A dynamic Asp–Arg interaction is essential for catalysis in microsomal prostaglandin E₂ synthase

Significance

Microsomal prostaglandin E₂ synthase type 1 (mPGES-1) is an integral membrane protein that produces prostaglandin E₂ (PGE₂), a mediator of inflammation, fever, pain, and tumorigenesis. Here we show that a serine residue implicated by the crystal structure is not required for function, whereas an arginine and aspartate residue in the active site, observed to be interacting within the crystal structure, are essential and mutually dependent during catalysis. We also demonstrate that a contact signaling network can interrupt the arginine–asparagine interaction and facilitate their participation in the chemical mechanism. Our work has broad implications for development of effective mPGES-1 inhibitors, potential drugs with clinical application in treatment of inflammatory diseases and cancer.

Abstract

Microsomal prostaglandin E₂ synthase type 1 (mPGES-1) is responsible for the formation of the potent lipid mediator prostaglandin E₂ under proinflammatory conditions, and this enzyme has received considerable attention as a drug target. Recently, a high-resolution crystal structure of human mPGES-1 was presented, with Ser-127 being proposed as the hydrogen-bond donor stabilizing thiolate anion formation within the cofactor, glutathione (GSH). We have combined site-directed mutagenesis and activity assays with a structural dynamics analysis to probe the functional roles of such putative catalytic residues. We found that Ser-127 is not required for activity, whereas an interaction between Arg-126 and Asp-49 is essential for catalysis. We postulate that both residues, in addition to a crystallographic water, serve critical roles within the enzymatic mechanism. After characterizing the size or charge conservative mutations Arg-126–Gln, Asp-49–Asn, and Arg-126–Lys, we inferred that a crystallographic water acts as a general base during GSH thiolate formation, stabilized by interaction with Arg-126, which is itself modulated by its respective interaction with Asp-49. We subsequently found hidden conformational ensembles within the crystal structure that correlate well with our biochemical data. The resulting contact signaling network connects Asp-49 to distal residues involved in GSH binding and is ligand dependent. Our work has broad implications for development of efficient mPGES-1 inhibitors, potential anti-inflammatory and anticancer agents.

Prostaglandin E₂ (PGE₂) is an abundant lipid mediator that signals via four receptors (EP1–4) to induce an array of important biological actions in physiology as well as pathophysiology (1). Under proinflammatory conditions, biosynthesis of PGE₂ proceeds from arachidonic acid, which is converted to the unstable endoperoxide PGH₂ by cyclooxygenase type 2 (COX-2). PGH₂ is further isomerized into PGE₂ by microsomal PGE synthase type 1 (mPGES-1) (2, 3). mPGES-1 is encoded by PTGES and is up-regulated by mitogens and cytokines in a pathway that is functionally coupled to COX-2 (2, 4). Because of its key role in PGE₂ synthesis, mPGES-1 has attracted attention as a potential drug target in the areas of inflammation, pain, fever, and cancer (5).

mPGES-1 is a member of the MAPEG (Membrane-Associated Proteins in Eicosanoid and Glutathione metabolism) superfamily of enzymes (6), which also includes two key proteins in the leukotriene (LT) cascade, viz. 5-lipoxygenase activating protein and LT C₄ synthase (LTC4S). All MAPEG members are integral, homotrimeric membrane proteins, and structural information on this family has been scarce. However, significant progress has recently been made in this area with several high-resolution structures being solved by X-ray crystallography (7–9). In particular, the crystal structures of human LTC4S provided detailed structural information, including an arginine residue that was later shown to activate the glutathione (GSH) thiolate (10, 11). We have previously proposed that this conserved arginine residue is also essential for enzymatic activity in mPGES-1 as Arg-126 (12). The recent structural determination of mPGES-1, however, at an exceptionally high resolution of 1.16 Å, uncovered several unanticipated structural features (13). The active sites, found at the three monomeric interfaces, show that Ser-127 is positioned near the GSH thiol group, indicating that it may act as a hydrogen-bond donor to assist in thiolate formation during catalysis. Furthermore, Arg-126 and Asp-49 participate in a charge interaction that could also contribute to catalysis. This structural information is supported by a mesophase crystal structure of an engineered version of mPGES-1 (14) and several inhibitor complexes that have recently been published (15).

Here, we initially confirmed the necessity for GSH thiolate during catalysis via incubations with the analogous tripeptide γ-Glu–Ser–Gly (GOH). We then used site-directed mutagenesis to analyze the functional roles of active site residues. We found that Ser-127 is nonessential for catalysis, whereas Arg-126 and Asp-49 are crucial and mutually dependent for native isomerase activity of the enzyme. Because the latter codependence of activity could be rationalized by a dynamic functional role of these residues, we turned to the high-resolution X-ray data (13, 15) deposited in the Protein Data Bank (PDB; www.rcsb.org) (16) to provide evidence of their dynamic motion within the crystal structure. Several recent studies have shown that the information present in such data is often underestimated and that it is possible to refine multiple conformations of residues simultaneously, each with individually refined occupancy and B factors, without overfitting the data (17–22). By such sampling of low-level electron density, discrete, “hidden” conformations are revealed, facilitating a more quantitative representation of dynamic motion within the crystal lattice. Furthermore, it has been shown that this information is often essential for understanding enzymatic function (23) and mechanism (24, 25) and successfully achieving structure-based drug design (26).

We quantified the dynamic conformations of active site residues using the software qFit (20) and CONTACT (27). This method generated significant improvements in the quality indicators for PDB ID codes 4AL0 and 4AL1 [1.16 and 1.95 Å, respectively (13)] and revealed a ligand-dependent contact network that corroborates the mechanism suggested from biochemical data. These van der Waals interactions within the binary complex with GSH (PDB ID code 4AL0) reveal an extensive network of correlated side-chain motions within the cytoplasmic “C-domain” that forms the bottom of the active site and confirm a dynamic role of Asp-49 in catalysis. In comparison, the much smaller networks found in PDB ID code 4AL1, with a bisphenyl–GSH analog and detergent molecule bound in the active site, indicate that ligand binding can influence network signaling. This finding is also supported by our analysis of recently published inhibitor complexes (15).

Our results suggest that the positively charged Arg-126 stabilizes transient thiolate formation and that its dynamic interaction with Asp-49 is essential for catalysis. We also observed a crystallographic water molecule that is ideally situated to act as a proton acceptor during this process. Furthermore, we found a striking contact signaling network within the active site that effects the conformation of residues in a ligand-dependent fashion.

Results and Discussion

GSH Thiolate Is Essential for Catalysis.

Fig. 1 depicts a substrate-limited assay that measured absolute product formation (nanograms) as described in Materials and Methods. Incubations with native enzyme in the presence of GSH resulted in near total conversion of PGH₂ to PGE₂ and are concurrent with the specific activity previously reported in the literature (∼4 µmol⋅min⁻¹⋅mg⁻¹ at 0 °C) (28). A negative control involved microsomal preparations of WT mPGES-1 resuspended in buffer containing the GSH analog, GOH, that differs from the native cofactor only by the replacement of the thiol moiety by a hydroxyl group and did not produce product above background levels. This finding provides strong evidence that thiolate anion is the chemical species of the cofactor essential for catalysis.

Fig. 1.

Mutagenic analysis of mPGES-1 active site residues. Aliquots of WT, S127A, D49N, R126Q, and R126K mPGES-1 were incubated with 12 µM PGH₂ and analyzed for PGE₂ formation by GC-MS, as described in Materials and Methods. The total amount (nanograms) of PGE₂ formed is shown from both purified and microsomal preparations of enzyme. PGE₂ formation was also monitored for microsomal preparations of WT enzyme incubated with GOH. Values represent the combined measurements from at least three different preparations of enzyme (n = 3), with error bars representing their SD. The levels of PGF₂α formed were also measured by this method as shown in Fig. S1.

Ser-127 Is Nonessential for Catalysis.

Judging from the orientation of Ser-127 in the recently published crystal structure (13), the authors’ hypothesis that its hydroxyl group may act as a hydrogen-bond donor to stabilize a GSH thiolate is apt, because this mechanism of thiol activation is a common theme within the evolution of soluble GSH transferases (29). However, because we had previously proposed Arg-126 as a strong candidate for this role (12), we investigated the function of Ser-127 in the conversion of PGH₂ into PGE₂.

To detail the role of Ser-127 in mPGES-1, we exchanged this residue for an alanine by site-directed mutagenesis. After expression in Pichia pastoris and purification, aliquots of recombinant protein were incubated with PGH₂. Formation of PGE₂ was analyzed by GC-MS. The combined measurements obtained from at least three different preparations of enzyme are depicted in Fig. 1 and show that Ser-127–Ala exhibits the same level of PGE₂ synthase activity as WT mPGES-1. This finding was true for both purified and microsomal preparations of the enzyme (Fig. 1). In addition, the dual conformations observed for this residue in the crystal structure indicate the absence of a strong hydrogen-bonding interaction. Conversely, Arg-126 is observed in a single conformation with an Nη-GSH thiol distance of 3.4 Å. We believe that this active site geometry also substantiates strong evidence for a mechanism of GSH thiol activation by an Arg-126 guanidinium interaction (30). Hence, despite compelling structural evidence, Ser-127 does not play a critical role in mPGES-1 catalysis.

Mutation of Arg-126 and Asp-49 Compromises PGE₂ Synthase Activity, but Allows PGH₂ Reduction to PGF₂α.

In light of the new structural data (13), we wanted to reexamine the functional role of Arg-126 and mutated this residue into both a glutamine and a lysine residue using site-directed mutagenesis. According to the crystal structure of mPGES-1 (13), Arg-126 and Asp-49 participate in an intermonomeric charge interaction. Therefore, we also mutated the negatively charged counterpart, Asp-49, into an asparagine residue. We anticipated that the size and charge conservative mutations of these residues could serve in probing their role in the enzymatic mechanism, while minimizing steric and electrostatic repulsion effects, such as disruption of the monomer interface. Although we also attempted to create the charge conservative mutant Asp-49–Glu, the resulting transformed construct failed to express, presumably because it resulted in an unstable quaternary structure.

After solubilization with detergent and purification via Ni-affinity chromatography, these mutants were assayed for PGE₂ synthase activity as described above. For three different purifications of each isoform, we found that the mutated enzymes did not convert PGH₂ into PGE₂ above background levels. After preparations of microsomal fractions, however, we found that the charge conservative mutation Arg-126–Lys still retained a low level of isomerase activity, indicating that a native membrane environment and a formal positive charge at position 126 are important factors for catalysis (Fig. 1). From these results, we conclude that both Arg-126 and Asp-49 are key to the PGE₂ synthase activity of mPGES-1.

That both of these residues are essential for catalysis is intriguing, because one could expect Arg-126 to be precluded from participating in thiolate stabilization if it was already engaged in a stable salt bridge interaction with Asp-49. Analysis of the relative torsional angles, however, shows the out-of-plane angle of the Asp-49 carboxylate relative to the Arg-126 guanidinium to be 44.7° (Fig. 2A). This value is far in excess of the ∼8–10° found to be typical of bidentate interactions for structures of a similar resolution as reported in a recent comprehensive review (31). Therefore, the Asp-49–Arg-126 interaction cannot be classified as the energetically stable, bidentate interaction of a formal salt bridge and implies that the energetic barrier for its disruption would be low. This finding provides evidence for the capacity of these residues to dynamically participate in active site chemistry and is corroborated by conformational fitting with qFit (27), which reveals hidden conformations of both residues depending on the identity of the adjacent ligand (Fig. 2).

Fig. 2.

The active site architecture of mPGES-1. The active sites of PDB ID codes 4AL0 (A) and 4AL1 (B) are compared with post-qFit conformational fitting and refinement. The coordinates of PDB depositions are overlaid in translucent over the refined qFit ensembles shown as opaque conformers in stick representation. The 2mFo-DFc electron density corresponding to a mechanistically relevant solvent molecule is shown as blue mesh contoured at 1 rmsd, and the rotation plane of the R126 guanidinium is shown relative to the carboxylate of D49 (44.7°) (A). This water is absent within the qFit-refined bis-phenyl complex (B), potentially due to a reduced capacity for GSH thiol interaction. A molecule of octyl glucoside bound at the C-domain and low occupancy GSH (∼13%) within the active site of B have been omitted for clarity (cf. Fig. S3). Polar interactions are shown as dashed lines with distances given in Å.

As we had previously observed for Arg-126 mutants (12), we found that other catalytically inactive mutants assayed in this study displayed a promiscuous reductase activity, converting PGH₂ into PGF₂α. Notably, the most pronounced activity in this respect was again observed for microsomal preparations of the Arg-126–Lys mutant (Fig. S1).

Fig. S1.

Mutagenic analysis of mPGES-1 PGF₂α formation. Aliquots of WT, S127A, D49N, R126Q, and R12K mPGES-1 were incubated with 12 µM PGH₂ and analyzed for PGF₂α formation by GC-MS. The total amount (nanograms) of PGF₂α formed is shown by both detergent-solubilized and microsomal preparations of enzyme. Values represent the combined measurements from at least three different preparations of enzyme (n = 3), with error bars representing their SD.

We confirmed that all mutants possessed the same tertiary fold as native enzyme via comparison of circular-dichroism spectra (Fig. S2).

Fig. S2.

Overlay of CD spectra of purified mPGES-1 compared with active site mutants. Comparison of CD spectra of native mPGES-1 with those of active site mutants indicates no major structural changes have resulted from site-directed mutagenesis. Small differences in the spectra are likely due to experimental error during the measurement of protein concentration.

A Crystallographic Water Molecule Is Ideally Situated to Participate in the Mechanism.

Analysis of the active site architecture also suggests that the α-carboxylate of GSH is involved in thiolate formation, via a tightly bound crystallographic water (2F_o − F_c peak of ∼5 rmsd, ADP = 21.9 Ų) within the active site (Fig. 2A). By forming a hydrogen-bonding network from the α-carboxylate moiety of GSH to its thiol group, it is ideally placed to assist in deprotonation of the latter during catalysis. The pK_a of the α-carboxylate, in turn, is undoubtedly lowered by the side-on, out-of-plane interaction with the guanidinium of Arg-38 (torsion angle 45.9°), which is itself engaged in solvent-mediated interactions with the main-chain carboxyl groups of Ala-43 and Arg-60. This architecture is highly reminiscent of the “electron-sharing network” that is functionally conserved in all classes of soluble GSTs for the same purpose (32), and the use of a bridging water molecule to transfer the thiol proton to the α-carboxylate of GSH has been shown to be energetically favorable within an alpha class soluble GST (33). Crucially, density for this water is absent for the Phenix (34) refined bis-phenyl GSH complex (PDB ID code 4AL1), in which the relative occupancies to GSH were refined as 0.87:0.13, respectively. After conformational change of Asp-49, we hypothesize that Arg-126 can further decrease the GSH thiol pK_a via charge stabilization. The crystallographic water molecule could then function as the yet-unidentified base that accepts a proton from GSH during thiolate formation, concurrently forming a transient hydronium ion or shuttling the proton to the α-carboxylate. Once formed and stabilized by interaction with Arg-126, we expect attack of GSH thiolate upon the endoperoxide ring at the C-9 position, resulting in O–O bond cleavage and proton donation via the hydronium ion. Asp-49, liberated from its interaction with Arg-126, would now be free to function as a base within the resulting transition state, facilitating a decrease of the C-9 proton pK_a, and spontaneous decomposition to yield the product PGE₂ and regenerated GSH (Fig. 4). Although an alternative mechanism in which thiolate would act as a general base abstracting the C-9 proton has been suggested to be more energetically favorable in model systems (35, 36), the probability of either pathway would ultimately be determined by the precise orientation of substrate relative to cofactor within the enzymatic active site. Although the apparent ability of the active site mutants characterized here to produce PGF₂α via reduction of a putative sulphenic acid ester intermediate speaks in favor of the former (Fig. S1), this alternative mechanism is shown in Fig. S4.

Fig. 4.

Proposed mechanism of mPGES-1. For details, please see Results and Discussion.

Fig. S4.

Alternative mechanism of mPGES-1 PGE₂ synthase activity. For details, please see Results and Discussion.

A Contact Signaling Network Modulates Active Site Residues in mPGES-1.

We submitted the PDB entries associated with the recently published crystal structure of mPGES-1 (PDB ID codes 4AL0 and 4AL1) (13) to the qFit server (smb.slac.stanford.edu/qFitServer/) (20). This software automatically samples conformational heterogeneity that is interpretable by fitting partial occupancy conformational ensembles into low-level electron density. The CONTACT algorithm was then used to calculate resulting van der Waals contact networks that indicate a probable correlation of conformations at each site (27).

Post-qFit conformational fitting and subsequent refinement by Phenix (34) of the 1.16-Å binary complex with GSH (PDB ID code 4AL0) resulted in a small, but significant, improvement of structure quality indicators, including the decrease of R/R_free values from 12.2/13.0% to 11.6/12.8%, respectively. Subsequent analysis of the structure with CONTACT revealed an extensive network of correlated side chain interactions centered upon the short, cytoplasmic helix separating transmembrane helices I and II that was referred to by Sjögren et al. (13), and will be hereafter, as the C-domain. Of most interest is that the network facilitates signal transduction from residues involved in the recognition of GSH to the active site residue Asp-49. This process could facilitate the disruption of its interaction with Arg-126, facilitating the latter’s role in thiolate stabilization on a time scale specific to catalysis. Thr-34 and Leu-69, located on helices I and II, respectively, make hydrophobic contacts with the γ-glutamyl moiety of GSH and initiate series of correlated van der Waals overlaps that ultimately affect Asp-49, e.g., Leu-69 → Thr-34 → Cys-68 → Asp-64 → Lys-41 → Arg-40 → Leu-39 → His-53 → K42 → H53 → Asp-49 (Fig. 3A and Movie S1).

Fig. 3.

Contact signaling within mPGES-1. (A) The van der Waals contact network identified by qFit conformational fitting and subsequent CONTACT analysis of mPGES-1 PDB entries are shown with translucent molecular surface representations over alternate conformers and correspondingly colored node diagrams. The nodes are connected by edges whose width is weighted according to the number of networks involving the pair they connect. The network observed within PDB ID code 4AL0, which includes the active site residue D49, is shown in red, both from the perspective of the membrane plane (Left) and the cytoplasm (Right). (B) The corresponding contact networks identified from the qFit ensemble complex within the bis-phenyl GSH complex (PDB ID code 4AL1) are much smaller and are shown in cyan and red. The latter contains the active site residue R126, now observed in dual conformations, possibly due to negation of GSH thiol interaction. One possible pathway is show in more detail within Movie S1.

Conversely, Arg-126, with which it forms an intermonomeric interaction, is fitted as the single conformation observed within the crystal structure (13) (Fig. 2A). As discussed above, we believe this active site geometry is strong evidence of a GSH thiol–Arg-126 interaction. Although the pK_a of GSH thiol has been measured as 9.42 in solution (37), the dynamic interaction of GSH thiol with Arg-126, combined with the solvent restricted electrostatics of the active site, may allow GSH to transiently form thiolate during catalysis via a mechanism of charge redistribution. Specifically, the hydrogen-bonding network formed by a crystallographic water molecule between the α-carboxylate and thiol moieties of GSH may be crucial in this respect (Figs. 2A and 4).

This finding is corroborated by comparison with the qFit-generated structural ensembles of the bis-phenyl complex (PDB ID code 4AL1), in which Arg-126, now with a reduced potential for interaction with thiol, is observed to be in dynamic motion (Figs. 2B and 3B) (see below).

Contact Signaling Is Ligand-Dependent.

After an iterative fitting of alternate conformations with qFit (20), building of N-terminal residues into density, and subsequent refinement with Phenix (34) (described in Materials and Methods), a significant improvement of quality indicators was achieved for the 1.95-Å resolution mPGES-1 complex with bisphenyl–GSH (PDB ID code 4AL1), with a reduction of R/R_free values from 16.3/17.2% to 13.7/16.6%, respectively. Subsequent CONTACT analysis lacked the extensive signaling network found within the GSH complex, however, which were instead focused on opposing sides of the active site. The ensemble structure was found to contain two networks of four and five residues, respectively, the latter of which occurs in the cytoplasmic loop between helices III and IV and contains alternate conformations of Arg-126 (Fig. 3B). This finding suggests that the activation of dynamic contact networks in mPGES-1 may be dependent upon the identity of the ligand bound at the active site. Although the difference in structural information inherent in the two datasets (1.16 Å cf. 1.95 Å) should be considered when drawing comparisons between the qFit-generated ensemble structures, the resolution of the 4AL1 dataset is still significantly higher than the upper limit of 2.1 Å suggested by the software developers (smb.slac.stanford.edu/qFitServer/) (20). In addition, we performed a qFit/CONTACT analysis of high-resolution (1.41–1.52 Å) mPGES-1 inhibitor complexes recently published (15) (PDB ID codes 4YK5, 4YL0, 4YL1, and 4YL3). These four compounds are also observed to bind in the intermonomeric active site, making extensive contact with the C-domain. Although the four inhibitors are varied in structure and binding modes, they all share a common interaction with the C-domain and lack the extensive networks found in the holoenzyme complex with GSH (PDB ID code 4AL0). Intriguingly, the same interaction is also fulfilled by an octyl glucoside (n-octyl-β-d-glucoside) detergent molecule (not shown in Figs. 2 and 3 for clarity; cf. Fig. S3) within the bis-phenyl complex (PDB ID code 4AL1), whose polar head group also makes contact with the turn/helix C-domain motif and whose hydrophobic tail stacks against the bis-phenyl moiety of the GSH analog (13) (Fig. S3). This finding indicates that stabilizing contacts within this region may disrupt potential for signal transduction (Fig. S3). As shown in Fig. 3A and Movie S1, the dynamic conformations of Lys-41, Arg-40, Leu-39, and His-53 are essential to the transmission of the contact network within the C-domain, and ultimately to the active site residue, Asp-49. Therefore, it is possible that their mode of inhibition is mediated by favoring certain conformations of these residues from the structural ensemble and subsequent interruption of signaling (26).

Fig. S3.

Comparison of mPGES-1 inhibitor binding. The binding of bis-phenyl GSH and a detergent molecule (octyl glucoside) from PDB ID code 4AL1 ((14) is compared with those of four different inhibitors from the high-resolution (1.41–1.52 Å) crystal structures corresponding to PDB ID codes 4YL3 (A), 4YL0 (B), 4YL1 (C), and 4YK5 (D) (16). As is the case for the bis-phenyl complex, qFit/CONTACT analysis (28) of these datasets could not find the striking signaling network found for PDB ID code 4AL0 (14) (cf. Fig. 4). (A) The only network greater than four residues was found for PDB ID code 4YL3 and is represented as a cyan molecular surface with a correspondingly colored node diagram. This finding implies that inhibitor binding can alter conformational dynamics and signaling.

This mechanism could be a common theme of potent mPGES-1 inhibitors. In a recent analysis of binding sites via mass spectrometry hydrogen/deuterium exchange experiments (38), the authors found that the greatest differences common to the two most potent inhibitors were observed in residues 37–54, corresponding to the C-domain.

Conclusions

The combined results of site-directed mutagenesis, functional assays, structural ensemble, and contact network analysis presented herein provide strong evidence for a mechanism of PGE₂ synthesis by mPGES-1 that features an activation of GSH thiolate by Arg-126, modulated via its respective interaction with Asp-49. Furthermore, we show that conformations of the latter can be affected by a ligand-dependent contact signaling, connecting it to distal residues involved in GSH recognition, with the potential to dynamically alter the Asp-49–Arg-126 interaction during catalysis (Fig. 3).

We propose a previously unidentified mechanism of PGH₂ isomerization by mPGES-1 that features a prominent role of a water-mediated interaction with the α-carboxylate of GSH and an Asp-49–mediated thiolate stabilization by Arg-126 (Fig. 4). We hypothesize that the active site of mPGES-1 lowers the pK_a of GSH thiol and the C-9 proton of PGH₂ concurrently via respective interactions with Arg-126 and Asp-49, facilitated by their dynamic conformational change in response to contact network signaling. Charge conservation in this solvent-restricted environment could thus be achieved via proton shuffling by the crystallographic water/α-carboxylate hydrogen-bonding network (Fig. 4).

This work has broad implications for the pharmacological efforts to inhibit this enzyme, which are a current topic of discussion within the literature (39).

Materials and Methods

Protein Expression and Purification.

Recombinant wild-type (WT) and active-site mutants of human mPGES-1 were overexpressed in P. pastoris and purified by Ni-affinity chromatography before exchanging buffer to 0.1 M phosphate buffer, 0.03% dodecyl maltoside, and 2.5 mM GSH, pH 7.4. Microsomal preparations were prepared via ultracentrifugation of lysed cell supernatant and homogenization of the microsomal pellets in assay buffer (20 mM Tris⋅HCl, pH 7.8, 2.5 mM GSH). For further details, please refer to SI Materials and Methods.

Synthesis of GOH.

The oxygen analog of GSH, GOH, was synthesized in a three-step procedure based on a published method (40). For further details, please see SI Materials and Methods.

Enzyme Activity Assay.

Conversion of PGH₂ to PGE₂ by WT or mutated mPGES-1 were quantified by using GC-MS as described (12). For further details, please refer to SI Materials and Methods.

Analysis of Dynamic Contact Networks.

The qFit Web server (smb.slac.stanford.edu/qFitServer/) and the CONTACT algorithm (27) were used for the quantification of conformational ensembles and functional contact networks, respectively. Before analysis, the physiological trimer was generated from the asymmetric unit via crystallographic symmetry using the program COOT (41). For PDB ID codes 4AL0, 4YL0, 4YL1, 4YL3, and 4YK5, the coordinates were submitted to the qFit server and refined and prepared for CONTACT as described (27), by using Phenix-1.9-1692 (34) without manual intervention. For PDB ID code 4AL1, significant density improvement at the amino terminus allowed residues 4–9 to be built into density after qFit conformer fitting and refinement. After a second round of refinement, the resulting improvement in quality indicators such as the R_free value were significant, such that the improved phase estimates were anticipated to affect the conformational ensemble fitting. Hence, the improved coordinates were resubmitted to the qFit server before being refined and prepared for analysis with CONTACT as above. Settings for all CONTACT analyses were as follows: Tstress (percentile) = 0.4, max_path_length = 100, sc_only_flag = f (all atom), relief_threshold = 0.90.

SI Materials and Methods

Imidazole, Tris base, NaCl, KCl, Triton X-100, sodium deoxycholate, GSH, TCEP, and 2-mercaptoethanol were obtained from Sigma. Platinum Pfx DNA polymerase and deoxyribonucleotides were from Invitrogen. Dodecyl maltoside (DDM) was obtained from Anatrace. Anti–mPGES-1 antiserum was purchased from Cayman Chemical. PGH₂ (>95% purity) was obtained from Lipidox Co. All other chemicals were obtained from common commercial sources.

Protein Expression and Purification.

The expression vector for human recombinant His-tagged mPGES-1, pPICZ-hisMPGES, was transformed into P. pastoris KM71H cells by using the Pichia EasyComp Transformation kit (Invitrogen). Recombinant cells were cultivated at 27 °C with shaking in baffled flasks containing 2.5 L of buffered minimal yeast medium with glycerol until OD₆₀₀ reached 8–10. The cells were then resuspended in 0.75 L of buffered minimal yeast medium supplemented every 24 h with 0.6% methanol, and the pH of the medium was adjusted to 6–6.5 by using 8% (vol/vol) NH₃. The cells were harvested after 48 h by centrifugation (2,500 × g, 6 min) and resuspended in 50 mM Tris⋅HCl, pH 7.8, 100 mM KCl, and 10% (vol/vol) glycerol. The cells were homogenized with glass beads (0.5 mm) by using a Bead Beater (BioSpec Products, Inc.) operated 7 × 1 min on ice with a cooling rest between each. The slurry was filtered through nylon net filters (180 µm; Millipore) and centrifuged (1,500 × g, 10 min). Membrane-bound proteins in the supernatant were then solubilized with Triton X-100 (1%, vol/vol) containing 5 mM 2-mercaptoethanol and sodium deoxycholate (0.5%, wt/vol) for 1 h with stirring on ice. After centrifugation (10,000 × g, 10 min) the supernatant was supplemented with 10 mM imidazole and loaded on Ni-Sepharose Fast Flow (GE Healthcare). The column was washed with three column volumes (CV) of buffer A [25 mM Tris⋅HCl, pH 7.8, 10% (vol/vol) glycerol, 0.1% Triton X-100, and 5 mM 2-mercaptoethanol] supplemented with 20 mM imidazole and 0.1 M NaCl, followed by additional wash with buffer A containing 40 mM imidazole and 0.5 M NaCl. mPGES-1 was eluted with 300 mM imidazole, 0.5 M NaCl, and 0.1 mM GSH in buffer A. Buffer was then exchanged to buffer B (0.1 M phosphate buffer, 0.03% DDM, 2.5 mM GSH, pH 7.4) using a PD-10 column (GE Healthcare). After a fivefold concentration on an Amicon Ultra-15 30-kDa cutoff centrifugal filter device (Millipore), the purified protein was stored frozen at –20 °C. Protein concentration was measured by UV spectrophotometry, and purity was assessed by SDS/PAGE and Western blot.

Circular Dichroism.

For the circular dichroism (CD) characterization of mutants, the isoforms were expressed in P. pastoris cells and were homogenized and solubilized as above. After centrifugation (10,000 × g, 10 min), the supernatant was loaded onto a HisTrap HP 5-mL Ni column (GE Healthcare) by using a peristaltic pump. The column was then washed with 20 CV of buffer C [0.1 M Na phosphate, 10% (vol/vol) glycerol, 0.5 M NaCl, 0.03% (wt/vol) DDM, 1 mM GSH, 0.5 mM TCEP, and 40 mM imidazole, pH 7.4]. The protein was then eluted with 2 CV of buffer C containing 300 mM imidazole, and the sample purity was confirmed to be >95% via SDS/PAGE. The samples were subsequently exchanged into buffer D (20 mM Na Phosphate, pH 7.4, 0.03% DDM, 1 mM GSH, 0.5 mM TCEP) via repeated concentration and serial dilution (1:103) by using an Amicon 0.5-mL 50-kDa cutoff centrifugal filter (Millipore). The samples were finally concentrated to a concentration of 0.5–0.7 mg/mL, as measured by absorbance at 280 nm before measuring CD spectra using a Jasco-810 spectropolarimeter with a path length of 0.01 or 0.05 cm.

Microsomal Preparation.

Human recombinant mPGES-1 was overexpressed in P. pastoris as described above. Cell homogenates were passed through nylon filters (180 µm; Millipore) and centrifuged (5,000 × g, 10 min). The supernatant was ultracentrifuged (100,000 × g, 65 min), and microsomes were prepared from the pellet by homogenization in 20 mM Tris⋅HCl, pH 7.8, and 2.5 mM GSH in a glass homogenizer. The microsome suspension was either used fresh or stored in aliquots at −20 °C. For incubations with GOH, the microsomal preparation was homogenized in 20 mM Tris⋅HCl, pH 7.8. Ultracentrifugation and homogenization was repeated three times to remove residual amounts of endogenous GSH.

Western Blot.

Western blot was carried out as described (13). The proteins were separated by SDS/PAGE and transferred to PVDF membranes by using a Pharmacia Phast system. Antiserum against mPGES-1 (160140; Cayman) and HRP-linked anti-rabbit IgG (NA934, Amersham) were used together with an ECL Plus detection kit (Amersham) to visualize the proteins.

Synthesis of GOH.

The oxygen analog of GSH, GOH, was synthesized in a three-step procedure based on a published method (40). Briefly, the protected active γ-ester of l-glutamic acid was prepared by stirring a solution of N-t-BOC-l-glutamic acid α-benzyl ester and 4-nitrophenol in the presence of dicyclohexylcarbodiimide. The protected active γ-ester was then coupled to l-serylglycine and subsequently deprotected by dissolving in anhydrous trifluoroacetic acid and bubbling through anhydrous HBr gas. The product was washed three times with diethylether, and the white precipitate was purified by using AG1-X2 acetate resin (Bio-Rad), via elution with a 50 mM to 1 M acetic acid gradient. The presence of the product was identified by using ninhydrin spray (Acros Organics) on TLC plates. The fractions containing the product were lyophilized three times with UltraPure H₂O to yield a white hygroscopic powder. The purity of the product was verified on a 14.0-T Bruker magnet equipped with a Bruker AV-III console operating at 600.13 MHz. Both 1H- and 13C-NMR chemical shifts corresponded to the published literature values.

Enzyme Activity Assay.

Aliquots of WT or mutated mPGES-1 or microsome suspensions (0.02 mg/mL) in 0.1 M phosphate buffer, pH 7.4, supplemented with 2.5 mM GSH or GOH were stirred at 0 °C with 12 μM PGH₂ for 5 min and added to 6 mL of a saturated solution of FeCl₂ in absolute ethanol containing 2.5 μg each of [3,3,4,4-2H4]-PGE₂ and [3,3,4,4-2H4]-PGF₂α. After 5 min at 23 °C, the solutions were extracted with ethyl acetate, and materials were methyl-esterifed by treatment with diazomethane and trimethylsilylated by treatment for 20 min with 0.2 mL of trimethylchlorosilane-hexamethyldisilazane-pyridine (2:1:2, vol/vol/v). The derivatized material was taken to dryness under vacuum and dissolved in 0.15 mL of hexane. Products were quantified by GC-MS analysis using a Hewlett-Packard model 5970B mass selective detector connected to a Hewlett-Packard model 5890 gas chromatograph equipped with a phenylmethylsilicone capillary column (12 m, 0.33 μM film thickness, carried gas, helium). The temperature was raised from 120 °C to 300 °C at a rate of 10 °C/min. Amounts of PGE₂ and PGF₂α were determined by GC-MS operated in the selected ion-monitoring mode using the ions m/z 439 and 443 (unlabeled and d4-labeled PGE₂ derivatives) and m/z 423 and 427 (unlabeled and d4-labeled PGF₂α derivatives).

Preparation of Figures.

Figures were generated by using the software gnuplot (Version 5.0; www.gnuplot.info/), the PyMOL molecular graphics system (Version 1.7.6; Schrödinger, LLC), Marvin (Version 1.6.0; Chemaxon) and Adobe Photoshop (Version 13.0.6; Adobe Systems). The modeled patch of membrane shown in Movie S1 was generated by using the programs VMD (42) and NAMD (43) according to the “lipid tail melting” procedure described in the associated membrane protein tutorial (www.ks.uiuc.edu/Training/Tutorials/).