Dominant-negative effects in prion diseases: insights from molecular dynamics simulations on mouse prion protein chimeras

Mutations in the prion protein (PrP) can cause spontaneous prion diseases in humans (Hu) and animals. In transgenic mice, mutations can determine the susceptibility to the infection of different prion strains. Some of these mutations also show a dominant-negative effect, thus halting the replication process by which wild type mouse (Mo) PrP is converted into Mo scrapie. Using all-atom molecular dynamics (MD) simulations, here we studied the structure of HuPrP, MoPrP, 10 Hu/MoPrP chimeras, and 1 Mo/sheepPrP chimera in explicit solvent. Overall, ∼2 μs of MD were collected. Our findings suggest that the interactions between α1 helix and N-terminal of α3 helix are critical in prion propagation, whereas the β2–α2 loop conformation plays a role in the dominant-negative effect. An animated Interactive 3D Complement (I3DC) is available in Proteopedia at http://proteopedia.org/w/Journal:JBSD:4.


Introduction
The key event in prion diseases, or transmissible spongiform encephalopathies, is the conversion of the cellular prion protein (PrP C ) to its pathogenic scrapie form denoted PrP Sc or prion (Brandner et al., 1996;Bueler et al., 1993). The conversion is the result of a posttranslational process whereby most α-helical motifs are replaced by β-sheet secondary structures without any covalent modifications. PrP Sc catalyzes the conversion of PrP C to nascent PrP Sc and triggers prion propagation. Structurally, PrP C is a glycophosphatidylinositol-anchored glycoprotein composed of 209 amino acids (in human [Hu] numbering) including an N-terminal (N-) unstructured domain (residues 23-127) and a C-terminal (C-) globular domain (GD) of three α-helices and a short, two-stranded, antiparallel β-sheet (residues 128-231) (Zahn et al., 2000). In con-trast to PrP C , PrP Sc structure has significant β-sheet content and exhibits distinct biophysical and biochemical properties (Pan et al., 1993), albeit no atomistic-resolution structural information is available.
Although PrP C influences several processes in the central and peripheral nervous systems, its function has not been established with certainty (Linden et al., 2008). Key evidence exists supporting the link between mutations in prion protein (PrP) gene (PRNP in Hu or Prnp in other mammalian species) and the spontaneous generation of prion diseases (Dossena et al., 2008;Friedman-Levi et al., 2011;Jackson et al., 2009). Several insertion and point mutations linked to genetic forms of Hu prion diseases have been identified (Mastrianni, 2010). Polymorphisms in PrP may influence the etiology and neuropathology of the disease both in humans (Bishop, Pennington, Heath, Will, & Knight, 2009) and in sheep (Bossers, Schreuder, Muileman, Belt, & Smits, 1996). Transgenic (Tg) mice expressing both mouse (Mo) PrP and HuPrP did not exhibit abbreviated incubation times when infected with Hu prions. However, Tg mice expressing a chimeric PrP in which a 9-residue segment was replaced with HuPrP sequence was found highly susceptible to Hu prions and exhibited abbreviated incubation times (Telling et al., 1994). Recently, Tg mice expressing MoPrP chimeras (CMPrP) containing mutations from other species have been reported to exhibit both transmission barriers against certain prion strains (Sigurdson et al., 2010) and spontaneous conversion into prions (Sigurdson et al., 2009(Sigurdson et al., , 2011. Interestingly, a variety of CMPrP investigated in scrapie infected Mo neuroblastoma (ScN2a) cells showed diverse susceptibility or resistance to Mo scrapie infection (Kaneko et al., 1997). Some of these CMPrP also acted as "dominant-negatives" to inhibit wild type (WT) MoPrP from being converted to PrP Sc . The protective effects of amino acid substitutions on the PrP C have been reported and deeply investigated in humans and in sheep. In humans, the naturally occurring polymorphism E219 K was found to protect against sporadic Creutzfeldt-Jakob disease (CJD) when expressed in heterozygosis (Shibuya, Higuchi, Shin, Tateishi, & Kitamoto, 1998;Soldevila et al., 2003). In sheep, three polymorphisms are acutely linked to the occurrence of scrapie: A136 V, R154H, and Q171R/H. These generate five commonly observed alleles: ARQ, ARR, AHQ, ARH, and VRQ. ARR and AHQ are associated with resistance to scrapie; ARQ, ARH, and VRQ are associated with susceptibility to scrapie (Baylis & Goldmann, 2004;Sabuncu et al., 2003).
Taken together, these experimental data suggest that subtle changes in PrP C structure can alter prion propagation, possibly by (de)stabilizing the physiological folding of PrP C and/or affecting its interactions with some yet unknown factors. Here, we applied our well-established molecular dynamics (MD) protocol to study the atomistic structures of 11 CMPrP (Table 1), which were compared with the structures of WT Mo and Hu PrP. Among these 11 CMPrP, 10 CMPrP are Hu/Mo chimeras that contain 1 or 2 mutations corresponding to the Hu codons. Another CMPrP Q172R (in Hu numbering) is a natural ovine dominant-negative polymorphism (Goldmann, Hunter, Smith, Foster, & Hope, 1994;Westaway et al. 1994). All these 11 CMPrP have been investigated in ScNa2 cells for their response to Mo scrapie infection (Kaneko et al., 1997). We included Q172R to CMPrP because this mutation is located in the β2-α2 loop of PrP, which is an epitope crucial for prion susceptibility (Sigurdson et al., 2010(Sigurdson et al., , 2011. Although Q172R may not have the same significance as the Hu/Mo CMPrP, studying it provided additional information and confirmed our findings with Hu/Mo CMPrP. The origin of these mutations as well as their in vivo and in vitro effects is listed in Table 1. This study may explain the transmission barrier between Hu and Mo prions, as well as the dominant-negative effects of certain polymorphisms.

Preparation of the simulation systems
The structural models of MoPrP and the CMPrP GD are based on the NMR structure of MoPrP (PDB entry 1XYX) (Gossert, Bonjour, Lysek, Fiorito, & Wuthrich, 2005) resolved at pH 4.5, the only solution structure available so far for MoPrP. Note that 1XYX is the refined structure of an earlier PDB entry 1AG2 (Riek et al., 1996) resolved at the same pH. The best representative conformer (model 1) of the 1XYX ensemble (Gossert et al., 2005) was used. Following the procedure validated in references Rossetti, Giachin, Legname, and Carloni (2010) and Rossetti, Cong, Caliandro, Legname, and Carloni (2011), the mutations in CMPrP were introduced into WT MoPrP using Swiss-Pdb-Viewer (DeepView 4.0) (Guex & Peitsch, 1997). The model of HuPrP is based on the NMR structure of HuPrP GD (PDB entry 1HJN) (Calzolai & Zahn, 2003). 1HJN is the only solution structure of HuPrP resolved at pH 7, which corresponds to the pH of our MD simulations. The best representative conformer of 1HJN is not indicated, whilst the minimized average structure 1HJM.pdb may not represent well the structure ensemble (Sutcliffe, 1993). Therefore, a cluster analysis (Micheletti, Seno, & Maritan, 2000) was performed on the 20 models in 1HJN to identify a representative model as the initial structure for the MD simulations. The same procedure has been successfully applied in references Rossetti et al. (2010Rossetti et al. ( , 2011. Test simulations were carried out to determine the protonation states of histidine (H) residues in these simulation systems. At pH 7, D, E, K, and R residues were considered in their ionized form but H residues may be neutral or, to a smaller extent, positively charged. We observed in reference Rossetti et al. (2010Rossetti et al. ( , 2011 that in HuPrP and its naturally occurring mutants, H140 and H177 are solvent exposed while H187 forms a single intramolecular hydrogen bond (HB) that involves H187 Nɛ and R156 backbone carbonyl. When all four Hs (H140, H155, H177, and H187) were monoprotonated at the Nɛ atom, our calculations reproduced accurately the structural determinants of the available NMR structures. In the GD of MoPrP and the CMPrP, only H140, H177, and H187 are present. They have the same local environment as those in HuPrP. Our test calculations (see "Supplementary material" for details) suggest that these H residues take the same Nɛ-protonation neutral state as that in HuPrP. Other protonation states introduce instability to the protein structure.
Following the protocol of references Rossetti et al. (2010Rossetti et al. ( , 2011, the protein models were solvated in a box of explicit water, ensuring that the solvent shell extended for at least 16 Å around them. Each system contained 9540-10,720 water molecules and 0-2 Na + counterions. The total number of atoms ranged from 29,700 to 33,240. Periodic boundary conditions were applied. AMBER99SBidln force field (Lindorff-Larsen et al., 2010), TIP3P water model (Jorgensen, Chandrasekhar, Madura, Impey, & Klein, 1983), and adapted åqvist potential (Aqvist, 1990) were used to treat the protein, water, and counterions, respectively. Applying the periodic boundary conditions, long-range electrostatic interactions were treated with the particle mesh Ewald (PME) method (Darden, York, & Pedersen, 1993). A Fourier spacing of 1.2 Å combined with a fourth-order cubic spline interpolation was used. A 12-Å cutoff was used for van der Waals interactions as well as the real-space part of the electrostatic interactions. All bond lengths were constrained with the LINCS algorithm (Hess, Bekker, Berendsen, & Fraaije, 1997) to allow for a time step of 2 fs. The systems underwent 1000 steps of steepest-descent energy minimization with 5000 kcal mol À1 ·Å À2 harmonic position restraints on the protein, followed by 2500 steps of steepest-descent and 2500 steps of conjugate gradient minimization without restraints. For each MD run, the starting temperature was 2 K. The system was then gradually heated in 12 steps of 100-ps simulation (from 2 to 25, 50, 75, 100, 125, 150, 175, 200, 225, 250, 275, and 298 K). For each step, the velocity was generated consistent with a Maxwell-Boltzmann distribution at the corresponding temperature. After that, a 200-ps equilibration followed by 60-ns production MD run was performed under the NPT ensemble (T = 298 K, p = 1 bar). The NPT ensemble was realized by coupling the systems with a velocity rescaling thermostat (Bussi, Donadio, & Parrinello, 2007) and an Andersen-Parrinello-Rahman barostat (Andersen, 1980;Parrinello & Rahman, 1981). Finally, three independent MD runs were performed on each system, starting from different random seeds for the initial velocity generation.

Data analyses
All the analyses were performed with tools in Gromacs 4.5 and VMD (Humphrey, Dalke, & Schulten, 1996) except for the secondary structure analysis which was carried out using DSSP software (Kabsch & Sander, 1983).

Results
Three independent MD runs were carried out on HuPrP, MoPrP, and each of the 11 CMPrP (Table 1). All the systems appeared to reach equilibrium within 10 ns, as indicated by the Cα root-mean-square deviation (RMSD) as a function of simulation time ( Figure S1 in "Supplementary material"). Fifty nanoseconds of equilibrated trajectory were taken from each MD run for analysis. Overall, 1.95 μs of MD trajectories were collected. All CMPrP maintained the 3D fold in GD similar to MoPrP and HuPrP during the simulations. Large fluctuations were observed in β2-α2 loop (residues 167-171) and C-of α3 helix (residues 219-231) (Figure 1). These two fragments were poorly defined in MoPrP NMR structure, indicating structural disorder or increased mobility (Riek et al., 1998). Excluding the flexible terminal residues (125-127 and 220-231), the GD of CMPrP during our simulations showed Cα RMSD not more than 2.7 ± .4 Å from the MoPrP NMR structure (Riek et al., 1998) or 2.8 ± .7 Å from the HuPrP NMR structure (Zahn et al., 2000) (Table 2). Consistently, the Cα radius of gyrations averaged along the x, y, and z axes ranged from 13.3 ± .1 to 13.6 ± .2 Å for residues 128-219 of each CMPrP (Table 2). All CMPrP except K220R showed smaller Cα root-mean-square fluctuations (RMSF) than MoPrP. These fluctuations were similar to those of HuPrP ( Figure 1). K220R exhibited large fluctuations in residues 133-157, which contain the β1-α1 loop and the α1 helix, as well as in residues 190-199 which correspond to N-α3 (Figure 1(D)). These fluctuations are due to a loss of intramolecular interactions between α1 and N-α3, which are reported in detail in the following section.
Next, we examined the β2-α2 loop where all CMPrP and HuPrP exhibited smaller RMSF than MoPrP. We analyzed the secondary structure elements of this loop  (Calzolai & Zahn, 2003). The β2-α2 loop (residues 167-171, blue) and C-α3 (residues 219-231, orange) that experience the largest fluctuations are labeled.  (Table 4). In such a Mo-like pattern, 3 10 -helix covered over 45% of the simulation time while coil occurred in less than 3%. The 3 10 -helix conformation is stabilized by two HBs: one is formed between the backbone of Q168 N atom and P165 O atom, the other is between the backbone of Y169 N atom and V166 O atom. The other four CMPrP (Q168E, V215I_Q219E, Q217R_Q219E, and Q219E) manifested a more Hu-like pattern: the "bend/turn"-exchange domi-nates the loop conformation, while 3 10 -helix covered less than 20% whereas coil accounted for over 10% (Table 4). A HB formed between D167 backbone N atom and S170 sidechain O atom stabilizes the bend conformation. Alternatively, a HB between the E168 backbone N atom and M166 backbone O atom contributes to the turn  This analysis was carried out using DSSP software (Kabsch & Sander, 1983 conformation. Our analyses did not identify any salt bridge or hydrophobic interactions that particularly contribute to these specific loop conformations.
Hydrogen exchange, site-directed spin labeling, electron paramagnetic resonance spectroscopy, and solid-state NMR studies have suggested that the core of PrP amyloid fibrils is composed of the region spanning the α2 helix, the α2-α3 loop, and a major part of α3 helix of PrP C (Cobb, Sonnichsen, McHaourab, & Surewicz, 2007;Lu et al., 2007;Tycko, Savtchenko, Ostapchenko, Makarava, & Baskakov, 2010). These studies indicate a dissociation of the α2-α3 region from the rest of PrP during the conversion. Therefore, enhanced interactions between α1 and N-α3 might inhibit the dissociation and the conversion, as seen in our simulations on the scrapie-resistant CMPrP. However, some other studies suggest that amyloid fibrils can be formed by the structured regions preceding the α1 helix, likely via an extended β-structure with residues 102-124. This hypothesis is supported by the extensive studies (Damo et al., 2010;Helmus, Surewicz, Nadaud, Surewicz, & Jaroniec, 2008;Lim et al., 2006) on the pathogenic HuPrP mutation Y145stop which contains only residues 23-144. These studies indicate that in the absence of the C-terminal residues 145-230, the fibril core is composed mostly of residues 102-139. While the misfolding mechanism of PrP remains obscure, a growing body of evidence suggests that distinct aggregate structures can be formed through multiple pathways by different prion strains (Cohen & Kelly, 2003;Legname et al., 2006;Nekooki-Machida et al., 2009;Prusiner, 2001;Tessier & Lindquist, 2007). This may explain the discrepancy of our findings and some others derived from different PrP variants or prion strains. The β2-α2 loop of PrP conformation has been deeply investigated in different mammalian PrPs (Christen, Perez, Hornemann, & Wuthrich, 2008;Gossert, Bonjour, Lysek, Fiorito & Wuthrich, 2005;Perez, Damberger, & Wuthrich, 2010). It has been suggested that the loop plasticity is related to the transmissible barrier among prion strains (Agrimi et al., 2008;Christen et al., 2008;Christen, Hornemann, Damberger, & Wuthrich, 2009;Gossert et al., 2005;James et al., 1997;Lysek et al., 2005;Perez et al., 2010;Wen et al., 2010). Solution NMR studies on MoPrP and HuPrP structures found that at room temperature (293-298 K) a line broadening of NMR signals was reported for the β2-α2 loop in all the Mo and Hu PrP constructs due to local conformational exchanges (Calzolai & Zahn, 2003;Damberger, Christen, Perez, Hornemann, & Wuthrich, 2011;Gossert et al., 2005;Zahn et al., 2000). A recent solution NMR study discovered that at 310 K, the β2-α2 loop of MoPrP is characterized by 3 10 -helix, while Y169A/G mutations result in β-turn conformation (Damberger, Christen, Perez, Hornemann, & Wuthrich, 2011). This study suggests that at room temperature the β2-α2 loop of MoPrP is disordered, likely due to conformational exchange between 3 10 -helix and another conformation. These studies imply that the loop conformation may play an essential role by binding to an unknown factor required for prion propagation. Consistently, previous studies on pathogenic HuPrP mutants (Meli, Gasset, & Colombo 2011;Rossetti et al., 2010Rossetti et al., , 2011 indicated that the conformation of the β2-α2 loop was related to the spontaneous human familial prion diseases. In this study, we identified a "3 10 -helix/turn" conformational exchange pattern of the β2-α2 loop in MoPrP and a "bend/turn/ coil" pattern in HuPrP. The intramolecular interactions that contribute to the 3 10 -helix turned out to be the same HBs as those observed in the MoPrP NMR structure at 310 K, which involve residues P165, V166, Q168, and Y169 (Damberger et al., 2011). Another two HBs were identified to control the bend and turn conformations, which involve residues M166, D167, E168, and S170. These specific interactions in the β2-α2 loop may be the determinants of the loop conformation. Seven CMPrP (N171S, Q172R, V215I, Q217R, Q219 K, K220R, and K220R_Q219E) exhibited the Mo-like pattern in MD. Among these, two of them (N171S and K220R) turned out to be susceptible to Mo scrapie in ScN2a cells, whereas the other four (Q172R, V215I, Q217R, and Q219 K) showed the dominant-negative effect (Kaneko et al., 1997). K220R_Q219E was resistant to Mo scrapie (Kaneko et al., 1997) but no data have been reported about the dominant-negative effect. The other four CMPrP studied here (Q168E, V215I_Q219E, Q217R_Q219E, and Q219E) showed a more Hu-like pattern. They turned out to be resistant to Mo scrapie but not to have dominant-negative effect (Kaneko et al., 1997). The data indicate that the Mo-like conformational exchange pattern is necessary for the dominant-negative effect if the CMPrP is also scrapie-resistant.
Based on these findings, we propose the following mechanism for the dominant-negative effect: the Mo-like conformation at the β2-α2 loop can bind to an unknown factor that facilitates prion propagation. When the scrapie-resistant CMPrP have the Mo-like β2-α2 loop conformation, they can bind to this factor but cannot be converted to PrP Sc . They stay bound with the factor and keep its binding site inaccessible for the WT MoPrP. Therefore, WT MoPrP is not converted to PrP Sc . In the case of nonresistant CMPrP such as N171S and K220R, they can bind to the factor but are easily converted to PrP Sc and released from it. The binding site is then free again to bind new CMPrP or WT MoPrP. Therefore, WT MoPrP is also converted to PrP Sc and no dominant-negative effect is observed.
Evidence exists that prion propagation requires an accessory cofactor (Deleault et al., 2012a;Deleault, Lucassen, & Supattapone, 2003;Telling et al., 1995). Recently, Deleault et al. demonstrated that a small molecule cofactor can regulate the conformation, strains and infectivity of PrP Sc in vivo (Deleault et al., 2012b). However, some other studies argue against the intervention of any ancillary components (Geoghegan, Miller, Kwak, Harris, & Supattapone, 2009) and suggest that the dominant-negative effect is only due to the heterozygosis of PRNP or Prnp genes (Geoghegan et al., 2009;Hizume et al., 2009;Lee, Yang, Perrier, & Baskakov, 2007). This alternative model, denoted as "stone fence model," suggests that Hu individuals heterozygous for E/ K at codon 219 are protected by sporadic CJD because the PrPs deriving from two allelic variants interfere with each other in the conversion process due to their incompatible structures. The in vivo and in vitro evidences supporting this model are reviewed in ). However, an ultimate structural explanation of the dominant-negative effect is still missing in prion biology.
Considering the previous findings on CMPrP (Kaneko et al., 1997;Telling et al., 1994), which have inspired this work, we propose the above possible mechanism of the dominant-negative effect. Nevertheless, further investigations are needed to verify this hypothesis. In particular, the alternative hypothesis that the dominant-negative effects are due to PRNP or Prnp heterozygosis may suggest novel rationally designed MD experiments aimed to investigate the interaction between two different allelic variants to mimic the heterozygosis condition.

Supplementary material
The supplementary material for this paper is available online at http://dx.doi.10.1080/07391102.2012.712477.