Cyclic peptides target the aromatic cage of a PHD-finger reader domain to modulate epigenetic protein function

Plant homeodomain fingers (PHD-fingers) are a family of reader domains that can recruit epigenetic proteins to specific histone modification sites. Many PHD-fingers recognise methylated lysines on histone tails and play crucial roles in transcriptional regulation, with their dysregulation linked to various human diseases. Despite their biological importance, chemical inhibitors for targeting PHD-fingers are very limited. Here we report a potent and selective de novo cyclic peptide inhibitor (OC9) targeting the Nε-trimethyllysine-binding PHD-fingers of the KDM7 histone demethylases, developed using mRNA display. OC9 disrupts PHD-finger interaction with histone H3K4me3 by engaging the Nε-methyllysine-binding aromatic cage through a valine, revealing a new non-lysine recognition motif for the PHD-fingers that does not require cation-π interaction. PHD-finger inhibition by OC9 impacted JmjC-domain mediated demethylase activity at H3K9me2, leading to inhibition of KDM7B (PHF8) but stimulation of KDM7A (KIAA1718), representing a new approach for selective allosteric modulation of demethylase activity. Chemoproteomic analysis showed selective engagement of OC9 with KDM7s in T cell lymphoblastic lymphoma SUP T1 cells. Our results highlight the utility of mRNA-display derived cyclic peptides for targeting challenging epigenetic reader proteins to probe their biology, and the broader potential of this approach for targeting protein–protein interactions.

Enrichment of the cyclic peptide NNK10-12 library against KDM7B dual domain. The recovery of DNA was determined by qPCR, using DNA standards generated by reverse transcription of an unbiased mRNA NNK starting library. Each recovery was calculated by taking the total quantity of DNA present at the end of the selection round, following magnetic bead extraction of binding sequences, as a percentage of the total quantity of DNA present in the pool immediately prior to selection. The structures of OC3, OC4 and OC9 are shown, and some of the most enriched sequences are tabulated. All OC peptides are cyclic, with thioether bond between N-and C-terminal amino acids. '%A' in the table represents percentage abundance of a given peptide sequence in the overall library count after round 5 NGS, with qualitative assessment of binding to KDM7B domains by BLI stated. (+) = binding, (-) = no binding, (NS) = non-specific.   Figure S2. BLI for KDM7 proteins with cyclic peptides.
BLI traces for binding to KDM7 protein domains immobilised by C-terminal His-tag on Ni-NTA biosensors, against cyclic peptides. Response trace for each peptide concentration is represented by a different colour, with the fitted curves overlaid in black. See Figure S5 for results of fit and the top concentration plotted, with 2-fold dilutions thereafter. Dual domain = PHD+JmjC.   Residues conserved within all three PHD-fingers are highlighted in dark blue, those conserved within at least two of the three are in light blue, and present in only one of the three in white. Alignment and graphic generated from JalView. Constructs those used in binding assays. Figure S4. BLI for KDM7 proteins with H3  Figure S6. ITC titration curves for OC9 against KDM7 protein domains.
Top panels are raw heat of injection, bottom panels are normalised data fitted to 1:1 binding model. OC9 with KDM7A/B/C PHD-finger and dual domain proteins. For results of fit and conditions see Figure  S10. Top panels are raw heat of injection, bottom panels are normalised data fitted to 1:1 binding model. H3  K4me3 with KDM7A/B/C PHD-finger and dual domain proteins. For results of fit and conditions see Figure S10. Top panels are raw heat of injection, bottom panels are normalised data fitted to 1:1 binding model. OC9, H3 1-21 K4me3 and H3 1-21 with His-Trx tagged KDM7B PHD-finger. His-Trx = His 6 +thioredoxin. For results of fit and conditions see Figure S10. The His-Trx tag made very little difference to binding of OC9 and H31-21K4me3. Top panel is raw heat of injection, bottom panel is normalised data fitted to 1:1 binding model. OC3 with KDM7B dual domain. For results of fit and conditions see Figure S10.     Residue level view generated from overlapping peptide fragments of HDX-MS results from incubation of KDM7B PHD-finger with compounds. Mass changes are relative to a DMSO control.
BLI traces for binding to KDM7B PHD-finger with C-terminal his-tag immobilised on Ni-NTA biosensors.
Response trace for each peptide concentration is represented by a different colour, with fitted curve overlaid in black. See Figure S19 for results of fit and the top concentration plotted, with 2-fold dilutions thereafter.     Annotated with the assigned backbone amide NH residues and the W29 indole. Red circles indicate side-chain associated NH. OC9:PHD is 1:1. The TALOS+ software package was used to predict the likelihood of involvement in either alpha helix or beta sheet secondary structure elements for each residue in solution, based on their chemical shifts and sequence. A score of greater than 0.5 was considered probable. The data are colour mapped to the PDB structure 3KV4 for comparison. The cage residues' side chains are shown.

H-NMR spectra of OC9 variant titrations against KDM7B PHD-finger
The shift region containing the OC9 V10 γ-methyl groups centred around -0.5 ppm (marked in red) is shown. No signals in this region were observed for OC9 V10A. The titrations of OC9 V10A/F11A showed progressive peak broadening and weakening of multiple residues within 1 H-15 N HSQC spectra up to 1:1 protein:peptide ratios, indicating an intermediate exchange regime consistent with their lower affinities, with the least affected protein signals corresponding to residues primarily located in the α1 and α3 regions. Annotated with the assigned backbone amide NH residues and the W29 indole, by nearest neighbour assignment to the OC9 complex. Red circles indicate side-chain associated NH. The change in CSP relative to the OC9 complex is plotted and coloured by standard deviation, mapped onto PDB: 3KV4, indicating the OC9 L3 residue may bind adjacent to the cage within a pocket encompassed by α1, α2, β4 and β5. OC9 L3A:PHD is 1.6:1. Annotated with the assigned backbone amide NH residues, by nearest neighbour assignment to the OC9 complex and apo protein.
Red circles indicate side-chain associated NH. OC9 V10A:PHD is 3:1. Annotated with the assigned backbone amide NH residues, by nearest neighbour assignment to the OC9 complex and apo protein. Red circles indicate side-chain associated NH. OC9 V10DV:PHD is 5:1.  NMR correlations identifying the OC9 V10 γ-methyl groups to be shielded and centred around -0.5 ppm (TOCSY) with OC9 in complex with KDM7B PHD-finger protein, with NOE contacts to least the I21, E22 and W29 protein residues (NOESY). ex = signal from free exchange of OC9 peptide by ROESY.   Top: screening against the DIDO1 PHD-finger. Bottom: screening against other reader domains.
Response trace for each peptide concentration is represented by a different colour, with the fitted curves where they could be fitted overlaid in black. See Figure S5 and S19 for results of fit and the top concentration plotted, with 2-fold dilutions thereafter. KDM7B residues that experience significant CSP upon OC9-binding by NMR are marked (black triangle), and amongst these the amino acids which differ in identity in DIDO1 are marked (red line). Alignment performed using Clustal Omega, with amino acid colouring in JalView (dark blue is present in all 4 sequences, light blue is present in 2 or 3, white is present in 1 only    Canonical KDM PHD finger sequences from Uniprot were multiply aligned using Clustal Omega. Highly conserved residues are marked in blue using JalView. Little similarity is seen between sub-families. Highly conserved Cys and His residues are zinc binding.   Confirmation of the dose dependent competition of OC9 with OC9-Bt recovery of endogenous KDM7 through pull-down from SUP-T1 nuclear lysates. The same samples were used for proteomic analysis. The full blot images are given underneath.

Supplementary Methods mRNA Display Selection and Next-Generation Sequencing Analysis
Library preparation: DNA templates for transcription were generated by PCR as described 2 with minor modifications. mRNA was transcribed from DNA using T7 RNA polymerase, purified by phenol/chloroform extraction followed by ethanol precipitation. In the first selection round a 1:1:1 molar ratio of NNK10/NNK11/NNK12 (see below for example library sequence) was used. An exemplar round of selection is described: The mRNA library (10 μM) was ligated 3 in T4 RNA ligase buffer, with puromycin-linker (15 μM), RNase inhibitor (0.2 U), T4 RNA ligase (18.8 U) and MilliQ H2O was added to a final volume of 7.5 μL. The reaction was assembled in a PCR tube at 20 °C for 1 hour, then 70 °C for 10 minutes. All translation components were kept on ice during preparation and assembled in a lowbinding tube. The puromycin-oligo ligated RNA encoding for the peptide library (2.5 µM) was translated in vitro using NEB PURE express kit (ΔRF1, ΔMet) supplemented with 500 pmol of N-chloroacetyl-Ltyrosine-tRNA fMet CAU 4 in 5.9 µL for 1 hr at 37 °C, then 10 minutes at 60 °C. Reverse transcription (10 µL) was carried out at 42 °C for 1 hour using M-MLV RTase (50 U), Primer P2 (1.78 µM), M-MLV buffer (0.53X), dNTPs (0.27 mM) and input (translation reaction). On completion, the RT mixture was diluted with 10 μL of 2X selection buffer (selection buffer: PBS, 0.01% BSA, 0.05% Tween20, 2 mM betamercaptoethanol (BME)). 0.25 μL from the diluted RT mixture was taken and further diluted to 1000 μL in H2O (Sample 1). All subsequent manipulations were carried out at 4 °C using pre-chilled buffers.
Binding cyclic peptide selection: To remove non-specific binders, the mRNA/cDNA-cyclic peptide library was initially applied to streptavidin coated magnetic beads (80 µg, pre-washed with selection buffer and half of beads loaded with Biotin) and incubated gently with agitation (30 min, 4 °C). The beads were pelleted using a magnet, and the supernatant was transferred to a tube with fresh beads and the process was conducted three times in total. 0.25 μL from the screened library mixture was taken and further diluted to 1000 μL in H2O (Sample 2). Target protein was loaded to beads by taking 10 μL of KDM7BPHD+JmjC-biotin (36 μM) (thawed on ice) and diluting it with an equal volume of selection buffer (with added 10 μM ZnCl2), then 5 μL of the diluted solution was further diluted with selection buffer (no added ZnCl2) to a final protein concentration of 1 μM. 100 μg of pre-washed streptavidin beads were re-suspended in 20 μL of 1 μM KDM7BPHD+JmjC-biotin at 4 °C for 30 min with rotation, then beads were pelleted and washed five times with 120 µL of selection buffer. Bead loading was previously determined to be ~2 pmol protein per μL bead slurry. The protein loaded beads were re-suspended in the ~20 µL of library solution and incubated at 4 °C with rotation for 30 min, then pelleted and supernatant removed before re-suspension of the beads in 100 μL of selection buffer and transfer to a fresh low-binding tube. The re-suspension and transfer processes was conducted three times in total. The beads were finally re-suspended in 100 μL of PCR mix (NH4 buffer, 2.5 mM MgCl2, 0.25 mM dNTPs, 0.5 μM P1, 0.5 μM P2), then heated at 95 °C for 5 mins and the supernatant containing the liberated DNA immediately recovered to a fresh tube. Recovered DNA was amplified by PCR and purified by phenol/chloroform extraction and ethanol precipitation, and used as the template DNA for the next round of selection. The percentage recovery was quantitated by qPCR (input sample 1&2 vs recovery) for each round of selection. The recovery at the end of round 5 suggested sufficient enrichment and no further rounds of selection were undertaken.
Next Generation Sequencing: DNA pools from selection rounds were sequenced by Illumina sequencing as previously described 5 .

Protein NMR
Nuclear Magnetic Resonance (NMR) spectra were recorded using a Bruker AVIII 700 MHz NMR spectrometer equipped with a 5-mm inverse TCI cryoprobe using 3 mm MATCH NMR tubes (Cortectnet, or Hilgenberg, #2001724). Buffering was with 25 mM phosphate buffer, 100 mM NaCl, in 90% H2O, 10% (v/v) D2O, pH 6.3 and sample temperatures were regulated at 308 K, except where otherwise specified (e.g. for variable temperature studies). Compounds were added from a concentrated stock in DMSO-d6, such that final DMSO ≤ 5% (v/v), with ratios noted in respective figure legends. Data were processed with Bruker Topspin 4.0.8 software.
Backbone Assignments: Variable Temperature Studies: 3-(Trimethylsilyl)propionic-2,2,3,3-d4 acid sodium salt (TSP) was used as a reference point for 1 H chemical shifts and is assumed to be temperature independent. The KDM7B PHD sample used 90 μM 15 N-labelled protein with 10 μM TSP. OC9 was added at 450 µM. For each, eleven 1 H-15 N HSQC spectra were collected from 288 K to 308 K at 2 K increments. The Bruker pulse sequence hsqcetfpf3gpsi was used with an FID size of 2048 points ( 1 H) and 256 points ( 15 N), a spectral width of 16 ppm ( 1 H) and 80 ppm ( 15 N), and 8 scans.

Chemical shift perturbations:
CSP  values between free and complexed protein were calculated for amide proton and nitrogen shifts according to: Where H and N are differences in chemical shifts between the free and complexed forms for 1 H and 15 N respectively.

2D NMR for protein and peptide sidechain studies:
Protein samples (240 µM final concentration) were in phosphate buffer and OC9 (240 µM final concentration) was added. 1

2D NMR for bound OC9 assignments:
1H Chemical shifts assignments for bound OC9 were obtained from 2D TOCSY and NOESY spectra collected using double F1 and F2 13 C and 15 N filtered sequences (A.L. Breeze, Prog. NMR Spectrosc. 2000, 36, 323-372) (Bruker pulse programs: dipsi2gpphwgxf and noesygpphwgxf.2). In these, protons bound to 13 C or 15 N nuclei are filtered (removed) from the resulting spectra, such that only resonances derived from components that are lacking these isotope labels remain visible. These 2D experiments were therefore collected using uniformly 13 C and 15 N labelled KDM7B PHD-finger (240 μM) complexed with ~0.9 equiv. unlabelled OC9 (200 μM) in 25 mM phosphate, 10 mM NaCl at pH 6.3 with 10 % D2O at 308 K, prepared in 180 μL (3 mm tube). Filtered TOCSY and NOESY spectra were collected with data sizes of 2048 x 256 points and spectral widths of 14 ppm for both 1 H dimensions centred at 4.7 ppm, and 13 C and 15 N offsets of 43 ppm (O2P) and 117 ppm (O3P) respectively. Mixing times were 80 ms and 150 ms for TOCSY and NOESY respectively, with 32 (TOCSY) or 64 (NOESY) transients per increment. Data were collected on a Bruker AVIII 700 equipped with a 5 mm TCI cryoprobe.

General Procedures
All histone H31-21 peptides were purchased from GL Biochem (Shanghai) (>95% purity), received as solid and dissolved to stock solutions into H2O or DMSOd6, with mass identity verified by MALDI-TOF-MS. Biotinylated histone peptides were purchased from AnaSpec, with a GGK-biotin unit at the Cterminus of the sequence.
Rink amide resin loading: Rink Amide MBHA resin (1.0 equiv) was pre-swelled in DMF or DMF/DCM for at least 15 min, then washed with DMF. Fmoc-deprotection of resin was performed using 20 % (v/v) piperidine in DMF (3 mL) at 50 °C for 2-3 minutes, and repeated once more. The resin was washed with DMF (3 x 3 mL).
Iterative Fmoc amino acid coupling and deprotection: The activating agent diisopropylcarbodiimide (DIC, Fluorochem) and activating base Oxyma (Novabiochem) (sometimes with N,Ndiisopropylethylamine additive (DIPEA, Sigma Aldrich)) were used to couple Fmoc-amino acids (AAs) in 1:1:1 ratio in DMF, with a final concentration of amino acids in the reaction typically ~83 mM. Synthesis was typically performed with 5-fold excess of reagents over the molar amount of resin, using a single coupling cycle at 90 °C for 2 minutes as standard, but 50 °C and 10 minutes for cysteine or histidine, and a 75 °C double coupling for arginine. The resin was washed with DMF (3 x 3 mL) and Fmoc deprotected using 20 % (v/v) piperidine in DMF, typically at 50 °C for 2 minutes, performed twice for each coupling. The resin was further washed with DMF (3 x 3 mL) prior to the next coupling step.
Chloroacetylation of N-terminus of peptides: For peptides synthesised on Liberty Blue, peptide-on-resin was manually suspended in ~5 mL of DMF and treated with 10 equivalents of either chloroacetic anhydride (Sigma Aldrich) or acetic anhydride for 2 hours at 20 °C with manual agitation every 30 minutes and a further 5 equivalents added for 1 hour. The resin was subsequently washed in a phase separator (TELOS, 25 mL) under reduced pressure with DMF (10 mL), DCM (3x 20 mL) and dried to a free-flowing texture under reduced pressure. For peptides synthesised on Gyros, chloroacetic anhydride treatment with 10-fold excess in DMF was performed on-machine for 30 minutes at 60 °C.
Peptide cyclisation: Thioether cyclisation was performed either in: i) DMSO with triethylamine added to pH ~8/9 and incubation at 37°C for 60 minutes, or ii) in MeCN/H2O with 0.1 M NaOH added dropwise to pH ~8/9 and incubation at 60°C for 60 minutes, or iii) in 20 mM ammonium bicarbonate in MeCN/H2O at pH 8 for 60 minutes at 40 °C with sonicator agitation. MALDI-TOF-MS was used to confirm cyclisation. Reaction mixtures were then either diluted with MeCN/H2O (0.1 % TFA) for purification by HPLC, or in the case of bicarbonate solution cyclisations they were freeze dried before re-dissolution in appropriate solvent. All cyclisation conditions gave the same product, with the same activity (as confirmed by MS and NMR, and different batches of OC9 yielding same binding response (KD) against KDM7B PHD).

General Peptide Purification and Characterisation
Peptides were purified using a JASCO or Agilent 1260 HPLC system with a Phenomenex Gemini-NX 5 μm C18 110 Å 250x30 mm column, UV Detection at 220 nm. Typically, a solvent system of A = H2O, 0.1 % (v/v) TFA and B = MeCN, 0.1 % (v/v) TFA, was used with increasing B gradient elution, customised for each peptide, over 45-60 minutes. Product containing fractions were identified by MALDI-TOF-MS, with those corresponding to a single peak combined and lyophilized to a white powder, and identity confirmed by high resolution ESI-MS. Analytical HPLC of purified samples was run on an Agilent 1220 system with a Phenomenex BioZen Peptide XB-C18 2.6 μm, 150 x 4.6 mm column or Phenomenex BioZen Peptide PS-C18 3 μm, 150 x 4.6 mm column, using a 2-98 % gradient of MeCN (analytical grade) in MilliQ water, with 0.1 % TFA (analytical grade).
Note: the final amino acid coupling had additional end-steps (final deprotection and washes): Step OC9 was then precipitated off-machine by manual addition of diethyl ether (Sigma Aldrich) (40 mL, total volume ~45 mL in 50 mL falcon collection tube). The solution was mixed and white precipitate was observed. The suspension was pelleted by centrifugation, the supernatant removed to waste and OC9 crude pellet re-suspended in 40 mL of diethyl ether by vigorous vortexing and centrifuged. The pellet washing process was performed five times. The final pellet was air-dried within a fumehood for at least 2 hours, then the pellet was re-dissolved in 10 mL The OC9 containing fractions were identified by MALDI-TOF-MS (eluting at 17.2 min), and those of suitable purity from each injection were combined and lyophilised to dryness. Resulting purified pellets were combined by re-dissolution into a total of 500 μL of 1:1 MeCN:H2O, and transferred to a preweighed 1.5 mL Eppendorf, then re-frozen in liquid nitrogen and lyophilised to dryness. This process gave 17.65 mg of a white powder solid (20 % yield, >99 % purity by analytical HPLC).

Peptide High-Resolution Mass Spectrometry Method
Samples were prepared by dilution in H2O/MeCN to < 0.1 mg/mL and run on a Waters Acquity For Ni-NTA, the biosensors were stripped between compounds by immersion in glycine (0.5 M, pH 1) and regenerated with 10 mM NiCl2 before re-equilibration in buffer and protein re-loading. Data were processed using ForteBio DataAnalysis v9 to apply baseline alignment, inter-step correction if necessary and Savitzky-Golay filtering, then processed data were exported and fitted with an 'Association then Dissociation' 1:1 binding model using GraphPad Prism v8/9 to determine KD. For equilibrium steady-state analysis, the equilibrium response (Robs) was fitted using 'One site -Total and non-specific binding' model to determine KD. For SA biosensors, non-specific data were obtained using non-loaded biosensors with compounds.

Isothermal Titration Calorimetry (ITC)
Malvern where peptides were filtered: minimum intensity of 5000, minimum of 0.2 products per amino acid, a maximum MH + error of 5 ppm, and found in all of the undeuterated datasets. The automatic peptide assignment in DynamX was performed using the standard parameters, but charge state assignment and retention times were verified manually for all peptides. Data were not corrected for back-exchange, so are relative rather than absolute deuterium uptake values. The first two residues of each peptide are excluded from the analyses due to rapid back-exchange.
A Student's t-test was performed on the peptide-level data using the pooled standard deviation of all peptides across all samples and time-points (Pooled SD(DualDomain) = 0.08 Da; Pooled SD(PHD) = 0.05 Da). Peptides with a p-value <0.01 were considered significant and taken forward in the analyses. A custom MATLAB script was used to combine data from redundant peptides to generate a residuelevel view of the differences in deuterium uptake for each time point. H31-21K4me3-GGK-biotin was purchased from AnaSpec. The linear range of signal was determined for each His-tagged protein/biotinylated peptide combination to select suitable assay conditions. Concentrations are reported as those in the final 20 µL assay volume. All KDM7 and DIDO proteins were tested with OC9 and H31-21K4me3 twice in technical triplicate, and with OC9 derivatives at least twice in technical duplicate, whilst other proteins were tested at least once in technical duplicate. Compound dilution series were typically prepared in aqueous 0.4 % DMSO (0.1 % DMSO final) and all other components in buffer. Compound (5 µL) was incubated with protein (5 µL) for 5 minutes at 20 °C, followed by addition of 5 µL of biotinylated peptide for 20 minutes and then addition of 5 µL of pre-mixed AlphaScreen beads (Nickel Chelate kit, PerkinElmer, 6760619M, 250-fold final dilution), then incubated at 25 °C for 90 minutes. The plate was sealed with opaque foil and pulse centrifuged briefly after each addition. The plate was read with a BMG Pherastar FS/FSX with AlphaScreen detection module (680, 570). Data were analysed using GraphPad Prism v9 and raw counts were fitted with the model: 'log(inhibitor) vs. response, variable slope'. Graphed data are also presented as '% Displacement', where each compound set was normalised to respective minima and maxima where reasonably fitted, or DMSO control where it could not be, then '% Displacement' calculated as: 100normalised % and plotted with the model: log(inhibitor) vs. normalized response, variable slope.  Table S5. Conditions used for each AlphaScreen displacement assay. Trx = thioredoxin. GST = glutathione S-transferase.

AlphaScreen Demethylation Assay
KDM AlphaScreen activity assays were carried out similarly to those described by Hopkinson et al 8 and Rose et al 9 with the following modifications. Compound titration series were ECHO dispensed at least in duplicate (100 nL in DMSO) to a ProxiPlate (384-Plus, White, PerkinElmer) for a 10 µL total volume assay (1 % DMSO final) in assay buffer. The AlphaScreen beads (IgG-Protein A Detection Kit, Perkin Elmer, 6760617) were incubated with primary antibody for at least 30 minutes at 20 °C prior to use. Compounds were pre-incubated with 7.5 µL of enzyme for 10 minutes at 20 °C, then the demethylation initiated by addition of 2.5 µL substrate mix (2OG, ammonium iron(II) sulfate hexahydrate (Fe II ), sodium ascorbate (Asc) and biotinylated H3 peptide substrate in buffer). Concentrations are reported as those present in the 10 µL volume. All components of the substrate mix were prepared in assay buffer, except for Fe(II) which was initially diluted in 20 mM HCl, then MilliQ prior to mixing. The reaction was incubated at 20 °C under a foil seal, then quenched at an endpoint within the linear range of signal by addition of 5 µL of quench solution (30 mM EDTA, 1.5 M NaCl), followed by addition of 5 µL of IgG-antibody bead mix. The plate was re-sealed and further incubated at 20 °C for 1 hour, then read with a BMG Pherastar FS with AlphaScreen detection module (680, 570). Data were calculated as a % of maximum signal and analysed using GraphPad Prism v9 with the model: 'log(inhibitor) vs. normalized response, variable slope'.  Table S7. Summary of assay conditions used in KDM7 MALDI-TOF-MS dose-response demethylation assays.

KDM Selectivity Screening
Compound screening at single concentrations was performed in a 96-well PCR plate (4ti-0740, 4titude) on a ThermoMixer C (Eppendorf). Compounds were tested once in either technical duplicate or triplicate and typically diluted in aqueous 5 % DMSO (1 % final). Protein in assay buffer (4 µL) was added to 2 µL of compound and pre-incubated at 20 °C for 10 minutes. The reaction was initiated by addition of 4 µL substrate mix (2OG, Fe II , Asc, and histone peptide substrate (GL Biochem)). All components of the substrate mix were prepared in assay buffer, except for Fe II which was initially diluted in 20 mM HCl, then MilliQ prior to mixing. All wells were quenched (4 µL of 6 % (v/v) formic acid) within a time-course assay determined linear range of activity. The assay mix was analysed directly by MALDI-TOF-MS. The % demethylation was determined by analysis (Bruker flexAnalysis v3.4) of the monoisotopic peak intensities corresponding to affected methylation states. Data were further analysed with GraphPad Prism v9, using an ordinary one-way ANOVA test, assuming a normal distribution and common variance, followed by Dunnett's multiple comparison test (comparing the mean value for each compound result with the mean value of the DMSO-control). The significance thresholds of the resulting multiplicity adjusted P-values were used to label the data (noted in the figure legends).
grouped, and the data filtered such that 2 valid LFQ values were required for the condition without any OC9 competition, and at least one valid LFQ value in each of the remaining experimental groups. Missing values were imputed using default settings, and the data distribution visually inspected to ensure that a normal distribution was maintained. Statistically significant competition was determined through the application of P2 tests, using a permutation-based FDR of 0.05 and volcano plot visualisation.
Data deposition: The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository 11 with the data set identifier PXD027151.