|Home | About | Journals | Submit | Contact Us | Français|
The Galacto-, Homoserine-, Mevalonate-, Phosphomevalonate-kinase (GHMP) superfamily encompases a wide-range of protein function. Three members of the family (mevalonate kinase, phosphomevalonate kinase and diphosphomevalonate decarboxylase) comprise the mevalonate pathway found in S. pneumoniae and other organisms. We have determined the 1.9 Å crystal structure of phosphomevalonate kinase (PMK) from S. pneumoniae in complex with phosphomevalonate and AMPPNP·Mg2+. Comparison of the apo and ternary PMK structures suggests that ligand binding reverses the side-chain orientations of two anti-parallel lysines residues (100 and 101) with the result that lys101 is “switched” into a position in which its ammonium ion is in direct contact with the β,γ-bridging atom of the nucleotide, where it is expected to stabilize both the ground and transition states of the reaction. Analysis of all available GHMP kinase ternary-complex structures reveals that while their Cα-scaffolds are highly conserved, their substrates bind in one of two conformations, which appear to be either reactive or non-reactive. The active site of PMK seems spacious enough to accommodate interconversion of the reactive and nonreactive conformers. A substantial fraction of the PMK active site is occupied by ordered water, which clusters near the charged regions of substrate. Notably, a water pentamer that interacts extensively with the reactive groups of both substrates was discovered at the active site.
The isoprenoid family of compounds contains roughly 25,000 distinct natural products (1, 2), and the classes within the family are often chemically sufficiently diverse to provide the extensive functional overlap often seen in robust metabolic systems. Ubiquinones, the electron carriers of redox chains, are one such class (3); the side-chains that functionalize heme rings are another (4); the isoprenoid tethers that control localization and function of proteins (5, 6) and tRNA (7-9) are a third; carotenoids (10), cholesterol (11), steroid hormones (12)), bile acids (13) - all of these and more fall within the family.
Eukaryotes, archea and a small number of bacteria convert mevalonate to isopentenyl diphosphate, the “building block” of isoprenoids, via the mevalonate pathway (Figure 1) (2, 14), which consists of three enzymes - mevalonate kinase (MK), phosphomevalonate kinase (PMK), and diphosphomevalonate decarboxylase (DPM-DC). Each of these enzymes is a member of the GHMP kinase superfamily (15), the N- and C-terminal sequence-signatures of which appear in ~ 0.75 % of the sequences in the UniProt database (16).
Given the pervasive role of isoprenoids in metabolism, it is not surprising that imbalances in isoprenoid synthesis and utilization can lead to cellular dysfunction and disease. Controlling flux through these pathways is used to prevent and cure human disease — statins, which decrease cellular levels of mevalonate by inhibiting HMG-CoA reductase (17, 18), are used widely to control cholesterol levels in humans (19), and may prove useful in the treatment of other diseases (20-24).
The mevalonate pathway is essential for the survival of Streptococcus pneumoniae in lung and serum (25, 26). PMK produces an allostere, diphosphomevalonate, that feedback-inhibits mevalonate kinase. This regulation occurs in S. pneumoniae, but not in humans (27), and is it hoped that careful study of the mevalonate pathway enzymes will provide the opportunity to intervene therapeutically in novel ways to prevent the devastating effect of this organism, which is estimated to kill ~ 3700 people daily, worldwide (28). Here we present the 1.9 Å X-ray structure of PMK in a ternary complex with phosphomevalonate and AMPPNP and compare it to the structures of the non-liganded form of the enzyme and other members of the family to assess its mechanism of action and structural traits of the family.
AMPPNP, MES, MgCl2, polyethylene glycol 4000, were purchased from Sigma-Aldrich. Phosphomevalonate was synthesized enzymatically and purified as described previously (29). The expression, in E. coli, and purification of recombinant phosphomevalonate kinase from S. pneumoniae has been described in detail (30).
Crystals of PMK in complex with Pmev and AMPPNP were prepared under silicon oil using the sitting drop method by mixing 2 μl of 12 mg/ml PMK in 25 mM Hepes/K+, pH 7.5, 0.25 mM Pmev, 8 mM AMPPNP, 10 mM MgCl2, with 2 μl of 36 % (w/v) polyethylene glycol 4000, 100 mM (MES/Na+), pH 6.0. The crystals were briefly transferred to a solution containing 36% polyethylene glycol 4000, 100 mM MES, pH 6.0 and 20% ethylene glycol prior to cooling to 125K. The crystals belonged to space group P21212 (a = 69.0, b = 115.3, c = 40.0 Å) and diffracted x-rays to 1.9 Å resolution using an in-house Rigaku R-Axis IV++ image plate detector and RU-H3R rotating anode x-ray generator equipped with Osmic Blue optics and operating at 50 kV and 100 mA. The data were reduced with HKL (31).
The structure was solved by molecular replacement with EPMR (32), using as the search probe a single subunit of apoenzyme S. pneumoniae PMK (1K47). Model building and atomic parameter refinements were carried out with ARP/wARP (33, 34), O (35) and CNS (36), employing thermal factor restraints. The Ramachandran plot of main chain torsion angle pairs places Tyr 16 in the disallowed region and Asp 297 in the generously allowed region (37, 38).
The crystal structure of PMK was determined to 1.9 Å resolution in ternary complex with the substrate Pmev, the non-hydrolyzable nucleotide analog AMPPNP and Mg2+ (Fig. (Fig.2A,2A, ,3).3). Its overall conformation is very similar to that of the apoenzyme from the same source with the exception of several conformational changes that can be directly attributed to ligand binding. Unlike the apoenzyme, which crystallized in a hexameric arrangement, the enzyme studied here crystallizes as a monomer, in agreement with gel filtration results (data not shown).
A comparison of the ternary complex and apo structures of PMK reveals the conformational changes that occur upon binding of ligands. Loops L1-3 and helix αH change noticeably. The Pmev phosphoryl-group is hydrogen bonded to both Lys 9 and the main-chain amide of Ala 293, while its carboxyl group is hydrogen bonded Ser 147. The isopentenyl body forms extensive hydrophobic interactions with Tyr 11, Tyr 16, Ile 18 and Leu 19 (Fig. 3B). Initial-rate kinetic experiments indicate that ground-state interaction between ligand and Lys 9, a highly conserved residue in the GHMP kinase family, is focused nearly exclusively on the phosphoryl-moiety of Pmev; this residue is also involved in other, as yet undefined, kcat-related steps (30). Consistent with these initial-rate findings, the structure shows the side chain of Lys 9 interacting with a phosphoryl-oxygen of Pmev (2.8 Å) and the γ-phosphoryl group of AMPPNP. Upon formation of the ternary complex, αH, the so-called “mobile” helix (30), shifts to enclose Pmev, resulting in hydrophobic contacts with Pmev (Val 217), and hydrogen bonding (Ser 213) with the terminal phosphoryl oxygens of both substrates. A 2.8-Å displacement of the nearby L3 loop appears to stabilize the position of αH through interactions of the L3, Asn 145 side chain with Ser 214, Val 217 and Gln 218 of αH.
AMPPNP interacts with both stationary and mobile regions of the PMK scaffold. The stationary region includes Ile 65 and Leu 66 as well as the glycine/serine-rich region (F102GLGSSGLV110) of motif II. The mobile regions include loops L1 and L2, which reorganize considerably to accommodate the nucleotide. L1, which is disordered over a 9-residue stretch in the apoenzyme, assumes an extended conformation in the ternary complex that places Ser 48 in hydrogen-bonding contact with adenine N1, and transfers Pro 61 and Asp 62 from a position that blocks adenine binding to one that promotes a solvent-mediated interaction between the Asp 62 carbonyl oxygen and both adenine N3 and ribose O2. Interestingly, the adenine binding site is similarly disordered in the diphosphomevalonate-bound structure of mevalonate kinase, in which the nucleotide site is empty and the acceptor site is filled. Thus, acceptor binding appears to be insufficient to reorganize the ATP-binding pocket, which is consistent with the lack of synergy in the steady-state affinities of Pmev and ATP (29). Notably, other GHMP superfamily members (galactokinase (2cz9), 4-cytidine 5′-diphospho-2C-metyl-D-erythritol kinase (1uek), and homoserine kinase (1h72)) present an open, solvent-filled framework that remodels upon ternary complex formation (39-43). It should also be noted that the presence of a crystal packing interface comprising N- and C-terminal segments of the apo-PMK L1 loop tempers predictions regarding the extent of L1 disorder under solution conditions.
Ligand binding fosters an L2 conformational change that delivers the primary ammonium ion of Lys 101 into Van der Waals contact with the α,β-bridging atom of the nucleotide tripolyphosphate chain where it can electrostatically stabilize both nucleotide binding and the development of negative charge that is expected to occur at the bridging atom in the presumed dissociative transition state (44) (Fig. 4). In the apoenzyme, Lys 100 and Lys 101 reside on a small stretch of random coil that connects the αC helix to the αB helical-turn. The lysine side chains adopt extended conformations and point away from one another, as if to maximize separation of their ammonium groups (13.9 Å). In the apo form Lys 101 is predicted to obstruct nucleotide binding, and is tethered by a salt-bridge to the carboxylate of Glu 98. Upon binding of ligand, the salt bridge is broken, the αB helix unravels, and the lysine side chains “swing” past one another into a conformation that is roughly the opposite of that seen in the apo form - Lys 101 now projects into the binding pocket and is in direct contact with the β,γ-bridging atom of AMPPNP, and Lys100 points away from the pocket and appears to be stabilized by an interaction with the backbone carbonyl oxygen of Lys 208. Thus, ligand binding is linked to the switching of Lys 101, a highly conserved residue among bacterial phosphomevalonate kinases, between what appear to be catalytic and non-catalytic positions.
The juxtaposition of the γ-phosphoryl group of AMPPNP and non-bridging phosphoryl-oxygens of Pmev in the ternary complex suggests that these groups must reorient to undergo in-line displacement chemistry, Fig. 5A. It is possible to achieve a reactive geometry by adjusting P-O-P torsion angles; however, doing so appears to require that both reactive groups be displaced from the first-coordination sphere of the metal ion. While plausible, a comparison of the available ternary-complex GHMP kinase structures offers an alternative interpretation. The PDB lists a total of six ternary-complex structures in addition to that of PMK. Three of these exhibit the “non-reactive” configuration seen in PMK (Type I, Fig. 5A); in the remaining three, the reactants are well positioned to accomplish in-line displacement (Type II, Fig. 5B). Notably, erythritol kinase (1uek) exhibits both Type I and II configurations. The fact that either or both of the Type I and II conformers are found distributed roughly evenly across these seven highly structurally conserved active sites suggests that the ability to bind either Type I or II complexes is an intrinsic property of the family.
To assess the potential of PMK to accommodate the Type II conformer, the structure of the PMK active site was compared to that of the human GalNAc kinase (45) - an example of a Type II ternary complex. Consistent with the conservation of the protein regions with which they interact, the adenosine moieties of the Type I and II nucleotides exhibit similar conformations; the conformations of the tripolyphosphate·M2+ moieties, however, are very different (see Fig. 5C). When the adenine bases are superposed, it is clear how the C4′-pucker and tripolyphosphate dihedral angles must reorient to align the nucleotide, and that such an alignment causes the divalent cation to migrate (perhaps as part of dihedral rotation) “across” the nucleotide, ultimately to embed in the opposing wall of the active site (Fig. 5E). Conservation of the Type II residues in contact with PPPi-Mg2+ in the Type I PMK structure was evaluated by aligning the Gal-NAc kinase contact residues with the corresponding resides in PMK (Fig. 5D). The seven Gal-NAc kinase residues (grey) containing atoms within 3 Å of the PPPi-Mg2+ moiety (green) are linked to it by dotted lines. Of these seven, five are spatio-chemically conserved in PMK. Of the remaining two, Ser145, which forms a hydrogen bond with the γ-P, is not conserved spatially (Gly 105 occupies this position in PMK) but its function can be provided by PMK Ser213, which, although it approaches the γ-P from the “opposite” side, is also well positioned to H-bond to it. The PMK structure does not appear to provide a good substitute for Asn233 of GK. Thus, on balance, the capacity to bind the Type II conformer appears to have been well preserved in PMK.
It is particularly interesting that the ammonium ion of PMK Lys 101 (which is switched into contact with the β,γ-bridge as a consequence of ligand binding) sits nearly exactly in the position occupied by Lys234 in the Gal-NAc kinase structure (i.e., in β,γ-bridge contact (45)); yet, these residues reside in very different regions of their respective structures. Moreover, Lys 234 is reported to adopt two conformations that place it either in or out of the active site (45). Thus, the ability to switch lysine (101 or 234) between what appear to catalytic and non-catalytic positions has been maintained; yet, the lys has migrated significantly within the GHMP framework.
Interconversion of the Type I and Type II conformations can occur either on the surface of the enzyme, in which case the conformers could represent discrete stages of binding, or via departure and rebinding of ligands, in which case the Type I represents a nonproductive complex that competes with Type II for the enzyme. Whether the PMK active site might accommodate the Type II nucleotide·Mn2+ complex of Gal-NAc kinase was assessed by positioning the nucleotide (via Cα-alignment) at the active site of the PMK ternary complex (Fig. 5E). The figure demonstrates that the cation shifts from “wall-to-wall” as a consequence of the interconversion, and that the active site seems spacious enough to accommodate a surface-interconversion of the complexes.
Twenty-one ordered water molecules fill the interstices between the Van der Waals surfaces of the PMK active site and its ligands (Fig. 6, panel A). The majority of this water is organized into clusters that contact charged regions of the ligands. In addition to ordered water, the cavity exhibits significant voids (also primarily near charged regions) that are presumed to contain disordered or rapidly exchanging solvent and/or its contents. The percentage of the surface area of the charged moieties (the tripolyphosphate of ATP and the phosphate and carbolylate of Pmev) in Van der Waals contact with the protein, as opposed to water, voids, or Mg2+, is calculated at ~ 50% (46). Thus, a significant fraction of these reactive regions are coated in a shell of water, raising the issue of how the PMK active site manages to prevent β,γ-bond hydrolysis during its catalytic cycle. It appears that the active site of PMK, and likely, to varying degrees, members of its family, has evolved to bind charged ligand-regions indirectly, through clusters of ordered water.
Of particular interest is the water pentamer, shown in olive in Figure 6A. The first coordination sphere of oxygen in liquid water is defined typically by four protons arranged in the tetrahedral geometry defined by the sp3 hybridization state of the oxygen (47, 48). These tetrahedral building blocks assemble into closed polygons (H2O)n, and extensive experimental and computational work predict that liquid water is composed largely of five- and six-membered rings that organize into cage-like structures (48-50). In the pentamer, each oxygen, or vertex, donates one of its protons to a nearest neighbor and accepts one from the other; the second proton of each oxygen radiates outward from the pentamer, like a spoke (48, 50). The radial protons alternately lie slightly above or below the ring - their interconversion is extremely rapid and occurs via quantum tunneling (50). The pentamer geometry, which is slightly puckered, is unique among small water-clusters in that it allows perfectly linear hydrogen bonds between donor and acceptor oxygens, leading to minimum inter-oxygen distances and atypically high-energy, ring-proton hydrogen bonds (48, 50-52), which, in turn, is expected to acidify the radial protons, enabling them to act as general acids.
To begin to understand how a water pentamer might interact with the PMK ligands, the pentamer oxygens were replaced with water molecules (Fig. 6B). The protons were arranged in accordance with the predicted pentamer bonding geometry (50), and oriented as if hydrogen bonded to the ligands. While the proton positioning is likely to be inaccurate in detail, the Van der Waals proximity of the radial aspect of the pentamer to the anionic oxygens of the substrate (the carboxyl- and phosphoryl-oxygens of Pmev and the β- and γ-phosphoryl oxygens of AMPPNP) supports hydrogen bonding and suggests that such structures participate directly in the binding of ligand.
To assess whether the active-site pentamer forms in the absence of ligands, the six subunits of the apo-PMK asymmetric unit were queried for its presence by aligning the Cα backbones of each subunit with that of the ternary complex and inspecting the water structure in the vicinity of the ternary-complex pentamer. Subsets of the pentamer were observed in each of the apo-subunits (Fig. 6C) with the weakest occupancy observed at position four. The alignments support formation of a pentamer-like structure even in the absence of ligand, and argue that this structure forms incompletely/transiently in preparation for the receipt of ligands that, once docked, stabilize the water scaffold on which they bind.
Comparison of the structure of the PMK ternary complex with that of apo-PMK and the ternary complexes of other GHMP kinases offers several insights into the structural basis of function in the enzyme and its family. While the Cα-architecture is conserved well across the family, the active site ligands of the ternary complexes crystallize within that architecture in one of two forms (I and II) - one of which appears poised for catalysis (Type II), while the other seems unlikely to react (Type I). The PMK active site houses twenty-one ordered water molecules and numerous voids, and appears sufficiently spacious to allow the isomers to interconvert on the surface of the enzyme, which calls into question whether the non-reactive isomer represents a binding intermediate. A pentamer of water with its edge positioned to hydrogen bond to the reactive regions of both ATP and Pmev is observed in the PMK structure - its presence strongly supports a role in ligand-binding and perhaps catalysis. Notably, a partially formed pentamer appears in the apo-enzyme, indicating that the pentamer is not “delivered” to the active site by substrate, but rather that it is largely present prior to binding and is completed, or stabilized, by the substrates as they add to, and depart from the active site.
Finally, Lys 101 appears to be switched into what appears to be a catalytic position by the binding of ligand. In the resting state of the enzyme, Lys 101 is approximately antiparallel with Lys 100, which sits in the ATP-binding pocket. Upon binding of ligand, the residues “swap” positions with the result that the ammonium ion of 101 is placed in direct contact with what would be the β,γ-bridging oxgen of ATP, where it is expected to strongly stabilize the ground and transition states of the reaction. Remarkably, a Lys similar to that of 101 (Lys 234) is delivered from a very different region of the Cα-scaffold of Gal-NAc kinase. Thus, the Lys switch has been conserved and yet allowed to migrate over evolutionary time around the edge of the active-site of the GHMP-kinase Cα-scaffold.
‡Supported by the National Institutes of Health Grant AI068989