Subunit structure and organization of the genes of the A1A0 ATPase from the Archaeon Methanosarcina mazei Gö1.

The proton-translocating A1A0 ATP synthase/hydrolase of Methanosarcina mazei Gö1 was purified and shown to consist of six subunits of molecular masses of 65, 49, 40, 36, 25, and 7 kDa. Electron microscopy revealed that this enzyme is organized in two domains, the hydrophilic A1 and the hydrophobic A0 domain, which are connected by a stalk. Genes coding for seven hydrophilic subunits were cloned and sequenced. From these data it is evident that the 65-, 49-, 40- and 25-kDa subunits are encoded by ahaA, ahaB, ahaC, and ahaD, respectively; they are part of the A1 domain or the stalk. In addition there are three more genes, ahaE, ahaF, and ahaG, encoding hydrophilic subunits, which were apparently lost during the purification of the protein. The A0 domain consists of at least the 7-kDa proteolipid and the 36-kDa subunit for which the genes have not yet been found. In summary, it is proposed that the A1A0 ATPase of Methanosarcina mazei Gö1 contains at least nine subunits, of which seven are located in A1 and/or the stalk and two in A0.

Methanogenesis from H 2 ϩ CO 2 as catalyzed by the nonmarine methanogenic Archaea Methanosarcina barkeri or Methanosarcina mazei Gö1 is obligatorily coupled to the generation of two primary ion gradients at the same time: the reduction of the heterodisulfide of coenzyme M 1 and 7-mercaptoheptanoylthreoninephosphate is coupled to electrogenic translocation of protons across the membrane; in addition the penultimate step of methanogenesis, the transfer of the methyl group from methyltetrahydromethanopterin to coenzyme M as catalyzed by the corrinoid-containing multi-subunit enzyme methyltetrahydromethanopterin:coenzyme M methyltransferase, is coupled to vectorial sodium ion translocation across the membrane (1). M. mazei Gö1 uses both gradients directly as driving force for ATP synthesis but employs two different en-zymes for this purpose: an A 1 A 0 ATP synthase couples ATP formation to ⌬ H ϩ, whereas a F 1 F 0 ATP synthase uses ⌬ Na ϩ as driving force (2). So far, methanogens are the only organisms known to contain two structurally different ATP synthases.
The archaeal A 1 A 0 ATPase 2 shares properties with both, bacterial F 1 F 0 and eucaryal V 1 V 0 ATPases (3,4). It clearly functions as an ATP synthase, which is in accordance with F 1 F 0 but in sharp contrast to V 1 V 0 ATPases; the structure of the proteolipid (one of the subunits of the A 0 domain), which is in the range of 6 -8 kDa in A 1 A 0 and F 1 F 0 ATPases but 16 kDa in V 1 V 0 ATPases, was suggested to be at least one of the reasons for this difference (5)(6)(7)(8). On the other hand, the primary sequences of the subunits A and B of the catalytic A 1 domain are clearly more closely related to vacuolar V 1 V 0 ATPases (9 -12). 3 Therefore, the A 1 A 0 ATPase is regarded as a chimeric protein in which the membrane domain is closely related to F 1 F 0 but the catalytic domain closely to V 1 V 0 ATPases.
Although A 1 A 0 ATPase activity has been demonstrated in a number of Archaea, the subunit composition of this enzyme is far from being settled. It is generally accepted that the A 1 domain contains at least two large subunits of 62-80 kDa (subunit A) and 49 -55 kDa (subunit B) and the A 0 domain contains a least a 7-kDa subunit, the so-called proteolipid. Minor subunits co-purified with the enzyme (5,8,(13)(14)(15)(16)(17), but it has not been established whether these are genuine constituents of the ATPase. Furthermore, little is known about the genetic organization of the A 1 A 0 ATPases. In a first step toward a better understanding of the structure and function of this interesting archaeal enzyme, we purified the A 1 A 0 complex from membranes of M. mazei Gö1 and cloned and characterized the 3Ј end of the A 1 A 0 ATPase operon.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were reagent grade and were purchased form Merck AG, Darmstadt, Germany. K 2 -ATP, DES, N-ethylmaleimide, and NBD-Cl were from Sigma, Deisenhofen, Germany. DCCD was from Aldrich, Steinheim, Germany.
Purification of the A 1 ATPase-Cells were washed once in buffer A (50 mM Tris-HCl, pH 6.9, 10 mM MgSO 4 , 40 mM KCl, 5 mM NaN 3 , 10% (v/v) glycerol) and disrupted in a French pressure cell at a pressure of * This work was supported by grants from the Deutsche Forschungsgemeinschaft (to F. M. and V. M.) and by a grant from the Fonds der Chemischen Industrie (to F. M.). 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) U47274.
Alternatively, purification in small amounts in two days was achieved by EDTA/pyrophosphate treatment, ion exchange chromatography on DEAE-Sephacel, followed by exclusion affinity chromatography on immobilized polyclonal antibodies directed against methyl-CoMmethylreductase of Methanobacterium thermoautotrophicum Marburg. The coupling of the antibodies to activated tresyl-Sepharose was performed as described (22). SDS-PAGE and linear gradient gel electrophoresis were performed following established procedures (23,24).
Purification of the A 1 A 0 ATPase-Cells were harvested in the mid-log phase, washed and suspended in membrane buffer (Tris-HCl, pH 8.0, 20 mM NaHSO 3 , 5 mM MgSO 4 , 40 mM KCl, 0.2 mM dithiothreitol, 0.2 mM EGTA, 10% (v/v) glycerol, 1 M phenylmethylsulfonyl fluoride, 6 mM p-aminobenzamidine) and disrupted in a French pressure cell at a pressure of 138 MPa. Debris was removed by centrifugation (25,000 ϫ g, 10 min). Membranes were pelleted by ultracentrifugation at 134,000 ϫ g for 90 min, washed twice under the same conditions, and solubilized with CHAPS (0.6%) for 20 min under stirring at 35°C. After ultracentrifugation (120,000 ϫ g, 90 min) the supernatant was applied to a DEAE-Sephacel column, equilibrated with buffer B (25 mM Tris-HCl, pH 8.0, 0.2% CHAPS). Fractionation was performed by application of a linear gradient of 0 -0.5 M NaCl in buffer B. Fractions with ATPase activity were pooled and PEG combined, and the ATPase was precipitated by a fractionated PEG 400 precipitation (15% PEG 400 in the first and 30% PEG 400 in the second step). Finally, a saccharose-glycerin gradient centrifugation was performed with 10 -30% saccharose and 5-10% glycerin.
Purification of the Proteolipid-The proteolipid was either extracted from the membranes with organic solvents as described or from the A 1 A 0 complex by hydrophobic chromatography on Phenyl-Sepharose CL-4B and electroendoosmotic preparative gel electrophoresis using the ELFE system from Genofit.
Isolation of Subunits A and B, and Immunization of Mice-Subunits A and B were separated by SDS-PAGE, cut out of the gel, and used to immunize mice. Polyclonal antisera were collected after 65 days, and the immunoglobulins were purified by affinity chromatography on Protein G-Sepharose CL-4B.
Electron Microscopy-Samples containing isolated A 1 A 0 ATPase were negatively stained with 4% (w/v) aquaeous uranyl acetate, pH 4.7, as described (26) and depicted at calibrated magnifications by conventional transmission microscopy.
ATPase Activity-ATPase activity was measured in an assay mixture containing 100 mM MES, pH 5.2, 40 mM KCl, 40 mM NaHSO 3 , 10 mM MgSO 4 , 10% (v/v) glycerol and enzyme solution. After preincubation at 37°C, the reaction was started by addition of ATP to a final concentration of 4 mM. Activity was measured by the release of inorganic phosphate as described (27). Inhibitors were dissolved in water, ethanol, or Me 2 SO and preincubated with the enzyme for at least 20 min at 37°C; controls received the solvent only.
Molecular Procedures-Chromosomal DNA of M. mazei Gö1 was isolated as described (28), restricted, size-fractionated by gradient centrifugation and cloned into either pSE420 or pGem-7Zf-(ϩ). All procedures used were standard techniques (19). Oligonucleotides were synthesized on a Gene Assembler Plus apparatus as recommended by the manufacturer (Pharmacia Biotech Inc.). To clone the A 1 A 0 ATPase genes, a homologous probe of 800 bp was generated by polymerase chain reaction with oligonucleotides corresponding to amino acids   (11). Southern blot analysis of genomic DNA of M. mazei Gö1 restricted with different enzymes revealed a 11-kbp XbaI fragment that hybridized with the homologous probe. This fragment was cloned into pGem-7Zf-(ϩ) and the recombinant plasmid was named pCF1. The XbaI-ClaI fragment of pCF1 was used as a probe to clone the 5Ј upstream region of pCF1 out of a BglII genomic library of M. mazei Gö1. The isolated plasmid (pEW1) contains a 8-kbp BglII fragment ligated into the BamHI site of pGem-7Zf-(ϩ). DNA sequence was determined by the chain termination method of Sanger and analyzed on a VAX computer using the GCG package (29,30).

Characterization of the A 1 A 0 ATPase in Inverted Membrane Vesicles
Since M. mazei Gö1 was recently shown by determining rates of ATP synthesis under various conditions to contain a Na ϩtranslocating F 1 F 0 as well as a H ϩ -translocating A 1 A 0 ATP synthase (2), it was essential to analyze the membrane-bound ATPase activity in more detail. The ATPase activity in washed membranes of M. mazei Gö1 was optimal at pH 5.2. pH values lower than 5.2 decreased the activity very strongly (zero activity at pH 4.5) but at values higher than 5.2 the decrease was less pronounced (22% activity at pH 7.0). The enzyme was stimulated 3-fold by addition of sulfite (40 mM), 1.7-fold by ethanol (18% (v/v)) and 2-fold by glycerol (20% (v/v)). Addition of various divalent cations resulted in different degrees of stimulation. Stimulation by Mg 2ϩ was maximal at 10 mM, a value that corresponds to a Mg:ATP ratio of 2:1. Zn 2ϩ but not Mn 2ϩ stimulated ATPase activity but the degree of stimulation was 1.3 times higher than that measured for Mg 2ϩ stimulation.
Other divalent cations such as Ca 2ϩ , Ni 2ϩ , Cu 2ϩ , and Fe 2ϩ and the monovalent cations Na ϩ and K ϩ had only negligible effects.
The ATPase activity of M. mazei Gö1 membranes was not inhibited by NBD-Cl (1 mM) or by azide (1 mM), typical inhibitors of F 1 F 0 ATPases. Nitrate, a typical inhibitor of V 1 V 0 ATPases, inhibited the enzyme with an I 50 of 30 mM. DCCD and DES both inhibited the enzyme with DES (I 50 100 M) being a more potent inhibitor than DCCD (I 50 500 M). Taken together, the catalytic properties and the inhibitor sensitivity of the ATPase activity at membranes of M. mazei Gö1 are in full accord with the action of an A 1 A 0 ATPase. Apparently, the F 1 F 0 type enzyme does not show ATP hydrolysis activity under these conditions. This is in accordance with our previous notion that the F 1 F 0 type enzyme is active only in the presence of an electrochemical potential across the membrane (2).

Purification of the A 1 ATPase
The A 1 ATPase was released from the membranes by incubation in low ionic strength buffer (EDTA-pyrophosphate depletion) and purified by chromatography on DEAE-Sepharose, DEAE-Sephacel, and gel filtration with an enrichment factor of 16.7 and a yield of 6.4% (Table I. The purified A 1 ATPase had a molecular mass of 360 kDa as determined by gel filtration on Superose 6, and of 350 kDa as determined by native gradient gel electrophoresis (data not shown). When this sample was applied to SDS-PAGE, only two bands representing proteins with molecular masses of 65 (subunit A) and 49 kDa (subunit B) were detected. ATP hydrolysis by this preparation occurred with a V max of 90 milliunits/mg protein and a K m for ATP of 4 mM (data not shown). Unfortunately, the time-dependent loss of enzyme activity was drastic; this made subsequent biochemical studies very laborious.
By EDTA-pyrophosphate depletion and DEAE-Sephacel chromatography small amounts of A 1 ATPase could be enriched within 2 days with a yield of 27%. One additional band was detectable in native gels, which was identified immunologically as methyl-CoM-methylreductase. This contaminating protein was removed by affinity exclusion chromatography with immobilized polyclonal antibodies raised against methyl-CoM-methylreductase from M. thermoautotrophicum Marburg. As detected by SDS-PAGE, the A 1 ATPase purified by this approach contained an additional subunit with an apparent molecular mass of 40 kDa (Fig. 1). However, this approach is not suitable for purification of large amounts of ATPase.

Purification of the A 1 A 0 ATPase
The A 1 A 0 ATPase was solubilized from washed inverted vesicles by treatment with 0.6% (w/v) CHAPS at a protein concentration of 8 -10 mg/ml corresponding to a protein:detergent ratio of 1:0.75 to 1:0.6. After ultracentrifugation 78% of the protein and 106% of the ATPase activity were recovered in the supernatant. The recovery of 106% of ATPase activity is due to a stimulation of ATPase activity by the detergent. The enzyme was enriched 11-fold by DEAE-Sephacel chromatogaphy, PEG 400 precipitation and saccharose-glycerin-gradient centrifugation with a yield of 4.1% and a specific activity of 159 milliunits/mg of protein (Table II). When this sample was applied to a SDS-PAGE polypeptides with apparent molecular masses of 65, 49, 40, 36, 28, and 7 kDa were observed (Fig. 2).

Purification of the Proteolipid
The proteolipid or subunit c of F 1 F 0 , V 1 V 0 , and A 1 A 0 ATPases is very hydrophobic and can be isolated by extraction of membranes with organic solvents. After extraction of membranes of M. mazei Gö1 with chloroform:methanol (2:1), the proteolipid was purified to apparent homogeneity by repeated precipitation with diethylether. In a SDS-PAGE, a single subunit with a molecular mass of 7 kDa was detected. This value corresponds well to the molecular mass of subunit c from F 1 F 0 ATPases. This 7-kDa polypeptide was also present in the A 1 A 0 preparation and could be purified from it by hydrophobic chromatography.

Immunological Relationship to F 1 F 0 and V 1 V 0 ATPases
To further confirm that the purified enzyme represents the A 1 A 0 ATPase of M. mazei Gö1, polyclonal antisera directed against subunit A or B were raised and used as probes in Western blots to test immunological cross-reactivity between various ATPases. Immunological cross-reactivity of subunits A and B was observed with membranes from Methanosarcina frisia, Methanolobus tindarius, and Methanococcus maripaludis but also with vacuolar membranes from sugar beet and Saccharomyces cerevisiae (data not shown). No immunological cross reactivities were detected with the investigated F 1 F 0 ATPases of eubacteria such as E. coli, Clostridium thermosaccharolyticum, and Thiosphaera pantotropha or spinach. Furthermore, cross-reaction of the A 1 ATPase was observed with the anti-subunit B-specific antibodies of the A 1 A 0 ATPase of S. acidocaldarius but not with the anti-subunit ␤-specific antibodies of the F 1 F 0 ATPase of E. coli. These experiments gave further evidence that the ATPase purified from M. mazei Gö1 is of the A 1 A 0 type.

Ultrastructure of the A 1 A 0 and the A 1 ATPase from M. mazei Gö1 as Revealed by Electron Microscopy
In electron micrographs the A 1 A 0 ATPase appeared to be composed of a base, a stalk, and a head part (Fig. 3). Often, several A 1 A 0 ATPases were aggregated in opposite orientation, brought about by interaction of their A 0 parts. The isolated A 1 parts of the enzyme showed a rotational symmetry and a size similar to that observed for the mitochondrial F 1 ATPase and the A 1 ATPase from S. acidocaldarius.

Nucleotide Sequence and Structure of the aha Operon
Coding for the A 1 ATPase from M. mazei Gö1 Two fragments (pCF1 and pEW1) containing A 1 A 0 ATPase genes were cloned (see "Experimental Procedures"). Biochemical and molecular studies (see below) revealed the presence of the following genes (in that order) in one gene cluster: ahaE, C, F, A, B, D, and G (Fig. 4). The transcriptional analysis, which will be presented in a separate publication, showed that these genes are organized in an operon; this was named aha operon for "archaeal H ϩ ATPase". Upstream of ahaE is an AT-rich region, which contains potential archaeal consensus promotor sequences. Immediately downstream of ahaG is a potential rho-independent transcriptional terminator. The intergenic region is thymine-rich and contains seven tandem repeats of a 22-bp sequence (TTTTTCCTGAATTGTGTTATTG), which might function as a transcriptional terminator (Fig. 5). 497 base pairs downstream of the stop codon of ahaG is the start of an open reading frame whose deduced amino acid sequence is 39% identical and 61% homologous to HypF from E. coli, which is involved in maturation and/or assembly of the NiFe-hydrogenase. Therefore, it is assumed that this open reading frame represents the hypF homologous gene of M. mazei Gö1. Characteristic translational features of the sequence are summarized in Table III. Putative rho-independent transcription terminators are indicated by the bold arrows above the sequence. Tandem repeats and partial sequences thereof are boxed, and the first base is indicated by the arrow above. Putative promotor sequences (Box A and B) are boxed, and conserved residues are marked by asterisks. A putative SD sequence is underlined.

Identification of aha Genes
The program TESTCODE categorized the region from ahaE through G as a coding sequence above the 95% confidence level. The program CODONPREFERENCE also recognized all designated genes as authentic genes. Furthermore, the apparent molecular masses of the 65-, 49-, 40-, and 25-kDa subunit of the purified enzyme correspond well to the deduced molecular masses of AhaA, B, C, and D.

Properties of the Gene Products and Similarity to Polypeptides of V 1 V 0 and F 1 F 0 ATPases
The properties of the aha gene products are summarized in Table IV. Generally, ahaE through G code for hydrophilic polypeptides. Data base searches, primary sequence alignments, secondary structure predictions, and comparisons of the molecular masses and isolelectric points were used to identify homologous subunits in V 1 V 0 and F 1 F 0 ATPases.
AhaA-The deduced molecular mass of AhaD (66.106 kDa) corresponds well to the experimentally derived molecular weight of 65,000 of subunit A of the purified enzyme. The deduced gene product is 86% identical to the corresponding subunit from M. barkeri (11) and 63 and 52% of the residues are identical to subunit A of the A 1 A 0 ATPase from Haloferax volcanii (32) and S. acidocaldarius (9), respectively (Fig. 6). Still 50 -55% of the residues are identical to subunit A of the V 1 V 0 ATPase of Enterococcus hirae (33), Vma1p of S. cerevisiae (34), and subunit A of bovine (35) and human (36) V 1 V 0 ATPase. On the other hand, only 27% of the residues are identical to subunit ␤ of the F 1 F 0 ATPase of E. coli (37). AhaA contains the Walker motifs A and B (38), which are part of the nucleotide binding domain, indicating that it represents the catalytic subunit. This is in agreement with the finding that anti-subunit A specific-antibodies were 4 times more effective in inhibiting ATPase activity of the purified enzyme from M. mazei Gö1 than anti-subunit B-specific antibodies. From the multiple alignment, it is evident that all of the residues involved in nucleotide binding in the mitochondrial F 1 F 0 ATPase (39) are conserved in the M. mazei Gö1 A 1 A 0 ATPase, and therefore this general scheme seems also valid for nucleotide binding in subunits B and A of A 1 A 0 and V 1 V 0 ATPases.
AhaC-ahaC codes for a polypeptide with a molecular mass of 41.486 kDa, which corresponds well to the experimentally derived value of 40 kDa of subunit C of the purified enzyme. By data base searches, NtpC of E. hirae V 1 V 0 ATPase (22% identity; 49% homology) (42) and AtpC of H. volcanii A 1 A 0 ATPase (GenBank accession no. X79516) (35% identity, 58% homology) were found to be homologous. A multiple alignment of these polypeptides to the 39-kDa accessory protein (Vma6p) of V 1 V 0 ATPases (43, 44) revealed a number of homologous residues indicating that these proteins are homologous (Fig. 8). Nine residues seem to be invariant, five of these are located in the C-terminal region, which in addition has several conserved substitutions. Of these five residues, three are charged and positioned in a predicted ␤-sheet structure. Apparently, there is no homologous polypeptide in F 1 F 0 ATPases.
AhaD-Data base searches revealed the following proteins to be homologous to AhaD (23.900 kDa): NtpD of E. hirae V 1 V 0 ATPase (37% identity) (42), AtpG of S. acidocaldarius A 1 A 0 ATPase (28% identity) (45), the 34-kDa subunit of bovine (26% identity) (46), Vma8p of yeast (28% identity) (46) V 1 V 0 ATPase, and a deduced polypeptide from Caenorhabditis elegans (YMF2; GenBank accession no. Z27080; Ref. 47). 21 and 26% of its residues are identical to subunit ␥ of F 1 F 0 ATPases. Multiple alignments revealed that the N and C termini are well conserved (Fig. 9). Secondary structure analysis predicts that these regions are ␣-helical, which is a striking homology to subunit ␥ of F 1 F 0 ATPases (37,48). Thus, we suggest that AhaD and the corresponding subunits in V 1 V 0 and A 1 A 0 ATPases are the homologues of subunit ␥ of F 1 F 0 ATPases. However, there are no conserved residues in these polypeptides.
AhaE-AtpD (21.981 kDa) of H. volcanii A 1 A 0 (33% identity) (GenBank accession no. X79516), NtpE (22.900 kDa) of E. hirae V 1 V 0 ATPase (17% identity) (42), AtpD of S. acidocaldarius A 1 A 0 (which is only 13 kDa), and Vma4p and the homologous 31-kDa subunit of V 1 V 0 ATPases of various organisms (49,50)   were identified by BLAST as the homologues of AhaE. Secondary structure analysis predicts that AhaE and its homologues are largely ␣-helical. A multiple alignment shows that the region Gly-145 to Leu-149 is well conserved between prokaryotic and eukaryotic subunits (Fig. 10). According to its size, AhaE could be the homologue of subunit ␦ of F 1 F 0 ATPases. This is in agreement with the observation that the aforementioned residues are also conserved in subunit ␦ of E. coli F 1 F 0 (37). AhaF-Data base searches identified only AtpE of H. volcanii A 1 A 0 ATPase (GenBank accession no. X79516) to be homologous (11.582 kDa; 50% identity). However, its physical loca-tion within the operon and its size suggest that AhaF is the homologue of NtpG of E. hirae V 1 V 0 (42), the 14-kDa subunit of V 1 V 0 ATPases of tobacco hornworm (51), Drosophila melanogaster (Genbank accession no. Z26918), Vma7p of S. cerevisiae (52), and subunit ⑀ of F 1 F 0 ATPases (37). Multiple alignments showed that the N terminus is fairly well conserved (Fig. 11), and the polypeptide is predicted to have a small ␣-helical region at its N terminus and a ␤-sheet region at its C terminus.
AhaG-Of the polypeptides tested, AhaG is the only one which did not reveal a homologous polypeptide by data base search. However, according to its size (6135 kDa) and its location within the operon we suggest that AhaG, NtpH of E. hirae V 1 V 0 (7.138 kDa; 12% identity, 40% homology) (42) and AtpE of S. acidocaldarius A 1 A 0 ATPase (7.038 kDa; 17% identity, 41% homology) (45) are homologous. A multiple alignment of AhaG, NtpH of E. hirae V 1 V 0 , and AtpE of S. acidocaldarius A 1 A 0 ATPase (Fig. 12) revealed only one conserved residue . No homologue was found in eucaryotic V 1 V 0 ATPases.

Structure of the A 1 A 0 ATPase and Possible Function of Sub-
units-It is apparent from the electron micrographs that the A 1 A 0 ATPase consists, like the F 1 F 0 and the V 1 V 0 ATPase, of a hydrophilic (A 1 ) and a hydrophobic domain (A 0 ), which are connected by a stalk. From the combined biochemical and molecular results, we suggest that this structure is built up by at least 9 non-identical subunits used in different stoichiometry (Fig. 13). In view of the apparent similarities, the discussion of the structure/function prediction of individual subunits of the A 1 A 0 ATPase will be done on the basis of the known features of F 1 F 0 and V 1 V 0 ATPases (53)(54)(55)(56). Subunits A and B were both present in the purified enzyme; they share extensive homologies with Vma1p/subunit A and Vma2p/subunit B of V 1 V 0 ATPases and with ␤ and ␣ of F 1 F 0 ATPases. According to the apparent homology, it is suggested that subunits A and B of the A 1 A 0 ATPase are present in triplicate each and are arranged alternately in an orange-like structure. This is supported by the rotational symmetry determined in the electron micrographs. AhaA carries the catalytic nucleotide binding sites, whereas the non-catalytic sites are located on AhaB. All of the residues important for nucleotide binding in ␣ and ␤ of F 1 F 0 (39) are conserved in A and B of A 1 A O 3 . The proline-rich hydrophobic sleeves suggested to be involved in ␥-␤ interaction in F 1 F 0 are also conserved in AhaA (residues 344 -351) and B (residues 265-273). The catalytic subunit A and the regulatory subunit B are more closely related to the corresponding subunits of V 1 V 0 than F 1 F 0 ATPases. This has been observed before with the corresponding subunits from A 1 A 0 ATPases of other organisms (9 -12). However, the degree of identity was higher to the corresponding subunits from the halophilic Archaea H. salinarium and H. volcanii than to the thermoacidophile S. acidocaldarius.
AhaD is identical to the 28-kDa subunit of the purified protein; it is homologous to subunit ␥ of F 1 F 0 ATPases, which plays a crucial role in transmitting energy from the F 0 to the F 1 domain (57). The x-ray structure revealed that subunit ␥ is highly ␣-helical at its N and C termini, which both protrude in the orange-like head assembly of the catalytic and non-catalytic subunits (39). A secondary structure analysis strongly predicts ␣-helical N and C termini also in AhaD. A multiple alignment did not reveal conserved residues (Fig. 10). Furthermore, the region including Gln-270, Thr-274, Glu-276, and Glu-279, which were shown by mutagenesis studies to be important for function in E. coli F 1 F 0 ␥ (58) is not present in the counterpart of prokaryotic V 1 V 0 and A 1 A 0 ATPases, indicating that not particular residues but rather a secondary structure is important for function. The fact that this subunit is only present in the A 1 A 0 rather than in the A 1 preparation might suggest a structural (and functional?) interaction also with the membrane domain. On the other hand, it is conceivable that its non-appearance in the A 1 preparation might be due to preparation artifacts. However, the available data indicate that the general structure and function of subunits A, B, and D of A 1 A 0 and V 1 V 0 ATPases are comparable to subunits ␤, ␣, and ␥ of F 1 F 0 ATPases.
AhaC is identical to the 40-kDa subunit of the purified enzyme; apparently it has no counterpart in F 1 F 0 ATPases. The eukaryotic homologue of AhaC (Vma6p or the 39-kDa accessory protein) is hydrophilic as is AhaC. However, since Vma6p copurifies with the V 0 domain, it is assumed to bind to the membrane via interaction with a membrane-bound subunit (44) and probably is part of the stalk. Since AhaC was found in the A 1 prepaparation detached from membranes of M. mazei Gö1, we regard this subunit as part of the A 1 domain or the stalk.
AhaE through G were apparently not present in the purified A 1 A 0 ATPase, although it cannot be excluded that they did not stain well. From the multiple alignment, it is suggested that AhaE is the homologue of Vma4p or the 31-kDa subunit of eukaryotic V 1 V 0 and subunit ␦ of F 1 F 0 ATPases, and AhaF and its homologues are suggested to be the homologues of subunit ⑀ of F 1 F 0 ATPases. However, in both cases the structure and function of the subunits are not well understood. Whether AhaG plays a structural role in the A 1 A 0 ATPase remains to be established. AhaG has no counterpart in eukaryotic V 1 V 0 ATPases. In this context it is interesting to note that there is a small gene (uncI) as the first gene in the unc operon, which codes for the F 1 F 0 ATPase. The function of the gene product is obscure but it is not essential for an active F 1 F 0 ATP synthase complex. Based on size, AhaG and its bacterial homologues could very well be the homologues of the uncI gene product.
The fact that the 36-kDa subunit co-purifies with the complex only after solubilization of the membrane is indicative for its membrane localization. Therefore, we suggest that the A 0 domain consists of at least two subunits of M r 7,000 (the proteolipid) and 36,000. In yeast V 1 V 0 three subunits (excluding Vma6p) are present in the membrane domain: the 16-kDa proteolipid (59), a recently described 13-kDa protein with homologies to subunit b of F 1 F 0 ATPases (60), and a 115-kDa protein (61).
Structure of the aha Operon-There is a striking homology between the aha operon of M. mazei Gö1, the ntp operon of E. hirae (42) and the atp operon of S. acidocaldarius (45) and H. volcanii (GenBank accession no. X79516) (Fig. 14). Regarding the hydrophilic subunits, the arrangement of ahaE through G of M. mazei Gö1 is identical to the corresponding genes in the eubacterium E. hirae and the homologues of ahaE through B are also conserved in the same order in H. volcanii. However, the organization of the genes coding for the hydrophobic sub- units is different. A NtpJ homologue is not present in M. mazei Gö1. This is expected since the ATPase of E. hirae is believed to function as a K ϩ /H ϩ antiporter (62) and NtpJ is believed by sequence similarities to play a role in this exchange (42). There is no more aha gene downstream of ahaG, but preliminary expression studies indicate that there are more genes in the aha operon upstream of ahaE; sequencing of this region is currently under way.