Degradation of the Amyloid β-Protein by the Novel Mitochondrial Peptidasome, PreP*♦

Recently we have identified the novel mitochondrial peptidase responsible for degrading presequences and other short unstructured peptides in mitochondria, the presequence peptidase, which we named PreP peptidasome. In the present study we have identified and characterized the human PreP homologue, hPreP, in brain mitochondria, and we show its capacity to degrade the amyloid β-protein (Aβ). PreP belongs to the pitrilysin oligopeptidase family M16C containing an inverted zinc-binding motif. We show that hPreP is localized to the mitochondrial matrix. In situ immuno-inactivation studies in human brain mitochondria using anti-hPreP antibodies showed complete inhibition of proteolytic activity against Aβ. We have cloned, overexpressed, and purified recombinant hPreP and its mutant with catalytic base Glu78 in the inverted zinc-binding motif replaced by Gln. In vitro studies using recombinant hPreP and liquid chromatography nanospray tandem mass spectrometry revealed novel cleavage specificities against Aβ-(1-42), Aβ-(1-40), and Aβ Arctic, a protein that causes increased protofibril formation an early onset familial variant of Alzheimer disease. In contrast to insulin degrading enzyme, which is a functional analogue of hPreP, hPreP does not degrade insulin but does degrade insulin B-chain. Molecular modeling of hPreP based on the crystal structure at 2.1 Å resolution of AtPreP allowed us to identify Cys90 and Cys527 that form disulfide bridges under oxidized conditions and might be involved in redox regulation of the enzyme. Degradation of the mitochondrial Aβ by hPreP may potentially be of importance in the pathology of Alzheimer disease.

Several human disorders are associated with the deposition of aggregated peptides. One of them is Alzheimer disease (AD) 4 in which the polymerization of amyloid ␤-protein (A␤) into insoluble fibrils in brain seems to be a pathological event. An extracellular accumulation of A␤ has been the main focus of molecular studies associated with AD (1). However, increasing attention is directed toward intracellular events including the mitochondrial role in AD (2). There are many links between mitochondrial dysfunctions and AD (3,4). Impairment of mitochondrial energy metabolism and altered cytochrome c oxidase activity are among the earliest detectable defects in AD (5,6). It has been shown that Alzheimer amyloid precursor protein (APP) 695 is not only targeted to the plasma membrane but also to mitochondria (5). Accumulation of APP in the outer mitochondrial membrane caused dysfunctions and impaired energy metabolism. The active ␥-secretase complex including presenilin, nicastrin, APH-1, and PEN-2, which cleave APP to generate A␤, has been shown to be present in the mitochondrial outer membrane (7). Furthermore, the occurrence of A␤ in mitochondria of AD patients and its direct binding to ABAD (A␤binding alcohol dehydrogenase, also called ERAB) induces apoptosis and free radical generation in neurons (8). A recent study demonstrated that A␤ is present in the mitochondrial matrix in AD brains and in brains from transgenic mice overexpressing mutant human APP that impairs neuronal energetics and contributes to cellular dysfunction in AD (9).
We have recently identified a novel mitochondrial metalloendopeptidase in Arabidopsis thaliana, presequence protease AtPreP, which degrades targeting peptides that are cleaved off in mitochondria after completed protein import (10). AtPreP has been shown to degrade not only targeting peptides but also other unstructured peptides up to 70 amino acids residues in length but not small proteins (11,12). Furthermore, AtPreP is also present in chloroplasts, and it uses an ambiguous signal for targeting of the protein to both mitochondria and chloroplasts (13). hPreP is a metalloendopeptidase that belongs to the pitrilysin family (subfamily M16C) of oligopeptidases called also inverzincins, as they contain an inverted zinc-binding motif (14). Moreover, we have determined the crystal structure of AtPreP refined at 2.1 Å (15) that is the first structure of a protein belonging to the pitrilysin family with a bound sub-strate peptide (Protein Data Bank accession code 2FGE). The AtPreP structure revealed a unique totally enclosed large cavity of 10,000 Å 3 that opens and closes in response to peptide binding, revealing a novel catalytic mechanism for proteolysis. We have identified amino acid residues involved in catalysis and shown that amino acids located about 800 residues from the inverted zinc-binding motif contribute to the formation of the active site. As PreP-catalyzed proteolysis occurs in the cavity (like in the proteasome) we refer to the structure of PreP as peptidasome (15).
Interestingly, PreP is an organellar functional analogue of the human insulin degrading enzyme, IDE, that also belongs to the pitrilysin family. IDE has been implicated in AD as it cleaves A␤ before insoluble amyloid fibers are formed (16 -18). Moreover, a genetic deletion of IDE in mice leads to significant elevation in brain A␤ levels (19,20). Up-regulation of IDE in neurons prevented AD-type pathology in transgenic mice overexpressing APP (21). These findings led us to investigate the degradation of A␤ by human PreP, hPreP. hPreP comprises 1037 amino acids and shows 48% sequence similarity (31% identity) to AtPreP. It has been previously identified and referred to as a metalloprotease, hMP1 (22). Intracellular localization of hMP1 was not previously studied. hMP1 has been shown to be widely and highly expressed in different cell lines and tissues on the mRNA expression level.
Here we show that the human PreP is localized to the mitochondrial matrix in mammalian mitochondria where besides presequence and other unstructured peptide degradation it has a novel function, degradation of A␤. We have produced recombinant hPreP and characterized hPreP proteolytic activity both in situ and in vitro. We show that hPreP efficiently degrades different A␤ proteins. Immuno-inactivation studies in human brain mitochondria using anti-hPreP antibodies reveal complete inhibition of the proteolytic activity against A␤ showing that under circumstances when A␤ is present in the mitochondria, hPreP is the protease responsible for degradation of this toxic protein. Furthermore, we have investigated the degradation pattern of the A␤ protein using mass spectrometric analysis. The analysis reveals several unique cleavage sites on A␤-(1-40) and A␤ Arctic protein. Our findings contribute to studies of the mitochondrial component in AD.

EXPERIMENTAL PROCEDURES
Preparation of Human Mitochondria-Post-mortem brain material (cortex) (four cases, age: 58 -91 years; post-mortem delay: 14 -24 h) was obtained from the Huddinge Brain Bank, Karolinska University Hospital, Huddinge, Sweden, and mitochondria were isolated as described by Ankarcrona et al. (23). Ethical approval was received from the local Ethical Committee at Karolinska Institutet Huddinge Universitetssjukhus. 1 g of brain tissue was thawed and homogenized in buffer A (250 mM mannitol, 0.5 mM EGTA, 5 mM HEPES, 0.1% bovine serum albumin, pH 7.4) by 10 strokes at 1500 rpm with a high torque motor-driven pestle. All centrifugation steps were carried out at 4°C. Unbroken cells and cell nuclei were removed by two centrifugations at 600 ϫ g for 10 min. The mitochondrial pellet was obtained by centrifugation at 10,000 ϫ g for 30 min. Mitochondria were resuspended in 20 mM HEPES, pH 7.5, 50 mM KCl, 2 mM EGTA, and Complete TM protease inhibitor mixture (Roche Applied Science) plus 20% glycerol, flash-frozen in liquid N 2 , and stored at Ϫ70°C before use.
Preparation and Subfractionation of Rat Liver Mitochondria-Mitochondria were isolated from liver of 200-g body weight Sprague-Dawley male rats by differential centrifugation technique according to Lapidus and Sokolove (24). To obtain mitoplasts rat liver mitochondria were diluted 10 times with sonication buffer containing 20 mM HEPES-KOH, pH 7.5 (1 mg/ml final concentration) and incubated 20 min at 4°C following centrifugation at 8000 ϫ g for 5 min. The pellet containing mitoplasts was either used directly for determination of rat PreP localization or used for preparation of the inner membrane and matrix fractions, which were obtained after sonication of mitoplasts and subsequent centrifugation at 50,000 ϫ g for 1 h. Supernatant containing outer membrane as well as soluble inter membrane space proteins were further centrifuged at 50,000 ϫ g for 1 h. The pellet contained outer membrane, whereas soluble intermembrane space proteins were concentrated by precipitation with 50% saturated ammonium sulfate, centrifuged for 10 min at 15,000 ϫ g, and suspended to 1 mg/ml in sonication buffer.
Overexpression and Purification of hPreP and Its hPreP-(E78Q) and hPreP(C90S) Mutants-An Escherichia coli overexpression strain, BL21(DE3), was transformed with the pGEX-6P-2 vector containing the predicted mature part (amino acids 27-1036 of hPreP (AAH05025, gi:13477137) fused to glutathione S-transferase (GST) and was grown at 30°C in LB medium. After 4 h, 1 mM IPTG was added to the culture, and the incubation was continued for another 4 h. The cells were pelleted, re-suspended in phosphate-buffered saline buffer (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3), and lysed by addition of 0.5 mg/ml lysozyme and 10 g/ml DNase I followed by sonication for 3 ϫ 30 s. The lysate was centrifuged for 20 min at 15,000 ϫ g and filtered through a 0.2-m membrane. The supernatant was loaded onto a GSTrap FF 1-ml column (Amersham Biosciences) equilibrated with phosphate-buffered saline buffer. The hPreP was eluted after on-column cleavage with PreScission Protease according to the manufacturer's instructions (Amersham Biosciences). The eluted hPreP was applied to a Superdex 200 HR 10/30 (Amersham Biosciences) equilibrated with 20 mM HEPES-KOH, 10 mM MgCl 2 , pH 8.0. The protein content of eluted fractions from GSTrap FF and Superdex 200 HR10/30 was analyzed on 12% SDS-PAGE in the presence of 4 M urea (Laemmli 1970) (25) and stained with Coomassie Brilliant Blue. Mutants of hPreP(E78Q) and hPreP (C90S) were constructed using a QuikChange site-directed mutagenesis kit (Stratagene). The construct was verified by DNA sequencing using a DYEnamic sequencing kit (Amersham Biosciences). The hPreP mutant (E78Q) for production antibodies was overexpressed in E. coli as GST-hPreP fusion, induced with IPTG in the presence of 10 M zinc acetate, and purified as follows: anion exchange on 10-ml ResourceQ15 column was used to capture the fusion protein, then a second anion exchange step on a 6-ml ReSourceQ column was performed to obtain pure fusion protein. Then the fusion protein was desalted into cleavage buffer (Amersham Biosciences) and reacted overnight at 4°C with PreScission protease. The cleaved hPreP mutant was purified by anion exchange on the 6-ml ReSourceQ column and then was loaded onto a 26 ϫ 60 Superdex200 gel filtration column. The gel filtration peaks were concentrated by ultrafiltration to 5 mg/ml.
Degradation Assays-The degradation assay for studying the proteolytic activity of recombinant hPreP contained 1 g purified hPreP and following substrates: 1 g of A␤-(1-40) or A␤-  or A␤-(1-40) E22G in a degradation buffer containing 20 mM HEPES-KOH, pH 8.0, and 10 mM MgCl 2 . For the inhibitory studies, 1 mM PMSF (phenylmethylsulfonyl fluoride), 20 mM oPh (ortho-phenanthroline), 1 mM NEM (N-ethylmaleimide, 100 units ml Ϫ1 apyrase, 50 g/ml bestatin were preincubated with 1 g of hPreP on ice for 10 min before addition of the substrate. The samples were incubated for 1 h at 37°C, and the reactions were stopped by addition of 2 ϫ Laemmli sample buffer, analyzed on 10 -20% Tris-Tricine gels (Bio-Rad), and stained with Coomassie Brilliant Blue. To investigate the degradation of insulin and insulin B-chain by hPreP, 1 g of insulin or insulin B-chain was incubated with 1 g of purified hPreP in the degradation buffer for 1 h at 37°C. The reactions were stopped and analyzed as described above. Proteolytic activity of the PreP cysteine mutant was investigated under oxidizing conditions (2.5 mM K 3 Fe(CN) 6 ) and reducing conditions (2.5 mM DTT). The protease was incubated under the respective conditions at room temperature for 5 min prior to the addition of A␤. The proteolytic activity of the human mitochondrial matrix and the rat mitochondrial subfractions was tested by incubation of the samples with A␤ for 3 h at 37°C. Proteolytic activity in the presence of protease inhibitors was tested after preincubation of the matrix fraction (20 g) with the inhibitors for 10 min at 4°C before addition of 1 g of A␤-(1-40). The reactions were stopped by adding 2 ϫ Laemmli sample buffer, analyzed on 10 -20% Tris-Tricine gels (Bio-Rad), and stained with Coomassie Brilliant Blue.
Intracellular Localization of hPreP-Human brain mitochondria were isolated as described elsewhere by Ankarcrona et al. (23). Human mitochondrial membrane and matrix were obtained by resuspending isolated mitochondria in a sonication buffer. The resuspended mitochondria were incubated on ice for 10 min followed by sonication 5 ϫ 15 s on ice and centrifuged at 18,600 ϫ g to discard the unbroken mitochondria. The supernatant was thereafter ultracentrifuged at 70,000 rpm (Beckman TL-100 Ultracentrifuge, TLA 100.2 rotor) for 45 min. Human brain cytosol, mitochondria, membrane, and matrix were analyzed on 12% SDS-PAGE in the presence of 4 M urea. Immunological cross-reactivity was analyzed by Western blot analysis using nitrocellulose membrane Hybond TM (Amersham Biosciences) with antibodies raised against hPreP, F 1 moiety of the ATP synthase from N. plumbaginifolia, and Tim17 followed by detection with horseradish peroxidase-coupled secondary antibody and ECL (Amersham Biosciences).
Immuno-inactivation-Immuno-inactivation studies were performed as follows: 6 g of antibodies raised against hPreP or 18 g of antibodies raised against presequence of F 1 ␤ subunit of the ATP synthase from Nicotiana plumbaginifolia, pF 1 ␤-(2-54), were preincubated with human mitochondrial soluble fraction at 4°C for 30 min before addition of A␤- . The samples were incubated for 3 h at 37°C, and the reaction was stopped by addition of Laemmli sample buffer and analyzed on 12-20% SDS-PAGE in the presence of 4 M urea. For Western blot analysis, proteins were transferred onto nitrocellulose membrane Hybond TM (Amersham Biosciences) and incubated with anti-A␤ monoclonal antibody 6E10 (Signet Laboratories) at 1:1000. Immunoreactivity was detected with anti-mouse horseradish peroxidase secondary antibody and ELC (Amersham Biosciences).
Mass Spectrometric Analysis; High Performance Liquid Chromatography-MS/MS-The sample was concentrated and desalted using ZipTips (Millipore) and injected onto a 0.1 ϫ 150-mm C18 column (YMC Co., Ltd., Kyoto, Japan) using a nano-LC injector (Valco Instruments) with a 1-ml loop. Peptides were eluted using a water/acetonitrile gradient supplemented with 0.2% formic acid: from 10 -20% acetonitrile in 10 min, 20 -35% acetonitrile in 30 min, and 35-50% acetonitrile in 15 min. The flow rate was 400 nl/min delivered by an Agilent 1100 nanopump (Agilent Technologies), and the column was coupled to an Agilent ion trap mass spectrometer (Agilent) fitted with a nanospray interface. Mass spectra were recorded from m/z 240 to 1800, and the three largest peaks in each spectrum were subjected to MS/MS analysis. The cycle time was 12 s. Peak areas were obtained from the selected ion traces corresponding to the calculated masses of the intact and cleaved peptide.
Homology Modeling of hPreP and Flexible A␤ Docking-The homology model of hPreP was based on the 2.1-Å crystal structure of AtPreP (Protein Data Bank accession code 2FGE (15) using the internal coordinate mechanics energy optimization method (26,27) and a sequence alignment generated by CLUSTALW (28). The model comprises all atoms including hydrogens and was created by the zero end gap dynamic programming algorithm, where the backbone and conserved side chains adopt the same conformation as the template. Loop regions were subjected to search against a data base of loop structures from the Protein Data Bank, and loops with the closest matching sequences and loop end positions are inserted into the homology model. The bound A␤-(12-17) peptide was created by substituting the side chains of the unidentified peptide found in the AtPreP crystal structure followed by several rounds of energy optimization implemented by Internal Coordinate Mechanics to allow the peptide to adopt a conformation that fit the active site of the hPreP model. The coordinates of AtPreP have been deposited in the Protein Data Bank under the accession code 2FGE (15).

Degradation of A␤ Proteins by hPreP-
The recombinant hPreP was cloned as a fusion protein with GST, overexpressed in E. coli, and purified to homogeneity on GSTrap FF column after cleavage with PreScission protease (Fig. 1A). hPreP has a molecular mass of 114 kDa. The "light" band at about 70 kDa corresponds to DnaK indicating that small quantity of this bacterial molecular chaperone co-purifies with hPreP. The recombinant hPreP shows proteolytic activity against different A␤ proteins. Fig. 1B shows SDS-PAGE of amyloid ␤-proteins deg-radation by hPreP. hPreP completely degraded both A␤-  and A␤-(1-42) as well as A␤ Arctic protein (1-42 E22G), a protein that causes an AD-like pathology, with an approximate activity of 0.053 g of A␤/min/g of hPreP. To compare proteolytic activities of hPreP and IDE, the degradation of insulin and insulin B-chain by hPreP was investigated. hPreP degraded completely the insulin B-chain but was not capable of degrading insulin (Fig. 1C).
Investigation of the effect of different types of protease inhibitors on the degradation of A␤-(1-40) by the recombinant hPreP showed that neither PMSF nor bestatin (i.e. serine or aminopeptidase type protease inhibitors) affected proteolysis (Fig.  1D), whereas NEM, a cysteine-type protease inhibitor, showed a small inhibitory effect of 10%, and the metalloprotease inhibitor oPh completely inhibited degradation of A␤, demonstrating that hPreP is a thiolsensitive metalloprotease. This is in agreement with studies on hMP1, which showed that hMP1 was efficiently inhibited by oPh and p-chloromercuriphenylsulfonic acid (22). Apyrase had no effect on the proteolytic activity showing that A␤ degradation is independent of ATP. hPreP contains an inverted zincbinding motif, H 75 ILE 78 H 79 . The importance of the metal-binding motif for the proteolytic activity has been demonstrated by studying the recombinant hPreP mutant, hPreP(E78Q), in which the catalytic base Glu 78 was changed to Gln (Fig.  1E). We have overexpressed and purified the mutant and showed that hPreP(E78Q) could not degrade (only 10% was degraded) the A␤-(1-40) protein (Fig. 1F) confirming the importance of the inverted zinc-binding site for the proteolytic activity. The degradation pattern of A␤-(1-40) and A␤ Arctic was studied using liquid chromatography nanospray tandem mass spectrometry (LC-MS/MS) (Fig. 2). The base peak chromatogram shown in Fig. 2B also shows a schematic representation of the cleavage sites by hPreP (filled arrows) on A␤-(1-40) and A␤ Arctic Fig. 2, A and C). The overall degradation resulted in the production of several 4 -14 amino acid fragments that are  Fig. 2A). The major cleavage product is the short very hydrophobic peptide, Leu 17 -Val 18 -Phe 19 -Phe 20 . hPreP degraded 〈␤ Arctic at cleavage sites that generally overlapped with those found for wild type A␤. However, two new cleavage sites appeared, after Gly 29 and Gly 38 (gray arrows).
Mitochondrial Localization of hPreP and Its Function-The bioinformatic programs created to predict intracellular localization of proteins, Mitoprot, Predotar, and TargetP localize hPreP to mitochondria with a high score (0.9888 for MitoProt and 0.877 for TargetP). However, the predicted presequence of hPreP is much shorter (29 amino acids) than for the plant homologue AtPreP (85 amino acids). The yeast homologue of PreP, MOP112, is not predicted to be a mitochondrial protein but was shown to be located in the mitochondrial intermembrane space (29). To confirm the localization of hPreP to mitochondria, we isolated human brain mitochondria and the cytosolic fraction and tested immunological cross-reactivity with antibodies against hPreP and the F 1 moiety of the ATP synthase. Western blot analysis verified mitochondrial localization of hPreP (Fig. 3A). Furthermore, hPreP was shown to be located in the soluble fraction of human brain mitochondria, no cross-reactivity was found in the membrane fraction probed with anti-Tim17 antibodies (Fig. 3B). The soluble fraction of human brain mitochondria showed proteolytic activity against A␤ that was almost completely inhibited by oPh and partially (20%) by NEM but not by PMSF or apyrase (Fig. 3C).
To investigate intramitochondrial localization of PreP we have used rat liver mitochondria due to a limited amount of human material. Rat PreP shows 85% sequence identity (93% similarity) to hPreP and is predicted to be a mitochondrial protein with a 28-amino acid-long presequence. Consistent with human and rat PreP, the predicted presequences of PreP from mouse, dog, and orangutan (Pongo pygmaeus) are estimated to be 29 amino acid residues in length. We have prepared highly purified rat liver mitochondria free from microsomal contamination and we have osmotically ruptured the outer mitochondrial membrane to obtain mitoplasts. Mitochondria were fractionated into OM, IM, IMS, and Ma. Western blot analysis of mitochondria and mitoplasts after proteinase K treatment clearly showed intramitochondrial localization of PreP. Analysis of the mitochondrial subfractions with compartment-specific antibodies against porin (OM), Tim17 (IM), Omi (IMS), and GRP75 (Ma) clearly showed that PreP was localized to the mitochondrial matrix (Fig. 3D). Insignificant amounts of porin in the IM might be due to the presence of "contact points" between the OM and the IM. The "light" band in the IMS does not have the same molecular mass as porin. Cross-reactivity detected with anti-Omi in the Ma might reflect residual electrostatic association of Omi with mitoplasts and its dissociation to the Ma fraction during preparation.
The involvement of hPreP in A␤ proteolysis in organello was verified by immuno-inactivation assay using human brain mitochondria. When the soluble human brain mitochondrial fraction was preincubated with antibodies against hPreP, the degradation of A␤-(1-40) was completely abolished (Fig. 4), whereas in the presence of antibodies against the F 1 ␤ presequence no effect on the proteolytic activity was detected. These results show that hPreP is the protease responsible for degradation of A␤ in mitochondria under conditions when A␤ accumulates in mitochondria as reported in recent studies (8,9).
Structural Model of hPreP with A␤ Hexapeptide-The sequence of hPreP is 31% identical to AtPreP (11), and the protein has a similar function, thus their structures are likely to be similar. A model of hPreP was generated by homology with the program ICM (26,27) using the 2.1 Å resolution structure of AtPreP as a template (15). In a homology model the main chain of the core protein is normally correct as is the position of most residues. However, the conformation of loops and side chains are harder to predict since they often differ between closely related proteins such as the PreP family. PreP comprises four topologically similar domains organized in two halves connected by a hinge region, which enclose a large proteolytic chamber wherein the active site resides (Fig. 5A). The active site is highly conserved with very few amino acid substitutions compared with AtPreP. The N-terminal domain contains the inverted zinc-binding motif and the distal glutamate at position 176, H 75 ILE 78 H 79 X(97)E 176 . Tyr 878 in the C-terminal domain separated by 800 residues from the zincbinding motif completes the active site. A few small differences between human and plant PreP are found in the S3 and S1Ј pockets. The crystal structure of AtPreP comprises a six-residue peptide in the active site, which binds in an extended conformation forming a short antiparallel ␤-strand to the enzyme. We modeled the human substrate A␤-(12-17) based on this peptide, and its position was refined to fit to the active site of the hPreP model (Fig. 5B). However, it should be noted that hPreP degrades several different substrates at multiple sites, so the model simply illustrates where the substrate is located during proteolysis.
Interestingly, in the homology model of hPreP two cysteine residues at positions Cys 90 in the first domain and Cys 527 at the hinge region are in close proximity to each other. These cysteines are not present in the AtPreP structure. Measurement of the proteolytic activity of hPreP in the presence of K 3 Fe(CN) 6 shows complete inhibition under oxidizing conditions, indicating that these cysteines can form a disulfide bridge locking the enzyme in a closed conformation and thus hinder substrate binding (Fig. 5C). To confirm involvement of the listed cysteines in disulfide bridge formation, we have produced a mutant, in which Cys 90 was changed to Ser, hPreP(C90S). The cysteine mutant, in contrast to the wild type hPreP, was active under oxidizing conditions showing that Cys 90 was indeed involved in formation of the disulfide bridge locking the enzyme in a closed inactive conformation (15) (Fig. 5D). This finding is interesting and might be of physiological importance as it implies a possible inhibition of the enzyme under conditions of elevated ROS production in mitochondria.

DISCUSSION
Our findings demonstrate for the first time that the newly identified mitochondrial metalloendopeptidase, called hPreP peptidasome, is present in the mitochondrial matrix of human brain mitochondria and that the enzyme, in addition to its previously identified function, degradation of the mitochondrial presequences and other unstructured peptides, is also putatively responsible for the degradation of the amyloid ␤-protein.
Accumulation of A␤ in mitochondria under pathological conditions associated with AD has been recently reported (8,9). hPreP is a metalloendopeptidase that belongs to pitrilysin subfamily M16C. The pitrilysin M16A subfamily includes several peptide-degrading oligopeptidases, such as IDE, di-basic convertase Nardilysin (NRD), and bacterial pitrilysin (protease III), that are monomers of similar size as PreP (about 100 kDa). Members of the M16B subfamily are heterodimers formed by two 50-kDa subunits, e.g. the mitochondrial processing peptidase that accepts bigger substrates cutting off targeting peptides from precursor proteins. IDE controls levels of insulin and degrades a wide range of other physiological peptides including glucagon, transforming growth factor-␣, ␤-endorphin, and amylin (30). NRD convertase is processing a number of neuropeptides including opioid peptides. This indicates that both IDE and NRD convertase play an important role in modulation of regulatory peptides levels in vivo. IDE and pitrilysin have also been shown to cleave A␤ protein (17,31,32) therefore leading us to the investigation of A␤ degradation by hPreP.
hPreP efficiently degrades different A␤ proteins including A␤-(1-40), A␤-(1-42), and A␤ Arctic. Screening of the hPreP cleavage sites on the A␤ protein shows that in most cases hydrophobic and small uncharged amino acids and also a positively charged amino acid are found in P 1 Ј and P 1 positions. It is in agreement with our previous subsite specificity study for cleavage of mitochondrial presequences by AtPreP showing that the enzyme has preference for basic amino acids in the P 1 Ј position and small uncharged amino acids or serines in the P 1 position (10). Substrates to IDE show little or no sequence similarities, but the enzyme also exhibits preference for basic and/or large hydrophobic residues (33,30). Comparison of our results with the results previously described for IDE (34) shows, however, that only two cleavage sites produced by hPreP overlapped with those found for IDE, Phe 19 2Phe 20 and Phe 20 2Ala 21 (34). All the other cleavage sites on A␤-(1-40) seem to be unique for hPreP (Gln 15 2Lys 16 , Lys 16 2Leu 17 , Ala 30 2Ile 31 , Gly 33 2Leu 34 , and Leu 34 2Met 35 ). Additionally, two new cleavage sites appeared on A␤ Arctic (Gly 29 2Ala 30 and Gly 38 2Val 39 ). Interestingly, the main difference in subsite specificity between hPreP and IDE on A␤ proteins is that hPreP   1 Å (15). A, structural model of hPreP with A␤ (12)(13)(14)(15)(16)(17) bound to the active site. B, close-up of the active site with residues that differ in AtPreP shown in parentheses. C, proteolytic activity of hPreP against A␤ in the presence of DTT or K 3 Fe(CN) 6 . D, proteolytic activity of hPreP mutant hPreP(C90S) against A␤ in the presence of DTT or K 3 Fe(CN) 6 or oPh.
has several cleavage sites after Gly 29 in a very hydrophobic C-terminal 〈␤-(29 -42) segment of A␤. This segment is prone to aggregation adopting exclusively a ␤-sheet conformation (35). Despite the fact that both PreP and IDE belong to the pitrilysin protease family, the overall sequence similarity of hPreP and IDE is very low and might explain different recognition patterns. There is only a 28-amino acid residue stretch in IDE around the zinc-binding motif that shows 38% sequence identity (55% similarity) to hPreP. Another difference, evident between hPreP and IDE, is that hPreP is not capable of degrading insulin (cf. Fig. 1C), while the insulin B-chain was completely degraded. Insulin A-and B-chains are linked by two disulfide bonds, and an additional disulfide is formed within the A-chain. The A-chain consists of 21 amino acids and the B-chain of 30 amino acids. As PreP degrades peptides up to 70 amino acid residues, the size of insulin is appropriate for degradation but the folding status seems to limit accessibility to the active site. Neither the de novo peptide Ala-␣ 3 consisting of 66 amino acids that is tightly folded into a three helical bundle structure was cleaved by PreP (11). This further confirms that PreP is inactive against folded small proteins (10).
It can be also mentioned that hPreP is encoded by a single gene PITRM1 that is located in chromosome region 10p15.2 and contains 30 exons. Also IDE is located on chromosome 10 (10q) and genetic association between single nucleotide polymorphisms (SNPs) in IDE and late-onset of AD has been reported (36). Interestingly, preliminary SNP searches in the public data bases (ENSEMBL and NCBI) of the hPreP gene, PITRM1, revealed several SNPs in the vicinity of the active site. 5 The relevance of this finding for hPreP proteolytic activity is at present unknown, but localization of SNPs in exons, close to the active site, makes these mutants interesting for further studies.
What is the intramitochondrial localization of hPreP and what is its function within the organelle? hPreP is a soluble protein with an isoelectric point of 6.57. Our immunological cross-reactivity studies clearly show that hPreP is localized to the mitochondrial matrix. Matrix localization is also found in plants (11). 6 Interestingly, the yeast homologue of PreP, MOP112 has been shown to be located in the intermembrane space (29). On the other hand, the first genome-wide screen of protein complexes in budding yeast localizes the product of the Cym1 gene corresponding to the MOP112 protein in complexes localized to the mitochondrial matrix (37). Comparison of the N-terminal sequences of human, other mammalian, and yeast PreP shows differences that may explain varying intramitochondrial localization. Whereas the mammalian presequences are predicted to be 28 -29 amino acids long and clearly classified as the mitochondrial targeting peptides (MitoProt), the yeast PreP is not predicted to be a mitochondrial protein, and it contains only a short, seven-amino-acid long N-terminal extension in comparison with the predicted mammalian mature forms of PreP. Proteolysis is crucial for maintaining mitochondrial morphology and function as it removes misfolded or damaged proteins and peptides and protects organelles from the toxicity of potentially harmful peptides.
Mitochondrial ATP-dependent proteases, the AAA proteases (ATPases associated with a number of cellular activities), Lon and ClpP generate peptides of 5-20 amino acid residues, whereas the Oma1 protease produces fragments of ϳ20 kDa (38,39). These peptides are subsequently degraded to smaller peptides or free amino acids by the ATP-independent proteases, such as the matrix hPreP or intermembrane space yeast MOP. Also the cleaved targeting peptides, accumulated A␤, and other toxic peptides will be degraded by PreP. It indicates thus that PreP is a general mitochondrial peptidase clearing the organelle from different peptides susceptible for proteolysis. Immuno-inactivation studies showed that PreP is an A␤ degrading protease in mitochondria (cf. Fig. 4). Our previous immuno-inactivation studies with S. oleracea mitochondria showed that PreP is also responsible for degradation of targeting peptides in mitochondria (11). Interestingly, an isoform of IDE with an N-terminal mitochondrial targeting sequence generated by translation at an in-frame initiation codon upstream of the canonical translation start has been recently identified, localized to mitochondria in cell lines, and shown to degrade mitochondrial targeting peptides (40). Previously, IDE has been reported to be present in other subcellular compartments including peroxisomes, endosomes, and the nucleus (41). Relative levels of IDE expression in different organellar compartments have not been reported.
The recently solved crystal structure at 2.1 Å resolution of AtPreP allowed molecular modeling of hPreP and identification of amino acids involved in hPreP catalyzed proteolysis, as discussed above. Furthermore, the structure reveals closed conformation of the enzyme forming an enclosed chamber in which the proteolysis occurs, which explains how these type of unspecific proteases discriminate between substrate peptides and larger folded proteins. The chamber is formed by two enzyme halves that come together and the volume of the chamber limits the substrate size to peptides and prevents unintentional degradation of proteins. Based on the structure and mutagenesis experiments with double cysteine mutants whose activity is inhibited under oxidizing conditions, we have proposed a novel mechanism for PreP proteolysis involving hinge-bending motions in an opening and closing cycle that is controlled by substrate binding (15). Unexpectedly, we have found that hPreP contains two native cysteine residues at position Cys 90 in the first domain and position Cys 527 at the hinge region in close proximity to each other. Notably, these cysteine residues are conserved in all known mammalian PreP sequences as well as in the yeast PreP homologue. Oxidation of hPreP resulted in the complete inhibition of the proteolytic activity against the A␤ protein, indicating formation of a disulfide bridge between the cysteines that locks the enzyme in a closed conformation and hinder substrate binding. Indeed, the cysteine mutant, hPreP(C90S), was active under oxidizing conditions showing that Cys 90 was involved in formation of the disulfide bridge locking the enzyme in a closed inactive conformation (15). Furthermore, importance of the status of cysteine residues for the proteolytic activity of hPreP may explain partial inhibitory effect of NEM on hPreP activity. As NEM is a cysteine-modifying agent, its substitution may cause steric hindrance during proteolysis (15). Importance of the redox status of cysteine res-idues for the proteolytic activity implies a possible inhibition of the enzyme in mitochondria under conditions with elevated ROS production that would additionally increase mitochondrial dysfunctions. The physiological consequences of this finding will be studied further.
In summary, mechanisms through which intracellular A␤ impairs cellular properties causing neuronal dysfunctions are not fully understood. Besides an immense number of studies showing extracellular accumulation of A␤, it has been shown that A␤ also accumulates in the brain mitochondrial matrix of AD patients and of transgenic mice overexpressing A␤ precursor protein (8,9) and causes perturbations of mitochondrial functions. Recent studies in which A␤ was incubated with isolated rat brain and muscle mitochondria show the accumulation of monomeric A␤ within the mitochondria, an increase in mitochondrial membrane viscosity with a concomitant decrease in ATP/O ratio, respiratory chain complexes inhibition, a potentialization of ROS production, and cytochrome c release (42). On the basis of the results discussed above we propose that the newly identified mitochondrial peptidasome, hPreP, functions as a peptide scavenger in the mitochondrial matrix clearing mitochondria from toxic peptides and protecting them against pathogenic peptide intruders. We speculate that age-dependent reduction of PreP proteolytic activity in mitochondria caused by mutations or other factors affecting the hinge-bending opening and closing cycle of the enzyme may lead to accumulation of toxic peptides and age-dependent onset of mitochondrial dysfunction.