The Inhibitory Mechanism of the ζ Subunit of the F1FO-ATPase Nanomotor of Paracoccus denitrificans and Related α-Proteobacteria*

The ζ subunit is a novel inhibitor of the F1FO-ATPase of Paracoccus denitrificans and related α-proteobacteria. It is different from the bacterial (ϵ) and mitochondrial (IF1) inhibitors. The N terminus of ζ blocks rotation of the γ subunit of the F1-ATPase of P. denitrificans (Zarco-Zavala, M., Morales-Ríos, E., Mendoza-Hernández, G., Ramírez-Silva, L., Pérez-Hernández, G., and García-Trejo, J. J. (2014) FASEB J. 24, 599–608) by a hitherto unknown quaternary structure that was first modeled here by structural homology and protein docking. The F1-ATPase and F1-ζ models of P. denitrificans were supported by cross-linking, limited proteolysis, mass spectrometry, and functional data. The final models show that ζ enters into F1-ATPase at the open catalytic αE/βE interface, and two partial γ rotations lock the N terminus of ζ in an “inhibition-general core region,” blocking further γ rotation, while the ζ globular domain anchors it to the closed αDP/βDP interface. Heterologous inhibition of the F1-ATPase of P. denitrificans by the mitochondrial IF1 supported both the modeled ζ binding site at the αDP/βDP/γ interface and the endosymbiotic α-proteobacterial origin of mitochondria. In summary, the ζ subunit blocks the intrinsic rotation of the nanomotor by inserting its N-terminal inhibitory domain at the same rotor/stator interface where the mitochondrial IF1 or the bacterial ϵ binds. The proposed pawl mechanism is coupled to the rotation of the central γ subunit working as a ratchet but with structural differences that make it a unique control mechanism of the nanomotor to favor the ATP synthase activity over the ATPase turnover in the α-proteobacteria.

The F 1 F O -ATP synthase is the ubiquitous nanomotor that fuels life by producing vital chemical energy through the condensation of ADP and P i to form ATP during oxidative phosphorylation or photophosphorylation. This process occurs in the energy-transducing inner membranes of bacteria, mitochondria, and chloroplasts. The core structure of the F 1 F O -ATP synthase nanomotor is conserved through evolution. This nanomotor has a stator plus a rotor that gyrates clockwise (viewed from F O to F 1 ) to synthesize ATP. The process is fueled by the proton flow through F O and the trans-membranous proton gradient established by respiratory or photosynthetic electron transfer chains of energy-transducing membranes. However, during partial or total collapse of the transmembrane proton gradient, the nanomotor will turn in the opposite counterclockwise direction and therefore act to hydrolyze the ATP. This reversal of direction of the F 1 F O -ATPase activity is detrimental in mitochondria and chloroplasts, but it could be advantageous for bacteria in the absence of oxygen or alternative electron acceptors, where the F-ATPase pumps protons to energize secondary transporters under anaerobic or non-respiring conditions. In order to prevent the reverse F 1 F O -ATPase activity, a couple of natural inhibitors have evolved: the eubacterial ⑀ subunit and the mitochondrial inhibitor protein (IF 1 ) 2 (reviewed in Ref. 1). Both inhibitors act to hinder predominantly the counterclockwise rotor gyration of the F 1 F O -ATPase (F 1 -ATPase turnover) by interacting differently with the ␥, ␣, and ␤ subunits (2)(3)(4). A recent study found that the inhibitory domains of bacterial ⑀ and mitochondrial IF 1 bind at the same interface of their respective bacterial or mitochondrial F 1 -ATPases, contacting the ␥ subunit at a common ␣ DP /␤ DP /␥ interface named the "inhibition-general core region" (5).
We recently discovered a third, natural, and potent inhibitor of the bacterial F 1 F O -ATPase in P. denitrificans and related ␣-proteobacteria (6,7). This new inhibitor is different in structure from the bacterial ⑀ and mitochondrial IF 1 and is conserved exclusively in the ␣-proteobacteria class. We named this inhibitor the subunit because it is smaller than ⑀ and showed that the N-terminal side harbors the inhibitory domain of the protein. The other side of , containing four ␣-helixes, works as a globular anchoring domain (7). These studies also showed cross-linking of with the ␣ and ␤ subunits of the F 1 -ATPase stator and with the ␥ and ⑀ subunits of the rotor, indicating that blocks rotation of the central stalk in a similar way to the mitochondrial IF 1 , which also blocks the intrinsic rotation of the mitochondrial F 1 -ATPase (8). The subunit also has a low affinity ATP binding site that seems to control its inhibitory capacity (7,9). In order to resolve the inhibitory mechanism of before the structural data becomes available, we constructed a homology model of the PdF 1 -ATPase complex of P. denitrificans to dock the NMR structure of at its inhibitory binding site. Together with previous and new biochemical data, the final model shows how the subunit blocks rotation of the F 1 F O -ATPase of P. denitrificans and related ␣-proteobacteria, by a "pawl" mechanism on a ratchet (10) formed by the ␥ subunit. This is somehow a hybrid mechanism between mitochondrial IF 1 and bacterial ⑀ but with structural differences giving it a uniqueness on the control of the ␣-proteobacterial F 1 -ATPase nanomotor.

Materials and Methods
PdF 1 -Model Construction-A homology model of the PdF 1 -ATPase was constructed by using the most complete mitochondrial F 1 -ATPase structure available as a template. The mitochondrial F 1 -stalk structure (Protein Data Bank entry 2WSS) was chosen because it showed the highest identity after structural alignment of the sequences of PdF 1 -ATPase with several available bacterial, yeast, and mitochondrial F 1 -ATPase structures ( Table 1). The mitochondrial second stalk and ⑀ subunits were removed from the template before the construction of the PdF 1 -ATPase model. Thus, the final PdF 1 -ATPase model contained only the subunits ␣ 3 , ␤ 3 , ␥ 1 , ␦ 1 , and ⑀ 1 (with the indicated stoichiometries). A model of each subunit was constructed separately by the Swissmodel server, and subsequently all of the subunits were then assembled into a model of the full PdF 1 -ATPase complex in Swissmodel (deep view). The quality of each subunit model was confirmed by manual checking of the full alignment in Swissmodel and Chimera, with an upper limit of main chain root mean square deviation of around 0.2 Å. The raw model obtained was subjected to several rounds of three-dimensional fitting using the template 2WSS structure. Afterward, the model was refined by correction of clashes and incorrect atom positions by several energy minimizations in Chimera, Swissmodel, and by evaluation of the model with Molprobity (11)(12)(13). The final model fitted to each subunit of the 2WSS template ( Fig. 1) with an average root mean square deviation of Յ0.2 Å. Some small regions of the ␥ subunit that were not resolved in the template (2WSS) were modeled in order to obtain a higher accuracy. Special care was taken with the PdF 1 -⑀ subunit, which had the lowest identity; therefore, this model was further evaluated using a combination of biochemical data together with other bacterial ⑀ templates.
PdF 1 -ATPase, Subunit, and IF 1 Purifications-The PdF 1 -ATPase was purified by chloroform extraction and chromatography as described before (7). The recombinant and IF 1 subunits were overproduced and purified as described elsewhere (7,14).
Cross-linking Analyses-Cross-linking analyses were carried out as described previously with dithiobis(succinimidyl propionate) (DSP) by incubating the PdF 1 -ATPase with or without recombinant subunit (or the ⌬NT construct) (7) for 30 min at room temperature in 20 mM KH 2 PO 4 buffer at pH 7.4. The cross-linking reaction was arrested by the addition of excess L-lysine (20 mM), and the samples were immediately loaded into SDS-polyacrylamide gels (15) to analyze cross-linking products.
Limited Proteolysis of the PdF 1 -Complex-The PdF 1 -ATPase containing its endogenous ⑀ and subunits was incubated for limited proteolysis with trypsin in a ratio (w/w) of 1:20 (trypsin/PdF 1 ) at room temperature as described before (6,7). The limited proteolysis was stopped with excess PMSF (5 mM), together with a 4ϫ supplement of the complete mixture of protease inhibitors (Roche Applied Science). Sequencing of the protein fragments was done by mass spectrometry as described before (7).
Other Procedures-The PdF 1 -ATPase activity was measured by the coupled ATPase assay as described before (7,14,16) in the presence of 0.15% of lauryldimethylamine oxide detergent, 1 mM sulfite, pH 6.9. Protein concentrations were measured by the modified TCA-Lowry method (17), and Western blots were carried out as before (7).

Results and Discussion
Attempts to crystallize the PdF 1 -ATPase had resulted in non-diffracting or unstable PdF 1 -ATPase preparations that dissociate upon crystallization and produced diffracting crystals of a single ␣/␤ catalytic pair, thus lacking the ␥, ⑀, and subunits (18). In order to obtain a suitable structural model of the PdF 1complex, an initial homology model for the PdF 1 -ATPase was constructed using the most similar and complete crystallographic structure of the F 1 -ATPase available (Protein Data Bank entry 2WSS). According to the endosymbiotic theory (19 -22), it was found that the most similar F 1 -ATPase structure was that of the mitochondrial enzyme, rather than the eubacterial form, such as that of Escherichia coli (Table 1). By using the 2WSS structure to construct the PdF 1 -ATPase model, it was possible to obtain a homology model with a very high superposition of tertiary and quaternary structures of the ␣, ␤, ␥, and ⑀ subunits (see Fig. 1A). After energy minimizations and

-ATPase resolved by x-ray crystallography and the PdF 1 -ATPase
The identities of each subunit of the PdF 1 -ATPase resolved by x-ray crystallography compared with the subunits of the PdF 1 -ATPase (IDvsPd) were determined by structural alignment with Swissmodel. The highest identities were observed in the first line with the MF 1 -stalk structure (Protein Data Bank entry 2WSS). For this reason, the 2WSS structure was chosen as the template to construct the PdF 1 -ATPase model.  The structure of PdF 1 -ATPase was modeled by homology using the structure of the bovine F 1 -stalk structure resolved by x-ray crystallography (Protein Data Bank entry 2WSS) as a template. The model was constructed with Swissmodel, Chimera, and PyMOL as described under "Materials and Methods." A, the model (␣ (cyan), ␤ (blue), ␥ (green), and ⑀ (yellow)) matched very well the overall main chain structure of the template (Protein Data Bank entry 2WSS, orange) after superposition, with an average root mean square deviation of 0.128 Å. The positions of the ␣ E /␤ E and ␣ DP /␤ DP interfaces, together with ␥ and ⑀ subunits, are indicated; the ␣ TP ␤ TP interface is not indicated because it is located "behind" and therefore it is not clearly visible from this view. B, "side view" of the final PdF 1 -ATPase model with the characteristic folding of the central rotor containing the ␥/⑀ heterodimer at the center of the ␣ 3 ␤ 3 heterohexamer. The three conformations observed in most F 1 -ATPases of the catalytic ␣/␤ interfaces as the empty (␣ E /␤ E ), ADP (␣ DP /␤ DP ), and ATP (␣ TP /␤ TP ) heterodimers are conserved. No nucleotides are included in the model. C, cross-linking with DSP of the PdF 1 -ATPase and EcF 1 -ATPase, as revealed by anti-⑀ and anti-␤ Western blot (WB). The PdF 1 -and EcF 1 -ATPases were incubated for cross-linking with DSP as indicated under "Materials and Methods." The cross-linking products were resolved by loading the samples in the same SDS-polyacrylamide gel and revealed by Western blotting after immunotransfer and cutting of the PVDF membrane in the middle (indicated by the central line). The left half of the membrane was developed with anti-⑀ primary antibodies, and the right side of the membrane was developed with anti-␤ primary antibodies. The immunoblots showed the presence of ⑀-␥ (45 kDa) and ⑀-␤ (65 kDa) cross-linking products in the EcF 1 -ATPase (EcF 1 -DSP) and their absence in the PdF 1 -ATPase (PdF 1 -DSP). These results support the compact conformation of ⑀ of the PdF 1 -ATPase model shown above (see "Results and Discussion"). D, limited proteolysis of the PdF 1 -ATPase carried out with trypsin and developed by Western blot with anti-␤, anti-⑀, and anti-antibodies as described under "Materials and Methods." The subunit was totally resistant to limited proteolysis with the cleavage sites occluded in the PdF 1 -complex as shown in Fig. 4. The ⑀ subunit was cleaved initially at the C-terminal ␣-helices that are exposed in the PdF 1 -ATPase model above; the first fragment that appeared below ⑀ corresponded to ⑀(1-103), according to mass spectrometry analyses.

Template
1B). The folding of the Pd-⑀ model was fitted to the compact conformer of bacterial ⑀ resolved before (23), and this conformation was supported by limited proteolysis and cross-linking experiments together with its lack of inhibition (6, 7), as described below. Two experimental approaches supported the presence of the compact conformer of ⑀ in the PdF 1 -ATPase. First, in order to explore the putative compact conformation of Pd-⑀, it was assessed whether the endogenous ⑀ subunit was able or not to cross-link with the ␤ subunits in the PdF 1 -ATPase complex (see Fig. 1C). In the PdF 1 -ATPase model, the compact Pd-⑀ (Fig. 1B) is at a distance of about 29 Å to the closest ␤ C terminus, whereas the cross-linker (DSP) has a shorter distance of 12 Å. In the extended conformation, the two C-terminal ␣-helixes of ⑀ make close contacts with ␣, ␤, and ␥ subunits (2,5,24). Therefore, in the compact conformation, Pd-⑀ should be unable to form the ⑀-␤ and ⑀-␥ adducts with DSP, whereas in the extended conformation, Pd-⑀ should form ⑀-␤ and ⑀-␥ crosslinkages. In order to determine the presence or absence of the ⑀-␤ adduct in the PdF 1 -ATPase, control experiments were also carried out with the F 1 -ATPase isolated from E. coli. As expected, the latter enzyme showed both the ⑀-␤ (Fig. 1C, left and right, EcF 1 -DSP, 65 kDa band) and the ⑀-␥ cross-linkings (Fig. 1C, left, EcF 1 -DSP, 45 kDa) previously identified by twodimensional SDS-PAGE (7). These results show that in E. coli, the extended conformation of ⑀ promotes the ⑀-␤ cross-linking of 65 kDa, which immunoreacted with both anti-⑀ and anti-␤ antibodies (Fig. 1C, EcF 1 -DSP, left and right). In contrast, in the PdF 1 -ATPase, there are some faint ⑀-adducts (Fig. 1C, left, PdF 1 -DSP), but none of these are ⑀-␤ adducts, as shown by the absence of the 65 kDa band in the anti-␤ blot (Fig. 1C, right, PdF 1 -DSP). A 65 kDa anti-⑀ band in the control non-crosslinked sample (PdF 1 ) seems to be an ⑀ aggregate or a nonspecificity (Fig. 1C, left PdF 1 lane), which is also present in the PdF 1 -DSP cross-linked sample (Fig. 1C, left). Taken together, the absence of the ⑀-␤ adduct in PdF 1 -ATPase and its presence in F 1 -ATPase of E. coli (EcF 1 ) support the finding that the endogenous Pd-⑀ prefers to adopt the compact conformation, as shown in the PdF 1 -ATPase model (Fig. 1B). This preferred compact conformation of Pd-⑀ explains why this subunit does not work as an endogenous inhibitor of the PdF 1 -ATPase and putatively should not work in the other ␣-proteobacteria. The Pd-⑀ subunit does not seem to be able to acquire the extended inhibitory conformation, and it is therefore unable to block ␥ rotation. This leaves this inhibitory role to the subunit.
The two C-terminal ␣-helixes of ⑀ were accessible to limited proteolysis in the PdF 1 -ATPase (see Fig. 1D). This accessibility to the protease is more likely to take place in the compact conformation of ⑀, as shown in Fig. 1B, than in the extended conformation of ⑀, where the two inhibitory C-terminal ␣-helixes are inserted and mostly buried within the ␣/␤/␥ interface (2). In the extended conformation, the C-terminal ␣-helixes of ⑀ should be inaccessible to limited proteolysis in the F 1 -ATPase complex. Our limited proteolysis (Fig. 2) and MS results of the PdF 1 -ATPase (supplemental data) showed a tryptic ⑀ fragment of 11 kDa, among other ␣ and ␥ fragments, migrating just below the intact subunit and containing the N-terminal peptide ⑀ Met1-Arg13 (see Fig. 2, A-D, and supplemental data). Therefore, the protease cleaved the two C-terminal ␣-helixes of ⑀, leaving an 11-kDa fragment consisting of ⑀ Met1-Arg103 (Fig. 2D). These results strongly suggest that the two C-terminal ⑀ ␣-helixes are exposed in the PdF 1 -ATPase and therefore most probably in the compact ⑀ conformation (Fig. 1, B and D). These results seem to be opposite to those of Wilkens and Capaldi (23). They found that the isolated ⑀ subunit from the F 1 -ATPase of E. coli (Ec-⑀) is relatively resistant to trypsin in its compact soluble conformation as resolved by NMR (23). Therefore, we analyzed the trypsin sensitivity of the isolated ⑀ subunit from P. denitrificans (Pd-⑀), and in contrast to that of E. coli (23), we observed an extensive cleavage of the Pd-⑀ subunit at trypsin/Pd-⑀ ratios of 1:5-1:100 (w/w) (Fig. 2B). The trypsin/Ec-⑀ ratio used by Wilkens and Capaldi (23) was in the same range (1:80), and the isolated Ec-⑀ was resistant to the protease for about 8 min, whereas our Pd-⑀ subunit was extensively cleaved at the same 1:80 ratio after only 3 min of limited proteolysis, clearly producing the 11-kDa ⑀ Met1-Arg103 fragment among other peptides (Fig. 2B). These results indicate that the trypsin sensitivities of the isolated Ec-⑀ and Pd-⑀ subunits are different, probably due to intrinsic sequence differences or to different proteolysis conditions, and therefore, their trypsin sensitivities are not comparable with each other. In consequence, given the high sensitivity of the isolated Pd-⑀ to trypsin, the formation of the ⑀ Met1-Arg103 fragment by limited trypsinolysis of the full PdF 1 -ATPase indicates that the two C-terminal ␣-helixes of the endogenous Pd-⑀ are exposed and most likely in the compact conformation as shown in the PdF 1 -ATPase model (Fig. 1, A and B).
Once the PdF 1 -ATPase model was completed, the most represented average structure of the 20 conformers of Pd-that were resolved was docked with Chimera into the ␣/␤ interfaces following the previously observed high cross-linking interaction of with PdF 1 -ATPase (6, 7). The most probable ␣/␤ pair to provide the first interaction surface for was the open ␣ E /␤ E interface, because this allowed an easier docking of the subunit than the other two closed interfaces. We already showed that the N terminus of is the inhibitory domain (red in Fig. 3A) and that there is a high yield of cross-linking with ␣ and ␤ subunits and a lower yield with ␥ and ⑀ subunits (6, 7). Therefore, the disordered N-terminal domain of was oriented toward the ␥ subunit, and its globular part, which works as an anchoring domain, was oriented to the C termini of the ␣ E /␤ E interface (Fig. 3B). This orientation is also supported by crosslinking data because the cross-linkers used previously (2-iminothiolane and DSP) were specific for lysines, and most of these lysines (4 of 6) are on the globular domain, compared with only 2 lysines in the N-terminal side. This suggests that the high -␣ and -␤ cross-linking yields (7) result from interaction of the globular part of with the ␣/␤ interface, whereas the lower -␥ cross-linking yield (7) is due to the N-terminal inhibitory domain of interacting with subunit ␥. In order to test this putative orientation, we took advantage of our Pd-⌬NT construct lacking the first 14 N-terminal residues (7). The overall model predicts the interaction of the N-terminal inhibitory domain of with the ␥ subunit to inhibit rotation. If this is the case, the Pd-⌬NT construct lacking the inhibitory N-terminal domain should decrease its cross-linking yield with ␥ after its reconstitution in conditions that promote its effective binding to the PdF 1 -ATPase (7) and where we had previously shown that the reconstitution of the full WT subunit increases the -␥ cross-linking yield (7). For this experiment, most of the endogenous subunit of the PdF 1 -ATPase complex was removed by immunoaffinity columns as described before (7), and then the Pd-⌬NT construct was reconstituted. Excess free Pd-⌬NT was removed by gel filtration. As revealed by Western blot, the DSP cross-linking showed a decrease in the yield of the -␥ crosslink instead of an increase (Fig. 3C, right lane), as expected if the N-terminal extreme of is the domain interacting with ␥ blocking its rotation. However, we also observed a decrease in the -␤ adduct and an increase in high molecular weight unidentified adducts (Fig. 3C, right lane). The latter could result from extensive Pd-⌬NT -␣/␤ cross-linking derived from a lower accessibility of Pd-⌬NT to cross-link with . Alternatively, because the Pd-⌬NT construct was reconstituted in excess amounts relative to PdF 1 , the truncated subunit could probably bind to two or three ␣/␤ interfaces, similar to the reconstituted mitochon-drial F 1 -(IF 1 ) 2 and F 1 -(IF 1 ) 3 complexes (2-4). If this is the case with the Pd-⌬NT construct, the larger molecular adducts are probably derived from extensive -␣-␤ cross-linkages formed after reconstitution into the PdF 1 -ATPase. On the other hand, the low yield -␥ cross-link observed after Pd-⌬NT reconstitution (Fig. 3C, right lane) is probably due to small amounts of remanant endogenous Pd-WT that could not be removed by immunoaffinity columns, as observed previously (7). In summary, the lack of the N-terminal inhibitory domain in the Pd-⌬NT construct decreased the -␥ cross-linking yield, suggesting that this is in close proximity to or in direct contact with ␥ to block its rotary function.
After finding that the final PdF 1 -ATPase structure is compatible with the cross-linking and limited proteolysis data, we therefore proceeded to accommodate the NMR structure of the subunit into its putative inhibitory site on the PdF 1 -ATPase model. In order to achieve this, the N-terminal domain of was oriented toward the ␥ subunit, and the globular and C-terminal At t 0 , PMSF was present before the addition of trypsin. B, limited trypsinolysis of the isolated Pd-⑀ subunit. The recombinant Pd-⑀ subunit was subjected to limited trypsinolysis at the indicated Trp/Pd-⑀ ratios by 3 min at 25°C in reconstitution buffer, and the reactions were arrested as before loading the samples onto the SDS-polyacrylamide gel as shown. C, Western blot against the ⑀ subunit of P. denitrificans of the same samples shown in A but loaded with 3-fold lower amounts of protein. D, observed limited trypsinolysis fragment of the endogenous ⑀ subunit bound to the PdF 1 -ATPase. The model of ⑀ subunit of P. denitrificans is shown, indicating in green the N-terminal Met 1 -Arg 13 peptide sequenced by MS from the proteolytic ⑀ fragment of 11 kDa (A and B). Given this 11 kDa size, the most probable cleavage position is the first of three arginine residues (Arg 103 ) in the C-terminal ␣-helixes of ⑀, corresponding to the inhibitory domain of the protein in other eubacteria. The trypsin-cleaved segment is shown in orange, and therefore, the proteolytic fragment is ⑀ Met1-Arg103 . The C-terminal domain of ⑀ is therefore exposed and accessible to the protease in the intact PdF 1 -ATPase (A and Fig. 1) as well as in the isolated ⑀ subunit (B).
domains were bound to the ␣/␤ interface. The N-terminal domain and the C-terminal ␣-helix of the subunit extend together on one side of and make together a continuous contact surface to interact with the ␣/␤ interface (red N terminus and green C terminus in Fig. 3A). In accordance with limited proteolysis assays, these regions were protected from limited trypsinization in the intact PdF 1 -complex, indicating that N and C termini of are occluded in the native PdF 1 . In contrast, both proteolytic sites on the N and C termini of Pd-(red and green in Fig. 3A) are exposed in the isolated subunit (6). The subunit was therefore inserted in the ␣ E /␤ E interface, which opens the more accessible entrance site, with the N terminus approaching the ␥ subunit and the C-terminal ␣-helix of occluded at the ␣ E /␤ E interface (Fig. 3B).
This initial entrance of the subunit does not seem to be the final inhibitory position because the productive interaction of with PdF 1 -ATPase is enhanced by the catalytic turnover of the enzyme, similar to the mechanism described for the mitochondrial IF 1 . In this mechanism, some ATP hydrolysis turnovers are required for the proper binding of the inhibitor into its final inhibitory position (6,7,25). It has been found that two 120°p artial rotations of the ␥ subunit are required for the accommodation of IF 1 in its final inhibitory binding ␣ DP /␤ DP /␥ interface in a binding lock mechanism (3, 4). Therefore, it was con-

. Structural and functional evidence indicating that and IF 1 bind to the same ␣ DP /␤ DP /␥ interface to inhibit rotation and F 1 -ATPase activity.
A, the NMR structure of the subunit of P. denitrificans (Protein Data Bank entry 2LL0) (7,9,28) is shown in different colors; the disordered and highly mobile N-terminal inhibitory domain that is cleaved by trypsin (7) is shown in red; the C-terminal ␣-helix 4, which is also cleaved by trypsin, is shown in green, and the globular domain of that is resistant to trypsin containing ␣-helices 1-3 is shown in white. B, the initial entrance site of Pd-into PdF 1 -ATPase was more easily modeled by docking the C-terminal ␣-helix of (orange) into the ␣ E /␤ E interface, with the inhibitory N-terminal domain of (red) pointing toward the central ␥/⑀ rotor. Subunit color codes are the same as in Fig. 1. C, Western blot anti-revealing the cross-linking products of control PdF 1 -ATPase (left lane) and the PdF 1 -ATPase (lacking most of the endogenous ) reconstituted with the Pd-⌬NT construct (right lane). The same amounts of protein (10 g) were loaded on both lanes. The -␥ cross-linking yield decreased with the reconstitution of the Pd-⌬NT construct, whereas other higher adducts increased probably because of extensive -␣/␤ cross-linkages. D, the PdF 1 -ATPase model (green) was aligned and fitted to the structure of the mitochondrial F 1 -IF 1 complex (Protein Data Bank entry 2V7Q, light pink) with PyMOL; the alignment is not as good as in the non-inhibited MF 1 -PdF 1 -ATPase alignment shown above (Fig. 1) because of differences in the rotor positions induced by IF 1 or stalk subunits, respectively. IF 1 is shown in orange, and the slightly similar (7) inhibitory N-terminal domain of is shown in red with a semitransparent surface. The alignment indicates that both and IF 1 are probably bound to the same inhibitory site. sidered that could work through an IF 1 -like mechanism because we also found previously that the N-terminal inhibitory domain of has a limited but convergent similarity with the corresponding inhibitory domain of IF 1 (7). This hypothesis was further supported by the recent finding that the inhibitory ␣-helixes of ⑀ and IF 1 interact with the homologous ␣ DP /␤ DP /␥ interface of their respective F 1 -ATPases of E. coli, PS3, and mitochondria, at a common inhibitor-binding pocket named the "inhibition general core region," which includes residues from the C terminus of ␣/␤ subunits and some catch residues of ␥ (5).
Taken together, these antecedents supported the possibility that the inhibitory N-terminal domain of the subunit will also bind to the same ␣ DP /␤ DP /␥ interface as the inhibitory domains of ⑀ and IF 1 . In order to assess this hypothesis, the N-terminal inhibitory domain of (-NT) was aligned structurally with the mitochondrial IF 1 bound in its inhibitory position in the mitochondrial IF 1 . Afterward, the PdF 1 -ATPase model was aligned and superimposed on the MF 1 -IF 1 and -NT structures. The small similarity between both inhibitory domains of and IF 1 (7) allowed a structural alignment in PyMOL (Fig. 3D). This structure of the PdF 1 -ATPase model with the -NT (red in Fig.  3D) aligned to IF 1 was used as a guide for the fitting of the subunit into the ␣ DP /␤ DP /␥ interface of the PdF 1 -ATPase model. The final structural alignment showed that the N-terminal inhibitory domain of very likely binds to the common inhibitor-binding region of the PdF 1 -ATPase at the ␣ DP /␤ DP /␥ interface (Fig. 3D) in a similar fashion to the way mitochondrial IF 1 binds to MF 1 -ATPase (3,26).
One of the main predictions of this model is that if and IF 1 share a common binding site with their N-terminal inhibitory domains interacting similarly with their respective F 1 -ATPases, this raises the possibility that the mitochondrial IF 1 could exert heterologous inhibition on the F 1 -ATPase of P. denitrificans, albeit with a lower affinity than that of the Pd-during homologous reconstitution. Therefore, the putative inhibitory effect of mitochondrial IF 1 on the PdF 1 -ATPase was assayed, in parallel to the known inhibitory action of , by carrying out titration curves of IF 1 and on the PdF 1 -ATPase activity. In accordance with the present model of the interaction of with PdF 1 , it was clearly observed that the mitochondrial IF 1 exerted a partial inhibition of the PdF 1 -ATPase. In contrast, the subunit exerted a stronger total inhibition (Fig. 4A). As expected, the affinity of IF 1 to inhibit the PdF 1 -ATPase was lower than that of , because the IC 50 of IF 1 was about 50-fold higher than that of Pd- (Fig. 4A). We also observed that after adding IF 1 in amounts close to its IC 5O in order to produce 50% inhibition, the subsequent addition of showed an additive effect and reached total inhibition (Fig. 4B). This suggests that both inhibitory proteins interact at the same binding site. To our knowledge, this is the first evidence of inhibition of a bacterial F 1 -ATPase by the mitochondrial IF 1 . It is well known that IF 1 does not inhibit the EcF 1 -ATPase. We confirmed this and found that IF 1 showed no inhibition of the EcF 1 -ATPase at all (Fig. 4A); however, it produced a 93% inhibition of the mitochondrial F 1 -ATPase as a positive control (Fig. 4B). The endosymbiotic theory proposes that mitochondria arose from ␣-proteobacteria (19 -22). Given the close similarity between PdF 1 -ATPase FIGURE 4. Heterologous inhibition of the PdF 1 -ATPase by the mitochondrial IF 1 inhibitor. A, effect of (E) and IF 1 (•) on the PdF 1 -ATPase. The PdF 1 -ATPase (21 g) was preincubated in 20 l with the indicated concentrations of recombinant or IF 1 proteins purified as described under "Materials and Methods" in the presence of 1 mM MgATP in reconstitution buffer at pH 7.0. After 20 min preincubation at room temperature, the full mixture was added to a 1-ml reaction cells at 37°C containing the coupled ATPase assay buffer as described under "Materials and Methods." 100% of PdF 1 -ATPase activity corresponds to 4.99 mol/min⅐mg for the titration curve and 4.019 mol/min⅐mg for the IF 1 titration curve and 41.5 mol/min⅐mg for the EcF 1 -ATPase (Ⅺ). The plot shows the fitting to a non-linear Hill inhibitory equation giving an IC 50 of ϭ 4.76 M and IC 50 of IF 1 ϭ 208.6 M carried out with Origin version 7.5. B, the additive effects of IF 1 and were assayed by preincubation of the PdF 1 -ATPase (4 g) with 50 g of purified bovine heart IF 1 as indicated above to obtain about 50% inhibition of the PdF 1 -ATPase (bar 2 indicated as IF 1 ). Increasing amounts of the subunit (0.25, 0.5, 1, 2, and 5 g) were added in addition to IF 1 to samples 3-7 for preincubation as described above, before measurement of the PdF 1 -ATPase activity as described under "Materials and Methods." Control samples were preincubated in the absence of IF 1 and inhibitors and showed a PdF 1 -ATPase activity of 7.7 mol/min⅐mg protein. In the last samples (9 -11), the activity of BhF 1 -ATPase was measured, and the effects of preincubation of 3.8 g of BhMF 1 -ATPase with 25 g of IF 1 and 22 g of were assayed on the BhMF 1 -ATPase activity. The control activity of the BhMF 1 -ATPase was 14.7 mol/min⅐mg protein; this activity is relatively lower than those reported before, probably because the ATPase assay was carried out at pH 6.9, and the optimal pH of the mitochondrial enzyme is 8.0. Only IF 1 , but not , exerted a strong inhibitory effect on the BhMF 1 -ATPase. and MF 1 -ATPase (Table 1), it may not be a coincidence that the PdF 1 -ATPase is inhibited by IF 1 , whereas the EcF 1 -ATPase is not (Fig. 4A). In other words, the selective heterologous inhibition of PdF 1 -ATPase by mitochondrial IF 1 is in concordance with the ␣-proteobacterial origin of mitochondria because the N-terminal inhibitory domains of and IF 1 are slightly similar (7). In addition, the "inhibition general core region" at the ␣/␤/␥ interface should be more conserved between the P. denitrificans and mitochondrial PdF 1 -ATPases than between the mitochondrial F 1 -ATPase and EcF 1 -ATPase. This closest similarity between Paracoccus and mitochondrial F 1 -ATPases allowed the productive binding of mitochondrial IF 1 into the inhibitor general domain of the ␣/␤/␥ interface of the PdF 1 -ATPase (Fig. 4A). The reciprocal experiment to assess the effect of on the mitochondrial BhMF 1 -ATPase showed no inhibition (Fig. 4B). However, the experiment was carried out with 22 g of , so it remains to be explored with larger amounts of to confirm whether or not it inhibits the mitochondrial F 1 -ATPase. Taken together, these results strongly suggest that the inhibitory mechanisms and therefore the binding sites of IF 1 and are very similar and support the model described above. This indicates that the limited but converging similarity between the N-terminal inhibitory domains of the subunit and IF 1 promote their binding to the common inhibitor binding region at the ␣ DP /␤ DP /␥ interface to block rotation of the PdF 1 -ATPase.
Taking all of the results together, the overall inhibitory mechanism of binding of the subunit to the PdF 1 -ATPase is most likely as follows: 1) the subunit enters through the open ␣ E /␤ E interface with its N-terminal inhibitory domain pointing forward to interact with the rotary ␥ subunit, with its globular part working as an anchoring domain (Fig. 5, A and B); 2) one partial 120°counterclockwise rotation of ␥ as induced by ATP binding (27) transforms the ␣ E /␤ E interface into the ␣ TP /␤ TP conformation (Fig. 5, C and D); 3) a second ATP binding and 120°counterclockwise rotation of ␥ locks the N-terminal inhibitory domain of into the ␣ DP /␤ DP /␥ interface known as the "inhibitor general core region," whereas the globular domain of holds the inhibitor bound to the C terminus of the ␣ DP /␤ DP interface (Fig. 5, E and F). This model explains the mechanism of inhibition of ␥ rotation exerted by , which acts to block the PdF 1 -ATPase activity. Presumably, this is preserved in most if not all ␣-proteobacteria. In summary, works similarly to IF 1 and ⑀ by working as a pawl (i.e. by blocking preferably the counterclockwise gyration of ␥, which rotates as a ratchet) (10). In the final modeled inhibitory position of , the N-terminal inhibitory domain of the protein is partially inserted in the "inhibition general core region" of the ␣ DP /␤ DP /␥ interface (Fig. 5, E  and F), in a similar way to IF 1 in the mitochondrial enzyme. This position of explains why the Pd-⌬NT construct binds but does not inhibit the PdF 1 -ATPase (7). Removal of the N-terminal inhibitory domain will allow the rotation of the ␥-⑀ rotor because the Pd-⌬NT construct will not be able to work as a pawl to hinder ␥ rotation. A further induced fit of the PdF 1 -interface should lead to a deeper insertion of the N terminus of to reach ␥ as it is proposed in Fig. 3D. This would result in a closer interaction of with the ␥ subunit to block rotation. If the inhibitory mechanisms of and IF 1 are similar enough, the dis-ordered N-terminal domain of (Fig. 3A) will probably fold into an extended ␣-helical conformation upon reaching its final locked position in PdF 1 , as the mitochondrial IF 1 does in MF 1 -ATPase (4). Our model also considers that the subunit may turn its globular anchoring domain upon reaching its final locked position. Some of these finer details of the Pd-PdF 1 interaction will have to wait for the atomic resolution of the structure; nonetheless, the overall inhibitory mechanism of FIGURE 5. Model for the inhibitory interaction of the subunit to inhibit rotation of the PdF 1 -ATPase nanomotor. A and B, the open ␣ E /␤ E interface of the PdF 1 -ATPase model was the most accessible first interaction surface to dock the subunit structure. Because the N-terminal domain (red) of the subunit is the inhibitory domain and this should interact with the ␥ subunit to inhibit rotation, this domain was directed through the open ␣ E /␤ E interface by accommodating the C-terminal ␣-helix of the globular part of as a binding surface. C and D, after the initial entrance of at the empty interface, a 120°r otation of the central rotor induced by ATP binding changes the conformation of the ␣ E /␤ E interface to the ␣ TP /␤ TP conformer and promotes a closer interaction of the globular domain of with this interface, whereas the N-terminal inhibitory domain is presented to interact with the ␥ subunit. E and F, a final ATP binding step promotes a second 120°partial gyration of the ␥/⑀ rotor; therefore, the catalytic interface interacting with the subunit shifts to the ␣ DP /␤ DP conformation, making a closer interaction with the globular domain of , whereas the inhibitory N-terminal domain (red) is inserted through the ␣ DP /␤ DP interface or "inhibitor general region," similarly to mitochondrial IF 1 bound to MF 1 . The N-terminal side of is now in position to hinder the further rotation of the ␥ subunit, thus inhibiting fully the PdF 1 -ATPase activity. A deeper insertion of the N-terminal extreme of might occur in order to align completely with the IF 1 binding position of MF 1 -ATPase as in Fig. 2. Presumably, this interaction inhibits preferably the counterclockwise rotation of the PdF 1 -ATPase. See "Results and Discussion" for details.
was resolved here. On the other hand, the globular domain of has additional ⑀-like features, such as a low affinity nucleotide binding site that may regulate the inhibitory capacity of (7). Analyses are ongoing to assess this putatively regulatory ATP binding site and also the proposed pawl mechanism of on the ␥ ratchet.