Endogenous Proteolytic Cleavage of Disease-associated Prion Protein to Produce C2 Fragments Is Strongly Cell- and Tissue-dependent*

The abnormally folded form of the prion protein (PrPSc) accumulating in nervous and lymphoid tissues of prion-infected individuals can be naturally cleaved to generate a N-terminal-truncated fragment called C2. Information about the identity of the cellular proteases involved in this process and its possible role in prion biology has remained limited and controversial. We investigated PrPSc N-terminal trimming in different cell lines and primary cultured nerve cells, and in the brain and spleen tissue from transgenic mice infected by ovine and mouse prions. We found the following: (i) the full-length to C2 ratio varies considerably depending on the infected cell or tissue. Thus, in primary neurons and brain tissue, PrPSc accumulated predominantly as untrimmed species, whereas efficient trimming occurred in Rov and MovS cells, and in spleen tissue. (ii) Although C2 is generally considered to be the counterpart of the PrPSc proteinase K-resistant core, the N termini of the fragments cleaved in vivo and in vitro can actually differ, as evidenced by a different reactivity toward the Pc248 anti-octarepeat antibody. (iii) In lysosome-impaired cells, the ratio of full-length versus C2 species dramatically increased, yet efficient prion propagation could occur. Moreover, cathepsin but not calpain inhibitors markedly inhibited C2 formation, and in vitro cleavage by cathepsins B and L produced PrPSc fragments lacking the Pc248 epitope, strongly arguing for the primary involvement of acidic hydrolases of the endolysosomal compartment. These findings have implications on the molecular analysis of PrPSc and cell pathogenesis of prion infection.

pathogenesis of these diseases is crucially linked to the cellular prion protein (PrP C ) (1), a host-encoded glycoprotein attached to the membrane by a glycosylphosphatidylinositol (GPI) anchor, whose normal function is uncertain (2). Infected individuals accumulate an abnormal form of this protein (PrP Sc ), principally in their nervous and lymphoid tissues (3). Conversion of PrP C into PrP Sc seems to take place at the cell surface or along the endocytic pathway. It involves a profound conformational change, leading to the acquisition of new properties such as insolubility in nondenaturing detergent, a strong tendency to aggregate, and an increased resistance to protease digestion, properties that are commonly used to distinguish the two PrP isoforms. Incubation of TSE-infected tissue homogenate with proteinase K (PK) in conditions that completely degrade PrP C generates N-terminal-truncated fragments of PrP Sc , referred to as PrP res . Molecular analysis of the relative mass and glycoform ratio of these fragments allows the categorization of the clinicopathological heterogeneous TSE affecting humans and animals (4,5). This phenotypic diversity is due to the existence of multiple strains of prion in conjunction with host genetic factors including prnp gene polymorphism or mutation. Variation in structural organization of PrP Sc within multimers is thought to underlie prion strain diversity. These strain-specific, conformational differences in turn lead to exposure of distinct cleavage sites for PK. Proteolytic processing of PrP Sc has been shown to occur both in brain tissue and cultured cells. A well recognized event is N-terminal truncation leading to the production of PrP Sc species commonly referred to as C2, possibly a step toward its complete degradation. Cleavage to produce C2 takes place within the unstructured region of the molecule, distal to the tandem array of octarepeats, and upstream of the physiological cleavage site of PrP C (position 111-112, human numbering) leading to a fragment called C1 (6). C2 is PK resistant and assumed to be the in vivo counterpart of the C-proximal fragment generated by PK digestion of full-length PrP Sc . The presence of the N-terminal-truncated PrP Sc species in infected brain tissue has been reported in naturally affected species, humans and sheep (6 -9), as well as in mouse and hamster models (10 -14). Immunohistochemical studies in sheep combined with PrP peptide mapping have demonstrated the intracellular accumulation in the brain and lymphoid tissues of the various N-terminal-truncated PrP Sc species, some of which might correspond to C2 (15,16). More recently, a region-specific deposition of C2 fragments was reported in sheep brain (9). Altogether these findings provide some evidence that the endocel-lular processing of PrP Sc can be influenced by the agent strain but also possibly by the cell or tissue where it propagates.
Cell culture systems steadily infected with prions offer a convenient system in which PrP Sc processing can be studied. N-terminal-truncated, C2-like fragments present before any PK digestion have been observed to accumulate in several mouse cell models, including N2a, GT1, and SMB cell lines (12,13,17). This trimming can occur within a few hours after PrP Sc acquires its protease resistance, as revealed by the use of metabolically labeled or more recently, tetracysteine-tagged PrP (12,18). Although matrix metalloproteases have been ascribed a role in the generation of the C1 PrP C fragment (19), the identity and relative contribution of the cellular proteases acting in PrP Sc processing is less clear. Treatments of cultures by lysosomotropic compounds such as NH 4 Cl have been reported to inhibit the generation of C2 cleavage products (12,17), thus potentially involving hydrolases from the acidic endosomal cell compartment, a recognized site of PrP Sc accumulation (20 -22). Cysteine protease inhibitors have also been shown to affect PrP Sc clearance in cell culture (23). One detailed study has led to the proposal that endoproteolytic C2 cleavage of PrP Sc , and prion propagation, are calpain-dependent processes (13). Although lysosome inhibition appeared to prevent PrP Sc trimming without any major effect on its biosynthesis (12,17), cysteine protease inhibitors were shown to either increase (23), reduce (24) the PrP Sc steady-state level, or leave it unaffected (18) depending on the cell model, thus raising the possibility that cysteine proteases may indirectly control PrP Sc propagation.
In this study, we investigated the endogenous processing of PrP Sc in various cell cultures and mouse tissues infected by the same TSE agent. We found that the proportion of N-terminaltruncated versus full-length molecules varies considerably depending on the cellular environment. This process, in which hydrolases from the acidic cell compartment, not calpain, appeared to be primarily involved, did not or only marginally affected prion formation in the cell culture. We also show that the N terminus of naturally trimmed PrP Sc molecules can differ from those produced by PK digestion. Our findings bring new information on the natural processing of PrP Sc molecules, which is important for prion cell biology and molecular characterization or subtyping of TSE agents.

EXPERIMENTAL PROCEDURES
Cell Culture-Rov cells (clone Rov9) and Rom cells are derived from the RK13 epithelial cell line and express the ovine or mouse PrP, respectively, in a doxycycline-dependent manner (25,26). They were grown in Opti-MEM medium supplemented with 10% fetal calf serum (FCS), penicillin, and streptomycin, and split at a 1 ⁄ 4 dilution once a week. MovS cells (clone MovS6) are Schwann cell-derived, immortalized cells isolated from tg338 mice constitutively expressing the VRQ allele of ovine PrP (27). The cells were grown in Opti-MEM medium supplemented with 10% FCS plus antibiotics and split once a week at 1 ⁄ 10 dilution. CAD cells are issued from a clone of the CathA cell line, originally derived from mouse brain neurons (28). The cells were cultivated in Opti-MEM (Invitrogen) supplemented with 10% FCS, and split after mechanical resuspen-sion twice a week at a 1 ⁄ 20 or 1 ⁄ 40 dilution. Primary cultures of cerebellar granule neurons (CGN) and astrocytes (CAS) were isolated from 6-day-old tg338 or C57Bl/6 mice as previously described (29,30). For neuron cultures, isolated cells were seeded on poly-D-lysine-coated plates and cultivated in Dulbecco's modified Eagle's medium-glutamax I high glucose (Invitrogen) with 10% FCS, antibiotics, with the addition of 20 mM KCl plus N2 and B27 supplements (Invitrogen). The cultures were complemented weekly with 1 mg/ml of glucose and antimitotic agents, uridine and fluorodeoxyuridine (Sigma) were added to inhibit astrocyte proliferation. Astrocyte cultures were prepared as described above and left for 8 days. To eliminate the neuronal population, the cells were shifted to Dulbecco's modified Eagle's medium containing 10% FCS and antibiotics. The medium was changed weekly for 3 weeks.
Prion Infection of Cell Cultures and Preparation of Lysates-Rov and MovS cells were infected with the 127S scrapie strain as previously described (25,27) using 1% (w/v) brain homogenate of terminally ill tg338 mice (37). CGN and CAS primary cultures from tg338 mice were exposed to 0.01% homogenate of 127S-infected tg338 mouse brain for 3 days and the accumulation of PrP Sc was determined after 3 weeks as previously described (29). CGN from C57Bl/6 mice were exposed to 0.01% homogenate of 139A or 22L-infected C57Bl/6 mouse brain (30). CAD cells were incubated for 1 or 2 days with 0.5% brain homogenate of C57Bl/6 mice infected with 22L, 139A, or Chandler strains of mouse-adapted scrapie as described (38). Scrapie-infected cells were analyzed from passages 2 to 12 or more after infection.
Transgenic Mice, in Vivo Infections, and Tissue Homogenate Preparation-tg338 mice, expressing the VRQ allele of ovine PrP (37), were infected by intracranial inoculation with the 127S scrapie strain. tga20 mice overexpressing mouse PrP were infected with the 139A strain. Brains and spleens of infected animals were harvested at the terminal stage of the disease and homogenized with a ribolyzer (Hybaid, Middlsex, UK) and adjusted to 20 and 10% (w/v) in medium containing 5% glucose, respectively.
Cell and Tissue Lysate Preparation-The cells were washed twice in phosphate-buffered saline and whole cell lysates were prepared in TL1 buffer (50 mM Tris-HCl, pH 7.4, 0.5% deoxycholate, 0.5% Triton X-100) or TNT buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Triton X-100) supplemented with protease inhibitors (Roche Applied Science). Lysates were clarified by centrifugation at 1000 ϫ g for 2 min before use or freezing. For analysis, brain and spleen homogenates were diluted in TL1 or TNT buffers. The protein concentration of cells and tissue lysates was determined by BCA (Pierce). The tissues were frozen immediately after removal to minimize spontaneous proteolysis and precautions were taken during the preparation of lysates.
PrP Sc Sedimentation Experiment-250 to 500 g of protein from cell or tissue lysates of control or infected samples were sedimented at 22,000 ϫ g for 30 min. Supernatants were recovered and pellets were washed once with TL1 buffer and resolubilized in Laemmli sample buffer for SDS-PAGE analysis. In some experiments, pellets were further resuspended in TL1 buffer and treated with PK (see below) before denaturation with Laemmli sample buffer.
Cu 2ϩ -IMAC Hi-Trap Chromatography-The AKTA Purifi-er100 FPLC chromatographic system was used (GE Healthcare). A 1-ml Hi-Trap IMAC column (GE Healthcare) was charged with 0.2 M CuSO 4 . The column was equilibrated with TNT buffer containing 3 mM imidazole. One milliliter of cell (1 mg of protein) or 0.5 ml of brain lysates (0.75 mg of protein) were injected into the column. The flow-through fraction was recovered, the column was washed, and a 10-min linear gradient of 3-200 mM imidazole in TNT buffer was applied to elute column-bound proteins at a flow rate of 1 ml/min; 0.5-ml fractions were collected. As a final step, 6 M urea was used to remove tightly bound proteins. Column fractions were analyzed by Western blotting for PrP detection before and after PK treatment. Individual columns were dedicated to each type of lysate.
Proteinase K Digestion-Aliquots of 50 l of whole cell lysates were treated with 10 g/ml (10 g/mg of protein) of PK at 37°C for 1 h and denatured with Laemmli sample buffer. Mouse brain and spleen lysates were either treated with 50 g/ml (10 g/mg of protein) of PK at 37°C for 1 h or PK digested according to the Bio-Rad test protocol (39). To analyze cell lysates fractionated on IMAC columns, 400 l of eluted fractions were treated with 5 g/ml of PK and then precipitated with cold methanol after addition of 20 g of bovine serum albumin as a protein carrier. Methanol-precipitated proteins were sedimented at 22,000 ϫ g for 15 min and the pellets were solubilized in 40 l of 1ϫ sample buffer. To analyze brain lysates, 50 l of IMAC column fractions were digested with 10 g/ml of PK for 1 h at 37°C and then denatured with 4ϫ sample buffer.
Thermolysin Treatment-Cell and brain lysates were treated with thermolysin (Sigma) as previously described (40) with slight modifications. Thermolysin was used at a concentration of 10 g/ml (10 g/mg protein) for cell lines and CGN, and at 200 g/ml for brain lysates. After incubation at 37°C for 1 h, samples were denatured with an equal volume of ϫ2 sample buffer.
Cathepsins Treatment-Sedimentation pellets of infected Rov9 cell lysates were solubilized in cathepsin digestion buffer containing (i) dibasic sodium phosphate brought to pH 6 or 5 with citric acid; (ii) L-cysteine (18 or 195 mM) for redox conditions mimicking that in slightly acidic or acidic compartments, respectively (41). Human liver cathepsin B and L (Merck) were used at 50 g/ml overnight at 37°C. Reactions were terminated by addition of an equal volume of 2ϫ sample buffer.
Quantification of Deglycosylated PrP Sc Species-Samples were treated with PNGase F according to the manufacturer's instructions (New England Biolabs) and quantification of fulllength and C2 PrP Sc bands was determined by GeneTools software after acquisition of chemiluminescent signals with a GeneGnome digital imager (Syngene).
Removal of GPI Anchor-Removal of the GPI anchor with hydrofluoric acid was performed according to a published procedure (42) with one critical modification. Briefly, 1 volume of crude lysates from uninfected brain or cell materials was mixed directly with 6 volumes of aqueous 48% hydrofluoric acid (Merck), incubated for 24 h at 4°C, and then vacuum-dried material was solubilized in 2ϫ SDS-PAGE sample buffer. PrP C was immunodetected using ICSM4 mAb, specific for the unglycosylated form.
Immunoblotting-Either 12 or 4 -12% NuPage gels (Invitrogen) were used for SDS-PAGE. Transfer of proteins on nitrocellulose filters was performed using a semi-dry transblot system (Bio-Rad). For detection of PrP by immunoblotting, an enhanced chemiluminescence (ECL) detection system (Pierce or Roche) was used with goat anti-mouse IgG coupled to peroxidase as secondary antibodies.

PrP Sc Processing Greatly Varies Depending on Cell and Tissue
Types-We looked for possible quantitative and qualitative differences in the natural processing of PrP Sc depending on which cells support the propagation of the prion. We first examined the molecular profile of the immunoreactive PrP species in different cell systems and mouse tissues infected by the same prion strain, 127S, a well characterized strain of sheep scrapie agent. The cell systems studied comprised previously described Rov and MovS cell lines (27,43), primary CGN and astrocytes (CAS) derived from tg338 mice (29), and the tissues were whole brain and spleen homogenates from tg338 mice. All these materials are genetically engineered to express solely the same ovine PrP VRQ allotype. On immunoblot analysis, a typical profile, including a well individualized band of 26 kDa corresponding to unglycosylated PrP C , was observed in uninfected cell cultures and healthy tissues. In scrapie-infected Rov and MovS cells, additional PrP fragments shorter than 26 kDa were present, whose size matched those of PK-resistant PrP Sc fragments (Fig.  1A). Centrifugation of cell lysates corroborated the presence of detergent-insoluble, truncated PrP Sc , which was quantitatively recovered in the pellets, whereas normal PrP C did not sediment in these conditions. As expected, sedimented PrP Sc was also PK-resistant (data not shown). Therefore, the bulk of PrP Sc accumulating in Rov and MovS cells consisted of naturally truncated molecules, shown below to be bona fide C2 (i.e. N-terminal-truncated) fragments. Upon quantification after deglycosylation by PNGase the full-length (FL) species were found to represent only 5 and 15% of PrP Sc accumulated in Rov and MovS cells, respectively (see Fig. 12 for an overview of the quantification data). Rov cultures were maintained in serumfree medium and passed without using trypsin for splitting accumulated truncated PrP Sc -like cultures kept in standard conditions, arguing that the observed PrP Sc processing was actually endogenous to cells rather than caused by exogenous proteases (data not shown).
In sharp contrast, in infected CGN cell cultures, fragments corresponding to the truncated PrP Sc species were hardly detectable in whole cell lysates despite high PrP res levels ( Fig.  1A). Only weak signals potentially corresponding to monoand unglycosylated cleaved PrP Sc were observed in overexposed blots (data not shown). Moreover, sedimentation pellets from infected CGN cells contained essentially FL PrP Sc , which became truncated after PK digestion. The proportion of FL over total PrP Sc averaged ϳ75%, as determined after PNGase treatment (see Fig. 12). In CAS cell cultures, an intermediate situation was observed, in which both FL and truncated C2-like PrP Sc molecules were detectable, with a predominance of the latter (Figs. 1B and 12). Immunoblot analysis of whole brain homogenates from terminally ill tg338 mice revealed that PrP Sc accumulating in this tissue mostly consisted of FL species, i.e. a picture similar to that seen in CGN cells. This was not the case when spleen homogenates from the same mice were examined: indeed, the sedimented PrP Sc material mainly consisted of the truncated PrP species (see Figs. 1B and 12). Collectively these data revealed major differences in terms of PrP Sc processing both in different cultured cells and animal tissues.

PrP Sc Molecules That Are N-terminal Truncated Predominate in
Infected Cell Lines-As a means to further investigate the PrP Sc processing occurring in our cell systems and confirm the sedimentation data, we first used thermolysin, a protease that allows the detection of intact PrP Sc molecules while fully digesting PrP C (40). As shown in Fig.  2, the presence of thermolysin-resistant PrP Sc with the same molecular mass as FL PrP was detected in both CGN cells and the tg338 mouse brain, whereas in Rov cells (and in MovS cells, not shown) PK-and thermolysin-resistant PrP Sc displayed overlapping profiles.
We next asked whether it would be feasible to physically separate the different PrP Sc species, given the different binding activity of FL and 103-231 recombinant PrP molecules to immobilized copper ion (44). MovS cell PrP C was retained on an IMAC-Cu 2ϩ column and recovered mainly in 60 -80 mM imidazole-eluted fractions (Fig. 3A). In infected cells, PrP molecules including species with a molecular mass lower than 26 kDa were eluted at a lower imidazole concentration (Fig. 3B). Globally, these latter species were not recognized by an anti-octarepeat antibody ( Fig. 3C) but were resistant to PK digestion ( Fig. 3D). Similar results were obtained with lysates from infected Rov cells (supplemental Fig. S1). These results unambiguously demonstrate that most of the PrP Sc accumulated in Rov and MovS cells was N-terminal-truncated before PK treatment. On the contrary, PrP Sc produced in CGN cells was eluted from the IMAC-Cu 2ϩ column mainly as FL PrP molecules, as assessed by PK digestion of the eluted fractions (Fig. 4, A and B).
On analyzing infected brain homogenates, we found that FL PrP Sc binds to copper with an avidity close to that of PrP C (Fig.  4, C and D). Brain samples treated with PK before loading also led to an efficient retention of the PrP Sc PK-resistant core species on the column, although their elution occurred at a lower imidazole concentration than PrP C or FL PrP Sc (Fig. 4E), as also shown for naturally truncated PrP Sc accumulating in Rov and MovS cells (see Fig. 3 and supplemental S1). Altogether, these results fully corroborated the difference of processing PrP Sc in cell lines versus primary nerve cells and brain tissue.
PrP Sc Endogenous Trimming Involves Amino Acids Downstream of the Main PK Cleavage Sites-As noticeable on the immunoblot shown in Fig. 1, the electrophoretic mobility of PK-resistant fragments in cell culture and tissue appeared to differ, although the infecting prion was the same. To clarify this point, relevant samples were run on the same gels and immu- noblots were performed using ICSM4 antibody specific for unglycosylated PrP. These experiments confirmed the slower mobility of the PK-resistant fragments from CGN cells and brain tissue compared with Rov and MovS cells (Fig. 5A). A similarly slower migration of PrP res fragments from the brain versus spleen tissue was also reproducibly observed (see Fig.  1B). These experiments also revealed a variation in the mobility of PrP C in corresponding, non-infected samples. To determine which cell-specific, post-translational modification of PrP C was involved, lysates from uninfected cultures or brain tissue were treated with hydrofluoric acid to remove the GPI moiety (Fig.  5B). Because after this treatment PrP C -unglycosylated bands migrated uniformly, it was concluded that a variation in the composition of the GPI anchor is mainly responsible for the observed mobility differences of PrP C .
However, based on the relative mobility we calculated that the difference between PK-resistant fragments slightly exceeded that between PrP C species (i.e. 2.1 versus 1.0 kDa, respectively, data not shown), thus suggesting that the observed differential mobility was not ascribable solely to PrP C . Relevant monoclonal antibodies were used to probe epitopes of PrP Sc present in the different cell cultures and tissues after PK digestion, and typical results are presented in Fig. 6. Whereas antibodies 12B2 (position 93-97 in sheep sequence) and CT179 (position 218 -231) both produced strong signals in all cases, clear differences were observed with antibody Pc248. This antibody, previously described by our laboratory (32,44), binds PrP within the octarepeat region with a very high avidity (epitope mapping and comparison with other anti-octarepeat mAbs are shown in supplemental Fig. S2). PKresistant fragments from Rov and MovS cells were not recognized by Pc248, contrary to those from primary neurons or brain tissues (Fig. 6B), thus providing clear evidence that an octarepeat motif, still present in CGN cells and brain PrP Sc after PK digestion, was lost in these cell lines. PK-resistant fragments from CAS exhibited some reactivity to Pc248 (Fig. 6C). When immunoprobing was performed on sedimented, non-PK-digested materials (Fig. 6D), CAS cell lysates were shown to contain a mixture of Pc248-negative and -positive (FL) PrP Sc molecules, whereas in spleen homogenates FL PrP Sc (Pc248positive) was hardly detectable. Rov and MovS sedimented, undigested material failed to react with Pc248, confirming that a nearly exhaustive trimming of the PrP Sc molecules rather than PK digestion accounted for loss of the Pc248 epitope in these cells. Altogether, these results established that N-terminal trimming of PrP Sc molecules generated by ovine prion produces PrP fragments distinguishable from those resulting from the exogenous PK cleavage of FL PrP Sc .
Differential PrP Sc Trimming Occurs in Mouse Prion-infected Cells and Tissues-To see whether our findings would extend to other than ovine prions, we next focused our study on various models propagating mouse prions. The CAD cell line originating from brain tissue displays several features of neurons (28,38). CAD cultures steadily infected by 139A, 22L, or Chandler prion strains were shown to produce naturally N-terminaltruncated, Pc248-negative PrP Sc , which was predominantly monoglycosylated as is typically the case for PK-resistant PrP Sc for these strains (Fig. 7A). Additional species of lower mobility were observed in sedimented material using Sha31 antibody (Fig. 7A), suggesting accumulation of FL PrP Sc too. Indeed, sedimented, non-PK treated samples contained Pc248-reactive, FL PrP Sc material, with predominant monoglycosylated  1 and 2 of each panel) and 127S-infected (lanes 3-5) tg338 CGN, brain, and Rov cultures were digested with thermolysin (Ն4 experiments) using conditions in which PrP C was completely proteolyzed (lanes 2 and 4). Immunoblots using the Sha31 mAb show that thermolysin-resistant PrP Sc from CGN cells and the brain migrates with the mobility of FL species (lanes 4), not with PrP res generated by the PK treatment (lanes 5).
In contrast, the profiles of thermolysin-and PK-resistant PrP Sc from Rov cells are similar (compare lanes 4 and 5 in Rov panel). The positions of molecular mass markers are indicated. forms, and Pc248-reactive species were resistant to thermolysin digestion (Fig. 7B). The full-length form represented around one-third of PrP Sc in these cells (see Fig. 12). Infected cultures of Rom cells, similar to Rov but genetically engineered to express mouse instead of ovine PrP (26), were shown to accumulate PrP Sc essentially under the C2, Pc248-negative form (supplemental Fig. 3), as did the Rov cells. In contrast, truncated species were hardly detectable in 139A-or 22L-infected, primary cultured CGN mouse neurons (Fig. 7C).
In tissues of 139A-infected tga20 mice, PrP Sc was shown to accumulate as a FL species in a great majority in the brain (ϳ80%), whereas it appeared to be extensively trimmed in the spleen (Figs. 8 and 12), i.e. a situation similar to that seen in tg338 mice. Altogether these results indicated that the phenomenon of differential trimming is not a unique feature of ovine prions.

Inhibition of Lysosomal Hydrolases but Not of Calpains Restores Accumulation of Full-length PrP Sc -To test whether the endolysosomal compartment was involved in trimming of
PrP Sc in the cell systems studied here, we used NH 4 Cl, a compound that inhibits protease activity by raising the pH in this compartment, and/or leupeptin, a drug commonly used to impair lysosomal hydrolases. Treatment of infected Rov cultures modified the PrP immunoreactive profile, with a decrease of the C2 species and a concomitant increase of the FL PrP species, obvious for the unglycosylated band; overall, the accumulation of PrP Sc did not decrease or decreased only slightly (Fig. 9A, left two panels). Accumulation of FL PrP Sc in the lysosome-impaired culture was further substantiated by detection

. PrP Sc from CGN and the brain binds to the IMAC-Cu 2؉ column and elutes as full-length PrP. Total lysates (lane T) of 127S-infected CGN (A) or the tg338 brain (C) were applied to the IMAC-Cu 2ϩ -charged column. The flow-through fraction (FT) was collected and bound proteins were eluted by application of a 3-200 mM linear imidazole gradient and analyzed by immunoblotting for the presence of PrP molecules before (A and C) and after PK treatment (B and D).
In E, the lysate of the infected brain was first digested with PK prior to loading onto the column and the eluted fractions were analyzed. Immunoblots were revealed with Sha31 mAb. The binding and elution of FL PrP C and PrP Sc produced in CGN and in the brain are similar (compare A to B and C to D). PK-digested brain PrP Sc (E) is also retained on a copper column but elutes at lower imidazole concentrations than FL PrP Sc (compare D to E). The positions of molecular mass markers are shown on the right of each immunoblot.

FIGURE 5. Higher electrophoretic mobility of PrP C and PrP res from MovS and Rov cells compared with CGN or brain tissue.
A, the comparative mobility of PK-resistant PrP Sc (right panel) and PrP C (left panel) from tg338 mouse brain, and MovS, CGN, and Rov lysates is shown on immunoblots revealed with ICSM4, an antibody specific for the non-glycosylated form of PrP. To compare the mobility differences accurately, in these experiments the runs were performed so that the migration distances of unglycosylated species PrP C and PrP Sc were the same. B, treatment with hydrofluoric acid (HF) of non-infected brain tissue, Rov, and MovS cells generates a PrP species with identical mobility (immunodetection with ICSM4). The electrophoretic mobility differences shown were observed in three (B) or more (A) independent experiments. of thermolysin-resistant, detergent-insoluble, and Pc248-reactive PrP species (Fig. 9A, right three panels). Under lysosome inhibition the proportion of FL PrP Sc increased from 5 to 58% and from 15 to 78% in Rov and MovS cells, respectively (see Fig. 12). Intriguingly, in a recent study, PrP Sc trimming was reported to be impaired following inhibition of calpains, cysteine proteases that are not associated with the endolysosomal compartment, whereas inhibition of lysosomal proteases was ineffective (13). However, in neither Rov nor MovS cell cultures did treatment with Ն50 M calpain inhibitor III (MDL28170) (13) affect PrP Sc N-terminal trimming; at the highest dose, the inhibitor decreased the accumulation of PrP Sc , yet the C2/FL ratio, as measured on sedimented material, remained unchanged (Fig. 9B).
Importantly, PrP res fragments generated by PK digestion of lysates from NH 4 Cl-treated cultures were recognized by Pc248 antibody, unlike that in control, untreated cultures (Fig. 9C). Therefore, PK-resistant fragments produced by digestion of FL PrP Sc generated in endolysosome-impaired cells contained an octarepeat motif. Similar experiments were performed in mouse prion-infected cells. Following treatment with NH 4 Cl, the proportion of FL PrP Sc in 22L-infected cultures increased in a dose-dependent manner, and a 10 mM treatment led to the accumulation of an exclusively FL species in 139A-infected cells (Fig. 10A). When 22L-infected cells were cultivated for one passage in the presence of NH 4 Cl plus leupeptin, truncated PrP Sc disappeared almost completely (Figs. 10, B and D, and  12); such cultures accumulated as much PK-resistant PrP as untreated cultures, implying that FL and truncated PrP Sc had a similar resistance to PK digestion (Fig. 10B). Remarkably, blockade of N-terminal trimming was compatible with serial propagation of prion infection because the level of PrP Sc remained fairly stable under sustained treatment over at least 6 passages (Fig.  10C). Upon discontinuation of the treatment, production of truncated forms resumed to its original level within one passage (supplemental Fig. S4). In CAD cell cultures treated with calpain inhibitor III the accumulation of C2 fragments was diminished, but again without an increase of the FL versus C2 species ratio (Fig. 10D, left panel). In contrast, inhibition of cathepsin L or B led to an increase of the FL species at the expense of trimmed fragments (Fig. 10D, right panel). Altogether, these results led us to conclude that the N-terminal trimming of PrP Sc in ovine and mouse prion-infected cells primarily involves proteases other than calpains.
To further substantiate the nature of the enzymes involved in generation of the C2 fragment, FL PrP Sc obtained following NH 4 Cl plus leupeptin treatment of infected Rov cells was subjected to in vitro digestion by cathepsin B or L, under two different buffer conditions; one mimicking the slightly acidic endosomal compartment (pH 6) and the other the acidic lysosomal compartment (pH 5) (Fig. 11) according to Jordans et al. (41). In both cases cleaved PrP Sc fragments were obtained that showed a typical profile but lacked Pc248 reactivity (Fig. 11), as in Rov cell cultures whose endolysosomal compartment was unimpaired (see above).
In the mouse prion cell systems, CAD (Fig. 10B) and Rom cells (supplemental Fig. S3), endogenously trimmed, and PKcleaved PrP Sc fragments exhibited the same mobility. Indeed, contrary to that seen in ovine prion-infected cells, PrP res fragments produced by PK digestion of FL PrP Sc from lysosome- impaired cells showed no reactivity toward Pc248 antibody ( Fig  10B). Similarly, PrP res in brain tissues from mice infected by 22L or other mouse prions, which mostly accumulate FL PrP Sc , lacked Pc248 reactivity (data not shown). From these results it appears that, even though the main PK cleavage sites of sheep and mouse PrP Sc differed, endogenous proteolytic cleavage generated C2 fragments with similar N termini, despite different cell types and PrP sequences.

DISCUSSION
The present study was focused on the natural processing of PrP Sc leading to N-terminal-trimmed fragments commonly called C2. Several new findings emerged that will be discussed successively below.
Our study first revealed that the efficiency of trimming varies dramatically depending on which cell or tissue supported replication of the infecting prion (see Fig. 12). Thus, in cell systems infectible by sheep prion, the epithelial Rov and Schwann MovS cell lines, the bulk of PrP Sc consisted of molecules that were N-terminal-truncated prior to PK digestion. In primary cultured cells, PrP Sc produced by neurons was essentially fulllength, whereas in astrocytes an important proportion was trimmed. A contrasted situation was also observed in vivo (tg338 mouse model), where intact PrP Sc molecules largely predominated in the brain, although they were minimally represented in the spleen. The above mentioned materials all expressed the same ovine prnp allele and were infected by the same, biologically cloned sheep prion, therefore excluding any sequence or strain-dependent conformational effect. Examination of models propagating mouse-adapted prions led to similar conclusions, indicating that this may be a general phenomenon. Although in CAD neuronal cells, the proportion of C2 over FL molecules averaged two-thirds whatever the infecting prion strain, limited trimming occurred in primary cultured neurons. In the brain and spleen tissues of tga20 mice the FL/C2 ratios were roughly inverse, as seen in ovine PrP mice. It is reasonable FIGURE 7. Mouse PrP Sc is truncated in CAD cells but remains full-length in CGN. A, CAD cells were infected with mouse-adapted scrapie prion (139A, 22L, or Chandler strain) or mock-infected and analyzed for their PrP C and PrP Sc content. Immunoreactive PrP species were revealed in whole cell extracts, PKdigested and sedimented materials, as indicated, using either Sha31 or Pc248 mAbs. Note that both truncated and FL, Pc248-positive PrP species are recovered in sedimented material from infected samples, whatever the strain. B, whole cell lysate of 22L-or mock-infected CAD cells and thermolysin-or PK-digested samples were analyzed using Sha31 or Pc248 antibody. Note the Pc248-positive bands in thermolysin-resistant and sedimented materials, reflecting the presence of FL PrP Sc . C, CGN from C57Bl/6 mice were mock-infected or infected with 139A or 22L scrapie strains. Cell lysates performed 24 days post-infection were centrifuged or digested either with thermolysin or PK as indicated, and analyzed by Western blot using Sha31 mAb. The molecular profiles have been observed in multiple experiments on multiple blots. n.i., non-infected. FIGURE 8. Mouse PrP Sc is efficiently truncated in the spleen but not the brain. Analysis of PrP Sc in brain (A) and spleen (B) homogenates from noninfected or 139A-infected tga20 mice. Tissue lysates (from at least three animals) were centrifuged (spleen only) or digested either by thermolysin or PK as indicated, and analyzed by Western blot using Sha31 mAb. PrP Sc detected in the spleen is essentially truncated prior to PK digestion (lane 3). Note the presence of sedimentable PrP C in both infected and non-infected tga20 mouse spleen, which was consistently observed with mouse PrP C in these conditions. n.i., non-infected.
to anticipate that such a markedly tissue-dependent, differential PrP Sc trimming also exists in species naturally infected by prions.
Conceivably the endogenous cleavage of PrP Sc , and more specifically the generation of C2 fragments, might differ quantitatively between different nerve cell categories within the brain. Accumulation of C2 fragments in the brain tissue has been reported to occur in rodent models other than mice (13,14), as well as in naturally affected host sheep (9) and humans (6,45). Moreover, several studies have offered evidence potentially linking the C2 accumulation level in the prion-affected mouse, sheep, and human brains or its neuroanatomical deposition (7,9,14,45) to TSE strain variation. Our observation that PrP Sc trimming efficiency in primary cultured astrocytes and neurons markedly differed, in particular, supports the view that certain nerve cell subpopulations might be more prone to endogenous proteolysis than others. Thus, the ratio of C2/FL PrP Sc species in the brain tissue could also reflect the cellular tropism or "targeting properties" of a prion, rather than the solely conformational specificities of PrP Sc . In future studies, it would be interesting to determine whether generation of the C2 fragments occurs preferentially in specific cell types within the central nervous system, and whether or not "high C2-producing" strains share common biological features.
It is generally considered that the C2 fragments are the in vivo counterparts of the protease-resistant fragments produced by PK digestion in vitro (6,11,13). However, the findings presented here do not support this view. As revealed by antibody mapping, C2 fragments consistently failed to react with the potent anti-octarepeat antibody Pc248, whatever the infecting prion. To be underlined, all the agents studied here exhibited molecular features of type 1 human prions (4,5), i.e. PK-resistant PrP fragments around 21 kDa (unglycosylated band), being typically strongly reactive toward the 12B2 antibody (33). The observed lack of Pc248 reactivity of fragments generated by PK cleavage of FL PrP Sc appears to be a specific feature of mouse prions. Indeed, in other prion-infected species, PK cleavage of "type 1-like" PrP Sc tends to preserve C-proximal octarepeat residues. This is the case for sheep prions, as shown in this study by the strong Pc248 reactivity of PK-resistant fragments generated from FL PrP Sc (e.g. in brain tissue or primary neurons (CGN)). Consistent with our findings, the Trp-84 residue, located upstream of the Pc248 epitope sequence delineated here, was recently identified by mass spectrometry analysis as one of the major PK cleavage sites in brains of natural or experimental scrapie-affected sheep (46), whereas residues within or downstream of the Pc248 epitope were identified as the main PK cleavage sites in scrapie-infected mice (11,47,48). Recently, PK-resistant PrP Sc accumulating in the brain of human patients infected by type 1 prions (5, 49) was reported to be strongly reactive toward two newly introduced anti-octarepeat antibodies (50). We surmised that these antibodies (Pom2 and Pom12) have properties very similar to those of Pc248, which is characterized by high avidity for brain PrP res , whereas commonly used anti-octarepeat antibodies generally exhibit weak reactivity (for more information on Pc248 antibody, see supplemental Fig. S2). PK-resistant fragments retaining an octarepeat motif are produced upon infection by the 21K type of hamster prions also (51). 5 In conclusion, the equivalence between C2 and PK-cleaved fragments commonly observed with mouse prions is far from the general rule. Endogenous 5 M. Moudjou, unpublished data.

FIGURE 9. N-terminal trimming of PrP Sc is inhibited in lysosome-impaired Rov and MovS cells but is not calpain-dependent.
A, infected Rov cells were left untreated or cultivated for 7 days in the presence of 10 mM NH 4 Cl or 10 mM NH 4 Cl ϩ 100 g/ml of leupeptin, as indicated. The PrP profiles of undigested, PK-digested, and thermolysin-digested samples are shown after Sha31 mAb immunoblotting. Lysates were also sedimented without protease treatment and the pellets were revealed with either Sha31 or Pc248 mAbs (right two panels). B, infected Rov and MovS cells were treated with various concentrations of calpain inhibitor III or NH 4 Cl plus leupeptin, as indicated (0*: vehicle only). Sedimented materials from cell lysates were analyzed by Western blotting with Sha31 mAb. C, infected Rov cells treated or not with NH 4 Cl were digested with PK before immunoblot analysis using Sha31 or Pc248 mAbs. The above experiments have been repeated three times with consistent results.
processing events that precede in vitro proteolytic digestion determine the molecular features of PrP res to a variable extent. This notion might be worth accounting for while interpreting PrP Sc molecular profiles in biopsies from brain or other tissues that might markedly differ in their trimming capability.
PrP res fragments in mouse and sheep prion-infected cell lines typically have a higher mobility than their counterparts in the brain tissue of wild type or ovine transgenic mice, respectively (25,27,52,53). Re-inoculation to mice of the cell culture-propagated agent revealed no permanent change of its molecular profile and strain properties (26,27,(52)(53)(54)(55). The phenomenon has been proposed to simply reflect post-translational differences of PrP C molecules when expressed either in culture or tissue (52,53). Thus, unglycosylated PrP C in mouse cell lines migrate faster than mouse brain PrP C , putatively reflecting different compositions of the GPI moieties (52). New findings of the present study offer clarification of the cause of these observations. Our results established that, whereas in mouse prioninfected cells the N termini of endogenously and PK-cleaved PrP Sc are similar (i.e. both Pc248-negative and 12B2-positive), a more complex situation prevails in sheep prion-infected cells; indeed, two factors appeared to account for the molecular mass difference observed between brain and cell PrP res fragments: (i) a differential mobility of PrP C due to the variable composition of the GPI moiety, which was abolished after removal of the latter by hydrofluoric acid treatment, information lacking in the former studies; (ii) the production of distinguishable N termini produced by in vivo (endogenous) and in vitro (PK) cleavages. Likewise, not only PrP C tissue-specific differences, such as those reported in the human and sheep species (7,31,35), but also different trimming capabilities, as observed here in two mouse models (tg338 and tga20), might account for PrP res molecular mass discrepancies in neural and lymphoid tissues in naturally infected hosts.

FIGURE 10. Lysosome inhibition blocks PrP Sc N-terminal trimming but not prion propagation in CAD cells.
A, non-infected, 22L-and 139A-infected cells were either left untreated or treated with NH 4 Cl as indicated, and cell lysates were centrifuged. The pellets were analyzed for their PrP Sc content by immunoblotting using Sha31 mAb. B, 22L-infected CAD cells cultivated in the absence or presence of 10 mM NH 4 Cl plus 10 g/ml of leupeptin for two consecutive passages. Whole cell lysates and sedimented materials (right two panels) were PK-digested or not before immunoblotting using Sha31 or Pc248 mAb. C, 22L-infected CAD cells were cultivated in the absence or presence of 10 mM NH 4 Cl for 4 days and cells were passaged twice a week in the absence or presence of the lysosomal inhibitor. Cell lysates were immunoblotted using Sha31 mAb. Molecular mass markers in kDa are shown. D, infected CAD cells were treated with various concentrations of calpain inhibitor III, NH 4 Cl plus leupeptin, cathepsin L inhibitor I (cath.L), or cathepsin B inhibitor I (cath.B), as indicated. Sedimented materials from cell lysates were analyzed by Western blotting with Sha31 mAb. These experiments have been repeated three times with consistent results. n.i., non-infected. FIGURE 11. In vitro cleavage of full-length PrP Sc by cathepsin B and L remove the Pc248 epitope. PrP Sc from a lysate of infected Rov cells treated with NH 4 Cl and leupeptin was purified by sedimentation, resuspended in two different cathepsin buffers (pH 6 or 5), and digested or not (control lanes) with the indicated proteases as described under "Experimental Procedures." The same samples were revealed with either Sha31 or Pc248 mAbs.
Another unexpected finding in this study was the retention of nondenatured PrP Sc by the copper ion, because failure to bind to immobilized copper ion has been reported by others to be a distinctive trait of nondenatured PrP Sc (56,57). In our experimental conditions, the binding of both FL PrP Sc and PrP res was reproducible, and plainly ion-dependent (data not shown). It was observed irrespective of whether cell lysate or brain homogenate was used as a starting material. A strainspecific effect is implausible, because efficient binding was observed with hamster PrP Sc also, 6 the prion agent used in one previous study (56). Although we still have no clear explanation to offer for these discrepant data, we suggest reconsidering copper ion binding as a potentially useful means to prepare fulllength or naturally trimmed PrP Sc molecules in the absence of any treatment that would denature or cleave the PrP Sc species. From a theoretical viewpoint, our observation suggests that determinants of the N-terminal region are accessible within native PrP Sc multimers, because FL PrP Sc and PrP C exhibited close avidities for bound copper.
Studying the effect of lysosomotropic compounds and protease inhibitors on the generation of C2 fragments has led to somewhat controversial observations in the literature. In all three cell lines examined here (Rov, MovS, and CAD), the formation of C2 fragments was blocked efficiently by leupeptin treatment and NH 4 Cl-induced pH elevation, with little or no effect on short term PrP Sc accumulation. To be noted, PrP Sc in lysosome-impaired Rov and CAD cells consisted of true FL species, not partially protected species with a shorter size truncation (17). Our results, in line with earlier observations in two other cell lines (12,17), thus provided robust evidence for the involvement of the endolysosomal compartment in the N-terminal trimming of PrP Sc . Furthermore, we showed that in vitro cleavage of full-length PrP Sc by cathepsins B and L produced fragments that lacked the Pc248 octarepeat epitope, as observed for the endogenously produced C2 fragments. Conversely, cathepsin inhibitors were found to diminish the formation of C2 fragments in infected cell cultures. These data bring direct evidence for the involvement of these hydrolases into the generation of C2 PrP Sc . The above findings are in sharp contrast with those of a recent study, in which inhibitors of lysosomal proteases had no effect on the production of C2, whereas calpain inhibition adversely affected C2 accumulation, and also prion propagation, thus raising the possibility that C2 generation was critical (13). In none of the cell systems studied here, however, did inhibition of calpain lead to accumulation of the FL species at the expense of the C2 species, as it was consistently observed in lysosome-impaired cultures, therefore arguing that endoproteolytic cleavage of PrP Sc is not a calpain-dependent process. Prion propagation was stably maintained along the reiterated subpassages in NH 4 Cl-treated CAD cell cultures, in which PrP Sc endoproteolysis was essentially blocked, both quantitatively and qualitatively. Altogether, our findings support the view that hydrolases from the endolysosomal compartment rather than calpain are primarily involved in PrP Sc trimming and that C2 formation neither profoundly affects nor is critically required for efficient prion propagation in the cell culture.
The mechanisms underlying the neurotoxic effects engendered by prion infection remain largely elusive but most prob-6 M. Moudjou, unpublished results. , and brain and spleen tissues. Each of the three left panels show three samples left untreated and one sample treated with lysosomal inhibitors (10 mM NH 4 Cl plus 10 g/ml leupeptin); the four right panels each show two different samples. Sedimented material from lysates of cell lines and primary cultures were resuspended and treated with PNGase (similar results were obtained using samples that were digested with thermolysin before PNGase; not shown). Lysates from tg338 brain and spleen tissues were treated with thermolysin and further centrifuged in the case of spleen material to concentrate PrP Sc . Arrows indicate deglycosylated full-length (FL) and N-terminal-truncated (C2) PrP Sc species. B, the histogram shows the percentage of full-length species over total PrP Sc as determined by quantification of samples such as seen in A. Average values and standard deviation were determined from four to seven different experiments.
ably involve some form of PrP Sc . Could the trimming of PrP Sc influence its neurotoxic potential? The N terminus of PrP C appears to be important for anti-apoptosis and antioxidant functions (58,59), but what about PrP Sc ? Transgenic mice that express a deleted form of PrP mimicking C2 fragments can be infected and develop a lethal neurological disease, but this does not preclude the existence in the truncated region of determinants that could modulate the cytotoxicity of PrP Sc . Interestingly, the neuropathology in one such transgenic line was reported to be profoundly modified, i.e. no histopathological lesions could be detected in the brain or brainstem of terminally ill animals, and the cytopathic anomalies were essentially restricted to neurons of the cervical spinal cord (60). Altogether our ex vivo observations raise the question of whether trimming the FL PrP Sc could modulate its neurotoxic potential and thus exert some protective effect in the target cells. Although immortalized cells fail to recapitulate the deleterious effects occurring in vivo, primary cultured neurons show a gradually compromised survival upon prion infection (29). It is intriguing that in the latter, the FL PrP Sc form largely predominates over the C2 form, whereas the opposite is commonly observed in immortalized cells even of neural origin. In vivo, prion infection causes no overt cell death of astrocytes, which showed good trimming capacity in primary cultures, or of lymphoreticular cells, where PrP Sc could be extensively endoproteolyzed, considering our results with splenic tissue. The processing of PrP Sc depends on its access to the endocytic pathway. Thus, a plausible hypothesis would be that non-internalized PrP Sc is most harmful to the host cell. It is noteworthy that histopathological studies performed on scrapie-affected sheep brain have pointed to a relationship between cell-membrane or extracellular PrP deposition and vacuolation (61). Approaches using cell culture systems, including blockade of PrP Sc trimming without impairing prion multiplication, should make it feasible to further explore these aspects.