The Transmembrane Domain of the Amyloid Precursor Protein in Microsomal Membranes Is on Both Sides Shorter than Predicted*

The amyloid precursor protein is cleaved within its ectodomain by β-amyloid-converting enzyme (BACE) yielding C99, which is further cleaved by γ-secretase within its putative transmembrane domain (TMD). Because it is difficult to envisage how a protease may cleave within the membrane, alternative mechanisms have been proposed for γ-cleavage in which the TMD is shorter than predicted or positioned such that the γ-cleavage site is accessible to cytosolic proteases. Here, we have biochemically determined the length of the TMD of C99 in microsomal membranes. Using a single cysteine mutagenesis scan of C99 combined with cysteine modification with a membrane-impermeable labeling reagent, we identified which residues are accessible to modification and thus located outside of the membrane. We find that in endoplasmic reticulum-derived microsomes the TMD of C99 consists of 12 residues that span from residues 37 to 48, which is N- and C-terminally shorter than predicted. Thus, the γ-cleavage sites are positioned around the middle of the lipid bilayer and are unlikely to be accessible to cytosolic proteases. Moreover, the center of the TMD is positioned at the γ-cleavage site at residue 42. Our data are consistent with a model in which γ-secretase is a membrane protein that cleaves at the center of the membrane.

An early event in the pathogenesis of Alzheimer's disease is the generation of the amyloid ␤ peptide (A␤), 1 which is later deposited in the amyloid plaque. The 4-kDa peptide A␤ is a product of the complex proteolytic processing of the amyloid precursor protein (APP) (for a review, see Ref. 1). Some features of the APP processing are shared with the processing of the cell surface receptor Notch (for a review, see Ref. 2). Both are type I membrane proteins that undergo a first proteolytic cleavage within their ectodomain, leaving a C-terminal frag-ment within the membrane. These C-terminal fragments may then undergo proteolytic cleavage within their transmembrane domains (TMD), a process termed intramembrane proteolysis (for a review, see Ref. 3). Upon intramembrane proteolysis, the cytosolic portions of these proteins may move to the nucleus and stimulate the transcription of target genes (4 -9). In the proteolytic pathway that leads to the generation of A␤, APP is first cleaved within its lumenal domain by the recently identified aspartyl protease, ␤-amyloid-converting enzyme (for a review, see Refs. 10 and 11). The resulting 99-residue-long Cterminal fragment of APP (C99) can then be cleaved intramembranously by a protease activity called ␥-secretase, which leads to the release and secretion of the A␤ peptide (for a review, see Ref. 1). Despite its importance for the pathogenesis of Alzheimer's disease, ␥-secretase has not yet been unequivocally identified. However, the ␥-cleavage of C99 depends on the presence of the membrane protein presenilin 1, which itself could be ␥-secretase or part of a larger ␥-secretase complex (for reviews, see Refs. 2 and 12).
␥-Cleavage occurs predominantly after residue 40 of C99 and to a minor extent after residue 42, thus generating the 40-and 42-residues-long peptides A␤ 40 and A␤ 42 (13)(14)(15)(16). Little is known about the mechanism of ␥-cleavage and the factors that determine whether the cleavage takes place after residue 40 or 42 of C99. Intensive mutagenesis of amino acid residues within the TMD of C99 has shown that the generation of A␤ is not sequence-specific (17)(18)(19)(20)(21)(22)(23). In contrast, the specific position where ␥-cleavage takes place (i.e. cleavage after residue 40 or residue 42) strongly depends on the length of the TMD of C99 (18,21).
In addition to its broad substrate specificity, ␥-cleavage is remarkable because it occurs within the predicted TMD of C99. Recently, an increasing number of proteins have been reported to be cleaved intramembranously, but so far no protease has been proven to cleave within the membrane. Nevertheless, the metallo-protease S2P, the putative serine protease rhomboid, and presenilin 1 have their active site residues within their predicted TMDs (24 -27).
Because of the lack of direct experimental proof for the ␥-cleavage taking place within the lipid bilayer, alternative structures and membrane boundaries for the TMD of C99 have been proposed. These models suggest that the C-terminal half of the putative TMD and the ␥-cleavage site are not buried within the membrane but instead are exposed to the cytosol and accessible for a soluble or membrane-associated protease (18,19,28). However, the true mechanism remains elusive because a thorough characterization of the length of the TMD of C99 in biological membranes is still lacking.
To resolve these controversies and to better understand the mechanism of ␥-cleavage, we biochemically determined the length and the position of the TMD of C99. In a cysteine mutagenesis scan, we created several single cysteine mutants of C99 with a cysteine placed within or adjacent to its predicted TMD. In vitro, the single cysteine mutants of C99 were translated directly into microsomal membranes, which derive from the endoplasmic reticulum. The C99-containing microsomes were incubated with IASD (4-acetamido-4Ј-((iodoacetyl) amino) stilbene-2,2Ј-disulfonate), which specifically labels cysteine residues. Because IASD is membrane-impermeable (29), only those cysteine residues are labeled that are located outside the membrane.
Our results show that the apparent TMD of C99 inserted into ER-derived membranes is significantly shorter than previously assumed and comprises only 12 amino acids instead of the 24 residues that are predicted by computer algorithms. Importantly, we found that the ␥-secretase cleavage site after residue 42 is positioned exactly at the center of the newly defined TMD.
These data indicate that the exact positioning of the APP TMD with respect to the membrane may determine the cleavage site and, thus, that the cleavage may occur at the exact center of the lipid bilayer.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The single cysteine mutants of C99 were generated by PCR using the QuikChange site-directed mutagenesis kit (Stratagene, Amsterdam), suitable oligonucleotides, and the plasmid pBS/SPC99 as template (17). The sequence of all constructs was confirmed by DNA sequencing.
In Vitro Transcription and Translation-The experiments described were performed with C99 mutants expressed from the SP-C99 encoding plasmids by in vitro transcription and translation using the RiboMAX large scale RNA production system (Promega) for transcription and the rabbit reticulocyte lysate system (Promega) for translation. Alternatively, coupled transcription and translation reactions were performed with the TnT T7 master mix system (Promega). Reactions were incubated for 90 min at 30°C with [ 35 S]methionine in the presence or absence of canine microsomal membranes (Promega).
Immunoprecipitation-5 l of the in vitro translation reaction before or after labeling with IASD were diluted in 500 l of solubilization buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40) and incubated for 2 h with 20 l of protein A-Sepharose (100 mg/ml, Amersham Biosciences) and 5 l of the polyclonal antibody 22/13, which was raised against the C-terminal 13 residues of C99. For the analysis on isoelectric focusing (IEF) gels, 40 l of the in vitro translation reaction were used and immunoprecipitated as described above but in a volume of 1 ml. Protein was recovered in 40 l of sample buffer for PAGE or in 30 l of denaturing buffer for IEF gel analysis (according to the IPGphor system instructions; Amersham Biosciences).
Membrane Sedimentation and Sodium Carbonate Extraction-For the IASD labeling of the C99 mutants with a cysteine placed at the N-terminal lumenal side of the membrane, the microsomal vesicles were converted into membrane sheets prior to IASD reaction (30). 30 -50 l of the in vitro translation reaction were mixed with 750 l of 100 mM sodium carbonate, pH 11.5, and incubated at 0°C for 30 min. Membranes were recovered by ultracentrifugation in the Beckman 45 TLi rotor at 50,000 rpm and 4°C for 45 min. For the subsequent IASD labeling reaction, the membrane pellets were rinsed once with ice-cold distilled water and then dissolved in 5-10 l of 0.5 M sodium phosphate.
Cysteine Modification by IASD and Gel-shift Electrophoresis-Each radiolabeled single cysteine mutant of C99 (6 l of the in vitro translation reaction) was incubated in 0.5 M sodium phosphate, pH 7.5, containing 10 mM dithiothreitol (6 l) for 5 min at room temperature. Then 100 mM IASD (MoBiTec, Göttingen, Germany) was added to a final concentration of 20 mM IASD and 4 mM dithiothreitol (31). After incubation for 30 min at room temperature, the remaining IASD was inactivated with 40 mM dithiothreitol. For each labeling experiment a negative control, in which water was used instead of IASD, was run in parallel. C99 was immunoprecipitated and analyzed on expanded Tris/ Tricine gels or on isoelectric focusing gels with an immobile pH gradient (pH 3-10, 13 cm, IPGphor isoelectric focusing system; Amersham Biosciences). Radiolabeled proteins separated by electrophoresis were visualized using phosphorimaging (BAS 1000), whereas those separated by isoelectric focusing were subjected to autoradiography with the help of the Kodak intensifying screen (BioMax TranScreen LE).
For IASD labeling after membrane disruption, the in vitro transcription/translation reactions were incubated in an equal volume of 0.5 M sodium phosphate, pH 7.5, containing 10 mM dithiothreitol and 2% Nonidet P-40 for 10 min at room temperature.

RESULTS
Cysteine-scanning Mutagenesis of C99 -To biochemically determine the TMD length of C99, we performed a single cysteine mutagenesis scan of C99. Because C99 does not contain any cysteine residue, we replaced individual residues within and just outside of the TMD of C99 by a cysteine, such that each of the 23 C99 mutants contained one cysteine (Fig. 1). In a coupled in vitro transcription/translation experiment, these mutants were inserted into canine microsomes that derive from the membrane of the endoplasmic reticulum. The C99-containing microsomes were then incubated with the membrane-impermeable cysteine-labeling reagent IASD. Because of its hydrophilic character, IASD cannot diffuse into the lipid bilayer and therefore only labels cysteine residues that are located outside the membrane (29,31). IASD labeling was subsequently analyzed by one-and two-dimensional gel electrophoresis.
Membrane Insertion of SP-C99 -To ensure the correct insertion of C99 into the microsomal membranes, an N-terminal signal peptide (SP) was fused to the C99 constructs (SP-C99, Fig. 1). The signal peptide is cleaved by signal peptidase upon membrane integration, thus converting SP-C99 into C99 (32)(33)(34). The C99 proteins were immunoprecipitated from the solubilized membranes with antibody 22/13 that binds to the C terminus of C99 and analyzed by polyacrylamide gel electrophoresis. A single band with the expected apparent molecular mass of 12 kDa was detected (Fig. 2). In a control experiment without membranes, the signal peptide of SP-C99 was not cleaved and the immunoprecipitated SP-C99 had an apparent molecular mass of 11.5 kDa (Fig. 2). Furthermore, SP-C99 showed extensive aggregation, as seen by the numerous high molecular mass bands (Fig. 2). Because of the presence of the signal peptide, SP-C99 contains 17 more amino acids than C99. Nevertheless SP-C99 showed a lower apparent molecular mass than C99, which might be because of a more compact secondary structure of SP-C99 induced by the presence of the hydrophobic signal peptide. This observation is in agreement with previous studies (32,35). The difference in molecular mass between SP-C99 and C99 was observed for C99 wt and all cysteine mutants (shown in Fig. 2 for representative mutants). This verifies that all C99 proteins were completely inserted into the microsomal membranes and that the signal peptide was cleaved. In addition to the 12-kDa band of C99, a 6.5-kDa FIG. 1. Schematic representation of the SP-C99 single cysteine mutants. SP-C99 consists of the 17-residue-long signal peptide (SP) of APP followed by a two-amino acid spacer (Leu and Glu) and the Cterminal 99 amino acids (C99) of APP. Amino acids are shown in the one letter code. The residues of wild type C99 that were replaced by a cysteine are indicated in black. Every SP-C99 mutant contains only a single cysteine residue. The hydrophilicity plot of SP-C99 (using the scale of Kyte and Doolittle (47) over a window of 9 residues) indicates hydrophobicity by negative numbers (kcal/mol) and shows that residues 29 to 52 are the predicted TMD of SP-C99. protein could be detected (Fig. 2) that is likely to be the result of an internal transcription starting point at methionine 35. As a further control for the correct membrane integration of the mutants, we verified that the C99 proteins translated in the presence of microsomal membranes could be recovered by sodium carbonate treatment and subsequent membrane sedimentation of the microsomes. This method allows the recovery of integral membrane proteins, whereas soluble proteins and peripheral membrane proteins do not sediment together with the microsomes (30). C99 wt and the C99 mutants were mainly detected in the integral membrane protein fraction (data not shown), demonstrating efficient membrane integration of the C99 mutants.
Molecular Mass Shift after IASD Treatment-To determine whether the individual cysteine residues of the C99 mutants are located within or outside of the microsomal membrane, we used IASD, a reagent that specifically reacts with free sulfhydryl groups, i.e. the single cysteine residues in the C99 mutants. IASD contains two sulfonate groups (pK a ϳ-6.5), which are fully ionized under our reaction conditions (pH 7.5), preventing diffusion into the membrane (29,31). Thus, cysteine residues of C99 that are buried within the lipid bilayer are not labeled. Because the N terminus of C99 is located in the lumen of the microsomal vesicles, those residues located on the Nterminal (lumenal) side of the TMD are not directly accessible to the membrane-impermeable IASD. Therefore, when analyzing the mutants C99/24C-40C the membrane vesicles were converted to membrane sheets by sodium carbonate treatment prior to IASD treatment. To verify that this method did not change the positioning of C99 within the membrane, we controlled that the sodium carbonate treatment did not interfere with the C-terminal (cytosolic) borderline of the IASD-inaccessible domain (data not shown) as we have identified it in this study.
IASD modification adds a mass of 450 Da to C99, which may change the electrophoretic mobility of C99, as has been reported previously for other proteins (29,31). After incubation with and without IASD, the different C99 mutants were immunoprecipitated and the electrophoretic mobility was compared. Under these conditions only a minor, not significant, fraction of the negative control (C99 wt) showed unspecific labeling (Fig. 3).
IASD treatment of the C99 mutants with a cysteine positioned N-terminal of residue 33 and C-terminal of residue 49 led to a clear shift to a higher apparent molecular mass (Fig. 3 and Table I). This demonstrates that the individual cysteine residue in these mutants (at positions 24 -32 and 50 -57) had been labeled by IASD and thus was located outside the microsomal membrane. In contrast, for cysteines at positions 33-49 and C99 wt no gel shift was observed ( Fig. 3 and data not shown). For the mutants with a cysteine between residue 33 and 40, an IASD-induced shift could not even be detected in a control experiment in which the mutants were IASD-treated in a non-membrane-embedded state (data not shown but summarized in Table I). Therefore, we considered the possibility that IASD modification at these residues may have occurred without an alteration in the apparent electrophoretic mobility.

Analysis of IASD Modification by Isoelectric
Focusing-To answer this question we used IEF. The two negative charges added to C99 upon IASD modification lower its isoelectric point (IEP) from an IEP of 6.09 to 5.52 (as calculated by the computer program DNAStar), which should be detectable as a shift on isoelectric focusing gels with an immobilized, linear pH gradient. To validate IEF as a method for determining the IASD modification of C99, we first analyzed as positive controls C99 mutants that had shown a shift in molecular mass (C99-V24C, V50C, and M51C, Fig. 3) and as negative control C99 wt. C99 wt with or without IASD treatment showed a single band at the expected calculated IEP of pH 6 ( Fig. 4) and thus was not modified by IASD. In contrast, IASD treatment of C99/V24C, -V50C, and -M51C led to a clear shift from an IEP of 6 to 5.5, as expected (Fig. 4). This shows that IASD labeling can be detected well on IEF gels. In some experiments an additional band with an IEP of 5.8 was observed for unknown reasons. However, a protein with an IEP of 5.8 carries only one additional negative charge compared with C99, which therefore cannot be because of IASD labeling. This is further confirmed by the fact that this protein band with an IEP of 5.8 did not occur consistently in identically performed experiments and even in the absence of IASD (see Fig. 4, C99/L34C as an example). This verifies that this IEP isomer of C99 is irrelevant for our analysis because it does not interfere with the analysis of IASD modification. The identity of the protein bands for C99 was confirmed by two-dimensional gels on which the proteins showed the expected molecular mass of 12 kDa (without IASD modification) and 12.5 kDa (after IASD modification; data not shown, but compare Fig. 2). The clear shift seen on IEF gels after IASD treatment of C99 mutants validates IEF as a method for determining whether IASD modification of C99 has occurred.
We next analyzed whether those C99 mutants that did not show a gel shift in their molecular mass upon IASD treatment showed a shift in their IEP. IASD incubation of the C99 mutants L34C, M35C, V36C, and L49C (Fig. 4) resulted in a clear shift in the IEP from 6 to 5.5, indicating that IASD modification had occurred and that this modification can be detected on IEF gels. These data suggest that the corresponding cysteine resi- (after signal peptide cleavage (ϩ)) has a higher apparent molecular mass than SP-C99 (without cleavage of the signal peptide (Ϫ)). Numbering refers to the residue of C99 that was exchanged by a cysteine. 35 Slabeled C99 single cysteine mutants. In vitro translations in the presence of microsomal membranes of C99 wt and C99 single cysteine mutants were treated with IASD (ϩ) or water (Ϫ), immunoprecipitated, separated by high-resolving Tris-/Tricine PAGE, and visualized using phosphorimaging (shown for a selection of the C99 mutants). The appropriate section of the gel shows a shift to a higher apparent molecular mass for IASD-treated samples of some C99 mutants, whereas for other mutants no shift could be detected (summarized in Table I). dues at positions 34, 35, 36, and 49 were located outside the microsomal membrane. In contrast, C99 mutants with cysteines at positions 37,38,39,40,44,45,47, and 48 did not show a shift in IEP on IEF gels (Fig. 4), suggesting that they are buried within the membrane bilayer. Moreover, when the microsomal membranes were disrupted by the addition of a detergent (1% Nonidet P-40, for 10 min), the mutants showed the expected shift upon IASD labeling, revealing that they can be labeled with IASD when no longer shielded by the membrane. This finding confirms that IASD does not label membraneinserted protein portions, which is in good agreement with previous work of Krishnasastry et al. (29) that showed that a cysteine mutant of ␣-hemolysin could not be labeled with IASD when embedded into membranes, although it was rapidly labeled in the absence of membranes. Therefore, the absence of a shift in the IEP for a membrane-inserted single cysteine mutant indicates that this cysteine is located within the microsomal membrane. This experiment further shows that IEF provides a method to overcome problems in the detection of IASD labeling that sometimes occur with analysis on conventional PAGE and that IEF can be used to analyze IASD labeling of all our C99 single cysteine mutants.

FIG. 3. IASD modification analysis of membrane-bound,
In summary, the IASD labeling analysis of all membraneinserted C99 mutants shows that only 12 of the 24 residues of the putative TMD of C99 (residues at position 37 to 48) are not

middle column) and without membrane insertion (left and right column) by molecular mass (MW) and by isoelectric point (IEP)
FIG . 4. a, IASD modification of membrane-bound C99 mutants demonstrated by a shift in the IEP on isoelectric focusing gels with an immobilized, linear pH gradient. The in vitro transcription/translation sample was IASD-treated, immunoprecipitated, separated by its IEP, and visualized by autoradiography. Numbers indicate the position of the individual cysteine mutation in the C99 sequence. After IASD treatment, (ϩ)-labeled C99 mutants shifted from an IEP of ϳ6 for untreated samples (Ϫ) to an IEP of ϳ5.5 (marked by a rectangle), as expected. The absence of such a shift in IEP seen for C99 wt and several mutants indicates that IASD labeling did not occur (highlighted in gray). b, control of general accessibility of the cysteines to IASD after membrane solubilization. After incubation of the in vitro transcription/ translations in the presence of a detergent, all mutants show the IASDinduced shift in the IEP, confirming that the absence of the shift in Fig.  4a for some mutants is because of the position of their cysteines within the lipid bilayer. accessible to IASD modification and thus constitute that part of C99 that is buried in microsomal membranes and constitute the TMD of C99 in ER-derived membranes (schematically shown in Fig. 5).

DISCUSSION
The ␥-secretase cleavage of the C-terminal APP fragments is a remarkable proteolytic mechanism as it occurs around the center of the putative TMD of APP (for a review, see Ref. 2). Because direct evidence is lacking that the ␥-cleavage takes place within the lipid bilayer, alternative boundaries for the TMD of C99 have been proposed such that a cytosolic or membrane-associated protease could cleave C99 (18,19,28). However, these studies only analyzed one half but not the total length of the TMD. Thus it remained unclear whether the ␥-cleavage occurs within the membrane. To analyze whether the ␥-cleavage sites of APP are indeed located within the membrane or are exposed to cytosolic cleavage, we biochemically determined the whole length of the TMD of C99 in microsomal membranes. For this analysis, we carried out a cysteine mutagenesis scan of C99, which is a widely employed technique for topology studies of membrane proteins (for a review, see Ref. 36). The cysteine mutants of C99 were inserted into membranes and labeled with the membrane-impermeable, cysteinespecific reagent IASD.
In our analysis we found that IASD modification occurred for all cysteine residues of C99 located N-terminal of residue 37 and C-terminal of residue 48, but cysteine residues at positions 37-40, 44, 45, 47, and 48 were not modified with IASD. We conclude from this that the 12 residues, 37 to 48, of C99 are shielded from IASD modification by the microsomal membrane and thus represent the actual domain of C99 or APP, respectively, that is embedded within the ER-derived membrane. This domain is significantly shorter on both sides than the 24 hydrophobic residues that are predicted by computer algorithms (37).
IASD is highly charged and therefore cannot diffuse into or through the lipid bilayer, but it might enter the region of the polar head groups of the lipids on the surface of the membrane. Furthermore, the thiol-reactive iodoacetyl group of IASD is a few Å apart from the charged sulfonate groups. Although the sulfonate groups are supposed to stay outside the membrane and confer membrane-impermeability to IASD, the hydrophobic thiol-reactive group may slightly dip into the membrane and react with a cysteine positioned just below the membrane boundary. This suggests that the exact border of the membrane-embedded protein portion may slightly differ, depending on the hydrophilicity, charge, and structure of the molecule interacting with the protein. The 12 amino acids that are IASD-inaccessible are flanked on both sides by hydrophobic residues and may permit a perpendicular movement of the TMD with respect to the membrane, such that over a given time more than the 12 residues are in contact with the membrane. In such a scenario it would be expected that a partial IASD labeling of the neighboring residues occur. Because labeling with IASD is supposed to be a very fast reaction, this effect may be visible only under less severe labeling conditions. However, we never detected such partial labeling, not even with reduced IASD concentrations and very short incubation times (data not shown). Accordingly, our study clearly shows that C99 inserted into an ER-derived membrane has a domain of 12 residues (37 to 48) that is never accessible to IASDmodification and therefore represents the effective TMD of C99, which is expected to be inaccessible to cytosolic proteases.
This result is in good agreement with theoretical considerations of the expected number of residues needed to span the ER membrane as well as with previous experiments analyzing the membrane integration of APP. The thickness of cellular membranes, and thus the number of residues needed to span the membrane, increases along the secretory pathway from the ER toward the plasma membrane because of increasing cholesterol and sphingolipids in the corresponding membranes (38). A comparison of proteins typically localized at the Golgi or plasma membrane shows that the minimum number of hydrophobic residues in plasma membrane TMDs is 20 residues whereas that in Golgi TMDs is 15 (39). Because the ER membrane is thought to be even thinner than the Golgi membrane (38), less than 15 hydrophobic residues is expected to be sufficient to span ER-like membranes, which agrees well with the 12 residues of C99 that we have determined to be IASD-inaccessible. Because the bilayer thickness increases along the secretory pathway, presumably all 24 hydrophobic residues of the predicted TMD of C99 are used to span the plasma membrane (39,40).
Our results are also in agreement with previous mutagenesis studies that suggested that the TMD of APP might be shorter than predicted on its C-terminal side (18,19,28). One of these studies showed that the introduction of negatively charged aspartyl residues into the putative TMD of APP interfered with membrane integration and ␥-secretase processing, whereas aspartyl residues introduced into the C-terminal half of the putative TMD did not (19). The authors concluded that the Cterminal membrane boundary is placed around residues 46 or 47. Given that the introduction of a negatively charged residue near the membrane may easily shift the position of C99 with respect to the membrane, their finding is nearly identical to our biochemical analysis, which identifies the TMD boundary to be after residue 48.
We demonstrate in this study that in the relatively thin membrane of the ER residues 37 to 48 of C99 are shielded by the membrane. This has important implications for the mechanism of ␥-cleavage. Because the major ␥-cleavage sites after residues 40 and 42 of C99 are positioned around the middle of the IASD-inaccessible domain, they are also expected to be inaccessible to cytosolic or membrane-associated proteases, which had been considered a possibility in previous studies (18,19,28). This suggests that C99 undergoes intramembrane cleavage by a membrane protein having its active site residues within its TMD. A candidate for such a protease is the polytopic membrane protein, presenilin 1, which is required for ␥-cleavage (for review, see Refs. 2 and 12). In this context it would be interesting to know whether the aspartate residues in the putative active site of presenilin 1 are indeed located in an appropriate position. Alternatively, presenilin 1 may be an essential cofactor for ␥-secretase activity, e.g. by activating ␥-secretase or providing a channel for the addition of water during proteolysis. Moreover, this effective TMD of 12 IASDinaccessible residues is not located directly in the middle of the FIG. 5. Schematic representation of the IASD modification pattern for membrane-bound C99 mutants. Amino acids are shown in the one letter code. Residues exchanged to a cysteine in one of the 19 single cysteine mutants are indicated by color; those that were labeled by IASD are marked in gray; those, that could not be labeled, in black. Please note that the middle of the predicted transmembrane domain is positioned at ␥-cleavage site 40, whereas the middle of the TMD in microsomal membranes is positioned at ␥-cleavage site 42. stretch of 24 hydrophobic residues that were predicted as TMD of APP by hydropathy plot (37). Instead, on the N-terminal side the TMD is 8 residues shorter, whereas on the C-terminal side it is only 4 residues shorter than previously predicted. This implies that the center of the newly determined TMD is exactly between residues 42 and 43. In contrast, in the predicted 24residue-long TMD of C99, which may represent the TMD of C99 at the plasma membrane, the center is located between residues 40 and 41. This observation is striking because the center of these TMDs coincides with the C terminus of the major A␤ species found in the corresponding compartments: at the plasma membrane (and the trans Golgi network) mostly A␤ 40 is found, whereas in the ER A␤ 42 is observed almost exclusively (41). This finding is consistent with a model in which ␥-cleavage always takes place at the center of the actual transmembraneous part of C99. In agreement with this model, we and others have recently shown that in C99 mutants with an altered length of the TMD the preferred ␥-cleavage site is determined by the location of the ␥-cleavage site with respect to the hydrophobic domain (18,21).
Recently, it has been found that presenilin 1 is required not only for ␥-cleavage but also for an additional cleavage of C99 between residues 49 and 50 close to the C terminus of the predicted TMD (⑀-cleavage), leading to the release of the APP intracellular domain (42)(43)(44)(45). This C-terminal ⑀-cleavage site is at a similar position within the TMD to the so-called S3 cleavage site of the cell surface receptor Notch, which also is presenilin 1-dependent (46). The position of the ⑀-cleavage site of C99 is located just outside the IASD-inaccessible, membraneembedded TMD in microsomal membranes. Thus, it is topologically different from the ␥-cleavage site in the middle of the TMD and could be envisaged to be accessible to cytosolic or membrane-associated proteases. However, as the membrane thickness increases along the secretory pathway, the ⑀and S3 cleavage sites of C99 and Notch could be shielded by a thicker membrane typical for later compartments of the secretory pathway and thus become inaccessible to cytosolic cleavage. Nevertheless, a comparison of the newly defined TMD of APP with the TMD predicted for Notch shows that the APP intracellular domain and S3 cleavage sites are topologically more related to each other than to the ␥-cleavage site in the middle of the TMD (8). This would account for different or modified cleavage mechanisms of ␥-cleavage, on the one hand, and APP intracellular domain and S3 cleavage on the other.
In summary, our study strongly supports the existence of a real intramembrane cleavage for APP. Furthermore, it suggests a model of the mechanism of ␥-secretase that uses membrane thickness and thereby lipid composition to explain the mechanism of ␥-cleavage site selection.