|Home | About | Journals | Submit | Contact Us | Français|
Conformational change in the prion protein (PrP) is thought to be responsible for a group of rare but fatal neurodegenerative diseases of humans and other animals, including Creutzfeldt-Jakob disease and bovine spongiform encephalopathy. However, little is known about the mechanism by which normal cellular PrPs initiate and propagate the conformational change. Here, we studied backbone dynamics of the inherited pathogenic mutants (P101L and H186R), protective mutants (Q167R and Q218K), and wild type mouse PrP(89−230) at pH 5.5 and 3.5. Mutations result in minor chemical shift changes around the mutation sites except that H186R induces large chemical shift changes at distal regions. At lowered pH values, the C-terminal half of the second helix is significantly disordered for the wild type and all mutant proteins, while other parts of the protein are essentially unaffected. This destabilization is accompanied by protonation of the partially exposed histidine H186 in the second helix of the wild type protein. This region in the mutant protein H186R is disordered even at pH 5.5. The wild type and mutant proteins have similar μs conformational exchange near the two β-strands and have similar ns internal motions in several regions including the C-terminal half of the second helix, but only wild type and P101L have extensive ns internal motions throughout the helices. These motions mostly disappear at lower pH. Our findings raise the possibility that the pathogenic or dominant negative mutations exert their effects on some non-native intermediate form such as PrP* after conversion of cellular PrP (PrPC) into the pathogenic isoform PrPSc has been initiated; additionally, formation of PrPSc might begin within the C-terminal folded region rather than in the disordered N-terminal region.
Prion diseases are a group of rare but fatal neurodegenerative disorders that pathologically manifest accumulation of protease-resistant amyloid plaques of prion protein (PrP) in affected brain regions (1-4). These disorders appear as sporadic, dominantly heritable, and transmissible maladies that include Creutzfeldt-Jakob disease (CJD), Gerstmann-Sträussler-Scheinker syndrome (GSS), fatal familial insomnia (FFI) and kuru in humans, bovine spongiform encephalopathy in cattle, scrapie in sheep, and chronic wasting disease in elk and deer.
PrP is a highly conserved glycoprotein that contains two glycosylation sites and is linked to the external surface of the cell through a glycosyl-phosphatidyl-inositol (GPI) anchor. NMR studies revealed that the recombinant PrP consists of a largely unfolded N-terminal region and a folded C-terminal domain encompassing three α helices (α1, α2 and α3) and two short β strands (β1 and β2), with a single disulfide bond bridging α2 and α3 (5,6).
All prion diseases are thought to be caused by a profound conformational change, which occurs when the normal, cellular isoform (PrPC) is converted to the pathological form (PrPSc) in the absence of any detectable covalent modification. PrPC is rich in α-helical structure; in contrast, PrPSc forms aggregates and is β-sheet rich (7,8). Many but not all forms of PrPSc are resistant to proteolyis (9). The molecular mechanism by which PrPSc initiates and propagates a conformational change during prion propagation remains unclear. In particular, the conditions necessary in vivo for the formation of PrPSc such as pH, redox environment, post-translational modifications and cofactor(s) remain elusive. There is ongoing controversy regarding whether the structure of PrPSc is amyloid or an intermediate β-sheet rich oligomer on the pathway to amyloid formation. In vitro conversion of PrPC into amyloid does not, in general, lead to the formation of PrPSc; most of the converted molecules appear to be in a non-infectious amyloid conformation whose structure is quite different from that of PrPSc (G. Stubbs, personal communication).
Exposure of scrapie-infected neuroblastoma (ScN2a) cells to weak bases inhibited the formation of nascent PrPSc. Acidic endosomes seem to play an important role in intracellular trafficking of PrPSc from cholestrerol-rich microdomains on the cell surface to lysosomes (10-13). Additionally, the conversion of PrP in vitro into the protease-resistant form accelerates at acidic pH in the presence of denaturant (14) and a β-sheet rich unfolding intermediate of PrP is exclusively observed at low pH (15).
In order to explore the factors that may affect the rate or propensity for the conversion of PrPC into PrPSc, we compared the solution behavior of four mutant PrPs in terms of deformation and altered flexibility which should modulate the initiation of spontaneous conversion and/or the interaction with intermediates on the route of conversion (16,17). The positions of these mutation sites are shown mapped on the structure of the mouse prion protein (5) with a modeled N-terminal tail in Fig. 1. The human equivalents of the mouse P101L and H186R mutations cause familial prion diseases (18,19). By contrast, the Q167R and Q218K mutations manifest a dominant negative phenotype in sheep and humans, respectively (20-22). Since the conversion of PrPC into PrPSc requires an acidic endosomal compartment (13), we compared the structural and dynamic changes induced by acidic pH on the wild-type (wild type) and four mutant mouse PrPs (residues 89−230) by measuring and analyzing backbone 15N NMR relaxation. We found that the wild type and mutant PrPs exhibited many similarities and relatively few differences in their dynamic features.
Our findings raise the possibility that the conformational change, which features in the conversion of PrPC into PrPSc, begins in C-terminal folded region. Our data also suggest that some PrP mutations do not alter the initial steps in the conformational transition that PrP undergoes.
15N/13C-labeled and/or 15N-labeled recombinant mouse PrP(89−230) of wild type, P101L, Q167R, H186R and Q218K were expressed in Escherichia coli, purified, and refolded as previously described (6,23). NMR samples were prepared in 20 mM sodium acetate pH 5.5 or 3.5, 90% H2O/10% D2O, and 0.05% sodium azide. The protein concentration was 0.1−0.6 mM (Supplementary Table S1).
Backbone resonances Cα, Cβ, C’, N, and H of wild type mouse PrP(89−230) were assigned by using three-dimensional (3-D) HNCACB, CBCA(CO)NH, HNCO, 15N-edited NOESY-HSQC (τm = 150 ms), 15N-edited TOCSY-HSQC (τm = 80 ms), (24) and the deposited chemical shifts data of Syrian hamster PrP(90−231) (BMRB id: 4307) (25). Backbone assignments were transferred from the wild type to the four mutant PrPs and confirmed using 15N-edited NOESY-HSQC (τm = 150 ms) and 15N-edited TOCSY-HSQC (τm = 80 ms). No backbone resonances were observed for residues 168−174 in wild type and all four mutants PrPs as reported previously for wild type (5,26). Backbone resonances for residues 187−188 of the H186R mutant protein were not observed due to line broadening (Supplementary Fig. S1). The histidine imidazole resonances 1Hδ2 1Hε1, 15Nδ1 and 15Nε2 of wild type mouse PrP(89−230) were assigned by two-dimensional (2-D) (Hβ)Cβ(CγCδ)Hδ (27) and by 2-D long-range 1H-15N heteronuclear multiple-quantum coherence (HMQC) experiment (28). Protonation states of the histidines were examined using a 2D long-range HMQC in which the 15N carrier frequency and sweep width were set to 180 ppm and 30 ppm, respectively and the delay during which 15N and 1H signals become anti-phase was set to 4.5 ms (29). 15N T1, T2, and [1H]-15N NOE were measured on Bruker 500 and 600 MHz spectrometers at 298 K using pulse sequences described previously (30,31). 10 spectra were acquired using relaxation delays of 0.011*, 0.161, 0.33*, 0.495, 0.66*, 0.825 and 1.1 s for T1 and 6*, 18, 34, 50*, 66, 90, 114* ms for T2 (the asterisks denote duplicate measurements). The [1H]-15N NOE was measured from at least 2 sets of saturated and unsaturated spectra in which protons were saturated for 3.0 s or not after repetition delay. Repetition delay was 3.5 s, 3.5 s, and 3.0 s for T1, T2 and [1H]-15N NOE experiments, respectively. All spectra were processed using nmrPipe (32). For obtaining relaxation rates R1(=1/T1) and R2(=1/T2), peak intensity was fitted to a single exponential decay function, I(t) = I0 exp(-tR) in which t is the variable relaxation delay and R is the relaxation rate. The [1H]-15N NOE was derived from ratio of peak intensity measured in the saturated and unsaturated spectra (= Isat/Iunsat). Experimental errors were estimated from standard deviations of the peak intensities of duplicate experiments. Carr-Purcell-Meiboom-Gill (CPMG) 15N R2 relaxation dispersion was measured at 298 K on Bruker 500 and 800 MHz spectrometers with total CPMG duration of 40 ms and recycling delay of 3 sec (24). The 15N longitudinal 2-spin-order exchange rates (zz-exchange) for H186R were measured and analyzed as described previously (33). Transverse cross-correlation (ηxy) between 15N-1H dipolar interactions and 15N chemical shift anisotropy (CSA) was measured at 500 MHz and 298 K with dephasing delays (Δη) of 10, 25, 40* ms (the asterisks denote duplicate measurements) (34).
15N R1, 15N R2, and [1H]-15N NOE data sets acquired at 500 and 600 MHz were analyzed by an in-house computer program eMF (S.-H. Bae, unpublished), using the extended Lipari-Szabo formalism (35-37). An axially symmetric rotational diffusion tensor was optimized against the mouse PrP structure (PDB id: 1xyx) (26) using selected residues with [1H]-15N NOE > 0.65 which belong to α-helices and β-strands (Supplementary Table S1). A best model for each residue was selected by the Bayesian information criterion (38-40) after excluding unrealistic model(s) that have S2 > 1 or τe < 0 ns, or R < 0 (39,40). Minimum errors were set to 3 %, 3 %, and 0.03 for 15N R1, 15N R2 and [1H]-15N NOE, respectively. 15N CSA and amide NH bond length were assumed to be −160 ppm and 1.02 Å. Reduced spectral densities J(0), J(ωN), and J(0.89ωH) were mapped assuming that J(ω) is proportional to 1/ω2 at high frequency around ωH (41).
The 15N R2 relaxation dispersion data were fitted with the in-house computer program GLOVE (J.C. Lansing, unpublished) using the general equation for two-site exchange that encompasses all conformational exchange time scales (42):
where ψ = kex2 − Δω2, ζ = −2Δωkex(pA − pB), τcp is the delay between 180° pulses in the CPMG pulse train, the pa and pb are populations of the two states A and B, kex is the exchange rate constant (kex =kA→B + kB→A), and Δω is the 15N chemical shift difference between two states.
Translational diffusion coefficients were measured at 298K using the PFG-SLED (pulsed field gradient stimulated echo longitudinal encode-decode) NMR method (43). The PFG duration was 6.3 ms and the self-diffusion delay was 80 ms. The PFG strength was varied from 5 % to 90 % in triplicate series of 1D spectra. The integral of peak intensity (I) is related to a relative diffusion coefficient, d, and relative gradient strength, G by I = I(0) exp(-dG2). The diffusion coefficient was calculated relative to the diffusion coefficient of lysozyme (1.08 × 10−6 cm2 s−1) (43).
The P101L and H186R mutations are associated with familial prion disease in humans; P101 and H186 are both highly conserved between mammalian species. These mutations occur in very different regions of the protein: P101 is located in the unstructured tail of the PrP 27−30 sequence (modeled in Fig. 1), while H186 is packed into the core of the folded C-terminal domain, between α2, α3 and β2. It appears likely that the reasons why these mutations should favor conversion of PrPC to other forms, including pathogenic forms, might differ. The dominant negative mutations Q167R and Q218K both occur in regions of the protein that are solvent-exposed, in the loop between β2 and α2 for Q167 and towards the end of the α3 helix for Q218. Neither of these residues is highly conserved among mammalian species: Q167 is replaced by Glu in the human sequence while Q218 is replaced by Glu in primate sequences.
The chemical shift perturbations caused at pH 5.5 by the P101L, Q167R, and Q218K mutations are small and are strictly localized to the mutation site and immediate neighbor residues (Fig. 2A), consistent with the positions of these mutation sites in solvent-exposed areas of the protein that are low in secondary and tertiary structure (5,6). By contrast, the H186R mutation causes large chemical shift changes for the linker between α1 and β2 (Y156, Q159, Y162), α2 (Q185, T191, G194) and α3 (F197, E206), and minor chemical shift changes around β1. The relative magnitude of these changes is consistent with the position of residue 186 in the middle of α2, and the packing of the side chain into the core of the molecule. Nevertheless, even for H186R, the majority of the chemical shifts were virtually unaffected by the mutations, indicating that the overall structure of mouse PrP was not substantially altered in any of the mutations. Variations in the chemical shift upon changing conditions such as pH (3.5 − 5.5), KCl (0 − 150mM), urea (0 − 1.5M) and temperature (25 − 40°C) showed that structural perturbations were most significant when the pH was changed (Supplementary Fig. S2). Chemical shift differences between mutant and wild type proteins at pH 3.5 (Fig. 2B) showed similar trends as those at pH 5.5 Since the chemical shift differences between mutants and wild type at pH 5.5 and pH 3.5 were similar, we inferred that all of the proteins were affected by acidic pH in similar ways.
Since acidic pH increases the probability of conversion to the pathogenic form (10-15), we anticipated that the inherited pathogenic PrP mutants might display different structural perturbations from the dominant negative PrP mutant or wild type proteins upon acidification. However, the chemical shift differences between pH 5.5 and pH 3.5 were the same for wild type, P101L, Q167R and Q218K, all of which showed similar and large chemical shift changes for residues from K184 to T198 (Fig. 3). This region encompasses the C-terminal half of α2 and the connecting loop between helices α2 and α3. Among the affected residues, H186, T191, and K193 undergo the largest chemical shifts changes, suggesting that the C-terminal part of α2 plays a dominant role in the structural changes that occur at acidic pH. This region also has the largest chemical shift differences between pH 7.0 and pH 4.5 in human PrP(121−230) (44). In sharp contrast, the H186R mutant shows only small chemical shift differences for helix α2 (Fig. 3), implying that this mutant does not undergo pH dependent structural perturbations. These observations identify histidine 186 as responsible for the pH dependent changes in the NMR spectra of the wild type, P101L, Q167R and Q218K mutant proteins. In addition, wild type and all mutants share significant pH-dependent chemical shift changes for residues 141−143 and 156−160 at the N- and C-termini of the α2 and β2 and residues 205−214 in α3 (Fig. 3), implicating a second titratable group.
A further decrease of pH to ≈ 2.1 resulted in severe line broadening of all resonances except for the N-terminal unfolded region (data not shown); this behavior was similar to that of the β-oligomer form of human PrP under moderately denaturating conditions (1 M urea, 0.2 M NaCl, 20 mM sodium acetate pH 3.6) (45).
15N T1, 15N T2, and [1H]-15N NOE data sets were acquired at 298 K in 20mM sodium acetate (pH 5.5 and pH 3.5) (Supplementary Fig. S3). Due to the presence of ≈ 35 unstructured residues at the N-terminus in PrP(89−230), which affects the tumbling of the folded domain, the rotational diffusion tensor could not be derived directly from the structure, but was obtained from fitting the relaxation data sets to the wild type mouse PrP(121−230) structure (PDB id: 1xyx) (26) assuming that the overall structures of wild type and mutant PrPs are not substantially different. Rotational correlation times calculated from the relaxation data were 9.8−11.9 ns, much larger than those expected from the empirical Stokes-Einstein estimation (≈ 8.4 ns) (46), and reflect the influence of the N-terminal unfolded region on the rotational tumbling of the C-terminal folded domain (38). The molecular tumbling of wild type and mutant PrPs is slightly anisotropic (DD ≈ 1.4−1.7) so that axially symmetric diffusion models fit the experimental data much better than an isotropic diffusion model (Supplementary Table S1). In the fitting, the axis of the longest helix α3 coincides with the major principal axis of the rotational diffusion tensor. The dominant effect of α3 on the anisotropy of mouse PrP(89−230) is consistent with previous observations on the Syrian hamster proteins PrP(23−231 and 90−231) (47). Using fitted rotational diffusion tensors, the backbone dynamics of each residue were determined by model free analysis (35-37). Previous attempts at model free analysis of the Syrian hamster PrP (23−231 and 90−231) resulted in invalid order parameters (S2 > 1) for many residues (25,47). The current analysis uses a non-isotropic rotational diffusion tensor (48) and a Bayesian information criterion (39,40) for model selection in conjunction with elimination of unrealistic models, which gives physically meaningful S2 values for all of the fitted residues in wild type and mutants. Notably, a recent model free analysis of a truncated form of the mouse PrP(113−231) using the isotropic rotational diffusion tensor has also resulted in valid S2 values (49).
The motions of the N-terminal unfolded residues were analyzed with a local rotational diffusion model since their rotational tumbling should be independent of the C-terminal folded domain of the protein. The S2 values in this region are ≈ 0.4, consistent with the highly flexible nature of the N-terminal terminal region (6). However, for all PrPs, there is a cluster of residues around H95 for which S2 (≈ 0.6−0.8) ranges above the rest of the N-terminal unfolded region (Fig. 4).
At pH 5.5 (green bars in Fig. 4), the C-terminal folded region of all the proteins have an S2 larger than 0.85 indicative of restricted backbone motion. However, two broad regions from β1 to α1 (residues ≈ 134−144) and from the C-terminal half of the α2 to the beginning of α3 (residues ≈187−197) have lower S2 values, indicative of backbone flexibility. All of the proteins, both wild type and mutants, share a similar S2 pattern except that a short segment in the H186R mutant following residue 186 (≈ 187−193) shows a sharp decrease in S2 while in the other proteins, the decrease was more gradual (Fig. 4). At pH 3.5 (red bars in Fig. 4), all proteins except H186R showed a substantial decrease in S2 in the same region (residues 187−197) while the rest of the protein was essentially unaffected by a decrease of pH. S2 for the H186R mutant was similar to the values observed for the N-terminal disordered tail. The variations in S2 between pH 5.5 and 3.5 for the wild type PrP are shown in Fig. 5A, B and compared with those for the H186R mutant in Fig. 5C. These observations suggest that residues 187−193 are disordered in the H186R mutant protein even at neutral pH.
Internal motions on the ns time scale appear in both of the flexible segments (residues ≈ 134−144 and ≈ 187−197) (Fig. 6). The ns internal motions in the first segment propagate toward the N-terminus as far as A116. Reduced spectral densities J(ωN) and J(0.89ωH) could be used without the necessity for the assumptions made in the model free analysis, and provide further evidence for significant flexibility in these regions (Supplementary Fig. S4). The time scales of the internal motions are virtually unaffected by pH except for wild type and P101L. It is intriguing that only the wild type and P101L proteins have extensive ns internal motions throughout helices α1-α3 at pH 5.5 (Fig. 6), even though high S2 values indicate restricted backbone motion on the ps-ns time scale. Many of these ns internal motions disappear at pH 3.5 (Fig. 6). The combination of low amplitudes of ps-ns backbone motion with extensive ns internal motion in the same region has been reported in a number of cases (50). Partial aggregation of the wild type and P101L mutant proteins could in principle be responsible for these apparently anomalous internal motions (51). However, on the basis of translational diffusion measurements, we could exclude this possibility: at the same concentration as the NMR relaxation measurements (0.55 mM), wild type, P101L and Q218K have the same translational diffusion coefficients within experimental error (1.06 ± 0.01, 1.08 ± 0.01, and 1.08 ± 0.02 × 10−6 cm2 s−1, respectively, at 298 K).
All wild type and mutant proteins show sharp increases in the R2 relaxation rate near β1 and β2, at G130, V165, and D166 (Supplementary Fig. S3); such an increase is usually interpreted as evidence for conformational exchange on a μs-ms time scale. In order to separate the exchange contribution from R2, exchange-free R2 (R20) was determined from transverse cross correlation (ηxy) rates (52). Our findings show that these sites indeed undergo significant conformational exchange (Fig. 7A). The Carr-Purcell-Meiboom-Gill (CPMG) 15N R2 relaxation dispersion data of G130, V165 and D166 show distinct differences between R2 relaxation rates at 500 and 800 MHz even though each has small R2 dispersion (< 5 s−1) (Fig. 7B). The exchange rate and the populations of the states were estimated from simultaneous fitting of the 15N R2 dispersion data of G130, V165 and D166 using the measured R20 and a two-site exchange model. These residues undergo fast exchange (7000 ± 2000 s−1), where the population of the less favorable conformation is ≈ 0.4%. The fast conformational exchange of G130, V165 and D166 might be related to an intermediate time scale (μs to ms) conformational fluctuation in the loop connecting β2 and α2 (residues 168−174) which is most likely responsible for the severe line broadening of backbone resonances of these residues in wild type and all four mutant PrPs (5,26). The loop between residues 170−175 has been implicated in disease (53,54); overexpression of PrP(170N, 174T) causes spongiform encephalopathy disease in the mouse (55). Interestingly, this loop region of human, cow, mouse, dog, and cat PrP is flexible, whereas that of elk, Syrian hamster and bank vole is rigid (6,26,54,56-58), providing insights into species barriers for prion disease.
Unlike wild type and other mutants, the H186R mutant has two sets of resonances at residues Y162 and R163 in β2 at pH 5.5, with one set having chemical shifts similar to those of the other mutants and wild type proteins (Supplementary Fig. S1). M128, L129 and G130 in β1 do not exhibit this slow conformational exchange although they have slightly broader resonances than other residues. We attribute these observations to a slow exchange between two conformations (the rate of exchange is too slow to be detected in the 15N longitudinal 2-spin-order exchange (zz-exchange) experiments; data not shown). In the conformation with chemical shifts different from wild type and other mutants Y162 shows significantly more flexibility:([1H]-15N NOEs are 0.19 ± 0.02 and 0.29 ± 0.04 at 500 and 600 MHz, respectively (Fig. S3G) and S2 is 0.34 ± 0.01) than the conformation with chemical shifts similar to wild type and the other mutants: ([1H]-15N NOEs are 0.72 ± 0.09 and 0.61 ± 0.06 at 500 and 600 MHz, respectively and S2 is 0.92 ± 0.01). The H186R mutation appears to destabilize the β2 region, resulting in an equilibrium between ordered and disordered states. Interestingly, the Y162 peak originating from the more rigid conformation disappears at pH 3.5, while the other Y162 peak originating from the more flexible conformation remains (S2 = 0.30 ± 0.01), suggesting that the conformational equilibrium in the β2 region of H186R shifts toward the disordered state at lower pH.
Since our results indicated that H186 is involved in the destabilization of PrP that occurs as the pH is lowered, the protonation states of H186 were examined at pH 5.5 and 3.5 by 15Nδ1 and 15Nε2 chemical shifts (assignments for the side chain resonances of the 5 histidines are shown in Supplementary Fig. S5). The resonance frequencies of 15Nδ1 and 15Nε2 are at ≈ 168 and ≈ 250 ppm respectively in a neutral imidazole and they both resonate at ≈ 177 ppm when completely protonated (29). Intermediate chemical shift values are indicative of fast exchange between protonated and deprotonated states. Among the five histidines in wild type mouse PrP(89−230), H95, H110, H139 and H176 are exposed on the surface and H186 is partially buried and surrounded by hydrophobic residues (5). The 15Nδ1 and 15Nε2 chemical shifts indicate that H110 and H176 are protonated at pH 5.5; H95, H139 and H186 are partially deprotonated at pH 5.5 but become protonated at pH 3.5 (Supplementary Table S2 and Fig. 8), indicating that the pKas of H95, H139 and H186 imidazole side chain are substantially lower than the typical pKa value (6.6 ± 1); significantly lowered pKa values are frequently observed for histidines located in the interior of proteins (59-61). Theoretical pKa calculations also estimate a consistently low pKa for H186 (62), supporting the notion that the low pKa of H186 is due to its partially buried environment. In addition, the broad lines of the H186 imidazole resonances in the HMQC spectrum at both pH 5.5 and pH 3.5 (Fig. 8) suggest the presence of chemical exchange on a μs-ms time scale.
It is surprising that H95 and H139 display pKas as low as H186, despite their apparent surface exposure. There do not appear to be substantial chemical shift differences in the vicinity of these residues upon pH change, in contrast to H186 (Fig. 3). The low pKas of the H95 and H139 imidazoles may be due to dynamic effects such as transient hydrophobic or electrostatic interactions.
The initiation and subsequent processes of the conformational changes of prion proteins that will ultimately lead to onset of disease are not well understood. This study with wild type and four mutant PrPs at two pHs suggests that the segment spanning the C-terminal half of the α2 helix to the beginning of α3 (residues ≈187−197) may initiate conformational changes among inherited pathogenic, dominant negative mutants and wild type PrP. Many of the pathogenic mutations causing GSS and familial CJD (H186R, T187R, T187K, T187A, E195K, and E199K; residue numbers as in mouse PrP) are clustered in this region (63). Notably, some individuals carrying the H186R mutation develop neurological dysfunction in childhood (19). At acidic pH, the positive charge introduced by protonation of the partially-buried H186 disrupts the surrounding hydrophobic interactions, resulting in destabilization of the C-terminal half of α2. Low pH has been reported to trigger dramatically increased exposure of hydrophobic residues on the surface of PrP (64), and molecular dynamic simulations of human PrP suggest disruption of the C-terminal half of α2 under mildly acidic conditions (65). In addition, the amino acid sequence of helix α2 (DCVNITIKQHTVTTTTKG) includes many features that are atypical, including a sequence of β-branched side chains TVTTTT (187−192) that is rare in the context of a helical conformation; such stretches are usually found in β-strand and loop conformations (66). Even though we identified a dynamically labile site by applying a pH change, the conformation and dynamics of this site may also be susceptible to changes induced by other factors such as hydrophobic and electrostatic perturbations.
In accord with our results, molecular dynamics simulations of the helices from mouse PrP showed that the C-terminal half of α2 (residues 184−194) and parts of α3 (residues 200−204 and 215−223) undergo transitions from α-helical structure to a β and/or random coil state (66). High pressure NMR data also showed that a metastable conformer of PrP existing at ≈ 1% population has disordered α2 and α3 at pH 5.2 and 30°C (67). Further support is provided by a recent in silico screening based on the differential dynamics of this region: a compound was discovered that specifically binds the region from α2 (V189, T192, and K194) to the α2-α3 loop (E196) and inhibits formation of PrPSc (68).
The amino acid sequence of the C-terminal half of α2 is well conserved among mammalian PrPs from different species (69), but is divergent for non-mammalian PrPs such as chickens, turtles and frogs in which there are no reports of prion diseases (70). This region may therefore have a role in the cellular function of mammalian PrPs: deletion of the corresponding segment of the prion protein analog Doppel (Dpl) abolishes its neurotoxic effect (71). Dpl has ≈ 25% sequence identity with PrP and causes late-onset ataxia when over-expressed in the absence of cellular PrP (72). Dpl has the same topology as PrP but the C-terminal region of α2 is kinked at the residue that corresponds to T187 in mouse PrP (73); flexibility of this and following residues in Dpl closely matches our results for PrP (73).
Many of the inherited pathogenic mutations in the human PrP gene are located in residues 177−219 and the majority of them are associated with change of electrostatic charge (D178N, H187R, T188R, T188K, E196K, E200K, D202N, R208H, E211Q, Q217R, and E219K; residue numbers as in human PrP) (1-4). Since the proteins corresponding to many inherited disease-causing PrP mutations are as stable as wild type PrP (74,75), the effect of the pathogenic PrP mutation is unlikely to be related to the thermodynamics of PrP in the native state. In addition, the solution structure of a pathogenic human PrP mutant (E200K) is nearly identical to that of the wild type (76). In line with these observations, comparison of the chemical shift differences and backbone dynamics of wild type, P101L, Q167R, H186R and Q218K mutant of the mouse PrPs in this study shows that none of these mutations leads to major conformational and dynamical changes relative to the wild type. An exception is the H186R mutant PrP that displays inherent, pH-independent flexibility at the C-terminal half of α2, demonstrating the importance of the H186 protonation state and the consequent disruption of nearby hydrophobic interactions for the destabilization of PrP structure at this site. Our data raise the possibility that the pathogenic or dominant negative mutations exert their effects on some non-native intermediate form such as PrP* after conversion of PrPC into PrPSc has been initiated. Depending on the amino acid substitution, rearrangements of hydrophobic groups can be facilitated or inhibited depending on the nature of the charge disturbance.
We thank John Chung and Gerard Kroon for assistance with NMR experiments, and Professor Gerald Stubbs for helpful suggestions and a critical reading of the manuscript.
This work was supported by grant AG21601 from the National Institutes of Health. SHB was supported by a fellowship from the Korea Research Foundation Grant funded by the Korean Government (MOEHRD, Basic Research Promotion Fund) (KRF-2006-352-C00036).