D. S. Wishart, Departments of Biological Sciences and Computing Science, University of Alberta, Rm. 2-21, Athabasca Hall, University of Alberta, Edmonton, AB, Canada T6G 2E8 Fax: +1 780 492 1071 Tel: +1 780 492 0383 E-mail: firstname.lastname@example.org
Although highly conserved throughout evolution, the exact biological function of the prion protein is still unclear. In an effort to identify the potential biological functions of the prion protein we conducted a small-molecule screening assay using the Syrian hamster prion protein [shPrP(90–232)]. The screen was performed using a library of 149 water-soluble metabolites that are known to pass through the blood–brain barrier. Using a combination of 1D NMR, fluorescence quenching and surface plasmon resonance we identified thiamine (vitamin B1) as a specific prion ligand with a binding constant of ∼ 60 μm. Subsequent studies showed that this interaction is evolutionarily conserved, with similar binding constants being seen for mouse, hamster and human prions. Various protein construct lengths, both with and without the unstructured N-terminal region in the presence and absence of copper, were examined. This indicates that the N-terminus has no influence on the protein’s ability to interact with thiamine. In addition to thiamine, the more biologically abundant forms of vitamin B1 (thiamine monophosphate and thiamine diphosphate) were also found to bind the prion protein with similar affinity. Heteronuclear NMR experiments were used to determine thiamine’s interaction site, which is located between helix 1 and the preceding loop. These data, in conjunction with computer-aided docking and molecular dynamics, were used to model the thiamine-binding pharmacophore and a comparison with other thiamine binding proteins was performed to reveal the common features of interaction.
Prion proteins (PrP) are endogenous, highly conserved membrane-anchored proteins that are particularly abundant in the neuronal cells of vertebrates. The mature form of the normal cellular isoform of the prion protein PrPc is a ∼ 200-residue glycoprotein that is tethered to the cell surface via a glycosyl-phosphatidylinositol anchor at the C-terminus . A β-rich, misfolded isoform of PrP (generically designated PrPsc) is the major macromolecule present in preparations of infectious prions. Prions are known to cause a variety of fatal, transmissible and incurable neurodegenerative diseases in both animals and humans. These include scrapie in sheep , bovine spongiform encephalopathy in cattle , chronic wasting disease in deer and elk , as well as Kuru, Creutzfeldt–Jacob disease and fatal familial insomnia in humans [3,5,6]. Prions cause disease by converting from a soluble, helix-rich form (PrPC) to an infectious β-rich form (PrPsc) that is both insoluble and highly pathogenic .
After more than 30 years of study very little is known about the physiological role of prion protein. Most of the studies published to date have focused on the identification of PrP-interacting partners such as Cu2+, Ni2+, glycosaminoglycans, DNA, RNA and a number of signaling proteins . Based on these observations, the proposed physiological roles of PrPc range from copper internalization and homeostasis to a variety of anti-apoptotic activities. Other potential functions include protection against oxidative stress, cell adhesion, cell signaling and the modulation of synaptic structure and function [9,10]. More recent findings suggest that PrPc may play a role in maintaining the long-term integrity of peripheral myelin sheaths  or even function as an antibacterial protein .
Aside from the identification of Cu2+ (and other divalent metals such as Ni2+, Zn2+, Fe2+ and possibly Mn2+) and hemin  as high-affinity PrP ligands, there has been no published evidence that PrPc binds any other endogenous small molecules. That is not to say that PrP isoforms cannot, or do not, bind small molecules. Indeed there are numerous studies that have identified a variety of exogenous or xenobiotic ligands that bind to either PrPc or PrPsc with relatively high affinity. These include tetracyclines , quinacrines , curcumin , simvastatin , Congo Red  and others. However, these small molecules are not endogenous molecules and they were identified through chemical screens aimed at finding potential prion therapies rather than potential PrP functions.
In an effort to identify potentially physiologically relevant binding partners of the prion protein, we investigated the binding of small molecules that would be easily accessible to prion proteins. Taking into account the extracellular location and the enrichment of PrPc in the central nervous system, we chose to look at a collection of water-soluble metabolites that could easily pass through the blood–brain barrier and which are found in high abundance in central nervous system tissues or biofluids. More specifically, we decided to screen recombinant prion proteins against a subset of 149 water-soluble metabolites that were previously identified as being enriched in human cerebrospinal fluid . Using a combination of 1D NMR, fluorescence quenching and surface plasmon resonance, we found that thiamine (vitamin B1) was the only ligand that exhibited clear and specific binding to PrPc. The binding constant was determined to be ∼ 60 μm. Subsequent assessment of thiamine binding showed that this interaction is evolutionarily conserved, with similar binding constants being seen for hamster, mouse and human prion proteins. We also assessed the binding of other, more physiologically abundant thiamine derivatives (thiamine monophosphate and thiamine diphosphate) and determined the exact binding site for thiamine using a combination of 2D heteronuclear NMR experiments and computer-aided docking and molecular dynamics.
1D NMR screening
One-dimensional NOESY NMR screening was performed on a total of 149 water-soluble metabolites (Table S1) in the presence and absence of recombinant Syrian hamster prion protein [shPrP(90–232)]. Spectra were compared for chemical shift, linewidth and/or lineshape perturbations of the metabolite proton signals. The criteria for selecting potential binders included the presence of new proton chemical shifts or perturbations > 0.02 ppm, > 10% reduction in peak intensity and/or > 0.2 Hz broadening of the metabolite signals. The parameters for the compound only and compound + protein spectral collection were optimized for detecting small molecules. Through this NMR analysis, we identified two potential binders that fit our criteria: thiamine and cytidine (Fig. 1A).
SPR and fluorescence studies of the thiamine-prion complex
The binding of the two candidates identified from our 1D NMR screen to the prion protein, (Fig. 1A) were also characterized using SPR as a second, independent method. At ligand concentrations of 10 mm, binding is clearly observed for thiamine (Fig. 1B). By contrast, the affinity for cytidine to PrP as measured by SPR was found to be very weak and we decided not pursue further studies with this ligand. Caffeine, a negative control, exhibited nonspecific binding as manifested by the negative binding response after dissociation. As shown in Fig. 1C, the biosensor response arising from thiamine binding is concentration dependent. Values of the association and dissociation constant (KD = 116 × 10−6m) for this ligand were calculated using regression analysis of the binding data.
Steady-state fluorescence quenching is a valuable technique to study ligand–protein interactions if the ligand binds near tyrosine or tryptophan residues . Assessment of thiamine binding to PrP using fluorescence quenching indicated a decrease in the fluorescence signal of shPrP at both 295 and 280 nm after addition of thiamine (Fig. 2A) and the calculated Stern–Volmer plot was linear for the selected concentration range. Using the fluorescence data (see Materials and methods) the estimated values for the disassociation constants (KD = 65.36 × 10−6m) were determined (Table 1, row 1). Interestingly, a similar quenching profile was found for a longer version of the Syrian hamster prion protein, shPrP(29–232) with a histidine affinity tag (Table 1, row 2), the mouse prion protein, moPrP(90–231) (Table 1, row 3) and the full-length human prion protein, huPrP(23–230) with histidine affinity tag (Table 1, row 4). These results indicate that the binding of thiamine is conserved across a wide range of mammalian species. Furthermore, they also show that the binding is not affected by the presence of the His-tag or the unstructured, copper-binding N-terminus. Similar fluorescence quenching of shPrP(29–232) was observed for the mono and diphosphate analogs of thiamine and for the moPrP(90–232) in the presence of a 3 m excess of CuCl2. The calculated KA and KD values for these constructs were in the same range of nonphosphorylated thiamine with the C-terminal core of the shPrP(90–232) protein (Table 1, rows 5–9). The Stern–Volmer plots are shown in Figs S1–S5.
Table 1. Binding affinity (KD) and correlation coefficients (C) of thiamine for different prion protein (PrP) constructs according to equation (1).
a No His-tag. b Value calculated based on tyrosine quenching. c Three molar excess (3 ×) of CuCl2 in relation to the protein and ligands were used.
65.36 × 10−6
66.66 × 10−6
58.82 × 10−6
62.11 × 10−6
72.72 × 10−6
69.52 × 10−6
Thiamine + CuCl2c
56.82 × 10−6
Thiamine(PO4) + CuCl2c
65.79 × 10−6
Thiamine(PO4)2 + CuCl2c
59.88 × 10−6
ShPrP(90–232) pH 6.0
67.71 × 10−6
ShPrP(90–232) pH 8.0
68.90 × 10−6
Additional NMR studies were conducted to further validate the PrP–thiamine binding and to identify the site of interaction. A saturation transfer difference TOCSY spectrum (STD-TOCSY) was collected to identify the ligand protons that interacted with the protein . As observed from the reference spectrum (Fig. S8B), the majority of thiamine protons, apart from the methyl groups and the methylene signal at 5.3 p.p.m., were completely suppressed. Upon addition of the shPrP(90–232), recovery of the remaining signals was observed (Fig. S8C), whereas the methyl groups and methylene proton displayed significant signal enhancement, indicating that all of thiamine’s protons are involved in interaction with the prion protein.
The thiamine binding site was investigated using heteronuclear NMR spectroscopy on a 15N-labeled sample of the shPrP(90–232 with 6 × -His tag) protein construct. Figure 3 shows the signal attenuation and chemical shift changes in the 15N-HSQC spectrum of the shPrP after addition of thiamine (33 : 1, ligand/protein molar ratio). The majority of attenuated signals (M138, F141, G142 and D144) and those with significant chemical shift perturbations (H140, N143 side chain) cluster in an unstructured loop adjacent to helix 1, which contains two tyrosines and a tryptophan residue. These aromatic residues are within the required 10 Å of the interacting ligand to exhibit fluorescence quenching at both 280 nm (Tyr, Trp excitation) and 295 nm (Trp excitation). The perturbation of the N143 side-chain amide is also noteworthy. Although it is not in direct contact with thiamine, its binding results in slight changes of the hydrogen bond length between it and the E146 side-chain carboxylic acid group. In the majority of the NMR structures (PDB: 1B10), these side chains are involved with the N-terminal capping of helix 1. In addition, a second region of the protein exhibits conformation change, which is likely a consequence of small changes in hydrogen bond lengths or other minor conformational/dynamic changes. However, no visible NOEs could be observed between thiamine and this region, which consists of the ‘amylome’ residues F169-NNQNNY-175. Long-range effects between this region and the distal residues in helix 3 have been noted in previous studies . Whether these distal effects on the amylome region upon thiamine binding have any significance or are an inconsequential artifact has yet to be determined.
Although 15N-HSQC titration spectra can be useful in identifying residues directly perturbed by ligand binding, they are also sensitive to small perturbations in hydrogen bond length, solvent structure and electric field effects not directly related to ligand binding, resulting in a number of false-positive signals. Evidence of this can be seen in the HSQC data presented in Fig. 3. Also shown are a number of resonances that exhibited some signal attenuation, likely due to ligand-induced changes in the protein’s dynamics (i.e. intermediate conformational exchange). In light of these hard-to-interpret signal perturbations, we felt that additional data were necessary to corroboate the initial findings. Consequently, a NOESY experiment (Varian VNMRJ v2.1b: tnnoesy) was performed on a sample containing a 50 : 1 molar ratio of thiamine to the shPrP(90–232). This experiment provided five unique NOE signals that could be clearly assigned to protons within the putative binding pocket (Fig. 4). Confirmation of the proton chemical shift assignments was made from 15N TOCSY-HSQC and NOESY-HSQC data. Strip plots from the residues providing NOE data are also provided in the supporting information (Fig. S7). The attenuated amide signal for G142 in the 15N-HSQC spectrum tracks from 9.05 to 8.90 p.p.m. upon addition of thiamine. This perturbation is also observed in the TOCSY-HSQC data (Fig. S7), which dislays both free and bound states in slow conformational exchange. The tnnoesy spectum exhibits a very strong NOE between thiamine’s aliphatic proton shift at 3.83 p.p.m. and G142 amide proton (at 8.90 p.p.m.) in the bound state (Fig. 4B). Other strong NOE data were observed between thiamine’s pyrimidine and isoxazole ring protons and the protons for residues M138, M139, H140 and Y150 aromatic rings (Fig. 4A). From these NMR data, the orientation of thiamine in complex with PrP was modelled in-silico (Fig. 5). Analysis of the docking result revealed that it was in good agreement with the observed NMR and fluorescence quenching data. Specifically in this model, all NOEs were satisfied, the ligand made good contact with residues displaying NMR perturbations and the placement of the ligand is within 10 Å of W145, Y149 and Y150.
By analyzing the fluorescence data, some additional clues about the nature of PrP–thiamine interaction could be obtained. The linearity of the Stern–Volmer plot proves there is only one fluorogen that interacts with the quencher (thiamine). However, these data alone cannot distinguish between a dynamic or a static quenching mechanism. To clarify this issue, we observed only a slight decrease in the slope of the Stern–Volmer plot when the experiments were conducted at higher temperatures (40 °C, data not shown). This result is characteristic of static quenching. By using equation (1):
It is possible to calculate the apparent quenching constants Kq of the two metabolites assuming the lifetime τ0 of the biomolecules to be ∼ 10 ns. It is known that the maximal diffusion collision quenching constant of quenchers is 2.0 × 1010m·s−1 . In this case, the apparent quenching constant calculated for thiamine (1.53 × 1012m·s−1) is larger than the diffusion-controlled quenching constant suggesting that the quenching effect of thiamine on the prion protein is not due to dynamic collisions. This result supports the fact that the origin of the observed quenching effect is due to the formation of a ground-state complex between the protein and thiamine. We have also excluded a Forster energy transfer mechanism as the other probable cause of static quenching because in this case the fluorescence emission spectrum of the prion protein does not overlap with the UV absorption spectrum of the thiamine.
The results obtained from our SPR studies confirmed the thiamine affinity measurements made by fluorescence (∼ 65 μm). However, the KD value obtained from SPR (∼ 116 μm) is about half that obtained by our fluorescence experiments. The lower binding constant as measured from SPR was not unexpected. Indeed, given the fact that the prion protein was covalently coupled (via lysine and arginine residues) to the SPR chip, one would expect that this chemical cross-linking would lead to a portion of the PrP molecules having their binding sites shielded or inaccessible. It is also notable that three arginine residues surround the putative binding site (R136, R148 and R151). Furthermore, reduced binding constants are expected given that some denaturation of the native structure is often induced by the covalent modification .
As noted earlier, the prion protein does bind to other endogenous ligands, including copper, zinc, manganese and nickel. The binding sites for these metals are located between residues 29 and 98, in the unstructured, N-terminal octarepeat region [25,26], well away from where thiamine binds. In most cases, the metal binding constants are quite strong. The KD for copper is an astonishingly tight 8 fm, the KD for nickel is ∼ 20 nm, the KD for zinc is ∼ 100 nm, and the KD for manganese is relatively weak at 200 μm . The other endogenous ligand that appears to bind to PrP is hemin, a naturally occurring iron-binding porphyrin. Hemin appears to bind exclusively in the unstructured N-terminus from residues 34 to 94, which corresponds to the octarepeat region . However, the binding affinity of hemin has not yet been determined. Some non-natural porphyrin analogs of have also been shown to bind to PrP with low micromolar affinity  so it is likely that the KD value for hemin would be in the same range. Among nonendogenous or xenobiotic compounds that have been shown to bind the cellular form of PrP, only quinacrine has had both its binding site and binding affinity determined. In particular, quinacrine binds to the C-terminal region of helix 3 with a rather low KD of 4.3 mm . Overall, these data suggest that the affinity of the prion protein for thiamine is within the range of other known PrP ligands. Furthermore, the thiamine binding site appears to be unique and, so far as we know, does not overlap with the binding site of other known (endogenous or exogenous) prion ligands.
To assess whether the binding of thiamine to PrP might share features with other thiamine binding proteins, we compared our PrP–thiamine structure with other known thiamine binding proteins including: transketolase, thiamine pyrophosphokinase (from mouse and Bacillus subtilus), thiamine phosphate synthase and thiamine binding protein (Fig. 6). For all the thiamine binding proteins, the amino substituted pyrimidine ring occupies a hydrophobic pocket. Stabilization and orientation of the ring is further provided through hydrogen bond acceptor oxygen atoms. In the case of pyrophosphokinase, salt bridges are also found to the positively charged imidazolium nitrogen. Finally, electrostatic and hydrogen bonds from the backbone amides and side chain amide (Q, N) or hydroxyl groups (S, T) and charged residues (D, E, R, K and H) provide contacts for the thiamine hydroxyl group or various phosphorylated forms. For the prion protein, similar contacts are observed. Specifically, the pyrimidine ring of thiamine sits in a hydrophobic pocket created by Y150, Y141 and M138, and is also stabilized electrostatically by the backbone carbonyl of M139 and the Y150 hydroxyl group. In addition, the positively charged imidazolium nitrogen forms a salt bridge with the carboxylic acid side chain from D147, and finally, the thiamine hydroxyl group is located within a hydrogen-bond distance range (2.5 Å) from the backbone amide proton of G142. There are a number of additional residues (N143, D144, D147, R148) in this region that can offer either electrostatic or hydrogen bonding interactions to either the thiamine hydroxyl group or its phosphate groups. Interestingly, it was noted that the H140 side chain played no direct part in the interaction of thiamine. In order to validate this observation, we performed additional binding studies at pH 6 and 8, thereby titrating H140 through its full range of pKa values , as well as most known physiological pH ranges. The fluorescence quenching data yielded no difference in the calculated KD constants for these pH values, indicating that pH does not influence thiamine binding (Table 1, Fig. S6). Other prion protein constructs from nonmammalian species (turtle and frog) show variability in the amino acid at this position (N and Q respectively). These findings suggest that the amino acid in this position plays no significant role in the protein affinity for thiamine.
Potential biological consequences
Thiamine (vitamin B1) is an essential, water-soluble, B vitamin that plays a critical role in carbohydrate metabolism . It is endogenously synthesized by bacteria and plants, but animals cannot synthesize it, so thiamine must be obtained from the diet. Generally, the unphosphorylated form is transported in the body, whereas the phosphorylated forms of thiamine (thiamine monophosphate, thiamine diphosphate and thiamine triphosphate) are the active forms of the vitamin. The human body keeps stores of 25–30 mg of thiamine, with the greatest concentrations being in metabolically active organs, such as skeletal muscle, the brain, the heart, the kidneys and the liver . Thiamine is known to bind to serum albumin , to a hormonally regulated protein called thiamine binding protein  and to a thiamine/folate transporter . Thiamine binding protein and serum albumin have an affinity for thiamine of ∼ 1 μm [31,32]. Thiamine and its phosphorylated derivatives have also been detected in blood, cerebrospinal fluid, milk and several other biofluids. The typical concentration of all forms of thiamine (free and phosphorylated) in human blood is ∼ 200–300 nm . Thiamine is absorbed by active transport and by passive diffusion.
Given the relatively low concentrations of thiamine in the body, one might ask how a protein with a modest (∼ 60 μm) affinity to thiamine could potentially play a biologically meaningful role. One possibility is that the prion protein functions to retain or concentrate thiamine in tissues more through avidity rather than affinity. Many copies of a weak binding protein on a cellular surface can create a ‘Velcro’ effect for ligand binding. It has been estimated that prion proteins are expressed at a level of between 2000 and 4000 copies per cell in peripheral tissues and organs and as much as 50 000 copies per cell in cerebral tissue . Clearly, if the role of PrP is to concentrate thiamine in tissues, it would need to have a relatively weak binding constant so that the thiamine could be released for absorption and subsequent utilization by cells. It is also important to note that the thiamine binding constant reported here was determined for a soluble, unglycosylated form of PrP rather than the native, membrane-bound form. It may be that native, membrane-bound PrP could exhibit higher affinities for thiamine because of the presence of the lipid bilayer or other synergistic protein–protein interactions. It is intriguing that prion proteins are highly expressed in the brain, spinal cord, heart, kidney, lung, white blood cells and lymphoid tissues . These correlate well with the tissues typically needing the highest levels of thiamine in the body and the tissues with the highest levels of carbohydrate metabolism.
In conclusion, we have identified, from an initial screening of 149 water-soluble metabolites commonly found in cerebrospinal fluid that the water-soluble vitamin B1, thiamine, interacts with multiple constructs of the prion protein. Three independent methods were used to confirm the interaction. Two of them (SPR and fluorescence quenching) led to binding constants in the weak μm range, and NMR studies pinpointed the site of interaction. Further docking studies were used to ascertain the hydrogen bonds, electrostatic and lypophilic interactions that comprise the pharmacophore. The residues involved with these interactions are conserved across multiple mammalian species (Fig. 7). Additional experiments with mouse, human and various hamster prion constructs showed that this binding was conserved across these mammalian species.
Materials and methods
Protein expression and purification
The expression and purification of recombinant shPrP(29–232), shPrP(90–231) and shPrP(120–232), moPrP(90–231) and huPrP(23–230) all followed a similar protocol. Specifically, synthetic genes corresponding each construct including a 22-residue N-terminal fusion tag containing 6 × His and a thrombin cleavage site (MGSSHHHHHHSSGLVPRGSHML) were synthesized by DNA 2.0 (Menlo Park, CA, USA). The genes were cloned into a pET15b expression vector between XhoI and EcoRI restriction sites and heat shock transformed into Escherichia coli strain BL21 (DE3). For expression, the transformed cells were grown in 100 mL Luria–Bertani broth plus 100 μg·mL−1 ampicillin overnight to generate a starter culture. Between 1% and 2% of this starter culture was then used to inoculate 1 L of Luria–Bertani media (giving a starting D600 of 0.1). The cells were allowed to reach an D600 between 0.6 and 1.0 before induction with 1 mm isopropyl thiogalactoside. Twelve to eighteen hours later, the cells were harvested by centrifugation at 1600 g for 25 min at 4 °C. In addition, 15N-labeled shPrP(90–232) was also expressed and purified from M9 media (1.0 g·L−1 15NH4Cl) for collection of heteronuclear NMR data. The inclusion of the 6× His tag afforded a standardized nickel affinity purification strategy for all protein constructs. The details of the purification protocol are described elsewhere .
NMR spectra were acquired at 25 °C on a 500 MHz Varian Unity INOVA spectrometer fitted with a 5 mm HCN z-gradient pulsed-field gradient cryogenic probe except for the STD-TOCSY, which was collected at 25 °C on a 800 MHz Varian Unity INOVA spectrometer fitted with a 5 mm HCN xyz-gradient pulsed-field gradient cryogenic probe. All experiments were collected using Varian BioPack pulse sequences (VNMRJ v2.1B). Spectra were processed using nmrpipe  and analyzed with nmrpipe and nmrviewj  unless stated otherwise.
Small molecule metabolites used in the ligand screening were purchased from Sigma-Aldrich (St. Louis, MI, USA), Fisher Scientific (Waltham, MA, USA), Acros Organics (Geel, Belgium) and Alfa Aesar (Ward Hill, MA, USA) and used without further purification. The final (cerebrospinal fluid-compatible) metabolite library consisted of 149 different water-soluble chemicals. The complete list is available in the Table S1. Seventeen screening sets containing between six and nine metabolites were manually selected with the aid of Chemaxon’s jklustor v 5.0 (ChemAxon Kft, Budapest, Hungary) using the Ward algorithm. Selection of each chemical set was made on the basis of minimizing 1H NMR spectral overlap. Two sets of 1D NOESY spectra were collected on each set of metabolites (with and without PrP protein). To collect the first NOESY NMR spectrum, each metabolite set was dissolved in 20 mm potassium phosphate buffer at pH 6.5 giving a final concentration of 100 μm. Ten percent (v/v) D2O was added to each sample to maintain a spectral lock, 1 mm of 2,2-dimethyl-2-silapentane-5-sulfonate was added for chemical shift referencing . For each 1D NOESY spectrum, 48 000 points were averaged from 256 transients over a sweep width of 6000 Hz. A recycle delay of 0.01 s and an acquisition time of 4 s were used. The mixing time for the screening experiments was 100 ms. Immediately following collection of the reference compound spectra, the NMR sample was used to reconstitute prealiquoted, lyophilized shPrP(90–232) and a second spectrum was collected with the same parameters. The molar ratio of protein to each compound was 1 : 1. Matching spectra were superposed and analyzed with the Chenomx NMR suite v6.0 to assess chemical shift and linewidth perturbations of the metabolite signals.
Saturation transfer difference TOCSY experiments were collected on a 12.5 mm sample of thiamine (20 mm K2HPO4, pH 7.5, 10% D2O) before and after the addition of 500 μm shPrP(90–232). The spectra were collected using 4096 transients with a sweepwidth of 12000 Hz, a mixing time of 10 ms, a recycle delay of 1 s and an acquisition time of 2 s. On and off irradiation frequencies corresponded to −0.73 and 24.5 p.p.m. respectively.
To collect the 2D 15N-HSQC titration spectra, a reference 15N-HSQC spectrum of the shPrP(90–232), alone, was first collected (300 μm, 350 μL, 20 mm KH2PO4, pH 7 and 6.2). These 15N-HSQC reference spectrum was collected with 2048 complex points in the 1H dimension and 256 complex points in the 15N dimension using a recycle delay of 1.5 s (nt = 120, sw = 6000 Hz, sw1 = 1800 Hz). Thiamine HCl was successively added to concentrations of 1, 2, 5 and 10 mm and the 15N-HSQC spectra recollected using identical acquisition parameters. The amide chemical shift data for the prion-thiamine complex has been deposited into the BioMagResBank (BMR17834).
A 2D tnnoesy experiment (Varian VNMRJ v2.1b) was collected on a 500 μm shPrP(90–232) sample with 25 mm thiamine in 20 mm KH2PO4, (pH 7.0, 350 μL, 10% D2O) at 25 °C. The mixing time was 50 ms and a 1.5 s recycle delay was used. Sixty-four transients were collected with sweep widths of 6000 Hz in both the direct and indirect detected dimensions (np = 1024, ni = 128). Three-dimensional 15N edited TOCSY-HSQC and NOESY-HSQC  experiments were acquired with a 500 μm shPrP(90–232) sample prepared with 10 mm thiamine in 20 mm KH2PO4 (pH 7.0, 10% D2O, 350 μL). Sixteen transients were collected for each experiment with sweepwidths of 6000 Hz in the direct and first indirectly detected dimensions (np = 1024, ni = 64). A sweep with of 1800 Hz was used for the second indirectly detected dimension (ni2 = 32). Mixing times of 50 and 100 ms were used for the TOCSY and NOESY experiments respectively. A recycle delay of 1.5 s was used and the experiments were collected at 25 °C. All samples were transferred to Shigemi (Shigemi Inc. Allison Park, PA, USA) microcell NMR tubes (350 μL) prior to spectral acquisition.
Steady-state fluorescence quenching measurements
Fluorescence emission spectra were recorded on a PTI MODEL-MP1 spectrofluorometer using a 1 cm fluorescence cell for all measurements. Two different excitation wavelengths (295 and 280 nm) were used and the scan range was 310–450 nm. Prior to collecting the fluorescence spectra, the prion protein (20 μm) was dissolved in 100 μL of 20 mm potassium phosphate buffer at pH 6.0, 7.0 or 8.0 and incubated with increasing concentrations of the metabolites of interest (10–700 μm) for 30 min with shaking (800 rpm). To analyze the effect of copper on the binding of thiamine and thiamine-phosphate analogs to PrP, CuCl2 was added to the mixture in a threefold excess previous to the addition of the ligand. Data from these fluorescence experiments were used to determine the apparent binding constant according to equation (2):
Where KS = KA, is the formation constant of the donor–acceptor (quencher–fluorogen) complex. The concentration of the quencher [Q] after titration is taken to be its ratio to protein concentration [Pt]: [Q]/[Pt]. From the slope of the linear plot of Fo/F versus [Q]/[Pt] the binding constant and dissociation constant (1/KA) were estimated. The results are expressed as mean values ± SD (n = 5–7).
The SPR data was collected on a Biacore™ system 3000 (BIAcore, Uppsala, Sweden) equipped with a CM5 sensor chip. HBS-EP buffer (10 mm Hepes pH 7.5, 150 mm NaCl, 3.4 mm EDTA, 0.005% surfactant P20) was used for immobilization of the protein. CM-dextran on A CM5 sensor chip was activated by mixing equal volumes of 50 mm N-hydroxy succinimide and 200 mm 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide followed by injection of the mixture over the sensor chip surface for 7 min at a flow rate of 5 μL·min−1. The shPrP(90–232) to be immobilized was injected over the surface for 7 min. The unreacted sites on the sensor chip surface were blocked by injection of 1 m ethanolamine, pH 8.5 for 7 min. Thiamine was diluted in HBS-EP-0.05% and simultaneously injected over the PrP flow cell and the reference for 3 min at a flow rate of 30 μL·min−1. The dissociation phase was monitored for 2.5 min. The flow cell was washed with glycine/HCl (pH 1.7) at 60 μL·min−1 for 30 s between each sample injection. Included in the assay were positive controls (quinacrine HCl and congo red), compounds known to bind shPrP(90–232), and a negative control, caffeine. The resultant sensorgrams, a plot of binding response over time, were all double-referenced by first subtracting the binding to the reference surface from the binding to the active surface, and further subtracting out the binding response of the sample diluents from all sensorgrams. The dissociation constant (KD) for thiamine HCl was calculated from data at doses ranging from 50 μm to 3.2 mm. Dilution series sensorgrams were then evaluated using the steady-state affinity analysis protocol of the biaevaluation 4.1.1 software (GE HealthCare, Piscataway, NJ, USA), to obtain affinity constants.
In silico docking protocol
The in silico docking protocol was initiated by screening models of the shPrP NMR ensemble, PDB: 1B10 , for structures having the propensity to bind thiamine in orientations that are consistent with the NMR data. This step was carried out using autodock v4.2 . Coordinates of the thiamine molecule in PDB format were obtained from the Human Metabolome Database . Hydrogen atoms were added to the thiamine molecule using the openbabel program . Non-polar hydrogens in PrP and thiamine were identified and merged with heavy atoms by autodock tools 1.5.4 . autodock tools were also employed for calculating Gasteiger charges for both the thiamine and PrP models. Six rotatable torsion angles in thiamine molecule were identified by autodock tools and allowed to freely rotate to perform flexible docking. The protein model was treated as a rigid body during the docking simulations. A large grid box (25 × 25 × 30 Å) centered in the vicinity of NMR-mapped thiamine binding site (residues M138, M139, H140, F141, G142, D144 and W145) was generated. This box also included helix 1, the loop between β-sheet 1 and helix 1, the C-terminal half of helix 2, the loop between helices 1 and 3, and the N-terminal half of helix 3. Grid spacing was set to 0.375 Å. The atom-specific affinity map, electrostatic potential map, and desolvation potential map were generated by the autogrid 4 program from the autodock package. The initial dihedral offset of the ligand and the initial position of ligand with respect to the protein model were selected randomly for every docking run. A Lamarckian genetic algorithm was selected to perform the docking simulations. Two hundred and fifty docking runs were conducted using the default autodock 4.2 parameters, except the following: 250 individuals in the ligand population, 2 500 000 energy evaluations, 0.2 Å translation step, 5° quaternion step, 5° torsion step.
During the second step, semiflexible docking was performed with xplor-nih 2.27  on the shPrP NMR structure (PDB: 1B10 :model 13). This model showed good agreement with the NOE data within the 6 Å upper-limit in our autodock simulations. Ligand and protein side-chains were allowed to be flexible, while protein backbone was kept rigid. Thiamine topology and parameter files for xplor were generated with the acpype program (http://code.google.com/p/acpype/). After a short initial energy minimization (50 steps), the thiamine orientation was optimized with Cartesian dynamics for 6000 constant-temperature steps (T = 1000 K) and 6000 steps of simulated annealing to the final temperature of 300 K, followed by 400 minimization steps. A model that had no NOE violations was selected as the final docking pose. The coordinate data for this model has been deposited into the Protein Data Bank (PDB accession no. 2LH8).
This project was funded by PrioNet Canada, the Alberta Prion Research Institute and the National Institute for Nanotechnology (NINT). We would like to thank the Canadian National High Field NMR Centre (NANUC) for their assistance and use of the facilities. The operation of NANUC is funded by the Canadian Institutes of Health Research (CIHR), the Natural Science and Engineering Research Council of Canada (NSERC), and the University of Alberta.