Alteration of a mitochondrial outer membrane signal anchor sequence that permits its insertion into the inner membrane. Contribution of hydrophobic residues.

Tom70p is targeted and inserted into the mitochondrial outer membrane in the Nin-Ccyto orientation, via an NH2-terminal signal anchor sequence. The signal anchor is comprised of two domains: an NH2-terminal hydrophilic region which is positively charged (amino acids 1-10) followed by the predicted transmembrane segment (amino acids 11-29). Substitution of the NH2-terminal domain with a matrix-targeting signal caused the signal anchor to adopt the reverse orientation in the outer membrane (Ncyto-Cin) or, if presented to mitoplasts, to arrest protein translocation at the inner membrane without insertion. Physically separating the transmembrane segment from the matrix-targeting signal by moving it downstream within the protein resulted in a failure to arrest in either membrane, and consequently the protein was imported to the matrix. However, if the mean hydrophobicity of the Tom70p transmembrane segment was increased in these constructs, the protein inserted into the inner membrane with an Nin-Cout orientation. Therefore we have determined conditions that allow the Tom70p transmembrane domain to insert in either membrane, pass through both membranes, or arrest without insertion in the inner membrane. These results identify the mean hydrophobicity of potential transmembrane domains within bitopic proteins as an important determinant for insertion into the mitochondrial inner membrane.

Nuclear-encoded precursor proteins destined for import into mitochondria are sorted to one of four compartments in the organelle: outer or inner membrane, intermembrane space, or matrix. Since import of most of these proteins is mediated by a common protein translocation machinery (for reviews see Refs. 1 and 2), specificity for sorting must reside within topogenic domains present in the precursor protein. To date, four such domains have been characterized as follows: signal anchor sequences selective for protein insertion into the outer membrane (3)(4)(5), stop-transfer sequences that arrest and integrate proteins in either the outer or inner membrane (6,7), intermembrane space-sorting signals (8,9), and matrix-targeting signals (10,11). In addition, complex variations of these domains may well exist, especially for proteins that assume polytopic structures within the lipid bilayer of either the outer or inner membrane, e.g. porin (12) and uncoupling protein (13,14), respectively. With the exception of matrix-targeting signals, which are characterized by sequences rich in basic and hydroxylated amino acids with the potential to form an amphiphilic helix (10,15,16), the others contain stretches of hydrophobic residues capable of spanning a membrane lipid bilayer.
Stop-transfer sequences do not contain intrinsic membraneselective targeting information but rather they are passive transmembrane segments that are located downstream of matrix-targeting signals, causing an otherwise matrix-destined protein to arrest translocation and insert into the outer or inner membrane. This is exemplified by yeast cytochrome oxidase subunit Va which has been shown to reside in the inner membrane and whose sorting signals are consistent with an NH 2 -terminal matrix-targeting signal combined with a downstream stop-transfer (7,25). Introducing a heterologous stoptransfer sequence, derived from Vesicular stomatitis virus G protein (26), into pre-ornithine carbamoyltransferase (pOCT) 1 downstream of the matrix-targeting signal causes this otherwise matrix-destined protein to insert into the inner membrane (27). However, when it is placed contiguous to the matrixtargeting signal, the protein arrests in the outer membrane (28).
Likewise, intermembrane space-sorting signals contain a stretch of hydrophobic residues immediately downstream of a matrix-targeting signal (29), and indeed there is one example of a protein with such a sequence, NADH-cytochrome-b 5 reductase, that sorts to both the outer membrane and intermembrane space (30). Controversy remains, however, concerning the pathway followed by proteins bearing intermembrane space-sorting signals. In one model the hydrophobic stretch is suggested to arrest translocation during unidirectional import within the inner membrane translocation machinery; processing of the precursor on either side of the membrane then liberates the mature protein into the intermembrane space (29). Another model proposes that following removal of the NH 2 terminus in the matrix, the hydrophobic domain redirects the precursor protein back from the matrix compartment to the intermembrane space (31). It has been consistently observed, however, that the hydrophobic domain, while capable of translocation arrest within the inner membrane import machinery, does not integrate into the surrounding bilayer. Clearly, this domain is functionally different from the structurally analogous regions in signal anchor and stop-transfer sequences.
Here we have addressed the question of how apparently similar stretches of hydrophobic amino acids within precursor proteins can function to target proteins to different locations within the mitochondrion. To do so, we have determined conditions and modifications that result in the transmembrane segment of a signal anchor sequence being inserted into either the outer or inner membrane, passing across both membranes without being arrested, or being arrested across the inner membrane without inserting into the bilayer. This recapitulation of targeting of the hydrophobic domain to different compartments was found to depend on several factors: 1) its net hydrophobicity, 2) whether or not it is permitted to pass across the outer membrane, 3) its distance from a matrix-targeting signal, and 4) the relative strength of the matrix-targeting signal.

MATERIALS AND METHODS
General Procedures-Previous articles describe the routine procedures used in this study (Refs. 17 and 32 and the references cited therein). These include in vitro transcription of pSP64 plasmids, translation of the resulting mRNA in rabbit reticulocyte lysate in the presence of [ 35 S]methionine, purification of mitochondria from rat heart and of mitoplasts from rat liver, protein import in vitro, and analysis of import products by SDS-polyacrylamide gel electrophoresis and fluorography.
Mitochondrial Import-Reaction mixtures contained 10% (v/v) rabbit reticulocyte lysate transcription-translation products labeled with [ 35 S]methionine, mitochondria or mitoplasts (0.5 mg protein/ml), 0.125 mM sucrose, 32 mM KCl, 0.8 mM magnesium acetate, 9.0 mM Hepes, pH 7.5, 0.5 mM dithiothreitol, 0.5 mM ATP, 2.5 mM sodium succinate, 0.04 mM ADP, and 1.0 mM potassium phosphate, pH 7.5. Some reaction mixtures also contained 1.0 M carbonyl cyanide m-chlorophenylhydrazone (CCCP) as indicated in the figure legends. The import reaction mixtures were incubated at 4 or 30°C for 30 min. For post-trypsin treatment, reaction mixtures were incubated with trypsin (0.125 mg/ ml) for 20 min on ice after which soybean trypsin inhibitor (1.25 mg/ml) was added, and the incubation continued for 10 min. The mitochondria or mitoplasts were recovered by layering 50-l aliquots over a 750-l sucrose cushion (0.25 mM sucrose, 10 mM Hepes, pH 7.5, 1.0 mM dithiothreitol) and centrifuging at 12,000 ϫ g for 6 min. Pellets were prepared for SDS-polyacrylamide gel electrophoresis either directly or after suspending in freshly prepared 0.1 M Na 2 CO 3 , pH 11.5 (alkali), to a final concentration of 0.25 mg of protein/ml and incubating on ice for 30 min with periodic vortexing. Membranes were collected by centrifugation at 30 p.s.i. for 10 min in a Beckman Airfuge (Beckman Instruments).

RESULTS AND DISCUSSION
A schematic of the various protein constructs employed in this investigation is presented in Fig. 1. pOMD29 contains the NH 2 -terminal signal anchor domain of Tom70p (amino acids 1-29) fused to dihydrofolate reductase (DHFR) and is targeted and inserted into the outer mitochondrial membrane in the N in -C cyto orientation (Fig. 5A) (3,4,20). Replacement of the extreme hydrophilic NH 2 terminus of the pOMD29 signal anchor with the matrix-targeting signal of pOCT created pO-SA 36 (formerly pO-OMD), which inserts into the outer membrane in an orientation opposite that of pOMD29, i.e. N cyto -C in (Fig.  5A) (17,20). Deletion of the predicted transmembrane portion (amino acids 11-29) of the pOMD29 signal anchor abolishes the ability of the protein to target mammalian mitochondria in vitro (4), whereas the pO-SA 36 fusion construct containing the pOCT matrix-targeting signal but lacking the Tom70p transmembrane segment (i.e. pO-DHFR) is efficiently imported to the matrix (34, 35) (Fig. 5A).
A Downstream Signal Anchor Transmembrane Segment Does Not Arrest Transport of a Matrix-destined Protein-In pO-SA 36, the transmembrane portion of the Tom70p signal anchor is contiguous to the pOCT matrix-targeting signal. To investigate the consequences of physically separating these domains, a spacer region was introduced by replacing pOCT amino acids 1-36 in pO-SA 36 with pOCT amino acids 1-141 or amino acids 1-242 to create pO-SA 141 or pO-SA 242, respectively (Fig. 1). The import of pO-SA 141 was compared with that of pO-141⌬, which is pO-SA 141 lacking the Tom70p transmembrane segment. Both constructs contain the matrix signal sequence processing site, which occurs between amino acids 32 and 33 of the pOCT sequence (27). As shown in Fig. 2, A and B, the pattern of import of pO-SA 141 and pO-141⌬ into intact mitochondria was very similar to that of the control matrix protein, pO-DHFR. For all three constructs, sedimentation and processing was dependent on the presence of mito-  3 with lane 4). The processed, but not the full-length precursor, forms of all three polypeptides were resistant to external protease ( Fig. 2A, lane 5), and the appearance of the processed molecules was dependent on the electrochemical gradient across the inner membrane since it was abolished by CCCP ( Fig. 2A, lane 6). These results indicate that all three molecules were transported across the outer membrane and that at least their amino termini were located in the matrix compartment, which is the site of signal sequence cleavage. Of particular note, however, is that the processed forms of the three polypeptides were extractable by alkali (Fig. 2B,  lanes 5 and 6) indicating a failure to integrate into a membrane lipid bilayer. This suggests, but does not prove, that the polypeptides were translocated entirely to the matrix compartment. Identical results to those presented in Fig. 2, A and B, were obtained for pO-SA 242 and pO-242⌬ (pO-SA 242 lacking the transmembrane region, Fig. 1) (data not shown) suggesting that the failure of pO-SA 141 and pO-SA 242 to insert into mitochondrial membranes was unlikely to be the result of the immediate polypeptide context of the Tom70p transmembrane segment.
To examine the possibility that the transmembrane portion of the Tom70p signal anchor in pO-SA 141 arrested import of the polypeptide across the inner membrane but failed to permit release from the translocation pore into the surrounding lipid bilayer, import of pO-SA 141 was examined in mitoplasts and compared with various control polypeptides. The generation of mitoplasts (32) was monitored by the release of the intermembrane space marker, sulfite oxidase, which was over 90% complete as judged by Western blot analysis (not shown). As shown in Fig. 3 (top panel), pO-SA 141 was imported and processed (lane 4), and the processed form of the molecule demonstrated ⌬⌿-dependent resistance to exogenous trypsin (compare lanes  5 and 6 with lane 4) indicating complete translocation of the polypeptide chain to the soluble matrix compartment. Consistent with this conclusion, both pO-DHFR and pO-SA 141 were also protected from trypsin following import into intact mitochondria and subsequent hypo-osmotic shock of the organelle to disrupt the outer membrane (20) (not shown). As expected (see also Fig. 2), the imported product in mitoplasts was extracted by alkali (lane 9). Very similar import results were obtained for pO-141⌬ and pO-DHFR (Fig. 3, panels 2 and 3) and for pO-SA 242 and pO-242⌬ (not shown). Finally, pO-SA 141 was completely degraded by trypsin following incubation with intact mitochondria in the presence of CCCP (not shown), indicating that pO-SA 141 did not insert into the outer membrane even in the absence of an electrochemical potential across the inner membrane.
Previous studies have documented import and insertion of pOMD29 into the inner membrane of mitoplasts in the N in -C out orientation (3) (Fig. 5B). As shown in Fig. 3, panel 5, this  6 and 7) or absence (lanes 2-5, 8, and 9) of 1.0 M CCCP. In lanes 5 and 6 the mitoplasts were subjected to treatment with protease at the end of the import reaction (Post-Trypsin), whereas the alkali-resistant fraction was analyzed in lanes 8 and 9 (Alkali). p and m refer to precursor and processed (mature) polypeptides, respectively, and are designated by arrows.
results in acquisition of resistance to extraction by alkali (compare lanes 4 and 9) but leaves the bulk of the polypeptide exposed at the surface of mitoplasts where it is susceptible to degradation by exogenous trypsin (compare lanes 4 and 5). Thus, the transmembrane segment of Tom70p is competent for insertion into the inner membrane in the context of pOMD29 but not in the context of pO-SA 141 or pO-SA 242. However, of particular interest were the findings with pO-SA 36, in which the Tom70p transmembrane segment is immediately adjacent to the pOCT matrix-targeting signal. Import and processing of pO-SA 36 (Fig. 3, panel 4) was dependent on the presence of mitoplasts (compare lanes 2 and 4) with an intact electrochemical potential (compare lanes 4 and 7), indicating that the NH 2 terminus of the polypeptide reached the matrix. The bulk of the polypeptide, however, was accessible to trypsin (lane 5) and therefore was located outside the organelle. This is in distinct contrast to pO-SA 141, pO-141⌬, and pO-DHFR, for which the processed forms of the polypeptides were protected against trypsin (panels 1-3, compare lanes 4 and 5). In addition, and in contrast to pOMD29, the processed form of pO-SA 36 was extractable by alkali ( compare lanes 4 and 9). Taken together, therefore, these results suggest that the Tom70p signal anchor transmembrane segment caused pO-SA 36 to pause or arrest during translocation across the inner membrane, but it did not trigger release of the transmembrane segment into the surrounding lipid bilayer.
Increasing the Hydrophobicity of the Tom70p Signal Anchor Transmembrane Segment Permits Insertion into the Inner Membrane-In a previous study (22), mutations were introduced into the transmembrane segment of the Tom70p signal anchor, in which alanines at positions 14, 15, 17, and 18 were converted to isoleucine. These changes did not affect the ability of the signal anchor in the context of pOMD29 to select and insert into the mitochondrial outer membrane in vitro (22). Here, the identical changes were introduced into the Tom70p hydrophobic domain of pO-SA 141, to create pO-SA 141-I4 (Fig.  4A). They resulted in the mean hydrophobicity of this segment increasing from 1.17 to 1.74 (Fig. 4A) employing the hydrophobicity scale of Kyte and Doolittle (36) (Secondary Structure Prediction, Prosis Program, Hitachi Software Engineering Co., Ltd.). Import and processing of pO-SA 141-I4 was dependent on mitochondria (Fig. 4B, compare lanes 2 and 4), and the processed product, but not the full-length precursor, acquired resistance to external trypsin (lane 5) that was dependent on the presence of an electrochemical gradient (lane 6). Thus, the trypsin-resistant component of pO-SA 141-I4 had crossed the outer membrane and deposited its NH 2 terminus into the matrix. However, in contrast to pO-SA 141 (Figs. 2 and 3) and pO-DHFR (Figs. 2, 3, and the lower panel in Fig. 4B) that are imported to the matrix where they remain extractable by alkali, pO-SA 141-I4 was resistant to alkaline extraction (Fig.  4B, compare lanes 4 and 7). Therefore, the results show that pO-SA 141-I4 was inserted into the inner membrane. Since the polypeptide has a single downstream transmembrane segment, this means that this bitopic-processed polypeptide spans the inner membrane once, leaving the COOH terminus in the intermembrane space. This was confirmed following import into mitoplasts, where subsequent digestion of processed pO-SA 141-I4 by trypsin yielded a polypeptide fragment whose size was consistent with an N in -C out orientation (data not shown).
Conclusions-In this study, we have examined the function of the Tom70p signal anchor transmembrane segment when placed in different contexts relative to a matrix-targeting signal. The results are summarized in Fig. 5. When the Tom70p transmembrane segment is contiguous to the pOCT matrixtargeting signal (pO-SA 36) and presented to intact mitochondria in vitro, it inserts into the outer membrane in the N cyto -C in orientation, which is opposite that observed for the native Tom70p signal anchor (pOMD29) (Fig. 5A). When this same polypeptide construct, pO-SA 36, is presented to mitoplasts, the transmembrane segment causes arrest of the polypeptide across the inner membrane in the N in -C out orientation, but it does not trigger insertion into the membrane lipid bilayer (Fig.  5B). If placed at some distance downstream of the matrixtargeting signal (pO-SA 141 and pO-SA 242) and presented to either intact mitochondria or mitoplasts, the transmembrane segment is no longer capable of arresting translocation or triggering insertion into either mitochondrial membrane, and the protein is translocated entirely to the matrix compartment. Increasing the net hydrophobicity of the Tom70p transmembrane segment within the context of this latter construct, however, results in its insertion into the inner membrane in the N in -C out orientation (pO-SA 141-I4). These findings are similar to those found for bacteria, where it has been shown that protein insertion into the cell membrane requires a threshold hydrophobicity for the transmembrane segment (37).
The observed outcomes that were specified by the hydrophobic domain within the various polypeptide constructs examined in this study closely mimic those specified by hydrophobic domains that exist within native proteins: signal anchor sequences that direct insertion into the outer membrane, stoptransfer sequences that specify insertion into the outer or inner membrane, and intermembrane sorting sequences that cause translocation arrest across the inner membrane. The results, therefore, help to identify characteristics of these closely related hydrophobic segments that are likely important for function. In particular, our findings suggest that insertion of a potential transmembrane segment into the inner membrane requires a relatively high net hydrophobicity when this segment is located at some distance downstream of a strong matrix-targeting sequence. In the absence of an outer membrane (i.e. mitoplasts), however, a transmembrane segment of lower hydrophobicity will either insert into the inner membrane if located adjacent to a weak matrix-targeting signal (i.e. the native Tom70p signal anchor) or it will only arrest translocation if adjacent to a strong matrix-targeting signal. The ability of a hydrophobic domain to insert into the inner membrane, therefore, likely depends on four inter-related factors: 1) its net hydrophobicity, 2) whether or not it is permitted to pass across the outer membrane, 3) its distance from a matrix-targeting signal, and 4) the relative strength of the matrix-targeting signal. How the constituent components of the outer and inner membrane translocation machineries discriminate between these different contexts and control protein sorting, however, is not known. Presumably, it involves a complex interplay between the dynamic and reversible interactions that can occur between the two import machineries and, in addition, may result from the different requirements that individual precursor proteins may have for ATP, the electrochemical potential, and chaperone interactions.