A Mechanism of Proton Translocation by F,F, ATP Synthases Suggested by Double Mutants of the a Subunit*

Three amino acid residues in the a subunit of the Escherichia coli FIFo ATP synthase are essential for proton translocation: Arg”, GluZ1’, and His245. In this study, the essential glutamic acid has been relocated to position 252 with retention of function. It had been known that GlnZ5’ can be replaced by Glu without significant effect. To test whether Q252E would function in the ab-sence of G ~ u ~ ~ ’ , a “site-directed second-site suppressor’’ experiment was designed. Saturation mutagenesis was applied to residue GluZ1’, and 14 different amino acid substitutions were isolated, five of which permitted growth on succinate minimal medium at 37 “C: Asp, Lys, Gly, Ala, and Ser. These results indicate that Q252E can provide the essential carboxyl group normally provided by GluZ1’, but that strict requirements are placed on the residue at position 219. We interpret these results to mean that the Q252E must occupy, at least partially, the normal position of GluZ1’. We present a novel mechanism of proton translocation by FIFo ATP synthases that includes a rotating oligomer of c subunits, in which the Asp61 of two c subunits simultaneously interact with GluZ1’ andArg21° of the a subunit. This mechanism can be adapted for both mitochondrial and sodium-driven bac-terial ATP synthases.

in the middle of the second transmembrane region, and has been implicated in proton translocation. A nonpolar carbodiimide, NJV-dicyclohexylcarbodiimide (DCCD),' reacts specifically with this residue and inhibits ATP hydrolysis, ATP-driven proton translocation, and passive proton movement through F,. Recent NMR analysis of the isolated c subunit has provided evidence for two interacting a-helices, supporting the hairpin model (12, 13).
The b subunit contains a single hydrophobic region, located at the extreme amino terminus, that is long enough to span a membrane as an a-helix. The bulk of the protein extends into the cytoplasm and is thought to interact with F, subunits (14- 17). No evidence has been presented that would indicate a direct role for the b subunit in proton movement. Recent studies have shown that the polar domains of two b subunits form an elongated dimer with considerable a-helical content (18).
Structural information about the largest subunit, a, is less certain. Models have been offered ranging from four to eight transmembrane spans based on sequence analysis (14,(19)(20)(21)(22) and gene fusions (23,24). No determinations of secondary structure have been done, but two regions of about 21 residues in length can be modeled as a-helical, with conserved, polar faces (22). The a subunit has been subject to extensive mutagenesis, leading to the view that it has an essential role in proton translocation. One residue that has been identified as essential is Arg2l0, which cannot be replaced by Lys, Gln, Ile, Val, or Glu without complete loss of ATP-driven proton translocation (25)(26)(27)(28). Other residues that may have essential functional roles include Glu219 and His245 (25,27,29,30).
Proton translocation mechanisms have been discussed recently for F, systems (3, [31][32][33]. These include "proton wires," H,O-lined channels, and crown ether coordination of H,O+. Models proposed have included piston-like movement of bundles of three c subunits (3), and rotations of a and b subunits within a ring of c subunits (21, 34). So far, no detailed proposals of proton movement through F, have been offered. Based on insights provided by a series of double-mutants, and a wide range of other evidence, we propose a novel mechanism of proton translocation coupled to rotation of an oligomer of c subunits. EXPERIMENTAL PROCEDURES Materials-Restriction endonucleases, T4 DNA ligase, and DNA polymerase (Klenow fragment) were obtained from New England Biolabs. CP~~S-~ATP was obtained from Amersham Corp. Reverse transcriptase and silver sequence reagents were from Promega. 9-Amino, 6-chloro, 2-methoxyacridine (ACMA) was obtained from Molecular Probes. Other chemicals were of the highest quality commercially available. Synthetic DNA was purchased from Operon Technologies. E. coli Strains-XL1-Blue (F'::TnlOproA+B+ ladq D(lacZ)M15lrecAI endAl gyrA96 (Nul') thi hsdRl7 (r-rnd supE44 relAl lac) was used for subcloning and mutagenesis. Str& RH305 (35) (bglR thi-1 HfrPOl The abbreviations used are: DCCD, NJV"dicyclohexy1carbodiimide; ACMA, 9- Construction ofPlasmids and Mutagenesis-Standard techniques for the isolation and manipulation of plasmids were used (37). The construction of pSBV16 from pSBVl0 (38) is outlined in Fig. 1. pSBV12 (-3680 bp) was constructed by digestion of pSBV 10 (-3900 bp) with HindIII and AseI, followed by S1 nuclease digestion and religation. The restriction sites for HindIII and AseI were eliminated. pSBV13 (-3050 bp) was constructed by digesting pSBV12 with BamHI and PuuI and ligating the large fragment to the following synthetic DNA

GATCCTGAATGTGCCGTGGGCCATTTTGGCCGCGGGCCCGCGCGGGTCCTGACGAT GACTTACACGGCACCCGGTAAAACCGGCGCCCGGGCGCGCCCAGGACTGC
This regenerated the BamHI and PuuI sites and created new sites for SfiI, SacII, and Eco0109I. pSBV14 (-2400 bp) was constructed by digesting pSBV13 with AuaI and Ec047111, generating blunt ends with DNA polymerase (Klenow fragment) and religation. The AuaI site was regenerated, but the Eco47III site was lost. pSBV15 (-3000 bp) was constructed by digesting pSBV14 with BamHI and ligating with a 617-bp BamHI fragment from pSBV11. This regenerated the 5' end of uncB. pSBV16 (-3100 bp) was constructed by digesting pSBV15 with SfiI and Eco0109I and ligating with the following synthetic DNA

TCCACATATTAATCATTACGCTGNNNGCCTTCATCTTCATG AAAAGGTGTATAATTAGTAATGCGACNNNCGGAATAGAAGTACCAG
This produced the wild type plasmid pSBV16 and various substitutions at position 252. To construct the double mutants, pSBV16 (Q252E) was digested with EarI and Ec1136II (SacI isoschizomer) and ligated with the following synthetic DNA:

GCAACATGTACGCCGGCNNS GTAGATGCGGCCGNNS
This eliminated theEcZl36II (Sad) site and generated a new, unique site NgoMI, used in screening. To make the double mutant E219Q/Q252E, which did not appear in the first group, the following synthetic DNAwas ligated to pSBV16 (Q252E), previously digested with EarI and SacI:

GCAACATGTACGCCGGCCAG GTAGATGCGGCCGGTCGA
Following mutagenesis, the DNA sequences of plasmids were determined using W~~S -A T P a s described before (39), or by a silver staining technique (Promega) according to the manufacturer's instructions.
Analysis of Mutants-Growth yields were done in minimal A medium with 6 mM glucose or 1% succinate as described elsewhere (38). Preparation and analysis of membranes were performed as previously described (39). ATP hydrolysis assays of membrane fractions were performed as previously described (38) or essentially as described in Lotscher et al. (40) in 25 mM Tris acetate, pH 7.5,3 m " T , 5 mM MgCI,, 5 mM KCN, 2 mM phosphoenolpyruvate, 0.25 mM NADH, pyruvate kinase (6 units), and lactate dehydrogenase (8 units) at 37 "C using a Beckman DU70 spectrophotometer. Protein assays were done with the detergent compatible protein assay (Bio-Rad) using bovine serum albumin as a standard.

Saturation Mutagenesis of Gln252-Previous workers (27,36)
have established that the strictly conserved residue Ginz5' can be replaced by Glu with little effect, and that even replacement by Leu does not totally eliminate proton translocation. We extended these results by analyzing fifteen amino acid replacements for GlnZ5' produced by cassette mutagenesis. The results of growth experiments indicated that the following seven mutations, when expressed in uncB mutant strain RH305, are able to grow in minimal medium supplemented with succinate: Glu, His, Ser, Asn, Cys, Val, and Leu. The mutations that lead to a n inability to grow on succinate minimal medium were Gly, Pro, Tyr, Phe, Ile, "rp, Lys, and Arg. Growth yields in minimal medium with limiting glucose followed a similar trend (data not shown). In addition to Glu, other polar residues are tolerated at this position. Aromatic, positively charged, and conformational residues are not tolerated. These results indicate that charge, polarity, and size are important aspects of functionality at this position.
Analysis of membrane vesicles prepared from mutant strains containing Glu, Asn, Ser, or Val at position 252 showed that all were capable of ATP-driven proton translocation to the extent suggested by the growth yield measurements (Fig. 2 A ) (30) have shown that residue GluZ1' can be replaced by Asp with no loss of function, but that replacement by His results in very low rates of ATP-driven proton translocation and inability to grow on succinate minimal medium. Other replacements tested (30,411, Gln and Leu, resulted in no ATP-driven proton translocation. We sought to test the possibility that position 219 could tolerate other substitutions if position 252 were Glu. The starting plasmid for this experiment was the double-acid construct Q252E. Using unique restriction sites which flanked the 219 position, Ear1 and Ec113611, this position was subjected to saturation mutagenesis. After mutagenesis fourteen different amino acid substitutions at this position were identified. Mutants were analyzed for growth on minimal medium with succinate or limiting glucose ( Table I). Three mutants grew well: Asp, Lys, and Gly. Ala and Ser were marginal growers at 37 "C, but grew better at room temperature. Two others, Pro and Val, grew marginally at room temperature only. The rest were unable to grow on succinate minimal medium.
Analysis of membranes from strain RH305 containing these mutations showed that the magnitude of ATP-driven proton translocation mirrored the growth yields (Fig. 3). None of the mutants examined showed enhanced proton permeability by stripped membranes (data not shown). The results of ATP hydrolysis assays are shown in Table 11. The specific activities of the membranes tested were all similar, at both pH 9.1 and at 7.5. The fraction of ATPase activity that remained bound to membranes was also similar in all of the mutants, ranging from 88 t o 96% of the value seen in the wild type strain. Differences were seen in the sensitivity of the membrane-bound ATPase activity to DCCD. Membranes with E219WQ252E and E219G/ Q252E were similar to the wild type, with about 40% sensitivity. The mutants which were more marginal growers on succinate minimal medium, E219A/Q252E, E219S/Q252E, and Growth on succinate was characterized by the appearance of colo-Growth yields were determined by measuring A,,, and are expressed as a percentage of the wild type growth. Results are an average of at least two determinations that did not vary by more than 22%.

DISCUSSION
GlnZ5' is a strictly conserved amino acid residue that lies in the center of a very hydrophobic span of about 21 amino acids. Previous analysis (22) has shown that if this region were a-helical, Gln252 would be located on a polar, conserved face. Thus, it could face a water-accessible proton channel, such as have been found in bacteriorhodopsin (42). The results of extensive mutagenesis in this study confirm earlier findings (27,36), and indicate a likely polar environment for this residue. Furthermore, the proton permeability caused by the Asn substitution suggests that this residue is located at a key position with respect to proton movement through F,.
Glu'" is one of a pair of residues, along with that are essential in the FIFo ATP synthase of E. coli, and are also found in a wide variety of other species. Results presented here show that Q252E can fulfill, at least partially, the essential role of GluZ1'. Previous studies (30) had shown that only Asp could replace G~u~~~ with retention of normal function. The results * Ratio of specific activity at pH 9.1 to pH 7.5. Results are the average of two measurements that did not vary by more than * 0.2.
Results shown are the fractions of total ATPase activity bound to membranes, expressed relative to that found in the wild type (wt). The wild type value was 75%.
Results are expressed as the percentage of membrane-bound ATPase activity that is inhibited by a pretreatment with 50 PM DCCD at 37 "C. The wild type value (39%) is lower than we normally measure (60-75%), and may be related to the new expression vector pSBV16.
presented here show that when GlnZ5' is replaced by Glu, significant function is retained when Glu'" is also replaced by Asp, Lys, or Gly. It is not surprising that the double mutant E219D/Q252E is functional, since E219D is. The high activity of the E219G/Q252E double mutant, and the more limited function seen with Ala and Ser, suggest that size of the residue at position 219 is a key determinant as to whether Q252E can substitute for GluZ1'. The double mutant E219Q/Q252E, which switches the two residues, was found t o be nonfunctional. One possibility is that Q252E can most effectively replace GluZ1' if it can occupy its vacated space. Of course, less direct explanations are possible, but the dependence seen is on size alone, as opposed to polarity or structural analogy.
The effectiveness of the E219WQ252E double mutant is less easy to rationalize, especially since Q219K has been shown to be nonfunctional. One possibility is that the lysine side chain interacts with the Q252E residue, in such a way as to facilitate proton translocation. Alternatively the €-amino group might participate in another interaction, such as with membrane phospholipids, resulting in a more active conformation of the protein.
In mitochondrial enzymes, as well as some bacteria, the positions of GluZ1' and His245 are reversed (30). Those workers showed that the double mutant in E. coli E219WH245E is marginally functional, and grows slowly on succinate minimal medium, while the single mutants were unable to grow on succinate. One interpretation is that these residues interact directly, and therefore, switching the residues preserves the interacting pair. However, ATP-driven proton translocation by the E219HRI245E double mutant is not much greater than that by the single mutants E219H or H245E.
An alternative possibility is that these residues function more independently in proton translocation. In the E. coli a subunit both positions might only require amino acids that can protonate and deprotonate a t moderate pH. Sequence analysis indicates that 219 is the more important position (see Fig. 4A). All proton-translocating F,F, ATP synthases have Glu, Asp, or His at the equivalent of position 219 in the E. coli a subunit. In some bacteria, which have Glu at position 219, such as Bacillus subtilis, Bacillus megaterium, and Streptococcus pneumoniae, a conserved His is found in the putative third transmembrane spanning region (not shown, but equivalent to position 162 in the E. coli a subunit) and not at the 245 position. In chloroplasts and cyanobacteria, which always contain Asp or Glu at both positions 218 and 219, no conserved His can be found in a with Ala79, the carboxyl terminus. The conserved carboxyl-containing residues, AspG1 in E. coli, are shown in boldface. Note that among the 4 residues on either side of the boldface residues, only the P. modestum protein has more than one with oxygen or nitrogen atoms in the side chains (Q, S, and 2'). All sequences were retrieved from GenBankTM. hydrophobic sequence, although several other conserved acidic residues can be identified (not shown). In contrast, the a subunit of the Na+-driven FIF, ATP synthase from Propionigenium modestum contains Met at the 219 position and Asp at the 245 position.
We propose a model for proton translocation by the E. coli F, complex that includes a prominent role for G~u'~'. It is based primarily on the proximity of GluZ1' t o the critical residue Arg2", and the essential nature of residue 219, or a nearby residue, t o be able t o protonate and deprotonate. The features of this model are the following. 1) Two c subunits interact with the a subunit simultaneously via two unprotonated AspG1. 2) The remaining c subunits contain protonated Asp". 3) All c subunits form an oligomer that rotates with respect t o the a subunit. 4) Protonated forms of the a subunit residues and Glu219 interact directly with AspG1 residues from two adjacent c subunits. 5) His245 of the a subunit is located near the cytoplasmic surface of a water-accessible channel, and it facilitates protonation and deprotonation of G W ' indirectly.
This model is pictured in Fig. 5. There are two water-accessible channels: one extends from the cytoplasmic surface to the center of the membrane and includes His245 and Glu219 of the a subunit. If the region from His245 to GlnZ5' is a-helical, and if GlnZ5' is near Glu'", then it is likely that His245 and Glu219 are -10 A apart and do not interact directly. The second wateraccessible channel extends from the periplasmic surface to the center of the membrane and includes the Arg2Io ( a subunit)-AspG1 (c subunit) ion pair.
A more detailed view is shown in Fig. 6, where proton pathways during ATP synthesis are indicated. Other features shown here are that the AspG1 residue of only one c subunit is accessible to the aqueous medium, through a water-filled channel to the periplasmic side, and that the pathway of individual protons requires a complete rotation of the c subunit oligomer between uptake on one side of the membrane and release on the other side.

Proton Dunsloeation by FIFO ATP Synthases
With respect to energy transduction, the key step is the protonation of the AspG1 of the c subunit that is interacting with Arg2" of the a subunit, by a proton originating from the periplasmic side of the membrane. The accessibility of this AspG1 would be enhanced by local conformational changes in-  This protonation will weaken the interaction between the a subunit and the oligomer of c subunits. This would facilitate movement of the c subunits relative to the a subunit, if the ion pair were a major stabilizing interaction. The movement of c subunits relative to the a subunit, i.e. rotation, would be driven by the attraction between Arg210 and the nearest c subunit that has an ionized AspG1. The rotation of c subunits would be communicated to F,, perhaps via the movement of other subunits, in an unspecified fashion. Proton movement across the membrane would be mediated by two c subunits at the a-c interface: one c subunit takes up a proton as it leaves the interface and a second one gives up a proton as it enters the interface. The ratio of H+/ATP would depend upon the following two factors (among others): 1) How many steps of rotation are sufficient to provide the conformational changes that are necessary to communicate with F,. 2) How tight is the proton cycle. For example, does Asp 61 spontaneously deprotonate outside of the a-c interface?
This model is consistent with a wide variety of other data, in addition to the results presented here. For example, a c subunit mutant in which the essential AspG1 has been moved to position 24, A24D/D61G has been isolated, and it retains partial function (43). Recently it has been shown that for retention of function there are severe structural constraints at position 61, i.e. Asn and Gly permit function, but Ala does not (44). Such double   FIG. 6 . Model of proton pathways in F , . The model is shown for ATP synthesis, which requires steps A, B , C, and D, with counterclockwise rotation of c subunits, as depicted here. The actual direction of rotation during ATP synthesis would be determined by the relative positions o f h 8 " and GIu2" with respect to the c subunits. A , a proton originating from the periplasm protonates the c subunit Asp6' that is ion-paired with the a mutants are likely to preserve both the spacing between essen-important related issue is how such movement in the a-c comtial carboxyl side chains, and the overall conformation of the c plex might be transmitted to F,. The intriguing possibility of an subunits. These are essential features of the current model. important interaction between c and E has been suggested by Other mutations that improve the function of the double recent results (51) in which second-site amino acid substitumutant in the c subunit A24DD61G have been isolated in the tions in E partially suppress c subunit mutants. a subunit, and most of them (10/13) map to a region adjacent t o Glu219 (45). On the basis of these results Fillingame (33) has . , "~~~"~ Hoang9 Dr. and proposed an important helix-helix contact between this region of the a subunit and the c subunit.