Reconstitution in Vitro of the V1 Complex from the Yeast Vacuolar Proton-translocating ATPase

Oligomeric assembly is a fundamental aspect of many complex enzymes. Using our native gel technique for examining subcomplexes of the V-ATPase V1 sector, we have developed an in vitro reconstitution assay for assembly of this complex. Assembly of complex II, the soluble V1complex observed in native gels, is dependent upon the presence of divalent cations and physiological temperatures. Assembly of soluble V1 can occur in a stepwise fashion from smaller subcomplexes found in some strains deleted for V-ATPase subunits. Specifically, V1 can be assembled directly from complex III (subunits E and G) with complex IV (subunits A, B, D, and F) without prior disassembly of complex IV. The formation of complex III in vivo is also shown to be essential and could not be achievedin vitro. Assembly from simpler precursors is possible and is enhanced by added ATP. Assembly can be blocked byN-ethylmaleimide in a Vma1p (subunit A)-specific manner. From these data, we extend our previous model to consider an assembly pathway whose steps reflect the catalytic mechanism of the Boyer binding-change model.

Many enzymes are oligomeric and require the coordinated assembly of numerous individual subunits. The vacuolar proton-translocating ATPase (V-ATPase) 1 is such a multisubunit enzyme, comprised of over a dozen different proteins both in and on the membranes of vacuoles, lysosomes, tonoplasts, and related compartments (1). This enzyme is also found at plasma membranes in certain species and tissues. More complex than its evolutionary cousin, the F-type ATP synthase, the V-ATPase is composed of over 22 individual protein subunits that are the product of at least 12 unique genes, some of which have multiple isoforms in higher eukaryotes. Because interactions between these subunits are fundamental to both the assembly and the energy-transducing mechanism of this enzyme, we imagine that the pathway of assembly might be related to the mechanism of the enzyme.
We have utilized native PAGE to reveal a myriad of partial V-ATPase complexes (Table I) found at steady state in the cytoplasm of both wild-type yeast as well as in deletion mu-tants of specific ATPase subunits encoded by the VMA and VPH gene families (2). 2 In wild-type yeast, as well as strains lacking one or more integral membrane sector subunits (e.g. vma3⌬ and vph1⌬stv1⌬), we observe a large stable complex (II; 576 Ϯ 96 kDa) that we believe is the soluble cytoplasmic form of V 1 . This complex contains at least Vma1p (A subunit), Vma2p (B subunit), Vma4p (E subunit), Vma7p (F subunit), Vma8p (D subunit), and Vma10p (G subunit), with a probable stoichiometry of A 3 B 3 DEFG. This complex is also present in vma5⌬ (Vma5p is the peripheral subunit C), indicating Vma5p is not necessary for assembly of the V 1 . In all strains except vma4⌬ and vma10⌬, we observe a small complex (III; 96 Ϯ 28 kDa) consisting of, minimally, Vma4p and Vma10p. In vma4⌬ and vma10⌬, we observe an intermediate complex (IV; 317 Ϯ 49 kDa) that contains all of the above proteins except for Vma4p and Vma10p, and in vma7⌬, we observe a complex (I) of low stability that runs above complex II and is of indeterminate mass. Again, it appears to contain all of the examined subunits, except for Vma7p. Finally, in all strains except vma7⌬ and vma8⌬, we observe complex VI that contains Vma7p and Vma8p.
From these observations, we have formulated speculative models for the assembly of the V 1 complex. However, these models are limited by assumptions regarding the intermediate nature of the observed complexes. We have not known if the observed complexes were intermediates trapped along a bona fide assembly pathway, or if they were products of artifactual sidetracks that resulted from the deletion of a specific subunit.
In the present study, we have extended our analysis of the intermediate complexes. We demonstrate that these complexes are able to assemble in vitro to form the soluble V 1 sector observed as complex II, thus indicating that they are not terminal complexes. We show that an early assembly step is ATP-dependent and that this is associated with the catalytic site of the enzyme. Finally, we show specific assembly of complex II onto isolated vacuolar membranes containing the V 0 sector, indicating that complex II is capable of V 0 -dependent membrane association. Kinetic studies of these processes have begun to distinguish discrete steps along the assembly pathway. These results lead us to conclude that the assembly pathway of the enzyme may recapitulate the functional mechanism.

MATERIALS AND METHODS
Reagents, Antibodies, Strains, and Plasmids-All reagents were from Sigma unless otherwise stated. HEPES buffer was from Research Organics, Inc. Media components were from Difco. All single deletion strains were based on the wild-type W303-1B (MAT␣ leu2 his3 ade2 trp1 ura3) and were gifts from N. Nelson (Tel Aviv University). Double deletions were constructed as follows. Strain vma4⌬vma7⌬ was constructed starting from the vma7⌬ strain (vma7::URA3) and performing * This work was supported by U. S. Public Health Service Grant GM53396. 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 VMA7 and VMA7HA expression plasmids were constructed as follows. Primers to VMA7, which isolated the ORF and attached a 5Ј-BamHI and a 3Ј-XhoI site, were designed and used to isolate the VMA7 ORF from YPN1VMA7 (5) and the VMA7HA ORF from pLG20 (4) by polymerase chain reaction amplification. The polymerase chain reaction products were digested with BamHI/XhoI and cloned into the BamHI/SalI sites of pHADR. This plasmid was constructed by excising the multiple cloning site of pRS423 (containing a HIS3 marker) and replacing it with a single SphI site to form pRS423S. The ADH/GAL4 fusion expression cassette from pAЈ (2) was modified by deleting the GAL4 fusion portion located between two HindIII sites and replacing it with a multiple cloning site. The modified ADH expression cassette was moved into the SphI site of pRS423S to form pHADR. Both pHADR7 and pHADR7HA complemented the vma phenotype when transformed into the vma7⌬ strain (data not shown).
Reconstitution and Native Gel Electrophoresis-Preparation of cytosolic extracts and native gel electrophoresis have been described previously (2). Extracts were kept on ice except during reconstitution as described in the figure legends and text. Incubations were for 10 -60 min, as indicated. The following were used at these final concentrations, where indicated: metal cations (5 mM), ATP (1 mM), EDTA (10 mM), NEM (1 mM). Reconstitutions were always returned to ice before loading onto 6% native gels. Immunoblots were quantitated with Millipore BioImage software.
Preparation of Vacuolar Membranes and Membrane Association Assay-Cells were harvested at midlog phase, washed with water, and incubated in 20 ml of 10 mM Tris-SO 4 , pH 9.4, 0.1% ␤-mercaptoethanol for 20 min. Cells were pelleted and resuspended in spheroplasting buffer (50 mM Tris-HCl, pH 7.5, 2% glucose, 1 M sorbitol) and incubated with Zymolyase (ICN) for 30 min. Spheroplasts were collected by centrifugation at 2000 rpm for 5 min and resuspended in 15 ml of Buffer A (10 mM MES/Tris, pH 6.9, 0.1 mM MgCl 2 , 12% Ficoll 400) containing 20 l of protease inhibitor mixture (5 mg/ml each pepstatin A, L-1-tosylamido-2-phenylethyl chloromethyl ketone, leupeptin, ␣ 2 -macroglobulin, and 200 mM phenylmethylsulfonyl fluoride). The cells were lysed at 4°C by Dounce homogenization. The lysate was divided into ultracentrifuge tubes, overlaid with 4 ml of Buffer B (as Buffer A, but 8% Ficoll 400), and spun at 30,000 rpm for 30 min. Vacuolar membranes were collected off the top, homogenized in Buffer C (10 mM MES/Tris, pH 6.9, 25 mM KCl, 1 mM EDTA) and spun at 30,000 rpm for 15 min. The supernatant was discarded, and the pellet containing vacuole membranes was resuspended in native gel sample buffer and stored at Ϫ80°C until needed.
For membrane association experiments, thawed membranes were combined with vma3⌬ cytosolic extract and incubated in the presence of 5 mM MgCl 2 for 25 min at 30°C. The membranes were pelleted at high speed in a microcentrifuge for 1.5 min, and the supernatant fraction was moved to a separate tube. Some supernatant/pellet pairs were brought up in Laemmli buffer and examined by SDS-PAGE and immunoblot.

Divalent Cations Are Required for Assembly in Crude
Extracts-Strains bearing disruptions in single VMA genes assemble subcomplexes of the peripheral V 1 sector. To demonstrate that the intermediate subcomplexes previously characterized were on the assembly pathway, we developed an in vitro reconstitution protocol. Preliminary experiments consisted of mixing extracts from vma4⌬ and vma8⌬. These ex-tracts were chosen because the former accumulates complex IV and the latter accumulates complex III, complexes which contain mutually exclusive subunits. Mixing crude extracts containing these complexes resulted in formation of complex II in a divalent cation-dependent manner (Fig. 1A). We examined the effect of different anions and cations on complex II formation and observed that Mg 2ϩ , Mn 2ϩ , and Ca 2ϩ were all effective, whereas Zn 2ϩ was ineffective (Fig. 1B). In fact, ZnCl 2 caused the complete loss of complexes III and IV when combined with extracts containing these complexes (data not shown). Both MgCl 2 and MgOAc 2 were effective, suggesting that no specific anion is necessary for assembly. Occasionally, we observed weak assembly with NaCl, but the signal from complex II was faint and inconsistent. Because there is 1 mM EDTA in our extract buffer, we suspect that 5 mM Na ϩ may compete weakly and release enough endogenous divalent cations to effect some limited reconstitution.
Complexes III and IV Are Consumed in the Reconstitution Reaction-To show that the ATPase-specific subcomplexes were responsible for the appearance of complex II, we did a titration of vma4⌬ extract versus vma8⌬ extract (Fig. 2). The formation of complex II clearly depends upon both extracts, with maximal reconstitution occurring when the extracts are at near equal proportions. Likewise, complex III and IV both disappear as they become limiting. It is interesting to note that complex VI (Vma7p, Vma8p) does not decrease significantly until complex IV (Vma1p, Vma2p, Vma7p, Vma8p) has first been consumed, suggesting that complex IV is a more direct precursor to complex II than is complex VI.
Under the conditions where complex III becomes limiting, we see the formation of a new complex, designated complex Q. This complex only appears in reconstitutions where vma4⌬ is one of the partners and may be an intermediate complex between IV and II. It runs at a position nearly identical to complex III. This complex is only weakly and transiently detectable relative to the other complexes observed, making complete analysis of its composition difficult.
Kinetics of Complex II Assembly-Because the assembly was dependent on the presence of Mg 2ϩ , we surmised it would be possible to gather kinetic data by stopping the assembly with Extracts indicated were combined on ice and then metal salt was added to 5 mM. Reconstitutions were incubated at 30°C for 10 min and then returned to ice. Immunoblots of native gels were probed with the indicated antibodies. See Table I for complete complex constituents. the further addition of EDTA. Reactions were prepared by mixing two extracts on ice and then started with the addition of MgCl 2 (5 mM final) and the transfer to either ice or a 30°C waterbath. Aliquots were taken at given time points and transferred to iced tubes containing EDTA (10 mM final concentration). From quantitation of native gels (Fig. 3), it can be seen that the assembly of complex II is temperature-dependent and occurs with a half-time of Ͻ5 min at 30°C. We wished to know if some amount of in vivo assembly or modification was required for in vitro assembly to occur. To test this, we attempted reconstitutions with all possible combinations of extracts, and the results are summarized in Table II. All combinations were able to support formation of complex II, except for vma4⌬ ϫ vma10⌬ and occasionally vma1⌬ ϫ vma2⌬ (see below). Kinetics of assembly in all positive combinations were similar to the vma4⌬ ϫ vma8⌬ cross, with half-times of ϳ5 min (data not shown). The vma4⌬ ϫ vma10⌬ combination, which completely lacks complex III ab initio, does not support formation of complex II under any conditions tested. However, we have observed that the steady-state levels of Vma4p are extremely low in vma10⌬ cells and that Vma4p is rapidly degraded in this strain. 2 Thus, levels of Vma4p may be the limiting factor. However, assembly of this complex may depend upon conditions found only in the intact cell.
Assembly Occurs from Association of Complexes-Because we observed reassembly in most cases, even those where there were no pre-existing large complexes in either of the starting extracts, it remained an open question whether or not complex II assembles directly from subcomplexes or from individual subunits that became available as the subcomplexes dissociate. In particular, we decided to examine if complex IV is competent for further assembly without dissociation. To examine this, we employed the epitope-tagged Vma7p (4) to label complex IV as shown in the scheme in Fig. 4A. By creating tagged and untagged complex IV and then complementing it in vitro with vma8⌬ extract containing untagged or tagged Vma7p, respectively, we could determine if complex IV was dissassembled because its Vma7p would mix with that in the total reaction.
The presence of the tagged Vma7p causes a slight shift in the position of complex IV (Fig. 4B, right panel, lane 1 versus lane  2) and the position of complex II (Fig. 4B, right panel, lanes 5  and 8 versus lanes 6 and 7). There is also the appearance of an additional signal below complex VI. The tagged Vma7p is found in complex II only when it begins in complex IV (Fig. 4B, right  panel versus left panel, lanes 5 and 8). Tagged Vma7p is excluded from complex II when complex IV in the reaction contains only untagged Vma7p (Fig. 4B, right panel versus left  panel, lanes 6 and 7). Levels of tagged Vma7p were comparable in the extracts from the strains in which this construct was expressed (Fig. 4C). These results indicate that complex IV is a bonafide intermediate on the pathway to formation of complex II and that dissociation of complex IV does not occur to any appreciable extent prior to complex II assembly. We also note a similar result for complex VI; free Vma7pHA does not replace  3. Assembly of complex II is temperature dependent. Extracts of vma4⌬ and vma8⌬ were combined in equal portions on ice. MgCl 2 was added to 5 mM, and the tubes were moved to either 30°C or left on ice. Samples were taken at the indicated times and moved into tubes containing EDTA (10 mM final concentration) on ice. Samples were loaded onto 6% native gels, immunoblotted, and probed with antibodies to Vma8p. Similar results are seen with other antibodies (data not shown).
Vma7p that is present in a complex with Vma8p (Fig. 4B,  lane 6).
Complex II Associates with the Membrane in a V 0 -dependent Manner-To demonstrate that complex II was not a dead-end product itself, we undertook experiments to show it could assemble on isolated vacuolar membranes containing V 0 . We prepared membranes from vma2⌬, which have been shown to contain V 0 without peripheral subunits (6). We also prepared vacuolar membranes from vma3⌬, which do not contain V 0 , as a control. Cytosolic extract from vma3⌬, which contains large quantities of complex II, was combined with these membranes both in the presence (Fig. 5) and absence (data not shown) of MgCl 2 . We observed that MgCl 2 was necessary for association with the membrane. After a 25-min incubation at 30°C, membranes were pelleted and examined by immunoblot for the presence of V 1 subunits.
Association with the membrane is dependent upon both cytosol containing complex II and membranes containing V 0 . The membranes from vma2⌬ show the presence of representative V 1 subunits found in complex II. Membrane binding is specific as evidenced by the absence of complex II subunits from the vma3⌬ membranes. The membranes from vma2⌬ also bring down Vma13p, a protein that has been shown to be necessary for activity but not assembly of the V-ATPase (7). Examination of native gels with antibodies to Vma13p gave equivocal results (data not shown); it may be a component of complex II, but it does not appear to be associated with any other complex. Interestingly, Vma5p is not associated with the membrane, which may explain our inability to detect activity from these membranes (data not shown). The V-ATPase has been observed to assemble without this subunit in other systems although its presence does enhance the level of activity (8). In the yeast vma5⌬ strain, no V-ATPase activity is detected, and the V 1 subunits are found assembled but not attached to the membrane (2,9). Vma5p may be involved in a transient interaction, necessary for attachment and full activity but not necessary for maintenance of a stable V 1 V 0 association. Our data suggest that complex II is capable of assembly onto the V 0 sector of the V-ATPase.
ATP Is Necessary for Early Assembly Steps in a Vma1p-dependent Manner-Under our standard conditions for reconstitution, we found weak assembly of complex II from the vma1⌬ ϫ vma2⌬ combination. Upon further examination, it was found that assembly of complex II from vma1⌬ ϫ vma2⌬ could be enhanced by the addition of 1 mM ATP. Measurement of assembly kinetics with MgATP, Mg 2ϩ , Mn 2ϩ , and Ca 2ϩ reveal both a lag and lower overall assembly without added ATP (Fig.  6). Because the extracts have not been depleted for triphosphonucleotides, there may be residual ATP available to support limited reconstitution in the absence of added ATP. Attempts to deplete the extracts with apyrase and hexokinase did abolish assembly (data not shown), although subsequent add-back of ATP was impossible due to the presence of the scavenger. ATP analogues such as ATP␥S and AMP-PNP showed no significant effect by themselves and did not show any competition at concentrations from 0.1 to 10 times the ATP concentration (data not shown). Interestingly, the divalent cation dependence for vma1⌬ ϫ vma2⌬ is ion-specific and parallels reported preferences of divalent cations for ATPase activity (10,11,12,13).
To determine if ATP binding at the catalytic site was important to formation of complex II, we examined the ability of N-ethyl maleimide to inhibit assembly. NEM reacts specifically and irreversibly with a cysteine residue in the active site on Vma1p (14). A recombination reaction of vma1⌬ ϫ vma2⌬, which was first treated with NEM followed by ATP showed no reconstitution, whereas if first treated with ATP, weak reconsititution was observed (Fig. 7A). More significantly, if the deletion extracts were treated separately, pretreatment of vma2⌬ (containing Vma1p) with NEM abolishes reconstitution while pretreatment of vma1⌬ (containing Vma2p) with NEM is still capable of assembly (Fig. 7B). This strongly suggests that binding of ATP at a functional catalytic site is necessary for assembly of the V 1 complex since the cytosolic constituents of these two extracts should be identical except for the specific Vma proteins. As our revised model makes clear, this can be Ϫ Extract from vma3⌬ was combined with membranes purified from the strains indicated and incubated at 30°C for 25 min. Membranes were pelleted (P) and separated from the supernatant (S). Both were run on SDS-PAGE, transferred to PVDF membrane, and probed with antibodies to the indicated subunits. elegantly coordinated with the favored mechanistic model, the Boyer binding-change mechanism. DISCUSSION Biosynthesis and assembly of multimeric proteins into their quaternary structures must be regulated by both kinetic and thermodynamic parameters. The stability of complex II suggests that it is a thermodynamic trap, making it the dominant complex in wild-type cells and also allowing us to form it in vitro. We do not see complex IV form in vitro, even under conditions where we assume it must be a transient intermediate (e.g. the vma1⌬ ϫ vma2⌬ reaction) because its equilibrium favors dissociation into smaller complexes or individual subunits unless it can further assemble to complex II. Attempts to form complex IV in vitro (vma4⌬vma7⌬ ϫ vma4⌬vma8⌬) were unsuccessful (data not shown). Once formed, the attachment and detachment of complex II to the membrane sector would then have an equilibrium independent of the levels of individual subunits. Assembly and disassembly of the V 1 V 0 has been proposed as a mechanism of regulation (15) and is supported by the hyperassembly of V 1 onto V 0 in cells with point mutations in V 0 sector subunits that abolish function but not assembly (16). This equilibrium would be regulated by other subunits, e.g. Vma5p, which is not necessary for assembly of V 1 but is required for assembly on the membrane in vivo and is involved in glucose deprivation-mediated modification of the V 1 following detachment from the membrane sector. 3 Interestingly, cytosolic V 1 complexes, which apparently do not exchange with membrane-bound complexes, have been observed in tissue culture cells (17). What the role of such complexes might be and whether they exist in yeast remain to be determined.
In contrast, assembly steps from individual subunits to complex II depend on temperature, divalent cation, and nucleotide. At 30°C, our kinetic data indicate that most combinations of deletion strains assemble with roughly equal half-times (ϳ5 min) although there is a tendency for extracts containing higher order complexes to show faster assembly (compare Fig.  3 with Fig. 6). At lower temperatures, measurable differences in assembly rates may appear and reveal key steps where kinetic control assumes a larger role.
We have expanded our previous assembly models to consider these data (Fig. 8). In addition to our data which support the model, we favor it because of its similarity to the Boyer bindingchange model of F-type ATP synthase catalysis (18), a mechanism which has been proposed for the V-type enzymes as well (19). The Boyer model proposes three sites, catalytically equivalent over time but instantaneously nonequivalent. We hypothesize that these three possible states (MgATP bound (T), MgADP/P i bound (D), and open (O)) are not energetically equivalent. Thus, if assembly begins with the association of Vma1p and Vma2p to form the catalytic site (found predominantly on Vma1p but shared by both), one of these three states may be preferred, and an accumulation of dimers in one state would result. The suggestion that these two subunits associate to form a catalytic site is supported by recent evidence that these two subunits have low but detectable ATPase activity higher than either individual subunit (20). Also, our own two-hybrid data indicate that Vma1p and Vma2p can interact independent of other subunits. Here, we propose that this interaction nu-3 J. Tomashek, unpublished data. cleates around the catalytic site of the heterodimer and suggest that MgATP stabilizes this dimerization. Thus, Vma1p-T-Vma2p is the most favored of the three possible initial complexes.
Association with another subunit or subcomplex, in our model complex VI, would convert Vma1p-T-Vma2p into a new state stabilized by both hydrolysis of the nucleotide and incorporation of the new subunits into the growing complex. We prefer a model where association of the two subcomplexes (Vma1p-T-Vma2p and VI) would precede hydrolysis of ATP. In our model, complex VI acts as a nucleotidase activating factor, similar to GTPase activating proteins. After the catalytic step, complex VI (associated now with Vma1p-D-Vma2p) is situated to bind another Vma1p-T-Vma2p dimer from the cytosolic pool. The result is the formation of complex IV, which now has two catalytic sites each in a different energetic state defined by their nucleotide contents and interactions with complex VI.
Complex III (Vma4p-Vma10p) may now interact with the Vma1p-D-Vma2p portion of complex IV, and stimulate release of the nucleotide and phosphate. In this sense, complex III is a nucleotide exchange factor. Release of ADP and P i , in turn, allows hydrolysis to proceed at the second site, and thus the core subunits become positioned to receive the third Vma1p-T-Vma2p dimer. The final complex now has the three different sites in the correct energetic states necessary for catalysis to commence. The complete V 1 might be regarded as three ATPases rotating around a complex of hydrolysis activating and nucleotide exchange factors.
This model also accommodates an explanation for why ATP analogues have no apparent effect on assembly. The ATPbound site is, based on the Boyer model, predicted to be of extremely high affinity. However, this may be true only for the ATP-bound site in the complete enzyme, i.e. of the last site added during assembly (incorporation of which makes the site "tight"). The Vma1p-T-Vma2p free dimer may have a low affinity and high exchange rate for nucleotide. Thus, in the absence of ATP, the analogues would not support assembly because they cannot be hydrolyzed. In the presence of ATP, the analogues would compete at high enough concentration but not block assembly, even if they were incorporated into the final "tight" site (although this would poison the enzyme). In our experiments, we are not able to eliminate completely endogenous ATP. NEM, which binds covalently and irreversibly, does block assembly.
What about complexes I and V? Complex V, seen in significant steady-state amounts only from vma4⌬ and vma10⌬ extracts, disappears very rapidly at 30°C or when combined with other extracts, and thus could not be followed in our assays. Complex I appears in the absence of Vma7p, a subunit which we have implicated in early assembly events. Because complex I is very unstable, we believe it may be an artifactual complex that assembles incorrectly, accepting Vma8p alone in lieu of complex VI. As a result, the catalytic dimers would all be in the Vma1p-T-Vma2p configuration, or possibly a combination of Vma1p-T-Vma2p and Vma1p-O-Vma2p, if complex III is able to release ATP from the second site incorporated. Because the energetic states in complex I would be discoordinated, disas-sembly of this complex may be necessary for subsequent correct assembly to occur. Therefore, complex I may be the dead end of an artifactual pathway that occurs only because of the absence of Vma7p.
Our assembly system is different from previously described reconstitution systems from plants (21), animals (20,22), and yeast (6) in three significant ways. First, we do not rely on chemical treatment (e.g. nitrate, iodide, or urea) to break apart the V-ATPase, a treatment which may chemically modify the enzyme (23). Instead, we begin with subunits and partially assembled complexes in a physiological milieu. Second, we are working with crude extracts, whereas other reconstitution has been done with purified enzymes and heterologously expressed subunits. Thus, we cannot exclude the involvement of factors apart from the ATPase subunits themselves. Third, we have been as yet unable to recover activity, while other systems have obtained detectable yields of activity from dialyzed mixtures of subunits. However, we expect with further attenuation to be able to recover activity and identify the specific factors responsible. Reassembly in other systems is achieved in the absence of Mg 2ϩ and ATP and at 4°C, albeit after long incubation and dialysis. From the results in our system, we believe assembly of the V 1 complex is an ordered process that recapitulates its proposed functional mechanism. We hope that further experiments will provide greater insight regarding the validity of this provocative ATPase assembly model.