Reagents and antibodies

All buffers and media for cell culture, Hank's balanced salt solution (HBSS) and 10× PBS (pH 7.4) were purchased from Invitrogen Corporation (Carlsbad, CA, USA). Plasmid purification and DNA gel extraction kits were from Omega Bio-Tek (Norcross, GA, USA). Triton X-100 (TX100), deoxycholic acid (DOC), Triton X-114 (TX114), sodium hydroxide (NaOH), N-lauroylsarcosin (sarcosyl), chymotrypsin, and proteinase K (PK) were purchased from Sigma-Aldrich Co., LLC (St. Louis, MO, USA). Guanidine hydrochloride (GdnHCl) was from Promega Corporation (Madison, WI, USA). Site-directed mutagenesis kit was purchased from Agilent Technologies, Inc. (Santa Clara, CA, USA). Tween 20, Bafilomycin A1 (Baf) and acrylamide (40%, 37.5∶1) solution were from EMD Chemicals Inc. (Gibbstown, NJ, USA). Pentosan polysulfate was from Bene-Arzneimittel GmbH (Munich, Germany). 3-methyladenine (3MA) was from Thermo Fisher Scientific Inc. (Waltham, MA, USA). MG-132 was purchased from Cayman Chemical Co. (Ann Arbor, MI, USA). The anti-PrP monoclonal antibody (mAb) 4H11 was described before [33]. The anti-PrP monoclonal antibody (mAb) 3F4 which recognizes residues108–111 was purchased from Covance (Princeton, NJ, USA). Anti-PrP C16-S rabbit mAb, which was raised against the C-terminal part of human H3, was purchased from Novus Biologicals (Littleton, CO, USA). Anti-LAMP1 rat mAb was from BD Biosciences (San Jose, CA, USA). Anti-calnexin (CNX) rabbit polyclonal antibody was from Assay Design's Inc. (Ann Arbor, MI, USA). All the secondary antibodies, DyLight488-conjugated anti-rat IgG antibody, DyLlight594-conjugated anti-mouse IgG antibody and HRP-conjugated or DyLight488-conjugated anti-rabbit IgG (with minimal cross-reactivity with serum proteins from other species) were purchased from Jackson Immunoresearch (West Grove, PA, U.S.A.). The plasmids pEGFP-LC3, -Rab7 and -Rab9 have previously been described [34], [35].

Cell culture

The mouse neuroblastoma cell line N2a was purchased from American Type Culture Collection (Manassas, VA, U.S.A.) and persistently infected with the mouse prions strain 22L (22L-ScN2a) [36]. A cell population with a high and stable level of PrP^Sc was selected by single-cell cloning and used throughout for the experiments described here. The non-infected counterpart (N2a) was prepared by disinfecting 22L-ScN2a cells by pentosan polysulfate (2 µg/ml) treatment over seven passages.

Site-directed mutagenesis

All primers used in these experiments were ordered from Integrated DNA Technologies, Inc. (Coralville, Iowa, U.S.A.) and are listed in Table S1. Mutations were made by site-directed mutagenesis according to the manufacturer's instruction, using a 3F4-epitope (methionine at position 108 and 111) tagged mouse Prnp gene cloned into pcDNA3.1(+) as template; therefore all ΔPrPs were 3F4-epitope tagged. Created mutations were verified by sequencing from both strands at Clemson University Genomics Institute (Clemson, SC, U.S.A.).

Transient transfection for evaluation of expression levels or DNI effect of ΔPrPs and preparation of lysates for immunoblot analysis

For evaluating the expression levels of ΔPrPs, N2a cells on 24-well culture plates were transiently transfected with 0.32 µg/well of each ΔPrP construct. For evaluation of DNI effect, 22L-ScN2a cells on 24-well culture plates were transiently co-transfected with 0.25 µg/well each of (3F4)MoPrP and ΔPrP (total DNA amount 0.5 µg/well; 1∶1 molar ratio). For transfection of cells on 6-well culture plates, 4-fold larger amounts of plasmid and transfection reagent per well than for 24-well plates were used. 22L-ScN2a or N2a cells were plated on 24-well or 6-well culture plates and transfected with plasmid DNA with Lipofectamine LTX Plus kit (Invitrogen) according to the manufacturer's instruction. 2-mercaptoethanol was added to the culture medium (final concentration 50 µM) during transfection. Cells were kept in medium containing plasmid and transfection reagent for 24 hours and then harvested or the old medium was replaced with fresh full medium without plasmid until harvesting. For evaluating effects of various chemical compounds on ΔPrP levels, the old medium was replaced with fresh full medium with either DMSO (0.15%), bafilomycin A1 (120 nM), 3MA (10 mM) or MG132 (10 µM) at this point and the transfected cells were incubated further for 7 hours before harvest. Cells were harvested with phosphate-buffered (pH 7.4) 0.5% TX100, 0.5% DOC (TX100/DOC) lysis buffer, 40 µl/well for 24-well and 300 µl/well for 6-well plate. After removal of nuclear debris by centrifugation (microcentrifuge Eppendorf AG, Hamburg, Germany) at 21,130 g for 1.5 minutes, the supernatant was transferred to another tube as TX100/DOC postnuclear lysate. For samples tested for protease-resistant PrP cores, lysates were digested with chymotrypsin (at indicated concentrations) or proteinase K (PK) (25 µg/ml) at 37°C for 30 minutes. Digestion was stopped by addition of Pefabloc (Roche Applied Science, Indianapolis, IN, U.S.A.) at 2 mM, and 1/4-volume of 5× sample buffer (5×SB; 12% SDS, 250 mM Tris-HCl, pH 7.1, 40% glycerol and bromophenol blue). Finally, lysates were boiled for 10 minutes at 95°C.

SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting

Samples were resolved on 10–12.5% SDS-PAGE gels and electrotransferred to PVDF membranes (Milipore, Billerica, MA, U.S.A.) by semi-dry blotting method. PVDF membranes were blocked with 5% Blotto (non-fat dry milk from Bio-Rad., Hercules, CA, U.S.A.) in tris-buffered saline supplemented with 0.1% Tween 20 (TBST) for 30 minutes and then incubated with anti-PrP antibodies 3F4 (mAb; 1∶10,000 dilution), C16-S (mAb; 1∶5,000) or 4H11 (mAb; 1∶1,000) in 5% Blotto in TBST. Then membranes were washed in TBST, incubated with HRP-conjugated anti-mouse IgG for mAbs 3F4 and 4H11 or anti-rabbit IgG for mAb C16-S, 1∶10,000 in 5% Blotto in TBST for 1 hour, washed again in TBST four times, then the labeled proteins were visualized with Pierce ECL Plus Western Blotting Substrate (Thermo Scientific, Rockford, IL, U.S.A.). For detection, membranes were exposed to X-ray films (Thermo Scientific, Rockford, IL, U.S.A.) and developed. X-ray films were scanned and quantified by densitometry with image processing software Image J (http://rsb.info.nih.gov/ij/). Acquired data were analyzed with statistics software “R” (www.r-project.org) for two-tail paired t-tests with p<0.05 as significance level. When re-probing with mAb 4H11 was required, the PVDF membrane was incubated in 100% MeOH for 20 minutes to remove bound antibodies. Then, the membrane was washed in TBST to remove residual MeOH, followed by incubation with mAb 4H11 in 5% Blotto in TBST. The following steps including secondary antibody incubation and washes in TBST were as described above.

Detergent solubility analysis

N-lauryl sarcosyl was added to TX100/DOC lysates from N2a or 22L-ScN2a cells expressing ΔPrPs to a final concentration of 4%. Lysates were ultracentrifuged using an OptimaTL ultracentrifuge (Beckman Instruments GmbH, Munich, Germany) at 100,000 g, 4°C, for 1 hour. The supernatant was carefully removed, transferred to fresh tubes and subjected to methanol/chloroform (MeOH/CHCl3) precipitation (described below). The pellet fraction was washed once with 100 µl of TX100/DOC lysis buffer and ultracentrifuged again at 100,000 g for 15 minutes. After removal of supernatant, the pellet was sonicated in 1× SB (2.4% SDS; 50 mM Tris-HCl, pH 7.1; 8% glycerol; bromophenol blue) and boiled for 10 minutes.

Methanol/chloroform (MeOH/CHCl₃) precipitation

First, 1/5-volume of methanol and 4/5-volume of chloroform were added to lysates, followed by rigorous vortexing for 30 seconds. The mixture was incubated on ice for 20 minutes and then centrifuged at 16,100× g, 4°C, for 30 minutes. The denatured proteins make a sheet between the upper phase with MeOH and the lower phase with CHCl₃. After removal of the upper phase, MeOH of 9-fold volume of the lower phase was added and mixed again. The mixture was centrifuged at 16,100× g, 4°C, for 30 minutes. Subsequent to centrifugation, the supernatant was thoroughly removed and the protein pellet was dried, dissolved in 1× SB and finally boiled.

Endoglycosidase H (EndoH) or peptide: N-glycosidase F (PNGaseF) digestion

ΔPrPs were digested with endoglycosidase H (EndoH; New England Biolab Inc., Ipswich, MA, U.S.A.) according to manufacturer's instructions. Briefly, first 100 µl of TX100/DOC lysates were subjected to MeOH/CHCl₃ precipitation. Protein pellets were re-dissolved in 25 µl of 1× denaturation buffer provided with the enzyme and boiled for 10 minutes. Then, 20 µl of deionized water, 5 µl of G5 reaction buffer and 2 µl of EndoH were added and incubated at 37°C for 1.5 hours. After digestion, 1/4-volume of 5× SB was added and boiled for 10 minutes. For PNGaseF digestion, following MeOH/CHCl₃ precipitation, the pelleted proteins were reconstituted in dilute sample buffer (5× SB diluted with deionized water, 1∶15) and boiled with shaking at 1,400 r.p.m. for 10 minutes. The denatured proteins were supplemented with 1/10-volume of G7 buffer and 10% NP40 attached to PNGaseF (New England Biolab) and then PNGaseF was added and incubated at 37°C for 2 hours. After incubation, 1/4-volume of 5× SB was added and the samples were boiled for 10 minutes.

Triton X-114 extraction of ΔPrPs

N2a cells expressing ΔPrPs on 6-well culture plates were incubated in PBS with 3 mM EDTA for 3 minutes and then mechanically detached by pipetting. The cell suspension was collected in a 1.5-ml tube and centrifuged at 1,000× g for 5 minutes at 4°C. After the centrifugation, the supernatant was thoroughly discarded and the pelleted intact cells were resuspended in 400 µl of phosphate-buffered (pH 7.4) 2% TX114 lysis buffer (2% TX114; 137 mM NaCl; 2.7 mM KCl; 8 mM Na₂HPO₄; 2 mM KH₂PO₄) and incubated on ice for 30 minutes, with vortexing from time to time. Subsequently, the cell suspension was centrifuged at 16,100× g, for 2 minutes at 4°C and the supernatant was transferred to another tube as TX114 lysate. Phase separation of TX114 lysates was done by incubating the lysates at 37°C for 10 minutes and centrifugation at 21,130× g for 10 minutes. The aqueous phase was transferred to another tube and subjected to MeOH/CHCl₃ precipitation. The remaining detergent phase was diluted with 0.1% TX114 wash buffer (0.1% TX114; 137 mM NaCl; 2.7 mM KCl; 8 mM Na₂HPO₄; 2 mM KH₂PO₄) and subjected to MeOH/CHCl₃ precipitation. The precipitated proteins were then processed as described above for ‘MeOH/CHCl₃ precipitation’. For phase-separation after denaturation with GdnHCl, 120 µl of TX114 lysates were mixed with 120 µl of 6 M GdnHCl and incubated at room temperature for 45 minutes. Then, 150 µl of 2% TX114 lysis buffer along with 1,000 µl of 0.1% TX114 wash buffer were added to dilute out GdnHCl, and lysates were subjected to phase separation. After separation, the aqueous phase was transferred to a fresh tube for MeOH/CHCl₃ precipitation, the detergent phase (∼50 µl) was diluted with 0.1% TX114 wash buffer up to 200 µl and then subjected to MeOH/CHCl₃ precipitation. The following steps for sample preparation were as above. For the in vitro PIPLC digestion of GdnHCl-treated lysates, after the first phase separation after denaturation with GdnHCl, the aqueous phase containing GdnHCl was discarded and the detergent phase (∼50 µl) was diluted with 550 µl of 0.1% TX114 wash buffer to further reduce GdnHCl concentration to a sufficiently low level for not inhibiting PIPLC activity. After lysates were cleared on ice, 4 µl of PIPLC (2 U/µl) was added to ‘PIPLC+’ samples and both the lysates with or without PIPLC were incubated at room temperature for 2.5 hours. Then, the lysates were subjected to a second round of phase separation and the aqueous and the detergent phases were processed as described above, except that 3 µl of 10 mg/ml bovine serum albumin (BSA) was added to each phase as a carrier before MeOH/CHCl₃ precipitation.

Immunofluorescence analysis/confocal microscopy

Cells were plated on glass coverslips placed at the bottom of 24-well culture plates and transient transfection was performed as described above. Next day, the transiently transfected cells on the coverslips were rinsed twice with HBSS and then fixed with 4% paraformaldehyde (USB Corporation, Cleveland, OH, U.S.A.) at room temperature for 30 minutes. After fixation, cells to be permeabilized were treated with 0.2% Triton X-100 in PBS for 10 minutes at room temperature and rinsed with PBS three times. When antigen retrieval by GdnHCl treatment was required, the permeabilized cells were incubated with 6 M GdnHCl for 45 minutes and then the cells were rinsed with PBS four times before incubation with the primary antibody. Permeabilized and non-permeabilized cells were incubated with mAb 3F4 (1∶2,000) in 3% BSA in PBS for 60 minutes on a rocking platform, followed by washing with PBS four times, then incubated with DyLight488-conjugated sheep anti-mouse IgG (1∶1,000; Jackson Immunoresearch, West Grove, PA, U.S.A.) in 3% BSA in PBS for 45 minutes and finally washed in PBS four times. After immunolabeling, the coverslips with cells on them were taken out from the 24-well plates and mounted on slide glass with a drop of Permafluor mountant (Thermo Scientific, Rockford, IL, U.S.A.). When the mountant was dried, samples were analyzed on a laser scanning confocal microscope, Zeiss710 (Carl Zeiss Inc., Thornwood, NY, U.S.A.), in the Robert A. Jenkins Microscopy Facility of the University of Wyoming. Samples were studied with an objective lens, EC Plan-Neofluar 100×/1.3 Oil Pol M27, and the wave length of the excitation lasers were 488 nm for DyLight488 and EGFP and 594 nm for DyLight594. The acquired image data were processed with Image J (NIH, USA). For observation of colocalization of ΔPrPs with LAMP1, to minimize cross-reactivity of anti-mouse secondary antibody to anti-LAMP1 rat mAb cells were first labeled in the following order: mAb 3F4, DyLight594-anti-mouse, anti-LAMP1 and then DyLight488-anti-rat. If needed, transfected cells were treated with bafilomycin A1 for ∼6 hours before fixation to inhibit the degradation of ΔPrPs by lysosomal proteases. Only a single slice at the level of nucleus where the punctate fluorescence was most abundant was used for analysis of co-localization.

Subcellular fractionation of transfected cells

N2a cells on a 6-well plate were transiently transfected as described above. Twenty-four hours later, the old medium was replaced with fresh medium, with or without bafilomycin A1, and cells were cultured for 6 more hours. Cells were rinsed once with PBS without calcium or magnesium, then incubated for a few minutes in 3 mM EDTA in PBS until cells could be easily detached from the plate by pipetting. Detached cells in PBS were collected in 1.5 ml tubes and centrifuged at 200× g at 4°C for 5 minutes. After removal of supernatant pelleted cells were resuspended in 200 µl of homogenization buffer [8.5% (w/w) sucrose; 4 mM Tris-HCl, pH 7.1; 2 mM EDTA; 30 µg/ml cycloheximide (Sigma Aldrich) and complete protease-inhibitor cocktail (Roche Applied Science, Indianapolis, IN, USA)]. Before homogenization, 10 µl of cell suspension was taken for whole-cell lysate measurement. The rest of the cells were homogenized by passing through a 25G ultra-thin-wall needle (Terumo Medical Corporation, Somerset, N.J.) until >90% of cells were disrupted. The homogenate was centrifuged at 2,000× g for 5 minutes and the supernatant, ∼175 µl, was transferred to another 1.5 ml tube as post-nuclear fraction and mixed well with 240 µl of a 62% (w/w) sucrose solution to a final concentration of ∼42%. The post-nuclear fraction was placed at the bottom of a centrifuge tube and three layers of sucrose solutions, 36%, 33% or 30%, and homogenization buffer were overlaid from the bottom to the top. The gradient was ultra-centrifuged at 40,000 rpm at 4°C for 1.5 hours with a Beckman SW50.1 swing rotor in a Beckman L8-80M ultracentrifuge. After ultra-centrifugation, fractionated organelles were visible as milky-white bands in interphases between the layers and 350 µl was carefully pipetted from each interphase to collect organelles as completely as possible. Collected interphases were first diluted with 300 µl of PBS with 0.01 µg/µl of BSA and then subjected to MtOH/CHCl₃ precipitation, as described above. Pelleted proteins were dissolved in 1× SB and boiled to make “interphase” samples. For “whole-cell lysate” samples, 10 µl of cell suspensions were mixed with 30 µl of TX100/DOC lysis buffer. After centrifugation to precipitate the cell debris, supernatants were collected in another tube as whole-cell lysates. After addition of 10 µl of 5× SB, they were boiled to prepare “whole-cell lysate” samples.

Engineering mutant PrPs with internal deletions and expression in cultured cells

We decided to engineer a series of mutant PrPs (ΔPrPs) with different deletion sizes, starting with a deletion of a single residue (i.e. glutamine 159) immediately after the H1 region, and extending C-terminally towards the H2 region (Fig. 1A). We reasoned that such deletion mutants would be more straight-forward for analysis of structure-function relationships than other types of mutations. We had chosen this region in particular because of its postulated importance for prion conversion. In addition, dominant negative inhibition of ΔPrPs with deletion of the B2 region has been previously reported [27] and we could expect that some of our mutant PrPs with deletions near B2 would show efficient DNI. If the DNI is then attenuated at a certain point when the deletion is C-terminally extended, this transition would be suggestive of the region involved in PrP-PrP interaction. For convenience of detection, all ΔPrPs contain a 3F4-epitope tag. First, we studied the expression of ΔPrPs upon transient transfection into murine neuroblastoma (N2a) cells. All constructs were expressed at comparable levels (Fig. 1B, lanes 2–10), although substantially lower than a 3F4-tagged wild-type PrP [(3F4)MoPrP] used as control (Fig. 1B, lane 1). Immunoblot appearance of ΔPrPs were also uniform with more demarcated and narrower bands than that of (3F4)MoPrP and with the diglycoform (Fig. 1B, closed arrowhead) predominant over the mono- and non-glycoforms (Fig. 1B, open arrowheads). Interestingly, Δ159–167 provided another band over the presumed diglycoform band (Fig. 1B, lane 7 and arrow). This was proven to be an extra N-glycan attached due to creation of a third glycosylation sequon (Asn-Tyr-Ser169) by this deletion. Digestion with endoglycosidase H (EndoH) resulted in the expected non-glycoform band (Fig. 2B, lane 3, arrowhead). Further proof for this was obtained when serine 169 was deleted [Δ159–167(169)], resulting in the previous glycosylation pattern (Fig. 2B, lane 2 and 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Expression of ΔPrPs and dominant-negative inhibition (DNI) of Δ159 and Δ159–175 in persistently prion-infected cells.

A. Schematic illustration of prion protein secondary structure elements comprising two β-strands (B1 and B2) and three α-helices (H1–3) and post-translational modifications (two N-linked glycans and glycosylphosphatidylinositol anchor) along with position and extent of deletion of ΔPrPs constructs. Deletions in the H1∼H2 portion have a common N-terminal end (residue 159) and gradually extend into the C-terminal direction. All constructs have a 3F4 epitope tag (methionines at residue 108 and 111). Underlined residues denote B1. B. Comparison of ΔPrP expression levels. Representative immunoblot is shown for detection of (3F4)MoPrP and ΔPrPs in transiently transfected N2a cells using mAb 3F4. ΔPrPs are similar in expression level glycosylation appearance, except for Δ159–167 which has an extra fragment larger than the diglycoform (lane 7, arrow). C. ΔPrPs are not converted into PK-resistant PrP in prion-infected cells. 22L-ScN2a cells were transiently transfected with (3F4)MoPrP, Δ159 or Δ159–175, and cell lysates digested with PK or not. Immunoblot was done using mAb 3F4. D. Pilot study for dominant-negative inhibition (DNI) of Δ159 and Δ159–175. DNI was assessed by co-transfecting 22L-ScN2a cells with (3F4)MoPrP and Δ159 or Δ159–175, respectively, and testing for PK-resistant (3F4)MoPrP. Left panel shows representative immunoblot (mAb 3F4) and right panel the statistical analysis of quantified PK-res levels of a triplicate experiment. Empty pcDNA3.1 plasmid was used as control for co-transfection in lane 1. The error bars indicate standard deviation. *, p<0.05.

https://doi.org/10.1371/journal.ppat.1003466.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. ΔPrPs are non-complex glycosylated and retained intracellularly.

A. N-linked glycans of ΔPrPs are not complex-type. Immunoblot comparing EndoH- and non-digested lysates of ΔPrP159 and Δ159–175 transfected cells. Left panel was probed with mAb 3F4 mAb, detecting transiently expressed ΔPrPs. On the right panel, the membrane was re-probed with mAb 4H11, detecting total PrP. Arrowheads depict deglycosylated ΔPrPs. B. The additional band of Δ159–167 represents a third N-glycan sensitive to EndoH. Upon EndoH digestion, all bands converge into a non-glycoform band (lane 3 and 4). In Δ159–167(169) the third glycosylation site is not present anymore. C. ΔPrPs are mainly distributed in intracellular compartments rather than on the cell surface. Confocal microscopy images of N2a cells transiently transfected with (3F4)MoPrP, Δ159 or Δ159–175, either permeabilized (right) or non-permeabilized (left) with 0.2% TX100 after fixation. Cells were immuno-labeled with mAb 3F4. (3F4)MoPrP-expressing cells show bright immunopositivity on the cell surface, whereas ΔPrPs are not observed on the cell surface. Also intracellular staining differs strongly between (3F4)MoPrP and ΔPrP expressing cells Scale bars, 25 µm.

https://doi.org/10.1371/journal.ppat.1003466.g002

Taken together, all mutant PrPs were expressed in neuronal cells at comparable levels and with similar immunoblot patterns.

Confirmation of the dominant negative inhibitory effect of mutant PrPs

For evaluating the DNI of ΔPrPs, we co-transfected ΔPrPs with (3F4)MoPrP in a 1∶1 ratio and evaluated DNI on the co-transfected (3F4)MoPrP, rather than on endogenous PrP^Sc. As shown previously [17], this method is very sensitive in detecting differences in newly formed PrP^Sc levels, requires shorter transfection times and thereby minimizes possible side effects caused by aberrant properties of ΔPrPs. We initially engineered the two constructs with the shortest and longest deletions, Δ159 and Δ159–175, and went on to confirm their DNI. We co-transfected them with (3F4)MoPrP into N2a cells persistently infected with 22L prions (22L-ScN2a) or transfected them alone (Fig. 1C). Cells were lysed, lysates were subjected to proteinase K (PK) digestion, separated by SDS-PAGE and evaluated for PrP^Sc levels in immunoblot analysis using the 3F4 monoclonal antibody (mAb). Under conditions as used for PK digestion (25 µg/ml for 30 min), ΔPrPs expressed in 22L-ScN2a cells were completely digested (Fig. 1C, lanes 5 and 6; Fig. S1) and 3F4-immunopositive PrP^Sc in co-transfection represents only those derived from (3F4)MoPrP (Fig. 1D). When performing co-transfections there were significant differences between the two constructs. Δ159 reduced PrP^Sc levels of (3F4)MoPrP to 20% of the empty-vector control (Fig. 1D, lane 1 and 2), whereas Δ159–175 reduced only to ∼70% (Fig. 1D, representative immunoblot and densitometric analysis).

Taken together, we show that Δ159 and Δ159–175 are not converted into PrP^Sc but are able to exert DNI on co-transfected wild-type PrP. Interestingly, Δ159 containing only a deletion of a single residue showed a much more pronounced DNI.

Characterization of biochemical and cellular properties of ΔPrPs

Prion proteins with internal deletions in the H1∼H2 portion had been reported before to be aberrant in that they have EndoH-sensitive high-mannose-type N-glycans, a mainly intracellular localization, detergent insolubility, and protease resistance, all independent of the presence of PrP^Sc [27], [37], [38]. As such aberrant properties might affect DNI efficiency and since there was an obvious difference in DNI efficiencies between Δ159 and Δ159–175, we next tested whether there is a detectable difference in biochemical and cellular properties, explaining the difference in DNI between them.

ΔPrPs spontaneously form aggregates.

Internally-deleted PrPs, e.g. MHM2PrP(Δ23–88, Δ141–176) (often referred to as PrP106), have been reported to form oligomers and β-sheet structures, becoming protease-resistant and detergent-insoluble [39]–[41]. Likewise, our ΔPrPs were rather insoluble even in 4% sarcosyl solution and largely were found in the pellet fraction, whereas most of endogenous wild-type PrP^c was completely solubilized (Fig. 3A, lanes 5–8, upper panel transfected PrP (mAb 3F4), lower panel total PrP (mAb 4H11)). Of note, there was no difference between ΔPrPs from 22L-infected and non-infected cells (Fig. 3A, lane 3 vs. 7, 4 vs. 8), demonstrating that their detergent insolubility is independent of and unaffected by PrP^Sc. It was also found that ΔPrPs were more resistant to chymotrypsin digestion than endogenous wild-type PrP. Substantial amounts of full-length ΔPrPs (Fig. 3B, upper panel, closed arrowhead) were still remaining even at 8 µg/ml treatment (Fig. 3B, lanes 5 and 10), at which full-length endogenous PrP (Fig. 3B, lower panel, open arrowhead) was completely digested and only truncated fragments were present at low amounts (curly bracket).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. ΔPrPs spontaneously form aggregates that are detergent-insoluble and relatively chymotrypsin-resistant.

A. ΔPrPs are insoluble in detergent solution. Immunoblots showing partition of Δ159 and Δ159–175 in supernatant (Sup) and pellet fraction after ultracentrifugation of lysates supplemented with 4% sarcosyl. Upper panel is mAb 3F4, lower panel is membrane re-probed with mAb 4H11 detecting also endogenous wild-type PrP besides ΔPrPs. The majority of ΔPrPs is found in the insoluble fraction in both infected and non–infected cells (upper panel, lanes 3, 4, 7 and 8). Wild-type PrP almost completely partitions into the soluble fraction (Sup) in non-infected N2a cells with present conditions. PrP in pellet fractions of non-infected N2a cells is from ΔPrPs as can be seen from the banding pattern (Lane 7 and 8, square bracket). PrP in lanes 3 and 4 (lower panel) is PrP27-30. B. ΔPrPs are moderately resistant to chymotrypsin digestion. Immunoblots showing chymotrypsin-resistant fragments of Δ159 and Δ159–175 from transiently transfected N2a cell lysates digested with indicated concentrations of chymotrypsin (Chy). Note that full-length diglycoform of ΔPrP still remains at 8 µg/ml Chy (lanes 5 and 10, upper panel, closed arrowhead), whereas that of endogenous wild-type PrP is completely digested (lanes 5 and 10, lower panel, open arrowhead) and only small amounts of truncated fragments remained (curly bracket). Other smaller fragments in lane 5 and 10 (square-brackets) represent ΔPrPs.

https://doi.org/10.1371/journal.ppat.1003466.g003

In summary, ΔPrPs are spontaneously forming aggregates which are mildly chymotrypsin-resistant. There was no detectable difference between Δ159 and Δ159–175 in these properties.

ΔPrPs are GPI-anchored.

So far glycosylphosphatidylinositol (GPI) anchoring of internally-deleted PrPs was controversial. It was reported that PrP106 has a GPI anchor [37], whereas other internally-deleted PrPs were reported to lack it [38]. As GPI anchoring greatly affects properties of PrP, especially intracellular trafficking and interaction with PrP^Sc [42], [43], we wanted to solve this issue for our mutant PrPs using Triton X-114 (TX114) extraction. First, we phase-separated TX114 lysates of cells expressing Δ159 and Δ159–175, respectively, and evaluated partition in detergent and aqueous phase. Although more than half of ΔPrPs was sequestered to the detergent phase, substantial amounts remained in the aqueous phase (Fig. 4A, left panel). In contrast, the vast majority of endogenous wild-type PrP was extracted to the detergent phase (Fig. 4A, right panel). As such a pattern of separation can be also seen with transmembrane proteins, e.g. lysosome-associated membrane protein-1 (LAMP1) used here as a control (data not shown), adequate GPI anchoring was still questionable at this point.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. ΔPrPs have a GPI-anchor attached.

A. ΔPrPs are substantially soluble in detergent without GdnHCl denaturation. Immunoblots probed with mAbs 3F4 (left) or 4H11 (right, membrane re-probed) showing ΔPrPs and endogenous PrP in detergent and aqueous phases of TX114 lysates prepared from N2a cells transfected with Δ159 or Δ159–175. Det denotes detergent phase, Aq aqueous phase. B. ΔPrPs are highly hydrophobic after denaturation with GdnHCl. Representative immunoblot of TX114-treated lysates after denaturation with 3 M GdnHCl and developed with mAb 3F4 showing Δ159 and Δ159–175 almost exclusively in detergent phase (lanes 1–4). PIPLC treatment (+, even lanes) was done after denaturation with GdnHCl or not (−, uneven lanes). PIPLC treatment did not significantly influence ΔPrP partitioning under these conditions. C. Hydrophobicity of ΔPrPs combined with sensitivity to PIPLC digestion. After further diluting out GdnHCl by phase-separation and removal of the aqueous phase containing GdnHCl, PIPLC was able to digest ΔPrPs. Left two panels (upper mAb 3F4, lower 4H11) are shorter expositions, right panels longer expositions to show also difference in PrP levels in the detergent phase. Under these conditions a substantial reduction of ΔPrPs in the detergent phase was achieved by PIPLC digestion (e.g. upper-left panel, lanes 1, 3 vs. 2, 4). Migration of ΔPrPs in the aqueous phase of PIPLC-digested samples was slightly slower than that of PIPLC-undigested samples (−) (upper-right panel, lane 9 vs. 10).

https://doi.org/10.1371/journal.ppat.1003466.g004

Since we considered that the aggregate nature of ΔPrP might conceal their hydrophobic parts, including the lipid moiety of the GPI-anchor, we tried GdnHCl treatment of ΔPrPs before phase separation to disintegrate the aggregates. As expected, the majority of ΔPrPs were now extracted to the detergent phase (Fig. 4B, lanes 1 vs. 5, 3 vs. 7). To further confirm GPI-anchoring, in vitro digestion with PIPLC was done before phase separation. Although we diluted out GdnHCl to ∼0.5 M, there was no difference between samples with and without PIPLC treatment (Fig. 4B, lane 1 vs. 2, 3 vs. 4), presumably because GdnHCl concentration was still too high for PIPLC digestion. We further diluted GdnHCl by removing the aqueous phase after the first phase separation and diluted the detergent phase with fresh buffer. Eventually, GdnHCl could be lowered to sufficient levels that PIPLC was able to digest ΔPrPs and endogenous PrP (Fig. 4C, upper panel mAb 3F4, lower panel mAb 4H11 for total PrP). There was a clear reduction of ΔPrP levels of lysates digested with PIPLC in the detergent phase along with an increase in the aqueous phase (Fig. 4C, lane 9 vs. 10, 11 vs. 12). The same changes were observed for endogenous PrP (Fig. 4C, lower panel). The increment of ΔPrPs and endogenous PrP in the aqueous phase of PIPLC-digested samples was much smaller than the decrement in the detergent phase. In addition, PIPLC-digested PrPs migrated slower than undigested ones (Fig. 4C, lane 9 vs. 10). These features are consistent with GPI-anchored proteins after PIPLC digestion. Loss of GPI-anchors results in reduction of hydrophobic surfaces, consequently poorer adsorption to PVDF membrane and fewer SDS molecules attached to proteins in SDS-PAGE [44].

Taken together, our findings prove that ΔPrPs are GPI-anchored.

ΔPrPs are degraded by acidic compartments, presumably involving autophagic processes

Next, we investigated by which degradation systems ΔPrPs are degraded. Since DNI involves a direct physical interaction between PrP^c and PrP^Sc, the subcellular trafficking of ΔPrPs should lead into compartments where PrP^Sc molecules reside. First, we studied the effects of inhibitors of lysosomal or proteasomal degradation on levels of ΔPrPs, specifically the V-ATPase inhibitor bafilomycin A1 (Baf), the autophagy inhibitor 3-methyladenine (3MA), and the proteasome inhibitor MG132. Cells were incubated only up to 7 hours to prevent interference with additional degradation systems [45]. There was no difference between Δ159 and Δ159–175, suggesting that these mutant PrPs share the same metabolic pathway. Baf and 3MA most efficiently inhibited degradation of the diglycoform of ΔPrP (Fig. 5A, square bracket). Therefore, at least part of ΔPrPs seemed to be transferred to a 3MA-sensitive, presumably class-III PI3K-dependent pathway and eventually degraded in acidic compartments. The banding pattern of ΔPrPs from MG132-treated cells was clearly different from those of Baf or 3MA-treated cells, with a main increase in the non-glycoform of ΔPrP (Fig. 5A, bracket), as expected for proteins subjected to endoplasmic reticulum-associated degradation (ERAD) [46].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. ΔPrPs are degraded by both lysosomal and proteasomal degradation systems.

A. Degradation of ΔPrPs by lysosomal and proteasomal systems. (Left) N2a cells transiently transfected with Δ159 or Δ159–175 were treated for 7 hours with inhibitors either for lysosomal degradation (bafilomycin A1, Baf, 120 nM), autophagy (3-methyladenine, 3MA, 10 mM), or the proteasome (MG132, 10 µM) and cell lysates analyzed in immunoblot for expression of Δ159 and Δ159–175 (mAb 3F4). Square bracket and bracket denote diglycoforms and nonglycoforms of ΔPrPs, respectively. An increment in diglycoforms is observed in Baf- and 3MA-treated cells, whereas in MG132-treated cells also the non-glycoforms are increased. (Right) A graph showing quantification of all PrP bands by densitometric analysis. Data from 3 independent (one in duplicate) experiments for Δ159 and 2 independent (one in duplicate) for Δ159–175 were statistically analyzed for mean and standard deviation (error bars). B. ΔPrPs can be localized in LAMP1-immunopositive vesicles. Immunofluorescence analysis shows distribution of ΔPrP (mAb 3F4) and LAMP1 in cells with or without 6M GdnHCl antigen-retrieval treatment. N2a cells transfected with ΔPrP159 were fixed, permeabilized and incubated without (upper panel) or with 6M GdnHCl (lower panel). With GndHCl treatment ΔPrP is diffusely distributed (arrowheads) and co-localizes poorly with LAMP1. After GdnHCl treatment (lower panel), bright 3F4-immunopositive puncta (arrows) are observed which partly co-localize with LAMP1-positive structures (merge). Scale bars, 10 µm.

https://doi.org/10.1371/journal.ppat.1003466.g005

This data suggests that ΔPrPs undergo a mixed degradation and are subjected both to lysosomal/autophagic pathways and the proteasome.

A lysosomal degradation of ΔPrPs was unexpected, as the EndoH-sensitive N-glycans of ΔPrPs suggested that they would not reach the medial Golgi from which a trafficking pathway to lysosomes has been suggested [47]. To confirm lysosomal degradation, we next studied whether ΔPrPs co-localize with LAMP1, calnexin, Rab7, Rab9 and microtuble-associated protein 1 light chain 3 (LC3) in immunofluorescence analysis/confocal microscopy. Non-transfected cells (for LAMP1 and calnexin) or cells transiently transfected with GFP-Rab7, -Rab9 and -LC3, respectively, were pre-treated with Baf for 6 hours to prevent lysosomal degradation. Without GdnHCl pretreatment of cells we did not observe a significant co-localization of ΔPrPs and LAMP1 (Fig. 5B, upper panels, arrowheads), GFP-Rab7 or GFP-Rab9 (data not shown). A very weak to moderate co-localization was found for ΔPrPs and LC3 or calnexin, respectively (Fig. S2 and S3). Since ΔPrPs forms aggregates, we hypothesized that the epitope for anti-PrP antibodies might not be accessible, just as is the case for PrP^Sc [48]. Therefore, we applied the antigen retrieval procedure for PrP^Sc by treating the cells with GdnHCl. After treatment with 6 M GdnHCl, punctuate-vesicular fluorescent structures were observed and some of them clearly co-localized with LAMP1 (Fig. 5B, lower panels, arrows). Unfortunately, these harsh conditions were not appropriate for detection of GFP-Rab7, -Rab9 and -LC3 (Fig. S4), presumably due to denaturation of EGFP.

To obtain further evidence for lysosomal degradation of ΔPrPs we utilized subcellular fractionation of Δ159 transfected N2a cell lysates on sucrose gradients (Fig. S5). Part of the studies involved pre-treatment of cells with bafilomycin A1 to study effects of this drug on distribution of Δ159 in the gradients (Fig. S5B). We found that a substantial proportion of Δ159, importantly only the glycosylated form, was present in low-density fractions, between 8.5 and 30% (w/w), where ER components are scarce, whereas late endosomes, lysosomes and autophagosomes are enriched there. This fraction was also most enriched with LAMP1 and LC3-II, and the amounts of the Δ159 diglycoform were strongly increased by bafilomycin A1 treatment (Fig. S5B). Calnexin was most abundant in the high-density fraction, specifically in the interphase between 36 and 42% (w/w). Interestingly, the mono- and non-glycoforms of Δ159 were also mainly present in the interphase where calnexin was most enriched. In conclusion, these findings strongly corroborate the microscopic findings that substantial amounts of glycosylated Δ159 are present in amphisomes and lysosomes and are consistent with a lysosomal degradation pathway as suggested by increase of the diglycoforms of ΔPrPs by bafilomycin A1 or 3MA treatment.

Taken together, we demonstrate that a portion of ΔPrPs is subjected to lysosomal degradation by a 3MA-sensitive process.

DNI of ΔPrPs is inversely correlated with size of deletion

As there was no detectable difference in biochemical or cell-biological properties between Δ159 and Δ159–175, we speculated that their difference in DNI is most likely attributable to the size and positioning of the internal deletions. Therefore, we next studied DNI efficiencies of all ΔPrPs by co-transfection with (3F4)MoPrP (Fig. 6A). DNI efficiencies of Δ159 and Δ159–175 were reproduced with PrP^Sc levels ∼20% and 70–80%, respectively, and ΔPrPs with deletions between these two constructs showed a gradual reduction of DNI efficiency as deletions were extended, i.e. an inverse relation between DNI efficiency and deletion size (Fig. 6A). The relatively low DNI efficiency of Δ159–163 might be due to the unique primary structure with two prolines and only one intervening residue, making the regional structure less flexible and disadvantageous for inter-PrP interaction (Fig. 1A). On the other hand, Δ159–167 showed a comparable DNI effect to adjacent ΔPrPs, despite the extra N-glycan (Fig. 1B). Next, we assessed whether small deletions in the C-terminal part of H1-H2 exert DNI. PrPΔ171–175 (Fig. 7A) was expressed at similar levels as Δ159–175 (Fig. 6B, left panel) and, interestingly, its DNI efficiency was also comparable to that of Δ159–175 (Fig. 6B, right panel), suggesting that the length of H1∼H2 is not the sole determinant of DNI and that the positioning of the mutation is also important for efficient DNI. Taken together, all PrPs with deletions in the H1–H2 region exerted DNI, although more pronounced when the C-terminal part was preserved.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Inverse correlation of DNI and size of deletion and importance of positioning of deletion.

A. DNI efficiencies of ΔPrPs are inversely correlated to deletion size. Representative immunoblot showing PK-resistant PrP^Sc moiety of (3F4)MoPrP co-expressed with indicated ΔPrPs in 22L-ScN2a cells. Quantitative analysis is shown below as scheme. Relative PrP^Sc levels were calculated as proportion of PK-resistant (3F4)MoPrP co-transfected with each ΔPrP to that of co-transfected with empty vector control in the same experiment. Data from 4 (for Δ159–165, Δ159–167 and Δ159–167(169) only 3) independent experiments were statistically analyzed for mean and standard deviation (error bars). B. DNI of ΔPrP with a deletion in the C-terminal part of H1∼H2. Δ171–175 (see Fig. 7A) was transiently expressed in N2a cells (left panel) and DNI efficiency evaluated in 22L-ScN2a cells (right panel) (mAb 3F4). Despite the much shorter deletion size, Δ171–175 exerted only inefficient DNI, similar to Δ159–175 (right panel).

https://doi.org/10.1371/journal.ppat.1003466.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. ΔPrPs lacking in addition the N-terminal region still exert efficient DNI.

A. Schematic illustration of Δ171–175, Δ31–160 and its variants. Δ31–160 variants lack the entire part between the polybasic motif (PB, italic and bold) and residue 161. SS, signal sequence. B. Δ31–160 variants are expressed at comparable levels and have EndoH-sensitive and –resistant N-glycans. Immunoblot comparing EndoH-digested and non-digested samples from N2a cells transiently transfected with Δ31–160 variants, probed with anti-PrP mAb C16-S raised against the C-terminal portion of H3. Non-Tf, samples prepared from N2a cells without transfection. Arrowhead denotes deglycosylated Δ159 (lane 11), bracket deglycosylated fragments, square bracket EndoH-resistant fragments. Asterisks indicate non-specific bands. Δ31–160 variants are not completely deglycosylated by EndoH, resulting in EndoH-resistant fragments (square bracket) which are observed also in reducing conditions (right panel; +/− dithiothreitol (DTT) treatment). Note that endogenous wild-type PrP^c was not detected under used conditions. C. DNI of Δ31–160 variants is similarly inversely related to size of deletion and dependent on intact C-terminal H1–H2 portion. Immunoblot probed with mAb 3F4 showing PK-resistant (3F4)MoPrP co-transfected with indicated constructs into 22L-ScN2a cells (left panel). Δ31–160 was slightly more effective than Δ159. Right panel shows quantification of results as obtained from a triplicate experiment, using densitometry on ImageJ. Bars illustrate mean ± standard deviation. *, p<0.05. D. Δ31–160 forms spontaneous aggregates with moderate chymotrypsin (Chy) resistance. Immunoblots probed with mAbs C16-S (upper panel) or 3F4 (lower panel) for comparing Chy-resistant fragments of Δ31–160 with that of Δ159. Lysates from N2a cells expressing Δ159 or Δ31–160 were digested with indicated concentrations of Chy for 30 minutes and then digested with PNGaseF. A sample from non-transfected N2a cells is shown in the lane 5. Asterisks indicate non-specific C16-S bands. Arrowhead denotes deglycoform of Δ159, bracket deglycoform of Δ31–160.

https://doi.org/10.1371/journal.ppat.1003466.g007

The region N-terminal to H1∼H2 is not necessary for efficient DNI

The observed correlation between DNI efficiencies and the size and/or positioning of the internal deletion suggested that the H1∼H2 region is critically important for PrP-PrP interaction, presumably forming part of the interaction interface. However, there was still the possibility that ΔPrPs with smaller deletions had higher affinities, hence more efficient DNI, because their entire molecular structure was more similar to intact PrP^c, whereas those with larger deletions had lower affinities. Alternatively, the interaction interface might be a discontinuous epitope over N- and C-terminal domains with H1∼H2 being the hinge connecting the domains. In such scenarios H1∼H2 itself would not contribute to the interaction interface. To address this experimentally, we created ΔPrPs lacking the entire sequence from residue 31 to 160 (Δ31–160) and tested their DNI. We preserved the polybasic motif (residues 23–28) and the following glycine residues (Fig. 7A), because this sequence was reported to affect subcellular trafficking and DNI [47], [49]. In addition, we created variants of Δ31–160 with deletions in the C-terminal end of H1∼H2, namely Δ31–160(175), Δ31–160(171–175), and Δ31–175 (Fig. 7A). All constructs were expressed at similar levels upon transient transfection and banding patterns were similar before and when subjected to EndoH digestion, suggesting a similar subcellular trafficking and metabolism. Of note, their N-glycans were not completely EndoH-sensitive with an additional band (Fig. 7B, left panel, square bracket) not converged to the non-glycoform band (Fig. 7B, left panel, bracket). This feature is different from that of ΔPrPs presented above. These additional bands were even unchanged after treatment with 100mM of DTT (Fig. 7B, right panel), although they were absent in PNGaseF treated samples (Fig. 7D, lane 6). In summary, it was evident that a substantial part of N-glycans of Δ31–160 and variants of it were EndoH-resistant. This fraction of PrP might exit ER, pass medial Golgi and reach endosomal/lysosomal compartments [47] or the cell surface, being consistent with the cell-surface distribution of a mutant PrP with the similar structure [50].

Again, all constructs were expressed at similar levels and their efficiencies varied substantially (Fig. 7C, right panel). Δ31–160 showed efficient DNI, whereas DNI of Δ31–160(175) was significantly less, although only one additional residue was deleted (Fig. 7C). As before, there was an inverse correlation between size of deletion and DNI. DNI of Δ31–175 which almost completely lacks H1∼H2 was rather low as expected (Fig. 7C, lane 6). Compared to ΔPrPs with preserved N-terminal portions, PrPs lacking the N-terminal region were higher expressed (see Fig. 7B, lane 11, arrowhead), which might bias their DNI efficacy. We finally analyzed protease resistance of these constructs and found that they were mildly resistant to chymotrypsin digestion (Fig. 7D).

Taken together, ΔPrPs lacking the N-terminal part of PrP were higher expressed and showed a different EndoH resistance pattern. The N-terminal part of PrP was not necessary for exerting DNI and the inverse relationship between size of deletion within the H1∼H2 region and DNI was preserved.

Underlying mechanism of DNI exerted by ΔPrPs

A question of superior importance in the interpretation of our results is what the interaction target of ΔPrPs is to exert their DNI. Initially, protein or factor X has been postulated as target of mutant PrPs for exerting DNI [17]. Recent studies using in vitro conversion systems, e.g. protein misfolding chain amplification (PMCA), have denied the necessity for other factors and DNI in vitro can occur only between the substrate PrP, template PrP^Sc, and the inhibitory mutant PrPs [25], [28]. From a stoichiometric point of view, an interaction between PrP^Sc and inhibitory mutant PrPs was the most likely cause of DNI [25]. These in vitro findings in concert with missing experimental evidence for a factor X make it reasonable to consider that ΔPrPs exert DNI in prion-infected cultured cells by direct interaction with template PrP^Sc or substrate PrP^C. Of importance is also the vast difference in expression levels when substrate (3F4)MoPrP and ΔPrPs are co-transfected. Although ΔPrPs are much less expressed, they are nevertheless able to exert substantial DNI effects, as was also reported before [27]. This suggested that the interaction target of ΔPrPs is PrP^Sc and not PrP^C. However, the possibility that substrate PrP^C is a target still cannot be fully excluded, given that only a small moiety of the total PrP^C population might find its way into the cellular compartment of prion conversion, where ΔPrPs also have access as we report here.

The variation in DNI efficiencies between Δ159, Δ171–175 and Δ159–175 suggests a critical importance of the integrity of the H1∼H2 region and its positioning relative to helix 2 and 3. We did not expect to see a similar tendency in DNI variation among the group of ΔPrPs which contain deletions of almost the entire stretch N-terminal to H1∼H2, like Δ31–160, Δ31–160(175), Δ31–160(171–175) and Δ31–175. Unfortunately, we cannot directly compare DNI efficiencies of Δ31–160 and Δ159 to deduce the direct contribution of the N-terminal deleted region in DNI, e.g. by calculating a ‘DNI per PrP mutant’ of each. This is because their properties seem to be rather distinct from each other, especially with regard to EndoH sensitivity which reflects subcellular localization and trafficking. In addition, Δ31–160 lacks almost the entire N-terminal region, including pre-octarepeat and octarepeat regions, deficiencies of which were shown to impair interaction efficiencies of mutant PrPs with PrP^Sc and/or PrP^C [54]–[56]. However, the obviously very efficient DNI of Δ31–160 demonstrates that the DNI contribution of the region encompassing residue 32 to the end of H1 is very small to negligible in our experimental set-up.

In any case, the finding that Δ31–160 variants with additional deletions in H1∼H2 behaved very similar in DNI as the initial ΔPrPs strongly corroborates our findings and its interpretation. Among the currently postulated models for PrP^Sc, our results are best compatible with the “domain-swapping model” or the “parallel in-register extended β-sheet model” [32], [57]–[59]. Here, H1∼H2 forms intermolecular anti-parallel or parallel β-sheets and H2 and H3 contribute to the stabilization of the superstructure by interaction with H1∼H2, H2 and/or H3 of the other PrP molecule [32], [60]. With such models, the inverse correlation between deletion size and DNI efficiency of ΔPrPs can be explained by changes in the surface area of the interaction interface. In addition, the requirement for appropriate positioning of H1∼H2 relative to H2 and H3 is also explained. Interestingly, these models with H1∼H2 forming intermolecular β-sheet structures were mainly based on in vitro synthesized PrP fibrils [32], [57], [60]. So it is conceivable that the modalities of PrP-PrP interaction as postulated in fibril formation in vitro are relevant also in living cells and form the molecular basis for DNI as observed here. In case full-length PrP and ΔPrPs share similar DNI mechanisms, the implications from the present study might be more widely applicable to scenarios which involve full-length PrP molecules.

There is a caveat to this point of view. It is the transgenic mice expressing PrP106, which completely lack the H1∼H2 region and nevertheless develop prion disease following inoculation with scrapie prions, albeit with extended incubation periods [37]. This suggests that PrP-PrP interactions still occur in this case. On the other hand, PrP106 did not show any DNI in transgenic mice co-expressing endogenous PrP and PrP106 [37], which is again consistent with our hypothesis. One explanation is that PrP106 still maintains some affinity for PrP^Sc through other regions than H1∼H2. This point of view is supported by our findings that even Δ159–175 and Δ31–175 still maintain a certain degree of DNI, although much less pronounced. In addition, H2 and H3 have been reported to have a substantial aggregation tendency to form fibrils in vitro [61]. Such weak affinity might be sufficient for PrP106 to facilitate prion propagation in vivo. Once PrP106 has acquired a PrP^Sc conformation, such nascent PrP^Sc composed of PrP106 would convert normal-isoform PrP106 much more efficiently than would do wild-type PrP^Sc.

The B2-H2 loop is thought to be important for prion propagation especially in inter-species transmission situations [31]. Interestingly, our data suggest that H1∼H2 including the B2-H2 loop is a critical determinant of DNI efficiencies. If full-length PrP interacts with PrP^C or PrP^Sc in the same modalities as does ΔPrP, H1∼H2, along with the B2-H2 loop, might be important for prion propagation because it affects PrP-PrP-interaction efficiencies. Investigations to test whether ΔPrPs and full-length PrPs similarly interact with PrP^Sc are under way. Further investigation of ΔPrP effects would also have practical implications. Identification of the structure which is required for efficient PrP-PrP binding can provide novel therapeutic targets and might lead to the development of small-molecule compounds which recognize PrP structural elements and hamper PrP-PrP interaction.

ΔPrPs have access to locales of prion conversion although not reaching the mid-Golgi: a new trafficking route for prion proteins?

For exerting DNI which involves a direct physical interaction between ΔPrP and PrP^Sc, these two proteins have to meet, ideally in the cellular locale of prion conversion. On the other hand, ΔPrPs were not complex glycosylated and were obviously subject to the known cellular quality control mechanisms in the secretory pathway. They did not reach the plasma membrane and were retained in ER and early-Golgi compartments. To resolve this obvious paradox, another focus of our analysis was on the subcellular trafficking and degradation pathways of these prion proteins. This lead to the finding that ΔPrPs undergo a mixed degradation and have access to subcellular trafficking routes previously not assumed for prion proteins.

A first unexpected finding was that the ΔPrPs studied here have GPI anchors attached, which is of importance as it affects subcellular trafficking and metabolism. An even more surprising finding was that ΔPrPs are subject to both the lysosomal and proteasomal degradation system. The effects of proteasome inhibitors on ΔPrP levels were also unexpected as GPI-anchored PrPs have previously been reported to be unsuitable for ERAD and proteasomal degradation [46]. The lysosomal degradation of ΔPrPs is even more interesting because it explains how ΔPrPs can eventually encounter template PrP^Sc. We and others have previously reported a post-ER cellular quality control pathway which re-routes aggregated PrP or EndoH-resistant mutant PrPs from Golgi apparatus or TGN to acidic digestive compartments [47], [62]. Unlike such EndoH-resistant mutant PrPs which reach at least the medial or late Golgi compartment, ΔPrPs are obviously retained earlier and their trafficking pathway to lysosomes must be a different one. The pronounced sensitivity to 3MA inhibition suggests that this process is class III PI₃K-dependent and part of the macro-autophagy pathway. A similar degradation mechanism for ER-retained proteins, e.g. glycoproteins with high-mannose-type N-glycans [63], misfolded dysferlin [64] and procollagen aggregates [65], was already reported. It utilizes autophagic sequestration and eventually directs these proteins in autophagic vesicles to lysosomes. The same system might be operative in the lysosomal degradation of ΔPrPs. Importantly, before fusion with lysosomes and degradation of its contents, autophagosomes can fuse with late endosomes to form amphisomes which are immunopositive both for LC3 and LAMP1 [66]. Possibly, such amphisomes are the site where ΔPrPs encounter PrP^Sc template first. Of note, this observation might provide novel insights into the cellular biology of prion conversion and involved trafficking and re-cycling pathways. It is clear that substrate PrP^C is converted to PrP^Sc either at the plasma membrane [67] or after endocytosis on the way to lysosomes where it undergoes N-terminal truncation [9], [68], [69]. There is now good experimental evidence that a main intracellular locale of prion conversion is the ERC compartment [70], [71] which strongly implies that there is a re-cycling of PrP^Sc back towards the plasma membrane, in order to sustain continuous presence of PrP^Sc template in conversion-competent compartments. Of note, there is no classical trafficking pathway described from late endosomes back to ERC or early endosomes. On the other hand, a re-cycling of PrP^Sc molecules from late endosomes to TGN has been reported by us and others [35], [72]. From there, such re-routed PrP^Sc has access to either ERC or plasma membrane, closing the cycle.

In the context of our experimental findings, ΔPrPs re-routed to late endosomes via autophagy pathways have to reach template PrP^Sc or substrate PrP^c in a stage before conversion into bona fide PrP^Sc is accomplished. Whether this is the case in late endosomes fused with amphisomes is questionable, although not impossible. It is conceivable that the process of making bona fide and fully PK resistant PrP^Sc is a multi-step pathway which involves more than one cellular compartment. In addition, also ΔPrPs might be subjected to the above described re-cycling from late endosomes/amphisomes back to TGN and plasma membrane. Interestingly, the stoichiometry is not in favor of ΔPrPs compared to wild-type PrP, nevertheless they exert a very efficient DNI. Either only a certain minor subpopulation of PrP^c is prone to be converted into PrP^Sc or ΔPrPs have access to a locale which is extremely powerful in the process of cellular prion conversion.

Another implication of our work is that other proteins or factors residing in ER and early Golgi might also have access to the locale of prion conversion. Without using the known exocytic and endocytic pathways such factors could be involved in the course of prion conversion or PrP^Sc degradation. Given such a scenario, even ER-resident proteins, e.g. ER chaperons, might have the possibility to interact with substrate PrP^C or template PrP^Sc. Alternatively, the non-protein co-factors as mainly described in in vitro systems [73]–[76] might get access. The confinement of such co-factors in small vesicles might change their stoichiometry and their ability to negatively or positively interfere with prion conversion and propagation.

Overall, our data reinforce the notion that autophagy pathways can influence prion propagation [77]. We also show that the intracellular trafficking of PrP isoforms is much more complex than previously anticipated.