Generation of Alzheimer Disease-associated Amyloid β42/43 Peptide by γ-Secretase Can Be Inhibited Directly by Modulation of Membrane Thickness*

Background: γ-Secretase-mediated intramembrane proteolysis generates Aβ42/43, pathogenic peptides implicated in causing Alzheimer disease (AD). Results: Reconstitution of γ-secretase in model membranes reveals that Aβ42/43 generation can be lowered in thick membranes. Conclusion: Membrane thickness is a crucial factor influencing the activity of γ-secretase to generate Aβ42/43. Significance: Targeting the lipid environment of γ-secretase by increasing membrane thickness may provide a therapeutic strategy for AD. Pathogenic generation of amyloid β-peptide (Aβ) by sequential cleavage of β-amyloid precursor protein (APP) by β- and γ-secretases is widely believed to causally underlie Alzheimer disease (AD). β-Secretase initially cleaves APP thereby generating a membrane-bound APP C-terminal fragment, from which γ-secretase subsequently liberates 37–43-amino acid long Aβ species. Although the latter cleavages are intramembranous and although lipid alterations have been implicated in AD, little is known of how the γ-secretase-mediated release of the various Aβ species, in particular that of the pathogenic longer variants Aβ42 and Aβ43, is affected by the lipid environment. Using a cell-free system, we have directly and systematically investigated the activity of γ-secretase reconstituted in defined model membranes of different thicknesses. We found that bilayer thickness is a critical parameter affecting both total activity as well as cleavage specificity of γ-secretase. Whereas the generation of the pathogenic Aβ42/43 species was markedly attenuated in thick membranes, that of the major and rather benign Aβ40 species was enhanced. Moreover, the increased production of Aβ42/43 by familial AD mutants of presenilin 1, the catalytic subunit of γ-secretase, could be substantially lowered in thick membranes. Our data demonstrate an effective modulation of γ-secretase activity by membrane thickness, which may provide an approach to lower the generation of the pathogenic Aβ42/43 species.

Pathogenic generation of amyloid ␤-peptide (A␤) by sequential cleavage of ␤-amyloid precursor protein (APP) by ␤and ␥-secretases is widely believed to causally underlie Alzheimer disease (AD). ␤-Secretase initially cleaves APP thereby generating a membrane-bound APP C-terminal fragment, from which ␥-secretase subsequently liberates 37-43-amino acid long A␤ species. Although the latter cleavages are intramembranous and although lipid alterations have been implicated in AD, little is known of how the ␥-secretase-mediated release of the various A␤ species, in particular that of the pathogenic longer variants A␤ 42 and A␤ 43 , is affected by the lipid environment. Using a cell-free system, we have directly and systematically investigated the activity of ␥-secretase reconstituted in defined model membranes of different thicknesses. We found that bilayer thickness is a critical parameter affecting both total activity as well as cleavage specificity of ␥-secretase. Whereas the generation of the pathogenic A␤ 42/43 species was markedly attenuated in thick membranes, that of the major and rather benign A␤ 40 species was enhanced. Moreover, the increased production of A␤ 42/43 by familial AD mutants of presenilin 1, the catalytic subunit of ␥-secretase, could be substantially lowered in thick membranes. Our data demonstrate an effective modulation of ␥-secretase activity by membrane thickness, which may provide an approach to lower the generation of the pathogenic A␤ 42/43 species.
Alzheimer disease (AD) 3 is the most common neurodegenerative disease affecting the elderly. Neurofibrillary tangles as well as amyloid plaques are invariant, major pathological hallmarks of the disease. The principal component of the amyloid plaques is the ϳ4-kDa amyloid ␤-peptide (A␤), which represents a heterogeneous mixture of hydrophobic peptides consisting of 37-43-amino acid A␤ species, A␤ 40 being the major species (1). The longer variants A␤ 42 and A␤ 43 are highly aggregation-prone. A␤ 42 , and as demonstrated recently A␤ 43 as well (2), is highly pathogenic and believed to cause AD by triggering a pathogenic series of events, the so-called amyloid cascade that ultimately leads to neurodegeneration and dementia (1,3). A␤ is generated by proteolytic processing of the ␤-amyloid precursor protein (APP), a type I membrane protein, by the combined action of ␤-secretase and ␥-secretase (4). ␤-Secretase removes the large portion of the APP ectodomain and leaves a 99-amino acid C-terminal fragment in the membrane, which is subsequently cleaved by ␥-secretase in the transmembrane domain (TMD). ␥-Secretase cleavage occurs in a stepwise manner thereby liberating the APP intracellular domain (AICD) and the various A␤ species from the membrane (5,6). The initial AICDreleasing cleavage at the ⑀-site occurs close to the cytoplasmic border of the membrane and is followed by furtherand ␥-site cuts downstream the TMD ultimately releasing the different A␤ species, which become secreted (5,6).
␥-Secretase belongs to the few unusual proteases that cleave their substrates within the core of the membrane (7). The protease is a complex composed of four integral membrane proteins, which are essential and necessary for activity (8 -10). Pre-* This work was supported by the Deutsche Forschungsgemeinschaft senilin (PS) 1 or its homolog PS2, respectively, act as catalytic subunits and mediate the intramembrane cleavages of the APP C-terminal fragment substrate (11). The potential substrate receptor nicastrin, APH-1 and PEN-2, are subunits required for stabilization and activation of PS (4,12). Although drug targeting ␥-secretase to lower A␤ generation may prove difficult as ␥-secretase has many other substrates besides APP including Notch1, a major regulator of cell differentiation in development and adulthood (13), APP-selective inhibitors and modulators of ␥-secretase have been developed that are in clinical trials for AD treatment (14).
Apart from the common sporadic forms, rare familial forms of AD (FAD) exist, which are characterized by an early disease onset and in the vast majority of the cases caused by mutations in PS1 and to a lesser extent by mutations in PS2. PS mutations cause an increase in the relative amounts of A␤ 42 and/or A␤ 43 species that are generated by ␥-secretase cleavage (15,16). Few FAD mutations have also been found in APP, most of them are located in the APP TMD at the ␥-secretase cleavage sites and increase the ratio of A␤ 42 to total A␤ (15).
Because APP processing is a membrane-associated process, alterations in the lipid membrane environment would naturally have an impact on the generation of A␤ and may thus potentially constitute a risk factor for AD. Alterations of a large variety of lipids have been documented in AD brain (17)(18)(19). However, whether and how known AD-associated lipid alterations affect APP processing and to what extent this may affect A␤ generation have remained unclear. Apart from previous studies demonstrating that amyloidogenic APP processing occurs predominantly in lipid rafts, i.e. cholesterol and sphingomyelinrich membrane microdomains (20,21), and is modulated by cholesterol (21)(22)(23)(24), our knowledge on the lipid dependence of A␤ generation has remained limited (19,25).
Reconstitution of purified membrane proteins has been an established and powerful tool to study directly the mechanism of action and structure-function relationships of a variety of integral membrane proteins including receptors, channels, carriers, as well as membrane-bound enzymes and enzyme complexes under well defined environmental conditions such as the membrane lipids in which the protein is embedded (26,27). Recent advances in the purification of the enzymes involved in APP processing and their reconstitution in liposomes now provide a platform not only to identify lipids that modulate A␤ generation but also to study directly the mechanism by which lipids might affect A␤ generation. These studies already provided a first insight into the lipid classes, which affect ␤and ␥-secretase activities directly and identified, among other lipids, cholesterol as an activity-promoting lipid of both secretases (28,29).
As has been reported for a number of membrane proteins and membrane protein complexes (30,31), alteration of basic membrane properties such as hydrophobic thickness, chain order, and the packing of the lipids of the membrane core may likely modulate the activities of the secretases, in particular that of ␥-secretase being a large intramembrane-cleaving protease complex containing 19 TMDs. To understand better the role of the lipid microenvironment of ␥-secretase for the generation of A␤, in particular that of the pathogenic A␤ 42/43 species, we set out to investigate systematically whether and how acyl chain length as well as unsaturation of phosphatidylcholine (PC) lipids would influence ␥-secretase activity and cleavage specificity in reconstituted proteoliposomes (PLs). Overall, we found that the activity of ␥-secretase is directly affected by bilayer thickness, and we show that its modulation may be exploited to lower the generation of the pathogenic A␤ 42/43 species.
Preparation and Biophysical Characterization of Lipid Vesicles-12 mg of lipid was hydrated with 1.2 ml of distilled water for 1 h at room temperature. Hydrated lipids were sonicated with a Branson sonifier in pulsed mode (duty cycle 20%, output level 3) for 1 h at temperatures above the chain-melting temperature, T m , of the lipids under a stream of argon gas. The clear suspension was centrifuged for 1 min with a table centrifuge to remove Ti particles and analyzed with dynamic light scattering (Malvern Instruments High Performance Particle Sizer) to confirm that small unilamellar vesicles (SUVs) with a diameter of about 30 nm had been formed. SUVs were diluted in assay buffer (35 mM sodium citrate, 3.5% glycerol, pH 6.4) to a final lipid concentration of 3.45 mg/ml and stored in aliquots of 200 l at Ϫ20°C until further use. For certain batches of diC14:0 PC, 15 mg of lipid was hydrated in 1.2 ml of distilled water to achieve robust activity of reconstituted ␥-secretase.
Purification and Reconstitution of ␥-Secretase into Lipid Vesicles-Native human ␥-secretase was isolated from HEK293 cells by a multistep purification protocol as described before (36). The final enzyme preparations (Q-Sepharose eluates) contained 30 -150 g/ml protein in 5% glycerol, 1 mM EDTA, 1% CHAPSO, 25 mM caproic acid, 18.75 mM BisTris, 500 mM NaCl, pH 7.0. To reconstitute the enzyme into SUVs, 200 l of the lipid vesicles were thawed, sonicated for 1 min as described above, and diluted 5 times with buffer (35 mM sodium citrate, 3.5% glycerol, 30 mM DTT, pH 6.4). Next, 1 volume of the purified enzyme was diluted with 4 volumes of the above prepared lipid vesicle suspension and incubated at 4°C as described before (36) to allow the formation of PLs. The final PL mixture contained 6 -30 g/ml protein, 0.55 mg/ml lipids, and 0.2% CHAPSO (lipid/protein ratio (w/w): 20 -100; detergent/lipid ratio (w/w): 3.6, for all lipids used). In each experiment identical aliquots of the same enzyme preparation were taken. One aliquot was always used for the POPC control, and the other aliquots were used for the lipids under investigation.
␥-Secretase in Vitro Assay-Purified recombinant C100-His 6 APP substrate was prepared as described (9). In a standard assay, C100-His 6 was added to 20 l of PLs to a final substrate concentration of 0.5 M. Following overnight incubation at 37°C, in vitro generated A␤ and AICD substrate cleavage products were analyzed by immunoblotting and quantified as described (36) by measuring the respective chemiluminescence signal intensities using the Fluorchem TM 8900 detection system (Alpha Innotech).
Tris-Bicine Urea SDS-PAGE-To analyze individual A␤ species by immunoblotting, Tris-Bicine urea SDS-PAGE was used (37), which was adjusted for the separation of short A␤ species from substrate (36) by using a 11% separation gel.
For additional procedures, see supplemental Materials and Methods.
Statistical Analysis-For each lipid, independent enzyme reconstitutions using independent ␥-secretase preparations and independent PC SUVs were analyzed. Quantitative data were collected from at least three independent assays. For some lipids, the corresponding data sets were repeatedly presented when used in more than one experimental context. Statistical significance of quantitative data were calculated by two-tailed unpaired Student's t test with Bonferroni correction for multiple testing.

RESULTS
To investigate the role of the lipid environment of ␥-secretase on APP substrate processing we made advantage of our previously established cell-free reconstitution system comprised of purified enzyme, purified substrate, and lipid vesicles (36), allowing us to test the activity of the endogenous protease isolated from HEK293 cells in various defined model membranes. Because we had previously shown that our purified CHAPSO-solubilized ␥-secretase displays robust activity and physiological cleavage specificity when reconstituted in PC vesicles (36), we based our studies on this lipid focusing on the influence of hydrophobic membrane thickness using model membranes of synthetic PC lipids with various acyl chains. POPC was used as control lipid in our studies because it represents the major phospholipid of biological membranes. Serving as general reference, a more detailed analysis of ␥-secretase reconstituted in this lipid was undertaken, demonstrating robust enzyme activity, and as shown by electron microscopy and dynamic light scattering analysis, formation of PLs with a diameter of 30 nm (supplemental Fig. 1). We further found that consistent with earlier results (29), the reconstituted enzyme required the presence of residual detergent for activity, probably to allow the release of cleavage products from the PLs and/or to facilitate substrate availability (supplemental Fig. 2).
␥-Secretase Activity and Cleavage Specificity Depend on Membrane Thickness-The hydrophobic thickness of lipid membranes is determined by many parameters including temperature, the state of the lipids (liquid-crystalline or gel phase), the distance between the head groups, the length and the amount of double bonds of the acyl chains, the angle of the acyl chains with respect of the plane of the bilayer, and the intercalation of the acyl chains (38). We started to investigate ␥-secretase activity in PLs composed of PC with two monounsaturated acyl chains of increasing length allowing us to maintain the liquid-crystalline state of the bilayers at assay temperature. As in supplemental Fig. 1A, ␥-secretase activity was generally analyzed after overnight incubation at 37°C by immunoblot analysis of the formation of the cleavage products A␤ and AICD from purified recombinant APP C100-His 6 substrate added to the PLs. To assess changes in the whole profile of the individual A␤ species generated, Tris-Bicine urea SDS-PAGE was used, which allowed us to separate shorter (A␤ [37][38][39][40] ) and longer (A␤ 42/43 ) A␤ species from each other as well as from C100-His 6 (36).
As shown in Fig. 1A, only poor ␥-secretase activity was observed with the short-chain lipids diC14:1 and di16:1, remaining below 5% of the POPC control. ␥-Secretase activity increased sharply with diC18:1 PC, reached an optimum comparable with control with diC20:1 PC, and gradually decreased down to ϳ10% with the long chain lipids diC22:1 and diC24:1 PC. 4 Thus, ␥-secretase activity depends on the length of the lipid acyl chains, with an optimum at 20 carbon atoms for monounsaturated acyl chains corresponding to a hydrophobic thickness of ϳ29 Å. We next investigated the cleavage specificity of ␥-secretase with respect to the increase in the hydrophobic thicknesses of these bilayers. Assessment of the individual A␤ species generated revealed that the relative production of A␤ 42/43 was lowered with increasing acyl chain length (Fig. 1B). Remarkably, at bilayers thicker than di20:1 PC, i.e. thicker than a POPC bilayer, the A␤ 42/43 /A␤ total ratio dropped significantly below that found for the POPC control.
To substantiate further the modulatory influence of membrane thickness on the cleavage specificity of ␥-secretase, we next analyzed saturated and monounsaturated PC in the C16 -C18 acyl chain length range. As expected from our analysis above, substantial ␥-secretase activity was observed for the diC16 lipids compared with the mixed-chain C16/C18 POPC control lipid, which was comparable with the diC18 lipid (Fig.  1C). In further support of the findings above, we found that compared with the diC16:0 membrane, the relative production of A␤ 42/43 by ␥-secretase was increased in the thinner C16:0/ C16:1 PC bilayer (Fig. 1D). Thickening the membrane by prolonging one of the acyl chains using C16:0/C18:1 PC (i.e. POPC) reverted this effect. Consistent with the data above, further thickening the bilayer using C18:0/C18:1 PC resulted in the lowest A␤ 42/43 /A␤ total ratio measured for these bilayers, lower than that produced by ␥-secretase in POPC (Fig. 1D).
Taken together, we conclude that ␥-secretase has an activity optimum for APP substrate cleavage in membranes with a hydrophobic core of 26 -29 Å. The relative production of A␤ 42/43 species decreases with increasing thickness of the 4 To allow easy assessment of A␤ generation in relation to the thicknesses of the respective model membranes, the values of the hydrophobic thickness of the membranes were graphically represented for all data sets. These values were taken from the literature using well documented experimental data or were otherwise calculated (see supplemental Table 1). bilayer, suggesting that lipid chain length molulates not only total activity of ␥-secretase, but also its cleavage specificity. ␥-Secretase Activity Is Affected by Lipid Saturation State-We next analyzed the impact of the saturation state of the lipid acyl chain on ␥-secretase activity by comparing diC18:0, C18: 0/C18:1, C18:0/C18:2, and two C18:0/polyunsaturated fatty acid PCs, i.e. lipids with an increasing degree of unsaturation in one of the acyl chains ( Fig. 2A). ␥-Secretase displayed poor activity in bilayers composed of diC18:0 PC, which might be due its gel state at the assay temperature (T m of diC18:0 PC ϭ 55°C) (38). However, adding one or two double bounds in one of the acyl chains recovered ␥-secretase activity toward the POPC control level. Arachidonic acid (C20:4 -6) reduced the activity to 50% of control, whereas docosahexaenoic acid (C22:6 -3) dropped the activity to Ͻ10%. As shown in Fig. 2B, while the A␤ 42/43 levels were not much different from POPC for most of these bilayers, the lowest productions of A␤ 42/43 were found for diC18:0 and C18:0/C18:1 PC, i.e. for the thickest bilayers of this series. Interestingly, addition of one double bond to both acyl chains (diC18:1 PC) also rescued the strongly reduced activity observed with diC18:0 PC, whereas addition of two double bonds in each acyl chain (diC18:2 PC) dropped the activity again to Ͻ10% (Fig. 2C). Consistent with the decrease in membrane thickness, A␤ 42/43 production was elevated with the latter lipid (Fig. 2D). Thus, overall, ␥-secretase activity showed a trend of an increased relative production of A␤ 42/43 as the membrane thickness was reduced in consequence of an increasing degree of acyl chain unsaturation.
DiC14:0 PC Robustly Elevates A␤ 42/43 Generation, Which Can Be Reverted by Acyl Chain Length Elongation-We also studied the influence of bilayer thickening on ␥-secretase activity in SUVs composed of PC containing two fully saturated acyl chains of increasing length. ␥-Secretase activity was strongly reduced for diC12:0 PC, increased robustly for diC14:0 and diC16:0 PC, nearly reaching the activity of the POPC control, and as shown above, dropped strongly down again for diC18:0 PC (Fig. 3A). Whereas the profile of A␤ species produced by ␥-secretase for diC16:0 PC was similar to that of the POPC  control, a striking change was observed for diC14:0 PC (Fig.  3B). In these thin bilayers, generation of the pathogenic A␤ 42/43 species was markedly increased, whereas A␤ 38 and A␤ 40 levels were reduced (Fig. 3B). As shown in supplemental Fig. 3, A␤ 43 was the predominant pathogenic A␤ species generated by ␥-secretase in this bilayer. In addition, the relative production of A␤ 42/43 was reduced as the membrane thickness increased from diC16:0 toward diC18:0 PC bilayers.
Increased A␤ 42/43 Generation by PS1 FAD Mutant ␥-Secretases Is Attenuated in Thick Membranes-Having established that membrane thickness influences ␥-secretase cleavage specificity modulating the generation of A␤ 42/43 , we next investigated whether increasing membrane thickness could attenuate the pathogenic generation of A␤ 42/43 by FAD mutants within PS1, the catalytic subunit of ␥-secretase. To this end, ␥-secretase was purified from HEK293 cells stably expressing WT or FAD mutant PS1 and reconstituted in diC20:1-diC24:1 PC bilayers. FAD mutants of moderate (PS1 ⌬exon9) and strong (PS1 L166P) pathogenic activity (35) were chosen for analysis and changes in cleavage specificity were compared with WT PS1. Both FAD mutants showed the expected increases of A␤ 42/43 compared with the WT enzyme in POPC control conditions (Fig. 4). Comparable with their behavior in cultured cells, PS1 ⌬exon9 produced roughly similar amounts of A␤ 42/43 as A␤ 40 , whereas the PS1 L166P mutant generated substantially more A␤ 42/43 than A␤ 40 . In agreement with recent results (39), these FAD mutants generated higher amounts of A␤ 43 than A␤ 42 in the cell free assay system (supplemental Fig. 4). Increasing the lipid chain length caused a reduction of total activity for the PS1 ⌬exon9 mutant, which was accompanied by a change of the A␤ 42/43 /A␤ total ratios. In diC22:1 PC and diC24:1 PC lipids, A␤ 42/43 levels were lower than that of A␤ 40 . Even for the strong PS1 L166P mutant, changes in the A␤ 42/43 /A␤ total ratios became apparent, which were most prominent in the diC24:1 PC bilayer. In this bilayer, a clear reduction of A␤ 42/43 was observed. Thus, the pathogenic activity of PS1 FAD mutants is attenuated in thick membranes and the increased production of A␤ 42/43 can be substantially attenuated under such conditions.

DISCUSSION
Despite being both an intramembrane-cleaving protease and a key AD drug target, only few studies have addressed the role of lipids for ␥-secretase with respect to its activity so far (29, 40 -42). Studies by Fraering et al. (40) and others (41,42) showed that the purified detergent-solubilized protease is inactive unless it is reconstituted with membrane lipids. Investigating the influence of mixtures of PC with sphingolipids, or glyc- erophospholipids containing different head groups, on the activity of ␥-secretase, Osenkowski et al. (29) could further show that while the latter had little effect, sphingolipids could stimulate ␥-secretase activity, which was boosted by soluble cholesterol.
Building on these initial studies, we systematically studied the role of membrane thickness and lipid chain order by reconstituting the purified enzyme into defined model membranes made of synthetic PC lipids with acyl chains of various length and saturation thereby allowing us to identify the optima for total activity and cleavage specificity of ␥-secretase. ␥-Secretase displayed robust (i.e. Ͼ75% compared with POPC) activity in membranes with a hydrophobic thickness of about 26 -29 Å, almost regardless of whether PLs composed of C18:0/C18:1, C18:0/C18:2, diC18:1, or diC20:1 were used. The only exception was diC18:0 PC, a lipid in which ␥-secretase displayed very low activity. It should be noted, however, that diC18:0 PC is the only lipid that we used which is in the rigid gel state at 37°C. We thus conclude that a requirement for robust ␥-secretase activity is that the lipid is in the liquid-crystalline state. Interestingly, lipids with polyunsaturated fatty acids, which are common in the brain, were not beneficial. Also adding double bonds in both acyl chains was not beneficial for its activity. Thus, besides hydrophobic thickness being an important parameter for robust ␥-secretase activity, too many double bonds in the acyl chains of the hydrophobic core of the bilayer reduces ␥-secretase activity. ␥-Secretase displayed better activity when lipids were used with one saturated and one unsaturated acyl chain. POPC (C16:0/C18:1 PC) has one relatively stiff unsaturated and one more mobile saturated acyl chain (43) and apparently builds a bilayer with optimal hydrophobic thickness and packing of the lipids allowing a conformation of ␥-secretase with optimal activity.
An important observation of this study was that the A␤ 42/43 / A␤ total ratio profile observed for the various bilayers did not correlate with that of the corresponding total ␥-secretase activity. Although there was an optimum membrane thickness for total A␤ production for POPC PLs, the tendency with regard to the A␤ 42/43 /A␤ total ratio was that it decreased with increasing thickness of the bilayer. This was even observed for bilayers in which total activity of ␥-secretase was very low. Interestingly, a decrease of A␤ 42/43 species was typically accompanied by an increase of the shorter A␤ 40 and/or A␤ 38 species, consistent with the sequential-cleavage model (5).
To distinguish the effects of membrane thickness from changes in lipid membrane fluidity on the cleavage specificity of ␥-secretase, we used classes of phospholipids with similar membrane fluidity and lipid packing. Thus, we used PC with identical saturated acyl chains, saturated acyl chains with mismatching chain lengths, identical acyl chains having one cis double bond, and one saturated and one monounsaturated acyl chain. Within each class we found the same trend with regard to a reduced relative production of A␤ 42/43 for thicker membranes, justifying our conclusion that this effect is due to changes in thickness and not due to changes in fluidity. This is further illustrated in Fig. 5, which gives a summary of ␥-secretase activity and cleavage specificity with respect to membrane thickness of in all used lipids in which the activity was Ͼ5% compared with the POPC control.
The highest A␤ 42/43 /A␤ total ratio was found in diC14:0 PC bilayers. The A␤ 42/43 /A␤ total ratio for diC14:0 PC PLs was about two times higher compared with the POPC control, similar to ratios found for weak FAD mutants. The high A␤ 42/43 / A␤ total ratios found for diC14:0 PC PLs could be reverted to the values observed for the POPC control by increasing one of the acyl chains of the lipid. In these experiments, the total ␥-secretase activity hardly changed whereas the A␤ 42/43 /A␤ total ratios dramatically decreased. However, A␤ 42/43 /A␤ total ratios were not as much elevated as in diC14:0 PC in other thin bilayers of FIGURE 5. Summary of the dependence of ␥-secretase activity and cleavage specificity on membrane thickness. Data of Figs. 1-3 were combined to represent all lipids in which the activity was higher than 5% compared with the POPC control. Total A␤ generation by ␥-secretase in each lipid relative to the control POPC PLs is shown in the top panel. Activity is low for very thin and very thick membranes. Optimal activity is achieved for membranes with a thickness between 26 and 29 Å. Note that the activity varies for lipids with similar thickness probably due to differences in fluidity and lipid packing. Particularly lipids with one saturated and one unsaturated acyl chain have optimal activity. Relative A␤ 42/43 generation is shown in the middle panel. A␤ 42/43 /A␤ total ratios decrease with increasing membrane thickness, irrespective to variations of total ␥-secretase activity. The hydrophobic membrane thickness of the bilayers comprising the corresponding lipids is shown in the bottom panel. f, diC18:0 PC is the only lipid investigated in our study, which is in the rigid gel state at 37 º C. ᭜, estimation of the hydrophobic thickness of a bilayer composed of C18:0/C20:4 PC might be inaccurate due to the large mismatch of the C18:0 and C20:4 acyl chains (see supplemental Table 1).
comparable thickness, such as diC18:2, or C16:0/C16:1 PC. This suggests that for diC14:0 PC, in addition to its property of forming a thin membrane bilayer, its two short saturated chains may lack the mobility to adjust to hydrophobic domains of the ␥-secretase leading to a pathogenic change of the cleavage specificity of ␥-secretase.
Because ␥-secretase generated lower absolute and relative amounts of A␤ 42/43 when reconstituted in thicker membranes, we finally asked whether the same also applied for PS FAD mutant enzymes. As expected, when PS1 FAD mutant ␥-secretase complexes were reconstituted in POPC vesicles, the A␤ 42/ 43/A␤ total ratios were elevated compared with the WT PS1 ␥-secretase control. Strikingly, A␤ 42/43 /A␤ total ratios were strongly attenuated when FAD mutants were reconstituted in thick membranes. Absolute A␤ 42/43 production was dramatically reduced for these PLs showing that the pathogenic activity of FAD mutant ␥-secretase is strongly influenced by the thickness of its surrounding lipid bilayer and attenuated in thicker membranes. This was shown for both the PS1 ⌬exon9 and PS1 L166P FAD mutants, which display a moderate-strong elevation of A␤ 42/43 generation, respectively.
How does the lipid environment affect the activity of ␥-secretase? Lipid properties affecting activities of integral membrane proteins include the head group, the chain order, the lipid packing, and the thickness of the hydrophobic part of the membrane. Particularly the extent of hydrophobic matching between the hydrophobic thickness of the bilayer and the length of the membrane-spanning part of the protein determines the amino acid composition, packing, and tilting of the TMDs of the protein thereby also influencing enzyme activity (30,31). Interestingly, with respect to ␥-secretase, changes of the tilt and elongation of the helical TMDs in bilayers of different thickness have recently been revealed by molecular dynamics simulations of the PS1 C-terminal fragment (44).
Altered hydrophobic matching between the bilayer and the TMDs of ␥-secretase and/or of the substrate TMD and its surrounding membrane lipids might both influence the positioning of the substrate cleavage sites relative to the active site. As a result, an alteration of membrane thickness, which increases from early to late compartments of the secretory pathway (45)(46)(47), will directly affect hydrophobic matching and thereby not only influence total ␥-secretase activity but also the ratios of A␤ 40 versus the pathogenic A␤ 42 and A␤ 43 species generated. As shown in our study, the exposure of the ␥ 42/43 cleavage sites toward the catalytic site is apparently increased in thin membranes leading to an enhanced generation of A␤ 42/43 , whereas in thicker membranes the ␥ 40 cleavage site is more exposed to the active site. Our data may thus also provide a direct explanation for the previously observed preferential generation of A␤ 42 in the early compartments of the secretory pathway (48,49).
Taken together, using defined model membranes, we have shown that ␥-secretase activity and cleavage specificity toward its substrate APP are modulated by membrane thickness including inhibition of total A␤ generation and/or that of the pathogenic A␤ 42/43 species. Our findings thus suggest that alteration of membrane thickness to lower the generation of the pathogenic A␤ 42/43 species by ␥-secretase could represent a target for potential therapeutic intervention of AD.