Cell lines and cell culture

KBM7 cells (obtained from T. Brummelkamp) and KBM7 iCas9 cells (a gift from J. Zuber) were grown in IMDM (Thermo Fisher Scientific) supplemented with 10% heat-inactivated FBS (Sigma-Aldrich) and 1% penicillin–streptomycin (Gibco). RKO iCas9-GFP and iCas9-BFP (gifted by J. Zuber), K562 (purchased from ATCC) and NALM-6 (obtained from A. Villunger) cells were cultured in RPMI 1640 (Thermo Fisher Scientific) supplemented with 10% FBS and 1% penicillin–streptomycin. HEK293T lentiviral packaging cells (obtained from Clontech), HEK293T (purchased from ATCC) and Flp-In T-REx 293 (obtained from Invitrogen) cells were cultured in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS and 1% penicillin–streptomycin.

For competitive Kinobead pull-downs, Jurkat, MCF7, K562, COLO-205 and MV-4-11 cells were cultured in RPMI 1640 medium (Biochrom) supplemented with 10% (v/v) FBS (Biochrom). SK-N-BE(2) cells were grown in DMEM/Ham’s F-12 (1:1) supplemented with 10% (v/v) FBS and OVCAR-8 cells were cultured in IMDM medium (Biochrom) supplemented with 10% (v/v) FBS.

Cell lines were cultured at 37 °C and 5% CO₂ in a humidified incubator and were regularly tested for mycoplasma contamination.

Plasmids and cloning

All plasmid preparation, unless specified otherwise, was performed in stable competent Escherichia coli (NEB) or, in the case of destination vectors, in One Shot ccdB Survival 2 T1R Competent Cells (Invitrogen) according to the manufacturer’s instructions.

The pLEX305-ccdB-Nluc-3×Flag luminescent reporter vector was generated as a destination vector starting from pLEX_305-ccdB-dTAG destination vector (Addgene, 91798) by restriction digest with AgeI and MluI and T4 DNA ligation (NEB) of a synthesized gene block (TWIST) containing the Nluc sequence and a C-terminal 3×Flag tag (5′-CGGGCAAAACCGGTGTCTTCACACTCGAAGATTTCGTTGGGGACTGGCGACAGACAGCCGGCTACAACCTGGACCAAGTCCTTGAACAGGGAGGTGTGTCCAGTTTGTTTCAGAATCTCGGGGTGTCCGTAACTCCGATCCAAAGGATTGTCCTGAGCGGTGAAAATGGGCTGAAGATCGACATCCATGTCATCATCCCGTATGAAGGTCTGAGCGGCGACCAAATGGGCCAGATCGAAAAAATTTTTAAGGTGGTGTACCCTGTGGATGATCATCACTTTAAGGTGATCCTGCACTATGGCACACTGGTAATCGACGGGGTTACGCCGAACATGATCGACTATTTCGGACGGCCGTATGAAGGCATCGCCGTGTTCGACGGCAAAAAGATCACTGTAACAGGGACCCTGTGGAACGGCAACAAAATTATCGACGAGCGCCTGATCAACCCCGACGGCTCCCTGCTGTTCCGAGTAACCATCAACGGAGTGACCGGCTGGCGGCTGTGCGAACGCATTCTGGCGGACTACAAGGACCACGACGGTGACTACAAGGACCACGACATCGACTACAAGGACGACGACGACAAGTAGTAAACGCGTTGACGATGG-3′).

To generate a destabilized version of the vector, an identical gene block was synthesized with the addition of the PEST sequence (Promega) 5′-AATTCTCACGGCTTTCCGCCTGAGGTTGAAGAGCAAGCCGCCGGTACATTGCCTATGTCCTGCGCACAAGAAAGCGGTATGGACCGGCACCCAGCCGCTTGTGCTTCAGCTCGATCAACGTC-3′ upstream of the stop codon. pENTR223 Gateway entry vectors for the kinases were obtained from Hahn/Root Labs Human Kinases ORF Kit^59,60 (Addgene Kit, 1000000014) with the exception of FYN and MAPK4, which were purchased separately (BCCM, LMBP ORF81088-E05, LMBP ORF81100-B12). pRK5-HA-ubiquitin, pRK5-HA-ubiquitin_K48R and pRK5-HA-ubiquitin_K63R plasmids were provided by G. Versteeg. A pENTR221-GFP vector was generated by BP Gateway cloning (Invitrogen), starting from the PCR-amplified GFP sequence of pCAG-GFP⁶¹ (gifted by C. Cepko, Addgene, 11150) and insertion into the empty pDONR221 (Invitrogen, 12536017). Final luminescent reporter vectors were generated by LR Gateway cloning according to the manufacturer’s recommendations (Invitrogen, incubation was routinely run overnight at 25 °C before heat inactivation and transformation). The correct insert size was assessed by analytical digest and in-frame cloning was verified by sequencing (Microsynth).

Single point mutations, with the exception of BLK S5A and S6A, were generated from the respective pENTR223 plasmids using either the Q5 site-directed mutagenesis kit (primers are shown in Supplementary Table 1, method, SDM, NEB) or by Q5 (NEB) PCR amplification (primers are shown in Supplementary Table 1, method, PCR), followed by 1 h of DpnI digest (NEB) and direct transformation into DH5α E. coli (NEB).

Stability vectors were generated by digesting the previously published plasmid backbone pRRL_SFFV_empty_BFP_P2A_mCherry²⁹ with SalI and BamHI, before insertion of the PCR-amplified kinase of interest using the NEBuilder HiFi DNA Assembly Master mix (NEB) according to the manufacturer’s instructions. BLK G2I, G2L, L3A, L3G, L3V, V4S, S5A and S6A stability reporter plasmids were generated analogously using the corresponding mutated primer pairs. Vectors for domain-swap experiments, truncated versions of the BLK stability reporter (with the exception of 1–7) and the ABL1(C464W) stability reporter were generated similarly by amplifying the respective DNA sequences from each kinase and performing two- or three-part assemblies. For the aforementioned truncated BLK version 1–7, oligos (Supplementary Table 3) were annealed, phosphorylated and ligated into the corresponding vector. Primers were designed using the NEBuilder Assembly tool (the sequences are provided in Supplementary Table 2).

pRRL_SFFV_CSK* EF1a_iRFP670 and pRRL_SFFV_BLK*_GGGS_3×Flag EF1as_BFP were generated by restriction digest with SalI and XhoI (or BamHI (3×Flag)) of pRRL.SFFV.DACF16.EF1as.iRFP670²⁹ or pRRL_SFFV_empty_GGGS_3×Flag EF1as_BFP and insertion of the PCR-amplified CSK/BLK fragments using the NEBuilder HiFi DNA Assembly Master Mix (NEB) (the sequences are provided in Supplementary Table 2).

The gateway vectors pcDNA5_FRT_ccdB_3×Flag_miniTurbo and pSTV6_ccdB_ 3×Flag_miniTurbo (provided by A. Gingras), as well as pRRL_EF1a_ccdB_emGFP_IRES_HygroR (gifted by G. Superti-Furga) formed the basis for the generation of the remaining kinase reporter vectors via LR gateway cloning (Invitrogen). For the LYN and RIPK2 BioID dataset pENTR223_LYN, pENTR223_RIPK2 or pENTR221_GFP were cloned into pcDNA5_FRT_ccdB_3×Flag_miniTurbo, and, for the BLK BioID dataset, pENTR223-BLK or pENTR223-APH1A (BCCM, LMBP ORF81047-H06) and pENTR221-GFP were cloned into pSTV6_ccdB_3×Flag_miniTurbo.

sgRNAs were cloned into a single sgRNA vector pLenti-U6-IT-EF1a-Thy1.1-P2A-Neo or dual sgRNA vector pLentiDual-hU6-IT-mU6-IT-EF1a-Thy1.1-P2A-Neo (both gifts from J. Zuber) as previously reported²⁹. The sgRNA sequences are described in Supplementary Table 4 and were designed using VBC score⁶².

The saturated mutagenesis library for BLK was ordered directly from GenScript and cloned starting from the BLK stability reporter plasmid. No barcoding was applied.

All inserted DNA sequences were verified by Sanger sequencing (Microsynth).

Cell line generation by lentiviral transduction

With exception of the generation of the LYN and RIPK2 BioID cell lines (see the ‘Generation of cell lines through flp recombinase’), all cell lines were generated by transduction of lentivirus. For virus production, HEK293T lentiviral packaging cells were transfected at 70% confluence with the to-be-packaged plasmid in addition to the two packaging plasmids (pCMVR8.74 helper, pMD2.G envelope, both gifted by D. Trono (Addgene, 22036 and 12259)) using polyethylenimine (PEI MAX MW 40000, Polysciences). Viral supernatants were collected 60 h after transfection and cell debris was removed using a 0.45-μm poly-ethersulfone filter.

For transduction, 1 million cells (K562 for luminescent reporters, HEK293T, KBM7, KBM7 iCas9 or RKO iCas9-GFP/BFP for all other reporters) per 2 ml were transduced with 250 µl virus solution and 8 µg ml⁻¹ polybrene. If required, the virus volume was adjusted to achieve the desired transduction efficiency. Then, 24 h after transduction, cells were expanded. For luminescent reporter cell lines, selection was performed with puromycin (1 µg ml⁻¹, Gibco) starting 48 h after cell recovery. Subsequently cell pools were subjected to quality control by means of immunoblot analysis using the C-terminal Flag epitope tag, as well as assessment of luminescence levels using NanoGlo Luciferase Assay System (Promega). For the latter, 10⁵ cells were seeded for each reporter cell pool in 30 µl on a 384-well plate and the luminescence was measured on the Victor X3 2030 Multilabel Reader (Perkin Elmer).

The three cell lines generated for the BLK BioID experiment (performed in KBM7) were likewise selected with puromycin (1 µg ml⁻¹, Gibco). Generated cell pools were assessed after cell recovery for construct expression 24 h after doxycycline treatment (1 µg ml⁻¹, PanReac AppliChem). Both correct fusion size and biotinylating efficiency were tested by immunoblotting. The latter was conducted by an additional incubation of cells with 100 µM biotin for varying timeframes and blotting for the biotinylated proteome using an anti-biotin antibodies (see the ‘Immunoblotting’ section).

sgRNA-vector-containing cells were selected using G418/neomycin (1 mg ml⁻¹, Gibco) 72 h after transduction. Completion of selection or sgRNA transduction efficiency was assessed by staining with APC anti-mouse Thy1.1 antibody (1:400, 202526, BioLegend) in human TruStain FcX Fc receptor blocking solution (1:1,000, 422302, BioLegend) for 5 min at 4 °C, followed by two PBS washes and subsequent analysis by flow cytometry. Genetic KOs were generated by induction of the tightly inducible Cas9 cassette by doxycycline (0.4 μg ml⁻¹, PanReac AppliChem) for a timeframe of 48 h up to 1 week (the incubation times per sgRNA are provided in Supplementary Table 4) before analysis using immunoblotting, imaging or flow cytometry.

All fluorescent reporter cell lines were either used directly for flow cytometry or selected by FACS using the CytoFLEX SRT Benchtop Cell Sorter (CytExpert SRT-Software (v.1.1.0.10007). In the first round of sorting, pools of reporter-positive cells (Supplementary Fig. 2) were enriched. For selected cell lines, single cells were sorted, expanded and used for flow cytometry, FACS-based CRISPR–Cas9 screens or imaging experiments. Specifically, the main stability reporters generated in KBM7 iCas9 for LYN WT, BLK WT and RIPK2 WT were used as clonal cell lines. over, the BLK(WT)–GFP reporters in RKO iCas9-BFP cells were also used as clones. The remainder of the stability reporters generated for imaging purposes in RKO iCas9-GFP cells were used as sorted pools. The KBM7 iCas9 LYN(Y32A) stability reporter was used as a sorted cell pool, whereas the suite of KBM7 iCas9 BLK mutant stability reporters were used without sorting and instead analysis was performed on the reporter-positive cell gate. For the latter, a matched unsorted BLK WT stability reporter was used as a control in the corresponding datasets. All genetic KOs were performed in a pooled format after G418 selection as detailed above.

Generation of cell lines through flp recombinase

Flp-In T-REx 293 cells were transfected with 200 ng of LYN-mT, RIPK2-mT and GFP-mT plasmids and 2 µg pOG44 vector (Invitrogen, V600520) using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions in a 6-well format. Then, 24 h after transfection, cells were expanded to a 10-cm dish and after an additional 24 h, cell selection was initiated with 200 µg ml⁻¹ hygromycin B (Roth) and maintained for at least four weeks. Subsequently, expression and biotinylation capacity (as described for the BioID cell lines generated by lentiviral transduction) was performed.

Temporal luminescence drug screen

Compounds from the kinase inhibitor library including 10 PROTAC controls (see the ‘Compounds’ section) and respective transcription and translation compounds were dispensed through an Echo 550 system into white 1,536-well plates (PerkinElmer, 6004684) at the appropriate concentrations (0.5–10 µM; Supplementary Data 1; 10 µM for CHX and 1 µM for NVP-2). Plates were sealed and stored at −20 °C. On the day of the drug screen, the plates were equilibrated to room temperature. Next, 5 µl of 1:100 endurazine (Promega) in buffered RPMI (complete RPMI supplemented with 50 mM HEPES pH 7, Sigma-Aldrich, H0887) were prelaid into each well using a liquid dispenser (Thermo Fisher Scientific, Multidrop Combi). Then, 5 µl of cells at a density of 640,000 cells per ml in buffered RPMI were dispensed on top. Cells were transferred to an incubator (humidified chamber, 37 °C, 5% CO₂) and the luminescence signal was measured every 4 h from 2 h to 18 h after seeding using the EnVision plate reader (Revvity). Raw luminescence signals were subsequently normalized for intraplate effects. Finally, compound effects were quantified by calculating percentage of control (POC) based on averaged, outlier-corrected DMSO (100%) and positive control (CHIR-99021; 0%) wells, for each plate and timepoint individually (Supplementary Fig. 4a). Only compounds that passed an initial preselection step were used for the final drug screen. The prescreen was performed following the identical steps but only for the two control cell lines GFP-Nluc and dGFP-Nluc and at two concentrations (2.5 µM or 10 µM for 10 mM stock compounds and 0.5 µM and 2 µM for 2 mM stock concentrations). Compounds were eliminated if any of the normalized POC data were smaller than 48 or larger than 150 or if the relative change relative to the 2-h timepoint was bigger than 0.58. In cases in which only the higher concentrations fulfilled these criteria, the corresponding lower concentration was used. In total, 1,620 compounds including 10 PROTACs were used for the drug screen (Supplementary Data 1).

Data analysis of luminescence drug screen

After the normalization of the initial drug-screening data, we performed additional data processing to obtain a binary active/inactive classification for each compound–kinase pair.

First, we implemented a filter to exclude compound–kinase pairs exhibiting high variability across replicates (s.d. > 30; Fig. 1) consistently across all five timepoints, resulting in the removal of 138 compound–kinase pairs. Overall, the proportion of pairs having 0 timepoints with a high s.d. was 99.67%. We further filtered out compounds that exhibited high reactivity against all kinases, considering them false positives due to their low initial 2-h timepoint (35 compounds with a median POC across kinases <70). We further excluded the three non-small molecules disitertide (TFA), Pep2m myristoylated (TFA) and pm26TGF-β1 (TFA) from our analysis. To ensure comparability of time series and to eliminate bias towards absolute POC values, we centred the compound–kinase series around 100 POC relative to the 2-h timepoint. This centring process was first applied across kinases and then across compounds.

We used the time series of CHX, NVP2 and DMSO controls to assess whether compound–kinase pairs significantly deviated from each control. For each kinase and timepoint, we independently calculated the normalized compound z score. A compound was considered to significantly reduce the kinase readout if it exhibited a substantial reduction (2 sigma) compared with the null model of the controls. We used the same methodology to calculate z scores of compounds concerning the distributions of all other compounds. We also calculated the z scores normalizing against only the initial 2-h timepoint to capture significant changes relative to the initial conditions.

This process resulted in eight normalization schemes: against CHX, NVP2, DMSO and compounds, considering both timepoint-independent and initial-timepoint-dependent situations. Each normalization offered varying selectivity over the compound–kinase time series, and we then expressed the scores as the count of significantly decreased timepoints (2 sigma).

Finally, we conducted a parameter scan to define a query for selecting hit compounds by combining the scores and specifying the minimal number of significantly deviated timepoints for each normalization scheme and the overall total combined through ‘or’ operators. We determined the normalization score thresholds for the query by minimizing the false-discovery rate (FDR). This was achieved using the 10 PROTAC controls and their respective kinase targets as a reference for true positives as well as a manually curated inclusion list. The query that reflected our constraints is as follows:

meaning that all of the positive kinase–compound pairs have to globally score 10 or more, or having a normalized score above the determined threshold in at least one of the individual screens (Supplementary Fig. 4b). For the rare instance of a missing timepoint (mainly associated with the kinase reporters for CDK4, CDK7 and CDK9), the score was corrected by +1. One compound was excluded from further data analysis due to scoring in >10 instances. The final KinDeg scores are shown in Supplementary Data 1 (including the annotation of excluded compounds). The final hit kinase trajectories are shown in Supplementary Fig. 9.

The screening data were used to fit the half-lives of each kinase. This was performed by fitting the equation 100 × e^{(−x × tau)} in Python (v.3.7.6) and the package scipy (v.1.4.1) to each kinase’s CHX screening trajectory. t-SNE plots were generated with sklearn and matplotlib (v.1.0.1 and v.3.5.3, respectively) from ChEMBL drug-binding data processed as described in the Chemical Checker (CC)²⁴ and compounds were characterized with CC global bioactivity signatures. Chaperone client status was mapped from a previous study⁸ to the respective canonical kinases (Supplementary Data 1) and respective pairwise comparisons were calculated using a Fisher’s exact test, applying the fisher_exact function from Python’s scipy.stats module (v.0.12.2). The JD values between kinase hit profiles were calculated as 1 − Jaccard similarities (JS), where the JS is the size of the intersection divided by the size of the union of two compound (hit) sets. Kinome trees were depicted using http://www.kinhub.org/kinmap/index.html (ref. ⁶³).

To assess whether the data were enriched for type I, II or allosteric inhibitors, we manually annotated our hit compounds for their respective binding mode (Supplementary Data 1) using the available literature data, structural properties of the inhibitors as well as structural data where available. We further used the data available in the PKIDB database^64,65 to annotate the remaining compounds.

Finally, for scoring of the stabilization events, we used a similar approach to the degraders. However, we focused on compounds exhibiting a substantial readout increase (2 sigma) relative to the controls’ null models. Importantly, we excluded the CHX normalization scheme from this analysis. Following this query:

We further excluded compounds that scored in the GFP-Nluc-3×Flag control cell line or that scored in >10 instances. With these boundary conditions, we identified 204 stabilization events across 64 kinases and 128 compounds. The associated data are provided in Supplementary Data 1.

Immunoblotting

Cell pellets (1–2 million cells per treatment) were lysed in urea lysis buffer (8 M urea, 1% CHAPS, 50 mM Tris-HCL pH 8) for 30 min with shaking at 4 °C and 1,200 rpm. Next, the samples were cleared by centrifugation for 15 min (20,000g, 4 °C) and quantified using the Pierce BCA protein assay kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Finally, the samples were diluted with Bolt LDS sample buffer (4×) (Invitrogen) supplemented with final concentration (f.c.) 10% β-mercaptoethanol (Sigma-Aldrich) and denatured for 10 min at 70 °C. Then, 20 µg per protein sample was separated on the Bolt 4–12% Bis-Tris Plus Gel (10–17 wells) (Invitrogen) using the Colour Prestained Protein Standard, Broad Range (10–250 kDa, NEB) as a marker. After transfer to a nitrocellulose membrane, membranes were stained by Ponceau-S. Next, the membranes were blocked with 5% milk in TBS-T (30 min, room temperature) and then incubated with primary antibodies overnight at 4 °C in TBS-T. The next day, the membranes were washed three times with TBS-T followed by incubation for 1 h at room temperature with the respective secondary antibodies if required. Finally, the membranes were again washed three times before analysis on the Chemidoc system using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific). The following antibodies and dilutions were used: GAPDH (1:5,000, Santa Cruz Biotechnology, sc-365062), GAPDH (1:5,000; Santa Cruz Biotechnology, sc-47724), vinculin (1:500; Szabo Scandic, SACSC-25336), Flag (1:2,000; Cell Signaling Technology, 2368), LYN (1:1,000; Cell Signaling Technology, 2796), BLK (1:1,000; Cell Signaling Technology, 3262), RIPK2 (1:1,000; Cell Signaling Technology, 4142S), phosphorylated LYN (Tyr507) (1:1,000; Cell Signaling Technology, 2731), FIP200 (1:1,000; Cell Signaling Technology, 12436), CDK9 (1:1,000; Cell Signaling Technology, 2316), TMUB1 (1:1,000, Abcam, EPR14066), cCBL (1:1,000; Cell Signaling Technology, 2747), phosphorylated LYN (Tyr397) (1:1,000; Cell Signaling Technology, 70926), HRP-conjugated anti-biotin (1:1,000; Cell Signaling Technology, 7075), peroxidase-conjugated goat anti-rabbit IgG (1:10,000; Jackson ImmunoResearch 111-035-003), peroxidase-conjugated goat anti-mouse IgG (1:5,000; Jackson ImmunoResearch, JAC115035003). For quantifications the accompanying ChemiDoc ImageLab software (v2.4.0.03) was used, normalized to the respective loading control and plotted as fold changes with respect to each genotype’s DMSO control or 0-h timepoint. The data were plotted as the mean from three independent biological replicates ± s.d. Replicates and uncropped images are shown in Supplementary Fig. 1.

Compounds

Carfilzomib (Cay17554-5) and BafA1 (Cay11038) were purchased from Cayman, HSP90i (4-(4-(23-dihydro-14-benzodioxin-6-yl)-5-methyl-1H-pyrazol-3-yl)-6-ethylresorcinol), 385920) was obtained from Calbiochem. All other small-molecule inhibitors were sourced from MedChemExpress. These include the kinase inhibitor library (Supplementary Data 1; 1,996 compounds, HY-L009), MLN4924 (HY-70062), TAK-243 (HY-100487), TAK-285 (TAK285, HY-15196), Src inhibitor 3 (SI-3, HY-130254), RI-4 (HY-107978), AV-412 (HY-10346), neratinib (HY-32721), afatinib (HY-10261), WZ4002 (HY-12026), nintedanib (HY-50904), DAPT (HY-13027), alkynyl myristic acid (HY-140335), THAL-SNS-032 (dCDK9, HY-123937), NVP-2 (HY-12214A), dabrafenib (HY-14660), PLX 4720 (HY-51424), ibrutinib (HY-10997), ONO-4059 (HY-18951), R406 (HY-11108), dasatinib (HY-10181), asciminib (HY-104010), DPH (HY-12070) and GNF-2 (HY-11007). Cycloheximide (CHX) was purchased from Cell Signaling Technology (2112S).

All compounds were dissolved in DMSO (Sigma-Aldrich, D1435) as 1 mM, 10 mM, 20 mM, 25 mM or 100 mM stock solutions. Working dilutions were prepared as 1,000× or 2,000× stock solutions. The kinase inhibitor library was delivered as 2 mM or 10 mM stock solutions (Supplementary Data 1).

Flow cytometry

Cells were treated with the compounds at the concentrations and timeframes indicated in the respective figure legends, and the fluorescent channels of interest were subsequently analysed on a LSR Fortessa (BD Biosciences) using the BD FACSDiva software (v.9.0). The data were analysed using FlowJo (v.10.6.2) as outlined in Supplementary Fig. 2 and the resulting mean BFP and mCherry values were exported for further processing. BFP/mCherry ratios were calculated after background subtraction (from matched WT cells) and normalized to either each pretreatment or genetic variant (referred to as normalized BFP in the figure legends) or normalized to a specific condition as indicated in the respective subscripts, for example, DMSO in Fig. 3c. Decay functions were fitted using Y = (Y₀ − plateau) × e(−K × X) + plateau and dose responses fitted using Y = bottom + (top − bottom)/(1 + (IC₅₀/X)n) where n is the Hill slope using the in-built functions of GraphPad Prism (v.10.0.3) and nonlinear regression fitting. Matched mCherry flow histograms to Figs. 2a and 4a and Extended Data Fig. 6e are shown in Supplementary Figs. 6a, 7a and 8a, respectively.

FACS-based CRISPR–Cas9 screen

The screens were performed as previously described²⁹. First, cells were transduced at an multiplicity of infection (MOI) of 0.1–0.2 with lentivirus containing the respective sgRNA library, prepared as described in the ‘Cell line generation by lentiviral transduction’ section to achieve a 1,000× representation per sgRNA. For LYN, the previously published UPS-focused sgRNA library⁶⁶ (7,801 sgRNAs) and, for BLK and RIPK2, a genome-wide library^62,67 was used. Then, 72 h after transduction, the transduction rate was assessed by staining with APC anti-mouse Thy1.1 antibody (1:400, 202526, BioLegend) and human TruStain FcX Fc receptor blocking solution (1:1,000, 422302, BioLegend) for 5 min at 4 °C. Next, selection with G418 (1 mg ml⁻¹, Gibco) was initiated. Cells were maintained in G418-positive medium for at least 14 days, splitting cells every 48–72 h. For the screen, Cas9 expression was induced with doxycycline (0.4 μg ml⁻¹, PanReac AppliChem) and, after 72 h, cells were treated with DMSO or the respective inhibitors (SI-3, 156 nM, 8 h; TAK285, 6 h; RI-4, 2.5 µM, 18 h). Cells were centrifuged for 5 min at 500g and stained with APC anti-mouse Thy1.1 antibody (1:400, 202526, BioLegend), Zombie NIR Fixable Viability Dye (1:1,000, BioLegend) and human TruStain FcX Fc receptor blocking solution (1:1,000, 422302, BioLegend) for 5 min at 4 °C. Subsequently cells were fixed with BD fixation buffer 4% (Thermo Fisher Scientific, Pierce) for 45 min at 4 °C followed by two washes with PBS and resuspension in FACS buffer (PBS, 5% FBS and 1 mM EDTA) for storage at 4 °C. All staining steps were performed in the dark. Cells were sorted within 48 h of fixation.

Sorting was performed on a BD FACS Aria Fusion (70-µm nozzle, BD Biosciences, BD FACSDiva software, v.8.0.2). First, cells were strained through a 35-μm nylon mesh. Next, cells were sorted for the 5% highest and lowest BFP-expressing cells as well as 30% of the mid-fraction (the gating strategy is shown in Supplementary Fig. 2). For each replicate and condition, cells corresponding to at least a 500-fold (genome-wide) or 1,000-fold (UPS-focused) library representation were sorted.

After sorting, the high, low and mid fractions were pooled per replicate and lysed overnight (14 h) at 55 °C with shaking at 1,200 rpm in lysis buffer (10 mM Tris-HCl, 150 mM NaCl, 10 mM EDTA, 0.1% SDS) supplemented with proteinase K (New England Biolabs). The next day, RNase was removed with DNase-free RNase (Thermo Fisher Scientific) for 2 h at 37 °C. The lysates were stored at −20 °C until further processing.

For DNA extraction, two rounds of phenol extraction (UltraPure Buffer-Saturated Phenol, Thermo Fisher Scientific, 15513039) using phase Lock Gel tubes (VWR, 7332477) followed by isopropanol precipitation overnight at −20 °C were performed. Next, the samples were barcoded using a two-step PCR protocol (AmpliTaq Gold polymerase, Invitrogen, 4311818). After each PCR step, amplicons were cleaned up with Mag-Bind TotalPure NGS beads (Omega Biotek) according to the the manufacturer’s protocol for double-sided selection. Final NGS libraries were pooled at equimolar ratios and sequenced on the HiSeq 3000 or NovaSeq 6000 platform (Illumina).

The resulting reads were trimmed using fastx-toolkit (v.0.0.14) and subsequently aligned (Bowtie2, v.2.4.5) and quantified (featureCounts, v.2.0.1). The corresponding workflows are available at GitHub (https://github.com/ZuberLab/crispr-process-nf/tree/566f6d46bbcc2a3f49f51bbc96b9820f408ec4a3 and https://github.com/ZuberLab/crispr-mageck-nf/tree/c75a90f670698bfa78bfd8be-786d6e5d6d4fc455). Gene-level enrichment was calculated by comparing each high or low population to the corresponding mid population using the median-normalized read counts. The resulting log₂(FC) and P values as well as the number of scoring and total quantified sgRNAs per gene are provided in Supplementary Data 3. Essential genes were retrieved from DepMap (23Q4)⁶⁸.

FACS-based DMS

The screen was conducted similar to the CRISPR–Cas9 screen, first transducing cells at an MOI of 0.1–0.2 followed by FACS-based enrichment of double-positive cells. Cells were treated before the screen with DMSO, TAK285 (2.5 µM, 6 h) or IMP-1088 (1 µM 24 h). Cells were then fixed (see the ‘FACS-based CRISPR–Cas9 screen’ section) and sorted for 5% high or low or 30% mid BFP level cells. After DNA extraction, samples were barcoded by two-step PCR with customized primer sets. The samples were finally sequenced using the NovaSeq 6000 platform (Illumina) run in PE150.

For the analysis, we adapted our previously established pipeline⁶⁹. In brief, the raw sequencing reads were converted to fastq format with samtools (v.1.17). Demultiplexing of paired-end reads was performed using cutadapt (v.4.4), matching read 1 5′ barcodes were provided in a separate FASTA file, with no trimming applied (–action=none). Demultiplexed paired-end FASTQ files were converted to unaligned BAM format using Picard’s FastqToSam tool (v.3.0.0) and trimmed using Trim Galore (v.0.6.6) in paired-end mode with Nextera adapter trimming enabled. Short reads were aligned to the BLK unique domain sequence and SAM files were generated using the mem algorithm from the bwa software package (v.0.7.17). The SAM file was converted to BAM format using samtools (v.1.15.1) and mutation calling was performed using the AnalyzeSaturationMutagenesis tool from GATK (v.4.1.8.1). Next, the relative frequencies of variants were calculated for each position and variants that were covered by less than 1 in 30,000 reads were excluded from further analysis. Read counts for each variant were then normalized to the total read counts of each sample and log₂(FC) values comparing high/low-to-mid fractions of each condition were calculated. P values were adjusted for multiple testing using the Benjamini–Hochberg procedure to control the FDR. The resulting significant (adjusted P< 0.05) log₂(FC) low-to-mid and high-to-mid comparisons per condition are provided in Supplementary Data 4. For DMSO-normalized results, finally, log₂(FC) values were calculated with respect to the respective DMSO high-to-mid or low-to-mid log₂(FC). Heat maps were generated using the pheatmap (v.1.0.12) package in R (v.4.1.0).

Immunofluorescence staining

Cells were seeded in PhenoPlate 96-well microplates (Revvity) and subjected to drug treatment after 24 h of pre-attachment. After treatment, cells were fixed with BD Cytofix for 10 min at room temperature. After three PBS washes, cells were permeabilized with 0.2% sodium citrate, 0.1% Triton X-100 for 5 min at room temperature. Next, cells were washed three times with PBS after a 30 min block with BSA (0.024 g ml⁻¹) and incubated overnight at 4 °C with the respective antibodies diluted in blocking solution (1:500, TMUB1, Abcam, EPR14066). The next day, cells were washed three times with PBS followed by an incubation for 2 h at room temperature with secondary antibodies (1:500, Alexa-Fluor 647, Cell Signaling Technology, 4414) and 1:1,000 concanavalin A–Alexa Fluor 488 (Thermo Fisher Scientific, C11252). Finally, cells were washed three times and kept in 100 µl PBS. The samples were imaged within the next 24 h.

High-content confocal imaging and data analysis

Cells were imaged using the PerkinElmer Opera Phenix automated microscope run on the Harmony software (v.4.9 or later) and using the pre-set filter settings for DAPI (BFP), AF-488 (GFP), AF-647 (TMUB1), mCherry and brightfield. Exposure was set to <400 ms per channel. BFP and GFP, as well as AF-647 and mCherry channels were separated during acquisition. Cells were seeded 24 h before imaging into 384-well or 96-well (CellCarrier Ultra, Revvity) plates to achieve a final cell density of 40–60%. Drugs were added immediately before imaging as indicated in the figure legends. All of the experiments were acquired with a ×40 air objective, with exception of the immunofluorescence data, which were acquired with a ×63 water objective.

Cells were segmented using cellpose⁷⁰ (0.6.5-foss-2020b) using either the mCherry (RIPK2) or GFP (BLK) channel and an adjusted diameter of 38, 50 or 80, respectively. Next, relevant features and fluorescence were extracted using custom-built cellprofiler pipelines (4.1.3-foss-2020b).

In all instances, ConvertImageToObjects (convert to boolean image (no), preserve original labels (yes)) was used to generate the primary objects. Next, for RIPK2, EnhanceOrSuppressFeatures was applied (Operation = Enhance, Type = Speckles, Size 6, Speed and accuracy = Fast) followed by IdentifyPrimaryObjects (diameter = 2-20, thresholding strategy = global, method = manual, threshold = 0.0016, smoothing scale 1.3488, method clumped objects&draw lines between clumped objects = Intensity, automatic smoothing and distance calculation enabled, holes filled in after both thresholding and declumping). RelateObjects was applied to assign the resulting speckles per cell object. Finally, MeasureObjectIntensity and MeasureObjectSizeShape were applied for measuring the respective parameters across the speckles and cell objects, before exporting the data to a database for further processing through self-written Python scripts. For the RIPK2 data associated with Fig. 4h–j and Extended Data Fig. 8h, due to the different absolute BFP fluorescence values of the constructs for RIPK2 WT and RIPK2(∆CARD), two steps were added before EnhanceOrSuppressFeatures. Namely, ExpandOrShrinkObjects was applied to eliminate cell boundaries (Operation = shrink by a specified number of pixels, pixels = 4) followed by ImageMath, which was used to calculate the BFP to mCherry ratio. The thresholds in IdentifyPrimaryObjects were thus adapted to 0.2 instead of 0.0016. For RIPK2(I212D) and RIPK2 WT stability reporter data, an additional step of prefiltering cells with less than 0.01 mCherry signal was added before segmentation of the speckles. In the TAK243 dataset and the extended KO data for XIAP and BIRC2, the threshold for IdentifyPrimaryObjects was adjusted to 0.0024 and 0.002 respective to the total BFP signal per acquired dataset.

For the immunofluorescence staining and GFP co-localization experiment, a similar approach as above was conducted. After primary object identification, RIPK2 foci were again identified using EnhanceOrSuppressFeatures (Operation = Enhance, Type = Speckles, Size 6, Speed and accuracy = Fast) followed by IdentifyPrimaryObjects (diameter = 2-20, thresholding strategy = global, method = manual, threshold = 0.0015, smoothing scale 1.3488, method clumped objects&draw lines between clumped objects = Intensity, automatic smoothing and distance calculation enabled, holes filled in after both thresholding and declumping). For TMUB1, IdentifyPrimaryObjects was applied with a threshold of 0.035 and, for GFP, a threshold of 0.015 was used. Each speckle was first assigned to a corresponding cell using the RelateObjects function, followed by the RelateObjects function run on the foci per condition.

For the BLK–GFP cell clones, only the module MeasureObjectIntensity was applied after object classification. Corresponding data were exported to a spreadsheet for further processing.

In all instances, Python (v.3.7.6) was used to annotate the resulting data (condition, replicate) and normalize the data. Normalized data were then exported and depicted in GraphPad Prism (v.10.0.3). In all cases, data were averaged per biological replicate of the mean values per cell. The s.d. was correspondingly calculated across the biological replicates.

NanoBRET

The assay was performed as described previously⁷¹. In brief, full-length CSK and LYN were obtained as plasmids cloned in frame with an N-terminal Nluc-fusion (gift from Promega). Plasmids were transfected into HEK293T cells using FuGENE HD (Promega, E2312), and proteins were allowed to express for 20 h. Serially diluted inhibitor and NanoBRET K4 Tracer (Promega, TracerDB: T000037) at the Tracer KD concentration taken from TracerDB⁷² were pipetted into white 384-well plates (Greiner 781207) using an ECHO acoustic dispenser (Labcyte). The transfected cells were added and reseeded at a density of 2 × 10⁵ cells per ml after trypsinization and resuspending in Opti-MEM without Phenol Red (Life Technologies). The system was allowed to equilibrate for 3 h (37 °C, 5% CO₂) before the bioluminescence resonance energy transfer (BRET) measurements. To measure BRET, NanoBRET NanoGlo Substrate and extracellular Nluc Inhibitor (Promega, N2540) were added according to the manufacturer’s protocol, and filtered luminescence was measured on the PHERAstar plate reader (BMG Labtech) equipped with a luminescence filter pair (450 nm BP filter (donor) and 610 nm LP filter (acceptor)). Competitive displacement data were then analysed using GraphPad Prism (v.10.0.3) software using a normalized three-parameter curve fit with the following equation: Y = 100/(1 + 10(X − log(IC₅₀))).

Commercial recombinant binding/inhibitory assays

In vitro kinase inhibitory or kinase binding assays were performed using the SelectScreen platform (Thermo Fisher Scientific). TAK285 (BLK) and SI-3 (LYN) were screened using the Z′-LYTE assay, while RI-4 (RIPK2) was screened using the LanthaScreen Eu Kinase Binding Assay according to their respective assay availability. Threefold dilutions were performed starting from 30 µM and in presence of ATP, using its standard apparent K_M per kinase.

MST binding assay

Protein purification was performed as previously described⁷³. Purified γ-secretase complex, modified γ-secretase complex (PS1 fused to GFP) and full-length BLK protein (fused to GFP) were diluted in buffer containing 25 mM HEPES pH 7.4, 150 mM NaCl and 0.1% (w/v) Digitonin.

TAK285 serial dilutions were mixed with purified γ-secretase (GFP-tagged). The mixture was loaded onto MO-K022 capillaries at room temperature. Microscale thermophoresis (MST) analyses were conducted on the Monolith NT.115 (NanoTemper) system with 20% LED power and 60% MST power. The MST data were analysed using MO.Affinity Analysis (v.2.3).

Characterization of the γ-secretase and BLK interaction was performed using GFP-tagged BLK (with or without 20 μM TAK285) as the target protein and addition of serially diluted untagged γ-secretase. Samples were measured as described above.

Immunoprecipitation

Cell pellets (10 million cells) were lysed in 900 µl IP lysis buffer (50 mM Tris-HCL (pH 7.4), 150 mM sodium chloride, 0.1% Triton X-100, 1 mM EDTA and 5 mM magnesium chloride, 1× protease inhibitors) followed by lysate clearance, protein quantification and immunoprecipitation as described in the ‘Immunoprecipitation, on-bead TAMRA click and in-gel fluorescence’ section. After immunoprecipitation and sample washing, proteins were then directly eluted using 70 µl as final volume before analysis using immunoblotting. For blocking, 5% BSA in TBS-T was used instead of 5% milk in TBS-T and the phosphorylated LYN Tyr507 or phosphorylated LYN Tyr397 antibody was diluted 1:1,000 in TBS-T containing 3% BSA and 0.1% sodium azide.

Immunoprecipitation, on-bead TAMRA click and in-gel fluorescence

Cell pellets (15 million cells per condition) were lysed in 900 µl NP40 lysis buffer (DPBS with 1.5 mM magnesium chloride, 1% NP40, 1× protease inhibitors, 1× benzonase) for 30 min on ice. The lysates were cleared by centrifugation (20 min, 4 °C, 20,000g) and quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The samples were then normalized to 1 mg per input and preactivated anti-Flag magnetic beads (Sigma-Aldrich) were added, followed by the incubation for 3 h at 4 °C on a rotating wheel. The beads were washed three times with lysis buffer. After removal of the supernatant, 56 µl of click-mix (170 µM TAMRA (5-TAMRA-Azide, CLK-FA008, Jena Biosciences), 230 µM copper sulfate, THPTA 1.15 mM, HCl 5 mM, sodium ascorbate 5 mM, in PBS) were added per sample. Finally, 18 µl of elution buffer (4× Laemmli buffer supplemented with f.c. 10% β-mercaptoethanol) were added and the samples were boiled at 95 °C for 10 min before loading 20 µl of supernatant and analysis using SDS–PAGE. Before the transfer for immunoblotting and its analysis (see section ‘Immunoblotting’), the SDS–PAGE gel was imaged on the ChemiDoc system using the Alexa 546 channel and Alexa 647 for the ladder.

Ubiquitination assays

HEK293T LYN-Nluc-3×Flag cells were seeded into 10-cm culture dishes to reach around 70% confluency on the day of transfection. Transfections were performed using Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer’s protocol. In brief, 30 μl of Lipofectamine 2000 was diluted in 720 µl Opti-MEM (Thermo Fisher Scientific) and mixed gently. In parallel, 12 µg of plasmid DNA (pRK5-HA-Ubiquitin, pRK5-HA-ubiquitin(K48R) or pRK5-HA-ubiquitin(K63R)) was diluted in 720 µl Opti-MEM. Both solutions were incubated separately at room temperature for 5 min, combined and incubated for 10 min. The resulting solution was then added dropwise to the cells. Cells were split the next day and subjected to treatments 72 h after transfection. After the treatments, cells were collected with ice cold PBS and, after an additional wash with ice-cold PBS, snap-frozen on dry ice. Cell pellets were lysed in 1 ml of lysis buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.1% Triton X-100, 1 mM EDTA, 5 mM MgCl₂, 5% glycerol, freshly added protease inhibitors (Thermo Fisher Scientific) and Benzonase nuclease (Sigma-Aldrich).

For Flag pull-down assays, 500 µg of clarified lysate was incubated with 25 µl of anti-Flag magnetic beads (Sigma-Aldrich) for 3 h at 4 °C on a rotating wheel. The beads were washed three times with lysis buffer and bound proteins were eluted by boiling at 95 °C for 10 min in 70 μl of lysis buffer supplemented with 4× SDS sample buffer.

Data plotting and statistical analysis

All data are represented as the mean of technical or biological replicates ± s.d. or ±confidence intervals. Datapoints were calculated as described in the respective sections.

Imaging data, all data related to the drug screen, proteomics, CRISPR screen, as well as in vitro kinase binding/inhibitory assay were plotted with seaborn (v.0.12.2) and matplotlib (v.3.4.2) in Python (v.3.7.6). Standard packages such as numpy (v.1.21.5), pandas (v.1.0.1) and scipy (v.1.4.1) were correspondingly used for data handling, processing, normalization, statistical calculations and/or data fitting. Immunoblot quantifications, MST results and data associated with flow cytometry (except for flow histograms) were plotted in GraphPad Prism (v.10.0.3). Statistical tests and data fitting for the corresponding datasets were calculated directly with in-build functions as detailed in the respective sections. Flow histograms were exported from FlowJo (v.10.6.2). Representative images of microscopy experiments were prepared using Fiji (ImageJ, v.2.1.1/1.53i).

Preparation of BioID MS samples

Bait-mT expression was induced 24 h before initiating cell treatments using 1 µg ml⁻¹ doxycycline. The next day, the respective inhibitors (RI-4 2.5 µM, TAK285, 2.5 µM or SI-3, 156 nM) or vehicle control (DMSO, across all cell lines including the two GFP-mT versions) and 100 µM biotin (Sigma-Aldrich, B4501) were added to the cells for 1 h. For BLK-mT, an additional condition was generated including 2 h carfilzomib (1 µM) pretreatment before TAK285 addition. For RIPK2-mT, additional conditions were generated including the co-treatment with BafA1 (100 nM). After the treatments, 20 million cells per condition and replicate were collected by centrifugation followed by two washes in ice-cold PBS. The resulting cell pellets were snap-frozen on dry ice and stored at −80 °C until further processing.

The following protocol was adapted from a previous study⁷⁴. All steps were carried out with Protein LoBind tubes (Eppendorf) and HPLC-grade reagents. In brief, for lysis, cell pellets were resuspended in 250 µl lysis buffer (PBS supplemented with 1% SDS (Sigma-Aldrich, 71736), 2 mM magnesium chloride (Invitrogen, AM9530G), protease inhibitors (Thermo Fisher Scientific, 78437) and benzonase (Merck, US170746-3)). The samples were vortexed and incubated at 37 °C (300 rpm, 30 min) followed by centrifugation at 18,000g (4 °C, 30 min). The supernatant was transferred to fresh tubes and the protein concentration was measured using the Pierce 660 nm protein assay reagent (Thermo Fisher Scientific, 22660) according to the manufacturer’s instructions. Per sample, 1 mg of total protein was diluted up to a final volume of 300 µl with lysis buffer. Next, 30 µl of 50 mM TCEP (Sigma-Aldrich, 75259, diluted in H₂O) was added, the samples were vortexed and incubated for 1 h at 56 °C with shaking at 300 rpm. Then, 80 µl of 1 M HEPES (pH 7.5, AppliChem, A6916) was added, followed by 45 µl of 200 mM iodoacetamide (Sigma-Aldrich, I1149). The samples were again vortexed and incubated at 25 °C and 300 rpm. Pierce Streptavidin Agarose (Thermo Fisher Scientific, 20353) resin was prepared by centrifugation for 30 s followed by two PBS washes. Next the protein samples were added and incubated on a rotator for 1 h at room temperature in the dark. Finally, the samples were washed twice using 1× pre-washed Mini BioSpin columns (Bio-Rad, 7326207) with wash buffer 1 (0.2% SDS in 1× PBS), followed by 16 washes with wash buffer 2 (8 M urea in 1× PBS) and four washes with PBS. For elution, the slurry was resuspended in 2× digestion buffer (50 mM ammonium bicarbonate, 200 mM guanidine hydrochloride, 1 mM calcium chloride, in H₂O) and transferred again to a fresh tube. Subsequently, the supernatant was removed and 250 µl digestion buffer and freshly supplied 10 µl of trypsin solution (0.1 μg μl⁻¹, Promega, V5117) were added, before incubation overnight on a rotating wheel (14 h).

The next day, the beads were centrifuged briefly (30 s) and the supernatant was transferred into a fresh tube. Resin was washed once using 200 µl H₂O, which was added to the already separated supernatant. Peptides were cleaned up with self-made stage tips columns. These were prepared from 1 mm circles of an Epore C18 disc inserted into a 200-µl tip. On top of the C18 disc, 24 µl Oligo R3 solution (15 mg ml⁻¹ in acetonitrile (ACN)) was added before 1 min of centrifugation (1,000g). The column was then washed twice with 100 μl ACN (1,000g for 1 min) and equilibrated twice with 200 μl 0.1% TFA (3 min at 1,000g). The samples were acidified with 30% TFA (1% final concentration) before loading of the samples in two fractions onto the column (1,000g for 3 min). One wash with 200 μl 0.1% TFA (3 min at 1,000g) was followed by a double elution step using 50 μl elution buffer (90% ACN, 0.01% TFA, in H₂O) each. The eluted peptides were dried using a vacuum centrifuge (45 °C) and stored at −20 °C. Next, the samples were TMT-labelled with the TMTpro 18-plex Label Reagent Set (Thermo Fisher Scientific, A52045) according to the manufacturer’s instructions. Subsequently, labelled peptides were pooled and fractionated using on-tip high-pH fractionation. Then, 1 ml of 20 mM ammonium formate (pH 10) was added per 320 µl of pooled sample and added again to self-made C18 columns, prepared as stated above except for the final wash steps, which were performed with 200 μl of 20 mM ammonium formate pH 10 instead of ACN. The samples were loaded in fractions of 250 µl followed by a wash with 200 µl of 20 mM ammonium formate pH 10. Centrifugation at each step was carried out for 3 min and 1,000g. Elution was carried out in five fractions (2 min at 1,000g) with buffers containing 20 mM ammonium formate (pH 10) and different percentages of ACN (16%, 20%, 24%, 28%, 80%). First, 50 µl was used per respective buffer, followed by 20 µl per buffer (2 min 1,000g each). Next, all fractions were dried using a vacuum centrifuge at 45 °C and the resulting dried peptides were stored at −20 °C until data acquisition.

Sample preparation for full-proteome profiling

Per condition, 20 million cells were lysed in 300 µl lysis buffer (50 mM HEPES, pH 8 supplemented with 1 mM PMSF, protease inhibitor cocktail (Sigma-Aldrich) and 2% SDS). Cells were homogenized by pipetting and incubated at room temperature for 20 min. Next, the samples were sonicated (Covaris S2 high-performance ultrasonicator) for 150 s. The lysates were clarified by centrifugation at 20,000g for 5 min at room temperature. Extracted protein amounts were determined by BCA (Pierce BCA Protein Assay, 23227). For each sample, 200 µg (K562) or 100 µg (NALM-6) of protein was digested using a filter-aided sample preparation (FASP) protocol essentially according to published procedures⁷⁵.

In brief, proteins were reduced by addition of DTT (final concentration 83.3 mM), followed by incubation at 95 °C for 5 min. After cooling the samples to room temperature, the samples were mixed with 200 µl freshly prepared 8 M urea in 100 mM Tris-HCl at pH 8.5 (UA-buffer) and added onto FASP filter units (Merck Millipore). For buffer-exchange, the samples were centrifuged at 14,000g for 15 min at 20 °C and residual SDS was washed by an additional washing step with 200 µl UA-buffer. All of the subsequent centrifugation steps were done at 14,000g for 15 min at 20 °C. Proteins were alkylated by addition of iodoacetamide (50 mM final concentration) and incubated for 30 min at room temperature in the dark. The samples were washed three times with 100 µl UA-buffer followed by three washes with 100 µl TEAB buffer (Sigma-Aldrich). Proteins were digested by addition of sequencing-grade trypsin at a ratio of 1:50 at 37 °C overnight.

To collect peptides, 50 µl of 50 mM TEAB buffer was added and samples were centrifuged. Filters were additionally washed with 50 µl of 0.5 M NaCl and the flowthroughs of both washing steps were pooled. Peptides were cleaned-up by C18 with peptide desalting spin-columns (Thermo Fisher Scientific). The peptides of each condition were labelled with TMTpro 18plex reagents (K562) or TMTpro 6plex (NALM-6) according to the manufacturer’s instructions (Thermo Fisher Scientific). After 1 h of labelling, 1 µl of each channel was pooled together, quenched and cleaned-up by C18 and concentrated under reduced pressure. This test mix was measured by data-dependent acquisition (DDA) in the Orbitrap for both MS1 and MS2. Quantification was performed at the MS2 level. The test mix was used to calculate the median signal intensity of each TMTpro channel. The ratios to the lowest median channel intensity were derived and all channels were normalized to equalize the labelling efficiency. The pooled channels were quenched, and the samples were cleaned up by C18. As an additional quality control of channel normalization, another test pool was injected. After pooling all of the samples, an aliquot of 100 µl corresponding to roughly 450 µg was cleaned up by C18 and resuspended in 10 mM ammonium formate buffer pH 10. Peptides were separated on an C18 reversed-phase column (150 × 2.0 mm Gemini-NX, 3 µm C18 110 Å, Phenomenex) by liquid chromatography (LC) into 96 time-window-based fractions operating at 50 µl min⁻¹ constant flow rate. A total of 36 fractions were collected, using a previously described pooling strategy⁷⁶. The samples were fractionated into glass vials with 5 µl 30% TFA to acidify samples after fractionation. The fractions were dried under reduced pressure and reconstituted in 0.1% TFA for MS analysis. Additional information with regard to the reagents is provided in Supplementary Table 5.

LC–MS/MS data acquisition of BioID and full proteome samples

MS data were acquired on the Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific) coupled to the Dionex Ultimate 3000 RSLCnano system (Thermo Fisher Scientific) interfaced with the Nanospray Flex Ion Source (Thermo Fisher Scientific). Peptides were loaded on a trap column (PepMap 100 C18, 5 μm, 5 × 0.3 mm, Thermo Fisher Scientific) at a constant flow rate of 10 µl min⁻¹ with 0.1% TFA in HPLC-grade H₂O.

Next, the trap column was switched in-line, and peptides were separated on an analytical column (50 cm, 75 mm inner diameter) in-house packed with ReproSil-Pur 120 C18-AQ, 3 µm (Dr. Maisch HPLC) fitted to an ESI emitter fused silica (20 μm inner diameter × 7 cm length × 365 μm outer diameter; orifice inner diameter, 10 μm; CoAnn Technologies) kept at 50 °C. For the analysis, an analytical gradient of 190 min operated at a constant flow rate of 230 nl min⁻¹ was used. The HPLC was operated with buffer A (0.4% formic acid in HPLC-grade H₂O), and buffer B (0.4% formic acid in ACN).

The analytical gradient comprised the following steps: 0–4 min, constant 6% buffer B; 4–5 min, from 6 to 9% buffer B; 5–146 min, increase to 30% buffer B; 146–154 min, increase to 65% buffer B; and a flush at 100% buffer B. The column was re-equilibrated at 6% buffer B from 167–190 min. The samples were acquired in DDA mode using a maximum of ten dependent scans (TopN approach) with synchronous precursor selection (SPS) enabled. Peptides were ionized by applying a constant voltage of 1.8 kV. MS1 precursor survey scans for MS2 and MS3 levels were acquired with scan range of 400–1,600 m/z and a resolution of 120,000 (at 200 m/z) in the Orbitrap. The automatic gain control (AGC) was set to ‘standard’ with a maximum injection time of 50 ms. Precursor ions were filtered by charge state (2–5) excluding undetermined charge states with a dynamic exclusion (60 s with a ±10 ppm window), and monoisotopic precursor selection. The MS1 precursor intensity threshold was set to 5.0 × 10³. For MS data analysis, a charge-state filter was used to select precursors for data-dependent scanning. In MS2 analysis, spectra were obtained using one charge state per branch (from z = 2 to z = 5) in a dual-pressure linear ion trap (ITMS2). Ions were isolated using a quadrupole isolation window with an isolation window of ±0.7. Fragmentation was achieved by collision-induced dissociation (CID) with a fixed normalized collision energy of 35% and an CID activation time of 10 ms. For MS2 scans, the normalized AGC target was set to 200% with a maximum injection time of 35 ms. For MS3 scans, precursor ions were isolated using SPS waveform with varying isolation windows for charge stats: 1.3 m/z for z = 2, 1.2 m/z for z = 3, 0.8 m/z for z = 4 and 0.7 m/z for z = 5. Fragment ions were further fragmented by high-energy collision-induced dissociation (HCD) at a fixed activation energy at 45% collision energy. The AGC target was set to 300% with a maximum injection time of 100 ms. The Orbitrap scan range was set to 100–500 m/z at a resolution of 50,000. Xcalibur v.4.3.73.11 and Tune v.3.4.3072.18 were used to operate the instrument.

Processing of BioID raw MS-injections

MS-raw files were processed with the Proteome Discoverer software (PD, Thermo Fisher Scientific, v.2.4.1.15). For the LYN dataset, a subset of 9 TMT-channels (126, 127N, 127C, 132C, 133N, 133C, 134N, 134C, 135N) were used, whereas, for the RIPK2 experiment, and the BLK and APH1A experiment the full channel set was processed. The three BioID datasets were processed independently.

The peptide identification search was performed using Sequest HT, searching for fully tryptic peptides with a maximum of two missed cleavages and a minimum peptide length of 6 and a maximum of 144 amino acids. The precursor mass tolerance was set to 10 ppm and fragment ion mass tolerance was restricted to 0.6 Da. Spectra were searched against the canonical human protein database obtained from UniProtKB (download 5 November 2021, 20,304 sequences) appended with an in-house-generated list of common laboratory contaminants (298 sequences) and streptavidin. As variable modification methionine oxidation (+15.994 Da), deamidation (0.984 Da), phosphorylation on serine, threonine and tyrosine (+79.966 Da), N-terminal specific acetylation (+42.011 Da), methionine loss (−131.040 Da) and acetylation with methionine loss (−89.030 Da) with a maximum number of three variable modification of the same type per peptide. Carbamidomethylation (+57.021 Da) of cysteine residues and TMT 18-plex labelling of peptide N termini and lysine residues (+304.207 Da) were used as static modifications. PSM and peptide FDR were controlled by Percolator at 1% respectively. The obtained results were filtered to include only spectrum matches with a Sequest HT cross-correlation factor (Xcorr) larger or equal to 0.9. Phosphosites needed a minimum site-probability of 75, corresponding to the high-confidence threshold. For protein abundance inference, only high-confidence proteotypic peptides were included.

Protein and peptide intensities were derived from TMTpro reporter ion intensities. The reporter abundances were based on signal-to-noise (S/N) values if applicable, otherwise reporter ion intensities were used. Correction of isotopic impurities was enabled. A co-isolation threshold for isolation interference of precursors was set to maximum 80%. over, to remove noisy signals, an average TMTpro reporter ion S/N threshold smaller or equal to 10 was used with an additional SPS mass matches threshold of 65%, removing peptides with strong interferences. The obtained data were normalized using the sum total peptide amount and scaled to the average. For normalization and to derive protein abundances, all quantified peptides were used. Protein ratios and log₂(FC) values were directly calculated from the grouped protein abundances, without missing value imputation. Abundance changes were tested for their significance using ANOVA on individual proteins across biological triplicates. P values were corrected for multiple testing using the Benjamini–Hochberg procedure.

Data analysis and representation of BioID data

The protein-level PD output was used for further analysis. For the LYN BioID experiments, 4,325 UniProtKB accessions were identified; for RIPK2 BioID experiments, 3,962 accessions were identified; and, for BLK, APH1A BioID experiments, 2,962 UniProtKB accessions were found. From the datasets, proteins flagged as contaminates and proteins without quantification values were removed, resulting in 3,564 UniProtKB accessions for the LYN dataset, 3,334 UniProtKB accessions for the RIPK2 dataset and 1,870 UniProtKB accessions for the BLK, APH1A dataset. For subsequent data analysis, the PD-derived normalized intensities, log₂(FC) and the adjusted P value (Benjamini–Hochberg corrected) were used. For the LYN experiment, the analysis focused on significantly changed proteins after SI-3 treatment versus the vehicle/baseline control (DMSO). To identify enriched proximity interactions, a combined threshold of a log₂(FC) ≥ 2 and an adjusted P ≤ 0.01 against GFP controls were used. The same thresholds were used to identify differentially changed interactions in the LYN SI-3-treated samples against LYN DMSO-treated control. For the RIPK2 and BLK/APH1A datasets, additional scoring of proximity interaction partners was performed using SAINTq⁷⁷. For this, the total sum normalized protein intensities per replicate and condition were grouped together and scored against GFP-negative controls (DMSO). SAINTq was performed on protein level (parameters: normalise_control = false, input_level = protein, compress_n_ctrl = 3, and max score across bait replicates). For BLK/APH1A, proteins with a log₂(FC) ≥ 0.5, a SAINTq-score ≥ 0.99 and a BFDR ≤ 0.01 were considered to be high-confidence proximity interaction partners. For the RIPK2 dataset, a more stringent log₂(FC) cut-off of log₂(FC) ≥ 1 was used. As a further filter, the CRAPome⁷⁸ frequency was mapped to each prey protein, using for BLK/APH1A a 10% frequency and for RIPK2 a 20% frequency threshold. Bait proteins were excluded from the CRAPome filter. over, for BLK/APH1A, prey proteins that were annotated as kinases and type I transmembrane proteins in UniProtKB were filtered. These annotated interactors were further filtered for significant changes against GFP negative controls, using the adjusted P value (Benjamini–Hochberg corrected) from the ANOVA hypothesis test performed within PD. The obtained proximity interactors were intersected between BLK and APH1A, revealing on one hand bait-specific and on the other hand shared preys. For the RIPK2 dataset, protein interactions for each condition (RI-4, 1 h; RI-4, 4 h; and BafA1, 18 h) were used as the input for gene set enrichment analysis for GO molecular function (2023) terms using Enrichr. All proteins covered in significantly enriched terms (adjusted P ≤ 0.05) were subset from the comparison of treated versus control (RI-4, 1 h and 4 h versus DMSO; and RI-4 + BafA1, 18 h versus BafA1). All interactions found in at least two conditions versus the GFP negative controls were selected for further visualization. The subset of obtained proximity interactors were grouped into broader molecular function terms. For visualization the log₂(FC) and adjusted P value against treatment controls (DMSO or BafA1, respectively) were used. Data analysis and visualizations were generated employing the statistical software R (v.4.3.1). The resulting processed datasets are provided in Supplementary Data 5. Normalization results and additional individual volcano plots or scatter plots of SAINTq results are provided in Supplementary Fig. 6d–f (LYN), Supplementary Fig. 7b–d (BLK/APH1A) and Supplementary Fig. 8c–e (RIPK2).

Processing and data analysis of full proteome profiling data

The full proteome datasets were processed in Proteome Discoverer v.2.4.1.15, deriving protein intensities using the TMTpro 18 or TMT 6-plex reporter ion quantities.

Peptide identification search was performed using Sequest HT searching for fully tryptic peptides of a minimum of 6 to up to 144 amino acids length and allowing for a maximum of 2 missed cleavage sites. Precursor mass tolerance was set to 10 ppm and fragment ion mass tolerance was restricted to 0.6 Da. The search was performed against the canonical human protein database obtained from UniProtKB (download 12 November 2020) appended with an in-house-generated list of common laboratory contaminants and streptavidin.

As variable modification methionine oxidation (+15.994 Da) and N-terminal specific acetylation (+42.011 Da), methionine loss (−131.040 Da) and acetylation with methionine loss (−89.030 Da) were set. The maximum number variable modification of the same type was limited to 3. As a static modification, carbamidomethylation (+57.021 Da) of cysteine residues and tandem mass tag (TMT) 18-plex/6-plex labelling of peptide N termini and lysine residues (+304.207 Da) were set. PSM and peptide FDR were controlled with Percolator at 1% respectively. Obtained results were filtered to include spectrum matches with a Sequest HT cross-correlation factor (Xcorr) ≥ 1 and strict Percolator target FDR filters. For further analysis, only peptides scored with high confidence and proteins identified with at least 1 proteotypic peptide were used. Protein and peptide intensities were derived from TMTpro reporter ion intensities. The reporter abundances were based on S/N values if applicable, otherwise reporter ion intensities were used. Correction of isotopic impurities was enabled. Co-isolation threshold of 70% for isolation interference of precursors was used. over, to remove noisy signals, an average TMTpro reporter ion S/N threshold of ≤10 was used. A SPS mass match threshold of at least 65% was applied. For reporter-ion-based quantification, unique and razored peptides were considered. The obtained data were normalized using the sum total peptide amount. For normalization and to derive protein abundances, all quantified peptides per protein were used. Protein ratios and log₂(FC) values were calculated from the grouped protein abundances, without missing value imputation. To test for differentially abundant proteins, ANOVA for individual proteins across all biological replicates (n = 3) was performed. For further analysis of degradation selectivity in K562, the 7,665 protein groups with a high confidence score and quantitative values were used. For the NALM-6 cells, the 7,437 protein groups were used for further analysis. For each drug-treated condition the log₂(FC) and P values were derived against DMSO/baseline control conditions. Depending on the duration of the treatment, either the 8- or the 18-h negative control was used. The resulting processed datasets are provided in Supplementary Data 2.

Kinobead profiling

Cells were lysed in 0.8% IGEPAL, 50 mM Tris-HCl pH 7.5, 5% glycerol, 1.5 mM magnesium chloride, 150 mM sodium chloride, 1 mM sodium orthovanadate, 25 mM sodium fluoride, 1 mM DTT, protease inhibitors (SigmaFast, Sigma-Aldrich) and phosphatase inhibitors (prepared in-house according to phosphatase inhibitor cocktail 1, 2 and 3 from Sigma-Aldrich). The cell lysate mixes used for compound profiling were generated either from COLO-205, K562, SK-N-BE(2), MV-4-11 and OVCAR-8 cell lysates (standard 5 CL (cell line) mix) or Jurkat and MCF7 cells mixed at equivalent ratios; the protein concentration was determined using the Bradford assay.

Kinobeads pull-down experiments were performed as previously described⁷⁹. In brief, 2.5 mg of the cell lysate mixture was pre-incubated with increasing compound concentrations (DMSO, 3 nM, 10 nM, 30 nM, 100 nM, 300 nM, 1 µM, 3 µM, 30 µM) for 45 min at 4 °C in an end-over-end shaker in either of the two lysate mixes. Next, the lysates were incubated with Kinobeads (17 µl settled beads) for 30 min at 4 °C. The beads were washed and bound proteins were reduced with 50 mM DTT in 8 M urea, 40 mM Tris HCl (pH 7.4) for 30 min at room temperature. After alkylation with 55 mM CAA, proteins were digested with trypsin overnight at 37 °C. Peptides were desalted using C18 StageTips and dried down in a SpeedVac. Peptides were analysed using LC–MS/MS on the Dionex Ultimate3000 nano HPLC system coupled online to an Orbitrap Fusion Lumos (Thermo Fisher Scientific) mass spectrometer. Peptides were delivered to a trap column (100 µm × 2 cm, packed in-house with Reprosil-Gold C18 ODS-3.5 µm resin, Dr. Maisch, Ammerbuch) and washed at a flow rate of 5 µl min⁻¹ in solvent A (0.1% formic acid, 5% DMSO in HPLC-grade water). Peptides were then separated on an analytical column (75 µm × 40 cm, packed in house with Reprosil-Gold C18 3 µm resin, Dr. Maisch) using a 52-min gradient ranging from 4 to 32% solvent B (0.1% formic acid, 5% DMSO in ACN) in solvent A at a flow rate of 300 nl min⁻¹. The mass spectrometer was operated in a data-dependent mode, automatically switching between MS1 and MS2 spectra. MS1 spectra were acquired over a m/z range of 360–1,300 m/z at a resolution of 60,000 in the Orbitrap using a maximum injection time of 50 ms and an AGC target value of 4 × 10⁵. Up to 12 peptide precursors were isolated (isolation width of 1.7 Th, maximum injection time of 75 ms, AGC value of 5 × 10⁴), fragmented by HCD using 30% normalized collision energy and analysed in the Orbitrap at a resolution of 15,000. The dynamic exclusion duration of fragmented precursor ions was set to 30 s.

Peptide and protein identification and quantification was performed using MaxQuant (v.1.5.3.30) by searching the tandem MS data against all canonical protein sequences as annotated in the UniProtKB reference database using the embedded search engine Andromeda. Carbamidomethylated cysteine was set as a fixed modification and phosphorylation of serine, threonine and tyrosine, oxidation of methionine and N-terminal protein acetylation as variable modifications. Trypsin/P was specified as the proteolytic enzyme and up to two missed cleavages were allowed. The minimum length of amino acids was set to seven and all data were adjusted to 1% PSM and 1% protein FDR. LFQ and match between runs were enabled within MaxQuant.

For the Kinobeads competition binding assays, protein intensities were normalized to the respective DMSO control and IC₅₀ and EC₅₀ values were deduced by a four-parameter log-logistic regression using an internal pipeline that uses the drc package⁷⁹ in R. An apparent dissociation constant (K_d,app) was calculated by multiplying the estimated EC₅₀ by a protein-dependent correction faction. The correction factor of a protein is defined as the ratio of the amount of protein captured from two consecutive pull-downs of the same DMSO control lysate. Targets of the compounds are annotated manually. A protein is considered a target if the resulting binding curve shows a sigmoidal curve shape with a dose-dependent decrease of binding to the beads. over, the number of unique peptides and MSMS counts per condition as well as the protein intensity in the DMSO control are taken into account. The resulting fitted parameters in addition to the normalized intensities are provided in Supplementary Data 6.

TAK285 chemoproteomics

To generate the TAK285 affinity matrix, the terminally amine-tethered TAK285 probe (synthesis is described in the Supplementary Methods) was immobilized to Sepharose beads as previously described⁸⁰.

For the competition assay, NALM-6 cell lysates were prepared as previously described⁸⁰. The protein amount of cell lysates was determined using the BCA assay and adjusted to an Igepal concentration of 0.4% and protein concentration of 5 mg ml⁻¹ by diluting with Igepal-reduced lysis buffer. The cell lysate was pre-incubated with different doses of TAK285 or the DMSO vehicle control for 45 min at 4 °C on a shaker, followed by incubation with 18 μl TAK285 affinity matrix for 30 min at 4 °C on a shaker. The beads were washed (once with 1 ml of lysis buffer without protease inhibitors and with only 0.4% Igepal, twice with 1 ml of lysis buffer without protease inhibitors and with only 0.2% Igepal, three times with 1 ml of lysis buffer without protease inhibitors and without Igepal), and the captured proteins were denatured with 8 M urea buffer, alkylated with 55 mM iodoacetamide and digested with trypsin according to standard procedures.

The resulting peptides were desalted by StageTip desalting⁸¹. To construct a StageTip, five C18 discs were packed into a 200 μl pipette tip. The StageTips were activated with 200 μl ACN (all centrifugation steps at 250g), washed with 200 μl buffer B (0.1% formic acid in 50% ACN) and equilibrated with 200 μl buffer A (0.1% formic acid in double-distilled H₂O). The peptide samples were acidified to a final concentration of around 0.3% formic acid (pH > 2) and loaded on to StageTips. The loading step was repeated with the flow-through. Peptides attached to the C18 material were washed twice with 200 μl buffer A and eluted by adding twice 40 μl of buffer B and collecting the flow-through. The eluent was vacuum-dried and stored at −20 °C until LC–MS/MS measurement.

LC–MS/MS measurement of the TAK285 competition assay

For proteomic data acquisition, a nanoflow LC–ESI-MS/MS setup, comprising a Dionex Ultimate 3000 RSLCnano system coupled to a Fusion Lumos mass spectrometer (both Thermo Fisher Scientific), was used in positive ionization mode. MS data acquisition was performed in DDA mode. For proteome analyses, half of the competition pull-down peptides were delivered to a trap column (Acclaim PepMap 100 C18, 3 μm, 5 × 0.3 mm, Thermo Fisher Scientific) at a flow rate of 5 μl min⁻¹ in HPLC-grade water with 0.1% (v/v) TFA. After 10 min of loading, peptides were transferred to an analytical column (ReproSil Pur C18-AQ, 3 μm, Dr. Maisch, 500 mm × 75 μm, self-packed) and separated using a stepped gradient from minute 11 at 4% solvent B (0.4% (v/v) formic acid in 90% ACN) to minute 61 at 24% solvent B and minute 81 at 36% solvent B at a 300 nl min⁻¹ flow rate. The nano-LC solvent A was 0.4% (v/v) formic acid HPLC-grade water.

MS1 spectra were recorded at a resolution of 60,000 using an AGC target value of 4 × 10⁵ and a maximum injection time of 50 ms. The cycle time was set to 2 s. Only precursors with charge state 2 to 6 that fall in a mass range between 360 to 1,300 Da were selected and dynamic exclusion of 30 s was enabled. Peptide fragmentation was performed using HCD and a normalized collision energy of 30%. The precursor isolation window width was set to 1.3 m/z. MS2 spectra were acquired at a resolution of 30,000 with an AGC target value of 5 × 10⁴ and a maximum injection time of 54 ms.

Data analysis of the TAK285 competition assay

Protein identification and quantification was performed using MaxQuant (v.2.4.9.0) by searching the LC–MS/MS data against all canonical protein sequences as annotated in the Swiss-Prot reference database (downloaded April 2024) using the embedded search engine Andromeda. Carbamidomethylated cysteine was set as fixed modification and oxidation of methionine and amino-terminal protein acetylation as variable modifications. Trypsin/P was specified as the proteolytic enzyme, and up to two missed cleavage sites were allowed. Precursor tolerance was set to 10 ppm, and fragment ion tolerance was set to 20 ppm. The minimum length of amino acids was set to seven, and all data were adjusted to 1% peptide spectrum matches and 1% protein FDR. LFQ⁸² and match between runs was enabled.

To search the proteomics data for dose-dependently competed proteins, we submitted the data to the CurveCurator pipeline⁸³. This tool automatically calculates protein LFQ intensities at each competition concentration relative to the DMSO control, plots dose–response curves and applies customized statistics for calling proteins dose-dependently regulated. The associated data are provided in Supplementary Data 7.

BLK γ-secretase complex prediction and molecular dynamics simulations

The BLK–γ-secretase complex was predicted using AlphaFold3, with template information enabled for γ-secretase⁴³. BLK was modelled starting at Gly2. The complex was prepared using CHARMM-GUI Membrane Builder⁸⁴: the structure was automatically oriented using the PPM 2.0 method and inserted into a POPC bilayer using the replacement method. An N-terminal myristoylation was added at Gly2 during the setup process. The system was solvated with TIP3P water and neutralized with 0.15 M NaCl. GROMACS (v.2023.2) input files were generated according to CHARMM-GUI’s standard protocol, comprising energy minimization (step 6.0), six-step equilibration (steps 6.1–6.6), and production dynamics (step 7), which were extended to 50 ns. Simulations were repeated in triplicate with different initial velocities. MM/GBSA binding free-energy estimates were computed, and interface contacts were analysed using GetContacts (https://getcontacts.github.io/). Depictions were generated with VMD (v.1.9.4).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.