Characterization of mutations in the beta subunit of the mitochondrial F1-ATPase that produce defects in enzyme catalysis and assembly.

The ATP2 gene, coding for the beta subunit of the mitochondrial F1-ATPase, was cloned from nine independent isolates of chemically mutagenized yeast. Seven different mutant alleles were identified. In one case the mutation occurs in the mitochondrial targeting sequence (M1I). The remaining six mutations map to the mature part of the beta subunit protein and alter amino acids that are conserved in the bovine heart mitochondrial and Escherichia coli beta subunit proteins. Biochemical analysis of the yeast atp2 mutants identified two different phenotypes. The G133D, P179L, and G227D mutations correlate with an assembly-defective phenotype that is characterized by the accumulation of the F1 alpha and beta subunits in large protein aggregates. Strains harboring the A192V, E222K, or R293K mutations assemble an F1 of normal size that is none-the-less catalytically inactive. The effect of the atp2 mutations was also analyzed in diploids formed by crossing the mutants to wild type yeast. Hybrid enzymes formed with beta subunits containing either the G133D, E222K, or R293K mutations are compromised for steady-state ATPase activity. The display of partial dominance confirms the importance of Gly133 for structural stability and of Glu222 and Arg293 for catalytic cooperativity.

The F 1 -F 0 ATP synthase (ATPase) of mitochondrial, chloroplast, and bacterial membranes catalyzes ATP synthesis coupled to respiration (1,2). The enzyme is composed of a catalytic moiety, F 1 , attached peripherally to an integral membrane component, F 0 . Three types of subunits (designated a, b, and c in bacteria (3,4)) constitute an evolutionarily conserved core F 0 structure, and there is variation among species as to whether or not additional F 0 subunits are present (5). F 1 shows a remarkable degree of structural conservation, and in most organisms studied, exists as an oligomer of five different subunits with the stoichiometry ␣ 3 ␤ 3 ␥␦⑀ (1, 2). The x-ray structure of bovine mitochondrial F 1 shows the ␣ and ␤ subunits alternating in a hexagonal ring, with the amino and carboxyl termini of the ␥ subunit located in the central core of the ring (6). Six adenine nucleotide binding sites are located at the ␣/␤ interfaces. Three of these sites are catalytic and reside primarily in the ␤ subunit; the other three sites are noncatalytic and are contributed mostly by residues of the ␣ subunits (6). The three catalytic sites interact cooperatively during enzyme turnover; this process is probably mediated by conformational changes in the F 1 subunits (1,2). A principal focus in F 1 research is to understand the mechanistic details that underlie catalytic cooperativity. The study of mutants with defects in the F 1 subunits offers a means to test the catalytic function of amino acids and to identify the residues that propagate conformational changes in the enzyme. Several excellent reviews are available that document studies performed with Escherichia coli mutants to probe the structure/ function relationships in F 1 (7)(8)(9). The mitochondrial enzyme from Saccharomyces cerevisiae is equally amenable to genetic manipulation. Moreover, the analysis of yeast F 1 mutants is of particular value now that an atomic structure of the mitochondrial enzyme is available (6).
Genetic studies with yeast F 1 have centered on the ␤ subunit, which is coded for by the ATP2 gene (10) in chromosome X (11). The work reported to date has used site-directed mutagenesis to effect changes in amino acids that were previously shown to react with chemical labels (12,13) and also to change residues that constitute a sequence element conserved in several nucleotide binding proteins (the P-loop) (13)(14)(15). The present study constitutes the first report in which mutations induced chemically in the yeast ␤ subunit gene were characterized both biochemically and genetically. Two classes of ␤ subunit mutations were identified: those that alter catalysis without affecting the enzyme structure and those that produce defects in F 1 assembly.
Preparation of Yeast Mitochondria-Yeast were grown aerobically in liquid YPGal at 30°C to early stationary phase. The method of Faye et al. (18) was used to prepare mitochondria with the exception that zymolyase, instead of Glusulase, was used to digest the cell wall. Phenylmethylsulfonyl fluoride was added to 10 g/ml final concentration during the cell-breaking step to minimize proteolysis.
Extraction of F 1 -F 0 Proteins from Mitochondria-Mitochondria were suspended at 6.6 mg/ml in 10 mM Tris-HCl, pH 8.0, 2 mM EDTA, 4 mM ATP (TEA buffer) and incubated for 15 min with 0.25% Triton X-100 to extract the F 1 -F 0 complex. The suspensions were centrifuged at 50,000 rpm in a Beckman 70.1Ti rotor for 30 min, and the resultant soluble and particulate fractions were assayed for F 1 proteins by Western analysis.
* This research was supported by National Institutes of Health Grant GM48157. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U46215.

Sucrose Gradient Centrifugation Analysis of F 1 -ATPase Subunits-
Step sucrose gradients were used to analyze the properties of F 1 subunits in cases where the enzyme was found to be resistant to Triton X-100 extraction. Mitochondria (5 mg/ml) were permeabilized in TEA buffer by a brief exposure to sonic irradiation, and 0.2-ml samples were overlaid on a 4.8-ml discontinuous gradient of Tris-HCl, pH 7.5, buffered sucrose (20 -80%, w/v). The preparation of the step gradients and the centrifugation protocol were as described (19). Linear sucrose gradients were used to evaluate the size of the F 1 -F 0 complex in mitochondrial supernatants following Triton X-100 extraction. Solubilized protein (0.6 ml) was loaded onto a 4.4-ml 6 -20% sucrose gradient prepared in 0.1% Triton X-100-supplemented TEA buffer, and the gradients were centrifuged at room temperature for 1.5 h at 55,000 rpm in a Beckman SW55Ti rotor. For all experiments, sucrose gradient fractions were collected from the bottom of the tube.
Colony Hybridization-The mutant atp2 genes were cloned by the method of colony hybridization (20). Following digestion of yeast genomic DNA with SphI, 7-kilobase fragments were purified from a 1% agarose gel and subcloned into a SphI-cut yeast/E. coli shuttle plasmid (either pRS316, CEN (21) or YEp352, 2 (22)). The mini-SphI libraries were used to transform E. coli, and the colonies were hybridized at 65°C with a nick-translated DNA fragment containing the entire ATP2 gene.
Construction of atp2 Null Strains-The one-step gene replacement method (23) was used to disrupt the chromosomal copy of ATP2 in the respiratory competent haploid yeasts W303-1A and W303-1B. The disruption allele (atp2::LEU2) was constructed by inserting a 2.8-kilobase BglII DNA fragment, carrying the entire LEU2 gene, into the unique BamHI site of the ATP2 reading frame. Wild type W303 strains were transformed with atp2::LEU2 on a linear DNA fragment and the transformants were screened for the stable acquisition of leucine prototrophy that correlated with the loss of respiratory function. Respiratory-deficient, Leu ϩ integrants of both mating types were obtained. Genetic crosses to o tester strains indicated that the respiratory defect of the disruption strains (W303ٌG1) was due solely to a nuclear mutation and Southern analysis verified that the gene disruption occurred at the ATP2 locus (data not shown). Genetic crosses also showed allelism between the disrupted gene in the W303ٌG1 strains and atp2 present in the chemically mutagenized yeast strains.
Assays-Protein concentrations were estimated by the method of Lowry et al. (24). ATPase activity was measured by determining the amount of inorganic phosphate released as described previously (25), in the absence and presence of 10 g of oligomycin.
Miscellaneous Procedures-Restriction endonuclease digestion of DNA, purification and ligation of DNA fragments, E. coli transformation and preparation of bacterial plasmids, and nick translation followed standard protocols (26). The dideoxy method (27) was used to sequence the mutant atp2 genes; both strands of the gene were sequenced using overlapping oligonucleotide primers spaced ϳ180 -200 nucleotides apart. Yeast transformations were done with the LiAc procedure (28). For Western analysis, proteins were separated on 12% polyacrylamide gels run in the electrophoretic system of Laemmli (29), modified as described (25) to optimize resolution of the F 1 ␣ and ␤ subunit proteins. The procedure of Schmidt et al. (30) was used for Western blotting. Antibodies against F 1 ␣, F 1 ␤, and cytochrome c 1 were used at dilutions of 1:2000, 1:3000, and 1:100 respectively.

Genetic Analysis of Yeast atp2
Mutants-Chemically mutagenized yeast strains belonging to complementation group G1 (17) carry mutations in ATP2, the nuclear gene that codes for the ␤ subunit of the F 1 -F 0 ATP synthase. The mutant atp2 genes were cloned from nine G1 strains, and the mutations were sequenced (see "Experimental Procedures"). The position of these mutations in the nucleotide and protein sequences is shown in Table II. A single missense mutation was found in each of the mutants, which alters the sequence of ATP2 relative to the wild type copy present in the parental strain D273-10B/A1 (GenBank™ accession number U46215). Two of the mutations were found in more than one strain: the P179L substitution occurs in the strains C103 and E210, and the mutation E222K is present in the strains E430 and E144. 1 The mutation in strain E700 converts the initiator methionine to isoleucine. Since this mutation does not provide information on residues involved in F 1 biogenesis or catalysis, it will not be discussed further.
Mitochondrial ATPase Activity and F 1 -F 0 Assembly in Haploid atp2 Mutant Strains-The amount of oligomycin-sensitive mitochondrial ATPase activity present in the mutant haploid strains is shown in Table III. All of the mutations produce a near complete loss (Ն93%) of ATPase activity. Western blots ( Fig. 1) were used to determine the amount of F 1 ␤ subunit in each of the atp2 mutants; these values are also reported in Table III. The level of F 1 ␤ subunit protein ranges from 44 to 85% the control level and thus does not account for the reduc-  Yeast ATPase ␤ Subunit Mutants tion in mitochondrial ATPase activity observed in the mutant strains.
The effect of the mutations on the assembly of F 1 -F 0 was also studied. The manner in which the F 1 ␣ and ␤ subunits distribute between soluble and particulate fractions when wild type and mutant mitochondria are extracted with 0.25% Triton X-100 is shown in the Western blots of Fig. 2. In the wild type (D273) the ␣ and ␤ subunits distribute almost equally between the two fractions, indicating that this procedure solubilizes approximately 40 -50% of the F 1 -F 0 from the inner mitochondrial membrane. The strains P159, E430, and E588 show evidence for mutant ␤ subunits with solubility properties that are comparable to those of the wild type protein. These strains were studied further to determine if the mutant protein is assembled into an F 1 -F 0 complex of the correct size. Soluble Triton X-100 extracts of mitochondria prepared from D273, and from the strains P159, E430, and E588, were centrifuged through linear sucrose gradients and the position of the F 1 ␣ and ␤ subunit proteins was determined by Western analysis (Fig. 3). The F 1 proteins in the mutant and the wild type samples showed comparable sedimentation properties. This result suggests that the A192V, E222K, and R293K mutations do not interfere with the formation of the F 1 -F 0 structure.
In the mutant strains E210 (P179L mutation), P133 (G133D mutation), and D123 (G227D mutation) virtually all of the ␤ subunit, and most of the ␣ subunit, sediment in the particulate fraction after Triton solubilization of the mitochondrial membrane (data shown for strain E210 in Fig. 2). This pattern of extraction is indicative of a defect in F 1 -F 0 assembly (19). The nature of this defect was investigated further by centrifuging permeabilized mitochondria through 20 -80% step-sucrose gradients (see "Experimental Procedures") to determine if the insoluble F 1 ␣ and ␤ protein is membrane-bound or in an aggregated form (19). Fractions from the gradient were analyzed on Western blots for the F 1 ␣ and ␤ subunits and for cytochrome c 1 . The latter protein remains membrane-bound under the conditions used to permeabilize the mitochondria and serves to mark the position of the inner membrane in the gradient. Fig. 4 shows the results of this experiment for the wild type strain, D273, and for one of the atp2 mutants (E210) that harbors insoluble F 1 protein. In the wild type (D273) the F 1 ␣ and ␤ subunits co-sediment with cytochrome c 1 (peak at fraction 6), indicating that most of the F 1 remains bound to F 0 in the inner membrane under the conditions of the experiment (see also Ref. 19). In the mutant strain E210 the F 1 subunits migrate separately from cytochrome c 1 and peak in the region of high sucrose density (fractions 2 and 3). Results similar to those shown for E210 were obtained with mitochondrial samples from mutant strains P133 and N123 (data not shown). This analysis indicates that the ␤ subunit mutations P179L, G133D, and G227D produce a defect in the assembly of F 1 characterized by the intramitochondrial accumulation of ␣ and ␤ subunit in large aggregates (see also Ref. 19).
Demonstration of Partial Dominance for the atp2 Point Mutations-The haploid yeast of the complementation group G1 are complemented in crosses to the wild type, which indicates the strains have recessive mutations in the ATP2 gene. However, in view of the catalytic and structural properties of the F 1 enzyme, it may be argued that certain mutations in ATP2 should be partially dominant with respect to a wild type copy of the ␤ subunit gene present in the same cell. There are three ␤ subunits in F 1 that are catalytic and the three catalytic sites interact cooperatively during enzyme turnover (1, 2). Studies using chemical (31) and photoaffinity (32) labels have shown that inactivating one of the three ␤ subunits attenuates multisite catalysis in F 1 . Analysis of hybrid forms of E. coli F 1 , reconstituted in vitro from mixtures of mutant and wild type subunits (either ␣ (33) or ␤ (34)), also show that inactivating a single catalytic site severely impairs F 1 steady-state catalysis. These studies support the notion that the co-synthesis of mutant and wild type ␤ subunits can lead to the formation of hybrid enzymes in vivo in which the cooperative interactions are lost. Alternatively, the presence of a mutant ␤ subunit could exert a partially dominant effect by interfering with the assembly of a stable F 1 structure. a The amount of F 1 ␤ subunit protein was determined using an AMBIS 4000 imaging system to scan Western blots of total mitochondrial protein. The mean values from the analysis of two separate Western blots are reported; one of the blots used is shown in Fig. 1. b Results from the assay of two separate mitochondrial preparations are shown; one set of values is given in parentheses. To investigate if the atp2 mutations we identified are partially dominant over wild type ATP2, catalytic activity of the F 1 -F 0 complex was analyzed in diploid strains formed by crossing the atp2 mutants to the wild type yeast, W303-1A. The level of oligomycin-sensitive ATPase was first measured in mitochondrial samples prepared from each of the strains (Table  IV). The mitochondria were then extracted with 0.25% Triton X-100, and the amount of solubilized F 1 -F 0 was quantified from Western blots (Fig. 5, Table IV). Finally, the specific activity of the F 1 -F 0 complex was calculated by dividing the percent mitochondrial ATPase activity by the percent solubilized ␤ subunit protein to give the values reported in the last column of Table IV. Support for using these calculations to determine the level of F 1 -F 0 in the yeast diploids comes from the fact that nonassembled F 1 is not extracted with Triton X-100 (19) and that under identical extraction conditions, the percentage of assembled F 1 -F 0 that is solubilized should be the same in each of the samples. The reference value for F 1 -F 0 catalytic activity is that measured with the double wild type diploid, W303ϫD273. Interestingly, control amounts of F 1 -F 0 ATPase activity were measured in W303 ٌG1ϫD273, which is a diploid strain that harbors a single gene for the ␤ subunit (Table IV). This result suggests that expression of ATP2 from either one or two wild type loci provides the diploid cell with the similar complement of the F 1 -F 0 ATP synthase.
Assuming that random interactions occur among the F 1 subunits from both parental cells, there is the potential to form four types of enzyme oligomers in the wild type/mutant diploids: a 3␤ wt , a 3␤ mut , and two types of hybrids (2␤ wt 1␤ mut and 1␤ wt 2␤ mut ). All four enzyme types are expected to form in diploids if the resident mutant ␤ subunit does not affect F 1 assembly. With respect to the strains shown in Table IV, this situation applies to W303ϫP159, W303ϫE430, and W303ϫE588. In such strains, the 3␤ wt and 3␤ mut enzymes are each expected to comprise 12.5% of the total F 1 oligomer population; the hybrid forms constitute, cumulatively, the remaining 75% of enzyme (35). Since the P159, E430, and E588 haploid strains are devoid of ATPase activity (Table III), it is assumed that the 3␤ mut F 1 formed in the derivative diploid cells is inactive. In cases where the 3␤ mut F 1 is the only inactive form of the enzyme, the maximal level of ATPase activity expected for the diploid strain is 87.5% (cumulative activity from the 3␤ wt and hybrid enzymes). This value is within 5% error of the ATPase activity observed for W303ϫP159 (Table IV). Therefore, it would appear that this particular diploid forms fully active F 1 hybrids in addition to the 3␤ wt enzyme. The other two members of this diploid class exhibit 31 and 74% the control level of F 1 -F 0 ATPase activity. Both values are in between the maximum (87.5%) and minimum (12.5%) levels of enzyme activity expected in this analysis, which suggests the presence of partially active F 1 hybrids in addition to the fully wild type enzyme. The reduction in the activity of the hybrid enzymes is most striking for W303ϫE430, suggesting that the E222K mutation displays significant dominance over the wild type ATP2 allele present in the diploid cell.
The other three wild type/mutant diploid strains shown in Table IV (W303ϫP133, W303ϫE210, W303ϫN123) synthesize mutant ␤ subunits that do not assemble into an F 1 structure in the respective haploid strains (Figs. 2 and 4). Therefore, it is assumed that the 3␤ mut enzyme does not form in these particular diploids. It is also likely that mutations, which affect the ability of the ␤ subunit to assemble homogeneous mutant enzymes, will also prevent the formation of hybrid structures. In such cases the presence of the mutant atp2 gene would be no worse than having the null allele, since the pool of wild type subunits would not be used to form F 1 hybrids. As only the 3␤ wt enzyme would be synthesized in the diploid cell, one should see levels of F 1 -F 0 ATPase activity comparable with that of the W303ٌG1ϫD273 strain (Table IV). This result is observed with W303ϫE210 and W303ϫN123 (Table IV), suggesting that the mutant ␤ subunits synthesized in these strains do not interact FIG. 4. Western analysis of step sucrose gradient fractions. Mitochondria prepared from the wild type strain, D273, and from the mutant, E210, were permeabilized by sonic irradiation and centrifuged through 20 -80% step sucrose gradients as described under "Experimental Procedures." Ten fractions of equivalent volume were collected from the bottom of the tube, and 15 l of each fraction were run on a 12% SDS-polyacrylamide gel. Following transfer to nitrocellulose, the blot was exposed to antibodies against the F 1 ␣ and ␤ subunits and cytochrome c 1 (cytc 1 ). The positions of the ␣, ␤, and cytochrome c 1 protein bands are indicated in the left-hand margin.

FIG. 3. Western analysis of linear sucrose gradient fractions.
Mitochondria, prepared from the yeast strains shown, were extracted with 0.25% Triton X-100, and aliquots of the soluble fractions were centrifuged through 6 -20% linear sucrose gradients as described under "Experimental Procedures." Twenty fractions of equivalent volume were collected from the bottom of the tube, and 15 l of each fraction were analyzed to detect the F 1 ␣ and ␤ subunits by Western analysis as described in the legend to Fig. 1. The position of the ␣ and ␤ subunit bands are indicated in the left-hand margin.
with wild type F 1 subunits. In contrast, the diploid strain W303ϫP133 shows 69% the control level of F 1 -F 0 ATPase activity. The submaximal level of enzyme activity observed indicates that this diploid harbors hybrid forms of F 1 and that these enzymes are only partially active.
It is likely that the partial dominance of the E222K, R293K, and G133D mutations was not recognized from simple complementation assays, because the respective diploid strains display Ն30% the control level of the enzyme activity. It has been reported that yeast mutants, with as little as 15% the wild type level of mitochondrial ATPase activity, are competent for growth on nonfermentable carbon sources (36). DISCUSSION With the exception of the initiator methionine change to isoleucine, the atp2 mutations reported in this paper are located in the central ␣␤ domain of the protein. This portion of the ␤ subunit encompasses the adenine nucleotide binding site and contains the P-loop region, which is a sequence element found in many nucleotide-binding proteins, including adenylate kinase and p21 ras (37). Fig. 6 shows the alignment for part of the ␣␤ domains of the S. cerevisiae and E. coli F 1 ␤ subunits. The amino acids of the bovine ␣␤ domain, which participate directly in ATP binding, are known from the x-ray structure of bovine F 1 (6); these residues are identically conserved in the yeast and bacterial sequences and are indicated in Fig. 6 with arrows. The yeast amino acids that are altered by the atp2 mutations described are boxed in the yeast sequence. The black boxes denote amino acids (Gly 133 , Pro 179 , and Gly 227 ) where the mutation alters enzyme assembly. The open boxes indicate amino acids (Ala 192 , Glu 222 , and Arg 293 ) where the mutations affect the catalytic activity of the enzyme, while leaving the F 1 -F 0 structure intact. Notably, these latter three yeast amino acids align with residues in the E. coli sequence (circled amino acids, Fig. 6) that have been analyzed in mutagenesis studies. The complementary information derived from the current work and from studies with the bacterial enzyme is outlined below.
A192V-Chemical mutagenesis induced a change of the Ploop alanine to valine both in yeast (Table II) and in E. coli (38). The phenotype described for the bacterial A151V mutant (38) is similar to what we have determined for the yeast strain, P159; the F 1 -F 0 assembles correctly, yet is deficient in membranebound ATPase activity (Table III, Fig. 3). Our results are also in accord with Shen et al. (15) who screened mutations in the yeast ␤ subunit P-loop region and found that mutants with the A192V change were respiratory-deficient. The position occupied by alanine in the P-loop of the F 1 ␤ subunit is a glycine residue in p21 ras ; notably, a mutant Ras protein with diminished GTPase activity was isolated in which this glycine residue is changed to valine (39). Apparently, substitution of valine at the 3rd position of the conserved P-loop (GXXXXGKT/S) has a global effect on the catalytic activity of NTP hydrolyzing enzymes that contain this consensus sequence. In E. coli F 1 , the A151V mutation produces a dramatic decrease in promotion from unisite to multisite catalysis (38). Our results with the W303ϫP159 diploid (Table IV) indicate that hybrid forms of F 1 , which contain either one or two ␤ subunits with the corresponding A192V mutation, are fully active. Therefore, it would appear that the P-loop Ala 3 Val mutation only blocks F 1 catalysis in enzyme oligomers homogenous for the mutant subunit.
E222K-The phenotype of yeast harboring the chemically induced E222K mutation (strains E430 and E144, Table II) can be compared with the E. coli E181Q ␤ subunit mutant derived from site-specific mutagenesis (40). Mutation of this glutamate residue to either lysine or glutamine does not interfere with F 1 -F 0 assembly, but lowers the steady-state level of membranebound ATPase activity to 2% the wild type level in haploid strains (Fig. 3, Table III (40,41). The fact that mutagenesis studies in yeast (present work) and in bacteria (41) give importance to this glutamate residue in the F 1 catalytic mechanism is in accord with structural information obtained with the bovine enzyme (6). The equivalent glutamate residue in the bovine ␤ subunit (Glu 188 ) contributes to ATP binding in the catalytic site, and there is electron density in the map that could represent a water molecule hydrogen bonded to its carboxylate group (6). These observations have led the authors to suggest that Glu 188 activates the water, promoting an in-line nucleophilic attack on the ␥-phosphate of ATP during F 1 turn- a The amount of soluble F 1 ␤ subunit protein was determined using an AMBIS 4000 imaging system to scan Western blots of Triton X-100 mitochondrial extracts. The mean values from the analysis of two separate Western blots are reported; one of the blots used is shown in Fig. 5.
b Results from the assay of two separate mitochondrial preparations are shown; one set of values is given in parentheses. c Values were calculated by dividing the mean percent mitochondrial ATPase activity by the mean percent F 1 ␤ subunit protein solubilized with Triton X-100.
d The mean values are reported.
FIG. 5. Western analysis of Triton X-100 extracts from mitochondria of yeast diploid strains. Mitochondria, prepared from the yeast diploids shown, were extracted with 0.25% Triton X-100 and centrifuged as described under "Experimental Procedures." Equal volumes of each supernatant were analyzed by Western analysis to detect the F 1 ␣ and ␤ subunits as described in the legend to Fig. 1. The volume of the supernatant loaded onto the protein gel was equivalent to 20 g of the starting mitochondrial protein.
over (6). The dramatic effect on F 1 catalytic activity that occurs when this glutamate residue is mutated is also readily apparent in diploid yeast (Table IV) and in E. coli transformants (41,42) that synthesize hybrid forms of the enzyme. Our results with the W303ϫE430 strain (Table IV) suggest that F 1 catalysis is severely inhibited in F 1 hybrids containing the mutant subunit, which supports the idea that the affected glutamate residue is important for cooperative interactions in the enzyme.
R293K-The yeast R293K ␤ subunit mutation produces a near complete loss of ATP hydrolytic activity in the haploid strain E588 (Table III). Two different mutations of the corresponding arginine residue in the E. coli ␤ subunit (R246C (43) and R246H (34)) were obtained by chemical mutagenesis. The phenotypes of both bacterial mutants are similar and correlate well with the defect of the corresponding yeast mutant. The catalytic impairment observed when this arginine residue is mutated is not unexpected, since the equivalent amino acid in the bovine ␤ subunit (Arg 260 ) participates in binding the triphosphate moiety of ATP in the catalytic site (6). In the diploid that harbors one copy of the mutant allele (Table IV), the mutation R293K reduces the maximum steady-state level of ATPase activity by 17%. This result implies that there is only a modest effect on catalytic cooperativity in hybrid enzymes that contain this particular mutant ␤ subunit.
The other three ␤ subunit mutations (G133D, P179L, and G227D) reported in the present work are novel in that they alter amino acids different from those studied previously in yeast or bacteria. The overall phenotype of the haploid mutants (P133, E210, and N123) is defective F 1 assembly with the accumulation of the ␣ and ␤ subunits in large aggregates inside mitochondria (Fig. 4). However, analysis of the wild type/mutant diploid strains (Table IV) indicates two subphenotypes, distinguished on the basis of whether or not the mutant ␤ subunit can interact with other F 1 subunits. The ␤ subunits from E210 and N123 do not form hybrid F 1 enzymes in the derivative diploid strains, while the P133 ␤ subunit does. The position of the mutated residues was identified in the structure of the bovine mitochondrial F 1 (coordinates were kindly provided by A. Leslie and J. E. Walker, Medical Research Council, Cambridge, United Kingdom (UK)). In E210, the mutation alters Pro 179 ; in the bovine protein this residue (Pro 145 ) is located at the interface with one of the ␣ subunits, in close proximity to the nonexchangeable adenine nucleotide binding site (6). The mutation of N123 alters residue Gly 227 ; the corresponding glycine (Gly 193 ) in the bovine ␤ subunit is in close juxtaposition with the other adjacent ␣ subunit. The fact that the E210 and N123 mutations alter ␤ subunit amino acids that interact with neighboring F 1 subunits supports the notion that the mutated proteins do not assemble hybrid enzymes in diploid cells. The P133 mutation produces a change in Gly 133 ; in the bovine ␤, the equivalent residue (Gly 99 ) is located at the junction between the amino-terminal ␤ barrel and central ␣␤ tertiary structural domains in the ␤ subunit (6). This glycine residue is structurally distant from adjacent F 1 subunits, which may explain why a mutation of this amino acid does not preclude the mutant subunit from interacting with the other enzyme subunits. However, inspection of the x-ray structure of bovine F 1 shows that the G133D mutation of strain P133 is likely to disturb the three-dimensional structure of the ␤ subunit, since the space in the vicinity of the glycine residue is too small to accommodate the mutation to aspartic acid. The presumed structural perturbation might be responsible for the reduced level of F 1 -F 0 specific activity and for the smaller amount of soluble ␤ subunit detected in W303ϫP133 with respect to the other diploid strains (Table IV). Thus, in addition to assembling catalytically impaired hybrid enzymes, the P133 mutant ␤ subunit probably engages in nonproductive interactions that lower the steady-state level of assembled (soluble) F 1 -F 0 .
The type of F 1 assembly defect observed in the atp2 mutants described here is also seen in yeast mutants where the genetic lesion affects the Atp11p and Atp12p F 1 assembly factor proteins (19). The proposed function of the F 1 -specific "chaperonetype" proteins (Atp11p and Atp12p) is to ensure that productive associations between F 1 ␣ and ␤ subunits prevail over the formation of nonproductive ␣ n and ␤ n complexes (44). In light of evidence for direct binding between Atp11p and the F 1 ␤ subunit (45), the atp2 mutations that correlate with an F 1 aggregation ϩ phenotype are of particular interest, since they may affect amino acids that interact with the assembly factor.
FIG. 6. Sequence alignment for part of the adenine nucleotide binding domain of the E. coli and S. cerevisiae F 1 ␤ subunit. Numbering of the E. coli sequence is according to Refs. 3 and 4. The yeast sequence starts with the initiator methionine as amino acid 1. Arrows indicate the amino acids homologous to the bovine residues that were shown to participate directly in binding adenine nucleotides (6). The boxed amino acids in the yeast sequence, and the circled residues in the bacterial sequence are discussed in detail in the text. The highly conserved P-loop region is indicated by the bar in the figure.