Characterisation of utrophin modulator SMT C1100 as a non-competitive inhibitor of firefly luciferase

Firefly luciferase (FLuc) is a powerful tool for molecular and cellular biology, and popular in high-throughput screening and drug discovery. However, FLuc assays have been plagued with positive and negative artefacts due to stabilisation and inhibition by small molecules from a range of chemical classes. Here we disclose Phase II clinical compound SMT C1100 for the treatment of Duchenne muscular dystrophy as an FLuc inhibitor (K D of 0.40 ± 0.15µM). Enzyme kinetic studies using SMT C1100 and other non-competitive inhibitors including resveratrol and NF κ BAI4 identified previously undescribed modes of inhibition with respect to FLuc’s luciferyl adenylate intermediate. Employing a photoaffinity strategy to identify SMT C1100′s binding site, a photolabelled SMT C1100 probe instead underwent FLuc-dependent photooxidation. Our findings support novel binding sites on FLuc for non-competitive inhibitors.


Introduction
Firefly luciferase (FLuc) is widely used in academia and industry due to its excellent sensitivity and dynamic range, and its ease of use [1]. Firefly luciferase (Photinus pyralis) is bioluminescent, enabling sensitive detection because a light excitation source is not required, reducing the background signal [2]. Dynamic changes in luciferase abundance can be detected due to its short half-life of 3-4 hrs in mammalian cells [3]. FLuc is used in a broad range of often highthroughput applications including reporter gene assays, in vivo imaging, cell viability assays, cAMP sensing for cell signalling assays and more [2].
However, FLuc assays are sensitive to positive and negative artefacts arising from compound interferences [1]. Studies conducted by Auld et al. have shown that~5% of compounds in large libraries interfere with FLuc at 11 µM [4], while~6% of the GSK published protein kinase inhibitor set are FLuc inhibitors [5]. Furthermore, up to 98% of compounds identified as high throughput screen (HTS) hits from FLuc reporter assays have been found to be FLuc inhibitors [6,7]. Many commonly used tool compounds such as p53-inhibitor pifithrin-α [8], resveratrol [9] and aryl hydrocarbon receptor inhibitor β-naphthoflavone [10] have also been found to be FLuc inhibitors. Consequently, hit and tool compounds should be routinely assessed for FLuc inhibition to validate interpretation of results from FLuc assays.
FLuc interferences are typically due to direct binding of the compound to luciferase rather than through light attenuation or scattering [4,11]. Compound binding has been found to impart a thermal stabilisation on the FLuc enzyme, which increases its half-life independently of transcription or translation [3,12]. This stabilisation effect is amplified by the short half-life of FLuc coupled with long compound incubation times (12-48 h) typically seen with reporter gene assays [13]. Increases in luciferase abundance by compound binding can lead to false positives, while false negatives arise from inhibition of the bioluminescence reaction.
Firefly luciferase's catalytic bioluminescence reaction proceeds via a relatively well characterised multistep mechanism (Fig. 1a). FLuc binds substrates D-luciferin (D-LH 2 ) 1 and ATP to form the reactive intermediate luciferyl-adenylate (LH 2 -AMP) 2 [14]. Upon adenylation, FLuc T undergoes a conformational change [15]. Then, LH 2 -AMP reacts with molecular oxygen to form an electronically excited dioxetanone 3 which collapses to generate oxyluciferin 4 in an excited state and carbon dioxide [16]. Relaxation to the ground state of oxyluciferin releases a photon of light of wavelength 560-610 nm [16]. FLuc can also perform 'dark' reactions such as oxidation of the intermediate LH 2 -AMP to form the potent tight-binding inhibitor L-AMP 5 which is responsible for the flashing of firefly lanterns [11,14,17].

Mechanisms of FLuc inhibition
FLuc inhibitors are typically linear, planar small molecules and some related structure-activity-relationships have been established in the literature [4,11,18]. Interestingly, inhibitors comprise a broad range of chemotypes (Fig. 1b), revealing FLuc to be a highly promiscuous enzyme. This wide structural diversity complicates prediction of compounds as FLuc inhibitors, hindering recognition of assay artefacts. It also suggests that multiple binding sites might exist on the FLuc enzyme to accommodate such a range of inhibitors.
FLuc inhibitors which compete with the ATP substrate have been less frequently observed but include a series of quinoline containing compounds 9 which were also found to compete with luciferin [11]. The lower frequency of identification of inhibitors of this mechanism is likely due to high saturating intracellular concentrations of ATP (~1 mM) which masks the evidence of this inhibition mode. Multisubstrate adduct inhibitors (MAI) are single molecules comprising two or more substrates of the enzyme [34,35]. This approach often leads to highly potent and selective inhibitors, with examples for FLuc including the naturally occuring L-AMP inhibitor and PTC124 10 (Ataluren/Translarna), both of which contain carboxylic acids and form adenylate adducts after reaction with ATP and release of pyrophosphate [36]. PTC124 has been granted conditional approval in Europe as a treatment for Duchenne muscular dystrophy in patients with nonsense mutations [37]. It was originally discovered using a cellular FLuc reporter nonsense codon assay [38] but was subsequently found to inhibit FLuc with an IC 50 of 7 nM, correlating to the activity in the original assay [20].
Numerous examples of inhibitors of FLuc which are non-competitive with respect to luciferin or ATP or both exist in the literature. These include compounds such as resveratrol 11 [9] and NF-κB inhibitor 12 [39], which continue to be used in luciferase reporter assays without control for their artefactual behavior [40]. Other non-competitive inhibitors of different chemotypes have been identified with IC 50 values as low as 0.26 nM [11,12,18]. However, the binding mode of these inhibitors has not yet been reported. The LH 2 -AMP mimic DLSA 13 reported by Branchini et al. was found to be non-competitive with respect to the two substrates luciferin and ATP, but competitive with the LH 2 -AMP intermediate [41]. This was rationalised by the significant conformational change involving a~140°rotation of the C terminal domain that occurs after the activating adenylation reaction forming LH 2 -AMP, offering different binding environments within the same active site [15].
In this work, the small molecule SMT C1100 (ezutromid) 14 was identified as a non-competitive FLuc inhibitor. It inhibits FLuc via a novel mechanism, along with other reported non-competitive inhibitors.

SMT C1100 is an FLuc inhibitor
SMT C1100 was originally discovered using a FLuc reporter gene assay for utrophin upregulation. Although it was later shown to upregulate utrophin at both the mRNA and protein level in DMD mouse models and in human DMD patients' cells [42][43][44], in dose-response analyses in the original reporter gene assay it displayed a bell-shaped curve, distinctive of FLuc inhibitors which stabilise FLuc and inhibit the bioluminescence reaction at high concentrations (Fig. 2a). This led us to suspect that an interaction between ezutromid and FLuc may be at least contributing to the output. As anticipated, when subjected to an in vitro enzyme assay with recombinant FLuc, SMT C1100 was found to be a potent FLuc inhibitor, with an IC 50 of 0.033 µM (Fig. 2b). Since SMT C1100 does not absorb above 380 nm, it cannot quench the green tone FLuc luminescence ( Supplementary Fig. S1), so must be interfering with FLuc via a different mechanism. Conservation of SMT C1100′s inhibition of FLuc in the presence of a range of detergents ( Supplementary  Fig. S2) excluded small molecule aggregation as the mechanism of inhibition [45]. SMT C1100 is however highly fluorescent (excitation at 317 nm, emission at 390 nm), so its binding affinity for FLuc could be determined via quenching of the compound's fluorescence upon binding of the protein. Fitting of the fluorescence quenching with a four parameter logistic function gave an apparent K D of 0.40 ± 0.15 µM and Hill coefficient of 0.98 ± 0.24 (Fig. 2c). This apparent FLuc K D correlates well with the EC 50 from the FLuc reporter gene assay (0.36 ± 0.22 µM, Fig. 2a), indicating that the original result was likely biased by FLuc interference.

SMT C1100 does not compete with luciferin or ATP
SMT C1100′s benzoxazole core and planar, hydrophobic structure are characteristic of luciferin competitors. Docking studies using the benzothiazole inhibitor 7 bound FLuc structure (PDB: 4e5d [4]) as a model supported this hypothesis (Fig. 3a), with SMT C1100 overlaying the inhibitor pose as anticipated. However, kinetic experiments varying first luciferin then ATP (with the other substrate held at saturating concentrations) indicated that substrate competition was not the mechanism by which SMT C1100 operates. Instead, fitting of initial rates to the Michaelis-Menten equation under rapid-equilibrium steady state kinetics led to determination of non-competitive inhibition with respect to luciferin, and uncompetitive with respect to ATP (Fig. 3d-e). These results were supported by Lineweaver-Burk analysis of the results, which showed FLuc's K M for luciferin unchanged at 18.2 µM (close to the literature value of 16.9 µM [23]) as is expected for non-competitive inhibition (Fig. 3b) and the K M for ATP decreasing as seen in uncompetitive inhibition (Fig. 3c). The K I was calculated as 0.76 ± 0.43 μM from Dixon plot analysis of the results ( Supplementary  Fig. S3).
To assess whether SMT C1100 could be competing with LH 2 -AMP for binding to FLuc's second catalytic conformation, further kinetic experiments were carried out, varying LH 2 -AMP which was supplied as the sole substrate (initial velocity and Lineweaver-Burk plots in Fig. 4). DLSA was used as a positive control and was confirmed to compete with LH 2 -AMP ( Supplementary Fig. S4). SMT C1100 was found not to compete with LH 2 -AMP, but instead was shown to display partial mixed inhibition, meaning it binds to both the FLuc and the FLuc:LH 2 -AMP complex, and also does not fully inhibit light production. Other compounds that have been found to bind via mixed type inhibition have also been reported. A class of aryl triazoles were found to have mixed non-competitive inhibition with respect to substrate analogue aminoluciferin [46], while some pyrrolo [2,3-d]pyrimidines had mixed competitive inhibition with respect to luciferin [47]. To determine whether SMT C1100′s mode of inhibition is similar to that of other reported noncompetitive inhibitors, resveratrol 11, NF-κBAI4 12 and compound 15 were included in further experiments ( Supplementary Fig. S4). Resveratrol and structurally related 12 were found to have non-competitive and partial non-competitive inhibition respectively regarding LH 2 -AMP, whilst 15 had mixed partial inhibition similar to SMT C1100 (table , Fig. 4c). These findings indicate that these non-competitive inhibitors are binding via a previously undescribed mechanism. This indicates that FLuc has multiple inhibitor binding sites and conformations beyond that which is currently understood.
X-ray crystallography is a powerful method for determining proteinligand binding modes. Co-crystallisation of FLuc with SMT C1100 alone or in the presence of ATP, luciferin or DLSA was attempted in this work to determine the exact location of the SMT C1100 binding site. However crystallisation attempts yielded only apo structures matching those previously reported in the literature [30,31]. This suggests that SMT C1100 may bind to a higher energy conformation of FLuc which we have been unable to crystallise.

Design and synthesis of a photoaffinity probe
In the absence of X-ray crystal data, a photoaffinity strategy was employed to elucidate the binding mode of SMT C1100. This approach involves incorporation of a photoaffinity group into a probe compound then exploiting its affinity for its protein target (Fig. 5b). Once bound, UV irradiation to activate the photoaffinity group results in carbene generation, then covalent binding to nearby residues [48]. The site of incorporation can be established by mass spectrometry to reveal the molecule's binding site. A trifluoromethyl-phenyl diazirine (TPD) was selected for incorporation into SMT C1100′s structure, due to its small size and expected small impact on FLuc binding affinity. Indeed, the sterics and electronics of the diazirine moiety appear a good match SMT C1100′s sulfonyl group. Furthermore, TPDs are well-characterised and have been successfully used in the literature for binding site identification studies [49][50][51]. Photoaffinity probe 16 retained FLuc inhibition behaviour, although with a reduced affinity, IC 50 = 0.595 ± 0.112 µM ( Supplementary Fig. S5).

Oxidation of probe reveals an alternative mechanism
However, incubation of the probe with FLuc, followed by UV irradiation (365 nm) and LC-MS of the reaction mixture resulted in no observation of probe incorporation into the enzyme (Fig. 5d). Instead, two species were generated from the probe, one whose mass corresponds to the addition of H 2 O 2 assigned as hydrate 18 the other corresponding to the addition of water, assigned as 17 (Fig. 5a). This finding was also observed with the addition of substrates luciferin, ATP and DLSA. However, interestingly, when the probe was irradiated in the absence of FLuc (Fig. 5c), only the water adduct was observed, indicating the requirement of the enzyme for the oxidation of the probe. ESI MS/MS fingerprint analysis of the putative water 17 and H 2 O 2 18 adducts generated in the irradiation experiments corresponded with the spectra generated from authentic samples of the compounds (Fig. 5e, synthesis described in the SI). Addition of molecular oxygen to TPDs has been reported in the literature [52] and FLuc has oxygenase activity (Fig. 1a), indeed an oxygen/pantetheine tunnel to the active site that has been postulated by Branchini et al. [53,54]. Furthermore, the approach of oxygen to the active site has been modelled by molecular dynamics and mechanics, finding that oxygen can move inside FLuc without any energetic barrier [55]. The results of this experiment suggest that probe 16 might bind to an oxygen-bound FLuc conformer.

Conclusions
SMT C1100, although shown to genuinely increase utrophin at the mRNA and protein levels in vitro and in vivo [42][43][44], has also been identified as an FLuc inhibitor, requiring cautious interpretation of results from its use in FLuc reporter gene assays. SMT C1100 is a noncompetitive inhibitor with respect to luciferin and uncompetitive with respect to ATP. SMT C1100, alongside three other literature non-competitive inhibitors, was found to have previously undescribed noncompetitive and mixed inhibition mechanisms with respect to the intermediate LH 2 -AMP. Photoirradiation studies to identify the binding site of a SMT C1100 probe revealed the formation of an FLuc-dependent oxidation product, consistent with binding of the probe in the proximity of molecular oxygen. However, the binding site of these, along with many other non-competitive inhibitors, is still uncharacterised. This along with a more detailed understanding of the FLuc catalytic mechanism and the dynamic conformational changes involved may form the basis of future studies.
Analytical procedures and chemical synthesis of LH 2 -AMP, DLSA, 15, 16, 17 and 18 are described in the supplementary information.

Cell culture
H2K-mdx utrnA-luc cells [43,56] were maintained in DMEM (Life Technologies) supplemented with 20% Fetal Bovine Serum (Life Technologies), 2% CEE (SLI), 2 mM L-Glutamine (Life Technologies), 1% Penicillin Streptomycin (Life Technologies) and 2 µg/500 mL Mouse Interferon-γ (Roche). Cells were maintained at 10% CO 2 at 33°C. The same data presented in double reciprocal Lineweaver-Burk plots; (c) Table of the effects on FLuc kinetic parameters K M and V max induced by DLSA, SMT C1100, resveratrol, NF-κBAI4 and 15, and whether full or partial inhibition is observed. K M and V max were calculated from slope and the reciprocal of the intercept of Lineweaver-Burk plots (Supplementary Fig. S4b). Mechanism of inhibition with respect to LH 2 -AMP (red box) was determined after analysis using both Dynafit and Lineweaver-Burk methods.

Utrophin-A FLuc reporter gene assay
White flat bottomed 96 well plates (Corning) were seeded with 5000 H2K-mdx utrnA-luc cells. After 24 h at 10% CO 2 and 33°C, the cells were dosed with compound in triplicate, in the following concentration series: 0.01, 0.03, 0.1, 0.3, 1.0, 3.0, 10.0 μM from 10 mM solution stocks in DMSO (final DMSO concentration was 0.3%). The cells were incubated for a further 24 h, (10% CO 2 , 33°C). Relative luminescence readout after using the Luciferase Assay System (Promega, E1500) reagents was measured using a FLUOstar Optima plate reader (BMG Labtech). The means from the biological triplicates were fitted with a four parameter logistic function with least squares regression (Levenberg-Marquardt algorithm) to calculate EC 50 values.

Biochemical FLuc inhibition assay
Recombinant FLuc was assayed at a final concentration of 0.6 nM in a buffer containing 25 mM HEPES, 5 mM MgCl 2 , 1 mM EDTA, 5 mM DTT and 1 mg/ml BSA. Compounds were diluted in the following concentration series: 0 nM, 0.95 nM, 3 nM, 30 nM, 95 nM, 0.3 μM, 0.95 μM, 3 μM, 9.5 μM, 30 μM, from 10 mM stocks in DMSO (with a final assay concentration of DMSO at 0.3%). PTC124 was used as a positive control [36]. Luciferase substrates ATP and D-luciferin were used in a final assay concentration of 10 μM. Luciferase was pre-incubated with ATP and the query compound at 0°C for 15 min. D-Luciferin was dispensed and endpoint luminescence output immediately read using a FLUOstar Optima plate reader (BMG Labtech). Luminescence output was fitted with a four parameter logistic function with least squares regression (Levenberg-Marquardt algorithm) to calculate EC 50 values.

FLuc fluorescence quenching assay
Fluorescence quenching of SMT C1100 upon binding recombinant FLuc was monitored using a fluorescence spectrophotometer (Hitachi F-4500 FL spectrophotometer, 317 nm excitation/390 nm emission). 0.3 μM SMT C1100 was incubated with FLuc (various concentrations ranging 0 -3 μM) for 15 min in a buffer containing 25 mM HEPES, 5 mM MgCl 2 , 1 mM EDTA, 5 mM DTT and 1% DMSO. Fluorescence emission was measured in each sample and the K D calculated by fitting the curve with a four parameter logistic function.

Docking of FLuc luciferin pocket
Docking studies were carried out using AutoDock Vina using default settings [57]. The luciferin pocket of FLuc crystal structure PDB: 4e5d was prepared as follows: bound ligands were extracted using PyMOL, the protonation state relevant for pH 7.4 was applied using a PDB2PQR [58,59] server and the binding box was prepared using Auto-DockTools4 [60]. Ligand files were prepared from sdf files with 3D coordinates generated using ChemAxon tools [61] then prepared for docking using AutoDockTools4.

FLuc kinetics assay, varying ATP, luciferin or LH 2 -AMP as the sole substrate
Recombinant FLuc was assayed at a final concentration of 0.6 nM in a buffer containing 25 mM HEPES, 5 mM MgCl 2 , 1 mM EDTA, 5 mM DTT and 1 mg/ml BSA. SMT C1100 was diluted in a concentration series from 10 mM stocks in DMSO (with a final assay concentration of DMSO at 0.3%). Either D-luciferin or ATP was varied in the following concentration series: 1 μM, 5 μM, 10 μM, 50 μM, 100 μM, 500 μM in triplicate, with the other substrate held at the saturating concentration of 250 μM (added immediately prior to measuring luminescence output). For the LH 2 -AMP experiment, a concentration series of 0.5 μM, 1 μM, 2.5 μM, 5 μM, 10 μM, 25 μM LH 2 -AMP was used and no other substrate was supplied. Luminescence output was measured every 0.04 s for the first 4 s of the reaction using a FLUOstar Optima plate reader (BMG Labtech) and the initial rate was calculated by linear slope fitting. The initial rates data were plotted and fitted using Dynafit4 for inhibition model discrimination by Akaike and Bayesian Information Criteria [62].

FLuc diazirine irradiation study
20 μM diazirine probe 16 was incubated with 10 μM recombinant FLuc in a buffer containing 25 mM HEPES, 5 mM MgCl 2 , 1 mM EDTA, 5 mM DTT and 5% DMSO. Where applicable, 100 μM ATP, 50 μM luciferin or 20 μM DLSA were included. After 15 mins incubation at 0°C, the samples were irradiated (365 nm, 100 W UV lamp, UVP™ B-100AP) for 3 mins at 0°C from a distance of 6 cm. The samples were diluted 1:1 with milli-Q water and submitted for LC-MS analysis on a Waters Acquity 1525 μHPLC system coupled to a LCT Premier XE (Waters) mass spectrometer. A flow rate of 0.3-0.75 mL/min was applied to a mobile phase of A = water + 0.1% formic acid, B = MeCN with a gradient of %A: 0.0 min 97%, 1.0 min 97%, 5.0 min 2%, 6.0 min 2%, 7.0 min 97%, 10.0 min 97%. The electrospray source had a capillary voltage of 3.00 kV and cone voltage of 100 V. Nitrogen was the nebuliser and desolvation gas, flow of 300 L/hr. Total mass spectra were deconvoluted using the MaxEnt algorithm in MassLynx 4.1 (Waters) according to the default settings.
ESI MS/MS analysis was carried out on an Acquity-UPLC system (Waters) connected to a Xevo G2-XS Q-TOF mass spectrometer (Waters) equipped with an electrospray ion source. The analyte separated on an ACE equivalent 3 C18 analytical column (2.1 mm i.d. × 50 mm, 3 μm, 100 Å) using a linear gradient (length: 10 min, 5% to 95% solvent B (0.1% formic acid in acetonitrile)) at a flow rate: 0.4 mL/min (solvent A 0.1% formic acid in water). The separated peptides were electrosprayed directly into the mass spectrometer operating in a full scan method using a CID based method. A full scan MS spectrum was operated in electrospray positive mode with a scan range 100-900 m/z and a scan time 1.5 s. Lockspray was used during analysis to maintain mass accuracy. Data was obtained in continuum mode. The electrospray source of the MS was operated with a capillary voltage of 3.00 kV and a cone voltage of 40 V. CID fragmentation was performed at a ramp from 35 to 45 V of normalized collision energy. The MS and MS/MS analysis was processed using Masslynx. The mass window search of precursor ion was set as 5 ppm error. All spectra were manually checked and validated.

Data availability statement
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).