The assembly factor Atp11p binds to the β-subunit of the mitochondrial F1-ATPase

Atp11p is a protein of Saccharomyces cerevisiae required for the assembly of the F1component of the mitochondrial F1F0-ATP synthase. This study presents evidence that Atp11p binds selectively to the β-subunit of F1. Under conditions in which avidin-Sepharose beads specifically adsorbed biotinylated Atp11p from yeast mitochondrial extracts, the F1 β-subunit coprecipitated with the tagged Atp11p protein. Binding interactions between Atp11p and the entire β-subunit of F1 or fragments of the β-subunit were also revealed by a yeast two-hybrid screen: Atp11p bound to a region of the nucleotide-binding domain of the β-subunit located between Gly114 and Leu318. Certain elements of this sequence that would be accessible to Atp11p in the free β-subunit make contact with adjacent α-subunits in the assembled enzyme. This observation suggests that the α-subunits may exchange for bound Atp11p during the process of F1 assembly.

Atp11p is a protein of Saccharomyces cerevisiae required for the assembly of the F 1 component of the mitochondrial F 1 F 0 -ATP synthase. This study presents evidence that Atp11p binds selectively to the ␤-subunit of F 1 . Under conditions in which avidin-Sepharose beads specifically adsorbed biotinylated Atp11p from yeast mitochondrial extracts, the F 1 ␤-subunit coprecipitated with the tagged Atp11p protein. Binding interactions between Atp11p and the entire ␤-subunit of F 1 or fragments of the ␤-subunit were also revealed by a yeast two-hybrid screen: Atp11p bound to a region of the nucleotide-binding domain of the ␤-subunit located between Gly 114 and Leu 318 . Certain elements of this sequence that would be accessible to Atp11p in the free ␤-subunit make contact with adjacent ␣-subunits in the assembled enzyme. This observation suggests that the ␣-subunits may exchange for bound Atp11p during the process of F 1 assembly.
The ATP synthase of mitochondria, chloroplasts, and bacteria catalyzes ATP synthesis during respiration (1,2). The enzyme is a large protein complex composed of two oligomeric units: an integral membrane component (F 0 ) and a peripherally bound catalytic moiety (F 1 ). F 1 that is bound to the membrane sector catalyzes both ATP synthesis and ATP hydrolysis in reactions that are coupled to proton translocation through F 0 (1, 2). F 1 contains five different types of subunits in the stoichiometric ratio ␣ 3 :␤ 3 :␥:␦:⑀ (1, 2). X-ray diffraction studies have revealed the three-dimensional structure of mitochondrial F 1 from bovine heart (3) and from rat liver (4). The ␣and ␤-subunits alternate in a hexamer that surrounds a central helical structure composed of the N and C termini of the ␥-subunit (3,4). Three catalytic sites and three non-catalytic sites are located at the interfaces between ␣and ␤-subunits (3,4).
Previous work with respiratory-deficient strains of Saccharomyces cerevisiae has shown that the F 1 ␣-subunit aggregates in mitochondria of yeast lacking the ␤-subunit; likewise, the ␤-subunit aggregates in ␣-subunit null strains (5). In contrast, the ␣and ␤-subunits can be recovered in soluble fractions when mitochondria are prepared from strains harboring a disrupted gene for the ␥-subunit (6), ␦-subunit (7), or ⑀-subunit (8), despite the fact that F 1 does not assemble in the absence of any of these three subunits. Moreover, in a strain disrupted for the ␥-subunit, the ␣and ␤-subunits show evidence of forming hetero-oligomers (6). Hence, it would appear that if ␣␤ complexes can form, the proteins remain soluble. If, on the other hand, only one of the "partner" subunits is present (e.g. in an ␣or ␤-subunit null strain), the natural tendency of free ␣or free ␤-protein to aggregate becomes apparent. Notably, the ␣and ␤-subunits of F 1 also aggregate in mitochondria of yeast carrying a nonfunctional allele of the ATP11 or ATP12 gene (5). The fact that the biochemical properties of these mutant strains are similar to those of ␣or ␤-subunit null strains suggests that the products of the ATP11 (9) and ATP12 (10) genes are required to maintain the unassembled ␣and ␤-subunits as soluble proteins during the early phase of enzyme assembly. As might be expected for proteins that bind free F 1 subunits, which constitute a very small fraction of the total pool of F 1 protein (11), Atp11p and Atp12p are present at a very low level in yeast mitochondria (9,10). Based on the facts that Atp11p and Atp12p appear to function exclusively in F 1 assembly (5) and do not share sequence homology with other proteins of known function, they can be considered as "F 1 -specific" assembly factors.
Atp11p is a 31-kDa monomeric protein of the mitochondrial matrix with an overall basic charge (12). Previous work has localized the functional domain of Atp11p to a region in the middle of the protein that is characterized by two stretches of hydrophobic sequence (13). Here we report results from coprecipitation experiments and yeast two-hybrid screens that probed for direct interactions between Atp11p and the F 1 ␣and ␤-subunits and conclude that Atp11p binds to the ␤-subunit only. The binding site for Atp11p was mapped within a sequence of 205 amino acids in the nucleotide-binding domain of the ␤-subunit.
Plasmid Constructions-The plasmids used in this study are described in Table I. Atp11p is numbered from residues 1 to 318 (9); the F 1 ␤-subunit is numbered from residues 1 to 511 (15); and the F 1 ␣-subunit is numbered from residues 1 to 544 (16). In all cases, residue 1 is the initiator methionine in the primary translation products. The * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Coprecipitation Experiments-Mitochondria were prepared as described (13) from yeast strains aW303⌬G13/pG13/ST5 and aW303⌬G13/pG13/BTYEP, which produce native and biotin-tagged forms of Atp11p, respectively, from the 2 vector YEp352. The mitochondria were suspended at 7 mg/ml (300-l final volume) in TEA buffer (10 mM Tris-HCl, pH 8.0, 2 mM EDTA, and 4 mM ATP) supplemented with protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 1 g/ml pepstatin) and subjected to four 10-s bursts of sonic irradiation at full power (Branson Model 450 sonifier). The sonicated material was centrifuged at 50,000 rpm in a Beckman Ti-70.1 rotor for 30 min at 4°C to sediment the membranes. The soluble extracts (130 l) were mixed with 50 l of avidin-Sepharose beads and brought to a 200-l final volume in either buffer A (TEA buffer supplemented with 1% Triton X-100 and 0.14 M NaCl) or buffer B (TEA buffer supplemented with 1% Triton X-100, 0.14 M NaCl, and 2 mM biotin). The samples were rotated end-over-end at 4°C for 30 min and subsequently centrifuged for 5 min (Beckman Microfuge) to sediment the avidin-Sepharose beads. The beads were washed three times by centrifugation with 400 l of buffer A or buffer B (as appropriate), followed by an additional three washes with 400 l of plain TEA buffer. The washed beads were suspended in 40 l of Laemmli sample buffer in preparation for Western analysis. The proteins present in the initial post-bead supernatants were precipitated with 10% trichloroacetic acid and suspended in 80 l of Laemmli sample buffer.
Yeast Two-hybrid Screen-The yeast two-hybrid screen (19) employed yeast vectors pACT2 and pAS2-1 and host strain Y190 (described above) as supplied in the MATCHMAKER Two-Hybrid System 2 from CLONTECH. Yeast were grown in SD/ϪLeu,Trp,His medium. Assays for ␤-galactosidase activity followed protocols described in the CLONTECH manual. Qualitative assessment of expression from the lacZ reporter gene was made using X-gal as a chromogenic substrate for ␤-galactosidase in a colony-lift filter assay. For quantitative determination of ␤-galactosidase activity, the absorbance at 420 nm of onitrophenyl released from the substrate o-nitrophenyl ␤-galactoside was normalized to cell density measured at 600 nm and is reported in Miller units (20).
Miscellaneous Methods-Standard techniques were used for restriction endonuclease analysis of DNA, purification and ligation of DNA fragments, and transformations and recovery of plasmid DNA from E. coli (21). Yeast transformations employed the LiAc procedure (22). The method of Laemmli (23) was used for SDS-polyacrylamide gel electrophoresis. Western blotting followed the procedure of Schmidt et al. (24). Antibodies against Atp11p, F 1 ␣, and F 1 ␤ were used at dilutions of 1:1000, 1:2000, and 1:3000, respectively (5,12). Monoclonal antibody against the transcriptional activation domain of Gal4p was purchased from CLONTECH and used at 0.4 g/ml. Visualization of the protein bands in x-ray film was by chemiluminescence using the ECL system from Amersham Pharmacia Biotech. Avidin-Sepharose resin was prepared by coupling native avidin (Sigma) to CNBr-activated Sepharose following the protocol supplied by the manufacturer of the resin beads (Amersham Pharmacia Biotech). Protein concentrations were estimated by the method of Lowry et al. (25).

Affinity Precipitation of Biotinylated Atp11p (Bt-Atp11p)
with Avidin-Sepharose Beads-Yeast transformed with plasmid pG13/BTYEP produce a form of Atp11p that carries a sequence for in vivo biotinylation at the C terminus of the protein (12). The tagged protein (Bt-Atp11p) is fully functional in F 1 assembly (9, 12) and can be used in combination with avidin affinity resins in coprecipitation experiments aimed at identifying proteins that form complexes with Atp11p. For this purpose, soluble mitochondrial extracts, prepared from yeast cells that produce either native or biotinylated Atp11p, were incubated with avidin-Sepharose beads in the absence or presence of free biotin in the buffer (see "Experimental Procedures"). Western analysis of the post-bead supernatants (Fig.  1) showed that the F 1 ␣and ␤-subunits were present in approximately equal amounts and that native Atp11p was more abundant than biotinylated Atp11p. In the bead precipitate from mitochondrial extracts containing native Atp11p, only background levels of Atp11p were detected compared with the strong signal for this protein in the supernatant. Moreover, there was no evidence that F 1 ␣or ␤-subunits were precipitated by the beads. In contrast, the bead precipitate from mitochondrial extracts containing biotinylated Atp11p showed significant amounts of both Bt-Atp11p and the F 1 ␤-subunit. The specificity of this coprecipitation is attested by the fact that when precipitation was carried out in the presence of excess biotin, there was only a background level of Bt-Atp11p, and no F 1 ␤-protein was detected in the bead precipitate. The absence of the ␣-subunit in the Bt-Atp11p precipitate suggests that this F 1 protein either does not interact directly with Atp11p or forms only a very weak complex.
Identification of the ␤-Subunit Amino Acid Sequence That Binds Atp11p-The yeast two-hybrid assay was used to probe for binding interactions between Atp11p and the F 1 ␣and ␤-subunits. The plasmids employed in this work (Table I) encode only the mature sequences (i.e. without the mitochondrial leader sequences) of the Atp11p and F 1 proteins fused to either the DNA-binding domain or the transcriptional activation domain of Gal4p. Combinations of these plasmids were used to transform a yeast host (Y190) that harbors lacZ under transcriptional control of Gal4p; the presence of ␤-galactosidase activity in such strains indicates that the two Gal4p fusion proteins interact. Initial screens for ␤-galactosidase involved exposing lysed cells to X-gal on filter paper. The amount of ␤-galactosidase activity was quantified in cell lysates prepared from blue colonies, using o-nitrophenyl ␤-galactoside as the substrate, and normalized to the cell density.
The combination of plasmids producing Atp11p and the fulllength F 1 ␤-subunit (Ala 36 -Asn 511 ) gave a positive signal in the two-hybrid assay (Fig. 2). Negative control experiments showed FIG. 1. Western blots of mitochondrial extracts following affinity precipitation with avidin-Sepharose beads. Protein samples from yeast cells that produce native Atp11p are shown on the left; protein samples from yeast cells that produce biotinylated Atp11p are shown on the right. Aliquots of the initial post-bead supernatant (Sup) and avidin bead-precipitated samples (Bead ppt Ϯ biotin) were loaded on a 12% SDS gel, transferred to nitrocellulose, and probed with either a mixture of antibodies against the ␣and ␤-subunits or with Atp11p antiserum. To allow detection of even small quantities of F 1 ␣and ␤-subunits in the bead precipitates (*), a 4-fold excess of protein was loaded on the gel versus the amount used to visualize the same proteins in the supernatant fractions. Atp11p binds to the F 1 ␤-Subunit there was no ␤-galactosidase activity in yeast harboring the Atp11p or F 1 ␤-subunit plasmid alone or in combination with the appropriate partner vector lacking any insert. In similar experiments, we have not been able to find any evidence of binding between Atp11p and the F 1 ␣-subunit (data not shown). This result is in accord with the fact that the ␣-subunit was not coprecipitated with Bt-Atp11p by avidin-Sepharose beads (see above).
The two-hybrid system was used further to locate the binding site for Atp11p on the ␤-subunit. There are three domains in the mature ␤-subunit: a ␤-barrel at the N terminus, a central domain that contains the adenine nucleotide-binding site, and a helix bundle at the C-terminal end (3, 4) (see Fig. 2). The fragment Glu 101 -Asn 511 , which lacks almost the entire ␤-barrel, and the fragment Ala 36 -Asp 382 , which lacks the helix bundle, both scored positively for binding Atp11p in this assay. There was no evidence of Atp11p binding to the sequence for the ␤-barrel domain (Ala 38 -Ile 116 ) or the helix bundle domain (Asp 382 -Asn 511 ). Western analysis with antibodies against the Gal4p activation domain confirmed the presence of the noninteracting ␤-subunit fusion proteins in the yeast host (data not shown).
The results from initial mapping studies indicated that the binding determinants for Atp11p are contained entirely within the nucleotide-binding domain of the ␤-subunit. To define the boundaries of the Atp11p-binding site, fragments of the ␤-subunit, deleted for sequence from the amino-or carboxyl-terminal end of the adenine nucleotide-binding domain, were tested for binding Atp11p in the two-hybrid assay. This analysis disclosed positive results for the fragments His 211 -Asn 511 , Ala 273 -Asn 511 , Asp 289 -Asn 511 , Ala 36 -Leu 318 , Gln 38 -Ile 288 , Gln 38 -Ile 272 , Gln 38 -Gly 253 , Gly 114 -Ile 288 , Gly 114 -Ile 272 , Gly 114 -Gly 253 , and Ala 273 -Leu 318 . The largest segment of ␤-subunit residues that can be removed from the N or C terminus of the protein without disrupting binding interactions with Atp11p occurs proximal to Gly 114 and distal to Leu 318 , which suggests that the binding site for Atp11p is located between these 2 amino acids (Fig. 2, shaded box). The data also indicate that the structural elements recognized by Atp11p are distributed throughout this region since not all of the fragments that scored positive for binding overlap in sequence. Notably, there are a number of fragments that encompass portions of the sequence between Gly 114 and Leu 318 that did not show evidence for binding Atp11p. Western analysis confirmed that the Gal4p/␤-subunit fusion proteins, which scored negatively in the assay, were produced in the yeast host (data not shown). On this basis, we suggest that the fragments Gln 38 -Ala 210 , Gly 114 -Ala 210 , His 211 -Ile 288 , Ala 273 -Ile 288 , Asp 289 -Leu 318 , and Asp 153 -Ile 288 may not fold correctly, which could prevent their recognition by Atp11p. DISCUSSION We have used two different methods to detect binding interactions between Atp11p and the ␣and ␤-subunits of F 1 . First, avidin-Sepharose beads, which selectively bind biotinylated Atp11p in mitochondrial extracts, were shown to coprecipitate the F 1 ␤-subunit. Second, direct binding between Atp11p and the F 1 ␤-subunit was demonstrated by means of a yeast twohybrid screen. Neither the affinity precipitation assay nor the two-hybrid screen provided indications that Atp11p binds also to the F 1 ␣-subunit. Additional work with the yeast two-hybrid screen has mapped the Atp11p-binding site to a sequence of 205 amino acids (Gly 114 -Leu 318 ) located within the nucleotidebinding domain of the ␤-subunit (Fig. 2). It is of note that only certain fragments of this sequence bind the assembly factor. For example, binding was observed for the ␤-subunit sequence Gly 114 -Ile 288 , but not for its smaller fragments, Gly 114 -Ala 210 and His 211 -Ile 288 . Similarly, the sequence Ala 273 -Leu 318 scored positive for Atp11p binding, whereas the fragments Ala 273 -Ile 288 and Asp 289 -Leu 318 did not. The ␤-subunit fragments that did not show evidence of binding Atp11p were detected in the cell by Western analysis in amounts comparable to those of the other fragments that did show evidence for binding the assembly factor. Thus, it would appear that other factors such as a correct folding are required for recognition and binding by Atp11p. This observation, together with the fact that atp11 mutants accumulate only the mature ␤-subunit (5), whereas strains defective in mitochondrial protein folding (i.e. hsp60 (mif4) mutants) accumulate both the precursor and mature forms of the ␤-subunit (26), supports the view that Atp11p acts at a step downstream from Hsp60 and binds the folded form of the ␤-subunit.
It is conceivable that Atp11p prevents the aggregation of unassembled ␤-subunits by shielding sequence elements that would cause nonproductive ␤/␤ interactions. The availability of structural information for bovine F 1 from x-ray studies (3) allows us to identify by homology modeling candidates for the sequence elements in the free ␤-subunit that are most likely to be protected by the assembly factor. The sequence Gly 114 -Leu 318 of the yeast ␤-subunit, which we have shown to harbor binding determinants for Atp11p, is homologous to Ala 80 -Leu 285 of the bovine F 1 ␤-subunit (3). This region, which extends from ␤-strand 1 through the beginning of ␣-helix F in the nucleotide-binding domain of the bovine ␤-subunit, is colored in red in the ribbon diagram shown in Fig. 3A. Within this segment of the ␤-subunit, several amino acid side chains are involved in intersubunit contacts with the adjacent ␣-subunits. This feature is illustrated in Fig. 3B, which shows the secondary structural elements of the ␤ DP -subunit sequence between Ala 80 and Leu 285 rendered as a schematic in red, the C-␣ traces of the adjacent ␣ DP -and ␣ E -subunits in cyan, and the amino acids at the interfaces between these subunits as stick models colored in yellow (contribution from the segment Ala 80 -Leu 285 of the ␤-subunit) or blue (contribution from the adjacent ␣-subunits). The participation of the Atp11p-binding region in the formation of the contact surfaces between ␣and ␤-subunits suggests that during assembly of the oligomer, the ␣-subunits may bind to the ␤-subunit in exchange for bound Atp11p. In consideration of the high tendency of ␤-subunit monomers to aggregate (5), this type of action would ensure that the ␤-subunit is never present as a free protein in solution.
Recent findings provide evidence that a similar mechanism (binding of Atp12p to the ␣-subunit) is adopted to prevent aggregation of the ␣-subunit during normal assembly of the enzyme. 2 In this context, of particular interest is the observation that in the absence of Atp11p, also the ␣-subunit aggregates inside mitochondria (5). One possible explanation of this phenomenon is that whereas under normal conditions, the very small pool of unassembled ␣-subunit is maintained in solution via binding to Atp12p, under conditions in which the ␤-subunit aggregates (i.e. in atp11 mutants), a large amount of unassembled ␣-subunit is likely to accumulate in excess over the Atp12p protein. Thus, under these circumstances, also the ␣-subunit is expected to aggregate in the mitochondrial matrix.