Rotation of artificial rotor axles in rotarymolecular motorsMihori Babaa,1, Kousuke Iwamotob,1, Ryota Iinob,c,d,e, Hiroshi Uenob, Mayu Harab, Atsuko Nakanishia,Jun-ichi Kishikawaa, Hiroyuki Nojib,2, and Ken Yokoyamaa,2

aDepartment of Molecular Biosciences, Kyoto Sangyo University, Kyoto 603-8555, Japan; bDepartment of Applied Chemistry, Graduate School ofEngineering, The University of Tokyo, Tokyo 113-8656, Japan; cDepartment of Functional Molecular Science, School of Physical Sciences, The GraduateUniversity for Advanced Studies , Kanagawa 240-0193, Japan; dOkazaki Institute for Integrative Bioscience, National Institutes of Natural Sciences, Okazaki444-8787, Japan; and eInstitute for Molecular Science, National Institutes of Natural Sciences, Okazaki 444-8787, Japan

Edited by Arieh Warshel, University of Southern California, Los Angeles, CA, and approved August 5, 2016 (received for review April 8, 2016)

F1- and V1-ATPase are rotary molecular motors that convert chemicalenergy released upon ATP hydrolysis into torque to rotate a centralrotor axle against the surrounding catalytic stator cylinder with highefficiency. How conformational change occurring in the stator is cou-pled to the rotary motion of the axle is the key unknown in themechanism of rotary motors. Here, we generated chimeric motorproteins by inserting an exogenous rod protein, FliJ, into the statorring of F1 or of V1 and tested the rotation properties of these chime-ric motors. Both motors showed unidirectional and continuous rota-tion, despite no obvious homology in amino acid sequence betweenFliJ and the intrinsic rotor subunit of F1 or V1. These results showedthat any residue-specific interactions between the stator and rotorare not a prerequisite for unidirectional rotation of both F1 and V1.The torque of chimeric motors estimated from viscous friction of therotation probe against medium revealed that whereas the F1-FliJchimera generates only 10% of WT F1, the V1-FliJ chimera generatestorque comparable to that of V1 with the native axle protein that isstructurally more similar to FliJ than the native rotor of F1. This sug-gests that the gross structural mismatch hinders smooth rotation ofFliJ accompanied with the stator ring of F1.

rotary molecular motor | protein design | ATPase | F1 | V-ATPase

Molecular motors are representatives of elegant proteincomplex systems that dynamically modulate finely tuned

intermolecular interactions to conduct unidirectional motion.Molecular motor systems are principally composed of two parts,a motor protein that undergoes a power-stroking conformationalchange fueled by nucleotide hydrolysis, or ion flux, and a coun-terpart protein that acts either as a track for linear motors or arotary shaft in rotary motors. Structural studies have revealedstriking shape complementarity of the motor–counter proteininterface, a feature that has also been observed in other proteincomplexes (1, 2). However, molecular motors modulate affinityto counterpart proteins upon catalysis, inducing unidirectionalmotion. Therefore, the current consensus view of the field is thatthe interfaces of molecular motor systems have sophisticateddesigns at an atomic level through molecular evolution.Engineering of motor proteins have been key to elucidating the

molecular design and working principle of motor proteins. However,although there have been a few successes, most attempts to redesignmotor proteins with gained function, such as enhanced force gen-eration or accelerated kinetic power, have proven unsuccessful (e.g.,see ref. 3) (4–6). Indeed, extensive engineering of the interface be-tween molecular motors and their counterpart proteins, for examplereplacing the counterpart protein with an exogenous protein, has notbeen reported. In this study, we constructed an artificial molecularmotor system from a stator ring of a naturally occurring rotarymotor, the F1-ATPase (F1) or V1-ATPase (V1), and FliJ chimeras.F1 and V1 are evolutionally closely related rotary motor pro-

teins that generate a large torque upon hydrolysis of ATP (7, 8).The catalytic core of FoF1-ATPases and VoV1-ATPases aremultisubunit membrane protein complexes that mediate energy

interexchange between the phosphorylate transfer potential ofATP and proton or sodium motive force across the membrane.F1 and V1 ATPases are composed of a heterohexameric ring withpseudo-threefold symmetry (the α3β3 in F1 and the A3B3 in V1)and a rod-shaped rotor protein (the γ subunit in F1 and the DFcomplex in V1) (8–10). The catalytic reaction centers for ATPhydrolysis/synthesis reside at the α−β or A–B interface in the F1and V1 enzymes, respectively, with the main catalytic residuesbeing harbored in either the β or A subunit (9–11). The rotaryshaft of F1, the γ subunit, is composed of two distinct parts: acoiled-coil of two alpha helices (an N-terminal short helix and aC-terminal long helix) and a protruding globular domain with anα/β fold. In V1, the DF complex forms structure with a shapesimilar to that of the γ subunit; the D- and F- subunits assembleinto a coiled-coil structure and a globular α/β fold, respectively(12–14). In both F1 and V1, the coiled-coil structure is almostcompletely embedded in the central cavity of the stator ring,whereas the protruding globular α/β fold domain has only a verysmall contact region with the stator subunits (10).The rotational dynamics of F1 and V1 also have several simi-

larities. Both motors have been shown to rotate continuously andunidirectionally in the counterclockwise direction when viewedfrom the membrane side (i.e., Fo or Vo side) (15, 16). The unitarystep of rotation is 120°, as expected from the pseudo-threefoldsymmetry of the stator rings. Both motors apparently generatetorque against viscous friction (V1; 27–35 pN·nm, F1; 40–50 pN·nm)(17–19). The distinctive feature of F1 and V1 that discriminates

Significance

F1/V1-ATPases are sophisticated molecular machines that convertthe motion of a stator cylinder driven by sequential ATP hy-drolysis to rotation of a central rotor protein. Here, we revealthe rotation of artificial rotor proteins composed of exogenousrod proteins that show no apparent sequence similarity with thenative axles. The estimated torque by the artificial rotor in thestator ring of V1 was almost identical to that by the native axleprotein. These results demonstrate that the principle of rota-tional motion by these molecular motors relies solely upon thecoarse-grained interaction between the rotor and stator. Thesefindings imply that the ancient F1 or V1 motor domain hasevolved from a poorly designedmotor protein more readily thaninitially assumed.

Author contributions: H.N. and K.Y. designed research; M.B., K.I., M.H., and A.N. per-formed research; R.I. contributed new reagents/analytic tools; M.B., K.I., and J.-i.K. ana-lyzed data; and M.B., H.U., H.N., and K.Y. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1M.B. and K.I. contributed equally to this work.2To whom correspondence may be addressed. Email: hnoji@appchem.t.u-tokyo.ac.jp oryokoken@cc.kyoto-su.ac.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1605640113/-/DCSupplemental.

11214–11219 | PNAS | October 4, 2016 | vol. 113 | no. 40 www.pnas.org/cgi/doi/10.1073/pnas.1605640113

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

021

them from other molecular motors is the reversibility of the che-momechanical coupling reaction. When forcibly rotated in the re-verse direction, the motors are capable of catalyzing the reversereaction (i.e., ATP synthesis from ADP and inorganic phosphate)(20, 21). This implies high efficiency of energy conversion fromchemical reaction to mechanical work, and vice versa, as experi-mentally verified by controlling the external torque and the chem-ical potential of ATP hydrolysis (22).To clarify the molecular basis for the efficient chemomechanical

coupling of F1, the stator–rotor interface has been extensively studiedby truncation of the γ helices or extensive amino acid substitution ofcontact loop of β (23–25). Surprisingly, all of the F1 constructs testedwith engineered stator–rotor interfaces showed unidirectional rota-tion, suggesting that none of the γ residues is essential for torquetransmission (24). These findings suggest that all rotor-constitutingresidues are equally responsible for torque transmission via individualspecific interactions with the stator ring. An alternative idea is thattorque is fundamentally transmitted via coarse-grained interaction.Coarse-grained simulation studies on F1 suggest that specific in-teraction at the residue level is not essential for the directed rotation(26, 27). Taken together, these observations imply that a rod-shapedprotein could be accommodated in the stator motor ring, could po-tentially rotate, and could generate estimated torque.Experimental verification of this hypothesis became realistic

when high-speed atomic force microscopy of F1 revealed that the βsubunits in the isolated α3β3 stator undergo sequential power-stroking conformational change without the interaction with γ (28).Motivated by these findings, we previously tested a chimeric rotarymotor reconstituted from the A3B3 stator ring of V1 and an exog-enous coiled-coil protein, FliJ, part of the bacterial flagellar type IIIexport complex (29). Despite no obvious sequence homology be-tween D of V1 and FliJ, a part of FliJ was successfully reconstitutedinto the A3B3 ring, enhancing the ATP hydrolysis activity of theA3B3 (30). However, unidirectional rotation of the reconstituted

FliJ was not observed using single-molecule rotation analysis. Here,we generated a V1-FliJ chimera (ChV1) in which the FliJ sequencewas fused to the C-terminal tip of the unstructured region of subunitD to enhance the stability of the complex. To forcibly hold thexenogeneic rotor in the stator ring, we assessed F1-FliJ chimeras(ChF1) by genetically fusing the β subunit of F1 to FliJ using aflexible linker as previously described (24). We found both chimericmotors function as active rotary motors. These findings stronglysuggest that any residue-specific interactions between the stator ringand the rotary shaft are not a prerequisite for the smooth rotationalmotion against an external frictional drag.

ResultsConstruction of V1-FliJ Chimeras.We constructed the chimera V1 byincorporating FliJ from Salmonella enterica into A3B3 stator ringfrom Thermus thermophilus (T.th). As shown in Fig. S1, there is noapparent sequence similarity between FliJ and subunit D of V1. Inthe previous study, we coexpressed FliJ with A3B3 to obtain thechimera complex ChV1

J147 (J147 denotes that the rotor is com-posed of the full length of FliJ with 147 aa). Although FliJ waspart of the chimera complex, ChV1

J147 did not show clear rotation,at least partially due to instability of the complex. Sequencealignment of subunit D’s from different species shows that theC-terminal 21 residues of the T.th V1 D subunit are not conservedin other species, suggesting that it would not have an active rolefor force transfer from the stator ring to subunit D. The structuralanalysis of T.th V1 suggests that the 21 C-terminal amino acids ofT.th D are located at the bottom of A3B3, the opposite side fromthe protruding portion of the rotor toward the Vo domain (13). Inthe present study, FliJ was fused to the C-terminal 21 residuesof the T.th D subunit and was coexpressed with A3B3 of T.th V1 toobtain chimera complex ChV1

J147/DC21 with the expectation thatthe additional sequence would enhance the complex’s stability. Inthe schematic image of the fusion rotor of ChV1

J147/DC21 (Fig. 1A),

Fig. 1. Construct of xenogeneic subunits in A3B3. (A) Structure models of V1 and the fusion rotor proteins from subunit D and FliJ. V1DΔ58–113 represents V1 with

subunit D, of which the loop domain (58–113) was truncated to leave only the coiled-coil structure. The truncated D and the A and the B subunits are representedin yellow, red, and green, respectively. Cyan in the fusion rotors represents the parts of FliJ. The helical portions derived from V1-D are represented in yellow, andthose from FliJ are represented in blue. Dashed lines of the fusion rotors in ChV1

D30/J86/D68 and ChV1J147/DC21 represent disordered C-terminal 12-aa residues (13).

(B) SDS/PAGE analysis of ChV1. Proteins are visualized by Coomassie blue staining (Upper) or by immunostaining using anti-FliJ to detect the xenogeneic rotors(Lower). The specificity of anti-FliJ was described in ref. 29.

Baba et al. PNAS | October 4, 2016 | vol. 113 | no. 40 | 11215

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

021

the additional 21 residues are represented as a short helix anddashed yellow line. For comparative purposes, we constructedother chimera V1’s with fusion rotor protein constructs of FliJ:ChV1

D30/J86/D68, ChV1D30/J86/D29, and ChV1

J147/D21. The number-ing in the superscripts of the mutant name denotes the length ofthe polypeptide chain of subunit D and FliJ, respectively, in thefusion rotary shaft. ChV1

D30/J86/D68 represents a chimera V1 withthe fusion rotor in which 86 residues of FliJ subunit were insertedbetween the N-terminal 30 residues and C-terminal 68 residues ofsubunit D (Fig. S2 and Fig. 1A).Although ChV1

D30/J86/D68 with a longer portion of subunit Dshowed expression and incorporation level comparable to that ofthe reference V1 mutant, V1

DΔ58–113 (Fig. 1B and Fig. S2), thefusion rotor of ChV1

D30/J86/D29 and ChV1J147/DC21 was not de-

tected in Coomassie staining and was barely detected when usingimmunostaining with anti-FliJ antibody (Fig. 1B). This suggeststhat most of the purified complex lost the rotor portion. Con-sidering that isolated stator and rotor proteins should not hinderthe rotation assay, the chimera motors were subjected to rotationexperiments after biotinylation.

Rotation of V1-FliJ Chimeras. We investigated whether these chi-meric V1 constructs rotate by attaching the A subunits onto aglass surface through the histidine residues at the N terminus ofthe A subunit, and by subsequently binding polystyrene duplexbeads of 290-nm diameter onto the rotors via the two cysteineresidues in 61 and 67. All mutants, including the ChV1

J147/DC21,rotated the 290-nm duplex beads in a counterclockwise directioncontinuously (Fig. 2A). The probability of finding a rotating beadwas variable depending on the chimera used (Table S1), and thenumber of beads attached on the coverslip was very low, despitethe use of equal protein concentration, suggesting that most ofmolecules of chimeric V1 do not contain the biotinylated rotor.This observation is consistent with the result of SDS/PAGEanalysis (Fig. 1B). At 4 mM ATP, which is saturating for the WTV1 hydrolytic activity, chimeric V1 motors made ∼200 or more

consecutive revolutions at approximately the same average rate,5.1 ± 1.4 revolutions per second (rps) for the ChV1

D30/J86/D68

(n = 7), 6.7 ± 1.3 rps for the ChV1D30/J86/D29 (n = 8), and 4.4 ±

1.0 rps for the ChV1J147/DC21 (n = 8), compared with 12.0 ± 1.6 rps

for WT V1 (n = 10) (A3B3DF; Fig. 2A and Table 1). ChimericV1 enzymes paused at angles separated by 120°, as shown inFig. S3. In estimating torque from the rotation trajectories, thepauses were eliminated from the time course, and 30 consecutive120° rotations were overlaid to obtain an average trajectory forthe 120° rotation (Fig. 2B). We then determined the rotationspeed of the duplex beads from the slope. Torque (N) was cal-culated from the angular velocity (ω) and frictional load (ξ) of therotating bead using the following equation: N = ωξ. The friction ofeach duplex bead was estimated from Eq. S1, taking into accountthe slight variation of the radius of the revolution of the outerbead of the duplex. The results are plotted in Fig. 2C and sum-marized in Table 1. The average torque values were 30.9 ± 2.3pN·nm for the WT V1 (A3B3DF), 19.6 ± 2.8 pN·nm for theV1

DΔ58–113, 20.6 ± 1.8 pN·nm for the ChV1D30/J86/D68, 21.0 ±

3.1 pN·nm for the ChV1D30/J86/D29, and 20.2 ± 3.7 pN·nm for the

ChV1J147/DC21. The torque of the ChV1

J147/DC21, lacking any ofthe coiled-coil structure of subunit D, is very similar to that ofV1

DΔ58–113 having the native axle (subunits DF). This clearly in-dicates that the conserved amino acid residues in subunit D arenot essential for torque generation in the V1.

Construct of F1-FliJ Chimeras. Unlike for V1, we have never beenable to accommodate FliJ into the central pore of the α3β3 ringof F1 by coexpression. To hold the coiled-coil structure of FliJ inthe α3β3 ring, FliJ was genetically fused with the β subunit via aflexible linker peptide that included a thrombin cleavage site(Figs. S4 and S5). In these constructs, only one of three FliJhelices occupies the central pore of the α3β3 ring, whereas theremaining two helices are outside the cavity. The FliJ helicesoutside the cavity might not affect the rotation properties, asdemonstrated by Chiwata et al. (24). The linker sequence that

Fig. 2. Rotation and torque estimation of the chimeric or truncated V1. (A) Time courses of 290-nm duplex rotation recorded at 250 or 1,000 frames per s.Overall time courses of WT V1 (black), truncated V1

DΔ58–113 (red), ChV1D30/J86/D68 (green), ChV1

D30/J86/D29 (purple), and ChV1J147/DC21 (sky blue). (B) Thin colored

curves show 30 consecutive 120° steps overlaid on top of each other, with the thick blue lines representing the average. The straight red lines indicate linearfit to the cyan curve between 0° to 120°. The slope of the red line, the angular velocity in radians−1, gives the torque N. (C) Torque values estimated from theinstantaneous rotary speed in consecutive 120 steps.

11216 | www.pnas.org/cgi/doi/10.1073/pnas.1605640113 Baba et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

021

connects the protrusion domainless γ with β was exactly the sameas the peptide sequence used in ref. 24. To investigate the role ofthe C-terminal sequence of γ in the force transmission, we pre-pared a series of FliJ-γ fusion constructs; 0, 9, 17, and 21 residuesfrom the C terminus of FliJ were substituted with the equivalentnumber of residues from the C terminus of γ [ChF1(0γ),ChF1(9γ), ChF1(17γ), and ChF1(21γ) as shown in Fig. 3A andFig. S5]. This series of ChF1 complexes were expressed as stablecomplexes in Escherichia coli and purified. SDS/PAGE analysisdemonstrates that each chimeric complex contained the FliJ-γ-βfusion with an α subunit. When treated with thrombin, the fusionprotein was digested into native β subunit and the isolated FliJ.In contrast with a previous report (24), the fusion protein wasalmost fully digested, suggesting that the FliJ helix was less firmlyheld in the α3β3 ring than the γ lacking the protruding globularα/β fold domain In contrast, a faint undigested fraction was ob-served for ChF1(9γ), ChF1(17γ), and ChF1(21γ) (Fig. 3B). It ispossible that the inserted γ sequence stabilizes the coiled-coilstructure in the cavity, and that the thrombin site is partiallyburied inside.

Rotation of F1-FliJ Chimeras. The rotation assay was conducted at2 mM ATP as described previously, the saturating level for WTF1. All constructs of ChF1, including the fully xenogeneic rotorconsisting of the coiled-coil structure of FliJ, showed continuousrotation (Fig. 4A). All rotation profiles were, on average, uni-directional in the counterclockwise direction as seen for the WTF1, although the rotation traces were marred by significantlymore experimental noise. Brownian, back-and-forth fluctuation

affected the rotation trajectories, suggesting the capability offorce transfer from the stator to the rotor was markedly impairedin chimera F1s (Fig. S6). Consistent with the large rotationalfluctuations, the rotation rate of the chimeras was significantlyslower than that for the WT: 1.0 ± 0.3 rps for ChF1(0γ), 0.8 ±0.3 rps for ChF1(9γ), 1.2 ± 0.4 rps for ChF1(17γ), and 1.5 ±0.3 rps for ChF1(21γ) (Fig. 4B and Table 1). The rotation ratesrecorded were consistent with those expected from the ATPhydrolysis rate determined by the ATP hydrolysis assay (∼6 persecond). Unlike the ChV1 constructs, the ChF1 constructs didnot show clear and regular intervening pauses during rotation.This prevented estimation of rotary torque using a 120° steppingprofile (an approach that gives a more precise estimation oftorque). Therefore, we estimated the torque of the ChF1 con-structs from a comparison with the average rotation rate of theWT F1 in the same condition using duplex 290-nm polystyrenebeads (13.9 ± 2.3 rps). Considering that we measured the torqueof the WT F1 around 40 pN·nm in previous work (19), we esti-mated the torque from the equation

NðpN×nmÞ= vchimera�vwild × 40,

where vchimera and vwild represent the average rotation rate. Thedetermined torque was 2.9 ± 1.0 pN·nm for ChF1(0γ), 2.3 ±0.9 pN·nm for ChF1(9γ), 3.5 ± 1.3 pN·nm for ChF1(17γ), and4.3 ± 1.1 pN·nm for ChF1(21γ) (Fig. 4C and Table 1).

DiscussionIn this study we demonstrated that the coiled-coil structure ofFliJ of the S. enterica flagella functions as a rotor in both theV1-A3B3 and F1-α3β3 hexamer rings. All of the tested chimeracomplexes showed continuous rotation in the presence of ATP inthe counterclockwise direction, as seen in the intact native mo-tors. Although FliJ has a coiled-coil structure similar to that ofthe native rotor subunits, the amino acid sequences show little orno similarity (Figs. S1 and S4). Most notably, the conservedpositively charged residues in both V1-D and F1-γ are not foundin FliJ. Therefore, FliJ is not able to form residue-specific in-teractions with the stator hexamer via the formation of saltbridges or hydrogen bonds. This finding demonstrates that thespecific interaction between stator and rotor is not a requirementfor unidirectional rotation, suggesting that the principal rotationmechanism is far more robust than previously thought. Thesimple explanation for this finding is that rotor rotation is in-duced via coarse-grained interaction between the stator ring and

Table 1. The rotation rate and the torque of chimeric enzymes

Chimeric enzymes Rotation rate, rps Torque, pN·nm

ChV1D30/J86/D68 5.1 ± 1.4 20.6 ± 1.8

ChV1D30/J86/D29 6.7 ± 1.3 21.0 ± 3.1

ChV1J147/DC21 4.4 ± 1.0 20.2 ± 3.7

V1DΔ58–113 6.0 ± 1.0 19.6 ± 2.8

WT V1 12.0 ± 1.6 30.9 ± 2.3ChF1(0γ) 1.0 ± 0.3 2.9 ± 1.0ChF1(9γ) 0.8 ± 0.3 2.3 ± 0.9ChF1(17γ) 1.2 ± 0.4 3.5 ± 1.3ChF1(21γ) 1.5 ± 0.3 4.3 ± 1.1WT F1 13.9 ± 2.3 40*

*This value was cited from ref. 19.

Fig. 3. Construct of xenogeneic subunits in α3β3. (A) Construct of F1-FliJ chimeras. Side views showing the central FliJ [blue; Protein Data Bank (PDB) ID code3AJW] and pair of α (green) and β subunits (red) of thermophilic Bacillus PS3 F1-ATPase (PDB ID code 4XD7). The helical portions of F1-γ and those of FliJ arecolored in orange and blue, respectively. The C-terminal unstructured portions of FliJ are represented by dashed lines. The linker portion containing thethrombin recognition site is represented by the purple line. (B) SDS/PAGE analysis of ChF1. Proteins are visualized by Coomassie blue staining.

Baba et al. PNAS | October 4, 2016 | vol. 113 | no. 40 | 11217

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

021

the rotor. In support of this proposition, a recent theoreticalstudy by Mukherijee and Warshel (27) emphasizes the role of theelectrostatic-mediated solvation energy arising from the polar/nonpolar residues for the torque transmission. In addition, atheoretical study by Koga and Takada suggested steric repulsionfrom the stator ring induces the rotor rotation, and the water-entropy effect was suggested to be a key factor to shape therotary potential of rotor shaft in the stator ring (26, 31). Furtheranalysis is required to clarify the fundamental factor for thetorque generation in the rotary motors.Another surprising result is that the chimeric ChV1 rotates a

large probe of 290-nm duplex beads against viscous friction at avelocity comparable to that of V1

DΔ58–113, which contains thenative coiled-coil structure of the rotor. The torque the chimericChV1 generates was estimated to be 20 pN·nm from the viscousfriction. This result suggests that even in the WT V1, specificamino acid interactions are not fundamental to the generation oftorque. Interestingly, the turnover rate of ChV1 is at least 18 s−1,estimated from an average rotation speed of ∼6 rps using 290-nmduplex beads (Fig. 2A and Table 1). This turnover rate indicates ahigher ATP hydrolysis activity than for the isolated A3B3 (∼1 s−1),implying that the coarse-grained interaction of the FliJ with A3B3enhances the allostery of A3B3 undergoing sequential catalysis.This, accompanied by the power-stroking conformational change,as seen in the isolated α3β3 ring (28), could result in the increasedATP hydrolysis activity observed with the chimeric V1 enzymes.Compared with ChV1 enzymes, rotational motion of beads by

ChF1 enzymes is significantly lower compared with the WT F1.The estimated torque obtained from the average rotation rate,∼1 rps, is only 10% of the WT measured under the same con-ditions (Fig. 4B). The mutant F1, lacking the protruding globularα/β domain, while retaining the coiled-coil structure of the γ, wasreported to rotate 290-nm beads at 5–6 rps and generate torqueof 50% of the WT. Thus, even compared with the protrusiondomainless F1 as in Chiwata et al. (24), ChF1 is <20% of the F1with intact coiled-coil structure. In addition, we observed thatthe estimated torque of ChF1 is significantly slower than that ofChV1. These findings suggest that the torque by ChF1 is only10% or 20% of that the WT or the protrusion domainless F1generates. This is attributable to the fact that FliJ is much morestructurally similar to subunit D of V1 than the γ subunit of F1(17, 29, 30) (Fig. S7). Both FliJ and subunit D have relativelystraight structures whereas the γ subunit is more curved. Taken

together, it is likely that the gross structural mismatch betweenFliJ and α3β3 hinders efficient rotation of FliJ. The gross struc-tural mismatch would also explain why the reconstitution of FliJdid not enhance ATP hydrolysis activity of the α3β3 ring.Interestingly, the estimated torque of ChF1, around 4 pN·nm,

is comparable to that for the Brownian ratchet mechanism. Asimple Brownian ratchet mechanism estimates apparent torque(force) against viscous friction as 2kBT/δ where δ is the angulardisplacement (32). A possible Brownian ratchet model for ChF1is that the rotor diffuses a flat rotary potential in the α3β3 statorring and after 120° diffusion the α3β3 stator ring changes itsconformation to block the backward diffusion of the rotor. Inthis model, δ is 2π/3 and the apparent torque is 4 pN·nm, con-sistent with the observation. Characterization of the rotationbehaviors of ChF1 under external force or with a rotation probeof different size would elucidate the actual working mechanismof ChF1.In the crystal structure of bacteria V1, the coiled-coil structure

of the D subunit is supported by the A3B3 cylinder at the top andbottom of the cavity (13, 14). The sequence of subunit D is wellconserved among species, especially in the helical region (Fig. S1).Nevertheless, the xenogeneic rotor of J147/DC21 in A3B3, whichdoes not contain these conserved residues, rotates with torque of∼20 pN·nm, roughly equal to that of V1

DΔ58–113. As shown in Fig.1B, the stability of the xenogenic J147/DC21 or D30/J86/D29 ro-tors in the cavity of A3B3 is much lower than that of the truncatedsubunit DΔ58–113 or the xenogenic D30/J86/D68 rotor. This clearlyindicates that the entire C-terminus region of D68 (156–223) is allthat is required for complex formation. Thus, we conclude that theconserved residues in the rotor protein contribute solely to effi-cient complex formation with the stator A3B3, and rarely to theenzymatic rotation function. In previous studies, we suggested thatthe rotor evolved from a rod-shaped protein like FliJ (30). Theresults of this study support the suggestion that an ancient poorlydesigned rotor could generate torque solely through a coarse-grained interaction.

Materials and MethodsProtein Preparation. A series of expression plasmids for ChV1 were constructedby partially substituting the subunit D sequence in the expression vector of His-tagged V1 (A(His-10/C28S/S232A/T235S/C255A/C508A)3B(C264S)3D(E48C/Q55C)) with the geneencoding FliJ as depicted in Fig. S2. For the truncated V1

DΔ58–113, the loopregion of D (58–113) was deleted. Other constructs contain xenogeneic rotors

Fig. 4. Rotation of F1-FliJ chimeras. (A) Time courses of rotation of WT F1 (black), ChF1(21γ) (red), ChF1(17γ) (green), ChF1(9γ) (purple), and ChF1(0γ) (sky blue).(B) Rotation rates of WT F1 (n = 8), ChF1(21γ) (n = 13), ChF1(17γ) (n = 12), ChF1(9γ) (n = 8), and ChF1(0γ) (n = 10) estimated from the fastest portionsof consecutive revolutions over 10 s. Filled circles indicate rotation rates for individual molecules. Columns show the means. The error bars indicate SDs.(C) Torque values estimated from the averaged rotation rate of WT F1 and the series of ChF1 shown in B considering the torque of WT F1 is 40 pN·nm. Columnsshow the means. The error bars indicate SDs.

11218 | www.pnas.org/cgi/doi/10.1073/pnas.1605640113 Baba et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

021

that included parts of FliJ and D. For ChV1J147/DC21, the C-terminal region of the

D subunit (DC21) was fused to the C terminus of FliJ to stabilize the rotor–stator interaction. The chimeric V1 constructs were expressed in E. coli andpurified with Ni2+-affinity chromatography (Qiagen) followed by ion exchangewith a UNO-Q column (GE Healthcare). The purified His-tagged enzymes werebiotinylated at two cysteines (FliJ/Ile61Cys and Ile67Cys, indicated in Fig. S1 bythe red arrows) using 6-{N-[2- (N-maleimide) ethyl]-N-piperazinylamide}hexyl-D-biotinamide (Dojindo). The endogenous cysteine residue (FliJ/32Cys) wasmutated to threonine. These biotinylated enzymes were subjected torotation analysis.

The expression plasmids for ChF1 were constructed by partially or fullysubstituting the γ subunit sequence with the gene encoding FliJ in the ex-pression vector of His-tagged F1, α(His-6 /C193S)3β(His-10)γ (S107C/I210C) as describedin Fig. S5. The inserted FliJ sequence contained cysteine substitutions at γ52and γ59. A flexible linker sequence was inserted at the C-terminal end of FliJto fuse FliJ to β and hold the xenogeneic rotor protein in the α3β3 ring,preventing dissociation. The linker sequence was the same one used pre-viously by Kinosita and coworkers (24) and includes a thrombin cleavage site(LVPRGS) for removal of the linker sequence for SDS/PAGE analysis. A seriesof chimeras of F1 with the FliJ-γ fusion rotor were prepared by substitutingthe C-terminal helix of FliJ with that of γ, to produce ChF1(xγ), where x in-dicates the length of substitution; x = 0, 9, 17, or 21. These chimericcomplexes were expressed in E. coli and purified with Ni2+-affinity chro-matography (Qiagen) followed by size-exclusion chromatography withSuperdex 200 10/300GL (GE Healthcare) as previously reported (33). Thepurified enzymes were biotinylated at the two cysteines in the FliJ sequenceand subjected to rotation analysis.

Rotation Assays. Rotation assays were conducted principally as described (refs.18 and 30 for ChV1 and ref. 32 for ChF1). The basic procedure for the assaywas as follows. Biotinylated chimeric enzymes (1∼100 nM) were infused intoa flow chamber and immobilized via the His tag on the glass surfacesfunctionalized with Ni2+-NTA. After incubation for 15 min at 25 °C, unboundenzymes were washed out with basal assay buffer [50 mM Hepes-KOH(pH 8.0), 100 mM KCl, 2 mM MgCl2, and 5 mg/mL BSA for ChV1 and 50 mM

Mops-KOH (pH 7.0), 50 mM KCl, and 10 mg/mL BSA for ChF1]. Basal buffercontaining 290-nm polystyrene beads for ChV1 (Thermo Science) or 290-nmpolystyrene beads for ChF1 (Seradyn) was infused and the chambers in-cubated for 3–5 min, followed by washing with basal buffer to removeunbound beads. Rotation was initiated by infusing assay buffer (4 mM Mg-ATP, 2.5 mM phosphoenol pyruvate, and 0.5 μg·mL−1 pyruvate kinase inbasal buffer without BSA for ChV1, 2 mM ATP with 2 mM excess MgCl2 inbasal buffer for ChF1) into the chambers. Rotation of the bead duplexes wasobserved by bright-field microscope at 25 °C at 250–1,000 frames per s aspreviously described (18, 23, 34).

The torque (N) against viscous drag was estimated from viscous frictionusing the viscous friction coefficient (ξ) and angular velocity (ω); N = ωξ (18).Because ChV1 showed distinct 120° steps with rotational pauses, the angularvelocity for ChV1 was determined from the slope of the time trajectories ofrotation. ChF1 did not show obvious steps, and the torque against viscousdrag was determined from the comparison with the averaged rotation rateof the WT F1 under the same conditions.

Biochemical Assays. Protein concentrations of the chimeric enzymes weredetermined from UV absorption at 280 nm based on the absorption co-efficient calibrated from the amino acid sequence. ATP hydrolysis activity wasmeasured at 25 °C with an enzyme-coupled ATP regenerating system asdescribed previously (35).

ACKNOWLEDGMENTS. We thank Dr. Tohru Minamino and Dr. KatsumiImada for gifting the antibody of FliJ, Dr. Furuike for giving the technical advicefor single-molecular observation, and Dr. Takahiro Sagawa for careful discus-sion on thermodynamics of molecular motors. We also thank Dr. BernadetteByrne and Dr. Duncan McMilan for critical reading of the manuscript. This workwas supported by Ministry of Education, Science, Sports and Culture of JapanGrants-in-Aid 24370059 and 26650039 (to K.Y.) and 25251016 (to H.N.) andJapan Society for the Promotion of Science KAKENHI Grants JP15H04366,JP16H00789, and JP16H00858 (to R.I.). The funders had no role in the studydesign, data collection and analysis, decision to publish, or preparation ofthe manuscript.

1. Vakser IA, Matar OG, Lam CF (1999) A systematic study of low-resolution recognitionin protein–protein complexes. Proc Natl Acad Sci USA 96(15):8477–8482.

2. Zhang Q, Sanner M, Olson AJ (2009) Shape complementarity of protein-proteincomplexes at multiple resolutions. Proteins 75(2):453–467.

3. Yukawa A, Iino R, Watanabe R, Hayashi S, Noji H (2015) Key chemical factors of ar-ginine finger catalysis of F1-ATPase clarified by an unnatural amino acid mutation.Biochemistry 54(2):472–480.

4. Schindler TD, Chen L, Lebel P, Nakamura M, Bryant Z (2014) Engineering myosins forlong-range transport on actin filaments. Nat Nanotechnol 9(1):33–38.

5. Lebel P, Basu A, Oberstrass FC, Tretter EM, Bryant Z (2014) Gold rotor bead trackingfor high-speed measurements of DNA twist, torque and extension. Nat Methods11(4):456–462.

6. Nakamura M, et al. (2014) Remote control of myosin and kinesin motors using light-activated gearshifting. Nat Nanotechnol 9(9):693–697.

7. Forgac M (2007) Vacuolar ATPases: Rotary proton pumps in physiology and patho-physiology. Nat Rev Mol Cell Biol 8(11):917–929.

8. Yokoyama K, Imamura H (2005) Rotation, structure, and classification of prokaryoticV-ATPase. J Bioenerg Biomembr 37(6):405–410.

9. Maher MJ, et al. (2009) Crystal structure of A3B3 complex of V-ATPase from Thermusthermophilus. EMBO J 28(23):3771–3779.

10. Abrahams JP, Leslie AG, Lutter R, Walker JE (1994) Structure at 2.8 A resolution of F1-ATPasefrom bovine heart mitochondria. Nature 370(6491):621–628.

11. Nadanaciva S, Weber J, Wilke-Mounts S, Senior AE (1999) Importance of F1-ATPaseresidue alpha-Arg-376 for catalytic transition state stabilization. Biochemistry 38(47):15493–15499.

12. Makyio H, et al. (2005) Structure of a central stalk subunit F of prokaryotic V-typeATPase/synthase from Thermus thermophilus. EMBO J 24(22):3974–3983.

13. Numoto N, Hasegawa Y, Takeda K, Miki K (2009) Inter-subunit interaction and qua-ternary rearrangement defined by the central stalk of prokaryotic V1-ATPase. EMBORep 10(11):1228–1234.

14. Arai S, et al. (2013) Rotation mechanism of Enterococcus hirae V1-ATPase based onasymmetric crystal structures. Nature 493(7434):703–707.

15. Noji H, Yasuda R, Yoshida M, Kinosita K, Jr (1997) Direct observation of the rotationof F1-ATPase. Nature 386(6622):299–302.

16. Imamura H, et al. (2003) Evidence for rotation of V1-ATPase. Proc Natl Acad Sci USA100(5):2312–2315.

17. Kishikawa J, et al. (2014) F-subunit reinforces torque generation in V-ATPase. EurBiophys J 43(8-9):415–422.

18. Imamura H, et al. (2005) Rotation scheme of V1-motor is different from that of F1-motor.Proc Natl Acad Sci USA 102(50):17929–17933.

19. Yasuda R, Noji H, Kinosita K, Jr, Yoshida M (1998) F1-ATPase is a highly efficientmolecular motor that rotates with discrete 120 degree steps. Cell 93(7):1117–1124.

20. Itoh H, et al. (2004) Mechanically driven ATP synthesis by F1-ATPase. Nature427(6973):465–468.

21. Rondelez Y, et al. (2005) Highly coupled ATP synthesis by F1-ATPase single molecules.Nature 433(7027):773–777.

22. Toyabe S, Watanabe-Nakayama T, Okamoto T, Kudo S, Muneyuki E (2011) Thermo-dynamic efficiency and mechanochemical coupling of F1-ATPase. Proc Natl Acad SciUSA 108(44):17951–17956.

23. Furuike S, et al. (2008) Axle-less F1-ATPase rotates in the correct direction. Science319(5865):955–958.

24. Chiwata R, et al. (2014) None of the rotor residues of F1-ATPase are essential fortorque generation. Biophys J 106(10):2166–2174.

25. Tanigawara M, et al. (2012) Role of the DELSEED loop in torque transmission of F1-ATPase. Biophys J 103(5):970–978.

26. Koga N, Takada S (2006) Folding-based molecular simulations reveal mechanisms ofthe rotary motor F1-ATPase. Proc Natl Acad Sci USA 103(14):5367–5372.

27. Mukherijee S, Warshel A (2014) Dissecting the role of the g-subunits in the rotary-chemical coupling and torque generation of F1-ATPase. Proc Natl Acad Sci USA 112(9):2746–2751.

28. Uchihashi T, Iino R, Ando T, Noji H (2011) High-speed atomic force microscopy revealsrotary catalysis of rotorless F₁-ATPase. Science 333(6043):755–758.

29. Ibuki T, et al. (2011) Common architecture of the flagellar type III protein exportapparatus and F- and V-type ATPases. Nat Struct Mol Biol 18(3):277–282.

30. Kishikawa J, et al. (2013) Common evolutionary origin for the rotor domain of rotaryATPases and flagellar protein export apparatus. PLoS One 8(5):e64695.

31. Ito Y, Yoshidome T, Matubayasi N, Kinoshita M, Ikeguchi M (2013) Molecular dy-namics simulations of yeast F1-ATPase before and after 16° rotation of the γ subunit.J Phys Chem B 117(12):3298–3307.

32. Peskin CS, Odell GM, Oster GF (1993) Cellular motions and thermal fluctuations: TheBrownian ratchet. Biophys J 65(1):316–324.

33. Bald D, Noji H, Yoshida M, Hirono-Hara Y, Hisabori T (2001) Redox regulation of therotation of F(1)-ATP synthase. J Biol Chem 276(43):39505–39507.

34. Watanabe R, et al. (2011) Mechanical modulation of catalytic power on F1-ATPase.Nat Chem Biol 8(1):86–92.

35. Yokoyama K, et al. (1998) V-ATPase of Thermus thermophilus is inactivated duringATP hydrolysis but can synthesize ATP. J Biol Chem 273(32):20504–20510.

Baba et al. PNAS | October 4, 2016 | vol. 113 | no. 40 | 11219

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

8, 2

021