High-Precision,Room Temperature Screening Assay for Inhibitors of Microsomal Prostaglandin E Synthase-1

Materials

All chemicals were of pro analysi or of similar high quality. PGH₂ (Larodan Fine Chemicals [Malmö, Sweden], shipped as 2 mg/mL in acetone) was diluted in acetonitrile (Fluka [Buchs, Switzerlnad] 00709, puris, over mol. sieve, ≤0.001% H₂O), using dried glassware and equipment, and aliquoted in 15-mL argon- or nitrogen-filled polypropylene tubes (352096; Falcon Fischer Scientific, Gothenburg, Sweden), which were frozen on dry ice and stored at −80 °C for later use. Microsome preparations containing high mPGES-1 activity (~10 mg total microsomal protein/mL, 10 mM potassium phosphate [pH 6.8], 0.1 mM EDTA, 10% [v/v] glycerol, Complete Protease inhibitor [EDTA free; Roche, Basel, Switzerland, 1 836 153, 1 tablet/10 mL]) were produced according to a standard microsome preparation procedure ⁸ from SF9 cells expressing the human mPGES-1 gene (GenBank BC008280) by use of the baculo expression system (BaculoDirect; Invitrogen, Carlsbad, CA). mPGES-1 microsomes used as blank-rate controls in the determination of K_M for PGH₂ were heat inactivated by incubation at 80 °C for at least 20 min and subsequently stored at −80 °C before use. Compound AZ1 was found within an in-house research program and has been published previously. ⁹ Compound MK-886 (P1833aa) was from Affiniti Research Products Ltd (Mamhead, UK).

Final prostaglandin E synthase assay

A small volume (0.4 µL) of test compound in DMSO was added (Axygen tips, cat. no. FX-384R; Axygen, Union City, CA) to a 60-µL microsome preparation with human mPGES-1 in potassium phosphate buffer (50 mM), pH 6.8, with 2.5 mM glutathione (GSH) in the sample wells of a 384-well polypropylene microplate (781280; Greiner Bio-One, Monroe, NC) and preincubated for 15 min. Corresponding solutions without test compound were used in 16 positive-control wells, and corresponding solutions without test compound and without microsomes were used in 16 negative-control wells. The enzymatic reaction was then started by addition of 5 µL of 140 µM PGH₂ in dry acetonitrile. The reaction was stopped after 1 min by addition of 20 µL of an acidic and degassed solution of ferric chloride and citrate, pH 1.9 (to reach final concentrations of 7 mM and 47 mM, respectively, pH to 3–4), which reduced remaining PGH₂ to lipids—mainly, 12-hydroxy-heptadecatrieneoic acid (12-HHT). ¹⁰ 12-HHT is not detected by the subsequent PGE₂ detection step. The resulting solution was then pH neutralized by addition of 20 µL potassium phosphate buffer (1.2 M, pH 7.5), prior to two sequential 37-fold dilutions (in total 1350-fold dilution) in a weak potassium phosphate buffer (50 mM, pH 6.8) containing 0.2% bovine serum albumin (BSA; w/v, A-3294, Fraction V, protease free; Sigma, St. Louis, MO). The PGE₂ formed was quantified by use of a commercial HTRF-based kit (cat. 62PG2PEC or 62P2APEC) from Cisbio International (Cisbio International, Bagnols-sur-Cèze, France): 6 µL of the sample solution was transferred to a Corning 3676 plate (Corning, Inc., Corning, NY) and, after subsequent addition of 3 µL PGE₂-d2 and 3 µL anti-PGE₂-cryptate from the HTRF kit, this plate was incubated overnight at 4 °C. Labcyte’s MicroClime Environmental Lid (LL-0310; Labcyte, Sunnyvale, CA) filled with water was used to cover the detection plate during the incubation. The large dilution prior to quantification was needed due to the highest concentration of PGE₂ quantifiable by the kit being more than 1000-fold lower than the concentration of PGE₂ formed (a very low ratio of substrate concentration over K_M,PGH2 was not desired for pharmacological reasons). After incubation, the plate was read in an EnVision reader (PerkinElmer, Waltham, MA) according to Goedken et al. ⁷ and kit instructions.

The whole assay procedure described above was performed on a Biomek FX system (Beckman Coulter, Brea, CA) equipped with one 384-head and one 96-head. The latter was used as an alternative gripper for parallel transportation of plates and lids. This parallel processing was a prerequisite for enabling automation of a precise assay with a reaction time of only 60 s. Liquid transfers and mixing are described in Supplemental Figure S1 and Supplemental Table S1, and subsequent tip washes are described in Supplemental Table S2. For each step, a separate box of tips was used. Tip boxes were reused only after centrifugation (2 min at relative centrifugal force [RCF] 559 g). After addition of PGH₂ that was transferred from a reagent plate (Greiner 781280) covered with an acetonitrile-filled lid (MicroClime Environmental Lid; Labcyte, cat no. LL-0310), the reaction mixture was mixed continuously until addition of stop solution. Therefore, tips were not washed until after the latter addition. To maximize the mix in the reaction and stop steps, Biomek AP384 P30XL Tips (cat. no. A22287) were used, whereas Biomek AP384 P30 Tips (cat. no. 719222; Beckman-Coulter) were used in all other steps. Carryover of compound was prevented by washing tips with DMSO/water in step 1 (Suppl. Table S2). To ensure clean, dry tips in the PGH₂ dispensing, these tips were additionally washed with acetonitrile (Suppl. Table S2, steps 2 + 3). Formation of drops on the tips used for transfer of BSA-containing solutions was prevented by wash with 0.01% Triton X-100/water (Suppl. Table S2, steps 4–9).

The typical reaction condition of the enzymatic reaction was thus:

Test compound: ranging from 60 μM to 0.002 μM (diluted in half-logarithmic steps in DMSO), or zero in positive and negative controls; Potassium phosphate buffer pH 6.8: 50 mM; GSH: 2.5 mM; mPGES-1-containing microsomes: 2 µg/mL (total protein concentration, sample and positive controls) or 0 µg/mL (negative control); PGH2 (diluted in dry acetonitrile): 10.8 µM; Acetonitrile: 7.7 % (v/v); DMSO: 0.6% (v/v).

Temperature in cooled assay microplate

The temperature in individual wells of assay microplates was determined with a digital thermometer (ELFA CIE mod. 305 with thermoelement, Thermo-Electra 8010 [type K]; Thermo Electra BV, Pijnacker, the Netherlands).

Determination of K_M for PGH₂

Determination of K_M for PGH₂ was performed according to the mPGES-1 assay protocol with the following deviations. The PGH₂ solutions in the PGH₂ reservoir plate were diluted in dry acetonitrile to various concentrations so that final assay concentrations varied from 2 to 115 µM both in wells used for enzyme reactions as well as in blank-rate wells (triplicates, 50 mM potassium phosphate, 2.5 mM GSH). After subtraction of PGE₂ levels in blank-rate wells from their cognate enzyme wells, equation (1) was fitted to the resulting net PGE₂ levels. The experiment was repeated once with a lower microsome concentration (1 µg/mL) with a similar result.

Y = A + B * X h / ( K M h + X h ) .

Determination of K_D for GSH

Determination of K_D for GSH was performed similarly to the determination of K_M above. Assay solutions with (enzyme reactions) and without (blank-rate controls) mPGES-1–containing microsomes (2 µg/mL) were dispensed into wells of an assay plate, which were subjected to the mPGES-1 assay protocol as described above. Equation (1), with the variable h fixed to unity, was fitted to the net PGE₂ produced during a 1-min reaction.

Calculation of Z′

Z′ values were calculated according to the equation of Zhang and coworkers. ¹¹

Calculation of IC₅₀

Concentration-response curves were fitted according to equation (3), using a four-parameter logistic equation where x is concentration (M); y is effect (% inhibition); A and B are the bottom and top plateau, respectively, of the curve; C is IC₅₀; and D is the slope (Hill) coefficient. The effect y is sample (s) signal normalized to positive (p) and negative (n) control well medians of (16 wells each) according to equation (2).

y = ( s − n ) / ( p − n ) .

y = A + ( ( B − A ) / ( 1 + ( ( C / x ) D ) ) ) .

Calculation of HTRF ratio

Calculation was performed according to equation (4), where F₆₆₅ and F₆₂₀ are the fluorescence values for the 665-nm and 620-nm channels, respectively.

Ratio = ( F 665 / F 620 ) * 10 4 .

Determination of adequate mix

The importance of adequate mixing of reagents in microplates was determined according to Poulsen et al. Poulsen, E.; Smith, R.; Norman, M.; Leijon, J.; Otroka, M., (manuscript in preparation). Briefly, to measure whether a mixture of two different solutions (e.g., DMSO and aqueous buffer) is homogeneous, tartrazine was dissolved in the liquid about to be transferred. After the transfer and the following mix, three to five layers (depending on the volume in the plate) of the mixture were transferred to Greiner 781101 microplates, and absorbance was read with a Molecular Devices (Sunnyvale, CA) SpectraMax 340 (at 426 nm). If the different layers were differing less than 5%, then the mixture was deemed homogeneous. To quickly achieve acceptable mixing, the maximum tip volume, minus some volume reserved for washing, was used.

Accuracy and precision of liquid transfers

Accuracy and precision of dry and wet transfers were performed essentially according to Taylor and coworkers ¹² with the following exceptions. Absorbance of tartrazine (Sigma) was used instead of rhodamine green fluorescence. Dilution after transfer, when required, was done in water.

Room Temperature Automated Assay in 384-Well Format with PGH₂ in Acetonitrile

We sought to set up a practical high-capacity assay of high precision for use in robust SAR-driving IC₅₀ determination. The starting point for assay development was a commercially available HTRF PGE₂ detection kit ⁷ and microsome preparation from mPGES-1 overproducing cells rather than the purified enzyme used in the previous assays by Massé et al. ⁶ and Goedken et al. ⁷ The reaction catalyzed is challenging for assay development because, in parallel to the mPGES-1–catalyzed reaction, PGH₂ spontaneously isomerizes to PGE₂ at a significant rate in aqueous solution at ambient temperature. To minimize background and simplify the assay, we chose FeCl₂/citrate to quench PGE₂ production, instead of SnCl₂/HCl, which converts PGH₂ to PGF_2α. PGF_2α cross-reacts weakly with PGE₂ in the detection kit, which could potentially contribute to background and thus reduce signal window and Z′. The FeCl₂/citrate reaction reduces PGH₂ to mainly 12-HHT, ¹⁰ which does not cross-react with the PGE₂ on the HTRF antibody.

Several available mPGES-1 assay protocols keep PGH₂ in a water-miscible organic solvent until the start of the reaction to let blank-rate hydrolysis to PGE₂ start simultaneously with the enzymatic reaction. However, the choice of solvent was a concern for us: The vapor pressure of the solvents used (e.g., acetone or isopropanol^3,7) is high, with elevated risk for evaporation-related issues (liquid transfer precision, work environment, number of plates per batch). Here, we instead opted to use dry acetonitrile, which has fairly low vapor pressure, as a solvent for PGH₂. We confirmed that this solvent evaporated at a very low rate from the 384-well microplate that we would use as a reservoir (data not shown) and that loss of PGH₂ to the wall of propylene tubes when stored at −80 °C were negligible at stock concentrations of 140 µM and above. The 140-µM PGH₂ preparation contained little PGE₂ and was fairly stable in dry acetonitrile (Suppl. Fig. S2): After 90 min at room temperature, less than 0.2% had isomerized to PGE₂.

Our first attempt at a robotized protocol, which involved keeping the aqueous reaction components cooled locally on the robotic system by using a cooled ALP and a dito cabinet (4 °C), resulted in large difficulties with plate edge effects: The temperature in the wells of the assay plate differed by several degrees. To avoid these effects, we instead ran the assay at room temperature with the short enzyme reaction time of 60 s. This protocol (see Material and Methods, Final Prostaglandin E Synthase Assay) worked well in several aspects. First, it displayed low blank formation of PGE₂ from spontaneous PGH₂ hydrolysis (0.3 µM), which allowed for a signal window of ~7 for the uninhibited reaction. Second, the total PGE₂ production over time from addition of PGH₂ until termination was approximately linear ( Fig. 1 ). The graph indicates that 1 min of assay reaction with 2 µg microsomes/mL lies within the stable and fairly linear domain of the assay reaction with respect to enzyme concentration and time: Assay components seem to be stable up to 3 min in the reaction. Third, further characterization of the assay system showed that PGE₂ detection by HTRF was precise and without plate-border effects. PGE₂ was stable in the intermediate steps of the reaction (Suppl. Fig. S3).

OPEN IN VIEWER

Figure 1.

Analysis of time-dependent prostaglandin E synthase (PGES) activity of microsomal prostaglandin E synthase-1 (mPGES-1) microsomes in the automated assay. Plotting according to Selwyn ¹⁴ of the PGE₂ formation versus the product of enzyme concentration and reaction time is useful for identifying time-dependent instabilities of assay components and products. Here, connecting the data points of equal reaction time clearly indicates (arrow) that the curve of 5 min assay-reaction time displays a lower activity than the other curves. Microsome concentrations were 0.5, 1, 2, 4, and 10 µg/mL (total microsome protein concentration). The dotted line is an approximate tangent to the curves of the 1- and 1.5-min assay reaction time.

Basic Assay Characterization and Enzyme-Kinetic Properties

It has previously been shown with purified mPGES-1 that GSH is not consumed during the mPGES-1–catalyzed reaction and thus is a true cofactor. ³ With the microsome preparation used here, we obtained an apparent K_M for PGH₂ isomerization of 22 ± 5 µM at 2.5 mM GSH and a K_D for GSH of 0.6 ± 0.1 mM in accordance with earlier estimates. ³ The enzyme-catalyzed reaction rate in response to variation in PGH₂ concentration was slightly sigmoid with a slope coefficient of 1.8 ± 0.2 (Suppl. Fig. S4). This could be interpreted as cooperativity between the subunits of the mPGES-1 homotrimer ¹³ but should be investigated further before such conclusions can be drawn. In this experiment, heat-inactivated microsomes were used in the blank-rate control. This control displayed a blank rate indistinguishable from the rate of hydrolysis of a range of concentrations of PGH₂ in pure assay buffer and with mock infected microsomes (data not shown). Enzyme activity was stable overnight in 50 mM potassium phosphate (pH 6.8) at ambient temperature as compared with enzyme solution freshly made from frozen mPGES-1 microsome stock. Assay solution including 2.5 mM GSH was also stable overnight.

Pharmacological validation confirmed IC₅₀ for MK-886 to 7 ± 3 µM ( Fig. 2 ), which is similar to a published value of 3 µM for which direct determination of PGE₂ was used. ³ The slightly higher value obtained here likely derives from the fact that we calculated IC₅₀s from HTRF ratios: The HTRF ratio relates to PGE₂ concentration in a nonlinear manner. We therefore compared IC₅₀ calculated from HTRF ratios and from concentration of product formed, respectively, for four mPGES-1 inhibitors and found that IC₅₀ was offset by a constant factor of around 2 in favor of the former over a potency range from 3 to 400 nM (data not shown). This thus explains the twofold difference to the published MK-886 value.

OPEN IN VIEWER

Figure 2.

Inhibition of microsomal prostaglandin E synthase-1 (mPGES-1) by compound MK-886. The curve was fitted to 20% inhibition values calculated from homogeneous time-resolved fluorescence ratios from 10 inhibitor concentrations in duplicate on the microplate. The slope coefficient of MK-886 concentration-response curves was generally higher (mean 2.2 ± 1.0) than for other tested compounds. No correlation between slope coefficient and IC₅₀ for this compound was observed.

Investigation of Mixing Enabled Very High Precision and Reproducibility

All liquid transfers were validated for accuracy and precision. The result from this work indicated that precision of the assay needed further optimization ( Fig. 3 , circles; Fig. 4a ) and that suboptimal mixing rather than calibration of transfer volumes needed to be addressed. The relatively high volumes in the plates combined with the narrow wells of a 384-well plate were expected to lead to an inhomogeneous mixture if not mixed thoroughly, which would lead to less precision and reproducibility with respect to Z′ and IC₅₀. To address this, we endeavored to validate all liquid transfers for homogeneity.

OPEN IN VIEWER

Figure 3.

Signal to noise of positive and negative controls: Z′ before and after optimization of mixing. Circles: experiments before optimization of mixing. Triangles: experiments after optimization of mixing. Z′ values were calculated from homogeneous time-resolved fluorescence ratios.

OPEN IN VIEWER

Figure 4.

Microsomal prostaglandin E synthase-1 (mPGES-1) assay IC₅₀s of compound AZ1 (a) before and (b) after optimization of mixing. Percentage inhibition values were calculated from homogeneous time-resolved fluorescence ratios. Each curve represents a separate test occasion (day). Each concentration of test compound (in total, 10 concentrations) was run in duplicate in neighboring wells. The same batch of mPGES-1 microsomes and test compound was used in all experiments.

The homogeneity of mixing was investigated thoroughly using a novel method developed at our lab (Poulsen, E.; Smith, R.; Norman, M.; Leijon, J.; Otroka, M., manuscript in preparation). Our investigation showed that the mixing indeed could be drastically improved—an increase of the number of mixes from 3 up to 15 in some steps (Suppl. Table S1) was necessary to reach an acceptable homogeneity of within 5%. The increased number of mixes improved the coefficient of variation (CV) of controls and of pIC₅₀ drastically ( Fig. 3 , triangles). Thus, it proved critical, to reach maximum precision when developing an automated assay of inhibition, to assess the quality of pipette mixing after liquid transfer. The investigation further identified the major source of lack of precision to be the mixing of liquid transfer step 2 (Suppl. Table S1, PGH₂ addition): To achieve satisfactory homogeneity, 45-µL pipette tips had to be introduced and the number of mixes increased to 13, which meant continuous mixing during the enzyme-catalyzed reaction. In addition, introduction of an environmental lid during incubation of the detection plate removed edge effects that were observed when using a regular polystyrene lid (Greiner cat. no. 656101). Also important for precision was the observation that when only water was used for tip wash in liquid transfer steps 5 to 10, in which all transfer solutions contained BSA, a significant volume of liquid was retained on the outside of the tips in subsequent transfers, probably caused by BSA residues on the surface of the tips. After the introduction of wash with 0.01% Triton X-100 of the tips after each of these transfer steps, the tips no longer retained excess solution (Suppl. Table S2).

A critical parameter in pharmacological inhibitor screening is cost. By using adequate pipette tips and an appropriate liquid-handling protocol, very small volumes (0.4 µL) of test compound in DMSO could be transferred, enabling low final DMSO concentration and a small final reaction volume. Washing combined with centrifugation enabled reuse of pipette tips at least 20 times without diminishing quality of data. To prevent sticky compounds to be carried over between plates, however, it was necessary to exchange some tips regardless of extensive washing. Reducing volumes in the HTRF detection step by 40% down from the manufacturer’s recommendation, together with a significant reduction of dead volume to 4 µL by introduction of a low-volume source plate (Greiner 784201), reduced detection reagent cost significantly. This reduction did not affect CV of the assay controls (below).

Improvement of mixing of all steps in the assay led to a large decrease in variation in the pIC₅₀ of the reference compound AZ1: CV dropped from 3% to 1% ( Fig. 4 ). This improved CV meant that ±3 standard deviations around pIC₅₀ covers as little as 0.5 units on the pIC₅₀ scale, which guarantees good separation between compounds during the building of the SAR for a chemical series. The final assay displayed a Z′ ≥ 0.7, as well as a CV of <5% and <3% for positive-control wells and negative-control wells, respectively (average of two plates; data not shown), as measured by the HTRF ratio. Z′ above 0.5 could be reached with even lower (fourfold) enzyme activity (data not shown). Whether the assay protocol could be amenable for greater capacity for use in high-throughput screening was not investigated here. The capacity of the present assay, however, was six 384-well microplates per batch of reagent plates, which is adequate for compound optimization and characterization in medical-chemistry efforts. Thorough investigation of the quality of mixing in all liquid transfer steps thus proved crucial for reaching the high-precision performance of the present assay—an observation that may be of general relevance for assay development in 384-well microplates, which cannot be properly mixed by shaking.

We have presented here, to our knowledge, the first screening assay in a 384-well format for mPGES-1 inhibitors that uses acetonitrile as a dry solvent for PGH₂ and is devoid of cooled steps during the enzyme reaction. With these properties, together with its high-precision pIC₅₀ determination, the presented assay should be of utility in generating SARs of high quality and in driving optimization of mPGES-1 inhibitory compounds. Also, given the recent crystallographic progress in mPGES-1 research, ¹³ the reaction catalyzed by mPGES-1 demands further examination of its biochemistry in screening assays as well as in vivo. The present protocol should be of utility in detailed investigation of the enzyme-kinetic characteristics of mPGES-1 in vitro.

In the work of developing the present assay, it was learned that thorough investigation of mixing is a prerequisite for reaching very high precision in an assay for determining IC₅₀s of mPGES-1 inhibitors, at least with the type of protocol used here. It is our belief that the challenge of investigating the quality of mixing may have been underestimated in pharmaceutical assay development and that setup of future screening assays may generally benefit from increased focus on this important aspect of assay development.