Matrix Processing Peptidase of Mitochondria STRUCTURE-FUNCTION RELATIONSHIPS*

The mitochondrial processing peptidase (MPP) and the processing enhancing protein (PEP) cooperate in the proteolytic cleavage of matrix targeting sequences from nuclear-encoded mitoehondrial precursor pro- teins. We have determined the cDNA sequence of iVeu-rospora MPP after expression cloning. to

The mitochondrial processing peptidase (MPP) and the processing enhancing protein (PEP) cooperate in the proteolytic cleavage of matrix targeting sequences from nuclear-encoded mitoehondrial precursor proteins. We have determined the cDNA sequence of iVeurospora MPP after expression cloning. MPP appears to contain two domains of approximately equal size which are separated by a loop-like sequence. Considerable structural similarity exists to the recently sequenced yeast MPP as well as to Neurosporu and yeast PEP. Four cysteine residues are conserved in Neurosporu and yeast MPP. Inactivation of MPP can be achieved by using sulfbydryl reagents. MPP (but not PEP) depends on the presence of divalent metal ions for activity. Both MPP and PEP are synthesized as precursors containing matrix targeting signals which are processed during import into mitochondria by the mature forms of MPP and PEP. thus appears to be of considerable interest. On the one hand, the peptidase acts on hundreds or thousands of rather diverse presequences and cleavage sites; on the other hand, it makes a single and specific cut (Ou et al., 1989).
A comparison of the various presequences determined so far reveals very few common structural features (von Heijne, 1986;Roise et at., 1986;Vassarotti et al., 1987). All presequences have a relatively high content of positive charges and may have a tendency to form amphipathic a-helical structures when inserted into a membrane. In many presequences but not in all there is an abundance of hydroxylated amino acids (serine and threonine). Otherwise, they are rather diverse, both with regard to the sequences of the targeting peptides and to the sequences around the cleavage sites. A common theme of the cleavage sites, however, seems to be the presence of an arginine residue 2 residues upstream of the peptide bond to be hydrolyzed (Nicholson and Neupert, 1988;von Heijne, 1988;Hart1 et al., 1989).
The catalytic specificity of the matrix processing enzyme The activity of a matrix processing enzyme has been determined in mitochondria from different organisms and has been shown to be metal-dependent (Blihni et al., 1980;McAda and Douglas, 1982;Miura et al., 1982;Conboy et nl., 1982;Schmidt et al., 1984). The enzyme was first purified from Neurospora crassa (Hawlitschek et al., 1988). Two proteins are required for proteolytic activity, the mitochondrial processing peptidase (MPP)' which appears to be the catalytic component, and the processing enhancing protein (PEP). Neurospora MPP and PEP have apparent molecular masses of 57 kDa and 52 kDa, respectively. The enzyme was subsequently isolated from yeast (Yang et al., 1988). It turned out that MPP is the product of the gene MIF2 or MAS2 (Pollock et al., 1988;Jensen and Yaffe, 1988) and PEP the product of the gene MASl or MIFl (Witte et al, 19881.' MPP and PEP were found to be structurally related, with a sequence identity of 20% (Pollock et al., 1988;Yang et al., 1988). Moreover, core proteins 1 and 2 (also called subunit I and II), the products of the genes CORl (Tzagoloff et al., 1986) and COR2 (Oudshoorn et aL, 1987) of ubiquinol cytochrome c reductase, are members of the protein family which includes both MPP and PEP. In Neurosporcs, PEP and core1 turned out to be structurally identical to each other (Schulte et aL., 1989).
In an attempt to obtain further insight. into the role of MPP, we have cloned the cDNA from N. crassa and have compared it with the yeast MPP sequence. Several highly conserved regions are found which may have a particular role in the catalytic activity in MPP and which are not present in other members of the MPP/PEP/core family. Most interestingly, 4 cysteine residues were found in identical positions in the two MPPs. Experiments with sulfiydryl reagents show that reactive cysteines indeed have an important function in MPP but not in PEP. Moreover, MPP is the Mn'+ ionbinding component responsible for the metal ion requirement of the processing activity. Finally, the data suggest that MPP is comprised of two domains of roughly equal size which are divided by a loop-like structure with an unusual amino acid composition.  (Kleene ef al., 1987) with the cDNA synthesis kit including EcoRI-adaptors of Pharmacia LKB Biotechnology Inc. following the standard protocol. Three samples were combined, and eDNA was fractionated according to length by electrophoresis in an 0.8% agarose gel. The gel was divided and five fractions of cDNA corresponding to (a) 300-900 bp, (b) 900-1500 bp, (c) 1500-2000 bp, (d) 2000-3000 bp, and (e) >3000 bp were obtained following electroelution of the cDNAs. The isolated fragments were ligated with 2 pg (a and b) or 1 us (c-e) of Xztll-EcoRI arms (Pharmacia LKB Biotechnology I&j, respective&. Immediately after packaging (Gigapack Gold, Stratagene) the libraries were amplified-in Es&&hio coli strain Y1088. The number of different nhages before amnlification were (a) 1.5.106, (b) 5. 105, (c) 9. 104, (2) 2.8. 106, (e) g-10*. The libraries had a titer of l-5 pfu/pl and were stored in SM (0.1 M NaCl, 8 rnM MgSOr, 50 rnM Tris/C!l pH 7.5, 1% gelatin) containing 1% CHC& and 0.02% NaNa at 4 "C or at -70 "C with additional 7% Me&O.

Screening
Procedure and Sequencing Strategy-For immunoscreening 5' lo4 pfu in Y1090 were grown in top agarose (1% Bactoagar, 0.625% yeast-extract, 0.1 M NaCl, 10 mM MgS04, 0.7% agarose) on agar plates with a diameter of 140 mm. In total 2.5. lo5 pfu of library C were analyzed. Screening was performed as described previously by Young and Davis (1983). MPP antibody was diluted 1:lOOO. For detection of positive clones, we used anti-rabbit IgG antibody coupled to cu-peroxidase (Sigma; Tsung et al., 1983). Positive clones were picked and rescreened at a density of 200 plaques/94-mm plate.
In a second rescreen we tested them for homogeneity. X-DNA was isolated in small scale preparations and cleaved with EcoRI. The cDNA inserts were subcloned into the EcoRI site of nGEM3.
Screening for full-length clones was done by plaque hybridization using the 32P-labeled insert of the antibody-positive clone pNiw2 as a probe. 5.10' pfu were grown on five agar plates (94 mm) in Y1088 at 37 "C overnight. The plaque DNA was fixed in situ on Nylon membranes (Amersham Corp.;Benton and Davis, 1977). Hybridization and washing was carried out following the standard protocol. We detected several positive clones, and two of them (pMk1 and pMk2) were finally purified and subcloned into pGEM3. A genomic library cloned in pBR322 was screened by colony hybridization using the same probe.
Supercoil sequencing (Chen and Seeburg, 1985) was performed by the dideoxv-chain termination method (Saneer et al., 1977). Short-I ened clones were prepared by exonuclease III/nuclease Sl treatment (Henikoff, 1984), by digestion with restriction enzymes (Boehringer Mannheim) that cut both the polylinker and the cDNA (HindII, HindIII, SalI, SmoI) followed by religation, or by s&cloning of fragments of the EDNA into pUC19. Parts of the cDNA and genomic clones were sequenced by using MPP-specific synthetic oligonucleotide primers.
In Vitro Transcription, Translation, and Processing-Full-length MPP cDNA (pMk2) as well as other precursor protein cDNAs were cloned in pGEM3. ppreMPP160 was constructed by cleaving pMk2 with EcoRI and HindIII, blunt-end formation with Klenow enzyme, and cloning the fragment of the 5'-end into a pGEM3-vector, which had been digested with Hind111 and XbaI and blunt-ended with Klenow enzyme. By this way the TAG codon in the XbaI-site was placed into the MPP reading frame. The plasmids were transcribed with SpG-RNA polymerase (Promega). Precursor proteins were synthesized in rabbit reticulocyte lysate containing radiolabeled amino acids (Peiham and Jackson, 1976).
The processing peptidase assay was carried in the presence of 1% Triton X-100, 30 mM Tris/Cl, pH 8.2, 1 mM MnCh and 1 mM phenylmethanesulfonyl fluoride (PMSF) in a final volume of 15 ~1. Partially purified MPP (25 ng) and purified PEP (25 ng) were used. The reaction was started with 1.5 pl of lysate containing the precursor of the 6 subunit of F,-ATPase (pF,@) synthesized from the cDNA in plasmid pGEM as substrate. After 30 min at 25 "C, the processing reaction was stopped by addition of 15 ~1 of Laemmli buffer (2-fold concentrated; Laemmli, 1970). Processing products were analyzed by SDS-polyacrylamide gel electrophoresis, fluorography, and laser densitometry.

Purification of Processing
Peptidase-Purification of PEP was carried out as described by Schulte et al. (1989). For purification of MPP a total protein extract of N. crassa hyphae (100 g) was prepared according to Hawlitschek et al. (1988). The initial chromatographic steps, namely on DEAE-cellulose (Whatman), metal chelate affinity Sepharose 6B (Pharmacia LKB Biotechnology Inc.), and PEI-cellulose (Sigma) were performed as described before (Hawlitschek et al., 1988). The resulting fractions from the PEI-cellulose chromatography containing MPP were then applied to a hydroxyapatite column (Bio-Rad; 2.5 X 10 cm) and ~hromatographed with a linear O-200 mM sodium phosphate gradient. MPP eluted at a phosphate concentration of 160-180 mM. The pooled fractions were applied to a Mono Q column (Pharmacia HR 5/5). Proteins were eluted with a linear salt gradient from 0 to 300 mM and MPP eluted at a salt concentration of 110 mM NaCl. The vield of MPP was 1% (05 us of orotein). This preparation was free of PEP as judged by Western"blocting. All steps were carried out at 4 "C and monitored by SDS-gel electrophoresis and Western blotting (Burnette, 1981).
Immunodecoration was carried out with antibody against MPP and visualized using anti rabbit IgG antibody coupled to alkaline phosphatase (Blake et al., 1984).

RESULTS
eDNA Cloning, ~eq~~~i~, and Predicted Amino Acid Sequence of ~e~rQ~~ra MPP-A cDNA library of iV. crassa containing cDNA inserts in the range of 1500-2000 bp cloned in Xgtll (Young and Davis, 1983) was screened by using an antibody probe against Neurospora MPP. We examined 2.5. 10' phage plaques and obtained five positive clones. The clones were isolated and the cDNAs were subcloned into the plasmid vector pGEM3. Sequence analysis of the 5'-and 3'ends showed that all clones were identical and were termed pMw2. The cDNA-insert of pMw2 was 1902 bp in length. The library was then rescreened with the 3ZP-labeled insert of pMw2. Two further clones (pMk1, pMk2) were analyzed and subcloned into pGEM3. pMk1 started 198 bp downstream of the 5'-end of pMw2 and included a fragment of a poly(A) tail containing 3 adenine residues. pMk2, compared with pMw2, contained an additional 126 residues at the 5'-end, including the proposed start codon, and lacked oniy a few residues at the 3'-end. pMk2 most likely represented a fulllength clone. A genomic DNA-library in pBR322 was also screened and genomic clones were isolated.
The complete pMw2 insert was sequenced while the other clones were sequenced only in part. An outline of the sequencing strategy is given in Fig. 1. The cDNA consists of 2037 bp (Fig. 2) 3'-untranslated region consists of 265 residues and includes a putative polyadenylation signal (AATATA) 24 bp upstream of the poly(A) tail. The exact start of the tail was determined by sequencing of a genomic clone.
Comparison with Related Protein Sequences-The amino acid sequence of Neurospora MPP was compared with the sequences of the protein family including, so far, yeast MPP, PEP, and core proteins 1 and 2 of the cytochrome c reductase, and Neurospora PEP (Schulte et al., 1989). Compared with yeast MPP there is a sequence identity of 43.5%. An alignment of the two MPP sequences is presented in Fig. 3. We found homologies in all parts of the protein with the exception of a serine-rich region in the center of the Neurospora sequence, which has no counterpart in the yeast protein. Without taking into account this region the identity is 48.3%. To verify this unusual sequence, three independent cDNA clones and one genomic clone were isolated, and the sequence in this region was confirmed. Therefore, it is unlikely that there was erroneous cDNA synthesis. In fact, the protein obtained by in uitro transcription, translation, and processing had the same mobility upon SDS-polyacrylamide gel electrophoresis as the purified MPP (not shown). Computer analysis predicted a high level of flexibility for this stretch, and both Neurospora and yeast MPP show a high frequency of proline residues in this area. This sequence may have originated from an intron which has lost one of its splice sites during evolution. Possible 5'-ends of such a putative intron are at nt 793 (GTACTT) and at nt 828 (GTCTCT). The consensus sequence for Neurospora is GTAXGT (Bowman et al., 1988). The 3'-end would be at nt 1014 (TAG) (consensus, PyAG) with CTCAC (consensus, CTPuAC) corresponding to the branch site 15 bp upstream (Fig. 2).
Several striking similarities between the Neurospora and yeast enzyme exist: first, there is a stretch of 33 identical residues starting at amino acid position 372. Most remarkable are 4 glycine residues which are surrounded by uncharged amino acids. This rather hydrophobic area has a predicted high flexibility in the otherwise hydrophilic protein and may therefore be located in the interior of the molecule. It may be relevant that this motif is not present in PEP and in core protein 1 of cytochrome c reductase. This region therefore may be important for the activity of the protein. Second, a highly conserved region extends from amino acid 467 to 506. This region is hydrophilic and is a common element of the Boxed sequences indicate similarities to intron boundaries (shadowed) and to the branch site consensus (Ilght box). We found no striking similarities to other proteins when we made an alignment against data banks. In particular, we compared the MPP sequences of Neurospora and yeast with several known proteases, especially cysteine and metalloproteases (Kamphuis et al., 1985;Jongeneel et al., 1989). The only similarity we found was to a well-conserved sequence in cysteine proteases. The motif His-Ala-Val-Thr-Ala-Ile-Gly-Tyr in stem bromelain (Go10 et al., 1980) is similar to the motif His-Ala-Leu-Thr-Thr-Asp-His-Gly-Tyr (amino acids 452-460) in Neurospora MPP. This motif, however, is not present in yeast MPP. Therefore it remains doubtful as to whether it is significant.
Both MPP and PEP Precursors Are Processed by Their Combined Mature Forms-The two components of the processing enzyme, MPP and PEP, are encoded in the nucleus. Recently, it has been reported that PEP is synthesized as a precursor and is processed to the mature form during import into mitochondria both in Neurospora and yeast (Hawlitschek et al., 1988;Witte et al., 1988). For yeast MPP it is so far unknown whether there is a cleavable signal sequence (Pollock et al., 1988;Jensen and Yaffe, 1988). The amino terminus of Neurospora MPP contains a number of arginines as well as serine and threonine residues but no negative charges. These features are generally observed in mitochondrial targeting sequences.
We subcloned both MPP and PEP cDNA into pGEM3 vector and performed in vitro transcription with SpG-RNA polymerase followed by in vitro translation in rabbit reticulocyte lysate in the presence of [""S]methionine.
The labeled MPP and PEP precursors obtained in this way could be imported into isolated Neurospora mitochondria.
In both cases a protein with reduced molecular weight was formed and was found to be protected against protease added to the reisolated mitochondria (Fig. 4A). In the case of PEP, part of the unprocessed precursor was also protease-protected.
In the presence of EDTA and I,IO-phenanthroline, which chelate metal ions and in this way inhibit the activity of the processing enzyme in the mitochondria (Schmidt et al., 1984), proteolytic cleavage was blocked completely in the case of PEP and was reduced to about 50% in the case of MPP.
When the lysates containing labeled MPP and PEP precursors were incubated with purified MPP and PEP, defined molecular weight shifts were observed (Fig. 4B). Again the cleavage of the MPP precursor was more efficient than the cleavage of the PEP precursor. We conclude that MPP and PEP follow the general import pathway for nuclear-encoded FIo. 4. A, Import of MPP and PEP into Neurospora mitochondria. Import was performed as described under "Experimental Procedures." The mitochondria, suspended in bovine serum albumin buffer including cold Iysate and EDTA, l,lO-phenanthroline (o-Phe) were preincubated for 5 min at 25 "C. Then either ""S-labeled precursor of MPP or PEP, synthesized in reticulocyte Iysate, was added and incubated for 30 min at 25 "C. B, In vitro processing of precursors of MPP, PEP, and F,D. Reticulocyte Iysates containing in uitro synthesized precursors were incubated with approximately 50 ng of MPP and PEP in 30 mM Tris/Cl pH 8.2, 1% Triton X-100, 1 mM MnCI, and 1 mM PMSF for 30 min at 25 "C. Samples were analyzed by SDS-gel electrophoresis and fluorography. mitochondrial proteins; they are synthesized as precursors in the cytosol and then processed during import into mitochondria by cooperation of their own mature forms. The start of the mature MPP was determined by radiosequencing of the protein that had been (i) imported into and processed in mitochondria and (ii) processed by using purified processing peptidase. Since the full-length protein was labile during the sequencing procedure and no clear result was obtained, we used a truncated form of MPP (preMPP160). This contained the first 160 amino acids of the MPP precursor. preMPP160 was synthesized by in vitro transcription/ translation in the presence of various radiolabeled amino acids.
Radiosequencing data on the processed form got from import into mitochondria are shown in Fig. 5A. After labeling with [""Slmethionine, we observed a peak at position 11, which corresponds to a lysine; since the protein was coupled to the solid support via the t-amino groups of lysine residues, every lysine gives a peak. Labeling with ["Hlvaline resulted in two peaks at position 8 (Val-8) and position 11 (Lys-11). When using ["Hlglutamic acid we observed peaks at position 11 and 15 (Glu-15). The putative signal for the predicted Glu-I2 is likely hidden in the tail of the strong signal at position 11. When preMPP160 radiolabeled with ["Hlvaline and processed by the purified enzyme was analyzed, peaks at positions 8, 11, 26, and 28 were observed (Fig. 5B). They apparently correspond to  We conclude that the processing site is after amino acid position 35 of the precursor of MPP: Asn-Asn-Ala-Arg-Thr-V-Leu-Ala-Thr-Arg.
This cleavage site is in agreement with the consensus Arg-X-V-Y (Nicholson and Neupert, 1988). It is, in fact, also very similar to the processing site in PEP which is Arg-Arg-Gly-V-Leu-Ala-Thr (Hawlitschek et al., 1988). It is further concluded that MPP/PEP are efficient in correctly processing MPP to its mature size.
The calculated molecular mass of mature MPP is then 59,058 daltons; this is somewhat higher than the apparent mass of 57 kDa determined by SDS-polyacrylamide gel electrophoresis of the purified enzyme (Hawlitschek et al., 1988 specifically affects only one of the two components which are responsible for processing activity, purified PEP and MPP were separately treated with 10 mM NEM. The processing activity was strongly reduced if MPP was treated with NEM whereas the pretreatment of PEP with NEM had no effect on the catalytic activity (lanes 5 and 8). The hydrophilic sulfbydryl reagents iodoacetic acid and iodoacetamide also inhibited the catalytic activity of MPP, but the extent of inhibition was much less than that observed by the hydrophobic reagent NEM. 10 mM iodoacetate and iodoacetamide inhibited MPP by 67 and 35%, respectively, as compared with 95% inhibition with 10 mM NEM. MPP was completely inactivated by p-chloromercuric benzoate at a concentration of 0.01 mM (not shown). The NEM sensitivity would indicate that the thiol groups of one or more of the cysteine residues in MPP are necessary for catalytic activity. Specific reagents that inhibit enzymes of the class of cysteine proteases such as chicken cystatin (Barrett et al., 1986) and epoxysuccinyl-leucyl agmatine (E 64) from Aspergillus japonicus (Rich, 1986), however, did not affect processing activity.
The Activity of MPP Is Metal-dependent-The processing activity in mitochondria and of the purified enzyme has been reported to be dependent on divalent metals such as Mn')+ (Bahni et al., 1980;Hawlitschek et al., 1988). Which of the two components of the processing peptidase, MPP or PEP, is responsible for this metal dependence?
To investigate this we incubated each protein in the absence or presence of 1 mM MnC12. Then immunoprecipitation with antibodies against either MPP or PEP was carried out in the absence or presence of 1 mM MnClp. Processing activity was then determined by addition of desalted PEP to immunoprecipitated MPP and addition of desalted MPP to immunoprecipitated PEP. Without further addition of Mn'+ to the assay system, processing was only observed if MPP had been pretreated with Mn'+, but not if PEP had been pretreated with Mn')+ (Fig. 7, lanes 2 and 4). If MnY+ was absent during pretreatment of MPP and PEP, processing activity was not observed (lanes  Preparations of purified MPP and PEP were desalted with a Sephadex G-25 column. Then 0.2 pg of each protein was incubated separately in the presence or absence of 1 mM MnCIZ for 7 min at 25 "C. To remove unbound manganese ions, MPP and PEP were immunoprecipitated with specific antibodies. Processing activity of the resulting pellet was tested by adding either PEP or MPP and F,@ precursor as a substrate. p, precursor; m, mature F,fl. 1 and 3). When Mn"' was included in the enzyme assay all samples showed processing activity (lanes 5-8). These results indicate that binding of metal ions, which is necessary for processing activity, occurs to MPP and not to PEP.

DISCUSSION
The amino acid sequence of Neurospora mitochondrial processing peptidase, the catalytic component of the matrix processing enzyme, shows a number of striking similarities to that of the yeast counterpart (Pollock et al., 1988;Jensen and Yaffe, 1988). The sequence also shows similarities to that of the other component of the Neurospora matrix processing enzyme, the processing enhancing protein (Hawlitschek et al., 1988) which in Neurosporu is identical to core protein 1 of the respiratory chain complex cytochrome c reductase (Schulte et al., 1989). Thus, Neurospora MPP appears to be a member of the MPP/PEP/core family. Unlike yeast MPP, the Neurospora MPP has an unusual serine-rich stretch in the center which is not present in any other member of the MPP/PEP/core family. In yeast MPP, however, an exceptionally high number of proline residues are located in this area. We propose that the MPP molecule consists of two domains which are separated by a spacer or a hinge formed by this serine-rich stretch or several proline turns. When one postulates that MPP should have two functions, namely (i) interaction with PEP and (ii) cleavage of presequences, these may be located in the different domains. This would resemble the situation with Neurospora PEP/ core1 where the amino-terminal half is more similar to yeast PEP while the carboxyl-terminal half is more similar to yeast core1 (Schulte et al., 1989).
The observation of 4 conserved cysteine residues in Neurospora and yeast MPP led us to ask whether these are essential for processing activity. Inhibition experiments with sulfhydryl reagents yielded two interesting results. First, inhibition of processing activity occurs with all sulfhydryl reagents employed, the hydrophobic N-ethylmaleimide, the hydrophilic iodoacetamide and iodoacetate, and p-chloromercurie benzoate. Second, selective treatment of MPP leads to loss of processing activity, but treatment of PEP does not inactivate processing activity.
Specific inhibitors of cysteine proteases did not inhibit the processing activity. Thus, the NEM-sensitive cysteine residues appear not to take part directly in the proteolytic step, and the enzyme therefore appears not to be a member of the class of cysteine proteases. Sequence comparisons of MPP and cysteine proteases showed no regions of similarity.
The conserved cysteine residues may have a role in determining the conformation of MPP. The matrix processing peptidase from several sources has been found to depend on divalent metal ions (Bohni et al., 1980;McAda and Douglas, 1982;Miura et al., 1982;Conboy et al., 1982;Schmidt et al., 1984). Mn2+ has to be included in processing assays to obtain full activity. Here we describe that it is MPP that requires manganese ions for processing of precursor proteins, and, on the other hand, that the stimulation of the catalytic activity by PEP is independent of metal ions. Is MPP a metalloprotease as would be indicated by these experiments?
In general metalloproteases are not NEM-sensitive, and furthermore no significant homologies to this class of proteases have been observed. Thus, it seems possible that in the matrix peptidase metal atoms are necessary for structure and are not directly involved in the catalytic step.
In summary, the processing enzyme does not appear to belong to any of the known and characterized classes of proteases: not to cysteine proteases and metalloproteases, not to serine proteases because of its insensitivity to PMSF, and not to aspartyl proteases, because pepstatin does not inhibit processing (not shown). The catalytic mechanism therefore remains enigmatic.
The amino terminus of the MPP sequence deduced from the cDNA sequence contains a typical matrix targeting signal which is 35 amino acid residues long and contains an abundance of positively charged residues. When in uitro synthesized MPP and PEP were imported into mitochondria, processing to the mature-sized species occurred. Processing during import was reduced in the presence of chelating agents, which are known to inhibit matrix protease (Bohni et al., 1980;Schmidt et al., 1984). Mature MPP and PEP could also be generated by in vitro processing with the purified peptidase. Thus, the precursors of MPP and PEP are processed by their (combined) own mature counterparts. Obviously, the continuous presence of functional MPP/PEP is a requirement for MPP/PEP biogenesis. This emphasizes a general principle of mitochondrial biogenesis, namely that formation of new mitochondria depends on pre-existing mitochondria.