. 2g). This observation prompted us to investigate a potential influence of the N-terminal half of PrP^C on the membrane-proximate shedding. We therefore transfected PrP-KO N2a cells with PrP-WT or with an N-terminally truncated construct (PrP-C1) corresponding to the physiological C1 fragment (Additional file 5). Despite indicating that a preference for the shedding of diglycosylated forms also exists for the C1 fragment, shedding of the latter was significantly reduced compared to fl-PrP in cells transfected with PrP-WT. Although further analysis are clearly required and differences might partially result from transient overexpression and altered surface expression of the constructs, these findings point to a role of the PrP N-terminal domain in the C-terminal shedding event. Thus, the glycosylation state as well as proteolytic truncation seem to affect PrP^C shedding.

To support our findings obtained with PrP glycomutants (Fig. 2), we pharmacologically manipulated glycosylation in wild-type N2a cells. While the antibiotic tunicamycin (TM) inhibits N-glycosylation, the alkaloid swainsonine (SWA) is a known inhibitor of the further maturation of N-linked glycan structures resulting in non-mature high-mannose glycans. Treatment of cells with TM completely prevented N-glycosylation of PrP^C, whereas treatment with SWA resulted in a shift in the molecular weight of diglycosylated PrP^C indicating immature glycosylation (Fig. 3a). Digestion of lysates with Endo H revealed that SWA (partially) impaired complex glycosylation as shown by an altered glycopattern compared to controls (Fig. 3b). Further confirmation for the TM- and SWA-treatments and the enzymatic deglycosylation reactions is presented in a side-by-side comparison (Additional file 6). However, it should be noted that we did not reach a complete deglycosylation in the case of Endo H-digestion of lysates of SWA-treated cells. Presence of residual complex glycosylated PrP^C suggests that the incubation with SWA (8 h) was too short or that Endo H digestion was incomplete. Independent of the type of treatment, PrP^C was expressed at the cellular surface (Fig. 3c). Intriguingly, PrP^C shedding was almost completely abolished in cells treated with TM (mean: 0.07 ± 0.03; n = 3; SEM) and significantly reduced upon treatment with SWA (0.41 ± 0.10) compared to untreated controls (set to 1.00 ± 0.12) (Fig. 3d, e). Even though only little non-mature diglycosylated PrP^C was present in SWA-treated cells (Fig. 3b), this fraction seems to be the only relevant substrate for shedding. Again, these data support our previous findings (Figs. 1 and 2) showing that diglycosylated PrP is the preferential substrate for ADAM10 at the cell surface and that altered glycosylation influences PrP shedding efficiency.

An additional modification that influences membrane topology, biological functions and pathophysiological roles of the prion protein is the type of membrane attachment. With PrP^C being one of only few GPI-anchored substrates of ADAM10 [47, 87], we wondered how altered attachment and topology at the membrane would affect its shedding. To this end, we used mutants of PrP^C either comprising a transmembrane domain instead of the GPI-anchor (PrP-TM) or carrying the GPI-anchor signal sequence (and –as a likely consequence– the GPI-anchor [61] (Puig et al. submitted)) of Thy-1, a protein described to reside in the dense cores of detergent-resistant membranes (herein referred to as lipid rafts) [88]. Whereas interaction between ADAM10 and PrP^C is thought to occur at the interface between lipid rafts and non-raft regions (Fig. 4a), previous studies have shown that PrP-TM is relocated outside of rafts and turned signaling-incompetent [38, 63, 89] while PrP^GPIThy-1 remains in rafts yet therein shows a different localization than PrP-WT [61] (Puig et al. submitted).

Immunofluorescent stainings of non-permeabilized cells showed surface expression (Fig. 4b) while western blot analysis revealed comparable expression of all constructs transfected into PrP-KO N2a cells (Fig. 4c). Deglycosylation of samples showed that all PrP mutants are subject to α-cleavage and confirmed an increase in molecular weight for PrP-TM due to its transmembrane domain (Fig. 4d). No alterations in the banding pattern were observed upon treatment with Endo H indicating correct glycosylation and –again– surface transport of all mutants (Fig. 4e). Lack of signal in the media for PrP-TM upon incubation of cells with PI-PLC proved the absence of a GPI-anchor and its attachment via a transmembrane domain (Fig. 4f).

Of note, despite a conserved shedding site in all constructs, shedding was completely abolished for PrP-TM (mean: 0 ± 0.02; n = 3; SEM) and significantly reduced for PrP^GPIThy-1 (0.33 ± 0.06) compared to PrP-WT (set to 1.00 ± 0.05) (Fig. 4g,h). In conclusion, altered membrane attachment and, hence, changed membrane localization severely impact on PrP^C shedding.

As already shown in Fig. 1f, we consistently observed an increased cellular release of PrP via alternative routes whenever the proteolytic shedding was impaired (e.g. by inhibition of ADAM10 with GI; Fig. 5a). Since PrP^C is released via microvesicles (e.g. exosomes [72, 90, 91]), we wondered whether this mechanism could compensate for abolished shedding. Inhibition of ADAM10 with GI resulted in a significant rise of exosome release (mean: 1.48 ± 0.09; compared to controls set to 1.00 ± 0.09; n = 4; SEM; Fig. 5b) without affecting the typical size of these vesicles (mean: 128 ± 12 nm; compared to controls 115 ± 2 nm; n = 4; SEM; Fig. 5c; details for characterization of exosomes are shown in Additional file 7). When amounts of exosomes were normalized, we found more than a twofold increase in their average PrP^C load upon GI treatment (2.12 ± 0.08; compared to controls set to 1.00 ± 0.08; n = 3; SEM; Fig. 5d,e). PrP^C levels in the corresponding cells were only moderately, yet significantly increased (1.40 ± 0.07; compared to controls set to 1.00 ± 0.07; n = 3; SEM; Fig. 5d,f). Thus, as a consequence of impaired shedding, more PrP^C is packed into exosomes and more exosomes are released by N2a cells. Interestingly, such an alternative release of PrP^C was not observed in murine fibroblasts (MEF) lacking ADAM10 (Additional file 8) [71] which instead accumulated PrP^C to 3-fold the amount of wild-type MEFs (ADAM10 KO: 3.07 ± 0.12; WT set to 1.00 ± 0.12; n = 8; SEM). This may indicate that not all cell types possess the compensatory machinery ensuring PrP^C release and membrane homeostasis.

Finally, we investigated how degradation, as a third aspect involved in cellular PrP^C homeostasis, influences PrP^C shedding. We blocked lysosomal degradation by treatment of N2a cells with leupeptin. As expected, cytosolic proteins (e.g. β-actin, β-tubulin, β-catenin) known to be degraded by the proteasome rather than in lysosomes were not accumulated (Fig. 5g). Instead, increased secretion of sAPPα, the proteolytic fragment of APP, indicated successful lysosomal inhibition (Fig. 5g). Of note, cell-associated PrP^C levels remained rather stable despite this treatment (Leupt.: 0.97 ± 0.05; untr. Cells set to 1.00 ± 0.05; n = 3; SEM; Fig. 5g,h) yet shedding of PrP^C was significantly increased (Leupt.: 2.31 ± 0.24; untr. Cells set to 1.00 ± 0.24; n = 3; SEM; Fig. 5g,i). No obvious differences in alternatively released PrP (Fig. 5g) suggests that, in this condition, shedding is the main contributor avoiding increased cellular PrP^C levels.

Impaired lysosomal degradation and increased PrP^C levels have recently been shown in mice lacking the sorting receptor sortilin-1 [69]. As a consequence of hindered transport to lysosomes, these mice had shown increased PrP^Sc conversion and shortened survival when infected with prions. Given the increase in shedding upon lysosomal inhibition with leupeptin in cells shown before, we asked whether shedding of PrP^C is likewise affected by the impaired degradation due to lack of sortilin-dependent transport in vivo. In fact, we found an approximately 2-fold increase for sPrP in sortilin1-deficient mice compared to controls (Sort1 KO: 1.85 ± 0.32; Sort1 WT set to 1.00 ± 0.09; n = 4; SD; Fig. 5j). Rather than from up-regulation of sheddase activity, this increase seems to result from elevated cellular PrP^C levels caused by impaired degradation (Additional file 9). Nevertheless, this demonstrates the capability of ADAM10 to release increased amounts of substrate.

In summary, these data suggest ADAM10-mediated shedding as a relevant factor regulating PrP^C levels. Shedding, exosomal release and degradation of PrP^C may be interconnected mechanisms that act in a compensatory manner ensuring PrP^C homeostasis and allowing –if at all– only subtle changes thereof.