Lengthening the Second Stalk of F1F0 ATP Synthase in Escherichia coli *

In Escherichia coliF1F0 ATP synthase, the two bsubunits dimerize forming the peripheral second stalk linking the membrane F0 sector to F1. Previously, we have demonstrated that the enzyme could accommodate relatively large deletions in the b subunits while retaining function (Sorgen, P. L., Caviston, T. L., Perry, R. C., and Cain, B. D. (1998) J. Biol. Chem. 273, 27873–27878). The manipulations of b subunit length have been extended by construction of insertion mutations into the uncF(b) gene adding amino acids to the second stalk. Mutants with insertions of seven amino acids were essentially identical to wild type strains, and mutants with insertions of up to 14 amino acids retained biologically significant levels of activity. Membranes prepared from these strains had readily detectable levels of F1F0-ATPase activity and proton pumping activity. However, the larger insertions resulted in decreasing levels of activity, and immunoblot analysis indicated that these reductions in activity correlated with reduced levels of b subunit in the membranes. Addition of 18 amino acids was sufficient to result in the loss of F1F0 ATP synthase function. Assuming the predicted α-helical structure for this area of the bsubunit, the 14-amino acid insertion would result in the addition of enough material to lengthen the b subunit by as much as 20 Å. The results of both insertion and deletion experiments support a model in which the second stalk is a flexible feature of the enzyme rather than a rigid rod-like structure.

In Escherichia coli F 1 F 0 ATP synthase, the two b subunits dimerize forming the peripheral second stalk linking the membrane F 0 sector to F 1  gene adding amino acids to the second stalk. Mutants with insertions of seven amino acids were essentially identical to wild type strains, and mutants with insertions of up to 14 amino acids retained biologically significant levels of activity. Membranes prepared from these strains had readily detectable levels of F 1 F 0 -ATPase activity and proton pumping activity. However, the larger insertions resulted in decreasing levels of activity, and immunoblot analysis indicated that these reductions in activity correlated with reduced levels of b subunit in the membranes. Addition of 18 amino acids was sufficient to result in the loss of F 1 F 0 ATP synthase function. Assuming the predicted ␣-helical structure for this area of the b subunit, the 14-amino acid insertion would result in the addition of enough material to lengthen the b subunit by as much as 20 Å. The results of both insertion and deletion experiments support a model in which the second stalk is a flexible feature of the enzyme rather than a rigid rod-like structure. F 1 F 0 ATP synthases are multimeric enzymes that function through a complex mechanism similar to a rotary motor (1)(2)(3). Most of the mass of the enzyme is organized into two sectors, referred to as F 1 and F 0 , which are linked by two thin stalk structures (4). A channel composed of the a and c subunits in the F 0 sector conducts protons across the membrane down the electrochemical gradient (5). Energy derived from proton translocation results in the propagation of a conformational change through the stalks driving ATP synthesis at distant catalytic sites in the F 1 sector. A series of elegant experiments have clearly established that this conformational change takes the form of a rotation of the central stalk (6 -10). The central stalk consists of the F 1 ␥ and ⑀ subunits, which appear to be in direct contact with a ring of 12 c subunits in the F 0 sector (11)(12)(13)(14)(15). During catalysis, rotation of the c 12 ␥⑀ subunits sequentially alters the properties of the three catalytic sites located at subunit interfaces in the F 1 ␣ 3 ␤ 3 hexamer. Differing interactions between the ␥ subunit and each of the three ␤ subunits accounts for the asymmetry observed in the high resolution structure of bovine F 1 and the differing conformations and nucleotide binding affinities of the catalytic sites (16).
The apparent function of the peripheral second stalk is to serve as a stator holding the ␣ 3 ␤ 3 hexamer in place against the rotation of the central stalk. Thus, the emerging view is that the second stalk is the primary structural feature maintaining the integrity of the stator elements in F 1 F 0 ATP synthase. A b 2 dimer forms the second stalk making contact with the ␦ subunit and at least one ␣␤ subunit pair (17)(18)(19)(20)(21). Based on a positioning of the ␦ subunit atop the F 1 sector (22), it appears that the b 2 subunits contribute all of the material visualized in the second stalk. From the electron micrographs and composite images of the Escherichia coli enzyme produced by Capaldi and colleagues (23), it is clear that the second stalk spans a distance of 40 -45 Å from the surface of the membrane to the bottom of F 1 . The stalk appeared to be bent at an angle of about 20°in micrographs of single complexes (4). Recently, the structure of peptides modeling the single transmembrane domain of the b subunit dissolved in organic solvents have been determined using nuclear magnetic resonance spectroscopy (24). The evidence suggests that Trp-26 may be at the hydrocarbon-polar interface on the cytoplasmic leaflet of the bilayer, so the region of the b subunit that constitutes the second stalk visible in the electron micrographs starts in the vicinity of Ala-32 and extends to at least Gln-85 and probably a few amino acids beyond. This segment of the b subunit is predicted to be in an ␣ helical conformation, and substantial experimental evidence supports this secondary structure (25)(26)(27).
Previously, we performed a deletion analysis in the second stalk segment of the b subunit to determine the minimum length of the second stalk necessary to form a productive F 1 F 0 ATP synthase (28). Surprisingly, the b subunits could be shortened by as much as 11 amino acids or approximately 16 Å with the retention of function. Losses of activity resulting from the deletions were largely attributable to defective assembly of the enzyme complex. Once assembled, the F 1 F 0 ATP synthases with the shortened b subunits were functional. These observations suggested that the b 2 subunit dimer has an inherent flexibility and that the deletions removed slack present in the second stalk of the wild type enzyme. This interpretation appeared to be incompatible with models in which the second stalk was a rigid structural feature of F 1 F 0 ATP synthase.
In the present work, we have further tested the flexibility hypothesis by inserting additional amino acids into the second stalk regions of the b subunits. The prediction was that if the b 2 dimer was indeed a flexible unit, then insertions of substantial length could be accommodated in functional F 1 F 0 ATP synthase complexes. The additional material would not produce a devastating distortion in the enzyme as might be expected with a rigid structural unit. Insertion mutations were constructed in the uncF(b) gene duplicating segments of the b subunit sequence. Because the goal of these experiments was to determine the maximum extension of the b subunit second stalk region allowing F 1 F 0 ATP synthase function, the most important issue was retention of detectable enzymatic activity. We show that insertion mutants in which the second stalk region of the b subunits has been lengthened substantially and retained biologically and biochemically measurable levels of activity.

EXPERIMENTAL PROCEDURES
Strains and Media-The E. coli uncF(b) deletion strain KM2 (29) and the wild type b subunit expression plasmid pKAM14 have been described previously (30). Plasmid pSD59 (b sol ) was the generous gift of Dr. Stanley Dunn (University of Western Ontario) (25). Growth of strains on a nonfermentable carbon source was scored using Minimal A medium containing succinate (0.2% w/v). Cells for membrane preparation were grown in LB supplemented with 0.2% (w/v) glucose at 37°C. Isopropyl-1-thio-␤-D-galactoside (IPTG) 1 (40 g/ml), ampicillin (100 g/ ml), and chloramphenicol (30 g/ml) were included as needed. Culture medium components were purchased from Difco, and antibiotics were obtained from Sigma.
Recombinant DNA Techniques-Molecular biology enzymes were purchased from Life Technologies, Inc. and New England BioLabs, and the oligonucleotides were synthesized and purified by Gemini Biotech. Restriction endonuclease reactions, ligation reactions, and transformations were performed according to the recommendations of the manufacturers. Plasmid DNA was purified with the Qiagen Mini and Maxi-Prep kits from Qiagen, and DNA fragments were separated by agarose gel electrophoresis and purified using the Qiagen QIAquick kit. Sitedirected mutagenesis was performed either by cassette mutagenesis (31) or by using a Stratagene QuikChange kit. Nucleotide sequences were determined by automated sequencing in the core facility of the University of Florida Interdisciplinary Center for Biotechnology Research.
Mutagenesis and Strain Construction-Plasmid pKAM14 (b) was used to construct the insertion mutations in the stalk region of the b subunit. All of the insertion plasmids were constructed by ligation of synthetic double-stranded oligonucleotides into the PpuMI and PvuII sites in plasmid pKAM14 (b) (see Fig. 1). Except for the insertions, the final products of all constructions were identical to the positive control plasmid pKAM14 (b) placing the recombinant uncF(b) genes under control of the lac promoter. Presence of the inserted sequences were initially detected by digestion of recombinant plasmid DNA with either StuI or HindIII, and then the nucleotide sequences were directly confirmed. For clarity, the numbers of amino acids inserted are indicated behind the plasmid name in parentheses throughout the paper, such as plasmid pAUL19 (ϩ7). Plasmids encoding the b subunit insertions and control plasmids pKAM14 (b) and pBR322 were transformed into E. coli strain KM2 (⌬b) for study. Each strain was also transformed with the plasmid pKAM16 (lacI q ) (29) to provide improved regulation of expression of the uncF(b) genes.
Selected mutations were transferred to plasmid pSD59 (b sol ) for expression of model b sol polypeptides. Plasmid pAUL49 (b solϩ7 ) was constructed by replacing the 541-base pair PpuMI/BstEII fragment of pSD59 (b sol ) with the comparable segment of pAUL19 (ϩ7) containing the insertion. Two base substitutions were generated by using the Stratagene QuikChange kit on plasmid pSD59 (b sol ) at uncF gene nucleotides 96 and 99; this resulted in silent mutations in the codons for b sol Ala-32,Ile33 while generating the unique MunI site in plasmid pAUL45 (b sol ). The new MunI site and the existing BstEII site were then used to facilitate transfer of the uncF(b) gene deletion in pAUL 3 (⌬7) (28) to pAUL45 (b sol ) generating plasmid pAUL50 (b sol⌬7 ). In each case, the nucleotide sequence of each recombinant plasmid was verified, and the expression of b sol polypeptides was confirmed by immunoblot analysis using the b subunit-specific antiserum.
Preparative Procedures-Inverted membrane vesicles were prepared from 500-ml cultures of strain KM2 (⌬b) carrying the recombinant uncF(b) gene expression plasmids according to methods described previously (25). Membrane vesicles stripped of F 1 were prepared essen-tially as described previously (32). E. coli strain 1100 was transformed with b sol expression plasmids pAUL49 and pAUL50 and inoculated into 500 mL of LB supplemented with glucose medium. Preparation of b sol polypeptides was performed as detailed earlier (27).
Analytic Procedures-Protein concentrations were determined by a modified Lowry procedure (33). Membrane energization was detected by the fluorescence quenching of 9-amino-6-chloro-2-methoxyacridine (ACMA) (34), and ATP hydrolysis activity of membrane fractions was assayed by the acid molybdate method (35). Membranes were assayed in buffer (50 mM Tris-HCl, 1 mM MgCl 2 , pH 9.1) for determinations of linearity with respect to both time and enzyme concentration. Immunoblot analyses using the anti-b subunit antibodies were as described previously (28). The anti-rabbit immunoglobulin horseradish peroxidase-linked whole antibody (from donkey), electrochemiluminescence Western blotting detection reagents, and Hyperfilm were obtained from Amersham Pharmacia Biotech. Anti-b subunit antibodies were kindly provided by Dr. Karlheinz Altendorf (Universitä t Osnabrü ck) (36). Chemical cross-linking of b sol peptides with bis(sulfosuccinimidyl) suberate and analytic ultracentrifugation experiments were conducted according to procedures published in earlier work (27).

Construction and Growth
Characteristics of Mutants-To determine the maximum length of a b subunit that can be incorporated into a functional F 1 F 0 ATP synthase, a collection of insertion mutations were generated within the stalk region of the b subunit by site-directed mutagenesis. The three options available for design of the insertion sequences were strings of alanines likely to generate ␣-helical segments, sequences of amino acids known to form ␣-helices in other proteins, and duplication of a segment of the b subunit. We elected to take the last approach because it seemed to hold the most promise in terms of establishing appropriate interactions between the two b subunits necessary for dimerization (Fig. 1). The duplicated segments started at b Leu-56 and extended in the direction of the carboxyl terminus because this area of the b subunit had proven to be the least sensitive to the effects of deletions (28). The lengths of the insertion mutations were designed to model, as closely as practicable, full turns of an ␣-helix. The earlier deletion experiments suggested that helix orientation effects resulting from less than complete turns reduced assembly of the enzyme complex (28). In all, four insertion plasmids were constructed ranging from addition of 7-18 amino acids modeling insertions of from 2-5 ␣-helical turns (Fig. 1).
The effects of the insertion mutations were studied by complementation of uncF(b) gene deletion strain KM2 (⌬b). E. coli strains defective for F 1 F 0 ATP synthase cannot derive energy from nonfermentable carbon sources, so growth on succinate minimal medium served as a convenient qualitative measure of enzyme function in vivo. Because high levels of expression of altered b subunits can in some instances overcome the effects from assembly defects (30), experiments were conducted under conditions of both high and low level expression of the mutated uncF(b) genes under control of the lac promoter. In experiments using low expression conditions, strain KM2/pAUL47 (ϩ11) harbored the largest b subunit insertion capable of supporting visible colony formation on solid succinate-based medium (Table I). In the presence of saturating concentrations of IPTG, slow growth of strain KM2/ pAUL48 (ϩ14) was observed indicating biologically significant levels of F 1 F 0 ATP synthase activity (Table I). Strain KM2/ pAUL52 (ϩ18) failed all growth tests, indicating that the insertion exceeded the maximum length allowable for the enzyme.
Assembly of F 1 F 0 ATP Synthase-Membrane vesicles were prepared, and immunoblots were performed using an anti-b antibody to detect production and incorporation of the b insertion subunits into the membrane. Although a small reduction in the level of b subunit was observed from the low induction conditions cells, membranes from strain KM2/pAUL19 (ϩ7) grown in the presence of IPTG possessed levels of b subunit proteins approaching that of the wild type control membranes (Fig. 2). However, reductions in levels of b subunits were readily apparent in membranes from cells with longer insertions. As might be expected for a mutation resulting in a total assembly defect, no b subunit protein was found in the KM2/ pAUL52 (ϩ18) membrane preparations. F 1 has little affinity for the membrane in the absence of F 0 , and in particular the b subunit hydrophilic domain (37-39), so total membrane-associated F 1 -ATPase activity was used as a test of F 1 F 0 ATP synthase complex assembly. The amounts of total F 1 F 0 -ATPase activity observed in the membranes fell with increasingly longer insertions ( Table I). The KM2/pAUL19 (ϩ7) membranes had the highest level of F 1 F 0 ATPase activity of the insertion strains, and no significant activity was present in membranes prepared from strain KM2/pAUL52 (ϩ18). Importantly, the membranes from strain KM2/pAUL48 (ϩ14) grown under high expression conditions retained substantial amounts of enzymatic activity, indicating the presence of intact F 1 F 0 -ATPase complexes.
Proton Translocation-Membranes were assayed for F 1 F 0 ATP synthase-mediated ATP-driven proton pumping activity as an indication of coupled activity for the b subunit insertion enzymes. Acidification of inverted membrane vesicles was followed by monitoring the fluorescence of ACMA (Fig. 3). To demonstrate that the vesicles were intact and closed, the level of NADH-driven fluorescence quenching was assayed for all membrane preparations. In every case, the levels of NADHdriven fluorescence quenching were strong and comparable with both wild type and negative control membranes (data not shown). The levels of coupled activity in the membranes correlated very well with the ability of the strains to grow on succinate medium and the amount of intact F 1 F 0 ATP synthase present. A modest reduction in ATP-driven proton pumping activity below wild type controls was observed in the KM2/ pAUL19 (ϩ7) membranes from cells grown under conditions of low expression (Fig. 3A). A sharper reduction in activity was seen for the KM2/pAUL47 (ϩ11) membranes, and no activity was detected in membranes from the longer insertion mutants. As anticipated from the slow growth of strain KM2/pAUL48 (ϩ14) on succinate minimal medium, membranes prepared from this strain under high expression growth conditions possessed readily detectable levels of proton pumping activity (Fig.  3B). ATP-driven proton pumping was enhanced in vesicles derived from IPTG-induced KM2/pAUL47 (ϩ11) relative to membranes from uninduced cells, and KM2/pAUL19 (ϩ7) membranes appeared to be essentially indistinguishable from wild type control KM2/pKAM14 membranes with respect to coupled activity. In general, the reductions in activity observed in membranes from the insertion strains reflected the reduced amounts of assembled F 1 F 0 ATP synthase. Most importantly, the data established that intact F 1 F 0 ATP synthase complexes containing the b subunits lengthened by as many as 14 amino acids possessed coupled activity.
To consider the functional state of the proton translocation mechanism in the insertion mutant b subunit F 0 sectors, passive proton permeability was studied in membranes stripped of F 1 (Fig. 4). F 0 -mediated dissipation of an imposed proton gradient was monitored using ACMA fluorescence. Stripped membranes prepared from strain KM2/pAUL19 (ϩ7) were virtually identical to the positive control membranes from KM2/ pKAM14. Other insertion strain membranes had reduced rates of proton leakage commensurate with a reduced number of F 0 sectors. Like the uncF(b) deletion mutants studied previously (28), insertions in the second stalk region of the b subunit did not appear to have any direct affect F 0 -mediated proton conductance.
Effects of Deletions and Insertions on Dimerization-The properties of model b sol polypeptides with either a deletion or an insertion of 7 amino acids were characterized to consider whether altering the length of the second stalk region of the b subunit affected interactions between the two b subunits in the hydrophilic domains. Plasmids pAUL49 (b solϩ7 ) and pAUL50 (b sol⌬7 ) were used to direct expression of a recombinant b sol polypeptide, and the recombinant b sol polypeptides were purified to homogeneity. The yield of recombinant b sol polypeptides obtained from approximately 2.0 g of cells (wet weight) was approximately 5 mg of pure protein. Chemical cross-linking using the irreversible agent bis(sulfosuccinimidyl) suberate was employed to look for dimer formation by the b sol proteins (Fig. 5). Cross-linking was observed for both the b solϩ7 and the b sol⌬7 polypeptides, indicating the presence of dimers. The cross-linked bands had the expected mobility for dimers in SDS-PAGE, and an immunoblot analysis using the anti-b subunit antibody confirmed that the 34-kDa cross-linked product was indeed b sol polypeptide (data not shown). Sedimentation equilibrium experiments were also performed to study dimer formation of the b solϩ7 and the b sol⌬7 polypeptides (Fig. 6). Like the b sol control, both the insertion and deletion polypeptides were largely in the dimeric form, and the samples contained less than 2% monomer. Concentrations of tetrameric polypeptides observed in the b solϩ7 and the b sol⌬7 preparations were slightly higher than for b sol . The data suggested that neither the deletion nor the insertion had a large impact on dimer formation. This can be viewed as support for the dimerization domain hypothesis of McLachlin and Dunn (26), because the design of the insertion held this region of the protein intact impinging only slightly on the area of the b subunit defined as important for dimer formation.

DISCUSSION
The segment of the b subunit thought to form the portion of the F 1 F 0 ATP synthase visualized as the second stalk in electron micrographs extends from about Ala-32 to at least Ser-85. In the present work, we have characterized the effects of insertion mutations designed to lengthen the second stalk region of the b subunit. Insertion of 7 amino acids resulted in an essentially normal phenotype. Abundant fully assembled and functional F 1 F 0 ATP synthase was observed in the membranes, and experiments using the model b sol ϩ7 polypeptide indicated a propensity for dimer formation similar to wild type b sol . As the sizes of the insertions were increased to 11 and 14 amino acids, the amounts of intact F 1 F 0 -ATP synthase present in the membranes fell, but significant levels of enzymatic activity remained. The losses of function observed in the insertion mutants represented assembly defects rather than functional failures in intact F 1 F 0 ATP synthase complexes. The assembly defect resulting from insertion of 18 amino acids was so severe that F 1 F 0 ATP synthase activity was completely lost and no b subunit could be detected in the membrane fractions.
The most important parameter in these studies was the retention of coupled F 1 F 0 ATP synthase function with addition b ATPase activities were measured as described under "Experimental Procedures." Units of specific activity ϭ mol of PO 4 released per mg of protein/min Ϯ S.D. Units were calculated from the slope of the line based on five independent measurements with incubations for 15 min. of up to 14 residues. This insertion amounts to approximately four full turns of an ␣-helix corresponding to the capacity to extend the second stalk by up to 20 Å. Although we did not rigorously exclude the possibility of a change in secondary structure resulting from the insertions, the simple duplication of b subunit segments in these experiments and computer analysis using secondary structure prediction algorithms argue against this idea. Therefore, the maximum functional extension of the second stalk segment of the b subunit was roughly 50% longer than that normally used to span the distance be-tween F 1 and F 0 in the wild type enzyme. If the second stalk were a rigid rod-like structure, such a large insertion would be expected to result in a dramatic distortion of the complex leading to failure of the stator and loss of F 1 F 0 ATP synthase function. The insertion mutations studies reported here and the deletions characterized earlier (28) point to an interpretation that the second stalk of F 1 F 0 ATP synthases have considerable flexibility to accommodate large changes in length. The second stalk seen in the electron micrographs of Wilkens and Capaldi (4) was bent. Viewing the second stalk as a flexible feature of the complex allows shortening and lengthening the b subunit to be accomplished, at least in part, by straightening or accentuating the bend. In simplistic terms, the second stalk seems much more likely to be a flexible rope that can be pulled taut, providing the tensile strength required to function as the major structural feature of the stator, rather than an inherently stiff rod. In this way, the results can be rationalized as compatible with the transient elastic storage of energy hypothesis for F 1 F 0 ATP synthase described by Junge and colleagues (2,40).
The two structurally and functionally constrained residues, Arg-36 and Ala-79, conserved in nearly all bacterial F 1 F 0 ATP synthase b subunits are separated by exactly 43 amino acids (29,41). All of the insertions reported here, as well as the earlier deletions (28), were constructed between these two sites disturbing this apparently conserved spacing. Changes of up to 7 amino acids in either direction made very little difference in the activity of intact F 1 F 0 ATP synthase complexes containing the altered b subunits. The remaining question is, why does there appear to be conservation of the length bacterial b subunits in the second stalk segment? The answer probably lies in optimization of assembly of the oligomeric enzyme complex during evolution. The length of the b subunit may have remained under selective pressure for positioning the b subunit for interactions with F 1 needed for efficient formation of the F 1 F 0 ATP synthase complex. Even a small decrease in the efficiency of enzyme assembly might have constituted a selective disadvantage for procaryotic organisms competing by a strategy based on rapid proliferation during evolution. The solid lines are the theoretical best fit curves calculated from the known sequences of the polypeptides and assuming a nonequilibrium distribution of monomer, dimer, and tetramer (27).