|Home | About | Journals | Submit | Contact Us | Français|
The author(s) have made the following declarations about their contributions: Conceived and designed the experiments: LP DBH TAK TM. Performed the experiments: LP ÖY DBH. Analyzed the data: LP ÖY TM. Contributed reagents/materials/analysis tools: LP ÖY DBH. Directed the project and wrote the paper: TAK TM.
We solved the crystal structure of a novel type of c-ring isolated from Bacillus pseudofirmus OF4 at 2.5 Å, revealing a cylinder with a tridecameric stoichiometry, a central pore, and an overall shape that is distinct from those reported thus far. Within the groove of two neighboring c-subunits, the conserved glutamate of the outer helix shares the proton with a bound water molecule which itself is coordinated by three other amino acids of outer helices. Although none of the inner helices contributes to ion binding and the glutamate has no other hydrogen bonding partner than the water oxygen, the site remains in a stable, ion-locked conformation that represents the functional state present at the c-ring/membrane interface during rotation. This structure reveals a new, third type of ion coordination in ATP synthases. It appears in the ion binding site of an alkaliphile in which it represents a finely tuned adaptation of the proton affinity during the reaction cycle.
Like the wind turbines that generate electricity, the F1Fo-ATP synthases are natural “ion turbines” each made up of a stator and a rotor that turns, when driven by a flow of ions, to generate the cell's energy supply of ATP. The Fo motor rotates by reversible binding and release of coupling ions that flow down the electrochemical ion gradient across the cytoplasmic cell membrane (in the case of bacteria) or intracellular organelle membranes (in the case of eukaryotic cells). Here, we present the structure of a rotor (c-)ring from a Bacillus species (B. pseudofirmus OF4) determined at high-resolution by X-ray crystallography. This bacterium prefers alkaline environments where the concentration of protons (H+) is lower outside than inside the cell – the inverse of the situation usually found in organisms that prefer neutral or acidic environments. The amino acid sequence of the protein subunits in this rotor, nevertheless, has features common to an important group of ATP synthases in organisms from bacteria to man. The structure reveals a new type of ion binding in which a protonated glutamate residue in the protein associates with a water molecule. This finding raises the possibility considered by Nobel laureate Paul Boyer several decades ago that a hydronium ion (a protonated water molecule, H3O+), rather than a proton alone, might be the coupling species that energizes ATP synthesis. Also, it demonstrates the finely tuned adaptation of ATP synthase rotor rings and their ion-binding sites to the specific requirements of different organisms.
Most living cells depend upon the adenosine triphosphate (ATP) generated by F1Fo-ATP synthases that are energized by a proton- or a sodium-motive force (pmf, smf). These multi-subunit enzymes contain a cytoplasmic F1 catalytic domain (subunits α3β3γδε) that is connected with a membrane-embedded Fo domain (ab2c10–15 in bacteria) by a central (γε) and peripheral (b2δ) stalk. Energetically down-hill ion translocation across the membrane through the Fo complex is mediated by successive interactions between the stator a-subunit and a rotor ring (c-ring). Translocation involves ion binding to an unoccupied c-subunit, rotation, and subsequent ion release. The c-ring is attached to the γε stalk subunits so that c-ring rotation causes rotation of the stalk. The inherently elastic  and asymmetric γ-subunit extends into the α3β3 headpiece  and by rotation  induces conformational changes  in the catalytic β-subunits, which results in ATP synthesis.
In the Na+-binding c11 ring from Ilyobacter tartaricus  and the H+-binding c15 ring from Spirulina platensis , the translocated ions are bound within the groove of two adjacent c-subunits in a coordination network including a conserved carboxylate (Glu). In both cases, the ion is further coordinated by a precise network of residues, several of which are common to both organisms (Figure 1). The ion specificity of these two systems is determined by several factors including the geometry and distances of the ion coordination network, and a water molecule  providing a coordination site for Na+. The ion binding specificity of ATP synthases in various cells is adapted to the physiological requirements of the organism, and the different binding motifs observed presumably reflect these adaptations. The range of the ion-binding motif includes variations from complete Na+-binding signatures to c-subunits where the conserved carboxylate (E/D) of the C-terminal helix is the only residue that can be predicted with confidence to play a role in ion coordination (e.g., in Escherichia coli or Homo sapiens, Figure 1). On this basis, we here assign the name “E/D-only” to the c-subunits of this sub-class of proton-coupled ATP synthases.
Alkaliphilic Bacillus species are among the bacteria having proton-coupled ATP synthases  with E/D-only c-subunits (Figure 1). The extreme alkaliphile Bacillus pseudofirmus OF4 grows by oxidative phosphorylation with cytoplasmic pH values maintained 1.5–2.3 pH units below the high external pH (up to 11) of the medium . The existence of this reversed ΔpH poses a major thermodynamic problem, with which these cells must cope. Among a variety of adaptive strategies to resolve the energetic problem , some special adaptations of the ATP synthase itself have evolved: latent ATPase activity ,, a-subunit modification ,, and in particular, specific adaptations of the c-subunit sequence  resulting in a large c-ring width with more c-subunits . The adaptations to alkaliphilic conditions in an ATP synthase rotor, with a widely found but structurally uncharacterized E/D-only motif, made its c-ring an attractive candidate for an X-ray diffraction study.
Three-dimensional crystals of the c-ring from native Bacillus pseudofirmus OF4 F1Fo-ATP synthase were obtained and diffracted to 2.5 Å (Table 1). An asymmetric unit contains one complete c-ring, forming crystal contacts with three neighboring, laterally translated c-rings including two loop-to-loop contacts and one loop-to-C-terminus contact. One c-ring is formed by 13 identical c-subunits with a central pore (Figure 2A). Because of the unusually short N- and C-termini typical of Bacillus species (Figure 1), it does not extend into the periplasmic space but instead ends at the periplasmic membrane surface, whereas on the cytoplasmic side, the protrusion out of the membrane is comparable to that observed in other c-rings (Figure S1). In contrast with the distinctly hour-glass shape of the c11, K10, and c15 rotor rings ,,, this c13 ring resembles a “tulip beer glass” appearance, in which the only slightly curved periplasmic side transitions at about the depth of the ion-binding site into a more significantly flared cytoplasmic side (Figure 2B). The c13 ring has an outer diameter of 63 Å on the cytoplasmic side, 54 Å on the periplasmic side, and 52 Å in the middle. Each c-subunit consists of two α-helices that are connected by a short cytoplasmic loop (RQPE, residues 34 to 37). The N-terminal α-helices (residues 1 to 33) form an inner ring, surrounded by an outer ring of the C-terminal α-helices (residues 38 to 69) in staggered position. While the inner helices are remarkably straight (Figure 3A) and only slightly tilted by ~5° toward the c-ring axis (as seen from the periplasm), the outer helices are more curved and have a kink in the middle of the helix at the key carboxylate residue (Glu54). Above that kink, toward the cytoplasmic loop, the outer helices are bent ~30° to the membrane vertical and form a convex surface shape on the outside of the ring. Toward the periplasmic side the helices remain more straight but still tilt ~10° against the membrane vertical (Figure 3). The outer surface of the c13 ring (Figure S2A) is encircled by a large hydrophobic region forming membrane borders at the level of Phe47 on the cytoplasmic side and Phe69 at the very end of the ring on the periplasmic side with a height of 34 Å. This region is the contact region of the c13 ring with the hydrophobic part of a membrane, in good agreement with previously described membrane borders for other c- (or K-) rings ,,. Except for the cytoplasmic end, the surface in the inner pore of the c13 ring is overall hydrophobic and binds detergent molecules (Figure S2B), which apparently replace phospholipids. A second leaflet of phospholipids/detergent molecules at the periplasmic side of the c-ring could only be discerned partially from the electron densities, but its existence can be inferred. In support of this notion are data showing a plug of phospholipids  in the I. tartaricus c11 ring and also atomic force microscopy topographs from Bacillus sp. strain TA2.A1 c13 ring .
Special amino acid sequence motifs are found in the c-subunits of alkaliphilic Bacillus species (Figure 1). An altered glycine motif in the inner helices is shown to affect the structure and biochemistry of c-rings ,. The 63 Å diameter of the c13 ring width (Figure 2A) is larger than expected if compared with the diameters of the c11 and c15 c-rings ,, 50 Å and 65 Å, respectively, since it is closer to the diameter of the c15 ring. The more relaxed positioning of the helices in the alkaliphile ring is caused by the structural impact of the alanine motif AxAxAxA that replaces the glycine motif GxGxGxG most often found in the first helix coding region of c-subunits. Similarly, replacement of two glycines with serines also accounts for the larger than anticipated c-ring diameter observed in the ATP synthase of Bacillus TA2.A1 . On the outer helices of the c13 ring from an alkaliphile, two prolines located one helix turn below and above the ion-binding glutamate can be identified in a motif (PxxExxP)  in which the first proline, Pro51, is specific for alkaliphiles . Both prolines break the regular α-helix hydrogen bonding pattern and cause helix bends; these two motifs of this c-ring are important factors that have an impact on the c13 diameter and ion binding as well as the overall “tulip beer glass” shape of the complex (Figure S1).
Ion binding in all c-rings includes a conserved outer helix Glu (or Asp). In the c13 ring of B. pseudofirmus OF4, this residue (Glu54) is located ~6 Å above the middle of the membrane (Figure 3) toward the cytoplasmic side. At the pH (4.5) used in the cryo-protection buffer, Glu54 is protonated. Two outer helices from neighboring c-subunits form an ion binding site (Figure 3C). During structure refinement a sphere-shaped density in 2Fobs-Fcalc as well as in the omit map remained unassigned. In close proximity of this density center (2.8-3.2 Å), four atoms were identified. One of these belongs to the side chain carboxyl oxygen (Oε2) of Glu54, whereas the three others originate from the backbone carbonyls of Leu52 and Ala53 and from the backbone nitrogen of Val56. The observed distances and the arrangement of the four associated hydrogen donor/acceptor sites around this density are in almost pyramidal arrangement and the hydrogen atom positions lie on a plane. Such an arrangement resembles internal protonated water molecules (hydronium ion)  in other proteins.
Electron densities of certain cations (e.g., Na+) or oxygen atoms from water molecules are similar and difficult to distinguish by X-ray crystallography at the given resolution. However, several lines of evidence indicate the density seen in the binding pocket of this c13 ring should be interpreted as water rather than Na+. The hydrogen-bonding distance and angle geometry for the ligands are in the typical range for water molecules in proteins, as they are also found, for example, in carboxypeptidase (Figure S3) or in the protonated water cluster of bacteriorhodopsin . In contrast, the mean distances for Na+ coordination such as those in the Na+-binding c11 ring  are significantly shorter (~2.3 Å)  than observed in the c13 ring. For direct experimental evidence we used NCD-4, a fluorescent analogue of the ATP synthase inhibitor DCCD, which is known to react covalently with protonated glutamates/aspartates in c-subunits . Figure 4 illustrates the time-dependent labeling of detergent-solubilized c13 ring at pH 6.0. Consistent with dependence of labeling on protonation of the carboxylate, an increase of pH to 9.0 immediately and significantly decelerates the reaction. This observation has also been reported by others , and a control experiment showed that the labeling by NCD-4 continued to increase linearly when the labeling time was extended to 9,000 s without an alkaline pH shift (Figure S4). The marked deceleration upon imposition of an alkaline pH shift also resembles the DCCD labeling pattern of the proton-coupled c15 ring as a function of pH (unpublished data). Most importantly, the presence of 200 mM Na+ shows no dramatic influence on the labeling kinetics and addition of 200 mM K+ leads to comparable effects to Na+. Whereas the presence of salts, either Na+ or K+, presumably causes minor changes on, for example, detergent micelles, water availability, or fluorophore quenching , the lack of a major Na+-specific effect on NCD-4 binding rates contrasts sharply with the immediate and strong Na+-protective effect of a much lower Na+ concentration (15 mM) on the Na+-binding c11 ring . This key evidence for H+ rather than Na+ binding of the B. pseudofirmus OF4 c13 ring is fully consistent with biochemical studies on the pmf- (but not smf-) dependent B. pseudofirmus OF4 cells  and its H+-dependent ATP synthase . The result furthermore suggests that the ion binding site is highly selective for H+ over Na+ essentially under any (physiological) condition with a minimum concentration excess of ~108 Na+ (200 mM) over H+ (pH 9.0). Taken together, the findings indicate that the density observed in the binding site of the c13 ring of the B. pseudofirmus OF4 ATP synthase is an oxygen atom with four valences. The data show that the proton must be located in between the atom positions of Glu54(Oε1/2) and the water oxygen (O).
In contrast to other rotor ring structures ,,, no residues from the inner helices contribute to ion coordination in the c13 ring (Figure 3B and Figure S1B). In the c11 and c15 rings a proline (Pro28 and Pro25, respectively) is involved in kinking the inner helices (not shown), thereby allowing the hydrogen bonding of the glutamine (Gln32 and Gln29, respectively) with the glutamate on the outer helix. This proline and glutamine are replaced by a glycine and valine, respectively, in the c13 ring. Consequently, the inner helices form a complete α-helical hydrogen bonding pattern, retain a straight shape, and cannot hydrogen bond with the ion-binding glutamate.
A second notable difference of the ion binding site in the c13 ring as compared with the c11 and c15 rings is visible at the second oxygen of the glutamate (Glu54 Oε1). Whereas in c11 and c15 this oxygen forms a hydrogen bond with a tyrosine from the adjacent outer helix, such an interaction is missing in the c13 ring (Figure 3C and Figure S1B). The hydrogen bonding network of Glu54 in the c13 ring is therefore reduced to one bond only compared to the others. The additional freedom allows more rotameric flexibility of the glutamate carboxylate. This property becomes impressively visible in an overlay of 13 single c-subunits taken from one asymmetric unit (Figure 5). Notably, although the carboxyl group appears in different rotameric states in the crystal structure, the distance of the closest Glu54 oxygen to the water oxygen remains in bonding distance (2.6–3.1 Å) across all binding sites of the c13 ring.
The subtle but important differences in the H-bonding network geometry allow a fine-tuning of the pK a of the carboxylic acid  and serve to optimize the required solvation energy  which is necessary to unlock the site and allow ion release and reloading during the ion translocation mechanism in the Fo complex (Figure 6 and Text S1). Fine tuning of these parameters is of crucial importance within the a/c-ring interface, where the rotor binding sites pass a more hydrophilic environment  (and J. D. Faraldo-Gómez, personal communication with TM) that is somewhat unique because of the adaptations in both the a- and c-subunits of the alkaliphile ,,. By contrast, while the ion-binding site is in contact with the hydrophobic barrier of the lipid phase, during the long rotation cycle of the rotor ring, the glutamate is expected to be neutralized. The structural data suggest that under these conditions the ion binding site of the c13 ring remains in the ion-locked state, much in analogy with the Na+- and H+-locked states in the c11 and c15 ring, respectively ,.
The amino acid residues involved in the coordination of the water are exactly at the same positions of the c13 ring as those that coordinate the water in the Na+-binding c11 ring (Figure 1 and Figure S1B) . This commonality of binding pattern underlines the evolutionary and functionally conserved relationship between the pmf- and smf-driven systems. The smf-driven ATP synthases have been suggested to be evolutionary pioneers in the establishment of the modern ATP synthases . If this hypothesis is correct, the E/D-only type such as that seen in the B. pseudofirmus OF4 c13 ring and in the mixed type (c15 ring) primarily found in light-driven systems (chloroplasts, cyanobacteria) could be derivatives of the c11 basic structure from an evolutionary point of view. Rather small differences in the amino acid sequences of the c-subunits apparently account for the different ion binding types, which are phenotypically manifested in the differently coupled and differently environmentally adapted ATP synthases.
Hydrogen atoms have a very weak X-ray diffracting power and their electron density often does not match with the exact position of the nucleus. Therefore, from a crystallographic point of view, at current resolution, the data do not allow the distinction between a protonated glutamate associating with a water molecule and a hydronium ion as a separate species. The scenario of a hydronium ion as a possible coupling ion species in F1Fo-ATP synthases was proposed for consideration by Boyer more than 20 years ago . Later, experimental differences in the pH-dependent inhibition kinetics of Na+- and H+-ATP synthases were interpreted to be in support of this hypothesis , but recent high-resolution structure data on the cyanobacterial  and chloroplast  c-rings clearly conflict with this as a general hypothesis. The possibility raised by the structural data presented here that this E/D-only type of c-ring may ultimately conform to Boyer's suggestion awaits further experimental (and/or quantum mechanical) analyses.
This work shows a new type of proton coordination in an F1Fo-ATP synthase rotor ring. An additional electron density within the protonated ion binding site corresponds to a water molecule (but not Na+). It is evident that the coordination network of the water itself, in analogy with the stable water coordination network in the Na+-binding c11 ring, is a stabilizing and therefore a structural part of this c-ring. The presence of the water has been shown to enhance the Na+-binding affinity in the Na+-binding c11 ring . Given this observation we propose that the water in the c13 ring binding pocket also enhances the proton affinity. High affinity rotor binding sites are of central importance for all ATP synthases but are especially important for ATP synthases of bacteria that grow in alkaline environments ,.
Some of the novel details of this c-ring are likely to be specific to Bacillus species growing at high pH , especially those differences in shape that relate directly to alkaliphile-specific motifs. Perhaps the novel manner in which a water participates in proton binding is also a consequence of adaptation of the ATP synthase to alkaliphily. Further structural analysis of c-rings from the large groups of non-alkaliphilic species harboring the E/D-only motif is necessary to clarify the precise role of such water molecules in the ion translocation process. It may reveal the presence of this ion binding type throughout a broader subset of H+-coupled rotors, where it influences both ion affinity and selectivity during torque generation in the Fo motors of the H+-dependent F-type ATP synthases, and possibly also for the ion-driven motors known from V-type and A-type ATPases/synthases.
The ATP synthase was purified from B. pseudofirmus OF4 in which a six histidine tag was inserted after the N-terminal methionine in the chromosomal gene encoding the β-subunit of the ATP synthase. The complex was extracted from everted vesicles with 1% β-dodecyl maltoside in the presence of 3 mg/ml soybean asolectin and purified by affinity chromatography on NiNTA agarose. The isolation of the c-ring was carried out according to . To improve purity of the sample, the c-ring was concentrated by ultrafiltration with an Amicon tube with a molecular weight cut-off of 10’000 (Millipore GmbH, Schwalbach, Germany), incubated with 1.5% Foscholine-12 (w/v) at 45°C for 10 min, and run on a sucrosegradient ,. The c-ring containing fractions were concentrated by Hydroxyapatite (BioGelHT, Bio-Rad, Munich, Germany)  and dialyzed for 12 h (10 mM Tris/HCl pH 8.0) at 4°C. The c-ring was further concentrated using polyethylene glycol (PEG)  to a final concentration of 2.5 mg/ml (bicinchoninic acid assay, Pierce, Rockford, IL, USA). Crystals were grown by vapor diffusion in hanging drops at 18°C to a size of approx. 200×100×100 µm3. The c-ring sample was supplied with 1% (w/v) of β-undecyl maltoside and mixed with crystallization buffer (0.1 M sodium acetate, pH 4.3) and 20% PEG 400 (v/v). Before flash-freezing in liquid nitrogen, the rod shaped clear crystals were transferred for 2 min into a buffer containing 30% PEG 400 (v/v), 0.1 M sodium acetate pH 4.5, and 0.05% β-dodecyl maltoside (w/v).
A 60 µl sample (0.2 mg/ml) of purified c13 from Bacillus pseudofirmus OF4 in 12.5 mM MES-HCl (pH 6.0) buffer and 0.05% β-dodecyl maltoside (w/v) was used. Continuous increase of fluorescence was recorded with an F-4500 Hitachi Fluorescence Spectrophotometer (λex=342 nm, λem=465 nm). The reaction was started by the addition of 0.6 µl of NCD-4 (Invitrogen Inc.) from a 10 mM stock solution in dimethylformamide. After 2,000 s, the rate of reaction was greatly reduced by addition of 11 µl of 1 M Tris/HCl pH 9.0. The time required for the addition of these compounds (NCD-4 and Tris buffer) was approximately 5 s in both cases.
Data to 2.5 Å resolution were collected from a single crystal at the Max-Planck beamline X10SA (PX-II) at the Swiss Light Source (SLS, Villigen, Schwitzerland) and processed using the XDS package (Kabsch, 1993). The structure was determined by molecular replacement using PHASER (McCoy, 2007) with two bundles of six subunits from the structure of the c15 ring from Spirulina platensis  as search model. Model bias was removed by density modification and solvent flattening with RESOLVE . Iterative cycles of model building and refinement were performed using COOT  and phenix.refine of the PHENIX package , respectively. During refinement, no non-crystallographic symmetry operation was applied. The refinement resulted in electron density maps that were unambiguously interpretable and after chain fitting the Ramachandran plot shows no outliers. Figures were generated using Povscript , POV-ray (http://www.povray.org), and Pymol . Electrostatic potential distribution was generated using Pymol .
The atomic coordinates and structure factors of the Bacillus pseudofirmus OF4 c13 ring have been deposited with accession code 2x2v.
Comparison of c-ring structures from Ilyobacter tartaricus , Bacillus pseudofirmus OF4, and Spirulina platensis. (A) The c-subunits are shown in ribbon representation. Side views of the c-rings from I. tartaricus (yellow, 1yce and 2wgm), B. pseudofirmus OF4 (blue, 2x2v), and S. platensis (green, 2wie). The membrane border is indicated with grey bars. (B) View on the three types of ion binding sites in F-type ATP synthases. c11, I. tartaricus; c13, B. pseudofirmus OF4; c15, S. platensis. The hydrogen bonding network is indicated by dashed lines and the ion/water molecules are shown with small spheres.
(2.44 MB TIF)
Electrostatic potential distribution of the B. pseudofirmus OF4 c13 ring surface. (A) Side view on the surface. (B) Section through the ring, same view as in (A). Detergent molecules (Foscholine-12) attached to the hydrophobic inner surface are displayed in stick representation (yellow) and helices of the c-ring in ribbon representation. Colors: red, negative; blue, positive; white, neutral. The membrane border is indicated with grey bars.
(0.37 MB TIF)
Comparison of ion coordination in B. pseudofirmus OF4 c13 ring and carboxypeptidase. (A) Ion coordination in the c13 ring from B. pseudofirmus OF4. (B) Ion coordination in carboxypeptidase A1 (PDB code 3i1u). The water oxygen at the glutamate is shown as a red sphere. Distances are given in Å. In both cases, the water oxygen has four valences for hydrogen, either four (A) or three (B) of them are forming a hydrogen bonding network with the corresponding protein (complex).
(1.67 MB TIF)
Long-term kinetics of the modification of the ion-coordinating Glu54 of detergent-solubilized c13 ring from B. pseudofirmus OF4 with NCD-4 in the absence and presence of NaCl. The fluorescence of a sample containing B. pseudofirmus c13 ring was taken every 20 min at pH 6 in the absence (yellow circles) or presence (blue squares) of 200 mM NaCl. The reaction of NCD-4 with Glu54 was started by the addition of 100 µM NCD-4 and the increase of fluorescence at λ=438 nm was followed for 9,000 s. The arrow marks the time point, at which the rate of NCD-4 labeling was reduced by shifting the pH to 9 in the experiment shown in Figure 4 (see text). For experimental details see Materials and Methods section.
(0.59 MB TIF)
Text S1 contains Figures S1–S4; Discussion of ion translocation through the B. pseudofirmus OF4 Fo complex; and References for Text S1.
(0.07 MB DOC)
Werner Kühlbrandt is acknowledged for generous support of TM's research and helpful comments on the manuscript. The authors also thank José D. Faraldo-Gómez for sharing his unpublished simulation results and contributing to the development of the microscopic model of ring rotation. Denys Pogoryelov is acknowledged for providing models of the c15 ring used for molecular replacement trials. We furthermore would like to thank the staff of the Swiss Light Source (SLS, PXII) and the European Synchrotron Radiation Facility (ESRF) for their continuous support.
The authors have declared that no competing interests exist.
This work was supported in parts by the Cluster of Excellence “Macromolecular Complexes” at the Goethe University Frankfurt (DFG Project EXC 115), the DFG Collaborative Research Center 807 (to TM), and a research grant GM28454 from the National Institute of General Medical Sciences (to TAK). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.