1D ¹H- and 2D ¹H-¹⁵N-TROSY NMR spectra of two different constructs of the ¹⁵N enriched human prion protein, huPrP(23–230) and huPrP(121–230), were recorded at pH 4.8 (acetate buffer) at various pressures. We used solution conditions identical to those used for the structure determination 19. Particularly for huPrP(23–230) slightly acidic pH is necessary because the protein aggregates at neutral pH 35. The shorter construct huPrP(121–230) contains essentially the C-terminal folded part of the protein, the longer construct huPrP(23–230) in addition the unfolded N-terminus of the prion protein. Under theses conditions they showed nearly identical and completely reversible pressure responses in the range from 0.1 to 200 MPa. At 293 K we applied hydrostatic pressures in steps of 0.1, 50, 100, 125, 150, 175 and 200 MPa to both, huPrP(23–230) and huPrP(121–230). Increasing pressure results in chemical shift changes of all resonances. In addition, in the TROSY-HSQC spectra the increased pressure leads to a decrease of the signal volumes of a number of resonances. Some of the signals disappear completely from the spectra while others show no significant volume change as described below (see section "pressure response of cross peak volumes").

In general, the pressure response of proteins is anisotropic and the chemical shift changes induced by the local and global conformational changes are a non-linear function of the pressure. Traditionally, the pressure dependence is described by the zero-order coefficient δ₀(p₀, T₀), the first order coefficient B₁ and the second order coefficient B₂ of a Taylor expansion around the pressure p₀ and the temperature T₀ (see eq. 1). When the pressure effects are assumed to be composed of two contributions, an unspecific effect as observed in random-coil peptides and a specific, structure dependent effect, then the influence of these effects on the Taylor coefficients can be assumed as additive as long as the two events are not or only weekly coupled (see Methods). In this case, the corrected, conformation dependent pressure (Taylor) coefficients B₁* and B₂* can be obtained by subtracting the corresponding pressure coefficients from random-coil peptides. Random-coil pressure coefficients are available for the ¹H-shifts from a study of the model peptides Gly-Gly-X-Ala peptides 36 but not for the ¹⁵N-shifts.

The first order conformation dependent pressure |B₁*(H)| for the ¹H^N shifts and of the uncorrected pressure coefficient |B₁(N)| of the ¹⁵N^H chemical shift as a function of the primary and the secondary structure are displayed in Fig. 1 for the full length protein huPrP(23–230). Analogously, the second order coefficients |B₂*(H)| and |B₂(N)| are plotted as a function of the sequence position in Fig. 2. The mean first and second order pressure coefficients |B₁*(H)|, |B₁(N)|, |B₂*(H)|, and |B₂(N)| are 0.38 ppm/GPa, 3.35 ppm/GPa, 1.23 ppm/GPa², and 5.70 ppm/GPa².

A similar shift pattern is obtained for the truncated prion protein (data not shown) in the regions which are common to both proteins. However, the mean first and second order pressure coefficients |B₁*(H)|, |B₁(N)|, |B₂*(H)|, and |B₂(N)| are with 0.57 ppm/GPa, 3.35 ppm/GPa, 1.69 ppm/GPa², and 7.29 ppm/GPa² significantly larger than in the full length protein. This is different when corresponding regions are compared since the pressure effects are much smaller in unfolded regions of the protein which contribute much more to the mean values in the full length protein. For the folded part of the full length protein one obtains mean values for <|B₁*(H)|>, <|B₁(N)|>, <|B₂*(H)|>, and <|B₂(N)|> 0.57 ppm/GPa, 3.33 ppm/GPa, 1.60 ppm/GPa², and 6.62 ppm/GPa², respectively.

If one takes the region common in the two proteins then the averages of the pressure coefficients are almost identical, that is the N-terminus does not significantly influence the average pressure response of the folded part. The correlation coefficient for the first order proton and nitrogen coefficients B₁*(H) and B₁(N) of the two proteins is 0.71 and 0.73, respectively, indicating also a quantitative agreement of the parameters in the two proteins. For the correlation coefficients of, B₂*(H), and B₂(N) one obtains with 0.34 and 0.63 smaller correlations in the two proteins. This is mainly due to the relatively large errors involved in the calculation of the Taylor pressure coefficients. In contrast, the direct correlation of the experimental chemical shift changes Δδ(p) (see Methods) is much higher. The mean value of these correlation coefficients for each residue of huPrP(23–230) to its corresponding residue in huPrP(121–230) is 0.86 for the amide protons and 0.93 for the amide nitrogen atoms.

Although the general pressure response of the folded core is very similar in the full length and the truncated huPrP there are some residues where clearly different responses can be observed (Fig. 3). The chemical shift differences already existing for Glu168, His187, Thr192, Glu207, Glu211 and Tyr226 at ambient pressure become larger with increasing pressure.

In the structured core (amino acid 121–230) of full length and truncated huPrP most residues shift with increasing pressure (Figs. 1 and 2). For identifying residues that behave differently to the majority of the peaks one can select those which display first or second order pressure coefficients which are larger or smaller than the average by a standard deviation (SD). Since large chemical shift changes are strongly correlated with large conformational changes, residues with pressure coefficients larger than average plus one SD are most probably involved in strong conformational changes. Accordingly, residues with pressure coefficients smaller than average minus one SD represent regions of small inferred conformational changes. The residues M129, F141, Q186, H187, T188, V189, K194, E196, E207, V210, E211, E219 and R220 display first order pressure coefficients |B₁*(H)| and G131, I139, H140, C179, T183, T188, T191, K194, G195, N197, F198, T199, D202 and M205 |B₁(N)| which are significantly larger than the average in this region (by one standard deviation) for huPrP(121–230). The largest pressure induced chemical shift changes of amide proton resonances are observed in the loop between strand β1 (Y128 to G131) and helix α1 (D144 to M154), in the loop connecting helix α2 (N173 to K194) and α3 (E200 to R228), as well as in regions of helix α2 and α3. Generally, the differences of the pressure dependence of the chemical shifts along the protein sequence are much less pronounced for the amide nitrogen atoms than for the amide protons. The pressure-sensitive regions for both, ¹H^N and ¹⁵N^H, overlap quite well but not completely. Almost the same residues are also identified by the second order coefficients. The same is true for the residues with a very small pressure response, namely 129, 130, 148, 156, 174, 178, 182, 186, 187, 190, 192, 212, 219 and 221 (|B₁(N)| smaller than the mean value minus one sigma) and 121, 122, 123, 126, 127, 128, 130, 133, 135, 138, 140, 142, 160, 164, 172, 177, 180, 184, 193, 198, 213, 222, 225, 226, 227, 228 (|B₁*(H)| smaller than the mean value minus 0.75 sigma) in case of huPrP(121–230). As has been stated above and has been quantified by the correlation analysis the pressure response of the amide nitrogen atoms of residues 121–230 of huPrP(23–230) is very similar to that of huPrP(121–230). In conclusion, huPrP(121–230) is a good model for the study of pressure effects on the structured core of the whole protein.

The N-terminal part of the protein (amino acids 23 to 120) is assumed to be unstructured and thus should show only small pressure effects (Figs. 1 and 2). This is in general true, the mean first and second order pressure coefficients |B₁*(H)|, |B₁(N)|, |B₂*(H)|, and |B₂(N)| are with 0.18 ppm/GPa, 3.37 ppm/GPa, 0.68 ppm/GPa², and 4.20 ppm/GPa² significantly smaller than in the folded part of the protein. However, a number of residues show pressure coefficients that are above the average. Surprisingly, this is also true for the amide proton shifts, where the pressure coefficients of random coil peptides 36 were subtracted. Notably, K27, S43, G46, G55, G56, K101, S103, K104 and M109 show linear pressure coefficients that deviate significantly from the expectation. For the quadratic coefficients this applies to G40, S43, G55, G56, N108 and M109. If the N-terminus would be completely unstructured, only pressure coefficients in the magnitude of the random coil peptide values would be expected.

Some of the residues show strong non-linear pressure dependences of the chemical shifts together with an asymptotic behavior within the pressure range studied (Fig. 4) which would be typical for a two-state transition in fast exchange on the NMR-time scale. In this model, at ambient pressure two local conformational states N₁ and N₂ coexist that are sensed by the corresponding residues. At high pressure the local conformational equilibrium of the protein is completely shifted from the ground state to the low-lying excited state N₂. The molar free energy (ΔG₀) for this local transition can be derived from the chemical shift change by equation (2) (see Methods). As already discussed above for removing unspecific pressure effects from specific conformation dependent effects, prior to the calculation of the Gibbs free energies ΔG₀ the pressure-induced chemical shift change as determined for the random coil standard peptides 36 was subtracted from the measured chemical shift values. Since that was only possible for the proton shifts only these were evaluated quantitatively. However, qualitatively the pressure dependence of the nitrogen chemical shifts followed the same rules.

For the residues of the structured core (huPrP(121–230)) one obtains 3.0 ± 1.4 kJ/mol. The ΔG₀ value for the pressure-induced transition of the N-terminal residues of huPrP(23–230) is 3.0 ± 1.1 kJ/mol (Table 1). At ambient pressure about 71% of the protein are in the highest populated native "ground state" conformation, whereas 29% are in the second "excited" native conformation which is favored by applying pressure. From the fast exchange condition, a lower limit for the exchange correlation rate 1/τ_e of 22 s^-1 can be estimated for the N₁-N₂ transition.

At 200 MPa the amide proton resonances of residues Y128, G131, M134, R136, I139, F141, G142, S143, D144, Y150, R156, Q160, V161, Y163, N174, D178, I182, T199, E200, D202, V210, C214, I215, Q217 and E221 are not observable in case of huPrP(121–230). Especially, the cross peak of Gly131 disappears already at 125 MPa, while residues Ile139, Phe141, Gln160, Val161, Tyr163 and Asp178 become undetectable at 150 MPa. These residues mainly cluster to the loop between the strand β1 and helix α1, near helix α3 and close to the β-sheet. In case of huPrP(23–230) qualitatively the same process can be observed for the core region, however, since in the longer construct the resolution is reduced and becomes worse with increasing pressure only the disappearance of residues Gly131, Ile139, Phe141, Arg156 and Asp178 can be reliably confirmed. The N-terminal unstructured region displays a less pronounced chemical shift dependence and no significant volume change upon pressure (Fig. 5a). By releasing the pressure we observe the original spectra at ambient pressure again, thus the pressure-induced changes are completely reversible.

There are two possible explanations for this loss in intensity, increased loss of intensity during the INEPT-periods by increased average T₂-relaxation and/or a two site exchange which is slow on the NMR-time scale and where the second signal would be too weak to be observable 37. The increased transverse relaxation rate could be due to an exchange broadening or an increased effective molecular size by protein aggregation. The last mechanism could also lead to a disappearance of the second signal in the slow exchange regime. Under fast exchange conditions the increased transverse relaxation time can also be estimated from the broadening of the signals by pressure. The observed increase in line width by a factor 1.2 to 2 is too small to account for the observed loss in signal intensities if fast exchange conditions would prevail. Using the approximations derived by Maurer et al. 37 a maximum intensity decrease by 20% could follow from that additional, pressure dependent line broadening. For most of the residues under consideration the intensity loss during the polarization transfer process is considerably smaller, thus pointing towards a slow exchange process with an intermediate conformation. A line broadening of the same magnitude as observed for the amide groups is also observed in 1D-spectra for resonances of non-exchangeable protons. As an example the high-field shifted methyl resonances of Leu125, Leu130, Ile139 and Ile182 are shown in Fig. 6. As seen in the TROSY spectra, in the 1D spectra a certain number of peaks shows no or only weak line broadening, too. The line broadening is not caused by a change of the viscosity of water with pressure, since at 293 K the water viscosity shows only weak pressure dependence up to pressures above 200 MPa. Actually the viscosity even decreases till 80 MPa to about 98% and displays a moderate increase with increasing pressure above that. At about 140 MPa it reaches its original value again 38. Thus, the origin of the line broadening might be due to conformational heterogeneity and/or exchange.

Assuming two site model under slow exchange conditions the pressure induced changes of volumes and thus of the corresponding populations can be used to calculate a molar free energy (ΔG₀) for every single amino acid residue in the polypeptide chain. The equilibrium constant K between the native state and the pressure populated intermediate can then be derived from the ratio of the peak volumes (see Methods). Fig. 5c shows the molar free energies at ambient pressure obtained from a fit of the data as a function of the primary structure of huPrP(121–230). The secondary structure elements are indicated at the bottom. Due to signal overlap it was not possible to determine exact values for the cross peak intensities of every residue which leads to gaps in the residue specific determination of the pressure stability (marked with "O"). Residues which show no significant change of the signal intensity are marked with "#". The average for the molar free energies ΔG₀ and their standard deviations are indicated in Fig. 5c by a solid and two broken lines. The mean of ΔG₀ is 14.2 kJ/mol with a standard deviation of 4.5 kJ/mol. The distribution of free energies is depicted in Fig. 7a showing that the obtained energy distribution is clearly divided into two groups. Fig. 7b shows peak volumes and the ΔG fit of typical residues of both groups. In Fig. 8 the two groups are mapped on the structure of huPrP(121–230). Group one encompasses 33 residues and has a ΔG₀ of 10.8 ± 1.9 kJ/mol, the mean ΔG₀of the 25 residues of group two is 18.6 ± 2.9 kJ/mol.

The pressure response of huPrP consists of three different effects: 1 Small linear chemical shift changes corresponding to small first-order pressure coefficients, 2 Non-linear chemical shift changes which show for some residues a typical asymptotic behavior as it would be expected for a fast exchange between slightly differing conformations, and 3 a loss of signal volumes which occurs for the majority of the amino acid residues. A thorough analysis shows the appearance of at least two different conformational transitions between the ground state N₁, the low-lying excited state N₂ and the intermediate states 252839. At ambient pressure the chemical shift data allow us to postulate the states N₁ and N₂, and the analysis of the signal volume loss induced by pressure indicates additional intermediate states I₁ and I₂. Strong pressure responses are often associated with the existence of packing defects, or cavities in the direct neighborhood. The cavities calculated from the PDB deposited structure (code 1QM2) are also shown in Fig. 8. The found cavity at this position is surrounded by acidic side chains thus leading to a hydrophilic pore. Close to helices α2 and α3 similar packing defects are also seen.

The pressure dependent energy landscape of the human prion protein is shown in Fig. 9 schematically. At ambient pressure two native states N₁ and N₂ as well as the energetic unfavorable excited intermediate states I₁ and I₂ are visible. With increasing pressure the energy profile changes. At the onset of pressure N₂ becomes the energetically lowest and therefore highest populated conformation. At very high pressures the intermediate conformations I₁ and I₂ are strongly favored.

Application of equations (2) to (5) allows the calculation of individual ΔG₀ and ΔV values for all atoms in the molecule. When more than one parameter can be used as in our case the chemical shift change and the peak volume change more than one ΔG₀ and ΔV value can be calculated for the structural transition observed by a single amide group in the protein. The question arises what the physical meaning of these values may be. In principle, ΔG₀ and ΔV are global quantities that describe the difference of the Gibbs free energies and partial volumes of a state X_i to an arbitrary state X₁ of the system. Since the number of possible states of the system (that is the number of conformational states of the protein and the solvent molecules) is infinite one has to restrict to a smaller number of global states for obtaining meaningful results. In general one can separate the states of the bulk water from the rest of the system. Since pressure influences strongly the water shell close to the protein the effects on the shell and the protein conformation itself are difficult to separate.

Our analysis of N amide cross peak volumes and 2N pressure induced ¹H and ¹⁵N chemical shift changes gives us information about 6N different conformational states when the individual parameters follow a two-site exchange model. In case of multiple site exchange even more (sub)states would be expected. The primary evaluation of our shift and volume changes is done in this way. The same has been done by Kuwata et al. for the hamster prion protein but only volume changes were taken into account 34.

However, the structural substates in a protein-water system are usually coupled; the obtained atom specific ΔG₀ and ΔV values are in general not to be interpreted locally since large contributions are due to structural differences of other parts of the molecule compared to the ground state (or the set of closely spaced low-energy states). As a further complication the calculation of the differences as done here must not necessarily refer to the same ground state.

After correction for unspecific "random coil" pressure induced shifts, we can group the ΔG₀ and ΔV values in four main classes. This means in a minimal description of the system the substates X_i are reduced to the combination of two "native" states N₁ and N₂ and two "intermediate" states I₁ and I₂. These states are primarily defined in the data evaluation as localized structural states and no information is available about the structural states of not directly considered other parts of the protein. Especially, in the data evaluation of the signal volume changes defining the intermediate states a localized native structural state has to be assumed. Since for some of the residues undergoing volume changes also shift changes corresponding to the N₁ to N₂ transition are observed, it seems plausible to assume that this transition is independent of the intermediate states and occurs in the whole protein. The calculated ΔG₀ values then would refer to a state N where the other parts of the molecules are in the structural state N₁ or N₂. The simplest picture would assume an ordered reaction scheme of conformational states S.

N₁ ⇆ N₂ ⇆ I₁ ⇆ I₂

The quality of the data does not allow distinguishing between this model and more complicated possible models. The corresponding energy surface has been depicted schematically in Fig. 9. The abscissa here is a generalized arbitrary coordinate which represents the structural states of the proteins only schematically.

Since the data can be explained sufficiently well with four states only, we restrict in the following discussion on the existence of these four states. However, we cannot exclude the existence of additional structural states with indistinguishable free energies. The pressure dependences of the chemical shifts indicate the existence of a second native conformation of huPrP, which is favored under high pressure and is in fast conformational exchange with the "ground state". At ambient pressure 71% of the protein is in the main native state N₁ and 29% are in the "excited" native conformation N₂. The low energy difference (ΔG₀ = 3.0 kJ/mol) and the relatively small chemical shift differences between N₁ and N₂ implies, that only slight structural changes occur in the N₁ ↔ N₂ transition. The state I₁ and I₂ are derived from the analysis of the volume changes of those resonances that are too broad to be observable in one of the states I. At ambient pressures only less than 1% exists in I₁ or I₂. At 200 MPa most of the protein is in state I₁ or I₂.

The N-terminal region of huPrP(23–230) is known to show no well-defined three-dimensional structure at the used buffer conditions. An analysis of the chemical shifts corrected for neighbor effects 44 indicates no well-defined regions where secondary structure is induced by pressure. Nevertheless, the complete N-terminus has a tendency towards positive chemical shift values.

After subtracting the random-coil pressure coefficients within the limits of error vanishing first and second order pressure coefficients are expected for those parts of the structure that are in a random-coil conformation. Unfortunately, random-coil values do only exist for protons 36 but not for ¹⁵N. Therefore, only the proton shifts can be interpreted in detail. In the presumably unstructured part (amino acids 23 to 120) the expectation values of B₁* and B₂* are with 0.11 ppm/GPa and 0.68 ppm/GPa² close to zero (see Results). Most interestingly, the range of amino acids 90 to 170 where two-dimensional electron crystallography and molecular dynamics simulations 4546 predict β-sheet formation after polymerization displays a high variety of pressure sensitivity. The strands β1 and β2 seem to be very insensitive to the application of high pressure while the loop and the helix α1 in between these two strands exhibit some of the most pronounced changes in structure we observed by applying high pressure.

As indicated by the pressure induced chemical shift changes there are two native conformations N₁ and N₂ in fast conformational exchange, which coexist even at ambient pressure. The relatively small difference of the molar free enthalpies of both states as well as the rather small chemical shift changes indicate, that the N₁ ↔ N₂ transition is only associated with a small structural rearrangement. Especially the very small chemical shift differences show that N₁ and N₂ adopt very similar structures. Unfortunately, a complete structural characterization of N₂ cannot be performed by the available data.

The pressure dependent reduction of cross peak volumes in the TROSY-HSQC-spectra was associated with the N ↔ I transition. They can be due to increased T₂-losses during the INEPT polarization transfer periods and/or a slow exchange process where the resonances of state I are broadened beyond detection 37. In principle mainly two mechanisms can be responsible for an increased T₂, an increase of the rotational correlation time by protein-protein interaction and exchange broadening. In one-dimensional proton NMR spectra we observe a nearly general line broadening of approximately a factor two which could indicate a formation of dimers at higher pressure. In the TROSY-HSQC spectra also an additional line broadening with pressure is observed which varies locally but is clearly smaller than the factor of two. Additional line broadenings in the TROSY-spectra are mainly associated with chemical exchange processes. The observed increase of the transverse relaxation rates cannot explain the observed reduction of cross peak volumes as the only factor (see Results). Therefore, as additional factor a reduction of the peak volume by a slow exchange process is required as it has been done also for the hamster prion protein by Kuwata et al. 34. In the absence of exchange broadening the cross peak volume change with pressure can be described by equations (4) and (5). In the presence of a moderate additional exchange these equations still provide a good approximation but the presence of exchange broadening would lead to too large estimates of the apparent ΔG₀ values.

Within the limits of error the cross peak volume changes can be associated with two folding intermediates. In the folded core of the prion protein the region around positions Ile139, His140, and Phe141 shows rather low ΔG₀ values that may indicate a larger exchange contribution. This may be interpreted as a premature local melting which correlates with a packing defect in this region (Fig. 8). A non-perfect stacking of side-chain and main-chain atoms leads to a less restricted peptide main chain in this area allowing a high degree of conformational freedom and thus conformational variability/heterogeneity. As pointed out by Pratt and co-workers 47 and has been observed experimentally by high-pressure NMR spectroscopy on RalGDS by us 25, pressure-induced unfolding is caused by the penetration of water into the hydrophobic core due to the dissociation of electrostatic bonds and solvation of hydrophobic residues. The sidechain resonance of the Ile139 methyl proton exhibits the most pronounced up field shift observed in the human prion protein indicating a strong interaction with the hydrophobic core of this protein.

Single low ΔG₀ values are also found in the helical region encompassing α2 and α3 which is in agreement with the finding that helices are more prone to pressure than extended structures like β-strands. In a similar study Kuwata et al. 34 investigated in a high-pressure NMR study the Syrian hamster ShaPrP(90–231). They found the most pronounced effects in the helical region encompassing the helices α2 and α3 implying a complete unfolding of these two helices in a slow exchange process. In addition, our data show the occurrence of a pressure-stabilized "excited" state N₂. The existence of that state was not taken into account in their data analysis. Subsequently we find a slow exchange process to two intermediate conformations, which are characterized by a melting of the region around Ile139-Phe141 and a partial melting of the helical region encompassing α2 and α3. Our data suggest a dual mechanism for the structural changes under pressure at least in the human prion protein. Especially the observation in intermediate I₂ that the most stable region of the human prion protein is localized in the region of the two helices where the disulfide bridge is located corresponds very well with the hydrogen protection measurements by other groups 4849.

Vanik et al. 16 adopted the Y145 Stop variant for in vitro specific seeding of amyloid fibrils. In particular, they found that mouse or hamster PrP-specific amino acid substitutions at position 138 and 139 are sufficient to change essential amyloidogenic properties of huPrP(23–144), with the mutant proteins adopting seeding specificities of PrP corresponding to different species. Vice versa, M138I/M139I ShaPrP(23–144) adopted the properties of human PrP(23–144) 16. Our results show that the importance of individual amino acids in PrP as determinants of the species barriers are due to their local conformational variability which can be highlighted by applying high pressure and monitored on an atomic scale through NMR spectroscopy. Especially, local clusters of very low stability found in our experiments are prone to change their conformation upon interaction with the infectious conformer PrP^Sc. Indeed, permissible first entry-points of the scrapie isomer need a certain degree of local conformational freedom to initiate the refolding of the cellular isomer to an infectious one. Thus, the species and transmission barrier is conserved through the differences in the conformational freedom of local areas in the protein structure which can vary substantially with differing solution conditions necessary for the proteolytic degradation of proteins in the cell.

The N-terminal part of huPrP(23–230) exhibits regions with chemical shift responses that are not typical for random-coil peptides. Their first and/or second order pressure coefficients deviate significantly from the average (by more than a standard deviation). This is best visible for the proton shift coefficients since they could be corrected by subtracting the published random-coil values 36. One can define four regions where at least one of the coefficients deviates significantly from the average, namely region A encompassing K27, region B encompassing G40, S43 and G46, region C encompassing G55 and G56 and region D with K101, S103, K104, N108 and M109. The data indicate that these regions have a non-random population of structural states that coexist in solution and show a characteristic change of populations with pressure. Region C is located within the three octapeptide repeats which are known to adopt a defined structure upon binding of copper 35. Unfortunately, the amino acids of the three octapeptide repeats could not be resolved individually in the spectra. They seem to have a similar pressure response, too. Region D is in the region Kuwata et al. previously discussed the existence of transient helices 34. This is in good agreement with our observations of pressure responses typical for structured proteins. Interestingly, there is an overlap between region B and C of the pressure-sensible residues and a possible pH-dependent aggregation site (residues 45–66) as described by Ralph Zahn 35. The observed pressure-sensitivity emphasizes the importance of this region.

The probability that some residues of the unfolded part of PrP are in the time average in contact with the surface of the folded part is quite high and was discussed by Zahn et al. 19. If this would be true it could also influence the pressure response of the folded core in full length PrP. Indeed, such a differential response is observed clearly for Glu168, His187, Thr192, Glu207, Glu211 and Tyr226. They show also different chemical shifts in truncated and full length PrP. From C^α chemical shift differences in the full length and the truncated PrP Zahn et al. 19 identified as possible interaction regions His187 to Thr193 and Glu219 to Tyr226. The differences in the pressure response observed may indicate a transient interaction between the region D that is closest in sequence to the core and the surface around amino acids mentioned above.

Recombinant human prion protein huPrP(23–230) (residues 23 to 230) and huPrP(121–230) (residues 121 to 230) was prepared as described previously 1952. Note that due to cloning artifacts the used recombinant proteins have two additional residues at the N-terminus (G21 and S22 in case of huPrP(23–230), G119 and S120 in case of huPrP(121–230). For the NMR experiments a 1.1 mM solution of ¹⁵N-enriched huPrP(23–230) and a 1.2 mM solution of ¹⁵N-enriched huPrP(121–230) each in 10 mM sodium acetate buffer pH 4.8 was used. The measurements were performed in ¹H₂O with 8% ²H₂O added. The samples contained 0.1 mM of 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) as internal reference.

All NMR experiments were carried out on a Bruker DRX 600 spectrometer, operating at 600 MHz proton resonance frequency. To apply high pressure we used an on-line variable pressure cell system 5354 with a sapphire capillary of 1.72 mm inner and 3.14 mm outer diameter 5556. Spectra were recorded at hydrostatic pressures of 0.1, 50, 100, 125, 150, 175 and 200 MPa. At a temperature of 293 K and different pressures one-dimensional ¹H spectra and two-dimensional ¹H-¹⁵N-TROSY spectra were acquired. The TROSY spectra 57 had a digital resolution of 2048 * 512 points and a frequency width of 7184 * 1825 Hz. All spectra were recorded and processed using the XWINNMR package from Bruker. AUREMOL 5859 was used for peak picking, signal volume integration and analysis of the spectra.

The spectra were assigned on basis of the assignment at ambient pressure 19. The chemical shift values of the amide protons and nitrogen atoms in the TROSY spectra under the influence of pressure were fitted to equation (1), where δ₀ is the chemical shift at ambient pressure p₀. The first order pressure coefficient B₁ and the second order pressure coefficient B₂ of each amide proton were corrected by subtracting the pressure coefficients of the standard peptide (Gly-Gly-X-Ala) of the corresponding amino acid as published by Arnold et al. 36. For glutamate the coefficients of the C-terminal methylated peptide (Gly-Gly-Glu-Ala-methyl) were used (B₁ 0.53 ppm/GPa, B₂ -0.98 ppm/GPa²) 56. The corrected coefficients are referred as B₁* and B₂*.

δ(p,T₀) = δ₀(p₀,T₀) + B₁(p₀,T₀)(p - p₀) + B₂(p₀,T₀)(p - p₀)²     (1)

If there is an equilibrium between two conformations of the protein that is in fast exchange on the time scale of the NMR experiment, the observed chemical shifts are the weighted average of the chemical shifts δ₁ and δ₂ of the two states as described by equation (2), with an equilibrium constant for the conformational exchange K=N2N1 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGlbWscqGH9aqpdaWcaaqaaiabd6eaonaaBaaaleaacqaIYaGmaeqaaaGcbaGaemOta40aaSbaaSqaaiabigdaXaqabaaaaaaa@336F@.

δ = 1 1 + K δ 1 K 1 + K δ 2 = δ 1 + δ 2 exp ⁡ ( − Δ G R T ) 1 + exp ⁡ ( − Δ G R T ) = δ 1 + ( δ 2 − δ 1 ) 1 1 + exp ⁡ ( Δ G R T )       ( 2 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaaiiGacqWF0oazcqGH9aqpdaWcaaqaaiabigdaXaqaaiabigdaXiabgUcaRiabdUealbaacqWF0oazdaWgaaWcbaGaeGymaedabeaakmaalaaabaGaem4saSeabaGaeGymaeJaey4kaSIaem4saSeaaiab=r7aKnaaBaaaleaacqaIYaGmaeqaaOGaeyypa0ZaaSaaaeaacqWF0oazdaWgaaWcbaGaeGymaedabeaakiabgUcaRiab=r7aKnaaBaaaleaacqaIYaGmaeqaaOGagiyzauMaeiiEaGNaeiiCaaNaeiikaGIaeyOeI0YaaSaaaeaacqGHuoarcqWGhbWraeaacqWGsbGucqWGubavaaGaeiykaKcabaGaeGymaeJaey4kaSIagiyzauMaeiiEaGNaeiiCaaNaeiikaGIaeyOeI0YaaSaaaeaacqGHuoarcqWGhbWraeaacqWGsbGucqWGubavaaGaeiykaKcaaiabg2da9iab=r7aKnaaBaaaleaacqaIXaqmaeqaaOGaey4kaSYaaeWaceaacqWF0oazdaWgaaWcbaGaeGOmaidabeaakiabgkHiTiab=r7aKnaaBaaaleaacqaIXaqmaeqaaaGccaGLOaGaayzkaaWaaSaaaeaacqaIXaqmaeaacqaIXaqmcqGHRaWkcyGGLbqzcqGG4baEcqGGWbaCcqGGOaakdaWcaaqaaiabgs5aejabdEeahbqaaiabdkfasjabdsfaubaacqGGPaqkaaGaaCzcaiaaxMaadaqadiqaaiabikdaYaGaayjkaiaawMcaaaaa@7BEC@

The free energy at ambient pressure ΔG₀ can be determined by fitting the data to equation (3), which represents the first terms of a Taylor expansion of ΔG as function of pressure p at constant temperature T. Here, ΔV is the partial molar volume and Δβ the compressibility. For our calculations of the ΔG₀ values we neglected the quadratic term in equation (3).

Δ G = Δ G 0 + Δ V ( p − p 0 ) + 1 2 Δ β ( p − p 0 ) 2       ( 3 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqGHuoarcqWGhbWrcqGH9aqpcqGHuoarcqWGhbWrdaWgaaWcbaGaeGimaadabeaakiabgUcaRiabgs5aejabdAfawnaabmGabaGaemiCaaNaeyOeI0IaemiCaa3aaSbaaSqaaiabicdaWaqabaaakiaawIcacaGLPaaacqGHRaWkdaWcaaqaaiabigdaXaqaaiabikdaYaaacqGHuoariiGacqWFYoGydaqadiqaaiabdchaWjabgkHiTiabdchaWnaaBaaaleaacqaIWaamaeqaaaGccaGLOaGaayzkaaWaaWbaaSqabeaacqaIYaGmaaGccaWLjaGaaCzcamaabmGabaGaeG4mamdacaGLOaGaayzkaaaaaa@4EFB@

Is there a conformational exchange which is slow relative to the time scale of the NMR experiment, the signal volume of one state is proportional to its population. Assuming a two-state transition for the change of the conformational equilibrium with increased pressure, the equilibrium constant K can be calculated from the relative cross-peak intensities of the TROSY spectra by equation (4) 34. I(p) is the relative cross-peak intensity at pressure p and I₀ is the one, when 100% of the protein are in the native state. With this one obtains the free energy ΔG associated with the pressure-induced conformational change by equation (5). The free energy at ambient pressure ΔG₀ is defined by equation (3) and can be obtained from a three parameter fit with I₀, ΔV₀ and ΔG₀ neglecting the second order term in p.

K = 1 − I ( p ) I 0 I ( p ) I 0       ( 4 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGlbWscqGH9aqpdaWcaaqaaiabigdaXiabgkHiTmaalaaabaGaemysaK0aaeWaceaacqWGWbaCaiaawIcacaGLPaaaaeaacqWGjbqsdaWgaaWcbaGaeGimaadabeaaaaaakeaadaWcaaqaaiabdMeajnaabmGabaGaemiCaahacaGLOaGaayzkaaaabaGaemysaK0aaSbaaSqaaiabicdaWaqabaaaaaaakiaaxMaacaWLjaWaaeWaceaacqaI0aanaiaawIcacaGLPaaaaaa@413F@

I ( p ) = I 0 1 + exp ⁡ ( − Δ G R T )       ( 5 ) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGjbqsdaqadiqaaiabdchaWbGaayjkaiaawMcaaiabg2da9maalaaabaGaemysaK0aaSbaaSqaaiabicdaWaqabaaakeaacqaIXaqmcqGHRaWkcyGGLbqzcqGG4baEcqGGWbaCcqGGOaakcqGHsisldaWcaaqaaiabgs5aejabdEeahbqaaiabdkfasjabdsfaubaacqGGPaqkaaGaaCzcaiaaxMaadaqadiqaaiabiwda1aGaayjkaiaawMcaaaaa@4568@

When the chemical shift changes in two samples A and B (e. g. the same spin in two constructs of a protein or the hydrogen and nitrogen atoms of an amide group) are governed by the same physical process that can be approximated by a two-site exchange then the chemical shift changes can be described by the same free energy ΔG. In this case the pressure dependent chemical shift changes Δδ^Aand Δδ^B

Δδ^A = δ^A(p) - δ^A(p₀) = (δ₂^A - δ₁^A)f(p)     (6a)

Δδ^B = δ^B(p) - δ^B(p₀) = (δ₂^B - δ₁^B)f(p)     (6b)

follow the linear relation