A seven-residue deletion in PrP leads to generation of a spontaneous prion formed from C-terminal C1 fragment of PrP

Prions result from a drastic conformational change of the host-encoded cellular prion protein (PrP), leading to the formation of beta-sheet-rich, insoluble and protease-resistant self-replicating assemblies (PrP Sc ). The cellular and molecular mechanisms involved in spontaneous prion formation in sporadic and inherited human prion diseases or equivalent animal diseases are poorly understood, in part because cell models of spontaneously-forming prions are currently lacking. Here, extending studies on the role of H2 alpha-helix C-terminus of PrP, we found that deletion of the highly conserved 190 HTVTTTT 196 segment of ovine PrP led to spontaneous prion formation in the RK13 rabbit kidney cell model. On long-term passage, the mutant cells stably produced proteinase-K resistant, insoluble and aggregated assemblies that were infectious for naïve cells expressing either the mutant protein or other PrPs with slightly different deletions in the same area. The electrophoretic pattern of PK-resistant core of the spontaneous prion ( ∆ Spont ) contained mainly C-terminal polypeptides akin to C1, the cell-surface anchored C-terminal moiety of PrP generated by natural cellular processing. RK13 cells expressing solely ∆ 190-196 C1 PrP construct, in absence of the full-length protein, were susceptible to ∆ Spont prions. ∆ Spont infection induced the conversion of the mutated C1 into a PK-resistant and infectious form perpetuating the biochemical characteristics of ∆ Spont prion. In conclusion this work provides a unique cell-derived system generating spontaneous prions and provides evidence that the 113 C-terminal residues of PrP are sufficient for a self-propagating prion entity.


Introduction
Mammalian prions are responsible for transmissible spongiform encephalopathies (TSE) in both humans and animals. Prions result from the misfolding of the host-encoded prion protein (PrP). Under its normal conformation, the cell surface GPI-anchored PrP (PrP C ) presents a globular domain containing three alpha-helices and two short antiparallel beta strands, preceded by an unstructured N-terminal part (1,2). In contrast, prions are made from assemblies of beta-sheet-rich, insoluble, aggregative and mostly partially protease-resistant PrP conformers called PrP Sc in reference to their original identification in scrapieinfected sheep (3). Prion replication appears to proceed by conversion of the normal protein through templated polymerization (4), which explains not only their propagation in tissues but also their intra-or interspecific infectivity. The high-resolution structure of PrP Sc is not yet resolved, due to inherent difficulties to produce large amounts of purified insoluble assemblies that may nonetheless have some intrinsic heterogeneity with respect to size. Several amyloid models were proposed and coexistence of different candidate structures has even been suggested (5)(6)(7). The two-third C-terminal part of PrP forming the protease-resistant core of PrP Sc , the C2 fragment, constitutes the domain necessary and sufficient for prion replication (8,9). Prion strains are identified by their specific biochemical and/or neuropathological features in the same infected host-species (10,11). Strains result from structural differences in three-dimensional or quaternary structure of PrP Sc . The N-terminal border of the C2 fragment is strain-dependent and can vary around amino acid positions 80 to 100. A C-terminal fragment called C1 results from the natural cleavage of PrP C by a cellular protease at the alpha cleavage site, which is between residues 110 and 111 of human PrP (8,12). C1 is thus smaller than C2 and considered so far too short to be converted into prions although it encompasses the structured globular domain of the full-length PrP C and is also present at the cell surface (13,14). Prions can emerge spontaneously as in sporadic cases of human Creutzfeldt-Jakob disease (CJD), that is without evidence of infection or contamination. In this context, prion generation requires at first the formation of nuclei stable enough to initiate the polymerization process, which is expected to be a slow and rate-limiting step (15). Indeed, spontaneous prion disease is a rare event, the prevalence of sporadic CJD being of approximately 1.5 case per million and per year worldwide. Mutations in PrP can favor PrP spontaneous conversion into prions. Indeed, more than 30 mutations responsible for inherited human prion diseases, including genetic CJD, Gerstmann-Sträussler-Scheinker syndrome (GSS) or Fatal Familial Insomnia, were identified and these 3 dominant allelic mutations usually show a high penetrance (16). Disease-causing mutations might favor partial unfolding or transient denaturation of PrP C , which are required for refolding into PrP Sc , and might also increase stability of initial PrP Sc seeds. The cellular and molecular processes underpinning or preventing spontaneous prion generation remain poorly understood. Transgenic mouse models of spontaneous prion formation have proven difficult to obtain. This was achieved for human or mouse PrP bearing some mutations (17,18). Hallmarks of the disease were not always reproduced in mice and intriguingly in several instances, prions showed a rather low resistance to protease digestion (19). Prions spontaneously formed in mice overexpressing either anchorless mouse PrP or I109 allele of bank vole PrP (20,21). Currently, no cellular model for spontaneous prion formation has been reported. Toward this goal we focused here on an intriguing highly conserved threonine rich region of the alpha-helix H2 associated with several disease-causing mutations in human PrP (22). In a recent work, we demonstrated that deletion of the cluster of four threonines in the alpha-helix H2 Cterminus has no or marginal effect on ovine prions replication in RK13 cells expressing ovine PrP (23,24). We now show that specific deletion of the larger H2 C-terminal segment HTVTTTT, that removes three additional residues, causes the spontaneous conversion of the mutant ovine PrP into a new type of prion. This prion exhibits a main protease-resistant core shorter than usual, of C1 size, which was able to infect naïve RK13 cells expressing the mutant C1 segment alone. The potential importance of H2 C-terminus for maintenance of normal PrP C conformation in the cell, the specificities of the new mutant prion and the surprising conversion of the homologous mutant C1 fragment into a prion entity are discussed.

∆190-196 deletion does not alter the overall structure of PrP but reduces its stability
We focused here on ∆190-196 ovine PrP (VRQ allelic variant), a mutant PrP with a specific deletion of seven amino acids at the end of helix H2 (Fig. 1). We previously reported that a larger deletion of the H2 C-terminus (∆190-197) did not have a major impact on the structure of the protein, leaving intact the spatial organization of the three alpha helices in the globular domain of PrP (23). As with ∆190-197 PrP, structural analysis of recombinant ∆190-196 PrP by circular dichroism indicated a conservation of the overall alpha helical content compared to wild-type (WT) PrP ( Fig. 2A). This is in agreement with NMR analysis of the segment 113-214 of ∆190-196 PrP (C1113), which contains the entire sequence of the structured domain and is an equivalent to the natural C1 fragment studied hereinafter. The large dispersion of amide chemical shifts observed in the 1 H-15 N HSQC spectrum of 15 N 13 C-labeled mutant C1113 indicated that it maintained a globular core, in addition to its unstructured N-terminal region (Fig.  2B). Moreover, comparison with the spectra of ∆193-196 and ∆190-197 mutant PrPs previously obtained (23,24) versus WT PrP showed that chemical shift perturbations followed a similar trend, confirming that the structure of the core domain of ∆190-196 C1113 is structurally close to those of other mutants (Fig. 2C). Last, analysis of ∆190-196 C1113 13 Calpha chemical shifts yielded the position of the three alpha-helices within residues 147-159, 175-189 and 203-230, showing that the topology is conserved with respect to WT PrP (Fig. 2C). The HTVTTTT deletion removed a histidine at position 190, which is the equivalent of His-187 in human PrP (Fig. S1). The pH-dependent protonation of this histidine is thought to play an important role in the electrostatic network and the stability of the globular part of PrP (25,26). But the deletion of the other residues might also impact the thermodynamic stability of the protein. We thus tested whether the deletion affected the stability using a thermal shift assay to determine the melting temperature of WT and mutant PrP. The melting temperature of ∆190-196 PrP (49.9°C) was reduced by 7°C compared to the WT PrP in sodium acetate buffer (10 mM, pH 5.0) (Fig 2D). A marked reduction of 8°C was also observed in these assays, using a different condition, sodium phosphate buffer (250 mM, pH 5.1), that increases the thermal stability of both WT and ∆190-196 recombinant PrPs (Fig. S2). Altogether these observations 4 indicate that ∆190-196 mutant PrP conserves the overall structure of WT PrP but loses some stability.

Expression of the mutant PrP C in RK13 cells
Ovine ∆190-196 PrP was stably transfected in RK13 cells to generate sublines and clones referred to as ∆190-196 Rov. Mutant PrP was efficiently expressed as a glycoprotein in Rov cells. Western blotting showed that unglycosylated forms of ∆190-196 PrP C were underrepresented compared to WT or ∆193-197 PrPs generated previously (23) (Fig.  3A). Deglycosylation by PNGase F treatment corroborated the high glycosylation level of ∆190-196 PrP and indicated that the mutant protein was smaller than WT PrP C by about one kDa, as expected (Fig. 3B). Using an antibody with an epitope in the C-terminal part of PrP rather than in the N-terminal region allowed us to identify both the full-length protein and its natural C-terminal C1 fragment. PNGase treatment was required for accurate identification of PrP C and C1, as they are both highly glycosylated. The relative proportion of the full-length PrP versus the C1 fragment was roughly similar for the WT and the mutant protein (

Spontaneous generation of a self-sustained protease-resistant form of PrP in ∆190-196 Rov
The expression of ∆190-196 PrP was turned on by addition of doxycycline and we followed the fate of the protein over cell passaging by Western blotting, checking for the appearance of PK-resistant forms. While ∆190-196 PrP was sensitive to PK digestion during the first passages, a PK-resistant form systematically appeared, usually after the 4 th or the 5 th passage (Fig. 4A). This protease-resistant form termed ∆190-196 PrP res persisted for > 1 year of continuous culture (Fig. 5A). The electrophoretic profile of ∆190-196 PrP res was characterized by a large smear of glycosylated species and presence of a well individualized faint band migrating at 14 kDa (Fig. 4A). A second weaker band migrating at 15.5-16 kDa was detected upon overexposure of the blots or when enough material was loaded on the gel (Fig. 5A). Treatment with PNGase F allowed resolving the whole emerging PK-resistant species in two major bands: a main 14 kDa species that had the same size than the C1 fragment and a larger less represented peptide above the 15 kDa molecular weight marker that will be further referred to as 16 kDa PrP res (Fig. 4B). This indicated that the 14 kDa and 16 kDa bands identified without PNGase treatment (Figs. 4A and 5A) were non-glycosylated native forms of ∆190-196 PrP res . The spontaneous emergence of PK-resistant ∆190-196 PrP was reproducible and occurred systematically from bulk cultures of ∆190-196 Rov obtained from three independent transfections (Fig.  5B). Individual ∆190-196 Rov clones obtained by limiting dilution spontaneously produced ∆190-196 PrP res , except for one clone, clone 12, despite expression of the mutant PrP to levels similar to those of other clones (Fig. 5C). This 'resistant' clone was useful for infection studies described in the following section and its existence suggested that currently unrecognized cellular factors are key for the spontaneous generation of ∆190-196 PrP res . ∆190-196 Rov cells could be frozen and thawed, without affecting the generation of PK-resistant PrP species. This characteristic together with persistence in cell culture recalled that of prion infected cells. We next examined whether the biochemical properties of ∆190-196 PrP Sc resembled those of prions passaged in WT Rov cells. ∆190-196 Rov lysates were treated with increasing PK concentrations and analyzed by Western blotting. ∆190-196 PrP Sc resisted to higher concentration of PK (Figs. 6A and 6B) than 127S prions propagated in WT Rov (Fig. 6C). ∆190-196 PrP res was recovered by centrifugation at 20,000 x g after PK digestion indicating that it was insoluble and aggregated (Fig. 5A). The aggregation size of ∆190-196 PrP Sc was determined by sedimentation velocity. ∆190-196 PrP Sc formed assemblies with a size in the range of PrP Sc assemblies formed by subfibrillar prions (Fig. 6D) according to previous reports (29,30). 127S PrP Sc assemblies from Rov cells had slightly larger assemblies with respect to size (Fig. 6D). Whether the difference is due to the number of PrP-mers composing the assemblies or 5 to the density of their main core remains to be determined. To summarize, introduction of the 190-196 deletion in RK13 cells favored the spontaneous and persistent production of PK-resistant PrP Sc species with an atypical electrophoretic pattern. ∆190-196 PrP spontaneously adopted a conformation that showed hallmarks of a prion: insolubility, aggregation, protease resistance and cell perpetuation. We thus called this entity ∆ Spont prion.

∆ Spont prion is infectious for cells expressing homologous or closely related mutant PrP
The infectious potential of ∆ Spont prions was primarily tested by cell-assay using ∆190-196 or WT Rov. As control, these cells were infected by 127S prions propagated in WT Rov. Naïve ∆190-196 Rov were susceptible to ∆190-196 lysates containing ∆ Spont as they produced PrP res in large amount as soon as the second passage postinfection (Fig. 7), while mock-infected cells did not. In contrast, WT Rov were not infectible with ∆ Spont prions (Fig. 7). Conversely, 127S could propagate in WT Rov but not in ∆190-196 Rov, at least for the first passages, the spontaneous emergence of ∆ Spont prions on long-term passage obscuring the fate of 127S infection. We next challenged the 'resistant' ∆190-196 cell clone 12 and found it readily susceptible to ∆ Spont prion infection. PrP res was detected as soon as the second passage and up to passage 8 post-infection (Fig. 8A). ∆ Spont prions were thus de novo infectious for cells expressing the homologous mutant protein.
The atypical profile of PrP res , with the characteristic presence of 14 kDa and 16 kDa bands was faithfully conserved upon infection. Previously, we established a set of Rov cells expressing PrP with deletions of different sizes in the C-terminus of helix H2. None of these cells produced spontaneously PK-resistant forms of mutant PrP (23). In particular, the ∆193-197 and ∆192-197 cells were susceptible to several ovine prions, including 127S. In contrast, the ∆190-197 cells were resistant to all of them (23). These three mutant cell lines were challenged with ∆ spont prions to determine whether ∆ Spont prion replication was strictly dependent on ∆190-196 PrP. Cells expressing PrP with 190-197, 192-197 or 193-197 deletions were all susceptible to ∆ Spont infection ( Fig. 8A). The ∆ Spont PrP res pattern was mostly maintained in the infected cells. We knew from our previous work that 127S prions propagated on ∆193-197 Rov cells were still infectious for WT Rov cells, indicating a structural compatibility between this mutant and WT PrP for 127S prion conversion. It was thus appealing to determine whether ∆ Spont propagated on ∆193-197 Rov cells could similarly become infectious for WT cells. The results showed that this was not the case, although ∆ Spont prions propagated on ∆193-197 Rov cells were de novo infectious for naïve ∆193-197 cells (Fig. 8B). Similar results were obtained for ∆ Spont propagated on ∆192-197 Rov cells: it was de novo infectious for ∆192-197 Rov cells and infectious for ∆193-197 Rov cells but not for WT Rov cells (Fig. 8C). Altogether, these cell-assays demonstrate the infectivity of ∆ Spont prions and their ability to propagate on cells expressing homologous ∆190-196 PrP or closely related mutants but not WT PrP. This suggests that a certain degree of compatibility in the H2 C-terminal sequence of PrP is required for conversion by ∆ Spont prions.

Molecular typing of ∆ Spont prions
Most prion strains can be grouped into two broad categories with respect to their molecular pattern, depending on the N-terminal endpoint of the C2 PrP res fragment. Type 1 strains produce PrP res species starting around position 85 of ovine PrP, while type 2 strains PrP res species are shorter, beginning at position 100. The size of unglycosylated C2 PrP res is thus a generic way to discriminate between strains. Therefore, we compared the electrophoretic profile of ∆ Spont PrP res with 127S type 1 prion strain propagated on WT Rov cells and with type 1 (T1 Ov ) and type 2 (T2 Ov ) strain propagated on ∆193-196 Rov cells, which express a mutant PrP with a size closer to ∆190-196 PrP (Fig. 9). Based on the aglycosylated lower bands, we could ascertain that the main 14 kDa band of ∆ spont prion was shorter than any type 1 or type 2 C2 fragment and that the16 kDa PK-resistant band was just slightly shorter than the C2 fragment of the type 2 strain. The PrP res electrophoretic profile of ∆ spont prions was thus clearly atypical with simultaneous presence of two major bands shorter than the usual PK-resistant C2 fragment from 'classical' prion strains.

6
To characterize in more details the nature of the 14 kDa and 16 kDa fragments from ∆ Spont PrP res , we performed an epitope mapping with anti-PrP antibodies spanning the entire PrP C-terminus, after deglycosylation of PrP res species. Signals produced by Sha31 and 8F9 mAbs were similar, indicating that ∆ Spont PrP res species are C-terminal fragments of PrP (Fig. 10). Regarding the major 14 kDa species, the band was not detected by 12B2 and 8G8 mAbs. Part of the 14 kDa band of higher molecular weight was recognized by 6C2 mAb and thus contained the full epitope of this antibody (114-HVAAAGA). The other part of lower molecular weight was 6C2 negative, indicating the loss of at least His-114 (Figs. 10 and S3). Therefore the 14 kDa band contained PrP res polypeptides with different Nterminal endpoints at the vicinity of the main cleavage site reported for the C1 fragment (8,31). In non-infected WT Rov cells and in ∆ Spont -free ∆190-196 Rov cells (early passage), part of the C1-PrP C fragment is 6C2-positive (Fig. S3), suggesting a certain variability in N-terminal endpoints. The PK-resistant 14 kDa band from ∆ Spont PrP res could thus result either or both from protease-digestion of the misfolded full-length ∆190-196 PrP or from a conformational change of the C1 fragment. Regarding the 16kDa band, the epitope mapping indicated a recognition by the 6C2 but not the 8G8 mAb. Therefore the N-terminus of these truncated PrP res polypeptides starts between residues inside the 8G8 epitope (position 100 to 105) and residue 113 at the head of 6C2 epitope, most likely close to the 8G8 border based on the molecular weight of these fragments. A faint and fuzzy band was also observed just above the 16 kDa band with all the mAbs used. This indicated the presence of few ∆ Spont PrP res species of larger size close to the C2 fragments from conventional prions. Altogether these observations show that ∆190-196 PrP turned spontaneously into a misfolded form producing a complex pattern of PK-resistant species with a main core around 14 kDa. These PrP res fragments of different size reflect the formation of different structures or assemblies by the misfolded protein. The prominence of the 14 kDa band raises the question of whether it could or not correspond to the minimum and necessary portion of ∆190-196 PrP for ∆ Spont replication as is the C2 fragment for classical prions. As a corollary this opens the unorthodox question of whether the mutant C1 fragment itself could misfold, either spontaneously or after conversion by ∆190-196 PrP Sc species produced from the full-length mutant PrP.

Conversion of mutant C1 by ∆ Spont prions
To examine the possibility that mutant ∆190-196 C1 PrP C (∆C1) alone can misfold into PrP Sc , either spontaneously or following infection by ∆ Spont prions, we generated Rov cells expressing solely ∆190-196 C1 PrP C . The C1 fragment is generated by PrP C cleavage at the alpha cleavage site, right upstream the N-terminal hydrophobic region. In human brain, the main C1 fragment starts at His-111 and was referred as to C1-upper but it can also start at Met-112 and was referred to as C1lower (8). Varying degrees of proteolytically processed C1 fragments at equivalent positions were suggested for ovine PrP (31). We thus built two constructs to express the C1 part of ovine PrP flanked by the N and C-terminal signal peptides of ovine PrP. The WT and ∆190-196 C1113 constructs were designed to start at residues Lys-113, one residue upstream His-114 the equivalent of His-111 in human PrP, to warrant both a correct and efficient cleavage of the N-terminal peptide signal and the presence of the full 6C2 epitope. The constructs referred as C1115 were designed to start at Val-115 to produce an equivalent to the C1-lower fragment. As with WT and ∆190-196 full-length PrP C , the C1 fragments were highly glycosylated in RK13 cells (Fig. 11A), but the unglycosylated forms were expressed in lower proportion in the two ∆190-196 mutants than in the WT counterparts. The C1 proteins had the expected size (Fig. 11B) and were present at the cell surface ( Fig. 11C and 11D), indicating that signal peptides were functional and processed by the cells. In contrast to full length ∆190-196 PrP C , mutant C1 PrP C did not spontaneously convert into a PKresistant form over passages, even after several months of cell culture (Fig. 12). However, upon exposure to ∆ Spont prions, these forms were converted into self-replicating PK-resistant PrP Sc (Fig. 12). These PrP Sc species were termed ∆C1 Sc . C1 conversion process was specific to ∆ Spont prions and needed the ∆190-196 deletion. 127S prions were not able to convert WT or ∆190-196 C1 PrP C . ∆ Spont prions were unable to convert WT C1 ( Fig.  7 12), as it was previously unable to do so for fulllength WT PrP (Fig. 7). As with WT PrP, a certain degree of sequence compatibility between the cellular substrate and the infecting prion is required for C1 conversion by ∆ Spont prion.

∆C1 Sc are de novo infectious and preserve ∆ Spont PrP res signature
We finally investigated the infectious potential of ∆C1113 Sc and ∆C1115 Sc . We used, as inocula, cell lysates of ∆ Spont -infected ∆190-196 C1 at passage 8 (see above). Controls were made to exclude any remnant infectivity of ∆ Spont prions or spontaneous infectivity of the homogenates from cells expressing mutant C1 PrP C . As shown in Figure 13, both ∆C1113 Sc and ∆C1115 Sc were de novo infectious for cells producing the homologous mutant ∆C1 proteins. In contrast cells expressing their WT counterparts were not susceptible to the infection. Remarkably, ∆C1 Sc exhibited the same activity as ∆190-196 PrP Sc with respect to the panel of susceptible cells. PrP C from 'resistant' ∆190-196 clone 12 and from the two closely related mutants ∆193-197 and ∆192-197 were converted into PKresistant PrP Sc . The western blot signature of PrP res in these cells recalled that produced following ∆ Spont prion infection (compare Fig. 13 and Fig. 8A). Not only the characteristic 14 kDa band, but also the 16 kDa band were present, indicating additional conversion of the full-length proteins by ∆C1 Sc . The main strain-specific determinant of ∆ Spont prions was thus enciphered in ∆C1113 Sc or ∆C1115 Sc .

Discussion
This work focuses on the spontaneous, in-cell conversion of a deletion mutant PrP into a novel form of prion. ∆ Spont prions showed three main original features: an internal deletion of seven residues in the protease-resistant core, an unusual PrP res signature and a remarkable capacity to propagate on the homologous C-terminal C1 segment of PrP, thus generating mutant C1 prions. In addition, this work provides a unique cell model to get insights in cell processes and factors associated with spontaneous prion emergence.

Spontaneous misfolding of ovine PrP with deletion in alpha-helix H2
Although spontaneous conversion of PrP into prion is a rare event, modifications in PrP can considerably increase this occurrence as not less than about forty disease-causing mutations are identified in inherited cases of human PrP (16). Spontaneous prion conversion also occurred in transgenic mice expressing some of these mutations (17)(18)(19)32,33). Conversion of equivalent mutant PrP in cell culture has so far been disappointing maybe in part because there is still no easy cell model for efficient human prion replication. We thus considered another approach that was to use a well-characterized reverse genetic model of prion replication and to remove from PrP highly conserved residues that might be important for PrP stability, while minimally affecting the overall structure. We previously identified the C-terminal region of the H2 alpha-helix as an area that meets these prerequisites. We found that the four contiguous threonine residues at the end of the helix were not necessary for replication of several ovine prion strains in RK13 cells (23). Deletion of the threonine cluster (∆193-196) even favored the replication of strains difficult to propagate in cells expressing wild-type ovine PrP (23). Thus, we considered this deletion as a good starting point to facilitate the emergence and propagation of a spontaneous prion. As ∆193-196 deletion was insufficient to cause spontaneous conversion of mutant PrP, we extended it to the three upstream residues His-190, Thr-191 and Val-192. Similar residues in human PrP are associated with diseasecausing mutations H187R, T188K and V189I responsible either for GSS or CJD (34-37), highlighting their potential importance for preservation of the normal form of the protein. We found that the simultaneous deletion of residues HTVTTTT maintained the overall PrP structure, but strongly reduced the stability of the recombinant ovine PrP. Reduced stability is expected to facilitate partial or complete unfolding of the protein and its spontaneous misfolding (38). Thus the ∆190-196 deletion might have introduced perturbations in the charge equilibrium, saltbridges and/or hydrophobic interactions. His-187 in human PrP, equivalent to His-190 in ovine PrP, is thought to be a key residue of the electrostatic network stabilizing the globular helical domain of PrP (26,39). Protonation of histidine is pHdependent and the positive charge acquired at acidic pH is thought to be involved in the 8 conformational shift of recombinant PrP in vitro (25,40,41). The disease-causing mutation H187R is one of the rare human mutations that markedly reduces PrP stability and this is attributed to replacement of histidine by a permanently positively charged residue (42). The deletion ∆190-196 not only removed His-190 residue but also brought Lys-197 in position 190, mimicking the replacement of His-190 by another permanently positively charged residue, recalling H187R mutation. This might have contributed to destabilization of ∆190-196 PrP and likely explains why the deletion ∆190-197, which eliminates the lysine residue did not lead to spontaneous prion generation, while the mutant protein remained convertible into PrP res by ∆ Spont prions. In addition, replacement of each or all the four contiguous threonines (193)(194)(195)(196) by different residues had also affected PrP stability (43,44). Any of these perturbations alone or in combination may have contributed to the spontaneous conversion of ∆190-196 PrP into ∆ Spont prions. The conservation of HTVTTTT sequence in mammalian PrPs might well reflect preservation of an important region for the dynamics and the maintenance of the helical folding of the protein in a cellular context.

Transmissibility of ∆ Spont prions
∆190-196 PrP was sensitive to protease digestion. Yet after several cell passages, a PK-resistant form systematically emerged, indicating spontaneous conformational change of the mutant PrP. The spontaneous formation of PK-resistant PrP was verified in one occasion by an experiment of cell transfection and culture entirely carried out in prion free laboratory, excluding inadvertent contamination. The ∆190-196, or ∆ Spont PrP Sc entity as we called it, was shown to be a self-propagating, protease-resistant, insoluble and aggregated form that was transmissible to other cells. ∆ Spont thus showed all the usual hallmarks of prion replication in cell culture. A basic characteristic of prions is their ability to transmit their own conformational state to homologous PrP C and in some instance to heterologous PrP C . Persistence of ∆190-196 PrP Sc in cell culture indicated some replication but as we faced a spontaneous conversion that might be reiterated at each passage, we demonstrated de novo infectivity of ∆ Spont prion in several cell-assays: iearlier accumulation of ∆190-196 PrP Sc in exposed cells; ii-infection of the ∆190-196 Rov clone 12 that did not produce spontaneously a PK-resistant form of the mutant protein; iii-infection of Rov cells expressing the ∆190-196 C1 PrP C ; and ivinfection of Rov cells expressing three other mutant PrPs with close deletions in H2 C-terminus, notably the ∆190-197 mutant that was refractory to infection by conventional ovine prion strains (23). Sequence proximity and/or structural adaptability between ∆ Spont and ∆190-197 PrP might explain it. In contrast, conversion of WT PrP by ∆ Spont prions failed despite many attempts and the use of different cell populations including the historic Rov clone 9 (27) or more susceptible clones (45). In a previous work, we found that the 127S ovine prion strain propagated on WT PrP was easily propagated on ∆193-197 Rov cells and reciprocally (23), suggesting a good structural compatibility for conversion between the ∆193-197 mutant and WT PrP when this strain is concerned. Here, ∆ Spont prion propagated on ∆193-197 PrP was still not able to convert WT PrP, indicating that the ∆193-197 PrP Sc structure transmitted by ∆ Spont prion rather than only sequence of the mutant protein causes the transmission barrier.

Biochemical specificities of ∆ Spont PrP Sc
The PrP res pattern of ∆ Spont was faithfully maintained in the different cell assays. The pattern was complex and showed differentially represented bands corresponding to N-terminal truncations at different end points. It did not fit with any known prion strains regarding the 14 kDa and 16 kDa bands. The minor 16 kDa band was slightly more N-terminally truncated than the classic C2 fragment from type 2 prions. The major 14 kDa band had a size close or similar to that of the C1 fragment of ∆190-196 PrP. These two bands were found in individual clones of ∆190-196 Rov cells with similar relative proportion as in populations of transfected cells. A third, minor, 8G8/12B2 positive band was identified that might be close or similar to that of more classical type1 prions. Thus ∆ Spont PrP res included fragments close to C2 of type 1 and type 2 plus prominent C1-like species. C2-like protease-resistant fragments are most probably truncated forms of full-length mutant ∆190-196 PrP Sc , while C1-like PrP res species might result from conversion of either or both the full-length mutant PrP and its C1 fragment. The unconventional PrP res 9 profile of ∆ Spont prion strongly suggests an original structural organization but might be complicated by conversion of two substrates, the full-length protein and its C1 fragment, rather than only full-length PrP for more classical prions.

Generation of ∆190-196 C1 prion
PrP C is naturally cleaved into a C-terminal C1 fragment and its complementary N-terminal N1 at the alpha cleavage site. This cleavage, the efficacy of which is cell-dependent, was attributed to metalloproteases of the ADAM family, but the exact nature of the enzyme remains a subject of controversy and may depend on tissues or cell lines considered (46)(47)(48)(49)(50). Moreover, the cleavage is not dependent on a specific sequence and whether it occurs at the cell surface or inside the cell between the Golgi and the plasma membrane is not clearly established (51,52). In Rov cells, the C1 fragment appears as a relatively large N-terminally truncated band, beginning around position His-114 or Val-115 equivalent to that determined in human brain (8,31). A consensus is that the C1 fragment of PrP C is shorter than the PrP res domain and thus cannot be converted into prion. Furthermore, an apparent inverse correlation between C1 levels and cell susceptibility to prions was reported in cell lines as well as a dominant negative effect on prion replication in transgenic mice overexpressing C1 (14,53). Thus, the C1 fragment is often considered as a competitive inhibitor of PrP C conversion. We found here different outcomes. The main PrP res domain of ∆ Spont prions had a size close or similar to that of C1, and ∆ Spont prions were able to convert the mutant form of C1. Protease resistant ∆C1 Sc showed hallmarks of prion replication in cell culture and was transmissible de novo to cells expressing either the homologous C1 protein, the homologous full-length mutant PrP or closely related mutants. To the best of our knowledge there is no other report of confirmed conversion of C1 or C1-like polypeptides. Two presumptive bovine spongiform encephalopathy cases with C1-like PrP res signature were reported but they were finally found to lack prion infectivity after inoculation of brain material to cattle and to bovine PrP transgenic mice (54,55). The exceptional conversion of ∆C1 into prion after infection by the unconventional ∆ Spont prion is likely due to the conjunction of the sequence modification introduced in the polypeptide and adoption of a misfolded structure significantly different from that of the other prions. ∆C1 Sc as a driving force for the propagation of the ∆ Spont prion ∆C1 prions had templating activity on the fulllength mutant proteins and preserved the PrP res signature of ∆ Spont prion. Indeed, infection by ∆C1 prions produced not only C1-like PrP res but also C2like 16 kDa species that were larger than C1 and thus resulted from conversion of the full-length protein. ∆C1 Sc therefore behaves like a prion strain which maintains the structural information associated with the specific biochemical profile of ∆ Spont prion. This strongly suggests that the 14 kDa ∆C1-like PrP res core produced by spontaneous conversion of ∆190-196 PrP has at least contributed to ∆ Spont prion propagation and might even be the most important template for the normally folded mutant protein.
As ∆C1 did not spontaneously form a prion, the primary event in ∆ Spont prion formation may be the spontaneous conversion of full-length PrP. Then the reciprocal capacity of ∆190-196 PrP Sc to induce conversion of ∆C1 fragment and of ∆C1 Sc to induce full-length mutant PrP conversion would lead to ∆ Spont complex pattern. It was unexpected that the short ∆C1115 or ∆C1113 fragments, with lengths of 113 and 115 residues, respectively could be sufficient to adopt a specific transmissible prion structure, as the lack of WT C1 convertibility into prion was explained by its shortened size compared to C2 PrP res fragments. Whether ∆C1 prion adopts a structure close to C2 but without the 15 or 30 N-terminal residues or a different structural organization remains an unresolved question. ∆ Spont and even more ∆C1 prions recall somehow PrP106, a mouse PrP with a double deletion (∆23-88, ∆141-176), that produced a miniprion following infection by RML strain, but not spontaneously (56). The sequence of this mutant protein was very different from that of ∆190-196 ovine PrP or C1. The sequence of PrP106 contained about 20 residues upstream of the mouse C1 segment and a large internal deletion that removed both H1, the beta2 strand and five Nterminal residues of H2 (56), together with the loops between H1 and H2 that are considered to be highly important for prion conversion (57)(58)(59). While structurally different from ∆C1 prion, the miniprion is another example of prion entity 10 constituted from elements of the PrP sequence. A third more distant example of short PrP sequence associated with prion formation is the human PrP with Y145-Stop mutation associated with GSS (60). In contrast to the two precedent mutants, PrP Y145-Stop conserves only the N-terminal moiety of PrP and lacks the whole globular helical domain. ∆ Spont and ∆190-196 C1 prions generated in this work are therefore original and attractive prion entities which deserve to be studied further with transgenic mouse models in the future.

Conclusion
We report the generation of a novel spontaneous mutant prion propagating in cell culture and conversion into prion of both the full-length and the C-terminal C1 fragment of ∆190-196 mutant PrP. We demonstrated that only 113 or 115 residues of the PrP C-terminus are sufficient to constitute a self-replicating and transmissible prion entity. Our results also suggest also that the HTVTTTT conserved sequence in the H2 C-terminus of PrP is important for prion protein stability and that removal of H2 C-terminal residues is required for productive infection in cells challenged by the spontaneous prion. This work also provides a unique cell culture model for spontaneous prion formation to further study cell factors and molecular processes involved.

Cell culture and isolation of Rov cells
Rov cells are epithelial RK13 cells that stably express either wild-type or mutant ovine PrP in an inducible manner, by using a tetracycline-inducible system (27). They were obtained by transfection of cells by lipofectamine and puromycin selection. We used cell populations produced by a pool of puromycin-resistant cells, unless indicated otherwise. In some occasions, we used individual clones that were obtained by serial dilution of transfected cells in presence of the selecting agent. Cells were grown in Opti-MEM medium (Invitrogen) supplemented with 10% fetal calf serum (FCS) and antibiotics, and split at 1:4 after trypsin dissociation once a week. To express fulllength PrP C or C-terminal C1 polypeptides, cells were cultivated in the continuous presence of 1µg/ml of doxycycline (Sigma).

Prions and prion strains
The spontaneous mutant prion ∆ Spont was propagated on Rov cells expressing the ∆190-196 PrP mutant unless stated otherwise. The 127S ovine prion strain was isolated through serial transmission and cloning by limiting dilution of PG127 field scrapie isolate to tg338 transgenic mice expressing the VRQ allele of ovine PrP (66,67). 127S was propagated on Rov cells expressing the WT PrP for comparative infection test or determination of PrP res profile. T1 Ov and T2 Ov prions were originally isolated from serial transmission of a human sporadic CJD case (MM2 type) to tg338 mice (45). For comparison of PrP res profiles with ∆ Spont , T1 Ov and T2 Ov strains were previously propagated on Rov cells expressing ∆193-196 mutant PrP (23).

Prion infection of cell cultures
To test the infectivity of cell cultures, cells were pelleted, frozen and thawed three times, sonicated three times for 30 seconds and the resulting homogenates were used as inocula to infect naive cell cultures as previously described (23,68). Homogenates were left three days on the challenged cells. They were then washed with PBS, trypsinized and seeded at 1/10 dilution in fresh culture medium. Cells were then split at 1:4 dilution after one week of culture as for each other successive passage. For infection by ∆ Spont prion, homogenates of ∆190-196 Rov cells harvested at least 9 passages after the addition of doxycycline were used. To test for infectivity of ∆ Spont propagated on Rov cells expressing other deletion mutant PrP or C-terminal C1 polypeptides, inocula were made from cells harvested 8 passages post exposure.

Cell lysis, protease digestion, and PNGase F treatment
Cells were washed twice with cold phosphatebuffered saline and whole-cell lysates were prepared in TL1 buffer (50 mM Tris-HCl [pH 7.4], 0.5% sodium deoxycholate, 0.5% Triton X-100). Lysates were clarified by centrifugation for 2 min at 800g, and protein concentrations were determined by microBCA assay (Pierce). For PrP res , lysates were incubated with 8µg of PK per 1 mg of protein for 2 h at 37°C and then centrifuged for 30 min at 22,000 x g. Pellets were dissolved in Laemmli sample buffer and boiled for 15 min at 100°C. When needed, 500 U of PNGase F (New England BioLabs, Massachusetts) and 1% Nonidet P-40 were added to denatured proteins that were further incubated at 37°C overnight.

Sedimentation velocity fractionation
Experiments were performed as previously described (30). Briefly, cells were solubilized by addition of a buffer containing 4% (w/v) dodecylβ-D maltoside and benzonase (0.4 unit/µl). After incubation for 30 min at 37°C, sarkosyl (N-lauryl sarcosine) was added to give a final concentration of 2% (w/v) in the samples. The incubation was pursued for 30 min at 37°C. 150 µl of solubilized samples were carefully loaded on a 4.8 ml continuous 10-25% iodixanol gradient (Optiprep, Sigma-Aldrich), with a final concentration of 25 mM HEPES pH 7.4, 150 mM NaCl, 2 mM EDTA, 0.5% Sarkosyl. The gradients were centrifuged at 285 000 g at 4°C for 45 min in a swinging-bucket SW-55 rotor. 30 fractions of 160µL were collected and PK-treated at a final concentration of 50 µg/ml for 1 hour at 37°C. PrP res contents were analyzed by Western blotting using a biotinylated anti-PrP Sha31 monoclonal antibody. Signal intensities were quantified using ImageLab software (Bio Rad) and converted into arbitrary units after normalization. A fixed quantity of human recombinant PrP was employed to calibrate the PrP signals in different gels.

Immunoblotting and detection of PrP C and PrP res
Either 4 to 12% NuPage Bis-Tris precast polyacrylamide gels (Invitrogen) or 12% Criterion XT Bis-Tris gels (Bio-Rad) were used for sodium dodecyl sulfate polyacrylamide gel electrophoresis. For PrP C analysis, 50 µg of protein per sample was loaded on the gel. For PrP res , unless otherwise indicated, the samples corresponding to PKresistant PrP contained in 25 or 50 µg of cell lysate protein were loaded onto the gel. The transfer of proteins, their detection, and their revelation were described previously (28,69).

Immunofluorescence, image acquisition and treatment
Cells were grown on plastic dishes or on glass coverslips in regular medium, washed twice with PBS before fixation for 10 minutes with 4% paraformaldehyde. Cells were then washed and incubated with the required monoclonal primary antibody (4F2 or Sha31, at 1:5000) in a blocking reagent buffer containing 0.5 % crystallin (Roche diagnostic) and 0.1% Tween 20 in PBS. After washing, cells were incubated with secondary Alexa Fluor 488-conjugated anti-mouse IgG antibodies (Molecular probes, Invitrogen) used at a 12 1:500 dilution, as previously reported (28,70) and nuclei were stained with 4',6-diamidino-2phenylindole (DAPI). To specifically label the cell surface, Rhodamine-conjugated wheat germ agglutinin (WGA, Invitrogen) was incubated for 5 minutes with living cells, washed once with PBS and cells were further fixed and processed as described above. Images were acquired either with an Axio observer Z1 microscope (Zeiss) equipped with a CoolSnap HQ2 camera (Photometrics) and driven by the Axio-vision imaging system software.

Expression and purification of recombinant PrP
Recombinant proteins were produced and purified from Escherichia coli as published previously (23). Briefly, by site-directed mutagenesis the deletion ∆190-196 was introduced inside the sequence of the full-length ovine PrP (residues 25 to 234, VRQ allele) cloned into a pET28 expression vector. Wild-type and mutant proteins produced by E. coli were purified by immobilized metal affinity chromatography (IMAC) on Nickel columns.

Circular dichroism
The secondary structure of recombinant PrP produced by E. coli was analyzed by circular dichroism (CD). Measurements were carried out on Jasco-810 spectropolarimeter. Far-UV CD spectra of full length ovine PrP and the deletion mutant 190-196 were recorded from 260 to 180 nm at 25ºC in 1 µm path-length quartz cuvette at a protein concentration of 50 µM in 10 mM Na-acetate buffer at pH 5.0. Each CD spectrum was obtained by averaging 6 scans collected at a scan rate of 200 nm/min. Baseline spectra obtained with buffer were subtracted for all spectra.

Nuclear Magnetic Resonance
2D 1 H-15 N HSQC and 3D HNCA spectra of 250 µM recombinant 15 N 13 C-labeled ∆190-196 C1113 in 10 mM Na acetate pH 5 buffer were acquired on a Bruker NMR AVANCE III spectrometer equipped with a cryoprobe at a magnetic field of 18.8 T and a temperature of 298 K. 13 Calpha chemical shifts were analyzed by TALOS-N software by excluding similar sequences (71)

Fluorescence-based thermal shift assay
Reaction mixtures containing 10 or 20 µM of recombinant PrP in 10 mM Na acetate pH 5.0 and SYPRO orange (diluted 500-fold from a 5000-fold stock solution; Invitrogen), were made in duplicate in a 96-well fast PCR plate at a final volume of 20 µL. The experiments were also reproduced in buffer 250 mM Na Phosphate pH 5.1 (supporting information Fig. S2). The temperature gradient was carried out in the range of 10 °C to 95 °C, at 3 °C/min with a StepOnePlus real-time PCR system (Applied Biosystems) as previously described (72). Fluorescence was recorded as a function of temperature in real time (excitation with a blue LED source and emission filtered through a 5carboxy-X-rhodamine [ROX] emission filter). The melting temperature (Tm) was calculated with the StepOne software v1.3 (Applied Biosystems) as the maximum of the derivative of the resulting SYPRO Orange fluorescence curves.

Data availability
All data in this study are contained within the manuscript. Figure 1. ∆190-196 PrP sequence. On the top is the schematic representation of mature ovine PrP C (23 to 234). Secondary structures building the globular part of the protein, the two short beta strands forming a beta sheet and the three alpha-helices are indicated. Post-translational modifications such as N-glycan chains (black dots), disulfide bridge (S-S) and the GPI anchor are also shown. Below, the amino acid sequence of alpha-helix H2 is highlighted in lavender color and the first residues of H3 in purple. Residues 190-196 that were removed from the WT PrP are colored in red and replaced by a dotted line for the deletion mutant.         (73) is indicated by horizontal double arrow dotted line. PrP res fragments resulting from PK digestion of ∆ Spont prions are represented below the scheme, the 14kDa C1-like PrP res is in red, the 16kDa in orange and the faint 17kDa C2-like species in pink. B. Representative western blots of PNGase F-treated ∆ Spont PrP res and WT PrP res revealed with the different anti-PrP mAbs. For the comparative analysis, the same couple of PrP res samples was loaded several times on a same gel, separated by stained molecular-weight markers and then transferred on a membrane that was split into six parts, each being incubated with a different primary antibody, as indicated. Figure 11. Expression of WT and ∆190-196 C1 polypeptides in RK13 cells. A. Immunoblot analysis of stably transfected C1 constructs, before or after PK treatment. The expression of C1-like polypeptides (residues 113-234 and 115-234 of ovine PrP, C1113 and C1115, respectively) and equivalent fragments with the HTVTTTT ∆190-196 deletion (∆C1113 and ∆C1115) was analyzed. B. Immunoblot analysis of the same C1 constructs before and after PNGase F treatment. 25µg of cellular protein was loaded. C. C1 PrP C expression pattern by immunofluorescence. PrP C (green) and nuclear marker 4',6-diamidino-2phenylindole (DAPI, blue) staining of non-permeabilized fixed cells. Scale bars 50µM. D. Confocal microscopy imaging of ∆C1113 Rov cells co-stained for white germ agglutinin (WGA, red) and PrP (green). Nuclei are stained with DAPI (blue). Merged confocal images or individual channels are shown. Scale bar 10µM.  . Infectivity of ∆C1 Sc prions. Rov cells expressing different forms of mutated, full-length PrP and WT/mutated C1 PrP were left uninfected (none) or exposed to ∆C1 Sc (left panel, ∆C1113 Sc ; right panel ∆C1115 Sc ) obtained from cell homogenates at passage 8 (see figure 11). As controls, non-infected cells (ni ∆C1113, ni ∆C1115) and ∆ Spont -infected RK13 cells (RK13 ctrl) were used. The challenged cells were ∆190-196 Rov cell clone 12 that did not produce spontaneously ∆ Spont prions, Rov cells expressing closely related full-length deletion mutants (∆193-197, ∆192-196), populations of cells expressing either WT or ∆ version of C1113 (∆C1113) and of C1115 (∆C1115). Immunoblots of PK-treated cell lysates at passage 8 post-infection are shown (Sha31 mAb).