Isolation and Identification of a Novel Mitochondrial Metalloprotease (PreP) That Degrades Targeting Presequences in Plants*

Most of the nuclear encoded mitochondrial precursor proteins contain an N-terminal extension called the presequence that carries targeting information and that is cleaved off after import into mitochondria. The presequences are amphiphilic, positively charged, membrane-interacting peptides with a propensity to form α-helices. Here we have investigated the proteolysis of the presequences that have been cleaved off inside mitochondria. A presequence derived from the overexpressed F1β subunit of the ATP synthase and specific synthetic fluorescent peptides (Pep Tag Protease assay) have been shown to undergo rapid degradation catalyzed by a matrix located protease. We have developed a three-step chromatographic procedure including affinity and anion exchange chromatography for isolation of the protease from potato tuber mitochondria. Two-dimensional gel electrophoresis of the isolated proteolytically active fraction followed by electrospray ionization-mass spectrometry/mass spectrometry and data base searches allowed identification of the presequence peptide-degrading protease inArabidopsis thaliana data base as a novel mitochondrial metalloendoprotease with a molecular mass of 105 kDa. The identified metalloprotease contains an inverted zinc-binding motif and belongs to the pitrilysin family.

Most mitochondrial proteins are encoded by nuclear genes, synthesized on cytosolic polyribosomes as precursor proteins, and imported into mitochondria post-translationally. Precursor proteins contain a cleavable peptide extension at the N terminus, called presequence or targeting or signal peptide (for reviews see Refs. 1 and 2). The presequences contain information necessary for targeting and are required at all steps along the mitochondrial protein import pathway (3,4). In the cytosol, nascent polypeptide chains and presequences interact with molecular chaperones. Presequences are further recognized by specific receptors on the outer membrane of mitochondria (5,6). The protein import machinery of mitochondria, the translocases of the outer and the inner membrane, contain a series of proteins that successively interact with the presequences (4,7). The mitochondrial Hsp70 is the primary target protein for the presequence protruding into the matrix (8). Finally, the presequence is cleaved off by the mitochondrial processing peptidase (MPP) 1 (9,10), and the mature protein is folded and assembled upon action of molecular chaperones into oligomeric protein complexes.
The great majority of mitochondrial presequences (Ͼ80%) are in the range of 20 -60 residues; however, the length varies substantially from 6 residues for ATP synthase F chain from yeast to 136 residues for Glycine max cytochrome c oxidase subunit 2 (Cox2) (11,12). Presequences do not show amino acid sequence similarity, but they have characteristic physicochemical properties. They are enriched in positively charged, hydroxylated, and hydrophobic residues and have the potential to form an amphiphilic ␣-helical structure (13,14). Synthetic peptides derived from mitochondrial presequences have been shown to bind to lipid vesicles containing anionic phospholipids as amphiphilic helical structures (15)(16)(17). The NMR structure of the Tom20 receptor in complex with a presequence peptide revealed that the bound presequence forms an amphiphilic ␣-helical structure with a hydrophobic surface that interacts with the hydrophobic patch in the Tom20 groove (18).
Several research groups (19 -21) have reported severe effects of presequence peptides on the structural and functional integrity of mitochondrial membranes. Presequence peptides can perturb natural and artificial lipid bilayers. Addition of presequences to mitochondria results in membrane lysis, uncoupling of respiration, and dissipation of the membrane potential (14,22,23). The mechanism of action for membrane-permeabilizing peptides is not clear, but it has been proposed that the presequence peptides induce a channel (24) or that the peptides themselves form a pore (25). Furthermore, peptides derived from mitochondrial presequences have been shown to be lethal toward Bacillus megaterium at micromolar concentrations (26). To avoid the potential harmful interactions of presequences with mitochondrial membranes, presequences cleaved off from precursor proteins inside mitochondria should be removed either by degradation inside mitochondria or by export out from the organelle.
By using chemical quantities of the presequence of the ATP synthase F 1 ␤ subunit, we have reported previously (27) experimental evidence for a rapid degradation of the presequence in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed. organello after in vitro import and processing. There are a few reports suggesting that chemically synthesized presequence peptides are degraded inside mitochondria (28,29). The synthetic presequence peptide derived from CoxIV was imported into rat mitochondria and found to be associated with the membrane; however, the authors suggested that the peptide was further translocated to the matrix and degraded over time (29). Another chemically synthesized presequence peptide derived from cytochrome P-450 was accumulated inside mitochondria only in the presence of a chelating agent, o-phenanthroline (28 -30). Subunit 9 of the bovine cytochrome bc 1 complex corresponds to the presequence of the bovine Rieske iron-sulfur protein and is the only known example of a presequence that is retained in the mitochondria and integrated as a subunit of an oligomeric membrane-bound protein complex (31). The precursor of the bovine Rieske iron-sulfur protein is targeted and assembled into the bc 1 complex before the presequence is cleaved off, presumably by the core proteins in the bc 1 complex (31)(32)(33).
Several ATP-dependent proteases have been identified in mitochondria that selectively remove non-assembled or misfolded polypeptides (34). The quality control of the mitochondrial inner membrane is ensured by the ATP and metaldependent AAA proteases, Yme1p, Yta10p, and Yta12p, homologues of the bacterial FtsH proteases (35). Yme1p, Yta10p, and Yta12p have been identified in Neurospora crassa, yeast, and mammalian mitochondria (36 -38). The PIM1 protease, a homologue of the bacterial Lon protease, is located in the mitochondrial matrix and has overlapping substrate specificities with the AAA proteases (39,40). Lon-like proteases have been found in the mitochondria of yeast and mammals (41,42). Maize mitochondria contain a Lon protease that is loosely associated to the inner membrane (43). Homologues of the bacterial ClpP protease have also been identified in the matrix of mammalian and plant mitochondria (44,45). However, no proteolytic function or natural substrates have been identified for ClpP in mitochondria. Both PIM1 and ClpP are ATP-dependent serine-type proteases. All the ATP-dependent proteases in mitochondria are assembled into large multimeric protein complexes and harbor not only the proteolytic but also chaperone activity (45,46). Studies from our laboratory (27) revealed that the presequence of F 1 ␤ undergoes rapid degradation catalyzed by a matrix located protease(s). Here we have developed a threestep chromatographic procedure for isolation of a protease involved in presequence degradation. The proteolytic activity was measured by immunological detection of the F 1 ␤ presequence degradation and by cleavage of synthetic fluorescent peptides. Two-dimensional gel electrophoresis of the isolated fraction followed by ESI MS/MS and data base searches allowed identification of the presequence peptide-degrading protease in Arabidopsis thaliana data base as a novel mitochondrial metalloprotease. The identified metalloprotease contains an inverted zinc-binding motif and belongs to the pitrilysin family.

EXPERIMENTAL PROCEDURES
Purification of a Novel Metalloprotease from Solanum tuberosum Mitochondria-Potato tuber mitochondria were isolated as described previously (47). The mitochondria were diluted with 20 mM HEPES-KOH (pH 8.0) and 30 mM MgCl 2 to a protein concentration of 8 mg/ml and sonicated for 3 ϫ 15 s at 4°C. Matrix was separated from membranes by centrifugation at 200,000 ϫ g for 30 min. The supernatant was filtered through a 0.2-m membrane and applied to an arginine-Sepharose (Amersham Biosciences) affinity column equilibrated with 20 mM HEPES-KOH (pH 8.0). Bound proteins were eluted from the column with a linear gradient of 0 -0.5 M NaCl. The proteolytically active fractions were pooled, desalted on a PD-10 column, loaded onto a Mono Q HR 5/5 (Amersham Biosciences), and equilibrated with 20 mM HEPES-KOH (pH 8.0). Bound proteins were eluted from the Mono Q anion exchanger with a linear gradient of 0 -0.4 M NaCl. The active fraction was desalted on a PD-10 column and applied to a Mini Q PE 4.6/50 (Amersham Biosciences), equilibrated with 20 mM HEPES-KOH (pH 8.0). Bound proteins were eluted from the Mini Q anion exchanger with a linear gradient of 0 -0.4 M NaCl. The active fraction eluted from the Mini Q column was desalted on PD-10 and concentrated.
Two-dimensional Gel Electrophoresis-Samples containing proteolytic activity (150 l, 1.8 mg/ml) were diluted to a final volume of 250 l with a solution containing 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 100 mM DTT, and 2% (v/v) carrier ampholytes (IPG buffer (pH 3-10), Amersham Biosciences) and applied to a nonlinear IPG strip, pH 3-10. The strips were allowed to rehydrate overnight at 20°C and were then transferred to IPG strip holders and covered with paraffin oil. Proteins were focused for 1 h at 150, 500, and 1,000 V followed by a stepwise increased gradient to 8,000 V. The isoelectric focusing was then completed at a constant voltage of 8,000 V until 80,000 V-h were reached. Subsequently, the strips were first equilibrated for 15 min in 50 mM Tris-HCl (pH 6.8), 6 M urea, 30% (v/v) glycerol, 2% SDS (w/v), and 10 mg/ml DTT and then for 10 min in the same buffer without DTT but with 25 mg/ml iodoacetamide and a trace of bromphenol blue. In the second dimension, 12% SDS-PAGE was carried out according to Laemmli (48). Proteins were stained with Coomassie Brilliant Blue R-250.
In-gel Digestion, Sample Preparation, Mass Spectrometry, and Data Base Search-Protein spots were excised from the Coomassie-stained gel and put in a 96-well PCR plate. In-gel digestion was performed according to Lanne et al. (49) with some modifications as follows: gel pieces were washed with water and 35% acetonitrile followed by dehydration in a SpeedVac vacuum evaporator. The gel plugs were digested with trypsin and extracted by addition of 1% formic acid, 1% acetonitrile. The digests were then analyzed by MALDI-TOF MS on a Voyager-DE STR mass spectrometer in reflector mode (Perspective Biosystems) with ␣-cyano-4-hydroxycinnamic acid as matrix. To achieve accurate mass determination, calibration was automatically performed using tryptic auto-digestion products as internal standards for each sample. The obtained mass spectra were annotated in Data Explorer, and the latest versions of NCBI nonredundant data bases were searched with the resulting peptide mass list, using the search engine MASCOT (www.matrixscience.com) or Profound (prowl.rockefeller.edu/cgi-bin/ ProFound). The remaining peptide extracts were purified and concentrated on a column of Poros 20R packing material (Applied Biosystems). About 5 l of sample was applied onto the column and washed with 5% formic acid, and the peptides were eluted with 2.5 l of 65% acetonitrile and 0.5% formic acid directly into the nanospray needle (Protana A/S). The samples were analyzed by electrospray MS using a quadrupole time-of-flight mass spectrometer (Micromass) equipped with a nanospray interface. The MS/MS spectra were either used to search NCBI nonredundant data bases with MASCOT directly or interpreted manually using MassLynx and Pepseq (Micromass).
Substrate Preparations-The precursor of the F 1 ␤ subunit of the mitochondrial ATP synthase, pF 1 ␤, from Nicotiana plumbaginifolia was overexpressed in Escherichia coli BL21(DE3). The expressed protein was insoluble and present in inclusion bodies. Isolated inclusion bodies were solubilized in 70% formic acid and chemically cleaved with CNBr after methionine residues. The sample was evaporated and redissolved in 10 mM Tris-HCl (pH 8.0), 4 M urea, 1 mM DTT and loaded onto a Mini S column. Elution was performed with a linear gradient of 0 -0.5 M NaCl. Fractions containing the 15-kDa N-terminal fragment of pF 1 ␤, N 15 pF 1 ␤, were collected. This polypeptide contains all information necessary for mitochondrial targeting, import, and processing (47). N 15 pF 1 ␤ (0,01 nM) was cleaved with purified MPP⅐bc 1 complex from Spinacia oleracea to generate the presequence of N 15 pF 1 ␤. The presequence is 54 amino acid residues long, has a molecular mass of 5.7 kDa, and is referred to as N 5.7 pF 1 ␤-(2-54) (27). The presequence with a modified C-terminal carboxylic group was prepared from a mutant pF 1 ␤, in which a methionine was inserted at position ϩ1 downstream of the MPP-processing site. The mutant protein was overexpressed in E. coli and treated with CNBr to cleave the presequence after the introduced methionine. During the cleavage reaction the methionine was converted into a homoserine lactone (hsl) group to generate N 5.7 pF 1 ␤-(2-54)-hsl (50). Purification of N 5.7 pF 1 ␤-(2-54)-hsl was performed as described above for N 15 pF 1 ␤.
Protease Activity-The proteolytic activity was determined by degradation of N 5.7 pF 1 ␤-(2-54) that was generated by cleavage of N 15 pF 1 ␤ with MPP⅐bc 1 complex. For the proteolytic reaction, precleaved N 15 pF 1 ␤ (0,01 nM) was incubated with mitochondrial matrix (33 g) or isolated chromatographic fractions (30 l) in a buffer containing 20 mM HEPES-KOH (pH 8.0), 1 mM MnCl 2 , and 1% (v/v) Triton X-100 for 30 min at 30°C. The reaction was stopped by addition of sample buffer and analyzed on 12-20% SDS-PAGE in the presence of 4 M urea (48). Immunological cross-reactivity was analyzed by Western blotting using Hybond ECL that was immunodecorated with antibodies raised against the C-terminal part of the presequence (residue 38 -54) of the pF 1 ␤, followed by detection with horseradish peroxidase-labeled secondary antibodies. Degradation of N 5.7 pF 1 ␤-(2-54)-hsl was studied to assess the role of the free C terminus in proteolysis. The degradation assay contained N 5.7 pF 1 ␤-(2-54)-hsl (0,01 nM) and the isolated fraction from the Mini Q column (30 l) and was carried out for 30 min at 30°C. The reaction was stopped, and products were analyzed as described for the N 5.7 pF 1 ␤- . To characterize the protease, degradation of synthetic peptides was also studied. The degradation assay contained 0.6 g of the synthetic fluorescent peptides, P1 or P2 (Pep Tag Protease assay, SDS, Promega), and the isolated fraction from the Mini Q column (30 l). Degradation was carried out at 30°C for 30 min, and 80% glycerol was added. The samples were analyzed directly on a 1% agarose gel, and the fluorescent peptides were visualized by UV light. The effect of protease inhibitors and ATP dependence of the proteolytic activity was investigated using 1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, 10 mM ortho-phenanthroline, 1 mM carboxypeptidase inhibitor (catalogue number 217359, Calbiochem), or 40 units/ml apyrase. The isolated fraction from the Mini Q column was preincubated for 10 min at 4°C with protease inhibitors before addition of substrates. Both N 5.7 pF 1 ␤-(2-54) and the P1 peptide were used as substrates. To test proteolytic activity in the presence of inhibitors, samples were incubated and analyzed as described before.
Protein Content-The polypeptide content of the mitochondrial matrix and isolated chromatographic fractions was analyzed by 12-20% SDS-PAGE in the presence of 4 M urea (48). Proteins were stained with silver.

Purification of a Mitochondrial Protease That Degrades
Mitochondrial Presequences-In order to assay degradation of mitochondrial presequences, we have prepared chemical amounts of a truncated precursor derived from the overexpressed full-length precursor of the F 1 ␤ (pF 1 ␤) subunit of the ATP synthase. The 15-kDa N-terminal fragment of pF 1 ␤, designated as N 15 pF 1 ␤, consisted of a 53-amino acid long presequence and the first 82 amino acids of the mature form and was detected by antibodies directed against the C-terminal part of the presequence (47). The precursor protein was incubated with the isolated spinach MPP⅐bc 1 complex (51) to generate the presequence peptide of 5.7 kDa, N 5.7 pF 1 ␤-(2-54) (Fig. 1B, lane  1). The proteolytic activity was found to be localized to the matrix of mitochondria (27). We have developed a three-step procedure for purification of a protease involved in presequence degradation. The procedure was based on affinity chromatography, followed by anion exchange chromatography. Mitochondrial matrix from S. tuberosum was applied to an affinity column, arginine-Sepharose. The majority of proteins did not bind to the column (flow-through). Bound proteins were eluted with a linear gradient of 0 -0.5 M NaCl and recovered in three protein peaks, A1 and A2 at 0.1-0.25 M NaCl and A3 at 0.25-0.35 M NaCl (Fig. 1A). When the isolated fractions were analyzed for proteolytic activity using N 5.7 pF 1 ␤-(2-54) (Fig. 1B), the proteolytic activity was found in fractions A1 and A2 (Fig.  1B, lanes 4 and 5). The protein content of eluted fractions analyzed by SDS-PAGE revealed a similar polypeptide pattern in fractions A1 and A2 (not shown) that were pooled, desalted, and loaded onto a Mono Q column. Proteins eluted from the   Mono Q with a linear gradient of 0 -0.4 M NaCl ( Fig. 2A, upper panel) exhibited two protein peaks, MoQ1 and MoQ2. The proteolytic activity was eluted at 225-275 mM NaCl in fraction MoQ1 (Fig. 2B, lane 3). SDS-PAGE of MoQ1 revealed a major band of 23 kDa as well as several other bands in the high molecular mass region (Fig. 2C). MALDI-TOF analysis identified the 23 kDa as Mn 2ϩ -containing superoxide dismutase (Mn-SOD). In order to separate Mn-SOD from potential protease candidates, an anion exchanger, Mini Q, with a higher resolution capacity than Mono Q was used in the purification procedure. Proteins bound to the Mini Q were eluted with a linear gradient of 0 -0.4 M NaCl and collected in two protein fractions MiQ1 and MiQ2 ( Fig. 2A, lower panel). The proteolytic activity was completely eluted at 310 -340 mM NaCl and found only in fraction MiQ2 (Fig. 2C, lane 6). As revealed by SDS-PAGE in one dimension, fraction MiQ1 contained the 23-kDa Mn-SOD band (Fig. 2C), whereas MiQ2 contained 7 polypeptides: 2 high molecular mass proteins (110 and 108 kDa), 4 proteins in the range of 55-66 kDa (66, 65, 55, and 54 kDa), and residual amounts of Mn-SOD (23 kDa) (Fig. 2C). No proteolytic fragments of the presequence of pF 1 ␤ were detected after incubation with isolated chromatographic fractions (Fig. 1B, lanes 4  and 5; Fig. 2B, lanes 3 and 6). In order to confirm complete degradation of the presequence, we used N 5.7 pF 1 ␤-(2-54) that was labeled with sulfosuccinimidyl biotin at lysine residues and detected using the avidin peroxidase system (27). There are 2 lysine residues in the presequence of F 1 ␤ at positions 32 and 44. Experiments with the sulfosuccinimidyl biotin-labeled F 1 ␤ presequence have not revealed any presequence fragments indicating complete degradation of the presequence (not shown).
Identification of the Protease, Two-dimensional Gels, and Mass Spectrometry-To identify the protease among the polypeptides in the MiQ2 fraction, two-dimensional gels followed by MALDI-TOF and ESI-MS/MS were applied. The twodimensional gel of the MiQ2 fraction revealed nine individual protein spots; two high molecular mass proteins (105 and 91 kDa), five proteins in the range of 55-70 kDa, a 35-kDa protein, and Mn-SOD (23 kDa) (Fig. 3). The molecular masses of proteins separated in one or two dimensions differ slightly from each other. This is due to the presence of urea in the one-dimensional SDS-PAGE system. MALDI-TOF followed by data base search allowed identification of the proteins in the 55-70-kDa range as the F 1 ␣ and F 1 ␤ subunits of ATP synthase (54 and 55 kDa), the flavoprotein from succinate dehydrogenase (66 kDa), dihydrolipoamide dehydrogenase (57 kDa), and mitochondrial Hsp70 (69 kDa) ( Table I). The F 1 ␣ and F 1 ␤ subunits are a part of the extramembranous catalytic core of the membrane-bound ATP synthase (52). Western blot analysis revealed small contaminations of the F 1 ␣ and F 1 ␤ subunits in the matrix preparations, Ͻ5%. Both the flavoprotein from succinate dehydrogenase and dihydrolipoamide dehydrogenase have been classified as soluble matrix proteins in studies of A. thaliana mitochondrial proteome (53). Hsp70 is one of the most conserved proteins and is described as a general molecular chaperone (54). In mitochondria, Hsp70 is localized to the matrix and drives mitochondrial import in an ATP-dependent manner (55). Mn-SOD has been identified in the matrix of mitochondria (53) and functions to neutralize produced oxygen radicals into H 2 O 2 (56). All the proteins identified by MALDI-TOF analysis are conserved, abundant mitochondrial proteins with a well characterized function. The three-dimensional structures have been solved for the MALDI-TOF identified proteins or their protein homologues, and no proteolytic sites or activities have been reported; therefore, it allows the exclusion of them as protease candidates (56 -61). No MALDI-TOF spectra could be obtained from spot number 9, neither could this spot be stained with silver.
In order to identify the 105-and 91-kDa proteins, we performed ESI-MS/MS analysis to obtain sequence information in combination with molecular masses. ESI-MS/MS identified the 91-kDa protein as a putative Hsp90 from A. thaliana (Table I). Subcellular prediction programs identified the putative Hsp90 as a mitochondrial protein. Hsp90 has also been found in human mitochondria (62). No proteolytic activity has been reported for Hsp90. The 18-amino acid residues sequence of the peptide derived from the 105-kDa protein ( Fig. 4 and Table I) matched four proteins in the NCBI data base, a zinc metalloprotease (insulinase family) (BAB02957), a putative zinc metalloprotease (AAG13049), and two putative hydrogenases (AAL67002 and NP_175386). The sequences of the putative zinc metalloprotease and the putative hydrogenases display 99% sequence identity and only differ in a stretch of 10 amino acids located in positions 189 -199 from the predicted cleavage site. These protein sequences are derived from the same gene (At1g49630) on chromosome 1 and therefore refer to the same protein reported three times in the NCBI data bases that will be referred to here as a putative zinc metalloprotease. The gene for the zinc metalloprotease (At3g19170) is located on chromosome 3. The putative zinc metalloprotease and the zinc metalloprotease display 89% sequence identity and 95% sequence similarity (Fig. 4), with most differences in the predicted targeting sequences. The targeting peptide of zinc metalloprotease is 30 amino acids shorter than the targeting peptide of the putative zinc metalloprotease. However, an EST clone of zinc metalloprotease from A. thaliana containing an additional 30 amino acid N-terminal region has been reported from the RIKEN Genomic Science Center and the Kazusa DNA Research Institute, Japan. The overexpressed A. thaliana zinc metalloprotease revealed proteolytic activity against the presequence peptide. 2 By using MitoProt, both proteases are predicted to contain a mitochondrial presequence with an arginine in position Ϫ2 upstream of the processing site, a common characteristic of mitochondrial presequences. However, using prediction programs that select preferable organellar localization, Predator and TargetP, the zinc metalloprotease with the 2 A. Ståhl and E. Glaser, unpublished data. additional 30-amino acid N-terminal region is predicted to be localized to mitochondria, whereas the putative metalloprotease was suggested to be a chloroplastic protein. The inverted zinc-binding region is located in the N-terminal portion of the protein and classifies the proteases to the pitrilysin family. We have named the experimentally identified metalloprotease that degrades targeting sequences as Presequence Protease, PreP.
Characterization of PreP, a Novel Mitochondrial Metalloprotease-We investigated if the purified protease was capable of degrading substrates other than the F 1 ␤ presequence. For this purpose we used synthetic, cleavage-specific peptides, P1 and P2, labeled with a fluorescent dye. Both peptides contained typical presequence elements like basic and hydrophobic residues but no acidic residues (Fig. 5A). In the native state both peptides are positively charged (Fig. 5, B and C, lane 1), but upon cleavage, fragments of different sizes and charges are generated that correspond to different cleavage sites. Incubation of the fraction containing PreP and the P1 peptide produced a negative charged fragment, indicating that the P1 peptide was cleaved after Pro, Leu, or Ser (Fig. 5B, lane 2). In contrast, the P2 peptide was not degraded by the PreP-containing fraction (Fig. 5C, lane 2). These results may reflect the requirement of a minimal peptide length or amino acid content to initiate and pursue degradation.
The effect of protease inhibitors on degradation of N 5.7 pF 1 ␤-(2-54) and synthetic peptides by PreP was evaluated (Fig. 6). Degradation of N 5.7 pF 1 ␤-(2-54) was inhibited by o-phenanthroline and carboxypeptidase inhibitor (Fig. 6A, lanes 5 and 6). The same protease inhibitors also abolished degradation of the P1 peptide (Fig. 6B, lanes 5 and 6). No effect was observed with phenylmethylsulfonyl fluoride, N-ethylmaleimide, or apyrase. In order to determine whether a free carboxyl group on presequences is required for their degradation, we used a presequence with a blocked carboxyl group, N 5.7 F 1 ␤-(2-54)-hsl. Despite the modified C terminus, N 5.7 F 1 ␤-(2-54)-hsl was degraded by the PreP-containing fraction (Fig. 7). The carboxypeptidase inhibitor is a 38-amino acid residue peptide isolated from potato that contains presequence characteristic amino acids, 5 positively charged and 11 hydrophobic amino acids (63). Therefore, the carboxypeptidase inhibitor peptide might function as a competitive substrate for the protease. DISCUSSION Mitochondrial targeting presequences are potentially harmful to the structure and function of mitochondria as they can disrupt biological membranes, dissipate membrane potential, and cause severe mitochondrial dysfunctions (19,21,22). Therefore, it is generally believed that after protein import and processing, the presequences are degraded inside mitochondria. However, virtually nothing is known about the degradation process or proteases involved in the degradation. Difficulties studying the fate of targeting peptides have come from the fact that small quantities of in vitro transcribed and translated precursors have been used in protein import studies. We have used chemical quantities of a precursor protein, and we were able for the first time to visualize degradation of a presequence inside mitochondria. We have shown that the presequence of the F 1 ␤ subunit of the ATP synthase was rapidly degraded by a matrix located protease(s) (27).
The current investigation aimed to identify and characterize mitochondrial protease(s) involved in presequence degradation. We have developed a simple chromatographic procedure for the isolation of the protease(s). The first step of the procedure included purification on an affinity column that contained arginines coupled to Sepharose (arginine-Sepharose). Because mitochondrial presequences are enriched in positively charged residues, arginines and lysines, the idea was to use a positively charged chromatographic gel matrix to bind potential protease candidates that would bind to presequences, in situ. Further purification was achieved by ion exchange chromatography, on Mono Q and high resolution Mini Q anion exchangers. The proteolytic activity of the isolated fraction was enriched by several hundredfold in comparison to the matrix. The twodimensional gel of the purified proteolytically active fraction from the Mini Q column revealed nine individual proteins, six of which were identified by MALDI-TOF as mitochondrial Hsp70, flavoprotein of succinate dehydrogenase, dihydrolipoamide dehydrogenase, the F 1 ␣ and F 1 ␤ subunits of the ATP synthase, and Mn-SOD. These proteins were excluded as protease candidates because they have well characterized function and structures with no indication of a proteolytic catalytic site. The 35-kDa protein could not be identified by any means. It was visualized with Coomassie Blue but not with silver staining, and no MALDI-TOF spectra could be obtained. These results may indicate a non-proteinaceous nature of this spot. Two residual spots were identified by ESI-MS/MS in order to obtain sequence information in combination with molecular masses. ESI-MS/MS identified the 91-kDa protein as a putative Hsp90 molecular chaperone, whereas the 105-kDa protein was identified as a metalloprotease in the A. thaliana data bases. The identified metalloprotease harbored an N-terminal mitochondrial targeting sequence, suggesting that it is localized to mitochondria. The metalloprotease contained an inverted zinc-binding motif (HILEHX 74 E) and belongs to the pitrilysin family. Members of the pitrilysin family can be divided into the pitrilysin subfamily, and the mitochondrial processing peptidase subfamily. The identified metalloprotease exhibits high overall homology with proteases of the pitrilysin subfamily. Members of the pitrilysin family are oligopeptidases of 100 kDa, such as insulinase from mammals and its homologue in bacteria, protease III (64). In the cell, these peptidases degrade small peptides of the similar size as mitochondrial presequences (65). Interestingly, insulinase was also found to degrade the cleaved leader peptide present in a peroxisomal protein (66).
Homologues of the zinc metalloprotease are present in human and in yeast. Subcellular prediction programs suggest that both homologues are localized to mitochondria. Alignments of the A. thaliana zinc metalloprotease and non-plant eucaryotic homologues revealed 30% sequence identity and a fully conserved inverted zinc-binding motif. The human homologue of 110 kDa (hMP1) was found to be sensitive to metal chelators and thiol reagents and was highly expressed in mitochondrial enriched tissues such as heart and skeletal muscles (67). The homologue identified in yeast is a hypothetical 989amino acid long protein (Ydr430cp). In yeast gene deletion studies, Ydr430cp was not found to be essential (68). No lethal phenotypes have been reported for mitochondrial proteases or their bacterial homologues (46,68,69). The absence of severe phenotypes for mitochondrial proteases suggests overlapping substrate specificities between the proteases, for example, shown for the PIM1 and the matrix-AAA proteases (40). The identification of the zinc metalloprotease homologues in various types of mitochondria might indicate that it is a general protease involved in the degradation of cleaved presequences peptides inside mitochondria.
The mitochondrial ATP-dependent proteases with different sub-organellar locations generate oligopeptides upon catalysis. Proteolysis by the membrane-bound AAA proteases generates small peptides with molecular masses of 2100 to 600 Da (70). In the matrix, the PIM1 protease degrades proteins to products containing 5-15 amino acid residues (71). In E. coli, ClpP hydrolyzes a number of proteins by endoproteolysis to generate short acid-soluble peptides (69). ClpP has been identified in the matrix of mitochondria; however, no native substrates or ClpP proteolytic activity have been demonstrated (44,45). Incubation of PreP with the presequence of F 1 ␤ revealed no proteolytic fragments. It was evidenced both upon immunodecoration of N 5.7 pF 1 ␤-(2-54) with antibodies directed against its C-terminal part or when the N 15 pF 1 ␤ was labeled with sulfosuccinimidyl biotin at lysines residues for detection with avidin peroxidase (27). It has been reported that cleavage of chloroplastic transit peptides in the stroma by stromal processing peptidase generates trimmed peptide fragments (72).
Further characterization of PreP was performed using synthetic cleavage-specific fluorescent peptides, P1 (11 amino acids) and P2 (7 amino acids). Like mitochondrial presequences, the peptides contained basic and hydrophobic residues and no negatively charged residues. Only the longer P1 peptide was degraded by PreP. This indicates that a minimal peptide length or specific amino acids might be required for efficient proteolysis by the PreP protease. The exact prerequisites for proteolysis still need to be determined. The proteolytic activity was investigated in the presence of inhibitors affecting different protease classes such as metallo, carboxy, serine, and cysteine proteases. The chelator o-phenanthroline completely inhibited the degradation of the F 1 ␤ presequence and the P1 peptide, demonstrating the need of metals for proteolysis. A partial inhibition of the activity was seen with a carboxypeptidase inhibitor; however, the presequence with a modified C-terminal carboxyl group was completely degraded by PreP indicating that there was no requirement of a free C terminus for degradation. The carboxypeptidase inhibitor is a peptide of 38 amino acids residues purified from potato and can thus function as a competitive substrate in presequence degradation. Classical protease inhibitors affecting serine and cysteine proteases did not abolish the activity. No ATP dependence was shown for PreP. Most proteases in the cell cleave substrates in an ATPindependent manner. ATP is not required for peptide bond hydrolysis per se but rather for unfolding or remodeling target substrates. The lack of ATP dependence for presequence degradation might be due to the fact that cleaved presequences are short, soluble peptides without structure in an aqueous environment (15). In conclusion, we characterize PreP as a metalloendopeptidase.
Can presequences be exported from mitochondria? A signal peptide was found to be exported from the endoplasmic reticulum membrane and interacted with calmodulin (73). The binding of signal peptides to calmodulin suggests that signal peptides contribute in regulation of cellular reactions (73). In yeast the ATP-binding cassette protein Mdl1 was shown to be involved in export of peptide fragments generated by proteolysis of the inner membrane AAA proteases (70). However, the majority (70%) of the released products from mitochondria were recovered as free methionine (70). Due to high energy costs and low export efficiency of peptides to the outside of mitochondria, presequences are more likely to be degraded inside mitochondria. Amino acids recovered from the degraded presequences can be used for mitochondrial protein synthesis or further catabolized via the urea cycle. Mitochondrial presequences are enriched in arginines that can be a precursor for synthesis of nitric oxide, polyamines, and other biologically active compounds. The arginine-converting enzymes like arginase and nitric-oxide synthase have been identified in mitochondria (74,75). Arginines that are released upon presequence degradation can be converted into nitric oxide and regulate transcription of nuclear encoded mitochondrial genes, thereby providing cross-talk between mitochondria and the nucleus.