Rebound growth of BRAF mutant pediatric glioma cells after MAPKi withdrawal is associated with MAPK reactivation and secretion of microglia-recruiting cytokines

Introduction Patients with pediatric low-grade gliomas (pLGGs), the most common primary brain tumors in children, can often benefit from MAPK inhibitor (MAPKi) treatment. However, rapid tumor regrowth, also referred to as rebound growth, may occur once treatment is stopped, constituting a significant clinical challenge. Methods Four patient-derived pediatric glioma models were investigated to model rebound growth in vitro based on viable cell counts in response to MAPKi treatment and withdrawal. A multi-omics dataset (RNA sequencing and LC-MS/MS based phospho-/proteomics) was generated to investigate possible rebound-driving mechanisms. Following in vitro validation, putative rebound-driving mechanisms were validated in vivo using the BT-40 orthotopic xenograft model. Results Of the tested models, only a BRAFV600E-driven model (BT-40, with additional CDKN2A/Bdel) showed rebound growth upon MAPKi withdrawal. Using this model, we identified a rapid reactivation of the MAPK pathway upon MAPKi withdrawal in vitro, also confirmed in vivo. Furthermore, transient overactivation of key MAPK molecules at transcriptional (e.g. FOS) and phosphorylation (e.g. pMEK) levels, was observed in vitro. Additionally, we detected increased expression and secretion of cytokines (CCL2, CX3CL1, CXCL10 and CCL7) upon MAPKi treatment, maintained during early withdrawal. While increased cytokine expression did not have tumor cell intrinsic effects, presence of these cytokines in conditioned media led to increased attraction of microglia cells in vitro. Conclusion Taken together, these data indicate rapid MAPK reactivation upon MAPKi withdrawal as a tumor cell intrinsic rebound-driving mechanism. Furthermore, increased secretion of microglia-recruiting cytokines may play a role in treatment response and rebound growth upon withdrawal, warranting further evaluation. Supplementary Information The online version contains supplementary material available at 10.1007/s11060-024-04672-9.

All cells were cultured in a humidified incubator at 37 °C and 5% CO2, dissociated using 0.05% Trypsin-EDTA (cat.no.25300054, ThermoFisher Scientific) and counted with a Vi-CELL XR (Beckman Coulter; Software v2.03) using the settings described in Table S1.
Seeding densities for each cell line for different experiments are listed in Table S2.
Cell lines were authenticated through SNP or STR profiling using Multiplex Cell Authentication by Multiplexion GmbH (Heidelberg, Germany).The purity of cell lines was evaluated using the Multiplex cell Contamination Test by Multiplexion GmbH (Heidelberg, Germany).

Drug treatments and withdrawal
Drug concentrations for MAPKi were chosen based on effect in vitro as measured by a metabolic activity assay in BT-40 (Fig. S1) or a MAPK reporter assay [1] in DKFZ-BT66 [2].
For treatment withdrawal, cells were washed three times with PBS (cat.no.D8537, Sigma-Aldrich), incubated for 15 min in medium at 37 °C, followed by three PBS washes and lastly addition of fresh medium.Time of withdrawal is counted from the timepoint fresh medium is added after the last PBS wash.

Metabolic activity assay for IC50 determination
For metabolic activity assays, BT-40 were seeded in RPMI containing 2% FCS.One day after seeding, cells were treated using the D300e Digital Dispenser (Tecan).72 h after treatment start, metabolic activity was measured using CellTiter-Glo 2.0 (cat.no.G9241, Promega) according to the manufacturer's instructions.Luminescence signal was measured using the FLUOstar OPTIMA automated plate reader (BMG Labtech).IC50/IC75 values were calculated using GraphPad Prism (v8.0.2) with a 4-parameter dose-response model.

Cell counting for growth curve analysis
The mean viable cell number of two technical replicates for each time point and condition was calculated and plotted.Doubling time (DT) was calculated as follows: where t refers to the time between the first cell count and the last cell count and Xstart and Xend refer to the viable cell numbers at the first and last cell count respectively.
For the DT calculation of untreated cells, time-frames for calculation where chose to lie within the exponential growth phase of the cells (i.e.before plateau of the growth curve due to contact inhibition)

RNA isolation, cDNA synthesis and quantitative reverse transcription real-time PCR (RT-qPCR)
RNA extraction was performed using the RNeasy Mini Kit (cat.no.74104, Qiagen) with on-column DNase digestion according to the manufacturer's instructions.In case of tumor tissue samples, the TissueLyser II (Qiagen) was used for mechanical tissue dissociation and lysis according to the manufacturer's instructions.cDNA synthesis was done using the RevertAid First Strand cDNA Synthesis Kit (cat.no.K1622, ThermoFisher Scientific) following the manufacturer's protocol.qPCR was performed as described previously [8] using an ABI 7500 Real Time PCR cycler (Applied Biosystems) with ABI 7500 Software v2.3 (Applied Biosystems) and qPCR Mastermix for SYBR® Green I (cat.no.4309155, ThermoFisher Scientific).The ΔΔCt method was used for relative quantification.ACTB and TBP were used as housekeeping genes for all in vitro samples.
For in vivo xenograft tumor samples only ACTB was used.In case genes of interest were undetected in some samples, Ct values for these samples were set to 40 (max.number of cycles).

Protein extraction and immunoblotting
Cells were lysed in SDS-Buffer containing PhosSTOP phosphatase inhibitors (cat.no.49068450001, Sigma Aldrich) and cOmplete TM mini proteinase inhibitors (cat.no.11836153001, Sigma Aldrich).Protein concentration was measured with the Pierce TM BCA Protein Assay Kit (cat. no. 23227, ThermoFisher Scientific) using the FLUOstar OPTIMA automated plate reader (BMG Labtech).Gel electrophoreses was performed using 7% or 10% acrylamide gels.Proteins were transferred to a PVDF membrane using the Trans-Blot Turbo RTA Mini 0.45 µM LF PVDF Transfer Kit (cat.no.1704274, Biorad) with the Trans-Blot Turbo Transfer System (Biorad).

RNA sequencing and data processing
RNA sample integrity after isolation was assessed using the 2100 Bioanalyzer (Aligent).
Sequencing libraries were prepared using the Illumina TruSeq mRNA stranded Kit following the manufacturer's instructions.Briefly, mRNA was purified from 500 ng of total RNA using oligo(dT) beads.Then poly(A)+ RNA was fragmented to 150 bp and converted to cDNA.The cDNA fragments were then end-repaired, adenylated on the 3′ end, adapter ligated and amplified with 15 cycles of PCR.The final libraries were validated using Qubit (Invitrogen) and TapeStation (Agilent Technologies).2x 100 bp paired-end sequencing was performed on the Illumina NovaSeq 6000 according to the manufacturer's protocol.FASTq files were then submitted to the OTP RNAseq pipeline [9] for read-trimming, alignment to the hg19/GRCh37 human genome and calculation of raw read counts, TPM and FPKM values.Afterwards, data was sorted to only include protein-coding genes.Furthermore, lowly expressed genes (TPM<1 in all samples) were excluded.

LC-MS/MS proteomics and phosphoproteomics data generation and processing
Cell were lysed in 4% SDC buffer, boiled at 95°C and sonicated using a tip probe sonicator (1s pulses, 40% power for 1min).Protein concentration was measured with the Pierce TM BCA Protein Assay Kit (cat.no.23227, ThermoFisher Scientific) using the FLUOstar OPTIMA automated plate reader (BMG Labtech).
For proteomics analysis, 50 µg protein was used.400 mM 2-Chloracetamide, 100 mM Tris-(2carboxyethyl)-phosphin hydrogen chloride and 400 mM potassium hydroxide were used to reduce and alkylate cysteine residues.Proteins were digested using Trypsin/Lys-C mix (cat.no.V5072, Promega) overnight (16-18 h) at 37°C and 1400 rpm, followed by inactivation using 1% TFA.20 µg of digested protein was then used for peptide clean up on self-made SDB-PRS stage tips, prepared by stuffing a 200 µL pipette tip (without filter) with 3 layers of an SDB-RPS extraction disk (Merck) using a modified blunt-end syringe (5 mL, 14 gauge).For centrifugation, a 3D-printed adapter was used.Stage tips were equilibrated prior to peptide loading using acetonitrile, followed by 30% methanol, followed by 0.2% trifluoroacetic acid in water.After peptide loading, washing was performed using 1% trifluoroacetic acid in 2-propanol followed by 0.2% trifluoroacetic acid in water.Finally, peptides were eluted using 80% acetonitrile and 1.25% ammonium hydroxide in H2O, dried using a vacuum concentrator at 45 °C and suspended in 2.5% 1,1,1,3,3,3-Hexafluoro-2-propanol and 0.1% trifluoroacetic acid in H2O.Peptide concentration was determined using the Pierce™ Quantitative Colorimetric Peptide Assay Kit (cat.no.23275, ThermoFisher Scientific) after sonication in an ultrasound bath (15 min) and a total amount of 0.5 µg was subjected to LC-MS/MS analysis.
For phosphoproteomics analysis, 500 µg protein was used.10 mM DTT and 500 mM iodacetamide were used to reduce disulfide-bonds and alkylate cysteine residues.Samples were cleaned by acetone precipitation before protein digestion overnight (16-18 h) using Trypsin/Lys-C mix (cat.no.V5072, Promega).Trypsin was inactivated using formic acid and peptides were dried using a vacuum concentrator at 45 °C.Afterwards, phosphopeptides were enriched using the High-Select™ TiO2 Phosphopeptide Enrichment Kit (cat.no.A32993, ThermoFisher according to the manufacturer's instructions.After enrichment, peptides were dried using a vacuum concentrator at 45 °C, resuspended in 2.5% 1,1,1,3,3,3-Hexafluoro-2-propanol and 0.1% trifluoroacetic acid in H2O and subjected to LC-MS/MS analysis Prior to MS/MS analysis using a timsTOF pro mass spectrometer (Bruker), samples were separated by a NanoElute HPLC system using a 90 min gradient.Ion accumulation and ramp time were set to 50 ms, ions with a mobility ranging from 1/K0 = 0.85-1.3V s cm -2 were included.
Precursors reaching an intensity threshold of 1500 arbitrary units (a.u.) were classified as suitable.
Resequencing of low-abundance precursors was performed, taking dynamic exclusion of 40 s into account, until a value of 20000 a.u. was reached.Ions with a mass range = 100-1700 m/z were selected for MS/MS fragmentation.A 2 Th window or a 3 Th window was used for ions with m/z < 700 or mz > 700 respectively.Quadropole switching events were synchronized with the precursor elution profile for isolation.Collision energy for dissociation was lowered linear as a function of increasing ion mobility, ranging from 59 eV (1/K0 = 1.6 V s cm -2 ) to 20 eV (1/K0 = 0.6 V s cm -2 ).Single charged precursor ions were excluded using a polygon filter.
Raw MS data was processed by the commandline version of the MaxQuant software (v1.6.17.0) [10] on a Linux machine with 128 physical cores (AMD EPIC 75032 32-Core Processor) and 256Gb of RAM.Spectra were searched against the human Uniprot database of canonical protein sequences downloaded in March 2022.Parameters including enzyme specificity, FDR on peptide spectral match, protein level precursor as well as fragment ion mass tolerance remained on default settings.For whole proteome analysis the variable modifications Aminoacid deamidiation (NQ), Oxidation (M) and Acetyl (Protein N-term) and MaxQuant internal normalization algorithm MaxLFQ and the search algorithm "Match between runs" were turned on.For the Phosphoproteome analysis serine, threonine and tyrosine phosphorylation (STY) was included to the variable modification.Furthermore, MaxLFQ and the search algorithm between runs" remained off.
Data was filtered to exclude proteins only identified by a modification site, reverse hits or possible contaminants.Additionally, proteins not detected in at least 2/3 replicates per condition were excluded.Missing data was imputed through random draws from a gaussian distribution centered around a minimal value.
Phosphoproteomics data was further processed before analysis using the R package "PhosR" (v1.6.0)[12,13].Data was filtered to exclude reverse hits and possible contaminants.Additionally, peptides not detected in at least 2/3 replicates per condition were excluded.Missing data was imputed through site-and condition-specific as well as tail-based imputation.Lastly, phosphoproteomics data was batch-corrected using a set of stably phosphorylated sites [12,13] and implementation of the Removing Unwanted Variation-III method [14].

Luminex-based multiplex assay
Conditioned media (CM) for each condition and timepoint was collected from 3 dishes, yielding approx.9 mL of CM per condition.Cells were counted at the time of CM harvesting using the Vi-CELL XR (Beckman Coulter; Software v2.03) using the settings described in Table S1.CM was centrifuged at 1500 rpm for 15 min to clear cellular debris and stored at -80 °C until use.
Protein concentrations in pg/ml were calculated and then normalized to viable cell counts to account for differences in cell density across the different conditions and timepoints.
ssGSEA analysis was performed using the ssGSEA module (version 10.1.0)[18] on GenePattern [19] using default settings except for min.gene set size which was set to 1. Gene sets related to proliferation and cell cycle were taken from the MSigDB (v7.5.1)C2 subcollection CP (Broad Institute).
MAPK activity scores based on phosphoproteomics data were calculated using the ssGSEA module (version 10.1.0)[18] on GenePattern [19] using default settings except for min.gene set size which was set to 1.The MEK1 PTM-SEA signature, containing protein phosphorylation sites phosphorylated by MEK1, was taken from the PTMsigDB collection (v2.0.0) [21].ssGSEA scores were not normalized and are therefore shown as measured in arbitrary units.
The MPAS score, consisting of ten genes shown to be downregulated upon MEK inhibition, was calculated using FPKM values as previously described [22].MPAS scores were not normalized and are therefore shown as measured in arbitrary units.
MAPK activity scores based on proteomics data were calculated using the ssGSEA module (version 10.1.0)[18] on GenePattern [19] using default settings except for min.gene set size which was set to 1.As MAPK ssGSEA score a set of proteins shown to be downregulated upon MEK inhibition was used [23].ssGSEA scores were not normalized and are therefore shown as measured in arbitrary units.
Longitudinal k-means clustering, an adaptation of the k-means clustering method, was performed using the R package "kml" (v2.4.6) [24] with 20 iterations.Up to 20 clusters were tested and the number of clusters with the highest Calinski-Harabasz index [25] was selected.Only genes, proteins and phospho-peptides with an adjusted p-value < 0.01 for at least one timepoint were included in the analysis.Differentially regulated clusters, used for further analysis, were defined based on the cluster mean log2FC with a cut-off of ±1.5.
Six-week-old female NOD scid gamma mice (NSG), purchased from Charles River UK, were used to establish intracranial xenografts.BT-40 cells (1.5x10 5 per mouse) were injected into the forebrain using a Hamilton syringe, at the coordinates: bregma + 1 mm anterior, 1.5 mm lateral and 3 mm ventral.Bioluminescence imaging (Firefly D-Luciferin, s.c.150mg/kg -PerkinElmer # 122799; IVIS Lumina III In Vivo Imaging System -PerkinElmer) was used to monitor intracranial tumour growth.Treatment of animals was started for each mouse individually once bioluminescence signal reached a radiance of >10 7photons/sec/cm 2 /sr.Animals were treated