Sialylation of Glycosylphosphatidylinositol (GPI) Anchors of Mammalian Prions Is Regulated in a Host-, Tissue-, and Cell-specific Manner*

Prions or PrPSc are proteinaceous infectious agents that consist of misfolded, self-replicating states of the prion protein or PrPC. PrPC is posttranslationally modified with N-linked glycans and a sialylated glycosylphosphatidylinositol (GPI) anchor. Conformational conversion of PrPC gives rise to glycosylated and GPI-anchored PrPSc. The question of the sialylation status of GPIs within PrPSc has been controversial. Previous studies that examined scrapie brains reported that both sialo- and asialo-GPIs were present in PrPSc, with the majority being asialo-GPIs. In contrast, recent work that employed cultured cells claimed that only PrPC with sialylo-GPIs could be recruited into PrPSc, whereas PrPC with asialo-GPIs inhibited conversion. To resolve this controversy, we analyzed the sialylation status of GPIs within PrPSc generated in the brain, spleen, or cultured N2a or C2C12 myotube cells. We found that recruiting PrPC with both sialo- and asialo-GPIs is a common feature of PrPSc. The mixtures of sialo- and asialo-GPIs were observed in PrPSc universally regardless of prion strain as well as host, tissue, or type of cells that produced PrPSc. Remarkably, the proportion of sialo- versus asialo-GPIs was found to be controlled by host, tissue, and cell type but not prion strain. In summary, this study found no strain-specific preferences for selecting PrPC with sialo- versus asialo-GPIs. Instead, this work suggests that the sialylation status of GPIs within PrPSc is regulated in a cell-, tissue-, or host-specific manner and is likely to be determined by the specifics of GPI biosynthesis.

Prions or PrP Sc are proteinaceous infectious agents that consist of misfolded, self-replicating states of the prion protein or PrP C . PrP C is posttranslationally modified with N-linked glycans and a sialylated glycosylphosphatidylinositol (GPI) anchor. Conformational conversion of PrP C gives rise to glycosylated and GPI-anchored PrP Sc . The question of the sialylation status of GPIs within PrP Sc has been controversial. Previous studies that examined scrapie brains reported that both sialo-and asialo-GPIs were present in PrP Sc , with the majority being asialo-GPIs. In contrast, recent work that employed cultured cells claimed that only PrP C with sialylo-GPIs could be recruited into PrP Sc , whereas PrP C with asialo-GPIs inhibited conversion. To resolve this controversy, we analyzed the sialylation status of GPIs within PrP Sc generated in the brain, spleen, or cultured N2a or C2C12 myotube cells. We found that recruiting PrP C with both sialo-and asialo-GPIs is a common feature of PrP Sc . The mixtures of sialo-and asialo-GPIs were observed in PrP Sc universally regardless of prion strain as well as host, tissue, or type of cells that produced PrP Sc . Remarkably, the proportion of sialoversus asialo-GPIs was found to be controlled by host, tissue, and cell type but not prion strain. In summary, this study found no strain-specific preferences for selecting PrP C with sialoversus asialo-GPIs. Instead, this work suggests that the sialylation status of GPIs within PrP Sc is regulated in a cell-, tissue-, or host-specific manner and is likely to be determined by the specifics of GPI biosynthesis.
Prions or PrP Sc2 are proteinaceous infectious agents that consist of misfolded, self-replicating states of a sialoglycoprotein called the prion protein or PrP C (1,2). Prions replicate by recruiting and converting PrP C molecules expressed by a host into misfolded PrP Sc states (3). In the PrP Sc state, the prion protein can acquire conformationally distinct self-replicating states, referred to as prion strains, that elicit different, strainspecific disease phenotypes (4 -10). Accumulation of PrP Sc in the CNS leads to prion diseases, a family of transmissible neurodegenerative maladies that are 100% lethal (11).
PrP C is posttranslationally modified with two N-linked glycans and a glycosylphosphatidylinositol (GPI) anchor that is attached to the C-terminal residue Ser-230 (the residue number is provided for mouse PrP C ) (12)(13)(14)(15)(16). Upon conversion of PrP C into PrP Sc , the N-linked glycans and GPI are carried over, giving rise to glycosylated and GPI-anchored PrP Sc (17,18). Like amyloids formed by other amyloidogenic proteins or peptides, PrP Sc displays a cross-␤ folding pattern (19,20), a key structural feature of amyloid states. However, unlike other amyloidogenic proteins linked to neurodegenerative diseases, PrP Sc recruits PrP C anchored via the GPI to the cell surface and replicates its cross-␤ structure while also being attached to the membrane via its GPI anchors (21)(22)(23)(24). For that reason, the mechanism of PrP Sc replication appears to be unique.
Multiple lines of evidence indicate that the GPI anchor plays an important role in PrP Sc transmission and toxicity (reviewed in Ref. 25). Using the GPI anchor, PrP C could be efficiently transported from cell to cell (26). GPI anchoring of PrP Sc was found to be important for its binding to and replication on the cell surface (27) and might be also involved in intercellular transmission of PrP Sc via exosomes (28). Upon inoculation with prions, transgenic mice expressing GPI anchorless PrP C supported prion replication and accumulated prion infectivity; however, the disease onset in these mice was substantially delayed or was lacking (29). Mice that overexpressed GPI anchorless PrP C developed neurological dysfunction and generated infectious prions spontaneously in the absence of exposure to prions (30). Previous studies suggested that neurotoxic signaling is dependent on PrP C attachment to the plasma membrane via its GPI anchor (31,32). In particular, experiments using primary neuronal cultures suggested that toxicity triggered by PrP Sc is dependent on the sialylation status of the GPI anchor within PrP C , as clustering of PrP C molecules with sialylated GPIs led to activation of cytoplasmic phospholipase A2 and synapse damage (31).
PrP C is one of very few proteins in which GPI anchors are sialylated (33,34). In fact, previous studies revealed that GPIs of PrP C are structurally heterogeneous, with a subfraction of GPIs modified with sialic acids (12). GPIs are synthesized in the endoplasmic reticulum, and then posttranslational modification of polypeptide chains with preformed GPIs occurs in both the endoplasmic reticulum and Golgi (34,35). Because sialyltransferases are localized in the Golgi (36), sialylation of GPIs presumably occurs in the Golgi after their attachment to proteins.
Although the biological function of sialylation of GPIs is not known, learning whether PrP C with both sialo-and asialo-GPIs can be recruited into PrP Sc is important. If prion replication is supported only by PrP C with either but not both sialo-or asialo-GPIs, then new strategies for therapeutic intervention against prion disease that would control GPI sialylation could be considered. The question of the sialylation status of GPIs within PrP Sc has been controversial. The classic study by Stahl et al. (12) reported that GPIs within PrP Sc consisted of six structural isoforms, ϳ30% of which were sialylated. Moreover, the composition of GPIs within PrP Sc was found to be similar to that of PrP C (12). In contrast, a recent work by Bate et al. (37) claimed that only PrP C with sialylated GPIs could be recruited into PrP Sc , whereas PrP C with asialo-GPIs inhibits conversion of PrP C with sialo-GPIs into PrP Sc .
To resolve the controversy, we analyzed the sialylation status of GPIs within PrP Sc of five prion strains from two hosts. In addition, we also assessed the sialylation status of GPIs of brainand spleen-derived PrP Sc as well as PrP Sc generated in neuroblastoma N2a cells or terminally differentiated C2C12 myotube cells cultured in vitro. This study reports that PrP Sc can recruit PrP C with both sialo-and asialo-GPIs, as the mixtures of sialoand asialo-GPIs were observed universally regardless of prion strain or host, tissue, or cell type in which PrP Sc was generated. However, the proportion of sialo forms of GPIs was variable and depended on the host, tissue, and cell type but not prion strain. In summary, these findings suggest that the proportion of the two sialo forms of GPIs within PrP Sc is determined by cell-or tissue-specific rates of synthesis of sialo-GPIs but not strainspecific structural constraints.

GPI Anchors of Brain-derived Mouse PrP Sc Consist of Sialylated and Asialylated
Forms-To assess the sialylation status of GPI anchors, PrP Sc was denatured into monomers, and then individual PrP molecules were analyzed using two-dimensional gel electrophoresis followed by Western blotting. In the vertical dimension of two-dimensional blots, PrP molecules are separated according to their glycosylation status. Diglycosylated PrPs are of the highest and unglycosylated PrPs of the lowest molecular weight. In the horizontal dimension of two-dimensional blots that separates molecules according to their pI, all three PrP glycoforms display multiple charge isoforms. For diand monoglycosylated molecules, multiple charge isoforms are largely attributed to heterogeneity in the sialylation status of N-linked glycans and, to a lesser degree, heterogeneity of the GPI anchors (38,39). However, in unglycosylated molecules, the multiple charge isoforms report on the structural heterogeneity of the GPI anchors, including their sialylation status (12). Because the sialylation status of N-linked glycans was analyzed in previous studies (39,40), this work focuses on the GPI anchors.
To assess the sialylation status of GPI anchors, we analyzed the charge distribution of unglycosylated molecules using twodimensional blots. In brain material from mice infected with prion strain 22L, unglycosylated PrP showed three major spots (referred to as 1, 2, and 3; Fig. 1, A and B). To find out whether any of these spots could be attributed to sialo-GPIs, 22L brain material was treated with Arthrobacter ureafaciens sialidase, which exhibits a broad specificity with respect to sialylation FIGURE 1. Two-dimensional analysis of the sialylation status of brain-and spleen-derived PrP Sc from 22L-infected animals. A, representative twodimensional Western blots of 22L brain homogenate (BH), 22L BH treated with A. ureafaciens sialidase (BH ϩAU), 22L spleen homogenate (SH), and 22L SH treated with A. ureafaciens sialidase (SH ϩ AU). Black and white arrowheads mark diglycosylated and monoglycosylated glycoforms, respectively, whereas arrows mark the unglycosylated form. The homogenates were treated with PK, and Western blots were stained with Ab3531 antibody. B, intensity profiles of unglycosylated isoforms of brain-derived (top plot, solid thick lines) or spleen-derived PrP Sc (bottom plot, solid thick lines) from animals infected with 22L (n ϭ 3 independent animals). Intensity profiles of unglycosylated isoforms of 22L brain or spleen homogenates treated with A. ureafaciens sialidase are provided as a reference (dotted lines). Three major spots corresponding to unglycosylated isoforms are marked as 1, 2, and 3. Profiles were built as described under "Experimental Procedures" based on two-dimensional Western blots.
linkages. Cleavage of negatively charged sialic acid residues is expected to shift the distribution of charge isoforms toward basic pH (to the right in two-dimensional blots). Indeed, upon sialidase treatment, the relative intensity of spot 1 increased at the expense of spots 2 and 3 (Fig. 1A). The relative presentation of sialo-GPIs versus asialo-GPIs is difficult to estimate because the yields from cleavage of sialic residues from GPI anchors by sialidases are not known. As judged from the substantial shift of di-and monodiglycosylated PrP molecules toward basic pI, the treatment with sialidase was very effective yet less than 100%. Nevertheless, these shifts argue that spot 2 consists of unglycosylated PrPs with both sialo-and asialo-GPIs, whereas spot 3 consists predominantly of PrPs with sialylated GPIs. The lack of appearance of any new spots to the right of spot 1 upon sialidase treatment argues that spot 1 consists only of PrPs with asialo-GPIs. These results demonstrated that brain-derived 22L scrapie consisted of PrPs with both sialo-and asialo-GPIs.
Tissue-specific Differences in Sialylation Status of GPIs in PrP Sc -We showed previously that the N-linked glycans are more sialylated in spleen-derived than brain-derived PrP Sc (40). To investigate whether the sialylation status of GPIs in PrP Sc is also tissue-specific, we compared brain-and spleen-derived PrP Sc from animals infected with 22L. The profiles of unglycosylated charge isoforms were markedly different in spleen versus brain (Fig. 1, A and B). In contrast to brain-derived PrP Sc , the charge isoforms corresponding to asialo-GPIs (spot 1) was predominant in spleen-derived PrP Sc (Fig. 1, A and B). The isoforms corresponding to sialo-GPIs (spot 3) were low populated in spleen-derived PrP Sc (Fig. 1, A and B). Upon treatment with sialidase of spleen-derived PrP Sc , the relative intensities of both spots 2 and 3, which contain sialo-GPIs, decreased; however, to a lower degree than the decrease observed upon treatment of brain-derived PrP Sc . Together, these results show that, although spleen-derived 22L contained sialo-GPIs, the propor-tion of asialo-GPIs versus sialo-GPIs in spleen was substantially higher than that in brain-derived PrP Sc .
The Sialylation Profile of GPIs of Brain-derived PrP Sc Is Strain-independent-Previously, we showed that the sialylation status of N-linked glycans on PrP Sc are strain-specific, illustrating that prion strains recruit PrP C sialoglycoforms selectively according to their strain-specific structural constraints (39). To test whether the sialylation status of GPIs is strain-specific, brain-derived scrapie material for three mouse prion strains 22L, RML, and ME7, were analyzed ( Fig. 2A). As judged from two-dimensional blots and corresponding density profiles, the distribution of charge isoforms of non-glycosylated PrPs were very similar for all three strains (Fig. 2, A and B). In contrast, significant differences in sialylation patterns of N-linked glycans were observed for these three strains, as judged from the profiles of mono-and diglycosylated isoforms. In full agreement with previous studies (39,40), the highest level of sialylation of N-glycans was observed for ME7 and the lowest for RML. These results argue that, although sialoglycoforms of PrP C are recruited into PrP Sc selectively according to strainspecific structural constraints, the sialylation status of GPI is not a subject of strain-specific selection. These results also suggest that the sialylation status of GPIs of PrP Sc simply reflects that of PrP C , as has been shown previously by Stahl et al. (12).
Species-specific Variations in Sialylation Status of GPIs in PrP Sc -To determine whether the sialylation status of GPIs in PrP Sc is host-dependent, unglycosylated isoforms of brain-derived PrP Sc of two hamster strains, HY and 263K, were analyzed by two-dimensional blots (Fig. 3A). The profiles of unglycosylated charge isoforms were very similar for both hamster strains but different from those of mouse strains. In hamster strains, the charge isoforms corresponding to asialo-GPIs (spot 1) were predominant, whereas the isoforms corresponding to sialo-GPIs (spot 3) were the least populated (Fig. 3B). All three mouse FIGURE 2. Two-dimensional analysis of the sialylation status of brain-derived PrP Sc from animals infected with RML, ME7, or 22L. A, representative two-dimensional Western blots of BH from animals infected with RML, ME7, or 22L. Black and white arrowheads mark diglycosylated and monoglycosylated glycoforms, respectively, whereas arrows mark the unglycosylated form. The homogenates were treated with PK, and Western blots were stained with Ab3531 antibody. B, intensity profiles of unglycosylated isoforms of brain-derived PrP Sc from animals infected with RML (top plot) or ME7 (bottom plot) (n ϭ 3 independent animals). Three major spots corresponding to unglycosylated isoforms are marked as 1, 2, and 3. Profiles were built as described under "Experimental Procedures" based on two-dimensional Western blots.
strains showed clear prevalence of spot 2 (Figs. 3B and 2B). Interestingly, after treatment with sialidase, the profiles of mouse and hamster strains looked similar (Fig. 3B), supporting the claim that the species-specific differences in GPIs were due to different ratios of sialoversus asialo-GPI forms. These results argue that the sialylation status of GPIs was largely determined by the species or host, whereas strain-specific variations were very minor.
The Sialylation Status of GPIs in PrP Sc Depends on Cell Type-To test whether the sialylation status of GPIs depends on cell type, we analyzed unglycosylated charge isoforms of PrP Sc generated in N2a and myotube C2C12 cells, both infected with the RML strain. N2a is neuroblastoma cell line that can be stably infected with mouse strains, including RML (41). Myotubes were also susceptible to the RML strain when generated via differentiating C2C12 myoblast cells (42). The two cell lines displayed significantly different profiles of unglycosylated charge isoforms, suggesting that a cell-specific difference in the sialylation status of GPIs exists. In N2a, spot 1 was clearly predominant, suggesting that the majority of N2a-generated PrP Sc had asialo-GPIs (Fig. 4A). Indeed, upon treatment of N2a-generated PrP Sc with sialidase, the profile for unglycosylated isoforms did not change, supporting the claim that PrP Sc had predominantly asialo-GPIs (Fig. 4, A and B). Remarkably, the dramatic shift of di-and monoglycosylated isoforms toward basic pH upon treatment with sialidase is attributed to a removal of sialic acids from N-linked glycans and confirms that this enzymatic procedure for desialylation was highly effective (Fig. 4A). Meanwhile, in myotubes examined on day 15 postinfection, the unglycosylated charge isoforms showed at least five major spots (Fig. 4A). In addition to spots 1-3, at least two more spots (designated A and B, Fig. 4A) that were not readily observed in other samples were observed in myotube-derived material. Sialidase treatment of myotube-produced PrP Sc revealed that the unusually broad charge heterogeneity of GPIs was attributable to sialylation (Fig. 4A). Notably, both the N2aand myotube-specific sialylaton profiles of GPIs for RML were different from that of brain-derived RML (Fig. 4A). In summary, these results argue that the sialylation status of GPIs within PrP Sc is controlled by cell type but not strain.
Time-dependent Changes in the Sialylation Status of GPIs in PrP Sc in Differentiating Myotubes-To test whether the status of GPIs in PrP Sc changes over the course of PrP Sc formation, we first examined the kinetics of PrP Sc accumulation in myotubes. During the first week after infection, the amounts of PrP Sc dropped considerably, reflecting clearance of PrP Sc administered to the cells (Fig. 5A), in agreement with previous studies (42). In fact, on day 7 post-infection, PrP Sc was barely detectible by Western blot (Fig. 5A). However, during the second week post-infection, the total amount of PrP Sc increased steadily, FIGURE 3. Two-dimensional analysis of the sialylation status of brain-derived PrP Sc from Syrian hamsters inoculated with HY or 263K. A, representative two-dimensional Western blots of HY BH, HY BH treated with A. ureafaciens sialidase (BH ϩAU), or 263K BH. A Western blot of 22L BH is provided as a reference. Black and white arrowheads mark diglycosylated and monoglycosylated glycoforms, respectively, whereas arrows mark the unglycosylated form. The homogenates were treated with PK, and Western blots were stained with 3F4 antibody for 263K and HY and Ab3531 antibody for 22L. B, intensity profiles of unglycosylated isoforms of brain-derived PrP Sc from animals infected with HY (top plot, solid thick lines), 263K (center plot), or 22L (bottom plot) (n ϭ 3 independent animals). Intensity profiles of unglycosylated isoforms of HY or 22L brain homogenates treated with A. ureafaciens sialidase are provided as a reference (dotted lines). Three major spots corresponding to unglycosylated isoforms are marked as 1, 2, and 3. Profiles were built as described under "Experimental Procedures" based on two-dimensional Western blots.
reflecting PrP Sc newly produced by myotubes (Fig. 5A). Twodimensional analysis of cells collected on days 10 and 15 postinfection revealed significantly different profiles of unglycosylated charge isoforms (Fig. 5, B and C). On day 10, three main spots (1, 2, and 3) were clearly visible, with spot 2 being predominant. On day 15, the profile becomes more complex, with the relative contribution of spots 3, A, and B increased at the expense of spot 2 (Fig. 5, B and C). This experiment illustrates that although sialo-and asialo-GPIs are present in PrP Sc at different stages of PrP Sc accumulation and cell differentiation, a shift toward sialo-GPIs could be observed over time. A more complex composition of GPI isoforms was observed with the progression of cell differentiation.
The Sialylation Profiles of GPIs of PrP C Proteolytic Fragment C1 Resemble Those of PrP Sc -We found previously that PrP C charge isoforms were poorly resolved on two-dimensional blots, presumably because of peculiar properties of the octarepeat region (43). Multiple attempts to improve the resolution using different strategies have not been successful. However, sialoglycoforms of the C-terminal proteolytic fragment C1, which is produced by cleavage of PrP C at amino acid residue 111, were well resolved on two-dimensional blots (43). In the absence of a reliable method for resolving full-length PrP C isoforms, we used C1 as a reporter for assessing the sialylation status of the GPIs. In C1 fragments, multiple positively charged amino acid residues within residues 99 -110 are lacking in comparison with PK-treated PrP Sc , which consists of residues ϳ90 -230. As a result, the C1 charge isoforms are shifted toward acidic pH on two-dimensional blots relative to those of PK-treated PrP Sc (compare Fig. 6A to Figs. 1 and 2). Two-dimensional analysis of mouse brain-derived C1 revealed three spots for unglycosylated isoforms, with spots 1 and 2 being predominant (Fig. 6, A and B). This profile resembled those of brain-derived PrP Sc , in which spot 2 was predominant. In N2aderived C1, the profiles for unglycosylated isoforms showed one main spot (1) and was very similar to the profiles seen for N2a-derived PrP Sc (compare Figs. 6, A and B, and 4, A and B). In summary, although the sialylation profiles for GPIs of C1 and PrP Sc were not identical, N2a-derived PrP Sc and C1 were considerably less sialylated relative to the brain-derived PrP Sc and C1, respectively, reflecting the same ranking order.

Discussion
GPI anchors represent a structurally diverse class of posttranslational modifications that anchor proteins in the outer leaflet of the cellular membrane and are involved in a number of functions, including protein trafficking into lipid rafts, signal transduction, cellular communication, targeting of protein to apical membranes, and others (35,44). It is believed that GPIcontaining proteins are sorted toward glycosphingolipid-and cholesterol-rich rafts, a sorting that already occurs in the Golgi (34). PrP C is one of the GPI-anchored proteins associated with lipid rafts and detergent-resistant membranes. Sialylated GPI anchors are rare in mammals (33,34). Apart from PrP C , only pig kidney dipeptidase and human CD59 (a membrane inhibitor of reactive lysis) were found to carry sialic acids on their GPIs (33). The biological roles of sialylation of GPIs are not known.
Determining whether PrP Sc recruits PrP C with either sialo-GPIs, asialo-GPIs, or both is important for several reasons. Selective recruitment of PrP C with either sialo-GPIs or asialo-GPIs would limit the population of PrP C molecules that can support prion replication. Moreover, if molecules with sialo-GPIs and asialo-GPIs are trafficked to different sites on cell surfaces (reviewed in Ref. 25), then prion replication would be restricted to those sites. Finally, selective recruitment of PrP C as a function of the sialylation status of their GPIs would open new opportunities for therapeutic intervention against prion diseases; for instance, via targeting the enzymatic pathways that are in charge of sialylation of GPIs. However, this work demonstrated that PrP Sc recruits PrP C molecules with both sialo-and asialo-GPIs. Indeed, both sialo-and asialo-GPIs were found in PrP Sc formed in vivo, including brain and spleen tissues as well as differentiated and non-differentiated cells cultured in vitro. All five prion strains examined showed mixtures of sialo-and asialo-GPIs in PrP Sc .
Although the ability to incorporate both forms appears to be universal, the ratio of sialoversus asialo-GPIs was found to depend on host, tissue, and cell type. All three prion strains propagated in mice showed a higher proportion of sialoversus asialo-GPIs in comparison with hamster strains. Thus, these differences are likely to be attributed to the host-specific ratios of sialoversus asialo-GPIs in PrP C rather than strain-specific preferences for selecting a particular type of GPI. Within the same strain, brain-derived PrP Sc showed a significantly higher ratio of sialoversus asialo-GPIs than spleen-derived PrP Sc . This result argues that the sialylation status of GPI is controlled in a tissue-specific manner. Tissue-specific differences in the sialylation status of GPI could be due to differences in the expression levels of sialyltransferases and sialidases in cells of different origin and/or tissue-specific differences in the rates of trafficking of PrP C through the Golgi. In the spleen, only a small subpopulation of cells, specifically follicular dendritic cells, expresses PrP C and supports prion replication (45)(46)(47). The relative levels of expression of sialyltransferases and sialidases in follicular dendritic cells versus brain are not known. It is also not known which sialyltransferases are responsible for sialylation of GPIs. Twenty sialyltransferases have been identified in mammals (48 -50). The total sialyltransferase activity in the spleen is higher than in the brain (48). However, in part, the sialyltransferase activity in the spleen is attributed to extracellular sialyltransferases, which are involved in modulating the differentiation and activation of cells of the immune system (51-57), whereas sialylation of GPIs is believed to occur in the Golgi. Remarkably, opposite to the relative sialylation levels of GPIs in the spleen versus the brain, the sialylation levels of N-linked glycans were found to be much higher in spleen-than brain-derived PrP Sc (40). However, the enhanced sialylation of spleen-derived PrP Sc was attributed to post-conversion sialylation by extracellular sialyltransferases (40).
The results obtained for N2a and myotube cells emphasized that the differences in sialylation status of GPIs within PrP Sc is controlled by cell type and differentiation status. RML generated in myotubes on days 10 and 15 post-infection had GPIs with substantially higher levels of sialylation than RML produced in N2a cells (Figs. 4 and 5). It is unlikely that these differences are attributed to a higher level of expression of sialyltransferases in myotubes relative to those in N2a. In fact, sialylation of PrP Sc N-linked glycans did not show such dramatic differences. Moreover, the expression levels of sialyltransferases are known to be much higher in cancer cell lines, to which N2a belongs, than in differentiated cells such as myotubes (reviewed in Refs. 58,59). If the sialylaton status of GPIs within PrP Sc reflects that of PrP C , then the cell-specific differences in sialylation of PrP Sc GPIs might reflect differences in the speed of PrP C trafficking through the Golgi and/or the life time of PrP C in myotubes versus N2a cells. Notably, parallel changes toward more sialylated isoforms of GPIs and N-linked glycans were observed in myotubes with the progression of differentiation to a terminal stage (Fig. 5). These changes are consistent with an increase in the expression levels of sialyltransferases during differentiation of myoblasts (60). RML derived from brain, N2a, and myotubes showed differences in sialylation patterns of N-linked glycans, a topic that will be examined in detail in other studies.
We showed previously that prion strains recruit PrP C selectively according to the sialylation status of N-linked glycans (39,61). Some strains are able to accommodate PrP C molecules with heavily sialylated glycans, whereas others preferentially select PrP C with low levels of glycan sialylation at the expense of heavily sialylated ones (39,61). It appears that this is not the case with respect to the sialylation status of GPI. This study established that, within PrP Sc , the sialoforms of GPIs are determined by host, tissue, cell type, and cell differentiation status but not strain-specific structures of PrP Sc . Indeed, the GPI sialylation status was very similar if not identical in the three mouse strains or the two hamster strains. At the same time, the GPI sialylation status was significantly different for the same scrapie strain derived from brain or spleen (Fig. 4) or produced in N2a cells, myotubes, or brain (Fig. 4). In fact, within the same strain, the tissue-specific differences in GPI sialylation status were considerable enough to determine the tissue origin of PrP Sc . Notably, upon treatment with sialidase, the profiles of asialo-GPIs were found to be very similar between mouse and hamster PrP Sc or between brain-and spleen-derived PrP Sc . These results illustrate that the differences in the GPI profiles between different species or tissues mainly result from alterations in the ratio of sialo-GPIs versus asialo-GPIs.
Previous work by Stahl et al. (12) that employed mass spectroscopy demonstrated that PrP C and PrP Sc have a similar GPI composition (12). Because of the poor resolution of PrP C isoforms on two-dimensional gels (43), we were unable to assess the sialylation status of GPI within PrP C directly. Instead, proteolytic fragment C1, which is well resolved on two-dimensional blots, was used as a reporter of sialylation of GPIs in PrP C . Because the extent to which the sialylation status of C1 reflects that of PrP C is not known, these results should be considered with caution. The sialylation profiles of GPIs in N2a-derived C1 and PrP Sc were very similar, with asialo-GPI being predominant in both forms. However, the GPIs of brain-derived PrP Sc appeared to be more sialylated than brain-derived C1. This difference could be due to an increase in the expression levels of sialyltrasferases in scrapie-infected brain relative to normal brain (62). Nevertheless, N2a-derived C1 and PrP Sc were considerably less sialylated in comparison with the brain-derived C1 and PrP Sc , respectively, suggesting that the pattern of GPI sialylation in PrP Sc is reflective of GPI sialylation of a substrate.
The heterogeneity in isoelectric points for unglycosylated PrP Sc attributed to sialo-and asialo-GPIs in this study is con-FIGURE 6. Two-dimensional analysis of the sialylation status of C-terminal fragment C1. A, representative two-dimensional Western blot of mouse NBH (top blot) or non-infected N2a cell lysate (bottom blot). Western blots were stained with SAF-84 antibody. Black and white arrowheads mark diglycosylated and monoglycosylated glycoforms, respectively, whereas arrows mark the unglycosylated forms of the C1 fragment. The dashed arrow marks diglycosylated full-length PrP C . B, intensity profiles of unglycosylated isoforms of C1 fragments from mouse NBH (top plot) or non-infected N2a cell lysate (n ϭ 3 independent animals or cell cultures). The three major spots corresponding to unglycosylated isoforms are marked as 1, 2, and 3. sistent with the two-dimensional patterns for PrP Sc of diverse origin, including human prion diseases reported in previous work (39,40,(63)(64)(65)(66)(67). Multiple charge isoforms attributable to several structural isoforms of GPIs were also observed upon removal of N-linked glycans using PNGase F treatment (63)(64)(65)(66)(67). However, interpretation of charge isoform patterns on two-dimensional blots upon removal of N-glycan by PNGase F is difficult because the asparagine residues to which glycans are linked are converted to aspartic acid residues upon PNGase F treatment. Such a transformation adds negative charges to the proteins and charge heterogeneity to two-dimensional profiles.
A previous study by Stahl et al. (12) analyzed GPI structures of PrP Sc of two hamster strains, 139H and Sc237, where the latter is of the same origin as the strain 263K used in this work (68). Using mass spectroscopy analysis, Stahl et al. (12) identified six structural isoforms of GPIs in brain-derived PrP Sc that belong to two classes, sialo-or asialo-GPIs, where the sialylated forms contributed to ϳ30% of the total GPIs. In sharp contrast to the work by Stahl et al. (12), recent studies by Bate et al. (37) claimed that PrP C with asialo-GPIs were not converted into PrP Sc and, even more, inhibited conversion of PrP C with sialo-GPIs into PrP Sc . To arrive at this conclusion, Bate et al. (37) used N2a cells and primary neurons infected with PrP Sc of an unspecified strain, whereas a cell painting technique was used to administer PrP C with sialylated or desialylated GPIs to cultured cells. The results of this work are in excellent agreement with the work by Stahl et al. (12), as we demonstrated that PrP Sc consists of PrP molecules with both sialo-and asialo-GPIs. Moreover, our two-dimensional analysis revealed that, in hamster strains, the majority of PrP Sc anchors were asialo-GPIs, which is again in excellent agreement with the study by Stahl et al. (12). Our conclusions regarding the composition of GPIs are not limited to the brain-derived scrapie material but are also applicable to PrP Sc produced in terminally differentiated and non-differentiated cultured cells, including the N2a cell line used by Bate et al. (37). Remarkably, in sharp contrast to the results of Bate et al. (37), we found that, when N2a cells are stably infected with prions, the vast majority of GPIs within PrP Sc were asialo-GPIs. If the hypothesis introduced by Bate et al. (37) is correct, N2a would not support prion replication.
Several factors could contribute to the discrepancy between this work and the study by Bate et al. (37). First, it is not clear whether PrP C administered to cells using the cell painting technique recapitulates all the features of PrP C expressed by a cell (37). Adding PrP C exogenously can cause nonspecific aggregation of PrP C that depends on the sialylation status of its GPIs or altered cellular localization, leading to the exclusion of a subpopulation of PrP C from conversion. Second, the experiments with the cell painting technique employed PrP C that was deglycosylated by treatment with PNGase F (37). Not only does PNGase F treatment alter the amino acid sequence of PrP C by transforming asparagine residues to aspartic acid residues, the absence of glycans in PrP C is known to promote formation of self-replicating, PK-resistant states that are not PrP Sc (69 -73). Third, the question of whether PK-resistant material generated by the cell painting technique in cultured cells is infectious was not addressed in the study by Bate et al. (37). In contrast, the infectivity of PrP Sc produced in N2a cells stably infected with prions, the method used in this work, was documented in numerous previous studies (41, 74 -77). In summary, this study found no strain-specific preferences for selecting PrP C with sialoversus asialo-GPIs. Instead, this work suggests that the sialylation status of GPIs within PrP Sc is regulated in a cell-, tissue-, or host-specific manner and is likely to be determined by the levels of biosynthetic enzymes that control the sialylation status of GPI.

Experimental Procedures
Ethics Statement-This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The animal protocol was approved by the Institutional Animal Care and Use Committee of the University of Maryland, Baltimore (assurance no. A32000-01, permit no. 0215002).
Preparation of Brain and Spleen Material for Two-dimensional Electrophoresis-Brains or spleens were collected individually from each animal at the terminal stage of the disease. 10% (w/v) scrapie brain or spleen homogenates were prepared separately for each animal in PBS using glass/Teflon homogenizers attached to a cordless 12-V compact drill (Ryobi) as described previously (40). For two-dimensional electrophoresis of brain-derived material, an aliquot of 10% (w/v) homogenate was diluted with 9 volumes of 1% (v/v) Triton X-100 in PBS, sonicated for 30 s inside a Misonix S-4000 microplate horn (Qsonica), and treated with 20 g/ml PK (New England Biolabs) for 30 min at 37°C. For two-dimensional electrophoresis of spleen-derived material, 250 l of 10% (w/v) homogenate was diluted 1:1 with PBS, aliquoted into 0.2-ml, thin-wall PCR tubes, sonicated for 30 s inside a Misonix S-4000 microplate horn, and then combined into one tube, which was subjected to a 30-min centrifugation at 16,000 ϫ g at 4°C. The pellet was resuspended in 25 l of 1% (v/v) Triton in PBS and treated with 20 g/ml PK for 30 min at 37°C. The resulting brain or spleen samples were supplemented with 4ϫ SDS loading buffer, heated for 10 min in a boiling water bath, and processed for two-dimensional electrophoresis as described below. For twodimensional analysis of age matched non-infected mouse brains, 200 l of 10% mouse normal brain homogenate was diluted 2-fold in PBS buffer supplied with proteinase inhibitors (catalog no. 1836145, Roche) and sonicated for 30 s in a water bath at 37°C. The sample was subsequently centrifuged at 16,000 ϫ g at 4°C for 30 min. The supernatant was discarded, and the pellet was dissolved in 100 l of 1% Triton X-100 in PBS. 18 l of brain material was mixed with 6 l of 4ϫ SDS sample buffer, incubated for 10 min at 95°C, and subsequently used for two-dimensional gel electrophoresis.
Culturing of N2a Cells-N2a cells were cultured at 37°C, 5% CO 2 in minimum essential medium (catalog no. 10-010-CV, Corning) supplemented with 10% FBS (catalog no. 10437, Life Technologies), antibiotics (1% v/v penicillin/streptomycin, catalog no. 15140, Life Technologies), and 1% GlutaMAX. To produce scrapie-infected N2a cells, they were grown to 50% confluence and then incubated for 24 h with 0.5% scrapie brain homogenate from RML-infected mice. After incubation, the medium was removed and replaced with fresh medium. Cells were then split and passaged 8 times to obtain a stable infection. N2a cells stably infected with RML were collected in 200 l of BS, lysed by 30-s sonication (Misonix S-4000 microplate horn, Qsonica) in the presence of 1% v/v Triton X-100, and treated with 10 g/ml PK (New England Biolabs) for 30 min at 37°C. After adding 4ϫ SDS loading buffer, the sample was heated for 10 min in a boiling water bath and analyzed by two-dimensional electrophoresis. Non-infected N2a cells were cultured to 70% confluence, lysed with M-PER Mammalian Protein Extraction Reagent (catalog no. 78501, Thermo Fisher Scientific), supplemented with 4ϫ SDS sample buffer, incubated for 10 min at 95°C, and used for two-dimensional analysis.
Culturing of Myotubes-C2C12 cells were differentiated into murine myotube cells as described previously with minor modifications (42). Differentiation was initiated by replacing proliferating medium (DMEM (catalog no. 10-013-CV,Corning-Cellgro), 10% fetal bovine serum (catalog no. 10082-147, Life Technologies), and 1ϫ antibiotics (catalog no. 15240-062, Life Technologies)) to differentiating medium (DMEM (catalog no. 10-013-CV, Corning-Cellgro), 1% horse serum (catalog no. 26050-088, Life Technologies), and 1ϫ antibiotics (catalog no. 15240-062, Life Technologies)) in confluent myoblast cells. After differentiation for 4 days, cells were incubated overnight with 0.05% scrapie brain homogenate from RML-infected mice. Cells were then washed with PBS and cultured in differentiating medium for up to 15 days as indicated. The differentiating medium was changed every day. Cells were lysed in 1 ml of M-PER (catalog no. 78501, Thermo Scientific). 200 l of lysate was digested by proteinase K (10 g/ml) for 30 min at 37°C with gentle shaking and then concentrated by overnight acetone precipitation. The resulting pellet was dissolved in 50 l of 1ϫ SDS loading buffer, heated for 10 min in a boiling water bath, and analyzed by two-dimensional electrophoresis.
Treatment of PrP Sc with Sialidase-10% (w/v) scrapie brain and spleen materials were diluted 10-fold in 1ϫ PBS supplied with 1% (v/v) Triton X-100 and supplemented with 25 g/ml PK. After 30 min at 37°C, PK digestion was stopped by addition of 5 mM PMSF. Then the samples were denatured by incubating for 10 min at 95°C. The subsequent treatment with A. ureafaciens sialidase (catalog no. P0722L, New England Biolabs) was as follows. After addition of 10% (v/v) sialidase buffer GlycoBuf-fer1 supplied by the enzyme manufacturer, 200 units/ml sialidase were added, followed by incubation on a shaker at 37°C for 10 -12 h. After treatment, 4ϫ SDS buffer was added to the samples, and they were boiled for 10 min at 95°C.
Two-dimensional Electrophoresis-Samples of 25 l (50 l in the case of sialidase-treated spleen material or N2a-and myotube-produced PrP Sc ) volume were prepared in loading buffer as described above, solubilized for 1 h at room temperature in 200 l of solubilization buffer (8 M urea, 2% (w/v) CHAPS, 5 mM tributylphosphine (cat no. 1632101, Bio-Rad), and 20 mM Tris-HCl (pH 8.0)), alkylated by adding 7 l of 0.5 M iodoacetamide, and incubated for 1 h at room temperature. Then 1150 l of ice-cold methanol was added, and samples were incubated for 2 h at Ϫ20°C. After centrifugation for 30 min at 16,000 ϫ g at 4°C, the supernatant was discarded, and the pellet was resolubilized in 200 l of rehydration buffer (7 M urea, 2 M thiourea, 1% (w/v) DTT, 1% (w/v) CHAPS, 1% (w/v) Triton X-100, 1% (v/v) ampholyte, and a trace amount of bromphenol blue). Fixed, immobilized, precast immobilized pH gradient (cat no. ZM0018, Thermo Fisher Scientific) strips with a linear pH gradient of 3-10 (catalog no. ZM0018, Life Technologies) were rehydrated in 155 l of the resulting mixture overnight at room temperature in IPG Runner cassettes (catalog no. ZM0003, Life Technologies). Isoelectrofocusing (first dimension separation) was performed at room temperature with rising voltage (175 V for 15 min, then a 175-to 2000-V linear gradient for 45 min, then 2000 V for 30 min) on a Life Technologies Zoom Dual Power Supply using the XCell SureLock Mini-Cell Electrophoresis System (catalog no. EI0001, Life Technologies). The IPG strips were then equilibrated for 15 min consecutively in (a) 6 M urea, 20% (v/v) glycerol, 2% SDS, 375 mM Tris-HCl (pH 8.8), and 130 mM DTT and (b) 6 M urea, 20% (v/v) glycerol, 2% SDS, 375 mM Tris-HCl (pH 8.8), and 135 mM iodoacetamide and then loaded on 4 -12% BisTris Zoom SDS-PAGE precast gels (catalog no. NP0330BOX, Life Technologies). For the second dimension, SDS-PAGE was performed for 1 h at 170 V. Immunoblotting was performed as described elsewhere, and blots were stained using 3F4 antibody for hamster-derived material or Ab3531 antibody for mouse-or cell-derived material.
GPI Sialylation Profile Analysis-Two-dimensional Western blotting signal intensities were digitized for densitometry analysis using AlphaView software (ProteinSimple, San Jose, CA). Charged isoform profiles of the non-glycosylated glycoform were built using the "Lane profile" function in the AlphaView program. Normalization was performed with the highest curve signal value taken as 100%.
Author Contributions-E. K., S. S., and N. K. performed the experiments. E. K. analyzed the data. E. K. and I. V. B. conceived the study. I. V. B. and E. K. wrote the manuscript. All authors read and approved the final manuscript.