Catalytic Core of a Membrane-Associated Eukaryotic Polyphosphate Polymerase

See allHide authors and affiliations

Science  24 Apr 2009:
Vol. 324, Issue 5926, pp. 513-516
DOI: 10.1126/science.1168120


Polyphosphate (polyP) occurs ubiquitously in cells, but its functions are poorly understood and its synthesis has only been characterized in bacteria. Using x-ray crystallography, we identified a eukaryotic polyphosphate polymerase within the membrane-integral vacuolar transporter chaperone (VTC) complex. A 2.6 angstrom crystal structure of the catalytic domain grown in the presence of adenosine triphosphate (ATP) reveals polyP winding through a tunnel-shaped pocket. Nucleotide- and phosphate-bound structures suggest that the enzyme functions by metal-assisted cleavage of the ATP γ-phosphate, which is then in-line transferred to an acceptor phosphate to form polyP chains. Mutational analysis of the transmembrane domain indicates that VTC may integrate cytoplasmic polymer synthesis with polyP membrane translocation. Identification of the polyP-synthesizing enzyme opens the way to determine the functions of polyP in lower eukaryotes.

Inorganic polyphosphate (polyP) occurs in all life forms. In prokaryotes, polyP is a store of phosphate and energy and augments stress responses (1). In eukaryotes, polyP also acts in phosphate transport between mycorrhizal fungi and symbiotic plants (2), in osmoregulation (3), and in bone calcification (4). Large quantities of plankton-generated polyP form marine sediments (5). PolyP kinase generates bacterial polyP (6), but in eukaryotes this enzyme has only been reported in slime mold (7). Genetic screens in yeast yielded ≥250 alternative candidates whose deletion decreases cellular polyP (8, 9), but the identity of the polyP-synthesizing enzyme has remained elusive. Among the candidates is the yeast vacuolar transporter chaperone (VTC) complex, a membrane protein assembly whose deletion reduces polyP accumulation (8) but also affects membrane transport and vesicular traffic (1012). We have used x-ray crystallography to identify a polyP polymerase in VTC.

Four Saccharomyces cerevisiae Vtc proteins form hetero-oligomeric complexes (8, 10, 12). Vtc1p is a small transmembrane protein. Vtc2p, 3p, and 4p contain the transmembrane domain and additional large and sequence-related cytoplasmic segments (13). Within these segments, we proteolytically defined 35-kD-sized domains. We determined the 2.1 Å crystal structure of the fragment Vtc2p183-553 and the 2.6 Å structure of Vtc4p* (Vtc4p189-480) in the presence of Li2SO4 or adenosine triphosphate (ATP)–MnCl2, respectively (see supporting online material) (tables S1 and S2). The structures are similar (root mean square deviation ~ 1.7 Å between 241 corresponding Cα atoms) and structurally related to the RNA triphosphatase Cet1p (14). Like Cet1p, they comprise a tunnel-shaped domain formed by antiparallel β strands (fig. S1A) with the tunnel walls lined by conserved basic residues. Vtc2p183-553 and Cet1p both have a sulfate coordinated in the tunnel center (fig. S1B). In the case of Cet1p, the sulfate mimics the γ-phosphate of nucleoside triphosphates (14), which Cet1p hydrolyzes in the presence of Mn2+ (15). The Vtc4p* structure contained a long chain of electron density winding through the tunnel domain (Fig. 1A), which suggests that this module had generated phosphate polymers from ATP during dialysis or crystallization. Difference density in our structure accounts for 29 phosphate units that are bound by two neighboring molecules in the asymmetric unit (fig. S2A). Consistently, Vtc4p*-synthesized polyP can be detected in solution (Fig. 1B). Vtc4p* generates polyP from ATP in a phosphotransfer reaction releasing adenosine diphosphate (ADP) (Fig. 1C) (16). Our results define Vtc4p* as a polyP-synthesizing enzyme, an activity that rationalizes the loss of vacuolar polyP in yeast vtc4 deletion strains (8) and the concomitant increase in cellular ATP (9).

Fig. 1

Vtc4p* is a polyP polymerase. (A) Ribbon diagram of Vtc4p* with the phosphate polymer (in bonds representation) and an omit electron difference density map contoured at 4.5 σ included. The helical plug is shown in magenta. Mapping of the electrostatic potential [in kBT/e; as output by program GRASP (24)] on the surface of Vtc4p* highlights the basic tunnel center and the negatively charged polymer (B) Urea polyacrylamide gel electrophoresis of polyP generated by Vtc4p* from ATP and in the presence of Mn2+ ions. (C) High-performance liquid chromatography (HPLC) analysis of nucleotide products in the polymerization reaction. (Top) Elution profile showing the retention times for different nucleotides. (Bottom) HPLC analysis after incubating 30 μM Vtc4p* with 200 μM ATP for 0 (black line), 15 (blue), 45 (green), and 90 min (red).

We located the substrate binding site in the structure of Vtc4p* crystallized in the presence of the substrate analog adenosine-5′-[(β, γ)-imido]triphosphate (AppNHp) and MnCl2. Difference electron density accounts for the Mn2+-bound triphosphate moiety of AppNHp, which is coordinated by seven conserved basic residues and a tyrosine in the tunnel center (Fig. 2A, and fig. S3A). The nucleoside moiety appears disordered. We can define the directionality of the nucleotide by the topology of the tunnel domain that is closed by a helical segment on one end (Fig. 1A) (17). To validate the unusual substrate binding mode, we quantified interaction of Vtc4p* with different nucleotides by isothermal titration calorimetry. Vtc4p* tightly interacts with ATP, as well as with other nucleoside triphosphates [dissociation constant (Kd) ranging from 0.3 to 2 μM], with a 1:1 stoichiometry and high binding enthalpy (Fig. 2B). Vtc4p* does not discriminate ATP and its 2′-deoxy form, but the absence of a γ-phosphate substantially weakens the interaction (Fig. 2B). Mutation of Arg264 or Arg266, which contact the ATP α- and β-phosphates, to alanine decreases nucleotide binding (by a factor of 2.5 to 4) (Fig. 2B) and Vtc4p activity (Fig. 3, F and G). Their simultaneous mutation substantially impairs binding (by a factor of 20) (Fig. 2B) and inactivates Vtc4p* (Fig. 3F). Consistent with these observations, our structure reveals extensive protein contacts only with the triphosphate moiety of the nucleotide substrate (Fig. 2A).

Fig. 2

Unusual modes of nucleotide and metal cofactor binding in Vtc4p*. (A) Close-up view of Vtc4p* tunnel center bound to AppNHp (in bonds representation) and Mn2+ (gray sphere). The sugar and base portions were modeled stereochemically. The pentavalent coordination of Mn2+ is distorted square-based pyramidal. A phased anomalous difference map calculated from data collected at the Mn K edge and contoured at 15 σ is shown in green. (B) Isothermal titration calorimetry of wild-type Vtc4p* versus ATP, together with table summaries for the binding constants for different ligands and Vtc4p* mutants (versus ATP). Shown are experimental values ± fitting errors.

Fig. 3

Structural views of the Vtc4p* polymerase cycle. (A) Structure of Vtc4p* bound to orthophosphate. Coordinating residues are shown, along with the catalytic Lys458 (marked with an arrow) to facilitate comparison. (B) Structure of the AppNHp-Mn2+ complex. The Mn2+ ion is shown as a gray sphere; the position of the orthophosphate has been inferred from (A). (C) Rotated view of Vtc4p* bound to PPi. One PPi occupies the nucleotide binding site mimicking ADP; a second molecule occupies the acceptor pocket. (D) Close-up view of the polyP exit tunnel. (E) ATP turnover analyzed by HPLC at 30 μM Vtc4p* and 1 mM Mn2+ (filled circles), 1 mM additional PPi (open circles), or 10 mM EDTA (squares). (F) ATP turnover rates for various Vtc4p* mutants at 30 μM. (G) Cellular polyP content for Vtc4p point mutants expressed under the control of their native promoter in yeast BY4742 Δvtc4. Error bars represent the standard deviation derived from three independent measurements.

ATP turnover is impaired in the presence of EDTA (Fig. 3E), which indicates that Vtc4p* activity requires bivalent cations. There is moderate metal ion specificity with Mn2+ > Zn2+ > Co2+ > Mg2+ > Fe2+ > Ni2+ (fig. S4). The nucleotide α-, β- and γ-phosphates coordinate the Mn2+, which in turn makes a bidentate interaction with Glu426 (Fig. 2A). Mutation of Glu426 reduces ATP turnover and polyP synthesis (Fig. 3, F and G). Mn2+ coordination in Vtc4p* differs from the previously reported octahedral binding geometry in Cet1p and other tunnel enzymes (14, 18) (fig. S4).

We next addressed the catalytic mechanism employed by Vtc4p. A structure of Vtc4p* crystallized from Na+/K+ phosphate reveals an orthophosphate (Pi) bound close to the nucleotide binding site (Fig. 3A). Superposition with the AppNHp-bound structure shows the phosphate in this “acceptor pocket” in a binding geometry compatible with an in-line transfer of the ATP γ-phosphate in a nucleophilic displacement reaction (Fig. 3B). Consistently, we found that ATP turnover can be stimulated by “priming” the reaction with either orthophosphate (by a factor of 3 at 10 mM Pi) or pyrophosphate (by a factor of 90 at 1 mM PPi) (Fig. 3E and fig. S5). ATP cleavage is supported by Mn2+ that is positioned over the terminal scissile phosphoanhydride (Fig. 2A). Mutation of Lys458 originating from the “helical plug” and directly facing the γ-phosphate impairs ATP turnover (Fig. 3, B and F), which suggests a role in catalysis.

To further validate the key position of the acceptor phosphate, we solved the structure of Vtc4p* cocrystallized with PPi. The structure reveals difference density peaks in the substrate binding pocket and in the acceptor pocket. It may thus mimic Vtc4p* after the first round of ATP cleavage (Fig. 3C). Lys200 contacts the γ-phosphate of AppNHp (Fig. 3B) and the transferred phosphate in the acceptor pocket (Fig. 3C). Mutation of this residue to alanine reduces polyP formation and substantially enhances ATP turnover (Fig. 3, F and G), seemingly uncoupling these steps.

The substrate-bound structures suggest that the polyP complex (Fig. 1A) represents a product-bound state in the polymerization process where polyP occupies the entire tunnel domain, including the substrate binding and acceptor pockets (fig. S2B). This structure defines a path along which polyP may be transported. Several conserved basic residues align like bristles, guiding the polymer away from the active site (Fig. 3D). Notably, the tunnel domain in Vtc4p* presents itself more closed in the nucleotide-bound state, when compared with the Pi-bound structure (fig. S2C). We speculate that substrate binding and polyP transport in Vtc4p* may be driven by similar conformational changes.

We next investigated Vtc4p-catalyzed polyP formation in vivo. Yeast vtc4 deletion strains lack the entire vacuolar polyP pool (8) (Fig. 3G). Point mutations in the full-length Vtc4 protein that correspond to those found essential for Vtc4p* function in vitro (R264A, R266A, K281A, E426A, and K458L) (Fig. 3F) cannot complement this deficiency (Fig. 3G). In media with limiting phosphate concentration, where the internal polyP store is used, these point mutants also arrested growth clearly faster than wild-type cells (fig. S3B). Thus, VTC accounts for most of the polyP synthesis in yeast, and the catalytic activity resides in the tunnel domain of Vtc4p.

In fungi, polyP accumulates in the vacuole and in the extracellular space (19). Correspondingly, we detected two VTC subcomplexes composed of Vtc1/2/4p or of Vtc1/3/4p (fig. S6). Vtc2p and 3p are catalytically impaired and appear to be accessory subunits in VTC (fig. S7). We studied green fluorescent protein (GFP)–tagged Vtc2p and 3p in yeast cells by microscopy. GFP-Vtc3p predominantly stains vacuoles on Pi-rich media as well as under low Pi conditions (Fig. 4A). In contrast, GFP-Vtc2p is only partially on vacuoles but mainly around the nucleus and at the cell periphery along the plasma membrane, where it is highly enriched in patches. Upon transfer of cells to low-phosphate medium, these patches disappear and GFP-Vtc2p concentrates on vacuoles (Fig. 4A), perhaps to maximize synthesis of the vacuolar polyP store that buffers fluctuations in exogenous phosphate (20). We assume that the peripheral VTC pool may generate extracellular polyP (19). At both membranes, the catalytic tunnel domain in VTC faces the cytoplasm (13), and thus the polymer must pass the membrane. Notably, all Vtcs share related transmembrane domains that contain conserved basic residues. Vtc1 point mutations targeting these residues drastically reduce cellular polyP levels (Fig. 4B and fig. S8), while leaving VTC complexes stable and correctly sorted (fig. S6). We thus speculate that Vtc transmembrane domains participate in the transport of polyP across the membrane. Phosphate polymerization and membrane translocation would thus be orchestrated functions of VTC. Purification and biochemical reconstitution of the entire VTC complex will be necessary to test this hypothesis.

Fig. 4

Phosphate-dependent subcellular distribution of Vtc2p or 3p subcomplexes. (A) GFP-Vtc2p or GFP-Vtc3p integrated into strain BY4742 was expressed under the control of an ADH promoter. Logarithmically growing cells were transferred for 2 hours to phosphate-depleted yeast extract, peptone, and dextrose medium (YPD) replenished with 0 (-Pi) or 5 mM (+Pi) Na-phosphate pH 5.5 and analyzed by spinning disc microscopy. Nucleus (N), vacuole (V), and peripheral (P) concentrations are labeled. (B) Point mutations in the Vtc1p transmembrane domain impact cellular polyP levels. Vtc1 mutants were expressed under the control of the vtc1 promoter from pRS304 plasmids integrated into the TRP1 locus of BY4727 vtc1::HIS.

Our identification of a eukaryotic polyP polymerase now allows investigating polyphosphate metabolism in fungi such as Laccaria (21) (fig. S9) that deliver polyP to the roots of their plant hosts, in diatoms such as Thalassiosira (22) (fig. S9), whose polyP pools form marine sediments (5), and in parasites where VTC is essential for osmoregulation (23). Because the VTC complex appears not to be conserved in animals or plants, another class of enzymes remains to be discovered that catalyzes phosphate polymerization in these organisms.

Supporting Online Material

Materials and Methods

Figs. S1 to S9

Tables S1 and S2


  • * Present address: Medical Research Council Laboratory of Molecular Biology, Hills Road, Cambridge CB2 0QH, UK.

  • Present address: Novartis Pharma Schweiz AG, Monbijoustrasse 118, 3007 Bern, Switzerland.

References and Notes

  1. Using recombinant Vtc4p* in vitro, we did not detect regeneration of ATP from ADP and either a polyP 3-, ~12-, or ~65-nucleotide oligomer.
  2. Modeling of the nucleotide substrate in the opposite direction leads to severe clashes of the ribose and base portions with main chain atoms in the C-terminal helix in Vtc4p*. The position of this helical plug appears invariant in all crystal forms analyzed, making the helix unlikely to move upon substrate binding. Further, our assigned directionality of the nucleotide is consistent with the positions of the acceptor pocket and the polyP exit tunnel, respectively (Fig. 3, A and D, and fig. S3B).
  3. With the exception of the AppNHp-Mn2+ and PPi complexes (determined at the Salk Institute), all crystallographic work was carried out at the European Molecular Biology Laboratory. We thank J. Noel, T. Gibson, and T. A. Steitz for discussion and J. Chory for generous support. This work was funded by the Peter and Traudl Engelhorn Foundation (Penzberg, Germany) and a European Molecular Biology Organization long-term fellowship (M.H), the Boehringer Ingelheim Fonds (A.U.), the German Science Foundation and the Landesstiftung Baden-Württemberg (K.S.), the Swiss National Science Foundation (A.M, P.O.H), and the National Science Foundation (IOS-0649389 to J. Chory). We thank the staff at the European Synchrotron Radiation Facility, Grenoble, France, at the Deutsches Elektronen Synchrotron, Hamburg, Germany, and at the Advanced Light Source, Berkeley, for technical support; and C. Müller and S. Cusack for sharing beam time. Coordinates and structure factors for Vtc2p183-553 (PDB-ID 3g3o), Vtc4p*-polyP (3g3q), Vtc4p*-AppNHp-Mn2+ (3g3r), Vtc4p*-Pi (3g3t), and Vtc4p*- PPi (3g3u) have been deposited with the Protein DataBank (
View Abstract

Stay Connected to Science

Navigate This Article