Organization of Subunits in the Membrane Domain of the Bovine F-ATPase Revealed by Covalent Cross-linking

Background: Bovine mitochondrial F-ATPase has six uncharacterized membrane subunits. Results: Labeled lysines have been mapped onto its incomplete structure. Conclusion: Subunit ATP8 extends about 70 Å from the membrane into the peripheral stalk. The N-terminal regions of five others are in the mitochondrial matrix. Significance: Subunit ATP8 is in the stator. Five other subunits form a subdomain.

peripheral stalks. The F 1 domain is an assembly of three ␣-subunits and three ␤-subunits arranged in alternation in a spherical complex around an elongated ␣-helical structure in the ␥-subunit (3), and the three catalytic sites of the enzyme lie at interfaces between the ␣and ␤-subunits. The ␥-subunit extends from the ␣ 3 ␤ 3 -spherical structure to the membrane domain, where it is augmented by the ␦and ⑀-subunits in the contact region or "foot" with the membrane domain. Together, the ␥-, ␦-, and ⑀-subunits constitute the central stalk, and the foot makes extensive contacts with a hydrophobic ring of eight c-subunits in the membrane domain of the enzyme (4). The c 8 -ring and the central stalk constitute the rotor of the enzyme. Their rotation as an ensemble at ϳ100 Hz carries energy into the catalytic sites of the enzyme, and the turning of the rotor brings about conformational changes in the three catalytic sites in the F 1 domain that lead to the binding of substrates and the formation of ATP and its release into the matrix. The region of contact between the rotating c-ring and subunit a, another membrane protein of unknown structure, provides the transmembrane pathway for protons to pass from the intermembrane space into the mitochondrial matrix (5), allowing energy from the transmembrane proton-motive force generated by respiration to drive the turning of the rotor. Subunit a forms part of the stator and is in contact with the membrane domain of subunit b, which extends through the peripheral stalk, an elongated subcomplex of single copies of subunits b, d, F 6 , and oscp (6 -9) to the "top" of the ␣ 3 ␤ 3 domain (10). Thus, the stator consists of the ␣ 3 ␤ 3 domain, plus subunits oscp, b, d, F 6 , and a, and it remains static relative to the turning of the rotor. Atomic resolution structures have been established for the F 1 c-ring (4) and peripheral stalk domains (7,8), and a mosaic structure of the enzyme has been built within the envelope of a low resolution structure of the entire complex determined by electron cryo-microscopy (11).
Also associated with the membrane domain of the bovine enzyme are six small proteins, ATP8, e, f, g, DAPIT 3 (diabetes * This work was supported by the intramural program of the Medical associated protein in insulin sensitive tissues), and 6.8 kDa proteolipid (6.8PL), with molecular masses in the range 6.3-11.4 kDa (1,2,(12)(13)(14). Each has a single predicted transmembrane ␣-helix, and based largely on their staining intensities in gel analyses of the subunit composition of the enzyme, it is assumed that there is one copy of each protein per F-ATPase complex, but there are no definitive quantitative data to support this assumption. These proteins are usually referred to as the "supernumerary" subunits as there are no orthologs in bacterial F-ATPases, and they appear not to be involved directly in the synthesis of ATP. In the yeast enzyme the orthologs of subunits e and g are associated with the formation of dimers of the F-ATPase (15,16), and they probably play the same role in the mammalian enzyme.
In this study the identification of covalent cross-links introduced into the subunits of the bovine F-ATPase by reaction of exposed amino groups with isotopically labeled bifunctional reagents has been employed as a means of gaining information about the orientation and location of the supernumerary subunits within the bovine enzyme complex.

Experimental Procedures
Analytical Methods-Protein concentrations were estimated by the bicinchoninic acid assay (17). Protein compositions were analyzed by SDS-PAGE in 12-22% or 4 -20% polyacrylamide gradient gels or by blue native PAGE (18,19). Proteins were detected by staining with Coomassie Blue dye.
Chemical Cross-linking-Four deuterium-labeled cross-linking agents were employed. Disuccinimidyl suberate (DSS) and disuccinimidyl glutarate (DSG) are membrane-permeable, whereas di(sulfosuccinimidyl) glutarate (DSSG) and bi(sulfosuccinimidyl) suberate (BS 3 ) are membrane impermeable. In the isotope-labeled versions, DSS(d 0 /d 12 ), BS 3 (d 0 /d 12 ), DSG(d 0 / d 6 ), and DSSG(d 0 /d 6 ), (Creative Molecules), the numbers of deuterium atoms, d, in the two acyl arms of each reagent are indicated by the subscripted numerals and were supplied and used at molar ratios of the two isotopic forms of 1:1 (22,23). Freshly purified bovine F-ATPase was diluted to a concentration of 1.22 mg/ml in buffer consisting of 20 mM HEPES, pH 7.3, 0.05% n-dodecyl-␤-D-maltose-neopentyl glycol (w/v), 10% glycerol (v/v), 50 mM sodium chloride, 1 mM tris(2carboxyethyl)phosphine, 2 mM magnesium sulfate, 2 mM ATP, 1-palmitoyl-2-oleoyl-sn-glycero-phosphocholine (100 g/ml), 1-palmitoyl-2-oleoyl-sn-glycerophosphoethanolamine (33 g/ ml), and 1-palmitoyl-2-oleoyl-sn-glycero-phospho-(1Ј-racglycerol (33 g/ml). The reaction conditions were optimized with DSS(d 0 /d 12 ) dissolved in dimethyl sulfoxide added at concentrations of 0.1-5 mM. The reactions were carried out at pH values of 6.8, 7.3, and 8.0 for 2, 4, or 20 h and at temperatures of 20 or 37°C. The optimal cross-linking conditions were 2 M F-ATPase (1.16 mg/ml) at pH 8.0 reacted with 2 mM DSS(d 0 / d 12 ) for 2 h at 20°C. BS 3 (d 0 /d 12 ), DSG(d 0 /d 6 ) and DSSG(d 0 /d 6 ) were reacted with the enzyme at the same concentration and under the same conditions. The reactions were terminated by the addition of ammonium bicarbonate to a final concentration of 200 mM. The solutions were kept at 20°C for 30 min. Insoluble material was removed by centrifugation (10,000 ϫ g, 5 min, 20°C), and the supernatants were analyzed by SDS-PAGE and blue native-PAGE. The cross-linked F-ATPases were precipitated with ethanol for 18 h at Ϫ20°C, and the precipitates were recovered by centrifugation and redissolved in 6 M guanidinium hydrochloride in 20 mM Tris-HCl buffer. The proteins were reduced with tris(2-carboxyethyl)phosphine, alkylated with iodoacetamide (10 mM), and re-precipitated. These samples of S-alkylated cross-linked F-ATPases were analyzed in three different ways. First, the precipitates were dissolved in 100 mM ammonium bicarbonate and digested at 37°C for 16 h with trypsin (trypsin:protein, 1:100, w:w). Portions of the digests were diluted with 0.1% formic acid for mass spectrometric analysis. Second, other samples of S-alkylated and cross-linked F-ATPase were fractioned by SDS-PAGE, and gel bands or gel regions were excised and digested with trypsin (24). Third, an S-alkylated sample of F-ATPase cross-linked with DSS(d 0 /d 12 ) was dialyzed for 18 h against 100 mM ammonium bicarbonate and digested at 37°C with trypsin (F-ATPase: trypsin, 100:1, w:w). The digest was dried and fractionated on a column of PolySULFOethyl A (50 mm long ϫ 1.0 mm inner diameter; 300 Å pore size; PolyLC, Columbia, MD) equilibrated at a flow rate of 50 l/min in 10 mM potassium phosphate buffer, pH 2.7, containing 10% acetonitrile. Peptides were eluted with a gradient of 0 -1 M potassium chloride in 20 mM potassium phosphate, pH 3.0, containing 10% acetonitrile (see Fig. 1). Fractions were dried and dissolved in 0.3% trifluoroacetic acid, and salts were removed from the samples by passage through a reversephase C 18 ZipTip (Millipore).

Analysis of Cross-linked Peptides by Mass Spectrometry-
Tryptic digests of protein bands from SDS-PAGE gels were analyzed in a MALDI-TOF-TOF mass spectrometer (Model 4800, ABSciex, Warrington, WA1 1RX, UK) with ␣-cyano-4hydroxycinnamic acid as matrix. Peptides were fragmented by collision-induced dissociation with air at a collision energy of 1 kV. Trypsin digests of cross-linked F-ATPases (ϳ400 fmol), and fractions of peptides from similar digests that had been obtained by cation exchange chromatography were injected into a nano-scale C 18 reverse phase column. The eluate was introduced directly into the nanospray interface of a LTQ Orbitrap XL ETD mass spectrometer (Thermo-Fisher Scientific, Hemel Hempstead, UK), operated in data-dependent acquisition mode. Up to 10 of the most abundant precursor ions of known charge states, but not singly charged, were selected and fragmented by collision-induced dissociation. The m/z values of precursor and fragment ions were measured simultaneously in the Orbitrap and ion-trap analyzers, respectively.
Data Analysis-Data files were processed with Proteome Discoverer (Thermo-Fisher Scientific). The coverage of the sequences of the subunits of the F-ATPase by the tryptic peptides was determined by comparison of the sequences of peptides with the mammalian subsection of the non-redundant database of the National Center for Biotechnology Information with MASCOT (25). "Mono-linked" peptides, where only one arm of the bifunctional cross-linking agent reacted with a protein amino group, were identified by comparison of the data with a FASTA database containing the sequences of the subunits of the enzyme. In these comparisons the light and heavy isotopes of the hydrolysis and aminolysis derivatives of monolinked ␣-amino and ⑀-amino groups, oxidized methionines, and S-propionamido-and S-carbamidomethyl derivatives of cysteine were considered as possible modifications with the following parameters: peptide mass tolerance, Ϯ 5ppm; fragment mass tolerance, Ϯ0.5 Da; maximum missed cleavages, 2; peptide confidence, at least medium.

Identification of Cross-linked and Loop-linked Peptides-
Doublet peaks corresponding to the light and heavy isotopically labeled cross-linked peptides (including loop-linked peptides) were identified from their mass differences with Hekate (26). For the doublets for DSS(d 0 /d 12 ) and BS 3 (d 0 /d 12 ) the mass difference was 12.0741 Ϯ 0.0375, and for DSG(d 0 /d 6 ) and DSSG(d 0 /d 6 ) it was 6.0370 Ϯ 0.0375 Da. No limits were imposed on the intensity ratios of the doublet peaks. To be assigned as an isotopic doublet, peptide ions containing light and heavy isotopes were required to elute within 30 s of each other. The mass threshold for matching the masses of the light isotope peak of cross-linked peptides with their calculated theoretical values was 2 ppm. The accuracy threshold used in the annotation of fragment ions was Ϯ0.5 Da. The identified crosslinks were ranked according to a modified version of Andromeda (26,27) with fragment ion data from both the light-and heavy-labeled cross-linked peptides.
Mass spectra from cross-linked and loop-linked peptides with a score of Ͼ400 were inspected manually. A cross-link was accepted as being significant if both cross-linked peptides had more than five unique fragment ions assigned to them or if the cross-linked peptide had three-five unique fragments ions assigned to one component and more than five to the other. Accepted loop links had three or more unique assigned fragments. Twenty mono-links and six loop links were detected (data not shown). Peptides were discarded when the majority of high intensity fragment ions were not annotated with a theoretical fragment ion from the associated peptide or when fewer than three fragment ions of a peptide component of a cross-link were detected.
Localization of Cross-links in the Structure of the Bovine F-ATPase-A structure of the bovine F 1 -c 8 -peripheral stalk complex was generated from the structures of the bovine F 1 -c 8ring (PDB code 2XND (4)), the F 1 -peripheral stalk (PDB code 2WSS (7)), and the peripheral stalk (PDB code 2CLY (8)), and the position of the peripheral stalk was adjusted according a low resolution structure of the intact complex (11). The distances between C␣ atoms of lysine residues in the model of the bovine F 1 -c 8 -peripheral stalk complex were estimated with PyMOL and Coot (28,29). For DSS(d 0 /d 12 ) and BS 3 (d 0 /d 12 ), the maximum permitted inter-C␣ distance between connected lysines is 27.4 Å (the 11.4 Å spacer arm plus the length of two lysine side chains, each 6.5 Å, plus twice the coordinate error of 1.5 Å for mobile surface residues). For DSG(d 0 /d 6 ) and DSSG(d 0 /d 6 ) with spacer arms of 7.7 Å, the maximum inter-C␣ distance is 23.7 Å.
Analysis of Sequences of F-ATPases-Sequences of subunits of F-ATPases were aligned with ClustalW (30). Transmembrane ␣-helices were predicted with HMMTOP (31), and the secondary structures of membrane extrinsic regions were predicted with PSIPRED (32).   Fig. 2, it is evident that the a-and c-subunits were unreactive with DSS(d 0 /d 12 ), and in lane j, that cross-linking of the enzyme reduced its apparent molecular mass on blue native PAGE gels relative to unreacted enzyme (Fig. 2, lane k), presumably because the cross-links make the enzyme more compact.

Optimization of Reaction Conditions for Cross-linking the
Characterization of Cross-linked Peptides-The number of cross-links identified in each experiment depended upon how the tryptic peptides were generated and fractionated. Fewer cross-linked pairs were identified in samples processed by SDS-PAGE and in-gel digestion than from other samples that were digested with trypsin and analyzed without any further fractionation, probably because of losses of cross-linked peptides in the former procedure during their extraction from gels and subsequent processing. However, there was considerable overlap between the sets of cross-linked peptides that were identified by these means. The most effective procedure was that applied to the tryptic digest of the F-ATPase modified with DSS(d 0 /d 12 ), where cross-linked peptides were separated from unmodified, mono-linked, and loop-linked peptides by cation exchange chromatography at pH 2.7. At this pH value, crosslinked peptide pairs carry four positive charges (or more if histidine residues are present) and so they bind more strongly to a strong cation exchange resin than unmodified, mono-linked, and loop-linked peptides, which carry a minimum of two positive charges. Thus, the cross-linked peptides were recovered in a separate peak eluting after the less positively charged peptides ( Fig. 1). Of the 53 cross-linked peptides identified in this peak, (Tables 1 and 2), 30 contained intrasubunit cross-links, and 23 had intersubunit cross-links. In other experiments where the ion exchange step was omitted, 28 peptides cross-linked with DSS were recovered, and 24, 14, and 17 peptides cross-linked with BS 3 , DSG, and DSSG were recovered, respectively. There was no significant difference between the number of crosslinked peptides recovered from the membrane subunits of the F-ATPase reacted with the membrane-permeable reagents, DSS and DSG, and with the membrane-impermeable reagents, BS 3

and DSSG.
Cross-links in the Membrane Extrinsic Region of the F-ATPase-Forty-four cross-linked peptide pairs were identified as originating from membrane-extrinsic subunits of the enzyme. They are summarized in Table 1 together with the C␣ distances between the cross-linked lysine residues measured in the structure of the bovine F-ATPase, and the corresponding mass spectrometry data are summarized in Table 2. The lengths of cross-linkers are given in the experimental section. In 38 of them both lysines are resolved in the structure, and the formation of the cross-links is compatible with the inter-C␣ distances. For a 39th cross-link between ␣Lys-455 and Lys-24 in subunit d, the inter-C␣ distance of 32.3 Å is only slightly above the maximum value of 27.4 Å, and a minor conformational change in the determined structure of the enzyme would allow this cross-link to form. In two other cross-links between the F 1 domain and the peripheral stalk, one of the reacted lysine residues is in a region that is locally unresolved in the structure, although adjacent structural elements are well defined. One such cross-link bridges between unresolved residue ␤Lys-9 in the N-terminal region of the subunit and residue ␤Lys-78, which is resolved in the six-stranded ␤-barrel that forms part of the "crown" of the F 1 domain. The C␣ distances between the resolved residue ␤Thr-13, which is close to ␤Lys-9 in the sequence of the subunit, and ␤Lys-78 are 10 Å, 10 Å, and 10.2 Å in the ␤ DP -, ␤ TP -, and ␤ E -subunits, respectively. Therefore, the formation of a cross-link bridging between ␤Lys-9 and ␤Lys-78 is compatible with the structure. The second cross-link involves the unresolved residue Lys-2 in subunit F 6 and the resolved residue Lys-191 in subunit b. It is clear from the structure of the enzyme that residue Lys-2 in subunit F 6 and residue Lys-191 in the C-terminal region of subunit b are in close proximity (the estimated distance is ϳ25 Å), and the formation of this crosslink is consistent with this conclusion. Three other cross-links are incompatible with the structure of the monomeric bovine F-ATPase, and each was observed once only (see Table 1). One of them bridges between ␤Lys-413 and ␦Lys-114, whereas the C␣ distance between the two lysine residues in the structure is 57.3 Å. This cross-link is likely to have formed between two different F-ATPase complexes, and the presence of a small amount of the oligomeric complex in the native gel of the modified complex (Fig. 2, lane j) is consistent with this suggestion. Another incompatible cross-link bridges between ␣Lys-132 and ␣Lys-196, two surface residues in the nucleotide binding domain of the subunit. Although the inter-C␣ value given in Table 1 is apparently within the range of acceptable distances, this distance has been measured along the direct path between the two C␣-atoms, which is impeded by a loop involving residues 309 -321 of the ␣-subunit, and the unimpeded C␣ distance is considerably longer. The third structurally incompatible cross-link is between ␣Lys-384 and ␥Lys-4. The former residue is exposed at the entrances to cat-alytic interfaces between the ␣-subunits and the adjacent ␤-subunit, whereas the latter residue is in the N-terminal ␣-helix of the ␥-subunit. This ␣-helix is one of the two ␣-helical elements in the central shaft of the rotor found in the aqueous cavity in the core of the F 1 domain. In one of the three catalytic ␣-␤ interfaces, between ␣ DP and ␤ DP , there is a direct unimpeded line of sight between the two residues, but the gap narrows to ϳ6 Å, and the inter-C␣ distance is 38.4 Å. These last two structurally incompatible cross-links could only have formed by reaction with disrupted F-ATPase complexes. Despite these three aberrant cross-links, the overwhelming majority, 93%, of the observed linkages were compatible with the atomic structure of the F-ATPase. Their main value is that they provide reassurance that any cross-links observed in regions of the F-ATPase that are not represented in the current structural model are likely to have validity and that they represent contributions toward defining the organization of subunits in the unresolved region.

Cross-linked tryptic peptides in the membrane extrinsic region of the bovine F-ATPase
The cross-links were produced by reaction of the enzyme with DSS(d 0 /d 12 ), and the peptides were fractionated by ion exchange chromatography.

Subunits
Cross Cross-links in the Membrane Domain of the F-ATPase-Eight intersubunit and four intrasubunit cross-links were identified among the tryptic peptides from the F-ATPase that had been cross-linked with the four reagents. The majority of them were identified in the digest of the DSS cross-linked enzyme that was fractionated by ion exchange chromatography. They are summarized in Table 3, and the corresponding mass spectrometry data are summarized in Table 4.
Three of the intersubunit cross-links were between subunit ATP8 and components of the peripheral stalk domain. One of them links residue Lys-54 of subunit ATP8 to residue Lys-73 of the F 6 subunit. The two others involve links from residue Lys-46 of ATP8, one to residue Lys-120 of subunit b and the other to residue Lys-24 of subunit d (Fig. 3).
Five of the intersubunit cross-links involved lysine residues in the five supernumerary subunits e, f, g, DAPIT, and 6.8PL (see Fig.  4). All fall in regions of the proteins that are predicted to be membrane extrinsic. Three involve links between residue Lys-49 in the C-terminal region of the 6.8PL subunit with lysine residues in the C-terminal regions of the DAPIT, e, and f subunits. Other cross-links bridge between the C-terminal regions of subunits e and f and the N-terminal regions of subunits f and g.

Mass spectrometric characterization of cross-linked tryptic peptides in the membrane extrinsic region of the bovine F-ATPase
The cross-links were produced by reaction of the enzyme with DSS(d 0 /d 12 ), and the peptides were fractionated by ion exchange chromatography. b Cross-link also found in bovine F-ATPase modified with DSG(d 0 d 6 ). c Cross-link also found in bovine F-ATPase modified with BS 3 (d 0 d 12 ). d Cross-link also found in bovine F-ATPase modified with DSSG(d 0 d 6 ).

Involvement of Subunit ATP8 in the Bovine Peripheral
Stalk-Since its discovery as a subunit of the bovine and yeast F-ATPases (12,33,34), subunit ATP8 (also known as A6L in mammals and Aap1 in Saccharomyces cerevisiae) has remained a rather mysterious component of the enzyme complex. It is not found in eubacterial and chloroplast enzymes, and therefore, it was classified as a supernumerary subunit, apparently not required for the core ATP synthetic and hydrolytic functions of F-ATPases. The subunit is encoded in the mitochondrial DNA of many, but not all, eukaryotic species (35), and in mammals, the genes for ATP8 and ATPase-6 (or subunit a) overlap (12,36). The bovine protein is probably folded into a single transmembrane ␣-helix from residues 8 -29 followed by a hydrophilic extension up to its C terminus at residue 66 (Fig. 5A). This region from residue 30 to 66 is predicted to have an In the peripheral stalk of the complex (right), the membrane extrinsic region of subunit b is magenta, subunit F 6 is blue, and subunit d is orange. Subunit ATP8 is deep blue, and its predicted transmembrane ␣-helix from residues 8 -29 is depicted as a cylinder. Except for the c-ring, the region occupied by membrane components of the enzyme is not shown. The secondary structure of residues 29 -66 of subunit ATP8 is not known, and this region is shown in an extended form reaching up so as to place lysine residues 46 and 54 in the vicinity of the residues to which they are cross-linked, residues 120, 24, and 73 (indicated by the magenta, orange, and blue dots in subunits b, d, and F 6 , respectively). The colored circles around each dot have a diameter of 20 Å, and they represent the region in which each cross-linked residue in ATP8 is likely to be found in the structure.  extended conformation, except for residues 57-61, which may form a short ␤-sheet. The region consisting of residues 51-63 is well conserved in mammals (Fig. 5B). In contrast, subunit ATP8 in S. cerevisiae has a C-terminal extension that is predicted to be mostly ␣-helical. It is significantly shorter and poorly related to the same region of the mammalian proteins, although the sequences of ATP8 proteins are well conserved among the fungi (Fig. 5C). On the basis of cysteine scanning mutagenesis and reaction with fluorescein-5-maleimide, residues 1-14 of yeast ATP8 have been proposed to be exposed to the intermem-  brane space followed by a transmembrane ␣-helix from residues 15-35 (37). A short ␤-strand is predicted around residue Arg-42, and this residue is conserved throughout fungi (Fig.  5C). Despite these differences, proteolytic digestion studies conducted on bovine mitochondrial membranes (38,39) and cross-linking studies on the yeast enzyme (40) have shown that the bovine and yeast proteins have a common topography with their N-terminal regions in the intermembrane space and their C-terminal regions on the matrix side of the inner mitochondrial membrane.

TABLE 4 Mass spectrometric characterization of cross-linked tryptic peptides in the subunits of the membrane intrinsic region of the bovine F-ATPase
In the current work new information about the location of bovine subunit ATP8 in the F-ATPase has come from an extensive study of covalent cross-linking conducted on the intact purified enzyme with bi-functional agents. Among the many identified cross-links, residues Lys-46 and Lys-54 of ATP8 were found to be linked to three lysine residues, one in each of subunits b, d, and F 6 in the peripheral stalk of the enzyme. These cross-links confirm that the C-terminal region of bovine ATP8 is exposed in the matrix of the organelle, and they demonstrate that this region of ATP8 is in the vicinity of the peripheral stalk. The shorter and poorly related C-terminal region of yeast subunit ATP8 has been proposed to interact with subunits b and d in the yeast enzyme (37). Two of the cross-links in the bovine enzyme involve the structurally defined residues Lys-120 and Lys-24 of subunits b and d, respectively, both of which are found ϳ60 Å from the membrane domain of the enzyme. Therefore, residues Lys-46 and Lys-54 of ATP8 are nearby (see Fig. 3), and the C-terminal region of bovine subunit ATP8 extends ϳ60 -70 Å upwards from the membrane domain, probably along the axis of the peripheral stalk defined by the long ␣-helix in subunit b. The third cross-link is between residue Lys-54 in ATP8 and residue Lys-73 in subunit F 6 (Fig. 3). Bovine subunit F 6 is 76 amino acids long, and in the peripheral stalk, it is folded into two ␣-helices from residues 4 -25 and 33-51 linked by an extended loop region from residues 26 -32. Residue 51 is followed by an extended region, which is resolved up to residue Glu-69. The cross-linked residue Lys-73 of F 6 is nearby and is estimated to be ϳ70 Å from the surface of the membrane domain of the enzyme, in accordance with the position deduced from the structurally defined lysine residues in subunits b and d. The only known human pathogenic mutation specific to subunit ATP8 is associated with infantile cardiomyopathy, and it leads to the truncation of the subunit at residue 54 and impairment of ATP synthesis (41). Therefore, in mammalian mitochondrial enzymes subunit ATP8 appears to have a fundamental role in either the synthesis of ATP by mitochondrial F-ATPases or the assembly of the complex or both. If the C-terminal region of subunit ATP8 is, as proposed, an intrinsic component of the peripheral stalk, truncation of the subunit would disrupt the integrity of the peripheral stalk and impair ATP synthesis by uncoupling the catalytic sites in the F 1 domain from the proton motive force, generated across the inner mitochondrial membrane by respiration. Similar proposals have been made that the yeast ATP8 subunit is a component of the peripheral stalk of the yeast F-ATPase (37,40) and that it is required for the assembly of the complex (42,43). Thus, despite significant differences in the lengths and sequences of mamma-lian and yeast ATP8 subunits, it is likely that they have related roles in the enzyme complexes.
Organization of Supernumerary Subunits-The presence of the supernumerary subunits e, f, and g in the membrane domains of F-ATPases was demonstrated first in the bovine enzyme (2,13) and subsequently in the enzyme from S. cerevisiae (15,44). All three subunits are predicted to contain a single transmembrane ␣-helix (see Fig. 4), and the sequences of the bovine and yeast e, f, and g subunits are conserved or conservatively substituted in 26%, 34%, and 46% of their residues, respectively. The predicted secondary structures of the bovine and yeast orthologs are similar but not entirely concordant. Since their discovery, the roles of the yeast subunits have been studied much more extensively than their bovine orthologs. Yeast subunits e and g are involved in the formation of dimers of the enzyme in the mitochondrial cristae (15,45,46), and they are oriented with their N-terminal regions in the mitochondrial matrix and their C-terminal regions in the intermembrane space of the organelle (37,47). A GXXXG sequence motif in residues 14 -18 of the yeast e-subunit forms a homodimeric ␣-helical coiled-coil involved in holding two F-ATPase complexes together (45). However, the motif is not conserved in the bovine protein. Deletion of either subunit e or subunit g leads to severe disruption of the structure of the mitochondrial cristae, but neither subunit is required for the formation of an enzymatically active complex (15,48). Like subunits e and g, subunit f is oriented with its N-and C-terminal regions in the mitochondrial matrix and the intermembrane space, respectively (37,47,49), but in contrast to subunits e and g, deletion of subunit f disrupted both the assembly and the activity of the complex (44). In addition to subunits ATP8, e, f, and g, the membrane domain of the bovine enzyme also contains two additional subunits, DAPIT and 6.8PL (14,50), that have been identified also in the human enzyme (51). They are less tightly associated with the bovine complex than the other supernumerary subunits, and they require the presence of phospholipids to remain bound to the complex (14). Subunit DAPIT is encoded only in the genomes of metazoans (Fig. 6), and subunit 6.8PL is restricted to vertebrates (Fig. 7), and so neither subunit is a component of the yeast enzyme. Both have a single predicted transmembrane ␣-helix (Fig. 4). Conversely, the membrane domains of the yeast F-ATPase additionally contain subunits j (also called i) and k (15,52,53), and a recently discovered sub- unit l, a homologue of subunit k (54), that are unrelated in sequence to any of the bovine subunits and are not encoded in metazoan genomes. Again, each of subunits j, k, and l has a single predicted transmembrane ␣-helical span.
The present studies have revealed a network of cross-links between exposed lysine residues in the N-and C-terminal regions of bovine subunits e, f, g, DAPIT, and 6.8PL (see Fig. 4). Residue Lys-49 in the C-terminal region of 6.8PL is linked to the C-terminal regions subunits DAPIT, e, and f via cross-links to residues Lys-55, Lys-54, and Lys-79, respectively, and the C-terminal regions of subunits e and f are linked via a cross-link from Lys-47 in subunit e to Lys-85 in subunit f. Finally, the N-terminal regions of subunit f and g are cross-linked from Lys-15 in subunit f to Lys-65 in subunit g. This network of cross-links shows that the N-terminal regions of all five proteins are on the same side of the lipid bilayer in the intact F-ATPase and that, conversely, all of their C-terminal regions are located on the opposite side of the membrane. Although there is as yet no direct information about the orientations of the subunits in the bovine inner mitochondrial membrane, it is reasonable to assume that subunits e, f, and g have the same orientations as their yeast orthologs. In view of the cross-links between the C-terminal regions of the 6.8PL, DAPIT, and e subunits, it seems likely that 6.8PL and DAPIT also will be oriented in a similar fashion, as depicted in Fig. 4.
These studies on the supernumerary subunits of the bovine F-ATPase have little or no apparent relevance to the interpretation of the structure of the dimeric F-ATPase from the mitochondria of the alga Polytomella sp. at 6.5 Å resolution (55). It is known that the F-ATPase from this species and from the related alga Chlamydomonas reinhardtii contains the catalytic core subunits ␣, ␤, ␥, ␦, ⑀, oscp, a, and c plus nine "atypical" subunits (56). Atypical subunits Asa2, Asa4, and Asa7 are thought to be components of the peripheral stalk (57), which is much stouter and more elaborate than the peripheral stalk in the bovine and fungal F-ATPases. The stoichiometry of none of the atypical algal subunits is known, and there is no evident sequence relationship to any of the much more extensively studied peripheral stalk subunits in the bovine enzyme especially (6 -10), and there is no relationship either of atypical subunits with any of the supernumerary subunits described here, including ATP8. Indeed, it is not clear whether or not the algal enzymes contain an ATP8 subunit (58). The information in this paper is much more likely to be helpful in establishing relatively high resolution structures of the bovine and fungal enzymes.