A C-terminal Membrane Anchor Affects the Interactions of Prion Proteins with Lipid Membranes*

Background: Membrane attachment of PrP via a GPI anchor is critical for conversion into PrPSc. Results: Semisynthetic membrane-anchored PrP variants are generated and analyzed in different lipid environments. Conclusion: A membrane anchor drastically changes PrP interactions with membranes and enables pore formation. Significance: Membrane-induced conformational changes and pore formation of PrP play a role in the conversion into PrPSc and its toxic effects. Membrane attachment via a C-terminal glycosylphosphatidylinositol anchor is critical for conversion of PrPC into pathogenic PrPSc. Therefore the effects of the anchor on PrP structure and function need to be deciphered. Three PrP variants, including full-length PrP (residues 23–231, FL_PrP), N-terminally truncated PrP (residues 90–231, T_PrP), and PrP missing its central hydrophobic region (Δ105–125, ΔCR_PrP), were equipped with a C-terminal membrane anchor via a semisynthesis strategy. Analyses of the interactions of lipidated PrPs with phospholipid membranes demonstrated that C-terminal membrane attachment induces a different binding mode of PrP to membranes, distinct from that of non-lipidated PrPs, and influences the biochemical and conformational properties of PrPs. Additionally, fluorescence-based assays indicated pore formation by lipidated ΔCR_PrP, a variant that is known to be highly neurotoxic in transgenic mice. This finding was supported by using patch clamp electrophysiological measurements of cultured cells. These results provide new evidence for the role of the membrane anchor in PrP-lipid interactions, highlighting the importance of the N-terminal and the central hydrophobic domain in these interactions.

Prion diseases are fatal neurodegenerative disorders characterized by the conformational conversion of the normal, ␣-helical, cellular prion protein (PrP C ) 3 into the toxic, ␤-sheet-enriched isoform called scrapie PrP (PrP Sc ) (1)(2)(3)(4). PrP C is a glyco-sylphosphatidylinositol (GPI)-anchored protein that is found predominantly at the outer leaflet of the neuronal cell membrane (5). Although the biological function of PrP C , especially within a membrane environment, is not fully understood, the C-terminal membrane attachment of PrP C via a GPI anchor appears to play a key role in the conversion of PrP C into PrP Sc (6 -9). Indeed, several studies provide evidence that the conversion depends on posttranslational GPI anchor attachment of PrP to the outer leaflet of the cell membrane. Methods to release PrP C from the cell surface or to disrupt the transport of PrP C to the plasma membrane are shown to prevent or significantly reduce conversion (10 -12). Baron and Caughey and co-workers (13,14) found further proof that a membrane environment as well as a GPI anchor on PrP is required for conversion using cell-free conversion conditions. Recently, by developing a new cellular system in which epitope-tagged PrP C was expressed in a PrP knockdown neuronal cell line and then exposed to the RML (Rocky Mountain Laboratory) prion (15), Goold et al. (16) found that the conversion occurred primarily on the plasma membrane, especially in GM1-enriched lipid microdomains. At the same time the infection of such cells exposed to the RML prion was extremely rapid. Moreover, lipids are also one of the factors currently believed to be essential for efficient conversion in vitro and in vivo (17)(18)(19). In this context, studying the interaction of PrP carrying a membrane anchor with membranes under fully controlled conditions is of paramount importance, e.g. to identify the effects of membrane composition on protein conformation and conversion. However, GPI-anchored PrP C isolated from neuronal cells is typically heterogeneous with respect to N-glycosylation and the molecular structure of the GPI anchor. Furthermore, contamination with other compounds such as lipids, carbohydrates, and even misfolded PrP cannot be excluded.
Thus, generating membrane-anchored PrP using recombinant expression in combination with selective chemical modifications is a viable alternative, which has been used by several groups so far (9,20,21). Here, we utilize a strategy developed previously in our group that relies on expressed protein ligation to prepare PrP con-taining a C-terminal membrane anchor that mimics a GPI anchor and conveys strong membrane attachment (22). We demonstrated previously that PrP with the GPI anchor mimic performs similarly to recombinant PrP equipped with a synthetic GPI anchor in comparative aggregation assays (23). This finding implies that the effects of both membrane anchors on PrP conversion are very similar and that the lipidated peptide constitutes a useful and easily accessible GPI anchor mimic.
Here we used expressed protein ligation to produce three membrane-anchored PrP constructs: full-length PrP (residues 23-231, FL_PrP-MA), PrP missing its central hydrophobic region (residues 23-104-126 -231, ⌬CR_PrP-MA) and N-terminally truncated PrP (residues 90 -231, T_PrP-MA). The conformational and biochemical properties of the resulting PrP variants within different membrane environments as well as the influence of PrP-lipid interactions on the integrity of membranes are investigated.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-PrP-Mxe intein fusion proteins were produced and purified according to the procedure described previously (22,24). The plasmids (Fig. 1A) were constructed and transformed into Escherichia coli chemicalcompetent cells (BL21(DE3)RIL or Rosetta 2 (Invitrogen) for expression of T_PrP or FL_PrP and ⌬CR_PrP, respectively). The cells were cultured in 2YT medium containing 100 g/ml ampicillin and 30 g/ml chloramphenicol at 37°C and 200 rpm with shaking. Expression was induced at A 600 nm ϭ 1.0 by the addition of 1 mM IPTG to the medium and continued for another 4 h. Cells were harvested, dissolved in TBS-EDTA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5 mM EDTA), lysed with a cell disruption system (Constant Systems, TS series), and then centrifuged at 50,000 ϫ g for 30 min at 4°C to collect the pellet. The pellet was washed with TBS-EDTA buffer containing 0.5% (v/v) Triton X-100 and centrifuged at 50,000 ϫ g for 20 min a 4°C. Washing and centrifugation steps were repeated three times with TBS-EDTA buffer. Next, the pellet was dissolved in 8 M Gnd-HCl, 50 mM Tris-HCl, pH 8.0, by stirring at room temperature overnight. The solubilized pellet solutions were loaded on nickel-nitrilotriacetic acid resin equilibrated with 8 M GndHCl, 50 mM Tris-HCl, pH 8.0. The resin slurry was washed with 8 M GndHCl, 50 mM Tris-HCl, pH 8.0. Finally, PrP-Mxe intein proteins were eluted from the resin with 8 M GndHCl, 50 mM Tris-HCl, 250 mM imidazole, pH 8.0. The PrP-Mxe intein proteins were transferred into 8 M urea, 50 mM Tris-HCl, pH 8.0, using a PD-10 column (GE Healthcare). Purified PrP-Mxe intein fusion proteins in 8 M urea buffer were slowly diluted to 4 M urea with 50 mM Tris-HCl, pH 8.0, and then sodium 2-mercaptoethanesulfonate (Sigma-Aldrich) was added to reach a final concentration of 500 mM, and the mixture was gently stirred at room temperature overnight (Fig. 1B). The cleavage mixtures containing PrP-C␣-thioester proteins were loaded on a preparative C4 column (22 mm diameter, 250 mm length, Grace Vydac) equilibrated with 5% (v/v) of buffer B (acetonitrile and 0.08% (v/v) TFA) mixed with buffer A (dH 2 O and 0.1% (v/v) TFA) at a flow rate of 10 ml/min for 20 min. Next, a linear gradient of 30 to 60% (v/v) of buffer B mixed with buffer A over 60 min was applied to elute PrP-C␣-thioesters from the column (Fig. 1C). Fractions containing PrP-C␣-thioester proteins were pooled and analyzed by MALDI-MS and SDS-PAGE (Fig. 1D). Pure fractions were lyophilized and used for native chemical ligation (NCL) reactions.
Synthesis of Membrane Anchor-An Fmoc-based solid phase peptide synthesis was conducted according to the strategy previously described (22). The peptide was purified by RP-HPLC using a semipreparative C4 column (10 mm diameter, 250 mm length, Grace Vydac) with an elution gradient of 30 to 90% (v/v) of buffer B mixed with buffer A at a flow rate of 3 ml/min for 60 min. Pure fractions were identified by ESI-MS and checked by analytical HPLC (Fig. 2).
Native Chemical Ligation-0.5 mol of PrP-C␣-thioester proteins and 2.5 mol of membrane anchor were dissolved in 500 l of degased NCL buffer (6 M GndHCl, 300 mM Na 2 HPO 4 , pH 7.8), and then 2% (v/v) thiophenol was added and mixed vigorously under an argon atmosphere. The NCL reactions were incubated at 37°C for 24 h (Fig. 3B). The reactions were stopped by adding 15% (v/v) ␤-mercaptoethanol. Reaction yields were checked by SDS-PAGE; ligation products (lipidated PrPs) were purified by the same HPLC procedure used above and lyophilized (Fig. 3, C and D).
Folding of Lipidated PrPs-The lyophilized lipidated PrPs were dissolved in 8 M GndHCl, 50 mM Tris-HCl pH 8.5, 1 mM EDTA, and 3.0/0.3 mM GSH/GSSG. A stepwise dilution strategy was applied to dilute Gdn-HCl from 8 to 2.5 M using dilution buffer containing 20 mM sodium acetate, pH 5.0, 20 mM N-octyl-␤-D-glucopyranoside (SERVA), and 0.3/3 mM GSSG/ GSH on ice. Folding reactions were incubated at 10°C for at least 48 h to prevent precipitation in the following step in which lipidated PrPs were exchanged into storage buffer (20 mM sodium acetate, pH 5.0, and 20 mM N-octyl-␤-D-glucopyranoside) using D-tube dialyzers (Novagen). Ultracentrifugation at 100,000 ϫ g for 1 h at 4°C was applied, and the supernatant was collected for CD spectrometry analysis (Fig.  3E). A similar folding protocol was also applied for PrPs without membrane anchors.
Preparation of Phospholipid Liposomes-10 mg of phospholipids was dissolved in 1 ml of chloroform in a 25-ml roundbottom flask, and chloroform was evaporated by nitrogen flow to form a thin lipid film. The film was dried in a vacuum desiccator for at least 2 h or overnight. Hydration of the film was performed in 20 mM sodium phosphate, pH 7.0, for 1 h, with vortexing every 10 min in between to obtain an opalescent solution (large multilamellar liposomes were formed in this step). The multilamellar liposome suspension was frozen in dry icemethanol for 5 s and then vortexed rapidly to form a thin frozen layer of phospholipid on the wall of the tube. Next, the frozen phospholipid layer was thawed by warming up the tube under cool tap water and then under warm (42°C) water for 30 s. The freeze/thaw cycle was repeated five times. The lipid solution was passed 20 times through a mini-extruder with a 100-nm polycarbonate membrane (Avanti Polar Lipids). The resulting liposomes were collected and kept under argon at 4°C for the next steps. Liposomes were produced using non-charged phospholipid DPPC or DOPC, anionic phospholipid POPG, or phospholipid mixtures mimicking neuronal membrane (NM). These NM mixtures consisted of PC:PE:PI:SM:Chol:POPG (30: Flotation Assays-These assays were carried out as described previously (26). Briefly, PrP proteins and liposomes were mixed and incubated for 30 min at room temperature. The mixtures were loaded under a discontinuous iodixanol (Optiprep, Sigma-Aldrich) gradient containing, sequentially, 1 ml of 36% (v/v), 1 ml of 31%, and 0.4 ml of 5% iodixanol in a centrifuge tube. The gradient was centrifuged at 200,000 ϫ g at 4°C for 3 h (MLS-50 rotor, Optima ultracentrifugation system, Beckman Coulter Inc.). Next, 200-l fractions were collected from top to bottom of the centrifuge tube, and PrP in each fraction was precipitated by TCA and dissolved in 1ϫ SDS loading buffer for SDS-PAGE and Western blot analysis with polyclonal anti-PrP antibody M20 (Santa Cruz Biotechnology) (Fig. 4, C and D).
Proteinase K (PK) Resistance Assay-20 l of PrP-phospholipid mixtures were subjected to PK digestions at 37°C for 30 min with the indicated PK:PrP molar ratios of 1:4, 1:8, and 1:16. Reactions were stopped by adding 5 mM PMSF and kept on ice for 15 min. The mixtures were subjected to SDS-PAGE and Western blot analysis as described above.
Calcein Release-The assay was performed as described previously (27) with minor modifications. The release of calcein from the calcein-containing liposomes was monitored by calcein fluorescence at 512 nm over 120 min after mixing five parts of the liposomes and one part of protein solution (5:1 (v/v)). The excitation wavelength was 490 nm. Experiments were carried out at room temperature. The concentration of liposomes was 50 M, and protein concentrations varied from 10 to 500 nM. The percentage of calcein release (R f ) was calculated based on Equation 1, where F t is the fluorescence intensity recorded at time t after adding PrP to liposomes, F o is fluorescence intensity without PrP proteins, and F tot is the total fluorescence intensity recorded after total release of calcein by disrupting the liposomes with 1% (v/v) Triton X-100. Each measurement was repeated three times and measured in 96-well plates with a microplate reader (BioTek Synergy Mx).
NBD Fluorescence Quenching-NBD-phosphatidylethanolamine (PE) (Avanti Polar Lipids) was dissolved in chloroform and added to phospholipids at 1 mole % and dried together under nitrogen flow to form thin films. The films were used to generate the NBD-liposomes as described above. Samples for fluorescence measurements were prepared by adding PrP at a concentration of 1 M, or using buffers only (blank run), and 200 M NBD-liposomes to 20 mM sodium phosphate, pH 7.0. Fluorescence quenching measurements were initiated by adding the quencher sodium dithionite at 10 mM (Merck) from a 130 mM stock solution freshly prepared in 50 mM Tris-base buffer, ϳpH 11. NBD fluorescence kinetics were recorded on a Fluoromax 4 system (Horiba) for 900 s at 22°C using an excitation wavelength of 467 nm and an emission wavelength of 535 nm. The NBD fluorescence decay was calculated as a percentage of the initial NBD fluorescence measured before adding sodium dithionite (28).
Circular Dichroism (CD) Spectroscopy-Far-UV CD spectra were recorded at 22°C on an Applied Photophysics Chirascan Plus system. PrP proteins were diluted in 20 mM sodium acetate, pH 5.0, to final concentration of 0.1-0.2 mg/ml. CD spectra were acquired at a speed of 20 nm/min, a 1-nm bandwidth, and a response time of 4 s. All spectra were recorded in a 0.1-cm quartz cuvette with a wavelength ranging between 195 and 260 nm. Each spectrum was recorded 10 times and subsequently averaged.
Tryptophan Fluorescence Measurements-Tryptophan fluorescence measurements were performed by mixing the protein solution with an equal volume of lipid liposomes. After equilibration for 3 min, four fluorescence spectra from 300 to 450 nm were recorded and averaged on a Fluoromax 4 system (Horiba) with an excitation wavelength of 295 nm (2 nm bandwidth) at 22°C. Fluorescent background was also recorded and subtracted from the final spectra (27). Tryptophan fluorescence quenching measurements were performed by adding freshly prepared aliquots of 1 M acrylamide stock solution in 20 mM sodium phosphate, pH 7.0, to PrP in the presence or absence of liposomes (DOPC or POPG) at concentrations of 3 M PrP and 3 mM lipid liposomes. Tryptophan fluorescence spectra for each concentration of the quencher acrylamide were recorded and corrected (29). The fluorescence quenching data were analyzed by using the Stern-Volmer equation for collisional quenching as given in Equation 2, gives a linear plot that yields the K SV value (59).
Cryo-electron Microscopy (EM) Measurements-Two samples including POPG liposomes alone and POPG liposomes mixed with lipidated ⌬CR_PrP were prepared and visualized using a urea system (FEI Tecnai F30 Helium "Polara"). The resulting images were processed by using ImageJ software.
Electrophysiology-The pore-forming activity of FL_PrP and ⌬CR_PrP with and without a membrane anchor was measured in HEK293 cells using patch clamping in the whole-cell configuration. HEK293 cells were purchased from American Type Cell Culture (ATCC CRL-1573). Cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen), 100 units/ml penicillin, and 100 g/ml streptomycin (Sigma-Aldrich).
Whole-cell currents were acquired from HEK293 cells at room temperature (19 -22°C) 24 -48 h after plating using an Axopatch 200B amplifier and Digidata 1440A with pCLAMP10.2 software (Axon Instruments). Currents were recorded at 10 kHz and filtered at 5 kHz. Glass coverslips with the cultured cells were transferred to a chamber of 1 ml capacity, mounted on the stage of an inverted microscope (Zeiss, Axiovert 100). After gigaohm seal formation, the equilibration period of 5 min was followed by recordings at a holding potential (E h ) of Ϫ100 mV unless stated otherwise. During a control period of 5 min, no ion current activity was observed. Aliquots of indicated prion proteins were thawed on ice and added into the bathing solution, resulting in the desired concentrations. Subsequent membrane current activities were recorded for up to 15 min. Capacity transients were cancelled, and series resistance was compensated. Access resistance was monitored and 75% compensated. Data acquisition and storage were performed directly on a PC. Gigaseals were monitored continuously during the experiments to avoid inadequate voltage clamp. Current densities were calculated by dividing the total current generated in a given cell by the membrane capacitance of the cell.
For studying pore-induced divalent cation current, the extracellular bath solution contained 20 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, and 140 mM choline-Cl titrated to pH 7.4 with methane sulfonic acid. The borosilicate glass patch pipettes (Harvard Apparatus, Holliston, MA) with resistances of 2-4 M⍀ were pulled and polished using a DMZ universal puller (Zeitz Instruments, Martinsried, Germany) and were filled with pipette solution containing 145 mM CsCl, 3 mM MgCl 2 ,10 mM HEPES and 10 mM EGTA, titrated to pH 7.25 with CsOH. The composition of the bath solution for the measurement of monovalent cation currents was 145 mM NaCl, 2.7 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , 5.5 mM glucose, and 10 mM HEPES, adjusted to pH 7.4 with 1 M NaOH solution. The pipette solution contained 135 mM potassium methanesulfonate, 10 mM KCl, 6 mM NaCl, 1 mM Mg 2 ATP, 2 mM Na 3 ATP, 10 mM HEPES, and 0.5 mM EGTA, adjusted to pH 7.2 with 1 M KOH solution.
Strong and Specific Binding of PrP Variants to Phospholipid Liposomes via Their C-terminal Membrane Anchors-To assess the binding of prion proteins to liposomes, a flotation assay in a discontinuous iodixanol gradient (36,31, and 5% (v/v) iodixanol) was used (26,33). Two kinds of phospholipid liposomes, DPPC (a temperature-dependent gel-phase, noncharged phospholipid) and POPG (an anionic phospholipid), were mixed with PrP constructs and loaded underneath the iodixanol gradient. In the case of flotation assays using the noncharged phospholipid DPPC, PrPs without a membrane anchor were deposited at the bottom of the centrifuge tube (Fig. 4A,  black). In contrast, PrPs with a C-terminal membrane anchor displayed strong binding to DPPC liposomes. These were found in fractions 1, 2, and 3 from the upper volume of the centrifuge tube (Fig. 4A, red). All PrPs carrying the membrane anchor analyzed here bind to liposomes, as demonstrated previously for T_PrP-MA (22). The high affinity binding of prion protein to vesicles made of the anionic phospholipid POPG has been reported previously to induce structural changes from ␣-helical into ␤-sheet-rich structures (26,27,34). Therefore, we analyzed the interaction of PrPs with POPG liposomes in greater detail. In contrast to the behavior with DPPC liposomes described above, all PrP variants (with and without membrane anchor) interacted with POPG liposomes (Fig. 4B). However, there was a pronounced difference between PrP variants with and without membrane anchor after treatment of PrP-POPG liposomes with 10 mM NaOH, which can partially extract PrP from POPG liposomes by disrupting weak electrostatic interactions without denaturing the protein (26). For PrPs without a membrane anchor, the interactions with POPG liposomes were disrupted by NaOH, and large amounts of protein were deposited at the bottom of the iodixanol gradient (Fig. 4B, black). In contrast, the interactions of PrPs with membrane anchors and POPG liposomes were stable under NaOH extraction conditions, as indicated by PrPs present mainly in the top fractions of the iodixanol gradient (Fig. 4B, red). These results clearly demonstrated that all PrP variants interact with negatively charged liposomes but that only the C-terminal membrane anchor imparts a strong interaction.
Conformation of PrP Constructs in the Presence of Phospholipid Membranes-Far-UV CD spectroscopy was used to investigate whether PrP variants with and without a membrane anchor undergo conformational changes during interaction with phospholipid liposomes. PrP variants were incubated with liposomes at 25°C, and CD spectra were recorded. The CD data depicted in Fig. 5A and Table 1 for PrPs without a membrane anchor indicate that PrP mixed with non-charged DOPC liposomes (green dots) did not undergo a change in their predominantly ␣-helical conformation and were almost indistinguishable from PrPs in the absence of lipids (black dots). In contrast, incubation with anionic POPG liposomes leads to a significant change in the secondary structure of PrP, from a typically ␣-helical to a ␤-sheet-rich structure (Fig. 5A, red), accompanied by an increase in random coil content as well ( Table 1). The conformational changes of FL_PrP and T_PrP without a membrane anchor upon interaction with POPG liposomes found here are consistent with previous reports (26,27). In addition, we found the same effect for ⌬CR_PrP without a membrane anchor. Surprisingly, PrP variants carrying a membrane anchor did not convert into ␤-sheet-rich structures during interaction with POPG liposomes. In the case of T_PrP-MA and ⌬CR_PrP-MA, their secondary structures remained mostly ␣-helical (Fig. 5A), without any significant changes in secondary structure parameters (Table 1). For FL_PrP-MA attached to POPG liposomes, a slight loss of secondary structure was observed, whereas more ␤-sheet was observed for FL_PrP without membrane anchor ( Fig. 5A and Table 1). This finding suggests that the secondary structure of PrP attached to the lipid membrane via a membrane anchor differs considerably from that of PrP without a membrane anchor (Table 1). This possibly accounts for the distinct differences in biochemical characteristics such as PK resistance described below.

Proteinase K Resistance of PrP Constructs within a Membrane
Environment-PK resistance is one of the typical characteristics of PrP Sc and distinguishes it from PrP C , which can be digested completely. Here PK digestion was used to evaluate whether the binding of PrPs to anionic POPG liposomes, as demonstrated above, can lead to PK resistance. PK digestion was carried out with PK to PrP molar ratios of 1:16, 1:8, and 1:4, after which samples were analyzed by Western blot. One sample without  OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 PK treatment was always included as a control. FL_PrP without a membrane anchor showed PK-resistant bands after mixing with POPG liposomes, whereas FL_PrP incubated in the absence of POPG liposomes remained PK-sensitive (Fig. 5B,  upper left). This result is in good agreement with a previous study by Wang et al. (26).

Pore Formation by a Membrane-anchored Prion Protein Variant
In contrast, FL_PrP-MA incubated with POPG liposomes still remained PK-sensitive under similar conditions (Fig. 5B,  lower left). This observation supports our CD data, which indicates that binding of FL_PrP-MA to POPG liposomes differs from binding of FL_PrP without a membrane anchor. This different binding prevents formation of PK-resistant PrP. For T_PrP and ⌬CR_PrP no PK resistance was observed, either for variants with or without membrane anchor (Fig. 5B). These variants remained PK-sensitive and even though changes in secondary structure were observed, no fibril formation occurred (Fig. 5). This finding is consistent with the idea that both electrostatic (weak in the case of T_PrP due to the missing, positively charged N-terminal region) and hydrophobic (lacking in ⌬CR_PrP) interactions play important roles in the interaction of PrP with anionic POPG liposomes (27,33). To investigate this point further, we analyzed the interaction of different PrP variants with membranes in more detail using tryptophan fluorescence quenching.
Characterizing the Interactions of PrP Variants with Membranes via Tryptophan Fluorescence Quenching-Tryptophan fluorescence quenching by water-soluble quenchers such as acrylamide can provide information about the location of tryptophan residues in the complex structure of membrane-bound proteins (35). In the case of PrP, seven tryptophan residues are present in the unstructured N-terminal region (residues 23-99), which is supposed to play a key role in interaction of prion proteins and phospholipid membranes (26,27). Linear Stern-Volmer plots were created for the quenching of tryptophan fluorescence of anchorless FL_PrP in phosphate buffer and under denaturing conditions after the addition of 8 M urea (Fig. 6A). The resulting Stern-Volmer constants (K SV ) were found to be 5.96 Ϯ 0.18 M Ϫ1 and 6.16 Ϯ 0.26 M Ϫ1 , respectively ( Table 2). The good agreement between both values suggests that tryptophan residues of folded and unfolded FL_PrP without a membrane anchor are equally accessible to the quencher in solution.
Next, linear Stern-Volmer plots corresponding to the quenching of the tryptophan fluorescence of each PrP variant in solution and mixed with liposomes were measured (Fig. 6,  B-D). The resulting K SV values are summarized in Table 2. K SV values for PrPs without a membrane anchor in the presence of POPG liposomes were found to be significantly lower than those of PrPs in solution, indicating that the tryptophan residues have become less accessible to the quencher acrylamide due to their binding to POPG liposomes. This suggests either close contact with or even insertion of tryptophan residues into the lipid membrane (27). To further illuminate the insertion behavior, the same experimental procedure was applied to FL_PrP with and without a membrane anchor mixed with noncharged DOPC liposomes. The resulting K SV values were found to be 5.83 Ϯ 0.11 M Ϫ1 and 5.20 Ϯ 0.11 M Ϫ1 , respectively, which is very close to those values measured for PrPs in solution ( Table 2). The close agreement of K SV values measured in the presence and absence of DOPC liposomes indicates little protection of tryptophan residues from quencher. This is due to absence of or very weak binding of the N-terminal region to DOPC liposomes and is in good agreement with data from calcein release and NBD fluorescence quenching assays shown below. Thus the C-terminal membrane anchor attachment has no significant impact on the interaction of PrP N termini with membranes composed of DOPC.
In contrast to these results, the interactions of PrP variants with anionic POPG liposomes are strongly influenced by the membrane anchor, as indicated by K SV values. For T_PrP and ⌬CR_PrP constructs, the interaction of variants with a membrane anchor and POPG liposomes resulted in far lower K SV values than those found for variants without a membrane anchor (Table 2). Thus, the C-terminal membrane anchor enhanced the contact and/or insertion of tryptophan residues of T_PrP and ⌬CR_PrP into lipid bilayers of POPG liposomes.
Surprisingly, the membrane anchor limited the protection of tryptophan residues from FL_PrP by anionic POPG liposomes. The K SV values indicated that the tryptophan residues from FL_PrP-MA were more accessible to the polar quencher acrylamide than those from FL_PrP without a membrane anchor. These results suggest that the contact/insertion of N-terminal tryptophan residues of FL_PrP without a membrane anchor into the POPG bilayer is stronger than that of FL_PrP-MA. Overall the C-terminal membrane anchor influences the binding of the N-terminal region of PrP variants to POPG liposomes. Having analyzed the impact of the membrane anchor on protein conformation and flexibility, we also wanted to investigate the effect of PrP variants on membrane integrity.
Calcein Release from Liposomes-To investigate whether the binding of prion proteins can affect the stability of lipid membranes, a well established method based on the release of calcein from liposomes was applied. The amounts of calcein released from DOPC liposomes upon interaction with PrPs with and without a membrane anchor showed no significant difference compared with background controls (BG) in the absence of any PrP (Fig. 7A).

TABLE 1 Secondary structure parameters of PrP variants in buffer only and upon interaction with DOPC or POPG liposomes
The CD spectra depicted in Fig. 5A were analyzed and deconvoluted using CDNN software. This result suggested minimal binding of PrPs without a membrane anchor to DOPC liposomes, as already deduced from flotation assays. Interestingly the already established interactions of PrPs with a membrane anchor with DOPC liposomes also had no influence on the release of calcein. This result implies that the insertion of the C-terminal membrane anchor into liposomes has no destabilizing effect per se.

␣-Helices ␤-Sheets Turns Random coils
In contrast to the interaction of PrPs with non-charged DOPC liposomes, the interaction of PrPs with anionic POPG liposomes leads to a distinct release of calcein when compared with background control (Fig. 7B). The binding of PrP variants to POPG liposomes has a destabilizing effect on the membrane and supports the fact that electrostatic interactions are a pre-requisite for binding of PrP to lipid membranes. To test more physiologically relevant membranes using the calcein release assay, liposomes consisting of a phospholipid mixture that mimics neuronal membranes were produced. Two kinds of anionic phospholipids (POPG and POPS) were used to analyze whether the difference in the anionic phospholipid component can affect the interaction of PrPs with membranes. As shown in Fig. 7, C and D, no significant difference of calcein release was observed for binding of PrP variants to NM-PG and NM-PS liposomes. One interesting observation is that the total amount of calcein released from liposomes containing PrPs with a membrane anchor is always higher than that from liposomes interacting with PrPs without a membrane anchor. Only in the case of T_PrP is no significant difference observed (Fig. 7). These results suggest that attaching PrP constructs via their C-terminal membrane anchor helps the N-terminal polybasic region of PrP (residues [23][24][25][26][27], which is lacking in the T_PrP construct, to enhance the electrostatic interaction of PrP with anionic phospholipid membranes. Indeed, Wang et al. (33) already determined that this region initiates (electrostatic) interaction between PrP and anionic phospholipids. Moreover, Turnbaugh et al. (36) recently found that this polybasic region dictates the efficiency of prion propagation by binding to PrP Sc . A deletion of this region can dramatically reduce susceptibility of prion protein to prion infection and significantly elongate the survival of transgenic (⌬23-31 PrP) mice after scrapie inoculation (36).
The interactions of ⌬CR_PrP equipped with a membrane anchor with POPG, NM-PG, and NM-PS liposomes revealed the highest extent of calcein release followed by FL-PrP-MA (Fig. 7, B-D). Such an effect is likely explained by the formation of discrete membrane pores if gross destabilization and rupture of liposomes can be excluded as in this case (see cryo-EM data in Fig. 9B). In previous experiments based on electrophysiological measurements with neuronal cells expressing ⌬CR_PrP, Solomon et al. (37,38) already surmised that ⌬CR_PrP might form ion channel-like assemblies.
NBD Fluorescence Quenching-To further understand the effects of PrP binding and the role of the membrane anchor on the integrity of lipid bilayers, NBD fluorescence quenching was used (Fig. 8A). With this approach we detected how much and how fast the quencher sodium dithionite could penetrate into

TABLE 2 Stern-Volmer constant (K sv ) values of PrP variants in solution or mixed with POPG and DOPC liposomes
PrP-loaded vesicles and quench the NBD fluorescence. Fluorescently labeled PE (NBD-PE) with the fluorophore attached to the head group of phosphatidylethanolamine was incorporated into the liposomes at a ratio of labeled to unlabeled lipids of 1 mole %. NBD is localized close to the head group of the lipids and thus is a sensitive probe to detect the interaction of proteins/peptides with the lipid bilayer. In addition, the accessibility of the dithionite quencher to NBD moieties can also be correlated with bilayer perforation or disruption (39,40).   OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 30155
NBD-PE incorporated into DOPC (left) and POPG (right) after these liposomes were mixed with different PrP constructs. Fig.  8C shows similar quenching experiments with NM-PG (left) and NM-PS (right) liposomes. The control (SUVs only) is based on incubation of the liposomes with dithionite in the absence of any PrP. In all cases the addition of any PrP variant led to accelerated fluorescence quenching.
However, in the case of PrPs mixed with NBD-DOPC liposomes, no distinguishable difference in NBD fluorescence quenching was recorded for any of the PrP variants (Fig. 8B). This observation indicates that PrP variants with and without a membrane anchor have no significant effect on NBD-DOPC liposomes, similar to the results obtained in our calcein release assays. The interactions of T_PrP and FL_PrP with NBD-POPG liposomes also did not show any significant differences between proteins with and without a membrane anchor (Fig. 8B). Therefore the binding of these variants to NBD-POPG liposomes does not destabilize or permeabilize liposome membranes, even though a clear interaction was demonstrated by floatation and tryptophan quenching assays.
In contrast, the interaction of ⌬CR_PrP-MA with NBD-POPG liposomes led to a very significant quenching effect (Fig.  8B, right panel, cyan), which provides additional evidence that ⌬CR_PrP-MA induces pores in POPG liposomes. Moreover, ⌬CR_PrP without a membrane anchor did not permeabilize NBD-POPG liposomes (Fig. 8B, right panel, blue), indicating that the C-terminal membrane anchor is a crucial factor contributing to pore formation. Fluorescence quenching assays were performed with different concentrations (from 0 to 1 M) of ⌬CR_PrP-MA. The quenching rate has a fast and a slow component, with the fast one corresponding to quenching of fluorophores immediately accessible to the quencher on the outer surface of all liposomes. The slower component represents the average rate of penetration of the quencher through the bilayer, where it quenches fluorophores on the inner leaflet (Fig. 8D). Langner and Hui (41) suggest a single exponential function that combines both components of the quenching rate. Here, the plot of concentrations of ⌬CR_PrP-MA versus NBD fluorescence was fitted with an exponential function to clearly demonstrate that the extent of NBD fluorescence quenching increased with the amount of ⌬CR_PrP-MA added (Fig. 8D). In an extension of this assay, liposomes mimicking neuronal membranes (NM-PG and NM-PS, with the phospholipid components as described above) including 1 mole % NBD-PE were prepared and mixed with PrPs with and without membrane anchors. Although NM-PG and NM-PS liposomes were made of different anionic phospholipid components (POPG and POPS, respectively), the NBD fluorescence quenching data depicted in Fig. 8C for the interactions of both liposomes with PrPs reveal a close similarity and are in very good agreement with previous results obtained from calcein release assays (Fig. 8A). However, NM-PG and NM-PS liposomes were less stable than POPG liposomes (as seen during the preparation of calcein-loaded liposomes), and thus the fluorescence quenching processes of NM-PG and NM-PS for all samples including controls (Fig. 8C, SUVs only (black)) are much faster and show higher background reaction. In addition, the fluorescence quenching results in Fig. 8C for NBD-NM-PG and NBD-NM-PS liposomes indicated that the C-terminal membrane anchor increased the extent of NBD fluorescence quenching in liposomes mimicking neuronal membranes. This observation most likely reflects a preferred localization of PrPs with membrane anchor at the lipid-water interface, where the NBD probe attached to the hydrophilic head group of PE also localizes. Interestingly, similar to the observation from NBD-POPG liposomes, the interactions of ⌬CR_PrP-MA with NBD-NM-PS or NBD-NM-PG liposomes also resulted in the highest extent of NBD fluorescence quenching (Fig. 7D, cyan). Therefore the results from NBD fluorescence quenching strongly support the hypothesis that the membrane-anchored versions of ⌬CR_PrP and, to a lesser extent, of FL_PrP induce the formation of pores in liposomes containing anionic phospholipids, as found in our previous assays and in vivo.
Cryo-EM-Cryo-electron microscopy of SUV samples in the presence and absence of ⌬CR_PrP-MA was carried out to ensure integrity of POPG liposomes under all conditions used here and to prove that no disruption or visible damage of vesicles occurred during our measurements. Two samples, POPG liposomes only and POPG liposomes plus ⌬CR_PrP-MA, were prepared and visualized with a transmission electron microscopy (TEM) system. The resulting images (Fig. 9B) clearly show that ⌬CR_PrP-MA bound to POPG liposomes has no visible influence on the integrity of liposomes. Therefore the efflux of calcein and influx of acrylamide or dithionite did not occur due to nonspecific damage or disruption of liposomes by ⌬CR_PrP-MA. The formation of pores by ⌬CR_PrP-MA is a viable explanation for the observed effects. To further support this hypothesis, we performed patch clamp experiments on whole cells to detect ion flow through such pores.
Lipidated ⌬CR_PrP Induces Spontaneous Inward Currents in HEK293 Cells-Whole-cell currents from HEK293 were recorded to determine the pore forming activity of ⌬CR_PrP, with and without membrane anchor, and of FL_PrP-MA. As shown in Fig. 9A, neither ⌬CR_PrP without a membrane anchor nor FL_PrP-MA induced currents with a monovalent cation (Na ϩ ) or a divalent cation (Ca 2ϩ ). ⌬CR_PrP only infrequently generated small currents in the pA range (95 Ϯ 10 pA/pF, n ϭ 3). In contrast, the addition of ⌬CR_PrP-MA up to a concentration of 500 nM induced a large spontaneous inward current. ⌬CR_PrP-MA-induced pores conducted Na ϩ (2.1 Ϯ 0.1 nA/pF, n ϭ 5) and Ca 2ϩ (1.9 Ϯ 0.5 nA/pF, n ϭ 5) cations in all tested cells after a wash-in period of 5 to 7 min, similar to the known pore-forming toxin Panton-Valentine leukocidin used as a positive control here (data not shown).

The C-terminal Membrane Anchor Strongly Influences the Biochemical and Conformational Characteristics of PrP Variants in Different Membrane
Environments-PrP is a GPI-anchored protein that is localized at the outer leaflet of the plasma membrane (11), and its close association with this membrane already suggests that the conversion of PrP C into PrP Sc may be influenced by the lipid bilayer as demonstrated previously (14 -16). Several studies of the interactions of PrP with phospholipid membranes have been carried out, mostly by using bacterially expressed PrP constructs without any posttranslational modi-fications (7,26,27,33,34,42). We have successfully extended our previously developed expressed protein ligation strategy to generate multimilligram amounts of three different PrP variants. Our previous results demonstrate that homogeneous, folded PrPs equipped with a membrane anchor behave similarly to GPI-modified PrP in vitro (23), and they have allowed us to address several questions about the membrane interaction and conformational changes of PrPs.
As demonstrated above, the presence of a membrane anchor dramatically influences the behavior of PrP when it is "tethered" to membranes, in comparison with free, soluble PrP. In our flotation assays, weak (alkaline-sensitive) binding of nonmodified PrPs to anionic POPG liposomes was observed, as described previously by others (26,27,33,43). This interaction nevertheless induces conversion of the typical ␣-helical structure of PrPs into ␤-sheet-enriched structures for PrPs without membrane anchor, which agrees well with earlier findings (26,27,33,34,42). The results from these assays also revealed that PrPs with a membrane anchor bind very strongly to negatively charged liposomes, mainly via their C-terminal anchors. How-ever, PrPs with a membrane anchor did not undergo a similar conformational change into ␤-sheet-rich structures. T_PrP-MA and ⌬CR_PrP-MA still retained a predominantly ␣-helical structure, and FL_PrP-MA partially converted into a random coil. These findings are summarized in Scheme 1 and provide an overview of the changes in secondary structure measured by CD and the functional consequences discussed in more detail below. In general, we concluded that the membrane anchor modulates the interaction of PrP variants with negatively charged membranes in a way that inhibits conformational changes into ␤-sheet-enriched structures, thereby slowing PrP aggregation and fibril formation as demonstrated previously (22). Such processes seem to be accelerated in the absence of the membrane anchor, as can be caused by enzymatic GPI anchor removal in vivo. However, we could not exclude the early formation of smaller aggregates.
The conformational changes in FL_PrP without a membrane anchor in the presence of anionic POPG liposomes were also found to lead to increased PK resistance (26), one of the typical FIGURE 9. Electrophysiological and cryo-EM analysis. A, ⌬CR_PrP with a membrane anchor induces spontaneous inward currents in HEK293 cells. Representative whole-cell patch clamp recordings are shown. HEK293 cells were clamped from a holding potential of 0 mV to a test potential of Ϫ100 mV for 10 s every 30 s. After a control phase, the indicated proteins were added directly to the bath solution. In the presence of FL_PrP-MA (left) and ⌬CR_PrP without a membrane anchor (middle), no or only weak pore forming activity was observed, whereas ⌬CR_PrP-MA induced large sodium (top right) and calcium (bottom right) inward currents. B, cryo-EM images of POPG vesicles (arrows) mixed with ⌬CR_PrP-MA (left) and POPG vesicles alone (right). The scale bar is 100 nm.
biochemical characteristics of PrP Sc . A similar behavior in the presence of POPG liposomes has already been described by Wang et al. (17,26). However, T_PrP and ⌬CR_PrP remained PK-sensitive even though a conversion into ␤-sheet-rich conformations occurred. An explanation of this apparently inconsistent behavior is based on the missing protein domains in both variants. T_PrP is missing the basic N-terminal region and has much weaker electrostatic interactions with POPG; ⌬CR_PrP lacking the core hydrophobic region cannot engage in hydrophobic interactions with the membrane. We suggest that these differences are sufficient to maintain the PK sensitiv-ity of T_PrP and ⌬CR_PrP on the time scale of our experiments (up to 12 h). A straightforward interpretation of these findings is that the interactions of both of the latter constructs with anionic POPG are much weaker than that of FL_PrP, which can be protected from PK by its strong association with POPG liposomes. The addition of Triton X-100 to disrupt the liposomes countermands these effects and allows PK to digest FL-PrP as well (data not shown). Tryptophan fluorescence quenching experiments also support this interpretation. FL_PrP-MA, as well as the other two PrP variants with a membrane anchor, bound to POPG vesicles was sensitive to PK, suggesting a different binding mode that keeps the protein fully exposed to PK.
A suitable method for studying the binding of tryptophanrich domains, as found in the N terminus of PrP, to membranes is by monitoring the changes in intrinsic tryptophan fluorescence and the response to a fluorescence quencher. Our tryptophan fluorescence experiments show that the N-terminal region of PrP (residues 23-99) binds to anionic POPG liposomes but not to non-charged DOPC liposomes, as already observed by others (27). In general this effect does not depend on the presence or absence of a membrane anchor, as a similar effect was found for all PrP variants in interactions with POPG liposomes. Based on this observation a direct interaction of the PrP N-terminal region with phospholipid membranes always occurs, as described for other lipid-anchored proteins such as Src proteins and HIV-1 Gag protein (27,44). However, tryptophan fluorescence quenching occurred to a different extent for PrPs with a membrane anchor compared with PrPs without a membrane anchor. In the case of the T_PrP and ⌬CR_PrP constructs, the comparison of K SV values between PrPs with and without a membrane anchor revealed that the membrane anchor enhances the insertion of tryptophan residues into the bilayer. In contrast, the membrane anchor limits the extent of binding for FL_PrP. Scheme 1 provides a simplified illustration of the different conformational effects of binding of PrP variants with or without a membrane anchor to negatively charged liposomes. Similar conformational effects have been described previously, e.g. for the sensitivity of PrP C to treatment with GPIspecific phospholipase when compared with PrP Sc . Whereas PrP C was easily cleaved by GPI-specific phospholipase and removed from the membrane surface, PrP Sc was retained on the membrane under similar conditions (45).
Pore Formation by ⌬CR_PrP with Membrane Anchor-Spontaneous neurodegenerative illness has been observed in transgenic mice expressing PrP molecules carrying any one of several different internal deletions (46 -48). The most severe phenotype is produced by the ⌬CR mutation, which consists of deletion of 21 amino acids in the central, primarily hydrophobic region of PrP C (47). Transgenic mice expressing the ⌬CR_PrP mutant in the absence of endogenous, wild-type PrP suffered from lethality during the neonatal period. Interestingly, ⌬CRPrP was found to contain the full set of posttranslational modifications (N-glycosylations and GPI attachment) typical of wild-type PrP C (49). Expression of ⌬CRPrP in transfected cell lines was found to induce spontaneous ionic currents, and the same was true for PrPs harboring several point mutations in the central hydrophobic region that are linked to familial prion diseases in humans (37,38). These ionic currents can be silenced SCHEME 1. Schematic illustration of how the PrP variants FL_PrP, ⌬CR_PrP, and T_PrP with or without an MA may interact with anionic phospholipid membranes. The scheme illustrates a distinct binding mode of predominantly ␣-helical PrP constructs (large blue circle) carrying a membrane anchor to POPG vesicles when compared with PrP constructs without a membrane anchor. Tryptophan residues (small red circles), the N-terminal polybasic region (residues 23-27, green oval) and central hydrophobic region (residues 105-125, red line) are indicated. The position of the latter region has not been determined by our experiments and is therefore not accurately depicted in this scheme. In addition, the conformational changes of the globular domain of PrP (residues 125-231) from mainly ␣-helical (large blue circle) to random coil (orange oval) or ␤-sheet (red rectangle) are highlighted. by co-expression of wild-type PrP C . Relying on biophysical results from cell-based assays Biasini et al. (50,51) suggested two possible explanations for these ionic currents: either ⌬CR_PrP and the familial mutants in the CR influence endogenous ion channel(s), or these PrP constructs themselves form an ion channel or a pore in the cellular membrane.
To follow up on these studies and make use of our unique membrane-anchored PrP variants, we carried out calcein release assays from POPG liposomes. As shown above the largest amount of calcein was released upon binding of ⌬CR_PrP-MA to POPG liposomes. Gross liposome disruption by ⌬CR_ PrP was ruled out by analyzing liposome integrity using cryo-EM. This was necessary because such membrane-disrupting effects have been described for other transmembrane proteins or peptides (52)(53)(54). Additional support for the formation of pores within the lipid bilayer was obtained by NBD fluorescence quenching assays. Here binding of ⌬CR_PrP-MA to liposomes also induced significant quenching of NBD fluorescence, much higher than observed for all other PrP variants with or without a membrane anchor. Comparative electrophysiological measurements with different cations (Ca 2ϩ and Na ϩ ) on HEK cells incubated with ⌬CR_PrP-MA were carried out, and strong inward currents were observed. This finding is compelling evidence for pore formation by ⌬CR_PrP-MA and may explain the dramatic neurotoxicity that this mutant produces in transgenic mice (47). Based on our results, we hypothesized critical roles for both C-terminal membrane attachment and the N-terminal domain of PrP in pore formation, suggesting that this latter region dips into the lipid bilayer to create a pore (Scheme 1). Consistent with this idea, the N terminus of PrP C contains a positively charged region (residues 23-31) that is reminiscent of "protein transduction domains" capable of penetrating the lipid bilayer and inducing ion channels (55,56). We have shown previously that these nine amino acids are essential for the neurotoxic and neuroprotective activities of PrP (57,58). We surmise that the N terminus of membrane-anchored wild-type PrP C also interacts with the lipid bilayer and that this may play an important role in the physiological activity of the protein.