Functional and Structural Characterization of F1-ATPase with common ancestral core domains in stator ring

Extant F 1-ATPases exhibit diverse rotational stepping behaviors —3-, 6-, or 9 -step cycles—yet the evolutionary origin of these patterns remains unclear. Here, we used ancestral sequence reconstruction to infer the catalytic β and non-catalytic α subunits of a putative ancestral F1-ATPase. We then fused their functionally critical domains into the thermostable F 1 from Bacillus PS3, yielding a stable chimeric enzyme. Cryo-EM revealed two distinct conformational states—binding and catalytic dwell states—separated by a ~34° rotation of the γ subunit, suggesting a fundamental six -step mechanism akin to that of extant 6-stepping F1-ATPases. Single-molecule rotation assays with ATP and the slowly hydrolyzed ATP analog ATPγS demonstrated that the chimeric motor is intrinsically a 6 -stepper, pausing at binding and catalytic dwell positions separated by 32.1°, although the binding dwell is significantly prolonged by an unknown mechanism. These findings indicate that F 1-ATPase was originally a 6-stepper and diversified into 3-, 6- and 9-step forms in evolutional adaptation. Based on these results, we discuss plausible features of the entire F oF1 complex, along with potential physiological contexts in last universal common ancestor and related lineages. 180 words (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint

F-type ATP synthase catalyzes the terminal reaction of oxidative phosphorylation —namely, the synthesis of ATP from ADP and inorganic phosphate (Pi) driven by proton translocation down a proton motive force ( pmf) across biomembranes. This enzyme is among the most ubiquitous in nature, found in the plasma membranes of prokaryotic cells, the thylakoid membranes of chloroplasts, and the inner membranes of mitochondria. Recent comparative genomic analyses have revealed that cells of the last universal common ancestor (LUCA) already possessed F-type or relevant type ATP synthase[1][2]. These findings provide significant insights into the metabolic capabilities and ecological contexts of LUCA and its descendants, including the last bacterial common ancestor (LBCA) and the last archaeal common ancestor (LACA)[1][2][3]. F-type ATP synthase is composed of two rotary motors termed Fo and F1. Fo is the membrane- embedded portion, where the c-oligomer ring ( c-ring) rotates against the ab2 stator complex during proton translocation across the membrane. F 1 is the membrane -protruding portion and rotates the inner rotor complex against the catalytic stator ring during ATP hydrolysis. Fo and F1 are connected by the rotor complex and the peripheral stalk, so as to enable the interconversion of pmf and the free energy of ATP hydrolysis. Under ATP -synthesizing conditions—when the pmf is sufficient and the rotational torque of Fo exceeds that of F1—Fo drives the reverse rotation of F 1 (opposite to t he ATP -hydrolyzing direction), thereby inducing ATP synthesis on the catalytic stator ring[3][4]. Conversely, when the torque of F1 exceeds that of Fo and in the absence of regulatory elements , F 1 rotates the c-ring in F o, forcing F o to pump protons in the reverse direction and thus generate pmf. In this way, FoF1 interconverts the pmf and the chemical potential of ATP hydrolysis via mechanical rotation The minimal subunit composition of F 1, acting as the ATP -driven motor, is α 3β3γ1, with the γ subunit embedded inside a hetero-hexameric stator ring consisting of alternating α and β subunits. The α3β3 stator ring has three catalytic sites, each located at α–β interface. Owing to the structural asymmetry of α and β, there are two distinct types of interfaces in the α 3β3 ring. One-type—the α–β interface—contains the catalytic site, whereas the other type—the β–α interface—also binds ATP but does not hydrolyze it. Because most catalytically important residues reside on the β subunit, it is termed the “catalytic subunit.” Conversely, the “non-catalytic” ATP-binding site at (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint the β–α interface is formed mainly by the α subunit residues, so the α subunit is termed the “non- catalytic subunit.” During catalysis, the three β subunits undergo conformational changes in a coordinated manner, resulting in the unidirectional rotation of the γ subunit[5]. The rotary catalysis of F1 has been extensively studied via single-molecule rotation assays of F1- ATPase derived from the thermophilic bacterium Bacillus PS3 (hereafter TF1) due to stability and the ease of handling[4]. In consistent with the pseudo-threefold symmetry of F1, the basic step size of rotation is 120°. This 120° step was subsequently resolved into two discrete substeps of 80° and 40°, each initiated after ATP binding and hydrolysis, respectively. Accordingly, the dwell states prior to the 80° and 40° substeps ar e referred to as the “binding dwell” and the “catalytic dwell,” respectively[6]. Later studies showed that F 1 releases Pi from a β subunit in the catalytic dwell, but distinct from the one engaged in catalysis. It was also reported that F1 pauses at binding dwell angle during temperature-sensitive reaction intermediate[7]. Note that the detailed statistical analysis of the catalytic dwell revealed that ATP hydrolysis induces rotation during the dwell phase[8]. However, the angular displacement upon hydrolysis during dwell phase is subtle and within the angle distribution of the catalytic dwell (typically ±10 − 20° ). Thus, TF1 is principally a '6-stepper' motor, making three binding and three catalytic dwells per turn. Similar reaction schemes with six steps per turn have been reported for F1-ATPases derived from E. coli (EF1) and yeast mitochondria (yMF 1)[9][10]. However, recent studies have revealed variations in the number of substeps. F1-ATPase from human and bovine mitochondrial F1 (hMF1, bMF1) exhibits an additional pause—referred to as the “short dwell”—alongside the binding and catalytic dwells, resulting in nine steps per turn [11]. In contrast, F 1-ATPase from Paracoccus denitrificans (PdF1) shows only three steps per turn under all tested conditions, despite its close evolutionary relationship to mitochondria ( Paracoccus is an α-proteobacterium from which the mitochondrial ancestor is thought to have diverged) [12]. Thus, while most F 1-ATPases pause six times per turn (“6 -steppers”), mammalian mitochondrial F 1 is a “9 -stepper,” and PdF 1 is a “3 - stepper.” Notably, the number of steps per turn correlates with the number of the c-subunits in the c-ring: 6-steppers typically pair Fo with a c10-ring, 9-steppers have a c8-ring, and 3-steppers have a c12-ring. This trend suggests that F1 with more steps per turn has a c-ring containing fewer c subunits, implying certain mechanistic or physiological constraints on the total number of steps in the entire FoF1 complex[13]. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint To investigate which subunit determines the stepping pattern of F 1, we previously constructed various hybrid F 1-ATPases whose subunits originated from different species —TF1, bMF1, and PdF1 [14]. Analysis of these hybrids showed that rotational speed principally depends on the origin of the β subunit, as expected. However, we did not identify a single comprehensive rule governing the number of steps for all hybrids. We did find one conditional ru le: whenever a hybrid F 1 contains a subunit from PdF 1, it consistently exhibits three-step rotation per turn, just like PdF 1, regardless of the origins of the other subunits. Hence, although fundamental features —such as the 120° step coupled to a single ATP hydrolysis turnover and the rotation direction—are broadly conserved across species, the number and size of substeps vary. This naturally raises the question: What was the original stepping pattern of F 1-ATPase? In other words, how did the common ancestral F1-ATPase rotate? Evolutionarily, F 1-ATPase is closely related to V 1-ATPase[15], the catalytic portion of V -type ATPases that function as ATP -driven proton pumps in the vacuoles of mammalian cells or as ATP synthases in the plasma membranes of archaea (Fig. 1A). V1-ATPase is also an ATP-driven rotary motor, rotating counterclockwise (when viewed from the membrane side) in the same direction as F1 [16]. V1 consists of a rotor complex and a hetero -hexameric stator ring composed of A and B subunits, corresponding to the β and the α subunits of F 1, respectively. Like F 1, a recent study of V 1-ATPase from Enterococcus hirae (EhV1) resolved these 120° steps into 40° and 80° substeps [17], whereas such substeps have not been observed for Thermus thermophilus V1 (TtV1)[18]. Thus, although V 1-ATPase shares some basic characteristics with F 1-ATPase, its stepping behavior does not directly clarify how the ancestral F1-ATPase might have operated. One experimentally feasible way to explore the functionality of ancestral enzymes is ancestral sequence reconstruction (ASR)[19][20][21]. ASR uses multiple sequence alignment (MSA) of extant species to infer the most probable amino acid changes along a phylogenetic tree, thereby predicting ancestral protein sequences. A variety of ancestral proteins have been successfully expressed and characterized. The feasibility of sequence reconstruction by ASR methods depends on the degree of sequence conservation in the extant enzymes. Although the rotor subunits of rotary ATPases are less conserved, the non -catalytic and catalytic subunits of F 1 and V 1 show high sequence similarity, respectively, and even resemble each other. Molecular phylogenetic analyses suggest that the ancestral motors of F1-ATPase and V1-ATPase diverged from a common ancestral rotary motor with a hetero-hexameric stator ring comprising the ancestral non-catalytic (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint (NCF/V) and catalytic ( CF/V) subunits (Fig. 1 B). It has also been proposed that this common ancestral rotary ATPase diverged from the homo -hexameric ATPase from which Type III secretion system (T3SS) ATPase originated[15][22] (Fig. 1A-B). Divergence time estimates indicate that the ancestral of either or both F 1- and V 1-ATPases—had already emerged as the catalytic portion of ATP synthase in LUCA[1][2]. Thus, previous studies have established an evolutionary phylogeny of rotary ATPases that stems from the pre -LUCA era [2], laying a foundation for the ancestral sequence inference of these ancient rotary ATPases. However, ancestral sequences reconstruction of multi -subunit complex enzymes requires extremely precise estimation of the amino acids forming the subunit –subunit interfaces, and consequently, there have been only a few successful examples [23]. In particular, there have been no reports of ancestral sequence research on molecular machines like F1-ATPase, which shows large conformational transitions. In the present study, we reconstructed the amino acid sequences of the catalytic and non-catalytic subunits of the common ancestral F 1-ATPase and prepared these subunits for experimental investigation to understand how the ancestral proteins functioned. We also attempted to test the functionality of the reconstituted sequences. To address the technical challenge regarding the complex instability often observed in ancestral enzyme s with multi -subunit composition, we incorporated the functionally core parts from the ancestral sequences into extant thermostable F1, TF1; the non-catalytic and catalytic subunits were designed as chimeras of the ancestral and extant proteins. Specifically, the functionally critical regions —namely, the nucleotide-binding domain and the C-terminal helical domain—were derived from the ancestral sequences, whereas the N - terminal domains were taken from TF1 to serve as a structural scaffold. We then co -expressed these chimeric subunits with the γ subunit from TF 1, thereby forming a functional and stable F 1 complex, of which core parts are derived from ancestral sequence. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint

Ancestral Sequence Reconstruction Previous studies have elucidated the phylogenetic branching of subunits forming hexameric rings in T3SS ATPase, V1-ATPase, and F1-ATPase[15][22] . In this study, we focused on the five families of subunits represented in the phylogenetic tree (Fig.1B): FliI subunit comprising of the homo-hexameric ring of T3SS ATPase, the non-catalytic subunit and the catalytic subunits of the hetero-hexameric rings of V1- and F1-ATPase (the B and the A subunits for V1-ATPase, and the α and the β subunits for F1-A TPase). Representative sequence data were downloaded from National Center for Biotechnology Information (NCBI) database and used as query sequences. These query sequences were subjected to BLASTP searches[24] to collect sequences with a similarity above a certain threshold, which were then compiled into datasets for each subunit. Multiple sequence alignment was performed for each subunit . S equences with large deletions, insertions, or redundancies were removed. As a result, we curated sequence datasets with conserved regions aligned for each subunit. The final numbers of sequences used for phylogenetic tree construction were 94 for the T3SS ATPase FliI subunit, 128 for the A subunit and 100 for the B subunit of V1-ATPase, and 142 for the α subunit and 153 for the β subunit of F1-ATPase. Using these datasets, phylogenetic trees were inferred with IQ-TREE[25], a fast and widely used maximum-likelihood-based software that incorporates advanced model selection. The phylogenetic tree constructed was compared with those from previous research, focusing on the branching positions of phyla that were consistently present across the trees. The analysis revealed a high degree of concordance in the branching positions (Supplementary Figure 2). The phylogenetic tree generated by IQ-TREE[25] served as a scaffold for ancestral sequence reconstruction using the codeml program in the PAML package [26], a tool for maximum - likelihood-based evolutionary analysis , and GASP [27], a probabilistic method for ancestral sequence inference. Amino acid residues at gap positions in GASP-reconstructed sequences were removed to obtain the final ancestral sequences. The reliability of the reconstructed ancestral sequences was evaluated by the probability values: approximately 0.9 for the ancest ral non- catalytic subunits and catalytic subunits of F1-ATPase (αanc and βanc) and those for V1-ATPases (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint (Banc and Aanc), (Supplementary Figure 3), ensuring the high reliability of the reconstruc ted sequences. The common ancestral non -catalytic subunit and catalytic subunit of F1- and V 1- ATPase (NCF/V and CF/V) showed the probability values around 0.65 (Supplementary Figure 3). The common ancestor protein of NCF/V and CF/V, CancR, which is supposed to form homo - hexameric ATPase also shows a similar probability value, 0.67 (Supplementary Figure 3), suggesting the sequence reconstruction of the more upstream common ancestral proteins is less reliable. To ensure the validity of the ancestral sequence inference, we also reconstructed a phylogenetic tree using RAxML[28], a software optimized for rapid and efficient phylogenic tree searches, on which the ancestral sequence estimations was conducted. The reconstructed ancestral sequence confirms high consistency with the sequences reconstructed from the IQ-TREE-based phylogenetic trees, particularly the sequence region encoding the structurally interior parts of the subunits (Supplementary Figure 4). The sequence regions encoding subunit-subunit interfaces formed in the F1 complex also shows high consistency. The differences between sequences reconstructed on IQ -TREE- or RAxML -based phylogenic trees are mainly found in the structurally exterior parts and the N -terminal domains that are thought not to be crucial for functionality. Based on these results, we concluded that the ancestral sequences inferred here represent the most plausible reconstructions achievable with contemporary computational methods. In followings, we focus the sequences of αanc and βanc derived from the IQ-TREE-based phylogenic tree, while the reconstructed sequences for NCF/V, CF/V and CancR were also analyzed (Supplementary Figure 5-8). (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint Fig.1. Schematic phylogenetic trees of rotary ATPases. A. Phylogenetic tree of rotary A TPases. The diagram highlights key species whose rotation schemes have been elucidated through single-molecule rotation assays. Species belonging to the same phylum are represented in the same color. B. Conceptual phylogenetic tree of the subunits forming the hexameric ring of rotary A TPases. The node labeled Canc represents the common ancestor of the α, β, A, and B subunits. αanc and βanc indicate the ancestral forms of the F-type ATPase α and β subunits, respectively, with similar annotations for other subunits. In this study, αanc and βanc are collectively considered as the ancestral form of the F -type ATPase. The sequences of the ancestral non-catalytic and catalytic subunits, αanc and βanc were compared with ones of the extant F1-ATPases, TF1, PdF1, and bMF1. The identities between the ancestral and the extant F1-ATPases were approximately 70%, while sequence identities among extant F 1-ATPase ranged from 75% to 85% (Fig. 2 A). As predicted from the high sequence conservation, we found the perfect conservation of the catalytically crucial sequences among the ancestral sequence and the extant sequences : regions including arginine finger , catalytic glutamate, and Walker motif A (phosphate binding loop, p-loop) (Fig. 2B). Thus, the sequence comparison between the sequences of the ancestral subunits and the extant ones supports the plausibility of the reconstructed sequences of αanc and βanc, suggesting the functionality of the se subunits. Note that the catalytic residues are also fully conserved among NCF/V, CF/V and CancR (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint (Supplementary Figure 7). Fig.2. Sequence comparison between the common ancestor of F-type ATPases and extant species. A. Comparison of sequence identity between extant species and ancestral forms. The catalytic subunit β exhibits higher sequence identity across its full length compared to the non -catalytic subunit α. Furthermore, the sequence identity between ancestral forms and extant species shows no sub stantial difference compared to the sequence identity among extant species. B. Comparison of ATPase motifs. The conservation of key amino acid residues involved in ATP hydrolysis, including the arginine finger, catalytic glutamate, and p-loop, was analyzed. αanc and βanc represent the ancestral sequences of the F- type A TPase α and β subunits reconstructed in this study. The other three sequences correspond to the extant species TF 1, PdF 1, and bMF1. The ATPase motifs are well conserved across the compared sequences, suggesting that ancestral forms utilize the same key amino acid residues as extant species for ATP hydrolysis. Biochemical analysis To test the functionality of the ancestral subunits, we attempted to express F1 composed of the ancestral subunits. ASR of enzymes with multis-subunits is challenging, due to the difficulty of highly precise inference of residues forming subunit-subunit interface. To reinforce the (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint structural integrity, TF 1 was used as scaffold to accommodate the functionally crucial core domains of αanc and βanc; nucleotide-binding domain (NDB) and C-terminal domain (CTD) of αanc and βanc were genetically fused with the N-terminal domain (NTD) of TF1 that forms the stable hexameric scaffold (residues α1-94 and β1-78 in TF1). The chimeric α and β subunits were co- expressed with the γ subunit of TF 1 (Fig.3A). The resultant hybrid F1 with the ancestral core domains, referred hereafter to as F1anc_core for simplicity, was successfully expressed as a stable complex as shown in the elution profile of size -exclusion chromatography ; it showed the distinctive peak of F 1 complexes beside peaks presumably corresponding to partially formed complexes and monomers (Fig.3B). The ATPase activity of F 1anc_core was assessed using an ATP regeneration system. F1anc_core exhibited ATPase activity approximately one-tenth that of TF1 (Table 1). Despite the low activity level, ATP hydrolysis was clearly evident. Notably, the ATPase activity after LDAO addition remained unchanged. LDAO is known to relieve ADP inhibition, a state in which ADP tightly remains bound on the catalytic site to prevent catalysis and rotation [29]. These findings suggest that the low ATPase activity of F1anc_core reflects intrinsically low activity. Fig. 3. The design and purification of F1anc_core A. The design of F1anc_core. The hexameric ring consists of α and β subunits, where the N-terminal β- barrel domains (NTD), forming the structural foundation, are derived from the extant species TF 1, while the remaining regions C-terminal domain (CTD) and nucleotide-binding domain (NBD) utilize the reconstructed ancestral sequences of the F -type A TPase (αanc, βanc). The γ subunit forming the (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint central stalk is based on the sequence of TF 1. B. Size-exclusion HPLC chromatogram after Ni-NTA purification of F1anc_core. The peak positions were estimated using a calibration curve generated with molecular weight markers. The peak corresponding to the hexameric complex aligns with the peak position observed during the purification of TF 1, which has a comparable molecular weight. Peaks corresponding to partial complex and monomeric forms were also detected; however, the peak for the fully assembled hexameric complex is distinctly observed. Table.1. ATPase activity values. Ancestral F1-A TPase TF1 w/o LDAO w/ LDAO w/o LDAO w/ LDAO 6.7 ± 0.1 s-1 7.5 ± 0.3 s-1 78 ± 3 s-1 140 ± 1 s-1 The activity measured within the first 300 seconds of the experiment. Values are mean ± SD (n = 3). Structural analysis The structure of F 1anc_core was determined by cryogenic electron microscopy (Cryo -EM), following our previous studies[30]. The purified sample was applied to EM grids at 22°C, vitrified in liquid ethane, and imaged using single-particle analysis (SPA) at 300 kV . When the grids were prepared in the presence of AMP-PNP—a nonhydrolyzable analog of ATP—only about 5% of the molecules were observed to form intact F1 complex, suggesting that the sample readily dissociates under those conditions (Supplementary Table 2). To overcome this, we instead prepared EM grids without any added nucleotides. Under this condition, five distinct Cryo-EM maps were obtained: two fully assembled F1 complexes with different γ-subunit rotational angles, a hexameric ring lacking the central shaft, and we also partially assembled complexes (a tetramer with the central shaft and a tetramer without it; see Supplementary Figure 9-11. The resolutions of the cryo-EM maps were determined to be 2.5 Å and 2. 5 Å for the two fully assembled complexes, 2. 8 Å for the hexameric ring without the shaft, 2. 5 Å for the tetramer with the shaft, and 2.7 Å for the tetramer without the shaft, using the "gold standard" method. A comparison of the two fully assembled complexes with previously published c ryo-EM structures of TF1 showed a good overall match, allowing us to identify one as the catalytic dwell (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint and the other as the binding dwell (Fig. 4 A). The rotational angle difference of the γ subunit between the binding dwell and catalytic dwell in F 1anc_core was approximately 34°, smaller than that of TF 1[30] (44°). Structural comparison of the β subunits between F 1anc_core and TF1 showed that they were highly similar (Fig.4B, C). Some nucleotides bound to the catalytic sites differed from those in TF1, likely because no exogenous nucleotides were added during sample preparation (Supplementary Fig.12-13). This suggests that the observed structures represent a state in which the enzyme re -establishes equilibrium with nucleotides originally carried over in bound form, rather than an authentic catalytic intermediate. Consistent with this overall structural similarity, the three-dimensional arrangement of conserved ATPase motif residues in the nucleotide-binding site is also well preserved between the ancestral and extant bMF1 (Supplementary Fig.14), in line with the sequen ce alignment shown in Fig. 2 B. Although the catalytic glutamate and arginine finger are in slightly different conformations, this is likely just due to the limited resolution of the structural data. TF1 is known to operate via a six-step rotational catalytic mechanism[4][30]. Given that the β subunit structures of F1anc_core correspond one-to-one with those of TF1, we infer that F1anc_core also operates via a six-step rotational catalytic mechanism similar to TF1. Structural alignment of the hexameric ring components (chains A –F) between F1anc_core lacking the γ subunit and the TF 1 binding dwell state yielded an RMSD of 0.002 Å, indicating a high degree of structural conservation within the ring. Since the hexameric ring without the central γ subunit is considered unaffected by the central γ subunit derived from TF1, it is highly likely that the stator ring of F1anc_core is prone to adopt the binding dwell structure without the γ subunit. Regarding the catalytic dwell, all structural analyses of extant species to date have consistently observed the catalytic dwell structure. Taking these points into account, it is highly likely that F1anc_core had two distinct conformational states—binding dwell state and catalytic dwell state—as same as TF1. However, the angular orientation of the γ subunit in the binding dwell and catalytic dwell states is influenced by the species from which the γ subunit is derived[14]. Therefore, the magnitude of the rotational angle difference between the two states in the F1 complex fully composed of the ancestral subunits (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint may vary. Fig.4. Structure of F1anc_core. A. Structures of the two rotational dwells of F1anc_core. The structures from the top are shown, with subunits colored as in Fig. 3a. Comparison of the Catalytic dwell and Binding dwell suggests that the γ subunit rotates counterclockwise between the two dwells (rotation highlighted with white bars and a black arrow). Each dwell exhibits three distinct conformations of (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint the β-subunits, illustrating the six sub -states (termed β0° 𝐻𝐶′ (Empty) , β80° 𝐶 (DP) , β120° 𝐶 (DP) , β200° 𝐶 (TP) , β240° 𝐻𝑂 (DP · P) , β320° 𝑂 (Empty) ) through which the enzyme progresses during its hydrolysis cycle. (B-C) Comparison of the β Conformations of F1anc_core. The β subunit structure of F1anc_core was superimposed onto the corresponding β subunit structures of TF1, with a focus on the N- terminal 81 amino acid residues forming the β-barrel structure. F1anc_core is depicted in orange, TF1 in the open conformation in blue, TF 1 in the closed conformation in purple, and TF 1 in the half -open conformation in yellow. B. Superimposition of the three β subunits in the catalytic state. The TF 1 structure is based on PDB entry 7L1R. C. Superimposition of the three β subunits in the binding state. The TF1 structure is based on PDB entry 7L1Q. Rotation analysis Single-molecule rotation assays of F1anc_core were performed using 40 nm gold colloid particles as the rotational probe, observed with a laser dark-field microscope at 2000 fps (frames per second)[31]. F1anc_core rotated in a counterclockwise direction, exhibiting three distinct rotational pauses at mM level of [A TP] (Fig. 5 C). Single -molecule rotation assays under various ATP concentrations allowed determination of the Vmax and Km values through Michaelis -Menten kinetics analysis, which were calculated to be 15.4 rps and 2.6 μM, respectively (Supplementary Figure 15). These Vmax and Km values were approximately one -tenth of those observed for TF 1. The rate constant of ATP binding, kon, estimated as 3×Vmax / Km, was determined to be 1.6×107 M-1s-1, which is very close to that of TF 1 (1.8×107 M-1s-1). Thus, the major kinetic difference is the slow maximum rotation speed. The maximum rotational velocity 15.4 rps corresponds to 46/sec as kcat of hydrolysis, that is significantly faster than the rate of hydrolysis estimated from biochemical ATPase activity assays performed with ATP regeneration system (Table 1). Such a discrepancy between the single- molecule rotation assay and the biochemical assay has often been reported[11][12][32], and it is usually attributed to ADP-inhibition. In the presence case, in addition to ADP -inhibition, the (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint apparently lower hydrolysis activity estimated from biochemical analysis can be attributed to the heterogeneity of the sample. Cryo-EM analysis revealed that a significant fraction of the sample consisted of partial complexes lacking the γ subunit and αβ pair. Another distinctive feature of F1anc_core rotation is that it proceeds in three steps per turn at all of [ATP] examined (Fig. 5A-C), whereas TF1 is known to make 80° and 40° substeps during each 120° rotation , particularly around Km region making six steps per turn . The buffer exchange experiment where [ATP] was switched between high and low concentrations confirmed that the dwell positions observed at low and high [ATP]s were coincident with each other (Supplementary Figure 16). This observation can be interpreted to mean that F1anc_core performs both ATP binding and hydrolysis within the same dwell. However, such an interpretation contradicts the cryo-EM analysis, which shows that F1anc_core adopts two distinct conformational states —a binding dwell and a catalytic dwell —where the γ subunit’s orientation differs by 33.5° (Fig. 4 A). Another possibility is that F1anc_core pauses for a prolonged period at the binding dwell angles while waiting on a reaction step other than ATP binding, and that the long pause dominates the overall reaction time, making the shorter catalytic dwells effectively negligible. To test this hypothesis, we conducted the rotation assay in the presence of ATPγS, a slowly hydrolyzed ATP analog often used to identify the angular position of the catalytic dwell in the rotation assay of F 1[11][12][33]. As expected, F1anc_core rotated at significantly lower speed in ATPγS than in ATP; the maximum rate was 1.46 rps, which is around one-tenth of ATP-driven rotation (Supplementary Figure 17). When the rotation was observed in the presence of 3 μM ATPγS— near the Km range for ATPγS-driven rotation (0.68 μM)—six rotational dwells were identified, separated by 32.1°, consistent with the predictions from cryo -EM analysis. To correlate these ATPγS-based dwell positions with those observed under ATP , we performed solution exchange experiments, alternately supplying ATP and ATPγS. By comparing the dwell positions in both conditions, we found that three of the six dwell points observed with ATPγS matched those under ATP. Based on this result, we assigned the three shared dwell points , seen with both ATP and ATPγS, to the binding dwell, whereas the dwell points observed exclusively under ATP γS were identified as the catalytic dwells. The analysis of the angular differences of the binding dwell from the catalytic dwell revealed a histogram centered at 32.1°. This angular difference aligns (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint closely with the rotational angle difference of ~34° between the binding and catalytic dwells of the γ subunit observed in Cryo-EM structural analyses, supporting the above assignment. Thus, it was confirmed that F1anc_core has two distinct states pausing at binding dwell angles or catalytic dwell angles as expected from Cryo -EM analysis, while the motor pauses predominantly at binding dwell angles waiting for a long reaction step to occur under ATP conditions. TF1 is reported to make pauses at binding dwel l angles when observed at low temperatures , due to temperature sensitive reaction (TS reaction). To test the possibility of TS reaction, we measured ATP hydrolysis activity at low temperatures. However, Q10 factor estimated from ATPase assay for F1anc_core was 1.52, too low to attribute the rate-limiting step as TS reaction (Supplementary Table 4). Thus, the long pause found at binding dwell angles should be a new class of reaction or conformational state. Fig.5. Single-molecule rotation assay of F1anc_core. (A-D) Time courses of rotation. The angular position histograms (left), and the x-y plots of the centroid of a rotating gold colloid (right) are shown in inset. A. 1 μM ATP B. 3 μM ATP (Km). C. 3mM A TP. D. 3 μM ATPγS. E. The angular position histograms of 3 mM ATP (orange) and 3 μM A TPγS (green). The x-y plots are shown in inset. The arrows show the angular position of catalytic dwell (green) and binding dwell (orange). F. Histogram of angular differences between catalytic dwell and binding dwell positions. Values are mean ± SD (n A B C D E F (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint = 25, 10 molecules).

In this study, we focused on the phylogenetic tree of rotary ATPases and inferred ancestral sequences of the hexameric ring –forming subunits at multiple nodes: the ancestral F 1-ATPase, the common ancestral rotary ATPase of F1- and V1-ATPases, and an even earlier rotary ATPase with a homo -hexameric ring. As a result, we obtained subunit sequences for these ancestral rings—αanc, βanc, CF/V, NCF/V and CancR. Comparative analysis revealed that these ancestral rotary ATPases already possessed the key motifs required for ATP hydrolysis —namely, the arginine- finger, catalytic glutamate, and Walker motif A—that are almost universally found in extant F 1- ATPases. (Fig. 2 B, Supplementary Figure 7). Although we did not perform a quantitative structural analysis, the predicted structures of these ancestral subunits closely resemble those of modern subunits (Supplementary Figure 8). Considering that LUCA employed either the ancestral rotary ATPase corresponding to F1 or V1-ATPase (or both)[1][2], the common ancestor of F1 and V1 must have already arisen as a rotary ATPase with a hexameric ring in the pre-LUCA era. It is also plausible that the ancestral ATPase diverging from the Type III secretion system (T3SS) ATPase possessed a hexameric ring structure, given that extant T3SS ATPases likewise form hexamers[34] [22]. Before expressing F1anc_core, which carries αanc and βanc sequences in its core domains, we first attempted to produce several ancestral ATPases—namely CF/V, NCF/V, and CancR. However, none of these proteins formed stable complexes, indicating that it remains challenging to obtain functional complexes solely from ancestral sequences with our current knowledge of rotary ATPase and the present ASR methodology. To address this issue, we constructed a chimeric ATPase using an extant F1 (TF1) as a structural scaffold. Specifically, we combined the TF1 N- terminal domain (NTD), which forms the base of the hexameric ring, with the ancestral sequences for the functionally essential domains—the nucleotide-binding domain (NBD) and the C-terminal domain (CTD) of αanc and βanc. When expressed with the γ subunit of TF1, this design yielded a stable F1 complex, termed F1anc_core. F1anc_core was expressed as a stable complex exhibiting ATPase activity although the activity is significantly lower than TF 1. Cryo-EM structural analysis revealed the significant fraction of (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint molecules retained the α3β3γ complex although partial complexes such as α3β3, α2β2γ, or α2β2 subcomplexes were also found. The α3β3γ complex structure of F1anc_core exhibited the two conformations. Comparison with TF 1 structures identified the two conformational states as the binding dwell state and catalytic dwell state, indicating that F 1anc_core shares the same rotational catalytic mechanism as same as TF 1. Interestingly, the conformational states of the α3β3 subcomplex corresponds to the binding dwell s tate as found in the α3β3 subcomplex of TF 1, reinforcing the abovementioned idea that F1anc_core operates the similar rotational catalytic mechanism to TF1, alternating the conformational state between binding dwell state and catalytic dwell. The result of rotation assay of F 1anc_core was apparently against this expectation. F 1anc_cor showed a 3-step rotation at all tested [ATP], indicating that it consistently pauses at binding dwell angles. Presuming that this apparent 3 -step pattern arises from an extended binding dwell overshadowing a brief catalytic dwell, we conducted rotation as says with ATPγS. Under these conditions, we observed well -defined stepping, with six pauses per turn. A subsequent buffer - exchange experiment—switching between ATP and ATPγS—revealed that F 1anc_core pauses at both binding and catalytic dwell angles, which differ by 32.1°. This value closely matches the ~34° difference determined by Cryo-EM analysis. Thus, F 1anc_core essentially follows a reaction scheme similar to that of TF 1, featuring two stable conformational states (binding dwell and catalytic dwell), even though the binding dwell of F 1anc_core is significantly longer than its catalytic dwell (Supplementary Figure 18). The molecular mechanism behind this prolonged binding dwell remains unclear. We hypothesized it might stem from a temperature-sensitive process, akin to what has been observed in TF1 rotation assays at low temperatures. However, the Q10 factor for the ATPase activity of F1anc_core is only 1.52—too low for a typical temperature -sensitive reaction. A candidate reaction responsible for the long -lived dwell is ADP release step that was suggested at or near the binding dwell angles[6][35]. Further investigation remains to address this point. Extant F1-ATPases exhibit diverse rotary mechanisms, classified as 3-steppers, 6-steppers, or 9-steppers. Our findings suggest that the ancestral F1-ATPase was fundamentally a 6-stepper. One might argue that if the binding dwell is relatively longer than the catalytic dwell, the motor (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint could effectively appear to be a 3-stepper. Yet, PdF1, the only known 3-stepper motor, still displays 3-step rotation even when measured with ATP γS, differing it from F1anc_core described here. Consequently, it seems more appropriate to conclude that the ancestral F 1 was intrinsically a 6- stepper, and it diversified into more 3-steppers, 6-steppers, and 9-steppers (Fig. 6). We previously identified an intriguing correlation between the number of rotational steps of F1 and the number of the c-subunits in the c-ring[13], although more data is needed to confirm its generality. From this correlation, one empirical rule emerges: a 6-stepper F1 pair with a 10-stepper Fo (i.e., 10 c-subunits). This rule suggests that ancestral F1 likely paired with a Fo containing a c10- ring. If the free energy of A TP hydrolysis was comparable to that of extant cells, then LUCA and related lineages must have already possessed electrically tight plasma membranes capable of maintaining a sufficiently high pmf. This follows from the fact that the number of c-subunits is one of the key factors determining the H⁺/ATP ratio and the equilibrium pmf between ATP hydrolysis and synthesis. From an engineering perspective, this study highlights a new strategy: using TF1 as a scaffold for implementing various ancestral or designed catalytic domains, including the common subunits of F1- and V1-ATPase, CF/Vand NCF/V. By extending this idea to the entire FoF1 complex, one could consider incorporating more ancestral-sequence subunits into a stably structured FoF1—such as T FoF1 found in extant species—to more accurately estimate the function of ancestral FoF1. Currently, ancestral sequence reconstruction methods are largely restricted to highly conserved subunits, so beyond the α and β subunits of F1, they can only be applied to certain regions of the c-subunit or the a-subunit. Even so, studying chimeric or hybrid FoF1 complexes that include such ancestral functional units may yield insights into functions and dynamics that cannot be inferred from sequence analysis alone. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint Fig.6. Differences in Rotational Catalytic Mechanisms Among Species and Their Branching Positions. The branching positions and unique features of the rotational catalytic mechanisms for species with well-characterized mechanisms are presented. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint

Phylogenetic analysis and ancestral sequence inference. Protein sequences were retrieved from the NCBI (National Center for Biotechnology Information) database. In this study, the KF database ver.2021 [36] was further expanded for A TPase research, resulting in the KFS Database. The database comprised all protein sequences of 136 archaeal species, 594 bacterial species, and 77 eukaryotic species, and was employed for BLAST searches. Amino acid sequences for the α and β subunits of the F 1-ATPase from bovine mitochondria (ACCESSION: P19483, P00829), the A and B subunits of the V1-ATPase from Thermus thermophilus (Q56403, Q56404), and the T3SS FliI protein from Escherichia coli (NP_416451) were retrieved from NCBI as query sequences. The FliI family of the T3SS was incorporated into the phylogenetic tree as an outgroup to determine the root position of F 1-ATPase and V1-A TPase. These sequences were subjected to BLASTP searches [24] against the KFS Database to collect homologous sequences. Redundant sequences and those with significantly different lengths were removed. The remaining 617 sequences were realigned using MAFFT [37] with secondary structure considerations, followed by manual refinement to produce a multiple sequence alignment, which was subsequently used for phylogenetic analysis. Alignment trimming was performed using TrimAl [38] (version 1.4) in automated trimming mode ( - automated1). Phylogenetic trees were constructed using IQ -TREE[25] (version 2.1.3). The ModelFinder Plus [39] (MFP) module identified the LG+R10 model as the optimal amino acid substitution model. Ultra fast bootstrap analysis[40] with 1,000 replicates was conducted to evaluate branch confidence. For additional reliability assessments, phylogenetic analysis was also conducted using RAxML [28] (version 8.2.12). The LG+G+I model was selected, and bootstrap analysis with 1,000 replicates was performed. Phylogenetic analyses and ancestral sequence reconstruction were performed using the codeml program in the PAML package (version 4.9j, February 2020) [26] with the LG substitution model for amino acid sequences. Default parameters were used unless specified otherwise. The gap positions of ancestral sequences were inferred by GASP[27] (version 2.0.0). (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint The ancestral sequences generated by PAML were compared with those by GASP. Regions identified as gaps in the ancestral sequence of GASP were excised from the PAML ancestral sequences to produce the final ancestral sequence set. Preparation for F1anc_core Genuine TF 1 was prepared as described [41]. To visualize rotation, two cysteine residues were introduced as previously reported [14]. The plasmid encoding F1anc_core was constructed using the TF 1 plasmid as a vector and PCR products encoding the C -terminal and nucleotide-binding domains of αanc and βanc (both with N-terminal His-tags) as insert DNAs. After ligation, the recombinant plasmid was introduced into the F oF1-deficient E. coli strain JM103∆unc. F1anc_core was then expressed in E. coli, purified, and biotinylated as described in [42]. The protein concentration was determined by UV absorbance using a molar extinction coefficient of 182,500 M⁻¹ cm⁻¹, calculated from its amino acid sequence using the ProtParam tool (ExPASy). Calibration Curve for Molecular Weight Determination by Size-Exclusion HPLC Molecular weight determination was performed using size -exclusion high -performance liquid chromatography (SEC-HPLC). A calibration curve was generated using the MW-Marker (HPLC) for Molecular Weight Determination (Oriental Yeast Co., Ltd.). Chromatographic analysis was conducted on an HPLC system equipped with a size -exclusion column maintained at 25°C. The mobile phase consisted of 50 mM HEPES-KOH (pH 7.5) containing 100 mM NaCl, delivered at a flow rate of 0.5 mL/min. UV absorbance was monitored at 280 nm. Retention times of the marker proteins were recorded, and a calibration curve was constructed by plotting the logarithm of molecular weight against retention time. A linear regression analysis was performed to establish the calibration equation, which was subsequently used to estimate the molecular weight of F1anc_core. Calibration was validated through triplicate measurements, ensuring reproducibility within an acceptable standard deviation. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint ATPase activity assay ATPase activity was measured at 25°C in a buffer containing 50 mM Hepes -KOH (pH 7.0), 50 mM KCl, 3 mM MgCl₂, and an A TP-regenerating system (0.2 mM NADH, 2.5 mM phosphoenolpyruvate, 200 μg/mL pyruvate kinase, and 50 μg/mL lactate dehydrogenase). Activity was calculated from the slope of NADH absorbance during the first 300 seconds of measurement. LDAO-stimulated activity was assessed by adding 0.3% LDAO to the reaction and calculating the slope of NADH absorbance thereafter. Cryo-EM grid preparation. A volume of 3.5 μL of purified F1anc_core was applied to a glow-discharged holey gold grid (Ultrafoils R0.6/1.0, 200 mesh). The grids were blotted for 4 seconds at 22°C and 100% humidity, then plunge- frozen in liquid ethane using a FEI Vitrobot Mark IV . Data collection. The grids were initially screened for ice thickness and particle density using a Thermo Fisher Scientific Talos Arctica transmission electron microscope (TEM) operating at 200 kV . Subsequently, the grids were transferred to a Thermo Fisher Scientific Titan Krios TEM operating at 300 kV , equipped with a Gatan BioQuantum energy filter (20 eV slit width) and a K3 camera. To mitigate orientation bias observed in an initial test sample, movie micrographs were recorded at tilt angles ranging from 20° to 40°, as s uggested by cryoEF [43], which indicated an optimal tilt angle of ~37°. Automatic data collection was performed using EPU (E Pluribus Unum, Thermo Fisher Scientific) at a nominal magnification of ×60,000 (displayed magnification of ×165,000 due to the energy filter), resulting in a pixel size of 0.84 Å. The total electron dose was set to 62 electrons per Ų, distributed over 80 frames with a total exposure time of 6.2 seconds. A total of 4373 movie micrographs were collected. Data processing. All image processing and refinement were performed using CryoSPARC v4.4.1[44]. Initially, micrographs were motion -corrected, and defocus values were estimated using the patch -based (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint workflow. Particles were automatically picked and subjected to two -dimensional (2D) classification to exclude "junk" particles, such as those from aggregates or minor contaminants. The 2D classes were split into hexamers and tetramers and processed separately. Initial models were generated by ab initio classification into 2 -3 classes. These ab initio maps were then used as inputs for heterogeneous refinement and 3D classifications, allowing further classification of particles into distinct structural classes. Each class was independently processed through homogeneous refinement and non -uniform refinements, yielding the final high-resolution maps. In regions where lower-resolution features were observed, DeepEMhancer [45] was applied to sharpen the maps, enhancing their interpretability in figures displaying the entire complex. Model building. Models were constructed for intact F 1 complexes and refined using Coot [46] and PHENIX [47], with PDB structures 7L1Q (TF1 binding dwell cryo-EM structure) and 7L1R (TF1 catalytic dwell cryo-EM structure) serving as templates. Supplementary Table 3. shows details of the refinement and validation statistics. Figures made using PyMOL. Single-molecule rotation assay Flow chambers were constructed using double-sided tape as spacers and two cover glasses (18 × 18 mm² and 24 × 32 mm²; Matsunami Glass). The bottom glass surface was coated with Ni-NTA. The basic assay buffer contained 50 mM Hepes-KOH (pH 7.5), 100 mM KCl, and 5 mM MgCl₂. When ATP was used as the substrate, an A TP-regenerating system (2 mM phosphoenolpyruvate and 100 μL/mL pyruvate kinase) was added. The flow chamber was first incubated with basic buffer containing 5 mg/mL BSA (BSA buffer) for 5 minutes. F1 molecules (200–500 pM) in BSA buffer were then introduced and incubated for 10 minutes, followed by washing with BSA buffer to remove unbound molecules. Next, 40 nm gold nanoparticles (prepared as described in [17]) were introduced, incubated for 10 minutes, and unbound particles were removed by washing with substrate-containing basic buffer. Rotational assays were performed as described in [17] at room temperature. Recorded videos were (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint analyzed using custom software. Data availability The cryo-EM maps from this study are deposited in the Electron Microscopy Data Bank (EMDB) under accession codes 49841, 49839, 49840, 49843 and 49842 for hexamer without stalk, hexamer with stalk in binding dwell, hexamer with stalk in catalytic dwell, tetramer without stalk, and tetramer with stalk, respectively. The models generated and analyzed for the two of the above maps are available in the Protein Data Bank under accession codes 9NVL for hexamer with stalk in binding dwell, and 9NVM for hexamer with stalk in catalytic dwell. 【Reference】 [1] Moody E. R. R, et al (2024) The nature of the last universal common ancestor and its impact on the early earth system. Nat Ecol Evol 8(9): 1654. [2] Mahendrarajah TA, et al (2023) ATP synthase evolution on a cross-braced dated tree of life. Nat Commun 14(1) [3] Weiss MC, et al (2016) The physiology and habitat of the last universal common ancestor . Nat Microbiol 1(9). [4] Noji H, Ueno H & Mcmillan DGG (2017) Catalytic robustness and torque generation of the F1- ATPase. Biophys Rev 9(2): 103. [5] Uchihashi, N., Iino, R., Ando, T. & Noji, H. High-speed atomic force microscopy reveals rotary catalysis of rotorless F1-ATPase. Science 333, 755–758 (2011). [6] Noji H, Ueno H (2022) How Does F1-ATPase Generate Torque?: Analysis From Cryo-Electron Microscopy and Rotational Catalysis of Thermophilic F1. Front. Microbiol. 13:904084. [7] Enoki, S., Watanabe, R., Iino, R. & Noji, H. Single-molecule study on the temperature-sensitive reaction of F1-ATPase with a hybrid F1 carrying a single β(E190D). J. Biol. Chem. 284, 23169–23176 (2009). [8] Li, CB., Ueno, H., Watanabe, R. et al. ATP hydrolysis assists phosphate release and promotes reaction ordering in F1-ATPase. Nat Commun 6, 10223 (2015). [9] Yanagisawa S, Bukhari ZA, Parra KJ & Frasch WD (2024) Eukaryotic yeast V 1-ATPase rotary (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint mechanism insights revealed by high-resolution single-molecule studies. Front Mol Biosci 11. [10] Steel BC, et al (2015) Comparison between single-molecule and X-ray crystallography data on yeast F1-ATPase. Sci Rep 5(1). [11] Kobayashi R, Ueno H, Li C -B, Noji H (2020) Rotary catalysis of bovine mitochondrial F 1- ATPase studied by single-molecule experiments. Proc Natl Acad Sci USA 117:1447–1456. [12] Zarco-Zavala M., Watanabe R., McMillan D. G. G., Suzuki T., Ueno H., Mendoza-Hoffmann F., García-Trejo J. J., Noji H. (2020) The 3 × 120° rotary mechanism of Paracoccus denitrificans F1-ATPase is different from that of the bacterial and mitochondrial F1-A TPases. Proc Natl Acad Sci USA 117(47): 29647–29657. [13] Noji H, Ueno H, Kobayashi R (2020) Correlation between the numbers of rotation steps in the ATPase and proton-conducting domains of F- and V-A TPases. Biophys Rev 12: 303–307. [14] Watanabe, R.R., et al. (2023). Rotary properties of hybrid F 1-ATPases consisting of subunits from different species. iScience, 26, 106626. [15] Gogarten, J.P., et al. (1989). Evolution of the vacuolar H+-ATPase: Implications for the origin of eukaryotes. Proc. Natl. Acad. Sci. U.S.A., 86, 6661–6665. [16] Ueno H, Suzuki K & Murata T (2018) Structure and dynamics of rotary V1 motor. Cell Mol Life Sci 75(10): 1789. [17] Iida T, et al (2019) Single-molecule analysis reveals rotational substeps and chemo-mechanical coupling scheme of enterococcus hirae V 1-A TPase. Journal of Biological Chemistry 294(45): 17017. [18] Imamura, H., Takeda, M., Funamoto, S., Shimabukuro, K., Yoshida, M., and Yokoyama, K. (2005) Rotation scheme of V 1-motor is different from that of F 1-motor. Proc. Natl. Acad. Sci. USA. 102, 17929–17933. [19] Matsui M (2021) Frontier of Inferring Phylogenies. JSBi Bioinformatics Review 2(1): 30-57. [20] Spence MA, Kaczmarski JA, Saunders JW, Jackson CJ (2021) Ancestral sequence reconstruction for protein engineers. Curr Opin Struct Biol 69: 131-141. [21] Dube N, Khan SH, Okafor CD (2022) Ancestral sequence reconstruction for evolutionary characterization of proteins. Trends Biochem Sci 47(1): 98-99. [22] Kibak, H., et al. (1992). Evolution of structure and function of V -ATPases. J. Bioenerg. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint Biomembr., 24(4), 415–424. [23] Schupfner M, Straub K, Busch F, Merkl R, Sterner R. (2019) Analysis of allosteric communication in a multienzyme complex by ancestral sequence reconstruction. Proc Natl Acad Sci USA 117: 346–354. [24] Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 25(17): 3389–3402. [25] Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ (2015) IQ -TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol 32(1): 268– 274. [26] Yang Z (2007) PAML 4: phylogenetic analysis by maximum likelihood. Mol Biol Evol 24(8): 1586–1591. [27] Edwards RJ, Shields DC (2004) GASP: Gapped Ancestral Sequence Prediction for proteins. BMC Bioinformatics 5: 123. [28] Stamatakis A (2014) RAxML version 8: a tool for phylogenetic analysis and post -analysis of large phylogenies. Bioinformatics 30(9): 1312–1313. [29] Hirono-Hara Y ., Noji H., Nishiura M., Muneyuki E., Hara K. Y ., Yasuda R., Kinosita K. Jr., Yoshida M. (2001) Pause and rotation of F 1-ATPase during catalysis. Proc Natl Acad Sci USA 98(24): 13649–13654. [30] Sobti M, Ueno H, Noji H & Stewart AG (2021) The six steps of the complete F1-ATPase rotary catalytic cycle. Nat Commun 12(1) [31] Ueno H., Nishikawa S., Iino R., Tabata K. V ., Sakakihara S., Yanagida T., Noji H. (2010) Simple dark-field microscopy with nanometer spatial precision and microsecond temporal resolution. Biophys J 98(9): 2014–2023. [32] Yasuda R., Noji H., Yoshida M., Kinosita K. Jr., Itoh H. (2001) Resolution of distinct rotational substeps by submillisecond kinetic analysis of F1-ATPase. Nature 410(6828): 898–904. [33] Shimabukuro K., Yasuda R., Muneyuki E., Hara K. Y ., Kinosita K. Jr., Yoshida M. (2003) Catalysis and rotation of F 1 motor: Cleavage of A TP at the catalytic site occurs in 1 ms before 40° substep rotation. Proc Natl Acad Sci USA 100(25): 14731–14736. (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint [34] Majewski DD, et al ., (2019) Cryo-EM structure of the homohexameric T3SS ATPase -central stalk complex reveals rotary ATPase-like asymmetry. Nat Commun 10(1). [35] Adachi K, Oiwa K, Nishizaka T, Furuike S, Noji H, Itoh H, Yoshida M, Kinosita K Jr. (2007) Coupling of rotation and catalysis in F 1-ATPase revealed by single -molecule imaging and manipulation. Cell 130(2): 309-321. [36] Furukawa R, Yokobori S, Sato R, Kumagawa T, Nakagawa M, Katoh K, Yamagishi A (2022) Amino acid specificity of ancestral aminoacyl -tRNA synthetase prior to the last universal common ancestor. J Mol Evol 90(1): 73–94. [37] Katoh K, Standley DM (2013) MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol 30(4): 772–780. [38] Capella-Gutiérrez S, Silla -Martínez JM, Gabaldón T (2009) trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25(15): 1972–1973. [39] Kalyaanamoorthy S, Minh BQ, Wong TKF, von Haeseler A, Jermiin LS (2017) ModelFinder: fast model selection for accurate phylogenetic estimates. Nat Methods 14(6): 587–589. [40] Hoang DT, Chernomor O, von Haeseler A, Minh BQ, Vinh LS (2018) UFBoot2: improving the ultrafast bootstrap approximation. Mol Biol Evol 35(2): 518–522. [41] Okuno D, Fujisawa R, Iino R, Hirono -Hara Y , Imamura H, Noji H (2008) Correlation between the conformational states of F 1-ATPase as determined from its crystal structure and single - molecule rotation. Proc Natl Acad Sci USA 105:20722–20727. [42] Rondelez Y ., Tresset G., Nakashima T., Kato-Yamada Y ., Fujita H., Takeuchi S., Noji H. (2005) Highly coupled A TP synthesis by F1-A TPase single molecules. Nature 433(7027): 773–777. [43] Naydenova, K. & Russo, C. J. Measuring the effects of particle orientation to improve the efficiency of electron cryomicroscopy. Nat. Commun. 8, 629; 10.1038/s41467 -017-00782-3 (2017). [44] Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017). [45] Sanchez-Garcia, R. et al. DeepEMhancer: a deep learning solution for cryo -EM volume post - processing. Commun. Biol. 4, 874; 10.1038/s42003-021-02399-1 (2021). [46] Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010). [47] Afonine, P. V . et al. Real-space refinement in PHENIX for cryo -EM and crystallography. Acta Crystallogr. D Struct. Biol. 74, 531–544 (2018). (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted May 2, 2025. ; https://doi.org/10.1101/2025.04.30.651372doi: bioRxiv preprint