Cleavage of amyloid precursor protein by an archaealpresenilin homologue PSHShangyu Danga,b,1, Shenjie Wua,b,1, Jiawei Wangc, Hongbo Lia,b, Min Huangd, Wei Hea,b, Yue-Ming Lie,Catherine C. L. Wongd, and Yigong Shia,b,2

aMinistry of Education Key Laboratory of Protein Science, bTsinghua-Peking Joint Center for Life Sciences, Center for Structural Biology, and cState KeyLaboratory of Bio-membrane and Membrane Biotechnology, School of Life Sciences and School of Medicine, Tsinghua University, Beijing 100084, China;dNational Center for Protein Science Shanghai, Institute of Biochemistry and Cell Biology, Shanghai Institutes of Biological Sciences, Chinese Academy ofSciences, Shanghai 200031, China; and eMolecular Pharmacology & Chemistry Program, Memorial Sloan–Kettering Cancer Center, New York, NY 10065

Contributed by Yigong Shi, February 10, 2015 (sent for review January 15, 2015)

Aberrant cleavage of amyloid precursor protein (APP) by γ-secretasecontributes to the development of Alzheimer’s disease. More than200 disease-derived mutations have been identified in presenilin (thecatalytic subunit of γ-secretase), making modulation of γ-secretaseactivity a potentially attractive therapeutic opportunity. Unfortunately,the technical challenges in dealing with intact γ-secretase have hin-dered discovery of modulators and demand a convenient substituteapproach. Here we report that, similar to γ-secretase, the archaealpresenilin homolog PSH faithfully processes the substrate APP C99 intoAβ42, Aβ40, and Aβ38. The molar ratio of the cleavage products Aβ42over Aβ40 by PSH is nearly identical to that by γ-secretase. The pro-teolytic activity of PSH is specifically suppressed by presenilin-specificinhibitors. Known modulators of γ-secretase also modulate PSH simi-larly in terms of the Aβ42/Aβ40 ratio. Structural analysis reveals asso-ciation of a known γ-secretase inhibitor with PSH between its twocatalytic aspartate residues. These findings identify PSH as a surrogateprotease for the screening of agents that may regulate the proteaseactivity and the cleavage preference of γ-secretase.

intramembrane protease | Alzheimer’s disease | PSH | gamma-secretase |beta-amyloid peptide

Amyloid precursor protein (APP) is initially cleaved byβ-secretase in the extracellular space, producing a mem-

brane-tethered, 99-residue carboxyl-terminal fragment known asAPP C99 (1). APP C99 then undergoes sequential cleavages byγ-secretase, first at the e sites, yielding Aβ49/Aβ48, and eventu-ally at the γ-sites, generating Aβ42/Aβ40/Aβ38 (2–4) (Fig. 1A).The 230-kDa γ-secretase contains four components: presenilin(PS), Pen-2, Aph-1, and nicastrin (NCT), of which PS1 is thetarget of most mutations derived from early-onset familial Alz-heimer’s disease (FAD) patients. Rather than abolishing theprotease activity of γ-secretase, these mutations are thought toincrease the molar ratio of Aβ42 over Aβ40 (2).Among the cleavage products of γ-secretase, Aβ42 is partic-

ularly prone to aggregation, leading to formation of β-amyloidplaque in the brain and presumably causing Alzheimer’s disease(3). Other FAD-derived mutations map to APP and PS2, lendingsupport to the causal relationship between formation of β-amyloidplaque and Alzheimer’s disease (5). Therapeutic intervention ofAlzheimer’s disease may directly benefit from in vitro investigationof γ-secretase and discovery of its potential modulators (6). Un-fortunately, such effort has been hampered by the difficulty inexpression, purification, and manipulation of the complex protease.This problem persists despite recent breakthroughs in structuralelucidation of γ-secretase and nicastrin (7, 8).The intramembrane aspartate protease PSH from the

archaeon Methanoculleus marisnigri JR1 shares 19% sequenceidentity and 53% sequence similarity with human PS1 (hPS1).The signature motifs for catalysis, ΦYDΦΦ (Φ for a hydro-phobic residue) on transmembrane segment 6 (TM6) andΦGΦGD on TM7, are identical between PSH and hPS1. PSHcan be readily overexpressed in Escherichia coli and purified in

large quantity (9, 10). Importantly, the crystal structure ofPSH is available and offers a valuable guide for the design andimprovement of protease modulators (10). Therefore, PSHmay represent an attractive surrogate of γ-secretase for mod-ulator screening if PSH cleaves APP C99 and such cleavagesrecapitulate those by the intact γ-secretase. Notably, due to thepresence of two different presenilin variants (hPS1 and hPS2)and two Aph-1 variants (Aph-1A and Aph-1B), there are atleast four distinct human γ-secretase complexes, which exhibitdifferent signature Aβ profiles (11). In this study, we comparePSH to the most prevalent γ-secretase complex comprising hPS1and Aph-1A.

ResultsCleavage of APP C99 by PSH. We reconstituted an in vitro bulkcleavage assay, using recombinant PSH as the protease and APPC99 as the substrate. To facilitate purification and detection,APP C99 was fused to a maltose-binding protein (MBP) at theamino terminus and tagged by an octa-histidine (8xHis) at thecarboxyl terminus (Fig. 1A). MBP can be conveniently releasedthrough thrombin cleavage. The MBP-tagged APP C99 serves asthe substrate for PSH throughout this study. Incubation of thissubstrate with PSH at 37 °C resulted in cleavage of APP C99 intoa large amino-terminal fragment and a small carboxyl-terminalpeptide, as judged by SDS-PAGE analysis (Fig. 1B, lanes 1–3).

Significance

Amyloid precursor protein (APP) is cleaved by β-secretaseto produce APP C99, which undergoes additional, sequentialcleavages by γ-secretase to generate amyloid-β peptides in-cluding Aβ40 and Aβ42. Increased ratios of Aβ42 over Aβ40 arethought to cause Alzheimer’s disease. Screening of γ-secretasemodulators is hindered by the technical challenges in expres-sion and biochemical manipulation of γ-secretase. In this study,we demonstrate that the archaeal intramembrane proteasePSH represents an excellent surrogate of γ-secretase in terms ofcleavage of APP C99, ratio of Aβ42 over Aβ40, and modulationof cleavage preferences by known modulators of γ-secretase.Our finding may facilitate discovery of γ-secretase inhibitorsand modulators.

Author contributions: S.D., S.W., J.W., H.L., W.H., Y.-M.L., C.C.L.W., and Y.S. designedresearch; S.D., S.W., J.W., H.L., M.H., and C.C.L.W. performed research; S.D., S.W., J.W.,H.L., W.H., Y.-M.L., and C.C.L.W. contributed new reagents/analytic tools; S.D., S.W., J.W.,H.L., M.H., W.H., Y.-M.L., C.C.L.W., and Y.S. analyzed data; and Y.S. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.pdb.org (PDB ID code 4Y6K).1S.D. and S.W. contributed equally to this work.2To whom correspondence should be addressed. Email: shi-lab@tsinghua.edu.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1502150112/-/DCSupplemental.

3344–3349 | PNAS | March 17, 2015 | vol. 112 | no. 11 www.pnas.org/cgi/doi/10.1073/pnas.1502150112

Dow

nloa

ded

by g

uest

on

Janu

ary

25, 2

020

Addition of the γ-secretase–specific inhibitor III-31C (12) led toconcentration-dependent inhibition of the protease activity ofPSH (Fig. 1B, lanes 4–14). Quantitation of the results revealedan inhibitory constant (IC50) of ∼10 μM for III-31C (Fig. 1C).The band that corresponds to the small carboxyl-terminal

peptide of APP C99 was excised from the SDS-PAGE gel,eluted, and subjected to amino-terminal peptide sequencing.This analysis revealed NH2-Leu-Val-Met-Leu-Lys-COOH as thefirst five amino acids of the major peptide species and NH2-Val-Met-Leu-Lys-Lys-COOH as the first five residues of the minorspecies (Fig. 1D). These sequences exactly match those in APPC99 and identify the first cleavage sites by PSH to be at thecarboxyl terminus of Thr48 and Leu49, coinciding with the e sitesby γ-secretase.Next, we subjected the substrate APP C99 to overnight in-

cubation with PSH, which led to more complete cleavage of APPC99. The carboxyl-terminal peptides of the digested productsand the uncleaved substrates were removed by Ni2+-NTA resin;the remaining fragments were liberated through thrombin cleav-age and the MBP moiety was removed by the maltose-bindingresin (Fig. S1A). Analysis of the remaining fragments by massspectrometry (MS) identified two major cleavage products, whosemolecular weights perfectly matched those of Aβ38 and Aβ40(Fig. S1B). This result demonstrates that, similar to γ-secretase,PSH cleaves APP C99 into Aβ38 and Aβ40. Our MS results alsoidentified Aβ42 and suggested Aβ40 to be a more abundantcleavage product compared with Aβ42 (Fig. S1C).

Cleavage Preference of PSH. To detect the less abundant cleavageproducts of PSH, we used an antibody-based, spectroscopicAlphaLISA assay (13–15). In this assay, the presence of a specificAβ peptide brings the donor and acceptor beads to close

proximity, triggering the transfer of an oxygen molecule from thedonor to the acceptor and consequent detection of emission lightfrom the acceptor (Fig. 2A). Using this sensitive method, wefound that the cleavage products of MBP-tagged APP C99 by3 μM PSH not only included Aβ40, but also Aβ42 (Fig. 2B, Left).The amount of Aβ40 is ∼4.8-fold more than Aβ42.In our hands, human γ-secretase exhibited a low level of

protease activity toward the MBP-tagged APP C99, which mightbe explained by a compromised ability for nicastrin to recognizethe amino terminus of the MBP-tagged substrate (16–19). Toovercome this problem, we expressed and purified the substrateAPP C99 with its native amino terminus exposed and its carboxylterminus attached to a poly-His tag. In sharp contrast to theMBP-tagged substrate, the amino terminus exposed APP C99was readily cleaved by 1 μM γ-secretase in the AlphaLISA assay,with considerably more Aβ40 than Aβ42 (Fig. 2B, Right). Im-portantly, the relative ratio of Aβ42 over Aβ40 by 3 μM PSH(0.1 mg/mL) is similar to that by 1 μM γ-secretase, further validatingPSH as a surrogate protease of γ-secretase. The Aβ42/Aβ40 ratiodecreased slightly with increasing concentrations of PSH (Fig.2C). Production of Aβ40 and Aβ42 by both PSH and γ-secretasewas also confirmed by Western blots using monoclonal anti-bodies against these peptides (Fig. S2).The FAD-derived mutation N135D in PS1 (N141I in PS2) was

reported to increase the relative ratio of Aβ42 over Aβ40 (20–22). We confirmed this result by quantifying the cleavage prod-ucts of wild-type (WT) γ-secretase and the PS1-N135D–containingγ-secretase (Fig. 2B, Right). Compared with WT PS1, the mainchange associated with PS1-N135D was a 93% reduction of theAβ40 level, although the Aβ42 level was also reduced by ∼52%.Next, we generated the corresponding mutation N46D in PSH,purified the recombinant protein PSH-N46D to homogeneity, and

Fig. 1. Cleavage of the amyloid precursor protein (APP) C99 fragment by the archaeal presenilin homolog PSH. (A) A schematic diagram of the APP C99substrate. The substrate is tagged by a maltose-binding protein (MBP) at the amino terminus and eight histidine residues (8xHis) at the carboxyl terminus.Some of the known γ-secretase cleavage sites are indicated. (B) APP C99 is cleaved by PSH, and the cleavage is inhibited by the γ-secretase–specific inhibitorIII-31C. Cleavage of APP C99 produced a 50-kDa amino-terminal fragment and a 10-kDa carboxyl-terminal fragment. Shown here is a SDS-PAGE stained byCoomassie blue. (C) Quantitation of III-31C–mediated inhibition of PSH cleavage reveals an inhibitory constant of ∼10 μM. (D) The carboxyl-terminal cleavagefragment of APP C99 was analyzed by amino-terminal peptide sequencing. The results revealed a major species, LVMLK, and a minor species, VMLKK. Thus,the first cleavages in APP C99 occur either between Thr48 and Leu49, producing Aβ48, or between Leu49 and Val50, generating Aβ49.

Dang et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3345

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

25, 2

020

examined its cleavage preferences by using the AlphaLISAassay. The result showed that, similar to γ-secretase, the mutationN46D in PSH markedly compromised its ability to generate Aβ40,resulting in an increased ratio of Aβ42 over Aβ40 (Fig. 2B, Left).These observations identify PSH as a faithful surrogate of γ-sec-retase in terms of the impact on cleavage preferences by themutation N46D.

Modulation of PSH Cleavages. The increased ratio of Aβ42 overAβ40 is thought to have a causal link with Alzheimer’s disease(23). Considerable effort has been dedicated to the discovery ofmodulators that, upon binding to γ-secretase, may decrease theAβ42/Aβ40 ratio. The small molecules GSM-1 and E2012 areknown to achieve this effect by inhibiting the production of Aβ42(24–27) (Fig. S3). This conclusion was confirmed by our analysisusing the AlphaLISA assay and the amino terminus exposedAPP C99 as the substrate (Fig. 3 A and B). For γ-secretase, thepresence of 125 and 250 μM GSM-1 led to marked reduction ofthe Aβ42 levels by ∼53% and ∼68%, respectively; production of

Aβ40 was also decreased by 26% and 37% (Fig. 3A). Conse-quently, the Aβ42/Aβ40 ratio by γ-secretase was decreased from1.0 to 0.65 ± 0.11 and 0.52 ± 0.08, respectively, in the presence of125 and 250 μM GSM-1 (Fig. 3A). In contrast, the decreasedAβ42/Aβ40 ratio by the modulator E2012 was mainly caused byincreased levels of Aβ40 and, to a lesser extent, by decreasedproduction of Aβ42 (Fig. 3B).Next, we investigated the impact of these modulators on the

cleavage preferences of PSH. The Aβ42/Aβ40 ratio was slightlyreduced in the presence of GSM-1 (Fig. 3C) but sharply de-creased in the presence of E2012 (Fig. 3D). The decreased Aβ42/Aβ40 ratio by E2012 was contributed by higher levels of Aβ40and lower levels of Aβ42 (Fig. 3D). These results indicate thatthe modulators GSM-1 and E2012 have a generally similar effecton γ-secretase and PSH.

Impact of Mutations on PSH Cleavage Activity. Our experimentalfindings predict that, in addition to N46D (Fig. 2B, Left), othermutations in PSH that correspond to FAD-derived mutations inhPS1 may also have a similar consequence on the cleavagepreference of PSH compared with γ-secretase. We examined thisprediction. The high degree of sequence homology between PSHand hPS1 allowed identification of a number of FAD-derivedmutations in hPS1 that affect invariant or conserved residuesbetween PSH and hPS1 (Fig. S4A). We chose three such rep-resentative hPS1 mutations: L85P (28), F237I (29), and L420R(30), which correspond to L17P, F142I, and V261R, respectively,in PSH. These three PSH variants were individually expressed,purified, and examined for their cleavage preferences.Both WT PSH and the variants were able to generate the

Aβ42 and Aβ40 cleavage fragments, as judged by Western blotsusing monoclonal antibodies against these peptides (Fig. S4B).To accurately determine changes of the Aβ42/Aβ40 ratio, weused the AlphaLISA assay. Compared with WT PSH, all threevariants exhibited varying degrees of compromised ability to gen-erate Aβ40 (Fig. 4A). The PSH variants F142I and V261R pro-duced more Aβ42, whereas the variant L17P exhibited a decreasedability to generate Aβ42. In all three cases, however, the Aβ42/Aβ40 ratio is elevated compared with WT PSH (Fig. 4B). Thesefindings further corroborate the notion that PSH is a bona fidesurrogate for γ-secretase in terms of substrate cleavage preference.

Structure of PSH Bound to an Inhibitor. In addition to III-31C, thereare a large number of other γ-secretase–specific inhibitors. Giventhe conserved biochemical properties between γ-secretase andPSH, we speculated that such inhibitors may also specifically tar-get PSH. Supporting this notion, two such inhibitors L-682,679(referred to as L679 hereafter) and JC5 (31), which share a similarchemical structure (Fig. S3), were able to inhibit the proteaseactivity of PSH in a concentration-dependent manner (Fig. 5A).L679 at 0.2 mM inhibited ∼50% cleavage activity of PSH towardthe substrate APP C99, whereas 1 mM L679 abrogated the activity(Fig. 5A, lanes 2–5). In contrast, 0.2 mM JC-5 inhibited nearly90% of PSH cleavage activity (Fig. 5A, lane 7).To understand the mechanism of inhibition, we sought to de-

termine the crystal structure of PSH bound to these small-mole-cule inhibitors. After numerous attempts, we crystallized PSHbound to L679 and solved the structure at 3.85 Å resolution (Fig.5B and Table S1). The presence of L679 was unambiguously de-termined by the Fo−Fc electron density that is located betweenTM6 and TM7. Determination of the relative orientation of theinhibitor was facilitated by the Br derivative of JC-5 (Fig. 5C,Fig. S5, and Table S1). The moderate resolution of 3.85 Å and thehighly anisotropic diffraction gave rise to electron density mapsthat are insufficient for accurate assignment of atomic interactionsbetween L679 and specific residues in PSH. Nonetheless, themechanism of inhibition is clearly suggested by the currentstructure: The two catalytic residues of PSH, Asp162 and Asp220,

Fig. 2. PSH produces a similar ratio of Aβ42 over Aβ40 as γ-secretase. (A) Aschematic diagram of the AlphaLISA assay. A monoclonal antibody attachedto the donor bead recognizes a common region in Aβ peptide. Upon bindingof the specific Aβ peptide by a second monoclonal antibody attached to theacceptor bead, an oxygen molecule is generated in the donor bead by a680-nm excitation wavelength and transferred to the acceptor bead, trig-gering an emission wavelength of 615 nm. The emission signal is detectedspectroscopically. (B) Production of Aβ40 and Aβ42 by PSH (Left) and γ-secretase(Right). Shown here are results of AlphaLISA assay. The spectroscopic read-ing of Aβ40 by wild-type (WT) γ-secretase is normalized as 1.0. Asn46 in PSHcorresponds to Asn135 in hPS1. (C) The ratio of Aβ42 over Aβ40 by PSH isclose to that by γ-secretase. The Aβ42/Aβ40 ratio by γ-secretase is normalizedas 1.0. The Aβ42/Aβ40 ratio by PSH deceases slowly with increasing concen-trations of PSH.

3346 | www.pnas.org/cgi/doi/10.1073/pnas.1502150112 Dang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

25, 2

020

are separated by a chunky electron density from L679, which wasassigned to a phenol group from the amide end of L679 (Fig.5B). This binding mode explains how L679 inhibits the proteaseactivity of PSH and suggests a similar mechanism for hPS1.Notably, no significant conformational changes were observedbetween free PSH and L679-bound PSH.

DiscussionIn this study, we demonstrate that the enzymatic properties ofPSH are strikingly similar to those of γ-secretase. PSH cleavesAPP C99 first at the e sites, producing Aβ48 and Aβ49 (Fig. 1),and then successively at the γ-sites, generating Aβ42, Aβ40, andAβ38 (Fig. 2 and Fig. S1). Importantly, PSH preferentially pro-duces Aβ40 over Aβ42, with a nearly identical ratio of Aβ42/Aβ40 compared with that by γ-secretase (Fig. 2). Despite their

chemical divergence, the γ-secretase modulators GSM-1 andE2012 exhibit a generally similar effect on PSH, decreasing theAβ42/Aβ40 ratio (Fig. 3). The detailed mechanisms by GSM-1and E2012 may differ for PSH versus γ-secretase. For example,GSM-1 decreases the production of Aβ42 by γ-secretase, but notby PSH; in contrast, E2012 markedly reduces Aβ42 productionby PSH, but not by γ-secretase (Fig. 3). Mutations in PSH thatcorrespond to FAD-derived mutations in hPS1 have a similarconsequence on the cleavage preference by PSH (Fig. 4). Together,these findings identify PSH as a surrogate for the investigation ofγ-secretase. PSH recapitulates nearly all enzymatic behaviors ofγ-secretase, including the Aβ42/Aβ40 ratio and its modulation byknown modulators (27, 32). These biochemical mimicries stronglysuggest a highly conserved active site arrangement and a similarpattern of substrate recognition between PSH and hPS1.

Fig. 3. Modulators of γ-secretase exhibit a similar effect on the cleavage preferences of PSH. (A) The γ-secretase modulator GSM-1 reduces the Aβ42/Aβ40ratio by γ-secretase. The production of both Aβ40 and Aβ42 was reduced in the presence of 125 and 250 μM GSM-1, although the extent of reduction wasmore pronounced for Aβ42 than for Aβ40. Shown here are results of AlphaLISA assay. (B) The γ-secretase modulator E2012 reduces the Aβ42/Aβ40 ratioby γ-secretase. The production of Aβ40 was increased slighted by 500 and 1,000 μM E2012, whereas Aβ42 was decreased. (C) GSM-1 slightly reduces theAβ42/Aβ40 ratio by PSH. At 125 and 250 μM GSM-1, the reduction is less pronounced for PSH than for γ-secretase. (D) E2012 drastically reduces the Aβ42/Aβ40ratio by PSH. At 500 and 1,000 μM E2012, the reduction is more drastic for PSH than for γ-secretase.

Fig. 4. Alteration of the Aβ42/Aβ40 ratio by PSH variants. (A) Quantitation of the cleavage products Aβ42 and Aβ40 by three PSH variants. The AlphaLISAreading of Aβ40 released by WT PSH was normalized as 1.0. (B) The Aβ42/Aβ40 ratio is increased for each of the three PSH variants.

Dang et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3347

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

25, 2

020

Despite generally similar biochemical properties, γ-secretaseand PSH exhibit subtle, but important differences. For example,the inhibitor III-31C is considerably more potent toward γ-sec-retase than for PSH (12) (Fig. 1C), which likely reflects the fineconformational differences between γ-secretase and PSH. In ad-dition, PSH is relatively insensitive to GSM-1 and only becomessensitive to E2012 at high concentrations of 500–1,000 μM (Fig. 3).By contrast, the cleavage preference of γ-secretase can be modu-lated by either GSM-1 or E2012 at 30–60 μM. We suggest thatPSH may be made more γ-secretase–like through dedicated pro-tein engineering effort in which select sets of amino acids in PSHcan be replaced by those in hPS1. Such protein engineering effortis likely facilitated by the crystal structure of PSH (10), and by theatomic structure of hPS1 in the future.It should be noted that the human γ-secretase comes in four

different forms, each with a different Aβ cleavage profile (11).hPS2 appears to reduce the endopeptidase activity of γ-secretase(e sites) compared with hPS1, and Aph-1 variants affect thecarboxypeptidase-like activities (γ-sites) (11). In this study, wedemonstrate that the enzymatic properties of PSH closely re-semble those of the most abundant γ-secretase complex thatcomprises hPS1, Pen-2, Aph-1A, and nicastrin. It remains to beinvestigated to what extent PSH can be compared with the otherthree distinct γ-secretase complexes. Regardless of the outcomeof these investigations, the use of a single protease PSH as asurrogate no longer allows assessment of the impact on Aβ gen-eration by two related γ-secretases, each with a different variantsubunit (such as Aph-1A versus Aph-1B) (11).In this study, we used five different methods for the detection

of APP C99 cleavage by PSH and γ-secretase. Direct visualiza-tion of cleavage products by SDS-PAGE (Figs. 1B and 5A),

amino-terminal peptide sequencing analysis (Fig. 1D), and theconventional Western blots based assay (Figs. S2 and S4B) offera qualitative assessment. In contrast, mass spectrometry (Fig. S1)and the AlphaLISA assay (Figs. 2 B and C, 3, and 4 A and B)tend to be more accurate and quantitative. Notably, in all cases,the cleavage reactions were carried out in detergent micelles, asopposed to lipid bilayers or liposomes. Published results showlittle difference for reactions performed in detergent micellesversus lipid bilayers (33).Although Aβ42 and the ratio of Aβ42/Aβ40 are treated as the

main culprits of Alzheimer’s disease, the real situation is con-siderably more complex. For example, Aβ43 was reported toexhibit more potent amyloidogenicity and pathogenicity (34).Existing data suggest that absolute levels of Aβ42 are not asimportant as the altered ratios of the Aβ peptides, most notablythe ratio of Aβ42/Aβ40 (22, 33, 35). These considerations havesignificant ramifications on the discovery of novel drugs that maytarget mutant γ-secretases. Nevertheless, the ease of expression,purification, and biochemical manipulation of PSH may greatlyfacilitate discovery of its enzymatic inhibitors and modulatorsthat may prove to have a similar effect on the intact γ-secretase.Such small-molecule inhibitors and modulators may not justserve as investigative tools for γ-secretase but also find potentialtherapeutic applications for Alzheimer’s disease.

Materials and MethodsPreparation of PSH and γ-Secretase Variants. All PSH variants were derivedfrom the PSH construct that was crystallized (10). The cloning and purifica-tion of PSH variants were performed as described (10). The variants weredesalted on a Hitrap desalting column (GE Healthcare) to PBS buffer, pH 7.4,with 0.02% (wt/vol) n-dodecyl-β-D-maltopyranoside (DDM; Anatrace). FAD-derived hPS1 mutations are from www.alzforum.org/mutations. The WT andthe N135D variant of γ-secretase were generated as described (7).

Activity Assay for PSH Variants and γ-Secretase. APP C99 was fused with anN-terminal MBP tag and a C-terminal 8xHis tag. A unique thrombin cleavagesite (LVPR/GS) was introduced between MBP and APP C99. The final constructwas cloned into pET-21b (Novagen), overexpressed in E. coli strain BL21(DE3),and purified. The PSH variants (0.1 mg·ml−1) were mixed with this substrate(1 mg·ml−1) in PBS, in the presence of 50 mM citrate, pH 5.1, and 0.02% DDM.APP C99 with a C-terminal 6xHis tag was used as the substrate for γ-secre-tase. In the reaction, γ-secretase variants (0.15 mg·mL−1) were mixed withAPP C99 (0.5 mg·mL−1) in 0.25% CHAPSO, 0.1% (wt/vol) phosphatidylcho-line, 0.025% (wt/vol) phosphatidylethanolamine, 25 mM Hepes, pH 7.4, and150 mM NaCl. The reaction was conducted at 37 °C for 16 h and stopped bySDS sample buffer. Inhibitor or modulator diluted by DMSO was added atindicated concentrations. The same volume of DMSO was added as a nega-tive control for each batch of assay.

Western Blotting. For detection of Aβ40 and Aβ42, the reaction samples wereapplied onto 16% SDS-PAGE and transferred to PVDF membranes. Mem-branes were blocked by 4% (wt/vol) BSA followed by incubation with 1:500primary antibody (Aβ1–40 or Aβ1–42 monoclonal antibody; Covance) in2% (wt/vol) BSA at 4 °C overnight. After washing, the membrane was incubatedwith 1:5,000 secondary antibody (Goat-anti-Mouse IgG; CWBio) for 1 h, in-cubated with the HRP substrate (Supersignal West Pico; Thermo Scientific),and exposed to film by using Universal Hood II Imaging System (Bio-Rad). Theraw images were processed by Image Lab software (Bio-Rad).

AlphaLISA Assay. The assay was performed as described in the AlphaLISA Kit(PerkinElmer). Briefly, 2-μL reaction products were incubated with 8-μLAlphaLISA Aβ1–40/42 Acceptor beads at 23 °C for 1 h. After another 30-minincubation with 10-μL AlphaLISA Aβ1–40/42 Donor beads in the dark at23 °C, the samples were read by using Envision-Alpha Reader (PerkinElmer).The readings were expressed in arbitrary unit.

Cleavage Sites Identification by Mass Spectrometry. The reaction productswere flowed through Ni2+-NTA affinity resin (Qiagen) to remove the C-ter-minal peptides of the digested products and the uncleaved products. Theflow-through was incubated with thrombin. The C-terminal fragments ofthrombin-cleaved products were collected by further flowing through chitinresin followed by incubation with 15% trichloroacetic acid. The pellets were

Fig. 5. Structure of PSH bound to a specific presenilin inhibitor L679.(A) L679 and JC-5 inhibit the cleavage of APP C99 by PSH. Shown here is aSDS-PAGE. (B) Structure of PSH bound to L679. The inhibitor is bound in thecleft surrounded by TMs 2, 6, 7, 8, and 9. The inhibitor is in the close proximityof the two catalytic residues Asp162 and Asp220. L679 is shown in stick, andthe 2Fo−Fc density is displayed at 3σ around the inhibitor. (C) A close-up viewon L679. The difference Fourier density for the Br atom in JC-5, displayed at 3σ,is shown in the corresponding location of L679-bound PSH. This informationhelps the assignment of relative orientation for L679.

3348 | www.pnas.org/cgi/doi/10.1073/pnas.1502150112 Dang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

25, 2

020

washed by acetone three times, dried, and dissolved in water:acetonitrile(95:5, vol/vol) with 0.1% formic acid, followed with 15-min sonication andbomb-loading to a fused silica column (100 μm i.d. × 8 cm) home-packedwith C4 (5 μm, 300 Å; Phenomenex) beads. The reaction products wereeluted to an Orbitrap Elite mass spectrometer (ThermoFisher) continuouslywith buffer of acetonitrile:water (80:20, vol/vol) with 0.1% formic acid.A distal 1.8-kV spray voltage was applied for ionization, and the resolutionwas set at 120,000 for full-scan data acquisition. MS scan functions were setat the following values: 1 micro scan, 10-ms maximum injection time, and1e6 AGC target were controlled by the Xcalibur data system (ThermoFisher).The resulting MS spectra were analyzed manually.

Synthesis of Brominated JC-5. Brominated JC-5 was prepared by using amodified procedure for the synthesis of JC-5. Briefly, the carboxylic acid wasmade in 11 steps from (S)-methyl 2-amino-3-phenylpropanoate. The bromin-ated dipeptide was made in four steps from (S)-2-amino-3-(4-bromophenyl)propanoic acid. Peptide coupling under standard conditions followed bycleavage of the tert-butyldimethylsilyl (TBS) group furnished the desired Br-JC-5in an overall 0.8% yield (for details, see Fig. S4).

Crystallization of Inhibitor-Bound PSH. Crystals of PSH bound to L679 weregrown in the same condition as free PSH (10) but diffracted X-rays weakly.With the addition of 1% (wt/vol) n-heptyl-β-D-thioglucoside (Anatrace)during crystallization and annealing at room temperature before flash-fro-zen for approximately 10 s, the diffraction improved dramatically. To confirmthe binding site of the inhibitor, crystals of PSH were grown in the presence ofbrominated JC-5 under the same condition. Both the native crystals andBr-derived crystals were flash-frozen in a cold nitrogen stream of 100 K.

Data Collection and Processing. Diffraction data of L679-bound and Br-JC-5–bound PSH crystals were collected on the beamlines Spring-8 BL41XU and SSRFBL17U, respectively. The data were integrated and scaled by using HKL2000(36). Further processing was carried out by using the CCP4 suite (37). Thestructures were refined with PHENIX (38). The anomalous difference Fouriermap was calculated with anomalous signal from bromide of Br-JC-5 by FFT.

ACKNOWLEDGMENTS. This work was supported by Ministry of Science andTechnology Grant 2014ZX09507003006, National Natural Science Founda-tion of China Grants 31130002 and 31321062, NIH Grants R01AG026660 andR01NS076117 (to Y.-M.L.), and Alzheimer Association Grant IIRG-12-242137(to Y.-M.L.).

1. De Strooper B, et al. (1998) Deficiency of presenilin-1 inhibits the normal cleavage ofamyloid precursor protein. Nature 391(6665):387–390.

2. Selkoe DJ, Wolfe MS (2007) Presenilin: Running with scissors in the membrane. Cell131(2):215–221.

3. De Strooper B, Iwatsubo T, Wolfe MS (2012) Presenilins and γ-secretase: Structure,function, and role in Alzheimer Disease. Cold Spring Harb Perspect Med 2(1):a006304.

4. Morishima-Kawashima M (2014) Molecular mechanism of the intramembrane cleav-age of the β-carboxyl terminal fragment of amyloid precursor protein by γ-secretase.Front Physiol 5:463.

5. Goate A, Hardy J (2012) Twenty years of Alzheimer’s disease-causing mutations.J Neurochem 120(Suppl 1):3–8.

6. Crump CJ, Johnson DS, Li YM (2013) Development and mechanism of γ-secretasemodulators for Alzheimer’s disease. Biochemistry 52(19):3197–3216.

7. Lu P, et al. (2014) Three-dimensional structure of human γ-secretase. Nature 512(7513):166–170.

8. Xie T, et al. (2014) Crystal structure of the γ-secretase component nicastrin. Proc NatlAcad Sci USA 111(37):13349–13354.

9. Torres-Arancivia C, et al. (2010) Identification of an archaeal presenilin-like intra-membrane protease. PLoS ONE 5(9):e13072.

10. Li X, et al. (2013) Structure of a presenilin family intramembrane aspartate protease.Nature 493(7430):56–61.

11. Acx H, et al. (2014) Signature amyloid β profiles are produced by different γ-secretasecomplexes. J Biol Chem 289(7):4346–4355.

12. Kornilova AY, Das C, Wolfe MS (2003) Differential effects of inhibitors on the gamma-secretase complex. Mechanistic implications. J Biol Chem 278(19):16470–16473.

13. Bielefeld-Sevigny M (2009) AlphaLISA immunoassay platform- the “no-wash” high-throughput alternative to ELISA. Assay Drug Dev Technol 7(1):90–92.

14. Chau DM, Crump CJ, Villa JC, Scheinberg DA, Li YM (2012) Familial Alzheimer diseasepresenilin-1 mutations alter the active site conformation of γ-secretase. J Biol Chem287(21):17288–17296.

15. Wang H, Nefzi A, Fields GB, Lakshmana MK, Minond D (2014) AlphaLISA-based high-throughput screening assay to measure levels of soluble amyloid precursor protein α.Anal Biochem 459:24–30.

16. Yu G, et al. (2000) Nicastrin modulates presenilin-mediated notch/glp-1 signal trans-duction and betaAPP processing. Nature 407(6800):48–54.

17. Shah S, et al. (2005) Nicastrin functions as a gamma-secretase-substrate receptor. Cell122(3):435–447.

18. Dries DR, et al. (2009) Glu-333 of nicastrin directly participates in gamma-secretaseactivity. J Biol Chem 284(43):29714–29724.

19. Goo JS, et al. (2013) Nicastrin overexpression in transgenic mice induces aberrantbehavior and APP processing. Mol Neurobiol 48(1):232–243.

20. Qi Y, Morishima-Kawashima M, Sato T, Mitsumori R, Ihara Y (2003) Distinct mecha-nisms by mutant presenilin 1 and 2 leading to increased intracellular levels of amyloidbeta-protein 42 in Chinese hamster ovary cells. Biochemistry 42(4):1042–1052.

21. Walker ES, Martinez M, Brunkan AL, Goate A (2005) Presenilin 2 familial Alzheimer’s

disease mutations result in partial loss of function and dramatic changes in Abeta 42/40

ratios. J Neurochem 92(2):294–301.22. Bentahir M, et al. (2006) Presenilin clinical mutations can affect gamma-secretase

activity by different mechanisms. J Neurochem 96(3):732–742.23. Wolfe MS (2013) Toward the structure of presenilin/γ-secretase and presenilin ho-

mologs. Biochim Biophys Acta 1828(12):2886–2897.24. Kretner B, et al. (2011) Attenuated Abeta42 responses to low potency gamma-sec-

retase modulators can be overcome for many pathogenic presenilin mutants by

second-generation compounds. J Biol Chem 286(17):15240–15251.25. Borgegard T, et al. (2012) First and second generation γ-secretase modulators (GSMs)

modulate amyloid-β (Aβ) peptide production through different mechanisms. J Biol

Chem 287(15):11810–11819.26. Golde TE, Koo EH, Felsenstein KM, Osborne BA, Miele L (2013) γ-Secretase inhibitors

and modulators. Biochim Biophys Acta 1828(12):2898–2907.27. Pozdnyakov N, et al. (2013) γ-Secretase modulator (GSM) photoaffinity probes reveal

distinct allosteric binding sites on presenilin. J Biol Chem 288(14):9710–9720.28. Ataka S, et al. (2004) A novel presenilin-1 mutation (Leu85Pro) in early-onset Alz-

heimer disease with spastic paraparesis. Arch Neurol 61(11):1773–1776.29. Sodeyama N, et al. (2001) Very early onset Alzheimer’s disease with spastic para-

paresis associated with a novel presenilin 1 mutation (Phe237Ile). J Neurol Neurosurg

Psychiatry 71(4):556–557.30. Shrimpton AE, et al. (2007) A presenilin 1 mutation (L420R) in a family with early

onset Alzheimer disease, seizures and cotton wool plaques, but not spastic para-

paresis. Neuropathology 27(3):228–232.31. Chun J, Yin YI, Yang G, Tarassishin L, Li YM (2004) Stereoselective synthesis of photo-

reactive peptidomimetic gamma-secretase inhibitors. J Org Chem 69(21):7344–7347.32. Crump CJ, et al. (2011) Piperidine acetic acid based γ-secretase modulators directly

bind to Presenilin-1. ACS Chem Neurosci 2(12):705–710.33. Fernandez MA, Klutkowski JA, Freret T, Wolfe MS (2014) Alzheimer presenilin-1

mutations dramatically reduce trimming of long amyloid β-peptides (Aβ) by γ-secre-tase to increase 42-to-40-residue Aβ. J Biol Chem 289(45):31043–31052.

34. Saito T, et al. (2011) Potent amyloidogenicity and pathogenicity of Aβ43. Nat Neurosci14(8):1023–1032.

35. Chávez-Gutiérrez L, et al. (2012) The mechanism of γ-Secretase dysfunction in familial

Alzheimer disease. EMBO J 31(10):2261–2274.36. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in os-

cillation mode. Methods Enzymol 276:307–326.37. Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Programs for

protein crystallography. Acta Crystallogr D Biol Crystallogr 50(Pt 5):760–763.38. Adams PD, et al. (2002) PHENIX: Building new software for automated crystallographic

structure determination. Acta Crystallogr D Biol Crystallogr 58(Pt 11):1948–1954.

Dang et al. PNAS | March 17, 2015 | vol. 112 | no. 11 | 3349

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

25, 2

020