Structural Snapshots of F1 ‐ATPase Catalysis in Drug Resistant Acinetobacter baumannii Reveal Novel Therapeutic Targets Authors: Khoa C M Le, Jan K Marzinek, Vikneswaran Mathiyazakan, Peter J Bond, Gerhard Grüber Published: 2026 ABSTRACT ABSTRACTGram‐negative Acinetobacter baumannii , one of the most critical priority pathogens, has developed multiple defense mechanisms, including upregulation of drug efflux pumps powered by the proton‐motive force (pmf) or ATP. As a nonfermenting bacterium, it relies on oxidative phosphorylation, in which the F‐ATP synthase, composed of an F1‐ATPase (α 3:β 3:γ:ε) and FO‐domain (a:b 2 :c 10), synthesizes ATP, the cell's currency of life. Here, we present cryo‐electron microscopy structures of the F1‐ATPase captured during ATP synthesis, revealing concerted conformational changes within the α 3:β 3‐catalytic head that trigger structural transitions of subunit ε and twisting of the γ:ε central stalk, coupled with spontaneous ATP formation. Molecular dynamics simulations validate and expand the ensemble of structural rearrangements underlying the rotary mechanism. The resolved intermediate states of the ε subunit reveal its role as a molecular gear, regulating ATP synthesis while simultaneously inhibiting ATP hydrolysis, essential for the pathogen to prevent wastage of ATP. These structural snapshots identify novel therapeutic targets and pave the way for rational design of selective inhibitors.Keywords: Acinetobacter baumannii , antibiotics, ATP synthase, bacterial pathogenesis, bioenergetics, membrane protein Introduction According to the World Health Organization (WHO), Acinetobacter baumannii is one of the top ten pathogens worldwide and is regarded as one of the most antibiotic‐resistant bacteria [1]. The Gram‐negative bacterium has developed resistance to almost all antibiotics. The associated antimicrobial mechanisms encompass expression of enzymes that neutralize drugs or modify their targets, alteration of cell wall permeability, upregulation of drug efflux pumps, and Type II‐, III‐ and IV secretion systems [2, 3]. The efflux pumps and secretion systems represent the main causative factors for A. baumannii's drug resistance. While most of the secretion systems are ATP‐driven, the efflux pumps are either ATP‐ or proton‐motive force (pmf) driven, energy forms generated by the oxidative phosphorylation (OXPHOS) pathway. The F1FO‐ATP synthase (F‐ATP synthase) is a central component of the A. baumannii OXPHOS pathway, catalyzing the conversion of ADP + Pi to ATP [4]. As a glucose‐ and lactose non‐fermenting bacterium [5], the pathogen derives most of its ATP from oxidative phosphorylation, rendering this path and enzyme a particularly attractive target space [6, 7]. The AbF‐ATP synthase (α 3:β 3:γ:δ:ε:a:b:c 10, [4]) shows substantial architectural divergence in overall subunit number and amino acid composition from its human mitochondrial counterpart (α 3:β 3:γ:δ:ε:IF1a:b:c 8 :e:f:g:ATP8:j:k:d:F6:OSCP) [8], a property that could be exploited to minimize host toxicity and improve target specificity. The water‐soluble AbF1‐ATPase (α 3:β 3:γ:ε) contains three alternating subunit αβ‐pairs, including the catalytic and nucleotide binding sites at their interfaces and the central stalk subunit pair γ‐ε [9]. The latter connects proton translocation via the FO‐domain subunits a‐c 10 and ATP synthesis in the α 3:β 3‐sector [4, 10, 11]. Subunits b 2 :δ build the peripheral stalk and facilitate the transmission of power between the rotary γ:ε stalk subunits and the c 10‐ring [4]. The AbF1FO‐ATP synthase [4] and F1‐ATPase [9] exhibit low ATP hydrolysis activity to prevent wastage of energy, thus controlling the pmf and the efficiency of pumps involved in drug uptake and efflux [2, 3]. Since the AbF1FO‐ATP synthase is a dual‐function catalyst, it needs a defined regulatory element to stimulate or inhibit ATP synthesis or hydrolysis depending on metabolic needs. Here we report cryo‐electron microscopy (cryo‐EM) structures of an ATP synthesizing AbF1‐ATPase, supported by molecular dynamics (MD) simulations, which reveal an ensemble of transition states, concerted actions between catalytic and rotary domains, and illuminate new targets for rational inhibitor design. The structures elucidate the transformations, rearrangements, and movements of the dual regulator subunit ε, triggering the enzyme into its active form, with relevance to F‐ATP synthases across a variety of antimicrobial resistant bacteria. Materials and Methods Purification and ATP Synthesis Measurement The wild‐type (WT) AbF1‐ATPase was purified as previously described [9]. ATP synthesis was measured using the previously described assay [9, 11], with modifications. Purified AbF1‐ATPase was diluted to a final concentration of 10 μg/mL in ATP synthesis buffer (50 mM 4‐Morpholinepropanesulfonic acid (MOPS), 8 mM MgCl2, pH 7.5), from which 25 μL was aliquoted onto 96‐well half‐area white plates. Then, 25 μL of substrate (50 mM MOPS, 8 mM MgCl2, 20 μM ADP, 500 μM KH2PO4) was added to the enzyme mixture and incubated for 30 s, 5 min, and 10 min at room temperature. Immediately after the incubation, 25 μL of CellTiter‐Glo Luminescent Cell Viability Assay (Promega, United States) was added and mixed well to stop the ATP synthesis reaction and start the luminescence reaction to quantify the amount of ATP synthesized. After a further 10 min incubation at room temperature in the dark, a luminescent reading was obtained using a BioTek Synergy H1 (Agilent, United States). These luminescent values were subtracted from a blank with the enzyme mix substituted with ATP synthesis buffer. Endogenous ATP amounts from the purified enzyme were also quantified (control), by mixing 25 μL of enzyme with 25 μL of ATP synthesis buffer, and then adding 25 μL of CellTiter‐Glo. The luminescence from the control sample was subtracted from the mixture of buffer and CellTiter‐Glo. Cryo‐EM Data Collection of Substrate‐Bound AbF1 ‐ATPase The purified enzyme was concentrated to 1.7 mg/mL and was then incubated with 12 μM ADP, 24 μM MgCl2, and 24 μM KH2PO4 to mimic the ATP synthesis assay [9, 11]. A 3.5 μL of the mixture with a final F1‐ATPase concentration of 1.5 mg/mL was applied to a holey carbon grid (Quantifoil 300 mesh R1.2/1.3 holey carbon grid; Quantifoil Micro Tools GmbH, Germany) that was glow‐discharged in air for 1 min. Grids were blotted twice using an FEI Vitrobot Mark IV (ThermoFisher Scientific (RRID:SCR_008452); Waltham, MA, USA) for 2.5 s at 100% humidity and 4°C before plunge‐freezing into liquid ethane. Cryo‐EM data of the substrate‐bound AbF1‐ATPase were acquired with an FEI Titan 300 kV Transmission electron microscope (TEM) (ThermoFisher Scientific) equipped with a Gatan K3 direct electron detector (Gatan; Pleasanton, CA, USA). Data were automatically collected using SerialEM 4.0 [12], from which 5093 movies were collected in super‐resolution mode with 40 frames per movie. All movies were recorded at 105000× magnification, giving a final pixel size of 0.858 Å/pixel. The total exposure for each movie was calculated to be 64.8 e − per Å2. Cryo‐EM Data Analysis All 5093 movies were imported into cryoSPARC V4.6.0 [13] and processed with the default pipeline: Patch Motion Correction, Patch CTF Estimation, Particle picking and extraction. Extracted particles were filtered to remove noise, then passed through multiple rounds of 2D classification to obtain sufficient views of the AbF1‐ATPase complex. The resulting 891 625 particles were then subjected to ab initio reconstruction to obtain a good initial EM map. This map was then used to classify the particle stack using Heterogeneous Refinement, resulting in 562 934 good particles for further analysis. Model Refinement and PDB Deposition Depending on the conformer, different models were used to fit into the respective maps. For states C1a and C1b, the previous WT AbF1‐ATPase model (PDB 7YRY [9]) was used, except for C1b, where β DP was duplicated from the β E chain. In the case of states C2a and C2b, the mutant models PDB 8ZI2 and 8ZI3 [14] were used, and the missing residues (ε135–139) were built manually in Coot [15]. States C3a and C3b were resolved using the structural model of the AbF1‐αβγε Δ134–139 mutant (PDB 8ZI4 [14]), with modifications to α DP of C3a. The chosen models were roughly fitted into the map using ChimeraX [16], refined in Real‐space‐refine in Phenix [17], followed by minor adjustments in Coot [15] and validated with MolProbity [18]. This process was repeated over multiple rounds to get an optimized final model. Model visualization and movie generation were performed with UCSF ChimeraX [16]. The structural models determined in this study are available in wwPDB at https://www.rcsb.org/ under PDB ID 24DH, 24DJ, 24DK, 24DL, 24DM and 24DN for conformers C1a, C1b, C2a, C2b, C3a and C3b, respectively. The cryo‐EM density maps are deposited into EM Data Bank at https://www.ebi.ac.uk/emdb/ under EMDB ID EMD‐69444, EMD‐69446, EMD‐69447, EMD‐69448, EMD‐69449 and EMD‐69450 for conformer C1a, C1b, C2a, C2b, C3a and C3b, respectively. MD Simulations—System Preparation The experimental cryo‐EM structure of AbF1‐ATPase in its C1a conformation was modeled using the CHARMM‐GUI [19]. Default protonation states at neutral pH were assigned to the protein using the CHARMM36m force field with charged termini [20]. Four ATP4−, two ADP3− and one PO42− with bound divalent Mg2+ cations were modeled assuming neutral pH. Each protein complex was placed in a cubic box of ~16.5 × 16.5 × 16.5 nm3 and solvated with ~130 000 TIP3P [21] water molecules. 150 mM of MgCl2 was added after neutralizing the overall charge of the system. The system was subjected to 5000 steps of steepest descent minimization followed by 50 ns of NPT equilibration with position restraints applied to the protein backbone atoms and heavy atoms of ATP4−, ADP3− and PO42− using a force constant of 1000 kJ mol−1 nm−2. Targeted Molecular Dynamics Simulations TMD simulations were performed using an RMSD‐based biasing potential, in which the RMSD to the target structure was decreased linearly over the course of the simulation using a harmonic restraint. We performed a series of consecutive TMD simulations, starting first from the C1a conformation, with each final structure serving as the starting point for biasing toward the subsequent state. These simulated transitions were as follows: (i) C1a to C1b, (ii) C1b to C2a, and (iii) C3a to C3b. Depending on the magnitude of the RMSD between initial and target structures, we used various TMD simulation lengths ranging from 200 to 600 ns to achieve an effective biasing rate of ~5 Å RMSD per 200 ns (Table S2). The RMSD harmonic potential force constant was linearly increased from 1000 kcal mol−1 Å−2 at 0 ns to 2000 kcal mol−1 Å−2 at the end of each simulation. Every 100 steps, the structure was fitted into the initial reference coordinates to remove global translation and rotation. The bias was applied only to selected subunits which were shown by cryo‐EM to undergo conformational changes. Each TMD simulation was followed by a short ~5–10 ns equilibration run without additional bias to relax the coordinates before the next transition. Simulation Details Equilibration and TMD simulations [22] were carried out using Gromacs 2018.8 [23]. Plumed 2.6.4 [24] was used in the case of TMD simulations. Both equilibration and TMD simulations used a temperature of 310 K and a pressure of 1 bar which were controlled using the v‐rescale thermostat and isotropic Parrinello‐Rahman barostat, respectively [25]. The particle‐mesh Ewald (PME) [26] method was used for long‐range electrostatic interactions with a cut‐off distance of 1.2 nm, and a switching function was applied after 1 nm. All bonds involving hydrogen atoms were constrained using the LINCS algorithm. Equations of motion were integrated using the leap‐frog algorithm with a time step of 2 fs. Periodic boundary conditions were applied in all directions. Each simulation employed 1 node with a 64‐core processor (AMD EPYC 7713 @ 2.0 GHz) and 2 GPUs (NVIDIA A100‐SXM4‐40 GB) on the National Supercomputing Centre (NSCC), Singapore. MD Analysis All simulation snapshots were generated using ChimeraX 1.11 [27] and VMD [28]. Simulation movies were generated using ChimeraX 1.11. To quantify the angular motion (C1a to C1b) of β DP C‐terminus relative to the stalk γ/ε over time, we used MDAnalysis [29]. Backbone atoms of the β DP C‐terminus and γ/ε were selected for each frame of the trajectory. Prior to analysis, the trajectory was superimposed onto the stalk to remove global translation and rotation. The center of mass (COM) for both β DP C‐terminus and the stalk was calculated at each time step. The vector between the COMs of β DP C‐terminus and the stalk was used to compute the angular motion by comparing it to the reference vector from the first frame. The resulting angle of rotation was plotted over time to visualize the relative motion of the C‐terminus with respect to the stalk. The number of contacts between β DP loop (residues 377–386) and ε was calculated based on a 0.4 nm cutoff distance. Per‐residue secondary structure analysis was performed using DSSP implemented in Gromacs [30]. To quantify torsional twisting of the εCTD, we computed the time‐resolved rotation of εCTD backbone atoms relative to the γ/ε stalk from TMD simulations using MDAnalysis. The γ/ε stalk (backbone atoms) was used to define a protein‐based rotational z‐axis, constructed in the reference frame as the unit vector connecting the centers of mass of the lowermost and uppermost backbone regions of the stalk. For each trajectory frame, global translation and rotation were removed by least‐squares superposition of the stalk backbone onto the first frame, and the same transformation was applied to εCTD coordinates. The relative rotation of εCTD with respect to its reference conformation was obtained by computing the optimal rotation matrix that superposes εCTD backbone atoms in each frame onto the reference εCTD. The instantaneous rotation matrix was converted to an axis–angle representation, and the signed twist angle θ(t) was calculated by projecting the rotation vector onto the stalk‐defined z‐axis. Angles were unwrapped to remove ±180° discontinuities and reported in degrees. Results Cryo‐EM Approach to Resolve the AbF1 ‐ATPase Incubated With MgADP and Pi Recently, we described ATP formation by the heterologously produced AbF‐ATP synthase within inverted membrane vesicles (IMVs) in the presence of about 12 μM MgADP and inorganic phosphate (Pi) [9, 10, 11]. Although the AbF1‐ATPase cannot undergo pmf‐driven rotational ATP synthesis in the absence of the FO domain, we studied whether ATP formation could occur under such nucleotide‐driven conditions and how this would be coupled with alterations within the F1‐ATPase subunits. Therefore, the purified recombinant AbF1‐ATPase was incubated with 12 μM MgADP + Pi. ATP synthesis measurements showed reproducible spontaneous ATP formation, which decreased over time due to slow cleavage of the formed ATP, confirming that the enzyme is active in the ATP synthesis direction. After ATP is formed, the enzyme switches to ATPase activity when no further pmf is provided (Figure 1A), which otherwise would drive the enzyme continuously in the ATP synthesis direction. To perform cryo‐EM studies, AbF1‐ATPase and MgADP + Pi were rapidly mixed and applied to a holey carbon grid before flash‐freezing (see Materials and Methods). Cryo‐EM data were recorded, resulting in 5094 movies. A summary of the data preprocessing, analysis steps and the cryo‐EM maps are illustrated in Figures S1–S3 and model statistics are presented in Table S1A,B. In total, six conformations of AbF1‐ATPase were resolved via cryo‐EM (Figure 1B), which, based on the conformation of ε's C‐terminal domain (εCTD), can be classified into three main groups (Figure 1B): an extended εCTD (C1a and C1b at 3.13 and 3.04 Å resolution), a partially extended εCTD (C2a and C2b at 3.05 and 2.95 Å resolution) as well as the outer transition states 1 and 2 of εCTD, called C3a and C3b at 3.08 and 2.88 Å resolution. While only an extended εCTD conformation could be observed in the WT AbF‐ATP synthase [4] and AbF1‐ATPase [9, 14], this is the first time that several conformations of stalk subunit ε could be observed in the entire WT AbF1‐ATPase. The biochemistry and structure of AbF1‐ATPase under work. (A) ATP synthesis of purified AbF1‐ATPase. Two conditions were tested: with substrate incubation (for 30 s, 5 min, 10 min) and without substrate (control). Both treatments were subtracted with their appropriate blanks (see Materials and Methods). Each measurement has three technical replicates across two biological replicates. Statistical analysis was performed to compare each condition with the spontaneous ATP synthesis (30 s) using an unpaired t‐test; ****p < 0.0001. (B) EM map of each conformer. Subunit α, β, γ and ε are colored dark green, orange, yellow and magenta, respectively. (C) Nucleotide occupancy of the β subunits from extended εCTD conformations C1a and C1b. In β TP and β E, similar nucleotides were found between the two conformers, binding to MgATP and ADP, respectively. However, β DP in C1a binds MgADP + Pi, while that of C1b binds only Pi. It's Swinging Time: The C‐Terminal Outside Movement of β DP The structure of conformer C1a, resolved at 3.13 Å resolution, comprises residues: α E, α TP and α DP 26–514, β E, β TP and β DP 2–464 (Figures S1 and S3). The three α subunits were occupied with MgATP, while MgATP, MgADP·Pi and ADP were observed in β TP, β DP and β E, respectively (Figure 1C). This differs from the MgADP (β TP), Pi (β DP) and empty (β E) arrangement described for the Ab‐F1FO ATP synthase (PDB ID 7P2Y, [4]), wild‐type (WT) AbF1‐ATPase (PDB ID 7YRY, [9]) and the AbF1‐αβγε Δ134–139 mutant (PDB ID 8ZI0, [14]), all of which have been resolved in the absence of any exogenous nucleotides. Such a nucleotide occupancy pattern of the MgADP + Pi and AbF1‐ATPase mixed structure confirms the spontaneous ATP formation observed (Figure 1A). In comparison, MgATP (β TP), Pi (β DP) and ADP (β E) could be clearly assigned in the density map of the structure of conformer C1b (Figure 1C). While nucleotide occupancy of β TP (MgATP) and β E (ADP) as well as their overall structures are similar in C1a and C1b (Figure S4), they differ in the occupancy of β DP (MgADP·Pi or Pi). Importantly, the C‐terminus of β DP of conformer C1b, including the critical 377DIIAILGMDE386‐loop (homologous to the E. coli DELSEED motif) that contributes to the angular velocity of rotary subunit γ [31, 32], swings outwards by 24° (Figures 2A and S5). This continuous movement could be resolved using 3D Flexible Refinement, providing a detailed map as revealed in Figure S6. In parallel, the phosphate‐binding loop (P‐loop) swings outwards (26°) and the interaction between the β‐catch loop residues (D307–310) and the γ‐tether residues γN268 and γN272, which is an important region in the rotary process [31], is weakened and may contribute to the Mg‐ADP release observed (Figure 2A). To balance this swinging motion, the central stalk slightly shifts its foot domain in the direction opposite to the movement of β DP (Figure 2B). However, despite the 24° movement of β DP, subunit γ twists by at most 6° in the CTD. The large movement of the C‐terminus in β DP does not affect its partnering α DP, as shown by the superimposition of α DP of the C1a and C1b structures (Figure S7), which align with a root mean square deviation (RMSD) of 0.403 Å. Independent motion of β DP. (A) Movement of β DP in C1a and C1b. The γ subunit (yellow) and β DP (orange) of C1a are shown. Subunit β DP of C1b (blue) is aligned to β DP of C1a by the N‐terminal domain β‐barrel. The P‐loop is highlighted in the red dashed box, with close‐up views of the aligned structures and their individual P‐loops. The P‐loops of C1a and C1b highlight residues involved in nucleotide stabilization, as well as the respective nucleotides in each subunit. The nucleotides are also shown with their EM map density. The β‐catch‐γ tether features are highlighted with a black dashed box, showing few changes in these features between the two conformers. (B) Structural alteration of C1b central stalk compared to C1a. (top) Movement of subunit γ's N‐ and C‐terminal α‐helices when all structures are aligned by γC1a CTD (γ261–289). (bottom) Comparison between subunit γ movement to β DP movement when the enzyme transitions from C1a to C1b. While the β P in C1b (blue) swings 24° compared to C1a (orange), C1b γ subunit (brown) only twists 6° compared to its C1a counterpart (yellow). Importantly, the large C‐terminal β DP movement from C1a to C1b disrupts the α DP:β DP:γ‐interface in state C1 (Figures 3A,B, 4 and S8) which is formed by polar interactions between βE386 and subunit ε residues εN107 and εD111, the hydrophobic interaction of βI381 with εA115, and subunit α residue αF407, as well as interactions of αF407 and αF410 with εN107. This provides space for the εCTD to transform the entire extended C‐terminal helix into smaller helices and reposition them closer to the εNTD (see details below). Importantly, comparison of states C1a and C1b reveals for the first time an “initiation state” of εCTD's transition from an extended (ε e) to a compact form (ε c), starting with the reorientation of the hook‐like A. baumannii ‐specific C‐terminus ε134–139, responsible for ATP hydrolysis inhibition, and the short α2‐helix (ε129–134) (Figures 3A,B and S8). In concert with these changes are reorientations of residues forming the interface of εNTD and N‐ and C‐terminal γ helices (Figures 3A,B and S8). The initiation state of the extended εCTD. In both (A) C1a and (B) C1b, the εCTD is extended into the catalytic site, however, there are changes at the (right) αβ DP‐γ‐ε interface, (top left) the hook‐like region ε129–139 and (bottom left) the εNTD‐γ foot interface. (top left) In C1a, at the εCTD hook‐like region, major interactions were formed between ε and their surrounding residues, namely α DPE285‐εQ139‐γQ277, εA138‐β TPV270, β TPT307‐εR137‐β TPT309, εN136‐γK269, εI134‐γL270, εK133‐β TPK373, εR132‐β TPD377 and γI7‐εI131‐γL263. When β DP swings outwards in C1b, this region reorients, strengthening certain interactions (green distances) while weakening others (red distances). (right) On the other hand, the αβ DP‐γ‐ε interface changes most drastically. In C1a, many residues in the εCTD (εN107, εA115, εD111, εL118, αF410) are involved in interactions with α DP (αF407 and αF410) and β DP (βI378, βI381, βE386). However, when β DP swings outwards by 24° in C1b, all stabilizing interactions with β DP subunit disappear. Moreover, αF407 now comes close to εA119 (4.8 Å). The AbF1‐ATPase undergoes dynamic rearrangement during work. Main structural alterations within the whole complex to enable the transition from up‐to‐down of εCTD. All structures are shown as tubes, with (top) showing side view and (middle) showing the top view, clipped into the closest εCTD. Only the subunits undergoing transition in response to εCTD rearrangements are shown for clarity. Dashed arrows with different colors indicate structural changes of the corresponding‐colored subunit that occur when the enzyme transitions to the next state. (bottom) View from the membrane showing the degree and direction of twisting of the central stalk. Dashed lines indicate the axis from which the degree of twisting was measured. For each state, the α‐β subunits are shown as transparent cartoons, and the central stalks are shown as solid surfaces, with γ colored yellow and ε colored magenta. Superimposed with that is the central stalk of the following state, shown as transparent surfaces. To confirm the 24° outward movement of β DP's C‐terminus and to investigate the intermediate rearrangements required to reach the position in state C1b, we employed targeted molecular dynamics (TMD) simulations. In these simulations, a harmonic RMSD‐based potential was applied to selected backbone atoms of the protein complex to bias their conformation toward the target structure. During the transition from state C1a to C1b (RMSD ~4.9 Å), this bias was applied to the C‐terminus of β DP over the course of 200 ns (Movie S1, Figure S9A). The C‐terminus of β DP moved in distinct steps away from residues ε107–118, swinging outward by 24° to reach the C1b conformation observed by cryo‐EM (Figure S9B). The loss of interactions between the β DP loop (residues 377–386) and ε was observed only after over 60% of the simulation time, indicating strong interactions between the two subunits in the C1a state (Figure S9C). One–Three–Two: εCTD Helix Formations in Transition A previously unresolved question concerning the ATP hydrolysis inhibitory C‐terminus of subunit ε has been how the transition from an ε e to the ε c form occurs, in which εCTD finally transforms into two helices oriented above the εNTD. This transformation has to occur in parallel with the translocation of the extended helix ε91–125, which is oriented between the α E:β DP pair and ε129–139, located at the α DP:β TP interface toward the bottom of αβ TP (Figure 4). Comparison of the AbF1‐ATPase states C1 and C2a reveals the transition from an extended α‐helix (ε91–125) into three smaller helices formed by residues ε91–98, ε102–108 and ε120–131 (Figures 4 and 5A). This transformation is accompanied by the translocation of ε91–108 at the α E:β DP interface and ε120–139 at α DP:β TP in state C1 to the bottom of αβ TP (state C2a; Figure 4). The three‐helix formation events in state C2a resemble those of the recently described conformation of the ATP hydrolytically active AbF1‐αβγε Δ134–39 mutant and reveal that, in the WT enzyme, the entire εCTD adopts this intermediate structure and position relative to γ ([14]; Figure 5B). Importantly, when state C2a switches to C2b, εCTD transitions from a three‐ to a two‐helical configuration, in which helices ε91–98 and ε102–108 fuse into a single helix spanning ε90–104 (Figure 5B). The latter helix is similar to the final α1 helix of the ε c form. The restructuring of a three‐ to two‐helix state is supported by a 14° twist of the εNTD (Figure S10A). These structural changes also alter the γ–ε interaction, with major changes to the side helix/helices (Figure S10A). In C2b, γ residues γ133, γ134, γ136, γ137, γ155 and γ158 are close to ε91, ε94, ε97, ε98, ε100, ε101 and ε104 of helix ε90–104. These interactions are likely important for the transformation from a two‐ (C2a) to one helix (C2b). Furthermore, the twisting of γ's N‐ and C‐terminal helices increases the distance between β DP and helix ε120–131 (Figure S10B,C), generating space for this helix to escape the β DP‐γ‐εCTD interface in state C2b. Structure comparison of partially extended εCTD from different conformers. (A) The side view of εC2a (magenta) and C2 ε Δ134–139 (pink, transparent, PDB 8ZI2 [14]) aligned by εNTD, together with γC2a (yellow) shows a similar three‐helix arrangement between the two structures. With the exception of the longer unstructured stretch in εC2a (ε132–139), the composition of the helices between two structures is also similar. (B) The side view of εC2a (magenta, transparent) and εC2b (pink) aligned by the εNTD, together with γC2a (yellow), shows major differences. First, C2a εCTD has three helices, composed of ε91–98, ε102–108 and ε120–131. Meanwhile, C2b εCTD arranges into two helices: ε90–104 and ε120–134. Notably, in the 90°‐rotated back view (right), ε120–134 from C2b is positioned further away from its εNTD compared to the ε120–131 of C2a. To facilitate the transition from an extended to a three‐helical and subsequently two‐helical arrangement, the β‐catch loop (24°) and the C‐terminus of β TP in the “tight” C1b conformation move outward (30°) as the β DP state is adopted in C2a (Figures 4 and 6A, Movie S2). This transition is accompanied by unfolding of the β TP helix T182‐S195, which is located close to the P‐loop (Figure 6B). This newly disordered part moves outwards together with the foot domain of α TP, favoring transformation of α TP to the α DP conformation (Figures 4 and 6B). These concerted outward rotations of the αβ TP to αβ DP pair open a pathway for the downward and outward movement of the εCTD (Figure 4, Movie S1). This transition also includes a twisting of the central stalk by 19° in the clockwise (ATP synthesis) direction, without rotation of the N‐ and C‐terminal domains of the γ subunit (Figures 4 and 6B). Concerted rearrangements following the εCTD initiation state. (A) The initiation state triggers β DP formation, making space for εCTD inner transition state. Side view of β DP between C1b and C2a. Red dashed box highlights the changes that induce the catalytic subunit transformation. After the initiation in C1b, interaction between β‐catch loop (orange) and εCTD (magenta) loosens in C1b (Figure 3B) and is lost in C2a, where the εCTD restructures into three helices (pink). The lost interaction is highlighted with a black dashed arrow. This then triggers the outward swinging of 377DIIAILGMDE386 loop by 30°, making space for the third helix of εC2a to form atop the γ's Rossman‐fold (magenta dashed arrow). (B) The restructured β DP182–187 loop in C2a triggers the formation of α DP. Red dashed box highlights the αβ interface. (middle) The restructured loop β DP182–187 moves downward (orange arrow) and pushes helix α DP340–350 and loop α DP376–383 downwards (dark green arrow). These changes induce the swinging of α DP by 20°, transitioning the α TP in C1b to α DP in C2a. Orange dashed box highlights this interaction. (left) The interaction between helix/loop β182–187 with their corresponding α subunit is shown. While nucleotide occupancy alters from state C1b (MgATP (β TP), Pi (β DP) and ADP (β E)) to C2a (ADP (β TP), MgADP (β DP) and ADP (β E)), no changes are observed between states C2a and C2b (Figure S11). Superimposition of C2a and C2b β TP, β DP and β E reveals similar β subunit structures across the two conformers, with pairwise RMSDs of 0.8, 1.9, and 1.2 Å, respectively (Figure S12). To confirm the rearrangements of Abε from an extended helix to smaller helices, and the movement in the foot domain required for the downward and outward motion of the εCTD, a 400 ns TMD simulation was performed. The RMSD between the initial and final states was ~11.7 Å, and the bias was applied exclusively to Abε, while the remaining subunits were allowed to evolve freely without any additional bias or restraints. The simulation (Movie S3, Figure S13A) shows the stepwise transition from an extended εCTD to smaller helices, which move out and below the α₃β₃ headpiece. Although not observed in the experimental structure of state C1b, a small unstructured region between residues 115–117 emerged during the TMD simulation from state C1a to C1b (Figure S13B). This suggests a propensity for this region to unfold, eventually forming the smaller helical segments observed in the subsequent state C1b. The two distinct helical regions (residues 91–108 and 121–131) observed at the end of the simulation were in agreement with the cryo‐EM data. During the downward and outward movement of εCTD, the β TP subunit was displaced. Although β TP was not directly biased during the simulation, it spontaneously adopted its target conformation by the end of the trajectory (Figure S13C). Providing Space at the Bottom of αβ DP for Helical Transformation and Translocation The third class of conformers C3a and C3b were determined at 3.08 and 2.88 Å resolution, respectively, representing the outer transition states in which εCTD had detached from the αβ‐pairs (Figure 1B). Both conformers contain the amino acids α DP, α TP and α E (26–514); β DP, β TP and β E (2–464); γ (2–289) (Figures S2 and S3). In the case of subunit ε in state C3a, residues ε2–106 were resolved, with ε90–106 forming a continuous helix next to γ's Rossman‐fold (Figure S14). This helix is likely stabilized via interactions with γV133, γA136, γT150 and γQ154 and tilted by 13° toward the foot domain of the enzyme when compared to state C2b. In comparison, in state C3b, only the εNTD (ε2–80) could be modeled into the cryo‐EM map (Figure S14). Now located outside and below the α 3:β 3‐hexamer, the very C‐terminal region of ε has fewer interacting partners, providing more flexibility during its transition to the final location above εNTD (ε c) (Figures 4 and 9), as observed in the Abε c solution NMR structure [9]. This increased flexibility may, in part, explain the absence of density needed for confident assignment. Using the Abε c and state C3b structures as references, we performed 600 ns TMD simulations, biasing the C2a coordinates toward these target conformations. The results reveal that ε's two C‐terminal regions move from their position close to the γ‐foot domain in state C3a further downward and above the εNTD, leading to formation of the ε c structure (Movie S4). Analysis revealed that the ε C‐terminal domain underwent a progressive torsional rotation of approximately 55° about the γ/ε‐defined stalk axis over the course of the simulation (Figure S14B). These movements are consistent with rearrangements of εNTD. Different εCTD states and their bioenergetic context. (A) Superimposition of different ε states by alignment at the εNTD (shown as magenta surface). The α‐β subunits are shown as transparent green‐orange surface. Other subunits are shown as tubes and colored as follows: γ (yellow), εC1a (magenta), εC2b (dark pink), εC3b (pink), ε ct (lime, PDB 8HGX [9]). A side view and back view highlight the difference in the structural arrangements between conformers. (B) The numerous Abε's CTD states act as different “gears” to regulate ATP synthesis and hydrolysis. The extended state represents a “low gear” mode to allow the cell's survival under unfavorable conditions by inhibiting ATP hydrolysis while allowing moderate ATP synthesis. In comparison, the inner (states C2a and C2b) and outer transition states (state C3a) represent a “mixed gears” mode as the bacterial cell experiences changes in its conditions and the εCTD undergoes structural changes in response. In favorable conditions, the εCTD transitions into a compact state (C3b and NMR structure [9]) to maximize catalytic activities. (top) The εCTD is depicted as tentpoles with low‐resolution maps of AbF1‐ATPase (PDB ID 7YRY [9]) in the background, showing its up‐to‐down transition. The lime color depicts “under transition” components. (middle) A car model to show different “gearbox” modes: Low gear, mixed gears, and high gear. (bottom) Cryo‐EM models of the AbF1‐ATPase under 12 μM MgADP + Pi, shown as surface. For state C1a‐b and C2a‐b, the models of C1a and C2a were used as representatives, respectively. The εCTD of the NMR structure is colored lime (PDB 8HGX [9]). The question that arises is: which structural changes are required to displace the remaining ε120–139 from the αβγ‐interface? A close look at the transition from state C2b to C3a reveals that the lower domain of β DP moves 30° inwards into its semi‐closed state (Figures 4 and 7A). This is accompanied by reformation of the β182–187 helix, abolishing its interaction with the α376–383 loop and the α340–350 helix, and moving the α DP C‐terminus inwards by 5° (Figure 7B). In addition, the γ‐foot domain swings by about 10 Å and twists the εNTD by 13° (Figure 7C). Interestingly, the direction of twisting is anticlockwise. This change in γ's twisting direction indicates that further twisting in the second intermediate state (C2b) is necessary to generate enough space (Figures 4 and 7C) and to restructure the εCTD from a three‐ to a two‐helical conformation in C2b (Figure 7C), before pulling the two‐helix εCTD downwards. Therefore, after the last εCTD helix has exited the catalytic cavity of αβγ, the central stalk returns to its previous state (C3a and C2a). Importantly, while nucleotide occupancy in β E and β TP did not change in the transition from states C2a to C3b, the unit β DP site changed from ADP‐bound (C2a‐b) to MgATP‐bound in C3a‐b (Figure S11), confirming that ATP was formed in one catalytic site. Concerted rearrangement as εCTD undergoes outer transitions. (A) Changes within the catalytic site β DP during the C2b → C3a transition. After the second helix of εC2b exits its β DP‐γ Rossman‐fold, β DP starts to adopt a “closed” conformation in C3a. The 377DIIAILGMDE386 swings inwards 30°. (B) The closing swing of β DP from C2b to C3a alters its interaction with α DP. Uniquely, α DP still maintains its “open” position. However, the structural rearrangement of β DP has abolished the deeper interaction from β182–187 loop, as it returns to its helical state. The loss of interaction causes α340–350 and loop α337–383 to move upwards, initiating the closing movement of α DP (by 5°). (C) Structural arrangement in εCTD and differences in central stalk between C2b and C3a. The central stalk of both conformers was aligned by the γCTD (γ269–289), which shows (left) a 10° swiveling of the γ‐foot domain and (bottom right) a 13° twist of the εNTD resulting from this movement. The εCTD is highlighted in green and zoomed‐in (top right), showing the helical arrangement of the two conformers. The first helix of εC3a (pink) takes a 13° turn from that of εC2b (magenta). The movements and restructuring of the εCTD during the C3a to C3b transition are accompanied by an inward movement of the α DP C‐terminus (25°), closing the αβ DP bottom domain (Figure 8A,B). This may also, in part, alter the geometry of the αβ DP active site, where the interaction of residues β181 and αR377 with MgATP in state C3b is much tighter compared to those in C3a (Figure 8B). Structural alterations of α DP between states C3a and C3b. The changes in α DP between C3a and C3b result in active site geometry changes in αβ DP. (A) Comparison between α DP of C3a (cyan) and C3b (forest green) by aligning their NTD β‐barrel domain. The bottom of C3a swings inwards by 25°. (B) Changes in geometry of the αβ DP active site. In C3a (right), the interacting residues hold MgATP more lightly compared to C3b (left). The C3a → C3b transitions decrease the distance between βR181 and its nucleotide (3.73 Å → 2.97 Å), and αR377 side chain reorients to bind MgATP (no interaction → 3.31 Å). Discussion Folding and Opening for New Endeavors Although showing low ATP hydrolysis activity, the AbF‐ATP synthase is a dual catalyst and therefore needs a defined regulatory element to stimulate or inhibit ATP synthesis and hydrolysis, depending on metabolic needs. Truncations of Abε's CTD have been shown to reduce cell growth due to decreased ATP synthesis, as confirmed by inverted membrane vesicles (IMVs), and to increase ATP hydrolysis by the AbF1‐ATPase, showing that the very C‐terminal hook‐like ε134–139 region is important for proper ATP formation and optimal regulation of ATP hydrolysis [9]. This fine‐tuned balance of dual enzyme function is relevant not only during periods of high metabolic demand in the pathogen, when ATP provides the major energy source for enzymatic and trafficking processes, but also under conditions of starvation. Furthermore, ATP/ADP homeostasis is fundamental for the regulation of key enzymes within catabolic and anabolic pathways, and ATP‐driven proton transport by the AbF‐ATP synthase contributes to the maintenance of the pmf. The latter drives and regulates influx and efflux pumps, involved in the transport of essential ions as well as drugs, which is often related to antibiotic resistance [33]. Because of its mechanistic and metabolic importance, as well as its relevance for drug discovery, the stepwise processes of stimulation and inhibition mediated by Abε's CTD must be understood. While the ε e form enables the hook‐like ε134–139 to anchor the long ε C‐terminal helix inside the α TP:β TP:γ‐interface to block ATP cleavage [4, 9], loosening these interactions switches the equilibrium to preferred ATP hydrolysis. The present study reveals the first WT AbF1‐ATPase in different states, driven by exogenous nucleotide conditions, capturing snapshots of conformations and structural alterations, particularly in Abε. Moreover, this study presents structural snapshots of an F1‐ATPase converting MgADP + Pi into ATP. Starting with the newly described “initiation state” between states C1a and C1b, where the ε134–139 region alters its interaction at the α TP:β TP:γ‐interface, residues at the εNTD‐γ interface reorientate and residues εN107, εD111 and εD115 lose their interactions with the β DP‐DELSEED region, which moves outward (Figures 3 and S8), as underscored by the MD simulations (Figure S9). Such movements create an exit pathway for the extended α‐helix (ε91–125) to move outside of the α DP:β DP:γ interface. Before that, the εCTD must be restructured from an elongated helix into smaller compact secondary structural elements to pass through the exit pathway, analogous to foldable tent poles (Figure 9A). States C2a and C2b visualize these steps for the first time, in which ε's extended helix reorganizes into three smaller helices (ε91–98, ε102–108 and ε120–131) in C2a (Figure 5B), followed by a two‐helical arrangement (ε90–104 and ε120–134) in C2b (Figures 5B and 9A). We propose that the two‐helical arrangement allows the εCTD to exit the α:β:γ‐headpiece, since these two helices form the CTD, which comes in proximity to the εNTD, consistent with the compact Abε solution NMR structure (Figures 4 and 9, [9]). The local structural transition from a single extended helix to a three‐ and subsequently two‐helical arrangement involves interactions with the rotary partner γ, which twists during these transitions (Figure 4) and brings specific residues of subunits γ and ε into proximity for interaction. Importantly, the TMD simulations provide mechanistic continuity between the experimentally resolved cryo‐EM states. While cryo‐EM captures discrete structural snapshots along the reaction pathway, MD simulations reveal the time‐resolved trajectories connecting these conformations. In each transition (C1a to C1b, C1b to C2a, and C3a to C3b), the simulations reproduced the experimentally observed angular displacements and secondary structural rearrangements of the ε C‐terminal domain, including the 24° outward movement of β DP and the progressive restructuring of the extended εCTD into compact helices. The simulations further demonstrated that these transitions occur in a stepwise and coordinated manner, with spontaneous responses of non‐biased subunits consistent with the cryo‐EM models. Thus, our MD simulations not only validate the structural interpretations derived from cryo‐EM but also provide dynamic insight into the sequence of events underlying εCTD remodeling and central stalk torsion during catalysis. Substrate Induced ATP Formation Intriguingly, the ATP formed in the AbF1‐ATPase in the presence of the substrates MgADP + Pi raises the question: where does the energy for the endergonic ADP + Pi to ATP reaction come from? In aqueous solution, the energy required for ATP synthesis is 31.5 kJ/mol [34], while in the tight binding pocket, that energy is −8.4 kJ/mol [35]. This makes ATP formation spontaneous within a tight binding pocket, while the pmf provides the energy to open the tight interface, release bound ATP, then take in ADP and Pi to continue the rotational ATP synthesis. Therefore, the energy required for ATP formation observed under ATP synthesis conditions comes from the force driving structural rearrangements of the catalytic interface: the central stalk reconfiguration. Evidently, after ADP binding to β DP through the subunit's individual swinging movement (C1a → C1b), the εCTD rearrangement is triggered, to which the entire central stalk twists in concert and eventually induces the tightening of β DP (C2b → C3a), forming ATP. This observation highlights an entirely new paradigm for ATP synthases: the binding of the substrates ADP and Pi induces concerted changes between the central stalk and catalytic site, forming ATP spontaneously and kickstarting the rotational ATP synthesis, which continues due to the energy provided by the pmf. Factors Triggering εCTD Conformational Changes The key question that arises is: what triggers these concerted actions? In the absence of a pmf, the added MgADP + Pi stimulates measurable (Figure 1A) and observable (Figures 1C and S10) ATP‐formation and may, like a starter battery, initiate the molecular engine. While the nucleotide occupancy of β E (ADP) and β TP (MgATP), as well as the overall structures of these catalytic subunits, are similar in C1a and C1b, β DP is occupied by MgADP·Pi in C1a and by Pi in C1b (Figure 1C). The difference in nucleotide occupancy stimulates the concerted movements of the P‐loop, β‐catch loop and β DP 377DIIAILGMDE386‐loop, thereby altering the overall arrangement and interactions of ε's C‐terminal helix and hook region within the α TP:β TP:γ‐interface. The interplay of movements during the transition of the αβ TP → αβ DP pair (C1b → C2a) and structural alterations of β182–187 helix act as a molecular “tin opener” to trigger the one‐to‐three‐to‐two helical transition of εCTD (Figure 6). The similar domains within the newly formed αβ DP pair as well as twisting of the γε‐foot domains promote the release of εCTD from the α 3:β 3‐headpiece during the C2b → C3 transition (Figures 4, 7, and 8). This enables the two C‐terminal helices to rearrange above εNTD to form the ε c conformation, coinciding with MgATP formation in β DP (Figures 8B and S7). Factors governing the transition from the ε c to the ε e state remain a matter of debate [36, 37, 38, 39, 40]. In E. coli , when the F‐ATP synthase is in hydrolyzing mode, the central stalk “tenses up”, bringing the εNTD and ε's first C‐terminal helix closer to the binding site of γ's Rossman‐fold and stabilizing the central stalk [41]. Meanwhile, when driven in the reverse direction, the central stalk relaxes and twists, allowing for a more flexible connection between the F1 and FO domains and favoring a proposed “down” conformation [41]. ATP‐binding to the Ecε [42, 43] and the thermophilic Bacillus ε subunit [44] has been described to regulate the upward and downward movements of its C‐terminal domain. However, Abε does not bind ATP and excludes nucleotide binding to Abε as a regulatory mechanism. An additional factor proposed to contribute to the coupling is the pmf [45]. A recent crosslinking analysis revealed that the ε c state of Ecε was more prevalent under culture conditions that generate a high pmf [46]. In contrast, in the absence of nutrients (low pmf), the Ecε CTD favored the extended state, thereby inhibiting ATP hydrolysis while maintaining moderate levels of ATP synthesis. These data indicate that the Ecε changes conformation in response to environmental conditions and the pmf, in addition to ATP concentration [41, 46], complementing the subunit's ATP‐sensing ability. Thus, εCTD may be described as a “gearbox”, switching between different “operating modes” of the molecular engine, depending on the environmental conditions in which the bacterium resides [46]. While Abε does not bind ATP [9], the A. baumannii enzyme still requires a mechanism(s) to sense environmental changes and respond accordingly. Evidently, the εCTD goes through a partial downward transition when the six C‐terminal residues are removed, allowing for ATP hydrolysis [14]. In the present study, the AbF1‐ATPase was found to respond to MgADP and Pi by initiating ATP formation, which dropped over time (Figure 1A) because of the absence of a pmf, which, like an alternator, keeps the engine in an operational ATP synthesis mode. Importantly, the AbF1‐ATPase in the presence of MgADP and Pi produced a library of in‐transition states, demonstrating that the up‐to‐down transition of Abε's CTD requires multiple structural rearrangements within the catalytic head and central stalk, resulting in a concerted transformation of the entire engine (Figures 4 and 9A, Movies [Link], [Link], [Link], [Link]). Furthermore, our recent studies of the AbF‐ATP synthase εΔ134–139 and AbF‐ATP synthase εΔ126–139 mutants demonstrated that the very C‐terminus is essential for ATP synthesis, with a direct impact on cellular growth [9]. Based on these findings, we propose that the nucleotide conditions at the catalytic sites (starter battery) and the pmf (alternator) drive concerted rearrangements of the A. baumannii enzyme, as revealed by our structures and simulations (Figure 4, Movies [Link], [Link], [Link], [Link]), resulting in extended, compact, and transition forms of Abε. By analogy to the E. coli enzyme, states C1a‐b of the AbF1‐ATPase featuring the extended Abε represent the “low gear” mode that inhibits ATP hydrolysis, conserves ATP, and maintains the pmf as well as ATP/ADP homeostasis (Figure 9B). States C2a‐b and C3a depict a “mixed gears” mode, in which the enzyme responds to changes in the cellular environment and undergoes structural alterations. Lastly, state C3b represents the “high gear” mode, in which nutrient conditions generate a pmf sufficient for continuous ATP synthesis. Such capacity for adaptation to physiological conditions is likely crucial as bacterial cells alternate rapidly between distinct bioenergetic modes during host infection [47] and in response to antibiotic treatment [48], thus making these final and intermediate structures valuable snapshots for novel target and hit identification. Furthermore, the energy from the central stalk twisting and εCTD rearrangement triggers ATP formation, identifying a new paradigm for ATP synthesis. These findings highlight the detailed structural alterations undergone by the dual regulator Abε and underscore the essentiality of the concerted actions of all subunits within the molecular engine of A. baumannii required to make this “gear” work. Transition States Representing Novel Targets Since A. baumannii relies on oxidative phosphorylation catalyzed by its F‐ATP synthase, the detailed atomistic insights into the transition states bring novel inhibitor targets into focus. In particular, the transition state C3a highlights for the first time a stepwise release of εCTD from the γ stalk, with helix 2 exiting before helix 1 (Figure 8). This is a crucial insight to aid the attempts at trapping these transient states and hampering efficient ATP production in A. baumannii . By anchoring both εCTD helices to γ's Rossman‐fold via a molecular glue as observed in CELMoD, a class of molecular glues [49], the critical transition step for ATP synthesis would be blocked. Importantly, since the amino acid composition of these A. baumannii γ‐ε epitopes differs significantly from that in the human counterpart, toxicity effects of such glue would be prevented. Author Contributions Experiment design: Khoa C. M. Le, Gerhard Grüber. Formal analysis: Khoa C. M. Le, Jan K. Marzinek, Peter J. Bond, Gerhard Grüber. Investigation: Khoa C. M. Le, Jan K. Marzinek, Vikneswaran Mathiyazakan. Writing‐original draft: Khoa C. M. Le, Jan K. Marzinek, Peter J. Bond, Gerhard Grüber. Writing – review and editing; Khoa C. M. Le, Jan K. Marzinek, Vikneswaran Mathiyazakan, Peter J. Bond, Gerhard Grüber. Funding acquisition: Peter J. Bond, Gerhard Grüber. Funding This work was supported by PMO | National Research Foundation Singapore (NRF), NRF‐CRP27‐2021‐0002, A*STAR | Bioinformatics Institute (BII). Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Processing pipeline for AbF1‐ATPase under 12 μM MgADP + Pi. (Left) A representative view of the micrographs collected for this dataset, with examples of selected particles circled in yellow. Representative 2D classes are also shown below. (Right) The processing pipeline of the cryo‐EM data is shown. From 5093 movies, 394 279 particles were extracted to refine into 6 conformers. State C1a and C1b were further refined to higher resolution using Flexible refinement (C1a 3.26 Å → 3.13 Å, C1b 3.37 Å → 3.04 Å). The foot domain of β DP from each conformer was highlighted to show the resolved movement. Conformer C2a and C2b show the inner transition state, where εCTD becomes partially extended and resides on the γ's Rossman‐fold. State C3a reveals the first outer transition state, where only the first helix of εCTD was resolved. In the original non‐uniform refined map, the density of this helix was poorly reconstructed. With focused refinement, the density was better resolved. Conformer C3b represents the second outer transition state, where the εCTD undergoes a dynamic restructuring into the compact state. The refined EM maps of each class are shown. Figure S2: Map quality of all conformers. Fourier shell correlation (FSC) curves with FSC‐mask auto‐tightening (left) of all refined density maps and the local resolution estimation computed via CryoSPARC [1]. Figure S3: Model‐to‐map fit of all conformers. Modeling of all six conformers suggests good conformity with the density maps generated. Figure S4: Similarity in β E and β TP subunits from C1a and C1b. When aligned together, the overall structure of these two catalytic sites is similar between C1a (orange) and C1b (blue), with an RMSD of 0.402 and 0.355 Å for β E and β TP, respectively. Figure S5: Comparison between β TP, β E and β DP of C1a and C1b. (top) Three β subunits when aligned by the globular NTD. In C1a, the β TP, β DP and β E adopt a closed, intermediate and open states, respectively. Meanwhile, in C1b, the β DP swings out further than β E. (bottom) Position of the P‐loop (indicated by dashed arrow) and its bound nucleotide from each β subunit. Due to the outwards‐swinging movement of β DP in C1b, the position of its P‐loop is also further away from the “canonical” β DP as observed in C1a. Figure S6: Flexible 3D refinement resolves continuous movement of β DP in C1b. Cartoon side view (left) showing C1b with the movement of β DP highlighted in dash box. Compared to non‐uniform refinement (top right) which resolves the moving β DP to about 6 Å, Flexible refinement (bottom right) resolves the region to 4–4.5 Å. The dashed box highlights the β DP movement in C1b. Density maps and Local resolution maps are shown to highlight the effectiveness of Flexible 3D Refinement [2]. Figure S7: The α DP does not change regardless of β DP swinging. When aligned together, the structure α DP of C1a (green) and C1b (cyan) are similar with an RMSD of 0.403 Å. Figure S8: Alterations in the εNTD and γ foot interface. The εNTD and γ foot interface also alters. Due to the low resolution of the region, residues in the β‐barrel of εNTD cannot be assigned side chains. Therefore, distances measured here are only approximate. In C1a, the following residues are in proximity with each other: γQ41 with εE13; γY45 with εV9‐V11‐L79; γN48‐R51 with εT77; γY208 with εP40‐L41‐V42‐E70; γI209‐εA39; γP212 with εT43; γL217 with εV68‐L69; γR224 with εD81; γY231 with εV11‐K12. When transitioned to C1b, some interactions strengthen (green distance) while others loosen (red distance). For instance, γQ41's side chain moves away from εE13, while that of γY231 moves closer to εV11‐K12. The side chain of γY45 also reorients and becomes closer to εV9‐V11 while moving away from εD81. Meanwhile, residues γR224 and γN48 move closer to their interacting partners, εD81 and εT77, respectively. Residue γY208 side chain also restructures to be closer to εE70, while moving away from εP40‐V42. The sidechains of other residues from γ subunit, γL217‐I209‐P212, all move away and potentially reduce interaction strength with their partners. Figure S9: Targeted MD simulations capture the transition pathway from state C1a to C1b. (A) Snapshots showing the movement of the catalytic β DP subunit (orange). Subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. Representative snapshots correspond to 0 ns (left), ~80 ns (middle), and 200 ns (right). (B) The angular movement of the β DP C‐terminus relative to γ and ε was calculated over the simulation time. (C) The number of contacts between the β DP loop (residues 377–386) and the ε subunit over the simulation time. The conformational changes between the two states were captured in the MD simulations, in agreement with our experimental cryo‐EM data. Figure S10: Structural alteration in the central stalk to enable the three‐two helical transition from C2a to C2b. (A) The side view showing the central stalks of both conformers aligned by γCTD (γ269–289). Subunits are colored as the following: γC2a (yellow), εC2a (magenta), γC2b (lime), εC2b (pink). The interactions between the side helices and γ's Rossman‐fold are highlighted in the red dash box. (left) Close‐up view of the interacting partners from γC2a with the two helices. Residues γV155, γQ154 and γD158 stabilize the first εC2a helix via close interactions with εE91, εI94 and εA97. Meanwhile, the second helix is stabilized by γV133, γA136 and γK127 interacting with εQ108 and εL105. (right) Close‐up view of the interactions between γC2b and εC2b to stabilize its first ε helix. A larger network of interactions between γ133–137, γV155 and γD158 interacts with εL104, εA101, εN100, εR98, εA97, εI94 and εE91. Side chains from γ residues are shown with their EM map densities, while residues on ε are only shown as C‐α due to the lower resolution. (bottom) The bottom view highlights the twisting of central stalk, leading to the rotation of εNTD by 14°. (B) (middle) The side view showing the central stalks of both conformers aligned by γCTD (γ269–289). Subunits are colored as the following: γC2a (yellow), γC2b (goldenrod). The γ subunit was observed to twist as the enzyme transitions from C2a to C2b. The foot domain of γNTD moves by 8°, while that of γCTD moves by 5°. (left and right) bottom view of the γCTD and NTD, respectively, shows the displacement of the foot of each domain as a result of the twisting. (C) The twisting of γ from C2a to C2b makes space between β DP‐ε interface. (top) The distance between C‐α of two furthest residues from the β's 377DIIAILGMDE386 (βL382) and εCTD (εL128) is measured, showing the increased distance as the central stalk twists from C2a to C2b, allowing the third helix of εC2b to escape from the interface. This space was also observed in the map view (bottom). Figure S11: Nucleotide occupation of C2 and C3 conformers. The models are colored as follows: α‐forest green, β‐orange, Mg‐light green. All nucleotides are shown with their corresponding EM map in mesh. Figure S12: Comparison between β TP, β E and β DP of C2a and C2b. (top) Three β subunits when aligned by the globular NTD. In both C2a and C2b, the β TP, β DP and β E adopt a closed, intermediate and open states, respectively. (bottom) Position of the P‐loop (indicated by dash arrow) and its bound nucleotide from each β subunit. Figure S13: Targeted MD simulations capture the transition pathway from state C1b to C2a. (A) Representative snapshots showing the movement of the CTD of Abε (magenta) at 0, 232, 254, and 400 ns (from left to right). The γ subunit is shown in yellow, and the remaining subunits are depicted as transparent gray surfaces. (B) Per‐residue alpha‐helix secondary structure propensity of the ε91–133 region over the two TMD transitions, from C1a to C1b and from C1b to C2a. (C) Initial (0 ns) and final (400 ns) simulation snapshots of the transition from C1b to C2a. The ε subunit is colored based on secondary structure (purple/blue—helix, cyan/white—unstructured). The β TP subunit is shown in orange, and its target structure is represented as a transparent gray cartoon. The dashed arrow indicates the direction of the movement of the ε subunit. Figure S14: The outer transitions C3a and C3b. (A) Model‐to‐map fit of conformer C3a and C3b. The model‐to‐map fit as viewed by two side views of C3a (left) and C3b (right). A green dashed box highlights the single ε helix, which is then zoomed‐in. In C3a, the composition of the side helix was annotated, and γ residues forming possible interactions with the helix were also shown. The density of the helix was able to fit residues ε90–106. These εCTD residues are in proximity to amino acids γV133, γA136, γT150 and γQ154. In C3b, only the ε N‐terminal β‐barrel domain was modeled, with residues ε2–80 fitting into the map. (B) Targeted MD simulations capture the transition pathway from state C3a to C3b. The angular movement of the Abε (residues 90–139) was calculated with respect to the z‐axis defined by the ε/γ stalk. Figure S15: Comparison between β TP, β DP and β E of C3a and C3b. (top) Three β subunits when aligned by the globular NTD. For both C3a and C3b, the β TP and β DP have highly similar “close” states, while β E takes an open state. (bottom) Position of the P‐loop (indicated by dash arrow) and its bound nucleotide from each β subunit. Table S1A: Cryo‐EM data acquisition and image processing of AbF1‐ATPase under 12 μM ADP, 24 μM MgCl2 and 24 μM H2KO4P. Table S1B: Cryo‐EM map statistics of 6 conformers resolved from AbF1‐ATPase under 12 μM MgADP + Pi. Table S1C: Model building, refinement and model statistics of 6 conformers resolved from AbF1‐ATPase under 12 μM MgADP + Pi. Table S2: Summary of all TMD simulations. Data S1: PDB file for conformer C1a. Data S2: PDB file for conformer C1b. Data S3: PDB file for conformer C2a. Data S4: PDB file for conformer C2b. Data S5: PDB file for conformer C3a. Data S6: PDB file for conformer C3b. Data S7: Validation report for conformer C1a, PDB ID 24DH. Data S8: Validation report for conformer C1b, PDB ID 24DJ. Data S9: Validation report for conformer C2a, PDB ID 24DK. Data S10: Validation report for conformer C2b, PDB ID 24DL. Data S11: Validation report for conformer C3a, PDB ID 24DM. Data S12: Validation report for conformer C3b, PDB ID 24DN. Movie S1: Targeted MD simulations capture the transition pathway from state C1a to C1b. The catalytic β DP subunit is shown in orange, while subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. The target structure of C1b is shown in transparent gray cartoon representation. Movie S2: Transition between all conformers based on atomic models. The different atomic models were colored as follows: α subunit (forest green), β subunit (orange), γ subunit (yellow), ε subunit (magenta). The modeled compact εCTD (PDB ID 8HGX) was colored lime. Morphing between two states was carried out using linear interpolation within ChimeraX software. Movie S3: Targeted MD simulations capture the transition pathway from state C1b to C2a. Subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. The target structure of C2a is shown in transparent gray cartoon representation. Movie S4: Targeted MD simulations capture the transition pathway from state C3a to C3b. Subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. The target structure of C3b is shown in transparent gray cartoon representation. Acknowledgments This research was supported by the National Research Foundation (NRF) Singapore, NRF Competitive Research Programme (CRP), Grant Award Number NRF‐CRP27‐2021‐0002 to G.G. J.K.M. and P.J.B. were supported by BII core funds. The simulation work for this article was fully performed on resources of the National Supercomputing Centre, Singapore (https://www.nscc.sg). The authors acknowledge the Cryo‐Electron Microscopy Facility at the Center for Bioimaging Science, Department of Biological Science, National University of Singapore, for the technical assistance in data collection as well as the use of the EM facility at the NTU Institute of Structural Biology, Nanyang Technological University, Singapore. Data Availability Statement All relevant data are contained within the article. Structural data that support the findings of this study are openly available from the Protein Data Bank (https://www.rcsb.org; PDB ID 24DH, 24DJ, 24DK, 24DL, 24DM and 24DN) and the EM Data Bank (https://www.ebi.ac.uk/emdb/; EMDB ID EMD‐69444, EMD‐69446, EMD‐69447, EMD‐69448, EMD‐69449 and EMD‐69450) for conformer C1a, C1b, C2a, C2b, C3a and C3b, respectively. Source data are provided with this paper. References Associated Data Supplementary Materials Figure S1: Processing pipeline for AbF1‐ATPase under 12 μM MgADP + Pi. (Left) A representative view of the micrographs collected for this dataset, with examples of selected particles circled in yellow. Representative 2D classes are also shown below. (Right) The processing pipeline of the cryo‐EM data is shown. From 5093 movies, 394 279 particles were extracted to refine into 6 conformers. State C1a and C1b were further refined to higher resolution using Flexible refinement (C1a 3.26 Å → 3.13 Å, C1b 3.37 Å → 3.04 Å). The foot domain of β DP from each conformer was highlighted to show the resolved movement. Conformer C2a and C2b show the inner transition state, where εCTD becomes partially extended and resides on the γ's Rossman‐fold. State C3a reveals the first outer transition state, where only the first helix of εCTD was resolved. In the original non‐uniform refined map, the density of this helix was poorly reconstructed. With focused refinement, the density was better resolved. Conformer C3b represents the second outer transition state, where the εCTD undergoes a dynamic restructuring into the compact state. The refined EM maps of each class are shown. Figure S2: Map quality of all conformers. Fourier shell correlation (FSC) curves with FSC‐mask auto‐tightening (left) of all refined density maps and the local resolution estimation computed via CryoSPARC [1]. Figure S3: Model‐to‐map fit of all conformers. Modeling of all six conformers suggests good conformity with the density maps generated. Figure S4: Similarity in β E and β TP subunits from C1a and C1b. When aligned together, the overall structure of these two catalytic sites is similar between C1a (orange) and C1b (blue), with an RMSD of 0.402 and 0.355 Å for β E and β TP, respectively. Figure S5: Comparison between β TP, β E and β DP of C1a and C1b. (top) Three β subunits when aligned by the globular NTD. In C1a, the β TP, β DP and β E adopt a closed, intermediate and open states, respectively. Meanwhile, in C1b, the β DP swings out further than β E. (bottom) Position of the P‐loop (indicated by dashed arrow) and its bound nucleotide from each β subunit. Due to the outwards‐swinging movement of β DP in C1b, the position of its P‐loop is also further away from the “canonical” β DP as observed in C1a. Figure S6: Flexible 3D refinement resolves continuous movement of β DP in C1b. Cartoon side view (left) showing C1b with the movement of β DP highlighted in dash box. Compared to non‐uniform refinement (top right) which resolves the moving β DP to about 6 Å, Flexible refinement (bottom right) resolves the region to 4–4.5 Å. The dashed box highlights the β DP movement in C1b. Density maps and Local resolution maps are shown to highlight the effectiveness of Flexible 3D Refinement [2]. Figure S7: The α DP does not change regardless of β DP swinging. When aligned together, the structure α DP of C1a (green) and C1b (cyan) are similar with an RMSD of 0.403 Å. Figure S8: Alterations in the εNTD and γ foot interface. The εNTD and γ foot interface also alters. Due to the low resolution of the region, residues in the β‐barrel of εNTD cannot be assigned side chains. Therefore, distances measured here are only approximate. In C1a, the following residues are in proximity with each other: γQ41 with εE13; γY45 with εV9‐V11‐L79; γN48‐R51 with εT77; γY208 with εP40‐L41‐V42‐E70; γI209‐εA39; γP212 with εT43; γL217 with εV68‐L69; γR224 with εD81; γY231 with εV11‐K12. When transitioned to C1b, some interactions strengthen (green distance) while others loosen (red distance). For instance, γQ41's side chain moves away from εE13, while that of γY231 moves closer to εV11‐K12. The side chain of γY45 also reorients and becomes closer to εV9‐V11 while moving away from εD81. Meanwhile, residues γR224 and γN48 move closer to their interacting partners, εD81 and εT77, respectively. Residue γY208 side chain also restructures to be closer to εE70, while moving away from εP40‐V42. The sidechains of other residues from γ subunit, γL217‐I209‐P212, all move away and potentially reduce interaction strength with their partners. Figure S9: Targeted MD simulations capture the transition pathway from state C1a to C1b. (A) Snapshots showing the movement of the catalytic β DP subunit (orange). Subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. Representative snapshots correspond to 0 ns (left), ~80 ns (middle), and 200 ns (right). (B) The angular movement of the β DP C‐terminus relative to γ and ε was calculated over the simulation time. (C) The number of contacts between the β DP loop (residues 377–386) and the ε subunit over the simulation time. The conformational changes between the two states were captured in the MD simulations, in agreement with our experimental cryo‐EM data. Figure S10: Structural alteration in the central stalk to enable the three‐two helical transition from C2a to C2b. (A) The side view showing the central stalks of both conformers aligned by γCTD (γ269–289). Subunits are colored as the following: γC2a (yellow), εC2a (magenta), γC2b (lime), εC2b (pink). The interactions between the side helices and γ's Rossman‐fold are highlighted in the red dash box. (left) Close‐up view of the interacting partners from γC2a with the two helices. Residues γV155, γQ154 and γD158 stabilize the first εC2a helix via close interactions with εE91, εI94 and εA97. Meanwhile, the second helix is stabilized by γV133, γA136 and γK127 interacting with εQ108 and εL105. (right) Close‐up view of the interactions between γC2b and εC2b to stabilize its first ε helix. A larger network of interactions between γ133–137, γV155 and γD158 interacts with εL104, εA101, εN100, εR98, εA97, εI94 and εE91. Side chains from γ residues are shown with their EM map densities, while residues on ε are only shown as C‐α due to the lower resolution. (bottom) The bottom view highlights the twisting of central stalk, leading to the rotation of εNTD by 14°. (B) (middle) The side view showing the central stalks of both conformers aligned by γCTD (γ269–289). Subunits are colored as the following: γC2a (yellow), γC2b (goldenrod). The γ subunit was observed to twist as the enzyme transitions from C2a to C2b. The foot domain of γNTD moves by 8°, while that of γCTD moves by 5°. (left and right) bottom view of the γCTD and NTD, respectively, shows the displacement of the foot of each domain as a result of the twisting. (C) The twisting of γ from C2a to C2b makes space between β DP‐ε interface. (top) The distance between C‐α of two furthest residues from the β's 377DIIAILGMDE386 (βL382) and εCTD (εL128) is measured, showing the increased distance as the central stalk twists from C2a to C2b, allowing the third helix of εC2b to escape from the interface. This space was also observed in the map view (bottom). Figure S11: Nucleotide occupation of C2 and C3 conformers. The models are colored as follows: α‐forest green, β‐orange, Mg‐light green. All nucleotides are shown with their corresponding EM map in mesh. Figure S12: Comparison between β TP, β E and β DP of C2a and C2b. (top) Three β subunits when aligned by the globular NTD. In both C2a and C2b, the β TP, β DP and β E adopt a closed, intermediate and open states, respectively. (bottom) Position of the P‐loop (indicated by dash arrow) and its bound nucleotide from each β subunit. Figure S13: Targeted MD simulations capture the transition pathway from state C1b to C2a. (A) Representative snapshots showing the movement of the CTD of Abε (magenta) at 0, 232, 254, and 400 ns (from left to right). The γ subunit is shown in yellow, and the remaining subunits are depicted as transparent gray surfaces. (B) Per‐residue alpha‐helix secondary structure propensity of the ε91–133 region over the two TMD transitions, from C1a to C1b and from C1b to C2a. (C) Initial (0 ns) and final (400 ns) simulation snapshots of the transition from C1b to C2a. The ε subunit is colored based on secondary structure (purple/blue—helix, cyan/white—unstructured). The β TP subunit is shown in orange, and its target structure is represented as a transparent gray cartoon. The dashed arrow indicates the direction of the movement of the ε subunit. Figure S14: The outer transitions C3a and C3b. (A) Model‐to‐map fit of conformer C3a and C3b. The model‐to‐map fit as viewed by two side views of C3a (left) and C3b (right). A green dashed box highlights the single ε helix, which is then zoomed‐in. In C3a, the composition of the side helix was annotated, and γ residues forming possible interactions with the helix were also shown. The density of the helix was able to fit residues ε90–106. These εCTD residues are in proximity to amino acids γV133, γA136, γT150 and γQ154. In C3b, only the ε N‐terminal β‐barrel domain was modeled, with residues ε2–80 fitting into the map. (B) Targeted MD simulations capture the transition pathway from state C3a to C3b. The angular movement of the Abε (residues 90–139) was calculated with respect to the z‐axis defined by the ε/γ stalk. Figure S15: Comparison between β TP, β DP and β E of C3a and C3b. (top) Three β subunits when aligned by the globular NTD. For both C3a and C3b, the β TP and β DP have highly similar “close” states, while β E takes an open state. (bottom) Position of the P‐loop (indicated by dash arrow) and its bound nucleotide from each β subunit. Table S1A: Cryo‐EM data acquisition and image processing of AbF1‐ATPase under 12 μM ADP, 24 μM MgCl2 and 24 μM H2KO4P. Table S1B: Cryo‐EM map statistics of 6 conformers resolved from AbF1‐ATPase under 12 μM MgADP + Pi. Table S1C: Model building, refinement and model statistics of 6 conformers resolved from AbF1‐ATPase under 12 μM MgADP + Pi. Table S2: Summary of all TMD simulations. Data S1: PDB file for conformer C1a. Data S2: PDB file for conformer C1b. Data S3: PDB file for conformer C2a. Data S4: PDB file for conformer C2b. Data S5: PDB file for conformer C3a. Data S6: PDB file for conformer C3b. Data S7: Validation report for conformer C1a, PDB ID 24DH. Data S8: Validation report for conformer C1b, PDB ID 24DJ. Data S9: Validation report for conformer C2a, PDB ID 24DK. Data S10: Validation report for conformer C2b, PDB ID 24DL. Data S11: Validation report for conformer C3a, PDB ID 24DM. Data S12: Validation report for conformer C3b, PDB ID 24DN. Movie S1: Targeted MD simulations capture the transition pathway from state C1a to C1b. The catalytic β DP subunit is shown in orange, while subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. The target structure of C1b is shown in transparent gray cartoon representation. Movie S2: Transition between all conformers based on atomic models. The different atomic models were colored as follows: α subunit (forest green), β subunit (orange), γ subunit (yellow), ε subunit (magenta). The modeled compact εCTD (PDB ID 8HGX) was colored lime. Morphing between two states was carried out using linear interpolation within ChimeraX software. Movie S3: Targeted MD simulations capture the transition pathway from state C1b to C2a. Subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. The target structure of C2a is shown in transparent gray cartoon representation. Movie S4: Targeted MD simulations capture the transition pathway from state C3a to C3b. Subunits γ and ε are represented in yellow and magenta, respectively. The remaining subunits are shown as transparent gray surfaces. The target structure of C3b is shown in transparent gray cartoon representation. Data Availability Statement All relevant data are contained within the article. Structural data that support the findings of this study are openly available from the Protein Data Bank (https://www.rcsb.org; PDB ID 24DH, 24DJ, 24DK, 24DL, 24DM and 24DN) and the EM Data Bank (https://www.ebi.ac.uk/emdb/; EMDB ID EMD‐69444, EMD‐69446, EMD‐69447, EMD‐69448, EMD‐69449 and EMD‐69450) for conformer C1a, C1b, C2a, C2b, C3a and C3b, respectively. Source data are provided with this paper.