V1-ATPases are highly conserved ATP-driven rotary molecular motors found in various membrane systems. We recently reported the crystal structures for the Enterococcus hirae A3B3DF (V1) complex, corresponding to the catalytic dwell state waiting for ATP hydrolysis. Here we present the crystal structures for two other dwell states obtained by soaking nucleotide-free V1 crystals in ADP. In the presence of 20 μM ADP, two ADP molecules bind to two of three binding sites and cooperatively induce conformational changes of the third site to an ATP-binding mode, corresponding to the ATP-binding dwell. In the presence of 2 mM ADP, all nucleotide-binding sites are occupied by ADP to induce conformational changes corresponding to the ADP-release dwell. Based on these and previous findings, we propose a V1-ATPase rotational mechanism model.
Ion-transporting rotary ATPases are divided into three types based on their function and taxonomic origin: F-, V- and A-type ATPases. F-ATPases function as ATP synthases in mitochondria, chloroplasts and oxidative bacteria1. V-ATPases function as proton pumps in acidic organelles and plasma membranes of eukaryotic cells2. A-ATPases function as ATP synthases similar to the F-ATPases in Archaea (the ‘A’ designation refers to Archaea), but the structure and subunit composition of A-ATPases are more similar to those of V-ATPases3. These ATPases possess similar overall structures consisting of a globular catalytic domain (F1, V1 or A1) and a membrane-embedded ion-transporting domain (Fo, Vo or Ao). These catalytic domains are similar rotary molecular motors, in which the central axis complexes rotate within pseudo-hexagonally arranged catalytic complexes powered by energy from ATP hydrolysis1,2,3,4,5,6.
The rotational catalysis of F1-ATPase has been investigated using structural analyses of bovine7,8,9,10,11, yeast12,13 and bacterial14,15 samples, and by single-molecule dynamics studies of bacterial samples16,17,18,19,20,21. However, contradictory findings have been obtained depending on the methods, conditions and species, leading to controversy regarding the general rotational model of F1 (refs 10, 18). Recently, a model was proposed that could consistently explain both the structural and single-molecule data obtained for mammalian F1-ATPase22,23. In this model, the central axis rotates at 120° per ATP molecule with three dwell states: waiting for ATP binding (ATP-binding dwell) at 0° (and 120°), waiting for Pi release (Pi-release dwell) at 65°, and waiting for ATP hydrolysis (catalytic dwell) at 90° (see Fig. 1).
Similar V1-ATPase experiments have been conducted using bacterial enzymes from Thermus thermophilus24,25,26,27,28 and E. hirae29,30,31,32,33. These enzymes are sometimes called A-ATPases. However, they are derived from Eubacteria, rather than Archaea3. Furthermore, E. hirae V-ATPase physiologically functions as an ion pump, similar to eukaryotic V-ATPases34,35,36,37, and is composed of nine subunits with amino acid sequences that are homologous to those of the corresponding subunits of eukaryotic V-ATPases6,38. Therefore, we believe the enzyme is a homologue of eukaryotic V-ATPases. We previously established the in vitro expression, purification and crystallization of E. hirae V1-ATPase (EhV1) from the A3B3 and DF complexes29,30. The crystal structures of the nucleotide-free and nucleotide-bound A3B3 (eA3B3 and bA3B3) and V1 (eV1 and bV1) complexes revealed conformational changes of the A3B3 complex induced by the binding of nucleotides and the DF axis (Supplementary Fig. 1), suggesting that the EhV1 structure corresponds to the catalytic dwell waiting for ATP hydrolysis in the rotary cycle31. We have also directly confirmed the unidirectional rotation of EhV1 with single-molecule observations32,33. EhV1 shows only three pausing positions separated by 120° at all ATP concentrations without distinct substeps, in contrast to that of F1-ATPase17,22. This suggests that the ATP hydrolysis step(s), for example, ATP binding, phosphate bond cleavage, ADP release or Pi release, is/are the rate-limiting step(s) in the three-pause rotation39,40. In this study, we performed experiments in which nucleotide-free V1 crystals were soaked with AMP-PNP (non-hydrolysable ATP analog adenosine 5′-(β,γ-imino)triphosphate), ADP or phosphate, and obtained two previously unidentified crystal structures corresponding to the ATP-binding dwell and ADP-release dwell states in the rotary cycle of EhV1. Our proposed rotational mechanism of EhV1 based on these crystal structures is apparently different from those previously reported for F1-ATPases18,19,20,21,22,23 (see Fig. 1).
Structure of the 2AMP-PNP-bound V1 complex
We previously reported that nucleotide-free EhV1 (eV1) is composed of three different conformations of AB pairs: ‘empty’ (AOBC), ‘bound’ (ACBO'), and ‘tight’ (ACRBCR)31 (Fig. 2a,b). A crystal structure of bV1 corresponding to the catalytic dwell state was also obtained by soaking eV1 crystals in mother liquor containing 200 μM AMP-PNP, which is sufficient to inhibit the ATP hydrolysis activity of purified EhV1 (Supplementary Table 1). Two AMP-PNP:Mg2+ molecules were bound to the binding sites of the ‘bound’ and ‘tight’ forms of eV1, which did not cause further conformational changes of the A (Eh-A) and B (Eh-B) subunits, and the overall structure was similar to that of eV1 (root mean square deviation (r.m.s.d.)=0.69 Å)31(Fig. 2c,d). These findings suggest that the binding affinities of AMP-PNP to the ‘bound’ and ‘tight’ forms are high, but that of the ‘empty’ form is low. In this study, in order to verify whether the third ‘empty’ form is able to bind AMP-PNP or not in the presence of a high concentration, we soaked the eV1 crystals with 2 mM AMP-PNP, and solved the crystal structure (denoted 2ATPV1) at a 2.7 Å resolution (Table 1). The structure showed two AMP-PNP molecules bound in the ‘bound’ and ‘tight’ forms and was almost identical to that of bV1 (r.m.s.d.=0.51 Å) (ref. 31) (Fig. 2e,f). No electron density peak for AMP-PNP was found in the ‘empty’ form (Supplementary Fig. 2), indicating that it has a very low affinity for AMP-PNP.
Structure of the 2ADP-bound V1 complex
Next, we soaked the crystals of eV1 in 20 μM ADP, and the crystal structure (denoted 2ADPV1) was solved at a resolution of 3.3 Å (Table 1). Two ADP:Mg2+ molecules were bound to the ‘bound’ and ‘tight’ forms of eV1, as in the case of 2ATPV1 (Supplementary Fig. 3), and induced conformational changes with the crystal packing rearrangements (Supplementary Fig. 4). The structure of 2ADPV1 was validated for possible model bias by generating omit maps of conformationally changed regions (Supplementary Fig. 5). ADP binding changed the structure of eV1, but the crystal packing force might have the potential to distort the actual conformational changes for 2ADPV1.
The structural differences between eV1 and 2ADPV1 that should have been induced by ADP binding are compared in Fig. 3 (see also Supplementary Movie 1). The eV1 ‘bound’ form did not show a conformational change upon ADP binding (r.m.s.d.=0.48 Å; Fig. 3c,d). However, the eV1 ‘tight’ form changed to a more open conformation (AC from ACR; BC’ from BCR) upon ADP binding (Fig. 3e,f). We designated the new ADP-bound ACBC’ pair of 2ADPV1 as the ‘ADP-bound’ form. The γ-phosphate contained in AMP-PNP of the bV1 ‘tight’ form interacted with the Lys238 residue of the P-loop (Pi-binding loop) and the Arg262 residue of the ‘arm’ region (fixed α-helix during the conformational changes: residues 261–275) in Eh-A, and the Arg350 residue (the so-called ‘Arg-finger’ in ATPases) in Eh-B to stabilize the ‘tight’ conformation31 (Fig. 4b), thus preventing any further conformational change. In contrast, ADP, which does not contain γ-phosphate, interacted with these side chains by binding to β-phosphate (Fig. 4a and Supplementary Figs 6 and 7). These different binding contacts induced an apparent conformational change to the ‘ADP-bound’ form (Fig. 4a–c), which seems to be a more stable conformation for the ADP-binding mode than the ‘tight’ form. According to the observed conformational changes, the DF axis became tilted towards the ‘ADP-bound’ form to maintain the extensive protein–protein interactions between DF and the ‘ADP-bound’ form (Fig. 3b–h).
The last conformation of the AB pair (eV1-‘empty’), which did not bind to ADP, also showed a cooperative conformational change. Specifically, Eh-A (AO) and Eh-B (BC) of the ‘empty’ form were attracted to the DF axis and the ‘ADP-bound’ form, respectively (Fig. 3b,g,h and Supplementary Movie 1). The wider conformation of the resultant AB (AO'BO'') pair was most similar to that of the eA3B3-‘bindable’ form (AO'BO: ATP-accessible state) among all AB pairs (r.m.s.d.=0.94 Å) (Supplementary Table 2), and was thus denoted a ‘bindable-like’ form. The structure at the nucleotide-binding site was also more similar to that of eA3B3-‘bindable’ (ATP-accessible state) than to that of eV1-‘empty’ (ATP-unbound state). Similar to the ‘bindable’ form, the topology between the Arg-finger (Eh-B-Arg350) and Eh-A-Arg262 of the ‘bindable-like’ form was more open than that of the ‘empty’ form (Fig. 4d–f; green boxes). Therefore, the ‘bindable-like’ conformation seemed to be able to bind a nucleotide and probably changes to the ‘bound’ form, as observed for eA3B3-‘bindable’31. Based on these findings, we inferred that the structure of 2ADPV1 corresponds to the state of waiting for ATP binding (that is, the ATP-binding dwell) in the rotation.
Structure of the 3ADP-bound V1 complex
Next, we soaked the eV1 crystals in a high concentration (2 mM) of ADP to verify nucleotide binding to the third ‘bindable-like’ form of 2ADPV1, and obtained the crystal structure (denoted 3ADPV1) at a resolution of 3.0 Å (Table 1). Three ADP:Mg2+ molecules were bound at all three nucleotide-binding sites (Supplementary Fig. 8) and induced conformational changes with the crystal packing rearrangements (Fig. 5a, Supplementary Fig. 4, and Supplementary Movie 2). The structure was verified by generating omit maps of conformationally changed regions (Supplementary Fig. 5), although it is possible that the crystal packing force will distort the actual conformational changes for 3ADPV1. The structural differences between the 2ADP-bound (2ADPV1) and 3ADP-bound (3ADPV1) V1 complexes, which are considered to be induced by ADP binding to the ‘bindable-like’ form of 2ADPV1, are compared in Fig. 5b–h (see also Supplementary Movie 3). ‘Bindable-like’ Eh-A (AO') changed to the half-closed conformation (denoted AHC) upon ADP binding, whereas the B subunit (BO'') did not (Fig. 5c,d). We thus designated the unique half-closed AHCBO'' pair of 3ADPV1 as the ‘half-closed’ form. A strong electron density peak for Pi or SO42− (a Pi analog) was observed at the nucleotide-binding site with ADP:Mg2+ (Fig. 6a, and Supplementary Figs 7 and 8). We assigned this peak to SO42− because 3 mM MgSO4, but not Pi, was contained in the crystallization condition. The adjacent ‘bound’ form was not affected by this conformational change to the ‘half-closed’ form. In contrast, the DF axis and ‘ADP-bound’ forms were slightly attracted to the AHC of the ‘half-closed’ form (Fig. 5b–h). The shifted ‘ADP-bound’ form was rather more similar to the observed ‘tight’ conformation (Fig. 5g,h). Furthermore, the nucleotide-binding site was also more similar to that of the ‘tight’ form than to that of the ‘ADP-bound’ form (Fig. 6b–d, and Supplementary Fig. 7). We, therefore, designated this shifted ‘ADP-bound’ form of 3ADPV1 as the ‘tight-like’ form. The distances between the β-phosphate of ADP and the interacting residues in the ‘tight-like’ form were slightly longer than those in the ‘ADP-bound’ form (Supplementary Fig. 6), suggesting that the binding affinity for ADP of the ‘tight-like’ form is lower than that of the ‘ADP-bound’ form. Consequently, an ADP molecule will be easily released from the binding site. Therefore, we inferred that the structure of 3ADPV1 corresponds to the state of waiting for ADP release (that is, ADP-release dwell) in the rotation. ATP hydrolysis activity of purified EhV1 was inhibited at a high (2 mM) concentration of ADP (Supplementary Table 1), which is significantly higher than the natural concentration in E. hirae cells. Therefore, the ADP-release dwell state might be a minor intermediate state, which might exist in the catalytic cycle with high [ADP] and low [ATP].
Structure of the Pi-bound V1 complex
Next, we soaked the eV1 crystals with 20 and 200 μM Pi, and solved the crystal structures at a resolution of 3.0 Å (0PiV1:20 μM) and 2.8 Å (0PiV1:200 μM), respectively (Table 1). However, no electron density peak for Pi, nor any conformational change was observed for either of the structures (Supplementary Fig. 9), and ATP hydrolysis activity of EhV1 was not inhibited, even in the presence of 20 mM Pi (Supplementary Table 1). These findings suggest that the binding affinity for Pi is lower than that of either AMP-PNP or ADP. We further soaked the crystals in a higher concentration (2 mM) of Pi, and the crystal structure (denoted 1PiV1) was determined at a resolution of 2.9 Å (Fig. 7 and Table 1). A Pi molecule with Mg2+ was found in the ‘tight’ form, which was fixed by the Arg-finger as observed for the binding of the γ-phosphate of AMP-PNP in bV1 (Fig. 7 and Supplementary Fig. 7). Importantly, no conformational change was observed upon Pi:Mg2+ binding (Fig. 7b); the overall structure was similar to those of eV1 (r.m.s.d.=0.46 Å) and 2ATPV1 (r.m.s.d.=0.47 Å), but not to that of 2ADPV1 (r.m.s.d.=2.05 Å) or 3ADPV1 (r.m.s.d.=2.78 Å). Thus, the soaking of eV1 crystals in Pi did not induce conformational changes, as in the case of 2ATPV1 after soaking with AMP-PNP. We also soaked the crystals of eV1 in the mixture of various concentrations of AMP-PNP, ADP and/or Pi to obtain other intermediate states. However, diffraction of these soaked crystals was not sufficient to solve the structure. Careful optimization of the ligand concentrations and crystal soaking times are necessary to improve the resolutions.
Binding affinities of nucleotide to V1 complex
We performed isothermal titration calorimetry (ITC) experiments to estimate the binding affinities of AMP-PNP to nucleotide-free EhV1. Exothermic reactions were observed upon the addition of AMP-PNP (Fig. 8a). The binding isotherm was saturated for titrations to an AMP-PNP/EhV1 molar ratio of 2.2 (∼14 μM AMP-PNP); no additional binding was observed for titrations up to the molar ratio of 1,400 (2.4 mM AMP-PNP) (Supplementary Fig. 10). The curve was fit by the two sets of sites model with the following parameters. The numbers (n1 and n2) of binding sites per EhV1 were 0.68 and 0.77, respectively. The Kd1 and Kd2 values were 9.4 and 40 nM, respectively. The ΔH1 and ΔH2 values were −9.3 and −9.7 kcal mol−1, respectively. The ΔS1 and ΔS2 values were 5.5 and 1.4 cal mol−1 per degree, respectively. These ITC data suggested that the binding affinities of AMP-PNP to the ‘bound’ and ‘tight’ forms were both high, and that of the third ‘empty’ form was very low (<2 mM), corresponding to the structural findings described above.
Next, we quantified the binding affinities of ADP to nucleotide-free EhV1 using ITC. The binding isotherm for ADP titration was remarkably different from that for AMP-PNP titration, and showed three distinct zones (Fig. 8b). The first zone, below an ADP/EhV1 molar ratio of 2, was characterized by a continuous decrease in the exothermic signal. The second zone, between ADP/EhV1 molar ratios of 2 and 2.8, exhibited the opposite trend, with an increase in the exothermic signal throughout the titration. Finally, in the third zone, the exothermic signal decreased as the ADP/EhV1 molar ratios increased from 2.8 to 5.6 (a ratio at which saturation was reached), and no additional exothermic signal was observed for titrations up to an ADP/EhV1 molar ratio of 1,400 (2.4 mM ADP) (Supplementary Fig. 10). Thus, the triphasic curve, which was likely to contain three different binding reactions, was analysed using the three sets of sites model41 with the following parameters. The numbers (n1, n2 and n3) of binding sites per EhV1 were 1.4, 0.82 and 0.65, respectively. The Kd1, Kd2 and Kd3 values were 6.7 nM, 13 nM and 3.6 μM, respectively. The ΔH1, ΔH2 and ΔH3 values were —4.3, 1.3 and −10 kcal mol−1, respectively. The ΔS1, ΔS2 and ΔS3 values were 23, 41 and -10 cal mol−1 per degree, respectively. Interestingly, ADP binding to site-2 showed an endothermic reaction, whereas those to site-1 and site-3 involved exothermic reactions. This implies a dynamic structural change with ADP binding to site-2, corresponding to the structural findings that ADP binding to the ‘tight’ form induces conformational changes to the ‘ADP-bound’ form and the adjacent ‘empty’ form then changes to a ‘bindable-like’ form in a cooperative manner. According to this interpretation, site-1, -2 and -3 correspond to the ‘bound’, ‘tight’ and ‘bindable-like’ forms of EhV1, respectively.
We also performed displacement ITC experiments of ADP-bound and AMP-PNP-bound EhV1 by addition of AMP-PNP and ADP, respectively. The titration experiment of AMP-PNP into 3ADP-bound EhV1 with 35 μM ADP (saturated concentration) showed that the exothermicity was remarkably lower than that into nucleotide-free EhV1 (Fig. 8a,c). No noticeable exothermic signal was observed for titrations up to an AMP-PNP/EhV1 molar ratio of 600 (2.4 mM AMP-PNP) (Supplementary Fig. 10). This suggests that AMP-PNP binding sites are already occupied by ADP in 3ADP-bound EhV1, and these nucleotides competitively bind to EhV1. In the competitive displacement experiment, the apparent Kd values for AMP-PNP to 3ADP-bound EhV1 were expected to be very high42, and were actually estimated to be very high. The weak binding of AMP-PNP yields a nearly horizontal trace in binding isotherm whether or not the real displacement of nucleotides takes place43. Therefore, it is difficult to investigate the exchange reaction precisely from the ITC data. Similarly, the titration experiment of ADP into 2AMP-PNP-bound EhV1 with 21 μM AMP-PNP (saturated concentration) also showed that the exothermicity was remarkably lower than that into nucleotide-free EhV1 (Fig. 8b,d), and no noticeable exothermic signals were observed for titrations up to an ADP/EhV1 molar ratio of 600 (2.4 mM ADP) (Supplementary Fig. 10); It is also predicted the exothermic signal should be very small because the apparent Kd values for ADP to 2AMP-PNP-bound EhV1 were estimated very high42,43. This finding suggests that ADP is not able to bind to the ‘empty’ form of 2ATPV1 owing to low affinity, as in the case of AMP-PNP. Therefore, the ‘half-closed’ form of 3ADPV1 seems to be obtained by ADP binding to the ‘bindable-like’ form of 2ADPV1, but not to the ‘empty’ form of eV1 in the soaking experiment of ADP to eV1 crystals, consistent with the structural findings in this study.
Tryptophan fluorescence change of V1 complex
Tryptophan fluorescence is very sensitive to conformational changes in proteins44. In order to verify the conformational change induced by ADP binding, which was observed by X-ray crystallography of ADP-soaked crystals, we measured the tryptophan fluorescence of EhV1 in the presence of AMP-PNP and/or ADP. Eh-A and Eh-B subunits of EhV1 have 8 and 1 tryptophan residues, respectively. Emission spectra of the intrinsic tryptophan fluorescence of EhV1 without added nucleotides showed a peak at 335 nm (Supplementary Fig. 11). The fluorescence intensity around 335 nm had distinct increase (2.3±0.2 a.u.) by the addition of 500 nM AMP-PNP (a higher concentration than estimated Kd values for AMP-PNP by ITC) (Fig. 8e, lane 1). However, the overall structure of 2ATPV1 was very similar to that of eV1 (Fig. 2). Therefore, we attributed this change in intensity to a side-chain shift of the Trp248 residue near the P-loop from AMP-PNP binding, rather than overall conformational changes of EhV1 (Supplementary Fig. 12). The fluorescence intensity was not affected by the re-addition of AMP-PNP (21 and 100 μM) (Fig. 8e, lane 2 and 3), consistent with the ITC data. On the other hand, the addition of 500 nM ADP (a higher concentration than the two Kd values for ADP and lower than the third Kd value for ADP) induced fluorescence change (3.1±0.2 a.u.), which was higher than that of AMP-PNP (Fig. 8e, lane 4). This fluorescence increase is consistent with the conformational changes observed for the 2ADP-bound crystal structure (2ADPV1) (Supplementary Fig. 12). Addition of 35 μM ADP (a higher concentration than the third Kd value for ADP) induced further changes in fluorescence intensity (Fig. 8e, lane 5), and the intensity was not affected by the addition of ADP at a higher concentration (that is, 100 μM ADP) (Fig. 8e, lane 6). This intensity change might correspond to the conformational changes to 3ADPV1 by ADP binding to the third ‘bindable-like’ form of 2ADPV1 (Supplementary Fig. 12).
We also performed competitive displacement experiments of ADP-bound and AMP-PNP-bound EhV1 by addition of AMP-PNP and ADP, respectively. When 2 mM AMP-PNP was added into 3ADP-bound EhV1 pre-incubated with 35 μM ADP (saturated concentration), the fluorescence intensity decreased rapidly (Supplementary Fig. 11), and reached an equilibrium within 5 min (2.3±0.1 a.u.; change from nucleotide-free EhV1) (Fig. 8f, lane 1), which was very similar to that of AMP-PNP bound EhV1 (Fig. 8e, lanes 1–3). Similarly, when 2 mM ADP was added into 2AMP-PNP-bound EhV1 pre-incubated with 21 μM AMP-PNP (saturated concentration), the fluorescence intensity increased slowly to 3.6±0.1 a.u. (change from nucleotide-free EhV1) (Fig. 8f, lane 2 and Supplementary Fig. 11), which was very similar to that of 3ADP-bound EhV1 (Fig. 8e, lanes 5 and 6). These findings suggest that EhV1 is able to bind AMP-PNP and ADP at two or three binding sites competitively and to reversibly change the conformations.
We previously reported that the structure of bV1 represents the catalytic dwell state (that is, a state of waiting for ATP hydrolysis) in which two ATP analogs (AMP-PNP) are bound, one in the ‘bound’ and the other in the ‘tight’ form. ATP hydrolysis is thought to occur in the ‘tight’ form due to induction caused by the approach of the Arg-finger31. However, the details of the reaction after hydrolysis remain unresolved. A new ATP molecule is unable to bind to the ‘empty’ form owing to its low affinity. In order for the reaction to continue, certain structural changes in the ‘tight’ form need to be induced via conversion to ADP and Pi. In this study, we solved the crystal structures of the 2ADP-bound V1 complex (2ADPV1) by soaking eV1 crystals in 20 μM ADP. The ‘tight’ form changed to the ‘ADP-bound’ form, and cooperatively induced conformational changes from the ‘empty’ to ‘bindable-like’ forms. In contrast, soaking with 20 and 200 μM Pi did not produce an electron density peak for Pi or any conformational change. Thus, ADP has a much higher binding affinity to the ‘tight’ form than does Pi. In the presence of 2 mM Pi, Pi:Mg2+ was bound to the ‘tight’ form, but this binding did not induce conformational changes. Therefore, we concluded that Pi is released first after ATP hydrolysis, which changes the conformation to the ATP-binding dwell state (2ADPV1). If ADP was released first, the conformational change required to continue the rotational reaction would not be induced, as observed for the 1PiV1 structure.
In contrast, the mammalian F1-ATPase reaction model proposes that ADP is released first after hydrolysis (Fig. 1). This ADP-release-first order has also been observed for single-molecule manipulation of thermophilic F1 (refs 19, 20) and in molecular dynamics simulations based on the crystal structures of eukaryotic F1 (refs 45, 46, 47). However, the release order is still uncertain owing to several inconsistent results among structural and single-molecule studies10,21. Regardless of the order, it is clear that the catalytic cycles of F1 and V1-ATPases differ substantially (see Fig. 1).
EhV1 showed three 120° steps in a 360° rotation without apparent substeps in our single-molecule dynamics studies32,33. This rotational cycle pattern is similar to the rotational steps observed for T. thermophilus V1 (ref. 25), whereas mammalian F1 rotation involves two rotational substeps for 120° steps (Fig. 1). In this study, we obtained the crystal structures of three different states in the rotational cycle, corresponding to the catalytic dwell (2ATPV1), ATP-binding dwell (2ADPV1), and ADP-release dwell (3ADPV1) states. There are several potential explanations for why these states were not distinguished in our singe-molecule analysis. First, there were clear structural differences between the catalytic dwell (eV1=2ATPV1) and ATP-binding dwell (2ADPV1). The DF axis of 2ADPV1 did not rotate significantly, but was instead tilted toward the ‘ADP-bound’ form owing to the conformational changes induced by the binding of ADP to the ‘tight’ form of eV1 (see Supplementary Movie 1). Such a tilt of DF without apparent rotation would be difficult to detect using the single-molecule observations as an additional substep32,33. Second, there were clear structural differences between the ATP-binding dwell (2ADPV1) and the ADP-release dwell (3ADPV1). The DF axis of 3ADPV1 was slightly bent towards the ‘tight-like’ form, but did not induce any rotational changes (see Supplementary Movie 3). This small shift in DF without apparent rotation would also be difficult to be detected using our single-molecule observations as additional substeps32,33. These findings suggested that EhV1 exists in at least three dwell states in the 120° rotation without any rotational substeps. Thus, although the number of dwell states in EhV1 and mammalian F1 appears to be the same, these V1 and F1 motors show clear differences in the release order of cleavage products, rotational arrest points, dynamics and conformational changes.
Three-dimensional structures for three rotational states of the whole V-ATPase complex of Saccharomyces cerevisiae have been obtained by electron cryo-microscopy48. The samples for this analysis were obtained in the absence of nucleotides during the purification procedures. Therefore, the V1 part of the three structures (PDB number: 3J9T, 3J9U and 3J9V) seemed to correspond to the nucleotide-free form. These three structures of yeast V1 are comparable to those of EhV1, and are the most similar to that of eV1 (the nucleotide-free form of EhV1 corresponding to the catalytic dwell state), although the tilts of these DF complexes are different (Supplementary Fig. 13). Thus, nucleotide-free V1-ATPases seem to form the catalytic dwell state, rather than the ATP-binding dwell and ADP-release dwell states. Recently, the crystal structure of S. cerevisiae V1-ATPase has been obtained at a 6.2 Å resolution49. This structure appears to be an inhibitory state wherein the subunit H inhibits the ATPase activity by stabilizing ADP binding to the catalytic site. We compared these structures of yeast V1 and EhV1, and found a lack of similarity between the yeast structure and EhV1 structures (2ATPV1, 2ADPV1, 3ADPV1), suggesting that the yeast V1 structure provides a unique view of an inhibitory state of a eukaryotic V-ATPase (Supplementary Fig. 13).
Finally, we propose a potential model to describe the rotation mechanism of E. hirae V1-ATPase based on the observed crystal structures and single-molecule observations32,33. Figure 9 shows the 120° rotation model starting from the catalytic dwell (Fig. 9a), in which the surface structure of the C-terminal domain of 2AMP-PNP-bound V1 (2ATPV1) is depicted, and the two ATP molecules are bound to the ‘bound’ and ‘tight’ forms (see Supplementary Fig. 14 and Supplementary Movie 4 for the 360° rotation model). The ATP that is tightly bound to the ‘tight’ form is hydrolysed to produce ADP and Pi31. The Pi molecule, which has a lower affinity than ADP, is released, inducing a change from the ‘tight’ to ‘ADP-bound’ form. Consequently, the DF axis tilts toward the ‘ADP-bound’ form, but this does not induce a rotational event of the DF axis. The adjacent ‘empty’ form (ATP-unbound state) then changes to a ‘bindable-like’ form (ATP-accessible state) in a cooperative manner (Fig. 9b: ATP-binding dwell). Next, a new ATP molecule binds to the ‘bindable-like’ form, which induces a conformational change to the ‘bound’ form, thereby releasing the bound ADP from the ‘ADP-bound’ form. If the ADP stays in the ‘ADP-bound’ form, the ‘bindable-like’ form will become ‘half-closed’ due to ATP binding (more specifically, due to ADP:Mg2+ binding with SO42−, as shown in Fig. 6a), which is accompanied by a small shift of the DF axis, but no apparent rotational substep. Consequently, the adjacent ‘ADP-bound’ form cooperatively returns to the ‘tight-like’ conformation, and the binding affinity for ADP is reduced, as described above (Fig. 9c: ADP-release dwell). Then, ADP is released from the ‘tight-like’ form, and the ‘half-closed’ form is converted to the ‘bound’ conformation. Following these conformational changes, the DF axis rotates 120° with a torque of ∼25 pNnm (ref. 33), and conformational changes from ‘tight-like’ to ‘empty’ and from ‘bound’ to ‘tight’ occur as a result of protein–protein interactions with the DF axis31. Finally, the enzyme resumes its initial catalytic dwell state, shown in Fig. 9a (Fig. 9d). Thus, the V1 motor achieves its rotational dynamics via several conformational changes that are generated by the binding of ATP, release of Pi and ADP, and cooperative coupling among these conformational changes. We will refine our model further by performing a combination of structural, single-molecule and computational analyses to fully understand the operational mechanism of EhV1.
An Escherichia coli cell-free protein expression system was used to synthesize the A3B3 and DF complexes using a mixture of plasmids harbouring the corresponding genes with a modified natural poly-histidine (MKDHLIHNHHKHEHAHAEH) affinity tag, tobacco etch virus cleavage site (EHLYFQG) and linker (SSGSSG) sequences at the N terminus29,30,31. The reacted cell-free lysate was loaded onto a HisTrap HP column (GE Healthcare, Little Chalfont, UK) equilibrated with buffer-A (50 mM Tris-HCl, 750 mM NaCl, 5 mM 2-mercaptoethanol, and 10 mM imidazole; pH 8.0), and bound proteins were eluted with buffer-B (50 mM Tris-HCl, 300 mM NaCl, 5 mM 2-mercaptoethanol, and 500 mM imidazole; pH 8.0). The sample buffer was exchanged to buffer-A using a HiPrep 26/10 desalting column (GE Healthcare, Little Chalfont, UK). To obtain non-tagged samples, the proteins were treated with tobacco etch virus protease at 4 °C for 12 h. The reaction solution was loaded onto a HisTrap HP column again, and the flow-through fractions containing the non-tagged proteins were pooled. The protein samples were dialysed against buffer-C (50 mM Tris-HCl, 10 mM NaCl, 5 mM 2-mercaptoethanol; pH 8.5), loaded onto a HiTrap Q HP column (GE Healthcare, Little Chalfont, UK) equilibrated with buffer-C and eluted with a linear gradient of 10–1,000 mM NaCl. Finally, the concentrated samples with an Amicon Ultra 30 K unit (Merck Millipore, Darmstadt, Germany) were loaded onto a HiLoad 16/60 Superdex 200 pg column (GE Healthcare, Little Chalfont, UK) equilibrated with buffer-D (20 mM Tris-HCl, 150 mM NaCl, and 2 mM dithiothreitol; pH 8.0) and eluted using buffer-D. The purified A3B3 and DF complexes were concentrated with an Amicon Ultra 30 K unit.
V1-ATPase (A3B3DF) was reconstituted and purified as follows: purified A3B3 and DF in buffer-D were mixed in a 1:5 molar ratio with the addition of MES (100 mM final concentration; pH 6.0) and incubated with 0.2 mM AMP-PNP and 5 mM MgSO4 for 1 h. Reconstituted V1-ATPase was purified using a HiLoad 16/60 Superdex 200 pg column equilibrated with buffer-E (20 mM MES, 10% glycerol, 100 mM NaCl, 5 mM MgSO4, and 2 mM dithiothreitol; pH 6.5). Purified complexes were concentrated with an Amicon Ultra 30 K unit.
Crystals of nucleotide-free A3B3DF (eV1) were obtained by mixing 0.1 μl of 8 mg ml−1 purified V1-ATPase in buffer-E (see previous section) with 0.1 μl of reservoir solution (0.1 M Bis-Tris propane (pH 6.5–7.5), 20–22% polyethylene glycol (PEG)-3350, and 0.2 M NaF), using the sitting-drop vapour diffusion method at 296 K. The crystals were soaked in the following conditions (i–vi), mounted on cryo-loops (Hampton Research, Aliso Viejo, CA, USA), flash-cooled, and stored in liquid nitrogen.
V1-ATPase soaked with 2 mM AMP-PNP (2ATPV1): The eV1 crystals were soaked for 6.5 h in 0.1 M Bis-Tris propane (pH 6.5), 21% PEG-3350, 2 mM AMP-PNP, 3 mM MgCl2, 0.28 M NaCl and 20% glycerol.
V1-ATPase soaked with 20 μM ADP (2ADPV1): The eV1 crystals were soaked for 4.5 h in 0.1 M Bis-Tris propane (pH 6.5), 21% PEG-3350, 20 μM ADP, 3 mM MgSO4, 0.28 M NaCl, and 20% glycerol.
V1-ATPase soaked with 2 mM ADP (3ADPV1): The eV1 crystals were soaked for 4.5 h in 0.1 M Bis-Tris propane (pH 6.5), 21% PEG-3350, 2 mM ADP, 3 mM MgSO4, 0.28 M NaCl and 20% glycerol.
V1-ATPase soaked with 20 μM Pi (0PiV1:20 μM): The eV1 crystals were soaked for 5.5 h in 0.1 M Bis-Tris propane (pH 7.5), 21% PEG-3350, 20 μM sodium phosphate, 3 mM MgCl2, 0.28 M NaCl and 20% glycerol.
V1-ATPase soaked with 200 μM Pi (0PiV1:200 μM): The eV1 crystals were soaked for 5.0 h in 0.1 M Bis-Tris propane (pH 6.5), 21% PEG-3350, 200 μM sodium phosphate, 3 mM MgCl2, 0.28 M NaCl and 20% glycerol.
V1-ATPase soaked with 2 mM Pi (1PiV1): The eV1 crystals were soaked for 5.0 h in 0.1 M Bis-Tris propane (pH 6.5), 21% PEG-3350, 2 mM sodium phosphate, 3 mM MgCl2, 0.28 M NaCl and 20% glycerol.
All X-ray diffraction data were collected from a single crystal at a cryogenic temperature (100 K) at the Photon Factory (Tsukuba, Japan). The collected data were processed using XDS50 or HKL2000 software (HKL Research, Inc., Charlottesville, VA, USA). The structures were solved by molecular replacement with Phaser51 or MOLREP52, using the crystal structures of bV1, eV1, 2ADPV1, bV1, 1PiV1 and bV1 as a search model for 2ATPV1, 2ADPV1, 3ADPV1, 0PiV1:20 μM, 0PiV1:200 μM and 1PiV1, respectively. The atomic models were built using Coot53, cross-validated by making various omit maps to minimize model bias, and iteratively refined using REFMAC5 (ref. 54) and Phenix55. TLS (Translation/Libration/Screw) refinement was performed in late stages of refinement. The refined structures were validated with RAMPAGE56. For the structures of 2ATPV1, 2ADPV1, 3ADPV1, 0PiV1:20 μM, 0PiV1:200 μM and 1PiV1, 99.9, 99.9, 99.9, 100, 99.9 and 99.9% of the residues, respectively, were in favoured or allowed regions based on a Ramachandran analysis. The crystallographic and refinement statistics are summarized in Table 1. All r.m.s.d. values were calculated using Cα atoms. The r.m.s.d. values for the superimpositions for each AB pair in the crystal structures are listed in Supplementary Table 2. Figures were prepared using PyMOL (The PyMOL Molecular Graphics System, Version 1.3, Schrodinger, LLC, New York, NY, USA).
Measurement of ATPase activity and protein concentrations
ATPase activity of the purified V1-ATPase in the presence of AMP-PNP or ADP was measured by the colorimetric method using molybdic acid57,58. The reaction was initiated by the addition of 1 mM ATP, after a 10 min pre-incubation with various concentrations of AMP-PNP or ADP, and terminated by the addition of 10% sodium dodecyl sulphate. The initial rate of the ATPase reaction at 23 °C was determined within 4 min, and the measurement was repeated three times. ATPase activities of the purified V1-ATPase in the presence of various concentrations of sodium phosphate were measured using an ATP regenerating system31,59. ATP hydrolysis rates at 23 °C were determined in terms of the rate of NADH oxidation, which was measured as a decrease in absorbance of 340 nm, and the measurement was repeated three times. Protein concentrations were determined using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Inc., Waltham, MA, USA) with bovine serum albumin as the standard.
Isothermal titration calorimetry (ITC)
V1-ATPase was prepared by mixing 12 μM A3B3 and 60 μM DF in 900 μl of buffer-D, and the suspended sample buffer was replaced with buffer-F (100 mM Tris-HCl, 100 mM NaCl, and 5 mM MgSO4; pH 7.5) using Spectra/Por 3 Dialysis Tubing (Spectrum Laboratories, Inc., Rancho Dominguez, CA, USA). ITC experiments were performed using the MicroCal iTC200 calorimeter (Malvern Instruments Ltd., Malvern, Worcestershire, UK), and the samples (7 μM) with/without 21 μM ANP-PNP or 35 μM ADP were loaded into the sample cell. Either 200 μM AMP-PNP or ADP in buffer-F was injected into the sample cell at 25 °C using one initial injection of 1.0 μl followed by 18 injections of 2.0 μl. Binding data were fitted to the two sets of sites model using Origin 7.0 (MicroCal) or the three sets of sites model using MATLAB41.
V1-ATPase was prepared by mixing 100 nM A3B3 and 500 nM DF in buffer-F in 1.2 ml. Fluorescence experiments were performed using the FP-6500 spectrofluorometer (JASCO, Tokyo, Japan) at 25 °C. Fluorescence spectra were recorded with excitation at 300 nm (slit width 1 nm) and emission between 310 and 450 nm (slit width 20 nm). Time courses of exchange reactions of AMP-PNP and ADP were measured every 0.5 s at 335 nm with excitation at 300 nm. Time courses were averaged for 20 data points around each point.
Coordinates and structure factors for the ADP- and Pi-bound V1-ATPase complexes have been deposited in the Protein Data Bank under the accession codes 5KNB (doi: 10.2210/pdb5knb/pdb; 2ADPV1 at 3.3 Å) (ref. 60), 5KNC (doi: 10.2210/pdb5knc/pdb; 3ADPV1 at 3.0 Å)61, and 5KND (doi: 10.2210/pdb5knd/pdb; 1PiV1 at 2.9 Å)62. The authors declare that all other relevant data supporting the findings of this study are available within the article and its Supplementary Information files.
How to cite this article: Suzuki, K. et al. Crystal structures of the ATP-binding and ADP-release dwells of the V1 rotary motor. Nat. Commun. 7, 13235 doi: 10.1038/ncomms13235 (2016).
This paper is dedicated to Late Prof Yoshimi Kakinuma (he died at the age of 64 in January 2016) as a memorial for his great contribution to the research of V-type ATPases. He studied Na+-translocating ATPase in E. hirae and found it to be a bacterial V-type ATPase. With his support, we succeeded in obtaining crystal structures of the V1-ATPase as described in this paper and are now close to understanding its bioenergy transduction mechanism. We also thank Drs Satoshi Arai, Shinya Saijo, Ryota Iino, Hiroshi Ueno, Yoshihiro Minagawa, Mitsunori Ikeguchi and Keitaro Yamashita for helpful suggestions on this study. The synchrotron radiation experiments were performed at Photon Factory (proposals 2011S2-005, 2012G-132, 2013R-22, 2014R-51). We also thank the beamline staff at BL1A and BL17A of Photon Factory (Tsukuba, Japan) for help during data collection. This work was supported in part by Grants-in-aid for Scientific Research (26291009 to T.M., 13J04652 to K. Suzuki, JP16H00811 to E.M.) from the Ministry of Education, Culture, Sports, Science and Technology, Japan (MEXT), and by the Platform Project for Supporting in Drug Discovery and Life Science Research (Platform for Drug Discovery, Informatics and Structural Life Science) from MEXT and Japan Agency for Medical Research and development (AMED).
Conformational changes of the nucleotide-free V1 complex (eV1) induced by the binding of 2 ADP molecules. The colours and view are consistent with those of Figure 3. The video was generated by morphing the X-ray crystal structures of nucleotide-free (eV1) and 2ADP-bound (2ADPV1) complexes using LSQMAN and PyMOL.
Conformational changes of the nucleotide-free V1 complex (eV1) induced by the binding of 3 ADP molecules. The colours and view are consistent with those of Figure 5. The video was generated by morphing the X-ray crystal structures of nucleotide-free (eV1) and 3ADP-bound (3ADPV1) complexes using LSQMAN and PyMOL.
Conformational changes of the 2 ADP-bound V1 complex (2ADPV1) induced by binding one additional ADP molecule to the ‘bindable-like’ form. The colours and view are consistent with those of Figure 5. The video was generated by morphing the X-ray crystal structures of 2ADP-bound (2ADPV1) and 3ADP-bound (3ADPV1) complexes using LSQMAN and PyMOL.
A model for the molecular mechanism of the V1 complex spanning 360° starting 3 from a state of waiting for ATP hydrolysis. The colours and view are consistent with those of Figure 9. The video was generated by morphing the X-ray crystal structures of 2ATPV1, 2ADPV1, and 3ADPV1 using LSQMAN and PyMOL.