WO2011083099A1

WO2011083099A1 - Crystals of mutant bace - 1

Info

Publication number: WO2011083099A1
Application number: PCT/EP2011/050004
Authority: WO
Inventors: David Banner; Joerg Benz; Armin Ruf; Daniel Schlatter; Martine Stihle
Original assignee: F. Hoffmann-La Roche Ag
Priority date: 2010-01-06
Filing date: 2011-01-03
Publication date: 2011-07-14

Abstract

The present invention relates to crystals of human β-secretase in apo-form and with an inhibitor, and to the three-dimensional X-ray crystal structure derived thereof.

Description

CRYSTALS OF MUTANT BACE - 1

The present invention relates to crystals of human β-secretase (BACE-1) in apo-form and with an inhibitor, and to the three-dimensional X-ray crystal structure derived thereof.

The most prominent neuropatho logical finding in Alzheimer's disease (AD) brains is the occurrence of cortical plaques containing beta-amyloid peptide. Genetic evidence points to a causal role of beta-amyloid during development of AD as it is postulated by the amyloid cascade hypothesis. Several independent approaches led to the identification of BACE-1 as the first enzyme of the amyloid cascade. BACE-1 cleaves the 695 or 751 aa amyloid precursor protein (APP) to a 99 aa long precursor peptide which is later trimmed by gamma-secretase to form the pathological 40 or 42 aa long beta-amyloid peptide. BACE-1 is an integral membrane protein with high homology to other aspartic proteases. The enzymatic activity does not depend on the presence of the transmembrane or intracellular domains. Given the lack of adverse effects in gene knock out mice BACE-1 is considered a prime target for the development of AD therapeutics. BACE-1 possesses a bilobal structure typical for eukaryotic aspartic proteases with the catalytical aspartate residues Asp93 and Asp289 located in the substrate binding cleft between the N-terminal and C-terminal lobes. The active site is partially covered by a flexible hairpin loop referred to as the "flap". In apo structures of BACE-1, the flap backbone is at a distance of about 12A from the side chains of Asp93 and Asp289 which coordinate the catalytic water molecule whereas in complexes with peptidomimetic inhibitors the flap moves about 4A closer to the catalytic center and the non-cleavable transition state isostere moiety of the inhibitors replaces the catalytic water molecule.

Assay conditions in the determination of ligand binding to a protein and crystallization conditions are rarely identical. With respect to high occupancy of the ligand in the crystal it might be very useful to bring both systems in agreement. Ligand binding to BACE-1 is typically measured at low pH and therefore we were searching for a new and soakable crystal form at low pH.

Therefore, there is a need for BACEl polypeptides allowing successful crystallisation and/or co-crystallisation and subsequent X-ray crystallograpy analysis.

In a first object, the present invention provides an isolated BACE-1 polypeptide comprising a BACEl amino acid sequence including amino acid 307 of Seq. Id. No. 1, wherein the amino acid at position 307 of Seq. Id. No. 1 is alanine. ferred embodiment the BACE-1 polypeptide comprises aa 14 - 454 of Seq. Id. No.

2.

Seq. Id. No. 1 shows the amino acid sequence of human BACEl protein and Seq. Id. No. 2 shows the amino acid sequence of human BACEl K307A protein i.e. lysine at position 307 of wildtype human BACE-1 polypeptide having Seq. Id. No. 1 has been replaced by alanine.

In a third object, the present invention provides a nucleic acid molecule encoding a BACE-1 polypeptide of the present invention.

In a fourth object the present invention provides a crystal of a BACE-1 polypeptide of the present invention, wherein the crystal has unit cell dimensions of a = 133 ± 3 A, b = 131 ± 3 A, c = 118 ± 3 A, β = 91.8° ± 1 ° and the crystal belongs to space group C2.

In a fifth object, the present invention provides a co-crystal of a BACE-1 polypeptide of the present invention and a ligand bound to BACE-1, wherein the crystal has unit cell dimensions of a = 134 ± 3 A, b = 131 ± 3 A, c = 118 ± 3 A, β = 91.8° ± 1° and the crystal belongs to space group C2. In a preferred embodiment of the co-crystal the ligand is [2-(4-tert-Butyl- phenylcarbamoyl)-propyl]-[l-(3-dipropylcarbamoyl-benzoylamino)-2-phenyl-ethyl]-phosphinic acid.

In a further object, the present invention provides a method for crystallizing or co- crystallizing a BACE-1 polypeptide of the present invention with a compound that binds to the binding site of said polypeptide, the method comprising:

a) providing an aqueous solution of the polypeptide,

b) optionally adding a molar excess of a ligand to the aqueous solution of the polypeptide, and

c) growing crystals. In another object, the present invention provides a method for identifying a compound that can bind to the binding site of a BACE-1 polypeptide comprising the steps:

a) determining an active site of BACE-1 from a three dimensional model of BACE-1 using the atomic coordinates of Fig. 2 or 3 ± a root mean square deviation from the backbone atoms of said amino acids of not more than 2 A; and

b) performing computer fitting analysis to identify a compound that can bind to the BACE-1 active site.

In a preferred embodiment, the method comprises the steps:

a) generating a three dimensional model of an active site of BACE-1 using the relative t coordinates of Fig. 2 or 3 of residues Gly72, Gln73, Gly74, Leu91, Asp93, Gly95, Ser96, Tyrl29, Vall30, Prol31, Tyrl32, Thrl33, Glnl34, Phel69, Ilel71, Trpl76, Ilel79, Ilel87, Alal88, Argl89, Tyr259, Ile287, Asp289, Ser290, Gly291, Thr292, Asn294, Arg296., ± a root mean square deviation from the backbone atoms of said amino acids of not more than 2 A; and

Crystals of the present invention can be grown by a number of techniques including batch crystallization, vapor diffusion (either by sitting drop or hanging drop) and by micro dialysis. Seeding of the crystals in some instances is required to obtain X-ray quality crystals. Standard micro- and/or macroseeding of crystals may therefore be used.

In a preferred embodiment of the invention, crystals and/or co-crystals are grown by vapor diffusion. In this method, the polypeptide solution is allowed to equilibrate in a closed container with a larger aqueous reservoir having a precipitant concentration optimal for producing crystals. Generally, less than about 10 μΐ, of substantially pure polypeptide solution is mixed with an equal or similar volume of reservoir solution, giving a precipitant concentration about half that required for crystallization. This solution is suspended as a droplet underneath a coverslip, which is sealed onto the top of a reservoir. The sealed container is allowed to stand, from one day to one year, usually for about 2-6 weeks, until crystals grow. The crystals and/or co-crystals of the present invention can be obtained by a method which comprises: providing a buffered, aqueous solution of 3.75 to 50 mg/ml of a polypeptide of the present invention, adding a molar excess of the ligand to the aqueous polypeptide solution, and growing crystals by vapor diffusion or microbatch using a buffered reservoir solution of 0 % to 30 % (w/v) PEG, wherein the PEG has an average molecular weight of 200 Da to 20000 Da. The PEG may be added as monomethyl ether. PEG may be used of an average molecular weight of 500 Da to 5,000 Da. The buffered reservoir solution further comprises 0 M to 2 M tri-ammonium citrate pH 7, 0 M to 1 M L-proline, 0 M to 1M trimethylamine-N-oxide, 0 M to 1 M ammonium sulfate, 0 M to 1 M lithium sulfate, 0 M to 1 M ammonium acetate, 0 M to 1 M sodium or magnesium formate and 0 M to 1 M D/L-malic acid pH 7. Said microbatch may be modified. Methods for obtaining the three-dimensional structure of the crystals described herein, as well as the atomic structure coordinates, are well-known in the art (see, e.g., D. E. McRee, Practical Protein Crystallography, published by Academic Press, San Diego (1993), and references cited therein).

The crystals of the invention, and particularly the atomic structure coordinates obtained therefrom, have a wide variety of uses. For example, the crystals and structure coordinates ein are particularly useful for identifying compounds that bind to BACE-1 as an approach towards developing new therapeutic agents.

The structure coordinates described herein can be used as phasing models in determining the crystal structures of additional native or mutated, as well as the structures of co-crystals of BACE-1 with bound ligand. The structure coordinates, as well as models of the three- dimensional structures obtained therefrom, can also be used to aid the elucidation of solution- based structures of native or mutated BACE-1, such as those obtained via NMR. Thus, the crystals and atomic structure coordinates of the invention provide a convenient means for elucidating the structures and functions of a BACE-1. The term "root mean square deviation" means the square root of the arithmetic mean of the squares of the deviations. It is a way to express the deviation or variation from a trend or object. For purposes of this invention, the "root mean square deviation" defines the variation in the backbone of a protein from the backbone of BACE-1 or an active binding site thereof, as defined by the structure coordinates of BACE-1 described herein. Molecular docking of large compound databases to target proteins of known or modeled 3- dimensional structure is now a common approach in the identification of new lead compounds. This "virtual screening" approach relies on fast and accurate estimation of the ligand binding mode and an estimate of ligand affinity. Typically a large database of compounds, either real or virtual is docked to a target structure and a list of the best potential ligands is produced. This ranking should be highly enriched for active compounds which may then be subject to further experimental validation.

The calculation of the ligand binding mode may be carried out by molecular docking programs which are able to dock the ligands in a flexible manner to a protein structure. The estimation of ligand affinity is typically carried out by the use of a separate scoring function. These scoring functions include energy-based approaches which calculate the molecular mechanics force field and rule-based approaches which use empirical rules derived from the analysis of a suitable database of structural information. Consensus scoring involves rescoring each ligand with multiple scoring functions and then using a combination of these rankings to generate a hit list.

As used herein, "apo crystal" refers to crystals of BACE-1 formed without a bound ligand. escription of the figures

Fig. 1A shows Electron density for [2-(4-tert-Butyl-phenylcarbamoyl)-propyl]-[l-(3- dipropyl carbamoyl-benzoylamino)-2-phenyl-ethyl]-phosphinic acid in complex with BACE-1. Final 2Fobs-Fcalc density contoured at 1.3 sigma superimposed onto the refined coordinates. Fig. IB: Electron density for [2-(4-tert-Butyl-phenylcarbamoyl)-propyl]-[l-(3-dipropyl carbamoyl-benzoylamino)-2-phenyl-ethyl]-phosphinic acid in complex with BACE-1. Initial difference density calculated from the initial molecular replacement model without any inhibitor bound and contoured at +/- 3.5 sigma. It shows strong negative and positive features indicating the movement of flap residues 128-138 (top) and positive density demonstrating the binding of the inhibitor.

Fig. 2 shows the coordinates of an apo-crystal of human BACE1 (amino acids 14 - 454 of Seq. Id. No. 2); the apo-crystal unit contains three BACE-1 molecules; the coordinates of amino acids 57 - 447 of Seq. Id. No. 2 are shown for the first BACE-1 molecule, the coordinates of amino acids 57 - 446 of Seq. Id. No. 2 are shown for the second BACE-1 molecule and the coordinates of amino acids 59 - 446 of Seq. Id. No. 2 are shown for the third BACE-1 molecule.

Fig. 3 shows the coordinates of a co-crystal of human human BACE1 (amino acids 14 - 454 of Seq. Id. No. 2) with ligand [2-(4-tert-Butyl-phenylcarbamoyl)-propyl]-[l-(3-dipropyl- carbamoyl-benzoylamino)-2-phenyl-ethyl]-phosphinic acid; the co-crystal unit contains three BACE-1 molecules; the coordinates of amino acids 57 - 447 of Seq. Id. No. 2 are shown for the first BACE-1 molecule; the coordinates of amino acids 57 - 446 of Seq. Id. No. 2 are shown for the second BACE-1 molecule and the coordinates of amino acids 59 - 446 of Seq. Id. No. 2 are shown for the third BACE-1 molecule.

Examples

Example 1 : Monoclinic BACE-1 Crystal: Apo Structure Protein production of B ACE- 1

Wild type BACE-1 ectodomain was cloned from Ala 14 to Thr454 (aa 14 - 454 of Seq. Id. No. 1) into the pETl la vector according to Hong [Lin Hong et al, Science 290, 150-153 (2000).

In search for new crystals forms protein engineering by site directed mutagenesis of surface residues was applied. Surface residues are involved in crystal packing and mutations can lead to new crystal forms. Out of eleven different mutations only the Lys 307→ Ala mutant protein delivered significantly better crystals. The protein was expressed in BL21 (DE3) cells. Since the protein was expressed forming insoluble inclusion bodies it had to be renatured. An lding and protein purification procedure was elaborated. Isolated and purified inclusion bodies were dissolved in a 0.05 M Tris/HCl buffer (pH 9.0) containing 8 M urea and 0.01 M β-mercaptoethanol. Protein refolding was initiated by a reduction in denaturant concentration and by altering the redox environment to enable disulphide bond formation. This was done by a dilution of the protein solution into a large volume of refolding buffer containing 0.02 M Tris/HCl pH 9.0, 0.001 M glutathione reduced, 0.0025 M glutathione oxidized. A few hours later the pH of the solution was lowered to pH 8.0 and the preparation kept at 4°C for at least two days. For BACE-1 with its low pi of 4.7 the most efficient procedure for concentrating the diluted refolding solution was an anion exchange chromatography using Q Sepharose Fast Flow resin. For this chromatography 0.05 M Tris/HCl pH 6.8, 0.4 M urea was used and therefore the pH of the solution was further lowered to pH 6.8 A gradient elution using 1 M NaCl resolved aggregated and correctly refolded protein which was further purified by a second anion exchange chromatography of the same type in 0.05 M Tris/HCl pH 6.8. Then the enzyme was activated by cleaving off the signal peptide with furin. Finally, the protein was chromatographed on a Superdex 75 column in 0.05 M Tris/HCl pH 7.2, 100 mM NaCl. As shown by SDS-PAGE >98% pure protein was obtained. It was monomeric as detected by analytical ultracentrifugation and highly active in the FRET assay.

Crystallization:

Protein BACE-1 K307A used for crystallization attempts has been purified as described above. Prior to crystallization the protein was transferred by ultrafiltration into a buffer containing 0.02 M Sodium Acetate pH 4.5, 0.15 M NaCl, 0.003 M EDTA. The protein was then concentrated to 20 mg/ml followed by centrifugation at 20000 x g for 10 minutes. Screening was performed against non buffered precipitant solutions. Crystallization droplets were set up at 22 °C by mixing 0.25 μΐ of protein solution with 0.1 μΐ reservoir solution and 0.06 μΐ crystal seed solution in vapour diffusion sitting drop experiments. Seeding was necessary to obtain crystals with reproducible size. Crystals appeared out of 0.1 M Sodium Acetate, 20 % PEG 1500 after 1 day and grew to a final size of 0.2 mm x 0.06 mm x 0.02 mm within 5 days.

Crystals were harvested with 20% Glycerol as cryoprotectant and then flash frozen in liquid N2. Diffraction images were collected at a temperature of 100K at the beam line XI OS A of the Swiss Light Source and processed with the programs XDS [Kabsch, W. Automatic processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J. Appl. Cryst. 26, 795-800 (1993)] and scaled with SADABS [obtained from BPvUKEPv AXS], yielding data to 2.2A resolution. This BACE-1 crystal belongs to the face- centered monoclinic space group C2 with cell dimensions of a=133.3 lA b=131.60A c=118.25A and β=91.83° and contains three BACE-1 molecules per crystallographic asymmetric unit (see Table 1). It is not isomorphous to any crystal of the human BACE-1 in the Protein Data Bank. d crystallographic programs from the CCP4 software suite were used to solve the structure by molecular replacement with hexagonal human BACE-1 as search model (PDB entry 3BRA or similar), to calculate the electron density, and to refine the x-ray structure [CCP4 (Collaborative Computational Project, N. The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D, 760-763 (1994)]. The structural models were rebuilt into the electron density using COOT (Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126-32 (2004)). Coordinates were refined with REFMAC5 (Murshudov, G.N., Vagin, A.A. & Dodson, E.J. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240-55 (1997)) and with autoBUSTER (Global Phasing Ltd.).

Table 1 : Data collection and structure refinement statistics for monoclinic BACE-1 apo- crystal

Data Collection

Wavelength (A) 1.0

Resolution¹ (A) 2.25 (2.35-2.25)

Unique reflections¹ 96570 (11776)

Completeness (% 99.9 (99.9)

R ί°/ Λ¹ '²

Emerge ' °) 9.03 (53.7)

<Ι/σ>¹ 10.8 (1.96)

Unit Cell (Space group C2) a=133.3 lA b=131.60A c=118.25A and β=91.83°

Refinement

Resolution (A) 2.25 (2.37-2.25)

p 1 ,3

J^cryst 20.74 (44.5))

Rfree 23.75 (47.8)

Number of Atoms in 9681

refinement

R.m.s. deviations from ideality 0.011 / 1.26

Bond lengths (A) / angles (°)

Main chain dihedral angles (%) 88.2 / 11.7 / 0.1 / 0.0

Most

favored/ allowed/ generous/

disallowed ⁵ ¹ Values in parentheses refer to the highest resolution bins.

2 I !-<!> I /∑! where I is intensity. ^■<F_C> | /∑F₀ where F₀ is the observed and F_c is the calculated structure factor amplitude.

⁴ Rfree was calculated based on 5% of the total data omited during refinement.

⁵ Calculated with PROCHECK [Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. PROCHECK: a program to check the stereochemical quality of protein structure. J. Appl.

Crystallogr. 26, 283-291 (1993)].

Example 2: Monoclinic BACE-1 Crystal: Complex Structure with [2-(4-tert-Butyl- phenylcarbamoyl)-propyl]-[l-(3-dipropylcarbamoyl-benzoylamino)-2-phenyl-ethyl]- phosphinic acid

Protein BACE-1 K307A used for crystallization attempts has been purified as described above. Prior to crystallization the protein was transferred by ultrafiltration into a buffer containing 0.02 M Sodium Acetate pH 4.5, 0.15 M NaCl, 0.003 M EDTA. The protein was then concentrated to 20 mg/ml and the protein solution centrifuged at 20000 x g for 10 minutes. The crystallization droplets were set up at 22 °C by mixing 0.25 μΐ of protein solution with 0.1 μΐ reservoir solution and 0.06 μΐ crystal seed solution in vapour diffusion sitting drop experiments. Seeding was necessary to obtain crystals with reproducible size. Crystals appeared out of 0.2 M Ammonium Citrate, 20 % PEG 3350 after 1 day and grew to final size of 0.2 mm x 0.06 mm x 0.02 mm within 5 days. For complex formation the crystals were soaked for one day in crystallization solution supplemented with 0.005 M ligand.

Crystals were harvested with 20% Glycerol as cryoprotectant and then flash frozen in liquid N₂. Diffraction images were collected at a temperature of 100K at the beam line XI OS A of the Swiss Light Source and processed with the programs XDS [Kabsch, W. Automatic processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J. Appl. Cryst. 26, 795-800 (1993)] and scaled with SADABS [obtained from BRUKER AXS], yielding data to 2.22A resolution. This BACE-1 crystal belongs to the face- centered monoclinic space group C2 with cell dimensions of a=134.0lA b=130.70A c=118.82A and β=91.78° and contains three BACE-1 molecules per crystallographic asymmetric unit (see Table 1). It is not isomorphous to any crystal of the human BACE-1 in the Protein Data Bank.

Standard crystallographic programs from the CCP4 software suite were used to solve the structure by molecular replacement with hexagonal human BACE-1 as search model (PDB entry 3BRA or similar), to calculate the electron density, and to refine the x-ray structure [CCP4 (Collaborative Computational Project, N. The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D, 760-763 (1994)]. The structural models were rebuilt and two inhibitor re added into the electron density using COOT (Ems ley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126-32 (2004)). In subunit A no electron density for the inhibitor was observed and consequently no inhibitor was added to this protein subunit. Coordinates were refined with REFMAC5 (Murshudov, G.N., Vagin, A.A. & Dodson, E.J. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240-55 (1997)) and with autoBUSTER (Global Phasing Ltd.).

The binding pocket of the ligand is defined by residues Gly72, Gln73, Gly74, Leu91, Asp93, Gly95, Ser96, Tyrl29, Vall30, Prol31, Tyrl32, Thrl33, Glnl34, Phel69, Ilel71, Trpl76, Ilel79, Ilel87, Alal88, Argl89, Tyr259, Ile287, Asp289, Ser290, Gly291, Thr292, Asn294, Arg296.

Table 2: Data collection and refinement statistics for monoclinic BACE-1 [2-(4-tert- Butyl-phenylcarbamoyD-propyl] - [ 1 -(3 -dipropylcarbamoyl-benzoylamino)-2-phenyl-ethyl] - phosphinic acid co-crystal

Data Collection

Wavelength (A) 1.0

Resolution¹ (A) 2.21 (2.30-2.21)

Unique reflections¹ 101038 (10544)

Completeness (% 98.9 (90.8)

R ί°/ Λ¹ '²

Emerge ' °) 14.8 (56.95)

<Ι/σ>¹ 13.3 (2.9)

Unit Cell (Space group C2) a=134.0lA b=130.70A c=l 18.82A and β=91.78°

Refinement

Resolution (A) 2.22 (2.37-2.25)

J^cryst 18.6 (25.2))

Rfree 21.9 (28.2)

Number of Atoms in 9823

refinement

R.m.s. deviations from ideality 0.011 / 1.26

Bond lengths (A) / angles (°)

Main chain dihedral angles (%) 88.4 / 11.6 / 0.0 / 0.0

Most

favored/ allowed/ generous/

disallowed ⁵ rentheses refer to the highest resolution bins.

² I I-<I> I /∑I where I is intensity.

³

I F₀-<F_C> I /∑F₀ where F₀ is the observed and F_c is the calculated structure factor amplitude.

4 Rfree was calculated based on 5% of the total data omited during refinement.

⁵ Calculated with PROCHECK [Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. PROCHECK: a program to check the stereochemical quality of protein structure. J. Appl. Crystallogr. 26, 283-291 (1993)].

While there are shown and described presently preferred embodiments of the invention, it is to be distinctly understood that the invention is not limited thereto but may be otherwise variously embodied and practiced within the scope of the following claims.

Claims

1. An isolated BACE-1 polypeptide comprising a BACE1 amino acid sequence including amino acid 307 of Seq. Id. No. 1, wherein the amino acid at position 307 of Seq. Id. No. 1 is alanine.

2. The BACE-1 polypeptide of claim 1 comprising aa 14 - 454 of Seq. Id. No. 2.

3. A nucleic acid molecule encoding a polypeptide of claim 1 or 2.

4. A crystal of a polypeptide of claim 1 or 2, wherein the crystal has unit cell dimensions of a = 133 ± 3 A, b = 131 ± 3 A, c = 118 ± 3 A, β = 91.8° ± 1° and the crystal belongs to space group C2.

5. A co-crystal of a polypeptide of claims 1 or 2 and a ligand bound to BACE-1, wherein the crystal has unit cell dimensions of a = 134 ± 3 A, b = 131 ± 3 A, c = 118 ± 3 A, β = 91.8° ± 1° and the crystal belongs to space group C2.

6. The co-crystal of claim 5, wherein the ligand is [2-(4-tert-Butyl-phenylcarbamoyl)- propyl]-[ 1 -(3-dipropylcarbamoyl-benzoylamino)-2-phenyl-ethyl]-phosphinic acid.

7. A method for crystallizing or co-crystallizing a polypeptide of claim 1 or 2 with a compound that binds to the binding site of said polypeptide, the method comprising:

a) providing an aqueous solution of the polypeptide,

c) growing crystals.

8. A crystal obtained by the method of claim 7.

9. A method for identifying a compound that can bind to the binding site of BACE-1 comprising the steps:

10. The method of claim 10 comprising the steps:

a) generating a three dimensional model of an active site of BACE-1 using the relative t coordinates of Fig. 1 of residues Gly72, Gln73, Gly74, Leu91, Asp93, Gly95, Ser96, Tyrl29, Vall30, Prol31, Tyrl32, Thrl33, Glnl34, Phel69, Ilel71, Trpl76, Ilel79, Ilel87, Alal88, Argl89, Tyr259, Ile287, Asp289, Ser290, Gly291, Thr292, Asn294, Arg296., ± a root mean square deviation from the backbone atoms of said amino acids of not more than 2 A; and

11. A crystal or co-crystal of a polypeptide of claim 1 or 2 and optionally a ligand bound to the BACE-1 having the structure defined by the coordinates of Fig. 2, Fig. 3 optionally varied by an rmsd of less than 2.0 A.

12. The polypeptides, crystals and methods substantially as hereinbefore described, especially with reference to the foregoing examples.