INTRODUCTION

The generation of the ~4-kDa -amyloid peptide is a central event in the pathogenesis of Alzheimer's disease (reviewed in Refs. 1 and 2). Therefore, the mechanistic details that lead to the generation of this peptide are of interest. It is known that the -amyloid peptide is produced in the course of normal cellular metabolism of the -amyloid precursor protein (-APP)¹ (3, 4, 5, 6). Also, increased levels of this peptide appear to be produced when excess and/or aberrant forms of -APP are expressed such as observed for a number of human situations linked to Alzheimer's disease, including naturally occurring mutations in -APP (7, 8, 9, 10, 11) and Down's syndrome (12). Similar examples can be found with -APP alterations produced in vitro (11, 13) and in transgenic mice (14, 15). We have previously demonstrated that the production of -amyloid results from the catabolic degradation of -APP (16) and that increased levels of this peptide are generated by the intracellular disposal of excess normal -APP or of abnormal -APP by the secretory pathway (13). Recently, multiple intracellular pathways have been identified that are capable of processing -APP to -amyloid, each of which appears linked to the secretory pathway but at different subcellular compartments involving the early endosome and the trans-Golgi (17, 18).

In addition to efforts focused on the identification of the intracellular sites that give rise to the -amyloid peptide, considerable attention has been directed to defining the proteolytic processing events of -APP that lead to -amyloid formation or preclude it. In general, there are three key -APP processing steps mediated by enzymes referred to as -, -, and -secretase. -Secretase appears to result from a collection of enzymatic activities that cleave within the -amyloid domain of -APP permitting the bulk of the precursor to be secreted (19, 20, 21, 22, 23). Cleavage of -APP by an -secretase precludes ~4-kDa -amyloid formation and ultimately yields a ~3-kDa peptide bearing the carboxyl terminus of -amyloid.

The combination of - and -secretase processing of -APP generates the amino and carboxyl termini of the ~4-kDa -amyloid peptide, respectively. A recent report has detailed the sequence requirements for -secretase (24). This study indicates that -secretase cleavage seems to be a consequence of the activity of multiple proteinases, together giving rise to a heterogeneous set of amino termini. Several major preferred -secretase cleavage sites were shown to correspond to the three most common amino termini of the -amyloid peptide (5, 23, 24). This situation is similar to -secretase processing, which also appears to be mediated by multiple proteinases acting at a few highly preferred cleavage sites (20, 21, 22, 23). In contrast, information on the details of -secretase processing has not been as extensive. Previously, we described several features of -secretase action, specifically that this step in -amyloid formation occurs as a consequence of -APP catabolism, that cleavage takes place in the early endosome, not the lysosome, and that -secretase activity can be blocked by a unique chemical inhibitor directed to, but not exclusive for, thiol proteinases (16). In this study, we extend our analysis of -secretase processing of -APP to examine the amino acid sequence specificity directing this cleavage. Because the sites that ultimately form the carboxyl termini of the -amyloid peptide are contained within the -APP transmembrane domain, our experiments also identified the position of the membrane boundary vis à vis these cleavage sites.

EXPERIMENTAL PROCEDURES

All mutations were made using the 695-amino acid isoform of -APP (-APP695) cDNA. For wild-type and all mutant expression, the neuron-specific enolase (NSE) promoter was employed. The promoter sequence and construction prepared with the -APP cDNA have been described by Quon et al. (14). Mutant constructs V40F, V40D, T48F, and T48D were prepared by using site-specific mutagenesis with single-stranded m13 DNA and oligonucleotides carrying the desired mutation according to standard methods. A SacI-HindIII fragment from the 3 terminus of -APP was subcloned into a SacI-HindIII-digested MP18 vector (Pharmacia Biotech Inc.). Single-stranded phage DNA was isolated and used as template for the oligonucleotide mutagenesis. Isolation of m13 phage harboring the mutation was done using conditions selective for hybridization of the mutant but not for the wild-type sequence with the mutant oligonucleotide as probe radiolabeled with [³²P]ATP by T4 kinase. The m13 replicative form of the mutant m13 DNA was isolated, digested with SacI-HindIII, and reinserted into the NSE--APP backbone plasmid. Each mutation was confirmed by DNA sequence analysis. The specific oligonucleotides used for these constructions were as follows: V40F, 5-GTCGCTATGAAAACACC-3; V40D, 5-GTCGCTATGTCAACACC-3; T48D, 5-ATCACCAAGTCGATGACGAT-3; T48F, 5-ATCACCAAGAAGATGACGAT-3.

The remaining mutant constructs were made by using single-stranded oligonucleotides that coded for amino acid changes. Complementary oligonucleotides were annealed and then directly ligated into the NSE:-APP expression vector. The DNA sequence of the NSE:-APP expression vector was modified in such a way as to introduce two new restriction sites. The first change added a StuI site at amino acid positions Val⁶³²-Gly⁶³⁴ (using -APP695 amino acid isoform numbering), the sequence changing from wild-type 5-GTGGGCGGT-3 to 5-GTT-3. The second change introduced a BsrI site at amino acid position Gln⁶⁵², the sequence changing from wild-type 5-CAGTACA-3 to 5-C-3. The oligonucleotides were designed such that when ligated into the NSE:-APP expression vector, they maintained the wild-type -APP coding sequence except for the desired mutation. In addition, the codon, but not the amino acid, for Leu⁶⁴⁵ was changed from TTG to CTG in all oligonucleotides that deleted the StyI restriction site. Removal of this restriction site was used to readily identify mutant constructs. The specific oligonucleotides used for these constructions were as follows: V44D (+)-strand, 5-CGGTGTTGTCATAGCGACAGACATCGTCATCACCCTGGTGATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATCACCAGGGTGATGACGATGTCTGTCGCTATGACAACACCG-3; I45D (+)-strand, 5-CGGTGTTGTCATAGCGACAGTGGACGTCATCACCCTGGTGATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATCACCAGGGTGATGACGTCCACTGTCGCTATGACAACACCG-3; V50D (+)-strand, 5-CGGTGTTGTCATAGCGACAGTGATCGTCATCACCCTGGACATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATGTCCAGGGTGATGACGATCACTGTCGCTATGACAACACCG-3; 39-42 (+)-strand, 5-CGGTACAGTGATCGTCATCACCCTGGTGATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATCACCAGGGTGATGACGATCACTGTACCG-3; 41-44 (+)-strand, 5-CGGTGTTGTCATCGTCATCACCCTGGTGATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATCACCAGGGTGATGACGATGACAACACCG - 3; 43-46 (+)-strand, 5-CGGTGTTGTCATAGCGATCACCCTGGTGATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATCACCAGGGTGATCGCTATGACAACACCG-3; 45-48 (+)-strand, 5-CGGTGTTGTCATAGCGACAGTGCTGGTGATGCTGAAGAAGAAACA-3;()-strand, 5-GATCTGTTTCTTCTTCAGCATCACCAGCACTGTCGCTATGACAACACCG-3; 49-52 (+)-strand, 5-CGGTGTTGTCATAGCGACAGTGATCGTCATCACCAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTGGTGATGACGATCACTGTCGCTATGACAACACCG-3; 46-52 (+)-strand, 5-CGGTGTTGTCATAGCGACAGTGATCAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTGATCACTGTCGCTATGACAACACCG-3; 38-47:EGF Rc (+) strand, 5-CATTTTCATGATGCTGGGCGGCACTTTTCTCACCCTGGTGATGCTGAAGAAGAAACA-3; ()-strand, 5-GTACTGTTTCTTCTTCAGCATCACCAGGGTGAGAAAAGTGCCGCCCAGCATCATGAAAATG-3; 39-56:EGF Rc (+) strand, 5-CGGTATTTTCATGATGCTGGGCGGCACTTTTCTCTACTGGCGTGGGCGCCGGATTCAGCA-3; ()-strand, 5-GTACTGCTGAATCCGGCGCCCACGCCAGTAGAGAAAAGTGCCGCCCAGCATCATGAAAATACCG-3.

COS-7 cells were propagated in Dulbecco's modified Eagle's medium/low glucose medium containing 5% fetal calf serum. For DNA transfections, 10-cm dishes of ~5 × 10⁵ COS-7 cells, a nearly confluent monolayer, were washed once in serum-free medium containing high glucose. Ten µg of plasmid DNA prepared by CsCl density purification was mixed with lipofectamine (Life Technologies, Inc.) according to the manufacturer's protocol in Dulbecco's modified Eagle's medium/high glucose medium and was added to each monolayer. Five hours after DNA addition, an equal volume of Dulbecco's modified Eagle's medium/high glucose medium with 20% fetal calf serum was added to each dish. Forty-eight hours after transfection, medium was removed and replaced with methionine-, cysteine-, and serum-free medium for 30 min. Fresh methionine-, cysteine-, and serum-free medium with ~150 µCi of [³⁵S]methionine/cysteine (Translabel; ICN) was added to each dish. Cells were radiolabeled for 4 h.

Immunoprecipitation of conditioned medium (3 ml total) for -amyloid peptide was performed as described previously (13, 16) except that a different antibody to -amyloid, number 6514, was used. The 6514 antiserum was raised in a rabbit using a synthetic 1-40 -amyloid peptide purchased from Bachem (Torrance, CA), which was conjugated to keyhole limpet hemocyanin. The epitope recognized by the 6514 antiserum was mapped to the amino terminus of the -amyloid peptide by radioimmunoassay with synthetic peptide subdomains of -amyloid and by its ability to immunoprecipitate ~4-kDa, but not ~3-kDa, -amyloid peptide. Preparation of cell lysates and immunoprecipitation of -APP and -APP carboxyl-terminal fragments with BC-1 antiserum were carried out as described by Higaki et al. (16) and Zhong et al. (13). Analysis of immunoprecipitates from conditioned medium or -APP carboxyl-terminal fragments from cell lysates was made using 16.5% Tris-Tricine SDS-polyacrylamide gel electrophoresis as described by Zhong et al. (13) and full-length -APP immunoprecipitates from cell lysates by 8% Tris SDS-polyacrylamide gels. Internal molecular weight protein standards were electrophoresed on each gel. Typical exposure times were ~16 h for -APP and -APP carboxyl-terminal fragments and 4 days for -amyloid. Quantitation of immunoprecipitated proteins was made using a PhosphorImager (Molecular Dynamics).

RESULTS

Processing of -APP by -secretase produces the carboxyl terminus of the ~4-kDa -amyloid peptide. Heterogeneity of this terminus has been observed for -amyloid produced in vitro (25) and in vivo (26, 27, 28, 29, 30), with position 40 of the -amyloid domain being the most common terminus and positions 38, 39, 41, 42, and 43 less frequently represented by wild-type -APP. We examined processing at this site with a variety of amino acid substitutions. Conservative and highly divergent substitutions were made and analyzed for -amyloid production using immunoprecipitation. Transient expression of wild-type -APP or -APP harboring each mutation was carried out in COS-7 cells. An antiserum, number 6514, that is directed to an epitope at the amino terminus of the ~4-kDa peptide was employed so that protein sequence alterations would not perturb detection of the peptide.

Hydrophobic substitutions at position 40 did not preclude -amyloid production. Fig. 1B and Table I illustrate one example of such a substitution, phenylalanine, in which levels of -amyloid were found to be nearly equivalent to that generated by transfection of wild-type -APP DNA. In addition, -APP production and maturation and carboxyl-terminal fragment production resulting from - and -secretases were unaltered for this mutant (Fig. 1, C and D). Identical results were obtained with alanine or methionine substitutions (data not shown). In contrast, substituting a negatively charged residue at position 40, such as an aspartate residue, resulted in a number of alterations, specifically the absence of -amyloid (Fig. 1B and Table I) and carboxyl-terminal fragments (Fig. 1C), as well as a lack of glycosylation of nascent -APP (Fig. 1D). Identical results were obtained with an arginine substitution at this site (data not shown). The effects of a charged amino acid substitution at position 40, in particular the lack of secretory processing and carboxyl-terminal fragment generation, indicated that this -APP mutant may not be capable of membrane insertion due to the addition of a charged residue within the transmembrane domain. This effect is consistent with previously determined dictums of protein membrane insertion (31, 32). The introduction of a charged residue, aspartate, was next used to map the boundary of the membrane on the cytoplasmic or carboxyl-terminal side of the bilayer. Two identical substitutions, phenylalanine or aspartate, were constructed at position 48 of the -amyloid domain (where position 1 is the amino-terminal aspartate of -amyloid). Upon expression of each -APP mutant, it was found that neither prevented the production of -amyloid (Fig. 1B). Also, no negative influence was exerted by these mutations on the generation of carboxyl-terminal fragments or glycosylated -APP, both reflecting secretory transport of the precursor (Fig. 1, C and D). Because the charged aspartate residue was tolerated at position 48, this suggested that the aspartate was not housed within the transmembrane domain of -APP. To more precisely map the membrane boundary in this region, a series of aspartate mutants were made spanning positions 40-50. The results of this ``aspartate walk'' are shown in Fig. 2. -Amyloid formation and secretory processing of -APP in terms of carboxyl-terminal fragment generation were evident with substitutions at positions 48 and 50. Substitutions at positions 40, 44, and 45 each precluded -amyloid formation, carboxyl-terminal generation, and -APP glycosylation despite roughly equal synthesis of -APP by all mutants. Table I summarizes the efficiency of -amyloid formation for all mutants shown in Figs. 1 and 2. Although peptide levels were somewhat reduced for all mutants at positions 48 and 50, in contrast, production of -amyloid was completely eliminated by the introduction of an aspartate residue between positions 40 and 45. The lack of ~4-kDa -amyloid production by these mutants was likely due to the absence of -APP secretory transport prevented by a lack of membrane insertion. Together these results suggest that the membrane boundary maps to position 46 or 47. 

Influence of substitution mutations on positions 40 and 48 of -amyloid domain.

Table I. -Amyloid levels produced by carboxyl-terminal substitution mutations in transmembrane domain 40 V I A T 44 V 45 I V I 48 T L 50 V  -amyloid levelsa (% wild type) Wild type 100 V40F F 81 V40D D 3 V44D D 0 V45D D 0 T48D D 57 T48F F 46 V50D D 48 a  -Amyloid levels were normalized for the amount of -APP produced by each transfected DNA relative to wild-type -APP. Quantitation of -amyloid and associated -APP for each mutant was made using a PhosphorImager after isolation of the proteins by immunoprecipitation and fractionation by PAGE. Each value represents the average of two separate experiments that gave similar results.

A series of deletions were introduced in the -amyloid domain of -APP that were inclusive of and adjacent to the -secretase cleavage sites for mature -amyloid. These deletions were made in order to examine the sequence specificity for the processing step yielding the carboxyl terminus of -amyloid. Five different deletions of 4 residues each were prepared and analyzed for their ability to form -amyloid. In toto, positions 39-52 were studied. A schematic representation of these nested deletion mutants is illustrated in Fig. 3A. Deleting successive blocks of 4 amino acids in this region did not eliminate -amyloid formation (Fig. 3B). In addition, -APP maturation was normal and identical to that for wild-type -APP (Fig. 3C). While an ~4-kDa -amyloid peptide was detected for each mutant, one mutant (39, 40, 41, 42) displayed a slightly slower gel mobility. This may be due to displaced cleavage or to an altered mobility due to peptide sequence. We favor the idea of displaced cleavage, since proline substitution mutants at position 40 or 42 do not alter -amyloid production but appear to change the site of cleavage in this domain.² A larger deletion was also made that encompassed positions 46-52 (see Fig. 3A). Unlike the shorter deletions, this mutant did not produce -amyloid or mature -APP (Fig. 3, B and C). The profile of this mutant is identical to those that precluded -APP membrane insertion. The native lysine, formerly at position 53 but moved to position 46 by this deletion, is likely to have blocked membrane insertion, which is consistent with mapping the face of the bilayer to position 46 or 47 as determined in the preceding experiments. The efficiency of -amyloid production for each deletion mutant was calculated (Table II). The large deletion spanning residues 46-52 did not produce significant amounts of -amyloid for reasons mentioned above. Four amino acid deletions collectively encompassing residues 45-52 did not substantially alter peptide levels. Some reduction in efficiency of -amyloid production was noted when residues spanning 39-42 and 41-44 were deleted, but a reproducible loss in efficiency of -secretase cleavage was noted when residues 43-46 were removed.

Deletion mutagenesis in carboxyl-terminal region of -amyloid domain.

Table II. -Amyloid levels produced by carboxyl-terminal deletion mutations in transmembrane domain 35 M V G G V 40 V I A T V 45 I   V I T L 50 V M L K K K  -amyloid levelsa (% wild type) Wild type 100  39-42 57  41-44 66  43-46 35  45-48 75  49-52 77  46-52 9 a  -Amyloid levels were normalized for the amount of -APP produced by each transfected DNA relative to wild-type -APP. Quantitation was made as described in Table I. Each value represents the average of two separate experiments.

Based on the results from series of four amino acid deletions, it appears that the protein sequence from positions 39-52 need not be exactly maintained for -secretase cleavage. It can be noticed, however, that each deletion mutant reconstructed a new hydrophobic sequence in this region of processing. Therefore, it is possible that the requirements of -secretase cleavage rely more on a hydrophobic stretch of residues than on the unique order of individual hydrophobic amino acids. To further investigate -secretase cleavage specificity, two replacement mutants were constructed using the transmembrane sequence from an entirely different protein, the human epidermal growth factor (EGF) receptor, HER2 (33). Because transmembrane domains are composed of a limited subset of amino acids, the EGF receptor transmembrane domain was selected for this experiment because its sequence is more divergent from the -APP transmembrane sequence as compared with a number of other type 1 membrane-spanning proteins. For one replacement mutant, 18 residues of -APP were removed, and 18 residues from a comparable position in the EGF receptor were added. For the second mutant, positions 38-47 of -APP were replaced with 10 residues of the EGF receptor again obtained from an analogous locale. Fig. 4A provides an illustration of both constructs. Transient expression of each -APP:EGF receptor chimeric mutant indicated that both efficiently produced a ~4-kDa -amyloid peptide and generated a typical pattern of -APP carboxyl-terminal fragments (Fig. 4, B and C). The efficiency of -amyloid formation was determined for the two chimeric mutants, and both were found to be acceptable substrates for -secretase: 37-47 -APP:EGF Rc and 39-56 -APP:EGF Rc for -amyloid with a 93% and 81% efficiency, respectively, relative to wild-type -APP. It should be noted, however, that cleavage efficiency may actually be lower should the processing of the chimeric substrates result in inclusion of the additional methionine residues donated from the EGF receptor moiety in this domain. In any case, this result confirms our previous observation that -secretase cleavage may favor a unique array of individual residues but that if these residues are conservatively altered, processing can nevertheless occur.

EGF receptor substitution mutagenesis of -APP -amyloid carboxyl-terminal region.

DISCUSSION

The enzymatic mechanism of -amyloid formation is not well understood, especially the cleavage that generates the carboxyl terminus of this peptide. Several factors confound this -APP processing step including 1) significant heterogeneity observed for the -amyloid carboxyl terminus spanning 6 residues (positions 38-43) (25, 26, 27, 28, 29, 30); 2) the influence of proximal mutations, such as at position 46 (or ``Hardy'' mutation), which alter primary cleavage site preference (7, 34, 35); and 3) the location of -secretase cleavage sites within the transmembrane domain of -APP.

The objective of this study was to define the parameters of -secretase processing. Specifically, we were interested in defining the sequence specificity of -secretase cleavage. Limited information has been available on the molecular details of this processing step. We have previously demonstrated that -secretase cleavage occurs in the early endosome, an organelle that can be derived directly from the trans-Golgi network and/or from the internalization of plasma membrane (16). In addition, we described a unique chemical inhibitor of -secretase activity that blocks -amyloid formation (16). This inhibitor acts on thiol proteinases but is not exclusive for this class of proteinases. While numerous reports have isolated putative -secretase activities, none has definitively identified a -secretase (37, 38, 39, 40). Hence, the enzyme class and cleavage specificity of -secretase are unclear. The data reported herein address the extent of sequence specificity required to produce the carboxyl terminus of the ~4-kDa -amyloid peptide.

Using DNA mutagenesis, a collection of -APP mutants was constructed, and they were analyzed for their influence on -amyloid formation. Together these mutants spanned the region of -secretase cleavage plus downstream sequences, i.e. residues 38-56 of the -amyloid domain (where position 1 is the amino-terminal aspartate of -amyloid). One set of -APP mutations were overlapping nested deletions, each of which removed 4 amino acids. None of these mutants prevented the formation of the ~4-kDa -amyloid peptide, suggesting loose sequence specificity preference for cleavage. A reduced efficiency of -secretase processing of -APP was found for mutations introduced in the region spanning residues 43-46 (TVIV), suggesting a preference for this amino acid sequence. A slight influence on cleavage efficiency was also noted for 4 residues flanking both sides of this domain (VVIATVIVITLV). Despite this apparent preference, our experiments with a more extreme pair of mutants indicate that -secretase processing has loose sequence requirements. One mutant of the pair replaced 10 residues of this -APP domain with a divergent sequence of 10 residues from an analogous region of the human EGF receptor transmembrane domain. The other mutant was a similar -APP:EGF receptor chimera but with 18 amino acids substituted into -APP from positions 39-56 of the -amyloid carboxyl-terminal domain. Neither of the two chimeras negatively influenced -amyloid formation. While the EGF receptor transmembrane sequence used in the chimeric constructs is composed of hydrophobic amino acids, it is significantly different with respect to the array of individual residues from that in -APP. Despite these major changes, carboxyl-terminal processing and -amyloid generation occurred.

In conclusion, these experiments indicate that -secretase does not cleave -APP by recognition of a unique sequence. Rather, -secretase appears to have a preferred primary sequence directed to hydrophobic (and possibly other) residues. If this cleavage site is conservatively altered, -secretase processing can still occur, indicating a loose preference or specificity. It is possible that multiple -secretases exist that collectively confer the observed loose cleavage specificity. It is also possible that individual -secretases may display greater specificity. The experiments described here do not address this issue, and not until pure -secretase(s) is available can this issue be definitively resolved.

The membrane vis à vis the generation of carboxyl terminus of -amyloid has been an additional issue. Two scenarios can be postulated to address the membrane and -amyloid processing: the membrane is destroyed or damaged, allowing -secretase free access to this region of -APP, or alternatively, the membrane remains intact and the carboxyl-terminal domain is accessible to -secretase action. No data exist that reconcile this issue. However, it is clear that the membrane is indirectly important to -amyloid generation. It has been shown by a number of investigators that transport of -APP through the secretory pathway is a requisite for -amyloid formation (3, 4, 5), and insertion of -APP into the membrane is essential for transport. Our results with substitution mutations within the -APP transmembrane confirm this point from a different experimental perspective. We found that introduction of charged residues in this region prevent membrane insertion of the precursor, block secretory transport, and as a result, preclude -amyloid formation. Furthermore, when we used this mutational approach, we were able to experimentally map the boundary of the -APP transmembrane domain on the cytosolic face of the bilayer. Analysis of sequence substitutions from positions 40-50 of the -amyloid domain identified position 46/47 as the membrane face (Fig. 5). Heretofore, the transmembrane domain was theoretically placed at position 52 by the Chow-Fasman hydropathy algorithm (41). It is interesting to note that point mutations associated with familial Alzheimer's disease (``Hardy'' mutation) map at position 46 (7, 34, 35). These natural mutations may further influence the cytosolic membrane boundary of -APP. Also, the core preference sequence that we identified for -secretase cleavage includes residue 46. Thus, the membrane position may direct carboxyl-terminal processing of -amyloid to this site. One might then speculate that carboxypeptidases continue to process this terminus, forming the heterogenous collection of carboxyl termini of -amyloid.

Illustration of the experimentally determined transmembrane position on the cytosolic face of -APP.

Our experimentally mapped membrane position for -APP is in agreement with recent observations that indicate endoplasmic reticulum membranes have substantially reduced cholesterol content as compared with the plasma membrane, resulting in shorter protein transmembrane domains (42). As a consequence of less cholesterol, the membrane thickness is reduced, and permeability is increased (42). Since -APP membrane insertion occurs in the endoplasmic reticulum, the step affected by the described substitution mutations, it is expected that the membrane does not extend to the lysine triplet commencing at position 53. It is important to note that analysis of brain tissue from patients with Alzheimer's disease shows a 30% reduction in cholesterol:phospholipid ratio compared with age-matched control samples (36). Therefore, the membrane position of -APP or the extent of its transmembrane domain in Alzheimer's disease brain may be in close proximity to the -secretase cleavage sites within the -amyloid domain and may, in fact, serve to direct cleavage to this locale. A combination of factors, such as reduced brain membrane cholesterol, which might potentially permit greater access of the -amyloid carboxyl-terminal domain to -secretase action, and loose sequence specificity for this processing step, may promote the formation of the -amyloid peptide in Alzheimer's disease brain.