Influence of pH on NMR structure and stability of the human prion protein globular domain.

The NMR structure of the globular domain of the human prion protein (hPrP) with residues 121-230 at pH 7.0 shows the same global fold as the previously published structure determined at pH 4.5. It contains three alpha-helices, comprising residues 144-156, 174-194, and 200-228, and a short anti-parallel beta-sheet, comprising residues 128-131 and 161-164. There are slight, strictly localized, conformational changes at neutral pH when compared with acidic solution conditions: helix alpha1 is elongated at the C-terminal end with residues 153-156 forming a 310-helix, and the population of helical structure in the C-terminal two turns of helix alpha 2 is increased. The protonation of His155 and His187 presumably contributes to these structural changes. Thermal unfolding monitored by far UV CD indicates that hPrP-(121-230) is significantly more stable at neutral pH. Measurements of amide proton protection factors map local differences in protein stability within residues 154-157 at the C-terminal end of helix alpha 1 and residues 161-164 of beta-strand 2. These two segments appear to form a separate domain that at acidic pH has a larger tendency to unfold than the overall protein structure. This domain could provide a "starting point" for pH-induced unfolding and thus may be implicated in endosomic PrPC to PrPSc conformational transition resulting in transmissible spongiform encephalopathies.

The prion protein (PrP), 1 a predominantly synaptic protein present in all higher organisms (1)(2)(3)(4), constitutes a major component of the infectious agent (prion) that causes transmissible spongiform encephalopathies (5,6). The normal cellular isoform of the protein, PrP C , is soluble and protease-sensitive, whereas the disease-associated ␤-sheet-rich form, PrP Sc , is insoluble, partially resistant to protease digestion (7,8), and thought to propagate by converting PrP C molecules into an alternative conformation (9 -11). Recently, it has been shown that the accumulation of even small quantities of misfolded PrP in the cytosol is strongly neurotoxic in cultured cells and transgenic mice (12,13). However, the subcellular localization of the conformational transition of PrP C into PrP Sc is controversial (14). There are indications that it takes place either at the cell surface, where the average interstitial milieu of the brain (15,16) has a pH of 7.3, or after internalization of PrP Sc into endosomes (17)(18)(19), where pH values range between 4.7 and 5.8 (20).
The in vitro conversion of human brain PrP C to a PrP Sc -like form is enhanced at acidic pH (21). Biophysical studies have shown that the free energy of unfolding of hPrP-(90 -231) is lower at acid pH than at neutral pH (22) and that in acidic guanidinium chloride hPrP-(90 -231) forms a folding intermediate that contains a large amount of ␤-sheet secondary structure. A ␤-sheet-rich folding intermediate has also been observed for mouse PrP-(121-231) at low pH in urea but is not seen at neutral pH (23). NMR structures are available for several recombinantly expressed mammalian prion proteins (24 -27) but only in acidic solution conditions between pH 4.5 and 5.5. A crystal structure of human PrP-(90 -231) has recently been determined from crystals grown in pH 8 solution, where two globular domains are linked through interchain disulfide bonds (28).
In an attempt to investigate the possible effects of pH on the structure of PrP C , we have studied the recombinant human prion protein globular domain of residues 121-230 in pH 7.0 solution. We describe a high quality NMR structure of monomeric hPrP-(121-230), amide hydrogen exchange experiments monitored by NMR, and thermal unfolding experiments monitored by CD. These results are compared with the previously published structural and thermodynamic data of hPrP-(121-230) obtained at pH 4.5 (26) and with the crystal structure of hPrP-(90 -231) determined at pH 8 (28).

EXPERIMENTAL PROCEDURES
Sample Preparation-15 N-and 13 C, 15 N-labeled recombinant hPrP-(121-230) was expressed and purified as described previously (29). NMR samples were 0.5-1.2 mM in protein concentration in buffer solution containing 10 mM sodium phosphate at pH 7.0 and 0.05% sodium azide. Dynamic light scattering and size exclusion chromatography measurements show that under these conditions samples are homogenous and monomeric. Samples were prepared either in 95% H 2 O, 5% D 2 O or in 99.9% D 2 O.
Circular Dichroism and Thermal Denaturation Experiments-Circular dichroism spectra were recorded with a Jasco J720 spectropolarimeter interfaced with a Peltier-type temperature control unit with 1-mm path length cuvette. Thermal denaturation experiments were performed by monitoring the circular dichroism at 222 nm while changing the temperature from 10 to 90°C or vice versa with a constant temperature gradient of 50°C/h. Reference spectra were collected at 10°C before and after each experiment. Denaturation curves were analyzed assuming a two-state unfolding model (30), where the temperature dependence of the Gibbs free energy change of denaturation, ⌬G U Ϫ N , is given by Equation 1 (31), where U is unfolded protein, N is native protein, ⌬H m is the change in enthalpy at the midpoint of the denaturation temperature, T m ; ⌬C p is the difference in heat capacity between denatured and native protein (⌬C p of hPrP-(121-230) was estimated by assuming 12 cal mol Ϫ1 deg Ϫ1 / amino acid (32); T is the absolute temperature in kelvin; R is the gas constant.
NMR Measurements and Structure Determination-NMR spectra were recorded on Bruker DRX600 and DRX750 spectrometers. The programs Prosa (33) and Xeasy (34) were used for data processing and spectral analysis, respectively. Sequence-specific resonance assignments were derived by adapting the assignments at pH 4.5 (26) to pH 7.0 and confirmed by standard triple-resonance NMR experiments (35).
Distance constraints for the structure calculation were obtained from three NOESY spectra recorded at a proton frequency of 750 MHz with a mixing time of 40 ms: a three-dimensional 13 (36) was used to convert NOE intensities into upper distance bounds according to an inverse sixth power volume-to-distance relationship (37). Final structure calculations using the torsion angle dynamics protocol of DYANA with 8Ј000 steps were started from 100 randomized conformers. The 20 conformers with the lowest final DYANA target function value were energy-minimized in a water shell with the program Opalp (38) using the Amber force field (39). Figures of molecules were prepared with the program Molmol (40).
Amide Proton Protection Factors-The exchange rate of amide protons in proteins with deuterium is generally analyzed in terms of a two-state equilibrium between the protected, closed form of the protein and the unprotected, open form. The exchange of amide protons with deuterium takes place from the open form with the intrinsic exchange rate constant k intr , where k o is the first order rate constant for the opening of the folded protein, k c the rate constant for its return to the closed state, and k intr the intrinsic rate constant for exchange with solvent for an unprotected amide proton that is known from model-peptide studies (41). Under folding conditions k c Ͼ Ͼ k o ; thus the observed exchange rate k obs becomes In the limiting case of k c Ͼ Ͼ k intr (EX2 exchange regime), Equation 1 can be simplified to In the EX2 regime the exchange data can be used to measure the local equilibrium constant between "open" and "closed" backbone hydrogen bonds. This equilibrium constant is given by the ratio k c /k o , which defines the protection factor (P) of each amide proton.
The protection factors for each individual residue can be related to the free energy of the apparent opening reaction that dominates exchange under EX2 conditions according to the equation, The amide proton exchange rates were calculated following the decrease of two-dimensional [ 15 N, 1 H]HSQC cross-peak integrals over time after dissolving the lyophilized protein in D 2 O. The resulting decay curves were fitted to a single exponential decay equation. The intrinsic exchange rates k intr were calculated taking into account the effect of neighboring side chains and corrected for temperature and pH effects (41). The structure calculation was performed with the program DYANA (36), and the relevant parameters are given in Table I. The small residual constraint violations show that the structure is consistent with the experimental constraints, and the global root-mean-square deviation values among the bundle of 20 energy-minimized conformers is representative of a high quality structure determination (Fig. 1). The analysis of the family of conformers, the NOE constraints, and the 13 C ␣ chemical shifts (Fig. 2a) indicate the presence of the following secondary structure elements: ␤-strand 1 with residues 128 -131, helix ␣1 with residues 144 -156, ␤-strand 2 with residues 161-164, helix ␣2 with residues 174 -194, and helix ␣3 with residues 200 -228. Although helix ␣2 and helix ␣3 are regular ␣-helices, helix ␣1 consists of a regular ␣-helix comprising residues 144 -152 followed by a short 3 10 -helix of residues 153-156. The amide proton of Met 134 forms a hydrogen bond with the carbonyl oxygen of Asn 159 that would be compatible with an elongation of the ␤-sheet toward the first ␣-helix with a ␤-bulge at residue 132. The loop region of residues 165-172, which connects ␤-strand 2 and helix ␣2, is poorly defined, and its resonance signals show line broadening because of dynamic processes in the millisecond time scale, which in some cases prevents the detection of NMR signals (42).

Resonance Assignment and
Thermal Stability and Amide Proton Protection Factors-The thermal stability of hPrP-(121-230) was measured follow-ing circular dichroism at 222 nm during heat denaturation and subsequent renaturation. At pH 7.0, hPrP-(121-230) undergoes a highly cooperative and reversible two-state transition with a melting temperature of 71°C and a free energy of unfolding of ⌬G UϪN ϭ 30 kJ mol Ϫ1 . This value is similar to ⌬G UϪN ϭ 28 kJ mol Ϫ1 of mouse PrP-(121-231) (43) as determined from urea equilibrium denaturation.
Exchange of backbone amide protons against deuterium was measured after dissolving lyophilized hPrP-  relatively small values of ⌬␦( 13 C ␣ ) in Fig. 2a.
For those peptide segments with P Ͻ 100, exchange occurs within the time required for recording the first two-dimensional [ 15 N, 1 H]HSQC experiment, i.e. within 15 min. For the amide protons with a high degree of protection, the calculated values of the free energy of exchange, ⌬G HX , is close to the value of ⌬G U-N , indicating that within these segments of secondary structure exchange occurs from a folding state that is very similar to the denatured protein (44).

Comparison of the Structure of hPrP-(121-230) at Neutral pH Versus Acidic pH-
The three-dimensional structures of hPrP-(121-230) at pH 7.0 and 4.5 show both global similarities and local differences. The global structure at neutral pH is similar to that at acidic pH (26), with a root-mean-square deviation value of 1.3 Å between the backbone heavy atoms of residues 125-228 in the mean structures determined at the two pH values (Fig. 4). Significant differences between the two structures are localized at the C-terminal ends of helices ␣1 and ␣2.
At pH 7.0, residues 153-156 at the end of helix ␣1 adopt a 3 10 -helix conformation, whereas at acidic pH the same residues show a less regular conformation (26). Interestingly, an elongated helix ␣1 has also been described for the structure of hPrP-(90 -231), which was determined from crystals grown in pH 8 solution (28). Helix ␣1 is extremely hydrophilic and has a low capacity to form hydrophobic contacts (45), indicating that its regular secondary structure must be stabilized by electro-static interactions. The protonation of His 155 at acidic pH, therefore, might contribute to the destabilization of helix ␣1 by introducing an unfavorable second positive charge close to Arg 156 .
Within the framework of the preserved global structure, the backbone atoms of residues 188 -194 at the C-terminal end of helix ␣2 show a better precision for the bundle of conformers calculated at pH 7.0 than at pH 4.5 (Fig. 4). The helical content within this region is also more populated at pH 7.0, as evidenced by the comparison of 13 C ␣ chemical shifts shown in Fig.  2b. The same helical segment in hPrP-(121-230) at pH 4.5 is in equilibrium with unfolded conformations (26). In this regard it is interesting to note that in the crystal structure of dimeric hPrP-(90 -231) residues 189 -198 of helix ␣2 serve as a hinge region for the rearrangement of helix ␣3 into the neighboring molecule in the crystal (28). The reduced stability of this region at low pH could be due to the protonation of His 187 , as positively charged histidine side chains in the middle of ␣-helices have been shown to have a destabilizing effect because of unfavorable interaction with the helix macrodipole (46). The surface of the NMR structure of mouse PrP-(121-231) determined at pH 4.5 is reportedly characterized by a markedly uneven distribution of positively and negatively charged residues (24), and the same uneven distribution is seen also in human PrP-(121-230) (26). It has been proposed that this dipolar arrangement of charges might stabilize the orientation of PrP C with its positively charged surface and also the nearby hydrophobic patches toward the cell membrane (24). The two glycosylation sites at Asn 181 and Asn 197 would then be located on the opposite, negatively charged site. The electrostatic surface potentials of the human PrP globular domain at pH 4.5 and 7.0 are compared in Fig. 5. At pH 4.5, there is an uneven distribution of positive and negative charges (Fig. 5, c and d), whereas at pH 7.0 the polarity of the electrostatic surface potential is less obvious (Fig. 5, a and b). Although the negatively charged surface is similar at the two pH values (Fig. 5, b  and d), the net charge at the opposite surface is less positive at the neutral pH value (Fig. 5, a and c). A major contribution to the positive net charge at acidic pH (Fig. 5c) comes from the protonated side chains of histidine residues 155 and 187. At the cell surface, where the average interstitial milieu of the brain has a pH value of 7.3 (15,16), the affinity of this side of PrP C to the negatively charged cell membrane therefore is not as stringent as expected from the PrP structures in pH 4.5 solution.
Influence of pH on PrP Globular Domain Local Stability-The globular domain of the human prion protein is more stable at neutral pH than at acidic pH. At pH 4.5, the thermal unfolding of hPrP-(121-230) is not completely reversible, but there appears to be a decrease in melting temperature by about 7°C (42). The measured amide proton exchange protection factors allow us to map at the amino acid level the pH-dependent destabilization of protein backbone structure. Fig. 6 illustrates the local differences in free energy of unfolding, ⌬⌬G HX , as calculated from amide proton exchange experiments at pH 7.0 and 4.5 (26). The ⌬G HX values are generally lower at acid pH, with the largest changes being clustered in two segments, helix ␣1 and the second ␤-strand. Within this region ⌬G HX Ͻ ⌬G U-N , suggesting a contribution of local unfolding events to exchange (44). This finding coincides with the results from an unfolding simulation of Syrian hamster PrP-(109 -219) at low pH (47), where the protein core, consisting of helix ␣2 and helix ␣3, was maintained, whereas a conformational transition and instability was particularly prominent at the N-terminal part of the protein and within helix ␣1 and the ␤-sheet.
Implications of pH Dependence for PrP C to PrP Sc Conversion-From the combination of structural and thermodynamic data, the decreased local stability of helix ␣1 and ␤-strand 2 at low pH thus might provide a "starting point" for the processes that ultimately leads to the conformational switch from ␣-helix secondary structure in PrP C into ␤-sheet secondary structure in PrP Sc during prion propagation. Consistent with this structural model for PrP conversion, a recent structural model of PrP Sc based on electron crystallography (48) suggests that helix ␣2 and helix ␣3 adopt a very similar conformation in both PrP C and PrP Sc , whereas the peptide segment comprising residues 90 -165 converts into a ␤-sheet-rich conformation. The impact of His 155 on the structure and stability of helix ␣1 is intriguing, as cell-free conversion experiments with chimeric mouse/hamster PrP have shown that the PrP Sc epitope of hamster PrP C includes Met 139 , Asn 155 , and Asn 170 (49). These results are consistent with the structural conversion of PrP C into PrP Sc taking place at acidic pH, i.e. along the endosomic pathway (17)(18)(19).