Molybdate pumping into the molybdenum storage protein via an ATP-powered piercing mechanism

Significance

This study on the cage-like molybdenum storage protein (MoSto) provides detailed insight into how nature realizes molybdenum biomineralization. Our data support the occurrence of molybdate kinase and pyrophosphatase reactions in MoSto to pump molybdate into the locked inner protein cage against a molybdate gradient. The high molybdate concentration in the cage causes a protein-assisted self-assembly process of molybdate to polyoxomolybdate clusters by which approximately 130 Mo are deposited in a compact manner. We believe that this molybdate pumping expands the known mechanistic repertoire of ATP-powered processes, since the chemical energy of hydrolysis of the phosphoric-molybdic anhydride intermediate would be conveyed onto the molybdate for penetration of the cage wall and not onto the protein for pore opening via conformational changes.

Abstract

The molybdenum storage protein (MoSto) deposits large amounts of molybdenum as polyoxomolybdate clusters in a heterohexameric (αβ)₃ cage-like protein complex under ATP consumption. Here, we suggest a unique mechanism for the ATP-powered molybdate pumping process based on X-ray crystallography, cryoelectron microscopy, hydrogen-deuterium exchange mass spectrometry, and mutational studies of MoSto from Azotobacter vinelandii. First, we show that molybdate, ATP, and Mg²⁺ consecutively bind into the open ATP-binding groove of the β-subunit, which thereafter becomes tightly locked by fixing the previously disordered N-terminal arm of the α-subunit over the β-ATP. Next, we propose a nucleophilic attack of molybdate onto the γ-phosphate of β-ATP, analogous to the similar reaction of the structurally related UMP kinase. The formed instable phosphoric-molybdic anhydride becomes immediately hydrolyzed and, according to the current data, the released and accelerated molybdate is pressed through the cage wall, presumably by turning aside the Metβ149 side chain. A structural comparison between MoSto and UMP kinase provides valuable insight into how an enzyme is converted into a molecular machine during evolution. The postulated direct conversion of chemical energy into kinetic energy via an activating molybdate kinase and an exothermic pyrophosphatase reaction to overcome a proteinous barrier represents a novelty in ATP-fueled biochemistry, because normally, ATP hydrolysis initiates large-scale conformational changes to drive a distant process.

Nature uses ATP binding/hydrolysis and subsequent ADP/phosphate release for driving manifold biochemical processes, including those in energy metabolism, active transport, DNA replication and maintenance, translation of genetic information, motility, and protein (un)folding. An unusual ATP-powered process is accomplished by the molybdenum storage protein (MoSto) offering some N₂-fixing bacteria a pronounced selection advantage against competitors for Mo (1–3), which is only variably available in their habitats. N₂-fixing bacteria continuously demand Mo in form of molybdate for synthesizing the FeMo cofactor of nitrogenases (4, 5). MoSto use ATP hydrolysis to deposit approximately 130 Mo over longer periods in a compact and polypeptide-fixed manner as discrete, structurally diverse, and rather instable polynuclear Mo(VI)-O or polyoxometalate (POM) clusters (6–8). In comparison, Fe is biomineralized by precipitating a large and highly stable but less defined iron-oxygen adduct inside the ferritin cavity by oxidizing Fe(II) to Fe(III) (9).

The MoSto of Azotobacter vinelandii is a heterohexameric (αβ)₃ cage-like structure (7, 8). The 3 α-subunits, related by a 3-fold axis, form one-half of the cage, and the architecturally similar β-subunits form the other half in an equivalent manner (Fig. 1). The interior of the cage serves as a container for up to approximately 12 POM clusters, whose structures are determined by specific pockets inside the cage, the 3-fold symmetry, and the preparation conditions (10, 11). The polynuclear Mo-O aggregates that are covalently or noncovalently linked with the polypeptide (8, 10) vary between 3 and 14 Mo atoms and are termed Mo3, Mo8, Mo5-7, hexagonal (bi)pyramidal Mo7-8, and Mo8-14 clusters (Fig. 1).

Fig. 1.

The (αβ)₃ hexameric MoSto structure. (A) MoSto is characterized by a cage-like architecture with a completely locked cavity, formed by 3 α-subunits (green) and 3 β-subunits (blue). (B) Removing 1 α- and β-subunit at the front side of MoSto reveals the inner cage, filled with POM clusters. Mo atoms of each POM cluster are depicted as different-colored spheres (Mo3, yellow; hexagonal bipyramidal Mo8, blue; Mo5, wheat; covalent and noncovalent Mo8, cyan and orange; Mo14, pink). The topologies of 4 POM clusters are also depicted as polyhedra (in the same color) formed by oxygens at the vertices that link the metals in a corner- and edge-sharing manner. Each subunit hosts a binding site for ATP, termed α-ATP or β-ATP (sticks), accessible from the outside of the cage.

Outside the cage, these POM clusters would be instable. On the other hand, molybdate and other transition metal oxide anions spontaneously polymerize to an enormous variety of POM clusters favorably in acidic solutions. Their investigation is an old but still productive research field in inorganic chemistry (12, 13). The occurrence of POM clusters in the cage is therefore based on an interplay between the inherent property of molybdate of self-assembly and the capability of proteins to bind/template/encapsulate them (8).

Both subunits architecturally belong to the amino acid kinase family, with UMP and acetylglutamate kinases as prominent members (14, 15) and host binding sites for ATP termed α- and β-ATP, respectively. Previous studies have indicated that α-ATP is more strongly bound and β-ATP significantly more weakly bound to MoSto, and that the hydrolysis of ATP to ADP and phosphate is strictly coupled to POM cluster assembly inside the cage (16). Mg²⁺ was detectable only in the α-ATP–binding site but never in the β-ATP–binding site. Consequently, due to the strict dependency on Mg²⁺ for ATP hydrolysis, the α-ATP–binding site has been considered as the motor for molybdate pumping. The MoSto of A. vinelandii occurs in the MoSto_zero, MoSto_basal, and MoSto_funct states containing neither ATP/ADP nor POM clusters, only ATP/ADP and both ATP/ADP and POM clusters, respectively (16). In the MoSto_funct state, the cell can be supplied with molybdate on request.

Several lines of evidence indicate that POM cluster storage is separated into a rapid ATP hydrolysis-dependent molybdate transport across the proteinous cage wall and a slow protein-assisted self-assembly of the POM clusters promoted by the high molybdate concentrations inside the cage (16). Despite establishing MoSto as a soluble, ATP-driven molybdate pump, the role of the 2 different ATP-binding sites, their potential communication, the entry site of molybdate, and in particular, the coupling mechanism of ATP hydrolysis with molybdate translocation remain unknown and are largely answered in this work.

Results

Functional Analyses of the Kα45S and Kβ42S Variants.

The variants Kα45S and Kβ42S are attractive candidates for exploring the unknown purpose of the 2 ATP-binding sites, since both lysine residues are placed adjacent to the β- and γ-phosphates of ATP in their respective binding sites and may influence ATP binding and/or hydrolysis and, consequently, molybdate pumping as well. Unexpectedly, variant Kα45S could not be expressed in significant amounts. We explained this finding by a drastically reduced affinity for α-ATP on lysine exchange, resulting in denaturation of the MoSto complex. This interpretation is in line with previous studies indicating that the strongly bound α-ATP can be removed only under denaturating conditions or enzymatically when the obtained MoSto_zero is stabilized by high phosphate concentrations (16).

In contrast, the variant Kβ42S can be prepared as a stable protein complex. However, it showed neither ATP hydrolysis activity nor POM cluster formation capability in the cage, according to the malachite green assay (phosphate detection), the dithiooxamide assay (molybdate detection) (16), and a 2.1-Å X-ray structure (SI Appendix, Table S1). The variant Kβ42S structure contains the completely occupied α-ATP–binding site as well as a strongly and homogeneously bound ATP in the β-ATP–binding groove (Fig. 2) with no undefined surrounding electron density as is found in most of the previously reported MoSto structures (16). The absence of the Lysβ42 side chain implicates a shorter hydrogen bond distance between Lysβ189 and the β-phosphate oxygen and, concomitantly, a 0.5-Å shift of the entire β-ATP away from the cage (Fig. 2). This result definitively proves a pivotal function of the β-ATP–binding site in Mo pumping.

Fig. 2.

β-ATP–binding site in the structurally characterized variant Kβ42S. Variant Kβ42S (orange) has lost the capability for ATP hydrolysis and POM cluster formation, as reflected in a homogeneous, highly occupied β-ATP (carbon in green). The 2F_o − F_c electron density is drawn in gray. Compared with β-ATP of native MoSto (yellow), β-ATP of variant Kβ42S is shifted slightly away from the cage wall to form a strong hydrogen bond with Lysβ189 in the absence of Lysβ42. The introduced serine at position 42 is linked by 2 water molecules with the γ-phosphate of ATP.

The Cryoelectron Microscopy Structure of MoSto.

Cryo-grids were prepared with a freshly purified MoSto solution supplemented with molybdate and ATP/Mg²⁺ to adjust turnover conditions such that the obtained MoSto structure reflects a functionally active state. This was uncertain for the P6₃22 crystal structure, because molybdate loading in the crystalline state is infeasible. From the 1,238 micrographs recorded on a JEOL 3200 FSC microscope, 137,558 particles were selected, resulting in a 3D reconstruction at 3.2-Å resolution (Fig. 3 and SI Appendix, Table S2 and Fig. S1). MoSto was found as a heterododecameric oligomer in essentially all particles of the cryo- electron microscopy (EM) and negative-stain EM images (the latter recorded at low protein concentrations of 0.01 mg/mL). The heterododecamer is composed of a dimer of 2 (αβ)₃ hexamers arranged upside down on each other (β₃α₃α₃β₃) with the interface formed by 3 α-subunits each (Fig. 3A). Before cryo-EM studies, the oligomerization state of MoSto was not assigned unambiguously, although hints of a dodecamer existed (17). The X-ray structure was pragmatically regarded as a heterohexamer (Fig. 1) because the cage, the apparent functional unit, was included and the biological benefit of an additional hexamer was not obvious. However, the EM dodecamer is also present in the P6₃22 crystal structure formed with a 2-fold crystallographic axis between the 2 hexamers.

Fig. 3.

The cryo-EM structure of MoSto at 3.2-Å resolution. (A) The overall architecture. MoSto was found as a heterododecamer with an interface formed between the 3 α-subunits (light blue, light green, light magenta) of 2 (αβ)₃ hexamers perpendicular to the 3-fold axis. Previous gel filtration, native PAGE, and preliminary ultracentrifugation data weakly argued for a heterododecamer but were not unambiguous (17). The 3 rather hydrophobic contact areas consist of the loops preceding strand α42:α47 and strand α73:α79 of the counter-hexamer and vice versa and the loops preceding strands α266:α271 of both hexamers. Except for the slightly displaced β-ATP–binding site and the rigidified αN-terminal arm, no notable structural differences between the cryo-EM and P6₃22 X-ray structures exist, as documented in an overall rmsd of 0.49 Å. (B) The POM clusters. The cryo-EM density revealed the Mo3, the covalent and noncovalent Mo8, the hexagonal bipyramidal Mo8, and the disordered Mo5 clusters (yellow and gray). Their densities are significantly greater than that of the polypeptide, which disappears at σ > 13 to 15. The order of their occupancy approximately corresponds to the ranking seen in the X-ray structures. (C) The fixed αN-terminal arm with density (gray). In contrast to the P6₃22 X-ray structure, residues α3 to α36 of the cryo-EM structure are found in a well-defined conformation above the β-ATP–binding site. Only Thrα19 is in van der Waals contact with the ribose and α-phosphate of β-ATP. Of note, the βN-terminal arm also interacts with the αN-terminal arm of a partner αβ dimer.

The cryo-EM map clearly reflects a MoSto_funct state. Density is visible for the Mo3 cluster, covalent and noncovalent Mo8 clusters, 2 bipyramidal Mo8 aggregates, and a disordered Mo5 cluster (Fig. 3B). ATP and presumably Mg²⁺ bind to the α-ATP–binding site in a virtually identical fashion as found in the X-ray structure. The β-ATP–binding site also contains ATP but encloses a density beyond the γ-phosphate moiety. In addition, the β-ATP together with the expanded segment β193 to β225, enveloping the adenosine moiety, is shifted toward the cage wall by approximately 1.5 Å compared with the P6₃22 X-ray structure (SI Appendix, Fig. S2). Most interestingly, the N-terminal arm (residues 3 to 36) of the α-subunit, disordered in the P6₃22 X-ray structure, is well ordered in the EM structure. The αN-terminal arm is arranged in a thread-like conformation (with 1 small helical segment) that wraps around the β-subunit and thereby shields the triphosphate of β-ATP from the only solvent-accessible side (Fig. 3C). The terminal residues α6 to α11 interact with the β-subunit and the α- and β-subunits of the adjacent dimer (Fig. 3C).

Crystal Structures of MoSto.

In the P6₃22 X-ray and cryo-EM structures, the αβ dimers of the hexamer are averaged and solely ATP, but neither ADP nor molybdate was found in the ATP-binding sites, although ADP was clearly identified by HPLC studies (16). Therefore, using X-ray crystallography, we searched for new states with the aim of trapping intermediates of the multistep Mo storage process.

Unexpectedly, 2 MoSto_funct structures of the P6₃22 crystal form (SI Appendix, Table S1) contained a Mo5 cluster (the Mo atoms substantiated by anomalous data) instead of ATP in the β-ATP–binding groove (SI Appendix, Fig. S3). Three molybdate units were found to approximately superimpose with the β-ATP triphosphates and interact with Arg83 and Arg168 of the β-ATP–binding site. This state (MoSto-Mo5) provided the first hint about a molybdate-binding site at the outer cage wall. In addition, a high electron density at and around the triphosphate moiety was found in multiple MoSto structures (16), suggesting the binding of a second molecule besides ATP inside the β-ATP–binding groove.

A 1.7-Å resolution structure of MoSto_funct crystallized under molybdate-loading conditions, determined from a new crystal form of space group P2₁2₁2₁ (SI Appendix, Table S1), contained well-formed covalent Mo8 and Mo3 clusters, a rather weakly occupied noncovalent Mo8 cluster, a hexagonal-pyramidal Mo7 cluster, and a tentatively modeled Mo10 cluster. While the 6 α-ATP–binding sites of the asymmetric unit are completely occupied with ATP/Mg²⁺, the β-ATP–binding sites, unexpectedly, contain a highly-occupied ADP, a Mg²⁺, and a molybdate (Fig. 4A). Even more surprisingly, the αN-terminal arm is in a well-ordered conformation, very close to that of the cryo-EM structure. Mg²⁺ is coordinated to the α- and β-phosphate of β-ADP and 3 to 4 water ligands (Fig. 4A). The water ligands form hydrogen bonds to 7 residues surrounding them and, most interestingly, also to Thrα19 and Glnα21 from the fixed αN-terminal arm. The mutation of the water ligand Lysβ42 to serine stops ATP hydrolysis, perhaps by impairing productive Mg²⁺ binding. Molybdate binds at the bottom of the β-ATP–binding groove close to the cage wall (Fig. 4A). This binding site is coated by 7 proton-donating groups (2 of which originate from the water ligands of Mg²⁺) and thus is designed to preferentially bind unprotonated molybdate (MoO₄²⁻) instead of the partially protonated phosphate (HPO₄²⁻ or H₂PO₄⁻) (Fig. 4A). In addition, the difference Fourier peak around the central atom of 3.4σ or 20.0σ after refining with either molybdate or phosphate, respectively, along with an anomalous difference map based on a dataset (SI Appendix, Table S1) collected at a wavelength of 1.738 Å, clearly argue for molybdate (SI Appendix, Fig. S4).

Fig. 4.

The β-ATP–binding groove. (A) The ADP/Mg²⁺-molybdate–binding site in the P2₁2₁2₁ X-ray structure. Mg²⁺ (green sphere) is octahedrally ligated with the α- and β-phosphates, 3 H₂O, and a variable binding site that appears to be occupied by a water, the γ-phosphate of a potentially bound ATP or a phosphate oxygen. The molybdate (Mo in cyan) sits at the bottom of the β-ATP–binding groove in contact with the outer face of the cage wall. Hydrogen bonds are formed between the molybdate oxygens and Tyrβ86, Argβ83, Glyβ79, Thrβ169, Argβ168, and 2 Mg²⁺ ligated waters, suggesting that the binding pocket is more favorable for MoO₄^2- than for HPO₄^2- (or H₂PO₄⁻). (B) Cryo-EM map. β-ATP is clearly visible in the cryo-EM density, which is slightly prolonged. After determination of the P2₁2₁2₁ X-ray structure, the extra density was interpreted as a molybdate (Mo as a yellow sphere) localized in front of Metβ149. A second molybdate (Mo as a cyan sphere) appears to sit inside the cage contacting Metβ149 and is connected to a disordered POM cluster, identified in some X-ray structures as an Mo5 cluster.

Leuα20 fixed by Leuα15, Tyrβ86, Argβ83, and Ileβ164 are in van der Waals contact with molybdate, which becomes completely encapsulated in the reaction chamber on fixation of the αN-terminal arm. We term this trapped state locked MoSto-β-ADP/Mg²⁺-molybdate. Some of the 6 β-ATP–binding sites appear to contain a weakly occupied phosphate in van der Waals contact to ADP that partially overlaps with the molybdate-binding site. This state is termed locked MoSto-β-ADP/Mg²⁺-phosphate. In the light of these findings, the EM density beyond the γ-phosphate of β-ATP might be interpreted as a result of the averaging of ATP, ADP plus molybdate, and ATP plus molybdate (Fig. 4B).

Furthermore, a 1.9-Å resolution X-ray structure of MoSto, cocrystallized with ATP/Mg²⁺ and molybdate, was determined from a crystal form adopting the space group P6₄22 (SI Appendix, Table S1). No POM clusters were found in the cage, and the α-subunits were again completely occupied with ATP/Mg²⁺. In contrast, the content of the 3 β-ATP–binding sites ranged from nearly empty to a weakly bound ADP/Mg²⁺ dependent on the respective β-ATP–binding groove. At the bottom of all 3 β-ATP–binding sites sits a phosphate/molybdate as found in the P2₁2₁2₁ structure (SI Appendix, Fig. S5), thereby supporting the assumption of a second anion-binding site. The αN-terminal arm is not rigidified over the β-ATP–binding groove but rather is attached until Glnα18 to another heterohexamer in the crystal lattice in an obviously artificial manner. The 3 active sites enable a view of the Mo storage process after ATP hydrolysis and POM cluster degradation (SI Appendix, Fig. S5).

HDX-MS Analysis of MoSto.

HDX-MS (18) was performed to explore the conformational dynamics of MoSto during Mo pumping and the roles of the different ligands in this process (19). We systematically tested different states of MoSto and compared the deuterium uptake among them, termed reaction and control (Dataset S1). The continuous molybdate storage process was followed by preparation of a control aliquot composed of 9 µM MoSto_basal solution, 1 mM MgCl₂, 1 mM MoO₄²⁻, and a reaction aliquot also containing 1 mM ATP (Fig. 5A). Deuterium uptake was slightly increased in the α-ATP–binding pocket, particularly in segment α198 to α250, which dominantly consists of the 2 rather flexible loops, α190 to α211 and α222 to α229, encapsulating the adenosine moiety (SI Appendix, Fig. S2). Deuterium uptake was substantially decreased in the β-ATP–binding groove under molybdate loading conditions (Fig. 5A), which can be directly explained by the protection of this area on the binding of ATP. This interpretation is in line with structural and HPLC data revealing no β-ATP binding in the MoSto_basal state but binding in diverse MoSto_funct states (16). A marked decrease in HDX was found, particularly in segment β196 to β228, containing the extended loops β190 to β209 and β219 to β228 that envelop the adenosine moiety of β-ATP (Fig. 5A). Deuterium uptake was also decreased between β36 and β57, most strongly in segment β40 to β48, which contains the ATP/Mg²⁺-binding residue Lysβ42, the exchange of which to serine stops Mo storage. Most strikingly, deuterium uptake in the αN-terminal arm (α3 to α36) was strongly decreased (Fig. 5A). This is interpreted as its fixation during the molybdate pumping process.

Fig. 5.

HDX-MS analysis of MoSto. Deuterium uptake differences between reaction and control samples were determined after 90 s of incubation and projected onto the P2₁2₁2₁ MoSto crystal structure, color-coded by gradients ranging from red (decreased uptake) to white (unchanged) to blue (increased uptake). Gray-colored regions indicate missing sequence coverage. The location of Kβ42 and the gating residue Metβ149 are highlighted by a sphere. ATP, ADP (C, in yellow), Mg²⁺ (lime), and MoO₄^2- (Mo, in orange) are in a ball- and-stick representation. (A) MoSto_funct was generated from MoSto_basal supplemented with MoO₄²⁻ and Mg²⁺ by adding ATP, as shown in overall (α/β-unit; Left) and zoom-in (β-ATP– binding groove; Right) representations. (B) MoSto_funct was generated from MoSto_basal supplemented with ATP and Mg²⁺ by adding MoO₄²⁻. Again, ATP hydrolysis and POM cluster formation occurred during the incubation. (C and D) ATP (C) or MoO₄^2- (D) was added to MoSto_basal supplemented with EDTA to withdraw residual Mg²⁺ and prevent background hydrolysis.

Considering the defined conformation of the αN-terminal arm in the EM and P2₁2₁2₁ X-ray structures and the disordered conformation in the P6₃22 (8) and P6₄22 X-ray structures, we identified both as real states of the reaction cycle. Interestingly, the N-terminal extension also plays a crucial role in the related acetylglutamate kinase by linking dimers to higher oligomers (20). Finally, a decreased HDX was identified in segment β115 to β131 containing the glycine-rich loop β127 to β132, involved in POM cluster binding, which becomes occluded during continuous POM cluster assembly (11). Mo storage was then investigated by a reversed order of ATP and molybdate addition. Thus, a MoSto_basal solution was incubated with 1 mM MgCl₂ and 50 μM ATP (control aliquot), along with 1 mM MoO₄²⁻ (reaction aliquot). Subsequent HDX-MS analysis (Fig. 5B) revealed no changes in the α-ATP–binding site. No uptake changes were observed at the β-ATP–binding site (Fig. 5B), indicating molybdate-independent ATP binding. The deuterium uptake of the αN-terminal arm (polypeptide fragment α3 to α43) decreased markedly (Fig. 5B). Thus, its fixation is a result of a concerted binding of both ATP/Mg²⁺ and MoO₄²⁻, as possibly seen in the EM map. The minor HDX decrease in the β-strand containing Lysβ42 independent of the substrate order is an indicator of Mg²⁺ binding (Fig. 4A) after the binding of both ATP and molybdate. Finally, we again observed decreased deuterium uptake for segment β127 to β142, indicating continuous POM cluster assembly (11).

To further deconvolute the effects of the individual components, the specific effects of ATP, Mg²⁺, and molybdate were explored. Accordingly, a MoSto_basal solution supplemented with 1 mM Mg²⁺ prepared without (control aliquot) or with 50 µM ATP (reaction aliquot) revealed no HDX in the α-ATP–binding site, reflecting its complete occupancy with ATP. The decreased HDX in the β-ATP–binding groove, again visible in segments β190 to β209, β219 to β228, and β40 to β48, confirms the binding of ATP (SI Appendix, Fig. S6A). The observed decrease in deuterium uptake of the αN-terminal segment on ATP addition in the absence of MoO₄²⁻ was interpreted by the small MoO₄²⁻ amounts ubiquitous in aqueous solutions, which might induce ATP hydrolysis analogous to that seen in the background hydrolysis activity of MoSto before the addition of MoO₄²⁻ (16). To suppress background hydrolysis, Mg²⁺ was sequestered by adding 1 mM EDTA instead of 1 mM Mg²⁺ to MoSto_basal in the absence and presence of ATP. Strong β-ATP binding, according to an HDX decrease, was detected even at low micromolar concentrations without Mg²⁺ (Fig. 5C). The αN-terminal segment displayed no decreased HDX in the absence of Mg²⁺, however.

To investigate the effect of MoO₄²⁻ binding, a MoSto_basal solution with 1 mM Mg²⁺ (SI Appendix, Fig. S6B) or 1 mM EDTA (Fig. 5D) was prepared with MoO₄²⁻ (reaction aliquot) and without MoO₄²⁻ (control aliquot) in the absence of ATP. Only minor uptake differences in restricted segments of the protein were observed, whereas the vast majority of the complex, including the αN-terminal arm, was unaffected by adding MoO₄^2-. These results demonstrate that both the ordered and disordered αN-terminal arms reflect real states of molybdate pumping, and that the binding of all 3 ligands—ATP, Mg²⁺, and MoO₄²⁻—is necessary to fixate the αN-terminal arm over the β-ATP–binding groove.

Discussion

Before this work, the α-ATP–binding site of MoSto was considered the site of ATP cleavage, mainly due to the finding that the functionally essential Mg²⁺ was exclusively present in the α-ATP–binding site but not in the β-ATP–binding site (16). The presented site-directed mutagenesis, cryo-EM, X-ray, and HDX-MS data completely changed our view. The importance of the β-ATP–binding groove was demonstrated by the incapability of the Kβ42S variant to cleave ATP (Fig. 2), by the newly found Mg²⁺- and molybdate-binding sites therein (Fig. 4A), as well by the rigidification of the αN-terminal arm (Fig. 3C). Significant structural changes during molybdate pumping, monitored by HDX measurements (Fig. 5), were essentially restricted to the β-ATP–binding groove. The α-ATP–binding site is exclusively filled with ATP (never with ADP or without nucleotide), whose γ-phosphate is shielded from a nucleophilic attack by H₂O. Moreover, the removal of the strongly attached α-ATP is only feasible by enzymatic cleavage, leading to unfolding in the absence of phosphate (16), and the Kα45S variant is presumably unstable due to its inability to bind ATP. Taken together, our results strongly argue that ATP-fueled molybdate pumping occurs in the β-ATP–binding groove, acting as a reaction chamber, while α-ATP has a passive role, essential for structural integrity. Likewise, in ATP synthases, only β-ATP and not α-ATP is catalytically competent (21).

Our experimental results allow us to outline a comprehensive mechanistic proposal for molybdate binding-induced β-ATP cleavage and the subsequent molybdate translocation into the cage on the basis of indirect evidence (Fig. 6A): 1) Molybdate in form of MoO₄^2- or a Mo5 cluster binds to the positively charged bottom of the β-ATP–binding groove (Fig. 4). 2) β-ATP binds afterward and in case of a Mo5 cluster prebound one of the molybdate units of the thereby destroyed POM cluster is placed into the established binding site (Fig. 4). 3) The binding of both ATP and molybdate enables Mg²⁺ binding (Fig. 4A). 4) The αN-terminal arm becomes fixed (Fig. 3C), and the Mg²⁺ ligation shell, the molybdate, and the triphosphate of ATP are thereby completely packed inside the protein matrix. Based on the MoSto-β-ATP structure (PDB ID code 6GUJ) and the MoSto-β-ADP/Mg²⁺-molybdate structure (PDB ID code 6RKE), we modeled a catalytically competent MoSto-β-ATP/Mg²⁺-molybdate state by which 1 γ-phosphate oxygen occupies 1 coordination site of Mg²⁺ (Fig. 6B). 5) The encapsulation is accompanied by several conformational changes in the β-subunit, such as the swinging of the Arg83 side chain toward molybdate and the Mg²⁺-water ligands. A postulated shift of the loops β190 to β209 and β219 to β228 together with β-ATP of approximately 1 to 2 Å (SI Appendix, Fig. S2) toward the bottom of the groove presses the negatively charged β-ATP against molybdate, which may transfer the active site into an activated state. 6) The obtained active site geometry strongly suggests a nucleophilic attack of the molybdate onto the γ-phosphate of ATP (Fig. 6 A and B) by which ATP and molybdate are converted into ADP and a mixed phosphoric-molybdic anhydride. This kinase reaction likely takes place close to the thermodynamic equilibrium. 7) The instabile phosphoric-molybdic anhydride might be hydrolyzed by a water ligand of Mg²⁺, as found for inorganic pyrophosphatases (22). We postulate that the released chemical energy is directly transformed into the kinetic energy of molybdate, which penetrates the cage wall at the adjacent Metβ149, considered the sole point for molybdate in the firmly locked reaction chamber, to release the built-up pressure (Fig. 6 A and C). We imagine that the electrostatic and steric repulsion between phosphate and molybdate press the latter outward, while phosphate and ADP have no space to move. Behind the Metβ149 side chain, the rather thin wall is already pierced through, and the molybdate migrates along, for example, residues Glyβ103, Serβ104, Alaβ107, Aspβ108, and Serβ147 (some of them conformationally mobile) into the cage. In the EM map, the putative density for a molybdate inside the cage, 4.5 Å apart from Metβ149, may indicate the endpoint of the route across the wall. This molybdate is directly linked with a highly disordered Mo5 cluster (Fig. 4B). 8) After molybdate has passed Metβ149, its strained side chain immediately springs back into the original position and closes the gap in the cage wall to prevent the efflux of molybdates. 9) After losing the interaction to the channeling molybdate, Mg²⁺ is released, and the αN-terminal arm detaches and becomes disordered. 10) ADP is liberated. In each cycle, 1 molybdate is pumped per 1 ATP hydrolyzed.

Fig. 6.

Mechanism of molybdate pumping. (A) Scheme of the reaction cycle (cage wall in green; β-ATP–binding site in grey and lightbrown). Molybdate is both the substrate for the activation reaction and the metabolite to be transported. (B) Modeling of the locked MoSto-ATP/Mg²⁺-molybdate structure. The modeled γ-phosphate oxygens interact with Mg²⁺, Thrβ169, Glnβ46, Glyβ77, Alaβ78, and Argβ83. The distance between the γ-phosphate phosphorous and the closest and potentially attacking molybdate oxygen is 3.0 Å. No space for a water molecule is available between molybdate and ATP to perform ATP hydrolysis. (C) The molybdate entrance site. The modeled (m) anhydride is shown in orange, and the modeled molybdates (green) are shown in a surface representation to highlight the exit pathway. MoO₄^2- enters the cage adjacent to the noncovalent Mo8 cluster. Metβ149 is shown in the closed (o) and modeled open (m) conformations. The space for penetration is essentially created by a conformational change of the Metβ149 side chain. This process is denoted by black arrows.

The annotation of MoSto as a molybdate kinase was inspired by the architecturally related amino acid kinase family, which uses a highly similar active site, involving conserved residues to catalyze the same reaction (SI Appendix, Fig. S7) (14). In UMP kinases, ATP is attacked by the phosphate group of UMP instead of molybdate in MoSto. This leads to the formation of ADP in both reactions but forms UDP in UMP kinase instead of a phosphoric-molybdic anhydride in MoSto. While anhydride formation is shared by all amino acid kinase family members, the formation of a phosphoric-molybdic anhydride is unusual in chemistry and biology but not without precedents. It is plausibly postulated as a short-lived intermediate for the molybdate-induced increase of ATP hydrolysis in aqueous solution by a factor of 150 (23, 24), in the nucleotide-assisted molybdenum insertion into molybdopterin (25), and in the molybdolysis of ATP sulfurylase (26). The assumed adenylyl molybdate formation instead of adenylyl sulfate formation in ATP sulfurylase convincingly demonstrates the feasibility of transforming a pyrophosphate into a mixed phosphoric-molybdic anhydride in a suitable polypeptide environment. The rapid dissociation of phosphoric-molybdic anhydrides has prevented their detection; however, a related anhydride between vanadate and pyrophosphate has been identified (27). Moreover, the estimated free energy of phosphoric-molybdic anhydride hydrolysis of 5 to 7 kcal/mol is considered sufficiently exothermic to turn aside the Metβ149 side chain (28). However, further site-directed mutagenesis and theoretical studies are needed to understand, in detail, the translocation of molybdate into the cage. Remarkably, the design of the molybdate entry site from an uridine-binding site in UMP kinase is essentially achieved by restructuring the linker between strand β143 to β146 and helix β169 to β179. Overall, MoSto is an impressive example of how a molybdate pump is developed from an ATP-consuming enzyme that underpins the postulated direction of evolution from binding proteins to enzymes and finally to complex machineries (29).

In ATP-fueled cellular processes performed by proteins, ATP binding/hydrolysis or ADP/phosphate release generally induces local conformational changes, which are often transformed into rigid-body movements that are transmitted over long distances to trigger energy-requiring protein docking/undocking, ligand binding/release, or active transport events (30). For example, ABC transporters use wide-reaching polypeptide rearrangements to create a passage for ions/solutes (e.g., molybdate) across the cell membrane (31). Formerly, for MoSto we considered a similar scenario, including allosterically modulated subunit rearrangements, as used by UMP kinase for regulation purposes (32). However, MoSto operates by a fundamentally different mechanism which to our knowledge is unique in biochemistry. The chemical energy of ATP cleavage is transmitted via a kinase reaction onto a reactive anhydride intermediate and, after its hydrolysis, is directly converted via a pyrophosphatase reaction into kinetic energy of the molybdate to be pierced through the cage wall. This action mode has striking analogies to the firing of a gun. A cartridge (molybdate) is put into a box (ATP-binding groove) and locked (fixation of the αN-terminal arm). The reaction is started by pushing the bolt (ATP) toward the cartridge, and the resulting chemical reaction (phosphoric-molybdic anhydride formation) induces an explosion (anhydride hydrolysis) and an acceleration of the bullet (molybdate) through the barrel to, for example, penetrate an object (cage wall). This gunshot-like mechanism requires a locked reaction chamber to ensure the directional movement of molybdate across the cage wall and thus prevent dissipation of the released energy into heat. For the same reason, the distance between the energy source and the energy-consuming event must be short (Fig. 6A), in contrast to the separation found in many other ATP-cleaving proteins.

Materials and Methods

Recombinant MoSto and MoSto Variant Production.

The production of recombinant MoSto of A. vinelandii was performed in the Escherichia coli strain BL21(DE3) CC5, in which a pET21a-based vector containing the genes encoding for the MoSto α- and β-subunits and a strepII-tag were incorporated (11). MoSto was purified using strep-tag affinity, anion exchange, and size exclusion chromatography and stored in the MoSto_funct state in 50 mM MOPS/NaOH pH 6.5 and 50 mM NaCl at −20 °C. The POM cluster-free MoSto_basal was prepared as described previously (6). Site-directed mutagenesis was performed using the Agilent QuikChange Lightning Site-Directed Mutagenesis Kit. Amplification of the mutagenesis product was done in E. coli DH5α cells, and extraction of the amplified plasmids ere extracted using the Qiagen QIAprep Spin Miniprep Kit.

Mo Content and Kinetic Analysis.

The Mo content in the cage was determined by chemical analysis (6, 33). ATP hydrolysis rates were determined by phosphate detection with a colorimetric malachite green assay (16, 34).

Single-Particle Cryo-Electron Microscopy.

Negative staining experiments with 1% uranyl acetate revealed clearly separable particles. For single particle cryo-EM, holey carbon grids (CF-MH-4C multi C-flat; ProtoChips) were incubated for 2 h with chloroform and then glow-discharged. After being loaded with 3 μL of 4 mg/mL MoSto_funct solution supplemented with 1 mM molybdate and 1 mM ATP/Mg²⁺, they were plunge-frozen in liquid ethane in a FEI Vitrobot Mark IV at 10 °C and 70% humidity, after blotting for 11 s. Images were collected with a JEOL 3200 FSC electron microscope at 300 kV. Beam-induced motion was corrected by MotionCor2 with correction for a magnification distortion (35), which resulted in a pixel size of 1.11 Å. A total of 1,238 EM micrographs remained, from which 174,681 particles were automatically picked using Relion 2.0 (36, 37). After 2D and 3D classification, a final dataset of 137,558 particles was refined in Relion 2.0, applying D3 symmetry. Ten frames with an accumulated dose of 16 e/Ų were selected for particle ensemble movements and resolution-dependent weighting.

Crystal Structure Analysis.

MoSto was crystallized under 3 different conditions; those of the P6₃22 crystal form were reported previously (8, 10). The crystallization conditions for the 2 new crystal forms, P2₁2₁2₁ and P6₄22, are given in SI Appendix, Table S1. Crystals of the latter forms were obtained after the removal of molybdate from MoSto and subsequent reloading of MoSto with molybdate by the addition of ATP/Mg²⁺ and molybdate. Data were processed with XDS (38), and the structure was determined by molecular replacement with PHASER (39) using MoSto with PDB ID code 4NDO as a model. Structural refinement was performed with PHENIX (40) and manual model building was done with Coot (41). No POM clusters were found in the P6₄22 structure, presumably due to the high pH of the crystallization conditions. Figures were generated with PyMOL (Schrödinger) and Chimera 1.13.X (42).

Hydrogen-Deuterium Exchange MS.

To monitor the binding of ATP/Mg²⁺ and MoO₄²⁻ and thereby the conformational rearrangements involved, a 9 pmol/µL solution of MoSto_basal (50 mM MOPS and 50 mM NaCl, pH 6.5), was split into 2 identical aliquots. The reaction aliquot was supplemented with a specific reactant, and the control aliquot was supplemented with an equal amount of buffer. HDX-MS analysis was performed on a Waters HDX setup as described previously (43). In brief, samples were 15-fold diluted with the corresponding deuterated buffer at 20 °C for defined times (0, 15, 30, 90, and 240 s). Because high MoO₄^2- concentrations inhibit pepsin (44), it was added only to the samples but not to the labeling buffers. The exchange reaction was rapidly quenched by 1:1 dilution with cooled (2 °C) quench buffer (75 mM KH₂PO₄ and 75 mM K₂PO₄, pH 2.5), and 18 pmol of protein was subjected to online peptic digestion. Peptides were trapped and washed for 3 min before chromatographic separation. Eluting peptides were analyzed on a quadrupole time-of-flight mass spectrometer. All experiments were conducted in technical quadruplicates of 2 purification batches. Peptides identified from undeuterated measurements (PLGS 3.0.2; Waters) were imported into DynamX 3.0 (Waters), and the assigned uptake spectra were curated. Statistical analysis was performed in R using an unpaired t test (P ≤ 0.05, 2-sided) as described previously (43). Deuterium uptake was tracked with 240 to 340 peptides (depending on the experiment), covering approximately 95% of the protein sequence. Subsequently, significant HDX differences observed between reaction and control were plotted onto the MoSto structure (PDB ID code 6RKE) to visualize changes in structural dynamics and solvent accessibility (43).

Data Availability.

The structure factors and coordinates of the MoSto variant Kβ42S, the MoSto–β-ADP/Mg²⁺-molybdate complex (P212121), and the MoSto–β-ADP/Mg²⁺ complex (P6422) are deposited under PDB ID codes 6RIS, 6RKE, and 6RIJ. The cryo-EM map is deposited in the Electron Microscopy Data Bank under the accession number EMD-4907, and the corresponding coordinates of the MoSto–β-ATP/Mg²⁺-molybdate complex under the PDB ID code 6RKD.