Dynamics of a truncated prion protein, PrP(113–231), from ¹⁵N NMR relaxation: Order parameters calculated and slow conformational fluctuations localized to a distinct region

Abstract

Prion diseases are associated with the misfolding of the prion protein (PrP^C) from a largely α-helical isoform to a β-sheet rich oligomer (PrP^Sc). Flexibility of the polypeptide could contribute to the ability of PrP^C to undergo the conformational rearrangement during PrP^C–PrP^Sc interactions, which then leads to the misfolded isoform. We have therefore examined the molecular motions of mouse PrP^C, residues 113–231, in solution, using ¹⁵N NMR relaxation measurements. A truncated fragment has been used to eliminate the effect of the 90-residue unstructured tail of PrP^C so the dynamics of the structured domain can be studied in isolation. ¹⁵N longitudinal (T₁) and transverse relaxation (T₂) times as well as the proton-nitrogen nuclear Overhauser effects have been used to calculate the spectral density at three frequencies, 0, ω_N, and 0.87ω_H. Spectral densities at each residue indicate various time-scale motions of the main-chain. Even within the structured domain of PrP^C, a diverse range of motions are observed. We find that removal of the tail increases T₂ relaxation times significantly indicating that the tail is responsible for shortening of T₂ times in full-length PrP^C. The truncated fragment of PrP has facilitated the determination of meaningful order parameters (S²) from the relaxation data and shows for the first time that all three helices in PrP^C have similar rigidity. Slow conformational fluctuations of mouse PrP^C are localized to a distinct region that involves residues 171 and 172. Interestingly, residues 170–175 have been identified as a segment within PrP that will form a steric zipper, believed to be the fundamental amyloid unit. The flexibility within these residues could facilitate the PrP^C–PrP^Sc recognition process during fibril elongation.

Introduction

Transmissible spongiform encephalopathies or prion diseases are a group of fatal neurodegenerative disease that can be sporadic, inherited, or acquired. The infectious agent of these diseases is devoid of nucleic acid and is almost entirely composed of the prion protein, PrP^Sc, which is a misfolded form of a cellular prion protein PrP^C.¹ Recombinant prion protein has now been shown to be capable of propagating prion disease.² It is yet to be established what structural relationship small toxic replicating oligomeric species have with the more mature amyloid fibers.³ It is clear the structure, flexibility, and stability of native PrP^C as well as folding intermediates of PrP are of significant interest to those studying the mechanism of prion misfolding and replication.

Post-translational processing of mammalian prion protein typically results in a ∼208 residue glycoprotein that possesses two N-glycosylation sites and one disulfide bridge. Mature PrP^C is then tethered to the cell surface via a glycosyl-phosphatidylinositol anchor at the C-terminus.¹ Three dimensional NMR solution structures of non-glycosolated PrP^C from a number of mammalian species have been determined, including mouse,⁴ Syrian hamster,^5–7 human,⁸ bovine,⁹ cats, dogs, pigs, sheep, chicken, turtles, frogs, and elk.^10–12 There has been some recent progress in defining the molecular architecture of prion amyloid fibers.^13,14

The mammalian prion proteins have a high homology at the sequence and structural level. All mammalian PrP^C studied consist of two structurally distinct domains. The N-terminal domain (residues 23–120) is highly disordered¹⁵ and is notable for its ability to bind Cu²⁺ ions.^16–21 The N-terminus contains a highly conserved octa-repeating sequence, PHGGGWGQ between residues 57 and 90. The flexible N-terminal tail of PrP^C has shown some ordering at a lower pH of 4.²² The C-terminal domain (residues 126–231) is predominantly helical, containing 3 α-helices as well as two short anti-parallel β-strands. For mouse, PrP^C helix A, B, and C include residues 144–154, 175–193, and 200–219, respectively,^4,7 and for the β-sheet residues 128–131 and 161–164.^4,7

The process of prion propagation to form amyloids has been described as a nucleation or template-assisted process, via an intermediate, PrP*. The native form of PrP is readily formed between pH 5 and 7, whereas amyloid fibers are favored under neutral pH, moderately denaturing conditions and agitation.²³ The third well characterized form of PrP is a soluble oligomeric species favored at low pH (∼4) under mild denaturing conditions with a high salt content.^24–28 CD indicates this low pH oligomer has a high β-sheet content, and solution NMR studies have shown that the N-terminal tail retains its mobility in this oligomeric form.²⁸

¹⁵N NMR relaxation measurements are an unique method for probing the main-chain motions of proteins on a per-residue basis.^29,30 Slow conformational fluctuations on a milli-to-microsecond time-scale can highlight regions of PrP^C that sample alternative main-chain conformations. These flexible regions could facilitate protein–protein docking. ¹⁵N relaxation measurements have previously been reported for full-length PrP(29–231)¹⁵ and the 90–231 fragment.^15,31,32 However, attempts to obtain order parameters (S²) for each NH-bond vector have been unsuccessful for the majority of residues in helix B and C for these PrP^C constructs.¹⁵ These fragments contain a significant amount of the unstructured tail. This long tail, residues 23–120, may have a profound influence of the relaxation behavior of the structured C-terminal domain of PrP^C. Thus, the aim of this study was to use a truncated version of PrP^C, residues 113–231, that contains almost exclusively the structured domain of PrP^C. The PrP(113–231) construct was chosen because this fragment is found to occur in vivo.³³ This shorter fragment will not have its dynamics dictated by the unstructured tail and will facilitate a more detailed analysis of per-residue motions of the structured core of PrP^C.

Results

Assignments of mouse PrP(112–231)

The protein fragment of mouse PrP(113–231) gives well resolved ¹⁵N-HSQC spectra. Assignment of the amide resonances was completed using 3D ¹⁵N HSQC-NOESY and TOCSY data. A strip-plot of the ¹⁵N HSQC-NOESY spectrum is shown in Figure 1 for the region of PrP^C containing the short stretch of anti-parallel β-sheet. Sequential and cross-strand NOEs are indicated. Although there is strong cross-strand H_N-H_N and H_α-H_N NOEs observed between M128 and Y162 and L129 and Y162, respectively, other cross-strand diagnostic NOEs are very weak. Similarly, sequential NH-NH and αH-NH are shown (Supp. Fig. S1) for a representative stretch of α-helix, between residues 171 and 182. Consistent with other PrP^C structures our construct contains three α-helices and a short, two-stranded β-sheet. Helical boundaries are as follows: helix A, D143 – N152; helix B, N171 – K193; helix C, E199 – Y225, based on the observation of characteristic NOEs. Helical boundaries defined by the published mouse PrP structure are more conservatively reported as follows; 144–154, 175–193, and 200–219.⁷ Although it is unnecessary in our case to solve a complete 3D structure, NOEs between secondary structure elements ensures their correct orientation in space. For instance, NOEs observed in R163 may originate from I181 and place the β-sheet adjacent to helix B. The majority of non-sequential NOEs were not assigned.

Figure 1.

Series of strips from ¹H-¹H_N slices of the 3D ¹⁵N NOESY-HSQC collected at 600 MHz (¹H) and 30°C. Each strip corresponds to a single residue in the two-stranded β-sheet, β1 (G125–S131), and β2 (Q159–R163). Solid lines connect sequential residues, dashed lines highlight connections occurring between the strands. The assignments are shown at the top of each strip.

Amide main-chain assignments are indicated for the ¹⁵N-HSQC shown in Figure 2, together with the mouse PrP primary sequence. All but eight main-chain amides have been assigned. Unassigned residues include: A115; A116; A119; and V120 from the N-terminus, as well as S169, F174, S230, and S231. The assignment process has been aided by backbone assignment reported for a number of other mammalian PrP's including, hamster, human, and bovine which have high sequence homology with mouse PrP.¹² TOCSY cross-peaks and NOEs allowing sequential assignment of residues forming a loop (residues 166–170) between β–sheet S2 and α–helix B are very weak (D166, Q167, Y168, N170) or missing (S169) and assignments for these residues are only tentative. Residues 166–170 were not used in detailed relaxation analysis because of spectral overlap for these residues in the ¹⁵N-HSQC.

Figure 2.

¹⁵N HSQC assignments for mPrP(113–231). (a). Mouse PrP sequence of the C-terminal domain used in this study. (b) Annotated ¹⁵N HSQC spectrum. Numbering relates to the sequence in (a). The C-terminal His-tag (232–239) is unassigned. Dashed lines connect side chains of unassigned Q and N sidechains. Boxes highlight folded arginine side-chain amine groups (w₁ = w_1,app – 30 ppm). The “indole” peak is from Trp 144. The A and S are Ser and Ala residues at the N-and C-terminus.

¹⁵N relaxation measurements

To gain insight into the motion of the backbone of PrP in solution ¹⁵N longitudinal (T₁) and transverse (T₂) relaxation times as well as ¹⁵N heteronuclear NOEs were measured at 600 MHz proton frequency for the truncated structured domain of mouse PrP residues (113–231). From a possible total of 115 main-chain amides (118 minus three prolines), we found 83 residues sufficiently dispersed to warrant relaxation analysis on a per-residue basis. In addition to eight unassigned amides, 24 more residues were not adequately resolved to permit full relaxation analysis. The residues not included in the analysis were: 115, 116, 119, 120, 122, 124, 139, 142, 145, 147, 155, 159, 166–170, 174, 177, 179, 184, 194, 195, 201, 202, 210, 212, 218, 225, 227, 230, and 231.

The R₁ (1/T₁) and R₂ (1/T₂) relaxation rates and heteronuclear NOEs for mPrP(113–231) are shown in Figure 3. The ¹⁵N heteronuclear NOE values for helix A, B, and C are very similar, with values between 0.75 and 0.85. There is a dip in ¹⁵N heteronuclear NOEs for the N-and C-terminus and residues between 190 and 197. Residues before 123 give negative heteronuclear NOEs indicating high mobility of the main-chain. The R₁ values are largely invariant for the bulk of the residues with typical relaxation rates between 1.3 and 1.6 s⁻¹ apart from six residues at the N-terminus, which exhibit reduced R₁ rates. The R₁ values for residues in Helix C are generally lower than in the other two helices. R₂ values show the largest range of variation with mean R₂ values for helix A at 11.2 s⁻¹, which are consistently less than residues in helix B and C with mean R₂ rates of 12.1 s⁻¹ and 13.5 s⁻¹, respectively. Following the trends in heteronuclear NOE values, R₂ rates are reduced for the N-and C-terminus and residues between 188 and 198. In addition, eight residues produce distinctly large R₂ rates when compared with the rest of the protein: G130, 16.3 s⁻¹; V165, 20.3 s⁻¹; Q171, 16.5 s⁻¹; N172, 16.5 s⁻¹; Q216, 15.2 s⁻¹; Y217, 15.1 s⁻¹; E220, 15.3 s⁻¹; and S221, 15.9 s⁻¹. The latter four residues are within helix C and are appreciably larger than the mean R₂ rate for helix C of 13.5 s⁻¹.

Figure 3.

Relaxation data for mPrP(113–231). (a) Heteronuclear NOE and the relaxation rates, (b)R₁ (=1/T₁), and (c)R₂ (=1/T₂) determined at 600 MHz. Sample conditions: 30°C; pH 5.3; 20 mM acetate buffer.

Reduced spectral densities

J(0.87ω_H)

The reduced spectral density function, J(ω), provides insights into the motion of the NH bond vector at three frequencies, zero, ω_N, and 0.87ω_H.^34,35 Spectra densities at these three frequencies are shown in Figure 4 on a per-residue-basis. The high frequency spectral densities J(0.87ω_H) are particularly useful in indicating fast pico-second motions typically observed in flexible regions of proteins. These are often associated with rapid hydrogen-bond rearrangements. J(0.87ω_H) spectra densities have been mapped on to the structure of PrP^C (see Supp. Fig. S2). It is clear that large J(0.87ω_H) are localized to the N-and C-termini of PrP(113–231), residues 113–127, and 222–231 and a loop region between helix B and C, residues 190–197. Much of the rest of the structured domain shows little fast pico-second time-scale motions. The extended polypeptide between helix A and β-strand two lacks the fast motions and has comparable J(0.87ω_H) to all three helixes. In addition, residues within and close to the short anti-parallel β strand-1 (residues 128–131) and strand-2 (residues 161–164) do not exhibit fast pico-second time-scale motions. It should be noted that J(0.87ω_H) and J(ω_N) are calculated independently of R₂ values and are therefore not influenced by slow R_ex milli-second to micro-second motions.

Figure 4.

Reduced spectral density functions: (a) J(0), (b) J(ω_N), and (c) J(0.87ω_H) for mPrP(113–231) at 600 MHz. The high frequency spectral densities, J(0.87ω_H), indicate fast pico-second motions typically observed in flexible regions of proteins. J(0) indicates sub-nanosecond flexibility of the NH bond vector, the smaller the value of J(0) the greater the flexibility. Uncharacteristically large J(0) indicate residues with slow conformational fluctuations. Sample conditions: 30°C; pH 5.3; 20 mM acetate buffer.

J(ω_N)

For small proteins, J(ω_N) will decrease with increased internal flexibility. As the protein size increases this effect becomes less apparent, and for very large proteins the effect is reversed and J(ω_N) will increase with increased flexibility.²⁹ For PrP(113–231), a relatively small protein fragment, the flexible N-terminal residues, between 113 and 125, exhibit a reduction in J(ω_N). A reduction in J(ω_N) spectral densities relative to the bulk of the protein is not apparent for the C-terminal residues. Residues in helix C (residues 200–219 with helical character up to residue 225) show a very slight reduction in J(ω_N) spectral densities. However, helix C does not exhibit increased flexibility according to both J(0.87ω_H) and J(0) spectral densities. This reduction in J(ω_N) reflects anisotropic tumbling of PrP^C. As discussed later, N—H bond vectors in helix C are aligned with the long axis of the protein and therefore have a longer rotational diffusion time relative to helix A and B.

J(0)

Assuming there are no internal motions, the spectral density function J(0) is equal to 2/5 τ_m, where τ_m is the rotational correlation time. A value for J(0) less than 2/5τ_m indicates internal flexibility of the NH bond; the smaller the value of J(0) the greater the sub-nanosecond flexibility of the NH bond vector.²⁹ There is a good deal of variation of the J(0) spectral densities. Residues in the three helices show some of the largest J(0) spectral density although there is some variation in the values, with helix A giving mean values of 3.0 ns, whereas helix B and C produce larger values for J(0) of 3.2 ns and 3.6 ns, respectively. Mirroring the J(0.87ω_H) spectral densities residues from the N-and C-termini and a loop region between helix B and C show a marked reduction in J(0) spectral densities also suggesting increased flexibility of the NH bond vector.

R_ex motions

In addition to sub-nanosecond flexibility of the NH bond vector, slow micro-to millisecond motions can also be reflected in J(0) spectral densities, as an increase in J(0) values. The slow ms-μs (R_ex) motions have no effect on J(ω_N) or J(ω_H) spectral densities and so may be used to isolate these motions observed in the J(0) spectral densities. Eight residues can be identified as having J(0) spectral densities significantly greater than the bulk of the protein with J(0) greater than 4 ns, shown in Figure 4. The cutoff of 4.0 ns was set as two standard deviations greater then the mean J(0) value for helix C. The mean J(0) for helix C (residues 200–215) is 3.56 ns with a standard deviation of 0.22 ns. These eight residues do not exhibit distinct J(ω_N) or J(ω_H) spectral densities suggesting that the large J(0) values are because of exchange (R_ex) motions on a milli-microsecond time scale. Typically, the slow motions that cause exchange broadening are due to large segmental rearrangements rather than single H-bond rearrangements. Interestingly, four of these residues are within or close to the anti-parallel β-sheet. The J(0) spectral densities for these four residues are as follows: G130, 4.4 ± 0.2 ns; V165, 5.5 ± 0.2 ns; Q171, 4.5 ± 0.1 ns; and N172, 4.5 ± 0.1 ns. Neighboring residues 129 and 131 are resolved (in the ¹⁵N HSQC spectra, Fig. 2) as are residues from the other side of the anti-parallel β-sheet, in particular, residues 160–163, whereas residue 164 is a proline. These residues do not show significantly distinct J(0) from the bulk of the protein. In addition, four other amides exhibit J(0) spectral densities greater than 4.0 ns and are located at the end of helix C. These residues have larger J(0) spectral densities (two standard deviations from the mean) than the other residues within helix-C. These four residues Q216, Y217, E220, and S221 represent a continuous face on the C-terminal end of helix-C. Residues in the loop between β-strand 2 and helix B (166–170) may also exhibit R_ex motions similar to that observed for V165, Q171, and N172. It has not been possible to assign this loop (166–170) because of a lack of adequate TOCSY and NOESY data combined with spectral overlap, consequently a detailed relaxation analysis of these residues has not been possible. However, weak or missing TOCSY and NOE peaks for these residues hint at exchange broadening as observed for the adjacent residues 165, 171, and 172.

Order parameters (S²) and rotational correlation time (τ_c)

τ_c

An effective rotational correlation time (τ_c,i) at each residue has been calculated from J(0) and J(ω_N) spectral densities using Eq. (1).^36,37 PrP(113–231) exhibits a good deal of variation in the rotational correlation time τ_c,i per residue, shown as Supporting Figure S3. This reflects both fast internal motions and slow segmental rearrangements in addition to the over-all rotational diffusion time. A trimmed mean rotational correlation time has been determined in order that only rigid amide nitrogens are used in the calculation of τ_c.³⁸ This has been achieved by using only residues that have heteronuclear NOEs greater than 0.72, so as to eliminate fast pico-second motions. Furthermore, the residues with unusually high R₂ rates, reflecting slow R_ex motions, have also been eliminated from the calculation of rotational correlation time. This trimmed rotational correlation time produces a mean value of 9.0 ns for residues within the three helices. There is still some variation in the rotational correlation time, residues in helix A give consistently lower τ_c values with a mean of 8.1 ± 0.2 ns, residues in helix B have a mean τ_c value of 8.4 ± 0.6 ns, whereas helix C gives consistently higher values with a mean of 9.9 ± 0.4 ns as is apparent from the Supporting Figure S3. If the protein does not approximate to a sphere, then individual τ_c values will vary according to the angle between the NH bond vector and the long axis of the protein.³⁹ The differences in τ_c from the three helices reflect anisotropic tumbling within the molecule. The variation in τ_c between helices have been shown to correlate well with asymmetric hydrodynamic radius calculated for PrP^C with the long axis, R_z of 24 Å and 16 and 12 Å for the shorter axis, R_x and R_y.¹⁵ It has previously been shown from calculations using the PrP^C solution structures show that the N—H bond vectors of helix C are parallel to the long axis, whereas helix A is perpendicular to it.¹⁵

S²

The model-free formalism^40,41 is often used to describe the motions of a protein. In particular, the order parameter (S²) describes the amplitude of the nanosecond time-scale motions and can range between 0 and 1.0, representing complete disorder and complete order, respectively. Approximate order parameters have been directly calculated from J(0) and J(ω_N) spectral densities using Eq. (2),^36,37 rather than a statistical selection of various models of motion.⁴² The correlation time τ_c for the protein has been allowed to vary on a per-residue-basis. The order parameters for mPrP(112–231) are shown in Figure 5(a), and the S² values have been mapped on to the structure of mPrP^C, Figure 5(b). Apart from the four residues identified as containing R_ex motions, all residues produce order parameters less than 1.0. The three helical regions give very similar order parameters, typical for a well ordered protein, with very similar mean S² values of 0.91, 0.92, and 0.92 for helices A, B, and C, respectively. Residues between the first β-strand and helix A exhibit very slightly increased flexibility (in particular I138) with S² values of ∼0.8. Considerable flexibility, with S² < 0.75, are observed for residues between 113 and 125 at the N-terminus, the loop region between helix B and C (residues N196 and F197) and residue R229 at the C-terminus of PrP^C.

Figure 5.

(a) Order parameters (S²) for mPrP(113–231). Order parameters describe the amplitude range of the nanosecond time-scale motions between 0 (flexible) and 1.0 (rigid). Residues that exhibit additional R_ex exchange motions are highlighted in red. (b) Order parameter mapped onto the structure (1 xyx) of mPrP^C: red S² > 1.0 (R_ex motions); blue 0.85 < S² < 1.0; cyan 0.75 < S² < 0.85; green 0.65 < S² < 0.75; and yellow S² < 0.65, residues in gray have no value calculated. (c) Slow conformational fluctuations are localized to a distinct region within PrP^C, the position of side-chains that exhibit R_ex motions are highlighted, dots are proton positions. (d) The molecular surface, in red, generated by the residues that exhibit R_ex motions, within the context of the whole structured domain. This figure was generated using GRASP2.

Discussion

Order parameters calculated for truncated PrP^c

¹⁵N relaxation measurements have previously been reported for full-length PrP(29–231)¹⁵ and the 90–231 fragment.^15,31 The full-length protein and truncated fragments contain a significant amount of the unstructured tail, residues 23–125. This long-tail attached to the C-terminal globular domain of PrP^C appears to have a profound influence of the relaxation behavior of the structured C-terminus. In particular, attempts to obtain an order parameter (S²) for each NH bond vector have been unsuccessful for the majority of residues in helix B and C, for both PrP(29–231) and the shorter fragment PrP(90–231),¹⁵ with S² driven to values considerably higher than the physical limit of 1.0. One explanation is that the long unstructured tail causes disproportionately short T₂ values relative to T₁ and heteronuclear NOEs making the determination of an order parameter intractable. This was somewhat of a surprise as other unstructured proteins have been readily fitted to model free formalism by using a variable rotational correlation time calculated per-residue.⁴³ Furthermore, the removal of a significant proportion of the unstructured tail of PrP^C residues 29–90, had no effect on the dynamics of the protein as ¹⁵N relaxation parameters for SHaPrP(90–231) and SHaPrP(29–231) are strikingly similar.¹⁵ Alternatively, we postulated that much of the structured domain might exhibit pronounced additional slow segmental exchange motions, and these would cause disproportionately short T₂ times. A large number of residues exhibiting significant exchange motions on a milli-second to micro-second time scale was a particularly appealing hypotheses to explain the unusually long transverse relaxation rates, bearing in mind the ability of PrP^C to misfold into a pathogenic isoform. With this in mind, we set about determining the ¹⁵N relaxation behavior of the truncated fragment of PrP^C containing only the structured domain residues 113–231, so eliminating the effect of the unstructured tail. Comparison of mPrP(113–231) with relaxation data for the longer fragment and full-length PrP highlights the influence of the long unstructured tail on the hydrodynamics of the whole protein. Relaxation data published for full-length Syrian-hamster PrP, and the 90–231 fragments were recorded at the same temperature and pH as this study¹⁵ and therefore permits direct comparison.

The magnitude and trend of R₁ rates and heteronuclear NOEs are strikingly similar for all three fragments; SHaPrP(29–231), SHaPrP(90–231),¹⁵ and mPrP(113–231). In particular, R₁ values for the bulk of residues are largely invariant and produce R₁ rates of ∼1.5 s⁻¹ for all three constructs. Similarly, heteronuclear-NOE values for residues within the helices of PrP^C exhibit very similar values for all three constructs – typically 0.80. In contrast, R₂ values for the short fragment that lacks the unstructured N-terminal tail exhibits strikingly different R₂ rates. The structured domain, particularly the three helices, give R₂ values between 16 and 20 s⁻¹ for both SHaPrP(90–231) and SHaPrP(29–231), whilst for PrP(113–231), R₂ rates are significantly reduced with average values between 11.2 s⁻¹ (helix A) and 13.5 s⁻¹ (helix C). The reduction in the R₂ rates between PrP(90–231) and PrP(113–231) can not simply be accounted for by the very small reduction in molecular weight that is the loss of 14 amino acids (from 141 amino acids to 118 + 9, for the His-tag). It is clear that the presence of a flexible tail causes the disproportionably large R₂ rates relative to R₁ and heteronuclear NOEs making the determination of an order parameter intractable. Removal of the flexible tail significantly reduces R₂ rates while leaving R₁ and heteronuclear NOEs largely unaffected. This means the R₂ values reflect the hydrodynamics of a globular protein and order parameters can readily be calculated.

It is clear from this study that even when the rotation correlation time is permitted to vary, per-residue, the calculation of an order parameter may not be appropriate for proteins with appreciable portions of unfolded polypeptide chain. Thus each N—H vector is better described by a distribution of rotational correlation times.⁴⁴ Our study suggests that large portions of PrP^C do not exhibit slow R_ex motions, as simply removing the flexible tail causes a dramatic reduction in R₂ values. The effect on the R₂ rates is sufficiently significant to facilitate, for the first time, the determination of order parameter (S²), for the PrP(113–231) fragment. It is notable that order parameters (S²) calculated do not require a term for R_ex motions and yet the S² values are typical for a globular protein for the overwhelming majority of residues. In contrast, the majority of residues for full-length PrP have S² values in excess of 1.0.

J(0) and S² can reflect internal flexibility of the NH bond vector, in particular, a value for J(0) less than 2/5τ_m indicates internal flexibility of the NH bond.²⁹ However, J(0) is strongly influenced by the rotational correlation time of the protein, and for this reason anisotropic tumbling can influence J(0) spectral densities. This is the case for PrP^C, where it has been shown that the structured domain is not isotropic with a the long axis, R_z of 24 Å and two shorter axis, R_x; 16 and R_y; 12 Å.¹⁵ It has been shown that helix C, the longer of the three helices, is aligned with long axis while helix A is perpendicular to it. This anisotropic tumbling results in mean J(0) values for helix C larger than helix A; 3.6 ns relative to 3.0 ns for mPrP(113–231). Calculations using the PrP^C solution structures show that the N—H bond vectors of helix C are parallel to the long axis while helix A is perpendicular to it. In contrast, S² order parameters indicate H—N bond motional amplitude distinct from a rotational correlation time. The value set for the rotational correlation time will clearly influence the order parameter.⁴⁵ However, here we allow τ_c to vary per residue, which therefore compensates for the effect of an anisotropic rotational diffusion within PrP^C, as shown in Supporting Figure S3. Inspection of the S² data [Fig. 5(b)] shows that all three helices have very similar mean S² values 0.91, 0.92, and 0.92 for helices A, B, and C, respectively. It is clear that an asymmetric rotational diffusion is reflected in J(0) values, but the values of the order parameters, calculated with a variable τ_ci are not influenced by asymmetric rotational diffusion.

The similarities between structures of mammalian PrP species are well documented.⁸ Sequence comparison of mouse and Syrian hamster shows a 94.2% homology in this fragment, there are six largely conservative substitutions. It is clear from this study that mouse and hamster PrP have very similar dynamical properties. Comparison of J(0.87ω_H) spectral densities between mouse and hamster show very similar trends. In particular, the flexible N-terminal tail stretches between residues 23 and 125, whereas the flexibility at the C-termini is limited to the last few residues. In addition, both proteins have a flexible loop between helix B and C, residues 190–197, which possesses fast pico-second time-scale motions.

PrP^C folding intermediates and stability

PrP*, the precursor for amyloid formation, could require the complete unfolding of PrP^C. Alternatively, PrP^C could be a kinetically trapped intermediate to the formation of PrP* and PrP^Sc. Finally, PrP* might represent a sampled conformation formed on the way to a more stable native PrP^C, as described in Supporting Figure S4. It is yet to be established whether PrP^Sc retains any of the structural elements found in PrP^C, although recent data suggests that residues 160–220 undergo significant conformation change to form amyloid fibers.^13,24 In contrast, modeling based on 2D crystals,⁴⁶ and epitope mapping studies⁴⁷ of PrP^C and PrP^Sc, as well as the presence of the intra-molecular disulphide in PrP^Sc,⁴⁸ suggests that much of helix B and C may be retained in the scrapie form. At face value, the J(0) values hint at the possibility that helix A of PrP^C has relatively more flexibility than helix B and C. However, the effect of anisotropic tumbling needs to be taken into account. This has been achieved by determining the order parameters, which give very similar S² values for all three helices. Hence, helix A does not have any more internal flexibility than the other helices. This observation agrees with measurements of PrP^C amide protection factors.⁴⁹ This study indicated that all three helices have similar hydrogen-bond protection factors, apart from a few amides adjacent to the single disulphide bond within PrP, Cys190–Cys214, which exhibit additional amide protection.⁵⁰ These observation support reports that, at neutral pH, PrP^C undergoes essentially a simple two state unfolding, from native to fully unfolded.^51,52 An intermediate in which, for example, helix A is unfolded and helix B and C remain intact is not suggested.

Studies using hydrostatic pressure⁵³ have been used to study the unfolding of PrP^C in an attempt to identify folding intermediates. These folding intermediates may represent conformations of the prion protein (PrP*) on a pathway to PrP^Sc and amyloid formation. Perhaps contradictory to the modeling based on 2D crystals,⁴⁶ these studies suggest that a largely intact helix B and C is unlikely. Indeed, the NMR studies suggest that helix A actually has slightly more stability than helix B or C.^32,53 Recent amide protection studies of prion fibrils (PrP^Sc rather than PrP^C) have indicated that a relatively short stretch of the sequence constitutes the β-sheet rich core of PrP^Sc between residues 169 and 214,²⁴ again supporting the assertion that helix B and C are not retained as helices in the PrP* precursor to the scrapie isoform.

Mutations in the prion protein linked with familial prion disease are found throughout the PrP sequence, some of these residues, in particular D178N, T183A, and Q217R involve disruption of salt-bridges or hydrogen-bond net works and seem to destabilize the native structure.⁷ Other mutations are found in the unstructured region of PrP^C, residues; 102, 105, 117 for example. These mutations may stabilize the formation of PrP* and PrP^Sc rather than destabilize PrP^C. Furthermore, it has been shown that residues 90–126, although unstructured in PrP^C, are essential for prion propagation. Removal of these residues (as with the construct studied here) does not destabilize PrP^C supporting the hypotheses that these residues stabilize the formation of PrP*. It is interesting to note that a genetic mutation, which has a stop codon at residue 145, generates a truncated PrP fragment residues 23–145. This fragment lacks residues from helix B and C, but still causes a familial prion disease.⁵⁴

Slow conformational fluctuations are localized to a distinct region within PrP^C

Slow motions that cause exchange broadening of NMR signals are typically because of large segmental motions rather than single hydrogen-bond rearrangements. Residues exhibiting exchange broadening sample alternative conformations in exchange with the native conformation on a milli-micro second time-scale. Exchange motions in the short anti-parallel β-sheet, in particular at reside G130, are interesting as polymorphism in human PrP is found at the adjacent residue 129M/L. This polymorphism dictates strain-type in familial prion disease and is a key determinant of susceptibility to sporadic and acquired CJD. In particular, only individuals homozygous for methionine at 129 (Met/Met) (with one exception) have contracted variant CJD.⁵⁵ Slow exchange motions observed for mPrP at G130 are also observed for Syrian hamster PrP at the same position and at the adjacent β-sheet V161.

Residues 165, 171, and 172 also show large R_ex motions while residues in the loop between them (166–170, between the β-strand S2 and the start of helix B) give weak NMR signals, which may be a result of broadening because of R_ex motions. NMR studies on a number of different mammalian prion proteins have exhibited similar weak signals in this loop.¹¹ Although for hamster PrP residues between 165 and 172 do not seem to show marked exchange broadening.¹⁵ The N170S mutation, hamster to mouse, seems to explain the differences in flexibility for residues 165, 171, and 172.

Residues Q216, Y217, E220, and S221 in mouse PrP show large J(0) spectral densities (greater than 4 ns, >2 standard deviations from the mean J(0) for helix C) indicating that these residues may also exhibit some R_ex motions. The analogous residues for hamster PrP are not identified as having pronounced J(0) values although the noise in these larger fragments is considerably increased and could well account for them going undetected.

Inspection of the mPrP^C solution structure (1 xyx)^4,7 indicates that the residues exhibiting some slow exchange line broadening, 216, 217, 220, and 221, lie along the same face of helix C and interact with residues exhibiting significant R_ex motion in the loop between V165 and N172. Figure 5(c) shows that G130 is close to Q216, whilst tyrosine Y217 sits in behind the 165–172 loop and forms hydrophobic packing interactions with V165 and Q171, S221 is also spatially close to V165. Thus, rather than simply the β-sheet exhibiting slow conformational fluctuations in PrP^C, the residues exhibiting R_ex motions can all be mapped to a localized patch or inter-linked region, as indicated in Figure 5(d). As we have been unable to define the R_ex motion of residues 166–170, they don't appear in Figure 5(d), but would extend the localized patch of residues involved in slow conformational fluctuations. Intriguingly, residues Q167, Q171, V214, and Q218 have been identified as the binding site for the so-called “protein X”.⁵⁶ These residues exhibit, or are sequentially adjacent to residues that exhibit, R_ex motions. The binding of an as yet, unidentified molecule might destabilize PrP^C or stabilize a PrP* intermediate. Furthermore, a mutation in humans that causes familial prion disease, Q217R (Q216 in mouse), has been identified and is likely to destabilize this region.⁷

Interestingly, Doppel protein which has both sequence and structural homology with PrP⁵⁷ is less stable than PrP^C50 but will not propagate prions. One key difference with Doppel is the presence of a second disulphide. This disulphide is found between the loop region and helix C (equivalent residues in the mouse sequence are residues 166 and 221).⁵⁸ This is the same region that exhibits slow conformational fluctuations in PrP^C. Perhaps, a key difference with Doppel is the stabilization of this region of the protein via an additional disulphide bond.

Profiling methods for determining fibril-forming segments of proteins have identified residues 170–175 (SNQNNF) within mouse and human PrP.⁵⁹ This hexapeptide will form fibrils and the closely related microcrystals. It is suggested that the fundamental unit of amyloids is the tightly interdigitated β-sheets, termed steric zippers, formed by the cross-β stacking of these residues.⁵⁹ This sequence with its high complimentary interface might nucleate fibrilization. It is notable that residues N171 and Q172 on the surface of mPrP^C exhibit slow motion conformational fluctuations. It is a possibility that the flexibility within these residues could facilitate the recognition process as PrP^C docks on to the exposed ends of amyloid fibers during fiber elongation. The region identified here as possessing conformational fluctuations might represent a possible therapeutic target, inhibiting the PrP^C–PrP^Sc recognition process or stabilizing PrP^C.

Materials and Methods

Expression and purification of recombinant mPrP(113–231)

C-terminally His-tagged full-length mouse PrP(23–231) was cloned into a pET-23 vector as previously described.⁶⁰ Codons for amino acid residues 23–112 were deleted by mutation of the sequence to include and NdeI site prior to the codon for residue 113. After restriction digestion with NdeI (Promega), the plasmid was religated and expressed a truncated PrP fragment, PrP(113–231) which lacks the natively unstructured N-terminal domain. This fragment retains a methionine residue prior to residue 113. The protein was expressed in 2-L flasks of Escherichia coli BL21(DE3) cells in ¹⁵N labeled minimal medium containing (1 g/L) ¹⁵NH₄Cl (Cambridge isotope laboratories) as the sole nitrogen source. Cultures were grown at 37°C to an A₆₀₀ of 0.9 and the protein expression was induced by the addition of isopropyl-β-d-thiogalactopyranoside to a final concentration of 1 mM. Bacterial pellets were harvested after 4 h and sonicated in a resuspension buffer containing 8M urea, 200 mM NaCl, and 50 mM Tris, pH 7.6. The resulting solution was cleared of bacterial debris by centrifugation at 6000g for 15 min. The protein was absorbed to a nickel-charged metal affinity column made from chelating Sepharose (Amersham Biosciences). The proteins were washed with at least 20 column volumes of resuspension buffer and eluted from the column using the same buffer supplemented with 300 mM imidazole. The protein was found to be greater than 95% pure by polyacrylamide electrophoresis and Coomassie staining. The protein was refolded by successive rounds of dilution in deionized water and concentrated in a Vivaspin centrifugal concentrator (Vivascience) with a 10,000 molecular weight cut off. The refolded protein was dialyzed twice at 4°C against deionized water to remove residual urea. The concentration of mPrP(113–231) was measured by its absorbance at 280 nm and confirmed by Bicinchoninic Acid (BCA) assay (Sigma).

Preparation of NMR sample

NMR samples contained; 0.6 mL of 3.15 mg/mL His tagged ¹⁵N mouse PrP(113–231) in 20 mM sodium acetate and 10% D₂O at pH 5.32.

NMR resonance assignments

All NMR spectra were acquired at 30°C on a 600 MHz Bruker Avance spectrometer equipped with a 5 mm inverse detection triple-resonance z-gradient probe. Phase-sensitive 2D ¹⁵N-HSQC spectra were acquired using Echo-antiecho gradient selection. ¹H acquisition parameters were 0.122 sec acquisition time, 1 sec fixed delay and 2048 complex (t₂) points, whereas 256 complex points were collected for the ¹⁵N dimension. Thirty-two transients were recorded for each t₁ interval. The ¹H and ¹⁵N dimensions possessed spectral widths of 14 ppm and 32 ppm, respectively. The ¹⁵N dimension was zero filled to 256 data points before squared-cosine apodization and Fourier transformation. Three-dimensional phase sensitive ¹⁵N-HSQC-NOESY⁶¹ and ¹⁵N-HSQC-TOCSY spectra⁶² were acquired using Echo-antiecho gradient selection. The 3D ¹⁵N-HSQC-TOCSY and 3D ¹⁵N-HSQC-NOESY spectra were obtained using a mixing time of 60 ms and 150 ms, respectively. For both experiments, 2048 complex points were collected in the direct ¹H dimension over a spectral width of 12 ppm, whereas 64 complex points and 160 complex points were collected for the indirect ¹⁵N and ¹H dimensions respectively over a ¹⁵N spectral width of 32.2 ppm and a ¹H spectral width of 12 ppm. Both spectra were zero filled to 2048 data points in the direct ¹H dimension, 128 data points in the indirect ¹⁵N dimension and 512 data points in the indirect ¹H dimension. All NMR data was processed in XWINNMR (Bruker) and analyzed in XEASY.⁶³ Resonance assignments were made manually using sequential ¹H NOE connectivities with comparison to assignments deposited in the Biological Magnetic Resonance Databank (www.bmrb.wisc.edu) for related mammalian prion proteins, in particular human and Syrian hamster PrP.

¹⁵N NMR relaxation measurements

All relaxation measurements were performed at a temperature of 30°C. The pulse sequences used to obtain T₁, T_2, and heteronuclear NOE spectra are those described by Farrow et al.,⁶⁴ which employ gradient selection and sensitivity enhancement, as well as minimal water suppression.

Three pairs of saturated and non-saturated²⁰-¹⁵N heteronuclear NOE experiments were recorded at 600 MHz (60.8 MHz ¹⁵N frequency). In the direct (¹H) dimension, the carrier frequency was set on the water resonance with a spectral width of 14 ppm; the indirect (¹⁵N) dimension was centered at 118 ppm with a 30 ppm spectral width. 2K complex data points with 128 complex increments were collected. Each of the-¹⁵N NOE experiments was acquired with 80 scans.²⁰ The ¹H saturation was achieved by the application of 120° ¹H pulses separated by 18 ms, for a 3 s period. In addition, a 4 sec recycle delay was used.

Twelve time points were collected for both T₁ and T₂ experiments, and each set included duplicate measurements to allow estimation of the uncertainty. The delays used were as follows:

T₁ Delays = 2, 2, 64, 128, 256, 384, 512, 768, 768, 1024, 1024, 1536 ms

T₂ Delays = 16, 16, 32, 48, 64, 96, 96, 128, 128, 192, 192, 256 ms

Repeated numbers indicate duplicate measurements. ¹⁵N T₁ and T₂ relaxation times were obtained from running phase-sensitive 2D ¹⁵N-HSQC spectra acquired with Echo/antiecho gradient selection as described above. Typical T₁ experiments were acquired with 16 scans and a repetition delay of 3 sec, whereas T₂ experiments were acquired with 16 scans and a 500 μs delay between 180 pulses in the Carr–Purcell–Meiboom–Gill (CPMG) pulse train and a 3 sec repetition delay.

NMR relaxation data processing

All spectra were processed using the Bruker NMR software. Spectra were zero filled to 4 K by 1 K points in the t₂ and t₁ dimensions respectively and were apodized using a phase-shifted sinebell squared function. Residual water signal was removed using a low frequency filter. Cross-peak intensity was measured as peak height rather than volume, to reduce errors associated with peak over-lap. Fourier transformed data was exported into the Tripos Sybyl 7.0 NMR TRIAD programmed. The peak height intensity was obtained by curve fitting to polynomial function for each peak. This is a preferred method rather than a searching for peak maximal which can introduce a systematic bias.⁶⁵ Weak intensity peaks, at the longer relaxation delay times are particularly prone to systematic bias when a search routine to pick peak maxima is used.⁶⁵

The heteronuclear NOE values were obtained from the ratio of peak intensity for ¹H-saturated and unsaturated spectra. The mean and standard deviation for the heteronuclear-NOE data was calculated from three pairs of experiments. To obtain mono-exponential decay curves for T₁ and T₂ relaxation data peak intensities (by height) were fitted to a two-parameter equation: I(t) = I₀e^−t/T_n, where I = intensity, t = delay in CPMG or longitudinal delay, T_n = T₁ or T₂ value.⁶⁶ This data was compared to relaxation values using a three parameter equation; I(t) = I_∞ + I₀e^−t/Tn, where I_∞ is the offset intensity from zero at infinite time. Comparison of relaxation values obtained using a two or three parameter fits confirmed that possible systematic errors and bias in the data had not been introduced by using a two-parameter fit.⁶⁵ The standard error for the R₁ and R₂ values was determined directly from the curve fit routine.

Spectral density functions and order parameters

Reduced spectral densities map the motion of each NH bond at frequencies ω = 0, ω_N, and 0.87ω_H.^34,35 Reduced spectral density assumes a single spectral density for J(ω_H + ω_N), J(ω_H), and J(ω_H−ω_N) of J(0.87ω_H). The spectral densities at these three frequencies were obtained using an in-house script; DUGGAN, from the relaxation rates R₁ (1/T₁), R₂ (1/T₂) and the hetero-nuclear NOE using four equations described by Farrow et al.^34,67 and shown as Supporting Information. The constant for chemical shift anisotropy was set at −170 ppm and the NH bond length was set to 1.02 Å.³⁹

The model-free formalism^40,41 can be used to describe the motions of a protein in terms of a rotational correlation time (τ_c) and an order parameter (S²). The most widely used model-free formalism assumes isotropic tumbling.^40,41 In this case, τ_c is identical for each amide. The approach selects various descriptors of the motion using statistical criteria as described by Mandel et al.⁴² We have chosen not to use this approach, instead here we have calculated an effective rotational correlation time (τ_c,i) at each residue directly from J(0) and J(ω_N) spectral densities according to Eq. (1).^36,37

1

Furthermore, an approximate order parameter (S²) has been obtained directly from J(0) and J(ω_N) using Eq. (2).^36,37

2

This approximation of S² does not require the statistical selection of various descriptions of the motion^42,68 as the calculation of the order parameter does not incorporate additional fast internal motions on the pico-second time scale (τ_e) or slower segmental motions on the milli- to micro-second time-scale (R_ex). Furthermore, the rotation correlation time is permitted to vary, per-residue, removing the possible effects of an anisotropic rotational diffusion tensor. Although an approximation of the order parameter, the values represent a reasonable description of the amplitude of the nanosecond time scale motions, as the order parameter values (S²) are largely insensitive to the presence or absence of a descriptor for fast pico-second time scale (τ_e) motions. In addition, any residues that contains significant milli-to micro-second time-scale (R_ex) motions can be identified from J(0) spectral densities.

Glossary

Abbreviations

BSE

bovine spongiform encephalopathy

CJD

Creutzfeldt-Jacob disease

GPI

glycosyl-phosphatidylinositol

huPrP

human PrP

mPrP

mouse PrP

PrP

prion protein

PrP^C

cellular isoform of PrP

PrP^Sc

scrapie isoform of PrP

SHaPrP

Syrian hamster PrP