EGFR Blockade Reverts Resistance to KRAS^G12C Inhibition in Colorectal Cancer

The small GTPase KRAS is mutated and constitutively active in 15% of all human tumors (1). This protein acts as a molecular switch, transitioning from an active (GTP-bound) to an inactive (GDP-bound) state. This transition is highly regulated, and physiologic feedback loops limit the duration of KRAS activation. Guanine exchange factors, such as SOS1, and GTPase activating proteins (GAP), such as NF1, directly mediate KRAS activation (2, 3). When mutated, KRAS loses the ability to interact with GAPs and therefore becomes constitutively activated, driving downstream MAPK and PI3K–AKT signaling pathways and enhancing cancer cell proliferation and survival (4). Different mutations in KRAS may modify the intrinsic activation (GTP-bound state) of the mutant protein and its interaction with downstream effectors, such as RAF proteins (5, 6). Moreover, KRAS signaling is also regulated by negative feedback mechanisms through the transcriptional induction of protein phosphatases (DUSPs and Sproutys) and SPREDs (2). The complexity of the signaling network and the heterogeneous features of the multiple KRAS-mutant alleles have contributed to the difficulty in developing molecularly targeted therapies against mutant KRAS, a target considered undruggable for decades.

However, in recent years, the unique characteristics of one of the KRAS-mutant alleles, KRAS^G12C, have been exploited for the design of covalent inhibitors that can specifically bind the cysteine on the mutant residue (7). These inhibitors can bind G12C-mutant KRAS beneath the effector binding switch II domain when KRAS is in the GDP-bound (inactive) state, trapping KRAS in its inactive form and taking advantage of the retained, intrinsic cycling of the G12C mutant from GTP-bound to GDP-bound states (7, 8). Given this particular feature of the KRAS^G12C-mutant protein, the efficacy of these compounds can be increased by concomitant receptor tyrosine kinase (RTK) inhibition that further maintains KRAS^G12C in a GDP-bound conformation, facilitating the binding of the covalent inhibitor (7–9).

Besides RTK inhibition, several other combinatorial strategies have already been proposed, with the goal of potentiating the effects of these inhibitors (10–13). Importantly, the unique feature of KRAS^G12C covalent inhibitors, which only bind the mutant allele, renders these drugs a potentially ideal partner for combinatorial strategies, as toxicities should be minimized and limited to the combinatorial partner.

Several clinical trials testing the efficacy of KRAS^G12C-targeted drugs are currently ongoing, and early results from the first two trials using AMG510 and MRTX849 have documented promising outcomes in the clinic. However, clinical studies unexpectedly reported that the response rate to these drugs is high in patients with non–small cell lung cancers (NSCLC) but limited in patients with colorectal cancer (14, 15).

In this work, we sought to understand the mechanisms underlying this lineage difference. We hypothesized that NSCLC bearing KRAS^G12C could be intrinsically different from KRAS^G12C-mutant colorectal cancer, despite harboring the same oncogenic mutation. We and others previously reported that BRAF^V600E colorectal cancer has limited sensitivity to single-agent RAF inhibitors, and this is a result of an adaptive response selectively triggered in colorectal cancer (16–18).

Therefore, we examined the adaptive response to KRAS^G12C inhibition and its kinetics in KRAS^G12C-mutant NSCLC and colorectal cancer cell models. We show that KRAS^G12C-mutant colorectal cancers maintain sensitivity to upstream RTK signaling, particularly EGFR. Using biochemical and pharmacologic approaches, we show that EGFR reactivation restricts the efficacy of KRAS^G12C inhibition in colorectal cancer. We find that KRAS^G12C and EGFR blockade is effective to overcome this adaptive resistance, both in cell lines and in patient-derived models. These findings have immediate relevance for clinical strategies to overcome unresponsiveness to KRAS^G12C blockade in patients with colorectal cancer.

To understand the mechanistic basis for differential clinical responses to treatment with the selective KRAS^G12C inhibitor AMG510 in colorectal cancer and NSCLC, we analyzed the effect of this drug in a panel of KRAS^G12C-mutant cell lines. These included two NSCLC (NCIH358 and LU65) and seven colorectal cancer (C106, RW7213, SW837, SNU1411, JVE015, and LIM2099) cell lines. Response to AMG510 was measured in a dose–response proliferation assay. Sensitivity to single agent was comparable among the models, showing no clear difference between colorectal cancer and NSCLC cell lines (Fig. 1A). Therefore, to explore the quality, duration, and adaptive signaling response to drug treatment, we compared the NSCLC cell lines with four colorectal cancer models, consisting of two representative models with low IC₅₀ to AMG510 (C106 and RW7213) and two representatives with high IC₅₀ values (SW837 and SNU1411). Cells were subjected to increasing concentrations of AMG510 for 1 and 24 hours (Fig. 1B). After 1 hour of drug treatment, KRAS^G12C inhibition led to downregulation of phosphorylated ERK in a dose-dependent manner in all cell models, regardless of tumor lineage. However, after 24 hours of drug exposure, we observed that both NSCLC cell lines further downregulated ERK phosphorylation, whereas all the colorectal cancer cell lines showed a rebound in phospho-ERK levels (Fig. 1C and D). We then examined the impact of AMG510 on the signaling network over time in NSCLC and colorectal cancer cell lines. At baseline, the colorectal cancer cell line C106 expressed lower levels of phosphorylated MEK and ERK with similar total protein relative to the NSCLC line NCIH358. Treatment with AMG510 decreased levels of active RAS-GTP, but the suppression was more pronounced in NCIH358 compared with C106. In C106, residual phospho-MEK and ERK were measurable already after 1 hour of treatment, whereas the NSCLC cell line (NCIH358) showed signs of rebound in phospho-MEK and phospho-ERK only at 72 hours after treatment (Supplementary Fig. S1A–S1C). Altogether, these data show that MAPK pathway inhibition upon AMG510 treatment is transient and less profound in colorectal cancer than lung models and suggest that this may be due to greater feedback suppression of MAPK signaling in colorectal cancer that is then released with KRAS^G12C inhibition.

KRAS mutation acts downstream of RTK signaling and leads to constitutive ERK activation; in turn, high ERK levels feedback inhibit receptor signaling (19). KRAS-mutant proteins are thought to be constitutively active; therefore, it would be expected that KRAS-mutant tumors do not rely on RTK activation. To characterize the basal level of RTKs in patient tumors, we assembled a series of colorectal cancer and NSCLC clinical samples carrying the KRAS^G12C mutation. We analyzed eight colorectal cancer and five NSCLC patient specimens (Supplementary Table S1), all collected before any targeted therapy treatment, and evaluated the expression of 72 different proteins, including RTKs, using a selected reaction monitoring-mass spectrometry (SRM-MS) assay (see Methods). Detectable levels of EGFR, HER2, and MET were present in both lineages without clear differences, whereas HER3 expression was more prevalent in colorectal cancer patient samples (Fig. 2A). We did not detect ALK, ROS1, RET, FGFR, or AXL expression in any of the samples. Two patients with colorectal cancer and 2 patients with NSCLC later went on the phase I trial of AMG510 (Supplementary Table S1). Despite similar basal total RTK levels in the tumors from trial-enrolled patients, both NSCLC cases achieved clinical benefit, whereas both patients with colorectal cancer progressed upon AMG510 treatment. These data indicate that even in the presence of KRAS^G12C mutation, tumors often express upstream RTKs, predominantly consisting of HER family and MET receptors. However, the levels of total RTK expression do not appear to underlie lineage-specific differential response to KRAS^G12C inhibition between patients with colorectal cancer and patients with NSCLC.

Although we were able to detect total RTK expression in the clinical samples, we could not assess activated RTK levels or evaluate the functional consequences of detectable RTK expression in these tumors. To test this, we measured the levels of active RTKs in colorectal cancer and NSCLC cell lines using phospho-RTK arrays. The NSCLC cell line, NCIH358, had no detectable basal RTK activation and low levels of the phospho-EGFR and phospho-IGF1R could be detected only with signal enhancement through long exposure. In contrast, both colorectal cancer cell lines, C106 and RW7213, showed a number of detectable phosphorylated RTKs (Fig. 2B). Marked phospho-EGFR signal was present in both colorectal cancer models and was accompanied by detectable levels of other phosphorylated receptors. Coimmunoprecipitation assays using cetuximab as a capture antibody confirmed that EGFR is bound to the other HER receptors, MET and FGFR, suggesting heterodimerization processes (Supplementary Fig. S2A and S2B). Several Ephrin receptors (EphR) were also found activated in colorectal cancer cell lines. Interestingly, EphR activation has been reported as a mechanism of resistance to cetuximab in colorectal cancer and head and neck squamous cell carcinoma, suggesting that the cross-talk between these two classes of RTKs can potentially amplify EGFR signaling (20).

We next investigated the functional effect of the high levels of basal RTK activation in the colorectal cancer cell lines by evaluating the signaling consequences of EGF stimulation over time. At baseline, signaling downstream of RAS (phospho-AKT, phospho-MEK, and phospho-ERK) was lower in the colon cancer cell lines compared with the NSCLC cell line, NCIH358. EGF treatment led to induction of phospho-EGFR in all cell models (Fig. 2C). However, activation of downstream effectors showed lineage-specific differences. The NSCLC cell line, NCIH358, did not show any modulation of AKT, MEK, or ERK phosphorylation upon EGF treatment, suggesting that KRAS^G12C does not get (further) activated by EGFR stimulation and can sustain downstream signaling without upstream inputs (Fig. 2C and D). In contrast, in the colorectal cancer cell lines, C106 and RW7213, downstream effectors followed the same kinetics of EGFR activation, showing that KRAS^G12C can be activated by upstream signaling components and that basal activation of RAS downstream effectors may be lower due to negative feedback loops (Fig. 2C and D). Altogether, these data indicate that even in the presence of the same KRAS-activating mutation, lineage-specific differences exist in the degree of RTK activation, contribution of upstream receptors to ERK activation (and thus rate of cycling of KRAS between the GTP- and GDP-bound forms), and feedback mechanisms that would be released from suppression by selective KRAS inhibitor–mediated ERK inhibition.

On the basis of the high levels of basally active EGFR, the contribution of EGFR to KRAS^G12C signaling outputs, and the known relevance of EGFR signaling in colorectal cancer, we next evaluated whether KRAS^G12C–EGFR combined inhibition could effectively suppress signaling and growth in the colorectal cancer models. Biochemical analysis and densitometric quantitation of phosphoprotein levels showed that concomitant KRAS^G12C and EGFR inhibition, with AMG510 and the anti-EGFR antibody cetuximab, respectively, led to sustained inhibition of phospho-MEK and phospho-ERK, as well as the ERK downstream targets DUSP6 and CyclinD1, indicating inhibition of ERK output (Fig. 3A; Supplementary Fig. S3A and S3B). Colorectal cancer cell lines were also tested in short-term proliferation assays. ATP content measurement showed that concurrent treatment with the EGFR inhibitor cetuximab synergized with AMG510 in all the cell lines tested (Fig. 3B; Supplementary Fig. S4A). Growth suppression was durable, as similar effects were also seen in long-term proliferation assays stained with crystal violet (Fig. 3C).

We next sought to determine whether combined KRAS and EGFR inhibition could lead to cell death together with cell proliferation arrest. We found that the combination treatment induced a pronounced cytotoxic effect, as measured by detection of DNA of cells with impaired membrane integrity, in all the models tested and across dose levels of AMG510 (Fig. 3D). These data indicate that combined KRAS^G12C and EGFR inhibition leads to increase cell death rate.

To further assess the role of EGFR signaling in AMG510 resistance, we used the KRAS wild-type, cetuximab-sensitive colorectal cancer cell line LIM1215 to generate isogenic models carrying the KRAS^G12C allele using an AAV-mediated knock-in approach (21). As expected, all KRAS^G12C-mutant clones (KI#1, KI#2, and KI#3) showed resistance to cetuximab as a single agent (Fig. 3E). KRAS^G12C-mutant clones were more sensitive to AMG510 than their wild-type counterpart (Fig. 3F). The addition of cetuximab in combination with KRAS^G12C inhibition was synergistic compared with the single agent alone (Fig. 3G; Supplementary Fig. S4B). These results provide further support that EGFR and KRAS^G12C should be concomitantly inhibited to overcome resistance to KRAS^G12C inhibition in colorectal cancer.

We and others previously reported that acquired KRAS mutations are responsible for secondary resistance to anti-EGFR therapies in colorectal cancer (22, 23). We reasoned that KRAS inhibitors could have clinical applications in patients who develop KRAS mutations at relapse after EGFR blockade. To directly test this possibility, we used a colorectal cancer cell line (HCA46-R Pmab), in which prolonged treatment with panitumumab led to the emergence of a KRAS^G12C allele (24). We examined whether AMG510 could be effective in this setting. Although single-agent AMG510 treatment was partially effective, its combination with cetuximab induced both cell proliferation arrest and cell death (Fig. 3H–J; Supplementary Fig. S4C). These data suggest that also in the secondary resistance setting, colorectal cancer cells could retain upstream signaling dependency.

To test whether and to what extent these findings were applicable to clinically relevant models, we collected biopsies and exploited established models from patients with colorectal cancer generated in our institutions (Supplementary Tables S2 and S3).

Taking advantage of our molecularly and therapeutically annotated colorectal cancer patient–derived biobank (25–27), in a cohort of more than 100 patient-derived organoids (PDO), we selected three different KRAS^G12C-mutant PDOs and one KRAS^G12D-mutant PDO (as negative control) and tested the effects of single agents AMG510 and cetuximab and their combination.

In the KRAS^G12C PDOs, single-agent AMG510 or cetuximab did not inhibit or poorly inhibited growth, whereas the combination suppressed proliferation with high synergy scores (Fig. 4A–C).

In parallel, we generated patient-derived xenografts (PDX) from 2 patients with KRAS^G12C-mutant colorectal cancer and tested AMG510 and cetuximab as single agents and in combination in vivo. Both models exhibited slow but persistent growth when mice were treated with either cetuximab or AMG510 single-agent treatment. Combination treatment induced massive or even complete tumor shrinkage in the treated animals (Fig. 4D and E). No tumor was visible after 10 weeks of treatment and no samples could be collected at the end of the PDX experiment (Fig. 4E). Interestingly, IHC staining for the proliferation marker Ki-67 was highly positive in tumor samples derived from vehicle-, cetuximab-, and AMG510-treated arms (Ki-67 IHC intensity of 97% with no necrosis in vehicle-treated mice, 92% with necrosis in cetuximab-treated mice, and 83% with necrosis in AMG510-treated mice), but strong phospho-ERK staining was observed only in the AMG510-treated samples, suggesting that strong MAPK pathway rebound occurs in patient-derived samples as well (Fig. 4F).

Recent clinical data have shown promising activity of agents targeting mutant (G12C) KRAS, a heretofore undruggable target. These trials, however, show substantial differences in the response rate between patients with NSCLC and patients with colorectal cancer (14, 15).

We found that KRAS^G12C inhibitors produce less profound and more transient inhibition of KRAS downstream signaling in colorectal cancer versus NSCLC models. Our data indicate that, despite harboring the same mutation, there are intrinsic differences in signaling between colorectal cancer and NSCLC. Our data suggest that in colorectal cancer the combination of (i) signaling rebound upon KRAS^G12C inhibition, (ii) the higher extent of RTK activation, and (iii) the stronger response to EGFR direct stimulation can be used to define the intrinsic dependency of bowel tumors on the EGFR–MAPK pathway. These results, together with the emerging clinical experience with single-agent KRAS^G12C inhibitors, indicate that combinatorial treatment is necessary to effectively treat KRAS^G12C-mutant colorectal cancer.

This is, to our knowledge, the first study aimed at specifically characterizing the effects of KRAS^G12C inhibition in colorectal cancer. Indeed, the main goal of this work was to understand and exploit colorectal cancer lineage–specific features to develop more effective pharmacologic strategies.

It is known that RTK inhibition in NSCLC preclinical models can potentiate KRAS^G12C covalent inhibitors by slowing down exchange factors favoring a GDP-bound status of KRAS (8). However, this combinatorial strategy is considered only marginal in this setting, as other combinatorial partners including PI3K, mTOR, and Aurora kinase inhibitors have been shown to have higher efficacy than EGFR inhibition in combination with KRAS^G12C inhibition in lung cancer (10, 12, 13).

We found that expression of EGFR and its dimerization partners is common in KRAS^G12C colorectal cancer and NSCLC clinical samples. Basal EGFR expression, as measured by IHC or mass spectrometry (28, 29), has not been associated with EGFR inhibitor sensitivity in colorectal cancer, and indeed we find no clear difference between receptor expression levels by lineage or in the subset of patients with colorectal cancer and NSCLC treated with AMG510. In cell line models, however, we are able to assess activated RTK levels, and we find that the degree of this dependency on RTK signaling across KRAS^G12C-mutant tumors can be very different. Specifically, we show that KRAS^G12C-mutant colorectal cancer cell lines still retain their dependency on EGFR for downstream signaling.

The clinical and signaling effects of AMG510 in colorectal cancer are reminiscent of the clinical experience in BRAF^V600E colorectal cancer, where the concomitant inhibition of upstream signaling reactivation results in improved clinical efficacy (30), which has led to the recent FDA approval of the BRAF inhibitor encorafenib in combination with cetuximab (BEACON colorectal cancer; NCT02928224).

Clinical trial results have shown limited efficacy of KRAS^G12C covalent inhibitors in colorectal cancer; only 1 patient with colorectal cancer treated with MRTX849 had a 47% tumor reduction by RECIST criteria (15). Our results show that in colorectal cancer cell models and in PDX, treatment with AMG510 single agent induces strong reactivation of MAPK pathway. These data indicate that even in the best scenario, up-front combination treatment is the best option for patients with KRAS^G12C-mutant colorectal cancer, as this approach would enhance the efficacy of KRAS^G12C inhibitors and would also limit the emergence of acquired resistance.

We showed that KRAS^G12C-mutant colorectal cancer still retains EGFR–MAPK dependency, as demonstrated by the signaling rebound upon KRAS^G12C inhibition, the more pronounced activation of RTKs, and the high response to EGFR direct stimulation. We can speculate that other KRAS-mutant alleles in colorectal cancer could also retain these characteristics, suggesting that other KRAS-mutant colorectal cancers could also potentially benefit from EGFR-based combinations. Therefore, direct or indirect KRAS blockade–based combinations could be considered as an option in these specific settings.

Combined inhibition of KRAS^G12C with SHP2, a phosphatase that serves as a common mediator of RTK signaling to RAS (31), has been proposed as another combination to overcome the feedback reactivation in KRAS^G12C tumors (32).

Available clinical data with SHP2 inhibitors is currently limited, and it is not yet known whether these inhibitors can be administered to patients in sufficient doses to effectively inhibit RTK signaling. Furthermore, preclinical models suggest growth suppression but not complete tumor regression with SHP2 and KRAS^G12C inhibitors, in contrast to the complete, durable tumor inhibition achieved in a PDX treated with EGFR and KRAS^G12C inhibitors.

Because RTK signaling in colorectal cancer is mostly dominated by wild-type EGFR, we propose to combine KRAS^G12C inhibitors with anti-EGFR mAbs to block receptor signaling rather than inhibiting EGFR kinase activity. Moreover, the differential activity of EGFR inhibitors in colorectal cancer versus NSCLC further indicates signaling differences in these two lineages (33). The anti-EGFR antibodies cetuximab and panitumumab are approved for the treatment of RAS/RAF wild-type colorectal cancer (34, 35), whereas EGFR tyrosine kinase inhibitors (TKI) are approved for the treatment of EGFR-mutant lung cancer (36). EGFR-targeting TKI have not been effective in colorectal cancer (37), probably because EGFR genetic alterations are rare in colorectal cancer and these drugs are more efficient in the EGFR-mutant setting. As release of wild-type receptor signaling from feedback inhibition is thought to mediate adaptive resistance to targeted therapies, the role of wild-type EGFR in activating signaling in colorectal cancer could render colorectal cancers more sensitive to this feedback release than NSCLC.

Taken together, our results demonstrate that biochemical characterization of response to KRAS^G12C inhibitors in colorectal cancer provides a mechanistic explanation for their limited activity in the clinic and, at the same time, provides a strong rationale to combine EGFR blockade with these agents in patients with colorectal cancer.

The cell lines NCIH358, LU65, SNU1411, JVE015, LIM2099, LIM1215, LIM1215 KRAS G12C KI, and RW7213 were cultured in RPMI (Lonza). The C106 cell line was cultured in Iscove's Modified Medium (Lonza). The SW837 cell line was cultured in DMEM/F12 (Lonza). The HCA46-R Pmab cell line was cultured in DMEM (Lonza). Each medium was supplemented with 10% FBS, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 mg/mL streptomycin. The SW837 and NCIH358 cell lines were purchased from the ATCC; the C106 and HCA46 cells were purchased from the European Collection of Animal Cell Cultures; the SNU1411 cells were purchased from KCLB; the LU65 cells were purchased from AcceGen Biotech; the LIM2099 cells were provided by Dr. F. Walker (The Walter & Eliza Hall Institute of Medical Research, Ludwig Institute for Cancer Research, Melbourne-Parkville Branch, Parkville, Victoria, Australia). LIM1215 cells were kindly provided by Dr. Whitehead (Vanderbilt University, Nashville, TN); JVE015 cells were provided by Prof. Tom Van Wezel (Leiden University Medical Center, Leiden, the Netherlands); and RW7213 cells were provided by Dr. Diego Arango (Institute of Biomedical Research of Lleida, Lleida, Spain).

All the cell lines were determined to be Mycoplasma free using the Venor GeM Classic Kit (Minerva Biolabs) and tested by short tandem repeats profiling at 10 different loci.

AMG510 was purchased from ChemGood (C-1499) and cetuximab was either gifted from Merck Serono under an academic material transfer agreement or purchased from the pharmacy at Memorial Sloan Kettering Cancer Center (New York, NY).

All the procedures performed with patient specimens were conducted under the approval of the local ethical committee of the institutions, after the written informed consent of the patients. CRC1589, CRC1360, and CRC0051 PDOs were derived from PDXs established from three distinct patients with KRAS^G12C colorectal cancer, whereas IRCC107A PDO was established directly from tissue biopsy obtained at the time of surgery. To generate PDX-derived organoids, PDX samples were first dissociated mechanically with gentle MACS Dissociator (Miltenyi Biotec) and then underwent enzymatic digestion of extracellular matrix using Human Tumor Dissociation Kit (Miltenyi Biotec). The derived single-cell suspension was centrifuged at 400 × g for 5 minutes and then resuspended in the organoids' basal medium. Cell suspension was filtered with 70-μm cell strainer to avoid cell cluster formation. The flow-through was centrifuged and the pellet was resuspended in Matrigel(Corning) and dispensed as 100 μL droplet in the center of 37°C prewarmed 12-well plate. After an incubation of 15 minutes at 37°C for Matrigel solidification, organoids were added with ENAS medium and incubated at 5% CO₂ and 37°C. ENAS medium was refreshed twice per week.

To generate PDOs directly from patient tissue biopsy, tumor tissue was first smashed into small pieces and incubated in PBS with collagenase A (0.5 mg/mL; Roche), hyaluronidase (20 mg/mL; Sigma), and 10 mmol/L ROCK inhibitor for 30 minutes at 37°C with shaking. After incubation, 5% FBS was added and the mixture was centrifuged at 400 × g for 5 minutes. The pellet was washed three times in PBS to remove debris and collagenase. At the end of the washing phase, the pellet was resuspended in Matrigel (Corning) and 50 μL of organoids–Matrigel suspension was dispensed into the center of each well of a 37°C prewarmed 24-well plate. Different densities of tumor cells were plated and left to solidify for 10–30 minutes at 37°C before ENAS medium was added and the cells were incubated at 37°C, 5% CO₂. Fresh medium was replaced every 2–3 days. Outgrowing organoids were passaged every 10–15 days after mechanical and enzymatic disruption. Organoid drug treatment was performed in each organoid-specific culture medium at 37°C and 5% CO₂.

The organoid basal medium was composed of Advanced DMEM/F12 Medium (Thermo Fisher Scientific) containing 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mmol/L Glutamax (Gibco), 10 mmol/L Hepes, and 50 μg/mL Primocin (Invitrogen). ENAS medium was prepared by adding 1× B27 supplement (Invitrogen), 1× N2 supplement (Invitrogen), 1.25 mmol/L N-acetyl-cysteine (Sigma Aldrich), 10 mmol/L Nicotinamide (Sigma Aldrich), 10 nmol/L Gastrin (Sigma), 50 ng/mL recombinant mouse EGF (Life Technologies), 100 ng/mL Noggin (PeproTech), 500 nmol/L TGFb type I receptor inhibitor A83-01 (Tocris), and 10 μmol/L p38 MAPK inhibitor SB202190 (Sigma Aldrich) to the organoid basal medium.

After seeding and drug treatments, cells were washed with cold PBS and lysed in RIPA buffer (Pierce #89901) plus phosphatase and protease inhibitors (Thermo Fisher Scientific #1861277 and #1861278). Lysates were cleared by centrifugation at 14,000 rpm at 4°C and quantified using BCA Method (Pierce #23224).

Samples were prepared using LDS+Reducing Agent Novex Buffers (Invitrogen #NP0008 and #NP0009). Ten to 20 μg of lysates were loaded and run on NuPageTM 4%–12% Bis-Tris Gels (Thermo Fisher Scientific #NP0321BOX) followed by transfer to nitrocellulose membranes (Bio-Rad #1620233). Membranes were incubated overnight with the indicated antibodies, washed, and incubated again for 45 minutes with anti-rabbit or anti-mouse secondary antibodies. Detection was performed using Immobilion Western (Millipore #WBKLS0500).

Primary antibodies were obtained from Cell Signaling Technology and were used at a concentration of 1:1,000: pAKT S473 (#9271), pMEK1/2 S217/221 (#9154), p-p44/42 MAPK T202/204 (#9101), MEK1/2 (#9122), ERK1/2 (#9102), AKT (#9272), EGFR (#2232S), phospho-EGFR Tyr1068 (#3777S), and Vinculin (#13901S). DUSP6 antibody was purchased from Abcam (#ab54940) and used at 1:500 dilution. CyclinD1 antibody was purchased from Santa Cruz Biotechnology (#sc-718) and used at 1:1,000 dilution. Densitometry analyses were performed using ImageJ software.

For immunoprecipitation assays, cells were seeded and starved for 24 hours and then stimulated with 20% FBS for 10 minutes. Cells were then washed in cold PBS and lysed using NP-40 buffer (150 mmol/L NaCl, 10 mmol/L Tris pH 8, 1% NP-40, and 10% glycerol). Proteins (20 μg) were used as total lysates. Protein lysates (1,500 μg) were incubated rotating at 4°C for 1.5 hours in the presence of 10 μg/mL cetuximab, Agarose protein G beads (Invitrogen #15920-010). Beads were centrifuged at 3,000 rpm for 1 minute and supernatant was removed. Similarly, beads were washed three times using NP-40 buffer and once using nuclease-free sterile water. Finally, immunoprecipitates were eluted using 1× sample buffer (NuPAGE LDS Sample Buffer, Thermo Fisher Scientific #NP0008 and NuPAGE Sample Reducing Agent, Thermo Fisher Scientific #NP0009) at 95°C for 5 minutes.

RAS-GTP pulldown assay was performed according to the manufacturer's protocol (Thermo Fisher Scientific #16117). Briefly, 500 μg of lysates were loaded into columns together with agarose beads and RAS-RBD bait and incubated for 1 hour at 4°C. After the incubation, beads were washed three times and resuspended in LDS+Reducing Agent Novex Buffers (Invitrogen #NP0008 and #NP0009). A fraction of lysates was used to measure total RAS amount. Pulldown and total lysates were subjected to Western blotting procedure as described above. The kit provided primary antibody against pan-RAS.

Human Phospho-RTK Arrays (R&D Systems, ARY001B) were used to detect activated RTKs according to the manufacturer's instructions. Cells were washed with cold PBS and lysed using the provided lysis buffer plus phosphatase and protease inhibitors. Lysates (200 μg) were incubated on membranes overnight. Membranes were subsequently washed and exposed to chemiluminescent reagent.

SRM-MS of 72 biomarkers of the tissue sections from formalin-fixed, paraffin-embedded (FFPE) blocks was conducted as described previously (38, 39). Briefly, two tissue sections (10 μm each) from FFPE blocks were placed on DIRECTOR slides, deparaffinized, and stained with hematoxylin. Tumor areas were marked by a board-certified pathologist, and then were microdissected and solubilized to tryptic fragments as per the manufacturer's instructions (Expression Pathology). Protein concentration of the tryptic peptides was calculated using microBCA. Stable heavy isotope-labeled internal standard peptides for 72 biomarkers were added to the solution and injected into the mass spectrometer (TSQ Quantiva, Thermo Fisher Scientific). On-column injection resulted in 5 fmol of isotopically labeled internal standard peptide and 1 μg of total tumor protein. Data analysis of the 72 biomarkers was conducted using Pinnacle Software (Optys Tech). All patients whose tumors were analyzed signed written informed consent for molecular profiling of their tumors (IRB 12-245) and approval was obtained for the mass spectrometry assay.

For short-term proliferation assays, 2,000 cells were seeded in 96-well culture plates in medium containing 10% FBS. After 24 hours, serial dilutions of AMG510 (alone or in combination with cetuximab) were added to cells in 10% FBS medium (ratio 1:1). Cell viability was determined at day 5 after the start of the treatment by measuring ATP content using CellTiter-Glo Luminescent Cell Viability Assay (Promega). DMSO only–treated cells were used as control. Excess over the Bliss (Bliss score) was calculated using effect of the combination (observed) versus what was predicted/expected on the basis of the effect of each single agent expressed as percentage. Negative values or values close to zero show no/very little synergy. Higher percentages indicate higher synergy.

For long-term proliferation assays, cells were seeded (5,000 cells per well for SNU1411, SW837, and HCA46R, and 80,000 cells per well for RW7213) in 24-well plates or in 6-well plates (C106 80,000 cells per well) in 10% FBS medium. The following day, plates were treated with different concentrations of drugs as indicated. Cells were fixed with 4% paraformaldehyde and stained with 10% crystal violet in a Methanol/Water Solution (Sigma Aldrich) 9–13 days after the start of treatment, according to the confluence reached by the untreated control. Every 7 days, media and drug were refreshed. In all the experiments, plates were incubated at 37°C in 5% CO_2.

Short-term assays were performed at least three times. All long-term assays were performed independently at least three times for HCA-46R, SW837, and SNU114, and two times for C106 and RW7213. Analyses and graphs were performed with GraphPad Prism Software using the function log (inhibitor) versus response-variable slope (four parameters).

For short-term cytotoxicity assay, 2,000 cell/well were seeded in 96-well plates and treated after 24 hours with AMG510 alone or in combination with cetuximab as indicated. After 5 days from the start of treatment, cell cytotoxicity was determined measuring the fluorescence of dead cells using CellTox Green Cytotoxicity Assay (Promega), according to the kit's instructions. The CellTox green results were normalized to results of CellTiter-Glo Luminescent Cell Viability assay conducted on the same plates. In all the experiments, plates were incubated at 37°C in 5% CO₂. All assays were performed independently at least three times.

The CLR113 PDX was derived from a liver metastasis under written consent. Tumor tissue was transplanted orthotopically into NSG mice to establish the PDX (institutional review board protocols 06-107 and 14-091). Once a tumor became visible in the first mouse, it was transplanted and expanded to other animals. Tumor tissue was implanted subcutaneously in the flank of 4- to 6-week-old NSG female mice, and treatment of the mice began when tumors reached approximately 100 mm³ in size. Mice were randomized (n = 5 mice per group) to receive drug treatments or vehicle as control. Studies were performed in compliance with institutional guidelines under an Institutional Animal Care and Use Committee (IACUC)–approved protocol. The animals were immediately euthanized as soon as the tumors reached the IACUC set limitations.

The CRC0051 PDX was derived from a colorectal cancer patient metastasis, under written consent. Patient tissue was transplanted into NOD/SCID mice to establish PDX and then transplanted and expanded to other animals. Tumor was implanted subcutaneously in the flank of 11- to 13-week-old NOD/SCID mice, and treatment of the mice began when tumor reached approximately 250 mm³. Mice were randomized (n = 7 per arm) and received drug treatment or vehicle control.

AMG510 was given 100 mg/kg daily by gavage. Cetuximab was administered 50 mg/kg twice a week, by intraperitoneal injections.

The entire experiment was performed in compliance with institutional guidelines of the Ethical committee of Candiolo Cancer Institute (Candiolo, Torino, Italy) and according to the Italian Ministry of Health–approved protocol. The animals were sacrificed on the basis of the rules defined in the approved protocol.

R. Yaeger is a consultant at Array BioPharma, reports receiving commercial research grants from Array BioPharma, Boehringer Ingelheim, and Novartis, and has an advisory board relationship with Array BioPharma. Y.R. Murciano-Goroff has received remuneration from AstraZeneca. N. Valeri has received speakers bureau honoraria from Bayer, Eli Lilly, Pfizer, and Merck. S. Thyparambil is a senior director R&D at mProbe. S. Siena is an advisory board member at Amgen, Bayer, BMS, CheckMab, Clovis, Daiichi-Sankyo, Merck, Roche-Genetech, and Seattle Genetics. L. Trusolino reports receiving commercial research grants from Symphogen, Servier, Pfizer, Merus, and Menarini, and has received speakers bureau honoraria from Eli Lilly, AstraZeneca, and Merck KGaA. B.T. Li is a consultant/advisor at Roche/Genentech, Biosceptre International, Thermo Fisher Scientific, Mersana Therapeutics, Hengrui Therapeutics, and Guardant Health, reports receiving commercial research grants from Roche/Genentech, Daiichi Sankyo, Amgen, Lilly, Hengrui Therapeutics, Illumina, Guardant Health, BioMedValley Discoveries, AstraZeneca, GRAIL, and MORE Health, and has ownership interest in two institutional patents at Memorial Sloan Kettering Cancer Center (US62/685,057, US62/514,661). N. Rosen is a SAB member at Beigene, MAPCure, AstraZeneca, Tarveda, Ribon, Chugai, and Zai Labs, is a consultant at Boehringer Ingelheim, reports receiving commercial research grants from Chugai and Boehringer Ingelheim, and has ownership interest (including patents) in Kura, Beigene, Zai Labs, Fortress, and Ribon. A. Bardelli reports receiving a commercial research grant from Neophore, has ownership interest (including patents) in Neophore, and has advisory board relationships with Horizon Discovery, Neophore, and Biocartis. S. Misale is a consultant at Boehringer-Ingelheim. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Amodio, R. Yaeger, N. Rosen, A. Bardelli, S. Misale

Development of methodology: V. Amodio, R. Yaeger, P. Arcella, M. Pinnelli, H.-Y. Zhao, L. Trusolino, F. Di Nicolantonio, S. Misale

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Amodio, R. Yaeger, P. Arcella, S. Lamba, S. Arena, M. Montone, B. Mussolin, Y. Bian, M. Pinnelli, Y.R. Murciano-Goroff, E. Vakiani, N. Valeri, W.-L. Liao, A. Bhalkikar, S. Thyparambil, H.-Y. Zhao, E. de Stanchina, S. Siena, A. Bertotti, L. Trusolino, B.T. Li, F. Di Nicolantonio, S. Misale

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Amodio, R. Yaeger, P. Arcella, S. Arena, Y.R. Murciano-Goroff, E. Vakiani, A. Bhalkikar, B.T. Li, F. Di Nicolantonio, A. Bardelli, S. Misale

Writing, review, and/or revision of the manuscript: V. Amodio, R. Yaeger, Y. Bian, S. Marsoni, S. Siena, A. Bertotti, L. Trusolino, B.T. Li, N. Rosen, F. Di Nicolantonio, A. Bardelli, S. Misale

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Cancelliere, A. Lorenzato, A. Whaley, H.-Y. Zhao, N. Rosen

Study supervision: V. Amodio, F. Di Nicolantonio, A. Bardelli, S. Misale

The authors thank Drs. Maurizio Scaltriti and David Solit for critical reading of the manuscript. The authors also thank the Molecular Cytology Core Facility at MSKCC for technical support. This work was supported by AIRC, Associazione Italiana per la Ricerca sul Cancro, Investigator Grants 20697 (to A. Bertotti), 21407 (to F. Di Nicolantonio), and 22802 (to L. Trusolino), AIRC IG 2018-ID 21923 project-PI Bardelli Alberto, AIRC under 5 per Mille 2018-ID 21091 program: principal investigator A. Bardelli, G.L. A. Bertotti, G.L. F. Di Nicolantonio, G.L. S. Marsoni, G.L. S. Siena, G.L. L. Trusolino, European Research Council Consolidator Grant 724748 – BEAT (to A. Bertotti), H2020 grant agreement no. 754923 COLOSSUS (to L. Trusolino), H2020 INFRAIA grant agreement no. 731105 EDIReX (to A. Bertotti), AIRC/CRUK/FC AECC Accelerator Award 22795 (to A. Bardelli and L. Trusolino), Fondazione Piemontese per la Ricerca sul Cancro-ONLUS, 5 × 1000 Ministero della Salute 2015 Project “STRATEGY” (to F. Di Nicolantonio), Fondazione Piemontese per la Ricerca sul Cancro-ONLUS, and 5 × 1000 Ministero della Salute 2015 Project “IMMUNOGENOMICA” (to A. Bardelli and L. Trusolino), 2014 and 2016 (to L. Trusolino). A. Bertotti and L. Trusolino are members of the EurOPDX Consortium. AIRC MFAG 2017 ID 20236 (to S. Arena). This work was also supported by the NIH R01 CA233736 (to R. Yaeger) and Cancer Center Core Grant P30 CA008748. This research is the responsibility of the authors and does not necessarily represent the official views of the NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.