Quantifying single-cell ERK dynamics in colorectal cancer organoids reveals EGFR as an amplifier of oncogenic MAPK pathway signalling

Abstract

Direct targeting of the downstream mitogen-activated protein kinase (MAPK) pathway to suppress extracellular-regulated kinase (ERK) activation in KRAS and BRAF mutant colorectal cancer (CRC) has proven clinically unsuccessful, but promising results have been obtained with combination therapies including epidermal growth factor receptor (EGFR) inhibition. To elucidate the interplay between EGF signalling and ERK activation in tumours, we used patient-derived organoids (PDOs) from KRAS and BRAF mutant CRCs. PDOs resemble in vivo tumours, model treatment response and are compatible with live-cell microscopy. We established real-time, quantitative drug response assessment in PDOs with single-cell resolution, using our improved fluorescence resonance energy transfer (FRET)-based ERK biosensor EKAREN5. We show that oncogene-driven signalling is strikingly limited without EGFR activity and insufficient to sustain full proliferative potential. In PDOs and in vivo, upstream EGFR activity rigorously amplifies signal transduction efficiency in KRAS or BRAF mutant MAPK pathways. Our data provide a mechanistic understanding of the effectivity of EGFR inhibitors within combination therapies against KRAS and BRAF mutant CRC.

Extended Data Fig. 1. CDK1/cyclinB phosphorylates EKAREV(Tq) during G₂- and M-phase

a, Typical mitotic EKAREV-FRET profile in HEK293 cells, including rising phase, steep increase at nuclear envelope breakdown (NEB) and sharp decline at anaphase (Supplementary Movie 1.) Corresponding snapshots of above cells with H2B-mScarlet support cell-cycle phases (20 cells; 1 experiment). 1, G₂; 2, NEB; 3, metaphase; 4, anaphase; 5, cytokinesis (c.k.); 6, G₁. FRET-signal relative to PMA saturation (150nM). Black, FRET-ratio of YPet(yellow)/Turq2(blue) intensities. b, As a, with cell-cycle stages recognized by EKAREV(Tq) biosensor exclusion from condensed chromosomes; consistently observed (53 cells, 4 experiments). c, As a, but EKAREV(Tq) biosensor lacking nuclear localization. Observed in 28 cells, 2 experiments. d, As a, but EKAREV(TA) control biosensor that cannot be phosphorylated. Observed in 5 cells, 1 experiment. e, EKAREV(Tq) FRET signal in mitotic arrested HEK293 cells (nocodazol, 0.83μM; 2hrs) is sensitive to CDK1 inhibitor RO-3306. In mitotic cells, recognized by absence of nuclear localization (NEB) of NLS-tagged biosensor (insert images), FRET decreased upon 10μM RO-3306 (9 cells; 2 experiments with similar results), or 3x 1μM (1 cell). e’, Loss of normalized FRET signal (ΔR, %) upon RO-3306 (10μM) or MAPK pathway inhibitors (sel+SCH, 5μM each). Box-and-whisker plots: boxes represent quartile 2 and 3, horizontal line represents median, whiskers represent minimum and maximum within 1,5x interquartile range. RO-3306: n=23 cells, sel+SCH n=16 cells. f, EKAREV(Tq) FRET signal is sensitive to CDK1-inhibition in G₂-phase. Synchronized cells were imaged before, during and after incubation with RO-3306 (10μM) and retrospectively analyzed if mitotic entry was observed <15 minutes after drug washout (n=23 cells). Right, similar experiment, here inhibiting MEK and ERK (n=20 cells). Graph shows mean±s.d. of baseline-normalized traces. g, As f, monitoring G₂-phase in HeLa cells (n=19) co-expressing ERK-KTR-mCherry and EKAREV(Tq). ERK-KTR biosensor suffers from same undesired CDK1-sensitivity. ***, two-sided student’s T-test, p<0.0005. Scale bar, 10μm.

Extended Data Fig. 2. Improving ERK specificity, generating EKAREN4

a, HeLa cells expressing EKAREV substrate variants Alt_ERK_Substr_1-6 (Extended Data Fig. 4b). PMA, 500nM; SCH, 10μM. Right, responses (mean±s.d.), normalized to EKAREV-GW4.0. b, HeLa cells expressing EKAREV-GW variants were arrested in mitosis (Extended Data Fig. 1e) to test CDK1/cyclinB sensitivity (RO-3306). Purple, mean of individual traces. Right, overview for several variants, mean ratio loss ± s.e.m. Best responder EKAREV-GW(Alt_substr._6) is compromised by RO-sensitivity. c, Repeat of Fig. 1e, quantifying FRET in G₂- and M-phase HEK293 cells (mean ± s.d), complemented with results from third-residue-substitution variants (purple). No improvements compared to EKAREN4/EKAREN5. d, Sensor dephosphorylation kinetics, assessed by instant ERK inactivation (sel+SCH, 5μM) after initial sensor saturation (PMA). For identical experimental conditions, HEK293 stably expressing EKAREV(Tq) or EKAREN4 were mixed. H2B-mScarlet selectively marked EKAREV(Tq) (left) or EKAREN4 cells (right) (insets: scale, 25μm). Cells analyzed individually and averaged after double normalization (baseline and PMA-plateau). e, Maximum FRET range (ΔR(%), approximated through saturation (PMA) in serum-starved HEK293 cells of widely variable expression levels. e’, Baseline and plateau ratios corresponding to cells in e. Increased FRET range of EKAREN4 likely results from elevated plateau ratios. Expression levels affect FRET range by differentially affecting baseline ratios (see slopes in a.u.). Experiment performed twice. f, As e, differential effect of expression level on FRET range is similar for ERK-insensitive control sensors EKAREV(TA) and EKAREN4(TA). g, Mean (± s.d.) ΔR per expression level category (see e). For panels a-g the n numbers represent cells and are indicated in the graph for each group. Box-and-whiskers: boxes represent quartile 2 and 3, horizontal line represents median, whiskers represent minimum and maximum within 1,5x interquartile range. Dots are outliers. P values in all relevant panels were calculated using a two-sided student’s T-test, * p<0.05; ***, p<0.0005. n.s., non-significant.

Extended Data Fig. 3. Multi-dimensional analyses comparing EKAREV, EKAREN4 and EKAREN5

a, FRET-range versus sensor expression level, as in Extended Data Fig. 2e (EKAREN4, n=80 cells; EKAREN5, n=75 cells). a’, baseline and plateau ratios corresponding to cells plotted in a. b, Means (± s.d.) of ΔR from a, calculated in three expression level categories (as in Extended Data Fig. 2g). c, Dephosphorylation kinetics of EKAREN5 were directly compared with EKAREV(Tq) (as Extended Data Fig. 2d). Retrospective unmixing was based on clustering plateau amplitudes (PMA) (see Fig. 2a). Experiment performed once. d, As in c, comparing phosphorylation and dephosphorylation kinetics of EKAREN5 with ~33-fold expression level difference. Co-seeded high and low expressors were simultaneously monitored. Experiment performed three times. e-h, Various automated analyses on autonomous ERK fluctuations of HeLa cells (dataset of Fig. 2f,g), registered simultaneously by ERK-KTR-mCherry and either of EKAREV/EKAREN FRET sensors. EKAREV, n=15 single-cell traces; EKAREN4, n=10 single-cell traces; EKAREN5, n=17 single-cell traces. e, Automated peak counting per individual cell. f, Temporal matching of rising phases in KTR versus FRET signals. g, Temporal matching of falling phases in KTR versus FRET signals. h, Counted ‘inflection’ points per trace, i.e. points where ERK changes accelerate or decelerate(see Methods). i, Correlation between EKAREN5-FRET and ppERK staining (mean nuclear signal). After various ERK manipulations, HeLa-EKAREN5 cells were FRET-imaged and fixed instantly after acquisition, yielding various ERK activity states between complete inhibition (MEKi+ERKi) and pathway saturation (>7 min EGF). Grey line, regression analysis (y=ax+b). Traces are mean ratios ± s.d. For panels a-i the n numbers represent cells and are indicated in the graph for each group. Box-and-whisker plots: boxes represent quartile 2 and 3, horizontal line represents median, whiskers represent minimum and maximum within 1,5x interquartile range. Dots, outliers. Scale bar, 50μm. Two-sided student’s T-tests: *, p<0.05; **, p<0.005; n.s., non-significant.

Extended Data Fig. 4. Biosensor sequences

a, Silent mutations introduced into Turquoise2 (insert) to minimize sequence homology with the YPet fluorophore in the same construct. Red ‘x’ marks silent mutations. Blue, residues discriminating Turquoise2 from parental eCFP (Goedhart et al., 2012). Green, residues rendering Turquoise2 prone to dimerize with YPet, in analogy to EKAREV design (Komatsu et al., 2011). Dark green, the V224L mutation was added to further enhance dimerization and, hence, FRET efficiency in ON-state (Vinkenborg et al., 2007). b, Alternative ERK substrate sequences were derived from ERK targets RSK1 (human) and ELK1 (human) using the Kinexus website and compared to parental EKAREV-GW-4.0 (GW = GateWay) with CDC25C substrate sequence. c, Overview of generated and tested point mutant variants of EKAREV. Red, central Threonine, target of ERK phosphorylation. Blue, the Lysine at position +4 mimics the general CDK1-consensus site, mutated to Proline (K->P). Purple, the Lysine at position +6 mimics the CDK1-consensus site and mutated to bulky Trp (K->W) to create steric hindrance with cyclinB. Underlined, ERK docking domain FQFP. Purple boxed characters, rational attempts to further eliminate CDK1-sensitivity with third amino acid replacements. V422T was aimed at favoring ERK over CDK1 consensus site; L427W was aimed at further augmenting sterical hinderance of cyclinB interaction; L427E was aimed at impeding cyclinB interaction through electrostatic repulsion. The Asp (D) at position+3 was left unchanged for its reported importance for the Pin1 affinity (Komatsu et al., 2011). d, Summary of available EKAREN4 and EKAREN5 plasmid constructs (including variable targeting motifs and Thr-Ala control versions), as well as adapted version of pInducer20 (Meerbrey et al., 2011) to initiate expression of HRAS^N17 and P2A-coupled reporter fluorophore mKate2-NLS.

Extended Data Fig. 5. Geometric effects on raw FRET signals in 3D organoid models

a, PDO model expressing the non-phosphorylatable (hence ERK-insensitive) control sensor EKAREV(TA) was FRET-imaged to assess non-biological geometry effects on raw signals in 3D organoid FRET microscopy. YFP/CFP ratios can differ between organoids situated either far away or close to the objective (distance difference ~150 μm). Experiment performed once, this direct comparison representing general observations. b, Performing FRET acquisition, Turq2 and YPet emissions were determined from all cells of a bulky organoid (>200 cells) and plotted against their z-coordinates. YFP/CFP ratios increase subtly with increasing depth within the organoid, likely due to differential scattering-induced loss of fluorescence between the two fluorophores. Experiment performed twice with same outcome.

Extended Data Fig. 6. Single-cell ERK responses in organoids to various doses of selumetinib

a, As in Fig. 3b, here using selumetinib at 50nM concentration. Shown are 33 single-cell analyses from two p9T organoids. Right, waterfall plot summarizing cellular recovery from ERK inhibition (see also Fig. 3b). b, Exact copy of Fig. 3b, using selumetinib at 200nM concentration. Shown are 59 single-cell analyses from three p9T organoids. Right, waterfall plot summarizing the cellular recovery from ERK inhibition. c, As in Fig. 3b, here using selumetinib at 1μM concentration. Shown are 42 single-cell analyses from three p9T organoids. Right, waterfall plot summarizing the recovery from ERK inhibition. d, Three example FRET traces to illustrate the power of time-resolved signal transduction analysis. Trace colours correspond to bars indicated with coloured arrows in waterfall plot of c. Top, cell displaying onset of recovery, interrupted by super-inhibition. Middle, selumetinib-induced inhibition is followed by a sustained phase without apparent recovery. Bottom, inhibitory effect is more prolonged, explaining negative outcome for Recovery (= ‘Recov’ – ‘Inh’). Grey, green and purple colours correspond with bars in waterfall plot of Extended Data Fig. 6b.

Extended Data Fig. 7. EGFR plays a central role in generating pulsatile ERK dynamics in PDO-KRAS^G12V

a, Fluctuating ERK dynamics in PDO-KRAS^G12V, abolished by afatinib at 50nM concentration. Shown traces are representative for 14 single-cell analyses from 3 PDOs. Experiment performed twice. b, As a, loss of ERK activity oscillations, observed in PDO-KRAS^G12V upon administration of pan-HER inhibitors lapatinib (1μM, 21 cells), dacometinib (500nM, 12 cells), or EGFR-specific inhibitors erlotinib (500nM, 17 cells) and gefinitib (500nM, 18 cells). Two independent experiments performed. ~5 organoids per condition. c, As a, loss of ERK oscillations, observed in PDO-KRAS^G12V upon anti-EGFR antibody cetuximab (500 ng/ml). Residual ERK activity was sensitive to trametinib (MEKi). Experiment performed once; 30 cells analyzed. d, Oscillating ERK dynamics persist in PDO-KRAS^G12V despite HER2-inhibitor CP-724714 (5μM). Three traces representative for 24 single-cell analyses, from one experiment with 4 organoids. e, FRET-trace demonstrating that afatinib (200nM) instantly interrupts rising phase of pulsatile ERK (arrow). Representative for >20 observations in various PDO-lines. f, Shp2-inhibitor SHP099 (5μM) abrogates autonomous ERK activity in PDO-KRAS^G12C, but not BRAF^V600E(#4). Shown are representative multi-cellular z-plane analyses. Experiment performed once; 3 organoids. f’ BRAF^V600E(#4) is similarly unresponsive to drugs targeting Src (KX2, 500nM), FAK (PND-1186, 500nM) and cKit/PDGFR/Bcr-Abl (dovitinib, 100nM; imatinib, 1.0 μM; masitinib, 100nM; pazopanib, 250nM). Shown are 2x-normalized ratios (mean±s.d.). n numbers represent PDOs and are indicated in the graph per group. Experiment performed once. g, Adapted pInducer20 for doxycyclin-inducible expression of HRAS^N17 and P2A-coupled reporter mKate2-NLS (TRE2, Tet-Responsive Element). Western blot demonstrating doxycyclin-mediated induction of HRAS^N17 expression (general anti-RAS antibody) in BRAF^V600E(#3)(EKAREN5+pInducer). S.E., short exposure; L.E., long exposure. Vinculin as loading control. Experiment performed once. h, Western blot analyses on indicated PDOs illustrating pan-HER inhibition on components of the linear EGFR-MAPK-pathway. Pharmacological treatments as in Fig. 4e. Experiment performed once.

Extended Data Fig. 8. Single-cell drug response upon long-term exposure to EGFR inhibitors

a, Single-cell analyses from data set presented in Fig 7b. After 72 hrs afatinib (1μM), EKAREN5 showed constant, non-oscillatory basal ERK signal. Rare activity spikes in few cells were observed in PDO-NRAS^Q61H and BRAF^V600E(#3). b, Single-cell drug response analysis after 8 days of afatinib treatment (1μM). ERK dynamics in EGFR-inhibited PDO-KRAS^G12C (63 cells in 6 organoids) and PDO-NRAS^Q61H (70 cells in 5 organoids) are constant, non-oscillatory in nature. Depletion of the continuous oscillatory dynamics unmasked a hidden pattern of (EGFR-independent) rare activity spikes in few cells (one per ~30 hr (19 pulses observed in 571 hours of single-cell signaling evaluation). Never were two pulses observed in one single-cell derivation. Plots show all single cell analyses from multiple organoids or from a single representative organoid. Dotted line: to indicate that the two pulses are from separate cells.

Extended Data Fig. 9. Examining clinical relevant drug therapy on BRAF mutant PDO

a, Growth assay performed as in Fig. 7c,d, here monitoring the BRAF^V600E mutant PDOs #1, #2 and #3 under treatment with drugs from clinical trial by Kopetz et al. (NEJM, 2019). Cetuximab, 1000 ng/ml; binimetinib (MEK-inh), 100nM; encorafenib (BRAF-inh), 300nM. Mean number of objects per time point: BRAF^V600E(#1), n=84; BRAF^V600E(#2), n=62; BRAF^V600E(#3), n=78. For the exact n-numbers of all presented points, see Source Data file. Data are represented as mean size ± s.e.m. Growth medium was supplemented with 200x reduced EGF concentration (0.25 ng/ml) to minimize competition between cetuximab and EGF ligand. Experiment performed once. b, In same PDO lines, ERK responses were recorded using EKAREN5-FRET. Data are represented as mean value ± s.d. Q uantification scheme based on double calibrated multi-cellular z-plane analyses (n numbers represent PDOs and are indicated in the graph for each group). ERK levels were maximally reduced in presence of triple combination. Box-and-whisker plots: boxes represent quartile 2 and 3, horizontal line represents median, whiskers represent minimum and maximum within 1,5x interquartile range. Dots are outliers.

Fig. 1. Minimizing cell cycle-dependent influences on EKAREV(Tq)-FRET

a, Insertion of Turquoise2 cDNA, including 169 silent mutations (Extended Data Fig. 4a), into EKAREV. FRET range of serum-starved HEK293 cells expressing parental EKAREV (n=55 cells) versus EKAREV(Tq) (n=61), approximated by PMA-induced sensor saturation (ΔR,%)), remained unaffected. 2 independent experiments. Right, EKAREV(Tq) reports spontaneous fluctuations in HeLa cells. b, Typical mitotic EKAREV profile in HEK293 cells persists under selumetinib (MEKi) and SCH772984 (ERKi) (both 5μM). Blue and yellow traces, Turq2 and YPet fluorescence intensities. Black, ratio YPet/Turq2. Loss of fluorescence indicates Nuclear Envelope Breakdown (NEB). c.k., cytokinesis. c, Top, crystal structure of CDK2•cyclinA2 complex. CDK2 (yellow) and cyclinA2 (grey), shown as space-filling model and substrate peptide PKTPKKAKKL (orange) in ball-and-stick representation. The structure contains the transition state mimic ADP•AlF₃ resembling ATP (blue). Below, sequence comparison of substrate oligomer from the CDK2•cylcinA2 complex, CDK1 and ERK1 consensus, EKAREV sequence and the PW-mutations. Red, threonine for phosphorylation. Blue, mutated Lysine (EKAREV(K424)). Purple, mutated Lysine that interacts with cyclinA2 (EKAREV(K426)). Underlined, ERK-docking site. d, Representative trace of EKAREV(PW) during G₂ and M-phase in HEK293. e, Quantification of M-phase peaks and e’, maximum FRET signal in late G₂, measured in HEK293 cells stably expressing EKAREV or point mutant derivatives. Insets: values were normalized to PMA-induced sensor saturation. p-values calculated from denoted cell numbers (n). 3 independent experiments. f, Assessing maximal FRET range (ΔR, %) of parental EKAREV(Tq), and mutant derivatives expressed at ‘medium’ levels. p-values calculated from denoted cell numbers (n). g, Ratiometric traces for FRET and KTR output of monoclonal HeLa cells co-expressing EKAREV(Tq) and ERK-KTR-mCherry biosensors (n=25 cells). Traces were synchronized on mitotic entry and depicted as mean ratio±s.d. Y-axis ratios: FRET (YPF/CFP) and KTR (cytosol/nucleus). Arrow, mitotic entry. Scale bar, 10μm. h, As in g, but HeLa cells co-expressing EKAREN4 and ERK-KTR-mCherry. p-values calculated from n=26 cells. Shown is mean±s.d. Two-sided student’s T-tests: *, p<0.05; **, p<0.001; ***, p<0.0005. n.s., non-significant. For all figures with box-and-whisker plots: boxes represent quartile 2 and 3, horizontal line represents median, whiskers represent minimum and maximum within 1,5x interquartile range. Dots are outliers.

Fig. 2. Comparative analyses of EKAREV, EKAREN4 and EKAREN5

a, EGF responses of >100 co-seeded (1:1), serum-starved monoclonal HeLa cells expressing EKAREV or EKAREN4 (<2 sec intervals). Clustered response amplitudes of EKAREV and EKAREN4 allowed post-analysis unmixing. Dashed line: 5% threshold as proxy for response onset time. Scale bar, 25μm. Experiment performed twice with identical results. b, As a, comparing response onset of parental EKAREV (red) versus (green) EKAREN4 linker length variants (number of amino acids denoted). Shown are averages of multiple single-cell analyses (n represents cells, denoted in graph) and zoomed-in rising phases. Response times in corresponding colors: time to exceed 5% ratio change (dashed line). Experiments b and b” were performed twice. n represents cells, denoted in graph. c, Fluctuations in HeLa cell are simultaneously detected by EKAREN5 and ERK-KTR-mCherry. Representative example from n>100. Inset: clonal HeLa cell expressing nuclear FRET sensor and KTR biosensor. Scale bar, 10μm. d, Scatter plot represents ERK fluctuations in low-density HeLa cells by their amplitudes in FRET and KTR signals. r²-values from linear regression analyses. n represents analyzed peaks, denoted in graph. e, Example of small fluctuation in KTR signal (purple), mirrored by EKAREN5 (blue) in high-density HeLa cell. Arrows, 5% KTR change; 0.5% FRET change. f, Matching scores (mean ratio ± s.d.) for co-detection of ‘small’ fluctuations (defined as 5-10% KTR-ratio changes) with FRET biosensors, scrutinized from single-cell recordings (18hr). p-values calculated from n single-cell traces : EKAREV, 67 events/n=8 traces; EKAREN4, 75 events/n=9 traces; EKAREN5, 116 events/n=14 traces. g, As f. Co-detecting ‘very small’ fluctuations (defined as 0-5% KTR-ratio changes), scrutinized from same recordings. p-values calculated from n single-cell traces : EKAREV, 197 events/n=15 traces; EKAREN4, 195 events/n=11 traces; EKAREN5, 549 events/n=15 traces. f,g, Two-sided student’s T-test: *, p<0.05. n.s., non-significant. h, 1) EKAREV shows steepest relationship (ERK activity versus FRET), with 2) fastest sensor saturation. 3) EKAREN4 shows latest sensor saturation towards high ERK. 4) EKAREN5 shows lowest detection threshold with optimal detection range. EKAREN4/5 have enhanced dynamic FRET ranges. i, Overview of biosensor characteristics. EKAREN5 is the sensor-of-choice, in particular at low (e.g. drug-inhibited) ERK dynamics.

Fig. 3. Single-cell ERK dynamics in CRC PDOs expose cell-to-cell heterogeneity in response to MAPK pathway inhibitors

a, CRC PDOs expressing the EKAREN5 ERK biosensor were imaged (XYZT) during 8-20 hours with 2-4 minute intervals. At the end of each experiment, minimum and maximum FRET signals were enforced by ‘super-inhibition’ (MEK-inh selumetinib (5μM) and ERK-inh SCH772984 (5μM)) and subsequent ‘super-activation’ (PMA (150nM) and subsequent okadaic acid (OA, 1μM)). Inhibitors were washed out prior to super-activation (asterisk). For bona fide analysis of single-cells moving in 3D, we developed a tool to ‘pick planes’, ensuring that resulting 2D (XYT) excerpts seamlessly follow trajectories of individual cells-of-interest over time, with optimal cross-sectioning of the sensor-filled nuclei (see Methods). b, Single-cell ERK responses to MEK inhibitor selumetinib (200nM, approximating clinical plasma concentrations) in EKAREN5-expressing PDO-KRAS^G12V, imaged and analyzed as outlined in a. Shown are single-cell responses from 59 cells in 3 simultaneously imaged PDOs. Below, exemplary responses of two neighboring cells displaying distinct recovery from ERK inhibition in the course of 2 hours. Corresponding cell 1 and cell 2 are indicated in depth-coded organoid image. See also Supplementary Movie 3. Experiment performed twice. b’, Waterfall plots summarizing the direct inhibition effect of selumetinib (green box, ‘Inh’) and the recovery effect (difference between green box ‘Inh’ and green box ‘Recov’). Colored bars (arrows) represent cell 1 and cell 2. c, PDO-BRAF^V600E(#3) organoids were subjected to pan-HER inhibitor LY3009120 (1μM). Most cells showed full inhibition (blue), but a subset of cells (red) showed rapid ERK reactivation, sensitive to afatinib (1μM). Scale bars, 10μm. Experiment performed once.

Fig. 4. pan-HER inhibitor afatinib eliminates ERK activity oscillations in KRAS, NRAS or BRAF mutant PDOs

a-b, EKAREN5 reveals ERK activity in CRC PDOs before and after afatinib administration (200nM). After afatinib-induced ERK suppression, basal ERK level is revealed by further FRET decrease upon super-inhibition (arrow). Unlike other PDOs, BRAF^V600E(#4) does not respond to afatinib. Dashed traces, z-plane analyses comprising multiple cells. Solid line, representative single-cell analysis from corresponding z-plane. BRAF^V600E(#3) (blue) also as Supplementary Movie 4. c, As in a. Drug response to afatinib (200nM) in PDO-WT(#1) and (#2). Unlike RAS/BRAF mutant PDOs, afatinib-induced ERK suppression leaves no residual basal levels (dashed circle). Characteristically, dynamics in WT PDOs showed either high-amplitude fluctuations (black) or relatively constant dynamics (grey). d, Following afatinib-treatment, pan-RAF inhibitor LY3009120 (1μM) abolishes basal ERK activity in KRAS-mutant (top) and BRAF-mutant (bottom) PDOs (dashed circle; no further reduction upon super-inhibition). Shown are z-plane analyses. e, Western blot analyses on indicated PDO-lines recapitulate above described EKAREN5 observations. In particular, afatinib abolished pERK in PDO-WT(#1), while in mutant PDOs pERK was largely affected, but not abolished. Afatinib (200nM), super-inhibition (sel+SCH, 5μM), super-activation (PMA+OA). S.E., short exposure; L.E., long exposure. Vinculin was used as loading control. Experiment performed twice. Blots for PDO-WT(#1) and PDO-KRAS^G12C were from experiment also depicted in Extended Data Fig. 7h. e’) Quantification of Western blots in e. pERK signal was normalized to Vinculin signal (relative to maximal induction (PMA+OA)). f, Doxycyclin-inducible expression of mKate2-NLS-P2A-HRAS^N17 (top) was generated in PDO-KRAS^G12C, PDO-NRAS^Q61H and PDO-BRAF^V600E(#3) via lentiviral infection. FRET-acquisition after 20hr of doxycyclin incubation. Analysis on HRAS^N17-positive cells (marked by mKate2-NLS), reveal loss of oscillatory ERK behavior. Super-inhibition (arrow), but not afatinib could further suppress basal ERK levels. Traces are single-cell analyses. Inset: mKate2-NLS (red) merged with EKAREN5 (green) in BRAF^V600E(#3) organoid; blue trace derived from cell indicated by blue arrow. Scale bar, 10μm. Experiment performed once. g, EGFR employs linear MAPK pathway to induce pulsatile ERK activation, involving both wild-type and mutant KRAS. g’) Model depicting ERK activity patterns in KRAS^G12X or BRAF^V600E mutant PDOs as composites of basal ‘oncogenic’ signaling and EGFR-mediated ERK activity oscillations.

Fig. 5. Mutant KRAS molecules engage in EGFR-mediated ERK activation

a, KRAS^G12C was specifically targeted by AMG-510 (250nM) in PDO-KRAS^G12C, inducing substantial suppression of ERK activity. Top, plane-analysis; bottom, single cells from same organoid. Unlike in afatinib responses (e.g. Fig. 4a), residual ERK activity remains characterized by pulsatile dynamics (of various amplitude), likely representing EGFR activity relayed by the unhindered wild-type RAS proteins. b, Three CRC PDOs containing either wild-type or different RAS mutants were scanned in parallel, excluding non-specific effects of AMG-510 (250nM) on ERK signaling. Shown are plane-analyses. c, Top, scheme for the analysis of EKAREN5 dynamics before and after drug administration on a double calibrated scale. Bottom, bar graph summarizes PDO responses before (green), to AMG-510 (orange) and to AMG-510 + afatinib (200nM) (red). Data presented as mean values ± s.d. N numbers represent individual PDOs (see individual data points) and are stated in the graph for each group. Representative traces in a and b. Experiment performed twice. Two-sided student’s T-test: *, p<0.05. n.s., non-significant. d, EKAREN5 was introduced into RASless MEFs harbouring RAS reconstitutions, being wild-type (top) or either of four mutants (bottom). The MEFs sister lines were serum-starved 24 hours prior to EGF stimulation, clearly indicating amplified signal transduction by mutant RAS upon EGFR activity. FRET-signal is normalized to super-inhibition (box). d’, Quantification of all single cell analyses. n numbers represent cells and are denoted in the graph for each group. Data presented as mean values ± s.d.

Fig. 6. Differential adaptation to long-term growth conditions without EGF

a, Culturing scheme to assess the role of EGF supplementation to the PDO culture medium. PDOs were cultured in medium containing standard EGF (50 ng/ml), 200x diluted EGF (0.25 ng/ml) or no EGF. After 8-13 weeks, PDOs were FRET-imaged during afatinib treatment (200nM) and z-plane analysis was performed (example FRET image shows ROI definition in purple). Right, example trace (representative for n>10): green dashed box, mean ERK before treatment; red dashed box, mean ERK after treatment. Afatinib-induced ERK suppression was quantified according to double calibration (circles: ERK_min and ERK_max). Scale bar, 10μm. b, After 8-13 weeks of culturing without EGF supplementation, the persistent ERK activity oscillations in PDO-KRAS^G12C (red) and PDO-KRAS^G12V (black) showed sensitivity to afatinib. Shown are representative single-cell analyses. b’) Effects of pan-HER inhibition on ERK activity levels, quantification as outlined in a. Green, before afatinib; red, after mock or afatinib (200nM) inhibition Indicated are PDO numbers (n), pooled from one or two experiments (exp). c, As in b, single-cell ERK traces from three PDO-BRAF^V600E mutant PDOs cultured without EGF. ERK dynamics in BRAF^V600E(#1) are now without oscillations. While BRAF^V600E(#3) (blue) still shows oscillatory ERK dynamics, the majority of cells has become unresponsive to afatinib. BRAF^V600E(#4) (grey) remains insensitive to afatinib. c’) Effects of pan-HER inhibition on ERK activity levels, quantification as outlined in a. Green, before afatinib; red, after mock or afatinib (200nM) inhibition. PDO numbers (n) as indicated. Experiment was performed at least twice (each comprising 10-25 organoids). b’,c’, boxes represent quartile 2 and 3, horizontal line represents median, whiskers represent minimum and maximum within 1,5x interquartile range. Dots are outliers. Two-sided student’s T-tests: *, p<0.05; **, p<0.005; ***, p<0.0005. n.s., non-significant. d, Western blot analysis of ERK activity status in PDO lines. ‘Low’, 200x reduced EGF (0.25 ng/ml) supplementation. ‘No’, without EGF supplementation. Vinculin was used as a loading control. Left, RAS mutant PDO lines; right, BRAF^V600E mutant PDO lines. Experiment performed once. d’), Quantifications of d.

Fig. 7. EGFR-mediated amplification of ERK signaling promotes tumor growth in vitro and in vivo

a, Representative FRET-traces (z-plane analyses) of EKAREN5 post afatinib treatment in KRAS^G12C, BRAF^V600E(#2) and BRAF^V600E(#3) shows sustained absence of EGFR-mediated amplification. b, As in a, but FRET-imaging initiated 72hr post afatinib treatment. Basal ERK levels remain unaltered during long-term drug response. c, Monitoring organoid growth of 5-day old PDO-KRAS^G12C during 8 days drug treatment, followed by 5 days without drugs to test growth re-initiation of viable organoids (asterisks, medium refreshments). Drug concentrations (μM) denoted. Representative brightfield images at days 5, 13 and 18. Right, EKAREN5 YPet fluorescence was used for object segmentation for size measurement (~103 PDOs per data point), depicted in corresponding graphs (mean ± s.e.m.). Experiment performed twice. Scale bar, 50μm. #, unreliable measurement (overgrowth) of 18 day-old DMSO-treated PDOs. d, As c, graphs depicting mean organoid size (+/-s.e.m.) during drug treatment on KRAS^G12V, NRAS^Q61H and BRAF^V600E(#3) PDOs. Experiment performed twice. Mean number of objects (PDOs) per time point: KRAS^G12V, 59; NRAS^Q61H, 84; BRAF^V600E(#3), 107. c,d, for exact n-numbers per presented data point, see Source Data. e-h, Frozen sections were taken from experiments performed for Bertotti et al.. e, Schematic of PDX model generation from metastatic CRC samples for prospective examination of cetuximab response. f, Changes in pERK levels were scored through immunohistochemistry on paraffin-embedded tumor sections after 2 weeks of cetuximab or placebo treatment in three independent mice bearing PDX from same tumor. Shown are representative sections from KRAS^G12D-mutant PDX model M245, illustrating strong reduction in pERK levels following cetuximab-treatment. Scale bar, 100 μm. g, Graph depicting quantified changes in pERK levels for 9 out of 10 PDX models of CRC (oncogenic mutations denoted). Changes in pERK levels were scored through immunohistochemistry as the average of 10 optical fields (40X), randomly chosen from each tumor (n=30). Minimally 2 mice per group were analyzed. h, Comparative re-analysis of tumor growth data per PDX model upon treatment with cetuximab (n=6 mice) or placebo (6 mice). Top, bar graph representing tumor size variation from baseline of untreated and cetuximab-treated animals. Bottom, bar graph representing tumor growth inhibition (%,treated/untreated) calculated from size variation.