Crystal structure of human prion protein bound to a therapeutic antibody

Abstract

Prion infection is characterized by the conversion of host cellular prion protein (PrP^C) into disease-related conformers (PrP^Sc) and can be arrested in vivo by passive immunization with anti-PrP monoclonal antibodies. Here, we show that the ability of an antibody to cure prion-infected cells correlates with its binding affinity for PrP^C rather than PrP^Sc. We have visualized this interaction at the molecular level by determining the crystal structure of human PrP bound to the Fab fragment of monoclonal antibody ICSM 18, which has the highest affinity for PrP^C and the highest therapeutic potency in vitro and in vivo. In this crystal structure, human PrP is observed in its native PrP^C conformation. Interactions between neighboring PrP molecules in the crystal structure are mediated by close homotypic contacts between residues at position 129 that lead to the formation of a 4-strand intermolecular β-sheet. The importance of this residue in mediating protein–protein contact could explain the genetic susceptibility and prion strain selection determined by polymorphic residue 129 in human prion disease, one of the strongest common susceptibility polymorphisms known in any human disease.

Prion diseases are associated with the accumulation in the brain of a misfolded protease-resistant glycoprotein known as PrP^Sc. The native cellular form of this protein (PrP^C) is ubiquitously expressed at high levels in the central nervous system, lymphoreticular tissue, and at neuromuscular junctions, and is tethered at the cell surface by a glycosylphosphatidylinositol (GPI) anchor (1). The disease-related PrP^Sc is derived from PrP^C by posttranslational processes involving conformational change and aggregation. According to the “protein-only” hypothesis (2), PrP^Sc is propagated by serving as a template for the autocatalytic recruitment of PrP^C and is the principal, possibly the sole, constituent of the transmissible agent or prion (3). Susceptibility to human prion diseases depends on the expression of PrP^C (4–6), and it has been shown that targeting PrP^C is an effective therapeutic strategy. In animal models of established neuroinvasive prion disease, abrogation of neuronal PrP^C synthesis, using transgenic manipulation (7, 8) or RNA interference (9), can prevent clinical disease onset and reverse early pathological changes. Hence, it is reasonable to focus on the PrP^C form owing to its role as the well-characterized “substrate” for prion propagation. Moreover, the fact that PrP^C is converted to a radically different conformation during prion propagation means that any molecule that binds specifically to PrP^C must stabilize the native fold, inhibiting conformation change and, consequently, prion propagation (10, 11) without the requirement for interference in PrP synthesis. We would therefore expect that antibodies specifically binding to PrP^C should inhibit prion propagation, with efficacies correlating to their PrP^C binding affinity.

Here, we show that, for a selection of anti-PrP antibodies, there is such a correlation between affinity for PrP^C and therapeutic potency in a cell model of prion infection. We also present the crystal structure of PrP in complex with the most effective therapeutic antibody, which gives us detailed information on the intermolecular interactions both between antibody and PrP and between PrP molecules.

Results

Antibodies with the Highest Affinity for PrP^C Are the Most Therapeutically Active.

The ICSM monoclonal antibodies we used were raised to either α-PrP or β-PrP, 2 different conformations of recombinant PrP [see supporting information (SI) Table S1]. The α-PrP species is formed by oxidative refolding of recombinant PrP and has the same, predominantly α-helical conformation as PrP^C, whereas the β-PrP species is a reduced and acidified form of the recombinant protein with features that resemble PrP^Sc, including a high content of β-sheet and propensity to aggregate (12, 13). The α and β conformations elicit different immune responses in mice (14), and, whereas antibodies raised against both recombinant species generally recognize both PrP^C and denatured PrP^Sc, only those raised against β-PrP can detect native PrP^Sc (15). To test the hypothesis that antibodies specifically binding to PrP^C inhibit prion propagation with efficacies that quantitatively correlate with their PrP^C binding affinity, we examined the relationship between dissociation constants for α-PrP (as measured by ELISA) and IC₅₀s for treatment of prion-infected ScN2a cells for a series of ICSM antibodies (Fig. 1A; see also Fig. S1 and Table S1). As expected, the monoclonal antibodies raised against α-PrP (ICSMs 4, 10, 17, 18, and 19) interact strongly with recombinant α-PrP, with K_d values in the 0.1–10 nM range, whereas those raised against β-PrP (ICSMs 33, 35, 37, and 42) have affinities some 4 orders of magnitude weaker, with K_d values typically in the 10 μM range. These antibodies raised against β-PrP are also ineffective in halting prion propagation, implying that targeting this PrP^Sc-like conformation is not an effective way of curing the infection.

Fig. 1.

The relationship between antibody affinity and ability to inhibit prion propagation for the ICSM antibodies. (A) IC₅₀s (M) for 3-day treatment of chronically infected N2a cells with antibody are plotted against K_d (M) for recognition of recombinant PrP by antibody as measured by ELISA (Fig. S1 and Table S1). Antibodies raised to α-PrP are shown in red and to β-PrP in blue. Equivalence between IC₅₀ and K_d is shown by the dashed diagonal line. (B) Median count of FACS measurements (n = 3) for the antibodies indicated, detected with FITC-conjugated anti-mouse secondary antibody on NS0 mouse cells.

Superimposed on the graph in Fig. 1A is the trend line that represents a quantitative equivalence between the affinity for α-PrP and the IC₅₀ for inhibiting prion propagation. The data for most of the antibodies lie reasonably close to this line, supporting the contention that the ability to inhibit PrP^Sc propagation is correlated with binding affinity for the PrP^C-type conformation. The antibodies ICSM 4, 17, and 19, however, clearly form a subgroup that does not follow this trend, i.e., despite their high affinity for PrP, they are poor inhibitors of PrP^Sc propagation. In one case, the explanation is straightforward; the ICSM 4 epitope on PrP spans the sites of N-glycosylation on mature PrP^C (15), consequently ICSM 4 binds tightly only to the unglycosylated protein (16). Also, data presented by Leclerc et al. (17) on mature PrP^C show that an epitope near to the recognition sites for ICSM 17 and 19 becomes inaccessible when the protein is at the cell surface.

To test for accessibility in situ, we used flow cytometry to probe the affinities of this subset of antibodies for mature, cell-surface PrP^C rather than recombinant PrP using ICSM 18 as a positive control (Fig. 1B). ICSMs 4, 17, and 19 gave weak signals in the FACS analysis compared with ICSM 18, indicating that their interaction is impaired and that the ability to recognize cell-surface PrP^C is important for inhibiting PrP^Sc formation in the cell system. The lower affinities of ICSMs 17 and 19 for mature PrP^C must result from the occlusion of their binding epitopes in mature membrane-associated PrP^C, as has been seen previously (17) and attributed to interaction with other ligands or to dimerization.

ICSM 18 was our most potent antibody, both as an inhibitor of prion propagation (IC₅₀ 0.7 nM) and in its affinity for PrP^C (K_d 0.1 nM). Significantly, this antibody is highly effective in treating prion-infected mice (18). In mice treated with ICSM 18, peripheral PrP^Sc levels and prion infectivity were markedly reduced, even when the antibody was administered as PrP^Sc reached maximal levels in the spleen. Furthermore, animals in which the treatment was continued remained healthy for >300 days after untreated animals succumbed to disease (18).

Crystal Structure of PrP in Complex with ICSM 18-Fab.

We visualized the interaction between PrP and ICSM 18 by determining the structure of truncated human PrP^119-231 complexed with a Fab fragment of the antibody to 2.9 Å resolution (Fig. 2 and Table 1; see also SI Text and Fig. S2 for the quality of the electron density of the structure). The overall fold of the human PrP globular domain in the complex reported here is similar to that of the C-terminal domain of human PrP in the NMR structure (19) but is different from that in the domain-swapped dimer (20), where helix 3 is swapped between the 2 molecules of the dimer, and an intramolecular disulfide bridge (Cys-179-Cys-214) is formed between helix 2 (residues 172–179) and helix 3 (200–223).

Fig. 2.

The complex between recombinant PrP^119-231 and the ICSM 18-Fab as determined by X-ray crystallography. (A) PrP^119-231 is shown in green with the heavy and light chains of the Fab in cyan and magenta, respectively. (B) Expanded view of the PrP/Fab interface. The participating PrP residues are labeled in black and those from the Fab heavy and light chains in blue and magenta, respectively. Potential hydrogen bonds are shown as dashed lines.

Table 1.

Summary of crystallographic data collection and refinement statistics

	FAB18	FAB18-PrP
Space group	P212121	P6322
Resolution, Å	22–1.57 (1.63–1.57)	42–2.9 (3.0–2.9)
Completeness, %	95 (70.0)	97.7 (83.9)
R merge	7.8 (38.3)	15.6 (79.2)
〈I〉/σI last shell	2.35	2.0
Redundancy	6.7 (3.1)	12.6 (6.9)
Overall reflections	1,080,709	482,099
Unique reflections	54,085	14,267
Wilson B-factor, Å2	17.8	62
Unit cell (Å, °)
a, Å	36.7	126.1
b, Å	84.7	126.1
c, Å	127.8	134.1
α = β, °	90	90
γ, °	90	120
Solvent, %	47.0	46.0
No. of FAB molecules	1	1
No. of PrP molecules		1
Final R crystal, %	18.3	21.0
R-free, %	21.7	26.9
ESU, Å, based on R value	0.094	1.4
No. of all atoms	4,083	4,185
No. of water molecules	728	55
No. of calcium ions	1
No. of sulfate ions		1
Average B-factor, Å2	18.3	40.0

The ICSM 18 epitope is formed by residues spanning the whole of helix 1 (H1, residues 143–156), confirming previous epitope mapping experiments (14). The Fab–PrP interface buries ≈900 Ų (15%) of the solvent accessible surface area of PrP, and there are several hydrogen bonds and salt bridges between PrP and both the light (L) and heavy (H) chains of the variable domain of the Fab molecule (Fig. 2B and Table 2). H1 has been proposed as a site for β-sheet transformation that may promote PrP^Sc formation (21, 22) and mutational analysis of PrP in cell-free conversion assays highlight this helix as the initiation site for the conversion of PrP^C to the proteinase resistant form (23). The extensive contacts observed in this crystal structure would provide a significant stabilizing effect on the helix and would restrict its involvement in secondary-structure changes. Indeed, the average temperature factor for H1 (36 Ų) in the complex suggests that intermolecular contacts stabilize this region of PrP relative to the overall structure (40 Ų; see Fig. S3), but it is worthy of note that this region is well defined in the majority of PrP^C structures derived from a range of species (19, 24, 25). Conversely, the most disordered part of the molecule is at the C-terminal end of H2, the N-terminal end of H3, and the short loop connecting them (residues 188–201, average B-factor 56 Ų, Figs. S2 and S3). This disordered segment of the human PrP^C structure has also been suggested as a possible site for propagating α → β transitions that promote PrP^Sc formation (20, 25, 26). In our complex, part of this segment, extending from Asn-197 to Met-205, creates an interface with the H chain of Fab that buries 140 Ų of the PrP surface and 200 Ų of the Fab surface. The physical proximity of the Fab H chain to a region of PrP that is a possible site for β-sheet formation suggests that the complex may provide structural inhibition by burying the “active” residues at this interface. Structural changes to the Fab molecule upon PrP binding are shown in Fig. S4.

Table 2.

Summary of close contacts between PrP and ICSM 18-Fab in the crystal structure

PrP	FAB18 light chain	Distance, Å	PrP	FAB18 heavy chain	Distance, Å
Ser143 N	Tyr31 OH	3.4	Tyr145 OH	Asp31 OD2	3.4
Asp144 N	Asp49 OD1	3.2	Arg148 NH2	Asp35 OD1	2.7
Asp144 OD2	His33 NE2	3.3	Arg148 NH1	Asp35 OD2	3.2
Arg151 NH2	Tyr95 OH	3.1	Glu152 OE1	Tyr52 OH	2.7
Arg151 NH1	Arg91 O	3.2	Glu152 OE2	Asn50 OD1	3.3
Arg151 NH2	Trp90 O	2.9	Lys204 NZ	Tyr101 OH	3.6

Discussion

Many anti-PrP monoclonal antibodies are therapeutically active in cellular models of prion propagation (27–35) and in animals (18, 36, 37). Our results indicate a clear correlation between ability to inhibit PrP^Sc propagation and binding affinity for a PrP^C-type conformation for therapeutic antibodies. This is consistent with observations that antibody efficacy is determined by cell-surface PrP recognition (32) and retention (31) but rather contrary to interpretations that efficacy is primarily determined by antibody recognition epitopes (28) or their ability to recognize PrP^Sc as well as PrP^C (33). Our most potent antibody in both prion inhibition and PrP^C affinity, ICSM 18, is also highly effective in treating prion-infected mice (18), and we have no evidence that the interaction has any distorting effect on the PrP^C conformation as has been observed in some other cases (38, 39). The therapeutic value of targeting PrP^C is further underlined by conditional knockout experiments showing that its abrogation has little detrimental effect on the adult animal (40).

Prion propagation requires contact between PrP molecules, and in this respect, it is interesting to note that symmetry-related PrP molecules in the crystal interact by forming a 4-stranded antiparallel β-sheet (Fig. 3A) made up of the existing 2-stranded β-sheets of each PrP molecule, residues 129–131 and 161–163. An identical interaction was observed in the crystal structure of sheep PrP obtained in the absence of antibody and in a different crystallographic space group (25), which is shown overlaid with our human PrP structure in Fig. 3B. The coincidence of β-sheet structure between PrP molecules in different crystal structures suggests that it is not an artifact of crystal packing but might have greater biological significance for prion propagation, especially in view of the powerful genetic susceptibility and prion strain selection determined by polymorphic residue 129 in human prion disease (41). The common human methionine–valine polymorphism at residue 129 has a profound influence on prion pathogenesis; for instance, heterozygosity at PRNP codon 129 is highly protective against sporadic and acquired prion diseases in humans (42–44), delays age of onset of some inherited prion diseases (45), and appears to have been selected during the evolution of modern humans (46). Residue 129 also plays a key role in conformational selection of prion strains (41, 47). The bovine spongioform encephalopathy (BSE)-like strain causing vCJD appears incompatible with the valine isoform of human PrP (48); to date vCJD has occurred exclusively in individuals homozygous for methionine at PRNP codon 129 (49, 50). Moreover, the controlling influence of this residue is not limited to humans; polymorphism at the corresponding codon protects against chronic wasting disease, a prion disease of cervids (51, 52).

Fig. 3.

The interaction of PrP chains in the crystal. (A) Illustration of the intermolecular 4-stranded antiparallel β-sheet formed between neighboring PrP chains (in cyan and green) emphasizing residue 129 at the molecular interface (see Inset). (B) Superimposition of the ovine [red (25)] and human (green) PrP dimers from the respective crystal structures. Note the common occurrence of the 4-stranded intermolecular β-sheet.

Interestingly, the stability and physical properties of PrP^C are not influenced by the residue 129 variation (53), so its effect must be realized in the physical properties and propagation of PrP^Sc. The only compelling explanation for the effects of this polymorphism is that the 129 region of the molecule is critical in determining the fold and the packing of protein chains in prion particles. Heterozygosity is protective owing to the requirement for sequence homogeneity in forming ordered, self-replicating particles (43) (analogous to protein crystallization), and prion strain susceptibility arises from the ease by which the methionine or the valine sequence adopts a given prion structure. The presence—in 2 completely different PrP crystal structures—of the same intermolecular 4-stranded sheet supports our proposal that such a structure is present in PrP^Sc and is propagated and stabilized by stacking to form the type of antiparallel zipper interactions that have been experimentally shown by Eisenberg and colleagues (54) to form cross-β amyloid spines. However, we point out that the formation of this β-sheet interface is not the only event necessary for PrP^Sc formation; it must also be coupled to gross rearrangement of the structure of component monomers (see Fig. S5). In this regard, it should be noted that antiparallel interactions are found in fibrils formed from short peptides (54) rather than those formed from proteins that, thus far, have been shown to be parallel (55). However, in some denaturing conditions large proteins have been seen to aggregate initially by antiparallel interactions that slowly rearrange to parallel on maturation (56). Nonetheless, if β-strand interactions in this region mediate PrP interactions during PrP^Sc propagation, then it is reasonable to propose that the exact geometry and packing of this segment of the chain varies between strains and that the nature of the 129 side chain has a strong selective effect on conformation. Such a mechanism would explain the potent effects of this human polymorphism at the level of fine structure within a prion particle.

Materials and Methods

Recombinant PrP and Antibodies.

Recombinant human PrP containing residues 91–231 (PrP^91-231) or 119–231 (PrP^119-231) was produced and purified as described (12). The ICSM antibodies were raised to recombinant PrP in either α- or β-conformation (12, 13), as listed in Table S1. The antibodies were characterized as previously described (14, 15) and were the gift of D-Gen Limited.

Determination of Antibody–PrP Affinities by ELISA.

Recombinant α-PrP^91-231 [2.5 μg/mL in bicarbonate buffer (pH 9.6)] was immobilized in 96-well ELISA plates. After washes and blocking with RF10/Tween-20, the plates were incubated with serial dilutions (0.1 mM to 10 μM) of the ICSM antibodies, HRP-conjugated goat anti-mouse secondary antibody (Sigma) and substrate TMB (Sigma). The color reaction was stopped by addition of 3 M sulfuric acid before reading intensities at 450 nm. Intensities were averaged across replicates (n = 3) and fitted to a single site binding curve to derive K_ds by using Grafit5 (Erithacus Software Ltd); see Fig. S1A.

Determination of IC₅₀s for ICSM-Treated Cells.

Scrapie susceptible N2a cells (subclone “PK1”) (57) were cultured in OPTIMEM (Invitrogen) supplemented with 10% FCS (Perbio) and penicillin/streptomycin (Invitrogen). Cells chronically infected with RML prions were cultivated in 24-well plates with 1 mL per well media supplemented with ICSM antibodies at various concentrations for 3 days before assay for levels of proteinase K-resistant PrP. Briefly, cells were lysed in ice-cold lysis buffer, blotted onto nitrocellulose membrane in a BioRad ELIFA apparatus, treated in situ with proteinase K (10 μg/mL for 1 h at 37 °C) and denaturant (3 M guadinidium thiocyanate) before immunodetection using ICSM 18 (0.2 μg/mL), HRP-conjugated anti-mouse secondary antibody (Sigma), Super Signal West Pico ECL reagent (Pierce), and exposure on X-ray film (Kodak Biomax). PK-resistant signal intensities were quantified by using a Kodak ImageStation 440 CF, and data were averaged across replicates (n = 4) and normalized against 100% intensity in untreated control samples. Half-maximal inhibition constants (IC₅₀) were derived by fitting of the averaged and normalized data by using Grafit5 (Erithacus Software Ltd); see Fig. S1B.

Flow Cytometry.

Suspensions of NS0 cells (2 × 10⁶ cells per milliliter) were incubated with 1–10 μg/mL ICSM antibodies, washed in PBS, and resuspended in FITC-conjugated secondary antibody (Sigma) in PBS. Cells were further washed and resuspended in PBS for FACS measurements by using a FACSCalibur (Becton Dickinson).

Crystallization, Data Collection, Refinement, and Model Building.

The Fab fragment of ICSM 18 was prepared by limited papain digest of the mature antibody and purification by gel filtration (58). Before crystallization, ICSM 18-Fab was additionally purified on a S200 gel filtration column by using 150 mM NaCl and 50 mM Tris (pH 8.0) and concentrated to 100 mg/mL. Crystals of ICSM 18-Fab were obtained by using the sitting-drop vapor diffusion technique; droplets containing 20 mg/mL protein in 0.1 M magnesium chloride, 0.05 M Tris·HCl (pH 8.0), 10% wt/vol PEG3350 were equilibrated over wells containing 0.2 M magnesium chloride, 0.1 M Tris·HCl (pH 8.0), 20% wt/vol PEG3350. Crystals grew in 1 day to 0.2 × 0.05 × 0.05 mm. ICSM 18-Fab and PrP^119-231 were mixed at a molar ratio of 1: 1 for preparation of the complex before crystallization. Crystals of the complex were obtained by using the hanging-drop vapor diffusion technique; droplets containing 20 mg/mL protein in 1.0 M ammonium sulfate, 0.05 M Tris (pH 8.0) were equilibrated over wells containing 2.0 M ammonium sulfate, 0.1 M Tris (pH 8.0). Crystals grew in 5 months to 0.1 × 0.05 × 0.05 mm.

Crystallographic Data Collection, Refinement, and Model Building.

Data were collected at 100 K with a MAR225 CCD detector by using beamline 10.1 at the Daresbury Laboratory Synchrotron Radiation Source, U.K. ICSM 18-Fab crystals were picked directly from the hanging drop for data collection, whereas for the ICSM 18-Fab–PrP complex, 3.4 M sodium malonate (pH 7.0) was used as a cryo protector. Data were processed with the program HKL2000 (59). Data collection statistics and cell parameters are summarized in Table 1. Structures were solved by the molecular replacement method by using the programs PHASER (60) for the complex and MOLREP (61) for ICSM 18 (see SI Text for ICSM 18 structure). An anti-PrP Fab structure (PDB ID: 2hh0) (62), was used as a starting model for ICSM 18-Fab. The variable and constant regions of the heavy and light chains were treated as independent rigid bodies before restrained refinement and model building cycles with addition of solvent. Subsequently, the refined structure of ICSM 18-Fab was used with the globular domain of sheep PrP (PDB ID: 1tpx) (21) as the starting point for solving the structure of the complex. All models were refined without NCS restrains by the maximum-likelihood method as implemented in REFMAC5 (63). Overall temperature factor refinement was used for the complex, and TLS refinement (64) was applied to both models, with each domain treated as a separate group, excluding the water molecules. This resulted in a drop in both R and R-free factors of nearly 5%. Model building was done with COOT (65). Solvent molecules were added to the model by using ARP (66), being accepted only when well-defined positive peaks were present in both 2F_o-F_c and F_o-F_c electron density maps and when they could form hydrogen bond(s) with either protein atoms or other water molecules. A summary of the structure refinement is given in Table 1. The final coordinates have been deposited in the Brookhaven Protein Data Bank.