RNA-Seq data

RNA-Seq data from primary and metastatic CRC samples, PDOs, and PDXs were used as available from the previous study [31]. This included, for primary CRC samples processed expression data from 88 tumor tissues and 22 normal liver tissues. For metastatic CRC samples data from 23 tumor tissues and 12 normal liver tissues. RNA-seq expression counts had been normalized as reads per kilobase of transcript per million mapped reads (RPKM). RPKM values were used for further analyses as done in the original publication. RPKM values of genes of interest were tested for normal distribution and then an either parametric or non-parametric test was used to test for differential expression between groups or genes within the same tissue/organ.

Tumor tissue

Tumor material was received with informed consent from 40 CRC patients from the St. John of God Hospital and the University Hospital Graz together with clinical information under approval by the ethics committees of the Medical University Graz and St. John of God Hospital (23–15 ex 1/11) [32]. Pathological processing was performed at the Department of Pathology, Hospital Graz II and the Diagnostic & Research Institute of Pathology of the Medical University Graz and pathological data were provided. All samples used for IHC analysis were also part of the patient cohort published by Schütte et al. [31].

Analyses of PDX and PDO data

Animal experiments as well as organoid culture was performed in the original study [31]. Data on tumor growth of PDX under the respective treatments or vehicle control were used for correlation analyses. For analysis of the resistance data of the PDO samples, log10 IC50 values, which can be found in Supplementary Data 14 of Schütte et al. [31], were used.

Immunohistochemistry (IHC)

IHC was performed on a Ventana Immunostainer XT (Ventana Medical Systems, Tucson, USA), using an ultra-VIEW universal DAB Detection Kit (Ventana Medical Systems, Tucson, USA) and cell conditioning solution for 30 min using heat induced epitope retrieval. The primary customized IGF2BP2/IMP2/p62 antibody [33, 34] was incubated for 30 min using a dilution of 1:1000.

Two observers (N. GS., J. H.), blinded to the clinical data related to the respective cases, evaluated all stains independently on a light microscope. IGF2BP2 expression was evaluated with respect to staining intensity (intensity score 0–3; 0 no staining, 1 weak, 2 moderate and 3 strong) and percentage of positive cells (proportion score; 0–100%). In discrepant cases, the average score was taken into account. The same scoring strategy has been used in previous published studies [35, 36].

Cell culture

HCT116, which are human colorectal carcinoma cells, were maintained in Dulbecco’s modified Eagle’s medium (DMEM, #21969-035, gibco). Medium was supplemented with 10% fetal calf serum (FCS, #P30-3306, PAN-Biotech), 1 mM glutamine (#X0551-100, Biowest), 100 U/mL penicillin, and 100 μg/mL streptomycin (#15070-063, Gibco). The cells were cultured at 37 °C and 5% CO2. Cell line authentication was conducted by STR/DNA profiling. For cell line authentification the Cell line Authentification Service by Eurofins Genomics was used. DNA isolation was carried out from cell pellet. Genetic characteristics were determined by PCR-single-locus-technology. 16 independent STR loci D8S1179, D21S11, D7S820, CSF1PO, D3S1358, TH01, D13S317, D16S539, D2S1338, AMEL, D5S818, FGA, D19S433, vWA, TPOX, and D18S51 were investigated using the AmpFlSTR® Identifiler® Plus PCR Amplification Kit (Thermo Fisher). In parallel, positive and negative controls were carried out yielding correct results. Results were compared to a reference sample and known profiles. Mycoplasma testing was performed regularly via PCR. For that purpose, cells were grown to confluence over several days in media without antibiotics. A PCR was performed with the supernatant of the cells (Primer forward 5´-3´: GGCGAATGGGTGAGTAACACG, Primer reward 5´-3´: CGGATAACGCTTGCGACCTATG). Positive samples reveal a distinct 500 bp band on the agarose gel.

CRISPR-mediated IGF2BP2 knockout cells have been previously published: all experiments in the current study were performed with cell clone KO#1 from the previous study [17]. Confirmation of IGF2BP2 KO was confirmed on a regular basis by Western blot analysis (Supplementary methods, Supplementary fig. S1). The overexpression plasmid and the control vector were previously published [15] and are deposited with Addgene (www.addgene.org; plasmid 42175, plasmid 42174). HCT116 cells were transfected with Lipofectamin 3000 (#L3000015, Thermo Fischer) according manufacturer instructions using 0.5 µl Lipofectamin 3000 and 500 ng plasmid DNA per well of a 24-well plate. For OCR/ECAR measurements cells were transfected 24 h post seeding to a 24-well plate. After two days of cultivation, cells were trypsinized and seeded to Seahorse XF96 V3 PS Cell Culture Microplates (#101085-004, Agilent) and experiments were conducted as described below.

2D and 3D Proliferation

The cells were treated with 5-fluorouracil (22.2 µM, stock 40 mM in DMSO, #F6627, Sigma Aldrich), oxaliplatin (2.6 µM, stock 4 mM in PBS, #Y0000271, European Directorate of the Quality of Medicines & HealthCare), gefitinib (23.5 µM, stock 50 mM in DMSO, #G-4408, LC Laboratories), regorafenib (8.1 µM, stock 50 mM in DMSO, #R-8024, LC Laboratories), selumitinib (7.4 µM, stock 50 mM in DMSO, #S-4490, LC Laboratories) or nintedanib (8.2 µM, stock 25 mM in DMSO, #N-9077, LC Laboratories). Stock solutions were stored in aliquots at -20 °C and used within one month after preparation. The concentrations used were determined as IC50 values for HCT116 wildtype by a resazurin assay. For this viability assay, cells were seeded and 24 h later treated with a concentration spectrum of the respective compound. After 48 h of treatment, cells were incubated with resazurin (0.02 mg/ml, stock 10× in PBS, #R7017, Sigma Aldrich) for 4 h. Detection was performed as previously described [37].

For the 2D kinetic proliferation analysis, 5000 cells were seeded per well into 96-well plates. The next day, cells were treated with the respective compounds, and cell confluency was monitored in an IncuCyte® S3 system for a period of 4 days. Cell confluency was analyzed using IncuCyte® basic analyzer software. The confluency was normalized to the time of treatment. The normalized values were substracted from the proper vehicle control.

For the 3D proliferation analysis, 3000 cells were seeded per well into low-attachment U-bottom 96-well plates. After spheroid formation for 3 days, spheroids were treated with the respective compounds, and monitoring in an IncuCyte® S3 system was started. The spheroid area was analyzed using the spheroid IncuCyte® software, and the area was normalized to the first measuring time point after treatment (0 h).

CLIP data analysis and gProfiler enrichment analysis

Publicly available enhanced CLIP (eCLIP) RNA-binding data from the ENCODE consortium for HepG2 liver cancer cells and K562 cells was used [38]. eCLIP peaks were obtained from the ENCODE data portal (https://www.encodeproject.org/) and analyzed as previously published [39]. Each annotated human gene in the Ensembl database, that had an eCLIP peak in at least one of the two cell lines, was denoted as an IGF2BP2 target gene (IGF2BP +). We used a Kolmogorov-Smirnow test on the distribution of all genes and IGF2BP + or IGF2BP- genes to compute significant changes in log fold changes using R. In order to obtain genes that are differentially expressed between normal and cancer or metastastic samples, we used the Mann-Whitney U Test. Resulting gene p-values were corrected for multiple testing using an FDR-based approach [40]. We used all differentially expressed target genes that had a log fold change of > 0.5 for the gene set enrichment analysis with g:Profiler [41].

OCR/ECAR measurement

The Mito Stress Test and the ATP Rate Assay were performed using an Agilent Seahorse 96XF device and respective kits. The assays were performed as described in the manufacturer’s protocol (#103592-100, #103015-100, Agilent). In brief, the cells were seeded 24 h before measurement. The medium was replaced one hour prior to measuring by seahorse XF DMEM assay medium (#103575-100, Agilent) supplemented by 10 mM glucose (#103577-100, Agilent), 1 mM sodium pyruvate (#103578-100, Agilent) and 2 mM glutamine (#103579-100, Agilent). For the Mito Stress Test, cells were treated with 2 μM oligomycin, 0.5 μM FCPP, 0.5 μM rotenone/antimycin A (Seahorse XF Cell Mito Stress Test Kit, #103015-100, Agilent) and 4 µM Hoechst (#62249, Thermo Scientific). The ATP Rate Assay included adding of 2 μM oligomycin followed by 0.5 μM rotenone/antimycin A and Hoechst to the cells. The data were analyzed by the Seahorse Wave Software (Agilent Technologies, Santa Clara, CA, USA). After measurements with the Seahorse XF Analyzer, cells were counted using the Bio Tek Cytation 5 Cell Imaging Multimode Reader for normalization by Hoechst staining.

qPCR

Total RNA was isolated using the High Pure RNA Isolation Kit (#11828665001, Roche). Concentration of isolated RNA was quantified by NanoDrop (Thermo Fisher Scientific), and RNA with an A260/A280 ratio higher than 1.7 was used for further experiments. RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (#4368813, Thermo Fisher Scientific) in the presence of an RNase inhibitor (#10777-019, Invitrogen) according to the manufacturer’s instructions. cDNA was analyzed using 5 × HotFirePol EvaGreen qPCR Mix (#08-25-00020, Solis BioDyne) with a CFX96 real-time PCR system (Bio-Rad). Primer sequences can be found in Supplementary table S1. qPCR data were analyzed with the Bio-Rad CFX Maestro 1.1 Software 2017 (Bio-Rad Laboratories). Reference genes were discriminated based on the GeNorm algorithm [42] which is included in the Maestro software. 18S was found to be more stable in its expression pattern compared to ACTB due to a lower M value (internal control gene-stability measure, M < 0.5). Consequently, data were normalized to 18S and are shown relative to the control.

Statistical analysis

Data analysis and statistics were performed using OriginPro. IC50 values were calculated using sigmoidal fitting with Origin pro version 19 software. Data are represented as means ± SEM if not indicated otherwise. Depending on whether the data were normally distributed and on the group size, statistical differences were calculated using one-way or two-way ANOVA, Student’s t-test, or Mann-Whitney U Test. Linear correlation was measured with the Pearson correlation coefficient or Spearman´s rank correlation coefficient. Fisher´s exact test was used in the analysis of contingency tables. #/*p < 0.05; **p < 0.01; ***p < 0.001. Biological replicates were represented as “n” and technical replicates were represented in brackets.