ATP synthase from Trypanosoma brucei has an elaborated canonical F₁-domain and conventional catalytic sites

Significance

Mitochondria generate the cellular fuel ATP to sustain complex life. Production of ATP depends on the oxidation of energy-rich compounds to produce the proton motive force (pmf), a chemical potential difference for protons, across the inner membrane. The pmf drives the ATP synthase to synthesize ATP via a mechanical rotary mechanism. The structures and functions of the protein components of this molecular machine, especially those involved directly in the catalytic formation of ATP, are widely conserved in metazoans, fungi, and eubacteria. Here we show that the proposal that this conservation does not extend to the ATP synthase from Trypanosoma brucei, a member of the euglenozoa and the causative agent of sleeping sickness in humans, is incorrect.

Abstract

The structures and functions of the components of ATP synthases, especially those subunits involved directly in the catalytic formation of ATP, are widely conserved in metazoans, fungi, eubacteria, and plant chloroplasts. On the basis of a map at 32.5-Å resolution determined in situ in the mitochondria of Trypanosoma brucei by electron cryotomography, it has been proposed that the ATP synthase in this species has a noncanonical structure and different catalytic sites in which the catalytically essential arginine finger is provided not by the α-subunit adjacent to the catalytic nucleotide-binding site as in all species investigated to date, but rather by a protein, p18, found only in the euglenozoa. A crystal structure at 3.2-Å resolution of the catalytic domain of the same enzyme demonstrates that this proposal is incorrect. In many respects, the structure is similar to the structures of F₁-ATPases determined previously. The α₃β₃-spherical portion of the catalytic domain in which the three catalytic sites are found, plus the central stalk, are highly conserved, and the arginine finger is provided conventionally by the α-subunits adjacent to each of the three catalytic sites found in the β-subunits. Thus, the enzyme has a conventional catalytic mechanism. The structure differs from previous described structures by the presence of a p18 subunit, identified only in the euglenozoa, associated with the external surface of each of the three α-subunits, thereby elaborating the F₁-domain. Subunit p18 is a pentatricopeptide repeat (PPR) protein with three PPRs and appears to have no function in the catalytic mechanism of the enzyme.

The ATP synthases, also known as F-ATPases or F₁F_o-ATPases, are multisubunit enzyme complexes found in energy-transducing membranes in eubacteria, chloroplasts, and mitochondria (1, 2). They make ATP from ADP and phosphate under aerobic conditions using a proton-motive force (pmf), generated by respiration or photosynthesis, as a source of energy. To date, studies of the subunit compositions, structures, and mechanism of the ATP synthases have been confined mainly to the vertebrates, especially humans and bovines, and to various fungi, eubacteria, and chloroplasts of green plants. These studies have established the conservation of the central features of these rotary machines. They are all membrane-bound assemblies of multiple subunits organized into membrane-intrinsic and membrane-extrinsic sectors.

The membrane-extrinsic sector, known as F₁-ATPase, is the catalytic part in which ATP is formed from ADP and inorganic phosphate. It can be detached experimentally from the membrane domain in an intact state, and retains the ability to hydrolyze, but not synthesize, ATP. The membrane intrinsic sector, sometimes called F_o, contains a rotary motor driven by pmf and is connected to the extrinsic domain by a central stalk and a peripheral stalk. The enzyme’s rotor constitutes the central stalk and an associated ring of c-subunits in the membrane domain. The central stalk lies along an axis of sixfold pseudosymmetry and penetrates into the α₃β₃-domain, where the catalytic sites of the enzyme are found at three of the interfaces of α- and β-subunits. The penetrant region of the central stalk is an asymmetric α-helical coiled coil, and its rotation inside the α₃β₃-domain takes each catalytic site through a series of conformational changes that lead to the binding of substrates and the formation and release of ATP.

During ATP hydrolysis in the experimentally detached F₁-domain, the direction of rotation, now driven by energy released from the hydrolysis of ATP, is opposite to the synthetic sense. Extensive structural analyses, mostly by X-ray crystallography at atomic resolution, have shown that the F₁-domains of the enzymes from bovine (3–23) and yeast (24–30) mitochondria, chloroplasts (31, 32), and eubacteria (33–39) are highly conserved. Not only is there conservation of the subunit compositions of the α₃β₃-domain and the central stalk (γ₁ε₁ in eubacteria and chloroplasts, and γ₁δ₁ plus an additional unique subunit, confusingly called ε, attached to the δ-subunit in mitochondria orthologs), but also the sequences of subunits are either highly conserved or absolutely conserved in many key residues. This extensive conservation includes residues in catalytic interfaces and in the catalytic sites themselves. In the β-subunits, they include a hydrophobic pocket where the adenine ring of ADP (or ATP) is bound; a P-loop sequence that interacts with the α-, β-, and γ-phosphates of ATP and provides residues involved either directly or indirectly via water molecules in the binding of a hexa-coordinate magnesium ion; and, in the adjacent α-subunit, an “arginine finger” residue, which senses whether ADP or ATP is bound to the catalytic site. Indeed, these catalytic features are common to a wide range of NTPases (40, 41), and together with conserved structural features are characteristic of the canonical ATP synthase.

Based on a structural model at 32.5-Å resolution derived by electron cryotomography (ECT), it has been suggested recently that the structure of the F₁-catalytic domain and its catalytic mechanism in the ATP synthase from Trypanosoma brucei have diverged extensively from the canonical complex in an unprecedented manner (42). It has been proposed that the structure of this F₁-domain is much more open than those described in other species, and that the “arginine finger” is provided not by the α-subunit, but rather by an additional p18-subunit found only in the euglenozoa (43–49). Here we examine this proposal in the context of a structure of the F₁-domain of the T. brucei ATP synthase determined by X-ray crystallography at 3.2-Å resolution.

Results and Discussion

Structure Determination.

The crystals of the T. brucei F₁-ATPase have the unit cell parameters a = 124.2 Å, b = 206.4 Å, and c = 130.2 Å, with α = γ = 90.0° and β = 104.9°, and they belong to space group P2₁, with one F₁-ATPase in the asymmetric unit. Data processing and refinement statistics are presented in Table S1. The final model of the complex contains the following residues: α_E, 20–125, 137–416, and 423–560; α_TP, 22–127, 137–414, and 421–560; α_DP, 22–125, 137–416, and 424–560; β_E, 6–492; β_TP, 7–494; β_DP, 8–488; γ, 2–58 and 66–285; δ, 5–16 and 32–165; ε, 1–66; and three copies of p18, residues 6–169, 6–167, and 6–170, attached to the α_TP-, α_DP-, and α_E-subunits, respectively (see below). An ADP molecule and a magnesium ion are bound to each of the three α-subunits and to the β_TP- and β_DP-subunits, whereas the β_E-subunit has a bound ADP molecule without a magnesium ion. A similar nucleotide occupancy of catalytic and noncatalytic sites has been reported in the bovine F₁-ATPase crystallized in the presence of phosphonate (20) and in the F₁-ATPase from Caldalkalibacillus thermarum (38). These structures are interpreted as representing a posthydrolysis state in which the ADP molecule has not been released from the enzyme.

An unusual feature of the T. brucei F₁-ATPase is that the diphosphate catalytic interface is more open than the triphosphate catalytic interface, similar to the F₁-ATPase from Saccharomyces cerevisiae (24), whereas the converse is observed in all other structures (Table S2). As usual, the empty interface is the most open of the three catalytic interfaces (Table S2). The rotational position of the γ-subunit (determined by superposition of crown regions of structures) is +23.1° relative to the bovine phosphate release dwell, which is at or close to the catalytic dwell at +30° in the rotary catalytic cycle (6).

Structure of the F₁-ATPase from T. brucei.

The structure consists of an α₃β₃-complex with α- and β-subunits arranged in alternation around an antiparallel α-helical coiled coil in the γ-subunit (Fig. 1). The rest of the γ-subunit sits beneath the α₃β₃-complex and is associated with the δ- and ε-subunits. Together, these three subunits form the central stalk. Thus, the overall structure of this catalytic domain of the ATP synthase complex is very close to structures of canonical F₁-ATPases determined in the mitochondria of other species, and in eubacteria and chloroplasts. For example, in a comparison of backbone atoms with the bovine F₁-ATPase crystallized in the presence of phosphonate (20), the rmsd is 3.24 Å. As in these other canonical structures, each of the α- and β-subunits in the T. brucei F₁-ATPase has three domains. The N-terminal domain (residues 1–103 and 1–88 in α- and β-subunits, respectively) consists of a six-stranded β-barrel in both α- and β-subunits, and these six β-domains are associated in a stable annulus known as the “crown”. The central domain (residues 104–389 and 89–365 in α- and β-subunits, respectively) provides the nucleotide-binding sites (Fig. S1). The C-terminal domain consists of a bundle of seven and four α-helices in α- and β-subunits, respectively. The crown stabilizes the entire F₁-domain, and, during rotary catalysis, the rest of the α- and β-subunits swing from this crown in response to the rotation of the asymmetrical α-helical coiled-coil region of the γ-subunit.

Fig. 1.

Structure of the F₁-ATPase from T. brucei. The α-, β-, γ-, δ-, ε-, and p18-subunits are shown in red, yellow, blue, green, magenta, and cyan, respectively. (A and B) Side (A) and top (B) views in cartoon representation. (C–E) Side views in surface representation rotated 180° relative to A. (C) The bovine enzyme (12). (D and E) The T. brucei enzyme. In D, p18 has been omitted, and only additional regions not found in the bovine enzyme are colored; the rest of the structure is gray. The two additional sections in the α-subunit (red) interact with the p18-subunit. (E) p18 is present and is shown interacting with the α-subunit.

The six bound ADP molecules occupy nucleotide-binding sites that are very similar in structure to those in other ATP synthases. They retain the conventional features of a hydrophobic pocket to bind the adenine ring, and a characteristic P-loop sequence (GDRQTGKT in the α-subunit, residues 182–189; GGAGVGKT in the β-subunit, residues 162–169) interacting with the α- and β-phosphates of ADP or ATP (Fig. 2). The five magnesium ions are hexacoordinated by a threonine residue (residues 189 and 169 in α- and β-subunits, respectively) and four water molecules in each case. In the canonical enzymes, the nucleotides bound to the β-subunits participate in catalysis and exchange during a catalytic cycle, whereas those bound to the α-subunits are permanently bound to the enzyme and do not participate in catalysis. The close similarity of the structures of the T. brucei and bovine F₁-ATPases suggests strongly that the α- and β-subunits in the T. brucei enzyme have the same, or very similar, roles to those in the bovine enzyme. Thus, the nucleotide-binding sites in the β-subunits are part of the catalytic sites of the enzyme, the other important catalytic feature being αArg-386, the arginine finger residue, which is positioned in the catalytic site in the β_DP-subunit from T. brucei, for example, in exactly the same position occupied by the equivalent residue, αArg-373, in the bovine enzyme (Fig. 2).

Fig. 2.

Conservation of the noncatalytic and catalytic nucleotide-binding sites in the F₁-ATPase from T. brucei. (A) The noncatalytic site in the α_DP-subunit superposed onto the equivalent site in the bovine enzyme (12). (B) The catalytic site in the β_DP-subunit superposed onto the equivalent site in the bovine enzyme. Residue αR386 is the catalytically essential arginine finger (equivalent to αR373 in the bovine protein). Residues contributed by α- and β-subunits are shown in red and yellow, respectively (with the bovine residues in muted colors), and the bound ADP molecules are in black in the T. brucei enzymes and in gray in the bovine enzymes. The green and red spheres represent magnesium ions and water molecules, respectively (in T. brucei only). The residue numbers in parentheses denote the equivalent bovine residues.

Despite the general conservation of the structure and mechanism of the T. brucei F₁-ATPase, the euglenozoan enzyme is elaborated relative to the bovine enzyme, for example. First, the α-subunit in T. brucei is cleaved in vivo by proteolysis at two adjacent sites, removing residues 128–135 (Fig. S2) (50). The cleavage of α-subunits has been noted in other euglenozoan ATP synthases as well (48, 51–53), although the sites of cleavage have not been characterized precisely. In the bovine enzyme, the equivalent region (residues 117–123) forms an external loop (Fig. S2). These cleavages have no evident impact on the stability of either the α-subunit or the F₁-ATPase complex itself. Second, the α-, β-, δ-, and ε-subunits of the T. brucei enzyme have additional surface features that are not found in the known structures of other F₁-ATPases (Fig. 1). The most extensive are residues 483–498 and 536–560 in the C-terminal region of the α-subunit, and their significance is discussed below. The additional surface features in the β-, δ-, and ε-subunits are residues 485–499, 1–17, and 39–50, respectively. Those in the β- and ε-subunits have no obvious functions. The resolved residues of the additional sequence in the δ-subunit increases its area of interaction with the γ-subunit from 1,000 Ų to 1,700 Ų. The C-terminal region of the γ-subunit from residues 286–304, although not resolved in the structure, is 19 residues longer than in the bovine enzyme, for example, and in the intact ATP synthase it could extend beyond the crown region, possibly making contacts, permanently or transiently, during rotary catalysis with the oligomycin sensitivity conferral protein (OSCP), a component of the peripheral stalk. In other species, the OSCP is bound to the F₁-domain by the N-terminal regions of the three α-subunits (19, 29, 37, 54).

Third, and most significantly from a structural standpoint, the T. brucei F₁-ATPase has an additional p18-subunit bound to each of its three α-subunits (50). The buried surface areas of interaction of the p18-subunits with their partner α_E-, α_TP-, and α_DP-subunits are 2,500, 2,600, and 2,500 Ų, respectively. All three p18-subunits are folded into seven α-helices, H1–H7, with an unstructured C-terminal region from residues 151–170. The subunit is bound via H2 and H4 to the surface of the nucleotide-binding domain of an α-subunit and via H5 and H6 to the surface of its C-terminal domain. H7 is not in contact with the α-subunit (Fig. S2) but is bound to H6. The unstructured C-terminal tail interacts with the C-terminal domain of the α-subunit, traveling toward, but not entering, the noncatalytic interface with the adjacent β-subunit (Figs. 1 and 3 and Fig. S3). In this region, the extended C-terminal element of the p18-subunit interacts with the two additional segments of sequence (residues 483–498 and 536–560) found in the T. brucei α-subunit (Fig. S3). The first additional segment is largely extended, starting with one α-helical turn (residues 483–485). The second additional segment starts with one α-helical turn (residues 536–539), is followed by an extended region (residues 540–544), and terminates with an α-helix (residues 546–558) that doubles back into the noncatalytic interface and interacts with the extreme C-terminal end of the p18-subunit.

Fig. 3.

Structure of the p18-subunit of the F₁-ATPase from T. brucei, and its relation to a PPR protein. (A) A p18 subunit (cyan) in cartoon representation, folded into α-helices H1–H7, with an extended C-terminal region from residues 151–170, bound to the α_DP-subunit in solid representation (red). The N-terminal, nucleotide-binding, and C-terminal domains of the α-subunit are indicated by Crown, NBD, and C-ter, respectively; the bound ADP molecule is in black. (B) Comparison of the p18-subunit with an example PPR protein, the PPR10 protein from Z. mays (yellow) (57). PPR10 has 18 PPRs; the structures of PPRs 11–14 are shown (Fig. S3). The orange region represents the backbone of an eight-residue ribonucleotide bound to PPR10. The three PPRs in the p18-subunit correspond to H1 plus H2 (residues 20–28 and 33–45), H3 plus H4 (residues 52–64 and 78–93), and H5 plus H6 (residues 99–112 and 115–126). PPR10 has an additional α-helix, labeled 8, which together with α-helix 7 constitutes a fourth PPR.

Role of the p18-Subunit.

As noted previously, the sequence of the p18-subunit is related to the pentatricopeptide repeat (PPR) proteins (55), which are found in association with RNA molecules primarily in mitochondria and chloroplasts, as well as in some bacterial species. These proteins are characterized by a 35-aa degenerate sequence motif related to, but distinct from, the motif in the tetratricopeptide repeat (TPR) proteins (56). The PPR repeat is folded into a helix-turn-helix motif, and PPR proteins usually contain several tandem repeats associated into a superhelix, with a concave groove on one face that serves as a binding surface for RNA ligands. The p18-subunit of the F₁-ATPase from T. brucei is predicted to be a PPR protein with three PPRs, whereas it was previously thought to have two PPRs (50, 55). Although the probability score (49%) is rather low, as reflected in the weak correspondence of the sequences of the three predicted PPRs to the PPR consensus (Fig. S4), the topography of p18 closely follows the topographies of other well-predicted and well-established PPR proteins, such as the RNA-binding PPR protein PPR10 from Zea mays (57) (Fig. S4).

This structural comparison (rmsd 2.3 Å) illustrates that, as predicted, the p18-subunit has three PPRs, consisting of H1 plus H2, H3 plus H4, and H5 plus H6. H7 could be a relic of the first element of a fourth PPR, in which H8 has evolved into the extended C-terminal tail region of the p18-subunit (Fig. 3B). However, p18 does not have a site equivalent to the RNA-binding site in PPR10, and other residues required for RNA binding in α-helix-7 of PPR10 have been substituted in H7 of p18. Therefore, there is no evidence suggesting that p18 has any role in binding an RNA molecule, and its role in the T. brucei F₁-ATPase remains obscure, although its presence is essential for assembly of the enzyme (50). The sequences of p18-subunits, including the PPR repeats, are highly conserved across the euglenozoa, suggesting that the structure and the mode of interaction of the various p18-proteins with their cognate F₁-ATPases are conserved as well (Fig. S5).

Structure of the T. brucei F₁-ATPase and the ECT Map.

The structure of the F₁-ATPase from T. brucei described above at 3.2-Å resolution was docked into the map of the ATP synthase complex from the same organism at 32.5-Å resolution derived by ECT of mitochondrial membranes (Fig. 4). The structure of the catalytic domain described herein closely fits the region of the map with the mushroom shape, characteristic of the catalytic F₁-domain of other ATP synthases. Thus, this correspondence is also consistent with the T. brucei ATP synthase having a canonical catalytic domain elaborated by the attachment of the three p18-subunits. It does not support the proposal in Fig. 4 C and D)), where the map has been interpreted as having a catalytic domain in which the nucleotide-binding and C-terminal domains of the α-subunits are displaced outward away from the central stalk, and the role of the p18-subunit, bound in an unspecified position, is to provide the catalytically essential arginine finger residue (42).

Fig. 4.

Relationship of the crystallographic structure of the F₁-domain of the ATP synthase from T. brucei to an ECT map of the intact ATP synthase in situ in mitochondrial membranes from T. brucei. The subunits of the F₁-domain are colored as in Fig. 1. (A and B) Top (A) and side (B) views of the ECT map (gray), determined independently at 32.5-Å resolution with the crystallographic structure of the F₁-domain determined at 3.2-Å resolution docked manually inside the ECT map, with subunits α_DP and β_TP proximal to the peripheral stalk. (C and D) A published interpretation of the same ECT map proposing a structure of T. brucei F₁-ATPase in which the α-subunit is opened away from the central stalk, with the p18-subunit (not shown) contributing to the catalytic sites by providing the arginine finger residue (red circle) in D (42). The catalytic sites are indicated by green asterisks in C and by a green circle in D. PS, peripheral stalk of the enzyme. C and D modified from ref. 42.

Two other features in the ECT map can also be interpreted in terms of the characterized structures of canonical ATP synthases. First, the uninterpreted region of density above the F₁-domain in Fig. 4 corresponds to the upper part of the peripheral stalk in other ATP synthases. This region is occupied by the OSCP in the eukarya, and by the orthologous δ-subunit in eubacterial and chloroplast enzymes. As the ATP synthase in T. brucei and other euglenozoa that have been examined contain orthologs of the OSCP (49, 58, 59), it is highly probable that the T. brucei OSCP provides this feature in the ECT map and that, as in the well- characterized ATP synthases, it is attached to the F₁-domain via interactions with the N-terminal regions of the three α-subunits, which extend from the “top” of the crown domain. The role of the peripheral stalk in ATP synthases is to provide the stator of the enzyme with integrity by connecting the α₃β₃-domain to the essential ATP6 (subunit a in eubacteria, subunit IV in chloroplasts) in the membrane domain. ATP6 and orthologs, together with the c-ring in the rotor, provide the transmembrane pathway for protons (2). To maintain the integrity of this pathway, and to keep ATP synthesis coupled to proton motive force, the static ATP6 and the rotating c-ring must be kept in contact by the action of the peripheral stalk.

The peripheral stalk is the most divergent of the essential features of ATP synthases (2). Apart from the OSCP and orthologous δ-subunits, the subunit compositions, sequences, and structures of the related and structurally simpler eubacterial and chloroplast peripheral stalks differ greatly from the more complex structurally characterized peripheral stalks in mitochondrial enzymes, although they are all dominated by approximately parallel, antiparallel, and apparently rigid long α-helical structures connecting the OSCP to the ATP6 subunit (and orthologs) running alongside the catalytic domains. The peripheral stalks of ATP synthases in the mitochondria of euglenozoa (59) and in the green alga, Polytomella (60), appear to be even more divergent than those in characterized mitochondrial enzymes. Their subunit compositions are more complex, and as is evident in the map feature to the left of the F₁-domain in Fig. 4 and in other published images, they are thicker and apparently more robust than structurally characterized peripheral stalks. More details are likely to emerge in the near future, most likely from the application of cryo-EM imaging of individual particles of these enzymes. These endeavors are driven by the imperative to use knowledge of the structure of the ATP synthase from T. brucei (61, 62) to aid the development of new drugs to treat patients with sleeping sickness by finding selective inhibitors of its activity.

Materials and Methods

Crystallization of F₁-ATPase from T. brucei.

The F₁-ATPase purified from T. brucei (50) was crystallized at 4 °C by the microbatch method under paraffin oil. The enzyme was dissolved at a protein concentration of 9.0 mg/mL in buffer consisting of 20 mM Tris⋅HCl pH 7.5, 100 mM NaCl, 10 mM MgSO₄ and 1 mM ADP. This protein solution was mixed in wells in microbatch plates with an equal volume of 7.7% (wt/vol) PEG 10,000 dissolved in a buffer containing 100 mM 2-(N-morpholino)-ethanesulfonic acid pH 6.0 under a layer of paraffin oil. The plates were kept at 4 °C. Crystals appeared after 48 h, and were harvested 8 d later. They were cryoprotected by the addition of 15 µL of a solution containing 10 mM Tris⋅HCl pH 8.0, 10 mM NaCl, 5 mM MgSO₄, 0.5 mM ADP, 50 mM 2-(N-morpholino)-ethanesulfonic acid pH 6.0, 5% (wt/vol) PEG 12,000, and 30% (vol/vol) glycerol to each well. After 5 min, the crystals were harvested with a MicroLoop (MiTeGen), flash-frozen, and stored in liquid nitrogen.

Data Collection and Structure Determination.

X-ray diffraction data were collected at 100 K from cryoprotected crystals with a PILATUS3 2M detector (Dectris) at a wavelength of 0.966 Å at the European Synchrotron Radiation Facility, Grenoble, France, using the MXPressE automated screening protocol (63, 64). Diffraction images were integrated with iMOSFLM (65), and the data were reduced with AIMLESS (66). Anisotropic correction was applied using STARANISO (staraniso.globalphasing.org). Molecular replacement using the α₃β₃-domain from the structure of the ground state structure of bovine F₁-ATPase [Protein Data Bank (PDB) ID code 2JDI] was carried out with PHASER (67). Nucleotides, magnesium ions, and water molecules were removed from the model. Rigid body refinement and restrained refinement were performed with REFMAC5 (68). Manual rebuilding was performed with Coot (69), alternating with refinement performed with REFMAC5. For calculations of R_free, 5% of the diffraction data were excluded from the refinement. Additional electron density features, adjacent to the α-subunits, were attributed to p18. Initially, poly-Ala α-helices were fitted into this additional density with Coot (69), and the assignment of the direction of the α-helices was guided by secondary structure predictions performed with PSIPRED (70). This prediction also detected structural homology of p18 with PPR10 from Z. mays (PDB ID code 4M59). Stereochemistry was assessed with MolProbity (71), and images of structures and electron density maps were prepared with PyMOL (72). Structural comparisons of T. brucei F₁-ATPase with bovine F₁-ATPase inhibited with dicyclohexylcarbodiimide (PDB ID code 1E79) (12), bovine F₁-ATPase crystallized in the presence of phosphonate (PDB ID code 4ASU) (20), bovine F₁-ATPase inhibited with ADP-AlF₄ (PDB ID code 1H8E) (16), and the ground state structure of yeast F₁-ATPase (PDB ID code 2HLD) (24) and of the p18-subunit from T. brucei with PPR10 from Z. mays (PDB ID code 4M59) (57) were done with Coot (69) and PyMOL (72). The p18-subunit was assessed for the presence of PPR and TPR sequences with TPRpred (73), and α-helices were assigned according to PyMOL.