Inactivation of NF1 Promotes Resistance to EGFR Inhibition in KRAS/NRAS/BRAF^V600-Wild-Type Colorectal Cancer Free

EGFR-targeting mAbs, cetuximab and panitumumab, have shown meaningful clinical efficacy as single agents and in combination with cytotoxic agents in patients with metastatic KRAS and NRAS exon 2 to 4 wild-type colorectal cancer (1–3). In contrast, 40% and 5% of colorectal cancer tumors that harbor activating KRAS and NRAS mutations, respectively (3, 4), exhibit primary resistance to anti-EGFR therapies, resulting in constitutive activation of the MAPK pathway and subsequent colorectal cancer cell survival and proliferation, despite upstream EGFR inhibition.

Nevertheless, even in patients with KRAS-wild-type tumors, the radiological response rate following treatment with anti-EGFR antibodies alone is limited to approximately 20%, with 25% of patients having progressive disease while on treatment (1). Other genetic aberrations may be implicated in anti-EGFR resistance, although these have not as yet been integrated in routine clinical practice. These include: activating mutations of the PIK3CA, BRAF, MAP2K1 genes, genetic and epigenetic loss of PTEN expression, acquired extracellular EGFR domain mutations, and HER2 amplification (5–7). Aside from genetic aberrations, compensatory signaling via activation of other receptors (e.g., via other members of ERBB family) and increased expression of EGFR ligands have also been suggested to contribute to anti-EGFR resistance in KRAS/NRAS-wild-type colorectal cancer (6, 8). Polyclonal resistance has also been reported to be a common feature in anti-EGFR–refractory colorectal cancer, with evolutionary changes typically leading to the selection of the “fittest” RAS-mutant resistant clones (9).

Targeted genome editing technologies, such as clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9, allow for the selective knock-out of genes and subsequent evaluation of the functional consequences of this. In the CRISPR/Cas9 system, the Cas9 nuclease is directed to the gene of interest through complementary binding of a single-guide RNA (sgRNA) to the target DNA. The Cas9 nuclease then cleaves DNA resulting in a double-stranded DNA break. This is then repaired by the error-prone nonhomologous end joining leading to irreversible gene disruption, through insertion or deletion of events. The CRISPR/Cas9 system has been shown to be superior to other commonly used high-throughput genetic approaches, such as pooled short hairpin RNA (shRNA)-based screens for performing functional genomic screens. In particular, CRISPR/Cas9 has been shown to have lower noise and less off-target effects resulting in a low FDR and a better consistency across different cell lines and reagents (10). Pooled CRISPR knock-out screens allow for the unbiased and simultaneous perturbation of multiple genes, at a whole genome scale, enabling the potential landscape of resistance mechanisms to be identified (11).

With the aim of improving our knowledge of mechanisms of resistance to anti-EGFR therapies in KRAS/NRAS/BRAF^V600-wild-type colorectal cancer, we conducted an unbiased pooled, genome-wide CRISPR screen using an anti-EGFR–sensitive colorectal cancer cell line under continuous exposure to the anti-EGFR tyrosine kinase inhibitor (TKI) gefitinib. Characterization of surviving cells and further validation identified inactivation of NF1, a known tumor suppressor gene that encodes for the RAS-GTPase–activating protein (RAS-GAP) neurofibromin, as a novel mechanism of resistance to both gefitinib and cetuximab. Our findings may aid personalized medicine approaches to direct therapeutic strategies for improved patient benefit.

Cell culture medium were obtained from Sigma Aldrich and Life Technologies. FBS Good was from PAN Biotech UK Ltd. The HEK293T and LIM1215 cell lines were obtained from the ATCC, HT115 cells were from Public Health England, DIFI cells were a kind gift from Professor Alberto Bardelli, and NCIH508 cells were a kind gift from Dr Anguraj Sadanandam. SNUC4 and SNU1040 cells were obtained from the Korean Cell Line Bank. Cell lines were cultured in medium recommended by the manufacturers for up to 3 months at a time, and cell line authentication was not performed. Cell lines were tested for Mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza). Small-molecule inhibitors were purchased from Selleck Chemicals. Cetuximab was supplied under a material transfer agreement from Merck.

After the desired treatment, cells were washed with PBS and lysed in 1% SDS lysis buffer (1% SDS, 10 mmol/L EDTA, 50 mmol/L Tris, pH 8). Bicinchoninic acid (Sigma) was used to determine protein concentration. Equal amounts of protein were separated by gel electrophoresis, using NuPAGE polyacrylamide gels (Life Technologies). Proteins were transferred to a nitrocellulose membrane using the iBlot 2 system (Life Technologies) or by wet transfer using a Bio-Rad Criterion Blotter and then blocked with TBS LI-COR blocking buffer (LI-COR Biosciences). Membranes were incubated with the primary antibodies overnight at 4°C, followed by IRdye-conjugated secondary antibodies (LI-COR Biosciences) and detected using an Odyssey Fc imaging system (LI-COR Biosciences). Quantification of Western blots was performed using Image Studio Lite (LI-COR Biosciences). Details of the antibodies used can be found in Supplementary Table S1.

For GI₅₀ determinations, cells were seeded in 96-well plates. The next day, cells were treated with increasing concentrations of inhibitor or with DMSO alone. After a 96-hour incubation period, cell proliferation was quantified by fluorescent detection of the reduction of resazurin to resorufin by viable cells and normalized to DMSO-treated wells. GI₅₀ values were calculated using nonlinear regression analysis in GraphPad Prism software. For longer-term colony assays, cells were seeded into 12-well plates (0.5–5 × 10⁴ cells/well), treated with inhibitors the next day and incubated for 10–14 days. Cells were fixed with 4% formaldehyde/PBS for 20 minutes, then stained with 0.5% crystal violet in 70% ethanol. Excess dye was removed by washing with water and plates were imaged with a GelCount (Oxford Optronix). Cell proliferation was quantified by solubilization of crystal violet dye in 10% acetic acid and absorbance measured at 595 nm using an EMax Plus plate reader (Molecular Devices), then expressed as a percentage of vehicle-treated cells.

DIFI cells were engineered to express Cas9 by centrifugation of 2 × 10⁵ cells with pXPR101-Cas9 lentivirus (The Broad Institute, Cambridge, MA), in the presence of 8 μg/mL polybrene for 1 hour at 37°C. Cells were incubated with fresh media overnight, before cells were trypsinized and pooled for selection with 10 μg/mL blasticidin for 7 days to select for successfully transduced cells.

To generate the lentiviral particles for the Brunello whole genome library (Addgene #73178) containing 77,440 sgRNAs targeting 19,110 genes (12, 13), HEK293T cells were seeded at 40% confluence in T225 flasks. After 24 hours, medium was replaced with 13 mL OptiMEM 1 hour before transfection with plasmids encoding the library. 20 μg of Brunello plasmid, 10 μg of pMD2.G, and 15 μg of psPAX2 were added to 4 mL of OptiMEM. 100 μL of Lipofectamine 2000 were added in another 4 mL of OptiMEM. After 5 minutes, these were mixed and added to the HEK293T cells. After 6 hours, the media were aspirated and replaced with 30 mL DMEM, 10% FBS. After 60 hours, medium was harvested and centrifuged. The supernatant containing lentiviral particles was then filtered through a 0.45-μm low protein-binding membrane and was stored in 1 mL aliquots at - 80°C. Viral titer was determined by transducing target cells with increasing volumes of virus (50, 100, 200, 300, 500 μL/well) and centrifugation at 2,000 rpm for 2 hours at 30°C with 8 μg/mL polybrene. 2 mL fresh media were then added to each well. The next day, the cells were trypsinized and replated in 6-well plates at a density of 3 × 10⁵ cells/well. Each virus dilution was performed in duplicate, with one well treated with 1 μg/mL puromycin. After 72 hours, cells were trypsinized and counted to determine the infection efficiency. This was calculated by dividing the cell number of puromycin-treated cells by the number for untreated cells for each of the virus dilutions. For individual sgRNAs (Supplementary Table S2), HEK293T cells were seeded at a density of 2.4 × 10⁶ cells/plate in 10-cm plates. The next day, cells were transfected with sgRNA plasmid (3 μg) and the packaging plasmids psPAX2 (2.1 μg) and pmD2.G (0.9 μg) using 30 μL Lipofectamine per transfection. Cells were incubated for 72 hours at 37°C, after which the supernatant was collected and stored in 0.5 mL aliquots at - 80°C for future experiments. Each batch of lentivirus was titrated on cells to determine concentration needed for 100% infection efficiency. To generate cell populations expressing each sgRNA, cells were transduced with the appropriate lentivirus and selected for using 1 μg/mL puromycin. Knock-out/suppression of the target gene was confirmed by Western blotting.

DIFI-Cas9 cells were seeded in 12-well plates at a density of 1.5 × 10⁶ cells/well in 2 mL medium. Cells were transduced with the Brunello pooled lentiviral library with a predicted representation of 750 cells/sgRNA at an infection efficiency of 50%. Cells were centrifuged at 2,000 rpm for 2 hours at 30°C in the presence of lentivirus and 8 μg/mL polybrene, followed by incubation in fresh medium overnight. Cells were pooled and seeded into T175 flasks at a density of 5 × 10⁶ cells/flask for selection with 1 μg/mL puromycin for 7 days and passaged as necessary. In parallel, cells were seeded in 6-well plates to determine infection efficiency. After 7 days of selection, DIFI-Cas9 cells were split into three arms and treated with either 0.1% DMSO or 240 nmol/L gefitinib. While we initially intended to use cetuximab in the screen, we found that cetuximab treatment did not produce the expected antiproliferative effects in the large-scale, longer-term cell culture format required for the screen. We hypothesize that cetuximab was binding to the plastic of the triple layer flasks and the free concentration of cetuximab in the culture medium was reduced, lessening the antiproliferative activity. We therefore elected to use gefitinib for the CRISPR screen, which behaved as expected under these culture conditions. We then validated our gefitinib screen results using both cetuximab and gefitinib.

Throughout the screen, cells were passaged as necessary, maintaining a total representation of 750 cells/sgRNA in each replicate (three per condition). After 8 population doublings, cells from each arm were collected and cell pellets stored at - 80°C. Genomic DNA was extracted using the QIAamp DNA Blood Maxi Kit (Qiagen). PCR amplification of sgRNA sequences and next-generation sequencing (NGS) was carried out by Cellecta (catalog no.: LNGS-900). The representation of sgRNA in each sample was quantified by counting the number of specific reads generated by NGS on an Illumina Instrument (e.g., NextSeq 500 or HiSeq; ref. 12).

The abundance of each sgRNA in each replicate was quantified by calculating the log₂ (sequencing reads/million; RPM), according to the formula below.

The log₂-fold change (LFC) from the pDNA sample was calculated by normalizing RPM for each sgRNA in each replicate to that in the pDNA sample. The LFC between the DMSO and drug-treated arms was calculated as the difference in average LFC across three replicates. This was used to rank individual sgRNAs according to their selective depletion or enrichment in the drug-treated arms. Top scoring genes were ranked according to the number of independent high scoring sgRNAs targeting the same gene, according to the STARS gene-ranking algorithm (12). To assess depletion of essential genes from the population, as a positive control for successful gene editing, the RPM for each sgRNA was normalized to the plasmid DNA to calculate the LFC from baseline. The list of 885 core essential genes was kindly provided by Dr. Marco Licciardello (The Institute of Cancer Research, London, United Kingdom) and was compiled from the genes that were consistently and significantly depleted in all cell lines tested from three previous publications (14–16).

We identified the DIFI cell line as a suitable candidate for use in a genome-scale CRISPR screen to discover genes that when suppressed, could promote resistance to EGFR-targeting therapies. The DIFI cell line harbors amplification of the EGFR gene, which is reflected in elevated protein expression and exquisite sensitivity to gefitinib, a small-molecule kinase inhibitor of EGFR and to cetuximab, an antibody that targets the extracellular domain of EGFR. To conduct a CRISPR/Cas9 screen, DIFI cells were transduced with a lentiviral expression vector for the RNA-guided nuclease Cas9. Both parental DIFI cells and the DIFI-Cas9 cells had near-identical GI₅₀ values for both gefitinib and cetuximab and the phosphorylation of EGFR, ERK, and AKT was suppressed to a similar extent by both treatments (Supplementary Fig. S1; Supplementary Table S3).

For the whole-genome CRISPR screen, we elected to use the Brunello lentiviral sgRNA library which is composed of 76,441 sgRNAs targeting 19,114 genes and 1,000 control, nontargeting sgRNAs. DIFI-Cas9 cells were transduced with the Brunello library and after selection and expansion of the cell population, cells were cultured in the presence of either DMSO or 240 nmol/L gefitinib (Fig. 1A), a concentration that inhibits EGFR and ERK phosphorylation and results in durable suppression of cell proliferation (Supplementary Fig. S2). Initially, gefitinib-treated cells did not proliferate but after 3–4 weeks, proliferation resumed at a rate similar to the DMSO-treated cells (Fig. 1B). Cells were passaged for up to 8 population doublings then genomic DNA was purified, sgRNAs were amplified by PCR, and the abundance of each sgRNA was determined by next-generation sequencing (NGS). The abundance of each sgRNA was determined relative to the Brunello plasmid DNA library as a reference. Comparison of two DMSO-treated replicates demonstrated good concordance between the two replicates (Pearson correlation 0.722), that nontargeting control (NTC) sgRNAs were largely unaltered in representation in the population and that sgRNAs targeting known essential genes were depleted from the population (Fig. 1C).

The LFC of the gefitinib-treated replicates was compared with the DMSO-treated replicates to identify sgRNAs that were enriched in the presence of gefitinib. Strikingly, all four sgRNAs targeting NF1 were the most highly enriched (ranked 1–4) in the presence of gefitinib relative to DMSO (Fig. 1D). STARS analysis was then used to rank genes for positive enrichment in the presence of gefitinib based on the sgRNA rankings, NF1 was the most highly ranked gene (Fig. 1E; Supplementary Table S4). In comparison, the next three most highly ranked genes BRMS1, TBL1XR1, and RNF7 showed inconsistent and relatively weak enrichment in the presence of gefitinib for individual sgRNAs and also across the three replicates (Supplementary Fig. S3). NF1 encodes the protein neurofibromin, a RAS-GAP which stimulates RAS proteins to hydrolyze bound GTP to GDP and consequently switch off RAS (17). Suppression of NF1 would therefore maintain RAS in the active, GTP-bound conformation and promote downstream signaling. We confirmed that targeting of NF1 with two independent sgRNAs not used in the genome-scale screen suppressed NF1 protein expression and resulted in increased levels of RAS-GTP in the DIFI-Cas9 cells (Fig. 1F).

We generated stable pools of DIFI cells which expressed either an sgRNA targeting GFP or two different sgRNAs targeting NF1. To validate the CRISPR screen findings, these cells were treated with increasing concentrations of gefitinib or cetuximab for 96 hours. Cell proliferation was determined by fluorescent detection of the reduction of resazurin to resorufin (Fig. 2A). Suppression of NF1 was associated with an increase in the GI₅₀ for gefitinib or cetuximab (Supplementary Table S3). Furthermore, longer-term colony formation assays demonstrated resistance to gefitinib and cetuximab in cells expressing NF1-targeting sgRNAs (Fig. 2B; Supplementary Fig. S4). Western blotting of lysates from cells treated with 300 nmol/L gefitinib or 1 μg/mL cetuximab demonstrated impaired inhibition of ERK phosphorylation in the NF1-targeted cells compared with the GFP-targeted controls, despite comparable inhibition of EGFR phosphorylation. In contrast, inhibition of phospho-AKT was unaffected by suppression of NF1 (Fig. 2C). Therefore, suppression of NF1 results in sufficient MAPK pathway activity to sustain cell proliferation in the presence of gefitinib or cetuximab, consistent with activation of RAS proteins.

We hypothesized that inhibition of MEK may be necessary to overcome loss of NF1-mediated resistance to EGFR inhibition. Hence, DIFI cells expressing NF1-targeting sgRNAs were treated with increasing concentrations of the MEK inhibitor trametinib for 96 hours and cell proliferation determined. As anticipated, cells with decreased NF1 expression were equally sensitive to trametinib compared with the GFP-targeting control (Fig. 3A; Supplementary Table S3). Western blotting of lysates from cells treated with increasing concentrations of trametinib for 24 hours demonstrated comparable inhibition of ERK phosphorylation relative to the GFP-targeting control (Fig. 3B). Prior reports suggest that the combination of EGFR and MEK inhibitors elicits synergistic activity in EGFR inhibitor–resistant models. We therefore tested the combination of gefitinib or cetuximab with trametinib. DIFI cells with NF1 suppression were treated with a matrix of gefitinib and trametinib or cetuximab and trametinib for 96 hours and cell proliferation was determined. Synergy was assessed using the Bliss independence model and demonstrated synergistic activity across both DIFI-sgGFP cells and DIFI-sgNF1 cells, affirming that this combination has potential therapeutic utility in NF1-suppressed models (Fig. 3C; ref. 18). In longer-term colony formation assays, modest resistance to trametinib was observed in DIFI-sgNF1 cells compared with DIFI-sgGFP cells but the combination of both gefitinib and trametinib or cetuximab and trametinib resulted in near-complete inhibition of colony formation (Fig. 3D; Supplementary Fig. S5). Western blotting of lysates from cells treated with the combination of gefitinib or cetuximab with trametinib for 24 hours demonstrated higher basal levels of ERK phosphorylation in DIFI-sgNF1 cells and incomplete inhibition of ERK phosphorylation following gefitinib or cetuximab treatment (Fig. 3E). The combination of trametinib with either gefitinib or cetuximab was required to achieve near-complete suppression of ERK phosphorylation. Therefore, inhibition of the MAPK pathway appears to be required for a robust antiproliferative effect in cells lacking NF1 expression.

To translate the findings from our genetic models of NF1-suppression to established cell lines, we consulted the Cancer Dependency Map for colorectal cancer cell lines that harbored mutations in NF1. Of 60 colorectal cancer cell lines, 17 (28%) were NF1-mutant. NF1 mutation was associated with a significant reduction in NF1 mRNA expression (Fig. 4A). Moreover, we utilized a recent proteomics-based profiling of 50 colorectal cancer cell lines, in which 12 (24%) were NF1-mutant (19). NF1 mutation was associated with significantly reduced expression of NF1 protein (Fig. 4B). Therefore, we selected three NF1-wild-type and three NF1-mutant colorectal cancer cell lines (all KRAS/NRAS/BRAF^V600-wild-type) and assessed NF1 expression by Western blotting. The NF1-mutant cell lines HT115 (NF1^R1241*), SNUC4 (NF1^T676fs), and SNU1040 (NF1^R461*) had little or no expression of NF1 protein (Fig. 4C).

The sensitivity of this panel of colorectal cancer cell lines to gefitinib and cetuximab was assessed and NF1 mutation was associated with reduced sensitivity to both agents (Fig. 4D; Supplementary Table S3). Inhibition of EGFR and ERK phosphorylation was also assessed in these cell lines following a 72-hour exposure to increasing concentrations of gefitinib or cetuximab (Fig. 4E). While inhibition of EGFR phosphorylation was generally observed across all cell lines treated with gefitinib, inhibition of ERK phosphorylation was less complete in NF1-mutant cell lines. Cetuximab treatment also resulted in decreased EGFR and ERK phosphorylation although interestingly, the association with cetuximab sensitivity was less pronounced.

As we have shown that DIFI cells with NF1 suppression were equally sensitive to the MEK inhibitor trametinib, the panel of colorectal cancer cell lines was assessed for sensitivity to trametinib (Fig. 5A; Supplementary Table S3). Only the HT115 NF1-mutant cell line showed sensitivity to trametinib similar to the NF1-wild-type cell lines DIFI, LIM1215, and NCIH508. To determine a suitable therapeutic strategy for those cell lines not sensitive to trametinib, we studied the mutational profile of the SNUC4 and SNU1040 cells for events that may drive resistance to trametinib. Both cell lines have damaging mutations in PTEN, which could promote resistance via activation of the PI3K pathway (www.DepMap.org). Notably, HT115 cells have mutations in PIK3CA, resulting in a triple mutant (PIK3CA^{E321D, R770Q, R88Q}), although this did not appear to be sufficient to drive resistance to trametinib. We first tested the cell line panel for sensitivity to the p110α-specific PI3K inhibitor BYL719 and all cell lines appeared to be resistant (Fig. 5B; Supplementary Table S3). We therefore tested the combination of gefitinib and trametinib or gefitinib and BYL719 for synergy as assessed by the Bliss independence model (Fig. 5C). Some synergy was observed in the HT115 cell line between gefitinib and trametinib or gefitinib and BYL719 but no synergy was seen with this combination in SNUC4 or SNU1040 cells. As both SNUC4 and SNU1040 cell lines lack expression of PTEN (20), it was possible that activation of the PI3K pathway via PTEN loss was sufficient to confer resistance to the dual EGFR/MEK inhibitor combination. We confirmed that both SNUC4 and SNU1040 cells exhibited high levels of AKT phosphorylation, relative to the other PTEN-wild-type cell lines in our panel, indicative of PTEN loss (Supplementary Fig. S6). Despite mutations in PIK3CA, HT115 cells did not exhibit high levels of AKT phosphorylation, which may explain their relative sensitivity to MEK inhibition.

In light of the above data, we hypothesized that the triple combination of gefitinib, trametinib, and BYL719 may inhibit cell proliferation more effectively than single agents or paired combinations. HT115, SNUC4, and SNU1040 cells were exposed to pathway-inhibitory concentrations of gefitinib, trametinib, and BYL719 alone or in combination (Fig. 5D). In HT115 cells, the combination of gefitinib and trametinib reduced colony formation by more than 90% and the combination of gefitinib and BYL719 inhibited colony formation to a similar extent. Therefore, dual inhibition of EGFR and MEK or EGFR and p110α appears to be synthetic lethal in HT115 cells. In contrast, the SNUC4 and SNU1040 cell lines were more resistant to the combination of gefitinib and trametinib but the addition of BYL719 further decreased colony formation (Fig. 5D; Supplementary Fig. S7). Analysis of lysates from cell lines treated with gefitinib, trametinib, or BYL719 for 24 hours indicated that dual suppression of both AKT and ERK phosphorylation by the triple combination was associated with greater antiproliferative activity (Fig. 5E). The above data suggest that the triple combination of gefitinib, trametinib, and BYL719 may yield greater antiproliferative activity than single agents or dual combinations. Given the high frequency of PI3K pathway mutations in KRAS/NRAS/BRAF^V600-wild-type colorectal cancer cell lines, these data raise the tantalizing prospect that the addition of a PI3K inhibitor to cetuximab and trametinib may enhance antitumor activity, as has been observed in BRAF-mutant colorectal cancers with the combination of BRAF, EGFR, and PI3K inhibitors.

The advent of targeted therapies has enabled the concept of precision medicine, whereby drugs are administered to a selected group of patients based on the expression (or lack of expression) of evidence-based molecular biomarkers. This approach not only maximizes efficacy and cost-effectiveness, it also prevents unwarranted toxicity. However, drug resistance whether through selection of subclonal populations that harbor preexisting resistant clones or by genetic or nongenetic adaptions to treatment, limit the duration of response and lead to disease progression. Understanding potential mechanisms of resistance to targeted therapies may therefore enable alternate therapeutic strategies such as combinations of drugs to overcome bypass mechanisms or targeting points of convergence of the primary target and resistance pathways.

CRISPR/Cas9 screens enable parallel assessment of the loss of thousands of genes on sensitivity to drug treatment. Typically, a synthetic-lethal approach is taken whereby loss of a particular gene may sensitize to treatment and sgRNAs targeting the sensitizing gene are depleted from the population. However, these “drop-out” screens require transduction of a greater number of cells to ensure sufficient sgRNA representation and greater sequencing depth, resulting in a reduced signal-to-noise ratio and potentially reduced confidence in hit calling. In this case, we sought genes that when lost, confer resistance to treatment and sgRNAs that target such genes are enriched in the population. The benefits of this approach are that sgRNA representation can be lower, requiring fewer cells for the initial sgRNA library viral transduction step, a lower sequencing depth, and greater signal-to-noise ratio, permitting more statistically robust hit calling.

Our genome-scale CRISPR screen identified suppression of NF1 expression as the most highly ranked and robust driver of resistance to EGFR inhibition. Because NF1 acts as a RAS-GAP, loss of this protein results in the activation of RAS, as it is stabilized in its GTP-bound form (17). KRAS and NRAS mutations are a well-established driver of resistance to EGFR-targeting therapy in colorectal cancer, therefore, reduced expression of NF1 may well phenocopy genetic alterations that promote RAS activity.

There have been recent publications that further support the notion that impairment of NF1 function may be implicated in anti-EGFR resistance in colorectal cancer. A study that analyzed the mutation profiles of 33 patients with metastatic colorectal cancer using next-generation sequencing before starting cetuximab suggested that patients with NF1 mutations (n = 4) had significantly shorter progression-free survival, following treatment with cetuximab when compared with those with NF1-wild-type tumors (21). In our own institution, a separate study investigated anti-EGFR therapy resistance in RAS-wild-type colorectal cancer by carrying out exome and RNA sequencing from biopsies taken at baseline and upon progression to cetuximab. Tumors from 2 of 35 patients with primary anti-EGFR resistance had inactivating mutations of NF1 (22). Recently, it was reported that in a limited study of CRISPR-mediated knockout of RAS-GAPs in human colorectal cancer organoids, NF1 knock-out conferred resistance to the EGFR TKI afatinib (23). Our genome-scale approach further demonstrated that NF1 was the most highly ranked loss-of-function event to drive resistance. Compelling evidence in support of NF1 loss as a possible resistance mechanism to anti-EGFR therapies was also reported by a study that used genome-wide siRNA screen in EGFR-mutant non–small cell lung cancer (NSCLC) cells that were exposed to another anti-EGFR TKI, erlotinib. The authors demonstrated that NF1 mRNA expression is downregulated in an EGFR-driven, erlotinib-resistant inducible model of NSCLC, potentially as an alternative resistance mechanism to mutation of the gatekeeper residue of EGFR itself. Furthermore, knockdown of NF1 expression by shRNA also conferred resistance to EGFR inhibition in the PC9 NSCLC xenograft model, demonstrating this effect translated from in vitro models to in vivo models. Reduced expression of NF1 was also associated with resistance to EGFR-directed treatment and lower survival in patients with NSCLC (24).

Our findings strongly demonstrate that relief of negative regulation of RAS via NF1 loss is a mechanism of resistance in NRAS/KRAS/BRAF^V600-wild-type colorectal cancer. Mutation of NF1 has been reported to occur in approximately 5% of colorectal cancer tumors (25). Aside genomic aberrations, loss of NF1 expression may occur via epigenetic mechanisms such as gene promoter methylation, therefore, loss of NF1 function may account for more than 5% of anti-EGFR resistance in KRAS/NRAS/BRAF^V600-wild-type colorectal cancer. Use of this information in clinical decision-making may spare these patients from unnecessary treatment.

Interestingly, our data point to activation of the MAPK pathway as being a major driver of resistance, as targeting of MEK overcomes resistance associated with suppression of NF1. This is in line with previous studies that suggested that escape routes from EGFR blockade in colorectal cancer biochemically converge on MAPK activation and that vertical suppression with concomitant blockade of EGFR and MEK could overcome primary and acquired resistance to anti-EGFR agents in KRAS-mutant colorectal cancer cells (26, 27). However, in our initial validation, we identified two cell lines that harbored NF1 mutations, SNUC4 and SNU1040, which were relatively resistant to trametinib or the combination of trametinib and gefitinib. These cell lines have cooccurring mutations in PTEN, which may contribute to the resistant phenotype, as combined inhibition of EGFR, MEK, and p110α suppressed cell proliferation as depicted in our simplified model in Fig. 6. Aside from PTEN inactivation, compensatory PI3K pathway activation may also arise from other mechanisms including activating PIK3CA mutations and nongenetic compensatory activation of the PI3K signaling pathway. The notion of compensatory PI3K pathway activation in cells with constitutive activation of the MAPK pathway is strongly supported by two studies that utilized KRAS-mutant colorectal cancer cell lines. In one study, KRAS-mutant colorectal cancer cells were rendered resistant to the combination of cetuximab and refametinib (a selective MEK inhibitor) after continuous exposure to increasing concentration of both drugs. It was suggested that resistance to this combination was due to secondary PI3K activation, following cooperative activation of multiple receptor tyrosine kinases: HER2, HER3, and IGF1R (28). Similarly, in another study, shRNA knockdown of KRAS expression in KRAS-mutant colorectal cancer cell lines led to ERK suppression but was not sufficient to downregulate the PI3K pathway. Instead PI3K activation was regulated via activation of other receptor tyrosine kinases rather than KRAS itself (29). Similar to KRAS-mutant colorectal cancer, we showed that in colorectal cancer cells with NF1 suppression and cooccurring PTEN mutation, bypass PI3K pathway activation may also occur. Our data agree with studies in colorectal cancer or other EGFR-driven cancers where the combination of EGFR inhibitors and MEK or PI3K inhibitors leads to greater antiproliferative activity, including in the context of RAS-driven resistance in both in vivo models (30–32) and early-phase clinical trials (33).

In conclusion, our results suggest that loss of NF1 function may be a predictive biomarker of response to anti-EGFR therapy in colorectal cancer and that in colorectal cancer cells with NF1 suppression concomitant inhibition of the MAPK and PI3K pathways may be required to overcome anti-EGFR resistance. Our data support the concept that higher-order drug combinations may be required to elicit robust antiproliferative effects in colorectal cancer (34). Further clinical validation of NF1 status as a predictor of response to anti-EGFR targeting antibodies in patients with colorectal cancer with KRAS/NRAS/BRAF^V600-wild-type tumors is warranted.

D. Cunningham reports receiving commercial research support from Amgen (REAL 3 Trial for the Royal Marsden NHS Foundation Trust), Sanofi (Sanofi Trial for the Royal Marsden NHS Foundation Trust), Janssen (Janssen IMYC Trial for the Royal Marsden NHS Foundation Trust), Merck (Merck ICONIC/POLEM Trial for the Royal Marsden NHS Foundation Trust), Merrimack (PLATFORM Trial for the Royal Marsden NHS Foundation Trust), AstraZeneca (FRGR Trial for the Royal Marsden NHS Foundation Trust), Celgene (PROSPECT R Trial for the Royal Marsden NHS Foundation Trust), MedImmune (MedImmune PLATFORM Trial for the Royal Marsden NHS Foundation Trust), Bayer (Bayer PROSPECT R Trial for the Royal Marsden NHS Foundation Trust), 4SC (EMERGE Trial for the Royal Marsden NHS Foundation Trust), Clovis (Clovis PLATFORM Trial for the Royal Marsden NHS Foundation Trust), Eli Lilly (Eli Lilly PLATFORM Trial for the Royal Marsden NHS Foundation Trust), and has ownership interest (including patents) in OVIBIO (for the Royal Marsden NHS Foundation Trust), is a consultant/advisory board member for OVIBIO (on scientific advisory board for the Royal Marsden NHS Foundation Trust). U. Banerji is an employee at The Institute of Cancer Research, has other commercial research support from AstraZeneca, Onyx, BTG International, Carrick Therapeutics Ltd, Chugai Pharma Japan, and Verastem, has received speakers bureau honoraria for Astellas, Novartis, Karus Therapeutics, Phoenix Solutions, Eli Lilly, Janssen, and Boehringer Ingelheim, and has ownership interest (including patents) in The Institute of Cancer Research. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Georgiou, U. Banerji, S.R. Whittaker

Development of methodology: A. Georgiou, A. Stewart, S.R. Whittaker

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Georgiou, D. Cunningham, U. Banerji, S.R. Whittaker

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Georgiou, A. Stewart, S.R. Whittaker

Writing, review, and/or revision of the manuscript: A. Georgiou, D. Cunningham, U. Banerji, S.R. Whittaker

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.R. Whittaker

Study supervision: U. Banerji, S.R. Whittaker

We thank Dr. Marco Licciardello, Institute of Cancer Research, for assistance with generating the Brunello pooled lentiviral library. This work was supported by a Cancer Research UK grant (awarded to A. Georgiou; clinical research fellowship at The Institute of Cancer Research). Authors acknowledge funding from the National Institute of Health (NIHR) Biomedical Research Centre (BRC), and Experimental Cancer Medicine Centre (ECMC) initiatives at The Institute of Cancer Research and The Royal Marsden NHS Foundation trust. U. Banerji is a recipient of a NIHR research professorship award (RP-2016-07-028).

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.