Identification and Characterization of a Novel 9.2-kDa Membrane Sector-associated Protein of Vacuolar Proton-ATPase from Chromaffin Granules*

Vacuolar proton-translocating ATPase (holoATPase and free membrane sector) was isolated from bovine chromaffin granules by blue native polyacrylamide gel electrophoresis. A 5-fold excess of membrane sector over holoenzyme was determined in isolated chromaffin granule membranes. M9.2, a novel extremely hydrophobic 9.2-kDa protein comprising 80 amino acids, was detected in the membrane sector. It shows sequence and structural similarity to Vma21p, a yeast protein required for assembly of vacuolar ATPase. A second membrane sector-associated protein (M8-9) was identified and characterized by amino-terminal protein sequencing.

Proton-translocating adenosine triphosphatases have fundamental roles in energy conservation, secondary active transport, the acidification of intracellular compartments, and cellular pH homeostasis. They fall into three broad classes, called F, P, and V (1), of which the vacuolar type (V-ATPases) 1 is both the most recently recognized and the least well characterized. ATPases of this class occur in endomembranes bounding the acidic compartments of animal, plant, and fungal cells (2) and also in the plasma membranes of some specialized cell types. They have been purified from several mammalian sources, including adrenal secretory vesicles (3,4), brain clathrincoated vesicles, (5,6), and kidney medulla microsomes (7), as well as from the vacuoles of fungi and higher plants. Most V-ATPases contain some 6 -10 different subunits (2), but subunit composition depends on the source of the enzyme, and tissue-specific isoforms exist (8). The V-type ATPases are structurally similar to those of the F-type, having a transmembrane proton-conducting sector and an extramembrane catalytic sector. By analogy with the two sectors of F-ATPases (9 -12), these are termed V 0 and V 1 , respectively. For a recent review, see Ref. 13.
In this work, the recently developed technique of blue native polyacrylamide gel electrophoresis (BN-PAGE; Refs. 14 -17) was employed to purify vacuolar ATPase holoenzyme (V 1 V 0 ) and free membrane sector (V 0 ) simultaneously from adrenal secretory vesicle membranes. Combined with high resolution Tricine-SDS-PAGE in the second dimension, the subunit composition, particularly with respect to small polypeptides, was determined. Two novel proteins, 8 -9 and 9.2 kDa in size, were found in the membrane sector. Here we report the detailed analysis of the larger of these two polypeptides.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and T4-DNA ligase were obtained from New England Biolabs. Taq DNA polymerase was from Stratagene, and TA Cloning Kit ® was from Invitrogen. Sequenase version 2.0 sequencing kit, [␣- 35 S]dATP and Hybond N ϩ membranes were obtained from Amersham Pharmacia Biotech. ABI Prism™ dye terminator cycle sequencing kit was purchased from Perkin Elmer. The cDNA library from bovine adrenal medulla was a kind gift from Leonora Ciufo (University of Edinburgh, Edinburgh, Scotland, United Kingdom). Human EST clone ID 143553 (GenBank™accession number R75754) was obtained from the IMAGE Consortium (18). Bovine tissues were frozen in liquid nitrogen several minutes after the death of the animal. The probe against human glyceraldehyde-3-phosphate dehydrogenase was a kind gift from J. Altschmied (Physiologische Chemie I, Universität, Wü rzburg). Anti-subunit G 1 antibody was kindly provided by Bill P. Crider (University of Texas Southwestern Medical Center, Dallas).
Isolation of V 0 and V 1 V 0 ATPase from Chromaffin Granule Membranes by BN-PAGE-Chromaffin granule membranes were prepared according to Apps et al. (19). The membranes (11 mg of protein in 1.5 ml of 10 mM Hepes/NaOH, pH 7.4) were solubilized at 4°C by addition of 1 ml of 1.75 M 6-aminohexanoic acid, 50 mM BisTris-Cl, pH 7.0, and 500 l of 10% dodecyl maltoside. After 30 min centrifugation at 100,000 ϫ g, 200 l of 5% Serva Blue G in 500 mM 6-aminohexanoic acid was added to the supernatant. One ml of supernatant was loaded onto each of three 3-mm-thick preparative 5-13% acrylamide gradient gels for BN-PAGE (14). After BN-PAGE, the blue bands were excised and the native complexes electroeluted. About 300 g of V 1 V 0 complex and 900 g of V 0 membrane sector were recovered from 11 mg of membrane protein.
Isolation of V 0 and V 1 V 0 ATPase from a Triton X-114 Extract-Triton X-114-extraction was used for enrichment of V 0 and V 1 V 0 -ATPase (20). Chromaffin granule membranes (1 mg protein in 0.17 ml of 10 mM Hepes/NaOH, pH 7.4) were centrifuged for 30 min at 100,000 ϫ g. The pellet was resuspended in 0.2 ml 150 mM KCl, 10 mM Tris, pH 7.5, and solubilized by addition of 50 l 10% (w/v) Triton X-114. V 0 and V 1 V 0 * This work was supported by Grant SFB 472 from the Deutsche Forschungsgemeinschaft (to H. S.). 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.
were precipitated by a 15-min incubation on ice. After a 10-min centrifugation at 100,000 ϫ g, the pellet was washed with 0.5 ml of 10 mM Na ϩ /Mops, pH 7.2, and centrifuged as before. The pelleted proteins were solubilized by addition of 90 l 1 M 6-aminohexanoic acid, 50 mM BisTris/HCl, pH 7.0, and 21 l of 10% dodecyl maltoside, and centrifuged again for 10 min at 100,000 ϫ g. After addition of 10 l of 5% Serva Blue G in 500 mM 6-aminohexanoic acid, 50 l were applied to 10-mm gel wells for analytical BN-PAGE. After blue native electrophoresis, individual lanes were cut from the gel and processed in a second dimension by Tricine-SDS-PAGE. Electrophoretic techniques, staining techniques, and densitometric quantification followed the protocols described previously (17,21).
Screening of a Bovine cDNA Library by PCR-The NH 2 -terminal amino acid sequences of bovine M9.2 and the sequences of two corresponding human cDNA clones, IMAGE consortium clone 143553 (Gen-Bank™ accession number R75754; Ref. 18) and murine MM85D12 (GenBank™ accession number D21772; Ref. 29) were used to deduce a pair of degenerate primers for PCR with the plasmid DNA of the whole bovine adrenal medulla cDNA library: VATPB9.2c, 5Ј-AT(C/T) GTG ATG AGC GTG TTC TGG GG-3Ј; and VATPB9.2n, 5Ј-GCC AAA AIA GAT AGC AGC AIA C-3Ј. PCR was performed in 50 mM KCl, 1.5 mM MgCl 2 , 0.1% gelatin, 200 M of each dNTP, 0.5 M of each primer, 10 mM Tris/HCl, pH 8.8. The temperature cycle was as follows: 94°C for 1 min, 42°C for 1 min, 70°C for 30 s for 30 cycles and a single step of 72°C for 10 min. The PCR product was cloned into the pCR II vector (Invitrogen) and sequenced. A third perfect match primer was deduced from this sequence: VATPB9.2a, 5Ј-GGG GCA TCG TCG GCT TCC TGG TGC-3Ј. The bovine cDNA library was then screened by PCR using the combination of primers VATPB9.2a and VATPB9.2n and the same temperature profile. Six pools, comprising a total of 1500 colonies, were taken as the template for PCR, and examined for the occurrence of a 106-bp PCR product. The positive pool was divided into subpools, and the procedure was repeated until a single clone (BVATPM9.2) was obtained.
Sequencing of Human and Bovine Clones-The insert from clone BVATPM9.2 was cut out with BamHI and cloned into pBluescript™ II SK(Ϫ). The new clone pBBM9.2 was subcloned by using the BstEII site at nt 69 and the XbaI site at nt 246. The insert of human cDNA clone 143553 was cut out with EcoRI and HindIII and cloned into pBlue-script™ II SK(Ϫ). Clones pBBM9.2 and pBHM9.2 were sequenced in both directions.
RNA Isolation and Northern Blotting-Total RNA was prepared according to the method of Chomczynski and Sacchi (30). RNA was separated by formaldehyde agarose gel electrophoresis using 5 mM sodium acetate and 0.1 mM EDTA in running and loading buffers, capillary-blotted on Hybond N ϩ membranes (31), and fixed by UV irradiation. DNA probes were labeled with [␣-32 P]dCTP by random priming, and QuikHyb ® solution from Stratagene was used for hybridization. A 900-bp cDNA was excised from pBBM9.2 with BamHI and used as a probe for M9.2. The probe against bovine V 1 V 0 -ATPase subunit c (proteolipid c), GenBank™ accession number J03835 (32), was made by PCR using primers BVch 5Ј-TCA GCC GCC ATG GTC TTC AG -3Ј and BVcr 5Ј-CGG CGA AGA TGA GGA TGA GG-3Ј. Using the bovine adrenal medulla cDNA library as a template, a 358-bp fragment corresponding to positions 190 -547 of V-ATPase proteolipid c was amplified by PCR, tested by restriction analysis, and cloned into pCR 2.1 (Invitrogen).

Characterization of V-ATPase by BN-PAGE and Two-dimen-
sional Electrophoresis-Analytical BN-PAGE ( Fig. 1) was used for separation of V 1 V 0 holocomplex and free V 0 -membrane sector from solubilized chromaffin granule membranes (lane M), and from a fraction prepurified by Triton X-114 extraction/ precipitation (lane P). The oxidative phosphorylation complexes from solubilized bovine heart mitochondria served as molecular mass standards (17,33,34). The molecular masses assigned to complex I and complex V are minimal values, inasmuch as the copy number of some subunits is not exactly known (35,36). The prominent band with an apparent mass around 440 kDa was identified as V 0 -membrane sector by the characteristic polypeptide patterns in two-dimensional electrophoresis, and by amino-terminal protein sequencing (see below). A faint protein band with an apparent mass of about 1000 kDa was identified as holo V 1 V 0 -ATPase. The position of free V 1 sector is also indicated in Fig. 1, although the amounts were too low for detection in BN-PAGE (see below).
Second-dimensional SDS-PAGE of lane M from BN-PAGE (cf. Fig. 1) revealed the characteristic polypeptide patterns of V 1 V 0 holoenzyme and V 0 membrane sector. Additionally, minor amounts of free V 1 sector and some contaminating mitochondrial F 1 F 0 -ATP-synthase could be detected (Fig. 2, upper panel). F 1 F 0 was removed by Triton X-114 extraction/precipitation (Fig. 2, lower panel). Using the migration distances of V 1 V 0 (Ϸ1000 kDa), F 1 F 0 (Ն550 kDa), and V 0 (Ϸ440 kDa) for calibration, the apparent mass of free V 1 sector was estimated at around 500 kDa. Several staining maxima of the 14-kDa proteolipid c in addition to those at the positions of V 1 V 0 and the major band of V 0 indicated the positions of minor amounts of V 0 sector in higher oligomeric states.
The molar ratio of V 0 /V 1 V 0 in isolated chromaffin granule membranes was deduced from Coomassie Blue-stained twodimensional gels by densitometric quantification of proteolipid c and 115-and 39-kDa proteins (data not shown). A 5-fold molar excess of free V 0 sector over assembled holocomplex was determined (cf. "Discussion"). Variation of the detergent/protein ratio for membrane solubilization by Ϯ50% had no effect on the V 0 /V 1 V 0 ratio, which was 5.1 Ϯ 0.3 (n ϭ 5).
Identification of Protein Subunits of V-ATPase and Membrane Sector-Analysis of polypeptide composition by SDS-PAGE was performed directly from lanes of BN-PAGE (Fig. 2) FIG. 1. Separation of V 1 V 0 ATPase and V 0 membrane sector by BN-PAGE. Solubilized chromaffin granule membranes (180 g of total protein; lane M) and a fraction enriched in V-ATPase, prepared from chromaffin granule membranes (420 g of total protein; lane P), were applied to a 5-13% acrylamide gradient gel and resolved by BN-PAGE. The five oxidative phosphorylation complexes from solubilized bovine heart mitochondria (150 g of total protein) were used for molecular mass calibration (lane BHM). The position of V 1 is indicated, although it was detected only after two-dimensional resolution (Fig. 2). and after electroelution of the complexes from blue native gels (Fig. 3).
Direct application of the two-dimensional technique had the disadvantage that the protein amounts were limited by the maximum load applicable to the first-dimension native gel. The presence of the novel M9.2 and M8-9 proteins (apparent mass in SDS-PAGE of 13 and 8 -9 kDa, respectively) in V 1 V 0 holoenzyme could therefore be detected only with prolonged silver staining (data not shown). However, because subunits of V 1 V 0 and V 0 complexes appeared as clearly recognizable columns of bands in the two-dimensional gels, a smearing 75-kDa band could be easily identified as the only major contaminant of both complexes (Fig. 2). This contaminant was identified as dopamine-␤-monooxygenase (GenBank™ accession no. P15101) by amino-terminal sequencing (data not shown). V 1 V 0 and V 0 complexes which were electroeluted from preparative blue native gels and resolved by SDS-PAGE (Fig. 3), revealed essentially the same polypeptide patterns as observed with the direct two-dimensional technique. However, at the higher protein loading, the low molecular mass bands in the holoenzyme now were easier to detect, and apparent molecular masses could be assigned using bovine bc 1 complex subunits for calibration (33).
The polypeptide number and masses in the 29 -115-kDa range (37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47) matched those for other V-ATPase preparations (3)(4)(5)(6). Smaller protein subunits in the range between 29 kDa and proteolipid c (32) have already been described, e.g. subunit G (48), which has also been termed M16 (49), and M20 (50) in the membrane sector. Anti-subunit G 1 antibody reacted with the 15-kDa band of V 1 V 0 , but did not detect any protein of the V 0 sector (cf. "Discussion"). The unidentified 16-kDa band of the V 0 sector therefore was tentatively assigned to M20. Only one protein component running below proteolipid c in SDS gels, namely subunit F, has been identified so far (51). We could detect more components in this low molecular mass range; M9.2 protein with an apparent mass of 13 kDa, and M8-9 protein, represented by a stack of four bands in the 8 -9-kDa range, were identified in the membrane sector and in the holoenzyme. According to the electrophoretic mobility, the unidentified 12-kDa protein was tentatively assigned to subunit F.
Comparison of 10%, 13%, and 16.5% acrylamide gels led to the identification of anomalous migration behavior of some subunits, which is often observed with hydrophobic membrane proteins. The M9.2 membrane sector-associated protein had an apparent mass around 10 kDa in 10% gels, which shifted to 13 kDa in 16.5% gels. In 16.5% acrylamide gels (Fig. 3), the M9.2 protein appeared as a diffuse background to the sharp 12-kDa band, but the 15-kDa subunit G and the 16-kDa band were resolved. Subunit G and the 16-kDa band comigrated in 13% acrylamide gels, but the M9.2 and 12-kDa proteins were separated (data not shown).
13% acrylamide gels were preferentially used for the densitometric quantification of Coomassie stain intensities of V 0 and V 1 V 0 subunits, and for determination of their staining ratios relative to the M39 subunit (Table I) Fig. 1 comprising separated multiprotein complexes from chromaffin granule membranes was resolved by Tricine-SDS-PAGE, using a 16.5% acrylamide gel. Coomassie Blue G 250 stain was used for detection (upper panel). Lane P from Fig. 1 starting from a fraction enriched in V-ATPase was resolved and silver-stained (lower panel). Apparent molecular masses were assigned to subunits of the V 0 membrane sector (right side), to V 1 sector subunits (upper panel, left side) and to all detectable subunits of the holoenzyme (lower panel, left side). In addition to a major form of the V 0 membrane sector, the positions of minor forms with higher oligomeric states are indicated.
FIG. 3. Subunit composition of V 1 V 0 ATPase and V 0 membrane sector. The multiprotein complexes recovered by electroelution from preparative blue native gels were resolved by Tricine-SDS-PAGE using a 16.5% acrylamide gel. Bovine heart bc 1 complex subunits (33) were used as molecular mass standards. Asterisks (*) mark subunits which were not verified by amino acid sequencing or Western blotting (cf. "Results").

9.2-kDa Protein of V-ATPase from Chromaffin Granules
M9.2 and M8-9 relative to M39 were almost identical in the V 0 membrane sector and in the V 1 V 0 holoenzyme which indicates almost identical stoichiometries in V 0 and V 1 V 0 . However, it was not possible to decide whether these proteins are present in stoichiometric or substoichiometric amounts, because Coomassie Blue staining intensities are not reliable indices of copy number. Subunit G and proteolipid c in the holoenzyme had comparable stain intensities, which might also indicate a higher copy number for subunit G.
Amino-terminal Protein Sequences-The complexes resolved by preparative BN-PAGE were electroeluted, and the protein subunits resolved by Tricine-SDS-PAGE and electroblotted onto PVDF membranes for direct amino-terminal protein sequencing (Fig. 4). Only a few of the proteins had free amino termini accessible to Edman degradation. Among these proteins were the major bovine brain subunit B, identified by the sequence MRGIVNGAAPELPV (39,40); M45, also called glycoprotein IV or Ac45 protein (41,42); and proteolipid c (32). However, more than 90% of proteolipid c appeared to be aminoterminally blocked, because the signal intensities of phenylthiohydantoin amino acids from cyanogen bromide fragments were up to 10-fold higher than after direct sequencing. The novel M9.2 and M8-9 proteins were also directly accessible to Edman degradation (cf. Table II). The amino-terminal sequences obtained from the four bands of the M8-9 protein (Fig.  3) suggested that M8-9 might be present in a "full length" form (largest band 1) and three amino-terminally shortened forms (smaller bands 2-4).
Subunit A (38) was identified after cleavage at tryptophan by iodosobenzoic acid, subunit C (46) after hydroxylamine cleavage between asparagine and glycine, subunits D (47), E (45), proteolipid c (32), and M45 (41, 42) after cyanogen bromide cleavage, M115 (37) after partial acidic hydrolysis, and M39 (43, 44) after use of deacylating conditions. The sequences obtained from V 0 subunits are summarized in Table II. We could not obtain internal protein sequences from the protein with an apparent mass of 16 kDa, because this protein tends to aggregate during electrophoresis, and is hardly transferred to PVDF membranes by electroblotting. These properties seem to indicate a hydrophobic membrane protein. We assume that this protein represents subunit M20 (50).
Primary Structure and Properties of the M9.2 Protein-TFASTA computer searching using the amino-terminal protein sequence of the bovine M9.2 protein revealed a high degree of homology to several human as well as one murine cDNA clone. The function of these proteins was not known. The M9.2 cDNA from one of the human cDNA clones, IMAGE Consortium Clone 143553 (GenBank accession no. R75754), was sequenced (Fig.  5). The sequence around the initiator codon matches exactly the optimal sequence for initiation by eukaryotic ribosomes ACCATGG as described by Kozak (52). The sequenced M9.2 cDNA clone from a bovine adrenal medulla cDNA library was incomplete (Fig. 5). However, the full bovine M9.2 protein sequence, except the amino acids at positions 4, 14, and 17, was obtained by Edman degradation (Fig. 4A), which also showed that the amino-terminal methionine residue was processed in the mature protein. The almost perfect conservation between human and murine proteins, which differed only at position 22, strongly suggests that the three unidentified residues may be conserved in the bovine protein as well. In this case, the bovine protein would be completely identical to the human protein.
The human M9.2 protein has a calculated molecular mass of 9.243 kDa if processing of the amino-terminal methionine is assumed, as in the bovine protein. It is an extremely hydrophobic membrane protein, with a polarity index of 22.5%, according to Capaldi and Vanderkooi (53) A computer search revealed a potential glycosylation site, NET, at positions 70 -72 (Fig. 4A), but glycopeptidase F (Sigma) had no effect on M9.2, whereas M45 was deglycosylated in a parallel experiment (data not shown).
A sequence motif CSVCC (positions 44 -48), similar to those of potential metal-binding proteins (54), is conserved in the protein from all known mammalian sources. It is located at the center of the second hydrophobic stretch.
Apart from a partial cDNA sequence for the rat protein (GenBank accession no. H32025), homologous sequences in unidentified reading frames on chromosome IV of Caenorhabditis elegans (accession no. Z68227) and on chromosome III of Drosophila melanogaster (accession no. L07835) were also found by computer searching (Fig. 4B). The presumed protein of C. elegans was deduced by translation of the joined segments 27704 -27827, 27876 -27934, and 28058 -28165 as indicated in the annotations to the sequence with accession no. Z68227. The presumed protein of D. melanogaster was deduced by translating the DNA-sequence from position 7957-8223 in reverse direction from the sequence with the accession no. L07835. The deduced C. elegans and D. melanogaster protein sequences share 44% and 35% identity, and 72% and 62% similarity, respectively, with the human protein. Transmembrane regions and hydrophobicity distribution (25,28,55) predicted for the deduced translation products of the unidentified reading frames from C. elegans and D. melanogaster are very similar to those for the mammalian M9.2 proteins (data not shown).
In the C. elegans and D. melanogaster sequences, a stretch with high similarity to the human CSVCC sequence motif is present. It comprises a doublet of cysteines, but cysteine 44 is not retained.
Data base searching using the TFASTA program revealed no further significant homologies; however, direct comparison with all known subunits of the bovine and yeast F and V-ATPases and with assembly factors for yeast V-ATPase led to TABLE I Ratios of Coomassie stain intensities of V 0 and V 1 V 0 subunits The data were obtained by densitometric quantification of different V 0 and V 1 V 0 preparations resolved by 13% acrylamide Tricine-SDSgels, and supplemented by data (in parentheses) from 16.5% acrylamide gels. Stain intensities (arbitrary units) were divided by the molecular masses of the individual subunits, and normalized to M39. Asterisks (*) mark subunits that were not verified by amino acid sequencing or Western blotting (cf. "Results").  (56), expressed by a score of 45% sequence similarity and 19% identity between the two proteins (Fig. 6A), and by the similarity of predicted transmembraneous helices (Fig. 6B). When the Vma21p sequence in turn was used for a data base search, no human cDNA clones were found, although a huge number of human EST-sequences are deposited in the data bases. Therefore, it seems likely that there is no human protein with significantly higher homology to the yeast Vma21p than M9.2.  DPSTTYNLAYKYNFEYPVVFNLVL a Fragments were not separated. The initial triple sequence could be arranged according to the fragments expected after cleavage of the 116-kDa polypeptide (37) at three aspartic acid-proline bonds.
b The amino-terminal sequence confirms that the coding region extends 231 nucleotides upstream from the initially reported start codon (43) as suggested by Bauerle et al. (44) by comparison with the yeast homologue Vma6p.
c The 16 kDa protein was tentatively assigned to M20 (cf. "Results"). d Less than 10% of proteolipid c was accessible to direct Edman degradation (cf. "Results"). e Band 1 is the largest of the four M8 -9 protein bands. The sequence stretch shared by bands 1-4 is underlined.

9.2-kDa Protein of V-ATPase from Chromaffin Granules
Tissue Distribution of M9.2 mRNA-A Northern blot using RNA from various bovine tissues was hybridized with a 32 Plabeled 900-bp cDNA probe against bovine M9.2, which was excised from pBBM9.2 with BamHI. A Ϸ900-bp transcript was present in all tissues, but in low concentrations in skeletal muscle, heart muscle, and cortex (Fig. 7A). The same blot was rehybridized with a probe against human glyceraldehyde-3phosphate dehydrogenase, which is present in every tissue (Fig. 7B), and with a probe against the bovine V 1 V 0 -ATPase proteolipid c (Fig. 7C). Comparable M9.2/proteolipid c signal ratios were observed in most tissues, including skeletal and heart muscle with weak hybridization signals, but not in brain. The proteolipid c signal in brain was strong, whereas the M9.2 signal was hardly detectable. Quantification indicated an approximately 100-fold lower M9.2/proteolipid c signal ratio in brain.

DISCUSSION
Two-dimensional electrophoresis (BN-PAGE/Tricine-SDS-PAGE) was used to identify the proteins associated with V 1 V 0 holocomplex and V 0 membrane sector. The novel M9.2 and M8-9 proteins were identified as proteins associated with the V 0 membrane sector for the following reasons. (i) Because BN-PAGE separates membrane proteins according to their molecular masses (17), contaminants of V 1 V 0 and V 0 complexes should also be multiprotein complexes or oligomeric forms of smaller complexes. However, in the 29 -115-kDa range, the only proteins detectable have already been identified in different V-ATPase preparations (3)(4)(5)(6), which makes the presence of significant amounts of contaminating protein complexes unlikely. (ii) The M9.2 and M8-9 proteins were found both in the membrane sector and in the holoenzyme. Their staining intensities relative to the M39 subunit were almost identical in the membrane sector and in the holoenzyme (Table I). (iii) Proteins that precipitate during BN-PAGE could contaminate the V-ATPase complexes. However, in two-dimensional gels, these contaminants would appear as smearing bands crossing the polypeptide columns of the complexes as was found with dopamine-␤-monooxygenase. This is not the case for the M9.2 and M8-9 proteins, as they are found only as discrete spots at the positions of the V 0 and V 1 V 0 complexes.
The assignment of subunit G to V 0 or V 1 is still a matter of debate. A protein homologous to subunit G was first discovered as a component of the yeast V-ATPase, encoded by the VMA10 gene (57). It was named M16, and was suggested to belong to V 0 on the basis of its sequence homology with subunit b of F-ATPase, from the characteristics of VMA10 knockouts, and from cold inactivation of the V-ATPase, which failed to release it from the membrane. Similar results were obtained on cold inactivation of the chromaffin granule V-ATPase (49); however, Tomashek et al. (58) have shown recently that Vma10p inter-

9.2-kDa Protein of V-ATPase from Chromaffin Granules
acts with subunit E and classified it as a stalk subunit, belonging to V 1 . Subunit G in the midgut V-ATPase of Manduca sexta could be released from the membrane by cold inactivation or by treatment with chaotropic anions (59). Cold inactivation studies suggested also that subunits G and H from bovine brain clathrin-coated vesicles (60), which were later shown to be isoforms and renamed G 1 and G 2 , belong to V 1 rather than to V 0 (48). In the present work, we could identify subunit G in the holo-V-ATPase, but not in the V 0 membrane sector, by using an anti-G 1 antibody (Western blot not shown). This direct approach again suggests that subunit G is a V 1 component.
The electrophoretic separation of the holoenzyme (V 1 V 0 ) from its subcomplexes (V 1 and V 0 ) allowed the determination of the molar ratio of the various species. We found a V 0 /V 1 V 0 ratio of 5 after solubilization of chromaffin-granule membranes and resolution by BN-PAGE. It is hard to exclude the loss of V 1

9.2-kDa Protein of V-ATPase from Chromaffin Granules
subcomplexes during membrane isolation, particularly as dissociation of V 1 V 0 is promoted by MgATP at low temperatures. Nevertheless, we consider this unlikely for the following reasons: 1) 2 mM EDTA was included in all buffers during membrane isolation; 2) release of subunit B, a component of V 1 , was not detectable by immune blotting of soluble fractions obtained during membrane isolation.
A large excess of V 0 over V 1 V 0 has been reported before in chromaffin granule membranes (61), although in this case the ratio was determined after a prepurification step, which may have selected for the membrane sector. There have, however, been several other reports of the occurrence of free V 0 and V 1 : After solubilization of stripped bovine brain clathrin-coated vesicles with the nonionic detergent C 12 E 9 , a V 0 /V 1 V 0 ratio of about 2 was found by glycerol-gradient velocity centrifugation (62), and free V 1 was detected in cytosol from bovine brain and from Madin-Darby bovine kidney cells (63). Convincing evidence for the regulation of V-ATPase activity by the reversible dissociation of V 1 V 0 has been presented. This occurs in the vacuoles of S. cerevisiae in response to glucose deprivation (64), and in goblet cell apical membranes of M. sexta during moulting or starving of the larvae (65,66). Whether reversible dissociation of V 1 V 0 might also have a regulatory role in chromaffin cells, or whether V 0 itself might have an independent function, for example in exocytosis, is still a matter for speculation (67,68). It is noteworthy that in synaptic vesicles V 0 appears to exist in a complex with the vesicle membrane proteins synaptobrevin and synaptophysin (69).
Coomassie staining intensities of V 0 -subunits (Table I) did not indicate a high copy number for any V 0 protein except for proteolipid c. Assuming 1:1 stoichiometries for all V 0 proteins except six copies of proteolipid c as determined by Arai et al. (70), and neglecting the extent of glycosylation, a total mass of 288 kDa was calculated from the masses of the proteins listed in Table I. It was impossible to assign a monomeric or dimeric state to the major band of the V 0 membrane sector, because it had an apparent mass of 440 kDa in BN-PAGE, which was between the calculated masses of a monomeric (288 kDa) and a dimeric state (576 kDa). There are no data at present on the effects of protein glycosylation on the apparent masses in BN-PAGE. However, we speculate that the major V 0 form was the monomeric form, because glycosylation of M115 and M45 subunits should increase the Stokes radius and the apparent mass.
The holoenzyme seemed to be present in monomeric form, inasmuch as the calculated mass of 815 kDa was close to the apparent mass of around 1000 kDa in BN-PAGE, assuming 3 copies each of subunits A and B (11,70). Furthermore, 3 copies of subunit G were assumed for calculation, because the normalized staining intensity of subunit G was about 3-4 times higher than that of M39 (cf. Table I).
In the mammalian M9.2 protein a CSVCC sequence resembles potential metal-binding motifs (54), but only a cysteine doublet is retained at corresponding positions in the C. elegans and D. melanogaster sequences. If the C. elegans and D. melanogaster sequences were equivalent to the mammalian sequences, this would argue against the presence of a functional metal binding site.
The sequence similarity of human M9.2 and the yeast Vma21 proteins is not very high (45% similarity, 19% identity), but corresponding proteins of yeast and mammalian origin can have low sequence similarity, as shown by comparison of the 6.4-kDa protein of bovine bc 1 complex and the homologous yeast 8.5-kDa protein (71). The sequence and structural similarities of M9.2 and Vma21p indicate that the two proteins are potential homologues, and that assembly of mammalian V-ATPase might follow a pathway similar to that of the yeast V-ATPase. However, yeast Vma21p, which is required for assembly of V-ATPase, is not a subunit of V-ATPase, but instead localizes to the endoplasmatic reticulum membrane (56), whereas M9.2 protein was found to be associated with V 0 and V 1 V 0 complexes in adrenal glands. Because antibodies against M9. 2 are not yet available, we cannot exclude that M9.2 additionally or mainly localizes to the endoplasmic reticulum membrane. It seems conceivable that the mammalian protein is integrated into the complex after exerting its function in assembly, whereas yeast Vma21p is not.
M9.2 mRNA was detected in all tissues, but the M9.2/proteolipid c transcript level was about 100-fold lower in brain than in other tissues. This tissue-specific variation is not yet understood, but could indicate altered translational control, or decreased M9.2 protein degradation in brain. Alternatively, one could speculate that an undetected brain-specific analogue of M9.2 exists. FIG. 7. Northern blot analysis of M9.2 from various tissues. A Northern blot with total RNA from different bovine tissues was hybridized with probes against bovine M9.2 (A), human glyceraldehyde-3phosphate dehydrogenase (B), and proteolipid c from bovine V-ATPase (C). A 0.9-kilobase pair M9.2 transcript was detected in all tissues, but with strongly differing signal intensities. In all tissues except brain, the ratios of the M9.2/proteolipid c hybridization signals were comparable. The M9.2/proteolipid c ratio in brain was about 100-fold lower, but the strong signal of proteolipid c indicated the presence of substantial amounts of V-ATPase.