The preferred stoichiometry of c subunits in the rotary motor sector of Escherichia coli ATP synthase is 10

Abstract

The stoichiometry of c subunits in the H⁺-transporting F_o rotary motor of ATP synthase is uncertain, the most recent suggestions varying from 10 to 14. The stoichiometry will determine the number of H⁺ transported per ATP synthesized and will directly relate to the P/O ratio of oxidative phosphorylation. The experiments described here show that the number of c subunits in functional complexes of F_oF₁ ATP synthase from Escherichia coli can be manipulated, but that the preferred number is 10. Mixtures of genetically fused cysteine-substituted trimers (c₃) and tetramers (c₄) of subunit c were coexpressed and the c subunits crosslinked in the plasma membrane. Prominent products corresponding to oligomers of c₇ and c₁₀ were observed in the membrane and purified F_oF₁ complex, indicating that the c₁₀ oligomer formed naturally. Oligomers larger than c₁₀ were also observed in the membrane fraction of cells expressing c₃ or c₄ individually, or in cells coexpressing c₃ and c₄ together, but these larger oligomers did not copurify with the functional F_oF₁ complex and were concluded to be aberrant products of assembly in the membrane.

The ATP made during oxidative and photo phosphorylation is synthesized by closely related enzymes located in the inner membrane of mitochondria, the thylakoid membrane of chloroplasts, and the plasma membrane of eubacteria (1). Recent evidence supports a rotary mechanism for ATP synthesis in which proton-transport-coupled rotation of an oligomeric ring of c subunits in the membrane is coupled to rotary movement of subunit γ between alternating catalytic sites in the α₃β₃ sector of the enzyme (refs. 2–8; Fig. 1). The proton-motive force is thought to drive rotation of the c-ring by a transport mechanism by using half channels that connect the aspartyl-61 proton-binding site of subunit c in the middle of the membrane to the aqueous compartments on either side of the membrane. After H⁺ binding from the entrance half channel at the a₁b₂ stator, the c subunit carrying the proton would rotate nearly 360° before encountering the H⁺ exit channel on the opposite side of the membrane. In such a mechanism, the number of H⁺ transported per revolution of the c oligomer would equal the number of subunit c in the ring. The number of subunit c would also determine the H⁺/ATP stoichiometry (i.e., H⁺ transported per ATP synthesized) and directly relate to the P/O ratio for oxidative phosphorylation, a fundamental parameter of biology (9).

Figure 1.

Rotary model for ATP synthase based on the subunit composition of the E. coli enzyme (8). The F₁ portion of the complex is bound at the cytoplasmic face of the membrane. The proton-motive force drives rotation of a ring composed of c subunits, the number (n) of which is uncertain. Protons enter the assembly through a periplasmic inlet channel and bind to the Asp⁶¹ carboxylate (open circle) of a subunit c at the a₁b₂ stator interface. The protonated binding site then moves from the stator interface into the lipid phase of the membrane, and after n steps the proton is released at an outlet channel on the F₁-binding side of the membrane. The γ and ɛ subunits remain fixed to the top of a set of c subunits so that rotation of the c oligomer also drives rotation of subunit γ within the α₃β₃ hexamer of F₁ to alternatively promote ADP + P_i binding and ATP product release in the β catalytic subunits. The b₂ and δ subunits of the stator hold the α₃β₃ subunits in a fixed position as the γ subunit turns inside them to drive ATP synthesis. The coupled reaction catalyzed by the complex is reversible so that ATP hydrolysis will drive proton transport in the reverse of the direction shown. This figure is modified from ref. 8.

The number of c subunits in F_o is controversial. Direct measurements of subunit ratios in Escherichia coli F_oF₁, after growth of cells on radioactive amino acids, indicated a range of 10 ± 1 c per 3 αβ pairs in the purified F_oF₁ complex or 12 c per 3 αβ in the membrane fraction in which F_oF₁ was overproduced (10). More recently, genetically fused subunit c were generated by insertion of a loop between the C-terminal residue of transmembrane helix-2 (TMH-2) of the first subunit and the N-terminal residue of TMH-1 of the next subunit (11). The genetically fused c₂ dimers and c₃ trimers were both active, which suggested that the stoichiometry was a multiple of two and three. Further, after introduction of Cys into these fused c₂ and c₃ subunits and crosslinking via disulfide bond formation, oligomeric ladders that maximized at c₁₂ (i.e., c₂ × 6 or c₃ × 4) were observed, which suggested a likely maximal stoichiometry of 12 c per oligomeric ring. Two recent observations led us to reexamine the stoichiometry question. First, Stock et al. (12) reported a 3.9-Å-resolution electron density map of a crystallized mitochondrial F₁/c-oligomer subcomplex, lacking the other subunits of F_o, in which there were only 10 c subunits. Second, Seelert et al. (13) reported an SDS-purified c-oligomer preparation from chloroplasts, which, according to atomic force microscopic analysis, organizes into a ring of 14 subunits after reconstitution into lipid bilayers. Explanations for these differences include the possibility of variable stoichiometry in different species, the possible loss of subunits on crystallization of the mitochondrial subcomplex (12), the possible association of c subunits into unnatural oligomeric rings during extraction and reconstitution (13), or the possibility that the expression of fused c subunits in E. coli forced an unnatural arrangement of the c-oligomer (11). In the experiments reported here, we show that the stoichiometry of c subunits in a functional F_oF₁ of E. coli can be varied, but that the preferred number is 10.

Experimental Procedures

Construction and Expression of Fused c₃ and c₄ Subunits.

Genetically fused trimers and tetramers of subunit c containing the I30C substitution in the first and last subunit were constructed by using restriction enzymes cutting at the following positions of the atp operon, with the numbering system of Walker et al. (14): PstI(1561), BamHI(1727), BsrGI(1911), AvaI(1976), HpaI(2162), EcoO109I(2342), PvuII(2376), and SphI(3216). The I30C substitution occurs at nucleotides 1973–1975 just before the AvaI(1976) site. The PstI/AvaI fragment from the plasmid encoding c₁(I30C) (11) was ligated with the 205-bp AvaI/BsrGI fragment from plasmid pC2A (11), which codes the c₂ dimer (the fragment coding respectively the C-terminal region from the first unit, the linker, and the N-terminal region of the second unit from the middle of the c₂ dimer), and ligated into the PstI/BsrGI sites of plasmid pNOC (15) to generate plasmid c2(I30C-1) with I30C substituted in the first or N-terminal subunit. The 681-bp PstI/AvaI fragment from a partial digest of plasmid c2(I30C-1) and 205-bp AvaI/BsrGI fragment of plasmid C2A were ligated into the PstI/BsrGI sites of plasmid c1(I30C) to generate plasmid c3(I30C-1/I30C-3) with I30C substitutions in the first and last subunits of the trimer. The 886-bp PstI/BsrGI fragment and 524-bp BsrGI/HpaI fragment prepared from a partial digests of plasmid c3(I30C-1/I30C-3) were ligated into the PstI/HpaI sites of plasmid pNOC (15) to generate plasmid c4(I30C-1/I30C-4) with I30C substitutions in the first and last subunit of the c4 tetramer. The c3(I30C-1/I30C-3) and c4(I30C-1/I30C-4) plasmids, which in addition code the a, b, and δ subunits of F_oF₁, proved capable of complementing a chromosomal strain from which the a, c, b, and δ subunits were deleted from the chromosome (11), i.e., the transformants were capable of slow growth on succinate minimal medium via oxidative phosphorylation. Derivatives of plasmid pACYC184 (16) carrying the gene for “c-only” (pCO plasmids) were constructed by ligating the BamHI/HpaI fragments from the c3(I30C-1/I30C-3) or c4(I30C-1/I30C-4) plasmids into the BamHI/HinCII sites of plasmid pACYC184. These “c-only” plasmids are designated pCOc3(I30C) and pCOc4(I30C) in the text. Whole operon plasmids were generated in a derivative of plasmid pBR322 (17) in which the rop gene had been deleted (18). In this plasmid, the atp operon from HindIII(870)-NdeI(9531) is ligated between the HindIII and NdeI sites of the pBR322 vector. The c₃(I30C) and c₄(I30C) genes were inserted as EcoO109I fragments between an EcoO109I site in the vector, 49 base pairs before the vector HindIII site, and the EcoO109I(2342) site in the atp operon. These “whole operon” plasmids are designated pWOc3(I30C) and pWOc4(I30C) in the text.

General Methods.

Membranes were prepared in 50 mM Tris-HCl, pH 7.5/5 mM MgCl₂/10% (vol/vol) glycerol [TMG buffer: 50 mM Tris-HCl, pH 7.5/5 mM MgCl₂/10% (v/v) glycerol] by rupture of cells in a French press (19). ATPase-coupled proton pumping was measured indirectly by the quenching of 9-amino-6-chloro-2-methoxyacridine (ACMA) fluorescence (18). Membranes were suspended in assay buffer at a protein concentration of 250 μg/ml and ACMA added to 0.3 μg/ml. To initiate the quenching response, ATP was added to 0.92 mM. To terminate the quenching response, the protonophore SF6847 was added to 0.25 μM. Membrane ATPase activity was measured as described (18), with 0.5% lauryldimethylamine oxide included in the assay buffer to remove the inhibitory ɛ subunit. To crosslink subunits, membranes at 5 mg/ml in TMG buffer were treated with 100 μM I₂ in TMG buffer for 1 h at 22°C. The reaction was stopped by addition of Na₂S₂O₃ and N-ethylmaleimide to 0.4 and 20 mM, respectively, and the membranes collected by centrifugation.

SDS gel electrophoresis and immunoblotting, by using an “ECL System” (Amersham Pharmacia Pharmacia Biotech.) for blot development, was carried out as described in Jones et al. (15). Enhanced chemiluminescence (ECL)-developed immunoblots were scanned in several instances to more quantitatively estimate the proportions of different oligomeric species. In these cases, multiple gel lanes were scanned after different development times to ensure that the scanned density was within the linear limits of the scanner. Control experiments were run to determine whether the efficiency of electrophoretic transfer varied for different sized oligomers. When the electrophoretic transfer time was altered from the routine time of 90 min to 45 or 135 min, all subunits from c₃ through c₁₂ transferred with nearly maximal transfer occurring by 90 min. The proportion of larger subunits did increase somewhat between 45 and 90 min but then remained constant at the 135-min interval. We interpret this as indicating that electrophoretic transfer of subunits from the SDS gel to the polyvinylidene difluoride membrane is essentially complete at 90 min, and that the blots accurately sample the total protein content of the gel.

Purification and Reconstitution of F_oF₁ Complex.

The complete F_oF₁ ATPase complex was purified by slight modifications of the procedure used by Foster and Fillingame (10, 19). Treated membranes were first suspended at 5 mg/ml and washed with 1 mM Tris-H₂SO₄, pH 8, buffer containing 0.5 mM Na₂EDTA, 10% (vol/vol) glycerol and 6 mM p-aminobenzamidine to remove extrinsic membrane proteins. Washed membranes were resuspended at 12–15 mg/ml in 50 mM Tris-HCl, pH 7.5/1 mM MgSO₄/10% (vol/vol) methanol buffer containing 1 mM phenylmethylsulfonylfluoride, 6 mM p-aminobenzamidine, and solid KCl then added to a final concentration of 1 M. The F₁F_o complex was then extracted by addition of Na-deoxycholate and Na-cholate to final concentrations of 0.5%. After 10 min on ice, the suspension was centrifuged at 186,000 × g for 60 min at 4°C and 0.4 ml of the soluble supernatant solution applied to a 4.4-ml linear 10–35% (vol/vol) glycerol gradient prepared in 10 mM Mes/10 mM tricine/1 mM MgCl₂ buffer made pH 7 with NaOH, which also contained 0.1% phosphatidyl-choline (Sigma no. P3644) and 0.28% Na-deoxycholate. The gradients were centrifuged at 237,000 × g_avg for 4 h at 4°C in a SW55Ti Beckman rotor. The gradients were fractionated into 0.5-ml fractions. The peak fractions shown in Figs. 4–6 correspond to those showing maximal dicyclohexylcarbodiimide-sensitive ATPase activity (19) and, after reconstitution into liposomes, exhibiting maximal ATP-driven quenching of ACMA fluorescence. Gradient fractions were reconstituted into acetone-precipitated soybean phosphatidylcholine (Sigma no. P3644) by the dialysis method of Kagawa and Sone (20). Aliquots of 0.2 ml were added to 8 mg of dried phosphatidylcholine and Na-cholate, Na-deoxycholate, and Na₂EDTA added to final concentrations of 1.6%, 0.8%, and 2 mM, respectively, in a final volume of 0.3 ml. Dialysis was carried out at 4°C versus 10 mM Tricine-NaOH, pH 8/2.5 mM MgSO₄/0.2 mM Na₂EDTA buffer.

Figure 4.

Crosslinked products generated in membranes from cells expressing c₃(I30C) and c₄(I30C) subunits. All samples were prepared after crosslinking of membranes with I₂. The oligomeric size of the crosslinked product is indicated on the vertical axis. (A) Membranes prepared from cells expressing pWOc3(I30C) or pWOc4(I30C) individually or cells expressing both and pWOc4(I30C) and pCOc3(I30C). The deoxycholate/cholate extract and glycerol gradient purified F_oF₁ fractions from the pWOc4(I30C)/pCOc3(I30C) membranes is also shown. (B) Glycerol gradient fraction shown in A after prolonged development. (C) Glycerol gradient F_oF₁ fractions prepared from the c₃ or c₄ membranes shown in A. (D) Scan of lanes of (c₄ + c₃) membrane and glycerol gradient shown in A.

Figure 6.

Cu(II)-phenanthroline catalyzed crosslinking of monomeric subunit c with double Cys substitutions. (A) Membrane fraction of c₁(Cys²¹/Cys⁶⁵) with (+) and without (−) crosslinking. (B) The crosslinked membrane, deoxycholate/cholate extract, and purified F_oF₁ from the c₁(Cys¹¹/Cys⁷⁵) mutant are shown. Crosslinking was carried out as described (15).

Results

Test of Preferred Stoichiometry by Coexpression of c₃ and c₄ Fused Subunits.

In the key experiment described here, we have coexpressed genetically fused c₃ trimers and c₄ tetramers in the same E. coli cell (Fig. 2A). If mixing of the subunits occurs in the membrane of these cells, it should be possible to form a c₁₀ oligomer (Fig. 2B). To detect formation of the possible c₁₀ oligomer, we have introduced a Cys at position 30 into transmembrane helix-1 of the first and third unit of the c₃ trimer and the first and fourth unit of the c₄ tetramer (Fig. 2B). The approximate position of the Cys-30 substitution at the hub of the c-ring, based on the oligomeric model of Dmitriev et al. (21), is indicated in Fig. 2B. If a c₁₀ oligomer forms on mixing of c₃ and c₄ in the same membrane, then c₆, c₇, and c₁₀ crosslinked products should be observed on crosslinking (Fig. 2B). If c₁₂ is the predominant oligomeric species, then crosslinked products corresponding to c₆, c₈, c₉, and c₁₂ should predominate, and the c₇ and c₁₀ species should be missing.

Figure 2.

Strategy for testing the preferred number of subunit c in E. coli F_oF₁ complex. (A) System used for coexpression of c₃(I30C) and c₄(I30C) substituted proteins in one cell. The eight genes of the atp (ATP synthase) operon are deleted from the chromosome of the cell shown. In the scheme shown, the c₄ subunit is expressed from a pBR322-derived plasmid carrying genes for the whole atp operon (pWOc4), with transformant cells selected on the basis of plasmid-encoded ampicillin resistance (Amp^r). The c₃ subunit is expressed by itself from a pACYC184-derived plasmid with transformant cells being selected for via chloramphenicol resistance (Cm^r). (B) Coexpression of c₃(I30C) and c₄(I30C) in the same cell should lead to formation of a c₁₀ decamer if that is the preferred structure in F_o. The figure shows the positions of the I30C-substituted cysteine in the first and last subunit of the c₃ trimer and c₄ tetramer. The approximate shape of a cross section of subunit c at the position of the I30C substitution is indicated (21). Cys–Cys crosslinking is expected to yield c₆, c₇, and c₁₀ products.

In the configuration shown in Fig. 2A, the c₄ tetramer is expressed from a pBR322 derived plasmid encoding all of the genes of the F_oF₁ complex [pWOc4(I30C)], where pWO denotes “whole operon” plasmid, and the c₃ trimer is expressed by itself from a pACYC184-derived plasmid [pCOc3(I30C)], where pCO denotes “c only” plasmid. On growth of these cells, both c₃ and c₄ were incorporated into the membrane with the amounts of c₃ exceeding the amount of c₄ by a 1.6 ± 0.1/1 molar ratio, based on scans (n = 11) of immunoblots of membrane fractions (see Experimental Procedures for controls on quantitation). When the subunits were expressed in the opposite configuration, i.e., pWOc3(I30C)/pCOc4(I30C), the c₃(I30C) and c₄(I30C) subunits were expressed in an equimolar ratio, i.e., 1.0 ± 0.1/1 ratio (n = 9 scans).

Function of Fused c₃ and c₄ Subunits.

The ATPase-coupled proton-pumping activity of F_oF₁ made from the c₃(I30C) and c₄(I30C) fused products is shown in Fig. 3. In both cases, ATP hydrolysis resulted in formation of a ΔpH, as indicated by the quenching of ACMA fluorescence, but the c₃(I30C) species of enzyme was considerably more active than the c₄(I30C) species. Activity was intermediate in membranes in which both types of subunits were expressed. Membranes prepared from cells expressing both subunit types exhibited somewhat greater ATPase activity than cells expressing c₃(I30C) or c₄(I30C) only, and the activity was inhibited by dicyclohexylcarbodiimide (Table 1). ATPase activity measured in the presence of the detergent lauryldimethylamine oxide is the best measure of the total amount of F_oF₁ incorporated into the membrane. The overexpression of c subunits, or coexpression of c₃ and c₄, may thus facilitate greater assembly of the F_oF₁ complex.

Figure 3.

ATP-driven proton translocation with membrane vesicles from cells expressing c₃ or c₄ by themselves, or cells coexpressing c₃ and c₄. Coexpression in the pWOc4/pCOc3 and pWOc3/pCOc4 configurations are indicated as (c₄ + c₃)I30C and (c₃ + c₄)I30C, respectively. The quenching of ACMA fluorescence was used as an indicator the pH gradient established by ATPase-coupled proton pumping. At the times indicated, ATP was added to 0.92 mM and, to terminate the quenching response, the protonophore SF6847 was added to 0.25 μM to make the membranes proton permeable.

Table 1.

Membrane ATPase activities from cells expressing c₃ and c₄ fused subunits

Plasmids expressed	ATPase activity*		Inhibition by DCCD†
	−LDAO	+LDAO
pWOc3(I30C)	0.61	3.48	73%
pWOc4(I30C)	0.81	4.40	70%
pWOc3(I30C)/pCOc4(I30C)	1.20	5.45	72%
pWOc4(I30C)/pCOc3(I30C)	1.45	7.10	55%

^*Membrane ATPase activity measured with and without 0.5% LDAO, which activates F₁ ATPase by disassociation of the inhibitory ɛ subunit. Activity expressed in units of μmol min⁻¹ mg⁻¹. 

^† Membrane ATPase activity was measured after preincubation of membranes at 10 μg/ml in ATPase assay buffer with 50 μM DCCD for 20 min at 22°C. 

Crosslinking Generates c₁₀ Oligomers.

To detect formation of the possible c₁₀ oligomer, we introduced a Cys at position 30 into transmembrane helix-1 of the first and third unit of the c₃ trimer and the first and fourth unit of the c₄ tetramer (Fig. 2B). The Cys substitution at this position was chosen because the monomeric I30C subunit c forms a high yield of dimers, i.e., >95% (15). Further, dimerization of the c₁(I30C) subunits resulted in minimal reduction of ATPase-coupled proton pumping and must therefore result in minimal structural perturbations.† The crosslinking of coexpressed subunits is shown in Fig. 4. The first two lanes show the crosslinking of c₃(I30C) or c₄(I30C), respectively, when each was expressed individually from the pBR322-derived whole operon plasmids, i.e., pWOc3(I30C) and pWOc4(I30C). Subunits were detected by immunoblotting of the plasma membrane fraction. With the c₃ membrane, c₆ and c₉ were the predominant crosslinked products with a smaller amount of c₁₂ also detected. With the c₄ membrane, the predominant products were c₈ and c₁₂, but a significant amount of c₁₆ was also observed. When the c₃(I30C) and c₄(I30C) subunits were coexpressed from the pWOc4(I30C)/pCOc3(I30C) configuration (shown in Fig. 2A) and membranes crosslinked, distinct products corresponding to c₆, c₇, c₈, c₉, c₁₀, c₁₁, and c₁₂ were observed. Note that the c₇ signal is slightly more intense than either c₆ or c₈.

A significantly different pattern was observed when the F_oF₁ complex was extracted from these membranes, by using a cholate and deoxycholate detergent mixture, and then purified by glycerol density gradient centrifugation. Crosslinked oligomers of sizes greater than c₁₀ were not observed as is most apparent from the overexposed lane of the purified glycerol gradient fractions (Fig. 4B), where there is no hint of oligomers greater than c₁₀. Note that the predominant dimeric species in the purified F₁F_o fraction was clearly c₆ (c₃ + c₃) and c₇ (c₃ + c₄), with lesser amounts of c₈ (c₄ + c₄) being found (Fig. 4 B and D). Nearly equal amounts of trimeric c₉ (c₃ + c₃ + c₃) and trimeric c₁₀ (c₃ + c₃ + c₄) were observed, so the tendency to form rings of 9 or 10 seems to be nearly identical. The F_oF₁ complex was also prepared from crosslinked membranes in which c₃ and c₄ were expressed individually (Fig. 4C). In the case of c₃, the largest crosslinked oligomer corresponded to c₉. In the case of c₄, the largest crosslinked oligomer corresponded to c₈. We conclude that the crosslinked oligomers in membranes of sizes greater than c₁₀ are not likely to be components of a functional F_oF₁ complex but rather may be an artifact of aberrant assembly in the membrane.

When coexpressed in the configuration opposite to that shown in Fig. 2A and Fig. 4, i.e., with plasmids pWOc3(I30C) and pCOc4(I30C), the c₃ and c₄ subunits were incorporated into the membrane in an equimolar ratio. After crosslinking in this case, c₇ and c₈ were the predominant dimeric products and lesser amounts of c₆ dimer seen (Fig. 5A). Prominent trimeric products of c₉, c₁₀, c₁₁, and c₁₂ were apparent in the membrane fraction. However, the c₉ and c₁₀ trimers were again preferentially extracted and purified with the F_oF₁ fraction (Fig. 5 A and B). In this case, a distinct c₁₁ trimer was also detected in the solubilized detergent extract but was much less evident in the purified F_oF₁ fraction (Fig. 5B). Although it is possible that trace quantities of c₁₁ and c₁₂ are present in the F_oF₁ fraction, the key point is that most of oligomeric species > c₁₀ are lost as the complete enzyme is purified. Membranes from cells coexpressing c₃ and c₄ in the two configurations show equivalent ATP-driven ACMA quenching activity (Fig. 3). Because relatively less c₆ and c₉, and relatively more c₇ and c₁₀, were found in the pWOc3(I30C)/pCOc4(I30C) membranes, most of the activity can be attributed to the c₁₀ form of the enzyme.

Figure 5.

Crosslinked products generated in membranes from cells expressing both the pWOc3(I30C) and pCOc4(I30C) plasmids. Membranes were treated with I₂ and the F_oF₁ complex extracted and purified by glycerol gradient centrifugation as described in the text. (A) Lanes from immunoblot. (B) Scan of crosslinked products ≥c₆ in the three lanes shown.

Monomeric Subunit c Forms a c₁₀ Oligomer in F_oF₁.

The conclusion that c₁₀ is the preferred stoichiometry actually agrees well with previous crosslinking experiments with a doubly Cys substituted c₁ monomers, where crosslinking usually generated products maximizing at the position of the c₁₀ oligomer (15). Such an example is shown in Fig. 6A with the Cys²¹/Cys⁶⁵ doubly substituted c₁ monomer. In the study of Jones and Fillingame (11), one doubly Cys-substituted c₁ protein (from the Cys¹¹/Cys⁷⁵ mutant) generated crosslinked products in the membrane that appeared to maximize at c₁₂. As a followup to the experiments reported here, the crosslinked F_oF₁ complex was purified from c₁(Cys¹¹/Cys⁷⁵) membranes. Products migrating at positions >c₁₀ were not observed in the F_oF₁ fraction (Fig. 6B); again, these products appear to be minor artifacts of assembly, formed in the membrane, but not assembling into a functional F_oF₁ complex.

Discussion

From the experiments reported here, we conclude that the preferred stoichiometry of c subunits in the oligomeric ring of E. coli F_o is 10. The experiments also indicate that c-oligomers composed of nine or eight subunits exhibit at least partial function. Although crosslinked products composed of more than 10 c subunits were observed in membranes, these oligomers did not copurify with the functional F_oF₁ complex and are probably artifacts of assembly. In the key experiment done here, we used genetically fused c₃ and c₄ subunits to test whether 10 or 12 was the preferred stoichiometry. After coexpression of the c₃ and c₄ subunits, crosslinking generated a prominent c₁₀ product, which was the maximum sized product seen in the purified F_oF₁ complex. Oligomers larger than c₁₀ were observed in the membrane fraction, but these did not extract with or copurify with the active F_oF₁ complex. We envision occasional assembly of oligomeric rings with >10 c subunits in the membrane, perhaps because of expression of c₃ and c₄ in nonstoichiometric ratios, but also envision that these larger rings do not assemble with subunits a or b to form a functional F_o or bind F₁. The differences in the structure of rings composed of 9–12 subunits are probably quite minor.‡ However, a cell very likely regulates the stoichiometry at a fixed number under normal physiological conditions, as discussed below.

The c₃ and c₄ subunits were coexpressed in two configurations, which resulted in slightly different ratios of subunit incorporation into the membrane and slight differences in the relative amounts of different types of crosslinked dimers and trimers in the membrane. Importantly, in either expression configuration, c₇ was the most prominent dimeric product and c₁₀ the most prominent trimeric product in purified F_oF₁, supporting the conclusion that c₁₀ is the preferred species. The relative amounts of c₆ and c₈ dimer varied depending on whether c₃ or c₄ was expressed in excess. Because c₃ or c₄, when expressed individually, form oligomers in purified F_oF₁ composed of c₉ (c₃ + c₃ + c₃) or c₈ (c₄ + c₄) only (see Fig. 4C), the membrane fraction of cells coexpressing c₃ and c₄ is likely to contain a mixture of F_oF₁ composed of c₈, c₉, c₁₀, and possibly c₁₁ rings. However, on purification of F_oF₁, the only well defined trimers seen were c₉ and c₁₀. Other oligomers, if present, could be present only in trace amounts.

One might argue that the c-ring formed from the genetically fused subunits is itself unnatural. However, the crosslinking data with monomeric subunits presented here also support a preferred stoichiometry of c₁₀ versus c₁₂. In the case of the c₁(Cys²¹/Cys⁶⁵), crosslinking in the membrane maximized at c₁₀ with no hint of a c₁₂ product (Fig. 6A). With c₁(Cys¹¹/Cys⁷⁵) membranes, some c₁₂ crosslinked product was observed in the membrane but not in the purified F_oF₁ complex (Fig. 6B). That crosslinked products with masses > c₁₀ do not copurify with functional complex suggests they are aberrant assembly products formed only in the membrane. The formation of oligomers > c₁₀ in the membrane clearly depends on the position of the double Cys substitutions in the protein. In the case of c₁(Cys¹¹/Cys⁷⁵), the 11 and 75 regions of the protein are clearly more flexible (15, 18) and may account for minor amounts of aberrant product formation. The position of the genetically introduced cysteine is clearly critical in the formation of different oligomers as c₉ products are formed in Cys²¹/Cys⁶⁵ and other membranes (15) but not with the Cys¹¹/Cys⁷⁵-substituted protein.

Despite the clear formation of a c₁₀ crosslinked product in purified F_oF₁ from either fused or monomeric subunits, one might suggest that c₉ is the preferred stoichiometry on the basis of the experiment shown in Fig. 3. The c₃ membrane vesicles show greater proton-pumping activity than membrane vesicles in which c₃ and c₄ are coexpressed. However, the low activity of the c₄ enzyme, also present in the c₃/c₄ hybrid membranes, should be considered in any accounting of an average lower total activity. ATP-driven ACMA quenching activity was observed with pWOc3(I30C)/pCOc4(I30C) membranes, which matched that of the pWOc4(I30C)/pCOc3(I30C) membranes. Because lesser amounts of c₉ crosslinked product were observed in the pWOc3(I30C)/pCOc4(I30C) membranes (Fig. 5), most of the activity observed can be attributed to the c₁₀ species of enzyme.

The mechanism by which the cell regulates assembly of the oligomer is of interest. The cell would be at a bioenergetic disadvantage if it were to assemble functional mixtures of c oligomers with different c stoichiometries. At thermodynamic equilibrium for ATP synthesis, the free energies of the proton-motive force (Δμ̄H⁺)/F and phosphorylation potential (ΔG_p = ΔG′^o + RTln[ATP]/[ADP][Pi]) are in balance, i.e. (H⁺/ATP)Δμ̄H⁺ = ΔG_p. With a mixture of two enzymes, e.g., c₉ and c₁₀, at least one of the enzymes would not be in thermodynamic equilibrium with Δμ̄H⁺. Depending on the displacement from equilibrium, the F_oF₁ complex with the higher c_n stoichiometry would preferentially synthesize ATP, whereas the complex with the lower c_n stoichiometry would preferentially hydrolyze ATP, resulting in an uncoupled system. It seems likely that cells precisely control c-oligomer assembly to prevent such mixtures and the potential problem of uncoupling.

The experiments presented here do not answer the question of whether there are differences in subunit c stoichiometry in different species. Indeed, the presence of a natural subunit c trimer in the Methanococcus jannascjii archaebacterial ATP synthase argues for such diversity (23), although the absence of a proton-binding carboxylate (equivalent to Asp⁶¹) in the first unit of the trimer suggests that the mechanism of rotary catalysis, if applicable, might be more complicated. The conclusion that 10 is the preferred stoichiometry in the c oligomer of E. coli does support the suggestion of a c₁₀ oligomer in yeast mitochondria, a suggestion based on the crystal structure of an F₁-c₁₀ subcomplex (12). A mitochondrial ATP synthase with a c₁₀ oligomer would use 3.3 H⁺ per ATP synthesized. If the consensus values for proton pumping by the mitochondrial respiratory chain are correct (9, 24), an H⁺/ATP stoichiometry of 3.3 is quite compatible with the P/O ratios measured for mitochondrial oxidative phosphorylation by Hinkle et al. (25) and would suggest a P/O ratio of 2.3 for NADH-linked substrates and 1.4 for succinate versus the classical textbook values of 3 and 2, respectively (9). The H⁺/ATP pumping ratio predicted for a c₁₀ oligomeric ring also fits well with a thermodynamically determined ratio of H⁺/ATP = 3, estimated for whole cells of E. coli by measurements of Δμ̄H⁺ and ΔG_p (26).

Abbreviation

ACMA

9-amino-6-chloro-2-methoxyacridine