Structure of the Protease Domain of Memapsin 2 (β-Secretase) Complexed with Inhibitor

See allHide authors and affiliations

Science  06 Oct 2000:
Vol. 290, Issue 5489, pp. 150-153
DOI: 10.1126/science.290.5489.150


Memapsin 2 (β-secretase) is a membrane-associated aspartic protease involved in the production of β-amyloid peptide in Alzheimer's disease and is a major target for drug design. We determined the crystal structure of the protease domain of human memapsin 2 complexed to an eight-residue inhibitor at 1.9 angstrom resolution. The active site of memapsin 2 is more open and less hydrophobic than that of other human aspartic proteases. The subsite locations from S4 to S2′ are well defined. A kink of the inhibitor chain at P2′ and the change of chain direction of P3′ and P4′ may be mimicked to provide inhibitor selectivity.

The accumulation of the 40- to 42-residue β-amyloid peptide (Aβ) in the brain is a key event in the pathogenesis of Alzheimer's disease (AD) (1). Aβ is generated in vivo through proteolytic cleavage of the membrane-anchored β-amyloid precursor protein (APP) by β- and γ-secretases. The γ-secretase activity, which cleaves APP within its transmembrane domain, is likely mediated by the transmembrane protein presenilin 1 (2–4). The β-secretase cleaves APP on the lumenal side of the membrane and its activity is the rate-limiting step of Aβ production in vivo (5). Both proteases are potential targets for inhibitor drugs against AD. Our group (6) and others (7) recently cloned a human brain aspartic protease, memapsin 2 or BACE, and demonstrated it to be β-secretase. Memapsin 2 is a class I transmembrane protein consisting of an NH2-terminal protease domain, a connecting strand, a transmembrane region, and a cytosolic domain (6, 7). Sequence homology with other aspartic proteases suggests that memapsin 2 has a pro sequence of about 48 residues at its NH2-terminal region. The protease domain of pro-memapsin 2, expressed recombinantly, hydrolyzes peptides from the APP β-secretase site but has a broad specificity (6). We have used this specificity information to design potent inhibitors against this enzyme (8). OM99-2, an eight-residue transition-state inhibitor (Fig. 1) has a K iof 1.6 nM for memapsin 2. To develop memapsin 2 inhibitors with therapeutic potential would require, besides good potency and pharmacokinetic properties, low molecular weight (<700 daltons) and high lipophilicity in order to penetrate the blood-brain barrier (9). We determined the three-dimensional structure of the memapsin 2 with an active site–bound OM99-2 at 1.9 Å resolution in order to define a template for the rational design of memapsin 2 inhibitor drugs.

Figure 1

The chemical structure of memapsin 2 inhibitor OM99-2 with the constituent amino acids and their subsite designations. The hydroxyethylene transition-state isostere is between P1-Leu and P1′-Ala.

Fully active recombinant memapsin 2, which contains 21 residues of the putative pro region, residues numbers 28p–48p (10) but without the transmembrane and intracellular domains (11), was crystallized as a complex with OM99-2 (12). We report here a crystal structure of this complex at 1.9 Å resolution. The crystal structure was determined using molecular replacement methods (12) with human pepsin (22% sequence identity) as the search model. The statistical data are shown in Table 1.

Table 1

Data collection and refinement statistics.

View this table:

The bilobal structure of memapsin 2 (Fig. 2A) has the conserved general folding of aspartic proteases (13). The inhibitor is located in the substrate binding cleft between the NH2- and COOH-terminal lobes (10) (Fig. 2A). Active-site Asp32 and Asp228 and the surrounding hydrogen bond network are located in the center of the cleft (Fig. 2, A and B) and are conserved (14). The hairpin loop known as the “flap” (10) partially covers the cleft. The active-site carboxyls are, however, not co-planar, and the degree of deviation (50°) exceeds those observed previously. Whether this is specific for OM99-2 binding has not been determined.

Figure 2

The crystal structure of memapsin 2 complexed to inhibitor OM99-2. (A) Stereo view of the polypeptide backbone of memapsin 2 is shown as a ribbon diagram. The N-lobe and C-lobe are blue and yellow, respectively, except the insertion loops, designated A to G (10) in the C-lobe are magenta and the COOH-terminal extension is green. The inhibitor bound between the lobes is shown in red. (B) The chain tracing of human memapsin 2 (dark blue) and human pepsin (gray) is compared. The light blue balls represent identical residues which are topologically equivalent. The disulfide bonds are shown in red for memapsin 2 and orange for pepsin. The COOH-terminal extension is in green. The active-site aspartic acids are shown in yellow.

Compared to pepsin (15), the most significant structural differences are the insertions and a COOH-terminal extension in the C-lobe. Four insertions, A, C, D, and F (10) (Fig. 2, A and B), as helices and loops are located on the adjacent molecular surface near the NH2-terminus of the inhibitor. Insertion F (10), which contains four acidic residues, forms the most negatively charged region on the molecular surface. Together, these insertions significantly enlarge the molecular boundary of memapsin 2 as compared to pepsin (Fig. 2B). These surface structural changes may function in the association of memapsin 2 with other cell-surface components. Insertions B and E are located on the molecular surface near the COOH-terminus of the inhibitor. Loop E is connected to a β-strand that is paired with part of the COOH-terminal extension. The active-site cleft of memapsin 2 is, in general, more open and accessible than that of pepsin, owing to structural differences near respective subsites P4, P2, and P1′ (see below) and the absence of six pepsin residues [P292TESGE297 (16)] at memapsin 2 residues Thr329/Gly330on a loop opposite the flap across the active-site cleft. The 35-residue COOH-terminal extension (10) unique to memapsin 2 consists mostly of highly ordered structure (residues 359–385). Residues 369–376 form a β structure with seven hydrogen bonds to strand 293–299, whereas residues 378–383 form a helix (Fig. 2A). Two of the three disulfide pairs (residues 155 and 359 and 217 and 382) unique to memapsin 2 fasten both ends of the extension region to the C-lobe. This COOH-terminal extension is longer than those observed previously for aspartic proteases and is conformationally quite different (17–20). The last eight residues (386–393) are not seen in the electron density map. Their mobility suggests the possibility of forming a short stem between the globular catalytic domain and the trans-membrane domain. Of the 21 putative pro residues present in the enzyme (10), only the last six, 43p–48p, are visible in the electron density map. The others are likely mobile, which is consistent with an unstructured pro segment being displaced from the active-site cleft by the inhibitor (21).

The interactions of the eight-residue inhibitor OM99-2 with memapsin 2 include four hydrogen bonds between two active-site aspartates and the hydroxyl of the transition-state isostere, and ten hydrogen bonds from different parts of the binding cleft and flap to inhibitor backbone (Fig. 3). Most of these hydrogen bonds are highly conserved among eukaryotic (14, 22, 23) and HIV (24) aspartic proteases, except hydrogen bonds to Gly11 and Tyr198. The protease residues in contact with individual inhibitor side chains (Fig. 3) are, however, quite different compared with other aspartic proteases (especially at S3, S1, and S1′). Some of these differences can be traced to various insertions and deletions around the cleft. Five NH2-terminal residues of OM99-2 are in an extended conformation and, with the exception of P1′-Ala, all have clear contacts (within 4 Å) which define protease subsites (Fig. 3). The protease S4 subsite is mostly hydrophilic and open to solvent. The inhibitor P4-Glu side chain is hydrogen bonded to P2-Asn and is also close to the Arg235and Arg307 side chains (Fig. 3), which may explain why deleting this residue from OM99-2, to give the shorter inhibitor OM99-1, causes a 10-fold increase in K i(8, 25). The protease S2 subsite is also relatively hydrophilic and open to solvent. The hydrophilic character of the memapsin 2 S4 and S2 subsites is not conserved in the corresponding subsites of human aspartic proteases, such as pepsin, gastricsin, and cathepsins D and E. This difference may be utilized to design selectivity into memapsin 2 inhibitors. The relatively small S2 residues Ser325 and Ser327 (Gln and Met, respectively, in pepsin) may accommodate an inhibitor side chain larger than P2-Asn. The memapsin 2 S1 and S3 subsites, consisting mostly of hydrophobic residues, have conformations very different from pepsin due to the absence of a pepsin helix at residues 111–114 (26, 27). The inhibitor side chains of P3-Val and P1-Leu are closely packed against each other and have substantial hydrophobic contacts with the enzyme (Fig. 3), especially P1, which interacts with Tyr71 and Phe108. In native APP, the P2 and P1 residues adjacent to the β-secretase cleavage site are Lys and Met, respectively. Swedish mutant APP has Asn and Leu in these positions, resulting in a 60-fold increase ofk cat/K m over that of the native APP (6) and an early onset of AD (28). The inhibitor P2-Asn side chain has hydrogen bonds to P4 Glu and Arg235(Figs. 3 and 4). Replacing P2-Asn with Lys would result in the loss of these hydrogen bonds and the positive charge would likely interact unfavorably with the Arg235 side chain. P1-Met would also likely have less favorable contact with the enzyme than P1-Leu (Fig. 4). No close contact with memapsin 2 was seen for P1′-Ala. An aspartic acid at this position, as in native APP, may be accommodated.

Figure 3

Stereo presentation of interactions between inhibitor OM99-2 (orange) and memapsin 2 (light blue). Nitrogen and oxygen atoms are marked blue and red, respectively. Hydrogen bonds are indicated in yellow dotted lines. Memapsin 2 residues which comprise the binding subsites are included.

Figure 4

Electron density of inhibitor OM99-2 and the differences in the binding of Swedish and native APP at P1and P2.The omit electron density map (the |F o| − |F c| map with the inhibitor excluded from the phase calculation), contoured at 2 σ, is superimposed onto the inhibitor model with carbon atoms in green, nitrogen atoms in blue, and oxygen atoms in red. Asn and Leu side chains are those for the Swedish mutant APP at P2 and P1, respectively. The hydrogen bonds between inhibitor P2 residue Asn and Arg235 are shown in magenta. The side chains of Lys and Met (in yellow) are those for the wild-type APP, and are modeled for comparison. The turn of the inhibitor backbone at P2′ is clearly visible.

The direction of inhibitor chain turns at P2′ and leads P3′ and P4′ toward the protein surface (Figs. 3 and 4). As a result, the backbone of these three inhibitor residues deviates from the regular extended conformation. The side chains of P3′-Glu and P4′-Phe point toward the molecular surface, but have little interaction with the protease, while the terminal COOH group of P4′ has a salt bridge to Lys224 and hydrogen bonded to the hydroxyl group of Tyr198. These two COOH-terminal residues have relatively high average Bfactors (56.7 Å2 for P′3-Glu and 71.9 Å2 for P′4-Phe as compared to 27.4, 22.6, 21.5, 23.7, 24.7, and 29.7 Å2 for residues P4–P2′, respectively) and poorly defined electron density, suggesting that they are relatively mobile. In contrast, the S3′ and S4′ subsites in renin-inhibitor (CH-66) complex (23) have a defined structure. The topologically equivalent region of these renin subsites (residues 293–298 in pepsin numbering) is absent in memapsin 2. The conformation of P2′ to P4′, including a kink at P2′ and the change of backbone direction at P3′ and P4′, is rare in aspartic protease inhbitors. The backbone turn at P2′ is likely caused by a hydrogen bond between P2′ carbonyl and hydroxyl of Tyr198, not seen in the inhibitor complexes of renin (23) and endothiapepsin (22). A similar hydrogen bond is present in pepsin and a similar P2′ kink has been observed for one of its inhibitors (27). The conformation of the three COOH-terminal residues of OM99-2, including the kink at the P2′ backbone, may be a way to direct a long protein substrate out of the active-site cleft.

The well-defined subsite structures spanning P4to P2′ provide a template for rational design of drugs against memapsin 2. The unusual conformation of subsites P2′, P3′, and P4′ may facilitate the design of inhibitors selective for memapsin 2.

  • * To whom correspondence should be addressed. E-mail: jordan-tang{at}


View Abstract

Stay Connected to Science

Navigate This Article