A high-throughput effector screen identifies a novel small molecule scaffold for inhibition of ten-eleven translocation dioxygenase 2

Ten-eleven translocation dioxygenases (TETs) are the erasers of 5-methylcytosine (mC), the central epigenetic regulator of mammalian DNA. TETs convert mC to three oxidized derivatives with unique physicochemical properties and inherent regulatory potential, and it initializes active demethylation by the base excision repair pathway. Potent small molecule inhibitors would be useful tools to study TET functions by conditional control. To facilitate the discovery of such tools, we here report a high-throughput screening pipeline and its application to screen and validate 31.5k compounds for inhibition of TET2. Using a homogenous fluorescence assay, we discover a novel quinoline-based scaffold that we further validate with an orthogonal semi-high throughput MALDI-MS assay for direct monitoring of substrate turnover. Structure–activity relationship (SAR) studies involving >20 derivatives of this scaffold led to the identification of optimized inhibitors, and together with computational studies suggested a plausible model for its mode of action.


Plasmid and cloning:
A TET2 construct was designed as previously described by Hu et al (Hu et al., 2013) spanning the C-terminal catalytic domain (1129-1936aa). To facilitate expression in E. coli, the low-complexity insert (1481-1843aa) which is predicted to be unstructured in solution was replaced by a GS-linker. The corresponding gene was codon-optimized for E. coli expression, synthesized and subsequently subcloned into a pET-15b vector succeeding a His6tag and thrombin cleavage site by using the NdeI and XhoI restriction sites (Geneart AG, Germany).

Protein expression and purification:
The plasmid was transformed into chemically competent BL21 (DE3) E. coli cells by heat shock and subsequently selected on LB agar containing 100 µg/mL ampicillin by incubation at 37 °C overnight. The next day, the transformed clones were washed off with 5 mL TB medium, transferred to additional 100 mL TB medium containing 100 µg/mL ampicillin and grown at 37 °C and 150 rpm overnight. Subsequently, 10 L of TB medium containing 100 µg/mL ampicillin were inoculated 1:100 with the starter culture and grown to OD600 = 0.8 at 37 °C, 150 rpm before target protein expression was induced by addition of 0.5 mM IPTG.
The expression culture was shaken for additional 16 h at 18 °C, 150 rpm followed by centrifugation at 4000 x g, 4 °C for 20 min. The cell pellet was snap frozen in liquid nitrogen and stored at -80 °C. The cell pellet was resuspended in buffer A (50 mM Tris, 500 mM NaCl, 1 mM DTT, 5% glycerol, pH 8.0) and lysed using a microfluidizer. The lysate was cleared by centrifugation at 4000 x g, 4 °C for 60 min and the resulting supernatant was loaded onto a hand-packed 20 mL Ni-NTA column (Ni-NTA Superflow, Qiagen) preequilibrated with buffer A. Afterwards, the column was washed with 10 column volumes (CV) buffer A and nonspecifically bound proteins were eluted with 4% buffer B (50 mM Tris, 500 mM NaCl, 500 mM imidazole, 1 mM DTT, 5% glycerol, pH 8.0). Finally, the target protein was eluted using a gradient from 4% up to 50% of buffer B over 30 min. Fractions containing the His6-tagged TET2 were pooled, diluted 1:20 with buffer C (50 mM Tris, 10 mM NaCl, 1 mM DTT, 5% glycerol, pH 8.0) and loaded onto an anion-exchange chromatography column (5 mL HiTrap Q FF, GE Healthcare) pre-equilibrated with buffer C.
The resin was washed with 10 CV buffer C and a gradient up to 30% of buffer D (50 mM Tris, 1000 mM NaCl, 1 mM DTT, 5% glycerol, pH 8.0) was used to elute bound proteins.
The protein was concentrated to approximately 5 mg/mL, snap-frozen in liquid nitrogen and stored at -80 °C. The identity and the purity of the protein was determined by electrospray ionization mass spectrometry (Velos Pro Dual-Pressure Linear Ion Trap Mass Spectrometer, Thermo Fisher Scientific).

LC-MS analysis of purified protein:
The His6-tagged TET2 protein at a concentration of 1.0 mg/mL was analyzed by ESI-MS using a Thermo Fisher Scientific Dionex UltiMate 3000 HPLC system connected to a Thermo Fisher Scientific Velos Pro (2d ion trap). 1 μL of the sample was injected and separated using a AdvanceBio Desalting-RP cartridge starting at 5% of solvent B for 5 min, followed by a gradient up to 80% of solvent B over 2.5 min with a flow rate of 400 μL/min with 0.1% formic acid in water as solvent A and 0.1% formic acid in MeCN as solvent B. A mass range of 700-2000 m/z was scanned, and raw data were deconvoluted and analyzed with MagTran software (Zhang and Marshall, 1998).

FRET assay for quantification of in vitro TET activity:
The FRET assay was performed in microplates ( Cy3 was set to 552 nm and the emission wavelength for Cy5 to 665 nm.

S23
For biochemical characterization, the hits from the screen and the compounds from the virtual screen were added to the DNA-TET2-solution as serial dilutions (8 concentrations, 2-fold dilution, 200 µM highest concentration) with the Echo 520 Liquid Handler. The reaction mixtures were preincubated for 1 hour before the iron solution was added.

High-Throughput inhibitor screening of small molecule library:
The screening of 31500 compounds was performed using the in-house RASPELD unit Tris-acetate, 10 mM magnesium acetate, 100 μg/mL BSA, pH 7.0) were added to each well and the Cy5 fluorescence was recorded on an Infinite M1000 plate reader (Tecan) immediately after (t0). After an additional incubation for 4 h at room temperature, the Cy5 fluorescence was recorded again (t240). Relative Cy5 fluorescence was calculated as the ratio of Cy5 fluorescence intensities at t240 and t0. The Z-factor, a statistical parameter reflecting assay robustness and its suitability for high-throughput screening (Zhang et al., 1999) was determined as follows: σ: standard deviation, μ: mean, pos: positive control, neg: negative control S24

Hit validation and IC50 determination using orthogonal semi-high throughput MALDI assay:
The MALDI assay was performed in microplates (384 well, flat bottom, white, polystyrene, small volume, Greiner). The pipetting was done with hand (for assay validation) or the help of the Multidrop™ Combi Reagent Dispenser (for determining dose response of the SAR compounds) and the Echo 520 Liquid Handler (for compound dilution and transfer). For desalting of the sample for MALDI analysis, 10 µL aqueous slurry (50%) cation exchange resins (BioRad) was added to the reaction mixture. Beforehand, the cation exchange resins were washed with deionized water (3 times) followed by equilibration with 1M ammonium bicarbonate solution. Finally, the equilibrated resins were washed with deionized water (3 times) before being added to the reaction mixture for desalting.
The desalted assay plate was briefly centrifuged and 1 µL of the supernatant was spotted onto a solidified matrix of 50 mg/mL 3-hydroxypicolinic acid in 50% acetonitrile aq / 0.1% TFA aq and 10 mg/mL diammonium hydrogen citrate on a ground steel MTP 384 target. Spectra were recorded in linear positive mode on a Bruker ultrafleXtreme MALDI.TOF/TOF system.

Processing of MALDI mass spectra and IC50 curve fitting:
The recorded MALDI spectra were processed with a custom script (part of the summerrmass R package available under https://zenodo.org/record/5501758; pipeline template-A01.R) to extract the desired peak heights and determine the IC50 for each compound with given additional metadata and internal controls. In brief, Bruker FID files were converted to mass spectrometry XML (mzXML) using CompassXport v3.0.4 (Bruker Daltonik GmbH, Bremen, Germany) and processed in R v4.0.5(R: A language and environment for statistical S25 computing., 2021) using functions implemented in the packages MALDIquant (Gibb and Strimmer, 2012) and MALDIquantForeign (Gibb, 2019). Specifically, the spectra baselines were removed using the statistics-sensitive non-linear iterative peak-clipping (SNIP) algorithm before spectra of replicate measurements of the same well (if present) were aligned using the default locally weighted scatterplot smoothing (

Re-synthesis of the SAR analogues:
General details: All reagents and solvents were purchased from Enamine, Acros, Activate Scientific, Alfa purified by column chromatography using VWR silica gel (40 -63 μm particle size) or flash chromatography on a Biotage Isolera One using Büchi Reveleris Silica Cartridges (4 -120 g) monitored by UV at λ = 210 nm and 280 nm. General scheme for the synthesis of sulfonic esters and amides

8-fluoro-3-iodoquinoline-5-sulfonic acid (2)
Water (1 mL) was slowly added to the stirred solution of quinoline-5-sulfonyl chloride (A) (30 mg, 0.1 mmol) in 1, 4-dioxane (2 mL) at 5 °C and the reaction was allowed to stir for 4-5 hours at room temperature. Excess solvent was evaporated and the crude was purified by CC to afford the title product as off white solid (72%

8-fluoro-3-iodoquinoline-5-sulfonamide (8)
A solution of sat. aqs NH4OH (1 mL) was added to a cold (0°C) solution of quinoline-5sulfonyl chloride (A) (30 mg, 0.1 mmol) in dioxane (1 mL) and the reaction mixture was allowed to stir for overnight at RT. Water (10 mL) was added and extracted twice with EtOAc (20 mL) and twice with DCM (20 mL). The combined organic fractions were dried over Na2SO4 and the solvent was evaporated to afford the title product which was purified by column chromatography provides the title compound as pale yellow solid (80%

General procedure for the preparation of sulfonic esters (Method A)
Excess equivalents of corresponding alcohol (0.5 mL) was slowly added to the stirred solution of quinoline-5-sulfonyl chloride (30 mg, 0.1 mmol) in dry DCM (2 mL) containing one drop of pyridine (catalytic) at 0 °C in a separate reaction vials and the reaction was allowed to proceed 4-8 hours at RT. Water was added after completion of the reaction and extracted twice with EtOAc (15 mL). The combined organic fractions were dried and the solvent was evaporated. Crude RM was purified by CC to afford the corresponding Quinoline sulfonic ester derivatives.
toggling and spinning was enabled. Automated binding site detection was disabled, and the binding site was restricted to all atoms in a 10 Å environment of the reference ligand atoms.
Per ligand, 100 GA runs were performed with a search efficiency of 2. We performed two docking runs for the docking with α-KG as reference ligand: a) the Fe 2+ ion was retained as metal ion in the binding site 2) the Fe 2+ ion was removed from the binding site. Apart from this, default settings were applied. The docking results of all docking runs were visually inspected. For most molecules, the three highest scored poses were highly similar. These poses were chosen for the investigation of potential binding modes of the hit compound 2.

Binding Site Comparison:
The preparation of the binding sites from the scPDB (Meslamani et al., 2011) was performed as described elsewhere (Ehrt et al., 2018). The same preparation steps were applied to the structure of TET2 (PDB-ID 5deu). The results of the comparison of the 2-oxogluatarate binding site of TET2 with all binding sites in the scPD with IsoMIF (Chartier et al., 2016) can be found in the SI. The IsoMIF-based alignment of the structure of HIF-1 (PDB-ID 3od4) was used to align FTO (PDB-ID 4ie4) to TET2.

DFT Calculations:
The complex model shown in Fig. S15 was geometry-optimized on ωB97X-D/dev2-SVP level of theory using the Wavefunction Spartan '18 software package and obtained structures subjected to ωB97X-D/dev2-TZVP single point energy calculations. To keep the electronic situation of the transition metal center as simple as possible and comparable for all herein compared coordination environments, comprising different ligand combinations, the evenelectron-count d 6 octahedral Fe(II) center was treated as a low-spin complex (spin multiplicity = 1) using restricted wave function computations for all examined complexes, even though spectroscopic studies on related water-substituted (but alpha-ketoglutarate-unbound) Fe(II) complexes in the resting states of alpha-ketoglutarate-dependent oxidases with one carboxylate (or halide) and two histidine ligands indicated a high-spin (S = 2) situation (Chang, 2018). In the following, the deprotonated 8HQ ligand was replaced by the 8fluoroquinoline ligand motif found in compound 2, further by 8HQ (non-deprotonated), quinoline and pyridine, to learn about the relative energies leading to formation of these coordination environments by ligand exchange. Therefore, complex formation reactions shown in Fig. S16 were computed by performing unconstrained gas-phase geometry optimizations on DFT ωB97X-D/dev2-SVP level for all components, followed by ωB97X-D/dev2-TZVP single point calculations.