|Home | About | Journals | Submit | Contact Us | Français|
Glutathione S-transferases (GSTs) constitute a family of detoxification enzymes that catalyze the conjugation of glutathione with a variety of hydrophobic compounds, including drugs and their metabolites, to yield water-soluble derivatives that are excreted in urine or bile. Profiling the effect of small molecules on GST activity is an important component in the characterization of drug candidates and compound libraries. Additionally, specific GST isozymes have been implicated in drug resistance, especially in cancer, and thus represent potential targets for intervention. To date, there are no sensitive miniaturized high-throughput assays available for GST activity detection. A series of GST substrates containing a masked luciferin moiety have been described recently, offering the potential for configuring a sensitive screening assay via coupled luciferase reaction and standard luminescence detection. We report on the optimization and miniaturization of this homogeneous method to 1,536-well format using GSTs from 3 different species: mouse isozyme A4-4, human isozymes A1-1, M1-1, and P1-1, and the major GST from the parasitic worm Schistosoma japonicum.
The importance of identifying poor candidates in the initial stages of drug discovery cannot be overstated in our present economic climate. As research and development costs increase exponentially, and high failure rates persist, it is more important than ever to weed out problematic molecules as early as possible as they progress through the drug discovery pipeline.1–3 Among the leading factors for a drug candidate’s attrition are poor absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties.3 While drug attrition due to ADMET issues has decreased recently, toxicity still remains one of the major reasons for failure3,4 due to the lack of accurate in vivo toxicity models.3 Recently, programs to characterize environmental toxins and minimize animal testing by using high-throughput screening (HTS) approaches to better predict the hazards of chemicals on the environment have been initiated.5 Having an ADMET profile of a compound collection based on in vitro screening results would represent a powerful tool, as it should lead to better informed decision making during early medicinal chemistry optimization of drug candidates.
Glutathione S-transferases (GSTs) constitute a family of phase II detoxification enzymes that catalyze the conjugation of glutathione with a number of hydrophobic compounds to yield water-soluble derivatives that are excreted in urine or bile.6,7 The human GST superfamily comprises at least 6 classes of isozymes: alpha (α or A), mu (μ or M), pi (π or P), omega (ω or O), theta (θ or T), and zeta (ζ or Z).8,9 In recent years, specific GST isoforms have been implicated in the clearance of drugs and environmental toxicants, such as carcinogens and pesticides, as well as in the metabolism of endogenously produced compounds such as lipid peroxidation by-products.9–16 A number of polymorphisms in the genes encoding GSTs have been described and associated with altered enzyme expression and/or activity.6,7 For instance, the GSTM1 gene is deleted in ~50% of the population (frequency of ~13%–55% in Caucasians, ~42%–55% in Asians, ~47% in African Americans, and 27% in Africans).17 Therefore, significant interindividual variation in clearance may exist for any toxicant found to be a substrate of GST. Furthermore, GSTs have been implicated as one of the causes for drug resistance, especially in cancer, and thus represent potential therapeutic targets.8,18 For these reasons, a fast and robust method for screening large compound collections against GST isozymes would provide an invaluable tool for identifying compounds liable to interfere with this class of detoxification enzymes.
To date, no miniaturized HTS methods are available for assessing the activity of GST isozymes. The most widely used assay measures turnover using the model substrate 1-chloro-2,4-dinitrobenzene (CDNB) coupled with UV–vis detection. However, it has only been practiced in cuvette and 96-well formats and suffers from small signal window (absorbance detection), large protein consumption, and susceptibility to compound interference (λ = 340 nm).19 Recently, a rhodamine-based fluorescent substrate for the M class was identified that may have the potential to be adapted for HTS, but to date that method has not been demonstrated in microtiter plates.20 Our starting point for assay development was a series of recently described GST substrates containing a masked luciferin moiety.21 These substrates are processed by GST isozymes, freeing up luciferin that is in turn quantified with a mixture of luciferase and ATP to generate light output, which can be analyzed using standard luminescence detection (Fig. 1).
We miniaturized the assay to 1,536-well format using purified recombinant mouse GST isozyme A4-4, human GST isozymes A1-1, M1-1, P1-1, and the Schistosoma japonicum GST as model enzymes. Enzyme concentration, reaction time parameters, and reagent stability were optimized in order to validate the assay for use in future unattended high-throughput screens. In addition, we investigated the utility of the assay by profiling a select group of inhibitors against the above isozymes.
HEPES pH 7.5 and HEPES buffer solution were purchased from Teknova (Holister, CA) and GIBCO/Invitrogen (Carlsbad, CA), respectively. Dimethyl sulfoxide (DMSO, certified ACS grade) was obtained from Fisher, Inc. Tween-20, 1-chloro-2,4-dinitrobenzene (CDNB), and glutathione (l-glutathione reduced, GSH) were obtained from Sigma-Aldrich (St. Louis, MO). Luciferase Detection Reagent was prepared by transferring the Reconstitution Buffer (V865B; Promega, Madison, WI) into the Luciferin Detection Reagent (V859B; Promega) followed by thorough mixing. Reaction buffer was prepared as a 50 mM HEPES pH 7.5 solution with 0.01% Tween-20 and stored at room temperature. Medium binding white solid bottom 384- and 1,536-well plates (assay plates) and 1,536-well polypropylene plates (compound plates) were purchased from Greiner Bio-One (Monroe, NC). The 384-well polypropylene V-bottom plates were purchased from Matrix/Thermo Scientific (Hudson, NH).
Purified recombinant mouse GST A4-4 (mGST A4-4) was prepared as a 35.7 μM solution as previously described.15 DNAs coding for the human GST (hGST), A1-1, M1-1, P1-1, and S. japonicum GST were cloned into Escherichia coli expression vectors. Recombinant GSTs were purified using glutathione affinity chromatography following routine protocols. hGST A1-1, M1-1, and P1-1 were stored as 42.9 μM solutions, while S. japonicum GST was stored as a 143 μM solution. Recombinant human GST O1-1 (hGST O1-1) was obtained from US Biological (Swampscott, MA) as a 112 μM solution. Human recombinant GST T1-1 (hGST T1-1) was obtained from Oxford Biomedical Research (Oxford, MI) as a 100 μM solution. Substrates PBI 1155, PBI 1158, and PBI 3773 (structures shown in Fig. 2A)21 were formulated as 20, 10, and 10 mM DMSO stock solutions, respectively, while 7-nitro-2,1,3-benzoxadiazole (NBD) inhibitors NBD-GS,22 NBD-8-OH,23 and bis-(NB-GS)24 (structures shown in Fig. 2B) were synthesized via published methods and prepared as 10 mM DMSO solutions. Additional GST inhibitors ethacrynic acid, Reactive Blue 2, and bromosulfalein were purchased from MicroSource Discovery Systems (Gaylordsville, CT), Polysciences (Warrington, PA), and Labotest (Niederschöna, Germany), respectively (structures shown in Fig. 2B), and were also prepared as 10 mM DMSO solutions.
Each isozyme was tested against each substrate to determine the best pair to optimize. Based on previously published methods,21 concentrations of 20, 100, and 10 μM were chosen for PBI substrates 1155, 1158, and 3773, respectively. After the addition of substrate to enzyme or buffer solutions, 20 μL aliquots were immediately placed in a 384-well assay plate, with final concentrations of 100 nM enzyme (or no enzyme, serving as the uncatalyzed reaction control) and 100 μM GSH. The plate was covered with a Kalypsys lid (San Diego, CA) and stored at room temperature. The reaction was stopped by the addition of equal volume of Luciferase Detection Reagent at various time points. The plate was centrifuged at 2,000 rpm for 15 s, remained covered for 15 min at room temperature, and then analyzed using a ViewLux CCD Imager (Perkin Elmer, Waltham, MA) using a luminescence protocol (clear emission filter). Due to the near-background readings obtained for hGST O1-1 and hGST T1-1 in the new assay (see Results), the enzyme activity of these samples was verified using the standard CDNB assay where 100 nM of hGST O1-1 or hGST T1-1 were incubated with 0.5 mM CDNB and 2 mM GSH, and the formation of product was detected using an Agilent 8453 UV–visible spectrophotometer (Santa Clara, CA) at an absorbance wavelength of 340 nm, following the reaction at 30-s intervals for 10 min.
After identifying a suitable substrate, an isozyme titration was performed with the corresponding optimal substrate, and in addition, a time-course study was performed by monitoring the reaction progress in 1,536-well format (Table 1).25 To begin, 3 μL of each isozyme concentration containing GSH, and buffer with GSH to serve as a no enzyme control, were dispensed into a 1,536-well assay plate using a BioRAPTR Flying Reagent Dispenser (Beckman Coulter, Fullerton, CA). The plate was covered and incubated for 15 min at room temperature, followed by a 1 μL addition of enzyme-specific substrate to start the reaction, for a final assay volume of 4 μL. The plate was then centrifuged at 1,000 rpm for 15 s, and while stored at room temperature, a 4 μL addition of Luciferase Detection Reagent was dispensed at each test point to stop the reaction. The plate was centrifuged at 1,000 rpm for 15 s, covered, and incubated for 15 min, and luminescence data were collected on a ViewLux reader.
After identifying a final enzyme concentration, overnight reagent stability tests were performed on the set of GST, substrate, and Luciferase Detection Reagent stock solutions. Fresh GST and substrate solutions at assay concentrations were prepared at time zero and stored at 4°C. Luciferase Detection Reagent, previously prepared and stored at −20°C, was thawed on the day of the experiment (time zero). Concentrations for hGST A1-1, hGST M1-1, mGST A4-4, hGST P1-1, and S. japonicum GST were 5, 5, 5, 125, and 5 nM, respectively. Concentrations for substrates PBI 1155, PBI 1158, and PBI 3773 were 20, 100, and 10 μM, respectively. Assays were performed as described above. Solutions were stored at 4°C and protected from light when not in use.
NBD-GS, NBD-8-OH, and bis-(NB-GS) were serially diluted (1:2, 16 points) column wise down to 305 nM in a 384-well polypropylene V-bottom plate. Using a CyBi-well pipetting system (CyBio, Boston, MA), the solutions were then transferred to a 1,536-well compound plate (n = 2 per dilution point, 1 column for each inhibitor titration, ½ empty column to act as the catalyzed control, ½ column for DMSO to examine for DMSO effect26). Ethacrynic acid, Reactive Blue 2, bromosulfalein, and CDNB were serially diluted (1:3, 8 points) using a JANUS automated workstation (Perkin Elmer, Waltham, MA). The compounds were assayed against isozyme concentrations for hGST A1-1, hGST M1-1, mGST A4-4, hGST P1-1, and S. japonicum GST of 5, 5, 5, 125, and 5 nM, respectively. Concentrations for substrates PBI 1155, PBI 1158, and PBI 3773 were 20, 100, and 10 μM, respectively. The assay was initiated with 3 μL dispense of reagents, consisting of isozyme and buffer, into 1,536-well assay plates. Compounds (23 nL) were transferred via Kalypsys pin tool equipped with a 1,536 pin array.27 The plate was incubated for 10 or 15 min at room temperature, followed by a 1 μL addition of isozyme-specific substrate to start the reaction, for a final assay volume of 4 μL. The plate was then centrifuged at 1,000 rpm for 15 s, and incubated for 40 min (60 min for hGST P1-1) at room temperature, followed by a 4 μL addition of Luciferase Detection Reagent. The plate was centrifuged at 1,000 rpm for 15 s, covered, and incubated for 10 or 15 min, and luminescence data were collected on a ViewLux reader. Percent inhibition was computed from the median values of the catalyzed, or neutral control, and the uncatalyzed, or 100% inhibited control, respectively, and concentration–response curve fitting was performed using GraphPad Prism 4.
We first evaluated an appropriate pro-luciferin substrate (Fig. 2A) for each of the GST isozymes studied. Figure 3A–3C present the reaction progress curves of each isozyme tested with substrate PBI 1155, PBI 1158, or PBI 3773, respectively. Figure 3D summarizes these results, representing the signal-to-background ratio defined as the luminescence signal from the catalyzed related to the uncatalyzed reaction at the 100-min incubation time point. Of the isozymes tested, hGST O1-1 and hGST T1-1 did not produce significant signal over the uncatalyzed background with any of the 3 substrates and were therefore not considered further. Testing these same 2 enzyme preparations in the standard CDNB assay confirmed that they possessed adequate activity (data not shown), indicating that the possible reason for their poor performance in the new assay was lack of compatibility with the 3 substrates. The lack of activity of these isozymes in the present assay notwithstanding, a glutathione peroxidase assay based upon hydrogen peroxide or cumene peroxide and glutathione can be utilized to measure at least the GST O and Z activity in a similar luminescence assay (J. Shultz, unpublished observations).
Substrate PBI 3773 appeared to be the most promiscuous molecule, being recognized by 4 out of 7 isozymes to produce a large assay signal. PBI 1158 produced low signal across all isozymes, consistent with its structure that differs slightly from that of luciferin, while PBI 1155 clearly displayed a preference for S. japonicum GST, yielding the highest signal ratio of ~250. The large signal and ability to be processed by most isozymes made substrate PBI 3773 the most amenable for HTS and a natural first choice when testing the assay on new GST isozymes. Isozyme–substrate pairs were chosen for further optimization based on the largest signal-to-background ratio observed, with substrate PBI 1155 selected for S. japonicum GST, substrate PBI 1158 for hGST P1-1, and substrate PBI 3773 for hGST A1-1, hGST M1-1, and mGST A4-4, respectively. Of note, this study identified a substrate for mGST A4-4, which had previously not been tested in this assay. When comparing mGST A4-4 to a similar isozyme, hGST A1-1, analogous profiles were observed with the 3 substrates.
After identifying isozyme–substrate pairs, isozyme titrations were performed while simultaneously migrating the assay into 1,536-well format (eg, plots for S. japonicum GST and mGST A4-4 presented in Figure 4A and and4B,4B, respectively, with the corresponding data for GST isozymes hGST A1-1, hGST M1-1, and hGST P1-1 provided in Appendix Fig. A1). Most concentrations tested for hGST A1-1, hGST M1-1, S. japonicum GST (Fig. 4C), and mGST A4-4 (Fig. 4D) exhibited robust assay performance, with Z′ factor28 values within the ranges of 0.40–0.89, 0.84–0.96, 0.79–0.97, and 0.10–0.97, respectively. While exact signal and Z′ ranges varied among isozymes, after a 30-min incubation all assays exhibited robust performance (Z′ factor values of >0.6) regardless of enzyme concentration. In general agreement with the initial tests performed in 384-well format, the miniaturized assay for hGST P1-1 yielded satisfactory signal and Z′ only at longer reaction times, generally after 40 min.
Based on these results, an enzyme concentration of 5 nM and an incubation time of 40 min were chosen for the 1,536-well reagent stability assay for all GSTs except hGST P1-1, where a concentration of 125 nM and incubation period of 60 min was used. The incubation period was chosen to balance the need to run the reaction under initial-rates regime, at the lowest isozyme concentration possible, and at minimal substrate conversion, and the desire for a robust luminescent signal difference between the catalyzed and uncatalyzed reactions for each isozyme. The low nanomolar concentrations of the GST isozymes used, except for hGST P1-1, sensitized the assay to inhibition and allowed multiple tests to be run with only a minimal consumption of enzyme.
With the basic assay parameters established, all reagent components were tested for stability. Excellent integrity was noted for at least 25 h of storage of the screening reagents formulated as working stocks: there was only minor activity loss for the duration of reagent storage, as the signal-to-background ratio decrease generally did not exceed 36% relative to that of the fresh solutions (Fig. 5). These results indicate that an unattended overnight screen using the present combinations of GST, substrate, and detection reagent is feasible.
We then used the assay (protocol outlined in Table 1) to profile several small molecules (structures provided in Fig. 2B) previously noted either as GSH mimetics (NBD-GS, bis-[NB-GS])22,24 or as diverse GST inhibitors or substrates (NBD-8-OH, ethacrynic acid, bromosulfalein, Reactive Blue 2, and CDNB).29–32 Individual concentration–response results are shown for ethacrynic acid, Reactive Blue 2, NBD-GS, and NBD-8-OH in Figure 6, panels A–D, respectively, with the corresponding plots for bromosulfalein, bis-(NB-GS), and CDNB provided in Appendix Figure A2, while Figure 6E summarizes the inhibitor profiles against all GSTs. The model GST substrate CDNB displayed partial inhibition only at the top concentrations tested (near 100 μM, Appendix Fig. A2C), consistent with its relatively high Km values, ranging between 700 μM and 1 mM. No DMSO effect was observed for any of the isozyme reactions tested, up to the concentration of 0.7% intended for use in HTS (data not shown).
The GSH mimetics exhibited similar inhibition profiles amongst themselves and against hGST A1-1, mGST A4-4, hGST P1-1, and S. japonicum GST, with IC50 values mostly in the low double-digit micromolar range, despite the utilization of different substrates and the difference in species origin among these 4 GSTs. However, a distinct inhibition profile was observed for hGST M1-1, with bis-(NB-GS) being the most potent at 0.191 μM, in contrast to NBD-GS, where an IC50 value of 24.4 μM (~130-fold difference) was observed. On the other hand, the 4 diverse inhibitors displayed a range of potencies that varied across GSTs. There was a significant difference in pattern between the related hGST A1-1 and mGST A4-4, where similar potencies were observed against hGST A1-1 but a nearly 100-fold spread in IC50 values was observed with mGST A4-4 for ethacrynic acid, Reactive Blue 2, and bromosulfalein. Interestingly, while a divergence in potencies was seen for the GSH mimetics against hGST M1-1, the effect of all 4 diverse inhibitors against that isozyme was very similar. Lastly, while several compounds exhibited low or no inhibition against some mammalian GSTs (eg, the lack of inhibition by bromosulfalein of hGST P1-1), the worm GST enzyme was inhibited by all compounds at ~10 μM or higher potency, with Reactive Blue 2 displaying the most potent inhibition at an IC50 of 63 nM.
While the present assay is robust and stable, it is not devoid of interferences that can confound the interpretation of inhibition experiments. For example, intensely colored compounds can affect the luminescence signal (centered at 559 nm)33 by acting as nonspecific light quenchers (ie, inner-filter effect).34 Indeed, one of the inhibitors tested here, Reactive Blue 2, is intensely blue-colored, while NBD-8-OH contains the yellow nitrobenz-2-oxa-1,3-diazol-4-yl chromophore, and they could therefore be suspected as having acted purely as luminescence quenchers. To evaluate this possibility, we performed a control experiment where Reactive Blue 2 and NBD-8-OH were added to the corresponding assay wells only upon the completion of the GST reaction and the luminescence detection reaction but before the luminescence read. A concentration-dependent decrease in signal was evident only above ~5 μM of Reactive Blue 2 and no significant attenuation was observed with NBD-8-OH up to the highest concentration tested (Appendix Fig. A3, panels A and B, respectively). Thus, luminescence attenuation by the chromophore on NBD-8-OH was unlikely to be the cause of the single-digit micromolar or better inhibitory activity of this compound against the GSTs tested. Similarly, for Reactive Blue 2 the onset of luminescence attenuation occurred at concentrations several hundred times greater than the corresponding IC50 values for S. japonicum GST and mGST A4-4, and ~60-fold higher than the IC50 value for hGST P1-1, strongly suggesting that the inhibitory effect of this colored compound against the 3 targets was not due to an inner-filter phenomenon. On the other hand, additional investigation into the effect of Reactive Blue 2 on hGST A1-1 and M1-1 may be warranted due to the relatively small difference between onset of luminescence decrease in the control test and the IC50 values for these isozymes (5- and 8-folds, respectively). A decrease in assay signal may also be due to the compound acting as a substrate and competing with the pro-luciferin reagent for GST-catalyzed conjugation to GSH. To account for a complex mode of action such as utilization of the test molecule as a GST substrate, secondary assays will need to be applied. Luciferase inhibitors, while not uncommon34 in compound libraries, do not represent a major source of false positives due to the application of excess luciferase and ATP relative to the upstream GST reaction. Finally, the assay utilizes the Ultra-Glo™ Luciferase Reagent, which has been noted for its lower rate of compound interference than the common Photinus pyralis luciferase.34,35
In summary, we have developed a homogeneous miniaturized luminescence assay for GST activity and have demonstrated its applicability to 1,536-well-based HTS by using several prominent members of the human GST family, as well as representatives from murine and parasitic worm species. Our profiles of a panel of inhibitors demonstrated the assay’s ability to report on a broad range of effectors. These results, combined with the simplicity of the assay protocol and the verified stability of the test reagents, set the stage for large-scale screening campaigns to identify inhibitors and substrates of this important target class.
This research was supported in part by the Molecular Libraries Initiative of the NIH Roadmap for Medical Research, the Intramural Research Program of the NHGRI, NIH, and NIH grant ES016623 (to E.A. and J.D.).
Adam Yasgar, NIH Chemical Genomics Center, National Human Genome Research Institute, National Institutes of Health, Bethesda, Maryland.
John Shultz, Promega Corp., Madison, Wisconsin.
Wenhui Zhou, Promega Biosciences Inc., San Luis Obispo, California.
Hui Wang, Promega Biosciences Inc., San Luis Obispo, California.
Fen Huang, Promega Corp., Madison, Wisconsin.
Nancy Murphy, Promega Corp., Madison, Wisconsin.
Erika L. Abel, Department of Carcinogenesis, The University of Texas M.D. Anderson Cancer Center, Science Park-Research Division, Smithville, Texas.
John DiGiovanni, Department of Carcinogenesis, The University of Texas M.D. Anderson Cancer Center, Science Park-Research Division, Smithville, Texas.
James Inglese, NIH Chemical Genomics Center, National Human Genome Research Institute, National Institutes of Health, Bethesda, Maryland.
Anton Simeonov, NIH Chemical Genomics Center, National Human Genome Research Institute, National Institutes of Health, Bethesda, Maryland.
The authors declare no competing financial interests.