Cargo sequences are important for Som1p-dependent signal peptide cleavage in yeast mitochondria.

The inner membrane protease (IMP) has two catalytic subunits, Imp1p and Imp2p, that exhibit nonoverlapping substrate specificity in mitochondria of the yeast Saccharomyces cerevisiae. The IMP also has at least one noncatalytic subunit, Som1p, which is required to cleave signal peptides from a subset of Imp1p substrates. To understand how Som1p mediates Imp1p substrate specificity, we addressed the possibility that Som1p functions as a molecular chaperone, which binds to specific substrates and directs them to the catalytic site. Our results show that cargo sequences attached to the signal peptide are important for Som1p-dependent presequence cleavage; however, no specific cargo sequence is required. Indeed, we show that a substrate normally destined for Imp2p is cleaved in a Som1p-dependent manner when the substrate is directed to Imp1p. These results argue against the notion that Som1p is a molecular chaperone. Instead, we propose that the cargo of some Imp1p substrates can assume a conformation incompatible with presequence cleavage. Som1p could thus act through Imp1p to improve cleavage efficiency early during substrate maturation.

The inner membrane protease (IMP) has two catalytic subunits, Imp1p and Imp2p, that exhibit nonoverlapping substrate specificity in mitochondria of the yeast Saccharomyces cerevisiae. The IMP also has at least one noncatalytic subunit, Som1p, which is required to cleave signal peptides from a subset of Imp1p substrates. To understand how Som1p mediates Imp1p substrate specificity, we addressed the possibility that Som1p functions as a molecular chaperone, which binds to specific substrates and directs them to the catalytic site. Our results show that cargo sequences attached to the signal peptide are important for Som1p-dependent presequence cleavage; however, no specific cargo sequence is required. Indeed, we show that a substrate normally destined for Imp2p is cleaved in a Som1p-dependent manner when the substrate is directed to Imp1p. These results argue against the notion that Som1p is a molecular chaperone. Instead, we propose that the cargo of some Imp1p substrates can assume a conformation incompatible with presequence cleavage. Som1p could thus act through Imp1p to improve cleavage efficiency early during substrate maturation.
Mitochondria consist of two hydrophobic compartments, the outer and inner membranes, and two hydrophilic compartments, the matrix and intermembrane space, which hold proteins that perform important cellular functions including oxidative phosphorylation, tri-carboxylic acid cycle, and fatty acid oxidation (1). Among the hundreds of proteins that reside within the mitochondrial compartments, only 13 (in human) or 8 (in yeast) are encoded by the mitochondrial genome. All other proteins are encoded within the nucleus and translated in the cytoplasm. To achieve mitochondrial targeting, many proteins have N-terminal signal peptides that must be cleaved following membrane translocation. The mitochondrial processing peptidase (MPP) 1 cleaves the signal peptides from most of these proteins (2). A smaller set of mitochondrial proteins has a signal peptide required for sorting to the intermembrane space and inner membrane following mitochondrial targeting and translocation. This signal peptide is cleaved by the inner membrane peptidase (IMP).
The third IMP subunit, Som1p, binds to Imp1p and is required for signal peptide cleavage from two Imp1p substrates, pre-cyt b 5 -red and pre-Cox2p (5,13). Other known Imp1p substrates, p-cyt b 2 and pre-Gut2p, can be cleaved in the absence of Som1p (5,9). This behavior indicates that noncatalytic subunit Som1p exerts a novel level of IMP substrate specificity, one that differentiates between Imp1p substrates. Here, we consider whether Som1p is a molecular chaperone that binds to specific substrates and delivers them to Imp1p.
Disruption of SOM1 Gene-Oligonucleotides containing sequences flanking the SOM1 and the LEU2 genes were used to amplify the LEU2 gene using PCR. PCR product was introduced into strain BY4741 with selection for Leu-positive colonies. Recombination of the LEU2 gene into the SOM1 chromosomal locus was then confirmed by diagnostic PCR. The disruption removed the entire SOM1 open reading frame from the genome of strain BY4741.
Construction of HA-tagged Proteins-A DNA segment encoding three consecutive HA epitopes (15) followed by a stop codon (TAA) was inserted between the EcoRI and KpnI restriction sites of plasmids pHF454 (2, TRP1, ADH1 promotor) and pHF455 (2, URA3, ADH1 promotor) (16). Genes encoding mutant and fusion proteins without translational stop codons were synthesized using two-step PCR (16). The genes were then inserted into the above mentioned plasmids between the ADH1 promotor (17) and the DNA segment encoding the triple HA epitopes. Sequences of oligonucleotides used in this study are available on request.
Pulse Labeling and Immunoprecipitation-Methods similar to those described previously were used (14). Briefly, cells were grown at 30°C to log phase (A 600 ϭ 1.5) in appropriate media. Cells were then shifted to media lacking methionine and cysteine for 1 h and pulse-labeled for 10 min with 60 Ci/ml [ 35 S]Met (PerkinElmer Life Sciences). Cells were harvested and broken by glass beads. Total proteins were precipitated by 10% trichloroacetic acid, resuspended in 20 l of SDS-PAGE sample buffer, and placed in a boiling water bath for 5 min. This mixture was diluted with 1 ml of phosphate-buffered saline/Triton X-100 (0.1%), and cell debris was removed by centrifugation. The protein solution was incubated overnight with 1.6 g/ml anti-HA antibodies at 4°C. This incubation was followed by incubation (2 h) with 20 l of protein G-agarose beads (Roche Applied Science). Protein G beads were washed twice with cold phosphate-buffered saline/Triton X-100 (0.1%) and then twice with distilled water. 20 l of SDS-PAGE sample buffer was added to the protein-bead slurry. The mixture was placed in a boiling water bath (5 min) and loaded onto an SDS-gel.

Cargo Sequences Influence Som1p-dependent Presequence
Cleavage-Som1p is needed for Imp1p cleavage of pre-cyt b 5red but not p-cyt b 2 (5). Based on this difference, we reasoned that pre-cyt b 5 -red has a sequence, which is not present in p-cyt b 2 , that is responsible for Som1p-dependent cleavage. To determine whether the putative sequence is located within the presequence or cargo of pre-cyt b 5 -red, we constructed a chimera that contained the cargo of pre-cyt b 5 -red fused to the presequence of p-cyt b 2 . The p-cyt b 2 presequence is actually composed of two signal peptides fused together in tandem (18). MPP cleaves a mitochondrial targeting signal peptide to generate i-cyt b 2 , and then Imp1p cleaves the second signal peptide to generate mature cyt b 2 . The p-cyt b 2 /cyt b 5 -red chimera, tagged with three HA epitopes at its C terminus, was expressed in strains BY4741 (wild type) and HFY501 (⌬som1). Yeast strain genotypes can be found under "Experimental Procedures." Cells were then subjected to a 10-min pulse with radiolabeled methionine, and proteins were precipitated from cell extracts with anti-HA antibodies (see under "Experimental Procedures"). As shown in Fig. 1, cleavage of i-cyt b 2 signal peptide from i-cyt b 2 /cyt b 5 -red was Som1p-dependent. Because this signal peptide can be cleaved from the cyt b 2 cargo independently of Som1p (5), the data support the notion that amino acid sequences located within the cargo of pre-cyt b 5 -red are important for Som1p-dependent cleavage.
Som1p Facilitates Imp1p Substrate Recognition without Requirement for a Specific Substrate Sequence-We next sought to identify specific cargo sequences involved in presequence cleavage from the p-cyt b 2 /cyt b 5 -red chimera. To this end, we introduced distinct internal and C-terminal deletions into the cyt b 5 -red cargo ( Fig. 2A) and expressed the constructs in strains BY4741 (wild type) and HFY501 (⌬som1). Cells were then examined by pulse labeling as described above. To our surprise, the mutated forms of i-cyt b 2 /cyt b 5 -red were cleaved in a Som1p-dependent manner (Fig. 2B). This result suggests that no specific cargo sequence is required for Som1p-dependent cleavage. Cleavage efficiency, however, was reduced when a region immediately flanking the Imp1p cleavage site was removed from the chimera (Fig. 2B, lanes 3 and 4). To examine this region in more detail, we designed a new set of constructs containing smaller internal deletions (10 or 11 amino acids) (Fig. 3A). As shown by pulse labeling, Som1p clearly was needed to cleave signal peptides from this new set of mutant chimeras (Fig. 3B). These results thus suggest that no specific cargo sequence is required for Som1p-dependent substrate cleavage, although it is plausible that deletion of cargo residues located near the cleavage site affect the conformation of the cleavage site and thus partially inhibit presequence cleavage in some protein contexts.
Next, we examined the role of cargo sequences using p-cyt b 5 -red, a natural Imp1p substrate. This protein is identical to the above-mentioned chimera with respect to the cargo sequence but differs, of course, in presequence composition. A series of internal and C-terminal deletions was constructed in the p-cyt b 5 -red cargo (Fig. 4A), and Imp1p cleavage was examined in cells that have or lack Som1p. Although most of the mutations did not prevent Som1p-dependent cleavage (Fig.  4B), residues located close to the cleavage site were important (Fig. 4B, lanes 3 and 4). These residues were nonessential for presequence cleavage in the chimeric protein context (Fig. 2B), suggesting that cargo residues located near the cleavage site can affect presequence cleavage in some protein contexts. Taken together, our data suggest that no specific cargo sequence is required for Som1p-dependent presequence cleavage.
p-Cyt b 5 -red Can Be Directed to Both the Mitochondria and ER- Fig. 4B depicts protein bands (*) with molecular weights that are greater than that of p-cyt b 5 -red. These bands disappear when yeast cells are treated with tunicamycin prior to the pulse (data not shown). Tunicamycin inhibits Asn-linked glycosylation in the ER (19), indicating that the bands with greater molecular mass represent p-cyt b 5 -red that has been directed to the ER and glycosylated. This conclusion is supported by our inspection of the p-cyt b 5 -red sequence, which identified four consensus Asn-linked glycosylation sites. We also observed differently sized glycosylated protein bands with different cargo deletions (Fig. 4B, compare lane 5 with lane 7). This result is consistent with deletion of one or more of the consensus glycosylation sites. The data presented in Fig. 4B thus demonstrate that p-cyt b 5 -red can be targeted both to mitochondria and the ER. Our data further suggest that the p-cyt b 5 -red presequence but not the p-cyt b 2 presequence exhibits this novel dual targeting, because glycosylated cyt b 5 -red cargo was not seen when the presequence of p-cyt b 2 was used to target the p-cyt b 2 /cyt b 5 -red chimera to mitochondria (Fig.  2B). The p-cyt b 5 -red presequence has a rather long hydrophobic stretch consisting of ϳ20 amino acid residues (8). This unusually long hydrophobic stretch is probably responsible for ER targeting, whereas distinct uncharacterized sequences in the p-cyt b 5 -red presequence are likely to be responsible for mitochondrial targeting. We do not know whether ER targeting occurs naturally in yeast cells that express wild type p-cyt b 5 -red, and it has not been reported previously. However, our overexpression of p-cyt b 5 -red using the ADH1 promotor to facilitate protein detection (see under "Experimental Procedures") may contribute to the ER targeting seen in this study.
A Mutant Form of p-Cyt c 1 Can Be Cleaved in a Som1p-dependent Manner-Previous work from our laboratory showed that Imp2p substrate p-cyt c 1 could be switched to Imp1p when -red constructs were expressed in strains BY4741 (WT, wild type) (oddnumbered lanes) and HFY501 (⌬som1) (even-numbered lanes), and strains were subjected to pulse labeling as described in Fig. 1. Proteins were resolved on a SDS-10% gel (lanes 1-10) or a SDS-12% gel (lanes 11-12).

FIG. 3. Deletion of cargo residues next to the Imp1p cleavage site in the p-cyt b 2 /cyt b 5 -red chimera does not prevent Som1p-dependent cleavage.
A, design of short internal deletions of cyt b 5 -red cargo is shown. B, HA-tagged p-cyt b 2 /cyt b 5 -red constructs were expressed in strains BY4741 (WT, wild type) (oddnumbered lanes) and HFY501 (⌬som1) (even-numbered lanes), and strains were subjected to pulse labeling as described in Fig. 1. Proteins were resolved by a SDS-10% gel.
Som1p Substrate Specificity 39398 the P 3 , P 1 , and PЈ 1 residues of p-cyt c 1 were changed to Ile, Asn, and Glu (INE), respectively (12). The presequence of p-cyt c 1 consists of a bipartite signal peptide (20). MPP cleavage of p-cyt c 1 generates i-cyt c 1 , and IMP cleavage of i-cyt c 1 generates cyt c 1 . To determine whether Som1p is needed for cleavage of i-cyt c 1 when it is switched to Imp1p, we prepared a DNA construct encoding a truncated form of p-cyt c 1 (p-cyt c 1 (INE)), lacking its C-terminal 87 residues, a region that includes the transmembrane segment of i-cyt c 1 (20). We employed this truncation to facilitate detection of the protein, because the truncated protein was expressed at higher levels than fulllength p-cyt c 1 in yeast mitochondria (data not shown).
The p-cyt c 1 (INE) was expressed in strains BY4741 (wild type), XCY101 (⌬imp1), HFY501 (⌬som1), and JN34 (⌬imp2)/ pXC3. Plasmid pXC3 carries a mutation that results in a S41A substitution in the Imp2p protein. This mutation inhibits Imp2p catalytic activity but allows Imp2p to bind to and stabilize Imp1p (11). As such, strain JN34/pXC3 has Imp1p activity but no Imp2p activity. As shown by pulse labeling (Fig. 5), i-cyt c 1 (INE) was processed by Imp1p but not by Imp2p (Fig. 5,  lanes 2 and 3). This result is consistent with our previous study, which showed that full-length i-cyt c 1 could be switched from Imp2p to Imp1p by incorporating INE residues at the cleavage site (12). Importantly, when Imp1p cleaved i-cyt c 1 (INE), it did so in a Som1p-dependent manner (Fig. 5, compare lane 2 with lane 4). As Imp2p does not need Som1p to cleave wild type i-cyt c 1 (5), it is highly unlikely that i-cyt c 1 (INE) would have a sequence that interacts with Som1p only when Imp1p cleaves the precursor artificially as a result of mutation at the presequence cleavage site. Thus, taken to- B, HA-tagged p-cyt b 5 -red constructs were expressed in strains BY4741 (WT, wild type) (odd-numbered lanes) and HFY501 (⌬som1) (even-numbered lanes), and strains were subjected to pulse labeling as described in Fig. 1. Proteins were resolved on a SDS-10% gel (lanes 1-10) or SDS-12% gel (lanes [11][12]. *, p-cyt b 5 -red molecules that were transported to the ER and glycosylated.
To further support our conclusion that cargo sequences are important for Imp1p substrate cleavage, we designed a fusion protein that contained the bipartite presequence of p-cyt c 1 (INE) fused to the cyt b 2 cargo. The signal peptide of i-cyt c 1 (INE) is cleaved in a Som1p-dependent manner (Fig. 5), whereas the signal peptide of i-cyt b 2 is cleaved in a Som1pindependent manner (5). We reasoned that, if cargo sequences dictate Som1p-dependence, the i-cyt c 1 (INE)/cyt b 2 chimera would be cleaved independently of Som1p. To test this idea, the chimera was expressed in strains BY4741 (wild type), JN34 (⌬imp2)/pXC3, XCY101 (⌬imp1), and HFY501 (⌬som1), and cells were subjected to pulse labeling. As shown in Fig. 6, the signal peptide of i-cyt c 1 (INE)/cyt b 2 was indeed cleaved in a Som1p-independent manner (lanes 2 and 4). These results support our argument that cargo sequences play a critical role in determining whether an Imp1p substrate needs Som1p to be cleaved.

DISCUSSION
Som1p is one of three known subunits, the others being Imp1p and Imp2p, of the yeast IMP. Imp1p and Imp2p are catalytic subunits with nonoverlapping substrate specificity (4), and both subunits are members of the type I signal peptidase family (6). Although the function of Som1p remains largely unknown, previous studies (5,9,13) have shown that Som1p is required to cleave two of the four known Imp1p substrates. Here, we have tested a model in which Som1p acts as a molecular chaperone that binds to the signal peptide or attached cargo of an Imp1p substrate and directs the substrate to Imp1p. We have tested this model by constructing and examining two chimeras. One chimera has the signal peptide of a Som1p-independent substrate fused to the cargo of a Som1pdependent substrate (Fig. 1). The second chimera has a configuration similar to the first except its signal peptide and cargo come from substrates that display the opposite requirements for Som1p (Fig. 6). When these chimeras are tested for Som1pdependent cleavage, results are consistent with the idea that cargo sequences play the critical role.
To identify cargo sequences that putatively bind to Som1p, we have examined a series of internal and C-terminal deletions in chimeric and natural Imp1p substrates (Figs. 2-4). To our surprise, the mutated proteins are cleaved in a Som1p-dependent manner, with one exception (Fig. 4, lanes 3 and 4). The exception occurs when a deletion removes cargo residues located next to the Imp1p cleavage site. This mutation inhibits Imp1p cleavage in the presence or absence of Som1p. However, deletion of cargo residues from a related protein does not prevent Som1p-dependent cleavage (Fig. 2, lanes 3 and 4). We therefore suggest that this deletion alters the conformation of the cleavage site, rendering it inaccessible to Imp1p. The possibility remains, however, that Som1p binds to sequences from disparate regions of the folded or partially folded protein cargo, and Som1p binding can still occur when the binding site is partially removed from the cargo. We do not favor this idea based on our analysis of a novel Imp1p substrate generated for this study. This novel substrate has been produced by changing the cleavage site residues of an Imp2p substrate so that the substrate can be cleaved by Imp1p. The substrate p-cyt c 1 is normally cleaved by Imp2p in a Som1p-independent manner, but when the substrate is switched to Imp1p, the substrate shows Som1p dependence (Fig. 5). We consider it unlikely that p-cyt c 1 would have a Som1p-binding site used only when it is artificially switched from Imp2p to Imp1p. For these reasons, we suggest that Som1p does not bind to Imp1p substrates and thus is not a molecular chaperone.
In a previous study, it has been shown that Som1p binds to Imp1p (5). This binding is required for stabilization of Som1p but not Imp1p in yeast cells. Because of this close association between Som1p and Imp1p, we suggest that Som1p binding could be needed to maintain Imp1p in a functionally efficient state. This model predicts that some Imp1p substrates (i.e. those substrates that require Som1p for signal peptide cleavage) can be processed only when Imp1p operates with high efficiency. Because our study clearly demonstrates that cargo sequences play the major role in dictating Som1p dependence, cargo sequences may interfere with presequence cleavage. Interference could occur when the cargo folds into a conformation that renders the cleavage site inaccessible following membrane translocation of the precursor polypeptide. In the presence of Som1p, Imp1p would, according to our interpretation, be able to recognize the precursor and cleave its presequence early during the folding process. It then follows that the conformations of different precursors would influence the Som1p requirement differently. This model thus suggests that, rather than binding substrates directly, Som1p acts through Imp1p to ensure cleavage of a structurally diverse set of protein substrates.