Loss of the first β-strand of human prion protein generates an aggregation-competent partially “open” form

Prion diseases, a group of incurable, lethal neurodegenerative disorders of mammals including humans, are caused by prions, assemblies of misfolded host prion protein (PrP). The pathway of PrP misfolding is still unclear, though previous data indicate the presence of a structural core in cellular PrP (PrPC), whose cooperative unfolding presents a substantial energy barrier on the path to prion formation. PrP is a GPI-anchored membrane protein, and a number of studies suggest that membrane interactions play an important role in the conversion of PrPC to its disease-associated form, including a transmembrane form of PrP in which a highly conserved region (residues 110 - 136) spans the ER membrane. Insertion of this region results in the detachment of the PrPC first β-strand from the structural core. The effect of this removal on the structure, stability and self-association of the folded domain of PrPC is determined here through a biophysical characterisation of a truncated form of PrPC lacking this region. Whilst markedly destabilised, NMR chemical shifts show that the truncated protein exhibits tertiary structure characteristic of a fully folded protein and retains its native secondary structure elements, including the second strand of the PrP β-sheet, but with altered conformational flexibility in the β2-α2 loop and first α-helix. The latter is destabilised relative to the other helical regions of the protein, with markedly increased solvent exposure. This truncated form of PrP fibrilises more readily than the native form of the protein. These data suggest a stepwise mechanism, in which a destabilised “open” form of PrPC may be a key intermediate in the refolding to the fibrillar, pathogenic form of the protein.

PrP C is a widely-expressed cell surface, glycosylphosphatidylinositol (GPI)-anchored glycoprotein that is sensitive to protease treatment and is soluble in detergents. PrP may have a role in cell adhesion or signalling processes, but its precise cellular function remains unknown. It consists of two structural domains, an unstructured N-terminus, spanning approximately residues 23 -125, which contains five repeats consisting of a single nonapeptide and four octapeptides, and a structured, mainly α-helical C-terminal domain, which includes a single disulphide bond and two glycosylation sites (1,14,15). A highly conserved section, (approximately residues 110-136) referred to as the conserved hydrophobic region (CHR) spans both domains. Prions, in contrast, are dominated by βsheet structure and are insoluble in detergents, with some displaying marked protease-resistance.
These have been classically designated as PrP "Scrapie" (PrP Sc ) and are found only in prion-infected tissue (16).
Prion diseases may be acquired (transmitted between animals or humans), inherited or sporadic (of unknown cause). Inherited prion diseases comprise around 10-15% of total prion disease cases, with over 30 different pathogenic somatic mutations and approximately 10 polymorphic variants having been identified in the human PrP gene (PRNP) (17). Amyloid formation in a range of proteins is associated with destabilisation of the protein native state through somatic mutations (18)(19)(20)(21).
Although destabilisation of PrP C does not correlate with specific disease phenotypes and no definitive link between PrP stability and disease has been established, the majority of PRNP pathogenic mutations reside within the structured C-terminal domain, with all fully penetrant pathogenic mutations showing significant destabilisation (22). In vitro experiments also show that destabilisation of native PrP C through the use of chemical denaturants, temperature, redox conditions, pH or nucleic acid-binding favours the formation of protease-resistant amyloid or fibrillar structures, some of which have been claimed to be associated with disease (23)(24)(25). The most general model proposed thus far for the process of nucleated protein polymerisation is that PrP C fluctuates between its dominant native state and other minor isoforms. These can self-associate in an ordered manner to produce a stable supramolecular structure composed of misfolded PrP monomers, which can convert other isoforms to the infectious isomer in an autocatalytic manner (2).
Many studies have thus focused on characterising partially folded intermediate states which may be involved in PrP self-association. A key observation in this regard is that the three α-helices and the second strand of the PrP C β-sheet form a stable assembly, which may be considered to be the core of the protein. This is because these structural elements have stabilities equivalent to the overall fold stability of the protein and the backbone amides of these structural elements exchange with solvent solely when PrP C completely unfolds (26). Minor isoforms which retain this structural core are thus energetically favoured over those in which this core region is disrupted. Notably, the first strand of the PrP β-sheet, which is also the most N-terminal structural element in the folded C-terminal domain, lies outside of this core region and displays markedly reduced stability in comparison to the other PrP secondary structure elements (approximately 30-fold less) (26). This raises the possibility that this secondary structure element may act independently of the core of PrP C , and indeed NMR chemical shift data indicate that the PrP C native state ensemble contains a population of conformers where the PrP N-terminus, including the first strand of the PrP C β-sheet, is detached from the PrP C core (27).
The detached region comprises up to residue 146, and includes the conserved hydrophobic region of PrP (CHR), which spans residues 110 -136. The CHR displays exceptionally high conservation across a wide range of species, indicative of an essential role in the endogenous function of PrP C (28,29).
The CHR controls PrP co-translational translocation at the endoplasmic reticulum (ER) during the biosynthesis of the protein (30). It can associate with the ER membrane in various topologies, with mutations in this region upregulating particular PrP isoforms, which can lead to neurodegenerative disease in mice and some heritable prion diseases (31). Its amino acid sequence also displays many of the characteristics of a transmembrane helix (32), and NMR studies show that residues within this region interact with membrane analogues (dodecylphosphocholine micelles) in aqueous solution both as part of PrP C (33) and as a distinct peptide (29). Here, it adopts a membrane-spanning helical conformation when associated with lipid micelles (29) and a radically different conformation to the reported solution structures of PrP C . In addition, antibody binding to the CHR, which would inhibit membrane insertion, also inhibits the propagation of proteinase K-resistant PrP Sc and prion infectivity, further implicating the CHR and its membrane association in prion disease (34).
Competitive antibody binding to the CHR has been proposed to block interactions between PrP and cofactor molecules thereby inhibiting PrP conversion and prion propagation (34).
Insertion of the CHR into the endoplasmic reticulum (ER) membrane necessarily involves detachment of residues 110 -136 from the folded C-terminal domain, including the first strand of the PrP β-sheet, from the PrP C core. Whilst the proportion of PrP molecules with a detached CHR region is low in solution (27), a high local concentration of membrane-binding sites for membraneincorporated PrP would be available, likely resulting in a significant bound population. Thus, it is reasonable to consider whether these membrane-associated forms of PrP C may destabilise the PrP C core region and facilitate its self-association. Our aim in this study was to investigate the effect of the removal of the CHR on the structure, stability and self-association of the remaining folded domain of PrP, through the biophysical characterisation of truncated PrP C lacking the CHR. We find that this form of the protein whilst retaining many of the characteristics and structure elements of native PrP C , has substantially increased solvent exposure relative to the native state, and is significantly more susceptible to fibrillisation, raising the possibility that this partially-folded state may be a principal precursor in the formation of ordered fibrillar structures.

Choice of PrP constructs studied
In order to assess the effects of removal of the CHR on the structured domain of PrP C we generated a recombinant PrP construct spanning residues 137-231, which we termed PrP 137 . Data for this construct was primarily compared with that of PrP 119 (residues 119-231 of PrP), which contains the structured elements of PrP C , and does not contain an extended unstructured N-terminus, which complicates the structural analyses. is broadly similar to that of PrP 119 , with two minima at approximately 208 and 222 nm, characteristic of a predominantly α-helical protein. The loss of β-sheet signal resulting from the loss of residues 119-136 is masked by the significantly more intense α-helical signal. PrP 137 thus appears to retain the mainly α-helical structure of the untruncated C-terminal domain. There is some reduction in the mean residue ellipticity (MRE) for the minimum at 208 nm, which is suggestive of some alteration in the packing of the α-helices however.
Both 1D and 2D NMR spectra of PrP 137 display a broad range of dispersed NMR signals ( Fig. 2 (b/c)).
The wide chemical shift dispersion indicates extensive tertiary organisation, and tight packing of amino acid side chains in the protein core characteristic of a fully folded, native-like protein. This is significant, as in so-called 'molten globule' intermediate states, truncated or destabilised proteins can display native-like secondary structure content, and compact hydrodynamic radius, despite the absence of well-defined tertiary structure (35). The characteristics found in molten globule states have been found in the transient intermediate states found during the folding of certain proteins (36). However, PrP 137 NMR spectra are markedly different to PrP 119 , indicating considerable perturbation of the folded domain, in particular residues in close proximity to the N-terminal deletion ( Fig. 2 (c)).

Structural perturbations in PrP 137 resulting from loss of CHR
The full extent of the structural perturbations resulting from the loss of the CHR was determined through analysis of backbone NMR chemical shifts. Differences between observed chemical shifts and their corresponding random coil (unstructured) values, in particular in Cα and carbonyl resonances, are highly correlated with protein secondary structure (37,38). Chemical shift data also provides accurate, quantitative and site-specific mapping of protein backbone mobility as well as Modelfree backbone order parameters (39,40).
NMR resonance chemical shifts (Cα/C'/Cβ/N/H N ) show that the backbone conformation of PrP 137 is similar to PrP 119 , with all three α-helices retained ( Fig. 3) (37,38). Remarkably, although some loss of structure is observed, the second β-strand also retains clearly observable extended structure despite the loss of the complementary β-strand and its stabilising hydrogen bonding contacts (Fig. 3). A notable change however is an increase of helical structure in the loop linking the second strand of the β-sheet and helix 2, the so-called β2-α2 loop (residues 165-172; Fig. 3). This region, which has been shown to affect prion cross-species transmissibility (41)(42)(43)(44)(45), is adjacent to the β-sheet, packing against residues N-terminal to the first β-strand, the β-sheet itself, and the C-terminus of helix 3 ( Fig.   1). This structural perturbation appears to have propagated to the C-terminus of helix 3 (residues 222 -227) which shows some indication of a loss of α-helicity, as does the C-terminus of helix 1 (residues 144 -154) (Fig. 3).
These alterations in secondary structure propensity are associated with changes in dynamics. For instance the β2-α2 loop displays a marked reduction in the Modelfree order parameter (S 2 ) ( Fig. 3 (c)) (39,40). S 2 reports on internal sub-nanosecond (ns) motions, and values range from 0 for highly flexible to 1 for rigid systems. The helical regions of PrP variants exhibit S 2 values of 0.8 -0.9, typical of structured regions of folded proteins. Residues of the α2-α3 loop (residues 194-199) and the Cterminus of helix 3 display reduced S 2 values, reflecting increased flexibility, commonly observed in loop regions of globular proteins (46,47). The reduction in S 2 for residues 164-168 in PrP 137 would suggest that the β2-α2 loop is more flexible in PrP 137 than PrP 119 . This would appear to be at odds with the predicted increase in α-helicity. However, this loop is subject to millisecond (ms) timescale motions proposed to be associated with a large-scale co-operative conformational change between a 3 10 -helix and a type I β-turn (48,49). Millisecond timescale motions can result in line-broadening of NMR signals and anomalous order parameter values (46,47). Loss of the CHR in PrP 137 does appear to alter the conformational dynamics of the β2-α2 loop and increase the population of the 3 10 helix.
The C-terminus of helix 3 (residue 219 -230) also displays reduced S 2 values (Fig. 3 (c)). A number of residues in this region of the protein interact closely with the β2-α2 loop. For example, Y218 and S222 with M166, and I215 and Y218 with Q172. Perturbation of the dynamics of the β2-α2 loop has previously been shown to affect the dynamics of the C-terminus of helix 3 (47), and in PrP 137 the loss of the CHR and contacts with the β-sheet appear to have propagated beyond the immediate site of the truncation. In addition, the C-terminus of helix 1 and following residues (residues 153-158) also display an increased level of flexibility ( Fig. 3 (c)).

PrP 137 stability
Despite the relative lack of secondary structure perturbations observed, removal of the CHR and the first β-strand resulted in a marked reduction of approximately 2 kcal mol -1 in the thermodynamic stability of PrP 137 ( Fig. 4 (b)). This equates to an approximately 20-fold reduction in the equilibrium between the native and unfolded states (K (N/U) ~3500) when compared to PrP 119 . As with PrP 119 , loss of secondary structure upon equilibrium denaturation in PrP 137 occurs in a single co-operative transition, without the formation of any populated intermediate species (Fig. 4 (a)). This unfolding transition is markedly less cooperative than in PrP 119 (27), but distinct from unfolding transitions demonstrated by molten globule states. The calculated m-values, which describe the sensitivity of the folded / unfolded state equilibrium to denaturant, and reflect the increase in solvent exposure of the hydrophobic core as the protein unfolds show that that the degree of hydrophobic exposure in PrP 137 on unfolding is significantly reduced in comparison to PrP 119 (Fig. 4 (c)).
In common with the denaturant-induced unfolding, the thermal unfolding of PrP 137 consisted of a single co-operative transition, without the formation of any populated intermediate species (Fig. 4 (d)). Both equilibrium unfolding transitions (GuHCl-or thermally-induced), were found to be completely reversible (Fig. 4 (d/f)). The thermal stability of PrP 137 was also markedly reduced in comparison with PrP 119 , with a significant (≈10 o C) reduction in the mid-point for thermal unfolding ( Fig. 4 (e)).

Stability of secondary structure elements and solvent accessibility.
An even more sensitive measure of local stability is hydrogen/deuterium exchange of backbone amides, observed by the decay of NMR amide signals. As with PrP 119 , protected amides in PrP 137 were located within the structured elements of the protein, in this case predominantly within helices 2 and 3 (Fig. 5). Protection factors (PF) for amides in these regions were equivalent to the free energy change for unfolding (i.e. PF = K (N/U) ). This behaviour is observed in PrP 119 , and also a number of other proteins (e.g. Barnase/Staphylococcal Nuclease) (50), where a substantial proportion of core residues can exchange only in the fully unfolded state. Protection factors for the second strand of the PrP β-sheet of PrP 137 were however below detectable levels. This is consistent with the loss of the first strand of the β-sheet and its hydrogen-bonding interactions with the second β-strand.
However, it was observed that the first α-helix in PrP 137 displayed markedly less protection than the remaining α-helices. In PrP 137 it does not display measurable protection (the lower limit for detection of protection factors was approximately 200), whereas in PrP 119 it has amide protection factors comparable to the other secondary structure elements, and equivalent to the equilibrium constant for unfolding ( Fig. 5 (a)). In order to extend the range of measurable protection factors we employed the CLEANEX-PM methodology, which uses phase-modulated chemical exchange NMR to specifically monitor water-amide proton exchange, allowing measurement of exchange rates of rapidly exchanging amide protons (51). This extended the number of measurable amide protons to include residues at the C-termini of helices 2 and 3, and two residues at the N-terminus of helix 1 (Fig. 5 (b)).
These displayed protection factors of 350 -450, approximately 10-fold lower than the average protection factors observed for helices 2 & 3. The remaining residues in helix 1 remained below the detectable limit, indicating that deletion of the CHR has selectively destabilised helix 1.

Propensity of PrP 137 to oligomerise
Given the increased exposure of hydrophobic residues, as indicated by the increased m value derived from the equilibrium denaturation data and the reduced hydrogen protection for the first αhelix and 2 nd β-strand, we sought to determine whether this increased the propensity of PrP 137 to self-associate. This was initially done using sedimentation velocity analytical ultracentrifugation (SV-AUC) (Fig. 6). A single species with a sedimentation coefficient of 1.6S, characterised by a frictional ratio (f/f0) of 1.25, giving a molecular mass of 11.8 kDa was observed for PrP 137 at pH 7.5.
Corresponding values at pH 6.5 were 1.67S, 1.23 and 11.95 kDa respectively. These molecular masses match closely the expected molecular mass for the PrP 137 monomeric protein (Fig. 6). For PrP 119 a single species was also observed at pH 6.5 and 7.5 with a molecular mass corresponding with those of the monomeric protein (13.6 kDa). PrP 137 displays a slightly reduced frictional ratio relative to PrP 119 indicating a more spherical character, consistent with removal of the relatively unstructured N-terminus (residues 119 -125) of PrP 119 . Thus it appeared that under the conditions used, PrP 137 is no more susceptible to oligomerisation than PrP 119 , despite the increased exposure of hydrophobic residues. This conclusion was tested further over a range of protein concentrations (15 -135 μM) at pH 6.5 and 7.5, with no multimeric species observed (Fig. 6).

Propensity of PrP 137 to fibrilise
When agitated at 42 o C under native conditions, PrP can be induced to form amyloid (52). Binding of the fluorescent thiazole dye thioflavin T to these β-sheet-rich fibrillar structures reports their formation, allowing a quantitative analysis of the kinetics of fibril formation (53). Although no association of native PrP 137 monomers was observed in solution, we found that PrP 137 can be induced to fibrilise much more readily than full-length WT PrP, with significantly shorter half-and lag-times ( Fig. 7).

Discussion
In this study, we have examined the effect of the removal of the N-terminus of the folded domain of PrP, on the structure, stability and fibrillisation characteristics of PrP C . This truncated form of PrP C mimics the structural effect of the association of the conserved hydrophobic region with the ER membrane on the folded C-terminal domain of PrP C .
We find that the truncated protein retains the native secondary structure elements of PrP C , including, remarkably, the second strand of the PrP β-sheet, but with increased conformational flexibility at the C-termini of α-helices 1 and 3 and altered dynamics in the conformationally-variable β2-α2 loop. It also retains a well-defined tertiary structure, with strong evidence for co-operative folding. There is a marked reduction in the stability and increased exposure of hydrophobic residues however, which would appear to facilitate its conversion to fibrillar states significantly. The structural elements of PrP C most affected by the deletion are the second strand of the β-sheet and the first α-helix, both of which display markedly increased solvent exposure and reduced stability relative to the other structure elements.
Previous hydrogen exchange data show that the core of PrP C , comprising the three α-helices and the second β-strand, is a stable entity, which has to completely unfold for exchange with the solvent to occur (26). These data suggest that for fully folded PrP C , the most likely route to PrP Sc is through the completely unfolded state, and that any partially-folded intermediate states involved in this conversion would most likely retain this core region.
Under partially denaturing, acidic conditions however, various regions of mouse PrP C have been shown to undergo sub-global unfolding, forming at least two distinct partially unfolded forms (PUFs) (54). These are in equilibrium with the native state and display increased solvent exposure relative to it. A key characteristic of the second of these PUFs is that the first α-helix, second strand of the βsheet and loop regions in between are disordered and solvent accessible. In this study we find that that removal of the PrP N-terminus to residue 137, mimicking the insertion of the CHR into the ER membrane, also markedly reduces the stability of α-helix 1 and the second β-strand, generating a partially unfolded form of PrP C which displays many of the characteristics of the PUF generated under partially-denaturing acidic conditions (54). Both of these open forms fibrillise much more readily than the native form of the protein, and resemble a conformation implicated to be an initial intermediate in the conversion of monomeric PrP into misfolded oligomer at pH 4 (55). α-helix 1 in particular has been implicated in the conversion to PrP Sc (56,57), with the anti-PrP therapeutic monoclonal antibody ICSM18 effectively curing prion-infected cells through binding to and stabilising α-helix 1 of PrP C (58). Indeed its humanised version has been used in the first-in-human treatment programme using an anti-PrP C monoclonal antibody (59).
The majority of the CHR itself is largely unprotected from amide exchange in PrP C (26), including the first PrP β-strand, which displays anomalously low protection factors in PrP C . This is intriguing given the full protection seen in the other paired β-strand. The detachment of the CHR from the PrP core thus raises two possibilities, firstly that the PrP β-sheet is maintained under native conditions and the loop regions not including the first β-strand (residues 125 -127, 132 -145) detaches, or that secondly, the residues within the β-sheet that are involved in hydrogen bonding within the PrP β-        The data shown are the mean readings from five replicates ±SD.
spectropolarimeter. The denaturation profile for each protein was measured in 3 separate experiments.

Calculating the equilibrium constant between folded and unfolded states (K & K w )
For the two-state equilibrium unfolding transitions, data were fitted to the following equation, where K and K (W) are equilibrium constants between the folded and unfolded states at a given denaturant activity (D) and in water, respectively, and m describes the sensitivity of the equilibrium to denaturant activity (62).
For visual representation of the data shown, data were converted to proportion folded, α F , using the following, α F = (K/(1+K)). Data fitting was carried out using GNUplot v. 5.2. The significance of the differences in free energy for folding and m values between PrP 119 and PrP 137 were determined by two-tailed student's t-Test.

Calculation of Denaturant Activity
Due to the non-linear relationship between denaturant concentration and the free energy for where D is the molar denaturant activity (62).

Equilibrium Thermal Denaturation Monitored by CD
The amide CD absorption of 6. where R is the gas constant (kcalmol -1 K -1 ). All equilibrium unfolding transitions, GuHCl-or thermally induced, were reversible.
Almost complete backbone assignments were determined, the exceptions being the amide protons and nitrogens of residues 167, 169-171, and 175, for which no resonances were detected. These residues occupy a loop region between β-strand 2 and α-helix 2, which is undergoing conformational exchange, resulting in line-broadening of NMR signals (47).

Prediction of secondary structure propensity and ModelFree Analysis
Secondary structure propensity was calculated from H N , Cα, Cβ, CO, & N random coil chemical shifts and average secondary shifts for protein secondary structure using the program TALOS-N (38).
TALOS-N is an artificial neural network based system for empirical prediction of protein backbone φ/ψ torsion angles, sidechain χ1 torsion angles and secondary structure using chemical shifts. The Modelfree order parameters S 2 for the backbone amide groups were predicted from backbone (H N , Cα, CO, & N) and Cβ chemical shifts using the Random Coil Index (RCI) approach as implemented within TALOS-N (38)(39)(40).

Amide Hydrogen-Deuterium Exchange
Hydrogen-deuterium exchange rates (k ex ) were determined by adding 140 µl 10 mM sodium phosphate, 1 mM sodium azide, pH 7.0, dissolved in 100% (v/v) D 2 O to the same volume of 1 mM PrP 137 in the equivalent protonated buffer. A series of 2D sensitivity-enhanced 1 H-15 N HSQC spectra (65,66) were acquired at 293K on a Bruker DRX-800 spectrometer. The decay curves of the 1 H-15 N HSQC cross-peaks were fitted to single exponential decays with offset, and protection factors (k ex /k int ) for observable amides were determined using intrinsic amide exchange rates (k int ) (67).
Acquisition of the first experiment began approximately 5 minutes after mixing.

Amide Hydrogen-Deuterium Exchange measured by CLEANEX
Fast exchanging amide proton rates were determined using the CLEAN chemical EXchange (CLEANEX-PM) approach (51) under the same conditions used for the hydrogen-deuterium exchange experiment described above.

Analytical Ultracentrifugation
Sedimentation velocity ultracentrifugation experiments (SV-AUC) were carried out using a Beckman Optima XL-I analytical ultracentrifuge. Samples were loaded into Beckman AUC sample cells with 12 mm optical path two-channel centrepieces, with matched buffer in the reference sector. Cells were spun at 50,000 rpm in an AnTi-50 rotor and scans were acquired using both interference and absorbance optics (at 280 nm) at 10 minute intervals over 16 hours. The sedimentation profiles were analysed using the software SEDFIT (v13b) (68). Partial specific volumes (ρ) for PrP 137 were calculated from the amino acid sequence using SEDNTERP software (69). Buffer densities and viscosities were measured using an Anton Paar DMA5000 density meter and an Anton Paar AMVn automated microviscometer, respectively. Sedimentation velocity data were analysed using the c(s) method of distribution (68)  To determine the half-and lag-times for fibril formation, data were fitted to an empirical function