Cell culture and treatment experiment
HCT116, Caco2 and DLD1 were obtained from ATCC (Manassas, Virginia, USA). Cell lines were confirmed to be mycoplasma free. All cell-lines were cultured in DMEM, high glucose, GlutaMAX Supplement, pyruvate (Gibco, Invitrogen), supplemented with 10% fetal bovine serum (FBS, Gibco), in a humidified incubator at 37 °C with 5% CO2. Experiments are performed at 80% confluence. Cells are harvested in ice cold PBS, by washing two times in ice cold PBS, and scraping from plate. The cells were rinsed again with 1× PBS and centrifuged at 250g for 5 min in a centrifuge maintained at 1 °C. The final PBS wash was removed and the resulting pellet was frozen in ethanol on dry ice and stored at −80 °C.
Treatment with Growth Factor and MEK Inhibitor
For the interaction experiment, we used the MEK inhibitor Selumetinib (AZD6244, #S1008, Selleckchem) at a final concentration of 10 µM, dissolved in DMSO. The growth factors used were HGF (Peprotech #100-39H, final concentration 0.025 µg/ml), FGF2 (Peprotech #100-18B, final concentration 0.005 µg/ml), EGF (Peprotech #AF-100-15, final concentration 0.025 µg/ml), and VEGF-C (Peprotech #100-20CD, final concentration 0.1 µg/ml), all of which were dissolved in 0.1% BSA. Depending on the specific experiment, these growth factors were either added individually or combined into a mixture.
Prior to treatment, the cells underwent a starvation period of approximately 18 hours using 0.1% FBS to synchronize their growth. Following this, the cells were treated with either MEKi or a control solution of DMSO for 3.5 hours. At 3.5 hours either a control solution (0.1% BSA) or the growth factor (mixture) was added to the cells for another 30 minutes. After a total incubation time of 4 hours, the cells were harvested for further analysis.
Selection and synthesis spike-in peptides
Phosphosites were selected for relevance to cellular signaling. The Kinase-Substrate, Phosphorylation-site and Regulatory-site datasets were downloaded from PhosphoSitePlus (February 2018). Phosphorylation sites that are present in all three datasets, annotated to regulate protein activity and annotated to selected KEGG signaling pathways. Phosphosites annotated to manually selected proteins (CREB1, ABL1, IGF1R, IRS1, RPS6KA1, PDGFRA, PIK3R1 and RPS6) or with more than 100 references were kept in regardless of activity and pathway annotation. This results in a list of ~1000 phosphorylation sites. These phosphorylation sites were mapped to an in-silico trypsin digested proteome, and the resulting phosphopeptides were filtered based on their MS properties. The peptide is not allowed to be found in the proteome >10 times and it can not contain >10 phospho-accepting residues. Phosphopeptides were filtered for synthesis feasibility, and singly phosphorylated peptides with a length between 7 and 21 amino acids and N-terminal K/R were selected for synthesis. The list with shorter peptides contains 524 phosphorylation sites, mapping to 485 unique phosphopeptides, on 277 proteins. Peptides were purchased from JPT, synthesized using FMOC solid-phase technology with crude purity and synthetic isotope–labeled c-terminal lysine (K) or arginine (R) and pooled. Lyophilized synthetic peptides were kept at -20.
Phosphoproteomics sample preparation
Cell pellets were lysed at 4 °C with urea lysis buffer (8 M urea, 50 mM Tris (pH 8),150 mM NaCl) supplemented with protease inhibitors (2 μg/ml aprotinin, 10 μg/ml leupeptin) and phosphatase inhibitors (10 mM NaF, phosphatase inhibitor cocktail 1 and 2, Sigma Aldrich). The cell lysate was treated with 5 mM dithiothreitol for 1 h to reduce proteins and then alkylated with 10 mM iodoacetamide for 45 min in the dark. Sequencing grade LysC (Wako) was added at a weight to weight ratio of 1:50. After 2 h, samples were diluted 1:4 with 50 mM Tris–HCl pH 8 and sequencing grade trypsin (Promega) was added at 1:50 ratio. Digestion was completed overnight. subsequently samples were acidified using FA and desalted with Sep-Pak C18 cc Cartridges (Waters). Lyophilized samples are diluted to 0.7 µg/µL in 80% ACN / 0.1% TFA. 200 fm heavy labeled synthetic peptides are added to 100ug sample and subjected to automated immobilized metal affinity chromatography (IMAC) phosphopeptide enrichment by the Bravo Automated Liquid Handling Platform (Agilent) with AssayMAP Fe(III)-NTA cartridges.68
Liquid chromatography mass spectrometry
Mass spectrometry raw data were acquired on a Bruker timsTOF Pro2 connected to a Thermo Fischer EASY-nLC 1200 system. Around 300 ng (⅓ of IMAC output) was injected. Samples were separated online on a 25 cm column packed in-house with C18-AQ 1.9 μm beads (Dr. Maisch Reprosil-Pur 120). A gradient of mobile phase A (0.1% formic acid and 3% acetonitrile in water) and mobile phase B (0.1% formic acid, 90% acetonitrile in water) was used to separate the peptides at a flow rate of 250 nl/min. Mobile phase B was ramped from 2% to 30% in the first 29 min, followed by an increase to 60% B in 3 min and a plateau of 90% B for 5 min. Temperature of the column was kept constant at 45 °C. The LC system was connected to Bruker timsTOF Pro2 hybrid TIMSQTOF mass spectrometer via a CaptiveSpray nano-electrospray source. The raw files were acquired in dia-PASEF mode, using the standard ‘long gradient’ method as supplied by the vendor. All spectra within a mass range of 400 to 1201 Da and an IM range from 1.6 to 0.6 V·s/cm2 were acquired using equal ion accumulation and ramp times in the dual TIMS analyzer of 100 ms each. The collision energy was lowered as a function of increasing ion mobility from 59 eV at 1/K0 = 1.6 V·s/cm2 to 20 eV at 1/K0 = 0.6 V·s/cm2. The estimated cycle time is 1.80s. The calibration status of the machine is monitored constantly and calibration of the ion mobility dimension is performed linearly using at least three ions from Agilent ESI LC/MS tuning mix (m/z, 1/K0: 622.0289, 0.9848 V·s/cm2; 922.0097, 1.1895 V·s/cm2; 1221.9906, 1.3820 V·s/cm2).
Benchmark sample generation and High-pH Reverse-Phase Fractionation for Library Generation
For the High pH library generation HCT116 was treated with combinations of phosphatase inhibitors, to increase the number detectable phospho-site relative to normal growth conditions. For inhibition of phosphatases, HCT116 (ATCC, #7) was treated with 1 mM pervanadate and 50 ng/ml calyculin A. To this end, cells were starved in 0.1% FBS for 3 hours. afterwards cells were treated with no serum, with 10% FBS, 10% FBS + calyculin A, or 10% FBS + calyculin A + Pervanadate. Samples were processed as described before, and desalted peptides were combined before drying down.
For library generation, the peptides are subjected to offline high pH reverse phase fractionation by HPLC on an Agilent 1290 Infinity II HPLC instrument. To this end, the dried peptides were reconstituted in high pH buffer A (4.5 mM ammonium formate, 2% ACN, pH 10), and loaded on a XBridge BEH C18 4.6 × 250 mm column (130Å, 3.5 μm bead size; Waters), and separated using a 96-min gradient with a flow rate of 1 ml/min. The gradient was performed by ramping high pH buffer B (4.5 mM ammonium formate, 90% ACN, pH 10) from 0% to 60% 68. The 96 fractions were collected and concatenated by pooling equal interval fractions. The final 48 fractions were dried down and resuspended for IMAC enrichment as described above. 100 µg of each pooled fraction were used for IMAC enrichment. Per IMAC enriched fraction 100 ng phospho-enriched peptides were measured on the timsTOF Pro2.
Same LC conditions as described previously were used. Data were acquired using default DDA-PASEF mode with a cycle time 1.1s and 10 PASEF MS/MS scans per topN acquisition cycle. All spectra were acquired within an m/z range from 100 to 1700 and an IM range from 1.6 to 0.6 V·s/cm2
Raw data was analyzed with MaxQuant (v2.4.0.0) and searched against the
human reference proteome database (downloaded from UniProt in 06/2023) and default protein contaminants included in MaxQuant. Fixed modifications were set to carbamidomethylation of C. Variable modifications included oxidation (M) and N-terminal acetylation and phosphorylation (STY). A maximum of 5 modifications per peptide and 2 missed cleavages were allowed. MaxQuant results are filtered to exclude reverse database hits, potential contaminants and phospho-sites with a localisation probability lower than 50%. MaxQuant results were transformed and were necessary combined to a DIAN-NN compatible library, including ion mobility information.
Synthetic peptides library generation
The synthetic phosphopeptides were measured to generate a library. The peptides were dissolved in Buffer A (3% ACN, 0.1% FA). To generate a library, 50 fm and 100 fm peptides were measured in DDA-PASEF mode in triplicates, with the same settings as for HpH-library generation acquisition. The resulting raw files were analysed in MaxQuant against a library specific .fasta file. MaxQuant settings and processing of MaxQuant output as above. A table with peptides in the library, along with main annotated phosphosite and other annotated phosphosites can be found in supplementary table 1.
Full phospho-proteome dilution benchmark
For dilution series with the SILAC labeled phosphoproteome HCT116 was cultured in Heavy or Light SILAC medium. SILAC medium consists out of arginine- and lysine- free DMEM, supplemented with 10% dialyzed fetal bovine serum (dFBS, Gibco) and either heavy (13C615N4 L-arginine or Arg10 an 13C615N2 L-lysine or Lys8) and light amino acids (Cambridge Isotope Laboratories) at 0.4 mmol/L and 0.8 mmol/L. Cell harvest and lysis as described above. Protein concentration in cell lysate was determined using BCA and the heavy labeled cell lysate was sequentially diluted into the light cell lysate. Subsequent sample preparation and phosphopeptide enrichment were performed as usual. For mass spectrometry 100 ng phospho-enriched peptides per sample were injected per dilution in triplicates and analysed in DIA-PASEF mode, as described above.
Cell line panel screen for MEKi-dependent receptor-mediated feedbacks
Bio-Plex data generation
Human colorectal cell-lines used in this experiment Colo205, Colo678, DLD-1, GEO, HCT116, HT29, LIM1215, RKO, SW403, SW480 and Caco-2 were provided by AG Sers Molekulare Tumorpathologie (Charité-Universitätsmedizin). All cell-lines were cultured in low glucose DMEM (D5546-6X500ML, Sigma-Aldrich) supplemented with 10% FBS, 10mM Ultraglutamine and Penicilin-Streptomycin and were incubated at 37°C and 5% CO2.
Before perturbation commenced cells were starved overnight in serum free medium. At4h before lysis the cells were treated with 1µM AZD6244 (Selleckchem, S1008) or solvent control DMSO and at 20 minutes before lysis cells were stimulated with ligands, full serum (10% FBS) or solvent control PBS/BSA (n=4 replicates). We used the following ligands (all Peprotech): EGF (25ng/ml), HGF (50ng/ml), IGF1 (100ng/ml), FGF2 (5ng/ml), PDGF (10ng/ml), VEGF-B (100ng/ml) and VEGF-C (100ng/ml). After treatment and incubation, lysates were collected and analyzed with the Bio‐Plex Protein Array system (Bio‐Rad, Hercules, CA) as described earlier using magnetic beads specific for AKTS473, ERK1/2T202,Y204/T185,Y187 and MEK1S217,S221. The beads and detection antibodies were diluted 1:3. For data acquisition, the Bio‐Plex Manager software and the R package lxb was used.
Bio-Plex data processing
First, obvious outliers among replicates exhibiting an absolute z-score >=3 for all three phosphosite measurements were removed. For each cell-line, data were processed separately for each of the 3 measured phosphoproteins. The value of the control (PBS/BSA+DMSO) was estimated as the mean value of the replicates and log2 fold changes with respect to the control were then computed for all conditions. The resulting fold changes x where then used to calculate the hyperactivation effect of GF and AZD on pAKT: AKTinteraction_FC = µ(xGF+AZD)-µ(xGF+DMSO) - µ(xBSA+AZD)-µ(xBSA+DMSO).
To estimate the significance of this hyperactivation we conducted a two-way anova analysis with interaction term and ascribed synergistic hyperactivation if meeting the following three criteria: (i) significance (p<=0.05), (ii) synergy (AKTinteraction_FC>0) and (iii) receptor dependency (AKTinteraction_FC(GF) > AKTinteraction_FC(PBS/BSA) in the same cell line).
SPIED-DIA analysis
Raw file processing
For the label-free analysis of the raw files, the raw files were processed using DIA-NN (v1.8.2 beta 11), with searches conducted against the library derived from the target peptides only (generated as described above) and reannotation enabled. Settings included methionine excision and in silico digestion at K/R, with cysteine carbamidomethylation as a fixed modification. Variable modifications included methionine oxidation, N-terminal acetylation, and phosphorylation on STY, with phosphorylation scored independently. The analysis allowed for one missed cleavage and a maximum of three variable modifications. The “report-lib-info” option was activated to facilitate raw data verification in subsequent stages. SILAC labeling with a mass delta of 0 at KR was applied as a fixed modification, SILAC channels L (K[0], R[0]), H (K[8.0142], R[10.0083]), and a decoy (K[16.0284], R[20.0165]) were registered.
Process DIA-NN output
Data are filtered to only include Heavy channel entries (spike-in), Channel.Q.Value < 0.05, PTM.Q.Value < 0.05, PTM.Site.Confidence > 0.5 and a Channel.H > 1000 (spike-in intensity). Furthermore, precursors need to have Channel.L (light, endogenous intensity) > 900 in at least 1 condition to ensure no noise-to-noise comparisons and the light precursor needs to be identified with a Channel.Q.Value < 0.5 in at least 3/12 samples. Subsequently, for the precursors passing the filters, a “rescalingfactor” is calculated using the median of Channel H intensities, and light intensities are rescaled to log10-transformed ratios of Channel L to Channel H, adjusted by this factor like such: log10((Channel.L/Channel.H)*rescalingfactor. This transformation is applied to mitigate intensity disparities and inspired by the RefQuant approach30
The rescaled intensities are normalized using the normaliseCyclicLoess function from the limma package. The differential abundance analysis is performed as described in the label-free data analysis pipeline. Precursors are grouped by unique phosphopeptide sequence and filtered for precursors with the lowest F-test p value. Precursors with an F test p value < 0.1 or 0.2 (indicated below relevant figures) are selected for visualization in a heatmap.
Visualization raw data
For the visualization of the raw MS/MS spectra, which facilitates validation of phosphorylation site localisation and the identification of stable isotope-labeled fragments, we employed the following procedure: the ScanID was retrieved from the DIA-NN output table. To determine the corresponding exact scan number from the Bruker raw file, we treated the approximate scan number as the absolute number of MS/MS scans within the run. The exact scan number was then directly derived from the raw data file itself. Subsequently, the MS/MS spectra were downloaded using the Bruker Data Analysis tool. Relevant peaks within the spectra were manually annotated in R using the spectral library as used for the DIA-NN analysis within a 10 ppm mass accuracy range and plotted.
Label-free DIA analysis
Raw file processing
For the label-free analysis of the raw files, the raw files were processed using DIA-NN (v1.8.2 beta 11), with searches conducted against the high pH library (generated as described above) and reannotation enabled. Settings included methionine excision and in silico digestion at K/R, with cysteine carbamidomethylation as a fixed modification. Variable modifications included methionine oxidation, N-terminal acetylation, and phosphorylation on STY, with phosphorylation scored independently. The analysis allowed for one missed cleavage and a maximum of three variable modifications. The “report-lib-info” option was activated to facilitate raw data verification in subsequent stages.
Filter and normalise DIA-NN output
DIA-NN output was processed in R (v4.3.0) filtered with Q.Value < 0.05, only phosphorylated precursors, PTM.Site.Confidence > 0.5 and PTM.Q.Value < 0.05 (Fig. S1B). Precursor intensities (Ms1.Area) were log10-transformed and collectively normalised to correct for loading bias between samples using loess (function: normalizeCyclicLoess) from the limma package (v3.56.1)69. No imputation was performed at any stage in the analysis. PCA was performed on precursors identified in every sample within a group (all cell lines together or individual cell lines).
Differential abundance analysis
Differential abundance analysis of phosphopeptides within cell lines, across conditions, was conducted using the limma package, employing a factorial analysis approach with MEKi and GFmix as factors in the linear model. Precursors were filtered to include only those with a maximum of five missing values. Within the factorial design, contrasts were strategically defined to investigate synergistic effects: the differential impact of the growth factor mix with and without MEKi ("GFmix w MEKi" and "GFmix w/o MEKi"), and conversely, the effect of MEKi with and without the growth factor mix ("MEKi w GFmix" and "MEKi w/o GFmix"). Potential synergistic interactions were explored through an "Interaction" contrast. A linear model was fitted to the data and Bayesian statistics (ebayes function limma) were then applied to estimate variance among the precursors, employing moderated t-statistics and moderated F-statistics. Results were extracted and aggregated for further analysis. Especially, precursors are grouped per unique phosphopeptide sequence and the precursor with lowest p-value from the moderated F-test is selected for downstream analysis.
Clustering significantly interesting phosphosites and investigation interesting clusters
Precursors with F p-values < 0.05 or 0.1 (as indicated below figure), indicating significant regulation, were selected and displayed in a heatmap. Hierarchical clustering was used to organize the heatmap, with the number of clusters determined manually to best represent the data. For each cluster, means of z-score normalized precursors were calculated. Clusters suggesting synergistic interactions between GFmix and MEKi were specifically identified for further analysis. Kinase signatures from PTMsigDB and iKIP-DB were used to perform kinase overrepresentation analysis via Fisher’s exact test, identifying enriched kinase activities linked to the treatment effects.
PTM-SEA analysis
PTM signature enrichment analysis (PTM-SEA, https://github.com/broadinstitute/ssGSEA2.0) was employed to infer kinase activity from regulated phospho-sites. Precursors with a p-value lower than 0.1 as derived from the limma moderated F-statistics (regulated phosphopeptides) were selected. As input we used signed (according to log2 fold change) -log10-transformed p-values per comparison, derived from moderated t-test. PTM signatures were sourced from PTMsigDB46 (v2.0.0) and iKIP-db47. As unique site identifiers, the 14 amino acid phospho-site flanking sequence window was used. Multiply phosphorylated peptides were split per phosphorylation site. PTM-SEA was run with sample.norm.type set to “none” and weight to “1”.
Growth curves inhibitor combination treatment HCT116 and DLD1
Combination Treatment
The effects of combination treatment were assessed by monitoring cell proliferation and death through live-cell imaging. In validation experiments, HCT116 and DLD1 cell lines were treated with combinations of a MEK inhibitor (AZD6244) and either a JNK inhibitor (JNK-IN-8, S4901, Selleckchem) or a PI3K inhibitor (Pictilisib, GDC-0941, S1065, Selleckchem). Cells were seeded at densities of 4000 cells per well for HCT116 and 2500 cells per well for DLD-1 in 96-well plates and cultured in the described growth medium. Twenty-four hours post-seeding, cells were treated with inhibitor combinations at concentrations of 0 (duplicated for double-negative controls), 0.2, 1, and 5 µM, using a quadratic mixing format. To mitigate edge effects, outer rows were left empty and filled with PBS. Cell growth was monitored for an additional three days post-treatment. Experiments were performed in biological triplicates. For treatments combining MEK and JNK inhibitors, the protocol included an additional condition where the medium was supplemented with growth factors (HGF, EGF, and FGF) at specified concentrations.
Incucyte Live Cell Imaging
Automated phase-contrast and green-fluorescent long-term imaging was conducted using an Incucyte instrument (dual-color model 4459, Incucyte Essen Bioscience) in a standard humidified incubator at 37°C and 5% CO2. Imaging occurred every four hours, capturing four frames per well using a Nikon 10x objective.
Image Processing
Images were processed using Incucyte ZOOM software (2018A) with the manufacturer's default masking settings. Confluence values (percentage of area covered by the confluence mask) were exported for further analysis. Image frame data were individually exported and processed in R.
Growth was assessed in multiple ways: raw growth curves were examined for outliers and excluded from further analysis. To determine changes in doubling time, the doubling time in the 48 hours post-treatment was compared to the baseline (0 µM concentration) within each replicate. The average doubling time for selected concentrations of interest was summarized across all replicates. The growth curves for selected concentrations were normalized to the confluence at time zero (treatment) within each well or image frame.