The intramembrane cleavage site of the amyloidprecursor protein depends on the length of itstransmembrane domainStefan F. Lichtenthaler*†, Dirk Beher‡, Heike S. Grimm§, Rong Wang¶, Mark S. Shearman‡, Colin L. Masters�,and Konrad Beyreuther§

*Massachusetts General Hospital�Harvard Medical School, Wellman Building, 50 Blossom Street, Boston, MA 02114; ‡Department of Molecular Biology,Merck, Sharp, and Dohme Research Laboratories, Neuroscience Research Centre, Terlings Park, Harlow, Essex CM20 2QR, England; §Center forMolecular Biology Heidelberg, University of Heidelberg, Im Neuenheimer Feld 282, D-69120 Heidelberg, Germany; ¶Department of HumanGenetics, Mount Sinai School of Medicine, 1425 Madison Avenue, New York, NY 10029; and �Department of Pathology,University of Melbourne, Parkville, Victoria 3052, Australia

Edited by Michael S. Brown, University of Texas Southwestern Medical Center, Dallas, TX, and approved December 3, 2001 (received for reviewJuly 27, 2001)

Proteolytic processing of the amyloid precursor protein by �-secre-tase generates C99, which subsequently is cleaved by �-secretase,yielding the amyloid � peptide (A�). This �-cleavage occurs withinthe transmembrane domain (TMD) of C99 and is similar to theintramembrane cleavage of Notch. However, Notch and C99 differin their site of intramembrane cleavage. The main �-cleavage ofC99 occurs in the middle of the TMD, whereas the cleavage ofNotch occurs close to the C-terminal end of the TMD, making itunclear whether both are cleaved by the same protease. Toinvestigate whether �-cleavage always occurs in the middle of theTMD of C99 or may also occur at the end of the TMD, we generatedC99-mutants with an altered length of the TMD and analyzed their�-cleavage in COS7 cells. The C terminus of A� and thus the site of�-cleavage were determined by using monoclonal antibodies andmass spectrometry. Compared with C99-wild type (wt), most mu-tants with an altered length of the TMD changed the cleavage siteof �-secretase, whereas control mutants with mutations outsidethe TMD did not. Thus, the length of the whole TMD is a majordeterminant for the cleavage site of �-secretase. Moreover, theC99-mutants were not only cleaved at one site but at two siteswithin their TMD. One cleavage site was located around the middleof the TMD, regardless of its actual length. An additional cleavageoccurred within the N-terminal half of their TMD and thus at theopposite side of the Notch cleavage site.

Regulated intramembrane proteolysis of type I membraneproteins has been shown to play an important role in the

pathogenesis of Alzheimer’s disease and in cell differentiation(for a review, see ref. 1). The corresponding proteins involved inthese biological processes are the amyloid precursor protein(APP) and Notch, respectively.

The intramembrane proteolysis of APP leads to the genera-tion of the amyloid � peptide (A�), which is deposited in thebrains of patients with Alzheimer’s disease (for a review, see ref.2). To undergo intramembrane proteolysis, APP is first proteo-lytically processed by one of the proteases termed �- and�-secretase (for a review, see ref. 3). The �-secretase seems tobe a member of the ADAM (a disintegrin and metalloprotease)-family of disintegrin-metalloproteases (4–6), whereas the re-cently identified �-secretase is a novel aspartyl protease calledBACE (beta-site APP cleaving enzyme; for a review, see ref. 7).The �- and �-secretases cleave APP within its ectodomain at ashort distance to the transmembrane domain (TMD), shearingoff most of the ectodomain (secretory APP). Subsequently, theremaining membrane-bound C-terminal fragments C83(through �-cleavage) and C99 (through �-cleavage) may un-dergo intramembrane proteolysis by a protease called �-secre-tase, which has not yet been unequivocally identified. However,the �-secretase cleavage depends on the presence of the poly-topic membrane protein presenilin 1, which itself may be �-secre-

tase (for reviews, see refs. 8 and 9). Because presenilin 1 is partof a high molecular weight protein complex, it may requireadditional proteins in this complex to carry out its proteolyticfunction. One such protein could be the recently identifiedNicastrin, which has been found to bind both presenilin 1 and thesubstrate C99 (10). The �-cleavage of C99 and C83 generates theN-terminal peptide fragments A� and p3, which are secreted bycultured cells (11–13), and the short-lived C-terminal fragmentp7 (CTF�), which may stimulate the transcription of so farunknown target genes (14). Because the �-cleavage occurs atdifferent peptide bonds—mainly after residue 40 and to a lowerextent after residue 42 of C99—the A� peptides have 40 or 42residues (A�40, A�42); the same C termini are found for the p3peptides (p340, p342) (15, 16). The �-cleavage seems to occur withbroad substrate specificity because A� is still generated whenresidues within the TMD of C99 are mutated (15, 17–21). Incontrast, the specific cleavage site used by �-secretase (e.g.,cleavage after residue 40 or 42) and thus the C terminus of A�strongly depends on the amino acid composition of the TMD (15,17, 18, 20, 21).

The proteolytic processing of APP is similar to the processingof the cell surface receptor Notch (for an overview, see ref. 8).On ligand binding, Notch is proteolytically processed within itsectodomain. The remaining C-terminal, membrane-bound frag-ment may then undergo intramembrane proteolysis within itsTMD at the so-called S3-site. As for the APP-derived fragmentC99, the intramembrane cleavage of Notch depends on thepresence of presenilin 1. Despite these similarities, the site withinthe TMDs of C99 and Notch, where the intramembrane cleavageoccurs, is different: C99 is mainly cleaved in the middle of itsTMD, whereas Notch is cleaved close to the C-terminal end ofits TMD (22). These differences of the cleavage site raise doubtabout whether both cleavage events share the same proteolyticmechanism and whether both are carried out by the sameprotease.

To address this question we analyzed in this study whichsequence elements of C99 direct the �-cleavage to occur mainlyin the middle of the TMD and whether the �-cleavage of C99

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: APP, amyloid precursor protein; A�40 and A�42, A� peptides with differentN termini ending at the C terminus with residue 40 or 42 of C99; C99, C-terminal 99 residuesof APP; TMD, transmembrane domain; SELDI-TOF, surface-enhanced laser desorption�ionization time-of-flight mass spectrometry; wt, wild type; SP, signal peptide; CTF, C-terminal fragment.

†To whom reprint requests should be addressed. E-mail: stefanL@molbio.mgh.harvard.edu.

The publication costs of this article were defrayed in part by page charge payment. Thisarticle must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.§1734 solely to indicate this fact.

www.pnas.org�cgi�doi�10.1073�pnas.032395699 PNAS � February 5, 2002 � vol. 99 � no. 3 � 1365–1370

CELL

BIO

LOG

Y

may also occur close to the end of the TMD similar to Notch. Byusing C99-mutants with an altered length of the TMD we showthat the �-cleavage site strongly depends on the length of thewhole TMD but not on sequence elements outside of the TMD.The cleavage occurs in the middle of the TMD but may evenoccur close to the N-terminal end of the TMD and thus at theopposite side of the Notch cleavage site.

MethodsCell Culture and Transfections. COS7 cells were cultured accordingto the protocol described previously (17). Cell culture mediawere obtained from Sigma. The pCEP4 vector (Invitrogen) withthe SPC99 cDNA inserts was transfected into COS7 cells byusing Lipofectin (GIBCO�BRL) as described in the manufac-turer’s protocol. For each construct, two or more independenttransfections were analyzed in cell culture with respect to�-secretase activity.

Antibodies. The monoclonal antibodies W02 (for the precipita-tion of all A� peptides regardless of the specific C terminus),G2-10 (specific for A� ending at residue 40), and G2-11 (specificfor A� ending at residue 42) were raised against syntheticpeptides (23).

Metabolic Labeling and Immunoprecipitation. After 45 min of pre-incubation in methionine-free MEM, stably transfected COS7cells were incubated for 16 h in methionine-free MEM contain-ing 10% FBS and 133 �Ci�ml [35S]methionine (AmershamPharmacia). The conditioned media were centrifuged at 4°C for1 min at 4000 � g and divided into three samples with a volumeof 1 ml each. A� and p3 were immunoprecipitated with antibodyW02 (5 �g�ml), G2-10 (12.5 �g�ml), or G2-11 (17.3 �g�ml) and100 �g protein-G Agarose (Boehringer Mannheim), respec-tively. The immunoprecipitated proteins were separated on 10%Tris�Tricine gels (24). The intensity of the bands was quantifiedby using a Fuji PhosphorImager (BAS 1000).

To determine whether the introduced mutations altered theamount of A� generated from C99, A� and C99 were immu-noprecipitated from the conditioned medium and the cell lysate,respectively, by using antibody W02 (concentration as above).After gel electrophoresis and Western Blotting using antibodyW02, the amounts of A� and C99 were measured. Each exper-iment was carried out in three or more independent experiments.For the inhibition of A�-generation, the experiments werecarried out as described above. The specific �-secretase inhibitor{1S-benzyl-4R-[1-(1S-carbamoyl-2-phenylethylcarbamoyl)-1S-3-methylbutylcarbamoyl]-2R-hydroxy-5-phenylpentyl}carbamicacid tert-butyl ester, referred to as Merck A in this study anddescribed in detail elsewhere (25), was dissolved in DMSO andwas used at a concentration of 10 �M. The control cells withoutinhibitor were treated with DMSO alone.

Plasmid Construction. Plasmids pUC18�SPC99-�NT, �CT, andCTI were generated as described earlier for pUC�C99-T43A(15), by using suited oligonucleotides and plasmid pSP65�SPC99(26) as template. The KpnI�SpeI fragments of the pUC18�SPC99-�NT and CTI constructs were cloned into the pCEP4vector digested with KpnI�NheI to generate the mutated expres-sion plasmids pCEP4�SPC99 plasmids. pCEP4�SPC99-�CT wasgenerated by cloning the NruI�SpeI fragment of pUC18�SPC99-�CT into the pCEP4 vector that was digested with PvuII�NheI.

Plasmids pBS�SPC99 with the mutations K28E, �SN, and�1725 were generated by using the Quik Change Site-DirectedMutagenesis Kit (Stratagene), suited oligonucleotides and pBS�SPC99 reverse as DNA-template (15). The KpnI�SpeI fragmentsof these constructs were cloned into the pCEP4 vector that wasdigested with KpnI�NheI. Plasmid pSP65�SPC99-NTI was ob-tained by using the Quik Change Site-Directed Mutagenesis Kit

(Stratagene), the corresponding sense and antisense primers,and pSP65�SPC99 as DNA-template (26). The SmaI�SpeI frag-ment of pSP65�SPC99-NTI was cloned into the pCEP4 vectorthat was digested with NheI�PvuII.

The identity of the constructs obtained by PCR was confirmedby DNA sequencing.

Surface Enhanced Laser Desorption�Ionization Time-of-Flight (SELDI-TOF) Mass Spectrometry. Immunocapture of amyloid peptidesfrom conditioned media by using monoclonal antibody 6E10(Senetec PLC, Napa, CA) covalently coupled to a SELDIprotein chip (Ciphergen Biosystems, Fremont, CA) was done asdescribed previously (25). Briefly, COS7 cells were incubatedover night in DMEM (Sigma), conditioned media diluted withequal volumes of 25 mM Hepes (pH 7.3) and 1 mM EDTA andincubated overnight at 4°C with a 6E10-coupled SELDI proteinchip. After extensive washing of the chip, a bovine insulinstandard for internal calibration was dissolved in a 1:5 dilutedsaturated �-cyano-4-hydroxycinnamic acid solution in 40% ace-tonitrile, 0.25% trif luoroacetic acid. The standard (0.5 �l, 25fmol bovine insulin) was applied to each spot of the SELDI chip,and, after air drying, samples were analyzed on a SELDI massanalyzer PBS II with a linear time-of-f light mass spectrometer(Ciphergen Biosystems) using time-lag focusing. All spectrawere calibrated internally by using the single and double posi-tively charged species of the bovine insulin peptide.

ResultsTo analyze whether specific sequence elements of C99 influencethe �-cleavage site, a systematic series of C99-mutants wasgenerated with an altered length of the TMD or with deletionscovering the whole ectodomain and cytoplasmic domain of C99.Mutations into the N-terminal 16 residues of C99 were notintroduced, because C83—which lacks the N-terminal 16 resi-dues of C99 because of cleavage by �-secretase—is cleaved by�-secretase at the same site as C99 (15–17), and, thus, theN-terminal 16 residues of C99 do not influence the �-cleavagesite. With four mutants, we analyzed possible sequence elementsin the ectodomain and the cytoplasmic domain of C99 (Fig. 1):a deletion of the whole cytoplasmic domain of C99 (�cyto), adeletion of nine residues in the ectodomain (Leu-17 to Gly-25,�1725), a deletion of two residues in the ectodomain (Ser-26 toAsn-27, �SN), and a substitution of residue Lys-28 by Glu(K28E). Because Lys-28 may function as a membrane anchor for

Fig. 1. SPC99 mutations used for the study of �-secretase cleavage. SPC99consists of the signal peptide (SP) of APP followed by Leu and Glu and theC-terminal 99 aa (C99) of APP. Amino acids are shown in the one letter codeand are numbered according to the C99-sequence (1 to 99). The vertical barswithin the C99-sequence indicate the cleavage sites of �-secretase after resi-dues 40 and 42. The shaded area represents the TMD. N and C, N- andC-terminal halves of the TMD.

1366 � www.pnas.org�cgi�doi�10.1073�pnas.032395699 Lichtenthaler et al.

C99, it was not deleted, but substituted by a residue of oppositecharge (Glu).

With four additional mutants, we analyzed whether the lengthof the TMD of C99 influences the cleavage site of �-secretase.In two mutants, two residues were inserted at the N-terminal(C99-NTI) or C-terminal (CTI) end of the TMD of C99 (Fig. 1).In two additional mutants, two residues were deleted at theN-terminal (C99-�NT) or C-terminal (�CT) end of the TMD.Larger alterations of the length of the TMD were not used,because such mutations could interfere strongly with proteintransport: proteins with short TMDs might not be transported tolater compartments of the secretory pathway (27), where at leasta part of the �-secretase activity occurs. In support of this notion,a recent study showed that several deletions and insertions offour or more residues strongly reduced the turnover of APP toA� (18).

Processing of C99-Mutants by �-Secretase. All C99-constructs werestably transfected as SPC99 (APP-signal-peptide-Leu-Glu-C99,Fig. 1) in COS7 cells. After signal peptide removal in theendoplasmic reticulum, the mature C99 starts at its N terminuswith the two additional residues Leu and Glu. Western Blotanalysis of A� immunoprecipitated from the conditioned me-dium shows that C99-wild type (wt) and all mutants wereprocessed to similar amounts of the 4.5-kDa A� peptide (Fig. 2A,shown for the mutants with an altered length of the TMD).Notably, A� generated from the mutant C99-�SN shows aslightly faster electrophoretic mobility than the other A� pep-tides. No A� was detected in vector-transfected control cells.This finding demonstrates that neither the specific residues thatwere mutated or deleted nor the specific sequence of residuesinto which amino acids had been introduced (NTI, CTI) arerequired for general �-secretase cleavage.

To confirm that the C99-mutants were processed by the samepresenilin-dependent �-secretase that cleaves C99-wt, a wellcharacterized potent and specific inhibitor of �-secretase MerckA (25) was shown to almost completely abolish the generation ofA� from wt and mutant C99 (Fig. 2B).

Ratio of A�42�A�40 Depends on the Length of the TMD of C99. Todetermine whether the C99-mutants were processed to the sameA� species as C99-wt (mainly to A�40 and some A�42), the COS7cells were metabolically labeled with [35S]methionine. A�40 andA�42 were then specifically immunoprecipitated from the con-ditioned medium and separated by gel electrophoresis; theiramounts were quantified by PhosphorImaging, and the relativeratios of A�42�A�40 were determined. An alteration of theA�42�A�40 ratio compared with C99-wt is a good indication foran altered �-cleavage of the corresponding mutant C99-protein.A�40 and A�42 denote the different peptides ending in residues40 and 42 of wild-type A�, but may be longer or shorter thanwild-type A� if they carry the indicated insertions or deletions.Both antibodies used for the immunoprecipitation also precip-itate the p3 peptides ending in the same residues (p340 and p342).

As with C99-wt, most C99-mutants were processed to A�40and p340 as well as to A�42 and p342 (Fig. 2C, data shown for themutants with an altered length of the TMD). However, C99-�SNand C99-NTI were merely processed to A�40 and p340 and notto A�42 and p342. Because both mutants were processed to A�(Fig. 2 A), this result shows that both mutants must mainly havebeen processed to A� peptides with a C terminus different fromA�40 and A�42. In COS7 cells that were transfected with thevector alone (control), A� was not detected, and p3 only in lowamounts (Fig. 2C). This p3 must be derived from the endogenousAPP of COS7 cells.

Next, we determined the A�42�A�40 ratios for C99-wt and theC99 mutants in four to six experiments (Fig. 3). The ratios ofA�42�A�40, as determined here, are expressed as percentagesand represent relative but not absolute ratios, because theantibodies have different affinities for the corresponding pep-tides. For C99-wt, the A�42�A�40 ratio is 4.7%. The mutants withmutations outside of the TMD (C99-�1725, K28E, and �cyto)did not alter significantly the A�42�A�40 ratios (�1725, 5.5 �0.9%; �cyto, 4.1 � 0.6%; and K28E, 5.4 � 0.9%), demonstratingthat the ectodomain and the cytoplasmic domain of C99 do notcontain specific sequence elements that influence the cleavagesite of �-secretase.

In contrast, most mutants with an altered length of the TMDled to a changed ratio of A�42�A�40 and thus an altered�-cleavage site. Both the C-terminal insertion (CTI) and theN-terminal deletion (�NT) of two residues led to a similarincrease of the A�42�A�40-ratio (increase: 4.1-fold (CTI), 4.3fold (�NT); A�42�A�40: 19.6 � 2.2% (CTI), 20.5 � 4.7% (�NT);P � 0.001 (CTI), P � 0.01 (�NT)). This increase corresponds toa shift of the �-cleavage site in the C-terminal direction. Incontrast, the C-terminal deletion of two residues (�CT) did notalter the A�42�A�40-ratio (4.6 � 1.5%). For C99-NTI andC99-�SN, no A�42 could be detected so that the calculated ratioA�42�A�40 was 0%.

Both antibodies used for the immunoprecipitation of A�40(G2-10) and A�42 (G2-11) also immunoprecipitated p340 andp342 (Fig. 2). A comparison with C99-wt revealed that the

Fig. 2. Influence of the SPC99 mutations on substrate turnover and on the�-secretase cleavage site. COS7 cells were stably transfected with the SPC99constructs indicated below the panels. (A and B) Western Blot analysis of A�

from the conditioned medium. Antibody W02 was used for the immunopre-cipitation and Western Blot detection. (A) The COS7 cells expressed similaramounts of C99. (B) The specific �-secretase inhibitor Merck A (10 �M) sup-pressed the generation of A� from wild-type and mutant C99. In this exper-iment, COS7 cells were used that did not express equal amounts of C99, whichexplains why different levels of A� were observed for the different mutants.(C) Generation of A�40 and A�42 from C99. Proteins from the conditionedmedium of COS7 cells labeled with [35S]methionine were immunoprecipitatedwith the antibodies specific for A� ending in residue 40 or 42 (indicated abovethe panels). Both antibodies also precipitate the 3-kDa p3 peptides. 40 and 42,antibodies specific for A� and p3 ending in residues 40 or 42. Con, control; cellstransfected with vector pCEP4 without an insert. Vertical black bars within thepictures indicate that the separated lanes were from the same gel but werenot directly next to each other.

Fig. 3. Relative ratios of A�42�A�40 in the conditioned medium of COS7 cellsexpressing C99. The amounts of A�40 and A�42 in Fig. 2 were quantified byPhosphorImaging in four to six independent experiments. Columns representthe mean values. Black error bars give the standard deviation. The asterisksindicate the significance (Student’s t test) relative to C99-wt (**, P � 0.01, ***,P � 0.001). a, the calculated ratios are 0% because no A�42 was detected.

Lichtenthaler et al. PNAS � February 5, 2002 � vol. 99 � no. 3 � 1367

CELL

BIO

LOG

Y

analyzed mutations altered the ratios p342�p340 to a similarextent as the ratios A�42�A�40 (data not shown).

Mass Spectrometric Analysis of A�. To determine exactly how themutations in the TMD of C99 affect the �-cleavage site, A� wasimmunocaptured by using monoclonal antibody 6E10, whichbinds to the N terminus of A�. Subsequently, the immunocap-tured A� was analyzed by SELDI-TOF mass spectrometry (Fig.4). C99-wt gave a prominent signal for A�40 (Fig. 4), confirmingprevious results by us and others (17, 28). The mutants �cyto and�CT showed spectra identical to C99-wt (not shown), and thusno altered �-cleavage. Similarly, the spectrum for C99-�1725showed that it was mainly processed after residue 40 (not shown).Taken together, the deletions outside of the TMD (�1725,�cyto) did not affect the �-cleavage site, confirming the resultsobtained by immunoprecipitation.

In contrast, most mutations within the TMD altered the�-cleavage site by leading to extended or shortened C termini of

A� (Fig. 4), demonstrating that the length of the whole TMD ofC99 is an important factor in determining the �-cleavage site.C99-CTI yielded peaks for A�40 and also for A�42. Because A�42had hardly been detected in the mass spectrum for C99-wt, thisresult indicates that C99-CTI was cleaved to a larger extent afterresidue 42 than C99-wt. Additionally, strong peaks were ob-served for A�39 and A�38. C99-�NT was processed to A�40 butalso gave a clear signal for A�42 and even A�43. This findingconfirms the result of the immunoprecipitation, which hadshown that, for C99-�NT, the �-cleavage site was shifted in theC-terminal direction. C99-NTI, for which hardly any A�40 and noA�42 had been detected by immunoprecipitation, was processedto A�37, but also to smaller A� peptides with a C terminus afterresidues 32, 33, and 34, corresponding to a shift of the �-cleavagesite in the N-terminal direction.

In general, the results of the mass spectrometric analysis arein good agreement with the results obtained by immunoprecipi-tation (Fig. 2C). However, some of the peptides with longer Ctermini, which were clearly detected by immunoprecipitation(Fig. 2C), were underrepresented or not detected in the massspectra (A�42 for C99-wt and �CT; A�40 for NTI and �SN). Thisresult may be due to their lower detection efficiency, as noted ina previous study (28).

The control mutant C99-�SN, which had no altered length ofthe TMD but a deletion of two residues just outside of the TMD,revealed an altered �-cleavage site relative to C99-wt (Fig. 4).C99-�SN was mainly processed to A� peptides with a shortenedC terminus (A�32, A�33, A�34, A�37, and A�38) and, hence, wascleaved at similar sites as C99-NTI. Thus, the deletion of tworesidues outside the TMD had a similar effect on �-cleavage asthe insertion of two residues into the TMD. This unexpectedresult may be explained as follows. In contrast to all othermutants, the deletion of the two polar residues Ser and Asn(�SN) just outside the TMD alters the N-terminal border andthus the length of the TMD (Fig. 5), as determined with thealgorithm of Kyte and Doolittle (29). In this case, Lys-28 isf lanked on both sides by hydrophobic residues and is part of theTMD although, in C99-wt, it is located just outside of the TMD.Therefore, C99-�SN may not be used as a mutant without an

Fig. 4. Mass spectrometric determination of the C terminus of the mutant A�

peptides. A� was immunocaptured from the conditioned medium of COS7cells expressing the indicated C99 mutants. The peaks in the spectra arenumbered according to the wild-type C99-sequence (1 to 99). Additionally, allpeptides contain the two residues Leu and Glu at the N terminus (see Fig. 1).All spectra were calibrated internally by using the single and double positivelycharged species of bovine insulin. The error of masses of all identified peptideswas � 400 ppm compared with the calculated average masses. Asterisksindicate A� peptides oxidized at methionine 35. Masses of these peptides areincreased by 16 mass units as expected for methionine sulfoxide derivatives.

Fig. 5. Major intramembrane �-cleavage sites in the C99-mutants with analtered length of the TMD. Amino acids are shown in the one letter code andare numbered according to the C99-sequence (1 to 99). The black bar belowthe sequence shows the length of the TMD for the individual C99-mutant witha vertical bar at the middle of the TMD. The black arrows indicate the major�-cleavage sites (A� peptides) as determined by mass spectrometry (Fig. 4).Open arrows indicate the additional cleavage sites that were detected only byimmunoprecipitation (Fig. 2C) but not by mass spectrometry (see text fordetails).

1368 � www.pnas.org�cgi�doi�10.1073�pnas.032395699 Lichtenthaler et al.

altered length of the TMD but, instead, as an additional C99-mutant with an increased length of the TMD.

DiscussionThe type I membrane proteins Notch and C99—derived fromAPP by �-secretase cleavage—both undergo a similar proteo-lytic cleavage within their TMDs, which is termed regulatedintramembrane proteolysis (1). This proteolytic cleavage reac-tion is poorly understood, and it remains unclear whether thecleavage of Notch and C99 takes place mechanistically in thesame way and whether it is carried out by the same protease. Thisconcern stems from the fact that Notch seems to be cleaved atthe C-terminal end of its TMD, whereas the �-cleavage of C99occurs exactly in the middle of the TMD. By using a systematicseries of C99-mutants with an altered length of the TMD or withdeletions covering the whole ectodomain and cytoplasmic do-main of C99, we analyzed which sequence elements of C99promote the �-cleavage to occur mainly in the middle of theTMD. Additionally, we investigated whether the introducedmutations affected the amount of A� generated from C99(cleavage efficiency).

Concerning the cleavage efficiency, we and others have shownpreviously that the amino acid sequence within the TMD of C99has little influence on the amount of secreted A� (15, 17–21),suggesting a broad substrate specificity of �-secretase. Here, weextend this analysis further to the ectodomain and the cytoplas-mic domain of C99. Our results show that neither domaincontains sequence elements that are required for the intramem-brane cleavage of C99. Interestingly, this broad substrate spec-ificity is shared with the �-secretase-like cleavage of Notch (30).Given this result, it is possible that �-cleavage occurs for moretype I proteins than just APP�C99 and Notch. Similar to Notchand APP, such additional proteins may first need to undergocleavage in their ectodomain to present a short ectodomain stub,which seems to be a requirement for intramembrane proteolysis(30). In support of this idea, the transmembrane receptortyrosine kinase ErbB4 has just been shown to also undergo a�-secretase-like cleavage, but only after an initial cleavage in theectodomain (31).

The broad substrate specificity clearly distinguishes the�-secretase cleavage of C99 from the similar intramembrane-cleavage of the sterol-regulatory element binding protein(SREBP), which is involved in the cellular cholesterol homeosta-sis (for an overview, see ref. 1). SREBP is cleaved by themetalloprotease S2P, which requires two amino acid motifswithin the SREBP sequence: first, the cytoplasmic residuesDRSR, which are located several residues away from the actualcleavage site (32), and, second, the two residues asparagine-proline around the middle of the TMD of SREBP (33). Thesecond motif is also required for the cleavage of ATF6, which isa second substrate of S2P (34).

A major challenge in understanding the mechanism of �-cleav-age results from the fact that the �-cleavage occurs within theTMD of C99 and that it is difficult to conceive how a proteasemay cleave within the lipid bilayer. Presenilin 1, which seems tobe �-secretase or at least part of a �-secretase complex, has twoaspartyl residues that are critical for its function located withinits TMDs so that �-cleavage could occur within the membrane(8). However, because no protease has been proven to cleavewithin the membrane, alternative cleavage mechanisms havebeen proposed for C99 that invoke the TMD to be C-terminallyshorter, such that a cytosolic or membrane-associated proteasecould cleave C99 (18, 19, 35). If the alternative models are true,we expect that deletions and insertions in the C-terminal half ofthe TMD have a weak or no effect on the �-cleavage site. Incontrast, if �-secretase is a membrane protein that cleaves C99within the membrane, we expect that deletions and insertions atboth ends of the TMD affect the �-cleavage site.

Our results identify the length of the whole TMD as animportant factor in determining the �-cleavage site: most C99-mutants with an altered length of the TMD changed the�-cleavage site compared with C99-wt, whereas the C99-mutantswith mutations outside of the TMD did not. Importantly,mutations at both ends of the TMD and not only in theN-terminal half affected the �-cleavage site, speaking against amechanism of �-cleavage outside of the membrane or at themembrane-boundary. This notion is further reinforced by pre-vious experiments that showed that mutations at every singleresidue within the C-terminal half of the TMD of C99 altered the�-cleavage site (17). Our results are consistent with a model inwhich �-secretase is a membrane protein and cleaves C99 withinthe membrane.

In contrast to the role of the TMD of C99 in determining the�-cleavage site, our study shows that the ectodomain and thecytoplasmic domain do not contain such signals. This result is ingood agreement with a previous study (36), which analyzed theproteolytic processing of a non-natural probe for �-secretase.The authors showed that a peptide consisting of the TMD ofAPP flanked on both sides by unique epitope tags is cleaved by�-secretase after residues 40 and 42 (as C99-wt) but also atadditional peptide bonds between residues 35 and 42. Becausethat study did not determine which peptide bond was the maincleavage site, it remains unclear whether that probe is cleaved inexactly the same way as the natural substrate C99, for which themain �-cleavage occurs after residue 40.

The �-cleavage of C99-wt mainly occurs at one peptide bond(after residue 40), which is exactly the middle of the putative24-residue-long TMD. In contrast, most of the mutants with analtered length and thus an altered middle of the TMD werecleaved at two sites within the TMD. One of the sites was locatedclose to the new middle of the TMD, whereas the other one wasnot. The C-terminal insertion (CTI) as well as the N-terminaldeletion (�NT), which both have the middle of their TMD moreC-terminal than C99-wt, shifted the �-cleavage site in theC-terminal direction toward less A�40 but more A�42 and evenA�43 (Fig. 5). In addition, a second, more N-terminal cleavagesite after A�38 was detected for C99-CTI. The two mutants NTIand �SN, which have the middle of their TMD more N-terminalthan C99-wt, shifted the �-cleavage site in the N-terminaldirection. As a result, less A�42 and A�40 were generated andmore A�37 and A�38 but also shorter peptides with a C terminusafter residues 32, 33, and 34. The only mutant that did not showa second major cleavage site is C99-�CT with a deletion at theC-terminal end of the TMD. Like C99-wt, this mutant wasmainly cleaved after residue 40, although the middle of its TMDis located after residue 39.

Particularly surprising is the result that the mutants NTI and�SN were not only processed around the middle of the TMD(A�37 and A�38) but also much more N-terminal (A�32, A�33 andA�34). Because a potent and specific inhibitor of �-secretasesuppressed the generation not only of the long but also of theshort A� peptides, it is clear that the short A� peptides must alsohave been generated in a �-secretase-dependent fashion. How-ever, it remains unclear whether the short A� peptides weregenerated by direct �-cleavage or whether the initial �-cleavageoccurred in the middle of the TMD and was followed by asecondary proteolytic event, e.g., an additional �-cleavage ortrimming by an exopeptidase. In this latter scenario, the Cterminus of A� would pretend a �-cleavage in the N-terminalhalf of the TMD although the cleavage had occurred in themiddle of the TMD. Interestingly, a similar observation hasrecently been made concerning the C-terminal fragment of C99that arises through �-cleavage (CTF�) and that had not beenidentified so far. Three studies reported the identification of aCTF�-like fragment (p6.5) in mouse brain and cultured cells thatstarts close to the C terminus of the TMD (predominantly

Lichtenthaler et al. PNAS � February 5, 2002 � vol. 99 � no. 3 � 1369

CELL

BIO

LOG

Y

residue 50 of C99; refs. 37–39). However, neither group foundthe slightly longer CTF� (p7) that would correspond to the�-cleavage after residues 40 and 42, which led to the conclusionthat �-cleavage may occur simultaneously at different siteswithin the TMD (after residues 40�42 as well as after residue 50).

Interestingly, the apparent N-terminal �-cleavage after A�33and A�34 of C99-NTI and �SN is the opposite side of thecleavage site of Notch, which seems to occur close to the Cterminus of its TMD (22). Similar to C99, only one of theproteolytic cleavage products of Notch (the C-terminal frag-ment) has been identified, so that we cannot rule out that theinitial �-like cleavage of Notch occurs in the middle of the TMDand is followed by a secondary proteolytic event. In this case, theintramembrane cleavage could be similar for both proteins, butthe identified cleavage products would pretend a differentcleavage mechanism. Determining the C terminus of theN-terminal cleavage fragment of Notch should answer thisquestion.

Our analysis shows that some of the mutations in the N-terminal half of the TMD of C99 (NTI, �SN) have a strongereffect on the cleavage site than some of the mutations in theC-terminal half (�CT), which is in agreement with a recentreport that analyzed the processing of full-length APP withdeletions, insertions, and point mutations within the TMD (18).The previous study, however, took this result as evidence for the�-cleavage site being determined by the length of the N-terminalhalf but not by the C-terminal half of the TMD of C99. Becauseour results show that the length of the whole TMD is importantfor determining the �-cleavage site, we favor a different expla-

nation. The N-terminal and the C-terminal half of the TMD ofC99 have very different amino acid compositions. The C-terminal half consists mainly of large residues (Met, Ile, Leu, andThr) that are evenly distributed around the axis of the helicalTMD. In contrast, the N-terminal half consists of small (Gly andAla) as well as of large residues (Met, Ile, and Leu), with thesmall residues on one side of the helix and the large residues onthe other side (not shown). One of these two helical sides couldbe an interface for a functional interaction of C99 with �-secre-tase or other proteins. Insertion or deletion of two residues in theN-terminal half of the TMD—as in our C99-mutants—abolishesthe amino acid arrangement of one helical side with largeresidues and one helical side with small residues. Thus, muta-tions in the N-terminal half of the TMD of C99 may have a muchmore pronounced effect on the cleavage site of �-secretase thanmutations in the C-terminal half.

In summary, our study shows that �-secretase has a very loosesubstrate specificity, which distinguishes �-secretase from S2P,another protease involved in regulated intramembrane proteol-ysis. Moreover, our experiments identify the length of the wholeTMD of C99 as a major determinant of the �-cleavage site andsuggest that �-cleavage takes place within the membrane. The�-cleavage site is located in the middle of the TMD, but anadditional cleavage may occur within the N-terminal half of theTMD and thus at the opposite side of the cleavage site of Notch.

This work was supported by the Deutsche Forschungsgemeinschaft(through SFB 317 and a postdoctoral Emmy Noether fellowship toS.F.L.) and the Fonds der Chemischen Industrie of Germany.

1. Brown, M. S., Ye, J., Rawson, R. B. & Goldstein, J. L. (2000) Cell 100, 391–398.2. Selkoe, D. J. (2001) Physiol. Rev. 81, 741–766.3. De Strooper, B. & Annaert, W. (2000) J. Cell Sci. 113, 1857–1870.4. Lammich, S., Kojro, E., Postina, R., Gilbert, S., Pfeiffer, R., Jasionowski, M.,

Haass, C. & Fahrenholz, F. (1999) Proc. Natl. Acad. Sci. USA 96, 3922–3927.5. Buxbaum, J. D., Liu, K. N., Luo, Y., Slack, J. L., Stocking, K. L., Peschon, J. J.,

Johnson, R. S., Castner, B. J., Cerretti, D. P. & Black, R. A. (1998) J. Biol.Chem. 273, 27765–27767.

6. Koike, H., Tomioka, S., Sorimachi, H., Saido, T. C., Maruyama, K., Okuyama,A., Fujisawa-Sehara, A., Ohno, S., Suzuki, K. & Ishiura, S. (1999) Biochem. J.343, 371–375.

7. Vassar, R. & Citron, M. (2000) Neuron 27, 419–422.8. Wolfe, M. S. & Haass, C. (2001) J. Biol. Chem. 276, 5413–5416.9. Steiner, H. & Haass, C. (2000) Nat. Rev. Mol. Cell. Biol. 1, 217–224.

10. Yu, G., Nishimura, M., Arawaka, S., Levitan, D., Zhang, L., Tandon, A., Song,Y. Q., Rogaeva, E., Chen, F., Kawarai, T., et al. (2000) Nature (London) 407, 48–54.

11. Seubert, P., Vigo Pelfrey, C., Esch, F., Lee, M., Dovey, H., Davis, D., Sinha,S., Schlossmacher, M., Whaley, J., Swindlehurst, C., et al. (1992) Nature(London) 359, 325–327.

12. Shoji, M., Golde, T. E., Ghiso, J., Cheung, T. T., Estus, S., Shaffer, L. M., Cai,X. D., McKay, D. M., Tintner, R., Frangione, B., et al. (1992) Science 258,126–129.

13. Haass, C., Schlossmacher, M. G., Hung, A. Y., Vigo Pelfrey, C., Mellon, A.,Ostaszewski, B. L., Lieberburg, I., Koo, E. H., Schenk, D., Teplow, D. B., et al.(1992) Nature (London) 359, 322–325.

14. Cao, X. & Sudhof, T. C. (2001) Science 293, 115–120.15. Lichtenthaler, S. F., Ida, N., Multhaup, G., Masters, C. L. & Beyreuther, K.

(1997) Biochemistry 36, 15396–15403.16. Citron, M., Diehl, T. S., Gordon, G., Biere, A. L., Seubert, P. & Selkoe, D. J.

(1996) Proc. Natl. Acad. Sci. USA 93, 13170–13175.17. Lichtenthaler, S. F., Wang, R., Grimm, H., Uljon, S. N., Masters, C. L. &

Beyreuther, K. (1999) Proc. Natl. Acad. Sci. USA 96, 3053–3058.18. Murphy, M. P., Hickman, L. J., Eckman, C. B., Uljon, S. N., Wang, R. & Golde,

T. E. (1999) J. Biol. Chem. 274, 11914–11923.19. Tischer, E. & Cordell, B. (1919) (1996) J. Biol. Chem. 271, 21914–21912.20. Maruyama, K., Tomita, T., Shinozaki, K., Kume, H., Asada, H., Saido, T. C.,

Ishiura, S., Iwatsubo, T. & Obata, K. (1996) Biochem. Biophys. Res. Commun.227, 730–735.

21. Suzuki, N., Cheung, T. T., Cai, X. D., Odaka, A., Otvos, L., Jr., Eckman, C.,Golde, T. E. & Younkin, S. G. (1994) Science 264, 1336–1340.

22. Schroeter, E. H., Kisslinger, J. A. & Kopan, R. (1998) Nature (London) 393,382–386.

23. Ida, N., Hartmann, T., Pantel, J., Schroder, J., Zerfass, R., Forstl, H.,Sandbrink, R., Masters, C. L. & Beyreuther, K. (1996) J. Biol. Chem. 271,22908–22914.

24. Schagger, H. & von Jagow, G. (1987) Anal. Biochem. 166, 368–379.25. Shearman, M. S., Beher, D., Clarke, E. E., Lewis, H. D., Harrison, T., Hunt,

P., Nadin, A., Smith, A. L., Stevenson, G. & Castro, J. L. (2000) Biochemistry39, 8698–8704.

26. Dyrks, T., Dyrks, E., Masters, C. & Beyreuther, K. (1992) FEBS Lett. 309,20–24.

27. Munro, S. (1995) EMBO J. 14, 4695–4704.28. Wang, R., Sweeney, D., Gandy, S. E. & Sisodia, S. S. (1996) J. Biol. Chem. 271,

31894–31902.29. Kyte, J. & Doolittle, R. F. (1982) J. Mol. Biol. 157, 105–132.30. Struhl, G. & Adachi, A. (2000) Mol. Cell 6, 625–636.31. Ni, C. Y., Murphy, M. P., Golde, T. E. & Carpenter, G. (2001) Science 294,

2179–2181.32. Duncan, E. A., Dave, U. P., Sakai, J., Goldstein, J. L. & Brown, M. S. (1998)

J. Biol. Chem. 273, 17801–17809.33. Ye, J., Dave, U. P., Grishin, N. V., Goldstein, J. L. & Brown, M. S. (2000) Proc.

Natl. Acad. Sci. USA 97, 5123–5128.34. Ye, J., Rawson, R. B., Komuro, R., Chen, X., Dave, U. P., Prywes, R., Brown,

M. S. & Goldstein, J. L. (2000) Mol. Cell 6, 1355–1364.35. Coles, M., Bicknell, W., Watson, A. A., Fairlie, D. P. & Craik, D. J. (1998)

Biochemistry 37, 11064–11077.36. Bunnell, W. L., Pham, H. V. & Glabe, C. G. (1998) J. Biol. Chem. 273,

31947–31955.37. Sastre, M., Steiner, H., Fuchs, K., Capell, A., Multhaup, G., Condron, M. M.,

Teplow, D. B. & Haass, C. (2001) EMBO Rep. 2, 835–841.38. Gu, Y., Misonou, H., Sato, T., Dohmae, N., Takio, K. & Ihara, Y. (2001) J. Biol.

Chem. 276, 35235–35238.39. Yu, C., Kim, S. H., Ikeuchi, T., Xu, H., Gasparini, L., Wang, R. & Sisodia, S. S.

(2001) J. Biol. Chem. 276, 43756–43760.

1370 � www.pnas.org�cgi�doi�10.1073�pnas.032395699 Lichtenthaler et al.