Preclinical efficacy of carfilzomib in BRAF-mutant colorectal cancer models.

Serine/threonine-protein kinase B-raf (BRAF) mutations are found in 8-15% of colorectal cancer patients and identify a subset of tumors with poor outcome in the metastatic setting. We have previously reported that BRAF-mutant human cells display a high rate of protein production, causing proteotoxic stress, and are selectively sensitive to the proteasome inhibitors bortezomib and carfilzomib. In this work, we tested whether carfilzomib could restrain the growth of BRAF-mutant colorectal tumors not only by targeting cancer cells directly, but also by promoting an immune-mediated antitumor response. In human and mouse colorectal cancer cells, carfilzomib triggered robust endoplasmic reticulum stress and autophagy, followed by the emission of immunogenic-damage-associated molecules. Intravenous administration of carfilzomib delayed the growth of BRAF-mutant murine tumors and mobilized the danger-signal proteins calreticulin and high mobility group box 1 (HMGB1). Analyses of drug-treated samples revealed increased intratumor recruitment of activated cytotoxic T cells and natural killers, concomitant with the downregulation of forkhead box protein P3 (Foxp3)+ T-cell surface glycoprotein CD4 (CD4)+ T cells, indicating that carfilzomib promotes reshaping of the immune microenvironment of BRAF-mutant murine colorectal tumors. These results will inform the design of clinical trials in BRAF-mutant colorectal cancer patients.

Non-canonical antigens are the largest fraction of peptides presented by MHC class I in mismatch repair deficient murine colorectal cancer.

BACKGROUND: Immunotherapy based on checkpoint inhibitors is highly effective in mismatch repair deficient (MMRd) colorectal cancer (CRC). These tumors carry a high number of mutations, which are predicted to translate into a wide array of neoepitopes; however, a systematic classification of the neoantigen repertoire in MMRd CRC is lacking. Mass spectrometry peptidomics has demonstrated the existence of MHC class I associated peptides (MAPs) originating from non-coding DNA regions. Based on these premises we investigated DNA genomic regions responsible for generating MMRd-induced peptides. METHODS: We exploited mouse CRC models in which the MMR gene Mlh1 was genetically inactivated. Isogenic cell lines CT26 Mlh1+/+ and Mlh1-/- were inoculated in immunocompromised and immunocompetent mice. Whole genome and RNA sequencing data were generated from samples obtained before and after injection in murine hosts. First, peptide databases were built from transcriptomes of isogenic cell lines. We then compiled a database of peptides lost after tumor cells injection in immunocompetent mice, likely due to immune editing. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) and matched next-generation sequencing databases were employed to identify the DNA regions from which the immune-targeted MAPs originated. Finally, we adopted in vitro T cell assays to verify whether MAP-specific T cells were part of the in vivo immune response against Mlh1-/- cells. RESULTS: Whole genome sequencing analyses revealed an unbalanced distribution of immune edited alterations across the genome in Mlh1-/- cells grown in immunocompetent mice. Specifically, untranslated (UTR) and coding regions exhibited the largest fraction of mutations leading to highly immunogenic peptides. Moreover, the integrated computational and LC-MS/MS analyses revealed that MAPs originate mainly from atypical translational events in both Mlh1+/+ and Mlh1-/- tumor cells. In addition, mutated MAPs-derived from UTRs and out-of-frame translation of coding regions-were highly enriched in Mlh1-/- cells. The MAPs trigger T-cell activation in mice primed with Mlh1-/- cells. CONCLUSIONS: Our results suggest that-in comparison to MMR proficient CRC-MMRd tumors generate a significantly higher number of non-canonical mutated peptides able to elicit T cell responses. These results reveal the importance of evaluating the diversity of neoepitope repertoire in MMRd tumors.

Tolerance to colibactin correlates with homologous recombination proficiency and resistance to irinotecan in colorectal cancer cells.

The bacterial genotoxin colibactin promotes colorectal cancer (CRC) tumorigenesis, but systematic assessment of its impact on DNA repair is lacking, and its effect on response to DNA-damaging chemotherapeutics is unknown. We find that CRC cell lines display differential response to colibactin on the basis of homologous recombination (HR) proficiency. Sensitivity to colibactin is induced by inhibition of ATM, which regulates DNA double-strand break repair, and blunted by HR reconstitution. Conversely, CRC cells chronically infected with colibactin develop a tolerant phenotype characterized by restored HR activity. Notably, sensitivity to colibactin correlates with response to irinotecan active metabolite SN38, in both cell lines and patient-derived organoids. Moreover, CRC cells that acquire colibactin tolerance develop cross-resistance to SN38, and a trend toward poorer response to irinotecan is observed in a retrospective cohort of CRCs harboring colibactin genomic island. Our results shed insight into colibactin activity and provide translational evidence on its chemoresistance-promoting role in CRC.

Dual VEGFA/BRAF targeting boosts PD-1 blockade in melanoma through GM-CSF-mediated infiltration of M1 macrophages.

The introduction of targeted therapies represented one of the most significant advances in the treatment of BRAFV600E melanoma. However, the onset of acquired resistance remains a challenge. Previously, we showed in mouse xenografts that vascular endothelial growth factor (VEGFA) removal enhanced the antitumor effect of BRAF inhibition through the recruitment of M1 macrophages. In this work, we explored the strategy of VEGFA/BRAF inhibition in immunocompetent melanoma murine models. In BRAF mutant D4M melanoma tumors, VEGFA/BRAF targeting reshaped the tumor microenvironment, largely by stimulating infiltration of M1 macrophages and CD8+ T cells, and sensitized tumors to immune checkpoint blockade (ICB). Furthermore, we reported that the association of VEGFA/BRAF targeting with anti-PD-1 antibody (triple therapy) resulted in a durable response and enabled complete tumor eradication in 50% of the mice, establishing immunological memory. Neutralization and CRISPR-Cas-mediated editing of granulocyte-macrophage colony-stimulating factor (GM-CSF) abrogated antitumor response prompted by triple therapy and identified GM-CSF as the cytokine instrumental in M1-macrophage recruitment. Our data suggest that VEGFA/BRAF targeting in melanoma induces the activation of innate and adaptive immunity and prepares tumors for ICB. Our study contributes to understanding the tumor biology of BRAFV600E melanoma and suggests VEGFA as therapeutic target.

Genetic and pharmacological modulation of DNA mismatch repair heterogeneous tumors promotes immune surveillance.

Patients affected by colorectal cancer (CRC) with DNA mismatch repair deficiency (MMRd), often respond to immune checkpoint blockade therapies, while those with mismatch repair-proficient (MMRp) tumors generally do not. Interestingly, a subset of MMRp CRCs contains variable fractions of MMRd cells, but it is unknown how their presence impacts immune surveillance. We asked whether modulation of the MMRd fraction in MMR heterogeneous tumors acts as an endogenous cancer vaccine by promoting immune surveillance. To test this hypothesis, we use isogenic MMRp (Mlh1+/+) and MMRd (Mlh1-/-) mouse CRC cells. MMRp/MMRd cells mixed at different ratios are injected in immunocompetent mice and tumor rejection is observed when at least 50% of cells are MMRd. To enrich the MMRd fraction, MMRp/MMRd tumors are treated with 6-thioguanine, which leads to tumor rejection. These results suggest that genetic and pharmacological modulation of the DNA mismatch repair machinery potentiate the immunogenicity of MMR heterogeneous tumors.

Targeting the DNA Damage Response Pathways and Replication Stress in Colorectal Cancer.

PURPOSE: Genomic instability is a hallmark of cancer and targeting DNA damage response (DDR) is emerging as a promising therapeutic strategy in different solid tumors. The effectiveness of targeting DDR in colorectal cancer has not been extensively explored. EXPERIMENTAL DESIGN: We challenged 112 cell models recapitulating the genomic landscape of metastatic colorectal cancer with ATM, ATR, CHK1, WEE1, and DNA-PK inhibitors, in parallel with chemotherapeutic agents. We focused then on ATR inhibitors (ATRi) and, to identify putative biomarkers of response and resistance, we analyzed at multiple levels colorectal cancer models highly sensitive or resistant to these drugs. RESULTS: We found that around 30% of colorectal cancers, including those carrying KRAS and BRAF mutations and unresponsive to targeted agents, are sensitive to at least one DDR inhibitor. By investigating potential biomarkers of response to ATRi, we found that ATRi-sensitive cells displayed reduced phospho-RPA32 foci at basal level, while ATRi-resistant cells showed increased RAD51 foci formation in response to replication stress. Lack of ATM and RAD51C expression was associated with ATRi sensitivity. Analysis of mutational signatures and HRDetect score identified a subgroup of ATRi-sensitive models. Organoids derived from patients with metastatic colorectal cancer recapitulated findings obtained in cell lines. CONCLUSIONS: In conclusion, a subset of colorectal cancers refractory to current therapies could benefit from inhibitors of DDR pathways and replication stress. A composite biomarker involving phospho-RPA32 and RAD51 foci, lack of ATM and RAD51C expression, as well as analysis of mutational signatures could be used to identify colorectal cancers likely to respond to ATRi.

A modified fluctuation-test framework characterizes the population dynamics and mutation rate of colorectal cancer persister cells.

Compelling evidence shows that cancer persister cells represent a major limit to the long-term efficacy of targeted therapies. However, the phenotype and population dynamics of cancer persister cells remain unclear. We developed a quantitative framework to study persisters by combining experimental characterization and mathematical modeling. We found that, in colorectal cancer, a fraction of persisters slowly replicates. Clinically approved targeted therapies induce a switch to drug-tolerant persisters and a temporary 7- to 50-fold increase of their mutation rate, thus increasing the number of persister-derived resistant cells. These findings reveal that treatment may influence persistence and mutability in cancer cells and pinpoint inhibition of error-prone DNA polymerases as a strategy to restrict tumor recurrence.

CD4 T Cell-Dependent Rejection of Beta-2 Microglobulin Null Mismatch Repair-Deficient Tumors.

Inactivation of beta-2 microglobulin (B2M) is considered a determinant of resistance to immune checkpoint inhibitors (ICPi) in melanoma and lung cancers. In contrast, B2M loss does not appear to affect response to ICPis in mismatch repair-deficient (MMRd) colorectal tumors where biallelic inactivation of B2M is frequently observed. We inactivated B2m in multiple murine MMRd cancer models. Although MMRd cells would not readily grow in immunocompetent mice, MMRd B2m null cells were tumorigenic and regressed when treated with anti-PD-1 and anti-CTLA4. The efficacy of ICPis against MMRd B2m null tumors did not require CD8+ T cells but relied on the presence of CD4+ T cells. Human tumors expressing low levels of B2M display increased intratumoral CD4+ T cells. We conclude that B2M inactivation does not blunt the efficacy of ICPi in MMRd tumors, and we identify a unique role for CD4+ T cells in tumor rejection. SIGNIFICANCE: B2M alterations, which impair antigen presentation, occur frequently in microsatellite-unstable colorectal cancers. Although in melanoma and lung cancers B2M loss is a mechanism of resistance to immune checkpoint blockade, we show that MMRd tumors respond to ICPis through CD4+ T-cell activation.This article is highlighted in the In This Issue feature, p. 1601.

EGFR Blockade Reverts Resistance to KRASG12C Inhibition in Colorectal Cancer.

Most patients with KRAS G12C-mutant non-small cell lung cancer (NSCLC) experience clinical benefit from selective KRASG12C inhibition, whereas patients with colorectal cancer bearing the same mutation rarely respond. To investigate the cause of the limited efficacy of KRASG12C inhibitors in colorectal cancer, we examined the effects of AMG510 in KRAS G12C colorectal cancer cell lines. Unlike NSCLC cell lines, KRAS G12C colorectal cancer models have high basal receptor tyrosine kinase (RTK) activation and are responsive to growth factor stimulation. In colorectal cancer lines, KRASG12C inhibition induces higher phospho-ERK rebound than in NSCLC cells. Although upstream activation of several RTKs interferes with KRASG12C blockade, we identify EGFR signaling as the dominant mechanism of colorectal cancer resistance to KRASG12C inhibitors. The combinatorial targeting of EGFR and KRASG12C is highly effective in colorectal cancer cells and patient-derived organoids and xenografts, suggesting a novel therapeutic strategy to treat patients with KRAS G12C colorectal cancer. SIGNIFICANCE: The efficacy of KRASG12C inhibitors in NSCLC and colorectal cancer is lineage-specific. RTK dependency and signaling rebound kinetics are responsible for sensitivity or resistance to KRASG12C inhibition in colorectal cancer. EGFR and KRASG12C should be concomitantly inhibited to overcome resistance to KRASG12C blockade in colorectal tumors.See related commentary by Koleilat and Kwong, p. 1094.This article is highlighted in the In This Issue feature, p. 1079.

Vitamin C Restricts the Emergence of Acquired Resistance to EGFR-Targeted Therapies in Colorectal Cancer.

The long-term efficacy of the Epidermal Growth Factor Receptor (EGFR)-targeted antibody cetuximab in advanced colorectal cancer (CRC) patients is limited by the emergence of drug-resistant (persister) cells. Recent studies in other cancer types have shown that cells surviving initial treatment with targeted agents are often vulnerable to alterations in cell metabolism including oxidative stress. Vitamin C (VitC) is an antioxidant agent which can paradoxically trigger oxidative stress at pharmacological dose. Here we tested the hypothesis that VitC in combination with cetuximab could restrain the emergence of secondary resistance to EGFR blockade in CRC RAS/BRAF wild-type models. We found that addition of VitC to cetuximab impairs the emergence of drug persisters, limits the growth of CRC organoids, and significantly delays acquired resistance in CRC patient-derived xenografts. Mechanistically, proteomic and metabolic flux analysis shows that cetuximab blunts carbohydrate metabolism by blocking glucose uptake and glycolysis, beyond promoting slow but progressive ROS production. In parallel, VitC disrupts iron homeostasis and further increases ROS levels ultimately leading to ferroptosis. Combination of VitC and cetuximab orchestrates a synthetic lethal metabolic cell death program triggered by ATP depletion and oxidative stress, which effectively limits the emergence of acquired resistance to anti-EGFR antibodies. Considering that high-dose VitC is known to be safe in cancer patients, our findings might have clinical impact on CRC patients treated with anti-EGFR therapies.

High-dose vitamin C enhances cancer immunotherapy.

Vitamin C (VitC) is known to directly impair cancer cell growth in preclinical models, but there is little clinical evidence on its antitumoral efficacy. In addition, whether and how VitC modulates anticancer immune responses is mostly unknown. Here, we show that a fully competent immune system is required to maximize the antiproliferative effect of VitC in breast, colorectal, melanoma, and pancreatic murine tumors. High-dose VitC modulates infiltration of the tumor microenvironment by cells of the immune system and delays cancer growth in a T cell-dependent manner. VitC not only enhances the cytotoxic activity of adoptively transferred CD8 T cells but also cooperates with immune checkpoint therapy (ICT) in several cancer types. Combination of VitC and ICT can be curative in models of mismatch repair-deficient tumors with high mutational burden. This work provides a rationale for clinical trials combining ICT with high doses of VitC.

Adaptive mutability of colorectal cancers in response to targeted therapies.

The emergence of drug resistance limits the efficacy of targeted therapies in human tumors. The prevalent view is that resistance is a fait accompli: when treatment is initiated, cancers already contain drug-resistant mutant cells. Bacteria exposed to antibiotics transiently increase their mutation rates (adaptive mutability), thus improving the likelihood of survival. We investigated whether human colorectal cancer (CRC) cells likewise exploit adaptive mutability to evade therapeutic pressure. We found that epidermal growth factor receptor (EGFR)/BRAF inhibition down-regulates mismatch repair (MMR) and homologous recombination DNA-repair genes and concomitantly up-regulates error-prone polymerases in drug-tolerant (persister) cells. MMR proteins were also down-regulated in patient-derived xenografts and tumor specimens during therapy. EGFR/BRAF inhibition induced DNA damage, increased mutability, and triggered microsatellite instability. Thus, like unicellular organisms, tumor cells evade therapeutic pressures by enhancing mutability.

Reliance upon ancestral mutations is maintained in colorectal cancers that heterogeneously evolve during targeted therapies.

Attempts at eradicating metastatic cancers with targeted therapies are limited by the emergence of resistant subclones bearing heterogeneous (epi)genetic changes. We used colorectal cancer (CRC) to test the hypothesis that interfering with an ancestral oncogenic event shared by all the malignant cells (such as WNT pathway alterations) could override heterogeneous mechanisms of acquired drug resistance. Here, we report that in CRC-resistant cell populations, phylogenetic analysis uncovers a complex subclonal architecture, indicating parallel evolution of multiple independent cellular lineages. Functional and pharmacological modulation of WNT signalling induces cell death in CRC preclinical models from patients that relapsed during the treatment, regardless of the drug type or resistance mechanisms. Concomitant blockade of WNT and MAPK signalling restrains the emergence of drug-resistant clones. Reliance upon the WNT-APC pathway is preserved throughout the branched genomic drift associated with emergence of treatment relapse, thus offering the possibility of a common therapeutic strategy to overcome secondary drug resistance.

Inactivation of DNA repair triggers neoantigen generation and impairs tumour growth.

Molecular alterations in genes involved in DNA mismatch repair (MMR) promote cancer initiation and foster tumour progression. Cancers deficient in MMR frequently show favourable prognosis and indolent progression. The functional basis of the clinical outcome of patients with tumours that are deficient in MMR is not clear. Here we genetically inactivate MutL homologue 1 (MLH1) in colorectal, breast and pancreatic mouse cancer cells. The growth of MMR-deficient cells was comparable to their proficient counterparts in vitro and on transplantation in immunocompromised mice. By contrast, MMR-deficient cancer cells grew poorly when transplanted in syngeneic mice. The inactivation of MMR increased the mutational burden and led to dynamic mutational profiles, which resulted in the persistent renewal of neoantigens in vitro and in vivo, whereas MMR-proficient cells exhibited stable mutational load and neoantigen profiles over time. Immune surveillance improved when cancer cells, in which MLH1 had been inactivated, accumulated neoantigens for several generations. When restricted to a clonal population, the dynamic generation of neoantigens driven by MMR further increased immune surveillance. Inactivation of MMR, driven by acquired resistance to the clinical agent temozolomide, increased mutational load, promoted continuous renewal of neoantigens in human colorectal cancers and triggered immune surveillance in mouse models. These results suggest that targeting DNA repair processes can increase the burden of neoantigens in tumour cells; this has the potential to be exploited in therapeutic approaches.

Emergence of MET hyper-amplification at progression to MET and BRAF inhibition in colorectal cancer.

BACKGROUND: Combined MET and BRAF inhibition showed clinical benefit in a patient with rectal cancer carrying BRAFV600E and MET amplification. However after 4 months, acquired resistance emerged and the patient deceased shortly after disease progression. The mechanism of resistance to this drug combination is unknown. METHODS: We analysed plasma circulating tumour DNA obtained at progression by exome sequencing and digital PCR. MET gene and mRNA in situ hybridisation analyses in two bioptic specimens obtained at progression were used to confirm the plasma data. RESULTS: We identified in plasma MET gene hyper-amplification as a potential mechanism underlying therapy resistance. Increased MET gene copy and transcript levels were detected in liver and lymph node metastatic biopsies. Finally, transduction of MET in BRAF mutant colorectal cancer cells conferred refractoriness to BRAF and MET inhibition. CONCLUSIONS: We identified in a rectal cancer patient MET gene hyper-amplification as mechanism of resistance to dual BRAF and MET inhibition.

Heat-shock protein 27 (HSP27, HSPB1) is synthetic lethal to cells with oncogenic activation of MET, EGFR and BRAF.

The small heat-shock protein of 27 kDa (HSP27) is highly expressed in many cancers and is associated with aggressive tumour behaviour, metastasis, poor prognosis and resistance to chemotherapy. We aimed at assessing the role of HSP27 in modulating responses to target therapies. We selected several oncogene-addicted cancer cell lines, which undergo either cell cycle blockade or cell death in response to agents that target the specific oncogene. Surprisingly, HSP27 suppression alone resulted in the apoptotic death of MET-addicted EBC-1 lung cancer cells, epidermal growth factor receptor (EGFR)-addicted colorectal carcinoma (CRC) DiFi cells and BRAF-addicted CRC COLO205 and OXCO-1 and melanoma COLO741 cells, all of which also undergo death when treated with the specific targeted agent. In other cell lines, such as MET-addicted gastric carcinoma MKN45 and EGFR-addicted CRC SW48 lines, where oncogene inhibition only blocked proliferation, HSP27 knockdown made targeted agents switch from cytostatic to cytotoxic activity. Mechanistically, the more the cells were susceptible to HSP27 suppression, the more they were primed for death, as demonstrated by increased levels of mitochondrial outer membrane permeabilization. Priming for death was accompanied by the increase in pro-apoptotic proteins of the BCL2 family and of active caspase-3 and lamin B. Together, these data suggest that oncogene-addicted cells require HSP27 for survival and that HSP27 might interfere with the effectiveness of targeted agents.

Nucleolin Targeting Impairs the Progression of Pancreatic Cancer and Promotes the Normalization of Tumor Vasculature.

Pancreatic cancer is a highly aggressive tumor, mostly resistant to the standard treatments. Nucleolin is overexpressed in cancers and its inhibition impairs tumor growth. Herein, we showed that nucleolin was overexpressed in human specimens of pancreatic ductal adenocarcinoma (PDAC) and that the overall survival significantly increased in patients with low levels of nucleolin. The nucleolin antagonist N6L strongly impaired the growth of primary tumors and liver metastasis in an orthotopic mouse model of PDAC (mPDAC). Similar antitumor effect of N6L has been observed in a highly angiogenic mouse model of pancreatic neuroendocrine tumor RIP-Tag2. N6L significantly inhibited both human and mouse pancreatic cell proliferation and invasion. Notably, the analysis of tumor vasculature revealed a strong increase of pericyte coverage and vessel perfusion both in mPDAC and RIP-Tag2 tumors, in parallel to an inhibition of tumor hypoxia. Nucleolin inhibition directly affected endothelial cell (EC) activation and changed a proangiogenic signature. Among the vascular activators, nucleolin inhibition significantly decreased angiopoietin-2 (Ang-2) secretion and expression in ECs, in the tumor and in the plasma of mPDAC mice. As a consequence of the observed N6L-induced tumor vessel normalization, pre-treatment with N6L efficiently improved chemotherapeutic drug delivery and increased the antitumor properties of gemcitabine in PDAC mice. In conclusion, nucleolin inhibition is a new anti-pancreatic cancer therapeutic strategy that dually blocks tumor progression and normalizes tumor vasculature, improving the delivery and efficacy of chemotherapeutic drugs. Moreover, we unveiled Ang-2 as a potential target and suitable response biomarker for N6L treatment in pancreatic cancer. Cancer Res; 76(24); 7181-93. ©2016 AACR.

BCAM and LAMA5 Mediate the Recognition between Tumor Cells and the Endothelium in the Metastatic Spreading of KRAS-Mutant Colorectal Cancer.

PURPOSE: KRAS mutations confer adverse prognosis to colorectal cancer, and no targeted therapies have shown efficacy in this patient subset. Paracrine, nongenetic events induced by KRAS-mutant tumor cells are expected to result in specific deregulation and/or relocation of tumor microenvironment (TME) proteins, which in principle can be exploited as alternative therapeutic targets. EXPERIMENTAL DESIGN: A multimodal strategy combining ex vivo/in vitro phage display screens with deep-sequencing and bioinformatics was applied to uncover TME-specific targets in KRAS-mutant hepatic metastasis from colorectal cancer. Expression and localization of BCAM and LAMA5 were validated by immunohistochemistry in preclinical models of human hepatic metastasis and in a panel of human specimens (n = 71). The antimetastatic efficacy of two BCAM-mimic peptides was evaluated in mouse models. The role of BCAM in the interaction of KRAS-mutant colorectal cancer cells with TME cells was investigated by adhesion assays. RESULTS: BCAM and LAMA5 were identified as molecular targets within both tumor cells and TME of KRAS-mutant hepatic metastasis from colorectal cancer, where they were specifically overexpressed. Two BCAM-mimic peptides inhibited KRAS-mutant hepatic metastasis in preclinical models. Genetic suppression and biochemical inhibition of either BCAM or LAMA5 impaired adhesion of KRAS-mutant colorectal cancer cells specifically to endothelial cells, whereas adhesion to pericytes and hepatocytes was unaffected. CONCLUSIONS: These data show that the BCAM/LAMA5 system plays a functional role in the metastatic spreading of KRAS-mutant colorectal cancer by mediating tumor-TME interactions and as such represents a valuable therapeutic candidate for this large, currently untreatable patient group. Clin Cancer Res; 22(19); 4923-33. ©2016 AACR.

Clonal evolution and resistance to EGFR blockade in the blood of colorectal cancer patients.

Colorectal cancers (CRCs) evolve by a reiterative process of genetic diversification and clonal evolution. The molecular profile of CRC is routinely assessed in surgical or bioptic samples. Genotyping of CRC tissue has inherent limitations; a tissue sample represents a single snapshot in time, and it is subjected to spatial selection bias owing to tumor heterogeneity. Repeated tissue samples are difficult to obtain and cannot be used for dynamic monitoring of disease progression and response to therapy. We exploited circulating tumor DNA (ctDNA) to genotype colorectal tumors and track clonal evolution during treatment with the epidermal growth factor receptor (EGFR)-specific antibodies cetuximab or panitumumab. We identified alterations in ctDNA of patients with primary or acquired resistance to EGFR blockade in the following genes: KRAS, NRAS, MET, ERBB2, FLT3, EGFR and MAP2K1. Mutated KRAS clones, which emerge in blood during EGFR blockade, decline upon withdrawal of EGFR-specific antibodies, indicating that clonal evolution continues beyond clinical progression. Pharmacogenomic analysis of CRC cells that had acquired resistance to cetuximab reveals that upon antibody withdrawal KRAS clones decay, whereas the population regains drug sensitivity. ctDNA profiles of individuals who benefit from multiple challenges with anti-EGFR antibodies exhibit pulsatile levels of mutant KRAS. These results indicate that the CRC genome adapts dynamically to intermittent drug schedules and provide a molecular explanation for the efficacy of rechallenge therapies based on EGFR blockade.