Experimental model and subject details
Saccharomyces cerevisiae cells (Euroscarf, Supplementary Data 16) were grown at 30 °C in YPD media to early log phase from a single colony picked from a fresh YPD plate. Cells were harvested by centrifugation and carefully washed three times with ice-cold lysis buffer (100 mM HEPES pH 7.5, 150 mM KCl, 1 mM MgCl2). Cell pellets were resuspended in lysis buffer, and cell suspensions were extruded from a gauge needle to produce drops that were immediately flash frozen in liquid nitrogen.
Botrytis cinerea (clone BO47), both wild-type and CK1 His-tagged, cells (Bayer Crop Science, Supplementary Data 16) were cultured in potato dextrose agar (39 g/L, Oxoid #CM0139) at 21 °C for 10 days. After 10 days growth the cells were suspended in 10 ml of GYPm liquid media (14.6 g/l d(+)-glucose monohydrate (VWR #24370.320), yeast extract (Merck #1.03753.0500), mycological peptone (Oxoid #LP0040)) and filtered (100 µm, Corning cell strainer) to harvest spores (final solution of 5 × 106 spores/ml). This liquid culture was incubated for 24 h at 110 rpm at 21 °C. Cell mycelium was pelleted by centrifugation (5 min, 16,000 g, 21 °C), media was removed and the pellet snap frozen in liquid nitrogen.
Whole-proteome preparation for MS analysis
Saccharomyces cerevisiae: Liquid-nitrogen frozen beads of S.cerevisiae cell suspensions in lysis buffer (100 mM HEPES pH 7.5, 150 mM KCl, 1 mM MgCl2) were mechanically ground in cryogenic conditions with a Freezer Mill (SPEX SamplePrep 6875). Cell debris was removed by centrifugation (10 min, 20,000g, 4 °C). The sample preparation procedure was performed at 4 °C.
HeLa and Botrytis cinerea cells (lysate): All biological samples were kept on ice through sample preparation. HeLa cell pellets (Ipracell, CC-01-10-10) (5 × 107 cells) and Botrytis cinerea mycelium (3 × 107 cells) were resuspended in 800 µl LiP buffer (100 mM HEPES pH 7.5, 150 mM KCl, 1 mM MgCl2) and lysed by passing completely through a BD Precision glide syringe needle (27 G) ten times, followed by 20 min incubation on ice. Lysate was cleared by centrifugation (16,000g at 4 °C) for 4 min. Supernatant was retained in a new Eppendorf tube and the pellet was resuspended in 400 µl of LiP buffer for repeated lysis under the aforementioned conditions, including incubation and centrifugation. After centrifugation, supernatants were combined and protein amount was determined using a Pierce BCA Protein Assay Kit (cat #23225) according to manufacturer’s instructions.
HeLa cells (in vivo culture): HeLa cells (Sigma-Aldrich, 93021013-1VL) were cultured in low-glucose Dulbecco’s Modified Eagle’s Medium (DMEM) (Sigma-Aldrich, #D6046) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Cells were passaged prior to confluency by detachment with 0.25% trypsin and subculture at a ratio of 1:8.
Wheat germ agglutinin beads membrane protein enrichment
HeLa cells were lysed as above (lysate) however 0.3% n-dodecyl-β-d-maltoside (DDM) was added prior to lysis. Four mg of lysate was rotated with 320 µl of wheat germ agglutinin agarose beads (Reactolab SA, AL-1023-2), pre-washed twice with LiP buffer, for 4 h at 4 °C. Beads were then centrifuged at 300g for 30 s to pellet and the supernatant was aspirated. Beads were then washed six times (300g, 30 s between washes) and finally eluted in 400 µl of 0.5 M N-acetyl glucosamine (in LiP buffer without detergent) at 4 °C for 30 min. Beads were pelleted at 300g for 30 s and eluate was collected and protein was quantified by BCA test. LiP-Quant assay was performed with membrane enriched eluate in the same manner as described for standard HeLa lysate.
Lysate and cell treatment for LiP-Quant in native conditions
Saccharomyces cerevisiae: Cell lysates from at least three independent biological replicates were aliquoted in equivalent volumes containing 100 µg of proteome sample and incubated for 10 min at 25 °C with the drug of interest. Proteinase K from Tritirachium album (Sigma Aldrich) was added simultaneously to all the proteome-metabolite samples with the aid of a multichannel pipette, at a proteinase K:substrate mass ratio of 1:100, and incubated at 25 °C for 4 min. Digestion reactions were stopped by heating samples for 5 min at 98 °C in a thermocycler followed by addition of sodium deoxycholate (Sigma Aldrich) to a final concentration of 5%. Samples were then heated again at 98 °C for 3 min in a thermocycler. These samples were then subjected to complete digestion in denaturing conditions as described below.
HeLa and Botrytis cinerea cells: 100 µg of protein lysate was aliquoted from a lysate pool for each of four independent replicates (n = 4 for all experiments) and incubated at room temperature (RT) with the compound of interest for 10 min. An 8-concentration dose–response was used for each experiment (seven compound dilutions plus a vehicle control) plus a single concentration rapamycin treatment as a positive control. For the rapamycin dose–response an additional concentration was used in place of the positive control. Proteinase K (1:100 ratio of enzyme to protein) was added and samples were incubated for a further 4 min. Samples were transferred to a heat block at 98 °C for 1 min, at which time proteinase K activity was quenched with an equal volume of 10% deoxycholate (to a final concentration of 5%) and incubated for a further 15 min at 98 °C.
HeLa cells: Near confluent 6-well plates (9.6 cm2 per well) were washed twice with DMEM minus FBS, followed by incubation with rapamycin (2 µM) or DMSO (0.2%) (n = 3 for each of biologically independent wells of cells) in DMEM minus FBS at 37 °C for 15 min. At the end of compound incubation cells were washed twice with ice-cold LiP buffer and then scraped into an Eppendorf tube in 100 µl of LiP buffer, which was immediately snap frozen. Cells were thawed at 4 °C and snap frozen again for a total of three freeze-thaw cycles. After the final thaw, proteinase K (1 µg per well) was added and samples were incubated at room temperature for 4 min. Samples were transferred to a heat block at 98 °C for 1 min, at which time proteinase K activity was quenched with an equal volume of 10% deoxycholate (to a final concentration of 5%) and incubated for a further 15 min at 98 °C.
Proteome preparation in denaturing conditions
Samples were removed from heat and reduced for 1 h at 37 °C with 5 mM tris(2-carboxyethyl)phosphine hydrochloride followed by a 30 min incubation at RT in the dark with 20 mM iodoacetamide. Subsequently, samples were diluted in two volumes of 0.1 M ammonium bicarbonate (final pH of 8) and digested for 2 h at 37 °C with lysyl endopeptidase (1:100 enzyme: substrate ratio). Samples were further digested for 16 h at 37 °C with trypsin (1:100 enzyme: substrate ratio). Deoxycholate was precipitated by addition of formic acid to a final concentration of 1.5% and centrifuged at 16,000g for 10 min. After transferring the supernatant to a new Eppendorf tube an equal volume of formic acid was added again and the centrifugation repeated. Digests were desalted using C18 MacroSpin columns (The Nest Group), or Sep-Pak C18 cartridges or into 96-well elution plates (Waters), following the manufacturer’s instructions and after drying resuspended in 1% acetonitrile (ACN) and 0.1% formic acid. The iRT kit (Biognosys AG, Schlieren, Switzerland) was added to all samples according to the manufacturer’s instructions.
High pH reversed phase fractionation
Equal amounts of peptides were taken and pooled from the final LiP reaction digests for each treatment (e.g. 7 µg from each replicate for each condition), resulting in approximately 200 µg of total digest. This digest pool was fractionated into 10-12 fractions using high pH reversed phase chromatography with a Dionex Ultimate 3000 HPLC (Thermo Fisher, Waltham, United States) and an ACQUITY UPLC CSH C18 column (1.7 µm × 150 mm) from Waters (Milford, United States). In brief, a 25% ammonium hydroxide solution was used to adjust the pH of the digest pool to 10. The lysate was run on a 30 min non-linear gradient, increasing from 1 to 40% ACN, at a flow rate of 0.3 ml per min and a micro-fraction size of 30 s. After drying the individual fractions were resuspended in 1% ACN and 0.1% formic acid and Biognosys’ iRT kit was added.
Mass spectrometric acquisition
For all samples generated from HeLa or Botrytis cinerea cells, for DIA (Data Independent Acquisition) runs, 2 µg of LiP reaction digest from each sample was analyzed using an in-house analytical column (75 µm × 50 cm). Samples were block randomized before acquisition. PicoFrit PicoTip Emitters (SELF/P Tip 10 µm) were packed with ReproSil-Pur C18-AQ 1.9 µm phase (Dr. Maisch, Ammerbuch-Entringen Germany) and connected to an Easy-nLC 1200. All experiments were run on a Q-Exactive HF mass spectrometer (Thermo Scientific) with the exception of the calyculin A dataset, which was acquired on a Q-Exactive HF-X. Peptides were separated by a 2 h segmented gradient at a flow rate of 250 nl/min with increasing solvent B (0.1% formic acid, 85% ACN) mixed into solvent A (0.1% formic acid, 1% ACN). Solvent B concentration was increased from 1% after 3 min according to the following gradient: 4% over 3 min, 5% for 3 min, 7% for 4 min, 9% for 5 min, 11% for 8 min, 16% for 19 min, 26% for 41 min, 29% for 9 min, 31% for 6 min, 33% for 5 min, 35% for 4 min, 38% for 4 min, 40% for 3 min, 44% for 3 min, 55% for 3 min and 90% in 10 s. This final concentration was held for 10 min followed by a rapid decrease to 1% over 10 s, which was then held for 5 min to finish the gradient. A full scan was acquired between 350 and 1650 m/z at a resolution of 120,000 (ACG target of 3e6 or 7 ms maximal injection time). A total of 37 DIA segments on HF were acquired at a resolution of 30,000 (ACG target of 3e6 or 47 ms maximal injection time) and 42 on the HF-X (ACG target of 3e6 or 55 ms maximal injection time). The normalized collision energy was stepped at 25.5, 27, 30. First mass was fixed at 200 m/z.
For DDA (Data Dependent Acquisition) runs from the same samples, peptides were separated by the same 2 h segmented gradient as utilized above for DIA runs with the exception that the final 1% solvent B flow was held for 4 min and 40 s (rather than 5 min). All experiments were run on a Q-Exactive HF mass spectrometer (Thermo Scientific) with the exception of the rapamycin (Q-Exactive HF-X) and FK506 datasets (Q-Exactive). A top 15 method was used across a scan range of 350 to 1650 m/z with a full MS resolution of 60,000 (ACG target of 3e6 or 20 ms injection time). Dependent MS2 scans were performed with a resolution of 15,000 (ACG target of 2e6 or 25 ms injection time) with an isolation window of 1.6 m/z and a fixed first mass of 120 m/z.
Peptide samples generated from Saccharomyces cerevisiae were analyzed on an Orbitrap Q Exactive Plus mass spectrometer (Thermo Fisher Scientific) equipped with a nano-electrospray ion source and a nano-flow LC system (Easy-nLC 1000, Thermo Fisher Scientific). MS data acquisition in DDA and DIA modes was essentially carried out as in Piazza et. al. 2018.
Mass spectrometric data analysis
DIA spectra were analyzed with Spectronaut X (Biognosys AG)40 using the default settings. In brief, retention time prediction type was set to dynamic iRT (adapted variable iRT extraction width for varying iRT precision during the gradient) and correction factor for window 1. Mass calibration was set to local mass calibration. The false discovery rate (FDR) was estimated with the mProphet approach41 and set to 1% at both the peptide precursor and protein level. Statistical comparisons were performed on the modified peptide level using fragment ions as quantitative input. The DDA spectra were analyzed with the SpectroMine (Biognosys AG) software using the default settings with the following alterations. Digestion enzyme specificity was set to Trypsin/P and semi-specific. Search criteria included carbamidomethylation of cysteine as a fixed modification, as well as oxidation of methionine and acetylation (protein N-terminus) as variable modifications. Up to 2 missed cleavages were allowed. The initial mass tolerance for the precursor was 4.5 ppm and for the fragment ions was 20 ppm. The DDA files were searched against the human UniProt fasta database (updated 2018-07-01) and the Biognosys’ iRT peptides fasta database (uploaded to the public repository). The libraries were generated using the library generation functionality of SpectroMine with default settings.
Machine learning-based training of the LiP-Quant classifier
All HeLa datasets were first analyzed for differentially regulated peptides between the highest drug concentration and vehicle using Spectronaut’s statistical testing (one sample two-sided t-test with Storey method correction) performed on the modified peptide sequence level using fragment ions as the smallest quantitative units. This candidate peptide list was filtered based upon q-value < 0.01 and an absolute log2 fold-change > 0.58. Each peptide in this filtered list was then subjected to dose–response correlation testing (using the drc package (https://www.r-project.org)) on all peptides (modified sequence with fragments ions as quantitative units) at every drug concentration to establish a sigmoidal correlation coefficient.
As the ground truth (target proteins) was known for the drugs tested in HeLa lysates, each protein identified in each dataset was annotated as either a known target or non-target and from this a contaminant database, or LiP-protein frequency library (PFL), was built. To do so, the same statistically filtered list of differentially regulated peptides as above was used and proteins that were present but not specific for the drug being tested were quantified and assigned a PFL (contamination) score. For example, a protein that showed differential regulation in 9 of 11 ground truth experiments (several experiments were performed more than once) was assigned a contamination score of 9/11 or 81.8% (Supplementary Data 14), proteins that never appeared as contaminants in any experiment were not included in the PFL-library. This library enabled the quantitative down-weighting of proteins that were frequently present in LiP experiments but not specific for the drug being tested. We observed high correlation between proteins identified as likely contaminants in the PFL of our LiP-Quant experiments (Supplementary Data 14) and those previously identified as common contaminants in affinity purification mass spectrometry (such as chaperone and structural proteins)42.
To establish the criteria that contribute to the identification of drug targets, we split our dose–response experimental data (filtered based on q-value and log2 fold-change and PFL annotated as mentioned above) into two independent datasets to train our classifier; training set A included the drugs calyculin A, rapamycin and staurosporine and training set B included FK506, selumetinib and fostriecin (Supplementary Fig. 4A). For each training set the data was combined and we used linear discriminant analysis (LDA) to build classifiers based upon all potential unique peptide/protein features (e.g. dose–response correlation, PFL frequency, protein coverage, etc). For each training set, known drug targets were selected as a positive training set, resulting in 95 modified sequences for training set A and 33 for training set B. We also randomly sampled 400 background modified sequences as a negative training set from each training set. The features were calculated and stabilized to a defined range between 0 and 1. The LDA-based machine learning was performed five times for each training dataset with resampling of the negative training set each time. The identified criteria were consistent across all LDA analyses (Supplementary Fig. 1B) and the contribution weights for each of the features from the five LDA analyses was averaged. The relative contributions of each parameter to the LiP-Quant score was very stable across the training sets (Supplementary Fig. 4B). We termed the linear classifier the LiP-Quant Score in this study. The weights were adjusted such that the combined linear classifier could reach a maximum value of 6. These weightings were incorporated into the analysis pipeline (see below) and verified independently on the other positive control datasets (i.e. training set A was verified on the datasets comprising training set B and vice versa) (Supplementary Fig. 4A). LiP rankings using both training set analysis parameters were similar across all datasets (Supplementary Fig. 4C).
Using this approach, we established four classifiers that contribute to positive drug target identification (Supplementary Fig. 1B): (I) correlation of fit with a dose–response binding model, (II) the presence of the identified protein in the LiP-protein frequency library, (III) the number of peptides from an identified protein showing regulation that are in the top ten percent of all peptides ranked by q-value in the Spectronaut filtered statistical test (see above) and (IV) the statistical significance (q-value) of the relative peptide abundances between drug and vehicle-treated samples. As training set A contained a larger positive training set (i.e. there were more known drug target peptides identified) the weightings calculated for this training set were used for all subsequent analyses.
Automated peptide/protein ranking of LiP-Quant experiments
Using the criteria and weightings established from our training datasets we wrote in-house scripts in R to calculate in an unbiased manner the individual peptide sub-scores for each LiP-Quant experiment. As these experiments contained on average over 100,000 peptides, peptides were first filtered based upon differential abundance from the Spectronaut statistical testing table (one sample two-sided t-test with Storey method correction, q-value < 0.01 and an absolute log2 fold-change > 0.46) using statistical comparisons against vehicle control for a range of drug concentrations (IC50 through 1000-fold the IC50, or the range closest to this). Each peptide in this narrowed down putative candidate list was then subjected to full LiP-Quant analysis using the four weighted criteria (Supplementary Fig. 1B) described above and a final LiP-Quant score for each peptide was calculated.
This final analysis pipeline enabled the selection and ranking of the most relevant peptides and proteins per experiment. The combined LiP-Quant score enables direct comparison of LiP peptides with each other and allows more robust discrimination of genuine targets from random hits. Ranking on the protein level was performed using the best LiP-score per protein, only. All half maximal effective/inhibitory concentrations (EC50/IC50) were calculated using the drc package (https://www.r-project.org). The necessary output files from Spectronaut are outlined in the docstring at the start of the R script.
Criteria used for establishing a LiP threshold score
Aggregating results from five positive control experiments (rapamycin, calyculin A, selumentinib, FK506 and fostreicin) conducted in HeLa lysate and analyzed with our LiP-Quant pipeline, we found that LiP scores show a bimodal distribution. Staurosporine was excluded from the threshold calculation as it shows a level of promiscuity (binding potentially hundreds of kinases) that is rare among drugs, making it difficult to ascertain if low scoring peptides are genuine targets that were not detected or kinases that are not bound by the drug. As this difficulty in interpreting non-target peptides could bias the threshold calculation the dataset was excluded. Peptides from known target proteins show a clear enrichment in the high-scoring peak of the distribution (LiP-Quant score > 1.5), whereas all other peptides are enriched in the low-scoring peak of the distribution with a median of approximately 0.8 (Fig. 1b). We defined a threshold score of 1.5 by taking the median LiP-Quant score from the aforementioned experiments, plus three standard deviations, to ensure minimal (<1%) non-target peptide presence (Supplementary Data 3). Although the approach ensures a strong enrichment for genuine targets, it should be noted that some peptides from these targets are expected below a LiP-Quant score of 1.5 as both LiP-Quant and non-LiP-Quant peptides can be expected from genuine target proteins.
Guidelines for interpreting LiP-Quant results
The purpose of the LiP-Quant score is to provide a candidate list of protein (and peptide) targets ranked by their likelihood of being a genuine drug interactor. The LiP-Quant scoring system covers the range of 0 to 6, with a peptide scoring 6 having the maximum probability of being a true target. In this way, these scores also assign ranks to proteins, by way of their peptides, enabling an unbiased prioritization of potential targets. This absolute scale is useful to make direct comparisons between experiments. For instance, the strongest peptide and protein candidate targets in LiP-Quant experiments are typically those with a score > 2.5 and rank in the top-scoring peptides of the whole proteome. Every LiP-Quant peptide has an EC50 value assigned, which corresponds to the inferred dose of drug necessary to observe half-maximum of the relative peptide intensity variation between drug and vehicle. This could also be, in principle, used as a discriminating factor under the assumption that drug targets with low EC50 values should be indicative of a strong binding interaction between protein and compound, and a compound that binds with high affinity (e.g. nM or lower) is more likely to be an effective drug in vivo. However, we normally do not exclude candidates with EC50 close to the uM range a priori, as compounds that weakly bind the target and have a phenotypic effect can be further refined during pre-clinical drug development.
The amino acid conservation in the structural model of calyculin A bound to the PP1-gamma catalytic subunit has been calculated using the ConSurf algorithm (Landau 2005). Structures of Botrytis cinerea protein–drug targets were modeled using homolog kinases with high sequence similarity for which experimental structural data was available using the Swiss-model. The Botrytis cinerea kinase models were then structurally aligned: Bc Bcin06g02870 with human kinase CSNK1A1 bound to the kinase inhibitor A86 (PDBID 6gzd) and Botyrris cinerea Bcin16g04330 with human kinase Abl in complex with imatinib and GNF-2 (PDBID 3k5v). Given the high structural and sequence similarity between Botrytis and human kinases, we used the experimental data relative to the holocomplexes between kinase inhibitors and protein kinases to assign the position of the ATP-binding (catalytic) site and allosteric sites of the Botrytis kinases.
Definition of PPV
We defined the PPV as the ratio between the number of true positive peptides and the sum of false positives (FP) and true positives (TP) identified by LiP-Quant (TP/ (TP + FP)). Unless specifically stated, these parameters refer to the top 50 LiP-Quant score ranking peptides.
Benchmarking of the LiP-Quant classifier
The LiP-Quant staurosporine experiment was selected to provide an estimation of the method FDR because of the known binding promiscuity of the compound. Published datasets of staurosporine proteome-wide binding profiles obtained with TPP6 and kinobeads23 were used to compare the predictive power of the three chemoproteomic methods. We used LiP-Quant scores for LiP-Quant, -log10(adjusted p-values) of the R2 correlation indices of the melting curves for TPP reporting both replicates analyzed in the original publication6, and the number of spectral counts (PSM) for kinobeads as ranking criteria. For the comparison of LiP-Quant, TPP and kinobeads in the venn diagram of Fig. 2b, we considered as TPP and kinobeads hits, those proteins defined as hits by the authors in the original publications. For kinobeads, these were the isolated proteins in the staurosporine-beads pull down. For TPP, Savitski et al.6 defined protein hits as those fulfilling the following criteria: (I) the minimal slope is below −0.06 in both biological replicates; (II) the minimal differential melting temperature in experiment 1 and 2 is higher than the same difference measured in the corresponding experiments with vehicle; (III) the differential melting temperature in experiment 1 and 2 have the same sign; (IV) the adjusted p-values of the R2 correlation indexes are below 0.05 in experiment 1 and below 0.1 in experiment 2 or the adjusted p-values of the R2 correlation indexes are below 0.1 in experiment 1 and below 0.05 in experiment 2.
Approximating drug binding sites from LiP-Quant data
We validated our predictive strategy to estimate the position of drug binding sites with the LiP-Quant experiments for which true targets are known (Fig. 3) and consider only the protein hits that have multiple high-scoring LiP-Quant peptides. We chose the top 3 LiP-Quant peptides of a protein–drug target candidate among the 15 highest-ranking peptides by LiP-Quant score of the whole proteome (Supplementary Data 15). Only one protein target candidate typically fulfilled these criteria in all analyzed experiments. LiP-Quant peptides measured with rapamycin, FK506, selumetinib, staurosporine, fostriecin and calyculin A in HeLa proteome extracts were analyzed. We analyzed the staurosporine dataset assayed using a 4 h long LC gradient. We calculated the position of drug binding sites using the center of mass of all atoms assigned to the aforementioned top 3 LiP-Quant peptides of the main candidate target. Structural models and geometric calculations were performed using PyMol 2.1.1 (Schrodinger).
Cellular thermal shift assay (CETSA)
Botrytis cinerea BO47 (CK1 His-Tagged) cell suspension was adjusted to 1 × 106 sp/ml GYPm and incubated for 24 h at 21 °C (110 rpm). 12.5 × 106 cells were treated with BAYE-004 (at various concentrations from 0.0001 to 67.5 µM) or control (1% DMSO) for the final 20 min of the 24 h growth period. Cells were harvested by filtration (100 µm) and rinsed with 15 ml of ice-cold HEPES buffer (0.1 M HEPES, 50 mM NaCl, pH 7.5). Harvested mycelium was resuspended in 3.5 ml HEPES buffer and kept on ice. 500 µl of each concentration was transferred to a 2 ml Eppendorf tube and heated to 56 °C on a thermoshaker for 3 min, an additional aliquot from each concentration was left unheated. After heating, cells were kept on ice for 3 min, snap frozen in liquid nitrogen, lyophilized overnight and then stored at −80 °C until protein extraction.
Lyophilized mycelium was lysed using a Retsch mixer mill (MM 400) with 3 mm tungsten carbide beads (30 Hz for 3 s, two cycles), then 500 µl of cold protein extraction buffer (50 mM HEPES, 50 mM NaCl, 0.4% NP-40) was added. Lysate was incubated for 10 min at 25 °C, centrifuged (10 min, 14,000g) and the supernatant was retained. The lysate was further centrifuged (20 min, 73,400g) to eliminate insoluble proteins. The supernatant was collected, and protein concentration was determined using the Qubit protein assay kit (#Q33211) and stored at −20 °C.
Target engagement was assessed by western blot. In brief, 17 µg of protein per treatment was loaded onto a TGX (4-20%) stain free gel (Bio-Rad, #4568094) and run at 250 V for 25 min. Proteins were transferred to a nitrocellulose membrane using the Trans-Blot Turbo system according to the manufacturer’s instructions (Bio-Rad, # 1704271). The membrane was probed using a monoclonal anti-polyhistidine-peroxidase antibody (1:2000, clone HIS-1, Sigma, A7058). The membrane (target protein) and gel (loading control) were imaged using a ChemiDocXRS camera and quantified using ImageJ43. The uncropped blot image is included in the source data file (see Data Availability).
Cell viability (IC50) assay
Botrytis cinerea BO5.10 (2 × 103 cells/ml) mycelium in GYPm liquid media (200 µl) was cultured at 21 °C without shaking in a micro-titer plate. Optical density was measured at 620 nm (Tecan M1000 plate reader) at the beginning of the culture period (day 0) and immediately inoculated with 2 µl of BAYE-004 to obtain final concentrations (µM) of 1.2234, 0.40745, 0.13582, 0.04527, 0.01509, 0.00168, 0.00056, 0.00019, and 0, respectively. The culture was grown for three days at 21 °C after inoculation at which point the optical density was measured again. Inhibition of cell growth was calculated at each concentration.
CK1 kinase assay
The LANCE Ultra time resolved fluorescence resonance energy transfer (Tr-FRET) kinase assay (Perkin Elmer, #TRF0300-C) protocol was used according to manufacturer’s instructions with adaptations made for the following conditions. Botrytis cinerea MBP-CK1 recombinant enzyme (89.7 kDa) was used at 12.5 nM with the following reagents: ULight-DNA topoisomerase 2-α(Thr1342) peptide (Phosphorylation motif: DEKTDDE, PerkinElmer, #TRF0130-M), Europium-labeled DNA topoisomerase 2-α(Thr1342) antibody (mouse monoclonal, PerkinElmer, #TRF0218-M) in a solution containing DMSO (1%). The reaction was performed in duplicates in the dark at 30 °C for 90 min. Plates were read in a Victor2 Perkin plate reader (excitation 340 nm/emission 665 nm).
Details of all materials used in these studies are provided in Supplementary Data 16. All chemicals, enzymes, peptides and compounds were purchased from Sigma-Aldrich unless specified otherwise. BAYE-004 was produced by Bayer Crop Science (Supplementary Note 1 and Supplementary Figs. 10, 11). MBP-CK1 recombinant enzyme was cloned in-house by Bayer Crop Science. Frozen HeLa cell pellets were purchased from Ipracell (Belgium) and live HeLa cells for culture were purchased from Sigma-Aldrich. Strains of S.cerevisiae were obtained from the European Saccharomyces Cerevisiae Archive for Functional Analysis (Euroscarf) or subcloned from them (Supplementary Data 16). B. Cinerea was provided by Bayer Crop Science (Supplementary Data 16), over-expression construct for CK1-His was cloned in-house by Bayer Crop Science. All software versions used and where they were obtained is outlined in Supplementary Data 16.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.