|Home | About | Journals | Submit | Contact Us | Français|
Mammalian prion diseases involve conversion of normal prion protein, PrPC, to a pathological aggregated state (PrPres). The 3D structure of PrPres is not known but infrared (IR) spectroscopy has indicated high, strain-dependent beta sheet content. PrPres molecules usually contain a glycophosphatidylinositol (GPI) anchor and large Asn-linked glycans, which can also vary with strain. Using IR spectroscopy, we tested the conformational effects of these post-translational modifications by comparing wild type PrPres with GPI- and glycan-deficient PrPres produced in GPI-anchorless PrP transgenic mice. These analyses required the development of substantially improved purification protocols. Spectra of both types of PrPres revealed conformational differences between the 22L, ME7 and Chandler (RML) murine scrapie strains, most notably in bands attributed to beta sheets. These PrPres spectra were also distinct from those of the hamster 263K scrapie strain. Spectra of wild type and anchorless 22L PrPres were nearly indistinguishable. With ME7 PrPres, modest differences between the wild type and anchorless spectra were detected, notably a ~2-cm−1 shift in an apparent beta sheet band. Collectively, the data provide evidence that the glycans and anchor do not grossly affect the strain-specific secondary structures of PrPres, at least relative to the differences observed between strains, but can subtly affect turns and certain beta sheet components. Recently reported H/D exchange analyses of anchorless PrPres preparations strongly suggested the presence of strain-dependent, solvent-inaccessible beta core structures throughout most of the C-terminal half of PrPres molecules, with no remaining alpha helix. Our present IR data provide evidence that similar core structures also comprise wild type PrPres.
The structure of the infectious, pathological form of prion protein (PrPres or PrPSc)1 has been an enduring mystery for decades. During the course of transmissible spongiform encephalopathies (TSEs) or mammalian prion diseases, PrPres is formed from the host’s normal prion protein, PrPsen or PrPC, in a process that involves oligomerization and massive conformational change. PrPC is largely alpha helical and disordered (1-3), while PrPres is an ill-defined, often fibrillar multimer that is rich in beta sheet (1;4-6). Whereas the three-dimensional structures of different PrPC molecules have been determined by solution NMR (7) and X-ray crystallography (8), detailed structural analyses of PrPres have been confounded by its insoluble multimeric properties. Rough secondary structure compositions of PrPres have been estimated by infrared (IR) (1;4;5;9) and circular dichroism spectroscopic studies (6). IR (9-11) and a variety of biochemical measurements (12-20) have also indicated the existence of prion strain-dependent conformations of PrPres. Electron crystallography on two-dimensional arrays present in certain PrPSc preparations has given additional low-resolution maps of an apparent nonfibrillar PrPres ultrastructure (21;22). The strain-dependent dimensions of fibrillar PrPres ultrastructures have been determined by electron microscopy and atomic force microscopy (23).
During biosynthesis, PrPC is usually modified by the addition of up to two structurally diverse Asn-linked glycans (24;25) and a glycophosphatidylinositol (GPI) anchor (26). These post-translational modifications are often retained upon conversion of PrPC to PrPres. One of the well-documented strain-dependent variables in PrPres structures is the relative proportion of singly glycosylated, doubly glycosylated, and unglycosylated molecules (i.e., glycoforms) in PrPres aggregates from a given source (12;13;17;18;27). Moreover, studies with transgenic mice or cells expressing PrPC mutants lacking one or both of the Asn-linked glycosylation sites have shown striking strain-dependent effects on prion disease susceptibility and pathogenesis (28-30). Such observations suggest that glycans might modulate PrPres conformations, or that PrPres conformers might vary in their propensities, or abilities, to incorporate differentially glycosylated PrPC molecules. There is evidence of the latter possibility from studies of PrPres formation in vitro (27;31) and in scrapie-infected cells (32). On the other hand, another report has suggested that glycosylation may not be necessary for strain-specific differences in PrPres conformation, based on the location of proteinase K digestion sites (33). In any case, no higher resolution structural comparisons have been made to directly investigate the impact of glycosylation on PrPres conformation for a given strain.
Removing the GPI anchor of PrP can also have dramatic effects on prion disease pathogenesis. Transgenic mice expressing PrPC that lacks the C-terminal GPI anchor signal peptide (anchorless PrPC) can be infected intracerebrally with prions, but the resulting disease (called fatal transmissible amyloid encephalopathy) is more like a cerebral amyloid angiopathy than a typical TSE disease (34;35). Interestingly, the absence of GPI-anchored PrPC in these mice also nearly eliminates prion neuroinvasion and neural spread from more peripheral sites of inoculation (36). Anchorless PrPC that is expressed in transgenic mice and in cell cultures is deficient in Asn-linked glycans (34;37). The reason for this is unclear, but presumably reflects an important role for membrane anchoring for proper glycosylation during biosynthesis (38). Consequently, when infected with mouse scrapie prions, these transgenic mice produce PrPres that is both lacking the GPI anchor and severely underglycosylated.
The recent availability of anchorless, as well as wild type, PrPres now permits analyses of the influence of glycans and GPI anchors on PrPres structures. Because IR spectroscopy is capable of discerning prion strain-dependent variations in PrPres secondary structures (9-11;39), it is probably the best available method for detecting conformational effects of these post-translational modifications on PrPres strains. The amide I region (~1600-1700 cm−1) of protein IR spectra reflects vibrational absorbances by the amide linkages that comprise the polypeptide backbone. The frequencies of these absorbances are modulated by transition dipole couplings and by hydrogen bonds between backbone amide protons and carbonyl oxygens of different amino acid residues, which, in turn, are determined primarily by various secondary structures. One practical obstacle to comparisons of anchorless and wild type PrPres by IR has been difficulties in obtaining wild type murine scrapie PrPres preparations of sufficient purity. In the present study we describe new improved purification protocols for different strains of anchorless and wild type PrPres and their conformational comparisons by IR spectroscopy.
The term detergent-resistant membrane (DRM) PrPres refers to the use of brain DRMs as a starting subcellular fraction from which the PrPres is subsequently purified. Wild type PrPres was purified from Syrian Golden hamsters infected with the 263K scrapie strain or mice infected with the 22L or ME7 scrapie strains. Brains were dissected from animals exhibiting clinical scrapie signs, rinsed in PBS, blotted on filter paper to remove excess blood, snap frozen in liquid nitrogen, and stored at −80°C until needed. To isolate DRM PrPres, brains were first thawed and rinsed in cold CBS (20 mM citrate/137 mM NaCl, pH 6.0). A 10% brain homogenate was then prepared from ~2 g of tissue by Dounce homogenization in CBS. In certain preparations the homogenate was centrifuged at 1000 g for 5 min at 4°C to remove large insoluble material. The supernatant was then subjected to Brij-96 (Fluka) extraction as described below. This optional centrifugation step had no noticeable effect on the purity of the final product nor on the IR spectra of DRM PrPres (data not shown) but could reduce the yield of DRM PrPres depending on the strain. The reduced yield was due to unexpected localization of the majority of the PrPres to the 1000 g pellet fraction for certain strains such as 22L (data not shown). Therefore, we recommend either omitting this centrifugation step or analyzing the partitioning of PrPres and optimizing Brij-96 extraction conditions for strains not described here. The homogenate was rapidly and thoroughly mixed on ice with one volume of ice-cold 1% Brij-96 (in CBS) and incubated on a rotisserie at 4°C for 30 min. After Brij-96 extraction, the sample was adjusted to 26% OptiPrep (in 10 mM citrate/137 mM NaCl, pH 6.0) and split equally across 6 pre-chilled centrifuge tubes. This fraction was overlaid with layers of 23% (12 ml) and 8% (8 ml) OptiPrep (in 10 mM citrate/137 mM NaCl, pH 6.0) followed by 1 ml of 10 mM citrate/137 mM NaCl (pH 6.0). The gradients were centrifuged in a pre-chilled Beckman SW32 rotor at 18500 rpm for 2 hr at 4°C. Prominent lipid bands were visible at the CBS/8% OptiPrep and 8%/23% OptiPrep interfaces, the latter representing the PrPres-enriched fraction. The layers above the 8%/23% lipid band were carefully aspirated and discarded prior to careful collection of the lipid band at the 8/23% interface, which we refer to as the DRM fraction.
The second portion of the procedure involved extraction and purification of PrPres from the DRM fraction. After equilibrating the sample to room temperature (RT), a high salt extraction was performed on the DRM fraction by adding ½ volume of 5 M NaCl (in 0.15 M Tris, pH 8.0), mixing thoroughly, and incubating for 15 min at RT. The high salt-washed DRMs were split to 4 tubes and fractionated on a second floatation gradient consisting of a bottom fraction of salt-extracted DRMs adjusted to 26% OptiPrep (from a 60% stock) and overlaid with ~1/4 volume (~7 ml) of 23% OptiPrep in TBS (10 mM Tris/137 mM NaCl, pH 8.0) and 1 ml of TBS. The gradients were centrifuged in an SW32 rotor at 18500 rpm for 2 hr at 20°C. This high salt extraction was intended to disrupt electrostatic interactions between PrPres and high molecular weight molecules such as glycosaminoglycans and ferritin (40) and remove some of these contaminants from the DRMs. The lipid band at the TBS/23% OptiPrep interface was carefully collected and adjusted to a final volume of 27 ml with TBS. To solubilize the membranes and digest nucleic acids, 3 ml of 20% sarkosyl/0.5 M Tris (pH 7.5) was added followed by Benzonase to a final concentration of 25 U/ml and incubation at 37°C for 30 min. The sample was then digested with 10 μg/ml PK for 1 hr at 37°C. PK digestion was terminated by addition of Pefabloc SC (2 mM final concentration) and incubation on ice for 15 min. To remove PK, the sample was first adjusted to 30 mM EDTA and incubated for 5 min at RT. Next, the sample was centrifuged at 1000 g for 5 min at RT to remove large insoluble material. The supernatant was adjusted to contain a final concentration of 1.8 M NaCl and the entire mixture was gently overlaid onto ½ volume of a sucrose pad solution (1 M sucrose/0.1 M NaCl/0.5% sulfobetaine 3-14/10 mM Tris, pH 7.5). After centrifugation in a Beckman Type 50.2 Ti rotor at 45000 rpm for 2 hr at 20°C, the supernatant was carefully removed without disturbing the pellet. The pellet was washed by resuspension in 3 ml of 0.5% sulfobetaine (in PBS) followed by centrifugation in a TLA100.3 rotor at 70000 rpm for 30 min at 4°C. The supernatant was thoroughly removed and the final DRM PrPres pellet was resuspended in 0.5% sulfobetaine (in PBS) by pipetting and brief cuphorn sonication as needed. Yields of DRM PrPres ranged from ~2.5-12.5 μg/g of brain for 22L and ME7 strains and ~25-50 μg/g of brain for 263K with the lower yields corresponding to preparations with homogenates subjected to the optional 1000 g clearing step above.
Brain homogenates from scrapie-infected tg44+/+ transgenic mice (35) expressing GPI anchorless PrPC were adjusted to contain 50 mM Tris/137 mM NaCl/2% sarkosyl (pH 8.0) (TBSS) and incubated for 20 min at room temperature (RT). After Benzonase treatment (25 U/ml, 30 min) to digest nucleic acids, the samples were centrifuged (Beckman Type 50.2 Ti rotor, 24 min, ~40500 gmax) to pellet the PrPres and other detergent-insoluble material. These centrifugation conditions provided a k factor comparable to centrifugation at top speed in a Beckman microcentrifuge. The pellet was resuspended in TBSS carefully avoiding the brown portion of the pellet. The sample was then digested with proteinase K (PK) at 10 μg/ml for 1 hr. After termination of the PK digest, a PK-stripping procedure similar to that described previously (41) was performed by addition of EDTA (30 mM final concentration) and NaCl (1.7 M final concentration), layering onto a sucrose cushion (1 M sucrose, 100 mM NaCl/0.5% sulfobetaine 3-14/10 mM Tris, pH 7.4), and recovery of the PrPres by centrifugation as above through the sucrose cushion. The pellet was then resuspended in 0.5% sulfobetaine 3-14 (in PBS) and centrifuged again in order to wash the pellet. The final PrPres pellet was resuspended in a small volume of 0.5% sulfobetaine 3-14 (in PBS) with brief cuphorn sonication. The yields of anchorless PrPres were ~168 μg/g of brain for 22L, ~123 μg/g of brain for Chandler (RML), and ~75 μg/g of brain for ME7. Control “mock” preparations from age-matched uninfected mice contained a small pellet of translucent material that was resuspended as described for the anchorless PrPres pellets. A more detailed description of the purification procedure for anchorless PrPres can be found elsewhere (42).
Samples were denatured by boiling 8-10 min in sample buffer (4 M urea/5% SDS/5% glycerol/3 mM EDTA/0.02% bromophenol blue/62.5 mM Tris, pH 6.8) with either DTT (50 mM) or β-mercaptoethanol (4%) as a reducing agent. As needed, samples were cuphorn sonicated in sample buffer to improve solubilization of highly stable protein aggregates. Proteins were separated on 10% Bis-Tris NuPAGE gels in MES running buffer. GelCode Blue staining was performed following the manufacturer’s recommendations (Thermo Scientific). Silver staining was performed as described elsewhere (43). Purified mouse ferritin was purchased from Lee Biosolutions. The 90-231 fragment of hamster PrP (recPrPsen 90-231) (predicted MW 16243 Da) was purified from Escherichia coli as described elsewhere (44) and used as a standard for quantitation of PK-treated anchorless PrPres. For immunoblotting, proteins were transferred to PVDF membranes by a semi-dry method. For PrP immunoblots, prior to blocking the membranes were treated with either 0.2 M NaOH (30 min) or 3 M GdnSCN (10 min) followed by 4 rinses with TBS. Primary antibodies used for PrP detection included 6D11 (Covance Research Products) and R20 (rabbit polyclonal antiserum against PrP residues 218-232). Primary antibodies to detect ferritin included chicken anti-mouse ferritin (Immunology Consultants Laboratory) and rabbit anti-heavy chain human ferritin (Abcam #16875). Secondary antibodies were purchased from Jackson Immunoresearch (alkaline phosphatase conjugates) or Invitrogen (Alexa Fluor 488 conjugates). Alkaline phosphatase conjugates were detected using Attophos substrate (Promega). Alexa Fluor 488 conjugates were visualized by direct fluorescence. All blots were imaged on either a Storm or a Typhoon scanner (GE Healthcare).
Samples were spotted onto Formvar and carbon-coated grids from Ted Pella (Redding, CA). For samples in Figure 1, grids were freshly glow-discharged. After allowing 30-60 min for adsorption, the grids were washed extensively with ultrapure water. The samples were then stained with either 1% methylamine tungstate (Nano-W from Nanoprobes, Yaphank, NY) or 1% ammonium molybdate where indicated. After drying, the samples were imaged at 80 kV on a Hitachi H7500 electron microscope equipped with an AMT XR-100 digital camera.
IR spectra were obtained using a Perkin-Elmer Spectrum 100 instrument equipped with a diamond attenuated total reflectance sample unit and an MCT detector as described (42). Briefly, PrPres aggregates in 0.5% sulfobetaine in PBS containing between 6-20 μg of protein were pelleted, washed with water, and resuspended as a 1-2 μl slurry. The slurry was applied to the diamond and dried to a film over a few minutes while following the disappearance of the broad liquid water band at ~1636 cm−1 and the appearance of the protein amide I and amide II bands. The spectra were background-subtracted and comprised of 64 accumulations (4 cm−1 resolution; 1 cm/sec OPD velocity; strong apodization) Those shown were collected as soon as successively collected spectra (each taking 80 sec) stabilized, indicating little further evaporation of liquid water. This approach was taken to essentially eliminate spectral contributions from free liquid water without desiccating the protein film any more than necessary. Second derivative spectra were calculated using the instrument software and 13 data points.
PrPres accumulates in extraneural tissues in scrapie-infected tg44+/+ transgenic mice expressing anchorless PrP (45). This provided several potential sources of tissue to use for the purification of anchorless PrPres. To determine the optimal tissue for purification, we first used semi-quantitative immunoblotting to compare the levels of PrPres in a selection of tissues reported to have the highest amounts of PrPres (45). When samples were normalized for total protein, brain and brown fat contained the highest reproducible levels of PrPres (data not shown). Since there is more brain tissue available per animal than brown fat, brain contained the largest total amount of PrPres and was therefore chosen as the starting material.
The purification procedure used for anchorless PrPres preparations analyzed in the present study was based on methods previously described for purification of wild type (41) and anchorless (23) PrPres with some important modifications as detailed under Experimental Procedures. Anchorless PrPres was prepared from the brains of tg44+/+ mice infected with the Chandler (RML), 22L, or ME7 strains of mouse scrapie. Analysis of the samples by transmission electron microscopy (TEM) showed that the PrPres preparations were almost exclusively comprised of amyloid-like fibrils (Figure 1) as expected from previous studies (23;34), the exception being rare individual collagen fibers that were only detected with extensive searches of entire grids (data not shown). Collagen fibers were also present but infrequent in control “mock” PrPres preparations from age-matched, uninfected tg44+/+ mice (data not shown). Occasional clusters of electron dense particles were observed in mock and scrapie PrPres preparations. Because these were also present in buffer control samples they were attributed to buffer components such as sulfobetaine (Figure 1, “Buffer control”). Extensive searches of anchorless and mock anchorless PrPres samples failed to detect ferritin, a known contaminant of wild type PrPres preparations (41;46) readily distinguished by its characteristic “fish-eye” appearance (Figure 1). Together, these observations established that the anchorless PrPres preparations were highly enriched for amyloid-like fibrils.
Anchorless PrPres preparations were further subjected to biochemical analyses to assess sample purity. Analysis by SDS-PAGE with silver staining showed that all anchorless scrapie PrPres preparations were of high purity (Figure 2A, lanes 4-6). As previously reported (34), predominantly unglycosylated PrP was present accompanied by a lower level of monoglycosylated PrP (Figure 2A, lanes 4-6, bracket). Interestingly, anchorless ME7 PrPres exhibited a higher apparent molecular mass compared to Chandler and 22L, suggesting anchorless ME7 PrPres contains a larger PK-resistant core. This is consistent with data from mass spectrometric analysis (42). The bands present in anchorless PrPres preparations were specific to scrapie-infected animals as shown by analysis of control mock preparations (Figure 2A, lanes 10-11). Ladders of higher molecular mass species observed by silver staining correlated with bands detected by immunoblotting with two different anti-PrP antibodies (Figure 2B, lanes 7-9 and 11-13), showing that the bands detected by silver staining primarily consist of SDS-resistant oligomers of anchorless PrPres. In higher exposure images, low molecular mass bands were also visualized that were unique to each strain and antibody, thereby revealing further strain-specific differences between the various anchorless PrPres aggregates (Figure 2B, brackets, lower panels). By comparison with rPrPsen 90-231 standards, we estimated the typical yields of anchorless PrPres to be ~168 μg/g of brain for 22L, ~123 μg/g of brain for Chandler, and ~75 μg/g of brain for ME7.
Immunoblotting with an anti-heavy chain ferritin antibody was used to more sensitively assay for the presence of ferritin. Ferritin was readily detected in a standard wild type 22L PrPres preparation (Figure 2C, lane 6). However, even when overloading the gel with ~4-7-fold more anchorless PrPres (~1.2-2 μg) vs. wild type (300 ng), only a faint pair of bands were detected in two independent anchorless PrPres preparations (Figure 2C). Similar results were obtained using a different anti-ferritin antibody, chicken anti-mouse ferritin (data not shown). Given the bands precisely co-migrate with anchorless PrPres monomers and the signal is not competed in a control immunoblot using anti-ferritin antibody pre-adsorbed with purified ferritin (Figure 2C, compare lanes 4-5 with 10-11), these data show there is at most a negligible concentration of ferritin present in anchorless PrPres preparations. Altogether, these data establish the very high purity of the anchorless PrPres preparations and allow their use in IR spectroscopy studies.
To develop an improved protocol for purification of wild type PrPres, we investigated potential sources for molecules that are co-enriched with PrPres in our standard preparation method (41) due to properties shared with PrPres (i.e. detergent insolubility, protease-resistance, high density). Key insight was obtained from characterization of “mock” PrPres preparations derived from uninfected, wild type animals analogous to the mock preparations from anchorless mice described above. First, ferritin-like particles were detected by electron microscopy (Supporting Figure S1A) independent of whether preparations were conducted with or without an optional PK treatment (41) during the preparation. Amorphous objects were also present but less frequent in PK-treated preparations (data not shown). Ferritin contamination was confirmed in mock, 22L, and 263K PrPres preparations by immunoblotting with anti-ferritin antibody (Supporting Figure S1B). These data are consistent with independent studies (40;46). In addition to ferritin, other lower molecular mass proteins were detected in the standard mock PrPres by GelCode Blue staining (Supporting Figure S1C) and silver staining (see below) with similar apparent molecular mass to proteins in scrapie PrPres preparations that did not correspond to known PrP bands (see below). This indicated that several endogenous brain proteins can contribute to contamination of PrPres preparations based on their inherent similarity to the biochemical and biophysical properties of PrPres. This finding was supported by independent mass spectrometry studies (46).
This prompted us to identify biochemical features that might distinguish PrPres from the mock PrPres contaminants. We determined that the predominant mock PrPres proteins lacked N-glycosylation and GPI anchors by digestions with PNGase F and phosphatidylinositol-specific phospholipase C (data not shown). Since PrPres is GPI-anchored and associated with brain detergent-resistant membranes (DRMs) for multiple rodent scrapie strains (47-50) and (G. Baron, unpublished observations), this suggested that DRM fractionation could be used to separate PrPres from many contaminants in PrPres preparations and thus generate PrPres preparations of higher purity.
Many factors such as type of detergent, detergent concentration, extraction temperature, and detergent:protein ratio can all influence the association of proteins with DRMs (51;52). Therefore, we tested two different detergents (Triton X-100 and Brij-96) and various extraction conditions and gradient profiles to optimize enrichment of PrPres in DRMs. Although both detergents gave similar results, extraction with cold 0.5% Brij-96 produced slightly better results in small scale experiments (data not shown). As shown in Figure 3A, DRM fractionation of brains from 263K-affected hamsters and 22L-affected mice led to floatation of virtually all PrPres in a lipid band of intermediate density consistent with DRMs (fractions 5-6). The distribution of total PrP in such gradients, as determined by immunoblotting samples without PK digestion, was identical to that of PrPres (data not shown). A lower density lipid band corresponding to fractions 2-3 was also observed, but this did not contain detectable PrPres. This membrane fraction may in part be derived from myelin membranes, which can exhibit some detergent-resistance (53). BCA protein assay of fractions from DRM floatation gradients showed that a substantial portion of the total protein was localized to the higher density bottom fractions with a much smaller amount corresponding to PrPres-containing DRM fractions (Figure 3A, graph). These data established that DRM fractionation led to a significant enrichment of PrPres over other brain proteins and biomolecules.
Because DRM fractionation alone would be insufficient to purify PrPres, we designed a new purification procedure (described in Experimental Procedures) that incorporated DRM fractionation with elements of a standard PrPres preparation method (41). For simplicity, we refer to preparations using this new method as DRM PrPres. Silver stain SDS-PAGE analysis of DRM PrPres preparations corresponding to the 263K, 22L, and ME7 scrapie strains showed dramatically reduced levels of the low molecular mass contaminants present in standard PrPres preparations and standard mock PrPres preparations (Figure 3B, brackets and arrows). Interestingly, a pellet of material is obtained in mock DRM PrPres preparations. However, mock DRM PrPres preparations essentially lacked any detectable proteins as compared with equal brain equivalents of standard mock PrPres preparations (Figure 3B, compare lanes 3 vs. 4 and 7 vs. 8). Mock DRM PrPres preparations are not devoid of protein as proteins can be detected in these preparations by loading much higher (~5-10-fold more) amounts, but the banding profile still differs from those in standard mock PrPres preparations (data not shown). Although the molecular composition of the mock DRM PrPres fraction remains to be determined, our data (here and see below) suggest it may contain lipids. Immunoblot analysis of DRM PrPres preparations with an anti-PrP antibody showed a strong correspondence between immunoreactive bands and bands detected by silver staining, including those of both monomeric (arrows) and SDS-resistant oligomeric (brackets) PrP (Figure 3C). Ferritin contamination of DRM PrPres preparations was also dramatically reduced compared to standard PrPres preparations as shown in Figure 3D. Although ferritin was readily detectable by immunoblot analysis of standard 263K PrPres (arrow), it was undetectable in a similar amount of 263K DRM PrPres. Semi-quantitative analysis of ferritin content in 22L DRM PrPres using two different anti-ferritin antibodies and ferritin standard curves estimated ferritin levels to be ~ 4.1 or 6.4 ng ferritin/150 ng PrP (lane 6 in Figure 3E and Supporting Figure S2). By contrast, a standard 22L PrPres prep contained either 78 ng or 26 ng of ferritin/150 ng PrP depending upon the antibody used (lane 5 in Figure 3E and Supporting Figure S2). This corresponds to a 4.1 to 19-fold reduction in ferritin content. Since the DRM PrPres preparations were PK-treated, we conducted control experiments to confirm that, as expected, pre-treating ferritin with PK digestion had a negligible effect on ferritin detection (lanes 1-4 vs. 7-10 in Figure 3E and Supporting Figure S2). TEM analysis of DRM PrPres preparations showed they consisted of amyloid-like fibrils with rare collagen fibrils as seen with the anchorless PrPres samples (Figure 4). The amyloid-like fibrils were absent from control mock DRM PrPres preparations, which contained only rare collagen fibrils and vesicle-like structures and were detected only by loading 10 times the sample equivalents versus the 263K or 22L DRM PrPres (Figure 4). Collectively, these observations provide strong evidence that the DRM PrPres purification procedure was successful in generating sufficient yields of highly-purified PrPres well-suited for IR spectroscopy and other applications.
IR spectra of multiple (at least two) independent 22L, ME7 and Chandler anchorless PrPres preparations were compared to one another as well as to spectra of recombinant murine PrPsen 90-231 and mock anchorless PrPres preparations from uninfected tg44+/+ mice. The upper half of Figure 5 shows the primary IR spectra and the bottom shows second derivative spectra calculated directly from the primary spectra. The second derivatives give negative peaks corresponding to maxima in the primary spectra and provide an unbiased enhancement of band resolution. Consistent with previously reported IR spectra of hamster (4;9-11) and murine PrPres preparations (54), the most prominent absorptions were in a spectral region typically assigned to beta sheets, i.e. 1615-1636 cm−1, indicating predominant beta sheet content. This high beta sheet absorbance was in striking contrast to the predominant absorbance maxima at ~1657 and 1650 cm−1 of the rPrPsen 90-231. The latter maxima were consistent with PrPsen’s high alpha helical and disordered secondary structures (2) and previously reported IR data (1;55). As has been reported for wild type PrPres from hamsters (4;9-11) and mice (54), clear differences were observed between the different mouse strains of anchorless PrPres, most notably in the beta sheet region of the second derivative spectra. These results provide direct evidence that strain-dependent differences in PrPres secondary structure exist in the absence of GPI anchors and, for the most part, N-linked glycans.
To investigate whether glycans and anchors modify the secondary structures of PrPres, we also collected IR spectra of fully glycosylated and GPI-anchored wild type DRM PrPres of the 22L and ME7 strains prepared by the improved purification protocol described above. In addition, an analogous mock preparation from uninfected mouse brain tissue was also analyzed. As shown in Figure 6, the mock preparation gave a spectrum in the 1600-1700 cm−1 region that was considerably less intense than the corresponding PrPres spectra. Moreover, the mock spectrum was broad and relatively featureless, and when subtracted from the PrPres spectra (ME7-mock), had little impact on the features of their second derivative spectra. Another potential complication in comparing the IR spectra of the wildtype and anchorless PrPres preparations is that N-acetyl glucosamine moieties in N-linked glycans and GPI anchors can also absorb near the beta sheet region (~1638 cm−1) (4). However, their overall contributions to PrPres spectra in this region should be minor (4). Indeed, the spectrum of the wild type 22L PrPres was nearly indistinguishable from that of the anchorless 22L PrPres, except for modestly enhanced absorbance centered at ~1670 cm−1(Figure 7). The latter band is in the spectral region commonly ascribed to turn structures (39). In the case of ME7 PrPres, differences between the wild type and anchorless structures were subtle, but more apparent than in the 22L PrPres preparations. Most notable was a shift in the low-frequency beta sheet band from ~1620 cm−1 to ~1622 cm−1 with the post-translational modifications. Altogether the IR spectra of the anchorless and wild type PrPres provide evidence that the N-linked glycans and GPI anchor do not grossly affect the strain-specific secondary structures of PrPres, but may have subtle effects on turn (or perhaps loop structures) and, in the case of ME7, certain beta sheet components.
The present work provides the first direct evaluation of the effects of the GPI anchor and Asn-linked glycans on PrPres conformation. Two recent developments have made these analyses possible: first, the recent availability of scrapie-infected transgenic mice expressing only anchorless PrP molecules which are also severely deficient in Asn-linked glycans (e.g. see predominance of lower, unglycosylated isoform in Figure 2A and 2B), and second, our new purification protocols for anchorless and wild type PrPres. These protocols greatly reduce the presence of contaminants that, in hindsight, must have confounded previous IR analyses of wild type murine PrPres (54) which, notably in the case of 22L PrPres, gave spectra that are significantly different than those reported here. Contamination has been much less of a problem with certain hamster PrPres preparations because of their much higher starting PrPres content per mass of brain tissue. Nevertheless, improvements were observed when the DRM PrPres method was applied to hamster 263K PrPres as shown by the reduction in ferritin contamination (Figure 3B and 3D).
Our rationale for purifying wild type PrPres starting with DRM membrane preparations rather than fully detergent-solubilized whole brain homogenates stemmed from the discovery that many contaminants present in preparations made using a standard method (41) seemed to correspond to endogenous brain proteins in mock PrPres preparations with biochemical and biophysical properties similar to PrPres. These properties included detergent-insolubility, protease-resistance, and high density. Working independently, another group has identified some of the proteins present in both mock and wild type mouse PrPres samples (46). We reasoned that some of these detergent-insoluble, high density contaminants might be separated from PrPres by floatation of PrPres-enriched DRMs on density gradients. To minimize co-enrichment for certain high molecular mass contaminants such as ferritin due to their direct association with PrPres via electrostatic interactions (40), we performed a high salt treatment of the DRM fraction. This should disrupt these interactions and remove these contaminants by recovery of the salt-washed DRMs on a second floatation gradient. The high salt treatment is a component of a standard PrPres purification procedure that is thought to remove some glycosaminoglycans (41). However, since this treatment is applied to a sarkosyl-extracted brain homogenate, high density molecules like ferritin that might become dissociated from PrPres would still co-sediment with the PrPres in the subsequent ultracentrifugation step. Because we have found that wild type PrPres of many scrapie strains associates with the DRM fraction, including strains prone to generating large plaque-like PrPres deposits (e.g. 87V, data not shown), it is likely our DRM PrPres purification method is broadly applicable to purification of wild type GPI-anchored PrPres.
Consistent with our rationale, at most very low levels of endogenous brain proteins were enriched in mock anchorless PrPres preparations (Figure 2A). The accumulation of anchorless PrPres in large amyloid deposits permitted the recovery of anchorless PrPres at substantially lower centrifugation speeds than are used for wild type PrPres. This could reduce the opportunity for co-purification of high molecular mass, dense, protease-resistant contaminants (e.g. ferritin). A further modification applied to both the new wild type and anchorless PrPres preparation procedures was the use of Benzonase to digest nucleic acids in place of sequential RNase A and DNase digestions. This avoided the use of RNase A which is partially resistant to PK digestion (56), forms aggregates under certain conditions (57;58), and thus could be a source of contamination.
Another group working in parallel has reported the use of DRM fractionation to enrich for PrPres with the goal of generating antigen for immunizations to create anti-PrP antisera (50). Although these authors used DRM fractionation in the initial purification step, subsequent steps to further enrich for PrPres were dramatically different from our method. Most importantly, the apparent purity of the PrPres from this method was substantially lower than that shown in the present study as indicated by the presence of prominent silver-stained bands that were not reactive with anti-PrP antibodies (50). Moreover, only portions of the gels corresponding to a narrow apparent molecular mass range from ~20-50 kDa are shown (50). Hnasko et al. (2010) argue that the lack of induction of antibodies against other brain DRM proteins besides PrP provides evidence of the purity of their PrPres preparations. However, this may simply represent murine immunological tolerance to these non-PrP rodent proteins just as they observed for PrP itself (i.e. no anti-hamster PrP antibodies were generated in immunizations of wild type mice). Although the PrPres preparations of Hnasko et al. are useful for immunization experiments, they would not be of sufficient purity for IR analyses presented here.
Our study clearly reinforces previous reports of the ability of IR spectroscopy to detect differences in conformation between PrPC and PrPres, as well as between different PrPres strains comprised of the same polypeptide precursor (9-11;39). Relative to those differences, the spectral effects of glycans and GPI anchors on PrPres were modest (Figures (Figures55--7).7). Clearly differences in beta sheet and apparent turn structures can be resolved between PrPres strains of the same amino acid sequence, but there is always potential for spectral overlaps in IR amide I spectra of different polypeptide secondary structures. Thus, an inability to distinguish IR spectra does not guarantee the presence of identical secondary structures, and, subtle differences in secondary structures, i.e., involving only a few residues, might also be difficult to detect. Nonetheless, the most straightforward interpretation of the present IR data is that glycans and anchors do not cause gross conformational changes that are commensurate with the differences between PrPres strains, especially in the case of 22L PrPres. A more noticeable effect of these post-translational modifications was seen with the ME7 strain, but much of the spectrum remained unchanged.
Given the apparently tight packing of PrP molecules in PrPres fibrils, as evidenced by their resistance to proteinase K digestion and H/D exchange (42), one can imagine that the presence of bulky post-translational modifications might not only affect monomer conformations, but also constrain the options for packing PrP monomers into ordered multimers and amyloid fibrils. The present data suggest that the presence of glycans and GPI anchor do not cause a major restructuring of the underlying polypeptide backbone secondary structures within the 22L and ME7 PrPres fibrils. However, the data do not negate the possibility that tertiary and quaternary structures might be affected significantly by glycans and anchors, or that the effects might be more significant with other prion strains.
Recent analyses of the rates of H/D exchange of the polypeptide backbone amide protons in infectious anchorless PrPres fibrils has revealed extreme resistance to exchange throughout the entire proteinase K-resistant C-terminal portions of the PrP subunits (i.e., from residues ~80-90 to ~223) (42). These results strongly suggest that, contrary to previous IR-based reports (1;4;5;9), there is virtually no alpha helix in PrPres. The previous reports employed different methods of estimating secondary structure compositions from IR spectra which were based on theoretical considerations and empirical calibrations using globular proteins of known structure. However, although such estimates are usually fairly accurate [e.g., (59)], they may at times be misleading due to ambiguities in band assignments or potential IR distortions due to poorly understood factors such as nonglobularity and/or oligomerization of particular protein samples (60). In any case, the fact that anchorless PrPres appears to lack any of the alpha helices that are present in PrPC raises the issue of whether wild type PrPres, with its large hydrophilic glycans attached to Asn residues that are located on, or between, residues comprising helices 2 and 3 in PrPC, also lacks those helices. Unfortunately, in the previous studies, the presence of the glycans prevented direct mass spectrometric analysis of H/D exchange in much of the proteinase K-resistant C-terminal region of wild type PrPres (42). However, the close similarity of the IR spectra of anchorless and fully glycosylated wild type PrPres that we have observed here strongly suggests that they have similar structures. Moreover, the spectra of the wild type PrPres preparations gave no evidence for enhanced alpha helical content which, in PrPC and many other proteins, absorb between 1650-1657 cm−1 (Figure 7). Indeed little, if any, banding in this latter spectral region was observed above the spectral noise for any of the PrPres preparations. The adjacent band at 1659 cm−1 was prominent in both anchorless PrPres preparations, which we have concluded has no alpha helices ; thus the 1659 cm−1 band is likely to be due to something other than alpha helix, such as turns. The prominence of the same band in wildtype PrPres, and the relative lack of bands in the 1650-1657 cm−1 region lead us to surmise that, like their anchorless counterparts, wild type 22L and ME7 PrPres fibrils have structures comprised predominantly of an extended solvent-inaccessible beta-sheeted core with occasional turns, but virtually no alpha helix.
Supporting Figure S1: Ferritin is enriched in standard mock PrPres preparations. (A) Apparent ferritin particles (arrows) observed by TEM in hamster mock PrPres preparations with (PK-treated) or without (Untreated) incorporation of a PK digestion step. Samples were stained with ammonium molybdate. Bar, 100 nm. (B) Immunoblot detection of ferritin in standard preparations of wild type mouse 22L, hamster 263K, and hamster mock PrPres with anti-heavy chain ferritin antibody. Approximately 300 ng of PrPres (22L, 263K) or comparable equivalents of hamster mock preparations were loaded per lane. Examples of preparations without (−) and with (+) incorporation of a PK digestion step are shown. Monomeric and SDS-resistant oligomers of ferritin are visible (arrows). (C) GelCode Blue-stained SDS-PAGE gel of standard prep of hamster mock PrPres. Note the primary effect of incorporating PK digestion into the purification protocol is to shift the overall banding profile (compare lanes 2 and 4). Lanes 1 and 3 show trypsin-digested samples after first denaturing the mock PrPres by boiling in 2% SDS/2 mM DTT followed by SDS neutralization with 1% NP-40. This formally demonstrates the bands in lanes 2 and 4 are proteins. Arrow indicates band corresponding to trypsin.
Supporting Figure S2: Immunoblot comparison of ferritin in standard (std) and DRM 22L wild type PrPres preparations using chicken anti-mouse ferritin antibody. Dilution series of purified mouse ferritin without (lanes 1-4) or with PK treatment (7-10) were loaded as indicated. Lanes 5 and 6 contain 150 ng of either std or DRM 22L PrPres, respectively. Arrow, ferritin monomer. SDS-resistant ferritin oligomers are also visible.
Animal experiments were performed in accordance with animal welfare guidelines under an animal study protocol approved by the Animal Care and Use Committee of the Rocky Mountain Laboratories, National Institute of Allergy and Infectious Diseases, National Institutes of Health. The authors thank Mikael Klingeborn, Roger Moore, and Lara Taubner for critical reading of the manuscript.
†This research was funded by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health.
SUPPORTING INFORMATION AVAILABLE Supporting Figures 1 and 2 as mentioned in the text. This material is available free of charge via the Internet at http://pubs.acs.org.