Genetic probing of the first and second transmembrane helices of the plasma membrane H(+)-ATPase from Saccharomyces cerevisiae.

Structural features of the putative helical hairpin region comprising transmembrane segments 1 (TM1) and 2 (TM2) of the yeast plasma membrane H(+)-ATPase were probed by site-directed mutagenesis. The importance of phenylalanine residues Phe-116, Phe-119, Phe-120, Phe-126, Phe-144, Phe-159, and Phe-163 was explored by alanine replacement mutagenesis. It was found that substitutions at all positions, except Phe-120 and Phe-144, produced viable enzymes, although a range of cellular growth phenotypes were observed like hygromycin B resistance and low pH sensitivity, which are linked to in vivo action of the H(+)-ATPase. Lethal positions Phe-120 and Phe-144, could be replaced with tryptophan to produce viable enzyme, although the F144W mutant was highly perturbed. ATP hydrolysis measurements showed that Km was not significantly altered for most mutant enzymes, whereas Vmax was moderately reduced with two mutants, F144W and F163A, showing less than 50% of the normal activity. Double Phe-->Ala mutations in TM1 and TM2 were constructed to examine whether such substitutions would result in a higher degree of enzyme destabilization. Mutant F116A/F119A was viable and gave a normal phenotype, while F159A/F163A was not viable. Other double mutants, F116A/F159A and F119AF/159A, which are predicted to lie juxtaposed on TM1 and TM2, produced non-functional enzymes. However, a viable F119V/F159A mutant was isolated and showed hygromycin B resistance. These results suggest that double mutations eliminating 2 phenylalanine residues strongly destabilize the enzyme. A putative proline kink at Gly-122/Pro-123 in TM1 is not essential for enzyme action since these residues could be variously substituted (G122A or G122N; P123A, P123G, or P123F) producing viable enzymes with moderate effects on in vitro ATP hydrolysis or proton transport. However, several substitutions produced prominent growth phenotypes, suggesting that local perturbations were occurring. The location of Pro-123 is important because Gly-122 and Pro-123 could not be exchanged. In addition, a double Pro-Pro created by a G122P mutation was lethal, suggesting that maintenance of an alpha-helical structure is important. Other mutations in the hairpin, including modification of a buried charged residue, E129A, were not critical for enzyme action. These data are consistent with the view that the helical hairpin comprising TM1 and TM2 has important structural determinants that contribute to its overall stability and flexibility.

The yeast plasma membrane H ϩ -ATPase is a typical P-type ion translocation ATPase that is related to the family of enzymes, which includes the mammalian Na ϩ ,K ϩ -ATPase, Ca 2ϩ -ATPase, and H ϩ ,K ϩ -ATPase; the plant H ϩ -ATPase; and the bacterial K ϩ -ATPase, Mg 2ϩ -ATPase, and Cu 2ϩ -ATPases (1-3). These enzymes couple ATP hydrolysis to ion transport and cycle between two principal conformational states during catalysis. They characteristically form an acylphosphate intermediate during catalysis and are sensitive to inhibition by vanadate (4). The plasma membrane H ϩ -ATPase from yeast is essential for growth (5), where it plays a critical role in the maintenance of electrochemical proton gradients and the regulation of intracellular pH (6). Significant sequence similarity exists between the various family members, with the greatest degree of sequence homology found within the cytoplasmic domain catalyzing ATP hydrolysis (7,8). The topology of these enzymes is similar, with the N and C termini residing in the cytosol (9,10), and most recent data are consistent with the presence of 10 transmembrane segments (11,12). There is general agreement on the identity and orientation of the first four transmembrane segments, with discrepancies occurring in the remaining C-terminal transmembrane elements.
The mechanistic nature of how the H ϩ -ATPase couples energy from ATP binding and hydrolysis within the cytosolicdomain to transport of ions within the membrane sector is not understood. Diverse studies involving drug interactions and immunological probing of higher eukaryotic enzymes (13)(14)(15)(16)(17)(18), as well as genetic modifications of the yeast H ϩ -ATPase (19 -21), support the involvement of long range conformational interactions. There is growing evidence that transmembrane segments 1 (TM1) 1 and 2 (TM2) are conformationally linked to the catalytic ATP hydrolysis domain (19,20,22).
Recently, we proposed a detailed structural model for TM1 and TM2 and used molecular dynamic simulations to assess potential conformational determinants in this region that help account for its functional role (23). TM1 and TM2 are predicted to form a helical hairpin structure that has a number of prominent structural features including a short 4 -6-amino acid turn linking the ␣-helices, a tightly packed head region, an N-cap structure stabilizing the turn region, a cluster of phenylalanine residues near the cytoplasmic face of the hairpin structure, and a flexible region consisting of Gly-122/Pro-123 that may kink TM1. The hairpin structure is hydrophobic, and only one charged amino acid, Glu-129, is present in TM1. The short turn region linking TM1 and TM2 was extensively probed by mutagenesis and found to be highly conformationally active with perturbations being manifested as alterations in catalytic function (21). In this report, we have examined the effects of amino acid substitutions on the 7 phenylalanines, the flexible proline kink region, and other putatively important residues within the hairpin region of TM1 and TM2. We provide evidence that the hairpin region is conformationally sensitive since viable mutations in this region yield hygromycin B-resistant and low pH-sensitive cellular phenotypes, and many of the mutant enzymes show altered catalytic properties. We further provide evidence that the cluster of phenylalanine residues near the cytoplasmic face of the bilayer may be important for structural stability.

MATERIALS AND METHODS
Yeast Strains and Cultures-All yeast strains utilized in this study are isogenic derivatives of Y55 (HO gal3 MAL1 SUC1) (24). Wild type control strain GW201 (HO ade6 -1 trp5-1 leu2-1 lys1-1 ura3-1 PMA1::URA3) was constructed by transplacing a 6.1-kb HindIII fragment containing intact PMA1 linked 3Ј to URA3 into SH122 (HO ade6 -1 trp5-1 leu2-1 lys1-1 ura3-1 pmal⌬::LEU2/PMA1), as described by Harris et al. (19). Wild type strain SN236 (HO ade6 -1 trp5-1 leu2-1 lys1-1 ura3-1PMA1) is a derivative of SN236 (20) in which the URA 3 marker has been lost. All yeast cultures were grown to early log-phase in YEPD medium (1% yeast extract, 2% peptone, and 2% dextrose, pH 5.7) at 22°C to an A 590 nm ϳ 3. Growth sensitivity to hygromycin B was monitored in YEPD agar plates containing 0, 100, 150, 200, and 300 g/ml hygromycin B. Growth sensitivity to low pH medium was determined at pH 2.5 in YEPD plates in the presence of 5 and 10 mM potassium acetate. Temperature sensitivity of growth was assessed by comparing growth at 30 and 40°C. (The wild type yeast strain, Y55-background, used in this study grows normally at 40°C unlike other wild type strains of Saccharomyces. It is important that the 40°C incubator should be saturated with water during cell growth.) Site-directed Mutagenesis-Site-directed pma1 mutants were constructed essentially, as described previously (20,21). PMA1 mutants F119A, F126A, M128A, E129A, F159A, G122N, M128C, and M128S were initially prepared in phagemid vector pSN54, which consists of a 2.1-kb Asp718 fragment from PMA1 subcloned into pGEM-3zf (20). Mutants prepared in pSN54 were first excised as part of a 0.7-kb BstEII-EcoRV fragment, and then purified by agarose gel electrophoresis. The purified fragment was reconstituted in PMA1 by subcloning into identical sites in alkaline phosphatase-treated pIV100 (derived from pSN57 containing an additional BamHI site in the BstEII-EcoRV exchange region of PMA1) (21). Mutants G122A and P123A were prepared in phagemid vector pDP100, which carried a new AvaII restriction site, also in BstEII-EcoRV exchange region of PMA1. These mutants were first screened by restriction digest analysis and then sequenced. A purified 0.7-kb BstEII-EcoRV fragment was transplaced into pSN57 to reconstitute intact PMA1, as described above. Other mutants were prepared in phagemid vector pGW201, which consists of a 6.1-kb HindIII fragment containing PMA1 marked with URA3 at the 3Ј non-coding end (19). The entire 6.1-kb region was excised and transplaced into yeast directly. All vectors containing reconstituted pma1 genes were sequenced prior to transplacement into yeast to confirm the primary site mutation and to eliminate potential secondary mutations in the target region. Isogenic pma1 mutants were prepared as described by Harris et al. (19). All pma1 mutations were reconfirmed after isolating meiotic segregants by polymerase chain reaction amplification of chromosomal DNA and sequence analysis, as described previously (21).
Plasma Membrane Isolation and ATP Hydrolysis Measurement-Plasma membranes were purified from wild type and pma1 mutant strains by centrifugation on a sucrose step gradient, as described previously (25). ATP hydrolysis measurements were performed in a microplate assay in a 100-l volume containing 10 mM Mes/Tris (pH 6.5), 25 mM NH 4 Cl, 5 mM ATP, 5 mM MgCl 2 , 0.5 mM NaN 3 , and 1 g of membrane protein, as described by Monk et al. (26).
Reconstitution and Proton Transport Measurements-A microsomal membrane fraction was prepared essentially by the method of Perlin and Brown (27), except that a lower centrifugal force (75,000 ϫ g for 30 min) was used for membrane recovery. The membrane vesicles were extracted with 0.5% (w/v) deoxycholate in Solubilization Buffer, consisting of 10 mM Hepes-KOH (pH 7.0), 0.1 M KCl, 45% glycerol, 0.2 mM EDTA, 1 mM dithiothreitol, and 1 mg/ml asolectin, as described previously (25). The detergent-extracted membranes were washed with an equal volume of 0.3 M KCl-Solubilization Buffer by resuspension and centrifugation, as above. The KCl-washed membranes (350 g) were resuspended in a 800-l volume containing 10 mg/ml asolectin, Solubilization Buffer, and 0.5% (w/v) deoxycholate (added dropwise with gentle stirring). The mixture was placed on ice for 5 min and then rapidly diluted into 25 ml of ice-cold Dilution Buffer containing 10 mM Hepes-KOH (pH 7.0), 100 mM KCl, and 1 mM dithiothreitol. The reconstituted vesicles were recovered by centrifugation at 250,000 ϫ g for 1 h. The pellet was resuspended in 400 l of the dilution buffer. Proton transport measurements were made according to the method of Perlin et al. (28). A fluorescence quenching reaction volume consisted of 1 ml of 10 mM Hepes-KOH (pH 7.0), 50 mM KCl, 5 mM ATP, 1 g/ml valinomycin, 1.5 M acridine orange, and 50 g of reconstituted vesicles. The reaction was initiated by the addition of 5 mM MgCl 2 . Fluorescence intensity was monitored on a Perkin-Elmer LS-5 spectrofluorometer.
H ϩ -ATPase Abundance Measurements-SDS-polyacrylamide gel electrophoresis and semidry electroblotting of plasma membrane proteins were performed, as described previously (26). Western blot analysis was performed with a polyclonal anti-H ϩ -ATPase antibody, described by Seto-Young et al. (21). Western blots were scanned with a UMAX color scanner (UMAX Data Systems, Inc.), and Adobe Photoshop and NIH Image software were used to quantitate the level of the intact H ϩ -ATPase (molecular mass ϳ100 kDa). Standard default settings were used for all measurements, and all mutant enzymes were compared to an internal wild type control on the same gel.
Other Procedures-Protein was determined by a modified Lowry method (28). Yeast transformants were prepared by the lithium acetate treatment method, as described in the alkali-cation kit (Bio 101, Inc.). DNA sequencing of plasmid DNA was performed either with Sequenase (version 2.0, United States Biochemical Corp.) or by polymerase chain reaction amplification of the target region and sequencing with the fmol sequencing system (Promega). Transmembrane helices 1 and 2 of the yeast H ϩ -ATPase were constructed with the molecular modeling program Insight II (version 2.2.1; Biosym Technologies) on a Silicon Graphics IRIS computer (model 4D/70GT).

Phenylalanine Mutagenesis of Hairpin Region-Aromatic
residues are suggested to play an important role in the folding and structural stability of many proteins (29). Seven Phe residues are contained within the TM1 and TM2 hairpin region of the yeast plasma membrane H ϩ -ATPase (Fig. 1). Five of the 7 residues are located toward the cytoplasmic end. They are predicted to be nested and overlapping (23), and are anticipated to be important for structural stability. The aromatic amino acids in this region are conserved among the fungal ATPases, apart from a single F163Y change in the Candida albicans enzyme (26). Scanning Ala mutagenesis was used to replace all of the Phe residues in the hairpin region with Ala by site-directed mutagenesis. Alanine was chosen because it is frequently found in helical structures, both inside and outside of the bilayer (30,31). Table I shows that Ala substitution of Phe at positions 116, 119, 126, 159, and 163 produced viable cells, indicating expression of a functional H ϩ -ATPase. (The H ϩ -ATPase is essential for growth and non-functional enzymes give a lethal phenotype.) F116A and F163A produced growth phenotypes that were essentially wild type; both mutations are predicted to lie at the base of TM1 and TM2, respectively, near the cytoplasmic interface of the membrane. Mutant strains carrying mutations F119A, F126A, and F159A showed growth resistance to hygromycin B, which has been correlated with a defect in membrane potential formation and high capacity proton pumping (32). Mutations F120A or F144A produced recessive and/or dominant lethal phenotypes, respectively. To determine whether an aromatic group was important, Trp was used in place of Ala. Both F120W and F144W produced viable mutants, with F120W showing a normal phenotype and F144W showing hygromycin B resistance and temperature sensitivity at 40°C. (It should be noted that the wild type yeast strain used in this study, which carries a Y55 background, grows normally at 40°C unlike other wild type strains of Saccharomyces.) It was previously shown that Trp was the only residue that could substitute for Phe at this position (21).
Plasma membranes were purified from wild type and all viable pma1 mutants. The abundance of intact H ϩ -ATPase (molecular mass of ϳ100 kDa) in the mutant membranes was assessed by SDS-polyacrylamide gel electrophoresis and Western blot analysis; it was found to exceed 70% of the wild type level (Table II). ATP hydrolysis measurements (Table II) showed that the K m for ATP was not significantly altered for most mutant enzymes; only enzymes from mutants F159A and F144W showed slightly lower K m values. In contrast, V max (adjusted for enzyme abundance) was significantly reduced in mutants F144W and F163A to 39 and 49%, respectively, of the wild type level of activity, while other mutants were reduced from 57 to 74%. The F144W mutation produced an enzyme with lower activity (39%) than the F120W mutation (68%).
The vanadate inhibition profiles (I 50 ) for most of the Phe-pma1 mutant enzymes were comparable to the wild type (Table  II). The pH dependence of ATP hydrolysis was assessed at pH 5.5, 6.5, and 7.5, and the hydrolysis activities at pH 5.5 and 7.5 were expressed as a function of activity at the normal pH optimum pH 6.5. All the Phe-pma1 mutant enzymes showed a near wild type-like activity ratio, which at most, was 20% less than wild type (data not show). The H ϩ transport properties of mutants showing hygromycin B-resistant phenotypes were examined in a reconstituted vesicle system. The mutations had no significant effect on ATP-mediated proton transport when equivalent amounts of ATP hydrolysis units were assayed. The F144W enzyme showed a somewhat lower initial rate of pumping, although the steady state pH gradient reached the same level as wild type (Table II).
Structural stability was assessed by measuring ATPase activity in sucrose gradient-purified plasma membranes at increasing temperature (30, 35, 40, 50, and 55°C). Mutant enzymes showed the same relative heat inactivation profile as wild type (data not show). However, when mutant and wild type enzymes were heated for 15 min at 45°C in the presence of increasing concentrations of urea (0 -2.67 M), mutant enzymes F163A and F144W showed enhanced heat inactivation (Fig. 2). Mutant enzymes F119A, F126A, and F159A were comparable to wild type (data not shown).
Double Phe Mutants in the Hairpin Region-Double Phe mutants were constructed to examine whether two Phe residues could be substituted, which would result in a higher degree of enzyme destabilization. Double mutant F116A/ F119A, lying nearby on TM1, was viable and gave normal growth phenotypes and enzymatic properties (Table II). This was a somewhat curious result because the individual mutant, F119A, showed abnormal growth and enzymatic properties (Tables I and II). It appears that the F116A mutation relieved the stress created by the F119A mutation. In contrast, double mutant F159A/F163A, lying nearby on TM2, was not viable. Other double mutants, F116A/F159A and F119AF/159A, which are expected to be juxtaposed on TM1 and TM2, also produced non-functional enzymes. The results suggest that a double substitution with Ala on TM1 and TM2 is not tolerated. However, a viable double mutant, F119V/F159A, was isolated. This double mutant with Val on TM1 at position 119 showed hygromycin B-resistant and low pH-sensitive growth phenotypes, indicative of a perturbed enzyme. However, the mutant enzyme ϩϩϩ ϩϩϩ Ϫ ϩϩ a F116/F159A produced two types of clones that, after dissection, yielded isolates displaying either dominant or recessive lethal phenotypes.
showed rates of ATP hydrolysis (Table II) and proton transport that were wild type in behavior, indicating that the in vivo perturbation was relieved upon plasma membrane purification.
Overall, these results suggest that eliminating two Phe from the base of each transmembrane segment is highly destabilizing. The ability of F119V to stabilize a second mutation at F159A may result from the side chain partially filling a cavity created by the loss of the aromatic residues from each helix. This would imply that, at the very least, a space-filling role is indicated for the phenylalanines.
Mutagenesis of Proline Kink Region-TM1 contains a Gly-122/Pro-123 combination that may form a highly flexible and helix distorting region. Pro-123 is conserved among all fungal H ϩ -ATPases (8, 23). Mutagenesis was used to investigate whether a potential Pro kink involving Pro-123 might play an important role in the function of the H ϩ -ATPase. A P123A mutation produced viable cells that showed growth phenotypes with hygromycin B resistance and low pH sensitivity (Table I). An additional substitution with a small Gly residue or a bulkier Phe residue resulted in hygromycin B-resistant phenotypes ( Table I). The P123G and P123A mutations showed 62% and 71%, respectively, of the normal abundance of enzyme in the membrane, although they showed only a moderate reduction in ATP hydrolysis rates (Table II). Gly-122, which precedes Pro-123, was substituted with Asn, Ala, and Pro. The G122N and G122A mutations showed growth resistance to hygromycin B, but only G122N showed low pH sensitivity (Table I). However, the G122N mutation showed normal biochemical properties, while the G122A mutation was significantly reduced in activity. A G122P mutation, which produced double Pro-Pro residues at positions 122 and 123, was not permissible. Doublets of Pro-Pro are rarely, if ever, found in ␣-helices (33). A double mutant, G122P/P123G, was constructed to shift the position of the Pro down the helix, but it resulted in a nonfunctional enzyme (Table I). A Pro was also introduced into TM2 by substituting Pro for Ala at position 155. The mutant was highly perturbed, showing mutant growth phenotypes (Table I) and ATP hydrolysis rates that were 20% of the wild type level (Table II). This result suggests that a Pro in TM2, potentially kinking or altering the helical structure, was not favorable. Enzymes from mutants G122A, G122N, and P123A showed small perturbations of H ϩ transport, with initial rates at 64 -76% of the wild type. In each case, steady state proton pumping levels approached that of wild type (Table II). These results suggested that a Pro kink, if it exists, in helix 1 is not required for H ϩ transport. On the other hand, the lethality of G122P, which creates a helix-destabilizing Pro-122/Pro-123 cluster, along with the prominent mutant growth phenotypes observed for viable mutations in this region, suggests that maintenance of an ␣-helix is important.
Other Mutants in the Hairpin Region-Glu-129 is predicted to lie in the middle of transmembrane segment 1. In the Neu-  rospora H ϩ -ATPase, this residue can be modified by DCCD (34), which inhibits enzyme activity. A previous study indicated that E129Q or E129L mutations had little effect on enzyme activity. Unfortunately, more subtle growth effects could not be examined in the expression system utilized (35). We explored whether Glu-129 plays a role in function by substituting a small Ala at this position. The E129A mutant had a weak effect on phenotype, producing only mild hygromycin B resistance ( Table I). As previously proposed (35), this result suggests that the bilayer-buried charged moiety Glu-129 is not important for catalytic function. The mutation has some effect on the initial rate of the H ϩ transport, but steady state pH gradient formation was comparable to the wild type (Table II). It is puzzling that a membrane-embedded charged moiety would be so highly conserved among the fungal enzymes (8), if it plays no apparent role in function.
According to the molecular structure model, Met-128 on helix 1 is predicted to lie within close proximity to Cys-148 on helix 2. An M128C mutation was created to examine whether a disulfide linkage could be established between the 2 residues. The M128C mutant yielded hygromycin B-resistant and low pH-sensitive phenotypes, and the mutant enzyme was significantly reduced in ATP hydrolysis and initial rate of H ϩ transport (Table II). However, there was no effect of sulfhydryl reagents on enzyme activity, which precluded an affirmation of a disulfide linkage. In addition, Ala and Ser substitutions at Met-128 were not viable, which further complicates the analysis, since it is not possible to distinguish between perturbations created by the mutation in helix 1 and a potential cross-linking of the hairpin structure. DISCUSSION A Role for TM1 and TM2-It has been suggested for P-type ATPases that the primary transported ion binds to a cytoplasmically-exposed site on the membrane surface of the enzyme, which is a considerable distance (50 -60 Å) (36, 37) away from the catalytic center engaged in ATP hydrolysis. Thus, long range structural interactions are necessary for coupling to occur. Emerging evidence suggests that local interactions in TM1 and TM2 provide important clues to the nature of such long distant energy coupling. In the yeast H ϩ -ATPase, the notion that the hairpin region encompassing TM1 and TM2 is conformationally linked to the catalytic ATP hydrolysis domain has arisen from genetic studies of the extracellular turn region (21), and from studies identifying second site suppressor mutations, which either partially or fully complement phenotypes produced by a primary site mutation in PMA1. For example, it was shown that the phenotype induced by a primary site mutation, S368F, near the site of phosphorylation (Asp-378), could be suppressed by second-site mutations in TM1 and TM2 (19). Conversely, the phenotypes induced by a mutation, A135V, near the extracytoplasmic face of TM1, could be suppressed by secondary site mutations within the catalytic core of the ATP binding domain (20). Additional evidence linking the TM1 and TM2 to the catalytic ATP hydrolysis domain in the yeast enzyme was provided by showing that modification of Cys-148 in TM2 with omeprazole was closely correlated with enzyme inhibition (38). In addition, a G158D mutation in TM2 was found to produce a partially uncoupled enzyme when assayed in vitro (28). Diverse studies on higher eukaryotic enzymes also support a potential linkage between TM1, TM2 and the catalytic ATP hydrolysis domain. Genetic modification of residues in TM1 and TM2, which alter ouabain sensitivity, was also observed to alter catalysis by the Na ϩ ,K ϩ -ATPase (39). A monoclonal antibody that recognizes an epitope in the extracellular turn region between TM1 and TM2 of the Na ϩ ,K ϩ -ATPase inhibited catalysis (16). Finally, the H ϩ ,K ϩ -ATPase antagonist SCH28080, which blocks ATP hydrolysis, appears to bind within the loop region linking TM1 and TM2 (13).
TM1 and TM2 are predicted to form a helical hairpin structure (13,23). We have used molecular dynamic simulations to predict how perturbations in the hairpin head region could be propagated throughout the structure (23). In addition, we used a detailed genetic analysis to explore limited conformational flexibility and tight packing in the head region (21), as predicted from the model studies. In this study, we systematically investigated residues comprising putatively important features of this structural region.
Importance of Aromatics-Interacting aromatic residues in proteins are believed to be important for structural stability and assembly of proteins (29,40,41). In addition, clustered aromatics may be important for translocation across the bilayer as has been recently proposed for sugar transport through the porin channel (42). The TM1/TM2 hairpin region contains seven Phe residues, with five residues predicted to form a clustered grouping at the cytoplasmic interface. These residues are predicted to be important for structural stability through the involvement of potentialinteractions (23). In fact, the sequence arrangement of Phe-116 and Phe-120 on TM1 and Phe-159 and Phe-163 on TM2 should place these residues on the same face of their respective ␣-helical segments. A similar cluster of aromatics is found in the ␣-subunit of the Na ϩ ,K ϩ -ATPase and in the Ca 2ϩ -ATPase (43,44). We substituted each of the seven Phe residues with Ala and found only two positions, Phe-120 and Phe-144 (described previously; Ref. 21), which could not support the loss of side chain mass (Table I). A Trp substitution at these positions produced viable enzymes, suggesting that bulky aromatic character was required. Of the remaining Phe residues that could be substituted with Ala (Phe-116, Phe-119, Phe-126, Phe-159, and Phe-163), the viable mutants showed varying growth phenotypes ranging from wild type to strongly hygromycin B-resistant and low pH-sensitive (Table I). Two mutants, F163A and F144W, showed enhanced thermal/chaotropic denaturation (Fig. 2) indicative of a stability defect. More severe effects were observed with double mutants constructed to remove a single Phe residue each from TM1 and TM2. Double mutants F116A/F159A and F119A/F159A were recessive lethal, as was a double mutant F159A/F163A on TM2 (Table I). However, an F116A/ F119A on TM1 produced a viable enzyme that appeared wild type in behavior. Unless aromatic residues on TM2 are interacting with Phe-120 on TM1, it appears that potential interactions between the helical segments are not as important as the presence of aromatic character on TM2. Of course, interactions with other elements or lipid cannot be ruled out. The bulkiness of the Phe side group is likely to be important. Deletion of these two residues from TM1 and TM2 could create a cavity that could be highly destabilizing, as has been observed with T4 lysozyme (45). The fact that a F119V/F159A mutation was viable (Table I) suggests that Val may partially substitute for the bulky Phe in this position. Overall, these data suggest that aromatic residues are important for the stability and/or folding the TM1,TM2 hairpin structure.
In many cases, reduced enzyme activity is correlated with growth phenotype, as previously observed (19,20,28). However, in some cases (e.g. F119V/F159A), cells showing prominent mutant growth phenotypes produced mutant enzymes displaying normal enzymatic properties when examined in vitro (Table II). One possible explanation may be that altering phenylalanines in TM1,TM2 influences the assembly efficiency of this region, which could result in an apparent growth irregularity. However, once the enzyme is assembled, it behaves normally. This suggestion is intriguing in view of the recently proposed role of TM1/TM2 as a catalyst in the membrane assembly of the H ϩ -ATPase in Neurospora (46). Alternatively, since the in vivo enzyme is displaced from equilibrium and experiences numerous potential constraints on catalytic activity such as pH gradients, membrane voltage, turgor pressure, and regulation due to phosphorylation, it may be that small perturbations are amplified and show more pronounced affects on cell physiology. In contrast, the in vitro enzyme operating at V max capacity would not be subject to these constraints, and would not be expected to show significant differences from the wild type enzyme under the same conditions. This latter explanation would be pertinent to all subtly perturbing mutant enzymes which show differential in vivo and in vitro properties.
Is a Proline Kink Important in TM1?-The helix in TM1 was predicted to kink toward TM2 about one third into the bilayer due to the presence of a flexible Gly-122 and a helix-breaking Pro-123 (23). Pro is an unusual amino acid in which its side chain is cyclized back on the backbone amide position and backbone conformation is restricted leading to a kinked structure (47). From a structural point of view, a Pro kink provides a way to make curved helices that could be packed into a funnel-like structure or into a cage-like structure (48). Such packing is important for membrane proteins such as bacteriorhodopsin in which structural interactions between helices can be maximized (49). In fact, Pro is frequently found in bilayerassociated structures of membrane proteins (50). To examine the role of Pro-123, substitutions were made with Ala, Gly, or Phe. The mutations produced viable enzymes with moderate affects on growth phenotype and ATP hydrolysis (Table I and  II), but a minor effect on the initial rate of H ϩ transport. The substitution of the preceding Gly-122 residue with either Ala or Asn produced very prominent hygromycin B resistance, and the G122N substitution also gave a low pH-sensitive phenotype. The G122A mutant enzyme was diminished in ATP hydrolysis, although surprisingly, the G122N mutant enzyme appeared wild type in ATP hydrolysis and H ϩ transport. Swapping the Gly-122 and Pro-123 positions was not tolerated, and the introduction of a double Pro-122/Pro-123 sequence, which should disrupt the helix, was also lethal. These results suggest that neither Gly-122 nor Pro-123 are essential for catalytic activity and the formation of a Pro kink, if it exists, is not critical. However, since helix-forming residues like Ala induce growth phenotypes indicative of a perturbed enzyme, it may be that a Pro kink is important for efficient assembly. The results do clearly suggest that TM1 is required to be helical because the double Pro insertion was lethal. In addition, the introduction of Pro at position 155 in TM2 was also lethal, suggesting that TM2 could not sustain a kinked structure.
Other Features of Hairpin Region-Glu-129 is predicted to lie in the middle of transmembrane segment 1. In the Neurospora H ϩ -ATPase, this residue can be modified by DCCD (34), which inhibits enzyme activity. In view of the role bilayer-associated DCCD-reactive Glu or Asp residues play in H ϩ transport by F 0 F 1 -type H ϩ -ATPases, we explored whether Glu-129 might play a similar role. The substitution of Ala for Glu at this position had a weak effect on phenotype producing only mild hygromycin B resistance (Table I). The mutant enzyme showed somewhat lower levels of ATP hydrolysis and initial rate of H ϩ transport, but the steady state pH gradient formation was comparable to wild type.
A prediction of the molecular structure model is that Met-128 on helix 1 should lie within close proximity to Cys-148 on helix 2 (23). An M128C mutation was created to examine whether a disulfide linkage could be established between the 2 residues. The M128C mutant showed perturbed growth phenotypes, and it was significantly reduced in catalytic activity (Table II). However, there was no effect of sulfhydryl reagents on this enzyme, which would be consistent with disulfide bond formation. This analysis was further complicated by the observation that Ala and Ser substitutions at Met-128 were not viable. Thus, it was not possible to distinguish between perturbations created by the mutation in helix 1, and a potential cross-linking of the hairpin structure.
Conclusion-The most significant finding in this study is that the aromatic side chains on TM2 appear most important for the structural stability and viability of the H ϩ -ATPase. In addition, the presence of aromatic side groups at positions 120 and 144 are essential for enzyme function. The Pro-123 on TM1 is not critical to the enzyme, but an intact ␣-helix is important. Overall, these results suggest that the helical hairpin model provides a reasonable description of the TM1/TM2 region of the H ϩ -ATPase. They provide additional evidence that perturbations within the helical hairpin, which are most significantly manifested as changes in cellular growth phenotypes, alter the efficiency of enzyme action.