EGFR controls transcriptional and metabolic rewiring in KRASG12D colorectal cancer

Abstract

Inhibition of the epidermal growth factor receptor (EGFR) shows clinical benefit in metastatic colorectal cancer (CRC) patients, but KRAS-mutations are known to confer resistance. However, recent reports highlight EGFR as a crucial target to be co-inhibited with RAS inhibitors for effective treatment of KRAS mutant CRC. Here, we investigated the tumor cell-intrinsic contribution of EGFR in KRAS^G12D tumors by establishing murine CRC organoids with key CRC mutations (KRAS, APC, TP53) and inducible EGFR deletion. Metabolomic, transcriptomic, and scRNA-analyses revealed that EGFR deletion in KRAS-mutant organoids reduced their phenotypic heterogeneity and activated a distinct cancer-stem-cell/WNT signature associated with reduced cell size and downregulation of major signaling cascades like MAPK, PI3K, and ErbB. This was accompanied by metabolic rewiring with a decrease in glycolytic routing and increased anaplerotic glutaminolysis. Mechanistically, following EGFR loss, Smoc2 was identified as a key upregulated target mediating these phenotypes that could be rescued upon additional Smoc2 deletion. Validation in patient-datasets revealed that the identified signature is associated with better overall survival of RAS mutant CRC patients possibly allowing to predict therapy responses in patients.

Keywords: CRC-organoids; EGFR; KRAS; Metabolism; Stemness-WNT.

Figure 1. EGFR deletion in KRAS^G12D tumor organoids does not affect proliferation but alters metabolite uptake and secretion.

(A) Schematic representation of intestinal-specific CRC organoid generation from GEMMs harboring deletion or oncogene activation in A (Apc^min/+), K (Kras^G12D/+), P (Trp53^Δ), or floxed E (Egfr^fl/fl) alleles, resulting in AKP, AKPE and AP organoids. (B) Doubling rate of AKP, AKPE, and AP organoids (n = 6, 8 or 3, respectively). (C) Flow cytometry analysis of mean fluorescence intensity of Ki67 (n = 3) or percentage of early (AnnexinV +) and late (Annexin V+ and 7AAD +) apoptotic (n = 6, 6 or 4) or EdU-positive (n = 3) cells of AKP, AKPE, and AP organoids. (D) Quantification of absolute glucose levels in supernatant by liquid chromatography–mass spectrometry (LC-MS) analysis (n = 3). (E) Quantification of absolute GLUT1 protein amount between AKP and AKPE organoids (n = 4, 3) (left). GLUT1 protein levels assessed by western blot (right). (F) Mean gray intensity GLUT1 expression levels in membrane or nuclear compartments assessed from immunofluorescence-stained organoids. (G) Paired pyruvate over lactate ratio of respective absolute quantities (n = 3). (H–K) Quantification of absolute pyruvate, lactate, glutamate and glutamine levels in supernatant by LC-MS analysis (n = 3). All data represent mean +/− SEM of at least three biologically independent organoids. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids) or one-way ANOVA. Source data are available online for this figure.

Figure 2. EGFR deletion fuels glutamine anaplerotic diversion.

(A) Experimental design and schematic overview of stable isotope tracing metabolomics studies of AKPE and AKP cultures with U-¹³C-glutamine for 8 h and schematic overview of glycolytic, oxidative glutamine catabolism and fractional contribution of labeled carbons of indicated metabolites. (B) Dot plot of fractional labeling from glutamine of respective metabolites. List of ranked metabolites according to percentage ¹³C-glutamine labeling provided as Dataset EV1. (C) Glutamine tracer metabolomics in AKPE (n = 3, biological) or paired AKP (n = 3, biological) organoids (technical replicates 4 per organoid line) showing absolute abundance and fractional labeling of glutamine, glutamate, 2-hydroxyglutarate (2-HG), alpha-ketoglutarate (aKG), succinic acid, fumarate, malate and citrate. M + 0 (all carbons unlabeled) to M + n isotopologues indicate number of ¹³C atoms present in respective metabolite. Total abundance normalized to protein (BCA) content. (D) Representative oxygen consumption rate (OCR) measurement of AKP and AKPE organoids (n = 3, biological) obtained by the Mito Stress test of the Seahorse XF analysis. Oligomycin, FCCP, Rotenone (Rot) and antimycin-A (A) were added at indicated timepoints. (E) Quantification of mean intensity per cell enzyme activity of glucose-6-phosphate dehydrogenase (G6PD) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and representative images of one representative biological replicate. Scale bars, 50 µm. (F) Simplified representation of glycolysis, TCA and glutaminolysis. Area of pie-charts represents percentage of abundance of respective metabolites. Purple, turquoise, gray represent fractional labeling coming from U-13C glucose, U-13C glutamine or other sources, respectively. aKG alpha-ketoglutarate, G6P Glucose-6-phosphate, 2-HG 2-hydroxyglutarate, TCA tricarboxylic acid. All data represent mean +/− SEM. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids). Source data are available online for this figure.

Figure 3. EGFR deletion invokes a distinct transcriptional signature in KRAS^G12D cells.

(A) Principal component analysis (PCA) of RNA-sequencing data of AKP, AKPE and AP organoids in steady state. Each subpopulation is depicted by indicated color. (B) Heatmap showing expression of differentially expressed genes (DEGs) in AKP, AKPE and AP organoids at steady state. Each column represents one organoid line derived from one mouse. (C) Volcano plot representing up or down-regulated gene expression for AKPE versus AKP organoids (n = 3). (D) Heatmap showing subset of differentially expressed genes of metabolic RECON3 curated gene list in AKP, AKPE and AP organoids. (E) Gene set enrichment analysis (GSEA) showing MSigDB KEGG pathways significantly enriched in AKPE versus AKP ranked by their normalized enrichment score (NES).

Figure 4. EGFR deletion results in distinct WNT and stemness transcriptional program.

(A) Heatmap showing expression of genes uniquely upregulated in AKPE compared to AKP and AP organoids categorized according to molecular function. (B) Inference analysis of RNA-seq data with PROGENy. Pathway activity scores are z-scaled und hierarchical clustering was performed by Euclidean distance and complete linkage. (C) Hierarchical clustering of the WGCNA derived significant gene modules: AP-associated genes (gray) and AKPE-associated genes (red). (D) Heatmap showing expression of upregulated significant AKPE-associated genes categorized by WNT pathways or stem cell signaling. (E) Dot plot representing up or down-regulated gene expression of AKPE-associated genes according to fold-change, colored by FDR. (F) Westernblot analysis of WNT pathway proteins of the three biologically derived organoids used in this study. (G) Quantification of LEF1 and β-catenin protein expression (n = 3). (H) Representative SMOC2 protein expression assessed by western blot. (I) RT-qPCR analysis of Egfr, Smoc2, Lef1 mRNA expression in AKP, AKPE and 24/48h- 4-OHT treated AKP organoids (n = 3, 3, 2 or 2, respectively). All data represent mean +/− SEM. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids) or One-way ANOVA. Source data are available online for this figure.

Figure 5. EGFR deletion alters cell size, while Smoc2 knockout rescues all phenotypes induced by EGFR loss.

(A) Kinetic quantification of organoid area from AKP and AKPE organoids cultured for 84 h (n = 3) and representative images at 78 h. Scale bars, 800 µm. (B) Flow cytometry of mean forward scatter area of AKP, AKPE, and AP organoids (n = 4, 6, or 7, respectively). Representative histogram and FSC-A versus SSC-A plots of AKP or AKPE (black or red outlined, respectively). (C) Mean forward scatter area assessed by flow cytometry of AKP, AKP plus 5 µM Erlotinib and AKPE (n = 3) and of (D) AKP, AKP + Smoc2^KO or AKPE and AKPE + Smoc2^KO organoids (n = 5). (E) Principal component analysis (PCA) of RNA-sequencing data of AKP + Smoc2^KO or AKPE and AKPE + Smoc2^KO organoids in steady state. Each subpopulation is depicted by indicated color (upper). Euclidean distance between samples of first and second principal component between annotated comparisons, depicting comparison of each sample to all other samples (lower) (n = 9). Horizontal lines denote the median and dots the mean. The box limits indicate 25th and 75th percentiles, whiskers extend to 1.5× of the interquartile range (IQR) from the 25th and 75th percentiles. (F) Heatmap showing expression AKPE-signature genes of AKP, AKP + Smoc2^KO, AKPE, and AKPE + Smoc2^KO organoids in steady state. (G) Flow cytometry analysis (n = 4 or n = 5, respectively) of mean fluorescence intensity of 2-NBDG of AKP, AKP + Smoc2^KO or AKPE, and AKPE + Smoc2^KO organoids. (H) Heatmap showing a subset of metabolic genes of AKP, AKP + Smoc2^KO, or AKPE and AKPE + Smoc2^KO organoids. All data represent mean +/− SEM. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids) or one-way ANOVA. Source data are available online for this figure.

Figure 6. EGFR deletion reshapes cellular heterogeneity and differentiation trajectories in KRAS^G12D tumor organoids.

(A) UMAP representation of single-cell transcriptomes of AKP and AKPE organoids colored according to genotype (n = 11,505 cells). (B) UMAP representation of single-cell transcriptomes of AKP or AKPE colored according to identified subcluster. (C) UMAP subcluster representation of single-cell transcriptomes of AKP or AKPE. (D) Violin plot shows heterogeneity assessed by centroid distance of AKP or AKPE cells computed based on principal components with significant variance contribution. Horizontal lines of integrated boxplot depict mean and box limits indicate 25th and 75th percentiles (n = 11,505 cells). (E) UMAP-based visualization of key marker Egfr, Ctnnb1, and Smoc2 expression or AKPE signature score in AKP and AKPE cells. (F) Bubble plot shows relative expression levels of selected differentiation and stemness genes separated by AKP and AKPE subclusters. Color shows expression and dot size represents the percentage of cells expressing the genes in each cluster. (G) Heatmap shows mean expression of AKP signature genes derived from bulk RNA data (Fig. 4) in AKP and AKPE subclusters. (H, I) UMAP representation of single-cell transcriptomes colored by Cytotrace2 relative potency score or potency category. Stacked bar plots indicating proportion of cells categorized by potency (differentiated, unipotent, oligopotent or multipotent). P-values calculated by Wilcoxon Rank-Sum test. Statistical significance: ****P < 0.0001.

Figure 7. EGFR deletion implicates synergistic sensitivity to KRAS^G12D inhibition by MRXT1133.

(A) Flow cytometry analysis (n = 3) representing percentage of EdU-positive cells from AKP, AKPE or AKP and AKPE treated with control vehicle or MRTX1133 (100 nM); One-way ANOVA. Data represent mean +/− SEM. (B) Violin plots of GSVA scores for AKPE signature in AKP, AKPE or AKP and AKPE treated with MRTX1133 (100 nM) (n = 3). (C) Inference analysis of RNA-seq data with PROGENy. Pathway activity scores are z-scaled and hierarchical clustering was performed by Euclidean distance and complete linkage. (D) Violin plots of GSVA scores for cell cycle genes in AKP, AKPE, or AKP and AKPE treated with MRTX1133 (100 nM) (n = 3). (B, D) Horizontal lines denote the median, box limits indicate 25th and 75th percentiles, and whiskers extend from the hinge to the lowest/largest value no further than 1.5x IQR from the 25th and 75th percentiles. Source data are available online for this figure.

Figure 8. WNT and stemness signatures are present in human colorectal tumors.

(A) Violin plots of GSVA scores in cells or patient samples with high or low EGFR expression for indicated publication (Fig. EV7A) (n = 10 organoids, 20 PDXs or 129 PDOs). (B) Venn diagram of mutually regulated genes between AKPE versus AKP and PDX treated with cetuximab versus non-treated from Leto et al (2023). (C) Violin plots of GSVA scores in cells or patient samples with up-or downregulated EGFR pathway activity for indicated publication (Fig. EV7A) (n = 10, 28 or 165 patients). (A, C) Horizontal lines denote the median, box limits indicate 25th and 75th percentiles and whiskers extend from the hinge to the lowest/largest value no further than 1.5x IQR from the 25th and 75th percentiles. (D, E) Kaplan-Meier survival curves of KRAS^wt or KRAS^mt patients comparing AKPE high or low expressors of indicated cohorts. (F) Correlation of EGFR expression with SMOC2 expression in KRAS^wt or KRAS^mt patients in the TCGA-COAD dataset.

Figure EV1. Establishment and molecular characterization of EGFR-deficient organoids isolated from GEMMs.

(A) Schematic depiction of genetic crossing strategy used for generating GEMMs, from which organoids were derived. (B) Representative FACS gating strategy of AKP (green fluorescent protein (GFP)-positive), Tamoxifen (4-OHT) or Adeno-Cre induced recombination in AKP organoids after passage one (tdTomato and GFP positive) or sorted AKPE organoids (tdTomato positive), two passages after recombination (left). Depiction of genetic dual-reporter cassette (right). (C) PCR verification of genomic DNA from AKP plus vehicle, AKP plus Tamoxifen (4-OHT) or loading control (ctrl) samples for wild-type Egfr^+/+, floxed Egfr^fl/+ or recombined Egfr^Δ alleles. (D) RT-qPCR analysis of Egfr mRNA in AKP, AKPE or AP organoids; n = 4, One-way ANOVA. (E) Absence of EGFR protein in the three independently derived EGFR-deleted AKPE organoids used in this study. (F) Quantification (left) of absolute GLUT1 protein amount between AKP, AKPE, and AP organoids assessed by western blot analysis (right), One-way ANOVA (n = 4, 3 or 4, respectively). (G) RT-qPCR analysis of Slc2a1/Glut1 mRNA expression in AKP or AKPE organoids (n = 7). All data represent mean +/− SEM. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids) or One-way ANOVA.

Figure EV2. Stable isotope tracing and metabolic analysis reveals differential glutamine and glucose metabolism in EGFR-deficient organoids.

(A) Heatmap showing fractional labeling of respective metabolites in AKP or AKPE organoids. The color scale corresponds to the z-score value of relative abundance of the metabolite. (B) Fractional enrichment in ¹³C-glutamine derived isotopologues of indicated metabolites as determined by LC-MS analysis. M + 0 (all carbons unlabeled) to M + n isotopologues indicate number of ¹³C atoms present in respective metabolite (n = 3 biological organoids with 4 technical replicates per organoid line). (C) Schematic overview of glycolytic and oxidative catabolism and fractional contribution of labeled ¹³C carbons of indicated metabolites. (D) Glucose tracer metabolomics in AKPE (n = 2, biological) or AKP (n = 2, biological) organoids (technical replicates 6 per organoid line) showing fold-change at timepoints 15 min and 8 h of fractional labeling of glucose-6-phosphate (G6P), fructose-6-phosphate (F6P), pyruvate. M + 0 (all carbons unlabeled) to M + n isotopologues indicate number of ¹³C atoms present in respective metabolite. Total abundance normalized to protein (BCA) content. (E) Representative extracellular acidification rate (ECAR) measurement of AKP and AKPE organoids (n = 3, biological) obtained by the Mito Stress test of the Seahorse XF analysis. Oligomycin, FCCP, Rotenone (Rot) and antimycin-A (A) were added at indicated timepoints. aKG: alpha-ketoglutarate. G6P: Glucose-6-phosphate. TCA: tricarboxylic acid. All data represent mean +/− SEM. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids).

Figure EV3. Distinct transcriptional changes induced by KRAS^G12D expression and EGFR deletion.

(A) Heatmap showing expression of selected cell cycle progression genes (based on (Avraham-Davidi et al, 2024)) of AKP, AKPE, and AP transcripts. (B) Heatmap showing expression of selected gene transcripts of receptor tyrosine kinases (RTKs) and EGFR ligands in AKPE, AKPE or AP organoids. (C) Principal component analysis (PCA) of metabolic gene subset of AKP, AKPE and AP organoids in steady state. Each subpopulation is depicted by indicated color. (D) Euclidean distance between samples of first and second principal component of metabolic gene subset between annotated comparisons, depicting comparisons of each sample to all other samples (n = 9). Horizontal lines denote the median and dots the mean. The box limits indicate 25th and 75th percentiles, whiskers extend to 1.5× of the interquartile range (IQR) from the 25th and 75th percentiles. (E) Heatmap showing expression of metabolic genes between AKP and AKPE organoids categorized according to molecular metabolic functional pathways. (F) Gene set enrichment analysis (GSEA) showing MSigDB Hallmark pathways significantly enriched in AKPE versus AKP ranked by their normalized enrichment score (NES). All data represent mean +/− SEM. P-values calculated by One-way ANOVA.

Figure EV4. Transcriptomic analysis reveals WNT pathway genes affected in EGFR-deficient colorectal organoids.

(A) Scatter plot of fold changes from differential expression of AKPE versus AKP plotted against AP versus AKP derived fold changes. (B) Pathway enrichment of gene set variation analysis (GSVA) in AKPE organoids ranked by their pathway score. (C) Dendrogram of marker genes obtained by weighted gene correlation network analysis (WGCNA) according to modules. (D) Volcano plot of pathway-level modules derived from WGCNA analysis. (E) Volcano plot representing up or down-regulated gene expression of AKPE-associated genes according to fold-change (n = 3). (F) Heatmap of GSVA analysis of indicated WNT-pathways. (G) Heatmap of selected WNT-receptor and ligand interaction genes. (H) RT-qPCR analysis of Smoc2 and Lef1 mRNA expression in AKP, AKPE and erlotinib treated organoids (n = 3). (I, J) RT-qPCR analysis of Egfr, Smoc2, Lef1 mRNA expression in AKPS or AKPSE organoids (n = 2). All data show mean +/− SEM, or for I data mean +/− SD. P-values calculated by paired, two-tailed t-test (between pairs of AKP and AKPE organoids).

Figure EV5. Characterisation of organoids lacking Smoc2.

(A) Forward scatter mean fluorescent intensity (MFI) and representative histogram of AKPS and AKPSE organoids assessed by flow cytometry. Data show mean +/− SEM. (B) RT-qPCR analysis of Smoc2 mRNA in AKP, AKP + Smoc2^KO or AKPE and AKPE + Smoc2^KO organoids (n = 2). (C) Heatmap showing subset of metabolic genes in AKP, AKPE or AKP and AKPE treated with control vehicle or 10 µM BPTES. (D) Heatmap showing expression of AKPE-signature genes in AKP, AKPE or AKP and AKPE treated with control vehicle or 10 µM BPTES. (E) Heatmap showing a subset of metabolic genes in AKP, AKPE, or AKP and AKPE treated with control vehicle or 3 µM CHIR99021 or 2 µM ICG-001. (F) Heatmap showing expression of AKPE-signature genes in AKP, AKPE, or AKP and AKPE treated with control vehicle or 3 µM CHIR99021 or 2 µM ICG-001.

Figure EV6. Distinct response to KRAS inhibition highlights EGFR loss as primary driver of AKPE gene signature.

(A) Heatmap showing expression of AKPE-signature genes in AKP, AKPE, or AKP and AKPE treated with control vehicle or MRTX1133 (100 nM). (B) Heatmap showing expression of selected cell cycle progression genes in AKP, AKPE, or AKP and AKPE treated with control vehicle or MRTX1133 (100 nM).

Figure EV7. Validation of EGFR–associated AKPE signatures across CRC datasets from mouse organoids and human tumors.

(A) Graphical overview of accessed, publicly available mouse and human CRC datasets. n indicate number of KRAS^mt of all samples. (B) Classification of single cells based on EGFR high or low pathways activity from Qin et al, (n = 10 organoids). (C) Classification of single cells based on EGFR high or low pathways activity from Lee et al, (n = 10 patients). (D) Venn diagram of mutually expressed genes between AKPE versus AKP, EGFR high versus low expressors from Herpers et al PDOs treated with cetuximab or control vehicle. (E) Stratification of TCGA-COAD KRAS^mt patients into high and low EGFR expressors (n = 165 patients). (F) Venn diagram of mutually expressed genes between AKPE versus AKP, EGFR high vs. low expressors from TCGA-COAD and cells with down versus upregulated EGFR pathway signature from Lee et al, . (G) Normalized counts of SMOC2, FZD2, FZD9, WNT6 and IDH3A of TCGA-COAD KRAS^mt patients of EGFR high and low expressors (n = 165 patients). Horizontal lines denote the median, box limits indicate 25th and 75th percentiles and whiskers extend from the hinge to the lowest/largest value no further than 1.5x IQR from the 25th and 75th percentiles. (H) Scatter correlation plots of EGFR versus SMOC2 expression in KRAS^mt single cells of the Joanito et al dataset. (I) Correlation of EGFR versus SMOC2 expression in KRAS^wt or mt patients of the Wetering et al dataset. (J) Correlation of EGFR versus SMOC2 in single cells of AKP organoids in the Qin et al dataset.