IPK represents the newest member of the AAK superfamily to be structurally determined. The overall fold, commonly referred to as the open αβα sandwich, was first discovered in carbamate kinase from E. faecalis
). IPK is architecturally most similar to fosfomycin resistance kinase (FomA) from S. wedmorensis
with a root-mean-square deviation (rmsd) of 2.0 Å for superimposed backbone atoms (NH−Cα−C) and a sequence identity of 22%. However, it shares the highest sequence identity, 25%, with uridine monophosphate kinase (UMPK) from A. fulgidus
. Two subdivisions of the AAK superfamily exist, referred to here as the phosphate and carboxylate subdivisions, respectively. Enzymes in the phosphate subdivision, including IPK, FomA, and UMPK, catalyze phosphorylation of a phosphate or phosphonate moiety. Enzymes in the carboxylate subdivision, including carbamate kinase, N
-acetylglutamate kinase (NAGK), aspartokinase, and glutamate-5-kinase, catalyze phosphorylation of a carbamate or carboxylate group ().
Like all other AAKs, IPK adopts a dimeric quaternary structure, and each monomer folds into structurally distinct N- and C-terminal domains (). The N-terminal domain, spanning residues 1−171, binds the nucleophilic phosphate group (IP in IPK). The C-terminal domain, spanning residues 171−260, coordinates a Mg2+ ion and binds the phosphate donor ATP. Although all attempts to crystallize M. jannaschii IPK with ATP, ADP, AMPPNP, and a variety of other analogs have thus far been unsuccessful, the location of the nucleotide-binding site is structurally conserved among all family members, affording a reasonable model for ATP binding. Each monomer of IPK contains 16 β-strands, 8 α-helices, and 1 310 helix. The core of the open αβα sandwich, represented by 8 β-strands, β14, β16, β15, β11, β1, β2, β8, and β5, resides between 4 α-helices on one side, αF, αH, αE, and αD, and 3 on the other, αG, αA, and αC (). Four β-hairpins, one α-helix, and one 310 helix (η1) decorate the periphery of the central β-sandwich. Three of the hairpins, β3−β4, β6−β7, and β9−β10, lie in the N-terminal domain and surround the back wall and one side of the IP binding pocket; the αB helix covers the remaining side of the isopentenyl-binding surface. The fourth β-hairpin, β12−β13, located within the C-terminal domain, resides in close proximity to the expected location for the adenine ring of ATP. Finally, the 310 helix links one end of the central β5 strand and the β6−β7 hairpin. Although the β1−αA junction is depicted as a loop in , it also adopts a helical structure in some cases.
Figure 2 Primary sequence, tertiary architecture, and active site snapshots of IPK. a) Primary sequence of IPK from M. jannaschii aligned with E. coli NAGK. The color coding of each motif correlates with its color shown on the three-dimensional model. b) Global (more ...)
The IPK crystalline dimer is consistent with its oligomeric state deduced by gel filtration chromatography (Supplementary
Figure 1). The dimer orients around a noncrystallographic two-fold axis. This dyadic axis sits perpendicular to the extended β-sheet spanning the length of the dimer (16 β-strands with 8 β-strands per monomer). Although every AAK family member utilizes a similar dimerization interface, each dimer is unique in that its monomers orient differently with respect to one another (13
). The IPK dimer closely resembles that of UMPK, with the αC helices from each monomer crossing one another at an angle of 190° (13
) (, panel b). In IPK, this interface is comprised of 8 charged hydrogen bonds and 29 noncharged hydrogen bonds with 14 residues participating in van der Waals interactions (14
). The majority of the hydrogen-bonding interactions stitch together three structural motifs: (i) the αC helices of each monomer; (ii) the αD helix of one monomer and the β9−β10 hairpin of the dyad related monomer; and (iii) the 310
helix of one monomer and β5 of its dyadic partner. Hydrophobic interactions between the two monomers include residues from the αC and αD helices, the 310
helix, and the β4, β5, β6, β8, β9, and β10 strands. These residues form an intimate hydrophobic interface further cementing the monomers together and burying a considerable amount of accessible surface area (1869 Å2
of buried surface area per monomer).
Active Site Architecture
The refined ‘apo’ structure contains two active site sulfate molecules bound per monomer. One sulfate superimposes onto the position of the monophosphate of IP in the IP-bound structure. The second sulfate is present in all structures determined thus far and lies in the approximate location of the β-phosphate of ATP observed in other structures of AAK family members (15
). This second sulfate ion is in close proximity to Gly9, Lys6, Lys221, and Thr179 (). The equivalent residues in other AAK family members stabilize the β-phosphate of ADP or ATP analogues (PDBIDs: 2hmf
)). Therefore, the sulfate ion appears to serve as a reasonable spatial mimic of the β-phosphate of ATP and is referred to as the β-sulfate ion.
Figure 3 Comparative close-up views of the nucleotide phosphate-binding region of the IPK and FomA active sites. a) Monomer A of the IPK−IPPβS complex depicting the β-sulfate ion and the surrounding residues. b) Monomer B of the IPK−IPPβS (more ...)
The structures of IPK in complex with IP and IPP define the secondary structural elements comprising the IP-binding pocket and include the β2−αB glycine-rich loop, the αB helix, the β3−β4 hairpin, the β4−αC loop, the N-terminal section of the αC helix, and the β9−β10 hairpin (, panel a). The β2−αB loop is one of two conserved glycine-rich loops present throughout the AAK family. It is thought to contribute to charge neutralization in the transition state during phosphoryl transfer (15
). Notably, the orientation of the αB helix is conserved only in the phosphate division of the AAK superfamily (including IPK, FomA, and UMPK). In FomA, the αB helix orders when substrate is present but is otherwise disordered (15
). In contrast, the αB helix in IPK is ordered in both ‘apo’- and IP-bound structures. Of even more limited familial distribution, the β3−β4 hairpin is present only in IPK and NAGK. In NAG-bound NAGK structures, the hairpin often exists in a closed conformation (17
); in contrast, all structures of IPK reveal the motif in an open conformation. Regardless, the hairpin may play a role in shielding the substrate-binding pocket from the surrounding solvent in both enzymes.
Figure 4 a) Close-up view of the N-terminal domain depicting the isopentenyl tail and the surrounding hydrophobic residues. The motifs surrounding the active site are colored as follows: β2−αB glycine-rich loop (red), αB helix (magenta), (more ...)
The branched C5 tail of the substrate resides in a pocket surrounded principally by hydrophobic residues, including Ala63, Phe76, Met79, Phe83, Ile86, Met90, Ile146, and Ile156, (, panel a). The arrangement of residues within the cavity suggests that transmutation of the isopentenyl binding pocket to accommodate longer hydrocarbon chains may be relatively facile. The phosphate moiety of IP occupies the active site region between three short motifs including His60 and the β2−αB (residues 54−56) and β10−αE (residues 157−159) loops (, panels a−c). In both monomers, the Nε2 atom of His60 and the N atoms of Gly55 and Gly159 when protonated could stabilize the three nonbridging O atoms on the phosphate of IP through hydrogen bonds (, panels b−c). In monomer B, Lys6 also interacts with one of these nonbridging O-atoms indirectly through hydrogen-bonding interactions with an intervening water molecule; in monomer A, there is no water molecule to facilitate this interaction.
Figure 5 a) Tertiary structure superposition of monomers A and B of the IPK−IP complex. The rmsd between the two monomers is 1.31 Å. b−c) Close-up views of residues proximal to and hydrogen bonding with the α-phosphate of IP in (more ...)
In monomer A, a loop at the β1−αA junction (Gly8−Leu12), residing near the active site, can adopt two distinct substrate-binding conformations based upon refinement of alternative conformations and the observed electron density. In one conformation, the loop lies near the active site β-sulfate ion, while in the other, the loop lies closer to the β2−αB loop (, panel b). None of the residues in this loop participate in hydrogen-bonding interactions with IP; however, the dual binding conformations are not observed for the ‘apo’ structure, suggesting that loop movement is partially dependent on the presence of substrate. In monomer B, the loop adopts one binding conformation roughly equidistant between the two modes present in monomer A.
Multiple Conformations of IPP in a Single Active Site
The crystal structure of IPK with its product bound reveals that IPP adopts two distinct conformers designated conformers A and B (, panels d−f). These conformers are similar except for the orientation of the β-phosphate group and the adjacent bridging O atom. In conformer A, these two moieties sit closer to the β10−αE loop, while in conformer B, they reside closer to the β2−αB loop (, panels e−f). In both conformers, a nonbridging O atom of the β-phosphate group hydrogen bonds to the protonated Nε2 atom of His60. A superposition of the two monomers (, panel d), reveals that His60 sits in a different location in each monomer, which may reflect conformations of this residue that are dynamically accessible as the phosphorylation reaction proceeds.
In monomer A only, a water molecule rests between a nonbridging O atom from the α-phosphate of IPP and the carboxylate moiety of Asp160 (, panel e). This water molecule is also found in substrate-bound structures of FomA kinase (PDBID 3d41
), E. coli
NAGK (pdbID 1gs5
), P. furiosus
UMPK (PDBID 2bmu
), and E. coli
UMPK (pdbID 2bne
), and it is stabilized in a similar fashion in each of these structures. Asp160 of IPK is highly conserved among the AAK family and has been suggested to function as an active site base and a central organizing residue (15
As discussed previously, the β1−αA loop occupies two conformations in monomer A of the IPK−IP complex: one that places it in close proximity to the β2−αB loop, and another that interacts with the β-sulfate ion (, panel b). In monomer A of the IPK−IPP complex, the latter conformation is the major binding mode observed. The former conformation can also be seen in the electron density, although this binding mode is so subtle that it did not refine well and was not built into the final structure. This minor binding mode may, however, hold some significance, as it places Gly9 of the β1−αA loop in close proximity to the β-phosphate group in conformer B of IPP. The β1−αA loop is often reported to interact with the β- and γ-phosphate groups from ATP analogs; however, there are also examples of this loop interacting with the β-phosphate of the product (in UMPK from E. coli
The catalytically relevant conformer for IPP is most likely conformer B. This hypothesis is supported by three pieces of information: (i) The β1−αA loop, which is thought to play a key role during phosphoryl transfer, accesses a minor binding mode that is in close proximity to the β-phosphate of conformer B (19
); (ii) A superposition of UDP-bound UMPK from E. coli
and IPP-bound IPK demonstrates that the phosphate moiety of UDP superimposes with conformer B of IPP (23
); and (iii) The ATPγS/IP/Mg2+
complex structure (discussed below) exhibits clear electron density for a IPPβS molecule bound in a single conformation that superimposes with conformer B of IPP from the IPP-bound structure. Conformer A of IPP may, therefore, represent a post-reaction enzyme−product (EP) complex.
Product-Bound Active Site Containing IPPβS
When a crystal of IPK was soaked in a stabilizing solution containing IP, Mg2+, and ATPγS, the product IPPβS was observed bound in the active site. This product resembles IPP except one of the nonbridging O atoms on the β-phosphate is clearly replaced with an S atom, as evidenced by the additional electron density associated with the β-thiophosphate. Notably, no electron density for the second product ADP is seen. This is the only structure determined where both substrates, IP and ATPγS, were soaked into the crystal resulting in a catalyzed reaction in the crystal lattice. Interestingly, this structure reveals only one binding mode for IPPβS consistent with the orientation of conformer B in the IPP-bound structure. In both monomers, Gly159 from the β10−αE loop stabilizes the α-phosphate group of IPPβS; the β-thiophosphate group remains in close proximity to His60. However, in monomer B (compared to monomer A), the substrate migrates to a position that shifts away from this residue and resides closer to the β-sulfate ion (). The intermediate location of the β-phosphate group in monomer B coupled with the inferred heightened dynamics of certain loops within this monomer suggests that the monomer B structure depicts an earlier phase in the transphosphorylation reaction compared to monomer A.
His60 Plays a Key Role in Binding and Catalysis
From the results discussed above, it is evident that His60 plays an important role in both substrate and product sequestration. This binding role is accomplished through a hydrogen-bonding interaction between the protonated Nε2 atom of His60 and a nonbridging O atom from the terminal phosphate group on either the substrate (IP) or the product (IPP). His60 was mutated to Asn, Gln, and Ala; the Asn and Gln mutations are isosteric with the protonated Nε1 and Nε2 groups on His, respectively. The three mutants were assayed at 25 °C using the pyruvate kinase/lactate dehydrogenase coupled reaction to detect kinase activity. Turnover for the H60A and H60N mutants was not detected. In contrast, the H60Q mutant, whose −NH2 side-chain moiety mimics the protonated Nε2 nitrogen of His60, catalyzed a measurable transphosphorylation of IP. Notably, the apparent Km,IP values for H60Q and wild-type IPK are 34.5 and 4.3 μM, respectively, while the apparent kcat values are 0.04 s−1 and 1.46 s−1, respectively. These experimental values yield an apparent catalytic efficiency, kcat/Km, 340 times lower for the H60Q IPK mutant ().
Kinetic Data for IPK-Mj Wild-Type and H60Q at 25°C
These steady-state kinetic results suggest several interpretations for the unique role of His60 in catalysis: (i) Since both H60A and H60N exhibit no measurable activity, while H60Q remains active (albeit at a fraction of wild-type activity), binding and/or catalysis appears to be dependent on the presence of a hydrogen bond donor that is isosteric with the protonated Nε2 nitrogen of His60; 2) Given that the H60Q mutant possesses a higher apparent Km than wild-type, His60 also appears to be important for ground-state binding. Additional flexibility in the Gln side chain relative to the imidazole group of His60 may hinder its ability to bind substrate as effectively as wild-type IPK; 3) The kcat/Km value is more than 300 times higher for wild-type compared to that of the H60Q mutant, which again suggests that His60, through its added charge and lowered conformational flexibility, plays a role in stabilization of the more negatively charged transition state accompanying phosphoryl transfer.
Through comparison of the solved IPK structures, it is evident that His60 shifts from stabilizing the α-phosphate on the substrate IP to stabilizing the β-phosphate on the product IPP. In FomA, His58 (the equivalent residue to His60 of IPK) indirectly stabilizes the substrate through an intervening water molecule that is within hydrogen-bonding distance of both His58 and fosfomycin (15
). In UMPKs, an arginine residue that aligns with His60 appears to serve two roles: (i) in bacterial UMPKs, this Arg interacts with GTP, which is an allosteric activator for all bacterial UMPKs (25
); and (ii) in both bacterial and archaeal UMPKs, this Arg stabilizes the phosphate intermediate throughout the course of the reaction (23
). In comparisons of crystal structures of E. coli
UMPK, Arg62 points in opposite directions to fulfill these two roles (25
), indicating that the length and the conformational freedom of this residue are important for its functional versatility. It is therefore not surprising that a mutation to histidine at this equivalent position results in loss of GTP activation, thereby reducing the enzyme’s catalytic activity (27
The phosphate division of the AAK family encompasses the three enzymes discussed above, which are the only currently known family members that contain a residue aligning with His60 of IPK; the four family members in the carboxylate division do not possess a residue or a motif that structurally aligns with this region of IPK. Regardless of other roles that this residue may play, His60 in IPK, His58 in FomA, and the aligning arginine in all UMPKs most likely play a role in substrate/product binding and/or transition-state stabilization.
IPK Mutants Can Phosphorylate Oligoprenyl Monophosphates
As mentioned previously, the tail of the IP substrate is sequestered within a hydrophobic binding pocket (, panel a). At the back of the pocket, several residues, Ile86, Ile146, and Ile156, can be mutated to smaller amino acids to accommodate the binding of ligands with extended carbon chains (such as GP and FP). With this idea in mind, several point mutants were generated including I86A, I86G, I146A, I146G, and I156A, and most of these were able to convert FP to FPP while the wild-type enzyme lacked an equivalent activity (). These observations support the idea that mutation to smaller residues widens the cavity to allow for the binding of an extended isoprenoid tail, while mutation to bulkier residues hinders this ability. It was found that several double and triple mutant combinations displayed improved FP to FPP conversion by an order of magnitude compared to the single mutants, providing evidence that a deeper or larger cavity is more effective for FP binding and catalysis (). It is also reassuring that these mutations are contextually dependent, meaning that mutations at the very back of the active site (at position 83) are not effective unless they are present with mutations closer to the front of the active site (at positions 86, 146, or 156).
Figure 6 Farnesyl phosphate (FP) phosphorylation by IPK chain length mutants. a) The coupled IPK−sesquiterpene synthase reaction used to test for FP transphosphorylation. b) Comparative bar graph depicting several IPK tunnel mutants qualitatively tested (more ...)
Modeling of an FP molecule within the active site of IPK suggests that a C15 isoprenyl tail can orient in several different directions without introducing a large number of steric clashes. These various orientations could be explored by mutating the appropriate amino acid side chains (to relieve putative steric clashes) and by utilizing a high-throughput coupled assay to test each mutant for its ability to convert FP to FPP. The coupled kinase/terpene synthase assay () is ideal for rapid and qualitative analysis, which is necessary here since a large number of mutants must be screened to explore all possible FP orientations and mutant combinations. Thus far, we have obtained active mutants that were designed to accommodate one putative FP orientation. Crystal structures of FP-bound IPK mutants will assist in the future design of more robust mutants. Finally, exploration of the above mutations in the context of a mesophilic ortholog should afford improvement in kinetic activity for new phosphate-bearing substrates; a more accurate picture of dynamic behavior can be painted for a mesophilic IPK being studied at ambient temperature rather than thermophilic IPK (used here) being studied at ambient temperature.