A monomer-dimer equilibrium of a cellular prion protein (PrPC) not observed with recombinant PrP.

Both the purified normal (protease-sensitive) isoform of the prion protein (PrP(C)) (Pergami, P., Jaffe, H., and Safar, J. (1996) Anal. Biochem. 236, 63-73) and recombinant prion protein (PrP) have been found to be in monomeric form (Mehlhorn, I., Groth, D., Stockel, J., Moffat, B., Reilly, D., Yansura, D., Willet, W. S., Baldwin, M., Fletterick, R., Cohen, F. E., Vandlen, R., Henner, D., and Prusiner, S. B. (1996) Biochemistry 35, 5528-5537; and this paper), and therefore PrP(C)-PrP(C) interactions were previously unknown. In this report we confirm recombinant PrP to be a monomer by analytical ultracentrifugation. However, by three lines of evidence (enzyme-linked immunosorbent assay (ELISA), cross-linking experiments, and size exclusion chromatography) we could also demonstrate that, under native conditions, at least part of the native bovine PrP(C) exists as a monomer-dimer equilibrium. A bovine PrP(C)-specific immuno-sandwich ELISA was developed and calibrated with recombinant PrP (Meyer, R. K., Oesch, B., Fatzer, R., Zurbriggen, A., and Vandevelde, M. (1999) J. Virol. 73, 9386-9392). By this ELISA we identified a distinct PrP(C) fraction and partially purified this protein. When serial dilutions of brain homogenate or partially purified PrP(C) were measured, using the peptide antibody C15S, a nonlinear dose-response curve was obtained. This nonlinearity was shown not to be due to an artifact of the procedure but to a monomer-dimer equilibrium of PrP(C) with preferential binding of the antibody to the dimer. From the curvature we could deduce the association constant (3.9 x 10(8) M(-1) at 37 degrees C). Accordingly, DeltaG degrees of the reaction was calculated (-48.6 kJ M(-1)), and DeltaH degrees (9.5 kJ M(-1)) as well as DeltaS degrees (0.2 kJ K(-1) M(-1)) were extrapolated from the van't Hoff plot. When serial dilutions of monomeric recombinant PrP were tested, only a straight line was obtained, supporting our hypothesis. Additional evidence of dimer formation was revealed by Western blotting of partially purified PrP(C) cross-linked by the homobifunctional cross-linker BS(3). Finally, size exclusion chromatography of partially purified PrP(C) fractions revealed an additional shoulder not observed with recombinant PrP. The difference in respect of dimer formation between native PrP(C) and recombinant PrP could be explained by the lack of glycosylation of the latter.

The prion protein (PrP) 1 was detected in attempts to identify the infective agent of transmissible spongiform encephalopathies (4). Later, several isoforms of this protein were described and named, in particular PrP C (5), either membrane bound (6) or soluble (7,8), and PrP Sc (9). All of these isoforms have essentially the same amino acid sequence but different biochemical characteristics. They are sialoglycoproteins (10) with two possible glycosylation sites, leading to diglycosylated, monoglycosylated, and nonglycosylated forms (11). Membranebound PrP C has a phosphatidylinositol anchor by which it is bound to the cell membrane (12). Most of the biochemistry of PrP C is known from recombinant PrP, because PrP C is comparatively rare even in the brain, and only a few micrograms have yet been purified (1,13). Recombinant PrP is a monomer (2); its structure has been elucidated by nuclear magnetic resonance (14). Membrane interaction (15), copper binding (16), and superoxide dismutase activity (17) have all been described. However, all recombinant PrPs have been cloned and expressed in bacterial expression systems. Therefore, they lack both glycosylation and a phosphatidylinositol anchor. The influence of these two posttranslational modifications on structure and function is largely unknown.
PrP Sc is part of, or even identical to, the prion, the infective agent of transmissible spongiform encephalopathies (18). It is not very soluble and mostly aggregated in prion rods or amyloid deposits (19). Whereas PrP C has a high ␣-helix content, ␤-sheets predominate in PrP Sc (20). Prions, including PrP Sc , are remarkably heat-and protease-stable, making infectivity difficult to destroy (21,22). In spongiform encephalopathies, PrP C is converted into PrP Sc by an unknown process (20,23). Some prion diseases, such as familial Creutzfeldt-Jakob disease (24), Gerstmann-Strä ussler-Scheinker disease (25), and fatal familial insomnia (26) of humans, are caused by germline mutations of the PrP gene, which facilitate conversion into the pathological isoform. Others, such as variant Creutzfeldt-Jakob disease and bovine spongiform encephalopathy, have been caused by accidental transmission of prions with contaminated food (27,28). The conversion of PrP C into PrP Sc in infected animals involves a conformational change within the N-terminal segment of the protein (29,30). This conformational change is induced by the presence of PrP Sc (31). Several hypotheses exist about the mechanism of this interaction (20). A seeding model was proposed, in which a spontaneous, reversible thermodynamically controlled conformational change of PrP C to PrP Sc was postulated. PrP Sc is stabilized only when bound to a crystal-like seed or aggregate of PrP Sc . Seed formation is extremely slow, but once a seed is present monomers can be added rapidly (20). However, increasing experimental evidence argues for a more specific interaction of PrP C with PrP Sc (32). The conversion of PrP C to PrP Sc was inhibited by antibody binding to PrP C in vitro and was interpreted as steric blocking of a binding site to PrP Sc (33). The site of this interaction was located on amino acid positions 91-146 using synthetic peptides (34). Such protein-protein interactions were absent in bovine recombinant PrP (34), and highly purified PrP C has not been shown by others to form dimers in vitro (1). Because both purified PrP C and recombinant PrP were found to be present in a monomeric form, the PrP C-PrP Sc interaction was thought to require additional factors. It was postulated that in an uninfected cell PrP C should exist in equilibrium in its monomeric ␣-helical state or bound to a hypothetical protein X (35). The hypothetical protein X, a PrP-binding protein present in brain homogenates, would enable dimerization (35,36) and could be a requirement for PrP C -PrP Sc interaction. The PrP C -protein X complex would then bind PrP Sc , creating a replication-competent assembly (36).
By antibody studies to monitor protein expression in native bovine brain tissues, we obtained convincing evidence of a monomer-dimer equlibrium of at least a fraction of PrP C . This evidence was further confirmed by cross-linking and by size exclusion chromatography of partially purified PrP C . Such protein-protein interactions were absent in recombinant protein, showing for the first time a biochemical difference in respect to the native, glycosylated form.

MATERIALS AND METHODS
Preparation of Brain Homogenates-Brain material (thalamus) was derived from normal Swiss cattle. Brain tissue from the fish Salmo truta and from PrP null mice was used as a negative control and for preparing dilutions. Fragments of brain tissue (Ն0.5 g) were homogenized in 10 ml of a 320 mM sucrose solution per g (wet weight) with an Ultra-Turrax T25 (Janke and Kungel, Staufen, Germany). The homogenate was cleared by a short (5 min) centrifugation at 7000 ϫ g.
Recombinant PrP-Recombinant PrP was obtained from Prionics Ltd. (Zü rich, Switzerland). Recombinant bovine PrP open reading frame was amplified by polymerase chain reaction from genomic DNA using the primers 5Ј-GGGAA TTCCA TATGA AGAAG CGACC AAAAC CTTG and 5Ј-CGGGA TCCTA TTAAC TTGCC CCTCG TTGGTA. The resulting product was cloned into pET11a (Novagen). The resulting plasmid (pBPrP3) was transfected into Escherichia coli BL21 (DE3). Recombinant bovine PrP was purified from inclusion bodies, after solubilization in 8 M urea, 10 mM 3-(N-morpholino)propanesulfonic acid, first on a carboxymethyl-Sepharose column and then by reverse-phase high pressure liquid chromatography (C4 protein column, Vydac Two sedimentation equilibrium runs were carried out at 0.15 and 0.05 mg/ml in the same cell as mentioned above. Both runs were performed at 20°C and 22,000 rpm. Records were taken at 230 nm. The molecular mass was calculated using a floating baseline computer program that adjusted the baseline absorbance to obtain the best linear fit of absorbance versus the square of the radial distance. For calculations, a partial specific volume of 0.714 ml/g, a buffer viscosity of 1.001 centipoise, and a buffer density of 1.001 g/ml were used. Anti-PrP Antibodies-For detection of PrP in Western blots and for ELISA, two different anti-PrP antibodies were used, one monoclonal antibody (6H4) and a rabbit antiserum (C15S). C15S was raised against a peptide of the bovine PrP sequence (37) (GQGGT HGQWN KPS). Both antibodies, 6H4 (30) and C15S, are described in detail elsewhere (3). Both could detect PrP C and PrP Sc in immunocytochemistry and Western blot (3). Western Blotting-Samples were first separated on either 10 or 12% sodium dodecyl sulfate polyacrylamide gels and then blotted on polyvinylidene difluoride membranes (Millipore). The membranes were then blocked for 1 h in PBS-Tween (137 mM NaCl, 2.7 mM KCl, 1.4 mM KH 2 PO 4 , 8.1 mM Na 2 HPO 4 , 0.01% Tween 20, pH 7.3) containing 10% dry milk. First and second antibodies were diluted 1:1000 to 1:5000 in PBS-Tween containing 3% dry milk and successively incubated with the membranes after thorough washing with PBS-Tween. Second antibodies were either swine anti-rabbit immunoglobulins or rabbit antimouse immunoglobulins (Dako, Glostrup, Denmark) labeled with horseradish peroxidase. Detection was carried out with ECL (Amersham Pharmacia Biotech) according to the provider.
ELISA for PrP-ELISA plates were coated by overnight incubation with 0.1 ml of carbonate buffer (15 mM Na 2 CO 3 , 35 mM NaHCO 3 , 0.02% NaN 3 , pH 9.6) containing 1 g of monoclonal antibody 6H4 at 4°C per well. The plates were blocked with 0.2 ml of RPB containing 0.01% Tween 20 (RPB-Tween) and 10% dry milk per well for 1 h at 37°C. The samples were prepared by diluting bovine brain homogenate (e.g. 200, 165, 130, 95, 60, and 25 l) or partially purified PrP to 400 l with RPB-Tween additionally containing either 5% dry milk or 1:2 diluted PrP null mouse or fish brain homogenate as indicated. All samples were incubated at 4, 25, or 37°C for 45-60 min, except the standard, which was incubated solely at 25°C. Samples were loaded in triplicates to the plates (0.1 ml/well). Incubation was for 1 h at the temperature used for sample incubation (if required a standard was run on a separate plate). After washing with RPB-Tween, the bound PrP was quantified by successive incubation with two other antibodies (1 h each at 37°C). One was C15S, and the other was a horseradish peroxidase-conjugated swine anti-rabbit antibody (Dako). The rabbit PrP antiserum was diluted 1:500 in RPB-Tween containing 3% dry milk, and the swine anti-rabbit antibody was diluted 1:300 in the same buffer. 0.1 ml/well was applied. After incubation the plates were washed with PBS-Tween. 0.2 ml of 2,2Ј-azinobis (3-ethylbenzthiazolinesulfonic acid) solution (Roche Molecular Biochemicals) per well was added for reaction with horseradish peroxidase. The plate was then measured at 405 nm in an ELISA reader. The calibration procedure with recombinant PrP is described elsewhere (3). To create a standard, bovine brain homogenate was diluted with fish brain homogenate to a concentration corresponding to 75 ng/ml. A 0.1-ml concentration per well of a 1:2 dilution in RPB-buffer of this standard was included in each plate to determine the length of the horseradish peroxidase reaction. The A of the plates was read when the standard reached an A between 0.900 to 1.100. To allow comparison of the results of different plates, zero values were subtracted first, and then all results were divided by the value of the standard. Its optical density (75 ng of PrP/ml) became 1.000 at this step.
Partial Purification of PrP C -Pieces of brain tissue (about 10 g) were homogenized in 10 ml of a 320 mM sucrose solution per g (wet weight) with an Ultra-Turrax T25 (Janke and Kungel). The homogenate was cleared by a short (5 min) centrifugation at 7000 ϫ g. The supernatant was separated from the pellet and diluted with RPB-Tween and guanidine thiocyanate to 0.1 M guanidine thiocyanate. The solution was transferred to a 200-ml glass bottle with an airtight metal cap. The bottle was incubated in a laboratory oven set at 150°C (actual temperature, 150 -160°C) for 20 min. After cooling to room temperature, the homogenate was centrifuged again for 15 min at 7000 ϫ g. The supernatant was dialyzed overnight against RPB-Tween diluted 1:5 with distilled water. Finally, the dialysate was concentrated in a vacuum evaporator to about one-tenth to one-fifteenth of the starting volume. Protein concentration was measured by the bicinchoninic acid (BCA) reagent kit obtained from Pierce, and the concentration of PrP C was measured by ELISA.
Dot Plot Assay-The assay was adapted according to a procedure published elsewhere (38). Recombinant PrP, partially purified PrP C , and thalamus homogenate were serially diluted in RPB-Tween containing 5% dry milk. Samples of 5 l were applied to nitrocellulose membranes (Bio-Rad) and air dried. The membranes were transferred to RPB-Tween containing 5% dry milk and incubated for 1 h at room temperature. 6H4 and rabbit anti-mouse immunoglobulins (Dako), labeled with horseradish peroxidase, were diluted 1:1000 to 1:5000 in PBS-Tween containing 3% dry milk and then incubated with the membranes for 1 h after thorough washing with PBS-Tween. Detection was carried out with ECL (Amersham Pharmacia Biotech) according to the provider.
BS 3 Cross-linking-The homobifunctional cross-linker bis(sulfosuccinimidyl)-suberate (BS 3 ) was obtained from Pierce. A solution of 5 mM BS 3 in 5 mM sodium citrate buffer (pH 5.0) was freshly prepared. Four 50-l samples of partially purified PrP C were diluted with 0, 1.9, 9.5, and 38.0 l of 5 mM BS 3 and water to 95 l. After 20 min 5 l of 1.5 M Tris buffer pH 6.8 was added to all samples to stop the reaction. Fifteen min later the samples were mixed 1:2 with sample buffer and analyzed on Western blots.
Size Exclusion Column Chromatography-A 30-cm column having a diameter of 1 cm was filled with Macro-Prep S.E. 1000/40 (Bio-Rad) according to the instructions of the provider. The column was washed and run with RPB-Tween. The typical flow rate was 0.47 ml/min. Fractions of 0.75 ml were collected. Calibration was done by running reconstituted gel filtration standards (Bio-Rad). For PrP analysis samples of 0.75 ml of partially purified PrP C were loaded. All 35 fractions collected were analyzed for protein by the BCA reagent kit (Pierce) and for PrP by ELISA.
Calculations-For calculations Microsoft Excel was used, and for statistics Statistix Analytical software was used. For calculation the molecular weight of both the recombinant PrP and PrP C was assumed to be 24,000.

A Native PrP C Fraction Detected by ELISA
A bovine PrP C -specific immuno-sandwich ELISA was developed and calibrated with recombinant PRP (3). To learn more about the nature of PrP C detected by this assay, PrP C was partially purified from normal bovine brain thalamus using the ELISA as a purification guide (Table I). In a first step, heatlabile proteins were precipitated by heat treatment. Most proteins, but little of the PrP under consideration, were precipitated by this procedure. After removing the precipitate by centrifugation, the supernatant was dialyzed and finally concentrated by vacuum evaporation. A Western blot of relevant samples is shown in Fig. 1. Bands having approximate molecular weights of 25,000 and 28,000 -35,000 were observed in untreated brain homogenate. In partially purified fractions the patterns of the bands at 28,000 -35,000 remained mainly unchanged by the heat treatment (compare lanes a and c in Fig.  1). However, their intensity was markedly reduced (compare lanes a and b in Fig. 1), and an at least 7-fold concentration was needed to restore it.
Quantification of the PrP bands of homogenates in Western blots (Fig. 1a) revealed no correspondence with the quantitative results obtained by the ELISA (Table I), suggesting that not all PrP C was detected. For further quantification, a dot plot assay was used (38). Recombinant PrP, partially purified PrP C , and thalamus homogenate were serially diluted, adsorbed to nitrocellulose membranes, and tested for PrP C by immunological methods using 6H4 as the detecting antibody (see "Materials and Methods"). The relative amount of PrP in the different samples was compared with the known amount of recombinant PrP. In brain homogenate, about 10 times more PrP C was detected with the dot blot assay than with the ELISA. But the amount of partially purified PrP C detected both by the dot plot assay and by the ELISA corresponded well with each other (Fig. 2). In conclusion, the ELISA detected only about 10% of the PrP C present in untreated thalamus homogenate, and only this fraction was actually partially purified. The nature of the remaining PrP C , not detected by the ELISA, was not analyzed further.

Dose-response Curve
When serial dilutions of PrP C were measured with the ELISA, a quadratic dose-response curve was obtained (Fig. 3). The shape of the curve did not depend on the concentration of the antibody used nor on the composition of the diluting agent. The dose response was linear when the concentration of the first, second, or third antibody (1:1000 to 1:100) was varied and the PrP C concentration was kept constant. With no difference in the result, fish brain homogenate, nonfat dry milk (10%), and brain homogenate of PrP null mice were tested as diluting agents. The bovine brain homogenate could be diluted even without added proteins. However, the results were quite inconsistent, and therefore the dilutions were routinely done with proteins included in the diluting buffer.
The dose-response curves of PrP C in untreated brain homogenate and partially purified PrP C were indistinguishable when tested at the same concentration range. But the size and curvature of the dose-response curve did depend on the temperature of the samples at the point of time when they were loaded to the ELISA plates (4, 25, and 37°C were tested; see Fig. 3). We concluded that the nonlinear dose response was a property of PrP C and not an artifact of the assay.
The quadratic nature of the dose-response curve suggested a monomer-dimer equilibrium of PrP C , with mostly dimers contributing to the measured A. To prove this hypothesis, we attempted to fit the experimentally obtained A by mathematical modeling based on the law of thermodynamics and of mass action.
Accordingly, six different PrP C concentrations between 0.2 and 1.6 nM were selected. A sample of each was incubated at three different temperatures, 0, 25, and 37°C, and measured in the ELISA. The averaged results of the optical density of at  Table I) on a 12% acrylamide slab gel. Lane a, brain homogenate. Western blot staining intensity does not correlate in this fraction with the 43 ng/ml PrP C detected by the ELISA. Lane b, supernatant of the heat-treated and centrifuged brain homogenate. 37 ng/ml PrP C were detected by the ELISA in this fraction, and Western blot staining intensity does correlate. Lane c, the same as in lane b but concentrated by vacuum evaporation to 262 ng/ml PrP C . The position and molecular weight (ϫ1000) of standards are given on the left. The bands were developed with 6H4 diluted 1:500. least three independent experiments were used for subsequent calculations (Fig. 3). In this respect, the total concentration (nM) of soluble PrP C in each of the samples was designated "c", the monomer concentration was designated "[PrP]," and the hypothetical dimer concentration was designated "[PrP 2 ]." c ϭ ͓PrP͔ ϩ 2 ⅐ ͓PrP 2 ] (Eq. 1) According to our hypothesis, A should obey equation 2, where ␣ is a proportionality constant representing the efficiency of PrP C binding to 6H4, of C15S binding to PrP C , of the swine antibody binding to the rabbit antibody, and finally the horseradish peroxidase color reaction. A/␣ ϭ K ⅐ ͑c Ϫ ͑2 ⅐ ͑A/␣))) 2 (Eq. 5) Using each set of A values obtained at the three different incubation temperatures, we searched for values of ␣ for which a plot of A/␣ versus (c Ϫ (2⅐(A/␣))) 2 yielded a straight line through the origin with the slope K (Fig. 4). For most values of ␣, bizarre nonlinear plots were obtained. However, when using 1.7 nM Ϫ1 (4°C), 2.5 nM Ϫ1 (25°C), and 3.1 nM Ϫ1 (37°C) for ␣, straight lines were obtained having slopes (K) of 0.25, 0.33, and 0.39 nM Ϫ1 , respectively (Fig. 4). As shown in Fig. 5, a plot of ln ␣ T versus 1/T yielded a straight line. In Fig. 3 the calculated dose-response curves are superimposed to the average of the experimental values.
When freshly prepared dilutions were incubated less than 30 min at the selected temperature before loading to the plates, the resulting A was usually higher than calculated (data not shown).

Calculation of the Change in Free Energy of the PrP Monomer-Dimer Equilibrium
The formula ⌬G°ϭ ϪRT ln K was used to calculate the free energy of the dimerization reaction. The result was a change in free energy of ⌬G°ϭ Ϫ48.6 kJ M Ϫ1 (with K ϭ 0.33 nM Ϫ1 at 25°C). To obtain the change in enthalpy (⌬H°) and entropy (⌬S°) of the equilibrium reaction, ln K T versus 1/T was plotted (van't Hoff plot) (Fig. 6). The respective values were calculated form the slope and the intercept of the linear regression. The values obtained were 9.5 kJ M Ϫ1 for ⌬H°and 0.2 kJ K Ϫ1 M Ϫ1 for ⌬S°.

Other Experiments Supporting PrP Dimer Formation
Recombinant PrP-Recombinant PrP has been reported to be a monomer by others (2, 39, 40). For verification, we investigated the aggregation of recombinant PrP by sedimentation velocity and sedimentation equilibrium. Both the resulting mo- FIG. 2. Semiquantitative dot plot assay of serial dilutions of recombinant PrP (rPrP), bovine brain homogenate (thalamus), and partially purified PrP C . 5 l of each solution were pipetted at each spot. All dilutions were done in 5% dry milk. The concentration of purified PrP C stock solution was 262 ng/ml (Table I). According to ELISA measurements the respective concentration was 86 ng/ml in the brain homogenate (brain hom.). The plot was developed with 6H4 diluted 1:500. lecular weight of 25,000 and the sedimentation coefficient of s 20,w ϭ 2.1 gave no proof of dimerization or aggregation. When serial dilutions of recombinant PrP were tested in the ELISA, a linear, comparatively weak signal was observed (Fig. 7, rectangles). With PrP C these assay conditions resulted in a quadratic dose-response curve (Figs. 3 and 7), suggesting that mostly dimers contribute to the signal (equation 2). Because dimers were not present in recombinant PrP, equation 2 was adapted to equation 6.
From the slope of the linear regression, the proportionality constant ␥ was calculated to be 0.1 nM Ϫ1 at 25°C. All the results with (bovine) recombinant PrP described above did not depend on the protein composition of the diluent in which the protein was dissolved. The same outcome was observed when either 10% dry milk or brain homogenate from fish or PrP null mice was used. Size Exclusion Column Chromatography and Cross-linking of PrP C -To get additional evidence of the dimerization reaction besides dose-response curves, partially purified PrP C was cross-linked by adding the homobifunctional cross-linker BS 3 and analyzed on Western blots. No high molecular weight bands were observed without the cross-linker (Fig. 8, lane a). Additional bands at the molecular weight of PrP C dimers became visible when the samples had been incubated with the cross-linker (Fig. 8, lanes b-d). The distribution and intensity of those bands was varied with the effective BS 3 concentration. In the untreated sample PrP bands were observed at molecular weights of about 28,000, 33,000, and 35,000. When 0.1 mM BS 3 was added one additional band appeared at M r 63,000. With 0.5 mM BS 3 two additional bands appeared, at M r 63,000 and 76,000. With 2.0 mM BS 3 four additional bands at M r 63,000, 76,000, 97,000, and at the top of the gel were observed with a significant reduction of the staining intensity of the PrP bands of lower molecular weight.
For additional verification of the PrP monomer-dimer equilibrium, size exclusion column chromatography was performed on a Bio-Rad Macro-Prep S.E. 1000/40 column. The column was calibrated with protein standards. Partially purified PrP C was run over the column, and each fraction was analyzed both for protein by BCA and for PrP by ELISA. As shown in Fig. 9, the main peak of PrP was between the position of ovalbumin (M r 44,000) and myoglobin (M r 17,000). However, about 30% of the PrP detected ran as a shoulder in front of the M r 44,000 marker but behind gamma globulin (M r 158,000), indicating the presence of PrP polymers. The PrP C fractions eluted from the column still behaved like a monomer-dimer equlibrium, as judged from the results of ELISA dose-response experiments. The size exclusion column profile with recombinant PrP was as published elsewhere (40), with no shoulder observed. DISCUSSION In the present study we used an antibody binding assay, cross-linking experiments, and column chromatography to investigate a monomer-dimer equilibrium of bovine brain PrP C . Such protein-protein interactions were absent in bovine recombinant PrP, indicating that this protein does not reflect all aspects of PrP C in animal tissues.
A distinctive PrP C fraction was identified by our PrP-specific   Table I ELISA procedure. The PrP C in question could be separated from other PrP C by heat treatment in the presence of 0.1 M guanidine thiocyanate and subsequent centrifugation. The use of bovine brain, instead of e.g. mouse brain, had the advantage that large amounts of precisely specified tissue (thalamus) were available. The quantitative difference in PrP C content between untreated and partially purified samples was revealed by a semiquantitative dot plot assay (Fig. 2) and by Western blots (Fig. 1). Accordingly, only about 10% of total PrP C was selectively purified by the heat treatment method, and only this fraction was detected by the ELISA in brain homogenates ( Table I). The remaining PrP C was removed together with membranes, cellular fragments, and other proteins. Most likely the purified PrP C was soluble, either secreted (41,42) or released during homogenization. A soluble form of PrP C was described for human cerebrospinal fluid (8). In our Western blots the main difference after purification was a missing band at M r 25,000 (Fig. 1). This band probably represents the unglycosylated PrP C usually enclosed in the cell lumen (11,43) and was precipitated together with other intracellular proteins. It is difficult to discriminate between native and denatured PrP C in the absence of an accepted assay for PrP function, but usually in the absence of added detergents proteins remaining in solution are not denatured. Therefore the PrP C detected seems not to be denatured by the treatment, and its heat stability was comparable with that of the infective PrP isoform PrP Sc (21,22).
Quantification of ELISA results is difficult with nonlinear dose-response curves. We observed such a curve in our PrP Cspecific ELISA (Fig. 3). We thoroughly investigated the cause of this nonlinearity. In a first series of experiments we excluded artifacts caused by protease digestion, by the diluting agent, or by other experimental procedures. We verified that it was not connected to the antibody concentration. We finally concluded it had to be an intrinsic property of the detected PrP C . The quadratic nature of the curve suggested a monomer-dimer equilibrium of PrP C as the most likely explanation, with most of the antibody binding to the dimeric form. Based on this hypothesis, we were able to describe the nonlinear dose-response curve by the law of mass action and by the law of thermodynamics. We predicted, for example, that according to the law of thermodynamics both the intensity and curvature of the results should be dependent on the temperature of the sample at the time of its incubation on the plates. This actually was observed and was a main argument for further investigations.
The conclusion was further supported by a linear dose-response curve obtained when recombinant PrP was tested under the same conditions (Fig. 7). If the nonlinear dose response was evidence of PrP C dimerization, then the linear dose response of recombinant PrP should be evidence of a monomeric form. We have shown by equilibrium centrifugation that recombinant PrP is indeed present as a monomer, as suggested by others (39,40). In addition, with recombinant PrP the proportionality constant (␥ in equation 6) was only 4% of that observed with tissue-derived PrP C (␣ in equation 2). Therefore, the affinity of the antibody to PrP monomers was only about 4% as compared with PrP dimers, allowing monomers to be omitted in equation 2. We do not know why C15S would preferentially bind to dimers. The most logical explanation for this phenomenon is that both antibody-binding sites are used because of the close proximity of two epitopes. Such a constellation would be present in a dimeric form of PrP C .
Mathematical replication of the quadratic dose-response curve revealed the dissociation constant K of the monomerdimer equilibrium reaction of PrP C . This value should be regarded as an approximation because the experimental error of the data was up to 15% (3). However, the value could be fitted accurately in a van't Hoff plot (Fig. 6). The calculation of the free energy (⌬G°ϭ Ϫ48.6 kJ M Ϫ1 ), enthalpy (⌬H°ϭ 9.5 kJ M Ϫ1 ), and entropy (⌬S°ϭ 0.2 kJ K Ϫ1 M Ϫ1 ) of the equilibrium reaction revealed ranges not unusual for protein-protein interactions. The temperature dependence of the proportionality constant ␣ (equation 2) represented the temperature-dependent binding of PrP to 6H4. Only at this step the temperature of sample and incubation was varied, and the affinity of all other antibodies was not supposed to change therefore. Accordingly, ␣ increased with increasing temperature and obeyed van't Hoff equations. A plot of 1/T versus ln ␣ yielded a straight line (Fig.  5).
The hypothesis of a monomer-dimer equilibrium of PrP C was further confirmed by cross-linking experiments and size exclusion chromatography. Addition of the cross-linker BS 3 to partially purified PrP C resulted in additional bands having the molecular weight of PrP dimers (Fig. 8). Without the crosslinker added such bands were not visible in Western blots (Fig.  8, lane a), probably because the samples are denatured by mixing with sodium dodecyl sulfate and heating before electrophoresis. Detergents and denaturation obviously inhibit dimer formation because previously purified PrP C was in monomeric form (1). BS 3 is a homobifunctional cross-linker, which crosslinks primary amines. In proteins this is predominantly lysine (Pierce, product description). Bovine PrP has 11 lysines, with only one of them located within the signal sequence, i.e. there are ample possibilities for BS 3 cross-linking. With a spacer length of 11.5 Å, only proteins rather close to each other are cross-linked. Even in the partially purified fractions, PrP C is outnumbered by more than 8000 (by weight) by unrelated proteins (Table I), thus arguing against unspecific cross-linking of PrP molecules simply by chance. With increasing BS 3 concentration, additional bands appeared, with the concomitant disappearance of the monomeric PrP. But even with a rather high BS 3 concentration (2 mM), PrP monomers were still observed, as would be expected for a monomer-dimer equilibrium, which always has some unbound monomers (Fig. 8, lane d).
We also performed size exclusion chromatography with partially purified PrP C on a calibrated column. The advantage of this technique is that it reveals the molecular weights of the native proteins. The shortcoming is a rather steep exponential molecular weight gradient. As expected, the peak of soluble PrP C appeared just behind ovalbumin (M r 44,000), as did recombinant PrP (40). But as much as 30% of loaded PrP C appeared in a broad shoulder located between gamma globulin (M r 158,000) and ovalbumin (Fig. 9). This shoulder was never FIG. 9. Size exclusion column chromatography of partially purified PrP C . Each fraction was measured for protein by the BCA protein assay (dotted line) and for PrP by the ELISA (solid line). The peak fractions of the molecular weight markers thyroglobulin (670,000), gamma globulin (158,000), ovalbumin (44,000), myoglobin (17,000), and vitamin B-12 (1350) are shown by triangles. Arrow, high molecular weight fraction of PrP C .nr., number. observed with recombinant PrP. Because of the monomerdimer equilibrium, faster-moving dimers will dissociate when they separate from the monomer pool, resulting in a deformation of the monomer peak toward higher molecular weights and not necessarily in the formation of an additional peak. Therefore, the existence of a PrP C monomer-dimer equlibrium was supported by two additional and independent experiments. The most obvious difference between tissue-derived PrP C and recombinant PrP is the lack of glycosylation of the latter. Probably, a specific PrP conformation is induced by carbohydrate addition, exposing the amino acids responsible for the dimerization reaction (33). However, highly purified PrP C has not been shown by others to form dimers in vitro (1). This obvious lack of dimerization could be explained either by denaturation during the purification process due to the use of high detergent concentrations. A PrP C dimerization has been described for neuroblastoma cells (44), but the PrP dimer observed by these authors was covalently cross-linked. We speculate that the cross-linking was enzymatically induced by the tissue culture cells used.
Our results showing a spontaneous PrP C monomer-dimer equilibrium support the concept of PrP dimer and heterodimer formation in prion propagation (32). Knowledge of the components involved in PrP interactions may not only allow the prediction of interspecies transmission of prion diseases but will also reveal possible points of intervention to interrupt the process of prion propagation.