The F, Subunits of the Escherichia coli FIF,-ATP Synthase Are Sufficient to Form a Functional Proton Pore*

The assembly of the F, sector of the Escherichia coli ATP synthase has been studied using both structural and functional criteria for assembly. Cross-linking E. coli minicell membranes containing only the F, subunits a, b, and c with dithiobis(succinimidy1 propio- nate) (DSP) produces bz and Ca dimers that are generated by cross-linking membranes containing the as- sembled holoenzyme. Five plasmids carrying the genes specifying the F, polypeptides in a bacterial strain lacking all of the unc (ATP synthase) genes show a good correlation between F, function and the amount of the membrane-bound F, polypeptides. In this report we revise a conclusion reached previously (Klionsky, D. J., Brusilow, W. S. A,, and Simoni, R. D. (1983) J. Bid. and present evidence that the F, subunits alone are sufficient to assemble a functional proton pore. Proton-translocating adenosinetriphosphatases (EC 3.6.1.3) are present in the energy-transducing membranes found in bacteria, plant, and animal cells. The H’-ATPase, or ATP synthase, catalyzes the synthesis of ATP from ADP and P; at the expense of a hydrogen ion potential. The Escherichia coli ATP synthase is composed of two domains termed the F1 and the F,. The F1 is the membrane extrinsic catalytic domain. The F,

protein, probably folds into a helical hairpin which traverses the membrane bilayer. The b polypeptide consists of a short hydrophobic amino terminus, which probably inserts into the bilayer, connected to a membrane extrinsic hydrophilic portion (8,9). Relatively little is known about the a subunit, which probably folds six or seven cy helical segments across the bilayer leaving hydrophilic portions of its length to either side of the membrane bilayer (9-12). Proton translocation requires all three subunits in vivo (13) and in vitro (14).
The assembly of the E. coli ATP synthase has been analyzed using mutations in the unc operon. Cox et al. (15) have observed that a mutation in uncD (coding for p) resulted in the absence of the b polypeptide from the membrane. An assembly sequence was proposed on the basis of the analysis of this and a number of other unc mutations (15). One feature of the proposed assembly sequence is coordinated assembly of the F, and F, sectors: completion of F, assembly (membrane integration of b) requires F, (@) subunit synthesis. We and others find that the b subunit inserts into the cytoplasmic membrane in the absence of the p subunit or other FI subunits (13,16,17). In addition, it is not clear that b exists or accumulates in the cytoplasm in the uncD mutant (15). It has been suggested that the membrane insertion of b in the absence of p may be an artifact of the overproduction of the b polypeptide from multicopy plasmids containing the uncF gene (18). In the present paper, we address some of these uncertainties in F, assembly using both structural and functional criteria for F, assembly.

Assembly of the F, Sector in Minicells Detected by DSP
Cross-linking-In our initial experiments, the combination DK3/pRPG45 (AuncB-Dlucb6) was unable to promote the formation of a functional F, in the absence of the F, subunits cy and /3 (16). 3 The cross-linking experiment described below was undertaken to establish whether or not the inability to measure F, function was due to lack of assembly of the F, Portions of this paper (including "Experimental Procedures," Fig.  1, and additional references) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are available from the Journal of Biological Chemistry, 9650 Rockville Pike, Bethesda, MD 20814.
Request Document No. 85M-347, cite the authors, and include a check or money order for $1.60 per set of photocopies. Full size photocopies are also included in the microfilm edition of the Journal that is available from Waverly Press.
Strain/plasmid combinations such as the strain DK3 harboring the plasmid pRPG45, will be noted as DK3/pRPG45 (AuncB-D/ acbcl). The polypeptides expressed by pRPG45 are listed according to gene order. The bacterial strain used for these preparations, strain DK3 (AuncB-D), lacks the chromosomal unc genes B, E, F, H, A, G, and D. Strain DK3 codes for the i and c subunits. subunits. Do the F, polypeptides form interactions in the absence of the F1 subunits?
We have previously used the cleavable cross-linking reagent DSP on the purified FIFO complex and on a minicell lysate to detect interactions between F, polypeptides (10). The crosslinking procedure in this report has been modified in two ways compared to our previous report. An unlabeled methionine chase (1 mM methionine-labeling period to permit synthesis of increased amounts of the F, polypeptides. Cross-linking is performed on membrane suspensions instead of minicell lysates. DSP cross-linking of the DK3/pRPG45 ( AuncB-Dlacbd) minicell membranes results in the detection of the bz dimer and the cz dimer at a DSP concentration of 0.5 mM (Fig. 2 A ,  spots1 and 2). The apparent molecular weights of the bz and cp dimers are 35,000 and 13,000, corresponding to the predicted values of 34,400 and 13,600, respectively (the apparent molecular weight for c in the gel system is 6,800). A DSP concentration of 0.05 mM detects only the b2 dimer (Fig. 2B, spot I). DSP cross-linking of the FIFO complex in minicell membranes from DK3/pRPG54 (AuncB-Dlacbdaypt) shows the b2 and cz dimers (Fig. 2C, spots 1 and 2) at a concentration of 0.5 mM DSP. Without DSP, no cross-links are generated (Fig. 20). The amount of the cross-linked F, polypeptides in Fig. 2, A and B, is low due to the nature of the cross-linking methodology but compares favorably with the amount of the cross-linked products in Fig. 2C, as well as with the amount of the cross-linked products obtained with purified FIFO holoenzyme (10). The formation of DSP cross-links between F, subunits has not been found to be quantitative in experiments performed by us as well as others (10, 11). Treatment of the cross-linked membranes with 20 mM iodoacetate and 20 mM N-ethylmaleimide in sample buffer (containing no reducing agent) for 20 min at 50 "C has no effect on the presence of the bp dimer spot (data not shown). This indicates that the b dimer detected with DSP cross-linking does not result from sulfhydryl group exchange or the formation of a cystine bridge between b molecules during electrophoresis.
An interaction between the b and c subunits has been detected using an immunoprecipitation technique. Antisera specific to the b polypeptide (10) co-immunoprecipitates the c polypeptide from an Aminoxid WS 35-solubilized membrane preparation from DK3/pRPG45 ( AuncB-Dlacbd) minicells (data not shown).
Although interactions between a and b have been detected in vivo (11) and with the purified FIF, (lo), it has been difficult to detect this interaction in the minicell system (10). The stoichiometry of a:b:c in membranes from DK3/pRPG45 (AuncB-Dlacbd) minicells labeled with [35S]methionine is 1.0:1.9:12.6, respectively. This stoichiometry was determined by electrophoresing a sample of the DK31pRPG45 minicell membrane, performing autoradiography, and quantitating the radioactivity present in each band densitometrically. The close correspondence of this ratio with the probable in vivo ratio of al:b2:~10~16 ( 5 , 6) suggests that the a subunit has properly integrated into the F, sector.
Dimerization of b Polypeptides in the Absence of Other F, Subunits-The interactions between the F, subunits detected above suggest that the information necessary for the assembly of the F, resides in the individual subunits and that the individual subunits may exhibit assembly interactions when isolated from one another. We have detected such an interaction in the analysis of the cross-linking properties of the b polypeptide in a membrane containing no other F, or F, polypeptides.
Minicell membranes from DK3/pRPG51 ( AuncB-D/bd) containing 35S-labeled b protein produce a prominent b d' lmer when treated with DSP ( Fig. 3A, spot I ) . Cross-linking also occurs in membranes from the strain DK6/pJPA17 (AuncB-C/b') between nonfunctional b' polypeptides carrying the mis-sense mutation which substitutes Asp-131 for Gly-131 ( Fig. 3B, spot 2). We have attempted to identify the residue(s) in the b polypeptide chain that are responsible for the 6-b interaction by cross-linking truncated b polypeptides with DSP. For this purpose we have used three b polypeptides that have missing carboxyl-terminal portions due to the introduction of nonsense mutations into the uncF gene (19). The three truncated b polypeptides studied have sizes of 13,700, 11,600, and 10,400 daltons corresponding to 122, 105, and 95 amino acids, respectively, compared to 156 amino acids residues in the fulllength b. These truncated b polypeptides are designated b( 14), b(12), and b(10), respectively. The b(14), b(12), and b(10) polypeptides in minicell membranes from DK3lpJPA14, DK3/pJPA12, and DK3/pJPA10, respectively, are crosslinked with DSP to form dimers (Fig. 4, spots [1][2][3]. The efficiency of cross-linking appears to diminish somewhat with the absence of the carboxyl-terminal residues. Two shorter truncated b polypeptides of sizes 7,900 and 9,300 daltons have been subjected to similar DSP cross-linking analysis and appear to form dimers as well (data not shown).
The a, 6, and c Subunits Are Sufficient to Form a Functional F,-On the basis of the results from the structural analysis of F, assembly above, we decided to re-examine the function of the F, sector in membrane suspensions using a fluorescence quenching assay. Membrane suspensions were carefully prepared and assayed as quickly as possible from fresh bacterial cultures without freezing of cell or membrane samples (see "Experimental Procedures"). The fluorescence quenching activity of a membrane sample from a given strain/plasmid combination varied only 5 to 10% between different membrane preparations.
Membranes from DK3/pRPG54 (AuncB-D/acbday@t) contain a complete ATP synthase complex and show high NADH-driven and ATP-driven fluorescence quenching activity (Fig. 5C). The amount of fluorescence quenching observed corresponds to the membrane potential generated by the electron transport chain, and the ATP synthase at the expense of the substrates NADH and ATP, respectively. The electron transport poison KCN and the protonophore CCCP abolish proton-dependent fluorescence quenching. The plasmid pDJK19 codes for the F, subunits a, b, and c and a partial a polypeptide (a'). The function of the F, from DK3/pDJK19 (AuncB-Dlacba') is measured in two ways in the fluorescence assay. Membranes show a low level of NADH-driven fluorescence quenching (Fig. 5A). This reduced level of quenching is due tp the dissipation of the NADH-dependent proton gradient by the functional proton-translocating F, sector. Membranes from DK3/pDJK19 (AuncB-Dlacba') do not contain a functional F, and produce no ATP-driven fluorescence quenching (Fig. 5A). Purified F1 may be added back to the DK3/pDJK19 membrane preparation to reconstitute ATPdriven fluorescence quenching (Fig. 5B). F,-reconstituted membranes also show restored NADH-driven quenching because proton translocation through the F, is blocked by the attached F,.
Five Plasmids with Different Levels of F, Activity-In order to investigate the basis for differences of F, function, we have examined five plasmids which produce a range of F, activity. The plasmids we have used are pDJK19 (acba'), pDJK20 (acb), pJPAl (acb), pRPG23 (acbda), and pRPG45 (acbd). Plasmids pDJK19 and pDJK20 are derivatives of pRPG54 The plasmids responsible for the range of F, activity may be ordered from least to most active: pRPG45 (acbb), pJPAl (nch), pRPG23 (achdn), pDJK19 (acbn'), and pDJK20 (acb). As expected, we find in general, the higher the reconstituted ATP-driven fluorescence quenching, the lower the NADHdriven fluorescence quenching.
The one exception occurs when comparing the pD.JK19 and pDJK20 samples (Fig. 6, A and H ) . Neither the 6 subunit, nor the combination of t h e 6 and CY subunits, appear to have an in oiuo effect on F, function.
The ATP-driven fluorescence quenching obtained with F1reconstituted membranes from DK6/pPJPA1 (AuncB-C/acb) was identical to that obtained with F,-reconstituted membranes from DK3/prJPA1 (.hncR-D/ach) (data not shown). This indicates that t.he t subunit, coded for by uncC, also does not play an obligate role in F,, assembly. Significantly, DK.?/pDdKlS (AuncR-Dlacbn') grew with a doubling time of 52 min, which compares favorably with the 49-min doubling time observed for the control combination DK.7/pDt1K4 (AuncB-Dlac). Likewise, DK3/pRPG23 (AuncR-D/nchdm) doubled in 50 min. Thus, the levels of F, function we have measured do not appear to be deleterious to the growth rat,e of the bacteria in rich medium.
Thc Amount of Membrane-bound F , Subunits Correlate with F,, Actioitv-In order to characterize this F, activity and demonstrate it,s authenticity, the relationship between the amount of the membrane-bound F,, subunits and the activity measured in the fluorescence assay was investigated. It was important to show, for example, that the F,, activity was not due to an excess of one or more of the F, subunits.
Membrane samples assayed by fluorescence were examined using a technique that allows the F,, proteins to he identified among the large number of other proteins in a membrane sample. Membrane suspensions (of equal protein content) were diluted with a low ionic strength buffer (STI) and sedimented. The resulting membrane pellets were resuspended in ST1 buffer, stirred in the cold, and sedimented as before. This procedure removes extrinsic proteins from the membrane. Electrophoresis was carried out on an aliquot of the resultant "stripped" membrane samples (Fig. 7 ) . Subunits a , 6, and c are identified by co-migmtion with the F,, pol.ypeptides present in a sample of purified ATP synthase and by comparison of membrane samples with and without the F,,  In order to more accurately quantify the amount of the b subunit present in the membrane, a procedure was developed for the extraction of b from low ionic strength washed membranes with the detergent Aminoxid WS 35. As described above, the membrane samples are diluted with ST1 buffer and sedimented. Extraction of b is achieved by resuspending the membrane pellet in ST1 buffer containing 10 mM Aminoxid. After sedimentation, the supernatant contains solubilized b subunit. We have characterized this procedure by comparing the polypeptides solubilized from DK3/pDJK4 ( AuncB-D/ac) and DKS/pDJK19 (AuncB-Dlucba') membranes. The solubilized b polypeptide can be seen in the Aminoxid supernatant from DK3/pDJK19, but not from DK3/pDJK4 (Fig. 8, lanes 4 and 5 ) . The b polypeptide was identified by co-migration with the b subunit in a sample of purified ATP synthase (data not shown). Using the Aminoxid procedure, electrophoretic quantification of the b subunit present in the membrane samples was performed (Fig. 9). The results from the electrophoretic quantification of all three F, subunits using the two procedures described above are summarized in Table I Table I, values for Q). This quotient, Q, is a rough measure of F, "assembly efficiency". Although the correspondence between activity and mass of F, polypeptides is not exactly quantitative, there is only a small amount of variation in the quotient Q (Table I) to different amounts of F,, sectors in the membranes and not to different efficiencies of F,, assembly. The Q value for DS410 may he increased relative to the values for the plasmid-bearing strain because the ATP-driven fluorescence quenching activity of DS410 membranes is due t,o the Fl synthesized in uiuo.
The ATP-driven fluorescence quenching activity from the plasmid-hearing strains is due to the FI added back to the membranes in oitro. We attrihute the different amounts of F, pol-ypeptides synt,hesized from these plasmids to differences in the promoters in the plasmid vector sequences (see "Experimental Procedures"). The stoichiometry of the a and 6 subunits may be calculated from the absolute staining intensity data obtained from the densit,ometer analysis (data not shown). (The data in Table I may he compared vertically hut not horizontally.) The a:b stoichiometries based on the densitometer scan data are: 1:1.6 (pRPG4.5), 1:2.1 (pJPAl), 1:2.5 (pRPG2.31, 1:1.4 (pDJK19), 1:l.g (pDdKZO), and 1:l.G (DS410). Although the relative staining efficiency for the F, suhunits is unknown, we believe it is significant. that these ratios are in agreement with the prohahle in vi00 ratio of al:bz ( 5 , 6).
The amount of t,he memhrane-bound c subunit shows no consist,ent relat.ionship to F,, function and even appears to be inversely related to activity in samples with intermediate levels of activity (Table I). Considering the extreme hydrophobicity of the c subunit, it is unlikely that c is lost from the membrane during manipulations. The discrepancy may be due to inaccuracies in the electrophoresis and staining of c relat,ed to the proximity of the c hand to the electrophoretic front..

The F, Subunits Synthcsiwd from I%wnids Do Not O L w -
accumulate-One possihle explanation for the ohservation of a functional F, in memhranes from bacteria containing multicopy plasmids is that the large numher of copies oft he genes coding for the F,, pol-ypeptides results in a n excess o f membrane-bound F,, subunits. An excess of F,, subunits in the membrane might allow a low percentage of the F., suhunits to assemble by a n inefficient, nonphysiological pathway. We have addressed this possibility in the experiments ahove. Electrophoretic quantification of the F,, suhunits from a wildtype hacterium without a plasmid, strain DS410 (uric+), was performed alongside the experiments with the plasmids (Figs.  7 and 9). None of the plasmids we have studied results in an excess of membrane-bound F,, suhunits, with the exception of the amount of c present in DK.3/pDqJK20. DS410 membranes contain more of the a and h suhunits than the memhranes from strain DK.3 (AuncR-D) containing the plasmids pIhJK20 (ach) or pDdK19 (achn') (Table I). Hence, overaccumulation of the F, suhunits in the membrane is not the reason we observe a functional F,.
We have attempted to delete the genes uncH, -A, 4 , -D, and -C, coding for the Fl polvpeptides, from the chromosomal operon in strain DS410, leaving onlv the unr promoter, uncl. and the genes coding for the F,. pdypeptides in the chromo-

Y -4
Assembly of the E. . coli F,, Subunits T h e role of the F1 in F., assemhlv was examined hy romparing the function of the F,, from cells svnthesizing onlv the F,, polypeptides to the F,. function from cells synthesizing hoth the Fl and F,, polypeptides. If the F,, resulting from DK3/ pRPG45 is assemhling inefficientlv due to the ahsence of the F,, then the addit.ion of F, during synthesis of the F,, should increase F,, function suhstantiallv. We compared memhranes from DK3/pRPG45 (AuncH-L)/arhd) t o memhranes from DK3/pRPC45/pD.JK35 (AunrH-Dlnchdldtrylij~ ). The membranes were treated wit.h low ionic strength (ST21 huffer to strip F, suhunits away from the F., sector. Purified FI was added hack to memhrane samples using the reconstitution procedure in order to measure ATP-driven fluorescence quenching. The F,, activity in the ahsence of the F, was 6 5 ' ; of the activity in the presence of the F, (Fig. IO, I ) and t,').
Similarly, the F,, activity from DK?/pD.JKIS (arhtr') is ahnut 65% of that from DK3/pRPG54 (achhtrydr) (Fig. 5). T h e 35"; increase in F,, activity due to the simultaneous svnthesis of F, and F,, suhunits is a relativelv small increase which may result from the protect ion of the F,, from degradation in L i L w . If the FI protects the F,,, then the Iahility of the F,, in t-itw should he detectahle in the ahsence of the F,. Consistent with this, we find that strain DK:l/pRPC4.5 incuhated fnr I 2 h in stationary phase loses F,, activity ( Fig. IO. H and ('). This result may explain our previous findings with the plasmid pRP(X5: we were unable to detect formation of a functional F,. unless the FI subunits were also svnthesized (16). In our previous experiments, hacterial cultures were incuhated at 37 "C fnr ahout 2 h in st,ationary phase hefore harvesting (16). In the ahsence of the F, suhunits, the lahile F,, may have heen degraded during the stationary phase, preventing the detection of F,, activity.

DlSCllSSlON
Cross-linking experiments demonstrate h-h and c-c interactions characteristic of the assemhled F,, sector. T h e c crosslinking pattern ohserved in membranes from cells hearing the F, plasmid pRPG45 (acb6) is identical to the pattern observed for minicell membranes containing an assembled ATP synthase as well as for the purified holoenzyme (10). In other experiments addressing the topology of the F,, the b-b interaction and the c-c interaction have been documented in vivo (7,11). The detection of a b-b interaction in membranes containing the b polypeptide isolated from other F, proteins supports the notion that the F, proteins are competent to form the interactions present in an assembled F,. This 6-b dimer interaction appears to involve the amino-terminal half of the b protein.
F,-mediated proton translocation has been measured in membranes from an E. coli strain containing only the genes uncB, -E, and -F, coding for the F, polypeptides a, c, and b. Since F, function is the most complete measure of F, assembly, these data argue that the F, polypeptides by themselves are sufficient to assemble a proton pore.
The quantification of the amounts of the F, subunits shows that the level of membrane-bound F, activity is proportional to the amount of the F, subunits in the membrane. Comparison of the wild-type bacterium with the plasmid-bearing strains indicates that plasmid-directed synthesis of the a, b, and c polypeptides does not result in an excess of membranebound polypeptides. This is probably because the promoter in the vector sequences of the five plasmids studied are not as strong as the unc operon promoter, which appears to be similar to the lac promoter in strength (20). Expression of the plasmid-borne uncB, -E, and -F genes from a very efficient promoter may have an inhibitory effect on growth rate that we did not observe. Fillingame (21) has observed such an effect in studies of the expression of the F, genes from a plasmid-borne lac promoter.
The F, subunits do not appear to contribute significantly to F, assembly, although the F, may protect the F, subunits from degradation in uiuo. In our previous examination of F, assembly, we concluded that the a and p subunits were essential for F, assembly (16). Currently, we interpret these results to mean that the and subunits are capable of protecting the F, sector, perhaps the b polypeptide specifically, from degradation. Similarly, the lack of protection of the b polypeptide in a strain carrying the polar uncD436 allele may account for the inability to detect the b polypeptide in the membranes from this strain (15). Hoppe et al. (22) have shown that in strains producing reduced amounts of the F1 polypeptides, or none at all, the amount of membrane-bound b subunit decreases with decreasing synthesis of the a and /3 subunits. Furthermore, the mass of subunit b was reduced in strains harvested in stationary phase compared to those harvested in log phase growth (22). Similarly, the presence of b in logarithmic growth phase and its disappearance in stationary phase in an unc mutant with an F, binding affinity of only 30% that of the wild type shows the sensitivity of b towards endogenous proteolysis (13). The susceptibility of b to degradation by exogenous protease has been well characterized and has revealed much about the structure and function of b (13,17,22,23).
In addition to the precautions taken to prevent or reduce degradation during membrane preparation, the bacterial strain used may have contributed to the measurement of proton translocation activity in this report. The minicell strain DK3 (AurzcB-D) transformed with the F, plasmids produced the highest F, activity values of the strains studied.
The strains 1100 and MC4100, deleted for the uncB-D genes and transformed with the F, plasmids, showed 25% or less of the F, activity measured in DK3 with identical plasmids.
Recently, Schneider and Altendorf (14) have reported the reconstitution of the F, in vitro from fractionated F, subunits.
Similarly, Perlin et al. (23) have been able to reconstitute the F, in vivo by mixing two membrane samples in the presence of asolectin. The ability to reconstitute a functional F, in the absence of the F, subunits supports our findings that the F, subunits are sufficient to assemble a functional F,. Klionsky and Simoni (24), in the accompanying report, show that the F, subunits are capable of assembling in the cytoplasm independent of the F, polypeptides. This suggests an assembly pathway in which the F, and F, sectors assemble separately and bind to one another only after each portion is complete. However, the demonstration that the F1 and F, subunits are sufficient to assemble functional Fl and F, domains does not preclude the interaction of subunits from each sector during assembly in vivo. An F, subunit, or the intact F1, may bind to a membrane integral F, subunit, perhaps b, before the attached F, subunit becomes integrated into the assembled F,. Although the assembly of the ATP synthase may involve interactions between the F, and F, sectors, it seems clear that the F1 and F, subunits possess the information necessary to specify the proper assembly of each sector, respectively, in the cytoplasmic and membrane compartments in vivo.