Cleavage at the amino and carboxyl termini of Alzheimer's amyloid-beta by cathepsin D.

Amyloid beta (A beta) is a 39-43-residue protein that originates from proteolysis of the beta-protein precursor (beta PP) and accumulates in senile plaques in brains of Alzheimer's disease (AD) patients. Mutant beta PP, which incorporates an AD-causing double mutation at positions 687-688, has been shown to enhance A beta production in transfected cells. In this work we investigate the susceptibility of the mutant beta PP sequence to proteolytic cleavage by proteinases from human brain. Internally quenched fluorogenic substrates were used that encompass the NH2-terminal sequence of A beta from wild-type beta PP, the double mutant, and the two single substitutions. Proteinase activity in brain extract cleaved the mutant substrate 100-fold faster than the wild-type substrate and the partial mutants 25-fold faster. The major cleavage site in all substrates was at the amyloidogenic Asp1 site. The brain activity appeared to be cathepsin D (CD), as indicated by similarities to purified CD in 1) the rate and site of substrates cleavage, 2) the pH optima, and 3) the sensitivity to pepstatin A. The increased activity against the mutant substrate was not shared by cathepsins B and C, pepsin, HIV proteinase, and Candida albicans Asp-proteinase. Furthermore, CD cleaved a substrate that incorporates the COOH terminus of A beta at positions equivalent to Thr43 and Ala42, at ratios of 68% and 32%, respectively. CD degraded A beta 1-40 into six fragments but A beta 1-42 was completely resistant to digestion, probably because of its aggregation characteristics. These results indicate that CD is capable of producing the cleavages resulting in A beta production and that it may prove to be a suitable therapeutic target.

number of mutations in or immediately flanking the AP region of PPP that are linked to an early onset form of the disease (see Refs. 2 and 3 for reviews). Therefore, understanding of the processing pathways of PPP in the brain is central to understanding the etiology of AD.
Two mutually exclusive pathways have been described for PPP. A secretory pathway generates a large secreted NH,-terminal region, which was observed in cerebrospinal fluid (4) and in the media of cell cultures transfected with PPP cDNAs. In various cell cultures (5) and in brains (61, the secretory cleavage occurs inside the AP region, which precludes the evolution of the AP peptide. A second pathway that involves the endosomal-lysosomal system has been shown to produce COOH-terminal fragments of 8-22 kDa that include potentially amyloidogenic forms with the 4-kDa AP region (7)(8)(9). Specifically, aspartyl proteinases seem to be involved in generating the COOH-terminal fragments, while cysteinyl proteinases are involved in degradation of the fragments (9). Whether these forms actually give rise to AP is still equivocal. A few observations suggest that PPP processing in lysosomes may be involved in the disease process and, therefore, in AP production. These include: lysosomal hydrolases accumulate around the amyloid plaques (lo), acidic pH between 3.5 and 6.0 promotes the aggregation of AP (ll), and potentially amyloidogenic fragments, similar to those generated in lysosomes, are neurotoxic (12,13). Conversely, more direct evidence suggests that AP production occurs in other acidic compartments such as endosomes and late Golgi or secretory vesicles. This suggestion is supported by experiments showing that lysosomotropic inhibitors failed to inhibit AP production (14) and by the observation ofAP production in I-cells, in which some lysosomal functions are deficient due to a genetic defect in mannose phosphorylation (15). In addition, leupeptin, a serine and cysteinyl proteinase inhibitor that inhibits 70% of lysosomal proteinase activities (16), failed to inhibit AP secretion (14,17). This last observation, however, does not address the possibility that AP is produced by an aspartyl proteinase. In one study, alkalinizing agents completely inhibited the generation of COOH-terminal fragments of PPP, but did not inhibit AP production, suggesting that AP is not generated in an acidic compartment (17). This observation, however, was not reproduced by others (14,18).
One of the PPP mutations that is associated with early onset AD in two related Swedish families consists of a double replacement of L y P 0 + and Met671 + Led7' (the "Swedish mutation") (19). When Swedish mutant PPP was transfected into neuroblastoma or kidney cells, 6-%fold more AP was secreted to the medium and &fold more potentially amyloidogenic COOH-terminal fragments were produced (20)(21)(22). The Swedish mutation is located just upstream of the NH,-terminal amyloidogenic cleavage and could affect substrate binding and catalytic efficiency of the proteinase producing the cleavage. Therefore, we compared the susceptibility of the Swedish and wild-type sequences of PPP to proteolytic cleavages by brain Cleavage of NH, and COOH Termini of AP by Cathepsin D 18423 proteinases. Aproteinase that cleaves the Swedish sequence at least &fold faster than the wild type may be the amyloidogenic proteinase and account for the observations in transfected cells. Indeed, we found that cathepsin D (CD) (EC 3.4.23.5) cleaved the Swedish NH,-terminal sequence (termed Swedish substrate) about 100-fold faster than the wild-type sequence (native substrate). Furthermore, CD produced the COOH-terminal amyloidogenic cleavages at positions equivalent to Ala4' and Thr43 of AP. Finally, CD degraded AP 1-40 into small peptides but not AP 1-42, suggesting that CD is capable of producing AP from its precursor protein.

MATERIALS AND METHODS
Biological Materials-Brain tissue samples (frontal cortex) were obtained from a donor with no evidence of neurological disorder. Samples were homogenized in a glassmeflon homogenizer for 1 min in 25 m~ MOPS/Tris, pH 7.5, 0.3 M sucrose, and 2.5 mM EDTA, at a 15 g/ml t i s s u e h f f e r ratio. The homogenized samples were centrifuged twice for 15 min at 5000 x g a t 4 "C.
Substrate Preparation-Fluorogenic substrates were prepared that span the amino acid sequence near the NH,-terminal cleavage site ofAP (see Fig. l), from G1ua8 to Phe675 of wild-type pPP770 (native substrate), and of mutant pPP from a Swedish family with early onset AD where the Lys-Met at positions 670-671 are replaced by Asn-Leu (Swedish substrate) (19). Two additional substrates were made, each incorporating one of the amino acid replacements: Lys to Asn ( h n substrate) and Met to Leu (Leu substrate). A substrate that spans the sequence near the COOH-terminal cleavage site ofAp (substrate C) from Val"' to Ile716 was also made. Polyether linkers were incorporated into the substrates to distance the fluorescent and absorbent moieties from the amino acid sequence and to enhance solubility.
Proteinase Assays-Proteinase assays were performed a t 37 "C, in a 300-pl reaction volume consisting of 40 m~ buffer, 10 substrate (from a 1 mM stock in dimethyl sulfoxide), and 5-10 pl of brain extract. Buffers used were malonate, pH 3.0; formate, pH 3.54.0; acetate, pH 4.5-5.0; MES, pH 5.5-6.0; MOPS, pH 7.0; and Tris, pH 7.5. The principle of operation of the fluorogenic assays employing fluorescence quenching have been described for the secretory and amyloidogenic site of pPP (23). Fluorescence measurements were made on a Shimadzu RF5000U spectrofluorimeter with excitation and emission wavelengths set at 340 and 490 nm, respectively.
Identification of Cleavage Sites-Substrates were incubated with brain extracts or purified enzymes under the conditions specified for proteinase assays with a substrate concentration of 20 p~ at 23 "C. Extract was diluted 50-100-fold into the reaction mixture. After incubation of 40 or 240 min, samples were analyzed by RP-HPLC using an Applied Biosystems C,, column (220 x 4.2 mm). Substrate fragments were eluted with a linear H,O/acetonitrile gradient at 3% acetonitrile/ min, containing 0.1% trifluoroacetic acid at a flow rate of 1 ml/min. Sequential absorbance and fluorescence measurements of the column eluant were performed. Absorbance was measured simultaneously at 260 nm for EDANS-containing peptides and at 495 nm for DABCYLcontaining peptides; fluorescence of EDANS-containing peptides was monitored with excitation and emission of 340 and 490 nm, respectively. The molecular weights of peptide isolated from substrate cleavage were analyzed using a Finnigan Mat ESI mass spectrometer or a Bruker Reflex laser desorption mass spectrometer.
Hydrodynamic Measurements-Sedimentation velocity measurements were performed on a Beckman XL-A analytical ultracentrifuge and analyzed as described by Holzman and Snyder (24). The experiments were carried out under the reaction condition specified for proteinase assays with Ap concentration of 200 p~ at 23 "C.

RESULTS
Cleavage of NH,-terminal Substrates of AP by Human Brain Extract-Four NH,-terminal substrates were used to assay human brain proteinase activities: a "native substrate" (wild type), a Swedish substrate that incorporates the double mutation of Lys-Met to Asn-Leu, the "Am substrate" that incorporates the Lys to Asn replacement, and the "Leu substrate" that incorporates the Met to Leu replacement (Fig. 1). With all substrates, maximum activity was observed at pH 3.0-3.5 (Fig. 2), suggesting that proteinases from a n acidic compartment are active against these substrates. Activity against the Swedish    5.5-7.5) were added to a total volume of 300 pl consisting of 40 mM substrate was 100-130-fold higher than the rate against the native substrate at pH 3.0-3.5 and the difference diminished as the pH increased. Both the Asn and Leu substrates resulted in about 25-fold increase in activity relative to the native substrate, at pH 3.5, suggesting that the 100-fold increase observed with the Swedish substrate resulted from a synergistic effect of both mutations.
The cleavage sites in all substrates were determined by sepa-  Cleavage of the Swedish, Asn, and Leu substrates, each produced a major peptide eluting a t 13.4, 12.6, or 14.2 min, respectively (Fig. 3). These peptides had M,s of 1095, 1114, or 1110, respectively, corresponding to cleavages at the amyloidogenic Met6-Asp7 or Leu6-Asp7 bonds. The small peaks eluting at 10.6 or 11.5 min had M,s of 982 or 996, corresponding to cleavages at the COOH-side of Asn' or Lys5, respectively.
Cleavage by Purified Human Cathepsin D-The brain activity against these substrates peaked at pH 3.0 suggesting that a proteinase from an acidic compartment, such as CD, may be involved. Indeed, the activity of purified human liver CD at pH 3.0-4.0 appeared quantitatively similar to that of human brain extract (Fig. 4). As in brain extract, activity against the Swedish substrate was -100-fold higher than activity against the native substrate, and activities against Asn and Leu substrates were -25-fold higher.
HPLC analyses of the cleavage sites produced by CD revealed that all the substrates were cleaved at the amyloidogenic Met6-Asp7 or Leu6-Asp7 bonds (Fig. 5), as was observed with brain extracts. Very low levels of cleavage at other sites were detectable when pure CD was used, suggesting that the minor cleavages observed with brain extract, may be produced by other proteinases that are present at low levels.
Activities by Other Proteinases-The activities of the Cys proteinases cathepsin B and C, and Asp proteinases pepsin, HIV proteinase, and Candida albicans proteinase were assayed to evaluate whether the susceptibility of the Swedish sequence to proteolysis is unique to CD or common to other enzymes (Table I). Relative to all other proteinases, the activity of CD was extremely high against both substrates and especially against the Swedish substrate. Of the additional proteinases tested, only pepsin showed a 10-fold preference for the Swedish substrate relative to the native. However, at pH 3.5 its activity was highest against the Asn substrate and not the Swedish (not shown). Thus, the results indicate that the strong preference for the Swedish substrate is unique to CD.
Effects of Inhibitors-The effects of pepstatin A on brain extract activity and on CD were compared in order to obtain additional evidence about the identity of brain activities against the native and Swedish substrates. Pepstatin A (at 9.6 nM) inhibited both CD and the brain activity against the Swedish substrate by 98% with IC,, values of 1.4 and 2.1 nM, respectively (Fig. 6A). These IC,, values are reasonably close, suggesting that the brain extract activity is indeed CD. Similar results were obtained with a novel aspartyl proteinase inhibitor (data not shown) with considerable specificity for CD.2 These results provide additional evidence that the brain activity against the Swedish substrate is that of CD.
Pepstatin A at 14.4 nM, inhibited the brain activity against the native substrate only by 50% and E64, a cysteinyl proteinase inhibitor, together with pepstatin inhibited the activity by 90% (Fig. 6B). However, CD activity against the native substrate was inhibited 95% with IC,, of 2 m. Thus, cysteinyl proteinases also hydrolyze the native substrate.
Cleavage of the native substrate by brain extract was further analyzed in the presence of pepstatin A and E64 (Fig. 7). Pepstatin A inhibited the cleavage a t Met6 by 96% and the cleav-J. Erickson, personal communication. ages a t other sites (Val4, Lys', Asp7, and Ala') by 3045%. Conversely, E64 inhibited only the cleavage at Lys' (by 70%), and increased the rate of cleavage at Asp7 and Met'. The increases in these cleavages could have resulted from inhibition of cysteinyl proteinase which may degrade the peptides produced by the aspartyl proteinase. Thus, an aspartyl proteinase, not a cysteinyl proteinase, cleaves the native substrate at the amyloidogenic site and therefore, is likely to be the same enzyme that cleaves the Swedish substrate. .. _ r .
Cleavage of Carboxyl-terminal Substrate-The presence of proteinase(s) capable of producing carboxyl-terminal cleavage of AP was studied using a substrate encompassing PPP sequence from position 711 t o 716 (substrate C). Brain extract activity against substrate C peaked at pH 3.5 (Fig. EM), suggesting the involvement of acidic proteinases. Cleavage site analysis revealed 3 NH,-terminal peptides. The peak eluting at 12.6 min had a M , of 1024 (Fig. 8B), corresponding to peptide G1y'-Thr' and cleavage at the Thr6-Va17 bond. This is equivalent to a cleavage at Thr43 of AP, which would yield an AP of 43 amino acids. The peak of 13.0 min had a M , of 923, which corresponds to peptide Gly'-Ala5 (cleavage at Ala4'); and the peak of 13.5 min, with M , of 852, corresponded to peptide G1y'-Ile4 (cleavage at Ile41). HPLC analysis every 40 min showed that initially the peak of 12.6 min, Thr6, accounted for 69% of the cleavage products, and peak Ala5 accounted for 31%. However, at longer time points, as the substrate concentration diminished, peak Thr' declined while peaks Ile4 and Ala5 increased, suggesting that Glyl-Ile4 evolved by exopeptidase activity (not shown).
Purified human CD produced two amino terminal peptides, Glyl-Thr', which accounted for 68% of the cleavage products, and G1y'-Ala', accounting for 32% (Fig. 8B). Even long incubation periods did not alter these ratios significantly, indicating that cleavage at Thr43 precluded the cleavage at position Ala4'. These results suggest that the major activity observed in human brain extract is CD, and that CD is capable of producing amyloidogenic cleavages at Thr43 or Ala4', but not at Ile41.
Digestion ofAmyloid-p 1-40 and Ap 1-42 by Human Cathepsin D-We show above that CD is capable of cleaving both the amino and carboxyl ends of AP. However, CD would be an unlikely amyloidogenic proteinase if it is capable of digesting the resulting amyloid into small fragments. To test whether CD degrades AP further, AP 1-40 and AP 1-42 were incubated with CD. AP 1-40 was quickly digested into 6 distinct fragments at pH 3.5 (Fig. 9A). At pH 5.0 AP was digested into the same  Fig. 6 except that 20 p~ substrate was added. The substrate was incubated with brain extract for 4 h in the presence of 14.4 nM pepstatin A or 300 nM E64, or both. The reaction was stopped by addition of trifluoroacetic acid to 2% and a 5O-pl aliquot was analyzed by HPLC. fragments but at a much slower rate (not shown). Conversely, AP 1-42 was not digested even after 16 h (Fig. 9B). This result was further confirmed by mass spectrometry analysis of the reaction mixture, which revealed the presence of AP with M, of 4511 and no other small peptides.
The resistance ofAP 1-42 to digestion by CD may result from the aggregation of AP 1-42, which could prevent accessibility of the enzyme to the cleavage sites (11). Therefore, the aggregation states of AP 1-40 and 1-42 were determined by analytical ultracentrifugation under the reaction conditions used. This analysis showed that the sedimentation coefficient of AP 1-40 was -11 S, which is consistent with a protein of a few hundred kilodaltons in mass (Fig. 10). The sedimentation coefficient of AP 1-42, however, was about 140 S, which is consistent with aggregate mass of tens of thousands of kilodaltons. Thus, the large aggregates of AP 1-42, or their different structure probably prevented access of the enzyme to the cleavage sites. DISCUSSION A mutation that was reported to cause early onset AD in a Swedish family, involves the substitution of the Lys-Met at position 670-671 of PPP with Asn-Leu (19). The mutant PPP, when transfected into cells, produced 6-8-fold more AP than wild-type PPP (20)(21)(22). The mutation is localized just upstream of the amyloidogenic cleavage site and could affect the susceptibility of the Met-Asp bond to proteolytic cleavage. To test this hypothesis, we used fluorogenic substrates that encompass the NH,-terminal amyloidogenic cleavage sites from wild-type PPP and from the Swedish mutant PPP. Using similar fluorogenic substrates we have previously reported that the amyloidogenic cleavage is unlikely to occur at neutral pH because the only endoproteolytic cleavage detected was not at the amyloidogenic site (23).

FIG. 8. Activity of human brain proteinases and human CD
effect of pH on proteinase activity from brain extract. Conditions were as in Fig. 2; Panel B, chromatograms of EDANS-containing peptides from a 40-min reaction with brain extract (upper) or from a n 8-h reaction with CD (lower). Conditions were as in Fig. 3 Consistent with our previous conclusion, we observed in this work that an acidic proteinase activity is present in human brains that cleaved the Swedish substrate at a rate about 100fold faster than the native substrates and that both substrates were cleaved at the amyloidogenic sites. A similar activity was observed in extracts from AD brains and from bovine brains (not shown). The brain activity is likely t o be CD as indicated by the following observations: 1) the brain activity showed similar cleavage rates with all four substrates and peaked at pH 3.0-3.5, as did CD activity; 2) the brain activity produced a cleavage pattern in three of the substrates that was similar to that of CD; and 3) pepstatin Ainhibited the brain activity against the Swedish substrate with IC,, similar to that of CD. The brain activity against the native substrate consisted of both cysteinyl and aspartyl proteinases, however, it appears that only the aspartyl proteinase produced the cleavage at the amyloidogenic Asp' site. These results suggest that CD produced the amyloidogenic cleavages in both the Swedish mutant PPP and in the wild-type PPP. Thus, the divergence from normal PPP metabolism in Swedish patients is likely to occur at the cleavage step by CD. Furthermore, CD appears to be the amyloidogenic proteinase not only in the Swedish patients with early onset AD, but in all AD patients. The brain extract also had activity that was capable of producing the COOH-terminal amyloidogenic cleavage, which also appears to be CD. Both the brain activity, and purified CD cleaved substrate C at two distinct sites, equivalent to positions Thr43 and Ala4' of AP. The major activity was at Thr43 and accounted for -68% of the total cleavage while the activity at Ala4' accounted for 32%. These ratios did not change even after a long incubation of purified CD with the substrate, indicating that the cleavage at Thr43 precluded the cleavage at Ala4'. However, an additional activity was present in brain extract that degraded the major product further to produce a cleavage at position Ile41. Thus, our observations can account for the ragged COOH terminus of AP. Unfortunately, our data did not provide any information concerning cleavages at positions Val4' or Val3' of AP. A cleavage at the COOH-side of Val2 (Val4' of AP) was not observed, but the absence of a peptide bond on the NH,-terminal side of this residue may have prevented such a cleavage. Longer COOH-terminal substrates that incorporate more upstream residues were synthesized, to allow analysis of cleavages at positions Val3' and Val4', but the solubility of the peptides was too low to permit reliable assays of proteolysis.
The COOH-terminal cleavage site of AP is embedded in the membrane spanning region of PPP which should limit the accessibility of water or the active site of proteinases to the cleavage site. Thus, the cleavage must be facilitated by membrane damage or by dislodging the PPP from the membrane. The possibility that membrane damage occurs prior to AP production is unlikely in some systems, as AP is produced as a part of the normal PPP processing in cultured or transfected cells, and is present in CSF of normal individuals (17,18,25,26). IfAp is produced by the same mechanism in transfected cells, in normal individuals and in AD patients, then PPP is dislodged from an undamaged membrane before the COOH-terminal cleavage can occur. How PPP may be dislodged from the membrane is not known; membrane anchorage and secretory processing of PPP have been shown to be facilitated by the triplet lysine residues at positions 724 t o 726, just down stream of the membrane spanning region (27). Thus, a cleavage at this site may release the PPP molecule into the endosomal milieu after internalization through a clathrin coated pit (7). Alternatively, PPP may be directed into lysosomes with a fragment of a mem- brane targeted for degradation. In this case, lysosomal phospholipases and lipases may degrade the membrane (28, 291, making the PPP accessible to proteolysis. This route does not necessarily involve prior presence of PPP on the plasma membrane, however, it is inconsistent with some data that suggest that AP is generated in an acidic but non-lyosomal compartment (14,15,18). CD is present in clathrin coated vesicles (30), endosomes (31), lysosomes (321, trans-Golgi vesicles (32,331, and transport vesicles (34). Thus, this enzyme would be accessible to PPP in various cell compartments, most likely early endosomes or trans-Golgi vesicles. The role of CD in the secretion of the membrane protein Galpl4GlcNAc a2-64alyltransferase from the trans-Golgi during acute-phase response (33) suggests a similar role in the secretion of PPP. This suggestion is supported by recent observations: 1) potentially amyloidogenic fragments were produced in PPP-transfected cells by an aspartic proteinase (9); 2) secreted PPP from cells transfected with the Swedish mutant, was cleaved at the amyloidogenic site (or P-secretory site (22)) that we find is very susceptible to cleavage by CD; and 3) potentially amyloidogenic fragments and AP are likely to be produced in an acidic compartment (see above). Thus, we propose that the NH,-terminal amyloidogenic cleavage by CD occurs in both wild type and mutant cases, during secretion in the trans-Golgi and gives rise to potentially amyloidogenic fragments. However, in mutant cases, the rate of cleavage at the amyloidogenic site may be higher, resulting the accumulation of AP. In other AD cases, the accumulation of AP may be mediated by different mechanisms such as binding to apolipoprotein-E, which may prevent the clearance of AP from the brain. Since potentially amyloidogenic fragments accumulate in cells transfected with mutant PPP, the COOH-terminal cleavage is likely to occur later, or at a substantially lower rate.
Other proteolytic enzymes present in these compartments may also come in contact with PPP. The susceptibility of the NH,-terminal substrates to proteolysis was tested with a few other lysosomal or aspartyl proteinases (Table I). However, only cathepsin E showed activity against the Swedish substrate that was greater than -100-fold the rate against the native sub-~t r a t e .~ Nonetheless, cathepsin E is present primarily in the stomach and is absent from brain tissue, making it unlikely to be the amyloidogenic proteinase (35).
Our results suggest that CD is capable of producing both the NH,and COOH-terminal cleavages in peptide substrates. If so, it may further digest AP into smaller fragments. Indeed, CD digested AP 1-40 into six distinct fragments, but Ap 1-42 was completely resistant to digestion. AP 1-42 formed aggregates that were about a 100-fold larger than the aggregates of AP 1-40. Thus, the large aggregates of AP 1-42, or their different structure, may prevent the accessibility of the enzyme to cleavage sites within AP.
Our suggestion that CD may be the amyloidogenic proteinase should be tested with additional experiments with "physiological P P P isolated from human brain membranes, in order to establish the role of CD in the formation of AP. If such a role for CD is indeed confirmed, this proteinase may prove to be a viable therapeutic target. A recent report suggesting that there is a balance between formation and clearance of AP (36), supports the notion that inhibition of amyloidogenic proteinases may clear some of the amyloid from patients brains.