Atp10p assists assembly of Atp6p into the F0 unit of the yeast mitochondrial ATPase.

The F(0)F(1)-ATPase complex of yeast mitochondria contains three mitochondrial and at least 17 nuclear gene products. The coordinate assembly of mitochondrial and cytosolic translation products relies on chaperones and specific factors that stabilize the pools of some unassembled subunits. Atp10p was identified as a mitochondrial inner membrane component necessary for the biogenesis of the hydrophobic F(0) sector of the ATPase. Here we show that, following its synthesis on mitochondrial ribosomes, subunit 6 of the ATPase (Atp6p) can be cross-linked to Atp10p. This interaction is required for the integration of Atp6p into a partially assembled subcomplex of the ATPase. Pulse labeling and chase of mitochondrial translation products in vivo indicate that Atp6p is less stable and more rapidly degraded in an atp10 null mutant than in wild type. Based on these observations, we propose Atp10p to be an Atp6p-specific chaperone that facilitates the incorporation of Atp6p into an intermediate subcomplex of ATPase subunits.

The bulk of cellular ATP of eukaryote cells is synthesized by the F 0 F 1 -ATPase complex of the mitochondrial inner membrane. The ATPase is composed of two functionally and physically coupled parts, i.e. the membrane-embedded F 0 sector to which the hydrophilic F 1 sector is attached from the matrix side. In the yeast Saccharomyces cerevisiae, the F 0 sector consists of eight subunits (for reviews, see Refs. [1][2][3]. An oligomer of Atp9p forms a ring-like structure that rotates in the membrane bilayer and, together with Atp6p, drives proton transfer across the inner membrane. Six other subunits, namely Atp8p, Atp4p, and the oligomycin sensitivity-conferring protein (OSCP) and its subunits d, f, and h, form the stator arm connecting F 0 to F 1 . The ␥, ␦ and ⑀ subunits of F 1 are part of a central stalk linking it to the Atp9p ring. ATP synthesis depends on a rotation of this stalk within the static catalytic F 1 hexamer made up of three ␣ and three ␤ subunits. Seven other proteins, stably associated with the F 0 F 1 -ATPase, function in the regulation or oligomerization of the yeast complex (1,4,5).
Three components of the yeast F 0 sector, Atp6p, Atp8p, and Atp9p, are encoded in the mitochondrial genome, whereas the other subunits are products of nuclear genes whose synthesis occurs in the cytosol. Current ideas of how the ATPase complex is assembled are based on the analysis of mutants and on what is known about the subunit topology of the complex (1, 6 -9). A model compatible with such information is depicted in Fig. 1. After synthesis on mitochondrial ribosomes, Atp9p oligomerizes to form a ring-like homo-oligomer composed of 10 -12 subunits (3). The F 1 sector assembles independently of the F 0 subunits and attaches to the Atp9p ring (10,11). This complex was reported to be further modified by the addition of Atp8p first (6), followed by Atp4p (8) and perhaps other components of the stator arm. Finally, assembly of the Atp6p subunit completes the process to yield a functional ATPase complex (6,12). The addition of Atp6p at a late stage of assembly may prevent the exchange of protons across the membrane through the partially assembled F 0 sector. This scheme is still very fragmentary and is lacking a host of intermolecular interactions of other F 0 subunits needed to stabilize the final complex (1,2,13).
Biogenesis of F 1 requires the general chaperone Hsp60p (14) and at least three factors, Atp11p, Atp12p, and Fmc1p (15)(16)(17). Atp11p and Atp12p function as molecular chaperones that probably stabilize unassembled ␣ and ␤ subunits by shielding their hydrophobic surfaces (16,18). The function of Fmc1p is unclear at present. Much less is known about assembly of F 0 . The nuclear gene ATP10 was identified in a genetic screen to be essential for the proper assembly of F 0 (19). In the absence of Atp10p, the mitochondrial encoded ATPase subunits are synthesized but not assembled into F 0 , because the F 1 -ATPase is only loosely attached to the membrane. Because Atp10p is not a subunit of the ATPase complex (19) and is not involved in post-translational processing or modification of Atp6p (20), it was proposed to be an Atp6p-specific chaperone (20). The ability of a mutation near the carboxyl terminus of Atp6p to suppress an atp10 null mutant (20) points at a role of Atp10p in folding, assembly, or maintenance of Atp6p. The matrix chaperone Hsp70p has also been shown to be necessary for biogenesis of the F 0 sector (21). This general chaperone binds to and facilitates the association of Atp6p with a partially assembled intermediate of F 0 containing the Atp9p oligomer.
To analyze the function of Atp10p in more detail, we studied the assembly of the ATPase complex following radiolabeling of the mitochondrially encoded subunits in isolated wild type and mutant mitochondria. Based on the results reported here and those of earlier studies, Atp10p is proposed to function as a chaperone that interacts physically with newly synthesized Atp6p and maintains it in an assembly-competent state. in ATP6) have been described (16,20). Yeast cultures were grown at 30°C in YP (1% yeast extract and 2% peptone) medium supplemented with 2% galactose or 0.1% KOH-buffered lactate (22). Mitochondria were isolated as described previously (22), except that the cultures were adjusted to 2 mg/ml chloramphenicol for 2 h prior to harvesting of the cells.

Yeast Strains and Growth
Labeling of Mitochondrial Translation Products-For in vivo labeling of mitochondrial translation products, cells were grown in YPGal medium (2% galactose, 1% yeast extract, 2% peptone). Cycloheximide (150 g/ml) was added to block cytosolic protein synthesis. After the addition of [ 35 S]methionine, cells were incubated with agitation for 15 min. Reactions were stopped by the addition of 40 mM puromycin and 25 mM cold methionine and incubation for 10 min. After different time periods, aliquots were removed. Proteins were extracted, precipitated by trichloroacetic acid, resolved by SDS-PAGE, and visualized by autoradiography.
Mitochondrial translation products were labeled in vitro as described previously (22,23). Mitochondria (40 g protein) were incubated in translation buffer (0.6 M sorbitol, 150 mM KCl, 15 mM KH 2 PO 4 , 13 mM MgSO 4 , 0.15 mg/ml all amino acids except methionine, 4 mM ATP, 0.5 mM GTP, 5 mM ␣-ketoglutarate, 5 mM phosphoenolpyruvate, 3 mg/ml fatty acid-free bovine serum albumin, and 20 mM Tris/HCl, pH 7.4) containing 0.6 units of pyruvate kinase and 10 Ci of [ 35 S]methionine. Typically, samples were incubated for 30 min at 30°C, and labeling was stopped by the addition of 25 mM cold methionine. The samples were further incubated for 5 min to complete synthesis, and mitochondria were isolated by centrifugation, washed in 1 ml 0.6 M sorbitol, 20 mM Hepes/HCl, pH 7.4, and lysed in 25 l of LiDS sample buffer (2% lithium dodecyl sulfate, 10% glycerol, 2.5% ␤-mercaptoethanol, 0.02% bromphenol blue, and 60 mM Tris/HCl, pH 6.8). Samples were shaken at 4°C for 10 min prior to loading on the gel. Heating was avoided to minimize dissociation of the ATPase oligomers.
Cross-linking and Immunoprecipitation-Cross-linking and immunoprecipitation were carried out essentially as described (24). Mitochondrial gene products were radiolabeled in isolated mitochondria in a translation buffer with bovine serum albumin omitted and Tris/HCl replaced by Hepes/KOH to prevent quenching of the amino groupspecific cross-linking reagents. Stock solutions (24 mM) of the crosslinking reagents dithiobis(succinimidyl propionate) (DSP) 1 and disuccinimidyl suberate (DSS) were freshly prepared in dimethyl sulfoxide and diluted 1:100 into the labeling reaction. An equal concentration of dimethyl sulfoxide was added to mock-treated samples. Following cross-linking, 25 mM glycine and cold methionine were added to quench unreacted cross-linkers.
For immunoprecipitation, mitochondria were lysed in 25 l of 1% SDS and 50 mM Tris-HCl, pH 7.5, cleared by centrifugation for 10 min at 16,000 ϫ g, and diluted in 40 volumes of 0.1% Triton X-100, 300 mM KCl, 5 mM EDTA, and 10 mM Tris/HCl, pH 7.4. For co-immunoprecipitation reactions, the SDS buffer was omitted, and the mitochondria were lysed directly in Triton buffer. After a clarifying spin for 10 min at 16,000 ϫ g at 4°C, the extract was added to 2 mg of protein A-Sepharose (Amersham Biosciences) and 4 -6 l of rabbit antisera. After incubation of 1 h at 4°C, the beads were washed twice in lysis buffer and once in 20 mM Tris/HCl, pH 7.4. Precipitated proteins were dissolved in LiDS sample buffer, separated by SDS-PAGE, and exposed to x-ray film.

Atp10p Is in Direct Contact with Newly Synthesized Atp6p-
Previous evidence for a genetic interaction of Atp10p and Atp6p (20) suggested that the two proteins may also interact physically. To detect a complex of the two proteins, mitochondrial translation products were labeled with [ 35 S]methionine in the presence of the non-cleavable cross-linker DSS. The labeled mitochondria were solubilized, and the extract was treated with antibody against Atp10p in the presence of protein A-Sepharose beads. Autoradiography of the proteins separated by SDS-PAGE indicated background signals of all the translation products, irrespective of the presence or absence of the crosslinker. This unspecific adsorption to Sepharose is due to the highly hydrophobic nature of the polypeptides. A band of an apparent size of 52 kDa was enriched by Atp10p antibodies (Fig. 2, asterisk). This band only occurred in the samples treated with the cross-linker and was absent in the ⌬atp10 mutant and the ⌬atp10 revertant (W303⌬ATP10/R1) (20). This indicated that one of the radiolabeled translation products is in the proximity of Atp10p during or directly following its synthesis in the mitochondria. The apparent size of the novel band (52 kDa) is consistent with it being an adduct of Atp10p (30 kDa) and Atp6p (22 kDa). Atp6p has a mass of ϳ28 kDa, but in the SDS-PAGE system used in this study it migrates as a protein of 22 kDa.
To confirm that Atp6p interacts with Atp10p, mitochondria were pulsed with [ 35 S]methionine in the presence of DSP. Proteins cross-linked through DSP resist cleavage during lysis of mitochondria and immunoprecipitation but are cleaved under the reducing conditions of the sample buffer used in SDS-PAGE. Following labeling of wild type mitochondria in the presence of DSP, samples treated with the Atp10p-specific antibodies (but not preimmune serum) were enriched for Atp6p (Fig. 3A, asterisk). Despite its low efficiency, cross-linking was specific, as there was no enrichment of Atp6p in mitochondria of the mutant or revertant (Fig. 3A, lower panel). Poor crosslinking is often observed with membrane proteins because of their relatively low content of charged amino acid residues.
The results obtained with DSP are consistent with the notion that the 52-kDa cross-linking product consists of an Atp6p/ 1 The abbreviations used are: DSP, dithiobis(succinimidyl propionate); DSS, disuccinimidyl suberate; LiDS, lithium dodecyl sulfate. Atp10p complex and indicate that Atp10p is in the proximity of newly synthesized Atp6p. The Atp6p released from the adduct appeared to migrate slightly faster than the Atp6p that had not been cross-linked (Fig. 3A, lanes marked T, loaded with total mitochondrial proteins). This mobility shift is caused by the immunoprecipitation procedure, because it is also seen when Atp6p is immunoprecipitated with Atp6p-specific antibodies from mitochondrial extracts that were not treated with a crosslinker (Fig. 3B).
The Interaction of Atp10p with Atp6p Can Occur Post-translationally-Proteins involved in membrane insertion of mito-chondrial translation products typically interact with their substrates co-translationally. This was shown to be the case for the Oxa1p translocase, which promotes insertion of the mitochondrially translated subunits of cytochrome oxidase into the inner membrane (25,26). On the other hand, factors that coordinate oligomerization of completed and folded subunit polypeptides are more likely to bind to their substrates posttranslationally. To assess if the Atp10p/Atp6p complex is formed co-or post-translationally, DSP was added to mitochondria either during or following the labeling reaction. No significant difference was noted in the amount of Atp6p associated with Atp10p under the two conditions (Fig. 4). This indicates a post-translational interaction of Atp10p with Atp6p but does not exclude that it can also occur co-translationally.
Assembly of the ATPase Complex in Isolated Wild Type Mitochondria-Assembly of the ATPase complex was reported to occur in a sequential and obligatory order (6 -8) as depicted in Fig. 1. To gain better insights into the role of Atp10p in this process, we studied the assembly of the ATPase complex in isolated mitochondria. Assembly in this in vitro system was enhanced in cells allowed to grow in the presence of chloramphenicol for 2 h prior to isolation of the mitochondria. Under these conditions, mitochondrial translation is blocked, leading to increased pools of nuclear gene products available for interaction with the newly synthesized subunits made in organello.
In agreement with earlier findings (21), all eight mitochondrially encoded polypeptides and two slower migrating bands with apparent masses of 48-and 54-kDa are detected in the autoradiogram of the labeled mitochondria (Fig. 5A, lane 1). The 48-kDa and 54-kDa products represent oligomers of F 0 subunits that are incompletely depolymerized by dodecyl sulfate (21) but are converted to the monomeric constituents when proteins are treated with trichloroacetic acid prior to SDS-PAGE (Fig. 5A, lane 2). Both oligomers are associated with the F 1 -ATPase, as they were co-immunoprecipitated by antibodies against the F 1 ␣ subunit (Fig. 5A, lane 3). The enrichment of the 54-kDa oligomer suggests that this form is more efficiently or more stably attached to the F 1 sector recognized by the F 1 ␣antiserum. An antiserum against the Atp4p subunit also coimmunoprecipitated the three mitochondrially synthesized A, mitochondria of wild type W303-1A (WT) and the atp10 null mutant aW303⌬ATP10 (⌬ATP10) were radiolabeled with [ 35 S]methionine for 30 min at 25°C in the presence or absence of 0.24 mM DSP. The cleavable cross-linker DSP was added during the labeling period. One aliquot of the solubilized mitochondria, equivalent to 10% of the total labeling reaction, was separated by SDS-PAGE without further treatment (T). The remaining sample was subjected to immunoprecipitation with either pre-immune serum (Ab Atp10p Ϫ) or antiserum against Atp10p (Ab Atp10p ϩ). The lanes showing the total labeled products (T) were exposed overnight, whereas those showing the products adsorbed on the beads were exposed for 4 days. Asterisk indicates the Atp6p protein that was immunoprecipitated with Atp10p-specific antibodies following cross-linking. B, wild type mitochondria were labeled as in section A without the cross-linker. They were subjected to immunoprecipitations with either pre-immune serum (Ab Atp6p Ϫ) or antiserum against subunit 6 of the ATPase (Ab Atp6p ϩ). The mitochondrially translated proteins are identified in the margin as follows: Var1, a subunit of mitochondrial ribosomes; Cox1, Cox2, and Cox3 are subunits of cytochrome oxidase; Atp6, Atp8, and Atp9 are subunits of the ATPase; Cyt b is a subunit of the bc 1 complex.

FIG. 4. Atp6p interacts with Atp10p post-translationally.
Translation products were radiolabeled in wild type mitochondria for 30 min at 25°C. In the lanes marked 0 min, DSP (0.24 mM) was present during the 30-min labeling period. Labeling was stopped by addition of excess unlabeled methionine. In the lanes marked 30 min, DSP was added after 30 min of labeling and the addition of unlabeled methionine. Cross-linking was allowed to proceed for 15 min. Translation products associated with Atp10p were enriched by immunoprecipitation and analyzed by SDS-PAGE as in Fig. 2. Asterisk indicates the immunoprecipitated Atp6p. The lanes showing the total mitochondrial translation products (T) were exposed to x-ray film overnight; those showing the radioactive proteins adsorbed on the protein A-Sepharose beads were exposed for 2 weeks.
ATPase subunits. In this instance, however, more Atp6p was co-precipitated because of its direct interaction to Atp4p (2). This indicates that, following synthesis in isolated mitochondria, Atp6p, Atp9p, and most likely also Atp8p assemble, at least in part, into complexes that contain subunits of the stator arm and F 1 .
The composition of the 48-and 54-kDa oligomers was examined by separating labeled wild type mitochondria by SDS-PAGE in one dimension, partially dissociating the oligomers in a strip of the gel with trichloroacetic acid, and separating the proteins in the acid treated strip by SDS-PAGE in a second dimension. The trichloroacetic acid treatment caused the release from the 48-kDa complex of a radioactive protein that migrated like Atp9p, as well as the release from the 54-kDa complex of two proteins, one corresponding to Atp9p and the other to Atp6p (Fig. 5B). Nuclear encoded ATPase subunits that might be present in the two oligomers are not labeled and, therefore, cannot be detected by this method. The presence of Atp6p in the 54-kDa oligomer is supported by Western blot analysis of mitochondrial proteins probed with an Atp6p-specific antiserum, which reacted with the 22-kDa Atp6p monomer and the 54-kDa oligomer (Fig. 5C). These data indicate that isolated mitochondria are capable of assembling two different subcomplexes of the ATPase, i.e. a 48-kDa complex lacking Atp6p and one of 54 kDa that contains Atp6p.
Atp10p Is Required to Form the 54-kDa Complex-Impaired biogenesis of the F 0 sector in the ⌬atp10 mutant was shown previously to markedly lower the steady-state levels of Atp6p (20). This was confirmed with an antibody against Atp6p, which failed to detect the monomeric or assembled form of Atp6p in the mutant (Fig. 6A). The F 0 defect in ⌬atp10 mitochondria was studied in more detail by examining the oligomerization state of newly synthesized Atp6p. Pulse labeling of mitochondria isolated from the ⌬atp10 mutant showed that all three of the mitochondrially encoded subunits of the ATPase were synthesized at ϳ50% of the rate in wild type (Fig. 6B). Furthermore, whereas the 48-kDa oligomer was also present in the mutant, the 54-kDa oligomer was absent (indicated by the top arrow in Fig. 6B). Thus, although the Atp9p oligomer assembles independently of Atp10p, formation of the larger complex containing Atp6p is blocked in the absence of Atp10p.
Earlier evidence indicated that the 48-kDa but not the 54- The gel slice of the labeled proteins was cut out and incubated in 12% trichloroacetic acid to partially dissociate the ATPase complexes. The pH of the gel slice was adjusted to 6.8 and positioned on top of a new SDS gel for separation in a second dimension. The 48-kDa complex (48K) partially dissociated into monomeric Atp9p, and the 54-kDa complex partially dissociated into Atp9p and Atp6p. Positions of the molecular mass standards for both dimensions are indicated. C, Atp6p is part of the 54-kDa complex. Mitochondria were lysed in LiDS sample buffer and subjected to SDS-PAGE. Proteins were transferred to nitrocellulose, and the monomeric and oligomeric forms of Atp6p were detected with an Atp6pspecific antiserum.

FIG. 6. Atp10p is required for assembly of the 54-kDa oligomer.
A, Atp10p is required for stable accumulation of Atp6p in mitochondria. Mitochondria (100 g) from the wild type strain W303-1A (WT) or the atp10 mutant aW303⌬ATP10 (⌬ATP10) were separated by SDS-PAGE, transferred to nitrocellulose, and probed with an Atp6p-specific antibody. B, translation products were radiolabeled mitochondria isolated from the wild type and the atp10 null mutant and were analyzed as described in the legend for Fig. 5B. C, mitochondrial translation products labeled in organello with [ 35 S]methionine were co-immunoprecipitated with antibodies against the mitochondrial Hsp70p chaperone and pre-immune serum (p.i.). The samples in the lanes labeled TCA ϩ were precipitated with trichloroacetic acid to dissociate the oligomeric ATPase complexes before depolymerization in sample buffer and separation by SDS-PAGE. D, translation products were labeled in wild type (WT) and the atp10 null mutant (⌬ATP10) mitochondria for 10 or 30 min and subjected to co-immunoprecipitation with Hsp70p antibodies. The amount of immunoprecipitated Atp9p was quantified by densitometry and normalized to the amount of the total Atp9p synthesized. kDa oligomer interacts with mitochondrial Hsp70p (21). This was confirmed by immunoprecipitation of mitochondrial translation products (Fig. 6C). Treatment of the isolated 48-kDa oligomer with acid caused it to collapse to monomeric Atp9p (Fig. 6C, see bands labeled TCA ϩ). This confirms the presence of Atp9p, but not of Atp6p or Atp8p, in the 48-kDa oligomer. Atp10p was not required for the interaction of Hsp70p with the Atp9p oligomer, as similar amounts of Atp9p could be precipitated with Hsp70p antibodies in wild type and ⌬atp10 mitochondria (Fig. 6D). Thus, Atp10p and Hsp70p appear to play distinct roles and interact with different pools of unassembled ATPase subunits, i.e. Atp10p with Atp6p and Hsp70p with a partially assembled complex containing the Atp9p oligomer.
Atp10p Partially Protects Newly Synthesized Atp6p against Proteolysis-In organello incorporation of [ 35 S]methionine into Atp6p is reduced in the ⌬atp10 mutant (Figs. 5A and 6B). This could be due to reduced synthesis or increased degradation of the protein by proteases. To distinguish between these possibilities, mitochondrial translation products were radiolabeled in yeast cells in the presence of cycloheximide for 15 min. Translation was stopped by the addition of puromycin, and the stability of the synthesized proteins was analyzed after various periods of time (Fig. 7A). No significant difference was found in the turnover of newly synthesized Atp8/9p in the wild type and the mutant during the 2 h period of chase. In contrast, turnover of Atp6p was higher in the mutant than in the wild type (Fig.   7, A and B). Atp10p could be protecting Atp6p by directly shielding it from degrading proteases or by promoting the assembly of Atp6p into a protease-resistant complex.

DISCUSSION
Formation of hetero-oligomeric membrane complexes depends on the assistance of specific chaperones that interact with their cognate subunits. This may be necessary in part because of the hydrophobic nature of unassembled subunits, which may have to be stabilized by accessory factors to allow them to progress though the assembly pathway and minimize non-productive side reactions. The results reported here, together with those reported earlier, indicate that Atp10p has such a chaperone function in assembly of Atp6p, as evidenced by the following four observations. 1) Atp10p can be crosslinked to newly synthesized Atp6p. This interaction occurs post-translationally and appears to be transient, as Atp10p is not a subunit of the functional ATPase complex. 2) Atp10p is required for proper assembly of Atp6p into the F 0 sector of the ATPase. A role of Atp10p for proteolytic maturation or posttranslational modification of Atp6p was excluded (20). Binding of Atp10p to Atp6p is more likely to be necessary to maintain Atp6p in an assembly-competent state. 3) Atp6p is more prone to proteolytic degradation in the ⌬atp10 mutant, indicating that it is directly or indirectly protected by Atp10p. 4) A point mutation in Atp6p can suppress ⌬atp10 mutants (20). The increased stability of Atp6p in the suppressed strain may reflect a shift of the equilibrium from degradation toward assembly of Atp6p.
Based on these observations, we propose that following synthesis on mitochondrial ribosomes, Atp6p is bound to Atp10p. This interaction prevents degradation and/or aggregation of Atp6p and channels it to associate with a preformed subcomplex of the ATPase (Fig. 8). The integration of Atp6p into this intermediate complex may be dependent directly on Atp10p. Alternatively, Atp10p may shift the equilibrium from degradation/aggregation to assembly. The latter seems more likely in view of the observation that ⌬atp10 mutants are leaky and  display slow growth on non-fermentable carbon sources (20). In the ⌬atp10 revertant the equilibrium might be shifted further toward assembly as a result of the increased stability of Atp6p. It is also possible, however, that the suppressor mutation helps to increase the efficiency with which Atp6p is assembled, thereby sparing a larger fraction of the protein from degradation. Assuming that Atp6p is one of the last subunits to be incorporated into the complex as proposed previously (12), the intermediate complex with which Atp6p assembles would be expected to contain most or all of the other ATPase subunits. Two oligomeric complexes are detected in mitochondria that are allowed to synthesize the three endogenously encoded subunits of the ATPase. The 48-kDa oligomer observed in wild type mitochondria is also detected in mitochondria lacking Atp10p. In the absence of Hsp70p function, the 48-kDa subcomplex loses its competence to interact with Atp6p, and assembly of the final ATPase structure is arrested (21). The interaction with the Hsp70p chaperone may stabilize this intermediate for further progression in the assembly pathway. Thus, the final step of the pathway appears to rely on two chaperones, namely the classical chaperone Hsp70p, which binds to the 48-kDa oligomeric intermediate, and Atp10p, which interacts with newly synthesized Atp6p and promotes its incorporation into a late assembly intermediate.
Members of the Hsp70p and Hsp60p families of chaperones participate in numerous cellular processes by facilitating folding and assembly of a large number of different substrates. In contrast, some reactions employ factors that have a single substrate polypeptide as their target. Although several such specific assembly factors have been reported in recent years, little is known about the mechanisms by which they work. One well studied example, Ump1p, is a specific factor that associates with and helps to complete the assembly of proteasome intermediates (27). Ump1p tightly regulates proteasome biogenesis and is removed at the final assembly step. The interaction with Ump1p is thought to prevent the harmful effects of proteolytically active but incompletely assembled proteasome intermediates. Despite the differences in the physiological role of the proteasome and the ATPase, subunit-specific chaperones like Atp10p may provide similar protective functions. The partially assembled F 0 sector of the ATPase could present potential problems for the cell because of its ability to dissipate the membrane potential. This might explain why the proton-conducting Atp6p subunit is integrated into the complex at a late assembly step (12). Atp10p might control this critical step by ensuring the maintenance of a pool of assembly-competent Atp6p subunits. Whereas the cellular levels of general chaperones that are not occupied by substrates vary greatly with different temperatures and physiological conditions, the pool of specific chaperones can be precisely controlled and adapted to the needs of a certain biological process. An important focus of future work will be to understand the regulation of substratespecific chaperones, such as Atp10p, and to clarify the molecular mechanisms by which they bind and release their substrate proteins.