Biophysical Characterization of a Thermoalkaliphilic Molecular Motor with a High Stepping Torque Gives Insight into Evolutionary ATP Synthase Adaptation*

F1F0 ATP synthases are bidirectional molecular motors that translocate protons across the cell membrane by either synthesizing or hydrolyzing ATP. Alkaliphile ATP synthases are highly adapted, performing oxidative phosphorylation at high pH against an inverted pH gradient (acidin/alkalineout). Unlike mesophilic ATP synthases, alkaliphilic enzymes have tightly regulated ATP hydrolysis activity, which can be relieved in the presence of lauryldimethylamine oxide. Here, we characterized the rotary dynamics of the Caldalkalibacillus thermarum TA2.A1 F1 ATPase (TA2F1) with two forms of single molecule analysis, a magnetic bead duplex and a gold nanoparticle. TA2F1 rotated in a counterclockwise direction in both systems, adhering to Michaelis-Menten kinetics with a maximum rotation rate (Vmax) of 112.4 revolutions/s. TA2F1 displayed 120° unitary steps coupled with ATP hydrolysis. Torque measurements revealed the highest torque (52.4 piconewtons) derived from an F1 molecule using fluctuation theorem. The implications of high torque in terms of extreme environment adaptation are discussed.

F 1 F 0 ATP synthases are bidirectional molecular motors that translocate protons across the cell membrane by either synthesizing or hydrolyzing ATP. Alkaliphile ATP synthases are highly adapted, performing oxidative phosphorylation at high pH against an inverted pH gradient (acid in /alkaline out ). Unlike mesophilic ATP synthases, alkaliphilic enzymes have tightly regulated ATP hydrolysis activity, which can be relieved in the presence of lauryldimethylamine oxide. Here, we characterized the rotary dynamics of the Caldalkalibacillus thermarum TA2.A1 F 1 ATPase (TA2F 1 ) with two forms of single molecule analysis, a magnetic bead duplex and a gold nanoparticle. TA2F 1 rotated in a counterclockwise direction in both systems, adhering to Michaelis-Menten kinetics with a maximum rotation rate (V max ) of 112.4 revolutions/s. TA2F 1 displayed 120°unitary steps coupled with ATP hydrolysis. Torque measurements revealed the highest torque (52.4 piconewtons) derived from an F 1 molecule using fluctuation theorem. The implications of high torque in terms of extreme environment adaptation are discussed.
ATP synthases are membrane-bound rotary molecular motors that are the major enzymes responsible for ATP synthesis in the majority of aerobic life (1,2). A key feature of the enzyme is the mechanism by which the translocation of protons (or Na ϩ ions) across the membrane barrier is linked to either synthesizing or hydrolyzing ATP (1)(2)(3). Bacterial F 1 F 0 -type ATP synthases are two-domain enzymes and are typically composed of eight subunits, ␣ 3 ␤ 3 ␥␦⑀ab 2 c 10 -15 (4,5). The soluble F 1 domain is composed of the ␣ 3 ␤ 3 ␥␦⑀ subunits in which the ␣and ␤-subunits are alternatively arranged in a hexameric ringshaped catalytic core. ATP hydrolysis and synthesis occur in the catalytic sites that are formed by each of the ␤-subunits (4,6). The core rotary shaft is composed of the ␥and ⑀-subunits. The ␥-subunit penetrates the center of the ␣ 3 ␤ 3 catalytic hexamer at one end and is associated with the a-and c-subunits of the F 0 domain at the other end (7). The F 0 domain is composed of ab 2 c 9 -15 subunits and is membrane-bound (8). The c-subunits form a ring parallel to the membrane (9); however, although stoichiometry is fixed within a species, it is variable between species (10 -15). The a-subunit forms a collar around the c-ring and is thought to conduct ions both into and out of the c-ring in a two-channel model (16 -19). The c-ring rotates and, together with a-subunit, conducts protons or Na ϩ ions across the cell membrane (9,20). The tightly associated ␥and ⑀-subunits together with the c-ring functionally form the rotor portion of the enzyme, whereas the ␣ 3 ␤ 3 ␦ab 2 subunits form the stator (4,5). The coupling of ion translocation through the ac-ring complex to the clockwise rotation of the rotor forces conformational changes within the ␤-subunits in the ␣ 3 ␤ 3 catalytic hexamer, catalyzing the synthesis of ATP from ADP and inorganic phosphate (21). Conversely, the ␣ 3 ␤ 3 catalytic hexamer can hydrolyze ATP, causing anticlockwise rotation of the ␥⑀c 10 -15 rotor, resulting in the pumping of either protons or Na ϩ ions into the bulk phase (21)(22)(23)(24). It is also noteworthy that ATP hydrolysis is suppressed in aerobic alkaliphilic organisms (25)(26)(27).
When the F 1 domain is separated from the F 0 domain it can be regarded as an ATPase and is useful for the study of the mechanical properties of ATP/ADP-P i catalysis and enzyme function (28). Indeed, the rotation of F 1 ATPases has been visualized using optical microscopy by attaching a probe to the ␥or ⑀-subunit in the relative area of where the F 0 domain would be located had it been attached (23,28,29). As this method of study has developed, the type of probe utilized has diversified from actin filaments (23,30) into gold nanoparticles/nanorods (31,32), magnetic beads (33,34), polystyrene beads (35,36), and fluorescent molecules (37)(38)(39). The two most heavily studied bacterial F 1 ATPases come from the thermophile Bacillus PS3 (TF 1 ) and the mesophile Escherichia coli (EF 1 ). Both enzymes are capable of ATP hydrolysis activity, rotating stepwise at an ATP concentration-dependent rate in an anticlockwise direction (23,40). Because the hexameric ␣ 3 ␤ 3 structure harbors three catalytic sites, the basic step size is 120°coupled to either the consumption or production of one ATP molecule (41). For both TF 1 and EF 1 , these 120°steps have been shown to consist of substeps of 80°and 40°and of 85°and 35°substeps, respectively (35,40,42). The 80°/85°substeps are proposed to be trig-gered by ATP binding and ADP release and are referred to as ATP-binding (ATP-waiting) dwells (37), whereas the 40°/35°s ubsteps have been shown to occur after ATP cleavage and release of inorganic phosphate and are referred to as catalytic dwells (43).
Upon catalysis, the generation of torque at the interface between the rotor (␥⑀ subcomplex) and the stator (␣ 3 ␤ 3 ␦ subcomplex) results in the rotation of the F 1 by using the chemical free energy from ATP hydrolysis. Due to the interconversion of chemical free energy and electrochemical potential via mechanical rotation, torque is an important factor affecting the energy conversion efficiency of a molecular rotor. Stepping torque has been estimated for TF 1 (41,44), EF 1 (40), the thermophilic Thermus thermophilus V 1 ATPase (TtV 1 ) (44,45), and the acidophilic Enterococcus hirae V 1 ATPase (EhV 1 ) (44), the last of which is proposed to act solely as a V 1 V 0 -type ATPase (24,44,46). Interestingly, of the F 1 -type enzymes described here, the torque appears to be higher in the thermophilic TF 1 and, according to some reports, lower in EF 1 . This implies that adaptation to temperature might result in an enzyme with higher torque. In agreement with this, of the V 1 -type enzymes, the TtV 1 torque is greater than that of EhV 1 ; however, the torque difference between TtV 1 and EhV 1 is much greater than that between TF 1 and EF 1 , suggesting that function may also have a role in defining torque, i.e. ATP synthesis versus hydrolysis (see Table 1).
In this study, we examine the F 1 ATPase from the thermoalkaliphilic Caldalkalibacillus thermarum TA2.A1. Alkaliphile ATP synthases are highly adapted, performing oxidative phosphorylation at high pH against an inverted pH gradient (acid in / alkaline out ), thereby challenging Mitchell's chemiosmotic model (25,27,(47)(48)(49). The environmental pressure of an alkaline environment, together with increased membrane permeability to ions at high temperature (50), results in a severe pressure to conserve energy. Unlike the ATP synthases/hydrolases described here, alkaliphilic enzymes have tightly regulated ATP hydrolysis activity (25)(26)(27), which can be relieved in the presence of LDAO 3 (26,(51)(52)(53). In addition to this, both the ␥and ⑀-subunits of the C. thermarum TA2.A1 ATP synthase (TA2F 1 F 0 ) are proposed to have regulatory roles in ATP hydrolysis (51,53). Because this enzyme is the antithesis of the EhV 1 V 0 (i.e. it only performs ATP synthesis (26)), it was of significant interest to examine both catalytic mechanism and torque. In addition, due to the thermophilic origin of TA2F 1 F 0 , we are also able to directly compare the effect of alkaliphily on torque. Here, we characterized the rotary dynamics of TA2F 1 with two forms of single molecule analysis, a magnetic bead and a gold nanoparticle.

Results
Expression of TA2F 1 and TA2F 1 ␥2c in E. coli-To examine rotary dynamics in F 1 ATPases, the enzyme must be capable of being labeled with a trackable probe such as a 200-nm magnetic bead duplex or a 40-nm gold nanoparticle. For other enzymes such as the TF 1 ATPase (23,31,34,42,54), this was achieved by the introduction of two cysteine residues in the ␥-subunit. The F 1 portion of the atp operon from C. thermarum TA2.A1, coding for TA2F 1 ␣ 3 ␤ 3 ␥␦⑀, with a hexahistidine tag at the N terminus of the ␤-subunit, and the same enzyme including two cysteines introduced at positions 107 and 210 in the ␥-subunit (TA2F 1 ␥2c; Fig. 1A) were cloned into the expression plasmid pTrc99A and overexpressed in the unc deletion mutant E. coli DK8. TA2F 1 and TA2F 1 ␥2c were extracted from E. coli strain DK8 cytoplasmic fractions and purified via a three-step purification procedure. SDS-PAGE analysis of the purified recombinant mutant enzyme (TA2F 1 ␥2c) revealed five subunits (viz. ␣, ␤, ␥, ␦, and ⑀) that were identical to those of the native F 1 purified from the overexpressed WT TA2F 1 previously by Keis et al. (51) (Fig. 1B, lane 2). Postpurification, TA2F 1 ␥2c was specifically biotin-modified via the cysteine residues at positions 107 and 210 in the ␥-subunit for attachment of rotation probes (Fig.  1B, lane 1). This was confirmed by Western blotting (Fig. 1B, lane 2), revealing that biotin was specifically incorporated into the ␥-subunit of TA2F 1 ␥2c.
ATP Hydrolysis of TA2F 1 ␥2c-Biotin-A feature of the TA2F 1 enzyme is its specific blockage in ATP hydrolysis activity and that this activity can be stimulated up to 30-fold with 0.05% LDAO (26,51). Like the recombinantly expressed native TA2F 1 , TA2F 1 ␥2c-biotin was blocked in ATP hydrolysis activity. Neither the WT nor the biotin-modified mutant displayed any measureable ATP hydrolysis activity (Fig. 1, C and D). ATP hydrolysis could be induced in TA2F 1 ␥2c-biotin at both 25 and 45°C with 0.1% LDAO (Fig. 1, E  We have further optimized the assay by introduction of 50 mM KCl to optimize pyruvate kinase activity (data not shown). In this optimized assay system, upon addition of ATP, we observe a small sharp drop indicative of a small amount of ADP present in the ATP. This is followed by a flat line indicating that the TA2F 1 ATPase has close to zero ATP hydrolysis activity at 25, 45, and 65°C (see Fig. 1, C-E, upon ATP introduction).
Rotation Velocity Is Dependent on ATP Concentration-Because TA2F 1 does not natively catalyze hydrolysis and we sought to examine TA2F 1 rotational dynamics with ATP hydrolysis-driven rotation, we required a system capable of activating F 1 ATPase activity. Fortunately, we have two methods to achieve this in the form of LDAO and the use of magnetic beads with magnetic tweezers, the latter of which is well documented to aid in relieving the TF 1 ATPase inhibited states (34,55,56). Although the use of magnetic beads is a powerful tool, to exam- 3 The abbreviations used are: LDAO, lauryldimethylamine oxide; TA2F 1  ine the true maximum rotation speed, a probe with no viscous drag is required. Therefore, to explore the kinetic properties of the TA2F 1 , a 40-nm gold nanoparticle was utilized. The hydrodynamic friction on a 40-nm gold nanoparticle is practically negligible, allowing the observation of unloaded rotation of TA2F 1 to be observed provided the bead is unobstructed and free to rotate in the assay solution (42). Considering our low flow chamber volume and requirement for accessibility to the chamber volume, assays at 65°C present a significant technical problem. However, because ATP hydrolysis at 25°C is 73% of that possible at 65°C (see Fig. 1E), we considered it appropriate to conduct all rotation experiments forthwith at 25°C. To observe single enzyme rotation, TA2F 1 ␥2c-biotin was immobilized on an NTA-modified glass surface in a flow chamber using hexahistidine tags at the N terminus of the ␤-subunit, and a 40-nm streptavidin-functionalized gold nanoparticle was attached at the top of the ␥-subunit as a rotation probe ( Fig.  2A). Initial experiments did not include an ATP regeneration system or LDAO. These experiments revealed very few rotating molecules, each of which only performed two to five revolutions before lapsing into an inactive state. Henceforth, we performed rotation studies using TA2F 1 ␥2c using a 40-nm gold  nanoparticle system in the presence of 0.1% LDAO and an ATP regeneration system. Rotation was observed at a variety of ATP concentrations using total internal reflection dark field microscopy (31) recorded at a rate of 10,000 frames per second (fps). In the presence of LDAO, TA2F 1 rotated counterclockwise for continuous revolutions that decreased in velocity upon decrease in ATP concentration (Fig. 2, B and C). Above 1 mM, rotation rates were essentially constant, and below 200 M ATP, the rotation rates were proportional to ATP concentration. This suggests that ATP binding may be rate-limiting at ATP concentrations below 200 M. Rotation velocities at various ATP concentrations collectively followed Michaelis-Menten kinetics with a V max of 112.4 Under V max conditions, each 120°step was completed within 0.2 ms; however, this time period was highly dependent on ATP concentration, underscoring that rotation rate is directly related to the duration of the intervening 120°pauses. At 2 mM ATP, considerably above the K m of 46.85 Ϯ 3.69 M, these pauses are proposed to represent catalytic dwells. To obtain the time constants and kinetic parameters, the duration of these pauses were examined. Unambiguous analysis was achieved by examination of the first 30 s of traces from 10 molecules distributed over 2 orders magnitude of substrate concentration. These were modeled using a three-consecutive reaction scheme representing ATP binding, hydrolysis, and ADP/P i release. All distributions of the durations showed a convex shape (Fig. 3, G- (35). This is achieved by closely examining rotational stepping behavior around the K m value. Unfortunately, we were unable to observe substepping using the full TA2F 1 construct with a gold nanoparticle; however, the 120°s teps were clearly broadened at 50 M ATP (K m ϭ 46.85 Ϯ 3.69 M; data not shown).

Rotation of TA2F 1 Using ATP and ATP␥S with Magnetic Beads Reveals a Six-dwell Rotation
Profile-To avoid using LDAO as an activation method, we replaced the gold nanoparticle with magnetic beads, allowing mechanical activation; this is a well documented method to aid in relieving TF 1 inhibited states (34,55,56). In place of a gold nanoparticle, a ϳ200-nm streptavidin-functionalized magnetic bead duplex was attached to the ␥-subunit as a rotation probe (Fig. 4A). Initially, rotation was observed at a saturating ATP concentration of 2 mM using phase-contrast microscopy at 30 fps. As with the gold nanoparticle rotation system, in the absence of LDAO or any mechanical stimulation, very few rotating molecules were observed. Each molecule in this system only performed two to five revolutions before lapsing into an inactive state. However, upon a single forced counterclockwise rotation using magnetic tweezers, TA2F 1 was capable of ATP hydrolysis (Fig. 4B)-driven rotation at speeds of up to 6.9 revolutions s Ϫ1 (Fig. 4, C and D). Rotation was observed at a variety of ATP concentrations that decreased in velocity upon decrease in ATP concentration, obeying Michaelis-Menten kinetics. The maximum rotary velocity measured under V max conditions was 6.5 Ϯ 0.4 revolutions s Ϫ1 with an apparent ATP K m of 2.72 Ϯ 0.45 M (Fig. 4, C  and D). The suppression of rotation velocity and lowering of apparent K m are a typical feature of using a large, high viscosity probe for rotation observation (31,33,44). The second-order rate constant (K on [ATP]) was determined using 3 ϫ V max /K m derived from Fig. 4D and was estimated to be (9.90 Ϯ 0.1) ϫ 10 6 M Ϫ1 s Ϫ1 , assuming that three ATP molecules are hydrolyzed per revolution of the ␥-subunit. This is very much in agreement with observations with the values for ATP of (7.20 Ϯ 0.1) ϫ 10 6 M Ϫ1 s Ϫ1 in the same assay system using a gold nanoparticle. However, in the magnetic bead system, these 120°steps were only clearly observed during rotation at very low ATP concentrations (Fig. 5, A-F). Because this is due to the combined effect of camera and probe response times coupled with the high viscous friction of the probe, we changed to using a fast framing camera at 5000 fps to take advantage of the ability to capture stepping. Initially, ATP was used as substrate; however, the 120°s teps were simply broadened at 3 M ATP. In light of this, we attempted to slow the catalytic steps to clarify the stepping behavior. One such method of achieving this is using the slowly hydrolyzable ATP analogue ATP␥S (35); however, using this analogue removes the possibility of harnessing the ATP regeneration assay to remove the buildup of Mg-ADP. To limit this unwanted effect, observation times in the rotation assay were limited to 10 min before flushing the flow cell with fresh ATP␥S at the desired concentration. Magnetic bead rotation observations at various ATP␥S concentrations revealed a maximum rotation velocity of 4.02 Ϯ 0.3 revolutions s Ϫ1 under saturating ATP␥S concentrations (2 mM; Fig. 6A). From the various ATP␥S concentrations tested, TA2F 1 has an apparent K m of 5.69 Ϯ 1.33 M (Fig. 6B). Due to the recording rate being at 5000 fps, discrete 120°steps were observed at most concentrations of ATP␥S tested (Fig. 6, C-H). However, at a concentration of 5 M ATP␥S, multistep (six-step) rotary profiles with 50°and 70°s ubsteps were observed (data not shown). Unfortunately, due to the rarity of the events measured and the less than discrete localization, this is a tentative estimate. TA2F 1 Torque Is the Highest Estimated Using Fluctuation Theorem-Because the conservation of energy is key to survival at alkaline pH, the stepping torque of TA2F 1 was measured to examine the power required to turn the rotor to drive ATP hydrolysis. One of the most accurate methods for this type of analysis is by utilizing fluctuation theorem (FT) (44). FT describes entropy production in small non-equilibrium systems, various colloidal particle systems, and more recently biological systems such as RNA hairpins and the F 1 ATPase (44). FT is of particular use as it estimates torque by use of steps isolated from the time courses of rotary angles (Fig. 7A) without the need for an accurate measurement of the frictional drag coefficient of the probe itself. Because this analysis requires perfect 120°stepping motion with close to equal dwell time at each step, we used rotational data of TA2F 1 hydrolyzing ATP␥S under V max conditions. Ten TA2F 1 rotation traces recorded in the presence of 2 mM ATP␥S were analyzed, revealing a rotary torque of 52.4 Ϯ 4 pNnm (Fig. 7, B and C). For comparison with another ATPase under the same conditions and using the same analysis, we also measured the torque for TF 1 . In our hands, TF 1 generated a torque of 41.4 Ϯ 2.2 pNnm, which is quite comparable with that reported previously (Table 1), strongly supporting the reliability of the estimated torque value reported.

Discussion
Here, we have conducted the first full biophysical interrogation of a thermoalkaliphilic F 1 ATPase in single molecule analysis with simultaneous comparison with biochemical methods. In this study, we comprehensively demonstrate that the TA2F 1 is a unidirectional rotary molecular motor. In previous studies with TA2F 1 , the small amount of hydrolysis activity observed was likely due to the presence of ␣ 3 ␤ 3 incomplete complex (51). However, the fully intact TA2F 1 used here does not natively hydrolyze ATP. However, in agreement with previous studies (26,49,51,52), TA2F 1 hydrolyzes ATP upon either mechanical or chemical induction. In TF 1 studies, the requirement for mechanical activation was proposed to be due to inhibition from a classical feedback loop of product inhibition by Mg-ADP (57); however, rotation was always observable without activation mechanically or with LDAO. Furthermore, TF 1 biochemically measured ATP hydrolysis activity was only enhanced 3-4-fold with LDAO (58). In contrast, TA2F 1 is activated from an inactive state with LDAO or mechanical activation, the mechanisms by which remain to be fully described. Interestingly, the K m and K on values obtained from bulk-phase and rotation experiments do not agree. The estimation of K m revealed by gold nanoparticle rotation was 5.95-fold lower than that determined by the bulk-phase biochemical assay, and the estimation of K on reflects this. However, this is unsurprising when considering that single molecule analysis specifically selects active molecules, whereas the bulk-phase assay does not, providing an ensemble measurement. The lowering of K m is a known hallmark for competitive inhibition, suggesting that inactive molecules are likely due to ADP inhibition. In support of this, it is well known that ATP hydrolysis in TF 1 is regulated by Mg-ADP inhibition to prevent wasteful ATP consumption (57), a phenomenon yet to be explored in TA2F 1 . In TF 1 studies, the difference between single molecule results and biochemical results has long been attributed to Mg-ADP inhibition. Under single molecule conditions using a 40-nm gold nanoparticle, TF 1 has a maximal rotation velocity of 200 revolutions s Ϫ1 ; however, in the ATP-regenerating assay under V max conditions, only 270 ATP/s/molecule were consumed. Because three ATP molecules are consumed per revolution, bulk biochemis-try results would imply only 90 revolutions s Ϫ1 are possible; however, TF 1 is 55% faster according to single molecule data (59). In line with this, according to TA2F 1 biochemical data, 124.3 ATP/s/molecule are consumed; therefore a maximum velocity of 41.44 revolutions s Ϫ1 is expected at 25°C. However, TA2F 1 single molecule results show a maximal velocity of 112.4 revolutions s Ϫ1 , a 63% faster V max velocity than that derived from biochemical data. This observation is strikingly close to the biochemical versus single molecule discrepancies reported for the TF 1 and EhV 1 (24,44,59).
At almost all ATP and ATP␥S concentrations tested above or below K m , TA2F 1 rotation profiles exhibited three dwells separated by 120° (Figs. 3 and 6) when recorded at 5000 fps. Upon using the slowly hydrolyzable substrate ATP␥S, six substeps were able to be resolved close to ATP␥S K m . This is in close agreement with observations of two other F 1 ATPases, TF 1 and EF 1 , that have both been reported to rotate with six dwells per revolution, close to their respective K m values (35,40). This observation together with the lack of six-step rotation observed in the EhV 1 ATPases (24) and structural differences between F 1 and V 1 ATPases (60,61) implies that the interactions between rotor and stator determine the presence or absence of substeps upon enzyme rotation. However, we report the angular spacing of these 50°and 70°substeps conservatively due to insufficient quality and quantity of the molecules observed. In addition, it was observed that in almost all cases, one of the six substeps was masked showing the "six-step profile." Conservation of energy is a constant challenge under thermoalkaliphilic conditions. In light of this, it is unsurprising that the stepping torque derived using fluctuation theorem for TA2F 1 is very high (52.4 Ϯ 4 pNnm); however, it was unexpected that it would be 20% higher than that for TF 1 (44). Although TF 1 and other F 1 ATPases studied in rotation studies come from organisms that grow at neutral (E. coli and Bacillus PS3) and acidic (E. hirae) pH, the C. thermarum TA2.A1 F 1 F 0 is highly adapted to function at alkaline pH ( Fig. 8A) (48,53,62). Given the difference in torque between TF 1 and TA2F 1 (Fig. 8B) despite both enzymes originating from organisms that grow optimally at 65°C and that the internal pH of both organisms is similar at 7.5-8.0 (63,64), we suggest that the differences in torque across species may be more related to evolutionary adaptation to external pH than temperature (Fig. 8, A and B). It is also noteworthy that both TA2F 1 and TF 1 torque was measured at 25°C, ruling out temperature as a potential evolutionary pressure on torque. Even if the torque of TF 1 measured at  65°C was the same as that of EF 1 measured at 25°C, TA2F 1 has 13 pNnm greater torque than TF 1 . Notwithstanding, a higher torque implies a much stiffer rotor. One of the key features of alkaliphilic adaptation is a large c-subunit ring rotor. In the case of C. thermarum TA2.A1, this is a 13-mer (15) with an F 1 torque of 52 pNnm in comparison with the E. coli 10-mer (65) with an F 1 torque of 30 pNnm (40) and the E. hirae 10-mer (46, 66) with a V 1 torque of 20 pNnm (44). This ion coupling ratio is critical in understanding the power of the molecular motor because although velocity and torque are important for determining enzyme efficiency the gearing (i.e. the number of protons coupled to a single revolution event) is also important when considering purpose and efficiency and that C. thermarum TA2.A1 must partially maintain internal pH by synthesis of ATP. However, it is clear that differences in torque cannot be fully ascribed to the number of c-subunits because E. coli and E. hirae both have 10 c-subunits, but E. coli has a 10 pNnm greater torque (40,44,65,66). This strongly suggests that although c-subunit ring size may have a mechanical influence it is the native direction of catalysis (i.e. TA2F 1 is a synthase, and EhF 1 is a hydrolase) that has the greater mechanical influence. This seems logical given that there is no torque without resistance; the greater the resistance, the greater the torque. Because we are examining TA2F 1 in hydrolysis, it is rotating against its native function; hence, one would expect more resistance.
However, in the field of ATPase dynamics, there are several methods by which to estimate torque values for enzymes (Table  1) (28,40,44,(67)(68)(69)(70). Early EF 1 torque measurements were estimated to be 50 pNnm using the drag on the curvature of an actin filament during rotation (67). However, more recently, when rotation of EF 1 was assessed using gold nanorods and analysis of torque was conducted using a "power stroke" drag force method, the torque was estimated to be 63 pNnm (68). Clearly, different experimental methods and systems can result in different torque estimations. Here, we use a very conserva-tive method of analysis, "fluctuation theorem," to directly compare TA2F 1 stepping torque with that of TF 1 and other enzymes that have been previously analyzed with similar methodology, thus allowing a very reliable comparison between enzymes. Although the stepping torque measured here reflects the shape of rotation potential, future measurement of stall torque will reflect the thermodynamic efficiency. Interestingly, the complete E. hirae V 1 V 0 ATPase has a higher torque than V 1 (44). This could be due to a "tighter" overall structure influenced by V 1 -V 0 contacts and/or the influence of specific subunit interactions, i.e. the observation of reinforced torque by genetic fusion of the F-subunit to the D-subunit in T. thermophilus V 1 (70). It remains to be examined whether this is a V-type ATPase feature or applies more generally. Future studies will focus on what features dictate torque, activity of the TA2F 1 F 0 complete with the c 13 -ring, the specific role of ADP inhibition, the substep size, and the mechanical roles of the ⑀and/or ␥-subunits in TA2F 1 F 0 regulation of hydrolysis.

Experimental Procedures
Bacterial Strains, Plasmids, and Growth Conditions-E. coli DH5␣ (71) was used for all cloning experiments, and E. coli DK8 (72) lacking the ATP synthase genes encoding the unc operon (⌬atp) was used to overproduce the TA2F 1 (63,73). For cloning and overexpression of TA2F 1 enzyme, the plasmid pTrc99A (Amersham Biosciences) was used. To overproduce F 1 ATPases, transformants of E. coli DK8 were routinely grown in 2ϫ YT medium (74) containing 2 g/liter glucose and 100 g/ml ampicillin at 37°C with shaking at 200 rpm.
Construction of pTA2F 1 ␥2c (␥H107C/␥S210C) and TA2F 1 ␣ 3 ␤ 3 ␥␦⑀ Complexes-In a previous study, the genes coding for the C. thermarum TA2.A1 F 1 ATPase were cloned into the expression vector pTrc99A, yielding the plasmid pTrcF1 (75). To substitute the native histidine 107 and serine 210 in the ␥-subunit of TA2F 1 to cysteine residues, double site-directed mutagenesis was carried out using the plasmid pTrcF 1 as template with a three-fragment strategy based on the overlap extension PCR method (76). To construct pTA2F 1 ␥2c, an 837-bp PCR product was generated using primers Tg-cysF and Tg-cysMutR, and a 1492-bp PCR product was generated using primers Tg-cysMutF2 and Tg-cysR. A 331-bp PCR product that overlapped the 3Ј-and 5Ј-ends of the 837-and 1492-bp PCR products, respectively, was generated using primers Tg-cysMutF1 and Tg-cysMutR2. The 837-and 331-bp PCR products were then joined by overlap PCR to generate a 1145-bp PCR product using primers Tg-cysF and Tg-cysMutR2. Lastly, the 1145-and 1492-bp PCR products were joined by overlap PCR to generate a 2615-bp fragment using primers Tg-cysF and Tg-cysR and digested with AflII and AfeI, and the resulting 2292-bp fragment was cloned with a 6817-bp fragment from pTrcF1 digested with the same enzymes (75). This new plasmid was designated pDMTA11. All PCRs were conducted with the Phusion high fidelity PCR kit (New England Biosciences) according to the manufacturer's instructions. Mutant inserts in both plasmids were confirmed to be correct by DNA sequencing, and the recombinant plasmid was transformed into E. coli DK8. Primer sequences and specific details are displayed in Table 2. Shown is a comparison of mild acidophilic (E. hirae), mesophilic (E. coli), thermophilic (Bacillus PS3), and thermoalkaliphilic (C. thermarum TA2.A1) A, optimal temperature and pH of growth conditions (temperature, solid symbols, left y axis; pH, empty circles, right y axis). B, torque of the ATPase. The torque value here for EF 1 was derived using a method similar to that used for TF 1 and TA2F 1 in this study. For methodology details and appropriate references, see Table 1.
Expression and Purification of TA2F 1 and TA2F 1 ␥2c-The procedure applied was a modified hybrid of that in McMillan et al. (48) and Keis et al. (51). DK8 harboring plasmid pTrcF1 or pDMTA11 was grown to an A 600 of 0.4. The culture was induced with 1 mM isopropyl ␤-D-thiogalactopyranoside, and incubation continued for 4 h. Cells were harvested, washed with precooled resuspension buffer (50 mM Tris (pH 8.0), 5 mM MgCl 2 ), and resuspended in the same buffer. Phenylmethylsulfonyl fluoride (PMSF) was added to 0.1 mM, and the cells were disrupted by two passages through a French pressure cell at 20,000 p.s.i. Pancreatic DNase I was added to 0.1 mg/ml, and the mixture was kept on ice for 30 min. Lysate was cleared of debris by centrifugation at 8000 ϫ g for 10 min, and the inverted membrane vesicles were pelleted from the supernatant at 180,000 ϫ g for 1 h at 4°C. To extract TA2F 1 or TA2F 1 ␥2c, the supernatant was heated to 60°C for 15 min to precipitate contaminating E. coli proteins, and the precipitate was pelleted from the supernatant at 180,000 ϫ g for 30 min at 4°C. The supernatant was then adjusted to 500 mM NaCl and 10 mM imidazole before application to a 20-ml column of nickel-Sepharose High Performance (GE Healthcare) that was previously washed with water and equilibrated with buffer A (50 mM Tris (pH 8.0), 2 mM MgCl 2 , 10% glycerol, 500 mM NaCl) containing 10 mM imidazole. To remove contaminating proteins, the resin was washed with buffer A containing 30 mM imidazole, and TA2F 1 , TA2F 1 ␥2c, or TA2F 1 ⑀2c was eluted with buffer A supplemented with 200 mM imidazole. The eluted TA2F 1 , TA2F 1 ␥2c, or TA2F 1 ⑀2c was pooled and concentrated to 10 -30 mg/ml using Amicon Ultra centrifugal filter devices (molecular weight cutoff, 50,000). The concentrated sample was then applied to a Superdex 200 10/300 (GE Healthcare) gel filtration column and fractionated with 50 mM Tris-HCl (pH 8.0), 2 mM MgCl 2 , 100 mM NaCl buffer. The eluted TA2F 1 -or TA2F 1 ␥2c-containing fractions (a peak between 370 and 380 kDa) were pooled and concentrated to 10 mg/ml using Amicon Ultra centrifugal filter devices (molecular weight cutoff, 50,000) in 50 mM Tris-HCl (pH 8.0) containing 2 mM MgCl 2 , 100 mM NaCl, and 10% glycerol.
Protein Assay-Protein concentrations were determined using a bicinchoninic acid (BCA) protein assay kit (Sigma) with bovine serum albumin as the standard.
Biochemical Assays-ATP hydrolysis activity was measured using a spectrophotometric ATP-regenerating assay at 25 and 45°C similar to that used in McMillan et al. (48,52). The assay mixture contained 100 mM Tris⅐Cl (pH 8.0), 1.5 mM MgCl 2 , 3 mM phosphoenolpyruvate, 50 mM KCl, 0.25 mM NADH, 5 units/ml pyruvate kinase, 6.4 units/ml lactate dehydrogenase, and variable concentrations of sodium ATP (Sigma) or lithium ATP␥S (Roche Applied Science). For measurements at 65°C, thermophilic enzymes, 10 units/ml pyruvate kinase HI1 and 6.4 units/ml lactate dehydrogenase 2, were used (Funakoshi Ltd.). The reaction was initiated by the addition of enzyme or ATP, as indicated, into 1 ml of assay mixture, and the rate of NADH oxidation was monitored continuously at 340 nm using a Jasco V-660 spectrophotometer. Approximately 5-10 g (as indicated) of recombinant TA2F 1 or TA2F 1 ␥2c subunit complexes was used for assays.
Single Molecule Rotation Assays-Single molecule rotation of TA2F 1 ␥2c was measured as described previously using either a ϳ200-nm magnetic bead duplex (high load probe) or a 40-nm gold nanoparticle (low load probe) (33,44,55,78). Briefly, a flow cell was constructed using a Ni 2ϩ -NTA-modified glass surface and an uncoated cover glass (Matsunami Glass Ltd., Japan), and biotinylated TA2F 1 ␥2c subunit complexes were immobilized on the Ni 2ϩ -NTA surface via a hexahistidine tag introduced at the N terminus of the ␤-subunit. After 5 min, non-immobilized TA2F 1 ␥2c subunit complexes were removed using buffer A (100 mM Tris (pH 8.0), 1.5 mM MgCl 2 , 50 mM KCl) followed by a 5-min incubation in 20 mg/ml bovine serum albumin in buffer A to prevent nonspecific interactions of probes with the surface. Either ϳ0.2-m high load streptavidin-functionalized magnetic bead duplexes (Seradyn Inc., Indi- anapolis, IN) or a low load 40-nm streptavidin-functionalized gold nanoparticle (British BioCell International, UK) were subsequently attached to the biotinylated cysteine residues in the TA2F 1 ␥-subunit in buffer A and incubated for 10 min. Nonbound probe was then removed from the flow cell using buffer A. Observation of TA2F 1 ␥2c rotation was initiated after the infusion of either buffer A containing Ն0.2 mM sodium ATP/ lithium ATP␥S or buffer B (100 mM Tris (pH 8.0), 300 M MgCl 2 , 50 mM KCl) containing Ͻ0.2 mM ATP/sodium ATP/ lithium ATP␥S. In line with previous studies, 3 mM phosphoenolpyruvate and 2.5 units/ml pyruvate kinase were used to prevent accumulation of ADP in solution; where necessary, 0.1% LDAO was also included in the assay mixture. When using magnetic beads, the rotation of the duplex attached to the TA2F 1 ␥2c ␥-subunit was observed using bright field or phasecontrast microscopy and, when necessary, manipulated with home-built magnetic tweezers (33) controlled with custommade software (Celery, Library) for activation of "stalled" enzyme. Images were recorded on an inverted microscope (IX-71, Olympus, Japan) at 30 fps (FC300M, Takenaka Systems, Japan) for all characterization experiments with the exception of the six-step rotation observation that was recorded using a high speed complementary metal oxide semiconductor camera at 5000 fps (FASTCAM-1024PCI, Photron). When using gold nanoparticle, the rotation of the duplex attached to the TA2F 1 ␥2c ␥-subunit was observed using a modified inverted microscope (IX-71) using total internal reflection dark field microscopy (31) with a lens capable of total internal reflection microscopy (APON 60XOTIRF, Olympus) (44). Images were recorded with a high speed complementary metal oxide semiconductor camera at 10,000 fps (FASTCAM-1024PCI). Torque Measurements-Stepping torque (N) was determined from the rotation trajectories of magnetic duplex beads (ϳ0.2 m in diameter) attached to TA2F 1 ␥2c by using the FT (44). For continuous rotation of TA2F1, the time evolution of (t) can be described by the Langevin equation (Equations 1 and 2).
where ⌫ is the frictional drag coefficient, is a random force representing thermal noise, k B is the Boltzmann constant, and T is the room temperature (298 K). N is torque and is assumed to be a constant value as described previously (44). Observation of rotation of TA2F 1 ␥2c was performed as described above except that observations were recorded at 5000 fps, and 2 mM lithium ATP␥S was used (FASTCAM-1024PCI). Only enzymes exhibiting clear and even dwell times of three steps (120°) were used for analysis where the area on each trace used for analysis was the period moving between one 120°step and the next. Because the rotary motor proteins rotate in one direction, the following expression is derived from FT.
ln ͓P͑⌬͒/P͑Ϫ⌬͔͒ ϭ N⌬/K B T (Eq. 3) where (t) is the rotary angle of the bead, ⌬ ϭ (t ϩ ⌬t) Ϫ (t), and P(⌬) is the probability distribution of ⌬. For molecules continuously rotating for at least 4 s, P(⌬) was calculated for the case ⌬t ϭ 5 ms, and then ln[P(⌬)/P(Ϫ⌬)] versus ⌬/k B T was plotted. The slope of the resulting plot corresponds to the torque. The torque of each molecule was defined as the maximum value obtained from the FT analysis when using a 4-s moving window with windows starting at 1-ms intervals.
Author Contributions-D. G. G. M. conceived the study, performed the experiments, analyzed the data, and wrote the article. R. W. and H. U. provided critical insight and experimental/analytical knowledge necessary for the magnetic bead and gold rotation assay platforms, respectively. G. M. C. and H. N. provided critical insight into the broader interpretation of results in the view of both physiology and mechanical mechanism.