PROTEIN IMPORT INTO MITOCHONDRIA

Transcription

1 Annu. Rev. Biochem : Copyright c 1997 by Annual Reviews Inc. All rights reserved PROTEIN IMPORT INTO MITOCHONDRIA Walter Neupert Institut für Physiologische Chemie der Universität München, Goethestrasse 33, D-80336, München, Germany KEY WORDS: mitochondria, protein transport, chaperones, membranes, protein folding ABSTRACT Mitochondria import many hundreds of different proteins that are encoded by nuclear genes. These proteins are targeted to the mitochondria, translocated through the mitochondrial membranes, and sorted to the different mitochondrial subcompartments. Separate translocases in the mitochondrial outer membrane (TOM complex) and in the inner membrane (TIM complex) facilitate recognition of preproteins and transport across the two membranes. Factors in the cytosol assist in targeting of preproteins. Protein components in the matrix partake in energetically driving translocation in a reaction that depends on the membrane potential and matrix-atp. Molecular chaperones in the matrix exert multiple functions in translocation, sorting, folding, and assembly of newly imported proteins. CONTENTS INTRODUCTION MITOCHONDRIAL TARGETING SIGNALS Matrix-Targeting Signals (Presequences) Internal Targeting Signals Multiple Targeting of Proteins to Mitochondria and Other Subcellular Locations MOVEMENT OF PREPROTEINS THROUGH THE CYTOSOL Cytosolic Factors with Presumptive Chaperone Functions Cytosolic Factors with Targeting Function Cotranslational vs Posttranslational Protein Import THE PREPROTEIN TRANSLOCASE OF THE OUTER MEMBRANE: TOM COMPLEX. 876 Constituents of the TOM Complex Receptor Function of the TOM Complex Translocation Pore Pathways of Preprotein Translocation Across the Outer Membrane THE PREPROTEIN TRANSLOCASE OF THE INNER MEMBRANE: TIM COMPLEX AND THE MITOCHONDRIAL HSP70 SYSTEM /97/ $

2 864 NEUPERT Components of the Preprotein Translocase of the Inner Membrane Translocation Function of the TIM Complex Mechanism and Energetics of Preprotein Translocation and Unfolding Dynamic Interaction Between TIM and TOM Complexes PROTEOLYTIC MATURATION OF PREPROTEINS IN THE MATRIX The Mitochondrial Matrix Processing Peptidase Mitochondrial Intermediate Peptidase (MIP) FOLDING OF PROTEINS IN THE MITOCHONDRIAL MATRIX Role of the Mitochondrial Hsp70 Chaperone System Role of the Hsp60-Hsp10 Chaperonin System Role of Peptidyl Prolyl cis-trans Isomerases (PPIases) Chaperoning of Folding and Assembly of Proteins Synthesized in Mitochondria SORTING OF PROTEINS TO THE VARIOUS SUBCOMPARTMENTS OF THE MITOCHONDRION Insertion of Proteins into the Outer Membrane Sorting to the Intermembrane Space Sorting of Proteins to and Integration into the Inner Membrane PERSPECTIVES INTRODUCTION Mitochondria arise by growth and division of pre-existing mitochondria. This growth occurs by insertion of newly synthesized constituents, resulting in expansion of the various compartments of each mitochondrion. The uptake of the protein components of mitochondria has attracted much attention during recent years. The present review is focused on protein targeting to mitochondria. Because the last review of this field in this series (1) appeared seven years ago, the bulk of the literature prior to 1990 is not readdressed here. Several reviews with varying formats and emphases have been published in the interim (2 39). In addition, comprehensive collections of methods applied to the study of mitochondrial protein import have been published (40, 41). Recent years have also brought considerable advances in the field of transport of other constituents into mitochondria, in particular lipids and RNA; this progress is not discussed here (42 45). Interesting connections between RNA, lipid, and protein transport will undoubtedly be made in the future. Mitochondria have an outer membrane, an inner membrane, a compartment between these two membranes (intermembrane space), and an internal compartment (matrix). The fundamental aspect of protein targeting to mitochondria is the transfer of nuclearly encoded, cytoplasmically synthesized proteins both across and into the mitochondrial membranes. Many hundreds of different mitochondrial proteins participate in these processes. In contrast, only a few protein components are encoded by mitochondrial DNA, synthesized on mitochondrial ribosomes, and inserted from the matrix side into the inner membrane. Significant advances in understanding mitochondrial protein import have been made in recent years: A number of components involved in protein targeting

3 PROTEIN IMPORT INTO MITOCHONDRIA 865 to mitochondria have been discovered; a general pathway directing preproteins into the matrix of mitochondria has been defined; and a series of pathways that lead mitochondrial precursor proteins to their sites of function have been elucidated. These achievements have come mainly from the study of yeast and Neurospora crassa. Recent evidence indicates, however, that transport in mammals occurs by very similar reactions (46). The individual steps of the general import pathway currently known are listed in Table 1. In addition, it has become clear that a variety of import routes deviate from the general pathway. Hence, Table 1 also presents specific routes that direct preproteins to the four mitochondrial subcompartments: the outer membrane, the intermembrane space, the inner membrane, and the matrix. Research over the past decade or so has significantly increased our knowledge of the mechanisms of the reactions listed in Table 1. A substantial number of new elements have been identified that function in mitochondrial protein Table 1 Steps of protein import into mitochondria General import pathway 1. Synthesis of proteins on cytosolic ribosomes as preproteins and release into the cytosol 2. Transfer to the mitochondria assisted by cytosolic factors that help to maintain import competence and prevent aggregation 3. Recognition of preproteins by interaction of mitochondrial targeting signals with receptors on the surface of the outer membrane of mitochondria 4. Initiation of transfer through the preprotein translocase complex of the outer membrane (TOM complex) 5. Interaction of preproteins with the surface of the inner membrane and insertion into the preprotein translocase of the inner membrane (TIM complex) triggered by the membrane potential, ψ 6. Completion of translocation through outer and inner membrane by the matrix-localized mt-hsp70-atp dependent driving system associated with the TIM complex 7. Proteolytic processing of preproteins with cleavable targeting signals in the matrix 8. Folding of proteins in the matrix assisted by the molecular chaperone systems, mt-hsp70, Hsp60, and associated co-chaperones Specific targeting pathways 1. Insertion of preproteins into the outer mitochondrial membrane facilitated by the TOM complex 2. Translocation of preproteins across the outer membrane into the intermembrane space through the TOM complex 3. Insertion of proteins into the inner membrane facilitated by both TOM and TIM complexes without passage through the matrix 4. Insertion of preproteins into the inner membrane after partial or complete passage through the matrix 5. Translocation of preproteins into the intermembrane space after partial or complete passage through the matrix 6. Translocation of cytochrome c across the outer membrane without mediation by the TOM complex

4 866 NEUPERT kinesis: constituents of the preprotein translocases in the two mitochondrial membranes, components of the machineries mediating sorting and facilitating folding in the matrix, and enzymes involved in processing and modification of preproteins. Furthermore, we now have a much better understanding of the energetics of import. The unraveling of the more specific import pathways is a subject currently attracting increasing attention. The rapid progress of this burgeoning research has led to some heterogeneity in the nomenclature for various components; recently, however, a unifying convention was adopted (47). MITOCHONDRIAL TARGETING SIGNALS Targeting signals are defined as sequences in preproteins that are both necessary and sufficient to direct the proteins to mitochondria. Additional signals for directing proteins to their proper destination within the mitochondrion are discussed in a subsequent section. Most targeting signals are contained within N-terminal segments (presequences), the majority of which are cleaved upon import into the mitochondria. Since they direct preproteins, at least partially, into the matrix space, presequences are also called matrix-targeting sequences/signals. Many preproteins, on the other hand, contain targeting sequences that reside not at the N terminus but more internally in the proteins. Matrix-Targeting Signals (Presequences) The structure and function of matrix-targeting signals were studied in early research on mitochondrial protein import. Fusions between such sequences and passenger proteins were found to be imported into mitochondria (48 50). Numerous observations of this kind, as well as mutational analyses, have confirmed the role of these matrix-targeting signals. STRUCTURE OF MATRIX-TARGETING SEQUENCES The basic features of matrixtargeting sequences are as follows: about amino acid residues with abundant positive charges, very few if any negative charges, and frequent hydroxylated residues. Targeting sequences are predicted to form amphipathic α-helices in membranes or in membrane-like environments, whereas in aqueous solution they show little structural organization (51 55). The amphipathic nature of these structures is thought to be important for their specific recognition by the protein import machinery. The structures of synthetic peptides corresponding to presequences have been analyzed using biophysical techniques to define their conformational states under various conditions and to study how they interact with lipid bilayers and with biological membranes, in order to understand targeting. The potential to form amphipathic structures is supported by experimental findings (56, 57).

5 PROTEIN IMPORT INTO MITOCHONDRIA 867 An in-depth analysis of the signal sequences of rat liver aldehyde dehydrogenase (paldh) and cytochrome oxidase subunit IV by two-dimensional nuclear magnetic resonance (NMR), fluorescence methods, and circular dichroism measurements yielded a detailed picture of their structure in lipid micelles (58 62). The prepeptides adopted an ordered conformation when the negatively charged phospholipid, cardiolipin, was present in unilamellar vesicles. The prepeptide of paldh is composed of two amphiphilic α-helical segments separated by a short linker. Both helices are necessary for the presequence to facilitate import; thus, it was proposed that an α-helix longer than 11 residues can take on the necessary conformation to be imported (61). Similar observations were made with N-terminal signal sequences of preproteins that are not cleaved after import into the matrix, such as rhodanese, 3-oxo-acyl-CoA thiolase, and chaperonin 10 (63 65). Peptides corresponding to the N-terminal sequences of these proteins can also adopt an ordered amphiphilic secondary structure in nonaqueous environments. INTERACTION OF MATRIX-TARGETING SEQUENCES WITH LIPID BILAYERS Several studies have addressed the question of whether the initial interaction of presequences could occur via direct association with membrane lipids. Binding of a synthetic mitochondrial presequence to unilamellar vesicles was observed, and electrostatic effects were proposed to be significant for this process (66, 67). Furthermore, import of a mitochondrial presequence into protein-free phospholipid vesicles, as well as into the bacterium Paracoccus denitrificans, was reported (68, 69). There is, however, some disagreement about the extent of import because deep penetration of presequence peptides into the lipid bilayer of vesicles was not seen in other studies in which only low concentrations of the peptides were applied (70). A peptide corresponding to the presequence of yeast cytochrome oxidase subunit IV, in addition to having the ability to insert into monolayers, was found to facilitate contact between the bilayers of unilamellar vesicles (71 73). It was proposed that contact-site formation between the outer and inner mitochondrial membranes may occur by such a mechanism and might play a role in the import process. Whether such model systems accurately represent the initial steps in the physiological import process remains to be determined. An alternate view is that the presequences interact directly with the receptor components of the translocase-of-the-outer-membrane (TOM) complex; the receptors presumably provide binding sites for both the hydrophobic and the positively charged faces of the presequence α-helix and direct the preproteins into a translocation channel. If so, the presequences would not interact with or even become inserted into the lipid bilayers of either of the two membranes (38, 74).

6 868 NEUPERT INTERACTION OF PREPEPTIDES WITH ISOLATED MITOCHONDRIA Interaction between isolated mitochondria and synthetic peptides corresponding to presequences has also been analyzed. Import of peptides corresponding to both cleaved and uncleaved targeting signals into isolated mitochondria was investigated (60, 75). Results suggest that the presequences initially interact with the surface and are then stably integrated into the hydrophobic membrane. Import of the model peptides, in some cases, was not dependent on surface receptors (76) or on the membrane potential across the inner membrane. Imported peptides were found in the mitochondrial membranes but not the matrix (77, 78). Synthetic presequence peptides were able to compete with authentic preproteins for import; however, this competition did not occur at the level of the surface receptors but apparently at a later stage (79, 80). In a similar study, binding and insertion of a synthetic presequence into yeast mitochondria was analyzed; reversible association with the surface of the outer membrane leading to partitioning into the lipid bilayer of the outer membrane was observed (81). In summary, it appears that prepeptides can be assimilated into isolated mitochondria, but whether the pathway of uptake is the same as that of authentic preprotein import is an open question. Uncoupling of mitochondria caused by high concentrations of these membrane-active prepeptides may lead to nonphysiological reactions (82). Internal Targeting Signals Although there is a wealth of information on targeting signals present at the N-termini of preproteins, little is known about internal signals. The analysis of internal signals, especially by genetic manipulation, is inherently difficult because alteration of sequences may modify the overall conformation of a preprotein and thereby the accessibility of its signal. The paucity of information about internal targeting signals is unfortunate, given that a large number of outer and inner membrane proteins, as well as proteins that reside in the intermembrane space, have such signals. For example, most of the components of the TOM and the translocase-of-the-inner-membrane (TIM) complexes have internal targeting signals. INTERNAL TARGETING SEQUENCES OF INNER MEMBRANE PROTEINS The inner membrane of mitochondria contains a large family of carrier proteins, of which the ATP/ADP translocase (AAC) is the most prominent member. The AAC is distinguished by the absence of an N-terminal cleavable targeting sequence. An early discovery showed that AAC and other carrier proteins harbor more than one internal targeting signal (83 85), yet little more has been learned subsequently. It has been shown, however, that addition of a matrix-targeting

7 PROTEIN IMPORT INTO MITOCHONDRIA 869 signal to the N termini of proteins with internal signals can override the internal targeting signal. For instance, both an outer membrane porin and the inner membrane uncoupling protein were delivered into the matrix by such a manipulation (86, 87). On the other hand, some carrier preproteins have N-terminal cleavable sequences (88 91); but these N-terminal sequences are not required for targeting to mitochondria (90, 91), although an enhancing effect of the presequence was observed, at least for the mammalian phosphate carrier (90). The presequence itself can target a passenger protein to mitochondria with low efficiency (90). An internal targeting signal was identified in the inner membrane protein, BCS1, which plays an essential role in the biogenesis of the Rieske iron-sulfur protein (92). BCS1 is anchored in the inner membrane by a single hydrophobic stretch with its N terminus facing the intermembrane space and its C terminus facing the matrix (92, 93). The hydrophobic anchor is followed by a positively charged stretch that has the characteristics of a classical matrix-targeting signal. In fact, when the N-terminal leader and the hydrophobic anchor were deleted, the protein was efficiently targeted to the matrix. The positively charged segment in this construct was even cleaved by the matrix processing peptidase (MPP), as its sequence contains a typical MPP cleavage site, although this site is not used in the context of the intact protein. It appears that the internal matrix-targeting signal, along with the hydrophobic segment, forms a loop during import, leading eventually to an N out -C in orientation (93). In this regard, data from studies of cytochrome oxidase subunit Va, COXVa, are also interesting. Deletion of the matrix-targeting signal of COXVa impaired import of the protein, but overexpression of presequence-deficient COXVa permitted import (94). This observation suggests that COXVa contains a second, weaker and internal targeting signal that normally does not play a role but can be recruited to mediate import. These data also show that the criteria for interaction between a mitochondrial presequence and the mitochondrial import machineries may not have very strict specificity. This conclusion is consistent with early reports showing that many randomly generated presequences can function as matrix-targeting signals (95 97). The matrix-located MTF1 protein in yeast, which is a transcription-stimulating factor (98), is an exceptional case. This protein lacks a recognizable matrix-targeting sequence, and its import is reported to be independent of outer membrane receptors,, and ATP. How specificity of targeting is achieved in this case and whether there is an entirely separate pathway for importing this protein remain to be clarified. INTERNAL TARGETING SIGNALS OF OUTER-MEMBRANE PROTEINS The targeting signals of outer-membrane proteins with a single N-terminal anchor have been

8 870 NEUPERT analysed in some detail with Tom70. Yeast Tom70 exposes approximately 10 N-terminal amino residues into the intermembrane space, which are followed by an 20-residue membrane anchor and a large 60-kDa C-terminal domain in the cytosol (99, 100). The information for both targeting and membrane integration is located in the first 30 residues (101). To study targeting of Tom70, a fusion protein was generated consisting of the N-terminal 29 residues fused to mouse cytosolic dihydrofolate reductase (DHFR) (102). Extending the length of the intermembrane space segment to 38 residues and increasing the number of positive charges from four to eight led to an inverse orientation of the fusion protein; reduction of the hydrophobicity of the transmembrane anchor restored this fusion protein to its normal orientation (103). Deletion of the N-terminal 10 residues did not impair targeting to the outer membrane (100). Apparently, a combination of the hydrophobicity of the membrane segment and the nature of its flanking sequences determines targeting and membrane insertion. How this information is decoded by the import machinery of the outer membrane is unclear. The specificity of suborganellar targeting is difficult to explain, given the observation that Tom70 can insert into the inner membrane of mitochondria that have a disrupted outer membrane (104, 105). In proteins anchored to the outer membrane by hydrophobic segments located close to the C terminus (29, 106, 107), the targeting signals appear to be located in the C-terminal region. The specificity of recognition and the mechanism of insertion are not understood in these cases. Multiple Targeting of Proteins to Mitochondria and Other Subcellular Locations Differential distribution of isoenzymes between mitochondria and other subcellular compartments is usually achieved by having two (or more) closely related genes in the nucleus. However, there are cases in which the product of a single gene is targeted to different locations. How is the same gene product delivered to two different subcellular destinations? The first example studied was histidinyl-trna synthetase in yeast. In this case, two different mrnas are produced from a single gene: One encodes a stretch representing a mitochondrial targeting sequence, and the other lacks this 5 -extension (108). Three trna-processing enzymes, encoded by the TRM1, MOD5, and CCA1 genes in yeast, are present in mitochondria, as well as in the nuclear and cytosolic compartments. The genes encoding these three enzymes have more than one in-frame ATG start codon. The choice of the ATG is dictated by transcription start site selection, by translational selection, or by an interplay between these reactions. The larger proteins are targeted to mitochondria because they contain mitochondrial targeting signals in their N-terminal extensions; the shorter proteins remain in the cytosol or are targeted

9 PROTEIN IMPORT INTO MITOCHONDRIA 871 to the nucleus by nuclear localization signals. The mitochondrial targeting information resides entirely within sequences that are removed by subsequent proteolytic cleavage, or may be present partly in the presequence and in portions that remain in the mature protein ( ). In N. crassa, a cyclophilin in the mitochondria and in the cytosol is encoded by a single gene that is transcribed into distinct mrna species. The larger translation product, after its transfer into the mitochondria, is cleaved by the matrix-processing protease (MPP) directly after the methionine that is encoded by the ATG that serves as the start codon for the cytosolic version (115). A different situation was found for targeting another single gene product, uracil- DNA glycosylase, to mitochondria, cytosol, and nucleus. A short form was found in both mitochondria and nuclei, and a long form in the cytosol. Thus, a cleaved presequence could mediate targeting to mitochondria, a shorter translation product might be translocated into the nucleus, and the long translation product residing in the cytosol could represent the precursor for both the mitochondrial and the cytosolic isoforms (116). The targeting of mammalian alanine glyoxylate aminotransferase 1 (AGT) is unusual because its location varies among species ( ). In some species, this protein is contained in mitochondria and is synthesized as a precursor with a 22-residue N-terminal extension. In other species, this protein is present both in mitochondria and in peroxisomes. The species in the peroxisomes lacks the N-terminal extension, owing to alternative transcription- and translationinitiation sites. In this case, a peroxisomal targeting signal (type 1) (120) directs the protein into peroxisomes. In humans, the mitochondrial targeting signal is entirely missing. In fact, in human patients with primary hyperoxaluria, the AGT gene contains mutations that appear to destroy the peroxisomal targeting information yet simultaneously create an (uncleaved) mitochondrial targeting signal at the N terminus ( ). In these patients, the enzyme ends up in the mitochondria and is nonfunctional. A variation on the theme of mistargeting to mitochondria by acquisition of a targeting signal has been described in yeast. An extragenic suppressor of a deletion in the nuclear gene encoding a yeast mitochondrial matrix protein, β-keto-acyl synthase, arose in another nuclear gene that encodes a protein normally located in the cytosol, fatty acyl-coa synthetase. The mutation generated an 18 amino acid residue N-terminal extension with the characteristics of a matrix-targeting sequence (121) that delivered the normally cytosolic protein to the mitochondria. Yeast mitochondrial and cytosolic fumarase are also intriguing because they are encoded by a single gene (FUM1) (122) that produces only one unique primary translation product, i.e. the cytosolic form has the same size as the mitochondrial form. The N terminus of this translation product contains a

10 872 NEUPERT cleavable mitochondrial targeting signal. The mature cytosolic form also appears to be processed by MPP and hence probably arrives back in the cytosol via release from the mitochondria (123). In plants and fungi, transport of proteins into mitochondria is very similar in most respects (124, 125); but in plants, it is not clear how mitochondrial targeting sequences differ from chloroplast transit (targeting) sequences that direct preproteins into the stroma of chloroplasts (126). Surprisingly, in some studies, import of chloroplast preproteins into mitochondria has been reported ( ), whereas mistargeting was not observed in another study (130). MOVEMENT OF PREPROTEINS THROUGH THE CYTOSOL It is widely agreed that the mitochondrial preproteins encoded by nuclear genes are released from cytoplasmic ribosomes after synthesis as completed chains (but see the section on Cotranslational vs Posttranslational Protein Import). Preproteins in the cytosol are generally in a more loosely folded state than their mature counterparts. In addition, preproteins may lose translocation competence, undergo aggregation (131), or be degraded by cellular proteases (1, 34, 38). There are cytosolic factors known as chaperones that can prevent these adverse reactions. The requirement for such factors may not be general, because some purified preproteins are efficiently imported into isolated mitochondria in vitro (132). A number of cytosolic components reportedly interact with nascent polypeptide chains, i.e. even before they are released from the ribosome, to mediate stabilization and (partial) folding (for review see 35). In addition, some proteins in the cytosol may exert a more specific function, namely guiding preproteins to the surface of mitochondria in a manner similar to the role of signal recognition particles (SRPs) for secretory proteins (38, 133). Indeed, several cytosolic factors that stimulate preprotein import have been described and characterized during the past decade (Table 2). Cytosolic Factors with Presumptive Chaperone Functions Preproteins synthesized in wheat germ extract were observed to be largely incapable of being imported into mitochondria, in contrast to preproteins translated in reticulocyte lysate or in yeast extracts (1, 34). Similarly, preproteins expressed in Escherichia coli and purified from bacterial extracts in the presence of high concentrations of urea were found in many cases to retain import competence only for a short time. Therefore, several attempts were made to isolate activities from yeast extracts, reticulocyte lysates, and rat liver cytosol that preserve import competence or stimulate import. Cytosolic Hsp70 (the eukaryotic homolog of bacterial DnaK) was discovered early on to be an important factor

11 PROTEIN IMPORT INTO MITOCHONDRIA 873 Table 2 Cytosolic components stimulating preprotein import into mitochondria Component Yeast Function References Import stimulating? 134 factor Cytosolic Hsp70 Ssa1p Molecular Ssa2p chaperone YDJ1 Ydj1p Co-chaperone (Mas5p) NEM-sensitive component? 136, 138? Presequence binding factor? 136, 142, 143 (PBF) Targeting factor Mitochondrial import Bmh1p? Targeting 34, 46, stimulating factor Bmh2p? factor (MSF) in intracellular protein traffic. For example, translocation of proteins into mitochondria was impaired in yeast strains in which three of the four SSA genes (which encode isoforms of cytosolic Hsp70) were deleted; preproteins accumulated in the cytosol fraction in these mutant cells (135). Similarly, transport of secretory proteins into the endoplasmic reticulum was defective, suggesting that cytosolic Hsp70 does not have an organelle-specific role. In addition, experiments in vitro further supported a function for cytosolic Hsp70 in protein import into mitochondria (136, 137). Interaction of synthetic prepeptides with isolated Hsp70 (Ssalp) was observed and depended on the amphiphilicity of the prepeptide. It was suggested that in vivo Hsp70 might bind to presequences and thereby promote import (131). The occurrence of Hsp70 on the surface of rat liver and plant mitochondria was reported (153, 154). In reticulocyte lysate, Hsc70 (a 70-kDa heat shock cognate protein) function appears to be required during the synthesis of preproteins; when preproteins were made initially in the absence of Hsc70, import competence could not be restored by subsequent addition of Hsc70 prior to assessing import (155). A relationship between the Hsc70 dependency and the size of import-competent forms of the preproteins was reported, suggesting that a limited number of Hsc70 binding sites exist on a preprotein and that aggregation of proteins is prevented by the chaperoning effect of Hsc70 (156). The Ydj1 protein (139), a farnesylated yeast homolog of bacterial DnaJ, also plays a role in mitochondrial protein import, as found by analysis of conditional mutants of Ydj1p (157, 158). The YDJ1 gene was also identified using an independent approach, namely by analyzing a collection of temperature-sensitive

12 874 NEUPERT mutants that displayed defects in growth and mitochondrial protein import. One such mutation (mas5) turned out to be an allele of ydj1 (140). In general, DnaJ homologs act in concert with DnaK homologs. Thus, Ydj1p may assist cytosolic Hsp70 to cycle on and off unfolded (or partially folded) preproteins, and might help to target preprotein-hsp70 complexes to the surface of the outer membrane of mitochondria where Ydj1 could be anchored via its prenyl chain. The involvement of heat-shock proteins in guiding proteins through the cytosol is illustrated by genetic experiments in which a mutation in the heat-shock transcription factor, Hsf1, led to a defect in protein import into mitochondria (159). Cytosolic Factors with Targeting Function Several of the cytosolic factors reported qualify for targeting factors in the sense that they bind preproteins in a specific manner and deliver them to the receptor components of the TOM complex. These factors were isolated by virtue of their ability to stimulate import of a presequence peptide or by interacting physically with such peptides. Using affinity chromatography with an immobilized prepeptide, a 28-kDa protein was isolated from rabbit reticulocyte lysate (145, 146). This factor was also found to be associated with the surface of rat liver mitochondria, and antibodies against it inhibited preprotein binding and import into mitochondria. This 28-kDa protein, called targeting factor, in many respects resembles MSF (see below), but whether they are identical is unknown. Another component was purified by using as an assay the dependence of import of a purified precursor on addition of reticulocyte lysate (136). A 50- kda protein was isolated and named presequence binding factor (PBF). PBF was reported to bind to preproteins in a presequence-dependent manner and to stimulate protein import. PBF activity did not depend on ATP hydrolysis and was not diminished by N-ethylmaleimide (NEM) treatment. It was suggested that PBF likely maintains the import competence of preproteins in cooperation with cytosolic Hsp70 (142, 143). The best-characterized factor is the mitochondrial import stimulation factor (MSF) (34). It confers import competence to preadrenodoxin precursor synthesized in wheat germ lysate. Purification from rat liver cytosol was achieved on an affinity matrix containing an immobilized presequence peptide. MSF is a heterodimer consisting of 30- and 32-kDa subunits. Purified MSF stimulates import of a variety of preproteins, even those which do not contain a matrix-targeting signal, such as porin precursor. MSF has two activities: It recognizes and forms stable complexes with presequences and mature parts of mitochondrial precursors, and it facilitates depolymerization and is thought to promote unfolding of preproteins. Aggregated preadrenodoxin synthesized

13 PROTEIN IMPORT INTO MITOCHONDRIA 875 in E. coli (160) became import-competent in the presence of MSF. These two activities can be separated because the first, but not the second, can be inhibited by treatment with NEM. This effect may explain the previously observed inactivation of import-stimulating cytosolic activities by NEM (see Table 2). The depolymerization and unfolding activities depend on ATP hydrolysis; they are inhibited by the nonhydrolyzable imidodiphosphate analog of ATP (AMP- PNP). Binding of preprotein or presequence peptides induces the ATPase activity of MSF; positively charged residues in presequences appear to play an important role in this stimulation ( ). Analysis of the relative contributions of cytosolic Hsp70 and MSF yielded some unexpected results (161). Hsp70 and MSF both mediate transfer of preproteins from the cytosol; the involvement of each depends on the nature of the preprotein. Although no ATP dependence was observed with Hsp70, the targeting of both MSF-preprotein complexes and MSF-Hsp70-preprotein complexes to mitochondria required ATP hydrolysis. In the presence of AMP-PNP, import mediated by MSF-Hsp70-preprotein ternary complexes was arrested, and the MSF-preprotein complex was found at the mitochondrial surface. MSF binding to the outer membrane occurred only in the presence of a functional matrix-targeting sequence. Targeting to mitochondria was proposed to occur via two different pathways: One pathway depends on the NEM-sensitive MSF and ATP, and the other pathway depends on Hsp70 and is ATP independent. This concept was then expanded (see section on The Preprotein Translocase of the Outer Membrane: TOM Complex) by showing that a MSF-preprotein complex first binds to the Tom37-Tom70 subunits of the TOM complex, from where the preprotein is transferred upon ATP-dependent release of MSF to the Tom20-Tom22 subunits and enters the translocation channel. In contrast, it was proposed that Hsp70-preprotein complexes interact directly with the Tom20- Tom22 subunits without an ATP requirement. Different precursors may use the two pathways differentially (150). MSF was identified as a member of the family of proteins (151). These proteins have a variety of functions and are found in virtually all organisms (162, 163). As yeast also contains proteins (152), genetic approaches could be used to confirm that the proposed functions in mitochondrial targeting, so far entirely deduced from experiments in vitro, do occur in intact cells. Cotranslational vs Posttranslational Protein Import Ever since mitochondrial protein traffic was first investigated, there have been opposing views on how preproteins are translocated into mitochondria. Cytosolic ribosomes were found to be associated with yeast mitochondria in vivo and in vitro under certain conditions, and some biochemical data were taken

14 876 NEUPERT as evidence of a cotranslational insertion of nascent polypeptide chains into mitochondria (14, 164, 165). On the other hand, kinetic analyses with intact N. crassa cells demonstrated the existence of cytosolic pools of newly synthesized completed mitochondrial polypeptides, which were rapidly turned over (166). The discussion has since been continued (14). Although it seems possible that some preproteins under certain conditions enter the mitochondria cotranslationally, there is no convincing evidence that this is true for the bulk of imported proteins. In fact, a simple calculation for yeast mitochondria (taking as parameters a generation time of 2 3 h, a concentration of the TOM complex in the mitochondria of 5 10 pmol mg 1 (167), and a step time for polypeptide chain elongation of 2 4 s 1 ) suggests that cotranslational translocation would be too slow to account for the measured rates of import. Polypeptide chain elongation is a relatively slow process; for example, translation/translocation of a polypeptide chain of 800 residues would occupy an import site for about 4 min. Therefore, release of polypeptide chains from the ribosomes followed by rapid translocation probably allows the whole process to occur at much higher rates. In contrast to the endoplasmic reticulum, in which the number of import sites is increased by expansion of the membrane when synthesis of secretory proteins is increased, the area of the outer membrane of mitochondria does not expand in cells with actively growing mitochondria, and therefore any increase in the number of import sites is likely to be limited. THE PREPROTEIN TRANSLOCASE OF THE OUTER MEMBRANE: TOM COMPLEX It has become clear that each of the two mitochondrial membranes possesses a translocation machinery for preproteins. These machineries cooperate in the translocation of preproteins into and across the inner membrane, but each translocation complex can also act separately. Using mitoplasts, translocation across the inner membrane can occur in vitro independently of translocation across the outer membrane (168, 169). Further support for the conclusion that the translocation complexes can act independently is the finding that a fusion protein harboring signals for targeting to both the intermembrane space and the matrix is imported in a stepwise fashion: first into the intermembrane space in the absence of, and then into the matrix when is reestablished (170). Furthermore, isolated outer membranes have the capacity to initiate import of matrix-targeted preproteins so that the presequence reaches the inner face of the outer membrane (171). The translocation complex in the outer membrane of N. crassa (172) and yeast (73) is composed of at least eight proteins. These components mediate recognition of preproteins (receptor function), transfer of preproteins through

15 PROTEIN IMPORT INTO MITOCHONDRIA 877 the outer membrane (general insertion pore), and insertion of resident outermembrane proteins. These various roles, initially suggested by functional studies (167), can now be assigned to the individual constituents of the TOM complex. Constituents of the TOM Complex COMPONENTS EXPOSING DOMAINS TO THE CYTOSOL The known components of the TOM complex are listed in Table 3. A uniform nomenclature was introduced recently, based mainly on the components identified in yeast and N. crassa (47); the previously used names are included in Table 3. Some of the TOM complex components (Figure 1) have hydrophilic domains that project into the cytosol. Tom20, Tom70, and Tom71 have N-terminal membrane anchors and hydrophilic C-terminal domains of 17 kda and 65 kda, respectively. Tom70 and Tom71 are closely related structurally (53% sequence identity, 70% similarity) (186, 187), are integrated into the outer membrane in the same orientation, and contain seven tetratricopeptide repeat Table 3 Components of the translocase of the outer membrane Essential for Component Other names Function viability in yeast References Tom20 MOM19 Surface receptor a Mas20p together with Tom 22 Tom22 MOM22 Surface receptor + 172, Mas22p and possibly part of channel Tom70 MOM72 Surface receptor Mas70p Tom71 Tom72 Surface receptor 74, 186, 187 Tom37 Mas37p Partner protein of 188 Tom70; contributes to receptor function Tom40 MOM38 Component of GIP b + 172, 189, 190 ISP42 Tom6 Isp6p Possibly component 191, 192 Mom8b of channel; interacts with Tom40 and preproteins Tom7 MOM7 Component of GIP? 193 Tom5 MOM8a Component of GIP?? 173, 194 a No growth on nonfermentable carbon sources. b General insertion pore.

16 878 NEUPERT Figure 1 Schematic representation of components of the TOM complex. Abbreviations: OM, outer membrane; IMS, intermembrane space. (TPR) motifs, which may have a role in their interaction with other Tom proteins. Tom70 has a tendency to form dimers, and the membrane anchor is responsible for, or at least contributes to, dimerization (100, 194, 195). Tom22 extends an N-terminal domain of 85 amino acid residues into the cytosol, has a single transmembrane segment, and has a smaller C-terminal domain ( 45 residues) facing the intermembrane space. The cytosolic domain of Tom22 is characterized by an abundance of negative charges ( ). Tom37 is also exposed at the cytosolic surface of the mitochondrion and has two predicted transmembrane segments (188). The various Tom proteins are in close proximity to each other. In yeast, Tom20 and Tom70 can be cross-linked and co-immunoprecipitated (196). In N. crassa, Tom22 can be cross-linked to Tom70 (197) and to Tom20 (198). Tom37 forms a 1:1 complex with Tom70 (188). Most of these interactions are labile and are preserved only when mild detergents such as digitonin are used for membrane solubilization (172), as expected if the complex is dynamic. Consistent with this possibility is the fact that only a fraction of the total Tom70 in N. crassa is found in a complex that contains the other TOM components (194, 199). Recently, a human homolog of Tom20 was identified by sequence similarity. This homolog can substitute functionally for the authentic yeast Tom20, suggesting that the structure of the TOM complex is conserved during evolution (178, 179, 200). MEMBRANE-EMBEDDED CONSTITUENTS The Tom40 protein is a component that appears to be deeply embedded in the outer membrane, although it does not have many obvious contiguous hydrophobic sequence segments. It is largely resistant to proteases added to isolated mitochondria (172, 190). It might well

17 PROTEIN IMPORT INTO MITOCHONDRIA 879 contain β-sheet like structures, much like mitochondrial and bacterial porins (201). The smaller Tom proteins were found in N. crassa and in yeast by coimmunoprecipitation with Tom20 or Tom40 (173, 193, 194). In yeast, Tom6, an integral membrane protein, is composed of 61 amino acid residues. This protein was identified in a genetic screen as a component that interacts with Tom40 (191). OTHER COMPONENTS RELATED TO THE OUTER MEMBRANE IMPORT MACHINERY A 40-kDa protein in the outer membrane of yeast mitochondria, termed Msp1, was identified in a genetic screen by virtue of its ability to cause mislocalization of an artificial preprotein to the inner membrane (202). A further characterization of this nonessential integral membrane protein is not available. Search for a high-copy suppressor of a conditional Tom40 mutation yielded a gene that codes for the previously identified mitochondrial RNase P, a ribonucleoprotein enzyme of the matrix (203). This finding might point to an obligatory interaction of the TOM complex with the translocase of the inner membrane to mediate import into the matrix, but a biochemical explanation for the genetic findings is lacking. In the case of rat liver mitochondria, components of 29 kda and 52 kda were purified using a prepeptide affinity column, and a role in recognition of preproteins was proposed (204, 205). The use of anti-idiotype antibodies to identify presequence-binding components has been reported in two studies. In a first step, antibodies were made against presequences, and in a second step, antibodies were made against the initial antibodies. In yeast mitochondria, these anti-idiotype antibodies recognized a 32-kDa protein and inhibited import of preproteins into isolated mitochondria. The 32-kDa protein turned out to be identical to the phosphate carrier, a protein normally found in the inner membrane ( ). It is not clear whether the 32-kDa protein has a role in both phosphate translocation and protein import nor whether it has a dual localization. A similar approach was taken with rat liver mitochondria, and several proteins reacting with the anti-idiotype antibodies were detected (210). These putative presequence receptors await further characterization. The cellular location of Mft1p (211), a yeast component identified by a genetic screen and required for preprotein targeting, has not yet been characterized, nor has the nature of its involvement in mitochondrial targeting. Receptor Function of the TOM Complex Although it was thought for some time that interaction of preproteins with mitochondria and membrane transfer does not require proteinaceous components, it is now clear that a whole set of receptor components is necessary for targeting of preproteins to the surface of mitochondria. Two receptor systems have been discovered, the Tom20-Tom22 subcomplex and the Tom70-Tom71 Tom37

18 880 NEUPERT subcomplex. The vast majority of preproteins, in particular preproteins with matrix-targeting sequences and proteins of the outer membrane, are recognized by the Tom20-Tom22 subcomplex. Furthermore, these components function also in the recognition of those preproteins that are initially recognized by the Tom70-Tom71 Tom37 complex. THE TOM20-TOM22 SUBCOMPLEX The components of the TOM complex were initially discovered in experiments in which antibodies against outer-membrane proteins were analyzed for their ability to inhibit protein import (174, 175). Antibodies against N. crassa Tom20 inhibited the import of preproteins, with the exception of a few, such as the ADP/ATP carrier, which was only moderately affected (174). Similar findings were made with yeast Tom20 (176, 177). Depletion of Tom20 in N. crassa leads to a severe growth defect. Mitochondria loose their cristae, cytochromes, and the ability to synthesize proteins (212, 213). Yeast cells lacking Tom20 are respiratory deficient, do not grow on nonfermentable carbon sources, and grow at a reduced rate on fermentable carbon sources (177, 214). Loss of Tom20 could be compensated by overexpression of Tom70, indicating that either Tom70 shares functions with Tom20 or that altered levels of Tom70 lead to induction of other TOM complex components (177). Both in N. crassa and in yeast, loss of Tom20 is accompanied by reduction in the levels of Tom22. Import of preproteins into Tom20-depleted N. crassa mitochondria in vitro is strongly reduced for most proteins of all submitochondrial compartments. In contrast, import of preproteins for the ADP/ATP carrier or cytochrome c 1 (which in wild-type mitochondria are either not dependent or only partly dependent on Tom20) is unaffected (or minimally affected) in vitro (213). Similarly, in isolated yeast mitochondria lacking Tom20, import of cleavable preproteins, but not of the ADP/ATP carrier, is heavily reduced (214). Experiments in vitro using antibodies against Tom22 showed inhibition of import of almost all preproteins analyzed (182, 198). Tom22 is an essential component both in yeast and N. crassa, and mitochondria lacking Tom22 are defective in import of most preproteins (180, 181, 215, 216). Increased levels of Tom22 in yeast can compensate for loss of Tom20 (180). There are a few preproteins whose import does not depend on the Tom20-Tom22 receptor subcomplex, in particular precursors of outer-membrane proteins such as Tom70, Tom20, and Tom6 (but not, for example, Tom22) (175, 197). For cytochrome oxidase subunit Va, contradictory findings have been reported (217, 218). Tom20 and Tom22 were shown to directly interact with preproteins (198, 219, 220). The subcomplex formed by these two components appears to represent the main structural element of the cis-site, defined as the initial binding of the

19 PROTEIN IMPORT INTO MITOCHONDRIA 881 targeting sequence at the surface of the outer membrane (171). The chemical nature of the binding reaction remains unclear. It has been proposed that the many negatively charged residues on the cytosolic domain of Tom22 provide a docking site for the positive charges on the hydrophilic face of the α-helical structure formed by matrix-targeting signals (182). Such electrostatic forces would be consistent with the observed lability of this interaction, which is especially pronounced under conditions of higher ionic strength (198, 221). Since not only matrix-targeting sequences but also preproteins such as porin, heme lyases, and ADP/ATP carrier (182, 222) are recognized by the Tom20-Tom22 subcomplex, additional as-yet-undefined recognition elements may exist. Most likely, the cis-site provides an extended binding area on which the various targeting signals can dock and are thereby guided into the outer membrane translocation pore (39). The receptors may not be absolutely necessary, at least not in vitro, but may serve to facilitate this insertion reaction. Bypass of the receptors has been observed with isolated mitochondria lacking the cytosolic domains of the receptors (223). THE TOM70-TOM37 SUBCOMPLEX Tom70 was identified in N. crassa as a receptor for the AAC family of carrier proteins (183, 184). In yeast, a previously identified 70-kDa outer-membrane protein of unknown function was found to accelerate import in vitro of certain preproteins such as AAC, the phosphate carrier, the F 1 -ATPase β-subunit, and cytochrome c 1 (185, 224). Antibodies against Tom70 inhibited the specific and productive binding of the AAC precursor at the mitochondrial surface (183), and mitochondria from a Tom70 deletion mutant showed reduced binding of AAC (185). Further in vitro studies were interpreted as suggesting that most preproteins can use Tom70 as an entry site (225). On the other hand, the overall effect of Tom70 deletion is rather mild; a reduction of growth on nonfermentable carbon sources is observed only at higher temperatures. In contrast, Tom20-deficient cells do not tolerate loss of Tom70, emphasizing the important role of Tom70; adaptation of tom20-tom70 double deletion cells led to restoration of slow growth by an unknown mechanism (226). Direct binding of AAC precursor was found by coimmunoprecipitation and by cross-linking experiments (183, 194). The purified cytosolic domain of yeast Tom70 interacts with preproteins, such as AAC and cytochrome c 1, and competes with Tom70 in the outer membrane for binding and import into mitochondria of these preproteins but not of most other preproteins (227). Tom71, a close relative of Tom70, is another constituent of the TOM complex in yeast. Deletion of the TOM71 gene yields a slightly more severe phenotype than deletion of TOM70, but even the double deletion mutant does not show a growth defect at a temperature of 25 C. It may be that these two closely related

20 882 NEUPERT Tom proteins have a similar or redundant function, although their precise roles and the function of Tom71 in relation to Tom70 is not yet clear (186). Import of preproteins into isolated mitochondria in the absence of Tom70 occurs via the Tom20-Tom22 pathway (184, 226). Tom37 by itself is not an essential component, as deletion of the TOM37 gene leads to the same rather inconspicuous phenotype as tom70 or tom71 mutations, namely temperature-sensitivity for respiration-driven growth (188). Apparently, lack of the individual components can be bypassed. On the other hand, double mutants of tom37 with tom70 or with tom20 are lethal (188). Import of certain preproteins into mitochondria from cells lacking Tom37 is reduced. Antibodies against Tom37 strongly inhibit import into isolated mitochondria of AAC and weakly inhibit import of precytochrome c 1. The precise biochemical function of Tom37 remains to be explained. Because it forms a complex with Tom70, it might contribute to the preprotein binding site present on Tom70-Tom71, or it may have some other role in preprotein targeting. As discussed above, the preprotein-msf complex appears to be targeted to the Tom70-Tom71 Tom37 complex of mitochondria. Tom37 may exert a specific function in this context, although direct interaction of Tom37 with preproteins has not been demonstrated (34, 150). Translocation Pore The existence of a general insertion pore (GIP) that guides preproteins through the outer membrane was proposed on the basis of functional studies showing that preproteins use different receptor systems but then pass through a common putative channel (167). The main candidate for a pore component is Tom40, a protein deeply embedded in the membrane (172, 189, 190). Tom40 is in contact with preproteins as they traverse the outer membrane (189). Tom40 is an essential component in yeast. A protein in plant outer mitochondrial membrane of similar size was found to cross-react with anti-tom40 antibodies from yeast and N. crassa and was shown to be involved in import (228). How Tom40 contributes to the structure of the translocation and insertion pore is unknown. The smaller TOM components seem to be part of this pore. Tom7 and Tom5 of N. crassa could be cross-linked to preprotein in transit across the outer membrane. Yeast Tom6 interacts functionally with Tom40, as demonstrated by a genetic study; it is a high-copy suppressor of a temperature-sensitive Tom40 mutant (191). Tom6 forms a complex with Tom40 in mild detergent. Tom6 seems to support the interaction of receptors with the GIP, and the release of preproteins from import components (192). Yeast Tom7 was reported to exert a destabilizing effect on the TOM complex; it was proposed to play a role in sorting and accumulation of preproteins at the outer membrane, to interact with Tom20, and to modulate the dynamics of the outer membrane preprotein translocase (193).

21 PROTEIN IMPORT INTO MITOCHONDRIA 883 Pathways of Preprotein Translocation Across the Outer Membrane CIS-SITE/TRANS-SITE MODEL In addition to the experiments described above, analysis of the interaction of preproteins with isolated outer membrane vesicles has provided important clues to the function of the TOM complex (171, 229). In this in vitro system, preproteins with matrix-targeting sequences can be arrested at the surface of the membrane, specifically in association with Tom20-Tom22; this form was defined as the cis-site for preprotein binding. To stall a preprotein at this site requires that it carry a stably folded domain immediately following its targeting sequence. This binding is very labile and salt-sensitive. If the folded domain can unfold, the preprotein moves on through the translocation pore, a process accelerated by prior interaction with the receptors. The next stage that can be distinguished is binding of the presequence to the so-called trans-site. In this position, the MPP cleavage site of the preprotein is exposed on the inner side of the outer membrane, as shown by using vesicles in which MPP was trapped in the lumen. Trans-site bound intermediates are cleaved by such trapped MPP, but not by MPP added to the outside of the vesicles. After cleavage by MPP, the processed preprotein is released from the vesicles into the supernatant solution; this result demonstrates that binding to the trans-site is presequence dependent. Trans-site intermediates are bound much more firmly than cis-site intermediates, and their binding is largely salt-resistant. Translocation into isolated outer-membrane vesicles proceeds only up to the trans-site. As discussed below, further movement of preproteins with matrixtargeting signals requires the involvement of the translocation system of the inner membrane. The molecular structure of the trans-site is not yet clear. It has been suggested that Tom22 not only contributes to the cis-site (via its N- terminal domain), but also is part of the trans-site (via its C-terminal domain) (220). A different conclusion was reached in two other studies, one with yeast and the other with N. crassa. The C-terminal domain of Tom22, which faces the intermembrane space, could be deleted without a significant reduction of the import of preproteins (202, 230). Among the known TOM complex proteins, a preferred candidate for a trans-site component is Tom40, which has exposed segments that are accessible to the intermembrane space. ENERGETICS OF TRANSLOCATION ACROSS THE OUTER MEMBRANE The cis-site/ trans-site model, in addition to explaining the movement of preproteins across the outer membrane, provides at least a partial explanation for the energetics of translocation across the outer membrane. The presence of two (or more) binding sites with increasing affinity for the presequence provides a driving force for the vectorial movement of the N terminus of preproteins. Furthermore, transsite binding can drive unfolding of a folded domain, such as mouse cytosolic dihydrofolate reductase fused to a matrix-targeting sequence (171). Apparently,

22 884 NEUPERT there is an equilibrium of the folded and a (partly) unfolded state; the latter form of the preprotein undergoes movement through the pore, and trans-site binding. Thereby the equilibrium can be shifted to the unfolded state. This hypothesis is supported by two observations: ATP is not required for this reaction, and addition of the folate antagonist methotrexate leads to refolding and release of the trans-site bound fusion protein back into the supernatant (171). RECONSTITUTION OF THE TOM COMPLEX IN LIPID VESICLES A major challenge for the future is to reconstitute TOM-complex functions into artificial lipid membranes. Some progress toward this aim has been reported (231). A peptidesensitive channel of large conductance was identified by reconstitution into lipid bilayers using three different techniques. This channel was found to be cation selective and blocked by positively charged peptides, in particular peptides corresponding to mitochondrial preprotein sequences. Therefore, the channel was named the peptide-sensitive channel (PSC). PSC was localized to the outer mitochondrial membrane ( ). Reported data show that the voltage-gated PSC may be the preprotein translocating channel of the outer membrane ( ). Further information on the identity of PSC and the relation between electrophysiological properties and preprotein conduction will require reconstitution of the purified TOM complex. THE PREPROTEIN TRANSLOCASE OF THE INNER MEMBRANE: TIM COMPLEX AND THE MITOCHONDRIAL HSP70 SYSTEM The components of the TOM complex were initially identified by biochemical means, whereas the components of the preprotein translocase of the inner membrane (TIM components) were identified by genetic approaches. One main reason for this difference is the paucity of TIM components in the inner membrane ( % of total protein), which makes direct purification very difficult. Discovery of TIM components began with the identification of four genes in yeast that were obtained in a positive selection procedure (239, 240). The URA3 gene product was directed into the matrix of the mitochondria, thereby rendering growth of these cells dependent on the addition of uracil to the culture medium. Mutants growing in the absence of uracil were selected with the expectation that they would harbor conditional defects in the mitochondrial import machinery. In a separate genetic approach, temperature-sensitive mutants that accumulate mitochondrial preproteins at the nonpermissive temperature were generated. This approach had previously led to the identification of the genes for the mitochondrial processing peptidase (241, 242) and Ydj1p (140). Two of the four genes mentioned above were identified in this way (243, 244).

23 PROTEIN IMPORT INTO MITOCHONDRIA 885 Components of the Preprotein Translocase of the Inner Membrane An overview of the presently known and functionally characterized components of the TIM machinery is presented in Table 4 and Figure 2. As for the TOM components, a uniform nomenclature has been introduced recently (47), and the table lists the TIM components and previously used synonyms. Tim23 is an integral 23-kDa protein of the yeast inner membrane (240, 243). The N-terminal half of Tim23 is hydrophilic whereas the C-terminal half is rather hydrophobic. The hydrophilic portion is exposed to the intermembrane space, while the hydrophobic portion is likely to span the inner membrane three or four times (249). Tim17 is also an integral membrane protein and shares sequence similarity with the hydrophobic portion of Tim23 (244, 248, 249). Therefore, Tim17 and Tim23 seem to have a similar orientation in the membrane, but Tim17 virtually lacks a hydrophilic intermembrane space domain. DNA sequences with homology to yeast Tim17 DNA have been found in data bases of human, Drosophila melanogaster, Caenorhabditis elegans, and Arabidopsis thaliana genomes. The human cdna was found to be able to encode Table 4 Components of translocase of the inner membrane of mitochondria Other Essential for Component names Function viability in yeast References Tim23 Mas6p Component of the + 240, 243, MIM23 import channel; dependent binding of presequence regulating channel Tim17 MIM17 Component of the + 244, 246, 248, 249 Sms1 import channel Tim44 Mpi1p Associates with import + 239, ISP45 channel; binds mt-hsp70 in ATP-dependent fashion; part of driving system for vectorial transport mt-hsp70 Ssc1p DnaK-type chaperone; + 253, 254, facilitates import, folding, assembly, degradation of proteins in the matrix; interacts with Tim44 MGE1 Mge1p Cochaperone: cooperates mgrpe with mt-hsp70 in most Yge1p if not all functions; nucleotide exchange factor in matrix

24 886 NEUPERT Figure 2 Schematic representation of components of the TIM complex. Abbreviations: IMS, intermembrane space; IM, inner membrane. a protein that can substitute for Tim17 in yeast (266), again suggesting that the import machineries are highly conserved during evolution. Tim44, in contrast, is a hydrophilic protein without a predicted membranespanning segment (239). Tim44 was also identified in a biochemical approach. Components exposed on the outer surface of the inner membrane were analyzed by the inhibitory effect of a collection of antibodies on protein import into mitoplasts and into right-side-out inner membrane vesicles. A 45-kDa protein was purified that could be cross-linked to preproteins in transit across the inner membrane; by sequence analysis, it turned out to be identical to Tim44 (250, 251, 254), although the equating of ISP45 with Tim44 was questioned on the basis of biochemical properties (252). Tim44 is associated with the inner membrane, although the precise nature of Tim44 association with the inner membrane is still not clear; it can be easily extracted from membranes at alkaline ph, suggesting that it is not an integral membrane protein ( ). A C-terminal segment was reported to be exposed to the intermembrane space (239, 250). On the other hand, Tim44 was found to be released from the membrane by treatment with moderate salt concentrations or very low concentrations of digitonin. It was found to be inaccessible to proteases from the intermembrane space side (246). Further analysis of the interaction of Tim44 with the inner membrane is clearly needed, but there is agreement that at least the majority of the protein is exposed on the matrix side of the inner membrane. Tim23 and Tim17 form a complex of about 1:1 stoichiometry (246, 267). This complex is stable in mild detergents such as digitonin but not in octyl glucoside or Triton X-100. Tim44 is associated with the Tim23-Tim17 couple, but this association is particularly labile and salt-sensitive. Tim44 can be extracted from