Shubhendu
Palei‡
,
Jörn
Weisner‡
,
Melina
Vogt
,
Rajesh
Gontla
,
Benjamin
Buchmuller
,
Christiane
Ehrt
,
Tobias
Grabe
,
Silke
Kleinbölting
,
Matthias
Müller
,
Guido H.
Clever
*,
Daniel
Rauh
* and
Daniel
Summerer
*
Department of Chemistry and Chemical Biology, TU Dortmund University and, Drug Discovery Hub Dortmund (DDHD), Zentrum für Integrierte Wirkstoffforschung (ZIW), Otto-Hahn Str. 4a, 44227 Dortmund, Germany. E-mail: guido.clever@tu-dortmund.de; daniel.rauh@tu-dortmund.de; daniel.summerer@tu-dortmund.de
First published on 2nd September 2022
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.
Recent efforts have afforded the first small molecule inhibitors of TETs17–19 (the cytosine-based compound Bobcat339 (ref. 21) has recently been shown to not be an inhibitor of TET1/2).22 However, a particular challenge of inhibiting TETs is their dependence on abundant substrates and cofactors that have additional key functions in cells and are used by a large number of related enzymes. This complicates the use of substrate analogues with high structural similarity, or the use of compounds that alter cofactor availability. For example, N-oxalylglycine (NOG) and its esterified derivative dimethyloxalylglycine (DMOG)16 are popular TET inhibitors that need to be applied in millimolar concentrations and differ only minimally in structure from α-KG, an abundant metabolite with numerous functions in cell physiology.23 Other inhibitors rely on the complexation of free Fe(II) (e.g. deferoxamine, DFOA24). Hence, there is a need to identify novel scaffolds that differ more substantially from the natural substrates and thus provide new avenues for the design of selective inhibitors. For example, an inhibitor was recently identified by a pure in silico screening that was proposed to bind the TET active site without competing with α-KG.17
To enable an effective experimental access to the discovery of novel small molecule TET inhibitors, we here report a high throughput pipeline for the screening and validation of TET inhibitors from large compound collections. We developed a homogenous one-pot FRET-assay for quantification of TET activity in vitro and employed it for screening a compound library comprising 31.5k chemically diverse molecules. We identified an inhibitor of human TET2 based on a novel tri-substituted, sulfonamide-derivatized quinoline scaffold. Careful hit validation and structure–activity relationships using a derivative library of this scaffold with an orthogonal MALDI-MS assay revealed a simple sulfonic acid derivative as low micromolar inhibitor of TET2 and together with computational studies provided clues to its mode of action.
We first developed an assay based on the ability of the methylation-sensitive restriction endonuclease MspI to cleave target DNA duplexes containing a CpG dinucleotide with an mC, hmC, or fC, but not caC in its recognition motif.28 Use of a DNA duplex terminally labelled with a FRET donor/acceptor dye pair (Cy3/Cy5) thereby enables measuring the amount of caC product generated from an mC duplex by TET (Fig. 1A). We expressed and purified the catalytic domain of TET2 (TET2-CD) with an N-terminal His6-tag followed by a thrombin cleavage site (Fig. 1B and S1†). We replaced the low-complexity (LC) insert of the catalytic domain by a GS linker for better expression in E. coli, since this has been shown to not affect its activity in vitro.29 We then designed a 12-bp oligonucleotide containing a central MspI recognition sequence (‘CCGG’) containing an mC nucleobase in the CpG dinucleotide, and hybridized it with the Cy3/Cy5-labeled antisense strand. We identified an universal buffer system (50 mM potassium acetate, 20 mM Tris-acetate, 10 mM magnesium acetate, 100 μg ml−1 BSA, pH 7.9) that was compatible with the activity of both TET2 and MspI, and thus was applicable to one pot reactions without intermediate purification steps. We incubated 0.5 μM of the DNA duplex with TET2 at different concentrations under aerobic conditions and in the presence of essential assay components, i.e. α-KG, ascorbate, dithiothreitol (DTT), and Fe(II) at 37 °C and pH 8 for 1 hour. We then added 20 U MspI enzyme to determine the caC level generated by TET2-mediated oxidation. We observed a high Cy5 fluorescence (high FRET) only for reaction mixtures containing TET2, indicating successful oxidation of mC and blockage of MspI. A 4-fold molar excess of TET2 enzyme (2 μM) was thereby sufficient for a quantitative oxidation of mC in 1 hour at 37 °C and subsequent inhibition of MspI activity (Fig. 1C).
For an assay upscale, we chose a concentration of 1 U MspI per reaction to achieve quantitative cleavage and baseline Cy5 fluorescence after incubation at RT for 4 hours (Fig. 1D, see Fig. S2† for optimization of duplex concentration).
As a final requirement for HTS, we found that the assay can readily tolerate up to 4% final DMSO concentration without altering TET2 or MspI enzymatic activity (Fig. S3†), making it highly compatible with the screening of compound collections stored in standard DMSO stocks. The Z-factor (Z′) of this optimized FRET assay in 384-well plates was 0.69, indicating a high robustness and compatibility with our screening conditions (Fig. S4†).30
We then evaluated the ability of our screening assay to detect and profile the inhibitory activity of previously reported broad-spectrum dioxygenase inhibitors with two different modes of action. We thereby obtained for the Fe(II) chelator DFOA and the α-KG analog NOG16 IC50 values of 46 μM and 11.3 mM, respectively (Fig. 1E).
For screening, we premixed the annealed oligonucleotides with TET2 and all other essential components except Fe(II) in 1536-well plates, followed by the acoustic transfer of test compounds into each well at a final concentration of 50 μM. We preincubated the compounds with the mixture of oligonucleotide and TET2 enzyme for 30 minutes at RT in a humidified chamber before the Fe(II) solution was added to start the reaction. We incubated the plate at 37 °C for 1 hour for TET2-mediated mC oxidation. Afterwards, we added MspI and incubated the plates for an additional 4 hours at RT. In each screening plate, we included DMSO as negative (0% inhibition) and 1 mM DFOA as positive (100% inhibition) controls to determine the assay quality (Z′). In order to distinguish the false-positive hits, we normalized the final Cy5 fluorescence of each well to its respective initial value immediately after MspI addition. We selected compounds as primary hits for further evaluation if they exhibited >50% inhibition of TET2 activity (Fig. 2B shows results of an exemplary 1536-well plate).
In total, we screened 24 individual 1536-well plates yielding a global Z-factor of 0.59 ± 0.05, which confirmed the robustness of the FRET assay in a 1536-well screening format (Fig. 2C). We identified six primary hits (corresponding to a hit rate of 0.02%) for further evaluation. We reordered all six primary hit compounds freshly from commercial sources and performed dose-dependent analyses using the FRET assay in 384-well plates. We thereby identified compound 1, a quinoline scaffold linked to an isoxazole ring via a sulfonamide moiety to be the only active hit compound inhibiting TET2 in a dose-dependent manner (IC50 = 13 ± 0.5 μM) (Fig. 2D). The observed reduction of Cy5 fluorescence was not due to quenching by 1 (no reduction in Cy5 fluorescence intensity in the absence of MspI enzyme was observed), indicating a specific inhibition of TET2 by 1 (Fig. S6†). Validation of hits with an orthogonal MALDI assay. In order to validate our primary screening results, we developed an orthogonal medium-throughput MALDI pipeline enabling the quantitative analysis of substrate turnover and formation of the TET2 products hmC and fC (Fig. 3A). Mass spectrometry-based methods have been previously employed to directly measure TET activity in vitro.19,25,32 We aimed to miniaturize such assays to adapt them to a plate-based, semi-HT assay setting. For this purpose, we used MALDI-TOF analysis and developed an ‘R’ pipeline for batch processing of the large amount of MALDI raw data. This allows automated quantification of the intensity of educt (mC) and products (hmC, fC) for rapid and accurate determination of IC50 values.
We used an annealed 8-mer palindromic oligonucleotide duplex (5′-CACmCGGTG-3′) as a substrate. We preincubated a mixture of 0.5 μM duplex and 1 μM TET2 along with the essential assay components with a serial dilution of the inhibitor in a humidified chamber at RT for 30 minutes. We then added Fe(II) and incubated the mixture for one additional hour at 37 °C for TET2-catalyzed oxidation. For MALDI analysis, we desalted the reaction mixtures with cation exchange resin. We transferred the desalted samples to a 384-well MALDI sample plate in parallel using 3-hydroxypicolinic acid and diammonium hydrogen citrate as the matrix (Fig. 3A). We measured MALDI-TOF MS of the dose–response samples in duplicates, and batch-analyzed the raw data to determine absolute ion intensities of [M + H]+ ions of mC, hmC and fC (intensity of the caC ion was neglected due to its low abundance). We assumed that oxidation at a single CpG of the oligo would not change its ionization efficiency.33 We quantified the TET activity as the percentage of oxidized products (hmC + fC) formed, and then normalized that value to the negative DMSO control. We then revalidated our primary hit compound 1 using the orthogonal MALDI assay and determined the dose response curve with a corresponding IC50 value of 2.1 ± 0.9 μM (Fig. 3B and S7†). We used the MALDI assay for determination of all further dose response experiments and IC50 values because of its ability to directly quantify mC oxidation products.
Indeed, resynthesized 2 showed a potent inhibition of TET2 activity, with an IC50 value of only 2.3 ± 0.6 μM (Table 1). 2 was thereby the only active sulfonic acid in our screen, suggesting that the specific scaffold of 2 was required for its activity (similarly, other sulfonic acids, such as o-anilinesulfonic acid, p-toluenesulfonic acid or 2-bromoisonicotinic acid showed no inhibition of TET2 in additional experiments, data not shown). The inactivity of 1 and the high activity of 2 indicated a key role of the substituent at the 5-position of the quinoline ring. We thus synthesized a focused library of derivatives with altered substituents at this position (and few derivatives with alterations at the fluorine and iodine position of 1; see Table 1). We first synthesized a series of sulfonic acid esters with gradually increasing steric demand of the alcohol moiety (3–7). Whereas these esters were all inactive, we observed that 3–6 showed activity after prolonged storage in DMSO, indicating hydrolysis to 2 (Fig. S10† shows data for 3). However, our library contained 27 additional sulfonic acid esters that were all inactive in our screen, further supporting the requirement of the specific scaffold of 2 for its activity. As a closely related compound to 2, we next tested the simple sulfonamide 8, and indeed observed activity (IC50 = 86.6 ± 5.6 μM). We were therefore interested, if other sulfonamides may also be active, and first tested the methyl-derivative 9 and the secondary pyrrolidine sulfonamide 10, which turned out to be inactive (Table 1). We however found that substitution with a phenyl ring resulted in even higher activity (IC50 of 23.9 ± 4.1 μM) as compared to the free sulfonamide 8 (Table 1). This activity required aromaticity of the ring, indicated by the inactivity of the cyclohexyl sulfonamide 12. Moreover, neither methylene nor ethylene bridges to the phenyl ring, nor further variations of the phenyl substituent were tolerated (13–17, Table 1). Similarly, a number of derivatives with alterations at the fluorine and iodine positions were inactive (Fig. S16†).
Hence, with a top-down systematic SAR approach we could streamline our initial screening results and identify 8-fluoro-3-iodoquinoline-5-sulfonic acid 2 as a novel scaffold that inhibits TET2 with a low μM IC50 value. This further highlights the importance of systematic resyntheses/validation and SAR for a successful discovery and development of potent inhibitors (see Fig. S8 and S9† for full inhibition profiles of the active compounds 2, 8 and 11).
However, ester derivatives of 2 (including the smallest methyl ester 3) were inactive in our MALDI assay (Table 1). Therefore, though this binding mode might interfere with DNA binding and base flipping, it is unlikely to be correct.
In the second scenario, the inhibitor binds in an α-KG-competitive manner. We thus analyzed potential binding modes in the α-KG binding pocket. To this end, we retained one water molecule and the metal ion (Fe(II)) in the active site. Intriguingly, the resulting poses of 2 and the two inactive esters 3 and 5 serving as negative compounds showed a similar interaction profile in the binding site. The oxygen atoms of the sulfonic acid moieties of all three compounds were engaged in coordination to the Fe(II) ion and formed hydrogen bonds to the R1261 side chain, whereas the quinoline nitrogen atoms formed hydrogen bonds with the side chain of R1896 (Fig. 4A). Though the deprotonated sulfonate of 2 is expected to coordinate Fe(II) better than the uncharged esters of 3 and 5, this proposed binding mode does not explain the observed complete inactivity of 3 and 5.
Fig. 4 Binding hypothesis for 2. A) Compound 2 and its inactive ester derivates 3 and 5 were docked into the TET2 α-KG binding site in the presence of Fe(II). The highest-scored binding pose of each compound is shown as light blue sticks. The protein is shown as grey cartoon with amino acid side chains within 4 Å of the docked compounds shown as grey sticks. The Fe(II) ion is shown as orange sphere. The docking-derived binding modes do not explain the observed SAR for the compounds. A 2D-plot of potential residues interacting with active compound 2 in presence of Fe(II) is shown on the right. B) Docking pose of active compound 2 docked in the absence of Fe(II) ion depicted as in Fig. 4A. The docking pose can only be found for the active compound 2 as opposed to the esters 3 and 5 whose substituents would clash with side chains in proximity of R1896. C) Comparison of the TET2 α-KG site with IsoMIF. D) Structural alignment of fat mass and obesity associated protein FTO (PDB-ID: 4ie4) shown as orange cartoon with Fe(II) as sphere (M) in complex with the α-KG analogue inhibitor 8-hydroxyquinoline-5-carboxylic acid (IOX1) and of TET2 (PDB-ID: 5deu) shown as blue cartoon in complex with Zn2+ (shown as sphere, M) and N-oxalylglycine (NOG). The structural similarity between 2 and IOX1 as well as the binding site similarity between the two dioxygenases corroborate the binding hypothesis of Fig. 4B. Figures were generated using UCSF Chimera.35 |
We thus performed another docking study in that we aimed to avoid a potential score bias with respect to metal coordination by using a complex without the Fe(II) ion as a working model. This time, we found a distinct binding mode of compound 2, which was significantly different from the binding mode of the two inactive esters. In this mode, the sulfonic acid group is engaged in a salt bridge with the R1896 side chain and in two additional hydrogen bonds to the side chains of H1416 and S1898 (Fig. 4B). The quinoline nitrogen atom is located at the position that was originally occupied by a coordinating water molecule (HOH2113, PDB-ID 5deu). The fluorine atom occupies the position of an oxygen atom of the coordinating carboxylic acid moiety of the α-KG cofactor. We did not find comparable binding modes for the inactive ester and amide derivatives of 2 (Table 1) due to a steric hindrance at the bottom of the α-KG pocket formed by R1896. Therefore, this binding mode – which is consistent with our SAR – is the mode that best explains the observed TET2 inhibition by 2 among the ones we obtained from our models.
In order to gain comparable insights into the metal-coordinating abilities of both compounds, we conducted DFT-based calculations. We generated a simplified structural model by merging features of the observed coordination environments of an Fe(II)- and α-KG-bound TET2-hmC complex (PDB-ID: 5deu) and an HIF1α complex, where a deprotonated 8HQ chelates the metal center (the latter being a square-pyramidal Zn(II) cation; PDB-ID: 3od4). In the model, we kept the 8HQ chelate with carboxylate trans to the quinoline-N donor as in the HIF1α structure, but with an Fe(II) cation in an octahedral coordination sphere complemented by a water ligand (Fig. S13†). By performing unconstrained gas-phase geometry optimizations, we computed a range of complex formation reactions with protonated and deprotonated 8HQ, 8-fluoroquinoline, quinoline and pyridine involving the replacement of one or two coordinating water molecules (Fig. S14†). While binding of the deprotonated 8HQ was found to be energetically strongly favored (−453.8 kJ mol−1; −4.8 kJ mol−1 for the unlikely protonated form), binding of 8-fluoroqinoline was energetically uphill (+14.9 kJ mol−1), suggesting that the latter is a significantly weaker ligand for the Fe(II) center, however, still being somewhat stronger than a quinoline lacking the F-substituent (+43 kJ mol−1).
Together with our SAR and hypothesized binding model, we thus speculate that 2 may act as an isostere of IOX1 that has lower affinity to the conserved Fe(II) center and could thus represent a novel starting point for the development of selective dioxygenase inhibitors.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2md00186a |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2022 |