|Home | About | Journals | Submit | Contact Us | Français|
The importance of bioluminescence in enabling a broad range of high-throughput screening (HTS) assay formats is evidenced by widespread use in industry and academia. Therefore, understanding the mechanisms by which reporter enzyme activity can be modulated by small molecules is critical to the interpretation of HTS data. In this Perspective, we provide evidence for stabilization of luciferase by inhibitors in cell-based luciferase reporter-gene assays resulting in the counterintuitive phenomenon of signal activation. These data were derived from our analysis of luciferase inhibitor compound structures and their prevalence in the Molecular Libraries Small Molecule Repository using 100 HTS experiments available in PubChem. Accordingly, we found an enrichment of luciferase inhibitors in luciferase reporter-gene activation assays but not in assays using other reporters. In addition, for several luciferase inhibitor chemotypes, we measured reporter stabilization and signal activation in cells that paralleled the inhibition determined using purified luciferase to provide further experimental support for these contrasting effects.
As high-throughput screening gains momentum in academia and public databases grow in size and scope, refining our understanding of target specific and non-specific effects within HTS assays will facilitate a more accurate interpretation of screening results. Cell-based reporter-gene assays are designed to measure the influence of a library compound on a cellular process or pathway through the modulation of the ‘reporter-gene’s’ transcription and expression levels. The level of reporter is a function of its transcription, expression and stability. However, enzymes can be stabilized by inhibitors (1) when an E•I complex is more resistant to degradation than the free enzyme. In cell-based assays this can lead to an accumulation of the enzymatic reporter independent of effects on transcription/translation, thus complicating the interpretation of HTS results (2). After characterizing and developing a comprehensive profile of luciferase inhibitors (3), we were able to search for these compounds in the list of compounds identified as active in the HTS assays found in PubChem. We show here that many of the compounds designated as activators of luciferase-based reporter-gene assays are luciferase inhibitors. Further luciferase inhibitors were not enriched in assays using other reporter types (e.g., GFP and β- lactamase), suggesting luciferase stabilization as the more likely activation mechanism, as opposed to targeted or general activation of gene transcription. Our findings thus show the utility of small molecule library bioactivity profiles and underscore the value of making such library characterization assays available in PubChem.
The Photinus pyralis luciferase is commonly used in cell-based reporter-gene assays because the luminescent response provides a sensitive assay signal with a wide dynamic range due to its relatively short protein half-life (4). Not surprisingly, an increase in luciferase half-life can have a substantial effect on an assay read-out. Using the model described by Hargrove and Schmidt (5), and assuming no effect on the rate of protein synthesis or mRNA levels, a modest increase in luciferase protein half-life (e.g.~30%) can lead to a 150% increase in luciferase levels within 12 hrs. Signal from the increased levels of luciferase would be detected as it is well within a reporter-gene assay response window, especially as many of these cell-based assays involve compound incubation times of 18 hrs or longer (6). Further, we noted in our previous study that ATP or luciferin competitive inhibitors demonstrated reduced inhibition or appeared inactive in the presence of luciferin-containing reporter-gene detection reagents which generally employ an excess of luciferase substrates (3). Therefore, in this scenario, it seems possible that luciferase inhibitors could interact with, and stabilize, the cellular luciferase enzyme during the long cell-based incubation times, but upon addition of luciferin-containing detection reagent, be effectively competed away by the excess substrate provided, and thus not inhibit the measured luciferase reaction. If this is the case, one may predict an increase in the reporter levels, and thus increased signal characteristic of activation.
We have previously described a cell-free profiling screen for inhibitors of the ATP-dependent luciferase (Figure 1a) from the firefly Photinus pyralis (PubChem AID: 411) using quantitative high-throughput screening (qHTS) that determined the concentration-response behavior for >70,000 samples in the Molecular Libraries Small Molecule Repository (MLSMR) (3). Approximately 3% of the library showed inhibitory activity while none of the compounds caused a direct activation of luciferase. This comprehensive profile allowed us to define the SAR for prominent luciferase inhibitor series (Figure 1b).
To investigate preexisting evidence for this mechanism in HTS we utilized our understanding of luciferase inhibitor SAR to analyze assays available in PubChem that were screened against the MLSMR. We first examined how luciferase inhibitors were distributed among PubChem assays. We queried PubChem to determine the types of assays associated with these luciferase inhibitors (Figure 2). Nearly 50% of the assays were luminescence assays that used P. pyralis firefly luciferase, the same variant present in our qHTS. Among these we found both biochemical-based assays (including our original luciferase profile qHTS as well as one from another center, AID: 1006) and cell-based reporter-gene assays designed to identify either activators or inhibitors. Further, we noted that all the reporter-gene assays were based on expression of P. pyralis luciferase. Luciferase inhibitors were also identified, although not overrepresented among hits (see below), in assays that typically show high hit rates such as those for cellular cytotoxicity and cytochrome P450 inhibition assays. The next assay category was fluorescence-based assays followed by a variety of other assay types.
We then compared the enrichment of luciferase inhibitors versus assay format for 100 assays in PubChem. Our luciferase qHTS identified a frequency of luciferase inhibitors of 3% within the MLSMR and therefore, active sets or ‘hit lists’ containing only 3% luciferase inhibitors would not be considered enriched above the expected background. However, an HTS active set found to contain, for example, 30% luciferase inhibitors is enriched 10-fold. We would thus expect that luciferase-coupled enzyme assays or reporter-gene assays designed to identify compounds that act as inhibitors would be enriched for luciferase inhibitors, and indeed we noted a high percentage of luciferase inhibitors in these assays (Figure 3). However, we also noted that reportergene assays targeting activators also displayed a similar percentage of luciferase inhibitors within active data sets. The enrichment of luciferase inhibitors in these assays varied with the compound incubation time. For example, in a dopamine receptor potentiation assay (see for example AID: 641) having a short compound exposure time (2.5 hrs) a low enrichment was observed (≤ 3-fold), while assays with prolonged compound exposure times showed large luciferase inhibitor enrichments of ≥10-fold (see for example, AID: 560). Furthermore, in one assay for activators of Steroidogenic Factor 1 (SF-1) approximately 60% of the hits selected for confirmatory concentration-response curve (CRC) determination were luciferase inhibitors (AID: 692). Enrichment for luciferase inhibitors was not observed in reporter-gene assays that used β-lactamase, GFP or other reporters despite compound exposure times for as long as 20 hrs and the use of similar hit cutoff criteria (typically between 30 and 50%). Thus, the prevalence of luciferase inhibitors within compound libraries, such as the MLSMR, and their enrichment in luciferase reporter-gene assays provides support for inhibitor-mediated stabilization of this enzyme reporter.
In our previous study we characterized structure-activity relationships (SAR) for several prominent chemical series including compounds that mimicked the luciferin substrate and acted as competitive inhibitors of the enzyme (3). An examination of the luciferase inhibitor SAR in relation to the SF-1 reporter-gene assay actives (Figure 1) revealed that the major chemical series previously recognized as containing potent luciferase inhibitors were among either the activators or inhibitors identified in the SF-1 luciferase reporter-gene assays (Figure 1b). For example, potent luciferase inhibitors whose inhibition is not easily relieved by detection reagents (3) were identified as inhibitors in the SF-1 inhibition assay (Figure 1c, blue shaded area). However, compounds that mimic the luciferase substrate were found to be associated with SF-1 reporter-gene activation, consistent with the ability of these compounds to form a stable E•I complex within cells that is later abolished in detection mixes containing excess substrate concentrations (Figure 1c, yellow shaded area). The portion of the luciferase subchemome containing diverse structures inactive in the SF-1 reporter-gene assays (grey areas of the chemome, Figure 1) could be due to multiple factors that affect small molecule activity, such as the achievable intracellular concentration, serum binding sequestration, or experimental variation between laboratories, which includes preparation of the compound sample – a highly variable step in HTS (7).
To further experimentally support an inhibitor-based stabilization mechanism we examined representative compounds in HEK293 cells expressing P. pyralis luciferase. Of note, one of the compounds we examined is a quinoline (Figure 1, ii) that was identified as a competitive inhibitor of firefly luciferase in our previous work (3) and as an activator in PubChem luciferase reporter-gene assays (Figure 4b). In these experiments we measured the CRCs for luciferase activity after treating cells with compound for 24 hrs. To rule out the possibility that these compounds influenced the rate of transcription or mRNA stabilization, we also examined the stability of the luciferase signal in compound-treated cells after the addition of cycloheximide (2), a small molecule that inhibits eukaryotic translation (8). The same compounds were also measured in a cell-free luciferase assay using purified luciferase and Km levels of substrates to confirm the inhibitory effect of these compounds. In these experiments, we observed apparent activation of the luciferase signal within the relevant screening concentration ranges (1–10 µM) upon addition of a reporter-gene detection cocktail containing excess luciferase substrates (see Figure 4a–c) When we examined the stability of the signal after cycloheximide treatment we noted a slower rate of decay in activity for wells treated with compound compared to wells without compound (Figure 4d). Further, plots of the relative amount of luciferase activity remaining after 24 hr treatment with cycloheximide (Figure 4a–c, red lines) showed a CRC that mirrored the inhibition of the purified enzyme. These parallel but opposite responses strongly support the observation that increased luciferase activity is due to inhibitor based-stabilization of the luciferase enzyme itself. Further, we found that stabilization can occur regardless of the mode of action of the compound. For example, we have previously shown that quinoline-like compounds (Figure 4b) exhibit competitive inhibition with respect to ATP and luciferin while a 1,2,4 oxadiazole (Figure 4c) is a non-competitive inhibitor. However both types of inhibitors – competitive and non-competitive - appear to stabilize luciferase in the cycloheximide treated cells.
The activation phenotype for all inhibitors tested was generally characterized by a bell-shaped CRC, with activation increasing from low to high concentrations of less than 10 µM, followed by a gradual decrease in activation with increases in inhibitor compound concentration. The complex bell-shaped CRCs observed in the cell-based assays is due to two opposing responses: activation in the reporter gene assay due to stabilization of the luciferase enzyme and inhibitory responses that include cytotoxcity or the amount of residual luciferase inhibition in the reporter-gene detection reagent. For example, we noted that the benzthiazole partially inhibited purified luciferase assayed with the reporter-gene detection cocktail at concentrations above 10 µM (Figure 4a; bottom graph; open squares) resulting in a decreased activity above 10 µM in both the cell and cell-free assays (open circles and open squares). Alternatively, the quinoline did not appear to significantly inhibit purified luciferase in the reporter-gene detection cocktail (Figure 4b; bottom graph, open squares), but exhibited a bell-shaped CRC in the cell-based assay (Figure 4b; top graph; open circles) suggesting cytotoxic effects. For compounds such as the 1,2,4 oxadiazole (Figure 4c), that behave as non-competitive inhibitors with respect to ATP and luciferin, the factors that influence the ability to observe the activation are more complex. For example stabilization is clearly seen for this compound when examining the amount of luciferase activity remaining after 24 hr treatment with cycloheximide. (Figure 4c, red line). Additionally, the rate of decay of luciferase activity in the presence of compound is diminished (Figure 4d). However, the activation effect was not observed at relevant screening concentrations, although it was found to be significant at very low concentrations (≤IC50) (Figure 4c; top graph; open circles).
Observation of apparent reporter-gene activation due to inhibitor-based stabilization of the reporter will therefore depend on several factors. These factors include the direct inhibition of the enzyme in the detection reagent, affects on cell viability, the degree of cell penetration/retention of the compound, affinity of the compound for the reporter, degree of stabilization, and the chosen screening concentration. In general, whether or not this increase will be detected as an apparent activation will depend on how much E•I is formed within the cells resulting in stabilization and how efficiently the inhibitor is competed off in the presence of detection reagents. Given these factors, this effect will be most readily observed when the amount of free enzyme is maximized during detection which can occur, for example, with competitive-type inhibitors and prolonged cell incubation times. These complexities help to explain why the luciferase reporter-gene assays mentioned above using a short (2.5 hr) incubation time (AIDs: 641, 642, and 647 all related to potentiation of the D1 receptor) did not show enrichment in luciferase inhibitors.
This study is an example of how information from compound profiling and PubChem can be employed, in this case, to make an informed connection between luciferase inhibitors and apparent gene activation in HTS reporter-gene assays. This work also illustrates the value of compound library profiling in identifying underlying mechanisms of reproducible ‘off-target’ assay responses that can confound the interpretation of the primary experimental results. The counterintuitive finding that inhibitors of reporters can appear as activators in cell-based reporter-gene assays is a prime example of an “off-target” response that can lead to erroneous interpretations if the underlying mechanism is not appreciated. While the SF-1 assay actives were subsequently re-tested in a related nuclear receptor counter-screen (RORα) using the same luciferase reporter to identify selective actives, we demonstrate an alternative ‘counter-screen database’ approach to aid in the efficient selection and prioritization of follow-up compounds, and ascribe a probable mechanism.
Luciferase assays are often the method of choice for HTS for many reasons, most notably the enormous signal above background these assays can exhibit (4). Although this study highlights an artifact inherent to luciferase-based assays, now that it is understood, and a profile of luciferase inhibitors has been characterized and described (3), researchers can use this information to prevent following un-interesting actives. All assays have artifacts, and many of these are far less well-understood than luciferase inhibitors. For example, fluorescence responses are non-linear and depend on the assay format and detector settings, making artifacts difficult to characterize and identify. To further complicate the matter, we have found that oftentimes “compound” fluorescence may actually be due to fluorescent impurities in the chemical sample (9). In contrast, interference with luciferase-based assays can be understood with more standard medicinal chemistry rules that define the SAR of the inhibitor series for the luciferase enzyme. The fact that this same SAR can be used to explain non-specific activation in luciferase reporter-gene assays underscores the tractability of luciferase-based artifacts compared to other methods. The use of an orthogonal assays (10), for example, based on β-lactamase reporters where inhibitors are most likely less prevalent (11), or substrate-independent reporters such as fluorescent proteins, expressed in a common cell line, would provide a complementary assay to the primary screen. An understanding of the SAR and effects of luciferase inhibitors in both cell-free and cell-based systems should allow more judicial development and application of this important category of bioluminescent assays.
As HTS in academia expands beyond the pharmaceutical industry to address the needs of chemical biology and translational research, the numerous sources of artifacts painstakingly discovered in the pharmaceutical sector will, for the most part, not transition beyond proprietary company databases. Broad and open access to a public chemical biology database can serve to mitigate reinvestigation of common HTS artifacts. The striking occurrence of luciferase inhibitor enrichment in assays designed to detect receptor agonists should reinforce the notion of inhibitor-stabilization as an important consideration in the interpretation of luciferase reporter-gene assays.
The luciferase sub-chemome dendrogram was generated by an in-house interactive visualization tool called Phylochem. Given the identified list of 1,879 luciferase inhibitors, Phylochem first applied a hierachical clustering algorithm (using a suitable similarity metric based on maximal common substructure) to organize the structures. A depth-first traversal of the resulting dendrogram was then performed to project each node onto a circle with the radius proportional to the node's depth. The embedding of each node in the dendrogram is similar to the layout used by the radial clustergrams of Agrafiotis et al. (12). The final layout was obtained by merging of overlapping non-terminal nodes.
Compounds tested in this study for luciferase stabilization were initially identified and described by Auld et al. (3), and included members of a benzthiazole series, a quinoline series, and a non-competitive luciferase inhibitor – 1,2,4 oxadiazole. Compounds were obtained from ChemBridge and reanalyzed for purity in house. Purity analysis was performed via LCMS analysis on a Waters ACQUITY reverse phase UPLC System and 1.7 M BEH column (2.1 × 50 mm) using a linear gradient in 0.1% aqueous formic acid (5% ACN in water increasing to 95% over 3 minutes). Compound purity was measured based upon peak integration from both UV/Vis absorbance and ELSD, and compound identity was based upon mass analysis; all compounds passed purity criteria (>95%). These compounds were prepared as DMSO solutions in 1536-well plates at initial concentrations of 10 mM to 1 nM in a 24-point two-fold titration across the plate. Each compound titration existed in duplicate on each plate, except for the benzthiazole and 1,2,4 oxadiazole compounds, with four titrations on the plate. Four rows of DMSO also existed on the compound plate.
A 20 nM luciferase (luciferase from P. pyralis; Sigma-Aldrich; L9506) stock was prepared in PBS pH 7.4 (Invitrogen; 10010) such that upon delivery of 3 µL to the assay well, the final concentration of luciferase was 10 nM in the 6 µL total assay volume. After dispensing 3 µL of this luciferase stock to assay plates (Greiner 1536-well white, tissue culture, sterile; 789173-F) using a BioRAPTR Flying Reagent Dispenser (FRD), 23 nL of inhibitor compounds were immediately transferred from the compound plate into the assay plate using a Kalypsys pin-tool transfer station resulting in a final compound concentration of approximately 38 µM to approximately 4.6 pM. Three microliters of Promega Steady-Glo Luciferase Assay Reagent (E2520) was dispensed into each well, again using the BioRAPTR FRD. Plates were read within five minutes of assay reagent addition using a PerkinElmer ViewLux CCD Imager with a clear filter and 10 second plate exposure time. Alternatively, the luciferase enzyme activity was measured using 10 µM D-luciferin (Sigma-Aldrich; L9504) and 10 µM ATP (Sigma-Aldrich; A7699), that represents substrate concentrations approximately = Km. These experiments were performed to re-confirm results described in Auld et al., 2008 (3), and data plotted from these experiments is the average of two to four compound titrations for a given compound.
HEK293 cells transiently transfected with the pGL3-Control Vector offered by Promega (E1741) that expresses the P. pyralis luciferase were plated at a density of 10,000 cells/well using a Multidrop Combi Dispenser (Thermo Electron Corp.) in a 4.5 µL volume. After incubation for an hour at 37 °C to allow a short recovery, 23 nL of inhibitor compounds were immediately transferred from the compound plate into the assay plate using a Kalypsys pin-tool transfer station, resulting in a final compound concentration of approximately 50 µM to approximately 6 pM. Cells were then incubated at 37° C for 24 hours. Subsequently, 23 nL of a 2.25 mg/mL cycloheximide (Sigma-Aldrich; C0934) stock in DMSO (or DMSO alone) was added into the assay plate using the Kalypsys pin-tool for a final concentration of 10 µg/mL of cycloheximide in a 4.5 µL total assay volume. Plates were incubated for various times (time 0, 3 hours, 6 hours, 12 hours, or 24 hours) at 37 °C before addition of 4.5 µL Promega Steady-Glo Luciferase Assay Reagent using the BioRAPTR FRD. After a 15 minute incubation at room temperature in the dark, plates were read using a PerkinElmer ViewLux CCD Imager with a clear filter and 10 or 30 second plate exposure time. Data plotted from these experiments is the average of four to eight compound titrations for a given compound.
Data was plotted using GraphPad Prism 4 and curves were fit to the data using the software’s built-in analysis to fit nonlinear curves to the data. To generate plots of the relative amount of luciferase activity remaining after 24 hr treatment with cycloheximide, the ratio of luciferase activity 24 hours post-cycloheximide treatment to luciferase activity at time zero was calculated and normalized to the luciferase activity obtained in the absence of compound at 24 hrs and then plotted for each concentration of compound tested.
This research was supported by the Molecular Libraries Initiative of the NIH Roadmap for Medical Research and the Intramural Research Program of the National Human Genome Research Institute, National Institute of Health. We thank N. Southall for critically reading this manuscript prior to submission and D. Leja for graphical artwork.