Research ArticleCancer

Blockade of EGFR and MEK Intercepts Heterogeneous Mechanisms of Acquired Resistance to Anti-EGFR Therapies in Colorectal Cancer

See allHide authors and affiliations

Science Translational Medicine  19 Feb 2014:
Vol. 6, Issue 224, pp. 224ra26
DOI: 10.1126/scitranslmed.3007947


Colorectal cancers (CRCs) that are sensitive to the anti–epidermal growth factor receptor (EGFR) antibodies cetuximab or panitumumab almost always develop resistance within several months of initiating therapy. We report the emergence of polyclonal KRAS, NRAS, and BRAF mutations in CRC cells with acquired resistance to EGFR blockade. Regardless of the genetic alterations, resistant cells consistently displayed mitogen-activated protein kinase kinase (MEK) and extracellular signal–regulated kinase (ERK) activation, which persisted after EGFR blockade. Inhibition of MEK1/2 alone failed to impair the growth of resistant cells in vitro and in vivo. An RNA interference screen demonstrated that suppression of EGFR, together with silencing of MEK1/2, was required to hamper the proliferation of resistant cells. Indeed, concomitant pharmacological blockade of MEK and EGFR induced prolonged ERK inhibition and severely impaired the growth of resistant tumor cells. Heterogeneous and concomitant mutations in KRAS and NRAS were also detected in plasma samples from patients who developed resistance to anti-EGFR antibodies. A mouse xenotransplant from a CRC patient who responded and subsequently relapsed upon EGFR therapy showed exquisite sensitivity to combinatorial treatment with MEK and EGFR inhibitors. Collectively, these results identify genetically distinct mechanisms that mediate secondary resistance to anti-EGFR therapies, all of which reactivate ERK signaling. These observations provide a rational strategy to overcome the multifaceted clonal heterogeneity that emerges when tumors are treated with targeted agents. We propose that MEK inhibitors, in combination with cetuximab or panitumumab, should be tested in CRC patients who become refractory to anti-EGFR therapies.


The development of secondary (acquired) resistance is a common feature of anticancer treatment. The identification of mechanisms underlying secondary resistance can lead to rational development of additional lines of therapy. However, recent evidence indicates that a variety of resistance mechanisms to targeted drugs can emerge. When BRAF-mutant melanomas are exposed to BRAF inhibitors, surviving cells display NRAS, mitogen-activated protein kinase (MAPK) kinase 1/2 (MEK1/2) or COT mutations, PDGFRB (platelet-derived growth factor receptor B) overexpression, BRAF gene amplification, or expression of BRAF truncated forms (15). Similarly, lung cancers carrying epidermal growth factor receptor (EGFR) mutations escape treatment with the highly specific inhibitors erlotinib and gefitinib through EGFR secondary mutations (T790M), MET amplification, HER2 amplification, MAPK1 amplification, PIK3CA mutations, AXL up-regulation, and other mechanisms (612).

These discoveries suggest that every patient with a malignancy that becomes refractory to a targeted therapy should ideally undergo a tumor biopsy, allowing the assessment of the mechanism(s) of resistance through extensive costly molecular analyses. Even if this level of sophistication could be eventually achieved in the clinical setting, advanced-stage patients often have multiple metastatic lesions in different organs. It is likely that the biopsied lesion would not be representative of the heterogeneity of the entire disease burden at relapse (13).

The realization that relapsed tumors are highly molecularly heterogeneous poses a formidable therapeutic challenge, as it would seem quite difficult to overcome the multiple resistance mutations that arise in individual patients. We hypothesized that the plethora of molecular determinants causative of secondary resistance would ultimately converge downstream in the signaling cascade at a limited number of critical points.

To test this assumption, we studied secondary resistance to the monoclonal antibodies (mAbs) cetuximab and panitumumab. These drugs are mostly effective in colorectal cancer (CRC) patients who do not carry mutations in KRAS, NRAS, or BRAF (14, 15). In these individuals, response to EGFR blockade can result in remarkable tumor regressions, but these are short-lived (16) and progression occurs within months (17).

We initially identified several cell lines with molecular features representing a subset of patients who respond to cetuximab and/or panitumumab. These were then treated with anti-EGFR antibodies until resistant derivatives emerged. In accordance with our recent report, resistant cells frequently displayed KRAS mutations (18). We noted, however, that the resistant cell population was heterogeneous, and not all cells carried KRAS mutations. This phenomenon is reminiscent of what we and others have found in clinical samples from patients who respond and then progress upon treatment with cetuximab or panitumumab (18, 19). This prompted us to define additional mechanisms of resistance to EGFR blockade and to assess the pathway they control because this is a prerequisite to overcome acquired resistance.


Distinct genetic events drive acquired resistance to anti-EGFR therapy in CRC cells

To generate models of acquired resistance to anti-EGFR–targeted therapies, we took advantage of a panel of CRC cell lines that we found to be highly sensitive to cetuximab and/or panitumumab. These included DiFi, LIM1215, HCA-46, NCIH508, and OXCO-2 cells. Mutational profiling showed that these cell lines are wild type for KRAS (exons 2 to 4), NRAS (exons 2 and 3), HRAS (exons 2 and 3), BRAF (exon 15), and PIK3CA (exons 9 and 20), with the exception of NCIH508, which carries a PIK3CA exon 9 mutation (p.E545K). These findings in cell lines are consistent with the clinical observations about the genetic status of CRC patients most likely to respond to cetuximab and panitumumab (20, 21).

Cells were exposed to cetuximab or panitumumab until resistant populations emerged. Derivative cells resistant to cetuximab (referred to as R-cetux) or panitumumab (which we called R-panit) displayed cross-resistance to both mAbs, consistent with what has been reported in patients treated with panitumumab after previous cetuximab-based treatment (22) (Fig. 1A).

Fig. 1. CRC cell lines resistant to cetuximab or panitumumab carry KRAS, NRAS, or BRAF mutations.

(A) DiFi, HCA-46, LIM1215, NCIH508, and OXCO-2 cetuximab (R-cetux)– and panitumumab (R-panit)–resistant cells were treated with cetuximab (Cmab, in black) or with panitumumab (Pmab, in gray) for 1 week, with increasing concentrations of the two drugs. Parental cells treated with cetuximab (red circles) or with panitumumab (red triangles) were included as controls. The name of the EGFR-targeted mAb used in each viability assay is indicated in parentheses. Cell viability was measured by the adenosine triphosphate (ATP) assay. Data points represent means ± SD of three independent experiments. (B) List of genes and molecular alterations detected in cetuximab- and panitumumab-resistant derivatives. LIM1215 R1- and R4-cetux, LIM1215 R-panit, NCIH508 R-panit, OXCO-2 R1-cetux, OXCO-2 R2-cetux, and OXCO-2 R-panit contain multiple mutations in the same cell population. The nomenclature -R1 -R2 -R3 -R4 indicate cetuximab-resistant populations, independently derived from the parental cell line.

To define the molecular mechanisms underlying acquired resistance, we initially performed Sanger sequencing of candidate genes, which are known to confer primary resistance to EGFR blockade in CRC (20). These included KRAS, BRAF, NRAS, and PIK3CA. Mutational profiling of the resistant populations revealed molecular alterations in KRAS, BRAF, and NRAS, but not in PIK3CA (Fig. 1B). OXCO-2– and LIM1215-resistant populations contained multiple mutations. Most of the changes affected residues known to activate the corresponding protein, such as KRAS position 12, BRAF position 600, and NRAS positions 12 and 61 (Fig. 1B and fig. S1, A and B). A recent study indicated that CRC tumors resistant to cetuximab may acquire a secondary mutation in the extracellular domain of EGFR (S492R), which would prevent drug binding (23). Sequencing of the EGFR ectodomain in genomic DNA and complementary DNA obtained from resistant clones ruled out the presence of this mutation. Previous reports showed that gene amplification of KRAS, HER2, or MET can also confer resistance to anti-EGFR–targeted therapies (18, 2427). To assess whether gene copy variations could have emerged during drug selection, we performed real-time quantitative polymerase chain reaction (PCR) on the same panel of candidates. This analysis revealed amplification of KRAS in DiFi R-cetux, HCA-46 R1-cetux, and NCIH508 R-cetux and, to a lesser extent, in NCIH508 R-panit cell lines. Amplification of the MET or HER2 genes was not observed (fig. S2, A and B). We did not detect an increase in KRAS gene copy number in any of the parental lines. However, we previously reported that very rare cells in the DiFi population display KRAS overexpression by immunohistochemical (IHC) analysis (18). We cannot exclude that a few KRAS-amplified cells (below the limit of detection of the real-time quantitative PCR assay) also exist in the other parental cell lines. To further address this aspect, we performed IHC (with KRAS antibodies) and fluorescence in situ hybridization (FISH) analysis (with a KRAS gene probe) of HCA-46 and NCIH508 parental and resistant cell populations. IHC and FISH confirmed the presence of KRAS-amplified cells in the resistant population and indicated that the parental cells do not display this molecular alteration (fig. S3, A and B).

To formally assess whether different genetic alterations occurred in a mutually exclusive pattern, we cloned the populations of resistant cells. We found that single colonies could not be efficiently isolated from HCA-46 and NCIH508 cell lines. We therefore focused on LIM1215 R1-cetux, LIM1215 R2-cetux, and OXCO-2 R1-cetux cells, which proved capable of generating individual clones upon limiting dilution. All LIM1215 R2-cetux derivative clones (70 of 70) carried the KRAS G13D mutation, confirming that the resistant population was homogeneous. On the contrary, of 13 clones derived from LIM1215 R1-cetux, 4 carried the KRAS G12R, 4 displayed the KRAS K117N variant (28), and the remaining 5 clones carried the NRAS G12C mutation. OXCO-2 R-cetux derivative clones (66 clones in total) showed either KRAS G13D (42 clones) or BRAF V600E mutations (24 clones). To obtain more accurate quantification of the percentage at which each allele was present in the resistant populations, we performed BEAMing analysis on a subset of the resistant derivatives (table S1).

The resistance protocol was then repeated starting from an independent batch of parental cells (LIM1215, OXCO-2, and HCA-46) to assess the reproducibility of the approach and to evaluate the emergence of additional variants. LIM1215 R4-cetux again displayed a mixed population of KRAS G13D and KRAS G12D. OXCO-2 R2-cetux contained NRAS G12C, G12D, and G13D. HCA-46 R2-cetux displayed KRAS G13D mutation.

Overall, these results suggest that (i) multiple genetic mechanisms can drive resistance to EGFR blockade and (ii) several subclones often coexist in the population that emerges after selection with anti-EGFR therapies.

Genetic alterations in EGFR signaling biochemically converge to activate MEK and ERK

As described above, escape from EGFR blockade in CRC cells is associated with the emergence of distinct alterations in several genes. We hypothesized that the plethora of molecular determinants causative of acquired drug resistance would ultimately converge at a limited number of signaling switches, which, in turn, could be rationally targeted by further lines of therapy. To test this assumption, we performed biochemical analysis of the resistant derivatives. For these studies, we elected to use LIM1215 R-panit (KRAS G13D, NRAS G12C), OXCO-2 R1-cetux (KRAS G12D, BRAF V600E), HCA46 R-panit (KRAS G12C), and NCIH508 R-cetux (KRAS-amplified) because they are representative of the distinct resistance mechanisms that emerged upon selection with cetuximab and panitumumab (Fig. 1B). We started by assessing levels of RAS activation using a CRAF pull-down assay. We found that resistant cells displayed increased guanosine triphosphate–bound KRAS compared to the matched parental cells (fig. S4). We next assessed the phosphorylation status of EGFR and its downstream effectors MEK and ERK (extracellular signal–regulated kinase) by Western blot analysis. We found that MEK and/or ERK were consistently activated in resistant cells compared to their parental counterparts (Fig. 2, A to D). Of relevance, sustained MEK or ERK activation was present in the resistant derivatives independent of the genetic mechanism of resistance. These results revealed that CRC cells become refractory to anti-EGFR therapies by increasing signaling output through the MEK-ERK pathway.

Fig. 2. Genetic alterations in the EGFR signaling cascade biochemically converge to activate MEK and ERK.

The indicated cell lines were analyzed for EGFR-MAPK pathway activation. Parental and resistant cells were treated with cetuximab (1.4 μM) or panitumumab (1.4 μM), after which whole-cell extracts were subjected to Western blot analysis and compared to untreated cells with phospho-EGFR (Tyr1068), total EGFR, total MEK1/2 and phospho-MEK1/2, and total ERK1/2 and phospho-ERK1/2 antibodies. Vinculin was included as a loading control. The genetic status of the resistant derivatives is listed below the blots.

Resistant cells are sensitive to concomitant silencing of the EGFR and MEK1/2 genes

We assessed whether the genetic alterations associated with the emergence of resistance to anti-EGFR drugs might underlie functional dependencies, which could be exploited to design therapeutic strategies for tumors with secondary resistance to cetuximab and panitumumab.

As a first step, we used a small interfering RNA (siRNA)–mediated functional screening to identify genes that, alone or in combination, could suppress the growth of resistant cells. The RNA interference assay was devised to interrogate candidate genes known to be involved in resistance to cetuximab or panitumumab and/or effectors of EGFR-initiated signaling. These included KRAS, NRAS, HRAS, BRAF, CRAF, MEK1/2, HER2, HER3, PIK3CA, and AKT1 (Fig. 3A). To confirm the specificity and the efficacy of the assays, siRNA-mediated suppression was followed by Western blotting with antibodies against the individual targeted gene products. As shown in fig. S5, individual siRNA effectively and specifically suppressed the designated target gene. The siRNA-mediated functional screen revealed that resistant cells were usually dependent on the expression of the genes whose alterations emerged during the selection procedure. For example, KRAS-mutant/KRAS-amplified cells (LIM1215-R panit, HCA46-R panit, and NCIH508R-cetux) were sensitive to KRAS suppression (Fig. 3A). Notably, LIM1215 R-panit did not show growth impairment upon NRAS silencing despite harboring an NRAS mutation. This result is consistent with our observations that NRAS-mutant cells are a minor subclone in the bulk population as assessed by BEAMing analysis (table S1). OXCO-2 R1-cetux cells, which are a mixture of KRAS- and BRAF-mutant cells, were greatly sensitive to combined suppression of KRAS + BRAF (Fig. 3B). The finding that siRNA-mediated suppression of MEK1/2 only marginally affected the growth of resistant cells was unexpected. We therefore assessed whether pimasertib, a selective allosteric MEK inhibitor, which is being tested in clinical trials (29, 30), might have similar impact. Consistently with the result obtained by siRNA analysis, we found that pharmacological inhibition of MEK was unable to successfully block the growth of resistant cells (fig. S6).

Fig. 3. Cell lines with acquired resistance to anti-EGFR antibodies are sensitive to suppression of EGFR and MEK1/2.

(A) siRNA suppression screening of genes involved in the EGFR pathway was performed using the indicated cell lines. The genetic status of the individual cell models is shown below the graph. Effective impairment of cell proliferation was arbitrarily set at less than 50% (dashed line). The survival fraction was determined by ATP assay. Data points represent means ± SD of three independent experiments. Black, HER family receptors; red, RAS family members; gray, RAS downstream effectors. (B) Dual silencing of EGFR together with individual EGFR pathway effectors. Effective impairment of cell proliferation was arbitrarily set at less than 50% (dashed line). The survival fraction was assayed by the ATP assay. Data points represent means ± SD of three independent experiments. Black, combination of EGFR + RAS effectors; gray, combination of KRAS + BRAF; red, combination of KRAS + MEK1/2.

Resistance to EGFR blockade is reversed by concomitant inhibition of EGFR and MEK

We reasoned that dependencies due to activation of signaling pathways associated with drugs inhibiting the EGFR might become apparent only when the receptor itself was concomitantly targeted. To directly test this hypothesis, we performed siRNA-mediated suppression of signaling effectors in association with silencing of the EGFR. Concomitant silencing of EGFR and MEK1/2 was the only combination capable of reducing the survival fraction below 50% in all four cell models (Fig. 3B). Parallel experiments in which concomitant targeting of the EGFR and MEK was achieved by combining siRNA suppression and pharmacological treatment confirmed these results (fig. S7).

The finding that cells that developed resistance to EGFR blockade display constitutive activation of MEK but are only modestly affected by MEK inhibition is intriguing. To study this further, we performed biochemical analyses in the presence of EGFR and MEK inhibitors (cetuximab and pimasertib, respectively) alone or in combination. To verify the kinetics of this effect, we performed time course pharmacological treatments. Notably, we found that in the absence of cetuximab, the MEK inhibitor pimasertib leads to efficient phospho-ERK suppression. This is transient, however, and after a few hours, ERK became phosphorylated again (Fig. 4). Concomitant with ERK reactivation, we observed increased phosphorylation of EGFR. Phosphorylation of the receptor increased over time and was maximal after 6 to 12 hours depending on the cell model. When the same experiment was performed in the presence of cetuximab, EGFR phosphorylation was suppressed, and this was accompanied by prolonged abrogation of ERK phosphorylation (Fig. 4). These results are consistent across multiple cell models of acquired resistance to EGFR blockade, irrespectively of their mutational status.

Fig. 4. MEK inhibition induces EGFR activation, resulting in ERK activation.

Resistant cells were treated with pimasertib (50 nM) with or without cetuximab (340 nM) at the indicated time points, after which whole-cell extracts were subjected to Western blot analysis. Whole-cell extracts were blotted with phospho-EGFR (Tyr1068), total EGFR, phospho-MEK1/2, total MEK1/2, phospho-ERK, and total ERK. Actin was included as a loading control. NT, not treated.

Concomitant inhibition of EGFR and MEK1/2 with cetuximab and pimasertib was effective not only in transient cell growth experiments but also when the drugs were tested in long-term proliferation assays (Fig. 5A). We next assessed the efficacy of the concomitant EGFR-MEK blockade in vivo. Pools of OXCO-2 and NCIH508 cetuximab-resistant cells were injected in nude mice. After tumors were established, mice were treated with vehicle, cetuximab and/or pimasertib alone, or the combination of the two agents. OXCO-2 R1-cetux tumors treated with cetuximab proliferated at a comparable rate to vehicle-treated xenografts. Pimasertib as a single agent delayed tumor growth without inducing regression. In contrast, the combination of cetuximab and pimasertib induced a remarkable reduction in tumor volume compared to baseline. Similar results were obtained in NCIH508 R-cetux xenografts (Fig. 5B and table S2). The two models confirmed that combinatorial treatment is effective in inducing tumor shrinkage in cetuximab-resistant cells irrespective of their mechanism of resistance.

Fig. 5. Resistance to EGFR therapy is reversed by pharmacological inhibition of EGFR and MEK.

(A) The indicated cell lines were treated for 2 weeks with increasing concentrations of cetuximab and pimasertib. At the end of the assay, cells were fixed and stained with crystal violet solution. The genetic status of each individual cell model is indicated. (B) Combinatorial treatment with cetuximab plus pimasertib is effective in inducing tumor shrinkage in vivo. OXCO-2 R1-cetux and NCIH508 R-cetux were injected subcutaneously in nude mice and then treated with cetuximab or pimasertib as monotherapy, with the combination of the two drugs, or with vehicle. n = 7 mice per group for OXCO-2 R1-cetux and 4 mice per group for NCIH508 R-cetux.

Plasma samples from patients who develop resistance to EGFR blockade display KRAS and NRAS mutations

The cell-based findings suggested that upon EGFR blockade, multiple resistant clones emerge, and that resistance is often driven by genetically distinct mechanisms. To assess the clinical relevance of these findings, we examined samples from metastatic CRC patients who received anti-EGFR antibodies. We hypothesized that distinct genetic events would also be observed in samples (biopsies) from CRC patients who initially responded and then became refractory to either cetuximab or panitumumab. Needle biopsies are particularly difficult to obtain in this setting for a number of reasons, including the intrinsic risk of this invasive procedure (31). The scenario is further complicated by the fact that patients with metastatic CRC usually have multiple lesions. Therefore, biopsies represent only a snapshot of the overall disease and, accordingly, are not well suited to monitor the emergence of resistant clones, which can be located in distant metastatic lesions. To overcome these limitations, at least in part, we have implemented a “liquid biopsy” approach to analyze circulating free tumor DNA because it is more likely to capture the overall genetic complexity of tumors in patients with advanced disease. We exploited the highly sensitive BEAMing technique to measure tumor-derived DNA mutations in the blood of patients (18, 19, 32, 33). We obtained plasma samples from four patients who responded and then became refractory to either cetuximab or panitumumab. BEAMing probes designed to identify the same somatic variants that were found in cell lines were used to monitor for KRAS, BRAF, and NRAS mutations in plasma. Notably, in two cases (patients #1 and #2), multiple KRAS variants were detected at relapse but not at baseline, suggesting the emergence of several independent clones during treatment (Fig. 6A). In patient #2, the concomitant presence of KRAS and NRAS mutations was observed in the relapse sample (Fig. 6B). The same occurred at relapse in patient #3 (Fig. 6C). In the last patient (#4), mutations in KRAS, NRAS, and BRAF were not found at progression (the baseline plasma was not available for patients #3 and 4). These results suggest that therapy with anti-EGFR antibodies selects multiple clones carrying heterogeneous patterns of mutations, a situation akin with what we observed in preclinical models. Analogous results were obtained by Bettegowda and colleagues as described in the companion manuscript (34).

Fig. 6. Tumors resistant to anti-EGFR therapy show a characteristic mutation pattern and respond to combined EGFR-MEK inhibition.

(A to C) BEAMing analysis of circulating tumor DNA of four patients with acquired resistance to cetuximab (patient #1) or panitumumab (patients #2 to #4) displays complex patterns of KRAS and NRAS mutations. (D and E) Mouse xenografts from a tumor at relapse after anti-EGFR therapy respond to combined EGFR-MEK inhibition. (D) Xenografted tumors retained the histopathologic characteristics of original samples. Hematoxylin and eosin stains of the biopsy taken from the original CRC lung metastasis and a histological sample derived from the same lesion engrafted in a nonobese diabetic–severe combined immunodeficient (NOD-SCID) mouse. Scale bars, 50 μm. (E) Waterfall plot showing the percent change in volume for the individual tumors in each arm at the time of sacrifice. Tumor volumes were normalized individually to their volumes at treatment day 1.

A patient-derived CRC xenograft with acquired resistance is sensitive to EGFR and MEK inhibition

Using an approach we previously optimized (27), we generated a mouse xenotransplant [patient-derived xenograft (PDX)] from a lung metastasis of a CRC patient who responded and subsequently relapsed upon anti-EGFR therapy (cetuximab). This tumor carried a KRAS mutation (A146T), which is identical to one of the KRAS variants we found in LIM1215 R3-cetux cells (Fig. 1B). After implantation and engraftment of the patient sample in a NOD-SCID mouse, the tumor was passaged and expanded for two generations. The morphological features of the biopsy obtained from the lung metastasis and a specimen from the xenotransplant grown in NOD-SCID mice were compared. As shown in Fig. 6D, xenografted tumors retained the histopathological characteristics of their original patient counterpart. We also confirmed the presence of the KRAS mutation in the xenopatient by Sanger and BEAMing analysis. The xenografted tumor was serially transplanted until production of four cohorts, each consisting of six mice. These were randomized to vehicle alone, cetuximab monotherapy, pimasertib monotherapy, and their combination. Notably, cetuximab or the MEK inhibitor pimasertib had limited effectiveness, whereas combinatorial (cetuximab-pimasertib) treatment prominently impaired tumor growth and induced moderate shrinkage (Fig. 6E, fig. S8, and table S2).


Recent evidence that cancer genomes (especially in the case of metastatic tumors) are highly heterogeneous (13) may explain why treatment with individual drugs (such as BRAF inhibitors or anti-EGFR antibodies) is only transiently effective. A cancer detectable by computed tomography scanning contains billions of cells. Mathematical models indicate that in a single tumor mass (for example, a CRC liver metastasis), there are hundreds to thousands of cells that are already resistant to a given targeted agent even before treatment is initiated (19). At first glance, overcoming resistance therefore appears to be an insurmountable task. We decided to focus on metastatic CRC (35) and the anti-EGFR mAbs cetuximab and panitumumab as a model system to study how, at least in principle, this problem could be tackled. We reasoned that although alterations in multiple genes could confer resistance to EGFR blockade, their activation may ultimately converge downstream in the signaling pathway that was initially targeted (in this case, the EGFR-MAPK pathway). Accordingly, this work was initiated on the basis two premises: (i) tumors that develop resistance to EGFR blockade would be genetically heterogeneous and (ii) activation of pathways conferring resistance would eventually converge on key downstream effector(s). Starting from several CRC cellular models that are highly sensitive to cetuximab and/or panitumumab, we obtained resistant derivatives by continuous drug exposure. In most cases, KRAS alterations (mutation or amplification) could be detected at resistance, in accordance with what has previously been reported (18, 19). Using several methodologies (sequencing and cloning), we found that the resistant populations were often highly heterogeneous and contained several KRAS alleles. Furthermore, we discovered that in addition to KRAS, NRAS and BRAF activating mutations could also be detected in several cellular models. This led us to postulate that the same might occur in patients. We reasoned that tissue biopsies would only offer a snapshot of the overall tumor mass and might therefore be ill suited to capture the multiclonal feature of the resistant disease. We therefore analyzed circulating free DNA, a form of liquid biopsy, which allows tumors to be genotyped using a blood sample from patients. Notably, we found that plasma samples of patients treated with anti-EGFR antibodies carried multiple distinct KRAS and NRAS alleles. These results, together with those described in a related manuscript in which circulating free DNA from patients treated with anti-EGFR antibodies was also analyzed (34), provide evidence that alterations in multiple genes are concomitantly associated with acquired resistance to EGFR blockade. The finding that, within a single patient, resistance to EGFR blockade can be associated with mutations in several genes (such as KRAS and NRAS) underscores the molecular heterogeneity of resistant tumors. It should be acknowledged, however, that a large number of samples from resistant patients will be required to establish the prevalence and the extent of this phenomenon.

These results also highlight a marked overlap between genes that, when mutated, drive the so-called primary (de novo) resistance and those that allegedly sustain secondary (acquired) resistance. It was previously reported that alterations in KRAS, BRAF, NRAS, MET, and HER2 can drive both de novo and acquired resistance (17, 20, 2427). Both inter- and intrapatient tumor genetic heterogeneities have been observed, with the extent of the latter likely reflecting the point at which the alteration was acquired during tumorigenesis. For example, de novo KRAS mutations occur early during tumor progression, whereas the emergence of KRAS mutations upon EGFR targeting occurs when the tumor is treated (usually at the stage of metastatic disease). In this case, the drug pressure is applied simultaneously, often to multiple metastases, which are genetically highly heterogeneous. Consequently, the spectrum of alleles driving secondary resistance appears evident even when only a few patients are examined. Overall, these results support the concept that primary and acquired resistances are driven by the same alterations.

We formulated the hypothesis that the molecular alterations emerging after treatment with a targeted agent would activate genes involved in the same pathway that was besieged by the selective pressure. We further assumed that the signals sustained by the mutant proteins would ultimately converge on a distinct downstream effector. Notably, these data mirror what was previously reported in cellular models and clinical samples of BRAF-mutant melanoma in which, independently of the mechanisms driving acquired resistance, reactivation of MAPK signaling is almost invariably detected (36, 37). Accordingly, clinical trials based on rechallenging melanomas with anti-MEK drugs have been designed (38). Collectively, biochemical analyses and reverse genetic experiments corroborate this hypothesis. We report that, regardless of the gene/mutation that confers resistance, the net output is invariably sustained activation (constitutive phosphorylation) of MEK and ERK, thus defining an example of convergent evolution. These data provide a rationale for overcoming resistance to EGFR antibodies using MEK inhibitors, many of which have already reached the clinic (30). We found that blockade of MEK is not sufficient to inhibit the proliferation of resistant cells. Using siRNA screening and pharmacological treatments, we found that cells that acquired resistance to cetuximab or panitumumab (through KRAS, BRAF, or NRAS mutations) are dependent on concomitant blockade of EGFR and MEK. The requirement for EGFR-MEK concomitant blockade is evident not only in vitro but also in vivo, as shown by mouse xenograft experiments. Biochemical time course analysis showed that MEK inhibition leads to transient ERK inactivation, which is followed by phosphorylation of the receptor. The latter is blocked by the addition of cetuximab, thus explaining why the double hit (EGFR-MEK) is required. The mechanism of EGFR feedback activation after MEK blockade is reminiscent of what is observed when BRAF-mutant CRCs are treated with BRAF or MEK inhibitors (39). Although our work suggests that MEK/ERK reactivation is linked to rephosphorylation of the EGFR receptor, the moderate intensity of EGFR phosphorylation suggests that additional signaling routes may also contribute. To overcome the limits of our study, further investigations are needed to elucidate the precise biochemical players underlying the effectiveness of the EGFR-MEK combination. However, the data generated in cell models were confirmed in a tumor xenograft derived from a patient who responded and then relapsed upon cetuximab therapy, thus supporting the clinical relevance of the findings.

Overall, the results presented in this manuscript define genetic alterations linked to secondary resistance to anti-EGFR mAbs in preclinical models and CRC patients. Considering that in the clinical setting anti-EGFR antibodies are often used in combination with chemotherapy (40, 41), it will be of high interest to explore whether and to what extent treatment with chemotherapy affects the emergence of genetic alterations upon EGFR blockade.

In conclusion, our study provides a rationale for clinical trials involving concomitant inhibition of EGFR and MEK in CRC patients who develop acquired resistance to cetuximab and panitumumab.


Study design

This study was designed to define the mechanisms of acquired resistance to EGFR blockade in CRC and to define pharmacological strategies to overcome acquired resistance to cetuximab and panitumumab. In the first section of this study, CRC cell lines that developed acquired resistance to anti-EGFR therapies were molecularly characterized. In the second part, a subset of these cell models was analyzed at a biochemical level and with siRNA screening to identify signaling nodes that sustain resistance and can be pharmacologically targeted. Western blot analyses were performed at least twice, starting with independent cell lysates. siRNA screening was performed using a pool of four different siRNAs for each target gene, and the results represent an average of three independent experiments. In the third section of the study, cell lines with acquired resistance to EGFR blockade were treated with drugs targeting proteins involved in signaling pathways. In vitro drug inhibition assays were performed in triplicate or quadruplicate. Cell lines were implanted in immunocompromised mice, which were then randomized to establish homogeneous treatment arms before drug administration. In the fifth section of the study, we used the highly sensitive BEAMing technology to detect genetically heterogeneous KRAS and RAS mutations in plasma samples from four CRC patients who relapsed after anti-EGFR treatment. In the last section, we established a tumor xenograft from a patient who relapsed after responding to EGFR blockade, which was then used to assess the combinatorial treatment initially tested in cell lines. The tumor sample was initially engrafted in one NOD-SCID mouse. After expansion in a large cohort of mice, xenografted tumors were randomized to establish homogeneous treatment arms for drug administration.

Statistical analysis

All data from the proliferation assay are presented as means ± SD of at least three independent experiments, each with three experimental replicates. siRNA screenings are presented as means ± SD of three independent experiments, each with four experimental replicates. In vivo experimental data points represent means ± SEM of the measurements of each mouse tumor.


Materials and Methods

Fig. S1. Mutational profiling of candidate genes in resistant cells.

Fig. S2. Gene copy number (GCN) analysis in resistant cells.

Fig. S3. KRAS amplification in parental and resistant NCIH508 and HCA-46 cell lines.

Fig. S4. Measurement of RAS activation in resistant cells.

Fig. S5. Biochemical validation of siRNA-mediated gene knockdown.

Fig. S6. Pharmacological inhibition of MEK in cells resistant to anti-EGFR blockade.

Fig. S7. Pharmacological inhibition of EGFR or MEK1/2 with cetuximab or pimasertib.

Fig. S8. Sensitivity to combinatorial EGFR and MEK inhibition in a mouse xenograft from a metastatic CRC patient who relapsed after anti-EGFR therapy.

Table S1. Frequencies of mutant alleles in resistant cell lines assessed by BEAMing.

Table S2. (separate Excel file) Tabular data for NCIH508 R-cetux and OXCO-2 R1-cetux xenograft measurements, as well as PDX measurements.

Table S3. List of primers for gene amplification and sequencing.

Table S4. List of primers and probes for BEAMing analysis.

Table S5. List of primers for gene copy number analysis by real-time PCR.

Table S6. List of siRNAs.

References (4245)


  1. Acknowledgments: We thank A. Cassingena and C. Cancelliere for technical assistance. The DiFi and OXCO-2 cell lines were gifts from J. Baselga in November 2004 (Oncology Department of Vall d’Hebron University Hospital, Barcelona, Spain) and V. Cerundolo in March 2010 (Weatherall Institute of Molecular Medicine, University of Oxford, UK), respectively. The LIM1215 parental cell line was obtained from R. Whitehead (Vanderbilt University, Nashville), with permission from the Ludwig Institute for Cancer Research (Zurich, Switzerland). Funding: Supported by the European Community’s Seventh Framework Programme under grant agreement no. 259015 COLTHERES (A.B. and S.S.); Associazione Italiana per la Ricerca sul Cancro (AIRC) IG grant no. 12812 (A.B.); AIRC MFAG no. 11349 (F.D.N.); “Farmacogenomica”—5 per mille 2009 MIUR—Fondazione Piemontese per la Ricerca sul Cancro—ONLUS (F.D.N.); AIRC 2010 Special Program Molecular Clinical Oncology 5 per mille, project no. 9970 (A.B. and S.S.); FPRC 5 per mille 2010 Ministero della Salute (A.B.); Ministero dell’Istruzione, dell’Università e della Ricerca, progetto PRIN (A.B.); Progetti di Ateneo-2011, Università di Torino (codice ORTO11RKTW to A.B.); Oncologia Ca’ Granda ONLUS (OCGO) Fondazione (S.S.); and Grant Terapia Molecolare dei Tumori from OCGO Fondazione. Author contributions: A.B., F.D.N., and S.M. designed the study; F.D.N. and A.B. devised the MEK-EGFR combinatorial treatment; A.B. conceived the idea of monitoring genetic alterations in patients’ blood. A.S.-B., K.B., A.A., and S.S. identified and provided clinical data on patients treated with anti-EGFR mAbs. S.M., S.A., S.L., G.S., A.L., S.H., M.R., M.B., L.L., C.L., and E.V. generated and analyzed laboratory-based data; A.B., F.D.N., and S.M. wrote the manuscript. All authors revised and approved the manuscript. Competing interests: S.S. is a clinical investigator of MEK inhibitors for CRC for Bayer, GlaxoSmithKline, and Novartis. A.S.-B. is a member of the advisory board for Amgen and has received speaking fees from Amgen and Bayer. A.B. is a shareholder and board member of Horizon Discovery (Cambridge, UK). A.B., F.D.N., and S.S. are co-inventors of a patent on “Methods of detecting and predicting acquired resistance to anti-EGFR treatment in colorectal cancer” (application number: 42823-502P01US). The other authors declare no competing interests. Data and materials availability: LIM1215 can be obtained through materials transfer agreement from the Ludwig Institute for Cancer Research (Zurich, Switzerland). HCA-46 cells are commercially available from the European Collection of Cell Cultures (ECACC catalog no. 07031601). The NCIH508 cells are commercially available from the American Type Culture Collection (catalog no. ATCC CCL-253). The DiFi and OXCO-2 cell lines were gifts from J. Baselga in November 2004 (Oncology Department of Vall d’Hebron University Hospital, Barcelona, Spain) and V. Cerundolo in March 2010 (Weatherall Institute of Molecular Medicine, University of Oxford, UK), respectively. Both cell lines can be obtained from the corresponding author.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article